ThermagVII PDF

REFRIGERATION SCIENCE AND TECHNOLOGY
PROCEEDINGS
Thermag VII
7th International Conference on

Magnetic Refrigeration at Room Temperature
11-14 September, 2016, Torino, Italy
INTERNATIONAL INSTITUTE OF REFRIGERATION

INSTITUT INTERNATIONAL DU FROID
Commissions A1,B2,E2
ISBN: 978-2-36215-016-6
ISSN: 0151-1637
Code: 2016/6
Thermag VII
7th International Conference on

Magnetic Refrigeration at Room Temperature
11-14 September, 2016

Torino Incontra Conference Center
Via Nino Costa 8, Torino, Italy
Editors
Karl Sandeman (chair), Brooklyn College (USA)
Ekkes Brück, Delft University of Technology (Netherlands)
Martino LoBue, CNRS Satie (France)
Jader Barbosa, Federal University of Santa Catarina (Brazil)
Afef Kedous-Lebouc, Laboratoire de Génie Electrique de Grenoble G2ELAB (France)
Andrew Rowe, University of Victoria (Canada)
Organized by Exhibitors
Under the auspices of Sponsor

177, Boulevard Malesherbes, 75017 Paris, France
i
Copyright and Disclaimer
The information provided in this document is based on the current state of the art and is designed to assist engineers,
scientists, companies and other organizations. It is a preliminary source of information that will need to be comple-
mented prior to any detailed application or project. Whilst all possible care has been taken in the production of this
document, the IIR, its employees, officers and consultants cannot accept any liability for the accuracy or correctness of
the information provided nor for the consequences of its use or misuse. Any opinions expressed herein are entirely those
of the authors.
For full or partial reproduction of anything published in this document, proper acknowledgement should be made to the
original source and its author(s). No parts of the book may be commercially reproduced, recorded and stored in a re-
trieval system or transmitted in any form or by any means (mechanical, electrostatic, magnetic, optic photographic,
multimedia, Internet-based or otherwise) without permission in writing from the IIR.
Copyright © 2016 IIR/IIF. All rights reserved.
Droit de reproduction et avertissement

Les informations présentées dans ce document sont basées sur l’état de l’art actuel et sont destinées à aider dans leurs
travaux les ingénieurs, scientifiques, sociétés et organisations. Elles constituent une source première de renseignements
destinée à être complétée avant la réalisation de tout projet ou application. Malgré le soin apporté pour réaliser ce
document, l’IIF, son personnel et ses consultants se dégagent de toute responsabilité sur la précision et l’exactitude des
informations fournies et les conséquences de leur bonne ou mauvaise utilisation. Les opinions émises n’engagent que la
responsabilité de leurs auteurs. Toute reproduction partielle ou totale du contenu du document ne pourra être faite que
sous réserve de citation précise de la source originale.
De plus, toute reproduction partielle ou totale du contenu du document sera soumise à l’autorisation écrite de l’IIF si
cette reproduction est faite dans une intention commerciale, quelle que soit la forme de cette reproduction, enregistre-
ment, stockage ou transmission (mécanique, électrostatique, magnétique, optique, photographique, multimédia, Internet
ou autres).
Tous droits réservés. © 2016 IIF/IIR.

INSTITUT INTERNATIONAL DU FROID
177, Boulevard Malesherbes, 75017 Paris, France
t. +33 (0)1 42.27.32.35 | f. +33 (0)1 47.63.17.98 | iif-iir@iifiir.org | www.iifiir.org
ii
Thermag VII
www.thermag2016.com
CONFERENCE CHAIR
Vittorio Basso, Istituto Nazionale di Ricerca Metrologica (Italy)
LOCAL ORGANIZING COMMITTEE (INRIM)

Elena Olivetti
Michaela Kuepferling
Luca Martino
Cinzia Beatrice
Marco Piazzi
Carmen Curcio
PROGRAMME COMMITTEE
Asaya Fujita, National Institute of Advanced Industrial Science and Technology (Japan)
Nini Pryds, Technical University of Denmark (Denmark)
Alessandro Pastore, Camfridge (UK)
Franca Albertini, Institute of Materials for Electronics and Magnetism, CNR (Italy)
Luca Antonio Tagliafico, University of Genova (Italy)
PUBLICATION COMMITTEE
Karl Sandeman (chair), Brooklyn College (USA)
Ekkes Brück, Delft University of Technology (Netherlands)
Martino LoBue, CNRS Satie (France)
Jader Barbosa, Federal University of Santa Catarina (Brazil)
Afef Kedous-Lebouc, Laboratoire de Génie Electrique de Grenoble G2ELAB (France)
Andrew Rowe, University of Victoria (Canada)
LOCAL ORGANISING SECRETARIAT (INRIM)

Elisabetta Melli
Silvia Cavallero
PUBLICITY AND WEBSITE (IIF/IIR)

Deonie Lambert
iii
KEYNOTE SPEAKERS
OLIVER GUTFLEISCH
TU Darmstadt (Germany)
Mastering hysteresis in magnetocaloric materials
ANDREJ KITANOVSKI
University of Ljubljana (Slovenia)
Future developments in magnetocaloric refrigeration and heat pumping
LAURA H. LEWIS
Northeastern University (USA)
Multivariable tuning and optimization in selected magnetostructural systems
ANDRÉ PEREIRA
University of Porto (Portugal)
Thermal switch in a magnetic cooling engine: state of the art, opportunities and open challenges
AKIKO T. SAITO
Toshiba (Japan)
Room-temperature magnetic refrigeration: from basic research to development for application
NEIL WILSON
Camfridge (UK)
Magnetic Cooling: Camfridge achievements and industry challenges
iv
FOREWORD
This volume contains the Proceedings of the seventh International Conference on Magnetic Refrig-
eration at Room Temperature, Thermag VII, held at Torino Incontra Conference Center, Torino, It-
aly from 11 to 14 September 2016. The conference was organized by the Italian National Institute
for Research on Metrology (INRIM) and by the International Institute of Refrigeration (IIR). After
Montreux, Switzerland (2005), Portoroz, Slovenia (2007), Des Moines, United States (2009), Bao-
tou, China (2010), Grenoble, France (2012), and Victoria, Canada (2014), Torino had the honor to
host the seventh gathering.
In accordance with the aims of the Thermag VII committees, the present Proceedings outline the
state of the art of magnetic cooling as a potentially high efficiency and environmentally friendly
technology. The Thermag VII conference demonstrates the evolution of magnetic refrigeration from
a low temperature academic research field with potentially interesting applications, to an innovative
technology that functions at room temperature using highly-studied refrigerants in increasingly af-
fordable magnetic cooling systems. Scientific research in the field continues to be very active and
inter-disciplinary, examining the detailed performance of magnetocaloric materials and investigat-
ing novel thermal engineering techniques. In addition, the Proceedings offer perspectives on other
solid state refrigeration technologies beyond the realm of purely magnetic refrigeration.
Vittorio Basso
Thermag VII Conference chair
Karl Sandeman
Thermag VII Publication Committee chair
v
DEDICATION
The works of the seventh Conference on Magnetic Refrigeration at Room Temperature have been
dedicated to the memory of
Karl Gschneidner Jr. (November 16, 1930 - April 27, 2016)
Karl Albert Gschneidner, Jr. was born in Detroit on

November 16, 1930 into a family of German immigrants.
He received a B.S. in Chemistry from the University of
Detroit in 1952 and a Ph.D. in Physical Chemistry from
Iowa State University in 1957. After a six-year stint as a
staff member and then a section leader at Los Alamos
Scientific Laboratory of the University of California
(now Los Alamos National Laboratory), Karl returned to
Iowa State University as an Associate Professor of
Metallurgy in 1963. He also became a Group Leader at
the Ames Laboratory. Ames, Iowa State University and
Ames Laboratory have remained his home for the next
53 years.
Rare earth science and technology were Karl’s life-long
passion. His seminal contributions to rare earth alloy
theory, creation and directorship of the renowned Rare
Earth Information Center, and his editorship of the well-
respected Handbook on the Physics and Chemistry of
Rare Earths are broadly recognized all over the world.
Karl became interested in the magnetocaloric effect in
the early 1990s when he developed a series of (Dy1‑xErx)
Al2 alloys to replace prohibitively expensive GdPd – then the only material available to liquefy
hydrogen using a magnetic refrigerator. Two of his greatest successes – the discovery of the giant
magnetocaloric effect in Gd5Si2Ge2, and the demonstration of a functional near-room temperature
magnetic cooler – turned out to be quite influential. Announced in 1997, not only these
breakthroughs crystallized numerous magnetocaloric research clusters worldwide, but they soon
lead to the establishment of the Thermag conference series in 2005. Realizing the importance of the
series, Karl co-chaired the third Thermag held in Des Moines, Iowa in 2009. Karl’s legacy lives on,
with Ames Laboratory now leading CaloriCoolTM – the Caloric Materials Consortium – a member
of the Energy Materials Network launched by the U.S. Department of Energy in 2016.
Karl has been a distinguished member of the Thermag community since the first (Montreux,
Switzerland) conference. He could not attend Thermag VI in Victoria – his illness did not allow him
to travel – but he was genuinely interested, asking to say “Hello” to everyone he knew and was
undeniably delighted to hear the same brought back to him. Regrettably, Karl did not have a chance
to do so this time, but if he did, he would as always have sent his warmest regards to everyone, and
would be wishing all of us yet another great – Thermag VII in Turin.
Vitalij Pecharsky,
August 2016
vi
TABLE OF CONTENTS
043 :: Substitutional and interstitial modulation of La-Fe-Si structure discussed with hydrogen sorption ki-
netics 1
X. Hai, C. Mayer, V. Nassif, F. Porcher, S. Miraglia
057 :: Multiobjective optimization of a reciprocating magnetic refrigerator using a genetic algorithm

7
S. Roy, S. Poncet, M. Sorin
059 :: A test stand to measure the magnetization work of magnetocaloric materials

11
L. Ferreira, C. Bessa, S. Gama, O. Horikawa
061 :: Thermal single-ionic magnetic properties of materials calculated with ATOMIC MATTERS MFA
computation system 15
R. Michalski, J. Zygadlo
070 :: Magnetocaloric bench: analytical model for gadolinium characteristics

20
A. Plait, S. Giurgea, C. Espanet, T. de Larochelambert, P. Nika
072 :: Performance comparison of La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8 magnetic refrigerant prepared by two

methods 24
J. Cheng, J. Huang, C. Liu, Y. Zhang, C. Zhang, P. Jin, C. Li, Z. Li
074 :: Scale up of MnFePSi materials in a broad temperature range and manufacturing of advanced heat
exchanger structures
28
F. Dötz, D. van Asten, D. Barrera-Medrano, C. Carroll, U. Kalck, S. Misra, B. Reesink, O. Rogge, F. Scharf, M.
Schwind, F. Seeler, M. Zeilinger, L. Zhang
075 :: Properties of La(Fe, Si)13 magnetocaloric compounds prepared by reactive Spark Plasma Sintering
31
V. Paul-Boncour, A. Patissier, K. Nakouri, L. Bessais
076 :: Performance analysis of an active magnetic refrigeration system with external loss mechanisms
35
I. Niknia, P. Trevizoli, P. Govindappa, O. Campbell, T. Christiaanse, R. Teyber, A. Rowe
077 :: A material screening technique for optimum performance of an AMR

39
I. Niknia, P. Trevizoli, P. Govindappa, O. Campbell, T. Christiaanse, R. Teyber, A. Rowe
078 :: Effects of magnetocaloric wire on increase in magnetic refrigeration cycle

43
K. Ueno, M. Kondo, K. Takeuchi, R. Nomura, T. Kizaki
079 :: New concept for magnetocaloric heat pumps based on thermal diodes and latent heat transfer
47
K. Bartholomé, T. Hess, A. Mahlke, J. König
084 :: Study of the performance coefficient of an active magnetic regenerative refrigeration system
51
Z. Meddeb, D. Bchiri, R. Ben Sidhom, M. Chrigui, A. Jeday
086 :: Study of heat transfer in magnetocaloric regenerator with oscillating flow

55
A. Meunier, J. Roy, R. Glises, Y. Bailly, T. de Larochelambert, P. Nika
vii
088 :: The operating principle of a fully solid state active magnetic regenerator
59
M. Zhang, A. Mehdizadeh Momen, O. Abdelaziz
090 :: Design and research of the room temperature magnetic wine cabinet
63
C. Zhang, J. Huang, Y. Hongwei, P. Jin, J. Cheng, C. Liu, Z. Li, Y. Zhang
096 :: Experimental evaluation of a single layer MnFeP1-xSix AMR

67
P. Govindappa, P. Trevizoli, I. Niknia, O. Campbell, T. Christiaanse, R. Teyber, A. Rowe
098 :: Transient simulation of the Active Magnetic Regenerator process

71
T. Christiaanse, P. Govindappa, I. Niknia, R. Teyber, P. Trevizoli, A. Rowe
099 :: Location dependent performance study on regenerators within a Halbach magnetic field source
75
T. Christiaanse, P. Govindappa, I. Niknia, R. Teyber, P. Trevizoli, A. Rowe
100 :: Effect of PET insulator layers on the conduction losses of plate Active Magnetic Regenerators
79
I. Mugica, S. Poncet, J. Bouchard
102 :: Influence of void volume and inlet flow maldistribution on the performance of thermal regenerators
83
P. Trevizoli, M. S. Capovilla, G. Peixer, A. Nakashima, J. Lozano, J. Barbosa
104 :: Modeling and simulation of a thermomagnetic reciprocating engine by gravity

A. Cocci de Souza, E. Evaristo, F. Clareth Colman, W. A. Dos Santos Conceição, G. L. Foleiss, W. Szpak, A. Car- 87
valho, C. Santiago Alves
105 :: Investigation of an active magnetic regenerator subject to different fluid flow waveforms
91
R. Teyber, P. Trevizoli, T. Christiaanse, P. Govindappa, I. Niknia, A. Rowe
106 :: Experimental evaluation of two-material active magnetic regenerators

95
R. Teyber, P. Trevizoli, T. Christiaanse, P. Govindappa, I. Niknia, A. Rowe
109 :: Magnetocaloric effect in severe plastic deformed Gd‐X (X=In, Ga, B, Y, Zr)
99
S. Taskaev, V. Khovaylo, K. Skokov, A. Pellenen, D. Karpenkov, M. Ulyanov, O. Gutfleisch
110 :: Generating the optimal magnetic field for magnetic refrigeration

103
R. Bjørk, A. R. Insinga, A. Smith, C. Bahl
112 :: Thermal loss estimation for active magnetic regenerator aiming at railway air-conditioner application
107
Y. Miyazaki, K. Ikeda, K. Waki, T. Kawanami
125 :: Analytical solutions of the magnetic field generated by two-pole nested Halbach cylinders
111
F. Fortkamp, J. Lozano, J. Barbosa
126 :: Optimization of Active Magnetic Regenerators with two and three layers of Gd and Gd-alloys
115
J. Cararo, J. Lozano, P. Trevizoli, R. Teyber, A. Rowe, J. Barbosa
127 :: Influence of particle size on the magnetocaloric properties of Mn1.30Fe0.65P0.5Si0.5 powders

119
A. Bartok, M. Kuepferling, C. Curcio, V. Basso, A. Pasko, K. Zehani, L. Bessais, F. Mazaleyrat, M. Lo Bue
viii
129 :: Influence of the heat exchangers thermal conductance on the thermodynamic performance of a mag- 123
netic refrigerator
R. Schneider Calomeno, J. Lozano, P. Trevizoli, J. Barbosa
131 :: Modified mean-field theory with phenomenological model of volume effects for 1st order transition
magnetocaloric materials 127
M. Risser, V. Hardy, S. Lionte, C. Muller
132 :: The optimization of the energy performances of a PMRR by using neural networks
131
C. Aprea, A. Greco, A. Maiorino, C. Masselli
134 :: High-temperature hydrogenation behaviors and magnetocaloric properties of LaFe11.6Si1.4 compound

138
Y. Tang, Y. Chen, H. Zheng, J. Wu, H. Wang
141 :: Improvement of the AMR's performance: multi-segment regenerator and by-pass flow
142
S. Lionte, M. Risser, C. Vasile, L. Elouad, C. Muller
142 :: A new magnetic refrigeration prototype with application in household and professional refrigerators
146
B. Monfared, B. Palm
143 :: Simulation of magnetic refrigeration systems with thermal diodes and axial conductive heat transfer
150
B. Monfared
145 :: Optimization of heat exchange in magnetic refrigerators and thermomagnetic motors

154
C. Bessa, L. Ferreira, S. Gama, T. Santos, J. Amaral, V. Costa, V. Amaral, O. Horikawa
147 :: Epoxy-bonded La(Fe,Mn,Si)13Hz as a multi layered active magnetic regenerator

158
H. Bez, K. Navickaite, T. Lei, K. Engelbrecht, A. Barcza, C. Bahl
149 :: Design and fabricating of a two-in-one compact magnet array

164
Q. Jiang, H. Li, X. Xu, J. Huang, D. Lu
150 :: A new concept refrigeration model based on thermal switch and numerical calculation
168
H. Li, Q. Jiang, X. Xu, D. Lu
151 :: Impact of the integration work on the performance measurements of the magnetocaloric cooling system
and its target refrigerated cabinet 172
M. Hittinger, J. Chaudron, L. Elouad, C. Muller, P. Haegel, C. Vasile
157 :: Theoretical description of the magnetocaloric effect in Mn-Fe-P-Si alloys

178
M. Piazzi, C. Bennati, C. Curcio, M. Kuepferling, V. Basso
158 :: Modeling of a novel digital hydraulic system for a magnetic refrigerator

182
P. Cardoso, M. Destro, M. Alvares, J. Lozano, V. De Negri, J. Barbosa
160 :: Thermal energy harvesting: thermomagnetic versus thermoelectric generator

186
M. Almanza, A. Pasko, A. Bartok, F. Mazaleyrat, M. Lo Bue
161 :: Dynamic temperature measurement for magnetocaloric materials

190
M. Almanza, W. Cherief, A. Kedous-Lebouc
ix
163 :: Peculiarities of the magnetocaloric effect in the vicinity of the ferromagnetic-antiferromagnetic transi-
tion in an iron-rhodium alloy 194
R. Gimaev, V. Zverev, A. Saletsky, A. Tishin, T. Miyanaga, J. Staunton
173 :: Electrocaloric polymers: different electric field cycles to increase their efficiency
198
S. Pruvost, V. Basso, L. Garbuio, A. Kedous-Lebouc, M. Almanza, J. Gérard
174 :: Theoretical modeling of heat transfer processes in Ni–Co–Mn–In magnetic wires

202
O. Pavlukhina, V. Sokolovskiy, V. Buchelnikov
175 :: Direct in situ study of magnetocaloric effect and martensitic twins structure of Ni-Mn-Ga Heusler alloy
in strong magnetic fields up to 14 T 206
Y. Koshkidko, E. Dilmieva, J. Cwik, A. Kamantsev, V. Khovaylo, V. Koledov, V. Kraposhin, A. Mashirov, V. Shavrov
176 :: Magnetocaloric properties of the stacked Ni50Mn18.75-xCu6.25+xGa25 (x = 0÷1) polycrystalline alloys

210
R. Wroblewski, K. Sielicki, M. Leonowicz
177 :: Properties of Gadolinium wire for magnetic refrigeration

214
R. Nomura, T. Kizaki, K. Ueno, M. Kondo, K. Takeuchi
178 :: On the rotating magnetocaloric effect in multiferroic RMn2O5 compounds

217
M. Balli, S. Mansouri, B. Roberge, S. Jandl, P. Fournier, D. Dimitrov
179 :: Active Caloric Regenerator Cycles: an analytic element model

222
A. Rowe, T. Christiaanse, P. Govindappa, I. Niknia, R. Teyber, P. Trevizoli
180 :: Magnetocaloric properties of Fe-Rh-(Z) (Z =Pd, Ni) alloys from ab initio and Monte Carlo calculations
228
V. Sokolovskiy, O. Pavlukhina, V. Buchelnikov, P. Entel
182 :: Fundamental study about an air conditioner which combined the magnetic refrigeration with humidity
regulation system 232
N. Hirano, K. Tanaka, M. Takahashi, R. Nagamine, T. Okamura
183 :: Features of magnetocaloric effect in rare-earth based R(Co-Fe)2 Laves phases, with R = Ho, Er
236
M. Anikin, E. Tarasov, N. Kudrevatykh, M. Semkin, A. Volegov, A. Inishev, A. Zinin
184 :: Experimental performance evaluation of sintered Gd spheres packed beds

240
A. Tura, K. Nielsen, N. Van Nong, N. Pryds, P. Trevizoli, T. Christiaanse, R. Teyber, A. Rowe
191 :: Integral optimisation of a magnetocaloric air-conditioning system for an electric vehicle

244
B. Torregrosa-Jaime, J. Payá, J. M. Corberán
194 :: A cascading model of an active magnetic regenerator system

248
M. Tahavori, K. Filonenko, C. T. Veje, T. Lei, K. Engelbrecht, C. Bahl
200 :: Giant magnetocaloric effect of compressible Ising and Heisenberg lattices

252
J. Amaral, N. Fortunato, C. Amorim, J. Gonçalves, V. Amaral
205 :: Reversible and non reversible magnetocaloric effect in phase separated manganites
258
M. Quintero, D. García Reparaz, D. Goijman, I. Irurzun, S. Passanante, A. G. Leyva, J. Sacanell
x
207 :: Investigation of magnetocaloric effect by the fast response infrared optical temperature probe
A. Kamantsev, V. Koledov, A. Mashirov, V. Shavrov, L. Butvina, I. S. Tereshina, A. S. Los, A. Gilewski, N. H. Yen, P. 263
T. Thanh, N. H. Dan
208 :: Magnetocaloric effect supression induced by mechanical milling on intermetallic DyCo2

V. G. de Paula, M. Gomes da Silva, L. Moreira da Silva, A. Oliveira dos Santos, R. Lang, L. Otubo, A. Coelho, L. 267
Cardoso
220 :: CFD Simulations of the magnetocaloric heat transfers between a gadolinium plate and an oscillating
liquid flow 271
J. Roy, T. de Larochelambert, P. Nika, Y. Bailly, R. Glises, A. Meunier
222 :: Ti substitution in MnBi rare-earth free magnetocaloric compound

275
C. Curcio, E. Olivetti, L. Martino, M. Kuepferling, A. Bartok, M. Lo Bue, V. Basso
227 :: Study on composite of thermoplastic loaded by magnetocaloric powders

279
J. Lanzarini, J. Gelin, T. Barriere, M. Sahli, A. Dubrez
243 :: Multidimensional numerical method for performance predictions of an AMR cycle using Ansys Fluent
software 283
K. Hamdani, A. Smaili
247 :: Simulation of a hydraulic circuit for a magnetic refrigerator

287
T. Ebel, J. Lozano, P. Cardoso, J. Barbosa
250 :: Future developments in magnetocaloric refrigeration and heat pumping

291
A. Kitanovski, U. Tomc, A. Poredoš
257 :: Room-temperature magnetic refrigeration: from basic research to development for application
297
A. T. Saito, H. Nakagome
258 :: Rotary magnetic regenerator design and assembly

303
F. Scarpa, L. A. Tagliafico, M. Gigante
AUTHOR INDEX 307
xi
SUBSTITUTIONAL AND INTERSTITIAL MODULATION OF La-Fe-Si
STRUCTURE DISCUSSED WITH HYDROGEN SORPTION KINETICS
X. Hai(a)(b)*, C. Mayer(b), V. Nassif(a), F. Porcher(c), S. Miraglia(a)

(a)
Institut Néel, Univ. Grenoble Alpes, 25 Avenue des Martyrs, 38042, Grenoble, France
(b)
Erasteel SAS, Tour Montparnasse, 33 Avenue du Maine, 75755, Paris, France
(c)
Laboratoire Léon Brillouin, UMR12 CEA-CNRS, Bât. 563 CEA Saclay, 91191, Gif sur Yvette, France
*Corresponding author. E-mail: xueying.hai@neel.cnrs.fr
ABSTRACT
Magnetocaloric materials (La,Ce)(Fe,Mn,Si)13Hx have great potential as commercializable candidate for

room-temperature magnetic refrigeration technology. It is important to have a better understanding of their
hydrogen stability for the industrial upscaling. The hydrogenation reaction is investigated by powder neutron
diffraction in order to have insights on the interaction between hydrogen and the host lattice. The reaction
kinetics of absorption and desorption of interstitial elements can be influenced by presence of other
interstitial and substitutional atoms. We demonstrate with neutron diffraction data the effects of C and Ce
atoms on the deuterium absorption and desorption kinetics. The influence of substitutions and interstitial
modification to the crystal structure are discussed at the level of interatomic Fe-Fe bond variations.
Keywords: magnetocaloric materials, hydrogen sorption kinetics, neutron diffraction

DOI: 10.18462/iir.thermag.2016.0043
1. INTRODUCTION
Magnetocaloric materials in the family of La(Fe1-xSix)13 exhibit tunable ordering temperature and
magnetocaloric performance from composition modification [1-5]. Currently the most effective method of
tuning 𝑇𝐶 to room temperature is the insertion of interstitial hydrogen (H) because of the preservation of
favorable magnetocaloric effect [6]. However, interstitial H shows unstable characteristics that manifest in
two ways. Fully hydrogenated compounds desorb H with time from even below 150°C [7], which can be
problematic for shaping processes. Partially hydrogenated material that exhibits strong first-order magnetic
transition characteristics shows age splitting of 𝑇𝐶 peak [8, 9], which is associated with the phase segregation
and coexistence of the paramagnetic and ferromagnetic phases [10]. Therefore, it is important to investigate
the H interstitial behavior inside the host lattice. Recently, we have demonstrated that a modest addition of
carbon (C) interstitial atoms in the La(Fe1-xSix)13 prior to hydrogenation can slow down the H insertion
kinetics and therefore, offers a prospect to have improved stability of hydrogenated materials in long-term
application [11]. In this paper, we present recent developments in H sorption kinetics, which will be
discussed with influences from both C interstitial insertion and Cerium (Ce) substitution.
2. MAIN SECTION
2.1 Experimental Conditions

The samples are elaborated by high-frequency induction melting appropriate amounts of starting materials.
The ingots are homogenized in a sealed quartz tube with partial pressure of Ar and Mo-foil as protection at
1373 K for 15 days. The ingots are mechanically grinded and sieved to irregularly-shaped powder between
40 and 200 μm in size for the neutron diffraction experiments. Deuterium (D) gas is used instead of H for
better signal-to-noise ratio in neutron. Details of the in-situ neutron diffraction experimental conditions are
described elsewhere [11]. High-resolution neutron diffraction patterns are recorded at the 3T2 diffractometer
at Laboratoire Léon Brillouin for annealed and hydrogenated samples. The incident neutron wavelength is
1.225 Å. The diffractograms are collected at room temperature from 4.50° to 115.35° at 0.05° step.
Seventh IIF-IIR International Conference on Magnetic Refrigeration at Room Temperature, Thermag VII
Torino, Italy, 11-14 September 2016
1
2.2 Effect of Partial Substitution of Ce to La
The partial substitution of Ce to La results in a lowering of the Curie temperature 𝑇𝐶 with decreasing lattice
constant. The change in 𝑇𝐶 is mainly attributed to the magnetovolume effect caused by the small radius of Ce
atoms [4]. With 0.3 Ce substitution, 𝑇𝐶 shifts from 206 K to 188 K under a magnetic field of 0.05 T. The
thermal hysteresis increases from 2 to 8 K [12]. Samples of nominal composition La0.7Ce0.3Fe11.44Si1.56 and
La0.7Ce0.3Fe11.44Si1.56C0.2 are studied at the D1B diffractometer. Neutron diffraction gives us access to
information at various scales. On the mesoscopic scale, the unit cell expansion and contraction is recorded
in-situ. At the atomic level, we are able to observe significant changes in the Fe-Fe distances (Section 2.3). A
full occupancy of the 24d site would correspond to the formula LaFe13-xSixH3, which is not attainable
practically because of H-H repulsive interactions. In the following analysis, H and D saturation refers to the
maximum amount of H or D interstitial atoms absorbed at given temperature and pressure conditions.
(a) (b)
11,57 11,57
11,56 11,56
Absorption
Absorption Desorption
11,55 11,55
Desorption
a (Angstr.)
a (Angstr.)
La0.7Ce0.3Fe11.44Si1.56
11,54 11,54
T = 540 K
11,53 11,53
LaFe11.44Si1.56
11,52 T = 540 K 11,52
11,51 11,51
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
Time (min) Time (min)
Figure 1. Refined unit cell parameter a during absorption and desorption for samples of composition LaFe11.44Si1.56
and La0.7Ce0.3Fe11.44Si1.56 vs. time at 540 K.
(a) 1,0
(b) 1,0
T = 540 K
0,8 0,8
LaFe11.44Si1.56
0,6 0,6
T = 540 K La0.7Ce0.3Fe11.44Si1.56
D/u.f.
D/u.f.
Absorption
0,4 0,4
Desorption
0,2 Absorption 0,2

Desorption
0,0 0,0
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200 1400
Figure 2. Refined concentration of D interstitials during absorption and desorption for samples of composition
LaFe11.44Si1.56 and La0.7Ce0.3Fe11.44Si1.56 vs. time at 540 K.
Ce substitution for La results in a reduced volume of the octahedral interstitial site (24𝑑) due to steric effect.
The progressions of unit cell parameter and the 24𝑑 site occupancy are tracked in the course of D insertion
and release at 540 K (Fig. 1 and Fig. 2). At the end of the available absorption time, the specimen has taken
in about 0.8 D/u.f. in Ce-substituted sample, slightly less than the Ce-free reference (1 D/u.f.). Considering
the kinetics, there is a clear delay for D absorption at 540 K in the case of Ce-substituted sample. It takes
specimen La0.7Ce0.3Fe11.44Si1.56 twice as long to finish the initial climb of the unit cell expansion. The effect is
even more evident by examining the D content per unit formula deduced from the 24𝑑 site occupancy. While
the unit cell parameter increases continuously to make room for interstitial atoms, the sequential filling of the
24𝑑 site is not continuous, rather step-like for La0.7Ce0.3Fe11.44Si1.56, as illustrated in Fig. 2. The evolution of
the two parameters is not as correlated as one could expect from a simple model. We observe several
changes of the slope for the progression of the unit cell parameter for La 0.7Ce0.3Fe11.44Si1.56, shown in Fig. 1
(b).
2
Based on the Kolmogorov-Johnson-Mehl-Avrami (KJMA) relation (Eq.1) for solid phase reaction, the
amount of absorption reaction occurred, 𝐹, is related to the reaction time, 𝑡, and the deuterium sorption
kinetics can be evaluated in terms of the reaction constant 𝑘.
F 1e(kt )
n
(1)
This disparity between the unit cell parameter and the 24𝑑 site occupancy is confirmed by modeling with
KJMA relation. Unlike samples LaFe11.44Si1.56 and LaFe11.44Si1.56C0.2, the Avrami exponent 𝑛 calculated from
the linearization of ln(− ln(1 − 𝐹)) = ln(𝑘) + 𝑛 ∙ 𝑙𝑛(𝑡) is 1.2 for the kinetics of unit cell expansion and 0.5
for that of interstitial filling. This underlines very different mechanisms for the evolution of these two
parameters in reaction. Upon desorption, the releasing of the deuterium atoms is almost instantaneous for Ce-
substituted specimen, showing a much faster desorption kinetics than that for compound LaFe 11.44Si1.56. For
all the samples measured, desorption reaction is much faster than absorption. This can be considered as a
consequence of decrepitation of the material due to D massive absorption during heat treatments, as we
observe that the effective size of the samples are smaller during desorption due to crack formation. The
mechanical strain from D absorption can partially contribute to the accelerated desorption kinetics in Ce-
substituted sample as the lattice has a smaller volume due to Ce substitution. But more importantly, Ce-
substitution modifies the D diffusion path by a selective contraction of certain interatomic distances (Section
2.3). Desorption is nearly complete for sample La0.7Ce0.3Fe11.44Si1.56, as shown by the emptying of the
interstitial site in Fig. 4. This suggests the important role of diffusion path on the facility to release interstitial
atoms.
2.3 Effect of C Interstitial Insertion

C interstitial insertion increases 𝑇𝐶 and reduces the thermal hysteresis in the magnetic transitions. In the
measurements under 0.05 T external field, the 𝑇𝐶 increase with 0.2 C inserted is 30 K for samples of
composition LaFe11.44Si1.56 and La0.7Ce0.3Fe11.44Si1.56 [12]. Previously, a noticeable obstructing effect of C on
the D absorption reaction was observed for sample of composition LaFe11.44Si1.56C0.2 compared with its C-
free counterpart [11]. However, C insertion does not modify the H insertion mechanism and the structural
modulation upon H introduction remains the same [11]. As mentioned above, Ce-substituted sample
La0.7Ce0.3Fe11.44Si1.56 shows a step-like evolution of D concentration during absorption. Therefore, it is
interesting to investigate the impact of C on the D sorption dynamics of Ce-substituted sample. Although the
Ce-substituted sample has increased difficulty in D absorption, the incorporation of 0.2 C/u.f. interstitials is
well succeeded through induction melting for sample La0.7Ce0.3Fe11.44Si1.56C0.2. C-doped sample
La0.7Ce0.3Fe11.44Si1.56C0.2 smooths the D insertion reaction, as seen in Fig. 3. D concentration stabilizes after
500 min, with C-doped sample showing slightly faster kinetics between 200 and 400 min of absorption. At
desorption, however, sample La0.7Ce0.3Fe11.44Si1.56 has emptied nearly all the interstitial D from the 24𝑑 site
while La0.7Ce0.3Fe11.44Si1.56C0.2 has retained about half of its D atoms.
(a) 1,0
(b) 1,0
Absorption La0.7Ce0.3Fe11.44Si1.56C0.2
T = 540 K
Desorption T = 540 K
0,8 0,8
0,6 0,6
La0.7Ce0.3Fe11.44Si1.56
D/u.f.
D/u.f.
Absorption
0,4 0,4
Desorption
0,2 0,2
0,0 0,0
0 200 400 600 800 1000 1200 1400 0 200 400 600 800 1000 1200 1400
Figure 3. Effect of C on the D absorption and desorption kinetics demonstrated by the D concentration per unit
formula for sample of composition La0.7Ce0.3Fe11.44Si1.56 and La0.7Ce0.3Fe11.44Si1.56C0.2 at 540 K.
3
The mechanism of light atom insertion into the interstitial site is associated with not only the available space
of accommodation, but also the facilitation of the diffusion path in the lattice. Fig. 4 summarizes variations
of the unit cell parameter due to C insertion and Ce substitution. The delayed kinetics seen in
LaFe11.44Si1.56C0.2 can be considered as a shortage of available space compared to LaFe 11.44Si1.56. If the
available space is the rate-limiting step, this phenomenon should be highlighted in Ce-substituted sample
since interstitial volume is reduced. However, the slightly accelerated absorption kinetics in
La0.7Ce0.3Fe11.44Si1.56C0.2 than in La0.7Ce0.3Fe11.44Si1.56 suggests that it might be more of a changed insertion
mechanism through modified interatomic distances. The arrival of D atoms into the interstitial sites could be
assisted by C insertion as it stretches the lattice volume. At the same time, the diffusion of interstitial atoms
in the lattice is governed by the interatomic distances, which needs to be considered as well. At the
desorption stage, C insertion clearly has a softening effect on the fast unit cell contraction in Ce-substituted
sample.
(a) (b)
11,60 11,60
T = 540 K LaFe11.44Si1.56
T = 540 K
11,58 11,58 LaFe11.44Si1.56C0.2
La0.7Ce0.3Fe11.44Si1.56
a (Angstr.)
a (Angstr.)
11,56 11,56 La0.7Ce0.3Fe11.44Si1.56C0.2
11,54 LaFe11.44Si1.56 11,54

LaFe11.44Si1.56C0.2
11,52 La0.7Ce0.3Fe11.44Si1.56 11,52
La0.7Ce0.3Fe11.44Si1.56C0.2
11,50 11,50
0 200 400 600 800 0 200 400 600 800 1000 1200 1400
Absorption time (min) Desorption time (min)
Figure 4. Evolution of unit cell parameter from “in-situ” neutron diffraction for D absorption at 540 K.
2.4 Discussion
Neutron powder diffraction allows us to look inside the structural breathing as the compound accommodates
interstitial atoms while maintaining the cubic structure. In La(Fe1-xSix)13 materials, the magnetocaloric effect
relies on the magnetic interaction between Fe sites. The Fe-Fe interatomic bonds are consisted of three bonds
that are formed by the Fe atoms in the same cluster and two inter-icosahedra bonds (Fig. 5). Based on the
refined results from hydrogenated samples measured at 3T2, H insertion results in the extension of the basal
plane of the 24d site (B4) and the contraction of the FeI-centered icosahedron (B2). By tracking the
interatomic distances at 540K in the course of absorption and desorption, we are able to confirm this
mechanism for D absorption and the D desorption reaction proceeds with the reversal process.
Figure 5. Structural representation of La(Fe1-xSix)13 phase showing Fe-Fe bonds and interstitial atoms (yellow dots)
on the 24d site.
Ce substitution reduces the lattice volume by reducing notably the B4 and B5 distances, directly affecting the
interstitial 24𝑑 site volume. Fig. 6 shows the evolution of B4, and B2 in the course of D absorption for
sample of composition La0.7Ce0.3Fe11.44Si1.56 and La0.7Ce0.3Fe11.44Si1.56C0.2. D absorption in sample
4
La0.7Ce0.3Fe11.44Si1.56 shows deviated interatomic bond evolution compared to the Ce-free sample. The first
200 min, the structure stretches B2 while reducing B4, which corresponds to an initial lattice rearrangement.
Only very small amount of D atoms are able to enter and the unit cell slowly expands. We refer to this period
as an incubation stage for absorption. Afterwards, the period from 200 to 600 minutes corresponds to an
accelerated absorption stage where D content jumps from 0.4 to 0.8 D/u.f. The mechanism in this stage is
similar to the classical D absorption mechanism observed for samples of composition LaFe11.44Si1.56 and
LaFe11.44Si1.56C0.2 under the same experimental conditions [11].
(a) 2,80 La0.7Ce0.3Fe11.44Si1.56 (b) 2,48 La0.7Ce0.3Fe11.44Si1.56

2,78 La0.7Ce0.3Fe11.44Si1.56C0.2 La0.7Ce0.3Fe11.44Si1.56C0.2
2,46
B2 (Angstr.)
B4 (Angstr.)
2,76
2,44
2,74
2,42
2,72
2,70
B2 2,40
B4
0 200 400 600 800 1000 0 200 400 600 800 1000
Absorption time (min) Absorption time (min)
Figure 6. Effect of C insertion on the D absorption kinetics demonstrated by variation of Fe-Fe distances B2 and B4
for samples of composition La0.7Ce0.3Fe11.44Si1.56 and La0.7Ce0.3Fe11.44Si1.56C0.2.
C insertion results in increased B4 and La-La distances, which constitute the volume of the 24d site. The C-
doped sample demonstrates typical D absorption mechanism by elongating B4 and reducing B2. The rate-
limiting step is the D atomic diffusion within the lattice and C insertion modifies the space within the
diffusion path by modifying the interatomic distances. C addition in Ce-substituted sample, therefore, works
as a springboard, which smoothes the lattice breathing that is restrained by Ce, as illustrated in the Fe-Fe
distance variation for sample La0.7Ce0.3Fe11.44Si1.56C0.2 in Fig. 6. The interstitial and substitutional
modulations on the structure provoke very interesting dynamics at the level of interstitial diffusion path,
which leads to further modification of the mechanism of D sorption.
Besides the octahedral 24𝑑 site, there exists two smaller tetrahedral sites that are likely to accommodate H or
D interstitials, 192𝑗 (0.216, 0.099, 0.14) and 64𝑔 (0.14, 0.14, 0.14). With the high-resolution neutron data
obtained at for La(Fe1-xSix)13 and La(Fe1-xSix)13Hmax samples, these secondary interstitial sites have been
tested for compounds saturated with hydrogen. However, adding secondary tetrahedral sites for H
accommodation leads to either unreasonable total H concentration or unacceptable distances between H
atoms. The small quantity of H atoms and their dispersive nature in the NaZn13-type cubic structure make it
quite difficult to distinguish secondary interstitial sites that might provide available volume. Even with high-
resolution powder neutron diffraction, one cannot conclude with certainty on the existences of secondary
interstitial sites.
Differential Scanning Calorimetry (DSC) can be used to measure the onset of H desorption, which
corresponds to an endothermic peak in the heat flow, as shown in Fig. 7. Fully hydrogenated pairs of samples
of composition La0.7Ce0.3Fe11.7Si1.3 and La0.7Ce0.3Fe11.7Si1.3C0.2 are measured in order to confirm the effect of
C on H desorption of Ce-substituted La-Fe-Si phase. The TC of the samples is controlled before and after
thermal desorption. Before hydrogenation, annealed sample of composition La0.7Ce0.3Fe11.7Si1.3 has a lower
TC (188 K) than La0.7Ce0.3Fe11.7Si1.3C0.2 (212 K). The difference in 𝑇𝐶 due to 0.2 C/u.f. inserted is about 24 K.
Fully hydrogenated samples of composition La0.7Ce0.3Fe11.7Si1.3 has a higher TC (328 K) than
La0.7Ce0.3Fe11.7Si1.3C0.2 (316 K). The insertion of carbon lowers slightly the maximum hydrogen absorption
capacity. Moreover, the difference in 𝑇𝐶 after H saturation is reduced to only 12 K. Interstitial modulations
by H only and by C and H combined lead to similar levels of 𝑇𝐶 after maximum H absorption. This suggests
a similar manner in how the Fe-Fe distances are modified for these two modulations, which is then
confirmed from the neutron diffraction data (3T2). The dehydrogenated TC corresponds to the TC of annealed
samples and confirms the total desorption of H atoms in the material. The onset of H desorption is shifted
from 521 K (248 °C) to 582 K (309 °C) for sample La0.7Ce0.3Fe11.7Si1.3 and La0.7Ce0.3Fe11.7Si1.3C0.2,
5
respectively. The stability of H atoms at high temperature is improved for fully hydrogenated
La0.7Ce0.3Fe11.7Si1.3C0.2.
Figure 7. DSC measurement showing the onset of H desorption temperature for sample of composition
La0.7Ce0.3Fe11.7Si1.3 and La0.7Ce0.3Fe11.7Si1.3C0.2.
3. CONCLUSIONS
We demonstrate with neutron diffraction data that the interstitial insertion is impeded by Ce partial
substitution. The analysis of the interatomic distances confirms that Ce substitution has significant
modification on the D absorption mechanism. Ce has two noticeable effects on the D sorption kinetics.
Smaller interstitial volume makes it difficult for the lattice to expand, therefore leading to a very slow
absorption reaction. However, the desorption reaction occurs almost instantaneously and Ce seems to
prohibit the retaining of deuterium atoms. A modest introduction of C atoms to the interstitial sites, on the
other hand, does not change the sorption mechanism but modifies the kinetics. By introducing C dopants in
Ce-substituted sample, the agitated lattice modulation is softened and the classical hydrogenation mechanism
is restored. Moreover, C addition in Ce-substituted sample can help retain 50% of the absorbed D during
desorption at given experimental conditions.
ACKNOWLEDGMENTS
This work is supported by project CIFRE n°2013/0827. The authors appreciate the French CRG project for
assistance in the neutron diffraction facility and experiment.
REFERENCES
[1] F. Wang et al., J. Phys. D.: Appl. Phys., 36, 1 (2003).

[2] X. Liu et al., J. Magn. Magn. Mater., 270, 305 (2004).
[3] M. Balli et al., J. Phys.: Cond. Matter, 19, 236230 (2007).
[4] S. Fujieda et al., J. Magn. Magn. Mater., 321, 3567 (2009).
[5] M. Krautz et al., J. Alloys Compd., 598, 27 (2014).
[6] Y. Chen et al., J. Phys.: Condens. Matter, 15, L161 (2003).
[7] C. Mayer et al., Phys. Status Solidi C, 11, 5 (2014).
[8] A. Barcza et al., IEEE Trans. on Magn., 47, 3391 (2011).
[9] C. B. Zimm and S. A. Jacobs, J. Appl. Phys., 113, 17A908 (2013).
[10] M. Krautz et al., J. Appl. Phys., 112, 083918 (2012).
[11] X. Hai et al., J. Magn. Magn. Mater., 400, 344 (2016).
[12] X. Hai et al., to be published
6
MULTIBOJECTIVE OPTIMIZATION OF A RECIPROCATING
MAGNETIC REFRIGERATOR USING A GENETIC ALGORITHM
S. Roy*, S. Poncet, M. Sorin

Department of Mechanical Engineering, Université de Sherbrooke (Sherbrooke, Canada)
*
Corresponding author. E-mail: Steven.Roy@usherbrooke.ca
ABSTRACT
This paper is intended to analyze the overall energetic and exergetic performance of an active magnetic
regenerative refrigerator (AMRR). A 1.5-D numerical model has been developed to simulate a system
composed of a parallel-plate regenerator, a magnetic source, a pump, heat exchangers, and control valves.
The behaviour in steady state is described and the basic thermodynamic equations are depicted. The effects
of several parameters like the magnetic field, the geometry, the mass, and the blowing time are studied.
Various types of liquid refrigerants are also tested thanks to CoolProp library. The main goals are to
maximize the COP, the exergetic efficiency, and the cooling power while respecting a precise set of
constraints. A multiobjective optimization based on a genetic algorithm (GA) is used to enhance the
performance regarding the Pareto efficiency. Finally, the tuning of the algorithm is discussed in order to
achieve optimal results in fair computational time.
Keywords: Multiobjective Optimization, Exergy, Efficiency, Genetic Algorithm.

DOI: 10.18462/iir.thermag.2016.0057
1. INTRODUCTION
Magnetic refrigeration is one of the most efficient and clean technologies for cold production. The operating
principle is based on magnetocaloric effect (MCE) which is related to the change of entropy that occurs from
the variation of a magnetic field. This generates a quasi-instantaneous temperature change typically about
2 K/T for gadolinium at 293 K [1]. With a refrigerant that acts as a heat transfer medium, a thermodynamic
cycle can be established to produce cold.
Nowadays, the scope of research is mostly focused on three axes: magnetocaloric materials (MCM),
magnetic fields, and cooling systems. This paper is aimed at studying the performance of the AMRR at room
temperature for the sake of improving cost and efficiency. A 1.5-D numerical model has been developed in
Matlab to simulate the behaviour of a reciprocating device that uses a parallel-plate regenerator. Compare to
others [2], this model takes into account convection, conduction, pressure drop due to friction, and
temperature dependant properties. The MCE is tackled from interpolation with experimental data given by
Dankov et al. [3]. The losses coming from the mechanical components, the Foucault currents and the
hysteresis are integrated with coefficients that arise from Kitanovski and Egolf [4]. Adiabatic conditions are
assumed along with equally distributed magnetic field and laminar flow. The validation has been done with
Roudaut’s data to confirm the order of magnitude [5]. The operating conditions and the geometry are also
motivated by many other prototypes that have been constructed over the years [6].
So far, most of the one-on-one variable relationships have been studied in-depth but their combined effects
are still unforeseeable. The real challenge is now about optimizing the entire system performance. When
several specific criteria need to be met, a multiobjective optimization (MOO) is required. In this case, the
choice of the computational algorithm is very important to find the best solution in the shortest possible time.
For this scope, few classical numerical approaches have already been compared with respect to their
effectiveness [7]. However, metaheuristic approaches like genetic algorithms (GA) have never been fully
explored yet. Accordingly, the results of this investigation show the effectiveness of such optimization
technic.
2. MODEL DESCRIPTION
2.1 Numerical modelling

The simplified structure of the studied device is pictured at figure 1 next to its schematic representation. This
system is made from two heat exchangers at each end, a source of magnetic field that can be removed with
the help of an actuator, a pump that drives the refrigerant and a regenerator which consists of parallel plates
7
of magnetocaloric material. The valves allow the fluid inside of the heat exchangers to flow only when the
temperature gradient is sufficient. Following the Brayton cycle, this system is emulated in four steps:
magnetization, cold blow, demagnetization and hot blow. Heat is absorbed from the cold source (CS) and
rejected to the hot source (HS) thanks to the work (W) performed by the pump and the actuator. The mass
flow circulating through the heat exchangers is automatically adjusted to fit the heat flow. Since the magnetic
field is applied in the positive z-direction, the demagnetizing field is unheeded.
Figure 1. Simplified representation of the AMRR

From a numerical point of view, the regenerator is converted to a single plate of equivalent thickness. Then,
the system is discretized in multiple nodes along the x-axis. The equations that govern the temperature
distribution in the fluid and solid parts during blows are obtained from the energy conservation law applied
to a differential element inside the regenerator [8]. These equations are solved simultaneously using first
order finite differences with an explicit scheme. The spatial and time discretization are thoroughly regulated
to ensure stability and convergence. Dirichlet and Neumann boundary conditions are used depending on the
fluid direction. Isothermal conditions are used for initial state. During the blows, the temperatures of the
reservoirs are calculated considering a perfect mixing and the displacement of the entire fluid contained
within. If it reaches the trigger temperature (TCHEX or THHEX), the valve opens, keeping the reservoir at a
constant temperature and allowing the heat to pass through the heat exchanger. This is expressed by the
following set of equations.
 1 blow
t
T f @ x = L ( t ) dt
tblow ∫0
Thot < THHEX : Thot =
Cold Blow  (1)

Thot ≥ THHEX : Thot =
THHEX
 1 blow
t
T f @ x =0 ( t ) dt
tblow ∫0
Tcold > TCHEX : Tcold =
Hot Blow 
T ≤ T (2)
 cold CHEX : Tcold =
TCHEX
2.2 Thermodynamic review

In steady state, the temperature of the sources is determined from the first law of thermodynamics.
Q abs
TCS= T0 −  p
mc (3)
Q rej
THS= T0 +  p
(4)
mc
where T0 is the room temperature, m the mass flow and cp the heat capacity of the fluid inside the heat
exchanger. The Carnot efficiency is used to calculate the exergy (E) of heat flows, as defined below. From
that, one can determine the exergetic efficiency as shown in equation 7.
T 
E abs Q=
= absθ abs Q abs  0 − 1 (5)
 CS
T 
 T 
E rej Q =
= rejθ rej Q rej 1 − 0  (6)
 THS 
E abs + E rej
ηex = (7)
W
8
2.3 Optimization
MOO involves more than one objective function (f) to be optimized. Those functions can be quite subjective
depending on the desired goals. Hence, there is no unique solution to be found but rather a combination of
parameters that will lead to an acceptable result. One way to quantify the outcome is the Pareto efficiency.
The best solution comes from the state where no objective functions could be improved without degrading
any others. In this paper, the emphasis is laid on three key thermodynamic objective functions that need to be
maximized: COP, exergetic efficiency, and cooling power. Any solution that respects all constraints is
accepted and considered just as good (see the problem statement at equation 8). The optimization is also
bounded by a set of constraints which define the allowable range of the variables. That is the intensity of the
magnetic field (B), the blowing time (tblow), and the mass of magnetocaloric material (m). All other variables
are kept fixed throughout the simulation.
 f1 = COP 10 1.0 ≤ B [T] ≤ 2.5
 
f =
 f2 = ηex > 40 % and 0.75 ≤ tblow [s] ≤ 2 (8)
  20 W 0.1 ≤ m [kg] ≤ 0.3
 f3 = Qabs 
3. RESULTS
Table 1 reports the default parameters used for the simulation. Trigger temperatures are set apart from the
ambient temperature considering the minimum allowable gradient in heat exchanger (ΔTHX). The regenerator
is discretized into 22 nodes and the maximum CFL is fixed to 0.30. This proved to be a good compromise
between speed and accuracy for multiple configurations. Steady-state is reached when the maximum
discrepancy on work for two consecutive cycles is less than 0.01 J.
Table 1. Parameters used for the simulation
Material T0 B Flow tblow tmag Length ΔTHX
0.2 kg Gd 293 K 1.5 T 10 x10-6 m3/s 1s 0.01 s 0.1 m ±1 K
Figure 2 shows the cooling power and the exergetic efficiency as a function of magnetic field and mass of
MCM for a specific blowing time. The grid size is 11x11x11 and it takes about 684 min to map all possible
combinations. Several trends can be observed from those graphs. First, increasing the magnetic field leads to
a greater amount of heat absorbed and therefore it boosts the exergetic efficiency as well. Second, the mass
has a similar effect, that is to say the top performance is achieved with more plates of MCM. Furthermore, it
is quite apparent that ηex is more sensible to the variation of B than m. However, these relations are strongly
nonlinear and it can be noticed that the repercussions of the mass are especially important at high intensity
fields.
Figure 3 reports the exergetic efficiency for numerous water-based mixtures at different concentrations. The
scale ranges from 0 to 60%, in other words from pure water to a more concentrated mixture. It is remarkable
that in all cases, performance goes down and particularly fast for low thermal capacity fluids like ethanol.
One could expect ammonia to outperform pure water since it has a higher thermal capacity (cp ~ 4.74 J/kgK)
but a higher viscosity and a lower density counterbalance the positive effects.
Figure 2. 3D map of cooling power (left) and exergetic efficiency (right)
9
Figure 3. Exergetic efficiency as a function of mass concentration (left) and the influence of crossover fraction (right)
The multiobjective optimization Matlab toolbox in concert with the function gamultiobj are used to
implement the genetic algorithm into the code using an initial population of 10 individuals (maximum
100 generations). Genetic diversity principally comes from two factors, mutation and crossover. The last one
represents the fraction of each population that is made of crossover children. Figure 3 shows the influence of
such variable regarding the normalized processing time. It turns out that the optimal range is located from
0.70 to 0.85, which arises with a solution about 9 times faster than mapping the entire grid. As one can see, it
is not an effective strategy to seek for one or the other extremity, because the algorithm is no more
sufficiently flexible to avoid falling into local optima.
4. CONCLUSIONS
This paper outlined the thermodynamic performance of a magnetic cooling device. A numerical model based
on experimental data of several prototypes has been developed to simulate and optimize the system. Among
numerous fluids, water proved to be the most effective, mainly because of its high thermal capacity. This
result is of particular practical interest, given that water is inexpensive and environmentally friendly. It has
also been observed that GA for multivariable and multiobjective optimization is a very versatile tool. The
example presented suggests that the best crossover fraction should be around 0.70. For future work, this
investigation will be extended to include mixtures and multilayers regenerator into the optimization process.
ACKNOWLEDGMENTS
This project is part of the research program of the Industrial NSERC Chair in Energy Efficiency, established
at the Université de Sherbrooke in 2014, with the support of Hydro-Québec, CanmetEnergie, Rio Tinto
Alcan, and the Natural Sciences and Engineering Research Council of Canada.
REFERENCES
[1] H. Bouchekara, “Recherche sur les systèmes de réfrigération magnétique : Modélisation numérique,
conception et optimisation”, Ph.D. Thesis, Institut National Polytechnique de Grenoble (2008).
[2] K. Nielsen et al., “Review on numerical modeling of active magnetic regenerators for room temperature
applications”, International Journal of Refrigeration 34, 603 (2011).
[3] S.Y. Dankov, A. Tishin, V. Pecharsky, K. Gschneidner, “Magnetic phase transitions and the
magnetothermal properties of gadolinium”, Physical Review B 57, 1 (1998).
[4] A. Kitanovski, P.W. Egolf, “Application of magnetic refrigeration and its assessment”, Journal of
Magnetism and Magnetic Materials 321, 777 (2009).
[5] J. Roudaut, “Modélisation et conception de systèmes de réfrigération magnétique autour de la
température ambiante”, Ph.D. Thesis, Université de Grenoble (2012).
[6] B. Yu, M. Liu, P.W. Egolf, A. Kitanovski, “A review of magnetic refrigerator and heat pump prototypes
built before the year 2010”, International Journal of Refrigeration 33, 1029 (2010).
[7] H. Bouchekara, A. Kedous-Lebouc, J. Yonnet, C. Chillet, “Multiobjective optimization of AMR
systems”, International Journal of Refrigeration 37, 63 (2014).
[8] P. Nikkola, C. Mahmed, M. Balli, O. Sari, “1D model of an active magnetic regenerator”, International
Journal of Refrigeration 37, 43 (2014).
10
A TEST STAND TO MEASURE THE MAGNETIZATION WORK OF
MAGNETOCALORIC MATERIALS
L. D. R. Ferreira (a)*, C. V. X. Bessa (a), S. Gama (b), O. Horikawa (a)
(a)
University of São Paulo, 2231 Av. Prof. Mello Morais, 05508-030, São Paulo, Brazil
(b)
Federal University of São Paulo, 1535 Manoel da Nóbrega, 09910-720, Diadema, Brazil
*
Corresponding author. E-mail: mec.lucas@usp.br
ABSTRACT
The experimental apparatus presented in this study is capable of precisely measuring the magnetic force
produced by a permanent magnets assembly over a magnetocaloric material at a regulated temperature. This
data is then used to calculate the magnetization work necessary to bring the given magnetocaloric sample
from zero field up to a 1 T field region over a given magnetic field gradient, for a range of temperatures
around the Curie temperature of the material. This magnetization work can be used to compare different
magnetocaloric materials, as well as to validate analytic methods to estimate the magnetization work of real
applications in magnetic refrigeration, and thermomagnetic motors. Construction details of the test stand are
presented as well as results for Gadolinium (Gd) and Manganese Arsenide (MnAs).
Keywords: Experimental Test Stand; Magnetic Systems; Permanent Magnets Assembly; Material
comparisons; Thermomagnetic Motors; Magnetic Refrigeration.
DOI: 10.18462/iir.thermag.2016.0059
1. INTRODUCTION
In the development of magnetic refrigerators, it is important to adequately calculate the coefficient of
performance (COP). This means finding the relation between the cooling capacity of the refrigerator and the
external magnetization work, i.e. the amount of work necessary to bring a magnetocaloric material into a
magnetic field. Also in the development of thermomagnetic motors, the magnetization work constitutes the
available work to be converted from thermal energy into useful kinetic energy. The magnetization work is
usually calculated using some form of the equation [1], in which w is the specific work of
magnetization (W/m³), μ0 is the permeability of vacuum, M is the magnetization of the material dependent
upon its temperature and thermal history (for first-order materials), and also dependent upon the magnetic
field H. However, one has to be careful of the type of thermodynamic cycle that is being realized and its
applicability to the type of system being analyzed. Also the equation does not take into account the many
irreversible processes that may occur during the application of the work, such as friction and the occurrence
of Eddy currents in the magnetic material due to the changing magnetic field.
This paper presents a test stand (Fig. 1) intended to directly measure the magnetization work by the field
produced by a permanent magnets assembly on a sample of magnetocaloric material, over a regulated
temperature range. The main purpose of such experiment is to provide a simple experimental base to validate
different thermodynamic/magnetic forces models, without the influence of any external irreversibility, and
also to provide a simple way to compare different magnetocaloric and thermomagnetic materials. The
following section presents the experimental apparatus built, and then some results for the materials
Gadolinium (Gd) and the Manganese Arsenide compound (MnAs), finally some conclusions and evaluations
of the system are presented.
2. EXPERIMENTAL SETUP
The test stand (Fig.1) is intended to measure only the effect of the magnetic force over the material. In order
to do that, it uses a linear flexure bearing to minimize measurement errors caused by friction in moving parts.
A precision linear actuator and a linear encoder control the relative position of the material; and a load cell
measures the varying magnetic force. To characterize the material’s thermomagnetic behavior, the force was
measured for a range of temperatures around the material’s Curie temperature (TC). The temperature of the
sample is controlled using a Peltier element and measured using nonmagnetic thermocouples (type N). The
sample has a coin shape with a diameter of 15 mm, and is supported by a Celeron piece that compresses the
11
material to one of the faces of the Peltier element. Thermal grease is applied between the two faces in order
to obtain a good thermal contact. The Celeron support also functions as thermal isolation between the
material and the environment.
(a) (b)
Figure 1: (a) A schematic view of the experimental setup. (b) A picture of the assembly.
The permanent magnets assembly used to produce the magnetic field, and with that inducing force to the
sample, is also presented in Fig.1.a. It uses two 20x20x50 mm NdFeB magnets (grade N50M) in the
orientation presented in Fig. 1.a. A steel support (AISI 1020) acts, both, to keep the magnets in position and
to concentrate the magnetic field in the desired direction. This assembly produces a field of up to 1.03 T, and
the magnetic field lines are oriented in a single direction.
In order to measure only the effect of the magnetic force over the material, the system must first be
calibrated to take into account the force produced by the linear flexure bearing. This is done by obtaining a
force curve of the system without the presence of any magnetic material. This curve has a linear behavior
and is subtracted from the measured force during the experiments. The measurement is realized in a static
manner, i.e. the magnetic force is registered only when the linear actuator is completely stopped. After the
force measurement, the linear actuator moves the magnets by a given distance, which is verified using the
linear encoder. By using this static measurement method, the system is not influenced by the Eddy currents
that may occur in a fast changing magnetic field. All materials used in the construction of the test stand are
non-magnetic. The base, the sample holder, and the load cell are made of aluminum. The linear flexure
bearing is made of non-magnetic stainless steel. This way, the only magnetic force occurring in the system is
caused by the magnetic material sample being analyzed.
This apparatus is then capable of measuring forces of up to 50 N, with a precision of ± 0.05 N, limited by the
load cell used. The position can be controlled with an accuracy of up to 0.07 mm, and the materials
temperature can be set from 278 K (5°C) up to 353 K (80°C) using the Peltier element. The permanent
magnets assembly used in the system can be changed for different configurations. The resulting magnetic
field is measured using a Hall Effect sensor. This way, the applied magnetic field can be directly obtained
from the magnets position. This test stand can then be used for different measurements, such as the
validation of theoretical force equations [2], and also to obtain the magnetization work for different
magnetocaloric materials working near room temperature, as presented below.
3. RESULTS
Using the test stand, a series of curves of the force as a function of the magnets assembly displacement is
obtained for varying material’s temperatures. The graph presented in Fig. 2, shows a comparison between the
measured force of a sample of 5.58 g of Gadolinium (with a diameter of 15 mm and a thickness of 4.12 mm)
at 283.15 K, and the simulation data obtained using a 3D finite element analysis (FEA) magnetic software
(Ansys Maxwell 3D). The simulation is produced using the same design and parameters of the system being
measured, and the data for the Gd sample was obtained through a Superconducting Quantum Interference
Device (SQUID), taking into account the internal field (Hint) of the sample, that is, considering the
12
demagnetization factor of the Gd sample analyzed in the SQUID (a sphere was used to ensure that the
correct demagnetization factor was applied).
Figure 2: A comparison between the measured data and a 3D magnetostatic simulation for Gadolinium at the temperature
of 283 K.
The result presented in Fig.2 shows a very good agreement between the simulation and the data obtained in
the experiment. This confirms that the simulation was done using adequate method and reliable data but also
shows that the built test stand is capable of precisely measuring magnetic forces. The measured data starts at
the position of 1.5 mm (the distance between the face of the magnets and the sample). This initial position is
due to thickness of the sample holder and the thermocouple located in the face of the material (Fig.1.b).
The resulting magnetic force curves are then integrated to obtain the magnetization work of the sample,
which is divided by the mass of the sample in order to obtain the specific magnetization work. The linear
displacements are then correlated to the measured applied magnetic field gradient ( ) produced by the
permanent magnets arrangement, obtaining a 3D graph surface of the sample specific magnetization work as
a function of the sample’s temperature and the applied magnetic field change from a zero field region. Fig. 3
presents the resulting specific magnetization work surface for Gadolinium (Gd).
Figure 3: The measured magnetization work of Gadolinium as a function of its temperature and the applied magnetic
field change from a zero field region.
The specific magnetization work is also measured for a MnAs compound, which is made into a 2.5 g and 15
mm diameter coin with a thickness of 3.4 mm, by agglomeration with epoxy (4% in mass). The MnAs
compound has a first order magnetic transition and a large hysteresis. This means that the magnetic state of
MnAs around the Curie temperature (TC) is dependent not only on the temperature itself but rather on the
thermal history of the sample. In order to adequately describe the material’s behavior, at least two surfaces
are required, one for the heating, and another for cooling the material, as presented in Fig. 4.
13
Specific Magnetization Work for MnAs
(a) (b)
Figure 4: The measured magnetization work of MnAs as a function of temperature and the applied magnetic field
change from a zero field region. (a) sample heated from room temperature to 338K; (b) sample cooled from 338K to
room temperature.
As shown in Fig.4.a, the MnAs, that is heated, has a Curie temperature (TC) around 318K, consistent with the
data obtained through other experimental methods [3]. When the sample is cooled from a temperature above
the initial TC, the material present a large thermal hysteresis and the magnetic transition will occur only at
308K. It is important to notice that sharp transition expected for a first order material is observed in the
specific work of MnAs, when the material is being heated and cooled.
4. CONCLUSIONS
This paper presented a test stand that is capable of precisely measure the magnetic force produced by a
permanent magnets assembly over a magnetic material sample with controlled temperature. This data is used
to obtain the specific magnetization work of magnetocaloric materials at a given temperature and a given
change in the applied magnetic field (over a known magnetic field gradient). This information can then be
compared to the expected magnetization work calculated using a thermodynamic method or different
magnetic force approaches. Through this comparison, different analytic methods can be validated and used
to model real magnetic refrigerators and thermomagnetic motors.
One important thing to notice is that the curves presented in this study should be correct regardless of the
magnetization direction. This means that the specific work that is obtained from bringing a sample at a given
temperature from zero field into a higher field is reversible. This way, the graphs can be used to calculate the
available work in a theoretical ideal thermomagnetic motor. Using the Gd as an example (Fig.3), the amount
of work that is produced by the magnetic interaction when bringing a Gd sample at 278K from zero field into
a 1T region is 45 J/kg. If the sample is then heated to 303K, the amount of work that will be necessary to
remove the material from the 1T region into zero field will be 11 J/kg, resulting in a theoretical maximum
available work for this field gradient of 34 J/kg. As this system does not take into account the work losses,
the available work in a real system working in the same conditions should be smaller.
ACKNOWLEDGMENTS
This work was supported by Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) grant
number 2012/09486-0 and Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq). The
authors also thank the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES).
REFERENCES
[1] A. Kitanovski, P.W. Egolf , “Thermodynamics of magnetic refrigeration”, International Journal of
Refrigeration. Vol. 29(1), 3 (2006).
[2] S. Gama et al. “Analytic and experimental analysis of magnetic force equations”, IEEE Transactions on
Magnetics, Vol. 52(7), 1 (2016).
[3] S. Gama et al. “A general approach to first order phase transitions and the anomalous behavior of
coexisting phases in the magnetic case”, Advanced Functional Materials, Vol. 19(6), 942 (2009).
14
THERMAL SINGLE-IONIC MAGNETIC PROPERTIES OF MATERIALS
CALCULATED WITH ATOMIC MATTERS MFA COMPUTATION
SYSTEM
R.Michalski(a)*, J. Zygadło(a,b)
(a)
Induforce, Pszona 41/29 31-462 Cracow, Poland
(b)
Jagiellonian University, Faculty of Mathematics and Computer Science,
Łojasiewicza 6 30-348 Cracow, Poland
*
E-mail: r.michalski@induforce.eu
ABSTRACT
We present the concept and scientific methods of a new computation system called ATOMIC MATTERS MFA.
This computation system is useful for magnetocaloric effect (MCE) estimations for defined materials.
ATOMIC MATTERS MFA provides magnetic, calorimetric and spectroscopic properties of atomic-like
localized electron systems under the influence of Crystal Electric Field (CEF), spin-orbit coupling, and
magnetic interactions, taken as both Mean Field Approximation (MFA) and the influence of external magnetic
fields. ATOMIC MATTERS MFA provides macroscopic properties of materials in different temperature
regions, especially around phase transition temperature, such as: magnetic moment (spin and orbit
components), localized electron specific heat, localized electron entropy (with a useful tool set for MCE
estimations ), evolution of energy level positions, and total free energy temperature dependences in a definable
temperature region. ATOMIC MATTERS MFA computation system is an extension of the ATOMIC
MATTERS application, a system for describing fine electronic structure and predicting basic magnetic and
spectral properties of materials.
Keywords: ATOMIC MATTERS, ATOMIC MATTERS MFA, Mean Field Approximation (MFA), crystal
electric field (CEF), spin-orbit coupling, Magnetocaloric effect (MCE).
DOI: 10.18462/iir.thermag.2016.0061
1. INTRODUCTION
The single ionic properties of paramagnetic ions are the core magnetic properties of magnetic materials.
Finding active materials dedicated to magnetic refrigeration at room temperature need tools for single ionic
properties provides. Therefore, we have created software to perform calculations of fine electronic structure
for paramagnetic ions. Our aim was to find and parameterize the most important interactions in atoms and ions
in ligand environments. Such parameterization, which comes directly from Crystal Electric Field (CEF)
theory[1-3], makes it possible to transfer the influence of local charge surroundings between different ions and
predict the magnetic properties of materials in isostructural series [4-6]. This system is called ATOMIC
MATTERS. Algorithms and technical details of ATOMIC MATTERS computation system were published at
[7] and are also available on the software web page [8], together with hundreds of examples of calculation
results for ions from different areas of the periodic table in real, well-known materials. All ATOMIC
MATTERS calculations can simulate materials under stable field conditions (both CEF and external magnetic
field) and interactions. ATOMIC MATTERS MFA can simulate phase transitions of ionic/atomic systems
according to dynamic calculations of the molecular field Bmol, simply defined as:
B mol (T)  nmol m(T), (1)
that interacts with ions to induce their magnetic moment. Such self-consistent calculations can be performed
after establishing the molecular field factor nmol, which is closely related to the phase transition temperature
(TC or TN)[9]. Our solution applies to situations in which an unclosed electron 3d/4d/5d/4f/5f subshell interacts
with the electrostatic potential of definable symmetry (taken as a CEF) and the external magnetic field and
molecular magnetic field of identical neighbors, calculated under Mean Feld Approximation (MFA)
methodology.
2. THEORETICAL APPROACH
Our approach, which is applied to both ATOMIC MATTERS and ATOMIC MATTERS MFA computation
systems, is deeply rooted in atomic physics. Taking into consideration the individual population of states of
fine electronic structure of ions/atoms different temperatures according L. Boltzmann statistics makes it
possible to define the temperature dependencies of properties such as:
15
 full structure of discrete electron levels Ei or Ei (T) for magnetization process in chosen calculation
space |J,Jz> or |L,S,Lz,Sz>
 Helmholtz free energy F(T), internal energy U(T), magnetic entropy Smag(T)
 formation of magnetic ordered state
 localized electronic contribution to specific heat cmol(T) for temperature dependent energy level states
Ei(T) around magnetic phase transition temperature.
 spectra simulations for different methods and conditions, including INS (Inelastic Neutron Scattering)
 magnetic moment and magnetization in ordered state
 spin and orbital contribution to angular momentum of the entire electron subshell
 magnetocrystalline anisotropy constants Ki(T)
For rapid calculations in the thermodynamically stable temperature region, ATOMIC MATTERS offers the
following CEF+Spin-Orbit+Zeeman Hamiltonian according to the chosen calculation space of ground
multiplet |J,Jz> or ground atomic term |L,S,Lz,Sz,> respectively[11]:
ˆ m (J, J )  g  J  B
H J  H CEF  H Zeeman   Bmn O (2)
n z L B ext
n m
or
ˆ m (L, L )   L  S   (L  g S)  B
H LS  H CEF  HS-O  H Zeeman =  BmnO (3)
n z s o B e ext
n m
For all Hamiltonians: Bmn denotes CEF parameters, Omn are Stevens operators, -is the spin–orbit constant, gL
and ge≈ 2.002324 are the gyromagnetic factors. For whole ion and for electron respectively, µB is the Bohr
magneton and Bext is the external magnetic field.
In all cases, calculations in the |L,S,Lz,Sz,> space are more physically appropriate due to their completeness,
but traditional calculations with base |J,Jz> can be also performed by our computation systems for comparisons
and rapid estimations [1],[2]. For calculating properties in temperatures around the magnetic phase transition
point (by ATOMIC MATTERS MFA), self-consistent methodology for molecular field calculation called
Mean Field Approximations(MFA) is applied. The idea of this method is the estimation of direction and value
of the magnetic field (molecular field) generated by ions at defined temperature, and calculating the influence
of this magnetic field for electronic states structures of ions. According to the chosen calculation space, we
define molecular field as:
B mol  nmol g L  B  J  (5)
or
B mol  nmol  B  L  g eS  . (6)
On the basis of calculated electronic structure Ei, the directional components of magnetic moments are
established for all identical ions. This means that Hamiltonian matrix diagonalization is performed for all
defined temperature steps recurrently, in contrast to simple ATOMIC MATTERS calculations, which
diagonalize matrices one time for a single run and deduce all thermodynamic properties from the stable discrete
energy level structure obtained. This self-consistent procedure provides temperature-dependent energy level
structure and has only one only free parameter, nmol, called the molecular field parameter. The value of nmol is
closely related to the phase transition temperature TN or TC of the macroscopic structure of ions. The formal
expression of the full Hamiltonian used by ATOMIC MATTERS MFA computation system, according cto the
hosen calculation space |J,Jz> or |L,S,Lz,Sz,> respectively, has the form:
ˆ m (J, J ) n g 2  2   J  J   1  J  2   g  J  B
H J mol  H J  H mol   Bmn O (7)
z mol L L ext
2
n B B
n m  
or
ˆ (L, L )   L  S  n  1 
H LS mol =  B O m
z
m
s o mol    (L  g eS)  L  g eS    L  g eS  2    B (L  g eS)  B ext
2
(8)
2
n n B
n m  
The eigenvectors of the Hamiltonian are described according to the selected calculation base by the total
momentum quantum numbers |J,Jz> or spin and orbit quantum numbers |L,S,Lz,Sz,>. Using the commutation
relations of the angular momentum operators, we obtain information about expected values of the projections
of magnetic momentum of all electronic states at a chosen temperature [2]-[5], [8].
gL B  E i (T )  B  E i (T ) 
m αJ ( T )    J  exp  
i
 (9) , m αLS ( T )    L  g S  exp  
i
e
i
 (10)
Z(T) i  k BT  Z(T) i  k BT 
Where: α indexes directional components, i - numbers the Hamiltonian eigenstates, while <Jα> represents the
expected value of the total angular momentum along the α-axis in the i-th state. All property calculations can
16
be done for 3D (x,y,z) real space by using complex Hamiltonian matrix elements defined by full expressions
of extended Stevens Onm operators[10]. Mostly for comparison with traditional calculation results, ATOMIC
MATTERS also offers, 2D (x,z) calculation methodology of a simplified model of CEF interactions defined
by Stevens Onm operators with real number matrix elements only [1-3],[7],[11].
3. THERMODYNAMICS
Taking into consideration the possibility of the thermal population of states, we automatically obtain thermal
evolution of single ion properties of whole compound[3],[4],[7-11]. Under the thermodynamic principle at a
temperature T = 0 K, only the ground state is occupied. In this situation, the magnetic moment of the ion is
exactly equal to the momentum of the ground state. If the temperature rises, the probability of occupying higher
states increases according to Boltzmann statistics. The number of ions with energy E i within a system at
temperature T is:
 E (T ) 
exp   i 
Ni ( T )  N0  k BT  ; (11)
Z(T )
where N0≈6.022.1023 mol-1 (Avogadro constant) and Z(T) is the statistical sum of states:
  H ˆ   E (T) 
Z ( T )  Tr exp       exp   i
 k T  i  k T  (12)
  B   B 
Knowing the statistical sum of the states, we can determine the Helmholtz free energy F(T):
F ( T )  k BT ln Z(T) (13)
According to thermodynamic principles, the contribution of localized electrons to the total specific heat of
materials can be calculated by numerical derivation of Helmholtz free energy:
 2 F ( T ) 
c mol ( T )   T 2
 (14)
 T 
It make it possible to calculate entropy according to definition:
c(T)
T
S( T )  S( 0 )   dT (15)
0
T
However, for general correctness, due to the fact that open subshell ions are in a self-organized molecular field,
their magnetic moment changes with temperature. Bearing in mind that each eigenstate of the CEF
Hamiltonian is related to its magnetic moment, the total moment of a strongly correlated electron system at a
given temperature is the resultant moment of occupied states calculated with the inclusion of the Boltzmann
weight and:
M ( T )  g L  B  N i (T)  Γi (T) | J  | Γi (T)  (16)
i
or
M ( T )   B  N i (T)  Γi (T) | (L  g eS ) | Γi (T)  (17)
i
where α-indexes directional components, i-numbers the Hamiltonian eigenstates, while <Jα> (from eq. 16) and
| (L  geS )  (from eq.17) represent the expected values of the total angular momentum operator defined
in different spaces. Such magnetic moment calculations, under influence of a freely defined external magnetic
field and internal dynamic calculated molecular field can be achieved by ATOMIC MATTERS MFA.
ATOMIC MATTERS MFA also provides single-ionic magnetocrystalline anisotropy calculations that include
full calculations (without Brillouin function approximation) of Ki(T) magnetocrystalline constants for defined
temperature ranges.
4. EXAMPLES OF RESULTS
According to the described methodology, we present the calculations results for well-known magnetic
magnetocaloric compound DyAl2. All used calculated parameters used for calculations are shown on the plots
apart from CEF parameters of cubic field taken from[12] : B4 =−5.5.10−5meV and B6=−5.6.10−7 meV.
17
Dy3+ ions in DyAl2 600 Dy3+ ions in DyAl2
400
|L,S,Lz,Sz> space |L,S,Lz,Sz> space
E (K)
E(K)
nmol=3.6T f.u./μB nmol=3.6T f.u./μB Bext=10T

Bext=0
300
400
200
200
100
0 0
0 20 40 60 80 100 0 20 40 60 80 100
Temperature (K) Temperature (K)
Fig1. Low energy electronic level structure, near ground state, calculated for Dy3+ ions in DyAl2, calculated by ATOMIC
MATTERS MFA in |L,S,Lz,Sz> space under influence of molecular magnetic field, external magnetic field B=0 and
B=10T and CEF with parameters taken from [12].
12 60
Dy3+ ions in DyAl2
Specific heat c (J/mol K)
B|| [100]
Magnetic moment m (µB)
|L,S,Lz,Sz> space
nmol=3.6T f.u./μB
8 40
4 B=0T 20 B=0T
B=1T B=1T
B=5T B=5T
B=10T B=10T
0 0
0 40 80 120 0 40 80 120
Fig2. Magnetic moment vs. temperature of Dy3+ ions in Fig3. 4f components of molar specific with Debye
DyAl2, calculated by ATOMIC MATTERS MFA in component (θ=380K) vs. temperature of Dy3+ ions in
|L,S,Lz,Sz> space under influence of molecular magnetic DyAl2, calculated by ATOMIC MATTERS MFA in
field, CEF with parameters taken from [12] and various |L,S,Lz,Sz> space under influence of molecular magnetic
external magnetic field. field, CEF with parameters taken from [12] and various
external magnetic field parallel to [100] axis
60 8
Specific heat c (J/mol K)
Dy3 in DyAl2 B=1T

B|| [110] B=5T
6 B=10T
-ΔSmag(J/K mol)
40
4
20
B=0T 2
B=1T
B=2T
0 0
0 40 80 120 0 25 50 75 100 125
Fig4. 4f components of molar specific with Debye Fig5. 4f entropy, calculated under influenxe of external
component (θ=380K) vs. temperature of Dy3+ ions in magnetic field parallel to [100] axis for Dy3+ ions in
DyAl2, calculated by ATOMIC MATTERS MFA in DyAl2, in |L,S,Lz,Sz> space under influence of molecular
|L,S,Lz,Sz> space under influence of molecular magnetic magnetic field, CEF with parameters taken from [12] and
field, CEF with parameters taken from [12] and various various external magnetic field parallel to [100]axis.
external magnetic field parallel to [110] axis.
18
5. CONCLUSIONS
Selected calculations of DyAl2 thermal dependent properties were performed with ATOMIC MATTERS
MFA. The calculation parameters were taken from references [12]. We obtained excellent agreement with
experimental data from [12]. We confirm agreement of our results with similar theoretical calculations made
in ‘narrow’ |J,Jz> space with infinite spin-orbit coupling [13,14], but we conclude that taking into consideration
higher multiplets (calculated in full |L,S,Lz,Sz> space with finite value of spin-orbit coupling constant λs-o)
gives more reliable results, especially in higher temperature regions. All spaces and simplifications described
above can be simulated by our computation systems for all naturally occurring ions. Our systems can transfer
parameters of CEF between different ions and different calculation spaces. User can freely define the strength
of spin-orbit coupling and estimate how large value of spin-orbit coupling can be treated as a ‘infinite’. The
fast comparison between calculations with free-ion value of spin-orbit constant λs-o and with freely defined
λs-o or even with ‘space of ground multiplet’ (with infinite s-o interaction), made it clear that ATOMIC
MATTERS MFA computation system is extremely useful. The visual form of calculation results and full 3D
interactive CEF potential visualization make it easy for the user to utilize the power of the application very
effectively. From a theoretical point of view, we conclude: In all cases the |L,S,Lz,Sz> base is better for precise
magnetism calculations and spectra recognition. Calculations on complex number matrices always produce
more precise magnetic simulations. Full, correct directional magnetic information is provided by calculations
on complex number matrices only. In general, base |J.Jz>, which is the most commonly used for calculation
of R ions, and real number matrix Hamiltonians should be chosen for comparisons and rapid calculations only,
because it offers approximate results with intermediate scientific value.
REFERENCES
[1] J. Elliot, K. W. H. Stevens, Proc. Roy. Soc. A 215 (1953) 437.
[2] M. T. Hutchings, Solid State Phys. 16 (New York 1964) 227.
[3] B. G. Wybourne, Symmetry Principles and Atomic Spectroscopy, J. Wiley and Sons, New York (1970).
[4] A. Abragam and B. Bleaney, EPR of Transition Ions, Clarendon Press, Oxford (1970).
[5] J. Mulak and Z. Gajek, The effective crystal field potential, Elsevier Science, Oxford, UK (2000).
[6] R. J. Radwański, R. Michalskia, Z. Ropka , A. Blaut., Physica B 319 (2002) 78–89
[7]Rafal Michalski, Jakub Zygadlo, Describing the Fine Electronic Structure and Predicting Properties of
Materials with ATOMIC MATTERS, 18th CPMCS 2016,Conference paper – in printing
[8] ATOMIC MATTERS computation system website: http://www.atomicmatters.eu/
[9] J.J.M. Franse, R.J. Radwanski; Magnetic properties of binary rare-earth… in Handbook Of Magnetic
Materials Vol 7. Edited K.H.J.Bushow (1993) 307-500.
[10]C. Rudowicz, J. Phys.: Condens. Matter 16 (2004) 5825–5847.
[11] R. J. Radwański, R. Michalski, Z. Ropka , A. Blaut., Physica B 319 (2002) 78–89
[12] A.L. Lima et al. Phys. Rev. B 72 (2005)
[13] P. J. von Ranke, N. A. de Oliveira,E. J. R. Plaza,V. S. R. de Sousa, B. P. Alho, A. Magnus G. Carvalho,
S. Gama and M. S. Reis, Journal Of Applied Physics 104, 093906 (2008)
[14] V.S.R. deSousa, E.J.R.Plaza, M.S.Reis, B.P.Alho, A.Magnus G.Carvalho, S.Gama, N.A.de Oliveira, P.J.von
Ranke, J. of Magn. Magn. Mater. 321 (2009) 3462–3465
19
MAGNETOCALORIC BENCH: ANALYTICAL MODEL FOR
GADOLINIUM CHARACTERISTICS
A. Plait(a)*, S. Giurgea(a), C. Espanet(b), T de Larochelambert(a), P. Nika(a)
(a)
FEMTO-ST Institute, Energy department, Belfort, France
(b)
Moving Magnet Technologies, Besancon, France
*
Corresponding author.E-mail: antony.plait@utbm.fr
ABSTRACT
Active magnetic refrigeration cycle analysis for different waveforms both magnetic field and heat transfer fluid
velocity is an essential issue to design and implement heating and cooling applications based on magneto-
caloric effect. Modelling the multiphysics phenomena that occur inside the magnetocaloric regenerators
requires the coupling of magnetostatic, magnetocaloric and thermofluidic models.
The global interest of the contribution presented here is to decrease the resolution time of the magnetostatic
model. Indeed, the multi-physics model integrating this fast magnetostatic model needs much less computation
time and ensures its compatibility with an optimization process.
In this context, an analytical modelling of magnetostatic phenomena is developed by taking into account the
nonlinear behaviour of the ferromagnetic external circuit as well as the active magnetocaloric material (MCM).
The analytical model calculates the values of the magnetic field and the magnetic flux density at each point of
the regenerator volume. Moreover, the model considers the dependency of magnetic model B(Hint) with the
temperature in the case of the magnetocaloric material characteristics obtained experimentally.
Keywords: magnetocaloric bench, analytical model, network reluctances, gadolinium, regenerator.

DOI: 10.18462/iir.thermag.2016.0070
1. INTRODUCTION
In the magnetic refrigeration domain, an increasing number of magnetic refrigeration prototypes based on
active magnetic regenerative refrigeration (AMRR) principle have been built and tested [1].
A specific experimental prototype has been developed and realised in our laboratory (Fig. 1, 2), including a
controlled source of pulsed magnetic field (powerful electromagnet) and a controlled hydraulic device
producing precise flow sequences through an active magnetocaloric regenerator block with two micro-heat
exchangers on its extremities. The regenerator consisting of pure gadolinium rectangular parallel plates (13 x
1 x 45 mm3) distant from 0.5 mm (Fig. 3) can be introduced inside a 21 mm wide air gap. The purpose of the
bench is to characterize the magnetocaloric regenerator and evaluate its maximum refrigeration performances.
Depth: 90mm
Length: 21mm D
Height: 50mm
Figure 1. Magnetocaloric test bench Figure 2. Simulated test bench Figure 3. Air gap test bench
A lot of numerical models of AMRR have been developed; some consider the magnetic field equal to the
applied field [2], some other consider the magnetic field constant in the whole material [3] or use simplified
approaches [4]. In a previous work, a magnetostatic finite element model (FEM) has been developed and
integrated in a multiphysic analysis of the refrigeration test bench, and a significant magnetic field
heterogeneity has been observed [5]. This heterogeneity can be explained by the strong dependence of the
internal magnetization with the temperature near the Curie point of the magnetocaloric material, and also with
20
3D phenomena in magnetic field distribution into the magnetocaloric test bench air gap. However, the main
drawback of the magneto static FEM model is the highly time consuming calculation.
In this work an analytical model is designed for the magnetic flux density and magnetic field calculation inside
the magnetic materials with a precision comparable to FEM, and takes into account the heterogeneous
distribution of the magnetic field in the active magnetocaloric regenerator. The 3D analytical model represents
the whole experimental system {electromagnet, regenerator, air gap} and allows to obtain adequate results
with much less calculation time than pure FEM.
2. PROPOSED MAGNETOSTATIC MODEL
As the main objective of the test bench is to characterize precisely the magnetic refrigeration AMRR cycle, it
is important to estimate the magnitude of the magnetic field distribution in the regenerator even if the external
magnetic circuit is saturated (non linear behavior).
Models using reluctances network have been used in electrical machine modeling and rarely in magnetic
refrigeration domain [6], [7]. The latter paper evaluates the influence of the Gd stack on the magnetic flux
inside the gap og Nd-Fe-B magnetic circuits. However, nonlinear behavior of ferromagnetic circuits is not
taken into account.
Firstly, the nonlinear behavior of the test bench external
ferromagnetic yoke is fitted on experimental measurements of
the magnetic field in the empty air gap; the calculation scheme
is presented in Fig. 4. During this step, the magnetomotive
force of the ferromagnetic circuit is calculated. The magnetic
flux density is precisely measured in the core of the air gap
with a Gauss-meter for currents ranging from 0 to 50 A with a
5 A step, leading to the magnetic flux (in Wb) Air gap = B0 .
SAir gap (air gap section S = 90×50 mm2). From there, the
reluctance R0 = L / (µ0S) of the air gap and the magnetomotive force (mmf) Um0 = R0.ΦAir gap are calculated
and further adjusted to FEM considering the leakage flux outside the regenerator air gap. So, the
magnetomotive force of the ferromagnetic circuit can be obtained using Kirchhoff's law UmFe = N.I - Um0 where
I is the intensity of the electric current flowing in the coils and N is the number of coil turns (Fig. 5).
Figure 5. NI, Um0 and Um Fe Figure 6. Decomposition of gadolinium plates
In the second step, the air gap with Np pure gadolinium rectangular parallel plates is considered (Np = 13 in
Fig. 6). The calculation scheme of air gap with gadolinium plates can be seen in Fig. 9.
For the flux of the air portion of the air gap, it is necessary to initially split the gap into a network of reluctances
in series and in parallel to calculate the equivalent reluctance Rair (Figure 3, 7 and 8). The air magnetic flux is
calculated by Φair = Um0 / Rair.
For the flux calculation in the regenerator, each plate i is divided into Ns equal segments k (k = 1 to Ns = 10 in
Fig. 6) since magnetic flux density is not homogeneous over the entire plate. The temperature of each segment
of each plate being known, the dependency B(H) curves are defined by the MCM characteristic B(H, T), which
allows to calculate the characteristic UmGd(Φ) along the (i, k) path, with Φi,k = Bi,k . S and UmGd = ∫ Hi,k dl (S
section and l width of the segment, S and l are constant, so UmGd = Hi,k . l).
A curve fit method is applied to obtain the different values of magnetomotive forces UmGd depending on the
current flowing through the coils. The sum of the fluxes through all segments according to the mmf UmGd
allows to obtain the overall flux in gadolinium which adds to the air portion flux related to mmf Um0.
21
Figure 7. Air gap front view Figure 8. Air gap top view
It remains to obtain NI for the total magnetic flux by adding

the mmf with the Kirchoff's law, leading to the dependence
Θ = f(Φ) where Θ = UmFe + Um0 = N . I. Thereafter, the
magnetomotive force UmAir gap will be estimated in return
knowing the current flowing in the coils along with the
magnetic flux in the air gap. For that, a subsequent interpolation is performed to obtain the magnetomotive
forces. Similarly, the flux is calculated for each segment using an interpolation based on the magnetomotive
forces calculated previously. The ultimate goal is to get the magnetic flux density B and magnetic field H
inside each segment (as the dependence Φi,k(UmGd) is known for each (i,k) segment).
3. RESULTS AND COMPARISON
3.1 Magnetic field and magnetic flux density

The results for magnetic field and magnetic flux
density calculation in a ten-fold segmented plate of
gadolinium with temperatures ranging between 293
K (cold side) and 296 K (hot side) is visualized in
Fig. 10.
The difference between the 10 segments can be
easily observed, the magnetic flux density and the
magnetic field depending clearly on temperature,
particularly around the Curie temperature (293K).
Figure 10. B(H) curves evolution for I [0:50, 5]
3.2 Multiphysics coupling model
Figure 12. Temperature of external segments
Figure 11. Calculation scheme of a magnetic refrigeration system Figure 13. One segment comparison
22
The model of a magnetic refrigeration system can be divided into three parts: a magnetostatic model, a
magnetocaloric model and a thermal model. The complete scheme can be seen in Fig. 11, in which the proposed
analytical model is used as a magnetostatic model [5]. The results are accurate, as seen on the Fig. 12 showing
the temperature evolution of the opposite segments of the central plate (hot and cold ends). The four phases
(magnetization, cold blow, demagnetization, hot blow) are displayed with a frequency 0.5 Hz.
3.3 Model comparison

In this part, the different possibilities to obtain or estimate intern magnetic field and magnetic flux density of
the material are compared. Each model has advantages and drawbacks:
- Finite element method (FEM): very precise magnetic values (3D phenomena) but highly time consuming
- Imposed maximum magnetic field: very fast calculation but very approximate magnetic values
- Analytical model: less time consuming than FEM and acceptable magnetic precision
The results of the different models for intern magnetic flux density and magnetic field applied to gadolinium
at 293K are shown in Fig. 13. The comparison can further focus on the magnetic flux distribution in one Gd
plate, as shown in Fig. 14a-c for magnetic flux density in the central gadolinium plate with imposed
temperatures at each end (cold side 290 K; hot side 299 K).
Figure 14a. FEM Figure 14b. Constant magnetic field Figure 14c. Analytical model
The most accurate results for the magnetic flux density inside the material are obtained with finite element
method (Fig. 14a), where an important magnetic flux density can be observed at plate ends, due to the
confinement of the field lines. In Fig. 14b, the magnetic field is kept constant for each segment with a maximal
value of 1.5 tesla, which represents the mean value obtained with FEM. Therefore, the results are non-precise,
particularly for the magnetization calculation, and don’t take into account all physical phenomena. A good
compromise between precision and time resolution can be observed in Fig.14c when the analytical model is
used. The maximum gap between FEM and analytical model results reaches approximately 4%.
4. CONCLUSIONS
An analytical model has been proposed to calculate the magnetic field H and magnetic flux density B in parallel
plates of magnetocaloric regenerators. The simulation time to get H and B can be considered very low (10
seconds) when applied to a gadolinium plate divided into Ns equal segments.
The analytical model is simple and directly applicable, and it can be extended easily to rotary magnetocaloric
devices. When associated with thermo-fluid model, it allows to complete a simulation of the whole final
multiphysics model within just 5 minutes, instead of 4 hours with 3D FEM (with a DELL T3600 computer).
The magnetic results obtained using our analytical model are very close to those of Flux3D® software
calculations, which confirms the efficiency of our analytical model.
REFERENCES
[1] B. Yu et al,. A review of magnetic refrigerator and heat pump prototypes built before the year 2010.
International Journal of Refrigeration 33 (2010) 1029-1060.
[2] G. Schroeder et al,. A flexible numerical model of a multistage active magnetocaloric regenerator,
International Journal of Refrigeration 65 (2016) 250-257.
[3] C. Aprea, A. Maiorino. A flexible numerical model to study an active magnetic refrigerator for near room
temperature applications. Applied Energy (2010).
[4] B. Torregrossa-Jaime et al,. An efficient numerical scheme for the simulation of parallel-plate active
magnetic regenerators, International Journal of Refrigeration 58 (2015) 121-130.
[5] A. Mira et al., Influence of computing magnetic field on thermal performance of a magnetocaloric cooling
system, European Journal of Electrical Engineering n°17/3-4/2014, 151-170.
[6] D. Vuarnoz T. Kawanami. Experimental validation of a coupled magneto-thermal model for a flat-parallel-
plate active magnetic regenerator, Applied Thermal Engineering 54 (2013) 433-439.
[7] W. Dai et al., Application of high-energy Nd-Fe-B magnets in the magnetic refrigeration, J. Magn. Magn.
Mat. 218 (2000) 25-30.
23
PERFORMANCE COMPARISON OF La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8
MAGNETIC REFRIGERANT PREPARED BY TWO METHODS
J.Cheng(a)(b)*, J.H.Huang(a)(b), C.L.Liu(a)(b), Y.D.Zhang(a)(b), C.Zhang(a)(b),

P.Y.Jin(a)(b), C.C.Li(a)(b), Z.J.Li(a)(b)
(a)
Baotou Research Institute of Rare-earths, Baotou, China
(b)
State Key Laboratory of Baiyunobo Rare Earth Resource Research and Comprehensive
Utilization, Baotou, China
*Corresponding author. E-mail: yigeorange@163.com
ABSTRACT
In this paper La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8 magnetic refrigerant were prepared by hot pressed method and
podwer bonded method respectively, the performance of the samples prepared by two methods were
studied. The structure, morphology and magnetic property were researched. The experiment resutls showed
that the max magnetic entropy for the boned sample is 11.8J/KgK which is larger than that of hot pressing
sample with the max magnetic property of 10.99 J/kgK at the magnetic field of 1.5T. The bending strength
and the thermal conductivity of hot pressing sample are both better than the boned sample.
Keywords: magnetocaloric effect, hot pressed, bonded, magnetic refrigerant

DOI: 10.18462/iir.thermag.2016.0072
1. INTRODUCTION
La(Fe1-XSiX)13 series compounds were found existing large magnetocaloric effect [1] by S. Fujied and
A. Fujida, the maximum |SM| and Tad is 28 J/kg·K and 8.1K respectively at the
magnetic field of 2 T when X=0.9. After that Hu Fengxia[2,3], Fujita、Fujieda[4,5] et al. have done a great
amount of work on the compounds. La(Fe1-XSiX)13 series compounds are considered as a kind of
magnetocaloric effect material with great potential, which has the advantages of high magnetic heat effect,
low price, good thermal conductivity, and continuous and adjustable Curie point. Absorbing hydrogen for
La(Fe,Si)13 is considered to be the optimal method which can effectively improve Tc to room temperature
while maintaining the large MCE. The La(Fe,Si)13 hydrides are easy to effloresce and have poor
mechanical properties. Hu.Zhang and Xai wei et al. prepared the hydrides to blocks by bonded
method using the epoxy resin as binder, the results showed the boned compound had large
magnetocaloric effect and good mechanical properties.[6,7] In this paper La0.8Ce0.2Fe11.47Mn0.23Si1.3
magnetic sample have been prepared by hot pressed method
24
and podwer bonded method respectively. The performance including microstructure, MCE and
mechanical properties prepared by the two methods have been analysed and contrasted by
synthesis.
2. EXPERIMENTAL
La0.8Ce0.2Fe11.47Mn0.23Si1.3 alloy were prepared by the industrial pure raw material. The alloy was
smelted in induction furnace under the atmosphere of high-purity argon. The as-cast compound
was annealed at 1363K for 144h under the high purity argon atmosphere, then qnenched into
water. For the hot pressed method the alloy were milled to powders and hot pressed into cylingdrical at the
temperature of 1173K under the pressure of 30Mpa for 15min, then cut into slices with the thikness of
1mm and hydrogenated at 593K at H2 atmosphere of 0.03Mpa for 6h. For the podwer bonded method the
alloy were hydrogenated first and milled to powders, then mixed with about 1.8 wt.% epoxy resin
adhesibe and additive and pressed into cylingdrical. The hydrogen concentration y of
La0.8Ce0.2Fe11.47Mn0.23Si1.3Hy was determined to be 1.8 analysed by oxygen and nitrogen hydrogen
analyzer (LECO ONH-836). The crystal structure was analyzed by X-ray diffraction (XRD,
Philips PW-1700). The microstructures were examined by scanning electron microscopy (SEM,
Hitachi S-3400N). The magnetic properties were performed in the vibrating-sample magnetometer
(VSM, Lakeshore 7407). The isothermal magnetic entropy changes of samples were calculated by
Maxwell relation . The bending strength and the thermal conductivity were
also measured by the Chinese Academy of Sciences.
3. RESULTS AND DISCUSSION
Fig 1 shows the SEM micrographs of La0.8Ce0.2Fe11.47Mn0.23Si1.3 H1.8 prepared by boned and hot pressed
method. There are a large number of pores and boundaries between different particles and epoxy resin on
the surface of the particles for the boned sample. The NaZn13 main phase (gray), α-Fe phase (black), La-
rich phase (white), pore (irregular black area), boundary and deciduous granule are clearly visible in the
hot pressed sample. Both of the boned sample and hot pressed sample show porous architecture, and the
pores for hot pressed sample are less than that of the boned sample. The structure of hot pressed sample is
more compact than the boned sample.
(a) boned sample (b) hot pressed sample

Figure 1. SEM images of the boned sample and hot pressed sample
25
Fig 2 shows the temperature dependence of the magnetization of the initial sample(after 6 days anneal), hot
pressed sample and the boned sample which was measured in a magnetic field of 0.05 T in a warming
process. The Curie temperature Tc defined as the minimum of dM/dT vs T cureves, is found to be 136K for
initial sample, 288K for hot pressed and boned sample. The moment declines sharply for the initial sample
around Tc indicated had larger magnetocaloric effect compared with the hot pressed sample and boned
sample. Above Tc the magnetic moment for the hot pressed sample is higher than that of the boned sample
indicate more impurity phase existing in the hot pressed sample. Fig 3 shows the temperature
dependence of ΔSM under a magnetic field change of 1.5T for the hydrogenated initial sample, hot
pressed sample and the boned sample respectively, which estimated on increasing magnetic field
from the magnetization data by using the Maxwell relation. The max value of ΔSM is 14.05 J/kgK
for hydrogenated initial sample, 10.9 J/kgK for hot pressed sample and 12.9 J/kgK for boned
sample. Both of ΔSm for the hot pressed sample and boned sample decrease a little compared with
the hydrogenated initial sample for the introduce of epoxy resin adhesibe and the increasment of α-Fe.
initial sample hydrogenated initial sample

hot pressed sample hot pressed sample
30 15
boned sample boned sample
-S/(J/kgK)
10
M/(emu/g)
20
5
10
0
120 160 200 240 280 320
270 275 280 285 290 295 300 305
T/K T/K
Figure 2. Temperature dependences of magnetization Figure 3. Temperature dependence of ΔSm for

for the initial, hot pressed and boned samples(0.05T) the hydrogenated initial, hot pressed and
boned samples(1.5T)
The hot pressed samples after hydrogen absorption didn’t appear cracks. Fig 4 shows the bending
strength of hot pressed samples and bonded samples. The bending strength of hot pressed sample
is 113.6Mpa which is higher than that of the bonded sample. The bending-strain curve of hot
pressed sample shows a yield stage and then drops sharply without further plastic deformation
which means the hot pressing La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8 is a kind of brittle material. The boned
sample shows a gentle decline with plastic deformation caused by the existing of glue which
means the toughening of the materials is improved. The thermal conductivity is 9.4w/mk for initial
sample, 7.3w/mk for hot pressed sample and 2.5w/mk for boned sample (Fig 5). Although the
thermal conductivity for hot pressed is lower than the initial sample, but it much larger than that of
boned sample. The high thermal conductivity of the hot pressed sample depends on the
compacting structure and non-glue.
26
12
160 hot pressed sample
boned sample
140 10
Thermal conductivity(w/mk)
Bending strength (Mpa)
120 initial sample

8
100
80 hot pressed sample

6
60
40 4
20
boned sample
0 2
0 5 10 15 20
Strain (%)
Figure 4. Bending strength of hot pressed and Figure 5. Thermal conductivity of initial, hot
boned sample pressed and boned sample
4. CONCLUSIONS
La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8 magnetic refrigerant were prepared by hot pressed method and podwer
bonded method respectively. The ΔSm is 11.8J/KgK for the boned sample (epoxy resin 1.8 wt.%) and
10.99 J/kgK (hot pressing temperature 1173K) for the hot pressed sample at the magnetic field of 1.5T. The
bending strength of hot pressed sample is better than that of the boned sample.The thermal conductivity is
7.3w/mk for hot pressed sample which is higher than boned sample with thermal conductivity of 2.5w/mk.
Compared with the boned samples the hot pressed samples have more advantages for practical usage.
ACKNOWLEDGEMENTS
This work was supported by the National Natural Science Foundation of China (51261001).
REFERENCES
[1] S Fujieda, A Fujita, K Fukamichi. Large magnetocaloric effect in La (FexSi1-x)13 itinerant-electron
metamagnetic compounds[J]. Appl Phys Lett, 2002, 81(7): 1276-1278.
［2］F. X. Hu, X. L. Qian, J. R. Sun, et al. Magnetic entropy change and its temperature variation in compounds
La urnal of Applied Physics, 2002, 92(7): 3620~3623
［3］F. X. Hu, Max Ilyn, A. M. Tishin, et al. Direct measurements of magnetocaloric effect in the first-order
system LaFe11.7Si1.3[J]. Journal of Applied Physics, 2003, 93(9): 5503~5506
［4］S. Fujieda, A. Fujita, K. Fukamichi, et al. Giant isotropic magnetostriction of itinerant-electron matamagnetic
La(Fe0.88Si0.12)13Hy compounds[J]. Applied Physics Letters, 2001, 79(5): 653~655
［5］A. Fujita, S. Fujieda, Y. Hasegawa, et al. Itinerant-electron metamagnetic transition and large magnetocaloric
effects in La(FeXSi1-X)13 compounds and their hydrides[J]. Physical Review B, 2003, 67: 104416
[6] Hu Zhang, YuJie Sun, et al. Enhanced mechanical properties and large magnetocaloric effects in boned
La(Fe,Si)13-base magnetic refrigeration materials[J]. Appl. Phys. Lett. 104 (2014) 062407.
[7] Xia Wei, Huang Jiaohong, Sun Naikun, et al. Influence of powder bonding on mechanical properties and
magnetocaloric effects of La0.9Ce0.1(Fe,Mn)11.7Si1.3H1.8[J]. Journal of Alloys and Compounds,2015,635:124-128.
27
SCALE UP OF MnFePSi MATERIALS IN A BROAD TEMPERATURE
RANGE AND MANUFACTURING OF ADVANCED HEAT EXCHANGER
STRUCTURES
F. Dötz(a), D. van Asten(b), D. Barrera-Medrano(a) , C. Carroll(a) , U. Kalck(c), S. Misra(c), B.
Reesink(b), O. Rogge(a), F. Scharf(c), M. Schwind(c), F. Seeler(c), M. Zeilinger(c) , L. Zhang(b)
(a)
BASF New Business GmbH, Benckiserplatz 1, 67059 Ludwigshafen, Germany
(b)
BASF Nederland B.V., 3454 PK De Meern, Netherlands
(c)
BASF SE, 67056 Ludwigshafen, Germany
*
Corresponding author. E-mail: florian.doetz@basf.com
ABSTRACT
We have developed a straightforward synthesis route of MnFePSi spheres via gas atomization yielding material
stocks ranging from Curie temperatures of 250 to 330K. These base materials can be incorporated into highly
stable heat exchanger beds of packed spheres allowing the consequent optimization of system parameters as
porosity, pressure drop, etc. in active regenerator tests. We will show initial results on advanced heat exchanger
structures displaying lower pressure drop and thereby giving room for higher efficiency devices.
Keywords: MnFePSi, gas atomization, Curie temperatures, heat exchanger beds, pressure drop
DOI: 10.18462/iir.thermag.2016.0074
1. INTRODUCTION
In 2015, we have demonstrated the world’s first magnetocaloric domestic appliance at CES Las Vegas [1]. As
magnetic cooling is entering the pre-marketing phase where first prototypes are currently being manufactured,
it is of outmost importance to have a reliable and scalable manufacturing process of magnetocaloric heat
exchanger beds at hand. This includes the industrial synthesis of active magnetocaloric MnFePSi [2] spheres
and the automated manufacturing of large numbers of heat exchanger beds. From an industrial perspective,
cost and performance of the beds are key factors for developing a suitable process route [3].
2. EXPERIMENTAL PROCEDURE
Mn x Fe 2-x P 1-y Si y is synthesized by gas atomization under inert atmosphere, a technique which is readily
available in the industry using pre-alloys/elements as starting materials. For our initial experiments, an atomizer
from Indutherm Gießtechnik was used with an induction furnace that can heat up to 2000°C. Different
atomization batch sizes were tested and validated.
The as-atomized spherical particles were collected from the atomizer and annealed in an Argon atmosphere at
temperatures ranging from 800 to 1200°C for several hours.
Cascades of packed beds were built by sequentially stacking epoxy-modified Mn x Fe 2-x P 1-y Si y layers in a bed
holder whereby each layer represents a certain Curie temperature. The beds were then thermally cured at
temperatures of 100-200°C to yield stable epoxy pucks from magnetocaloric material.
3. RESULTS
The atomization process proved to be a reliable way of manufacturing spherical particles of magnetocaloric
materials of high quality. Figure 1 shows an image of the magnetocaloric powder as-atomized.
28
Figure 1. Atomized magnetocaloric powder.
Table 1 shows the average magnetocaloric properties of the atomized particles after annealing. Note that the
entropy change and adiabatic temperature change are measured at different magnetic fields (1.5T vs. 1.1T).
Property Parameter Value Unit

Cp (0T) Apparent heat capacity peak 1.6-2.0 J/g K
ΔS(1.5T) Entropy change 12-15 J/kg K
ΔT(1.1T) Ad. temp. change 1.8-2.2 K
Table 1. Properties of MnxFe2-xP1-ySiy particles synthesized by gas atomization
The magnetocaloric material is then epoxy-modified and shaped into a solid, porous puck by thermal treatment.
Such a packed bed consists of a cascade of several layers of Mn x Fe 2-x P 1-y Si y material whereby the Curie
temperature of each layer is adjusted by the elemental stoichiometry (Figure 2). The overall temperature range
of the cascade is then spanned up by the coldest layer on one side and the hottest layer on the other side.
Figure 2. Visualization of the layered structure of a cascade.
The process for stacking epoxy-modified Mn x Fe 2-x P 1-y Si y layers was developed in a way that internal structure
throughout the cascade remained homogenous to allow for good fluid flow through the material. Figure 2
shows a CT scan of a packed bed.
29
Figure 3. CT scan of a packed bed of MnxFe2-xP1-ySiy particles
In our effort to increase device efficiency, we aimed at developing bed geometries that display a lower pressure
drop than beds from packed particles. One option with significantly lower pressure loss is a mesh-type structure
which can be produced by 3D printing (as shown in Figure 3).
Figure 4. SEM image of a mesh-type structure of MnxFe2-xP1-ySiy
Several structures have already been produced and are currently being tested for mechanical stability and
magnetocaloric performance. Major challenge will be to achieve good mechanical and magnetocaloric
properties over a broad range of Curie temperatures.
4. CONCLUSIONS
Mn x Fe 2-x P 1-y Si y materials have been produced on industrial scale over a broad range of Curie temperatures by
gas atomization. The spherical particles are converted into stable, solid pucks by epoxy-modifying them and
subsequent thermal treatment. By combining a series of layers with consecutive Curie temperatures, cascades
can be built up with large temperature spans. Latest developments are dedicated to new bed shapes with
reduced pressure drop to increase device efficiency.
5. ACKNOWLEDGEMENTS
The authors would like to acknowledge financial support from the FOM project IPP128.
6. REFERENCES
[1] Press release https://www.basf.com/en/company/news-and-media/news-releases/2015/01/p-15-100.html

[2] O. Tegus, E. Brück, K. H. J. Buschow, F. R. de Boer, “Transition-metal-based magnetic refrigerants for
room temperature applications”, Nature 2002, 415, 150.
[3] F. Scharf, L. Zhang, F. Seeler, F. Dötz, “Magnetic Cooling becoming a hot topic: How novel
magnetocaloric materials pave the way for first applications of energy-efficient ”, Chemie Ingenieur
Technik 2016, 88, No. 5, 553-562.
30
PROPERTIES OF La(Fe, Si)13 MAGNETOCALORIC COMPOUNDS
PREPARED BY REACTIVE SPARK PLASMA SINTERING
V. Paul-Boncour*, A. Patissier, K. Nakouri, L. Bessais
Institut de Chimie et des Matériaux Paris Est, CNRS-UPEC, 2 rue H. Dunant, 94320 Thiais, France
*Corresponding author. E-mail: paulbon@icmpe.cnrs.fr
ABSTRACT
La(Fe, Si)13 magnetocaloric compounds have been synthesized by high energy ball milling (BM) and sintered
by Spark Plasma Sintering (SPS), starting either from an already annealed sample or from a non-annealed
sample (reactive sintering). It is possible to form the NaZn13 phase directly by reactive sintering at only
1273K. The influence of the annealing temperature has been investigated in details as well as other
parameters. The properties of sintered compounds have been compared with those of samples which were
only prepared by BM and annealed.
Intermetallics; mechanical alloying; sintering; magnetocaloric compounds; magnetic measurements

DOI: 10.18462/iir.thermag.2016.0075
1. INTRODUCTION
Development of efficient magnetocaloric materials (MCM) has become a challenge for magnetic
refrigeration near room temperature, since the discovery of a giant magnetocaloric effect (MCE) in
Gd(Ge,Si)5 compounds [1]. Since that time, intensive studies have yielded discovery of several families of
materials, among which the La(Fe,Si)13 system, which displays a giant magnetocaloric effect (MCE) [2]. The
Curie temperatures of these intermetallics which are below room temperature (near 200 K) can be increased
near room temperature by Co for Fe substitution [3] or light element insertion like hydrogen [4,5]. The alloys
can be prepared by arc or induction melting, but an annealing treatment of several days or even weeks is
necessary to obtain single phase LaFe13-xSix compounds with NaZn13 type structure. Other methods like melt
spinning [6] or gas atomization [7] were successful to reduce significantly the time of annealing treatment. In
a previous work [8], we have also developed a method using high energy ball milling (BM) followed by a
heat treatment of only 30 min at 1423 K to obtain almost single phase compound (> 95 wt% of NaZn 13
phase). The magnetic refrigeration application requires also shaping the magnetocaloric material in
appropriate forms in accordance with the design of the magnetic refrigeration device. Several methods have
been already investigated like thermal decomposition and recombination (TDR) process [9] or extrusion after
mixing with a polymer [7]. Spark Plasma Sintering (SPS) method is already used to sinter and shape a
powdered material at an industrial scale as reviewed in [10]. This method is well appropriate to sinter
powders prepared by ball milling and has been already used to sinter different types of MCM. The purpose
of our work, was to synthesize and sinter La(Fe, Si)13 compounds in only one step by using reactive sintering
of BM powder through spark plasma sintering. This concept has been proved in ref [11], but we have
undertaken an optimization of the sample sintering by studying the influence of several external parameters
(temperature, time of sintering, atmosphere).
2.1 Experimental methods

The samples were prepared by ball milling of LaSi precursor with Fe (Alpha Aesar spherical < 10 µm, 99.9
%) and Si powder (bulk Si was crushed and sieved below 100 µm). The mass ratio of powder: ball was 1:14
(5g of powder for 5 balls). The powders were mixed and ball milled during 30 min at 200 rpm and then 1
hour at 600 rpm in a Premium P7 apparatus from Fritsch Company in an argon glove box. After milling, part
of the raw powder was wrapped in Ta sheet, introduced in a silica tube, sealed under argon atmosphere and
annealed 30 min at 1423 K.
About 1 g of powdered sample was pressed into carbon graphite dies of 8 mm inner diameter inside a glove
box under argon atmosphere and transferred to the SPS 515S Syntex machine from Fujii Company. A very
short exposition on air occurred during the transfer of the dies between the pistons, then, the system was
pumped under primary vacuum with intermediate argon purges. During all the heating period, an uniaxial
31
load was applied on the die. The temperature was raised with a rate of 100 K/ min and maintained 5 to 30
min at the maximum temperature before fast cooling.
The X-ray diffraction (XRD) patterns were measured with a Bruker D8 diffractometer using Cu K α radiation
and refined with the Rietveld method running with the fullprof code [12].
The samples sintered at 1273 K were analyzed quantitatively by Electron Probe microanalysis (EPMA) with
a SX 100 Cameca apparatus.
The magnetic properties were measured with a Physical Properties Measurement System (PPMS) from
Quantum Design (1.5 < T < 320 K, Bmax= 9 T). The magnetic entropy variation SM was obtained using the
Maxwell relation as described in [8].
2.2 Results
The piston displacement and the displacement speed of a BM LaFe11.5Si1.5 sample sintered by SPS which was
either pre-annealed at 1423 K (A-SPS) (left) or non-annealed (NA-SPS) (right) are compared in Figure 1.
When the BM powder is pre-annealed at 1423 K the piston displacement starts to increase at 1000±10 K and
the sintering is achieved at 1273 K. The speed of displacement shows a single broad peak with a maximum
around 1170±20 K. For a non-annealed powder, the sintering starts around 773±20 K and is ended at 1173
K. In addition the derivative of the displacement shows two maxima around 903±20 K and 970±30 K. To
identify the phase transformation versus temperature, the XRD patterns of NA-SPS samples sintered
between 873 and 1423 K have been registered (Figure 2) and compared to those of powders annealed at
different temperatures but not sintered (A). The sintering of a pre-annealed sample does not change the
weight percentage of the phases once NaZn13 phase has been formed. The XRD patterns were refined in
order to obtain the nature of the phases, their weight percentages and units cell parameters; the main results
are reported in Table 1. Some peaks of carbon, due to the graphite die were observed at low angle for the NA
SPS samples. The pattern of the sample annealed at 873 K is similar to the initial BM powder with a large Fe
peak, and some broad contribution attributed to amorphous phases. At the same sintering temperature a small
amount of NaZn13 and LaFeSi phases are already observed in NA-SPS sample. Upon increasing temperature,
the percentage of NaZn13 phase obtained either by annealing or reactive sintering increases. By SPS a new
increase of the piston displacement is observed above 1373 K, and a redecomposition of the NaZn 13 phase
into Fe is observed. This has been explained by a partial melting of the NaZn13 phase, which is not observed
when the powder is simply annealed. The optimum sintering temperature is therefore 1273 K for reactive
sintering, this temperature is lower than the 1423 K temperature optimized for BM samples [8].
0.4 0.3
Annealed BM LaFe11.5Si1.5 sample Non annealed BM LaFe11.5Si1.5 sample
0.3 1.5
Piston displacement (mm)
1.0
Piston displacement (mm)
0.2
Speed (mm/s)
0.2
Speed (mm/s)
1.0
0.5 0.1
0.1
0.0 0.5
0.0
-0.1 0.0
0.0
400 600 800 1000 1200
400 600 800 1000 1200 1400
T (K)
T (K)
Figure 1: Piston displacement and displacement speed measured during the sintering by SPS versus temperature for an
annealed at 1423 K (left) and a non annealed (right) BM LaFe11.5Si1.5 compound. The speed corresponds to the
derivative of the piston displacement versus time; the curves were smoothed for clarity.
These results confirm that the NaZn13 phase can be obtained at a lower temperature using reactive sintering
compared to classical annealing. In addition the weight percentage of NaZn13 phase can be slightly increased
either by a longer plateau (15 min instead of 5 min) and by working under argon (Table 1). These results
confirm that the NaZn13 phase is formed by a solid state reaction between -(Fe,Si) and amorphous LaSi
phase to form first LaFeSi and then La(Fe,Si)13 through a diffusion mechanism.
32
-Fe * *
NaZn13 ° La2O3 + LaFeSi Fe,Si NaZn13
1393 K 1393 K
1273 K
1273 K
Intensity (a. u.)

Intensity (a.u.)
1223 K
1173 K
° °° 1123 K
1073 K
° °
*
973 K 973 K
* * *
* *
873 K 873 K
* * + +
* * + +
40 60 80 100 120 40 60 80 100 120

2 (°) 2 (°)
Figure 2: XRD patterns of BM LaFe11.5Si1.5 samples (left) annealed in Si tube under argon atmosphere at the indicated
temperature and (right) sintered by SPS under vacuum and with a 5 min plateau at the indicated temperature.
Table 1: XRD results on annealed (A) and sintered (NA and A SPS) samples.
Sample Method atm. t (min) Tmax (K) NaZn13 Fe-Si LaFeSi La2O3
at Tmax
a (Å) wt% wt% wt% wt%
LaFe11.5Si1.5 A argon 30 873 11.459 100
LaFe11.5Si1.5 A argon 30 973 11.459 12.6 76.9 10.5
LaFe11.5Si1.5 A argon 30 1023 11.470 59.4 31.5 9.1
LaFe11.5Si1.5 A argon 30 1123 11.469 61.2 29.5 9.3
LaFe11.5Si1.5 A argon 30 1223 11.467 64.3 27.9 7.8
LaFe11.5Si1.5 A argon 30 1273 11.468 69.9 23.2 6.9
LaFe11.5Si1.5 A argon 30 1393 11.475 98.2 0.6 1.2
LaFe11.5Si1.5 NA-SPS vacuum 5 873 11.499 3.0 84 13.0
LaFe11.5Si1.5 NA-SPS vacuum 5 973 11.468 16.5 72.8 10.7
LaFe11.5Si1.5 NA-SPS vacuum 5 1073 11.503 75.4 24.6
LaFe11.5Si1.5 NA-SPS vacuum 5 1173 11.493 86.3 13.7
LaFe11.5Si1.5 NA-SPS vacuum 5 1273 11.497 92 6 2
LaFe11.5Si1.5 NA-SPS vacuum 5 1423 11.504 70 30
LaFe11.5Si1.5 NA-SPS vacuum 15 1273 11.489 95 5
LaFe11.5Si1.5 NA-SPS argon 5 1273 11.491 96 4
LaFe11.5Si1.5 NA-SPS argon 15 1273 11.491 97 3
LaFe11.5Si1.5 A-SPS vacuum 5 1273 11.497 89.7 6.9 3.4
The magnetic properties of samples prepared by reactive sintering have been investigated. The Curie
temperature for a field of 0.03T of the NA-SPS samples prepared under argon atmosphere was 221± for an
annealing plateau of 5 min and 210 K for a plateau of 15 min. The thermal hysteresis was around 1K,
whereas the isotherm magnetization curves present almost no hysteresis, although a first order transition is
still observed. The corresponding M(H) curves for a 5 min plateau is reported in Figure 3, as well as the
magnetic entropy variation calculated for different fields. The maximum entropy | SM max|=13 J/ K kg for
H=5T for this sample. For the sample sintered with a 15 min plateau a larger magnetic entropy variation is
observed with| SM max| =17 J/ K kg for H=5T. This last value is close to that obtained for a melted sample
annealed during one month (17.8 J/ K kg). The differences between the samples sintered during 5 and 15 min
can be related to the sample homogeneity and the Si content. Indeed, a lower Curie temperature and a higher
value of | SM max| were obtained for samples with a lower Si content [8]. The influence of carbon diffusion
inside the sample should also be clarified, as it can increase T C and decrease the magnetic entropy variation.
Therefore, further studies should be performed to protect the powder from carbon contamination.
33
NA-SPS 1273 K 5min under Ar
14
20 H (T)
180 K
12 1
2
10 3
15 T=2K 4
- SM (J/kg.K)
8 5
M (µB)
10 6
4
206 K
5 2
0
0 180 190 200 210 220 230 240 250
0 1 2 3 4 5 H (T)
H (T)
Figure 3: Evolution of the M(H) curves of the sample NA-SPS sintered at 1273 K with a 5 min plateau and
corresponding magnetic entropy variation.
3. CONCLUSIONS
This study has shown that it is possible to prepare sintered magnetocaloric La(Fe, Si)13 compounds with
NaZn13 structure by reactive sintering of BM powder at 1273 K in 15 to 30 min. The compounds formation is
achieved at a lower temperature than by a classical annealing treatment of the powder. The sintering of
La(Fe,Co,Si)13 compounds with various Co contents to have TC around 300 K is under study.
ACKNOWLEDGMENTS
Some XRD results on annealed powder were obtained during the PhD thesis of M. Phejar. We are thankful
to B. Villeroy for his help to perform Spark Plasma Sintering experiments.
REFERENCES
[1] V.K. Pecharsky, K.A. Gschneidner, Giant magnetocaloric effect of Gd 5(Si2Ge2), Phys. Rev. Lett. 78,
4494 (1997).
[2] S. Fujieda, A. Fujita, K. Fukamichi, Large magnetocaloric effect in La(Fe xSi1-x)13 itinerant-electron
metamagnetic compounds, Appl. Phys. Lett. 81, 1276 (2002).
[3] X.B. Liu, X.D. Liu, Z. Altounian, Phase formation and magnetocaloric effect in rapidly quenched La(Fe1-
xCox)11.4Si1.6, J. Appl. Phys. 98, 113904 (2005).
[4] S. Fujieda, A. Fujita, K. Fukamichi et al. , Giant isotropic magnetostriction of itinerant-electron
metamagnetic La(Fe0.88Si0.12)13Hy, Appl. Phys. Lett. 79, 653 (2001).
[5] M. Phejar, V. Paul-Boncour, L. Bessais, Investigation on structural and magnetocaloric properties of
LaFe13-xSix(H,C)y compounds, J. Sol. State Chem. 233, 95 (2016).
[6] A. Yan, Structure and magnetocaloric effect in melt-spun La (Fe, Si)13 and MnFePGe compounds, Rare
Metals 25, 544 (2006).
[7] C. Mayer, A. Dubrez, M. Pierronnet et al. , Towards the large scale production of (La 1-zCez)(Fe1-x-
yMnySix)(13)H-n products for room temperature refrigeration, Physica Status Solidi C. 11, 1059 (2014).
[8] M. Phejar, V. Paul-Boncour, L. Bessais, Structural and magnetic properties of magnetocaloric LaFe 13-xSix
compounds synthesized by high energy ball-milling, Intermetallics 18, 2301 (2010).
[9] M. Katter, V. Zellmann, A. Barcza in proceedings of Fourth IIF-IIR International conference on
Magnetic Refrigeration at room temperature, Thermag IV, (Baotou, China), (2010).
[10] R. Orrù, R. Licheri, A. Locci et al. , Consolidation/ Synthesis of materials by current activates/assisted
sintering, Materials Science and Engineering 63, 127 (2009).
[11] A. Patissier, V. Paul-Boncour, Fast synthesis of LaFe13-xSix magnetocaloric compounds by reactive
Spark Plasma Sintering, J. Alloys Compd. 645, 143 (2015).
[12] J. Rodríguez-Carvajal in Abstract of the Satellite Meeting on Powder Diffraction of the XVth Congress
of the International Union of Crystallography, (Toulouse, France), (1990).
34
PERFORMANCE ANALYSIS OF AN ACTIVE MAGNETIC
REFRIGERATION SYSTEM WITH EXTERNAL LOSS MECHANISMS
I. Niknia*, P. V. Trevizoli, P. Govindappa, O. Campbell, T. V. Christiaanse, R. Teyber,
A. Rowe
Institute for Integrated Energy Systems (IESVic),
University of Victoria, Victoria, BC, Canada
*
Corresponding author. E-mail: iniknia@uvic.ca
ABSTRACT
In this study, a one dimensional, time dependent model is used to study the performance of a magnetic
refrigeration system. The objective is to quantify system losses and identify the operating conditions that
maximize the performance. Parameters related to configuration such as external heat leaks and
demagnetization effects are included. A simple resistance network model is used to simulate external losses.
Through experiments, loss coefficient factors are estimated to determine conductances linking the cold side
of the system to the environment and to the hot side. It is found experimental results on AMRs may
underestimate intrinsic regenerator performance if losses are not considered. A detailed numerical study is
performed to investigate the impacts of operating temperature on the performance of the system. Results
indicate that the system efficiency is maximized when the average temperature of the matrix is slightly
higher than the Curie point of the refrigerant.
Keywords: AMR, Transient simulation, Magnetic refrigeration, Magnetocaloric effect, External losses.
DOI: 10.18462/iir.thermag.2016.0076
1. INTRODUCTION
The main challenges in developing active magnetic regenerator (AMR) devices can be classified in two
categories: material and system design [1]. From the system design point of view, an active magnetic device
consists of several parts besides the regenerators themselves. Many different designs exist and design choices
tend to weight some losses higher than others. Part of the engineering challenge is to quantify the trade-offs
for a system configuration and to optimize the system as a whole. Depending upon the operating temperature
span of the regenerator, the aspect ratio, and the design of the casing holding the regenerator, heat leaks
through the surrounding structure can lead to decreases or increases in performance. While not often
considered, imperfect thermal isolation may actually help a device obtain a larger temperature span when
operating above the environmental temperature. Other configuration losses can arise from dead volumes and
heat exchanger ineffectiveness [1]. In this paper, a one dimensional transient model is used to study the
performance of an AMR system with external losses. External losses are mechanisms resulting in imperfect
thermal isolation of the regenerator and temperature reservoirs. The term configuration losses is also used as
the external losses are a function of device configuration. Experimental measurements are used to validate
the model. The impacts of losses on the temperature span and cooling power of an AMR are reported. The
impacts of operating temperature on the performance of the system are investigated.
2. METHODS
The energy balance equations for a porous regenerative matrix for fluid and solid phases are defined as
follows:
mf Lcp T f T f    Ak f ,eff T f  m p
 mcp  *  * 
 heff Aw L Ts  T f   (1)
B t *
x *
x  L x   f x*
ms LcB Ts   1    Aks ,eff Ts  ms LcB Tad


 B t * x* 

L

x* 
 heff A
w L  T f  Ts  
 B t *
 L
 Qother (2)
35
where T is temperature, x* is the dimensionless spatial coordinate (normalized by regenerator length, x = x*L)
and t* is dimensionless time coordinate (normalized by blow period, t = t*τB). Subscript f refers to fluid and
subscript s refers to solid. mf’ is the entrained mass of fluid per unit length of the regenerator ms’ is the mass
of MCM per unit length, c is specific heat, Aw’ is the surface area of the solid matrix per unit length, τB is the
blow period, keff and heff are the effective thermal conductivity and convective heat transfer coefficient
respectively, A is the cross-sectional area of the regenerator, and, α is porosity. In Eq. (1) the terms from left
to right represent: energy storage, enthalpy flow by fluid displacement, thermal diffusion, convective transfer
between the solid and the fluid, and viscous dissipation. The terms in Eq. (2) are: energy storage in the solid,
axial heat conduction, convective heat transfer between solid and fluid, and magnetic work effects. Eq. (2)
includes a generic source term, Qother’ which is zero for the current study. The convection coefficient is
determined using the correlation suggested by Wakao and Kaguei [2]. The effective value is determined by a
degradation factor, Df, as described elsewhere [3-4]. For the fluid, both conduction and dispersion are
considered in the diffusion term such that for a one dimensional analysis, keff is given by [5-6]. Pressure drop
is determined using the modified Ergun equation as suggested by Macdonald [7].
The governing equations are discretized in space using Galerkin-Petrov method with second order accuracy
in space [8]. This is done using PDEPE function in Matlab. ODE15 suit is used to integrate the governing
equations in time [9].
A simple resistance network model is used to simulate external losses [10]. The heat leaks: (i) between the
cold side of the regenerator and the environment, (ii) between the warm and cold sides of the system are
included in the model by using simple thermal resistor method [11]. The magnitudes of these thermal
interactions are assumed to be determined by the specific thermal resistances in the device. The heat leak,
Q0C, between the cold end at TC and the ambient at T0 is given by,
Q0C  K0C T0  TC  (3)
where K0C is the effective thermal conductance. Likewise, the thermal interaction between the warm side of
the system and the cold side is determined by the conductance KHC,
QHC  K HC TH  TC  . (4)
The external heat leaks reduce the available net cooling power of the system (i.e the applied load), QNET,
QC  QNET  Q0C  QHC (5)
where QC is the gross cooling power calculated based on the cycle averaged enthalpy flow at the cold end of
the AMR.
3. PERFORMANCE ANALYSIS
Experimental measurements are carried out to validate the performance of the numerical model. The
experimental apparatus used in this study is PM II device, at the University of Victoria. The device has been
described in detail elsewhere [12]. The loss coefficient factors are calculated through a series of experimental
measurements from PM II device. For each set of experiments, the applied load and the hot side temperature
of the device are changed and steady state measurements are done when the device is not running. These data
are used to estimate the loss coefficients of the system and are included in the numerical model.
Performance of the model is compared against experimental measurements for several cases. Details of the
numerical simulations, tolerance, meshing and resulting fit between model and experiment are described in
detail elsewhere [10]. Simulations of experiments show that when the impacts of external losses are
neglected, the model over predicts the temperature span significantly, however, after including the loss
factors, the model predictions are in good agreement with the experimental measurements.
After validation, the model is used to study the performance under a broader range of operating conditions.
Fig. 1 shows the cooling power as a function of temperature span. The solid lines represent cases where the
external losses are neglected and the dashed and dotted lines represent the cooling power when the impacts
of external losses are included in the model. At higher temperature spans, where, the cooling power becomes
36
smaller, the impacts of external losses become more significant. For a fixed temperature span, decreasing the
frequency, also leads to lower cooling power and therefore, can increases the impacts of external losses on
the performance.
90
80 TH=307
frequency
70 1.0
Cooling Power [W]

60
2.0
50
3.0
40
30
20
10
0
0 10 20 30
Temperature Span [K]
Figure 1. Cooling power as a function of temperature span. The solid lines represent gross cooling power and the
dashed/dotted lines show the net cooling power (after including external losses effects).
To investigate the impacts of operating temperature on the performance of the system, several hundred
numerical simulations were performed. Table 1 shows the range of parameters used and the increments
between points.
Table 1 different operating conditions studied by the presented model. δ shows the increment used for each variable.
Material f [Hz] Hot Temperature [K] Cold Temperature [K] Displaced volume [cm3 ]
Gd 0.5-3 (δ = 0.5) 307-295 (δ = -4) 278-290 (δ = 2) 5.2-31.2 (δ = 5.2)
The efficiency at each operating point is calculated using,

Q  TH 
   1 . (6)
WM  WP  TC 
where Q is the cooling power, WM is the magnetic work and Wp is the pumping work. The denominator
(WM+Wp) is calculated using an energy balance on the AMR, QH – QC. For select hot side temperatures
(Th=307K, Th=303K) and various operating conditions (Table 1), using gross and net cooling power in Eq. 6,
efficiency is calculated and plotted versus gross and net cooling power (Fig. 2 (a) and (b) respectively). The
operating points are color coded to indicate the average temperature of the regenerator for each point.
(a) (b)
100 100
293.5 294.5 293.5 294.5
90 295.5 296.5 90 295.5 296.5
80 297.5 298.5 80 297.5 298.5
70 291.5 292.5 70 291.5 292.5
293.5 294.5 293.5 294.5
60 295.5 296.5 60 295.5 296.5
50 50
η
○ Th=307K ○ Th=307K
40 □ Th=303K 40 □ Th=303K
30 30
20 20
10 10
0 0
0 50 100 0 50 100
QGross [W] QNet [W]
Figure 2. (a) Efficiency versus gross and (b) efficiency versus net cooling power for different operating points. The
operating points are color coded to indicate the average (mean) temperature of the regenerator for each operating point.
37
In Fig. 2 (a), the efficiency is calculated using the gross cooling power and in Fig. 2 (b) the efficiency is
calculated using the net cooling power. The results show that the highest cooling power and efficiency can
be achieved when the mean material operating temperature is between 295 and 297 K which is slightly
higher than Curie temperature for Gd. Comparison of Fig. 2 (a) and (b), indicate that the maximum
achievable efficiency is highly impacted by the loss effects. As the cooling power decreases, the impacts of
losses on efficiency increase. This does not seem to impact the operating temperature for optimum efficiency
(i.e. for both cases, whether losses are included or not, the system works at maximum performance when the
operating temperature is slightly higher than the curie temperature of the refrigerant).
4. CONCLUSIONS
Device losses are important parameters that are usually neglected in performance studies. Although these
losses may be neglected in large scale systems when cooling power is significant, they should not be
overlooked when using small scale laboratory prototypes to characterize AMRs and the magnetocaloric
materials used. Numerical simulations of a test apparatus where external heat leaks are included show that
the effects of device losses should not be overlooked. This is particularly important for low applied loads
where the maximum temperature span is found. At large temperature spans the cooling power will approach
zero and device losses become more significant. It is also observed that the system operates at highest
efficiency when the mean temperature of the matrix is slightly higher than the Curie point of the refrigerant.
Losses have minimal impacts on the temperature where the system operates at highest efficiency.
ACKNOWLEDGMENTS
The support of the Natural Sciences and Engineering Research Council of Canada and BASF New Business
is greatly appreciated.
REFERENCES
[1] P. V. Trevizoli, T. V. Chistiaanse, P. Govindappa, I. Niknia, R. Teyber, J. R. Barbosa, and A. Rowe,

“Magnetic heat pumps: an overview of design principles and challenges,” Sci. Technol. Built
Environ., p. In press, 2016, DOI: 10.1080/23744731.2016.1171632.
[2] N. Wakao and S. Kaguei, Heat and Mass Transfer in Packed Beds. New York, NY: Gordon and
Breach Science Publishers, 1982.
[3] K. Engelbrecht, “A Numerical Model of an Active Magnetic Regenerator Refrigerator with
Experimental Validation by,” University of Wisconsin-Madison, 2008.
[4] K. L. Engelbrecht, G. F. Nellis, and S. a. Klein, “The Effect of Internal Temperature Gradients on
Regenerator Matrix Performance,” J. Heat Transfer, vol. 128, no. 10, pp. 1060–1069, 2006.
[5] M. Kaviany, Principles of Heat Transfer in Porous Media. New York, NY, 1995.
[6] T. Burdyny and A. Rowe, “Simplified modeling of active magnetic regenerators,” Int. J. Refrig., vol.
36, no. 3, pp. 932–940, May 2013.
[7] I. Macdonald, “Flow through porous media-The Ergun equation revisited,” Ind. Eng. Chem. Fundam.,
vol. 18, no. 3, pp. 199–208, 1979.
[8] R. D. Skeel and M. Berzins, “A Method for the Spatial Discretization of Parabolic Equations in One
Space Variable,” SIAM J. Sci. Stat. Comput., vol. 11, no. 1, pp. 1–32, Jan. 1990.
[9] L. F. Shampine and M. W. Reichelt, “The MATLAB ODE Suite,” SIAM J. Sci. Comput., vol. 18, no.
1, pp. 1–22, Jan. 1997.
[10] I. Niknia, O. Campbell, T. V. Christianse, P. Govindappa, R. Teyber, P. V. Trevizoli, and A. Rowe,
“Impacts of configuration losses on active magnetic regenerator device performance.,” Appl. Therm.
Eng., vol. 106, pp. 601–612, 2016.
[11] F. P. Incropera and D. P. DeWitt, Fundamentals of Heat and Mass Transfer. New York, 2002.
[12] D. S. Arnold, A. Tura, A. Ruebsaat-Trott, and A. Rowe, “Design improvements of a permanent
magnet active magnetic refrigerator,” Int. J. Refrig., vol. 37, no. January 2014, pp. 99–105, Oct. 2014.
38
A MATERIAL SCREENING TECHNIQUE FOR OPTIMUM
PERFORMANCE OF AN AMR
I. Niknia*, P.V. Trevizoli, P. Govindappa, O. Campbell, T.V. Christiaanse, R. Teyber,
A. Rowe
(a)
*
Corresponding author: iniknia@uvic.ca
ABSTRACT
The selection of an optimum magnetocaloric material is a challenge in the development of magnetic

refrigeration systems. In this study, two MnFeP1-xAsx alloys with the same transition temperature, but
different specific heat and adiabatic temperature change are investigated. The two alloys are used to create
active regenerators and subsequently tested, experimentally and numerically, over a wide range of operating
parameters. The results show that the material with smaller peak and wider magnetic entropy change curve,
outperforms the material with larger peak but narrower magnetic entropy change curve. To further
investigate impacts of material properties on performance, a range of material metrics are compared against
simulated performance. A strong linear correlation is observed between the maximum RCP and the
maximum exergetic cooling power of 12 materials tested in this study.
Keywords: Magnetocaloric material, AMR, Transient simulation, Magnetocaloric effect, Material screening.
DOI: 10.18462/iir.thermag.2016.0077
1. INTRODUCTION
Magnetocaloric materials can be classified into two types based on the transition behaviour from
ferromagnetic to paramagnetic phase: second order materials (SOM) and first order materials (FOM). SOMs
tend to order over a wider temperature range but they can be expensive and not always suitable for room
temperature applications. FOMs use readily available constituent elements; however, their useful
magnetocaloric effect is limited to a narrower operating temperature range. Other challenges exist with
FOMs such as a hysteresis [1]. For both FOMs and SOMs, the effective magnetocaloric effect (MCE)
declines as the operating temperature deviates from the ordering temperature. Layering materials with
different phase transition temperatures in an active magnetic regenerator (AMR) is one way to overcome this
problem and create a wider operating span [2]. In this paper, numerical and experimental measurements are
combined to investigate the magnetic properties of a material and how these properties impact AMR
performance.
2. SCREENING METRICS
Selecting materials for application in an AMR is a crucial step in designing high performance systems.
Magnetocaloric materials (MCM) are usually characterized by measuring their specific heat, cp, conductivity,
k, density, ρ, adiabatic temperature change, ΔTad. In addition, magnetization, M, can be used to estimate
adiabatic temperature change and magnetic entropy change, ΔSM [3]. From the early works of Wood and
Potter [4] to more recent study of Aprea et al. [5] the impact of adiabatic temperature change and magnetic
entropy change on the potential of a MCM for AMR applications has been investigated. Wood and Potter
define the refrigerant capacity (RC) as the reversible work to operate between Tcold and Thot for isothermal
heat absorption at Tcold, i.e.:
RC  Tcold S Tcold   Thot  Tcold  (1)
A modified definition was suggested as a useful parameter to measure cooling capacity q, [6]:
Thot
q   S (T , P, H ) p ,H dT . (2)
Tcold
39
Other metrics used to rank materials include peak entropy change SM ax , peak adiabatic temperature change
TMax , and combinations weighted by temperature span. Examples of the latter are RCP(s) and RCP(T) [6][7]
where,
RCP( S )  SM ,max   TFWHM
(3)
RCP(T )  Tad ,max   TFWHM
As discussed by Sandeman [8], a maximum RCP can be defined according to,


RCPmax   S (T )dT (4)
0
which is equivalent to MsatHmax. Although all of these parameters can be used to screen material for use in
magnetic refrigeration, it is not clear which parameter is a better metric to characterize a MCM. There has
been no rigorous study to determine how these metrics correlate with performance in an AMR cycle. In this
paper, we experimentally measure the performance of a material system composed of the same constituents,
but varying magnetocaloric properties. We then examine the metrics discussed above against AMR
performance using simulations of various hypothetical materials. Finally, the experimental data is used to
validate the expected performance based on suitable metrics.
3. EXPERIMENTAL MEASUREMENTS
Two samples of MnFeP1-xAsx materials with the same transition temperature and different adiabatic
temperature change and magnetic entropy change are identified (Fig.1 and Table 1). The properties of these
two materials are measured through VSM and DSC measurements. Material 1 shows a larger peak in
adiabatic temperature change, and magnetic entropy change. On the other hand, material 2 has a wider curve
with a smaller peak. The transition temperature and hysteresis for these two material are similar. Table 1
summarizes the transition temperatures for the warming curves at zero field as well as 1 and 1.5 Tesla. RCP
in terms of temperature and entropy are listed as well as the maximum magnetic entropy change and
adiabatic temperature change for each field.
Two sets of AMRs are constructed from these materials and their performance is compared through
experimental measurements. The experimental tests are carried out in PM II (The device has been described
in [9]). The performance of the two selected material are compared for different frequencies (0.5 and 0.7
Hz), displaced volumes (3.47, 6.95, 10.42 cm3) and heat rejection temperatures (286-304 K with increment
of 2K). A water-glycol mixture (80%-20%) on volumetric basis is used as heat transfer fluid.
(a) (b)
2.4 3.5
Material 1 Low field, B=0 [T] Material No. 1
Adiabatic Temperature Change [K]
2.2 Material 1 high field, B=1 [T] 3 Material No. 2

2 Material 2 Low field, B=0 [T]
1.8 Material 2 high field, B=1 [T] 2.5
Cp [J (Kg K)-1]
1.6 2
1.4
1.5
1.2
1 1
0.8
0.5
0.6
0.4 0
280 285 290 295 300 270 290 310
Temperature [K] Temperature [K]
Figure 1. Measured material properties (adiabatic temperature change and specific heat) for MnFeP 1-xAsx samples.
Specific heat values are presented for warming curves.
40
Table 1 measured material properties for MnFeP1-xAsx
Field Tc FWHM Hysteresis ΔS ΔTad RCP(T) RCP(S)

[T] [K] [K] [K] [J (kg K)-1] [K] [K2] [J kg-1]
Material 1 0.0 287.9 2.96 1.3
1.0 292.8 2.98 1.4 12.74 3.23 9.83 37.96
1.5 295.1 2.98 1.3 14.50 4.65 13.85 43.21
Material 2 0.0 288.6 6.72 1.3
1.0 292.9 6.54 1 7.48 2.24 14.64 48.91
1.5 295.7 6.63 1.5 10.29 3.35 21.90 68.22
4. NUMERICAL MODEL
Because the range of experimental measurements are limited by the operating parameter of test apparatus, a
numerical model is used to study the two samples over a wider range of operating conditions. In the current
study, a 1D model is used. The details of the model and validation are described elsewhere [10]. The model
is also employed to investigate the various screening metrics used for MCMs. In order to focus on the
impacts of the magnetocaloric properties, external loss effects are neglected, fluid properties, regenerator
volume, particle size, and porosity are fixed. The regenerator cooling power is calculated based on the cycle
averaged enthalpy flow at the cold end neglecting pressure variations. The exergetic equivalent cooling
power, ExQ is used to quantify performance [11],
T 
ExQ  Qc  hot  1 . (5)
 Tcold 
5. PERFORMANCE ANALYSIS
The performance of two identified samples are experimentally measured for various operating parameters
(such as displaced volumes, frequency and applied loads). For different operating conditions, the temperature
span and cooling power produced by each of the two samples are measured and compared. Fig. 2 (a) shows
an example of such comparisons for frequency of 0.7 Hz, displaced volume of 6.95 cm3 and zero and 10
watts of applied load. It is observed that for all the experiments, material 2 which has a wider curve and
shorter peak, produces higher exergetic cooling power compared to material 1 (i.e for the same cooling
power, heat rejection temperature and operating conditions, material 2 shows a larger maximum temperature
span). According to the metrics presented in (Table 1), there is not a clear picture of which material should
give the highest exergetic power (i.e material 1 has larger ΔTad and ΔSM but according to experimental
measurements, produces smaller exergetic power).
(b)
(a)
14 300
Material 1, load=0 [W]
12 Material 1, Load=10 [W]
Temperature span[K]
250
Material 2, Load=0 [W]
10 Material 2 Load=10 [W]
RCPmax [J kg-1]
200
Vd =6.95 cm3
8 f =0.7 Hz
150
6
100
4
y = 35.319x + 43.612
2 50 R² = 0.9969
0 0
283 293 303 0 5 10
Hot side temperature [K] Maximum ExQ
Figure 2- (a) temperature span as a function of hot side temperature for MnFeP1-xAsxsamples (experimental
measurements), (b) cooling potential versus the maximum exergetic cooling power, for different FOM and SOM
samples (Numerical modeling results).
41
In order to further investigate the accuracy of the identified metrics, the performance of 12 additional
samples of FOM and SOM material are compared through numerical simulations. The magnetization,
specific heat, and entropy of the identified SOM material (e.g. Gd, Dy) are modeled by combining mean
field theory (MFT) and Sommerfeld Model [12] with the Debye approximation [13]. For the FOM cases,
specific heat is modeled using a Lorentzian curve and integrated to determine isofield entropy curves. Low
and high field entropy are then interpolated to determine adiabatic temperature change as a function of
temperature. The two approaches to creating material properties ensure thermodynamic consistency in the
key magnetocaloric data. Whether or not the modeled materials represent an actual material is irrelevant, as
we are only interested in creating a range of thermodynamically consistent data for specific heat, entropy,
and adiabatic temperature change. For each case, a wide range of operating conditions are modeled and the
point of maximum exergetic cooling power is identified.
The maximum exergetic power of each material (including the two experimental samples and, identified
FOM and SOM materials) is plotted against RCPmax (Fig. 2 (b)). The results, show that the maximum
exergetic cooling power of each material, is linearly correlated to RCPmax, however, such strong correlation
was not observed for some of the widely used screening metrics such as RCP(T) and RCP(S).
6. CONCLUSIONS
In this study, performance of two MnFeP1-xAsx samples with similar transition temperature and different heat
capacity and adiabatic temperature change are compared. The experimental measurements, indicated that the
sample with a wider curve and shorter peak for ΔTad and ΔSH, outperforms the material 1 which has a larger
peak and narrower property curves. Twelve additional cases of FOM and SOM material are identified and
their performances are compared through numerical simulations. RCPmax is proposed as a material screening
technique. It is observed that the maximum exergetic power of all the tested material have a linear
correlation with the proposed metric. Such linearity was not observed for conventional screening techniques.
ACKNOWLEDGMENTS
REFERENCES
[1] A. Smith et al., “Material challenges for high performance magnetocaloric refrigeration devices,” Adv. energy
Mater., vol. 2, pp. 1288–1318, 2012.
[2] A. Rowe and A. Tura, “Experimental investigation of a three-material layered active magnetic regenerator,” Int.
J. Refrig., vol. 29, no. 8, pp. 1286–1293, Dec. 2006.
[3] A. Kitanovski and P. W. Egolf, “Thermodynamics of magnetic refrigeration,” Int. J. Refrig., vol. 29, no. 1, pp.
3–21, Jan. 2006.
[4] M. E. Wood and W. H. Potter, “General analysis of magnetic refrigeration and its optimization using a new
concept: maximization of refrigerant capacity,” Cryogenics., vol. 25, no. 12, pp. 667–683, 1985.
[5] C. Aprea, A. Greco, A. Maiorino, and C. Masselli, “A comparison between rare earth and transition metals
working as magnetic materials in an AMR refrigerator in the room temperature range,” Appl. Therm. Eng., vol.
91, pp. 767–777, 2015.
[6] K. a Gschneidner and V. K. Pecharsky, “Magnetocaloric materials,” Annu. Rev. Mater. Sci., vol. 30, pp. 387–
429, 2000.
[7] V. Franco et al., “The magnetocaloric effect and magnetic refrigeration near room temperature: materials and
models,” Annu. Rev. Mater. Res., vol. 42, no. 1, pp. 305–342, 2012.
[8] K. G. Sandeman, “Magnetocaloric materials: The search for new systems,” Scr. Mater., vol. 67, no. 6, pp. 566–
571, 2012.
[9] D. Arnold, “Design Principles and Performance Metrics for Magnetic Refrigerators Operating Near Room
Temperature by,” University of Victoria, 2014.
[10] I. Niknia, O. Campbell, T. V. Christianse, P. Govindappa, R. Teyber, P. V. Trevizoli, and A. Rowe, “Impacts of
configuration losses on active magnetic regenerator device performance.,” Appl. Therm. Eng. Under Rev.
[11] A. Rowe, “Configuration and performance analysis of magnetic refrigerators,” Int. J. Refrig., vol. 34, no. 1, pp.
168–177, 2011.
[12] C. Kittel, Introduction to Solid State Physics. New York: John Wiley & Sons Inc., 2005.
[13] A. M. Tishin and Y. I. Spichkin, The Magnetocaloric Effect and its Applications. CRC press, 2003.
42
EFFECTS OF MAGNETOCALORIC WIRE
ON INCREASE IN MAGNETIC REFRIGERATION CYCLE
K. Ueno(a)*, M. Kondo(a), K. Takeuchi(a), R. Nomura(a), T. Kizaki(a)

(a)
Fujikura Ltd./1440,Mutsuzaki,Sakura-shi (Chiba,Japan)
*
Corresponding author: kota.ueno@jp.fujikura.com
ABSTRACT
The phase-out of chlorofluorocarbon (CFC) and hydrochlorofluorocarbon (HCFC) refrigerants has recently
been underway with greater awareness of the environment. Given this background, magnetocaloric
refrigeration technology is considered a candidate among new refrigeration technologies that comply with
regulations regarding the substances. Magnetic refrigeration systems are more efficient in theory than vapor
compression type refrigeration systems and are free of CFC and HCFC refrigerants. The purpose of this
study was to increase the cooling power of magnetocaloric refrigerators by increasing the cycle frequency
and put them to practical use. Although other researchers reported that their experiments were conducted at a
frequency of several Hz, we assume that a frequency greater than 10 Hz is required to gain a power output of
several kW. However, when the cycle frequency is set to a higher value, the pressure losses in the flow paths
increase, resulting in insufficient cooling power to transfer heat efficiently. Consequently, we have
developed a wire-shaped magnetocaloric material (MCM) using gadolinium (Gd). The configuration of wires
has been considered suitable to reduce pressure losses and secure a large contact area in an active magnetic
regenerator (AMR). However, working the MCM has been considered difficult. The performance of the
wire-shaped MCM was measured using an experimental device. The result showed a specific cooling power
of 300 W/kg at 10.0 Hz. This result was four times larger than that of Gd particles.
Keyword: Magnetocaloric material, Twisted wire, High frequency, Gadolinum, AMR

DOI: 10.18462/iir.thermag.2016.0078
1. INTRODUCTION
In the conventional refrigeration technology for air conditioners and refrigerators, the refrigerant fluids often
contain chlorofluorocarbons (CFCs) or hydrochlorofluorocarbons (HCFCs). On the other hand, the phase-out
of these substances has been underway with greater awareness of the environment [1]. Consequently, new
refrigeration technologies are required to comply with regulations regarding the substances [2]. One of the
candidates is magnetocaloric refrigeration technology. Magnetic refrigeration systems are more efficient in
theory than vapor compression type refrigeration systems and are free of CFC and HCFC refrigerants [3].
The wine cooler manufactured by Astronautics [4] and the medical refrigerator manufactured by Cooltech
[5] have been announced in recent years as actual application examples of the magnetic refrigeration
technology. However, these magnetic refrigerators have a cooling power of only several tens of watts. There
are still challenges to develop magnetic refrigeration systems that are required of having high cooling power,
such as an air conditioning system. Increasing the cooling power of the magnetic heat pump is one of the
challenges for this new system to be put to practical use. The solutions include increasing the amount of the
magnetocaloric material (MCM), the magnetic flux, and the frequency of the active magnetic regenerator
(AMR) cycle. Among these solutions, we have focused on the speed-up of the AMR cycle. It is known that
the time response to temperature changes in an AMR cycle due to the magnetocaloric effect (MCE) is so fast
[6]. MCM are generally used in a form of particle, which have large contact areas. The particles, however,
make the flow paths more complex and cause large pressure losses between the MCM and the refrigerant.
The pressure losses limit the flow rate and the speed-up of the AMR cycle. For this reason, a material
configuration with a large contact area and low pressure losses is required to transfer heat quickly [7].
43
2. MAIN SECTION
2.1 Wire Made of Magnetocaloric Material

We have developed gadolinium (Gd) wires with diameters of 0.25 mm and 0.5 mm (Fig. 1-(a), (b)). When
single wires are cut into a same length and arranged side by side in the AMR bed, the porosity of the wires is
smaller than that of Gd particles with a diameter of 0.3 mm. This resulted in an increase in pressure losses
(Fig. 2-(a)). To solve the challenge, we made three-twisted wires, and the AMR bed was filled with the
wires (Fig. 2-(b), (c)). The use of the twisted wires allowed the securing of the flow paths between the wires
and the refrigerant and the control of the porosity. Figure 3 shows the comparison results of the nominal
contact area and the flow rate of each configuration of the MCM. There was no significant difference in the
flow rates between the Gd particles and the single Gd wires. On the other hand, the flow rate was increased
by using three-twisted wires with the same contact area as those of the particles and the single Gd wires.
Fig. 1 Gd wire Fig. 2 Twisted Gd wire
Fig. 3 Comparision of flow rate in each configuration of MCM at 0.03 MPa.
44
2.2 Experimental Methods
We measured the specific power of each MCM. Figure 4 shows a scheme (a) and a photograph of the
experimental apparatus (b). Table 1 shows a geometry of the AMR bed, and Table 2 shows measuring
parameters. The experimental apparatus is a reciprocating type, where the piston is movable in accordance
with the reciprocation of the Neodymium permanent magnets and creates the hot side and cool side on each
end of the AMR bed. A heater was inserted into the piping on the cold side so that this heater cancels the
temperature changes there. The output power of the heater was defined as the cooling power.
Fig. 4 Scheme (a) and photograph (b) of experimental apparatus
Table 1 Geometry of AMR bed

Gd Gd Gd Gd Gd
Material configrations
Particle Twisted wire Twisted wire Single wire Single wire
Diameter of MCM [mm] 0.3 0.25 0.5 1.0 2.0
Mass of MCM [kg] 0.130 0.112 0.112 0.122 0.122
Dimentions of bed [mm] 14 (height) x 14 (width) x 100 (length)
Measured porosity [%] 40 57 49 21 21
Contact area [m2] 0.33 0.23 0.12 0.062 0.031
Table 2 Measuring parameters

Operation frequency [Hz] 1.0 - 10.0
Magnetic field [T] 0 - 0.6
Flow rate [l/min] 0.288 - 4.32
Refrigerant Water
Ambient temperture [℃] 20
2.3 Results
Figure 5 shows the measurement results of the specific cooling power of the MCM. The specific cooling
power at an operating frequency of 3.0 Hz for Gd particles gradually decreased with increases in operating
frequency after peaking out at 61.9 W/kg. On the other hand, the specific cooling power at an operating
frequency of 5.0 Hz for Gd single wires with a diameter of 1.0 mm gradually decreased with increases in
operating frequency after peaking out. The measurement results of twisted Gd wires with diameters of 0.25
mm and 0.5 mm did not show the peak value in this measurement range. Although the porosities of all
samples were not theoretical minimum values and there might be a few preferential flow channels in some
samples, the largest specific cooling power in all the experimental data was achieved by using three-twisted
wires with a diameter of 0.25 mm at an operating frequency of 10.0 Hz. Moreover, there is a possibility that
the performance can be further improved by increasing the operating frequency. These results suggest the
technical feasibility of the wire-shaped MCM in a high frequency range.
45
Fig. 5 Performance of Gd wires in the exprimental apparatus
4. CONCLUSIONS
The results of the experiments have opened up new possibilities for increasing the operating frequency of the
magnetic refrigeration system using the wire-shaped MCM. If AMR cycles are operated using a wire-shaped
MCM at an operating frequency of above 10.0 Hz, the specific cooling power of wires will probably be more
than four times that of the particles. This study suggests the possibility of the wire-shaped MCM in applying
magnetic refrigeration technology. We will further develop a high-output magnetic refrigeration heat pump
by using the materials in the future.
REFERENCES
[1] United Nations Environment Programme Press, " The Montreal Protocol on Substances That Deplete the
Ozone Layer "
[2] B.F. Yu, Q. Gao, B. Zhang et al, " Review on research of room temperature magnetic refrigeration ",
International Journal of Refrigeration 26, 622-636 (2003)
[3] B. Yu, M. Liu, P.W. Egolf et al, " A review of magnetic refrigerator and heat pump prototypes built
before the year 2010 ", International Journal of Refrigeration 33, 1029-1060 (2010)
[4] V. Belusa, " Prototype of Magnetocaloric Wine Cooler ", BASF SE, CES 2015 (2015)
[5] Cooltech Applications Press," Premiere of Revolutionary Medical Refrigerator with Cooltech
Applications at Medica 2015 ", Medica 2015 (2015)
[6] N. Watanabe, M. Yorinaga, Y. Suzuki, et al, " FEASIBILITY STUDY OF HIGH FREQUENCY
MAGNETIC REFRIGERATION CYCLE BY FAST RESPONSE TEMPERATURE MEASUREMENT OF
MAGNETOCALORIC EFFECT OF La (Fe0.88Si0.12) 13 ", 6th IIF-IIR International Conference on
Magnetic Refrigeration, 7-10 September (2014)
[7] D.Vuarnoz, T.Kawanami, " Numerical analysis of a reciprocating active magnetic regenerator made of
gadolinium wires ", Applied Themal Engineering 37, 388-395 (2012)
46
NEW CONCEPT FOR MAGNETOCALORIC HEAT PUMPS BASED ON
THERMAL DIODES AND LATENT HEAT TRANSFER
K. Bartholomé*, T. Hess, A. Mahlke, J. König
Fraunhofer-Institute for Physical Measurement Techniques IPM
(Heidenhofstraße 8, 79110 Freiburg, Germany)
*
Corresponding author. E-mail: kilian.bartholome@ipm.fraunhofer.de
ABSTRACT
Since the beginning of the last decade, several dozens of magnetocaloric heat pump systems have been built
by different groups. Basically all of these systems are based on the Active Magnetic Regenerator (AMR)
concept, where a heat transfer fluid is actively pumped through a bed of magnetocaloric material in order to
transfer thermal energy to hot and cold side heat exchangers. Hereby several powerful systems were built,
generating large temperature spans of more than 50 K while others provided large cooling capacities of
several kW. However, up to now no system has been built which provides large temperature span and
cooling capacity while having a coefficient-of-performance (COP) better than standard compressor-based
cooling systems [1].
In this work a new concept and first experimental data of a magnetocaloric heat pump will be presented. In
this concept, the heat transfer is realized by the combination of magnetocaloric material with thermal diodes
which are based on latent heat transfer. Similar to thermosyphons, thermal energy is efficiently transported
by condensation and evaporation processes leading to heat transfer rates which are several orders of
magnitude larger than in conventional systems. At the same time, no additional pumps are required for
transporting the heat exchange fluids, enabling systems which large temperature spans and competitive
COPs.
DOI: 10.18462/iir.thermag.2016.0079
1. INTRODUCTION
The efficient transfer of heat between the magnetocaloric material and the heat exchanger is a decisive factor
in the overall efficiency of a cooling system based on magnetocalorics. Since the discovery of the giant
magnetocaloric effect by Pecharsky and Gschneidner in 1997 [2], many different magnetocaloric prototype
systems for room temperature cooling have been built, demonstrating the capability of this technology to
generate large temperature differences [3] as well as large cooling capacities [4]. Typically all of these
prototypes are based on the concept of “active magnetic regeneration” (AMR), in which a liquid is pumped
through the magnetocaloric material to transfer heat. The main challenges in the AMR-approach are
restricted heat transfer coefficients, large pressure drops within the regenerator associated with internal
friction of the liquid, resulting in low cycle frequency as well as large amounts of pumping energy required.
Current research activities on the AMR-concept are therefore focused on the optimization of the
magnetocaloric heat exchanger structures in order to minimize friction as well as pressure losses in the
regenerator and thereby maximize the system efficiency. Nevertheless, up until today no magnetocaloric
system has demonstrated the large potential of magnetocaloric cooling in terms of maximum system
efficiency.
In this work an alternative system design based on thermal diodes in combination with latent heat transport
using evaporation and condensation is presented. Hereby, the transport of thermal energy is in analogy to the
heat transfer in heat pipes [5]. By evaporating a fluid such as water or ethanol at the magnetocaloric material
and subsequently condensing it at the heat sink, it is possible to achieve heat transfer coefficients that are
several orders of magnitude higher than those achieved in traditional heat transfer by means of thermal
conduction or convection.
47
2. CONCEPT
2.1. Heat Pipes for Efficient Thermal Transport

The concept of efficient heat transfer using latent heat of evaporation and condensation is realized in heat
pipes and thermosyphon since almost 180 years. Thereby, a fluid is contained in a hermetically sealed tube
usually made of copper, in which all non-condensable gases have been removed. The fluid is present in the
liquid as well as the gaseous phase; the equilibrium between the two phases adjusts itself according to the
vapor pressure and the temperature. Inside this tube, a small temperature increase, e.g. induced by the
magnetocaloric material, leads to an evaporation of the liquid and an increase in vapor pressure. This in turn
results in a condensation of fluid on the condenser parts inside the heat pipe. Heat is passively transported
from the magnetocaloric material into the condenser section. The heat transfer coefficient of such an
evaporation process reaches values of up to 100 kW / (m2 K), which are orders of magnitude larger than for
conventional heat transfer processes, in which the heat is transferred by actively pumping a fluid.
Depending on the design of the heat pipe, efficient thermal diodes can be realized: The simplest are
gravitational-based thermosyphons, where the fluid is fed back from the condenser to the evaporator by
gravitational force. Reflux of the fluid does not occur against gravity, and therefore in this direction the heat
transfer is blocked. Alternative approaches or the realization of thermal diodes are the use of valves or
superhydrophobic surface treatments [6] for producing highly efficient thermal diodes.
2.2. System Design

The central component of the magnetocaloric cooling unit system is shown in Figure 1. It consists of
individual segments (1) with magnetocaloric material, which are connected in series and are cyclically
heated and cooled by rotation of a magnetic system (2). By designing these magnetocaloric segments as
thermal diodes, the heat generated in these segments is “pushed” in one direction, resulting in cooling one
side while heating up the other side.
Figure 1: Left: Magnetocaloric cooling unit replaces compressor system in refrigerator.

Center: Magnetocaloric cooling unit consisting of magnetocaloric segments (1) and magnetic system (2).
Right: Magnetocaloric segment consisting of magnetocaloric heat exchanger (3) and pressure relief valve (4).
In each of the segments is magnetocaloric material, designed as a heat exchanger unit (3). The segments are
separated by pressure relief valves (4). The valves are passively opened and closed by the pressure changes
which are induced by temperature changes inside the individual segments due to the magnetization and
demagnetization process. Thus, a directed flow of the gaseous fluid is ensured, and the magnetocaloric
segments act as thermal diodes (Figure 2).
Each segment creates a temperature difference of a few Kelvin. In order to achieve a large temperature lift of
some 10 K, several of these units must be connected in series. A parallel connection of the same units results
in a higher heat pumping capacity.
The main advantages of this concept are the following:
48
 Increase in systems efficiency: For the transfer of the thermal energy from the magnetocaloric
material to the heat exchanger, no additional pumps and therefore no additional energy is required,
which increases the systems efficiency.
 Larger cooling power: Heat transfer based on latent heat is several orders of magnitude larger than
heat transfer using sensible heat by pumping of fluids. This opens the possibility to realize
magnetocaloric systems being capable of system frequencies > 10 Hz, which results in an increase of
cooling power.
 Reduced system costs: Due to an increase in the system frequency, the amount of magnetic as well
as magnetocaloric material required to produce a specific cooling power can be significantly
reduced, which in turn reduces the perspective costs of a magnetocaloric cooling system.
1 2 3 4
Figure 2: Concept of thermal diode

Phase 1: Magnetic field applied to magnetocaloric heat exchanger: 1. The magnetic field is applied. Thereby, the
magnetocaloric material is heated and liquid is evaporated. 2. The pressure increases due to the evaporation of the
liquid. The valve to the right opens, fluid in the gas phase transmits latent heat to the next segment.
Phase 2: Magnetic field is turned off by rotation of the magnet: 3.The magnetocaloric material cools down due to the
decrease in the magnetic field strength. 4. Fluid from the gas phase condensates on the magnetocaloric material, the
vapor pressure decreases below the value in the foregoing segment. Now the valve to the left opens, gaseous fluid from
the left enters and heat is absorbed from the foregoing segment.
3. EXPERIMENTAL RESULTS
In order to demonstrate the latent heat transfer from magnetocaloric material to a condenser, a glass heat pipe
containing granular magnetocaloric material (Calorivac-H from Vacuumschmelze GmbH, Curie-
Temperature at 21.2 °C) as evaporator and degassed ethanol as fluid was built. A thermocouple of type K
was used as a condenser, capable of measuring the temperature TCond. The temperature of the magnetocaloric
material TMC was measured using a second type-K-thermocouple. The thermocouples were inserted into the
glass tube via a butyl-rubber-plug, which hermetically sealed the glass tube. The glass tube was then attached
to a vacuum pump by using a vacuum cannula in order to remove all non-condensable gases from the tube.
The glass heat way was placed between the two poles of an electromagnet. During change of the magnetic
field strength, the temperature changes of the condenser and the magnetocaloric materials were monitored
using a Keithley 2700.
Figure 3: Experimental data on latent heat transfer from magnetocaloric material.

Left: Glass heat pipe prepared with magnetocaloric material, thermocouples for measuring the temperature of the
magnetocaloric material (TMC) and condenser (TCond), placed between the poles of an electromagnet.
Right: Generated temperature difference of magnetocaloric material and condenser as well as strength of B-field plotted
versus time.
49
When the magnetic field is turned on, the temperature rise of the condenser sums up to > 90% of the
temperature rise of the magnetocaloric material, and the time shift between both curves is smaller than 0.1 s.
This shows that heat transfer by means of latent heat is capable of transferring heat from the magnetocaloric
material to the condenser fast and efficiently.
When the magnetic field is turned off however, the cooling of the condenser is much smaller than the cooling
of the magnetocaloric material. Apparently, the thermal coupling between magnetocaloric material and
condenser is now reduced, which presumably originates from a reduced wetting of the condenser with the
fluid.
4. CONCLUSIONS
In this work a new concept for magnetocaloric cooling systems based on latent heat transfer and thermal
diodes has been presented. The experimental data indicate that efficient and fast thermal transport using this
concept is viable. For the successful implementation of this concept into highly efficient magnetocaloric
cooling systems, it is crucial to enhance the diodicity of the thermal diodes. Therefore, two concepts will be
investigated in more detail: The first concept is the integration of pressure relief valves between different
magnetocaloric segments in order to guarantee a unidirectional thermal transport. The second concept is the
manipulation of the surface wetting of the fluid on magnetocaloric material and condenser by
superhydrophilic and superhydrophobic surface treatment [7]. Hereby the effect of a reduced thermal
transport in reverse direction of the thermal diode can greatly be reduced, resulting in diodicities of more
than 100 [6,7]. In combination with a matched magnetic system, a magnetocaloric cooling system with
greatly enhanced COP is feasible.
Besides the application to magnetocaloric materials, this concept of latent heat transport is also possible for
electrocaloric and mechanocaloric materials.
ACKNOWLEDGMENTS
The authors would like to acknowledge the Sustainability Center Freiburg for the support of the pilot project
“ActiPipe” as well as the Ministry for Finance and Economics Baden-Württemberg for the support of the
project “MacCool”.
REFERENCES
[1] A. Kitanovski, J. Tušek, U. Tomc, U. Plaznik, M. Ozbolt, and A. Poredoš, Magnetocaloric Energy
Conversion 2015).
[2] V. K. Pecharsky and K. A. Gschneidner, Physical Review Letters 78, 4494 (1997).
[3] D. Arnold, A. Tura, and A. Rowe, International Journal of Refrigeration 34, 178 (2011).
[4] S. Jacobs, J. Auringer, A. Boeder, J. Chell, L. Komorowski, J. Leonard, S. Russek, and C. Zimm,
International Journal of Refrigeration 37, 84 (2014).
[5] A. Faghri, Journal of Heat Transfer-Transactions of the Asme 134, 123001 (2012).
[6] J. B. Boreyko, Y. Zhao, and C.-H. Chen, Applied Physics Letters 99, 234105 (2011).
[7] T. Hess, A. Mahlke, O. Herm, and K. Bartholomé, in International Conference on Magnetic
Refrigeration at Room Temperature, Thermag VII, Turin, (2016).
50
STUDY OF THE PERFORMANCE COEFFICIENT OF AN ACTIVE
MAGNETIC REGENERATIVE REFRIGERATION SYSTEM
Z. Meddeb (a)*, D. Bchiri (a), R. Sidhom (a), M. Chrigui (b), M. R. Jeday (a)
(a)
Gabes University, National Engineering School of Gabes, R. U. Energy and Environment, Street Omar
IbnElkhatabZrig, 6072 Gabes, Tunisia.
(b)
Institute for Energy and Power Plant Technology, Technische Universitat Darmstadt, U.R. department of
Mechanical Engineering,Patersenstrasse 30, D-64287 Darmstadt, Germany
*Corresponding author. E-mail: zinameddeb1@gmail.com
ABSTRACT
This work seeks to study the coefficient of performance (COP) of an active magnetic regenerative
refrigeration system. The refrigeration system is consisted of a fluid circulating between a cold and a hot
source, active regenerator is formed by parallel plates of gadolinium, a circulation pump, a rotating magnet
and two heat exchangers to transfer heat between the fluid and the two sources. Thermodynamic, fluidic and
magnetic models were carried out. The supplied works within the system were determined to calculate the
COP such as the work of the circulating pump (Wp), the work of engine turning the magnet (We) and the
magnetic work (Wm). Contrary to Wp and We calculations, Wm calculation is not classic. To study the COP,
different exchanged heats within a cycle were calculated: the heat (Qc) recovered from the cold source, the
heat (Qh) rejected to the hot source and the heat regenerated by the gadolinium (Qr). In this study, Qc was
given by hypothesis, and Qh is determined by the first thermodynamic principle. While Qr was numerically
determined after an optimization of system operating parameters. A resolution of the continuity equation, the
amount of movement equation and the heat equation were carried out in order to study the temperature
profile in the regenerator and the fluid. Then, we deduce the temperatures at the inlet and the outlet of the
heat exchangers which help us to do the thermal balance. In this studied, the COP values of the magnetic
refrigeration system are considered as significant.
Keywords: Magnetic refrigeration, AMR cycle, Magnetocaloric effect, Gadolinium, COP

DOI: 10.18462/iir.thermag.2016.0084
1. INTRODUCTION
Magnetic refrigeration is a cooling technology based on the magnetocaloric effect (MCE). The MCE
describes the resulting change in temperature of a material due to the application of an external magnetic
field. From an environmental point of view, this technology is so fruitful because it does not involve the use
of greenhouse gases. The study of magnetic refrigeration was begun with the discovery of MCE [1]. Then, it
has been used in cryogenic refrigeration. It is maturely used in liquefaction of hydrogen and helium. Models
are paramount in order to both design and size prototypes. In this work, MCE is used as a parameter variable
which is an analytical expression according to the function of the material nature, the local regenerator
temperature and the applied magnetic field. In general, this parameter is considered as constant which the
case is not always.
The COP is a dimensionless quantity that depicts the performance of the refrigeration cycle. The Variation of
the COP, as a function of Qc, will be studied. The other originality of this work lies in the fact that the
expression of the magnetic working (Wm) is distinct from those used in the literature [2].
2. STUDIED SYSTEM
The cooling system consists of four generators constituted by parallel plaques of gadolinium [3], tow heat
exchangers, a pump and a moving magnet. The heat transfer fluid is water, gallium or nanofluid. The
regenerator is periodically magnetized and demagnetized. The system fluid passes through the hot heat
51
exchanger to transfer heat Qh to the hot source. Then, the fluid passes through the demagnetized regenerator
to provide heat. This cold fluid cools the cold source by exchanging heat with cooling load Qc. The fluid is
then heated by the magnetized regenerator, and it continues the cycle around.
3. PERFORMANCE COEFFICIENT
3.1. Thermodynamic System

The existing energies in the system are defined as follow: Qc is the heat transferred with the cold source. Qh
is the heat transferred with the hot source, and Qr is the heat transferred with the regenerator 1 and 2 or with
3 and 4. Qlosses is the lost energy by the system. Wp is the work of pump. We is the engine work of the magnet
rotation, and Wm is the work received by the material during magnetization.
The thermodynamic system is the fluid. It transfers heats and work with the hot source, the cold source and
the circulation pump. Therefore, Wp embodies the thermodynamic cycle work.
“fig. 1” shows a diagram representing the thermodynamic cooling system.
Demagnetized
regenerator
Wp We
- Qrd
System:
Cold Qc -Qh Hot
Heat transfer
source fluid source
Wm
Qra Qlosses
Magnetized
Magnetic field regenerator
Figure 1. Thermodynamic cooling system
The COP is calculated by the ratio of the cooling load (Qc) and the supplied work to the system (Ws).
So the COP is calculated by:
Gain Qc Qc
COP   
Supply Ws Wp  We  Wm
(1)
3.2. Exchanged Works
The pomp work, which is (Wp=Dv P.e ), and the engine work, which is (We =Ce.w.c ), are similarly ex-
used in other works. The magnetic work (Wm) is calculated by the magnetic force acting on an element. For
volume dV and x direction, we can formulate the following:



F  m.grad .B  (2)
B B
dFx   m dV
 x (3)
 m 1 B ²
dFx  dV
 2 x (4)
 1 B²
Fx  m V dV
 2 x (5)
52
 m .V 1 B ²
 2  x
dWm  dx
(6)
 m .V B ²
Wm 
 2 (7)
3.3. Exchanged Heats
-The heat transferred with the cold source (Qc) is:
Tfcs
Qc  Tfce m f Cp f (T )dT
(8)
-The heat transferred with the regenerator (Qr) is:
Tf h Nx1
Qr  Tf m f Cp f (T )dT   m .em .leq .x  Cp m (T ).MCE (T )
(10)
c
1
MCE(T) is determined by a semi-empirical method.

-The heat transferred with the hot source (Qh) is:
Tfhs
Qh  Tfhe m f Cp f (T )dT
(11)
According to the first law of thermodynamics, we can write:
Qc  Qh  2(Qra  Qrd)  Qlosses  Wp  0 (12)
Qlosses are the whole parasitic losses such as eddy currents in the magnetic material, viscous dissipation due to
fluid pressure drop and ambient heat leaks.
Therefore, to determine the different exchanged heat, we need to determine the temperature
profiles.
4. MODELING THE REGENERATOR TEMPERATURE
4.1. Physic Equations

In this proposal, we take into consideration the effects of the applied magnetic field on the fluid. The
governing equations of the AMR model equations are presented as following:
Fluid:
(13)
(14)
( ) ( ) ( ) (15)
Material:
( ) (16)
4.2. Numerical Analysis

A numerical code is developed to deduce the effects of the parameters (Lm, T, em, ef, B, Dv, Nc, fluid…)
on the system. We choose the optimized geometry. Then, we use a developed numerical tool
applying the FLUENT software as an attempt to resolve the equations. The boundary conditions of the
modeled regenerator are presented in “fig.2”.
C.L.1: VELOCITY_INLET (inlet fluid);

C.L.2: PRESSURE_OUTLET (outlet fluid);
C.L.3: SYMMETRY;
C.L.4: WALL (adiabatic);
Figure 2. The Boundary Conditions of the modeled regenerator.
53
4. RESULTS
It is deduced that there are several parameters affecting on the operating system. “Fig. 3” shows that the
material temperature changes depending on the regenerator length. Thereupon, we must select an accurate
length. “Fig. 4” represents the regenerator temperature as a function of position for various periods
for the same flow rate with different temperature between cold and hot side. We note that the
temperature profiles are different. Hence, an optimization is done to determine the proper settings. There
are other developed works on the optimization similar to our used parameters [4].
Figure 3. Regenerator temperature as

a function of position for different
periods for the same flow rate 6 ml/s.
case a: length is 0.1 m; case b: length
(a) (b) is 0.15 m
Figure 4. Regenerator temperature as a

function of position for different
periods for the same flow rate 6 ml/s.
case a:T=10K;case b: T=20K
(a) (b)
After sizing the geometrical system, we determine numerically the temperature profiles. Then, we
determine the different heat and works. As a result, we can deduce the value of the COP. “Fig.5” shows
the variation of the COP as a function of Qc, the quantity of heat transfer with the cold source. The COP
values are significant.
Figure 5. Variation of the COP as a

function of Qc at Dv=6ml/s, T=7K
5. CONCLUSIONS
In this paper, we have proposed a method to calculate the COP of magnetic refrigerating system. To
calculate the COP, we have to analyze, size as well as optimize the parameters in advance. Therefore, we
studied the entire system. It has been revealed that there existed an interaction among the parameters.
In fact, the temperature profile in the regenerator and in the fluid are determined. Hence, the amount of
regenerated heat by gadolinium is deduced. Afterwards, we determined Qh and those different works which
we were analytically established. Ultimately, the COP was deduced. The COP values are considered as
significant. A detailed, an energetic and economic analysis will be the aim of our future works.
REFERENCES
[1] B.F. Yu, Q. Gao, B. Zhang, X.Z. Meng, Z. Chen, Review on research of room temperature magnetic
refrigeration, International Journal of Refrigeration, 26 (2003).
[2] Andrew Row, Therodynamics of active magnetic regenerators, Crygenics 52 (2012).
[3] L. Rosario, M. M. Rahman, Analysis of magnetic refrigerator, Applied Thermal Engineering, 31 (2011).
[4]H. R. El-Hana Bouchekara, Afef Kedous-Lebouc, Jean-Paul Yonnet, Christian Chillet, Multiobjective
optimization of AMR systems, 37 (2014).
54
STUDY OF HEAT TRANSFERS IN A MAGNETOCALORIC
REGENERATOR WITH OSCILLATING HEAT FLOW
A. Meunier a,b,* , J.C. Roy b , R. Glises b, Y. Bailly b, T. de Larochelambert b, P.Nika b
(a)
NexT°PAC, 14 Impasse de la Nied, 57455, Seingbouse, France
(b)
Femto-ST, Energy department (CNRS UMR6174),2 avenue Jean Moulin, 90000, Belfort, France
*
Corresponding author. E-mail: alexandre.meunier@nextpac.fr
ABSTRACT
This paper propounds a modeling approach of the heat transfer in a magnetocaloric regenerator and the
experimental method to evaluate correlations of the heat transfer coefficient between the magnetocaloric
material and the coolant with an oscillating flow. This model includes a specific resolution for the fluid flow
equations which takes into account a specific case of oscillating flows, the annular effect. These correlations
can be used in the numerical model of a magnetocaloric heat pump to estimate the efficiency of this pump.
This study involves the demonstrator development which allows to improve the numerical model.
Keywords: heat transfer, heat pump, magnetocaloric, oscillating flow.

DOI: 10.18462/iir.thermag.2016.0086
1. INTRODUCTION
First studies on the magnetocaloric effect have been carried out in the late 18th century and its first
application concerns the cryogenics fields and the liquefaction field in the thirties. Since the 1970s,
the magnetocaloric effect has been used in several prototypes operating near room temperature that
have shown the potential of this effect for heat transfer systems with the use of Active Magnetic
Regeneration cycle [1] [2]. Nevertheless, the integration of the magnetocaloric effect in a complex
system such as a heat pump is not easy because several physic phenomena are simultaneously
involved (magnetism, thermodynamics, mechanics, fluid mechanics) in a transient regime. Indeed,
in a magnetocaloric system, the material is subjected to a magnetization/demagnetization cycle and
two steps of heat transfer with coolant: the cold blow after the magnetization and the hot blow after
the demagnetization where the coolant flows in opposite direction compared to the flow of the hot
blow step. An annular effect can be created with this reverse flow [3] and influences the heat
transfer coefficient. To evaluate the impact of the annular effect on the heat transfer, the numerical
model has to be developed in two dimensions [4]. Our model describes the evolution of the
temperature of the coolant fluid and the magnetocaloric materials during an Active Magnetic
Regeneration cycle. The results are compared to the measurements performed with a
magnetocaloric demonstrator developed in the laboratory.
2. MODEL AND DEMONSTRATOR
2.1 Development of 2D a numerical model
Definition of the study system and velocity profile
This numerical model shows different physical interactions in the magnetocaloric regenerator of a
heat pump. The regenerator is represented as a surface (magnetic material) in contact with an
oscillating coolant flow. The velocity of the coolant flow is defined by the function [5]:
( , )= ∗ℜ ( ) (1)
√ √
cosh − cosh 2
( )= (2)
√
1 − cosh 2
with the angular frequency, the Womersley number, the plate spacing and the imaginary
unit.
55
This system is discretized in two dimensions by a nodal network in different cells.
Insulation
Symmetry condition
(Measured face)
N solid nodes layers
Velocity
Wall
a/2
Raduis
M fluide nodes layers
x Symmetry condition
3 4 2 1 y
Figure 1: Flow velocity as a function of Flow
radius and time and mesh of study system.
Fundamental equations and resolution method
Equations (1) and (2) respectively describe the thermal behaviour of every material and fluid nodes.
(3)
(4)
with λ, T, , Cp and u respectively the thermal conductivity, the temperature, the density, the heat
capacity of the material or fluid and the axial flow velocity defined by the function (1);
,-./ is the volumetric magnetic production term such as:
(5)
with 3 the magnetization, % the internal magnetic field, 2 the free space permeability.
In the first approximation, the term is considered constant and is evaluated through data on
gadolinium. This model uses a finite-difference method in order to solve the heat transfer equations
between the fluid and the magnetocaloric material. The equation system is solved with a Crank-
Nicholson method through a matrix expression.
2.2 Development of an experimental demonstrator
An experimental demonstrator has been developed to measure the temperature evolution in the
channel of the magnetocaloric regenerator.
Magnetic system
The magnetic system consists of a magnetic circuit with an air gap and a magnetic rotor including
permanent magnets driven by a motor, allowing to generate a variable magnetic field in the air gap
with a trapezoidal time sequence reaching 10 Hz.
56
0,6
0,4
Magnetic field B(T)

0,2
0,0 0 45 90 135 180 225 270 315 360
-0,2
-0,4
angle (°)
-0,6
Figure 2: Magnetic system Figure 3. Evolution of the magnetic field in the air gap
Regenerator magnetocaloric and fluid circuit
The regenerator consists of a single 1 mm wide channel between two magnetocaloric plates (40 × 4
× 0.6 mm3). The magnetocaloric material used is a composite provided by Erasteel. Its Curie
temperature is 20°C. One of the two plates is fixed on a sapphire window to measure the
temperature of the material with a thermal camera. The oscillating flow (water) is generated by a
modification of the fluid circuit through electrovalves ensuring frequencies up to 10 Hz (Fig. 4).
The fluid (water) enter in the channel with a temperature close to the Curie temperature of
magnetocaloric material and the flow rate is calculated with this relation:
89 = 2 ∗ :; ∗ 0 ∗ (1)
With 89 is masse flow, :; is volume of the channel, is the frequency of

magnetization/demagnetization.
Figure 4: Measured face on the plate of

magnetocaloric material
Figure 5: Fluid circuit design

Results
This demonstrator allows to follow the instantaneous temperature’s evolution of the magnetocaloric
material. After a few minutes, we observe the variation of a stationnary longitudinal temperature
gradient along the magnetocaloric plates before and after magnetization around the Curie
temperature in no-load condition with a 0.04 m.s-1 average velocity at 0.5 Hz (Fig.6) and a 0.32 m.s-
1
at a magnetization/demagnetization frequency of 4 Hz (Fig.7),.
57
22,0 20,8
21,5 20,6
20,4
Temperature (°C)
21,0
Temperature (°C)
20,2
20,5
20
20,0
19,8
19,5
19,6
19,0 19,4
18,5 19,2
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)
Cold side Hot side Cold side Hot side
(a) (b)
Figure6: Instantaneous temperature of hot and cold sides of the magnetocaloric material at 0.5Hz
on the measured face (a: simulation, b: experimental)
21,5 21,6
21,4
21,0
21,2
Temperature (°C)
Temperature (K)
20,5 21
20,8
20,0
20,6
19,5 20,4
20,2
19,0
20
18,5 19,8
0 0,2 0,4 0,6 0,8 1 0 0,2 0,4 0,6 0,8 1
Time (s) Times (s)
Cold side Hot side hot side cold side
(a) (b)
Figure7: Instantaneous temperature of hot and cold sides of the magnetocaloric material at 4Hz
on the measured face (a: simulation, b: experimental)
At a 4 Hz frequency and more, the thickness of the plate is too big to observe heat exchanges in the
cold blow and hot blow steps because the characteristic time of thermal conduction is too important
in our system. For the next system, to be able to measure the temperature evolution on these steps,
the thickness of the plate must be reduced.
3. CONCLUSION
The numerical model presented in this paper which integrates fluidic phenomena such as annular
effect to solve the equation of heat transfer in a magnetocaloric regenerator. This heat transfer
model is coupled with a numerical modelling of the magnetocaloric effect developed in the
laboratory. Our demonstrator allows to perform measurements of the magnetocaloric material with
a thermal camera and allows to generate a longitudinal temperature gradient. This demonstrator is
the first step toward a comprehensive study of magnetocaloric regenerator for magnetocaloric heat
pump.
REFERENCES
[1] K.A. Gschneidner, Jr, V.K. Pecharsky, “Thirty years of near room temperature magnetic cooling: Where
we are today and future prospects”, in Int. J. Refr., vol 31, pp 945-961, 2008.
[2] F.C. Chenet al.“Thermodynamic Analysis of Four Magnetic Heat-Pump Cycles”, in J. Eng. Gas Turb.
Pow., vol 114, pp 715-720, 1992.
[3] G.E.Richardson, E. Tyler, “The transverse velocity gradient near the mouths of pipes in which an
alternating or continuous flow of air is established”, in Proc. Phys. Soc., vol 42, pp 1-15, 1929.
[4] P.A. Oliveira et al., “A 2D hybrid model of the fluid flow and heat transfer in a reciprocating active magnetic
regenerator”, in Int. J. Refr., vol 35, pp 98 -114, 2012
[5] P.C. Chatwin, “On the longitudinal dispersion of passive contaminant in oscillatory flows in tubes”, in
Journal of Fluid Mechanics, vol 71, pp 513-527, 1975.
58
THE OPERATING PRINCIPLE OF A FULLY SOLID STATE ACTIVE
MAGNETIC REGENERATOR
Mingkan Zhang, Ayyoub Mehdizadeh Momen, Omar Abdelaziz*
Building Equipment Research, Oak Ridge National Laboratory
Oak Ridge, TN, USA
*
Corresponding author. E-mail: abdelazizoa@ornl.gov
ABSTRACT
As an alternative refrigeration technology, magnetocaloric refrigeration has the potential to be safer, quieter,
more efficient, and more environmentally friendly than the conventional vapor compression refrigeration
technology. Most of the reported active magnetic regenerator (AMR) systems that operate based on the
magnetocaloric effect use heat transfer fluid to exchange heat, which results in complicated mechanical
subsystems and components such as rotating valves and hydraulic pumps. This paper presents an operating
principle of a fully solid state AMR, in which an alternative mechanism for heat transfer between the AMR
and the heat source/sink is proposed. The operating principle of the fully solid state AMR is based on moving
rods/sheets (e.g. copper, brass, iron or aluminum), which are employed to replace the heat transfer fluid.
Such fully solid state AMR would provide a significantly higher heat transfer rate than a conventional AMR
because the conductivity of moving solid rods/plates is high and it enables the increase in the machine
operating frequency hence the cooling capacity. The details of operating principle are presented and
discussed here. One of the key enabling features for this technology is the contact between the moving
rods/sheets and magnetocaloric material, and heat exchange mechanism at the heat source/sink. This paper
provides an overview of the design for a fully solid state magnetocaloric refrigeration system along with
guidelines for their optimal design.
Keywords: Active Magnetic Regenerator; Solid State; Operating Principle;

DOI: 10.18462/iir.thermag.2016.0088
1. INTRODUCTION
The magnetocaloric regenerator in simple words is a device that utilizes the magnetocaloric effect (MCE) to
induce an adiabatic temperature change of a magnetocaloric material (MCM) during the magnetic field
change of the material. To overcome the low maximum adiabatic temperature change of most MCM at
moderate magnetic fields (up to 5 K in 1.5 T) at room temperature, the concept of AMR was invented by
Barclay and Steyert [1] in 1982. In this technique, a regenerative cycle is applied to create temperature spans
high enough comparable to traditional cooling systems. From then on the AMR has become an interesting
research topic in magnetic refrigeration community for several decades since it provides an alternative
refrigeration technology with significant potential energy savings compared to conventional vapor
compression refrigeration technology at room temperature. A large number of experimental and numerical
studies have been reported about it, which are summarized by the review papers [2]–[8].
Although the performance of AMR has been highly improved and some AMR prototypes have been
developed in the last 30 years, some challenges still hinder its commercialization, e.g. the high cost due to
the need of a pump or rotating valve for fluidic system and low heat transfer rate because of the low thermal
conductivity of the most working fluids (e.g. water). We propose to replace the heat exchange fluids by solid
state materials as a one of the potential solutions to address these challenges. However, AMR has so far been
restricted to applications using fluids and very few works have been reported about solid state AMR. Silva et
al. and Silva et al. [9], [10] published a design of a solid state AMR based on a new kind of material whose
thermal conductivity changes with magnetic field as a gate to control the direction of heat exchange instead
of controlling the direction of fluid flow. However, their designs are ideal because they highly depend on the
advantage in material science to discover the material they need. Our group recently provided an alternative
design of fully solid state AMR, in which high-conductivity moving rods/sheets made of regular materials
(e.g. copper, brass, iron or aluminum) are utilized to replace heat transfer fluid enhancing the heat transfer
rate as well as
59
cooling/heating capacity significantly. This paper discusses the operating principle of the fully solid state
AMR, and provides an overview of the design for this technology.
2. Operating Principle of the Solid State AMR
The proposed patented [11] Solid State AMR is based on the replacement of the heat transfer fluid with high-
conductivity moving sheets or rods to exchange heat. This will provide exceptional heat transfer
characteristics, which cannot be achieved otherwise with the fluidic system, and eliminate the need for a
pump or rotating valve. A typical Nusselt number (which describes the ratio between convection and pure
conduction) in a mini-channel with laminar flow is approximately 4. If a fluidic system to be replaced with
high conductivity solid sheets, theoretically the new system offers more than 1 order of magnitude, increase
in heat transfer rate (~KSolid/(4×KLiquid)). Although some restrictions (e.g. contact surface roughness) could
limit the system performance, our preliminary analysis shows one order of magnitude improvement in the
cooling capacity is expected [12]. In order to lubricate the contact surface and minimize the thermal
resistance of the gap, the gap between the body of MCM and the moving sheet is full of low-viscosity
thermal grease. Our simulation results indicates that with a high conductivity grease[12], the heat exchange
through the gap is enough to generate a high cooling power. The enhanced heat transfer and low friction
between the high-conductivity moving rods and the MCM will allow approximately an order of magnitude
increase in the machine operating frequency, which is proportional to the cooling/heating capacity. Thus for
a fixed cooling/heating capacity demand, the system can achieve the target using smaller amounts of MCM
hence smaller magnets. It may eliminate not only the need for expensive pump/valve systems but also the
problem of sealing fluids. As a result, the system can effectively be very compact.
The regenerator is comprised of high conductivity moving sheets/rods instead of a working fluid sandwiched
by the body of MCM. To avoid performance loss due to the longitudinal conduction of the high thermal
conductivity sheets/rods, the moving sheets/rods are comprised of small pieces of high conductivity metal
connecting using low conductivity materials. The operating principle of the solid state AMR is shown in
Figure 1. The solid state AMR is comprised of a MCM cylinder or sheets with a plurality of channels
extending between the two ends, a matrix of copper rods (or sheets) positioned within each channel of the
MCM cylinder, a magnet surrounding the cylinder to provide a uniform magnetic field over the cylinder. The
MCM cylinder is graded along the cylinder axis with a range of MCMs each with a different Curie
temperature to enhance the solid state AMR performance. The gaps between the MCM cylinder and the
copper rods are filled with thermal grease. Moreover, a well-designed motor system is employed to provide
both magnet and rods actuations. A simple cycle analogous to the Carnot cycle is performed here, in which
the magnetic field strength increases and decreases instead of the pressure changing in the Carnot cycle.
Because of the temperature change of MCM by exposing it to a changing magnetic field, a heat transfer will
occur through the MCM and solid state material, then creating a temperature difference in cold and hot ends
to conduct cooling and heating capacity, respectively. Figure 2 depicts the variation of the copper rods
velocity and the magnetic field during the AMR thermodynamic cycle, in which the cycle is divided into
four processes. Following are the four steps of the Brayton-like cycle, which will be repeated to generate
high cooling/heating energy.
Step 1: The MCM cylinder is placed in a thermal insulated environment. The magnet is moved to surround
the MCM, leading to an increasing magnetic field (+H), which causes the magnetic dipoles to align, and
material's magnetic entropy to decrease. Since this process only lasts a short time, the heat conducted from
the MCM to solid rods is neglected. Due to the adiabatic process, that the overall energy is not reduced,
hence the MCM temperature increases to conserve total entropy. This temperature change is called adiabatic
temperature change (ΔTad) and only happens when the magnetic field changes. This step is called adiabatic
magnetization process.
Step 2: After the MCM temperature reaches T + ΔTad, the solid rods are moved towards the hot end at a
velocity of –u (analogous to movement of the fluid in the conventional systems). When the copper rods
travel through the hot end, heat exchange occurs between the copper rods and regenerator and the extended
parts of the solid rods exchange heat to ambient at the hot end to remove the heat (-Q) from MCM. Due to
60
the higher conductivity of solids compared to liquids, the heat could be removed much more quickly than the
conventional AMR employing fluid to exchange heat. During the step 2, the magnetic field is held constant
to prevent the dipoles from reabsorbing the heat.
Step 3: When the MCM is sufficiently cooled (i.e. temperature is back to T again), an adiabatic
demagnetization process is performed by moving the magnet away from the MCM. During the
demagnetization, the magnetic dipoles become disordered, thereby increasing the magnetic entropy, as such
the temperature drops to conserve the total entropy. This results in an adiabatic temperature change (-ΔTad).
Step 4: In the last step of the cycle, the magnetic field is kept same as in step 3 to prevent reheating the
MCM. The solid rods move towards the cold end to refrigerate the target. In this step, energy transfers from
cold end to the MCM (+Q).
It is important to note that in the steps above, T is a function of the location within the regenerator. We
assume that the adiabatic temperature change for each stage is different; as such, the temperature within the
regenerator is a function of the axial position.
Figure 1. An example of a solid state AMR.
Adiabatic
Magnetization: +H
Heat Addition from Cold End: T Adiabatic Magnetization: T + ΔTad
-u
Hot end Cold end Hot end Cold end

+Qcold
-Qhot
Adiabatic
Adiabatic Demagnetization: T - ΔTad Demagnetization: -H Heat Rejection from Hot End: T
Hot end Cold end Hot end Cold end

Figure 2. A thermodynamic cycle of the solid state AMR.
61
3. CONCLUSIONS
The operating principle of a novel solid state AMR is introduced in this paper, in which an alternative
mechanism for heat transfer between the AMR and the heat source/sink is proposed. The solid state AMR
can potentially provide an enhanced heat transfer between the high-conductivity moving rods/sheets and the
MCM, which would allow a machine operating frequency to enhance hence the cooling and heating
capacities.
ACKNOWLEDGMENTS
This work was sponsored by the U. S. Department of Energy’s Building Technologies Office under Contract
No. DE-AC05-00OR22725 with UT-Battelle, LLC. We would like to acknowledge Mr. Antonio Bouza the
Technology Manager for the HVAC & Appliances for his support.
REFERENCES
[1] J. A. Barclay, W. A. Steyert, “Active magnetic regenerator”, Patent US4332135, (1982).

[2] K. Engelbrecht et al.,“Review Article: Recent Developments in Room Temperature Active Magnetic
Regenerative Refrigeration”, HVAC&R Res., 13, 525, (2011).
[3] B. Yu et al., “A review of magnetic refrigerator and heat pump prototypes built before the year
2010”, Int. J. Refrig.,33, 1029, (2010).
[4] K. Gschneidner, V. Pecharsky, “Thirty years of near room temperature magnetic cooling: Where we
are today and future prospects”, Int. J. Refrig.,31, 945, (2008).
[5] V. Franco et al., “The Magnetocaloric Effect and Magnetic Refrigeration Near Room Temperature:
Materials and Models”, Annu. Rev. Mater. Res., 42, 305, (2012).
[6] J. Romero Gómez et al., “Magnetocaloric effect: A review of the thermodynamic cycles in magnetic
refrigeration,” Renew. Sustain. Energy Rev., 17, 74, (2013).
[7] B. Yu et al., “Review on research of room temperature magnetic refrigeration”, Int. J. Refrig., 26,
622, (2003).
[8] K. Nielsen et al., “Review on numerical modeling of active magnetic regenerators for room
temperature applications”, Int. J. Refrig., 34, 603, (2011).
[9] D. Silva et al., “Solid state magnetic refrigerator”, Appl. Energy, 93, 570, (2012).
[10] D. Silva et al., “Maximizing the temperature span of a solid state active magnetic regenerative
refrigerator”, Appl. Energy, 113, 1149, (2014).
[11] A. Momen et al., “Magnetocaloric Refrigeration Using Solid Working Medium”, US Provisional
Patent Application, ID 3263.1, (2014).
[12] M. Zhang et al., “Preliminary Analysis of a Fully Solid State Magnetocaloric Refrigeration”,
Proceedings of the 16th international Refrigeration and Air Conditioning Conference at Purdue,
(2016).
This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of
Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United
States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this
manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to
these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-
access-plan).
62
DESIGN AND RESEARCH OF THE ROOM TEMPERATURE
MAGNETIC WINE CABINET
Zhang Cheng(a)(b)*, Huang Jiaohong(a)(b), Yan Hongwei(a), Jin Peiyu(a)(b)

Cheng Juan(a)(b), Liu Cuilan(a)(b), Li Zhaojie(a)(b), Zhang Yingde(a)(b)
(a)Baotou Research Institute of Rare-earths, Baotou, China

(b)State Key Laboratory of Baiyunobo Rare Earth Resource Research and Comprehensive
Utilization, Baotou, China
*Corresponding author. E-mail: jiaohongh@sina.cn
ABSTRACT
Anovel permanent magnet composite magnetic refrigerator on room temperature which used the double sets
of two concentric quadropole Halbach array magnet was designed. The refrigerant usedGd and GdEr alloy
particles and was divided into four parts according to the curie temperature then filled into the regenerator
with the quality of 1.5kg. The magnetic refrigerator was applied on the wine cabinetwith the cold room
capacity of 92l and the max cooling power was 147W. The cooling capacity of the wine cabinet was
measured. The results showed that the max temperature span could exceed 25K and the lowest temperature
at the cold end was lower than 267K. The experiment results showed that the wine cabinet had met the
requirement of cold storage and could be practical used preliminary.
Keywords:magnetic refrigerator, Halbach array magnet, cooling power, temperature span
DOI: 10.18462/iir.thermag.2016.0090
1.INTRODUCTION
In recent years,more and more researchers fixed attention on room temperature magnetic refrigerator and
many magnetocaloric devices and prototypeshave beendesigned. However, significant challenges still need
to be overcome before the technology could replacethe conventional vapor compression and enter into
people's life[1].One of the most important problems is the ability to exploit the potential for high
temperature spans.In this paper we hadcreated a novel permanent magnetic refrigerator with making
maximum cooling power and high temperature spans as our basic objective in system.
2. METHODS AND RESULTS
2.1 Thermal design of AMR

The magnetic system was comprised of the doublesets of two concentric quadropole Halbach array
magnet. Fig.1 shows a picture of the magnetic system. Each magnetic system was nested
63
by two concentric Halbach array magnet,which contained two parts,that were inner container, and outer
container.Fig.2 shows picture of the two concentric Halbach array magnet. The regenerator was deposited
in the middle of the inner container. The air-gap between inner container and regenerator acts as insulation
and can reduce the magnetized volume available for MCM[2]. Through simulation analysis,the dimension
of the air-gap was confirmed to 1.5mm.The inner container was rotated by a brushless DCmotor with a
gearbox while the outer container was stationary.The rotation of theassembly could affect the
regeneratorwith a periodic magnetic flux density with a period of 180°and a peak flux density as high as
1.5T in the magnetized region and as low as 0.1T in the demagnetizedregion. we had considered multi-layer
AMR made of the Gd and GdEr alloy particles with the Curie temperature interval of about 5K, that were
278K,283K,288K and 293K, respectively. The length of the regenerator is 200mm,with the internal
diameter is 32mm. The diameter of the alloy particles was 0.4mm-0.6mm and the quality of each
regenerator by filling MCM for about 0.75kg,two a total of about 1.5kg. Demineralizedwater was used as
the heat transfer fluid, and a commercial pump, able to return a maximum volumetric flow rate of
5.3Lmin-1, was adopted. The pump and valve combination created an reciprocating fluid flow through the
regenerators, achieving heat exchanger. Fig.3 showsa scheme diagram of the hydraulic system of the
device.
Fig.1 The double sets of two Fig.2 The Halbach Fig.3 The scheme diagram of
concentric Halbach array magnet array magnetic system the hydraulic system
2.2Experimental procedure
We had successfully contrived a wine cabinet with the cold room capacity of 92l. As is shown in
Fig.4.During the tests, special attention had been given to the maximum temperature span of the
refrigerated cabinet. As is known to us,the operating frequency is an important parameter to the
refrigerator,which can influence the refrigeration ability of it. So before testing the maximum temperature
span of the refrigerated cabinet, we determine the optimal operating frequency through repeated
experiments, that is 0.2Hz. At the same time,we have measured the cooling power.when measuringthe
cooling power, wehave the safe temperature compensation to agree with the room temperature,the power is
147W. Fig.5 shows the experimental results of maximum temperature span between cold and hot end
temperatures from multi-layer AMR madeof Gd and GdEr alloy particles materials at 291.6K.From the
temperature curve,wo can know that the maximum temperature span is about 25K,while the temperature of
the cold-end is lower than 272K.The temperature of the safe is about 6K. In order to make the refrigerated
cabinet enter the market as early as possible, we tried our best to design optimization for it. Mainly includes
two aspects, one is the gear transmission system,
64
and the other is the heat exchange system. We had measured the mass flux of the exchange fluid flowing
through the regenerator, and the heat exchange system was designed reasonably. After optimizing, the
result of experiment is shown inFig.6. From the graph,we can gain that the lowest temperature at the cold
end is lower than 267.5K,and the maximum temperature span exceeds 25K.It is worth noting that the safe
temperature,at the end of the experiment,the safe temperature approaches to 3.5K,which meets the cold
storage needs.
Fig.4 Wine cabinet
Fig.5 Experimental results of temperature Fig.6 Experimental results of temperature

span when the room-temperature is span when the room-temperature is
291.6K 290.4K after optimizing
3. CONCLUSION
In order to realize performance and cost competitive withconventional vapor compression technology in
the future, importantfields were summarized.(1) Design of layered AMR can improve the efficiency of the
refrigerated cabinet.(2) Experimentalresults show that the frequency is not the higher or the lower the
better. In my opinion,the higher the frequency, the more difficult heat exchange, so the optimal frequency
is 0.2 Hz .
ACKNOWLEDGMENTS
This work was supported by the National Natural Science Foundation of China(51261001).
65
REFERENCES
[1] D. S. Arnold, A. Tura, A. Ruebsaat-Trott, and A.Rowe, “Design improvements of a

permanentmagnet active magnetic refrigerator,” Int. J. Refrig.,2014,vol. 37:99-105.
[2] A. Tura, A. Rowe “Permanent magnet magnetic refrigerator design and experimental
characterization.” Int. J. Refrigeration, 2011,34 (3): 628-639.
[3] J. Romero Go´mez, R. Ferreiro Garcia, J. Carbia Carril, M. Romero Go´mez “Experimental
analysis of a reciprocatingmagnetic refrigeration prototype.” Int. J. Refrigeration , 2013,36 (4):
1388-1398.
66
EXPERIMENTAL EVALUATION OF A SINGLE LAYER MnFeP1-xSix AMR
P. Govindappa*, P. V. Trevizoli, I. Niknia, O. Campbell, T. V. Christiaanse, R. Teyber,
A. Rowe

*
Corresponding author. E-mail: premg@uvic.ca
ABSTRACT
First order magnetocaloric materials (FOM) have desirable characteristic such as a high magnetocaloric
effect, large heat capacity and use inexpensive constituents. However, irreversibilities associated with
thermal and magnetic hysteresis, the strong dependency of the specific heat with temperature and magnetic
field, and the narrow temperature range of useful magnetocaloric effect require further investigation.
Among the the MnFeP1-xSix system is one of the promising materials for magnetic heat pumps
applications near room temperature. The present paper describes the experimental investigation of a
single-layer MnFeP1-xSix active magnetic regenerator (AMR), under different test conditions and
following a protocol of heating and cooling processes. The results for the FOM are compared with a Gd
AMR, that is experimentally tested following the same protocol, with the objective to study the
irreversibilities associated with FOM.
Keywords: magnetocaloric effect, active magnetic regenerators, first order material, thermal hysteresis, PM
I. DOI: 10.18462/iir.thermag.2016.0096
1. INTRODUCTION
First order magnetocaloric materials (FOM) have been extensively researched by several groups around the
world [1, 2]. Although some of FOMs have desirable characteristics such as a high magnetocaloric effect,
large specific heat and use inexpensive raw components, they can also have some drawbacks such as
irreversibility associated with thermal and magnetic hysteresis, the strong dependency of the specific heat
with temperature and magnetic field, and the narrow temperature range of the magnetocaloric effect that
constitute challenges for implementation [3, 4]. The hysteresis, for instance, can reduce the usefulness of a
material in a refrigeration cycle by decreasing the reversible adiabatic temperature change and by decreasing
efficiency [5]. The present paper describes experimental characterization of a single-layer MnFeP1-xSix
AMR. The material was experimentally characterized via DSC measurements and thermal hysteresis of
5oC was measured. The AMR experimental tests are performed in the permanent magnetic test apparatus
(PM I) [6]. Fixed displaced volume of 5.09 cm3 and frequency of 1.0 Hz are used, while the rejection
temperature is varied from 10 to 40oC, at no heat load conditions. In order to study the irreversibilities
associated with this FOM material, heating (increasing the rejection temperature) and cooling
(reducing the rejection temperature) measurements are performed. In addition, a Gd AMR with
similar geometric parameters is also tested using the same experimental protocols. In contrast to
FOMs, it is well known that Gd does not present significant hysteresis and its adiabatic temperature
change is high over a broad range of temperature. In this way, a comparison between the FOM and Gd
are performed in order to study irreversibilities associated with FOM. The results are presented in terms of
the maximum temperature span as a function of the rejection temperature.
2. EXPERIMENTAL
The experimental tests are performed using the test apparatus known as PM I at the University of Victoria
[6]. This device uses two rotary nested permanent magnet Halbach arrays to generate a time-varying
magnetic field in the bores. Two AMR beds are tested with one in the bore of each magnet. As the
magnets rotate, fluid is pumped from the cold to the hot heat exchanger of the magnetized bed, and from hot
to the cold heat exchanger of the demagnetized bed. The hot heat exchangers has its temperature controlled
by a thermal bath, while the cold heat exchangers impose a thermal load using an electrical heater. Check
valves are used in the cold heat exchangers to minimize dead volume in the cold section of the device.
67
In all the experimental tests the displaced volume (Vd) of 5.09 cm3 and the operating frequency (f) of 1 Hz
are fixed. The rejection temperature (TH) is varied in a range from 10 to 40oC. In the present work zero
applied load (parasitic heat leaks only) tests are reported. The heat transfer fluid is mixture of water and
ethylene glycol in a volume fraction of 80/20%. The AMR matrix is housed in a G10 tube with 16 mm
diameter. Table 1 presents the main properties for the Gd and the MnFeP1-xSix AMRs.
Table 1. Properties for the Gd and the MnFeP1-xSix AMRs beds.
Properties G MnFeP1-xSix
Transition temperature (oC) 20
0
Mass (g/bed) 51.5 50.6
Porosity (-) 0.36 0.42
Geometry Spheres Crushed irregular particles
Particle size (µm) 250-600 300-425
Regenerator length (mm) 50 56.6
Both types of AMR beds (Gd and MnFeP1-xSix) are tested considering heating and cooling procedures.
The heating procedure begins by setting the rejection temperature ~10oC lower than the respective Gd or
MnFeP1-xSix transition temperature. For instance, for the Gd bed the first rejection temperature is ~10oC.
Then, the test is performed until periodic stead-state is reached, which may take up to 3 hours
depending on the test conditions. After recording the experimental data, a new rejection temperature is set
~ 3 to 5oC higher than the previous point. This procedure is repeated up to rejection temperatures of
35-40oC. The cooling procedure, on the other hand, begins at the highest rejection temperature used in the
heating procedure and, after the periodic stead-state, the rejection temperature is reduced 3-5oC. Once
steady-state is reached, the rejection temperature is decreased.
3. RESULTS
Fig. 1(a) compares the adiabatic temperature change directly measured, for 0-1.1 T magnetic field change,
for the Gd and the MnFeP1-xSix. For the present MnFeP1-xSix alloy, Gd has a larger adiabatic temperature
change over a broader temperature range. On the other hand, the MnFeP1-xSix presents a considerably
larger specific heat capacity than the Gd, as presented in Fig. 1(b), where the heat capacity at 0 T as a
function of the temperature is plotted for heat and cooling measurements. From Fig. 1(b), a thermal
hysteresis of about 5oC is observed for the present FOM.
3 1600
Gd (a) M nFeP1-xSix - Heating (b)
0 H = 1.1 T
1400 M nFeP1-xSix - Cooling
Specific Heat Capacity [J/kgK]
2.5 0 H = 0 T
1200
2
1000
Tad [K]
1.5 800
M nFeP1-xSix - Heating
600
1
M nFeP1-xSix - Cooling 400
0.5
200 Gd
0 0
270 275 280 285 290 295 300 305 310 315 270 275 280 285 290 295 300 305 310 315
Temperature [K] Temperature [K]
Figure 1. Gd and MnFeP1-xSix properties: (a) Direct measured adiabatic temperature change as a function of the
temperature, for a magnetic field variation of 1.1 T; (b) Specific heat capacity as a function of the temperature at 0 T.
Fig. 2 presents the measured maximum temperature span (ܶ௦௣௔௡), at no load conditions, as a function of the
rejection temperature for the heating and cooling curves performed using the Gd bed. The results show the
expected trend: as the rejection temperature increases the temperature span increases [6, 7] and a
maximum temperature span of 26oC at a rejection temperature of 36oC is found. Comparing results for the
heating and
68
cooling processes shows that a single temperature span curve is produced. As expected, for typical
second order materials no irreversibilities associated with hysteresis are observed.
30
25 Heati
eati ng
Co
Cooolin
ing
g
20
TSpan [oC]
15
10
0
10 15 20 25 30 35 40
o
TH [ C]
Figure 2. Heating and cooling processes maximum temperature span as a function of the rejection temperature for the
Gd AMR, at zero applied load conditions.
Fig. 3 shows the heating and cooling temperature span, at no load conditions, as a function of the
rejection temperature for the MnFeP1-xSix bed. For this case, a peak of 12oC temperature span is found
around 30oC (0W). Comparing the heating and cooling curves above the rejection temperature of 30oC,
a significant difference between the heating and cooling curves is observed. This result suggest that the
FOM presents history dependent performance even after undergoing thousands of cycles at a single
rejection temperature.
14
12
Heat
eatii ng
10 Coo
Cooling
ing
TSpan [oC]
0
15 20 25 30 35 40
o
TH [ C]
Figure 3. Temperature span as a function of rejection temperature for the MnFeP1-xSix FOM AMR, at no load
conditions under heating and cooling test protocols.
69
4. DISCUSSIONS
Comparing first the results for Gd (Fig. 2) and MnFeP1-xSix AMR (Fig. 3) one can see that the Gd bed
outperforms the FOM bed under zero applied load. This is expected given the amplitude and width of the
adiabatic temperature change is considerably larger for Gd than for the FOM. For the range of temperatures
presented in Fig. 1(a), Gd has an average ΔTad of about 2oC, while for the Mn-based FOM material it is
~0.5oC. However, even with a much lower average magnetocaloric effect, the FOM develops a significant
temperature span. One of the main differences between the two regenerators is the thermal mass where the
FOM specific heat is much larger than Gd. Hence, for similar NTU, the thermal effectivenss of an FOM
regenerator can be higher than a Gd bed for the same operating conditions.
Looking now at Fig. 3, we observe two different performance curves for the heating and cooling processes,
in which the differences in the temperature span are more visible at higher rejection temperatures (> 30oC).
For instance, at a fixed TH of 34oC, a span of 10.4oC is found for the heating curve, while the span is 7.3oC
for the cooling curve. This may be associated to a combination of different physical phenomena such as the
thermal hysteresis and high dependence of the specific heat with temperature and magnetic field. However,
the results presented here are preliminary and additional testing is underway. A more detailed analysis will
be published in the near future.
5. CONCLUSIONS
In the present paper two different AMR beds are experimentally characterized in a permanent magnet
active magnetic regenerator apparatus. One test uses regenerators composed of Gd and the second
experiments use MnFeP1-xSix FOM. The tests are carried out using a fixed displaced volume and
frequency, but at different rejection temperatures. Two different process are considered: a heating
process, where the rejection temperature is increased after steady-state is reached, and a cooling
process using the reverse protocol. As expected, the Gd bed produced a single temperature span curve as
a function of TH because of negligible hysteresis. The MnFeP1-xSix AMR, on the other hand, presented two
different heating and cooling performance curves, in which major differences in temperature span are
observed above a rejection temperature of 30oC. This behavior may be associated with thermal hysteresis,
but additional analysis is needed.
ACKNOWLEDGMENTS
REFERENCES
[1] E. Bruck. “Developments in magnetocaloric refrigeration”, J. Phys. D: Appl. Phys., 38, R381-R391 (2005).
[2] A. Kitanovski. et al., Magnetocaloric Energy Conversion: From Theory to Applications. Springer
International Publishing (2015).
[3] V. Basso et al. “Effect of materials hysteresis in magnetic refrigeration cycles”, Int. J. Refrig. 29(8),
1358- 1364 (2006).
[4] K. Engelbrecht et al. “Material properties and modeling characteristics for MnFeP1-xAsx materials
for application in magnetic refrigeration”, J. Appl. Phys., 113, 173510 (2013).
[5] V. Provenzano et al. “Reduction of hysteresis losses in the magnetic refrigerant Gd5Ge2Si2 by addition
of iron”, Nature, 429, 853-857 (2004).
[6] A. Tura and A. Rowe. “Permanent magnet magnetic refrigeration design and experimental
characterization”, Int. J. Refrig. 34, 628-639 (2011).
[7] D. Arnold et al. “Design improvements of a permanent magnetic active magnetic refrigerator”, Int. J.
of Refrig. 37, 99-105 (2014).
70
TRANSIENT SIMULATION OF THE ACTIVE MAGNETIC
REGENERATOR PROCESS
T.V. Christiaanse(a)*, P. Govindappa(a), I. Niknia(a), R. Teyber(a), P.V. Trevizoli(a), A. Rowe(a)
(a)
University of Victoria, (3800 Finnerty Rd., V8P 5C2, Victoria, BC, Canada)
*
Corresponding author. E-mail: theochri@uvic.ca
ABSTRACT
A simplified modeling scheme is proposed to model hours of operation of AMR prototype systems. The
modeling scheme could allow for quick assessments of AMR sizing based on system response time
requirements. Hours of experimental operation were computed <1min on a laptop. The scheme is based on the
use of a lumped mass model. The model includes the heat load, heat leaks and AMR cooling power. We have
found that the model works well within room temperature rejection temperatures under loading conditions of
10W. However, under no load conditions the model over predicts the temperature span of the system. This
could be due to improper characterization of the heat leaks with the system. The accuracy of the model shows
potential for future development as region of validity is close to the region of interest for commercial appli-
cations.
Keywords: time dependent simulations, AMR, transient, magnetic refrigeration

DOI: 10.18462/iir.thermag.2016.0098
1. INTRODUCTION
The operation of the Active Magnetic Regenerator (AMR) is often characterized by its steady state points.
These are measured at set rejection temperatures and cold side loads. In real applications the system is
subjected to variable loading and variable rejection temperatures over numerous hours. The rate of cooling
might be important to study for some commercial applications. Medical applications for example where certain
medicine would need to be cooled down within a certain timeframe. We need to understand how fast the
system would be able to do this.
Throughout this cooling process the temperature span develops and different dynamics of the AMR are at play.
The dynamics of the cooling changes due to the non-linear behaviour of the material and regenerator
properties.
Various time dependent simulations of AMR have been presented in the literature however, these simulations
employ 1D or 2D numerical simulations [1]. Modeling the response time over many hours of operation is
difficult for these type of simulations because of computational cost [3]. Semi-empirical models on the other
hand lets us compute the cold load pumping power knowing hot and cold side temperatures of the regenerator
[2, 4]. Coupling this with a simple numerical scheme many hours of operation can be computed in matter of
seconds.
In this paper an explicit method is described to simulate the operation of the AMR process. The model is based
on a lumped mass approach. The lumped mass approach reduces the dimension of the cooling problem to a
zero dimensional problem. A control volume is set around the cold heat exchanger and an energy balance is
defined. A simplified semi empirical model is used to calculate the cooling power of the AMR at each time
step. The model is validated against two experiments done in the Permanent magnet prototype device (PM1)
build by Tura in 2010 [5]. Within the model fitting parameters are introduced to refine initial assumptions.
2. MATHEMATICAL MODEL
2.1 Energy Balance
A control volume is define around the cold heat exchanger of the PM1 device. In a refrigerator this is the
refrigeration compartment with all its content. The energy balance takes into account the capacitance of all the
components, 𝐶, within the refrigerator compartment or heat exchanger. The heat fluxes taken into account are
71
the heat leaks (𝑄𝑙𝑒𝑎𝑘𝑠 ), the applied cooling load (𝑄𝑙𝑜𝑎𝑑 ) and the cooling power of the AMR (𝑄𝐴𝑀𝑅 ). 𝑇𝑐𝑜𝑙𝑑 is
the average temperature of all the components.
𝑑𝑇𝑐𝑜𝑙𝑑
𝐶 = 𝑄𝑙𝑜𝑎𝑑 + 𝑄𝑙𝑒𝑎𝑘𝑠 − 𝑄𝐴𝑀𝑅 (1)
𝑑𝑡
The cooling power of the AMR is given by the empirical formula Eq. 2 [2, 4]. Using the implementation
method described by Burdyny [3] we are able to find the cooling power as a function of cold (𝑇𝑐𝑜𝑙𝑑 ) and hot
side (𝑇ℎ𝑜𝑡 ) temperatures of the AMR.
𝑚𝑠 𝑐𝑠 𝑈 𝑈 −1 𝑇ℎ𝑜𝑡
𝑄𝐴𝑀𝑅 (𝑇𝑐𝑜𝑙𝑑 , 𝑇ℎ𝑜𝑡 ) = 𝑈𝜉𝑇𝑐𝑜𝑙𝑑 [1 − ( + ( 𝜉) ) ( − 1)] (2)
𝜏𝐶 2 Κ 𝑇𝑐𝑜𝑙𝑑
The state variables (𝑈, 𝜉, 𝜏𝑐 , 𝑚𝑠 , 𝑐𝑠 , 𝐾) are directly connected to the material and regenerator properties. Where
𝑚𝑠 , regenerator mass , 𝑐𝑠 average specific heat conduction coefficient, 𝜏𝑐 blow time, 𝐾 effective conduction,
U the utilization and 𝜉 is the minimal adiabatic temperature change over temperature term of the material [6].
The material properties for Gadolinium (Gd) are calculated with Mean Field Theory [7]. The heat leak (𝑄𝑙𝑒𝑎𝑘 )
is determined from the empirical relationship Eq. 3 [3].
𝑄𝑙𝑒𝑎𝑘 (𝑇𝑐𝑜𝑙𝑑 , 𝑇𝑎𝑚𝑏 ) = 0.284 ⋅ (𝑇𝑎𝑚𝑏 − 𝑇𝑐𝑜𝑙𝑑 ) (3)
2.2 Numerical approximation
The forward march in time, denoted by 𝑖, of the simulation will be determined by using an explicit method.
Eq. 1 is discretized using a first order approximation. By rearranging the terms, the next time step cold side
temperature, 𝑇𝑐𝑜𝑙𝑑 (𝑖 + 1), can be expressed by known values from the previous time step, 𝑖.
𝑇𝑐𝑜𝑙𝑑 (𝑖 + 1) = 𝑇𝑐𝑜𝑙𝑑 (𝑖)

Δ𝑡 (4)
+ [𝑄 (𝑖) + 𝑄𝐴𝑀𝑅 (𝑇𝑐𝑜𝑙𝑑 (𝑖), 𝑇ℎ𝑜𝑡 (𝑖)) − 𝛾𝑙 𝑄𝑙𝑒𝑎𝑘 (𝑇𝑐𝑜𝑙𝑑 (𝑖), 𝑇𝑎𝑚𝑏 (𝑖))]
𝛾𝑐 𝐶 𝑙𝑜𝑎𝑑
The capacitance, 𝐶, and time step, Δ𝑡, are empirically determined for each device and experiment. 𝛾𝑐 and 𝛾𝑙
are introduced as fitting parameters to refine initial assumptions of the capacitance and heat leak.
2.3 Pseudo algorithm
Before the forward march in time the state variable from Eq. 2 are calculated. These can be considered constant
for second order materials. Initial conditions for the temperature of the cold side is at Room temperature. The
hot and ambient temperature are used as input for every time step to find the 𝑄𝑙𝑒𝑎𝑘 and 𝑄𝐴𝑀𝑅 . The loading
may also be used as a time dependent variable. With the use of 𝑄𝑙𝑒𝑎𝑘 , 𝑄𝐴𝑀𝑅 , 𝑄𝑙𝑜𝑎𝑑 and 𝑇𝑐𝑜𝑙𝑑 the next iteration
of 𝑇𝑐𝑜𝑙𝑑 can be determined with Eq. 4.
Figure 1: Pseudo algorithm of the transient simulation.
72
2.4 Experimental description and capacitance assumption
The model is validated against experiments

performed in the PM1 [5]. The capacitance of the Table 1: Capacitance estimation of the PMMR 1 device
cold heat exchanger of the PM 1 is estimated by a Material ρ[kg/m^3] c [J/kgK] V[m3] C[J/K]
summation of the capacitance of all the different
components of the heat exchanger assembly. An Metal 2780 880 2.448 10
. -5
599
estimated of the volume, density and specific heat of G10 1400 935 1.729.10-4 22.6
the different components is made (See Table 1). The Water 3800 960 2.210 10
. -4
806
total capacitance for the PM1 device is estimated at
𝐽 Acrylic 1200 960 5.780 10
. -5
66.6
1.49 ⋅ 103 [ ]. The regenerators are constructed by
𝐾
packing Ø=500µm Gd spheres. The spheres are Total 1.49.103
packed in a cylindrical housing. The final size of the
regenerator is 16mm OD and 55mm lengthwise. The total weight of the material is 55g of Gd spheres. The
experiment is run with a relatively large displaced volume, 5.3cm3, to reduce the effects of dead volume [8].
3. RESULTS
The experiments consist of two phases. The first phase is the heating phase where the device is ran a low
frequency (0.2-0.3Hz) and a load is applied of 20W. This collapses the temperature span after which the
sweeping phase is started. The device is ran at 1Hz, a constant load is applied and rejection temperature is
automatically changed by the control system. The rejection temperature is changed when the temperature span
changes less than 0.1C over the past 12min. The rejection (hot), cold and ambient temperature is monitored
throughout the experiment. Two experiments are performed. In the first experiment after the initial heating
phase the load is turned off. In second experiment after the initial heating phase the load is changed to 10W.
Figure 2: Transient experimental and simulation results for the 0W experiment.
Figure 3: Transient experimental and simulation results for the 10W experiment.
The simulated data is plotted in the Fig. 2 and 3 with the purple solid line. For both simulations the
𝛾𝑐 = 1 and 𝛾𝑙 = 2.6 was set. These parameters were fit using the data of the heating phase of both
73
measurements. Fig 3. shows the simulation fits well at 10W loading and between 308-298K rejection
temperatures. The simulation showed to be very fast at computing hours of operation. The loading of the
experimental data cost < 30seconds, the calculation of the regenerator properties was <15 seconds and the
computation of hours of operation was < 1 second.
4. CONCLUSIONS
After fitting the heat leak to the heating phase the simulation was run over the hours of experimental data. The
10W simulation heating and sweeping phase fit well between 308-298K rejection temperatures. The 0W
simulation fits the heating phase well but over performs during the sweeping phase. The validation shows that
with a simple numerical scheme the transient of an AMR experiment can be estimated under loading
conditions. The speed of the model is very high and the model does well under loading condition within
rejection temperatures that are close to room temperature. Various improvement to the model could be made.
The scaling of the heat leak indicates that this has to be properly measured. This includes the decoupling of
the heat exchange liquid and the heat exchanger. The cold side temperature is measured in the heat exchange
fluid. The heat exchanger is exposed to the environment and by measuring the heat exchanger temperature
independently from the heat exchange liquid a better assessment of the heat leak could be made.
ACKNOWLEDGMENTS
We would like to thank the Natural Sciences and Engineering Research Council of Canada and BASF New
Business for their support.
REFERENCES
[1] K. K. Nielsen, J. Tusek, K. Engelbrecht, S. Schopfer, a. Kitanovski, C. R. H. Bahl, a. Smith, N.

Pryds, and a. Poredos, “Review on numerical modeling of active magnetic regenerators for room
temperature applications,” Int. J. Refrig., vol. 34, no. 3, pp. 603–616, 2011.
[2] A. Rowe, “Thermodynamics of active magnetic regenerators: Part I,” Cryogenics (Guildf)., vol. 52,
no. 2–3, pp. 111–118, 2012.
[3] T. Burdyny, D. S. Arnold, and A. Rowe, “AMR thermodynamics: Semi-analytic modeling,”
Cryogenics (Guildf)., vol. 62, pp. 177–184, 2014.
[4] A. Rowe, “Thermodynamics of active magnetic regenerators: Part II,” Cryogenics (Guildf)., vol. 52,
no. 2–3, pp. 111–118, 2012.
[5] A. Tura and A. Rowe, “Permanent magnet magnetic refrigerator design and experimental
characterization,” Int. J. Refrig., vol. 34, no. 3, pp. 628–639, 2011.
[6] T. Burdyny, A. Ruebsaat-Trott, and A. Rowe, “Performance modeling of AMR refrigerators,” Int. J.
Refrig., vol. 37, no. 1, pp. 51–62, 2014.
[7] J. M. D. Coey, Magnetism and Magnetic Materials. New York: Cambridge University Press, 2006.
[8] I. Park, Y. Kim, and S. Jeong, “Development of the tandem reciprocating magnetic regenerative
refrigerator and numerical simulation for the dead volume effect,” Int. J. Refrig., vol. 36, no. 6, pp.
1741–1749, 2013.
74
LOCATION DEPENDENT PERFORMANCE STUDY ON REGENERATORS
WITHIN A HALBACH MAGNETIC FIELD SOURCE
T.V. Christiaanse*, P. Govindappa, I. Niknia, R. Teyber, P.V. Trevizoli, A. Rowe
University of Victoria, (3800 Finnerty Rd., V8P 5C2, Victoria, BC, Canada)
*
Corresponding author. E-mail: theochri@uvic.ca
ABSTRACT
This research describes the characterization of a concentric Halbach magnetic refrigerator using Gadolinium.
Experiments are performed with the active regenerator located at three different axial locations in the magnet.
Results show that the performance of the regenerator is lower when located near the ends of the Halbach
magnets as compared to the center. Numerical simulations are performed using COMSOL to analyze the effects
of field distribution, magnitude and demagnetization, in addition to the dead volumes on either side of the
regenerator. The numerical results show a large difference between the temperature span of the active part of
the regenerator and the end points of the regenerator housing where measurements are taken. These results
suggest that proper characterization of the magnetic field and the dead volume is needed to explain the
operation of the device.
Keywords: dead volume, void space, magnetic field dependence, 1D simulation, experiments, AMR, magnetic
refrigeration
DOI: 10.18462/iir.thermag.2016.0099
1. INTRODUCTION
The permanent magnet prototype device PM1 [1] is designed in a fashion that allows for easy exchange of
regenerators. Quick investigation of different materials in regenerator form is achievable. However, limited
material might be available from different manufactures. Therefore, small regenerators as described by this
paper could be used. This work investigates the impact of dead volume and field on those regenerators using
Gd, a well understood substance.
2. EXPERIMENTAL
Three different experiments were performed. Each experiment used a single regenerator with the dimensions
16mm OD, 23mm L and porosity of 0.39. A single bed is placed or at the center (z = 0) or at the edges of the
magnet (z = -5; 5), as in Fig. 1. Three different setups, hence, are subject to different applied field strengths
(Fig. 2) and different dead volume sizes are required to fill the void space between the regenerator bed and the
HEX [2]. At the center of the Halbach magnets the field wave form is higher than at the edges (See Fig. 2). By
shifting the regenerator from the cold side closer to the hot side we are also changing the void volume of the
regenerator at either side. This void volume is packed with glass spheres to reduce the void space [3].
Figure 1: Regenerator placement vs. Halbach magnets. Figure 2: Applied field as a function of Halbach angles
at set location.
75
The PM1 has a magnetic field source constructed from two Halbach magnets, 100mm wide, which rotate with
respect to each other. Regenerators are placed within the bore of the magnets within a regenerator housing that
is 19.05mm OD and 16mm ID. The total length of the housing from cold heat exchanger to hot heat exchanger
is 200mm. Three different sets of regenerators are constructed. Each regenerator is composed by 23 g of
Ø300µm Gd spheres, in which the overall dimensions are 16-mm diameter and 23-mm length, and porosity of
0.39. The packed Gd spheres are held in place by using Ø3.175mm glass spheres separated by thin stainless
steel meshes. The regenerators are placed on either side of the Halbach magnets and at the center (See Fig. 1).
This results in different applied field curves during the cycle (See Fig. 2). Tests are performed at a fixed
rejection temperature 21.1±0.5 C and fixed frequency 1Hz. The fixed rejection temperature is maintained by
using a chiller and keeping the hot heat exchanger at said temperature. The cold side hex is kept floating with
a heater attached to add additional heat.
The dead volume (DV) is determined from the placement of the regenerator, and the known total length
of the housing 200mm. The DV is reduced by filling the space with the glass spheres assumed to have 0.5
porosity (𝜖). (See Fig. 3).
Figure 3: Placement of the regenerators within the regenerator housing.
3. COMSOL MODEL
A COMSOL model is developed to understand the impact of the dead volume on the internal temperature span.
In the simulation fluid and solid phases are independently governed by their governing equations and are linked
by non-equilibrium heat transfer equation. The same closure relationships are used for both glass and Gd
spheres however, the properties due vary. The fluid phase energy equation is given by:
𝑑𝑇𝑓 𝑑𝑇𝑓 𝑑 𝑑𝑇𝑓

𝜖𝐴𝑐 𝜌𝑓 𝑐𝑝,𝑓 ( + 𝑢̅ )+ (𝜖𝐴𝑐 κdyn ) = 𝑄ℎ + 𝑄𝑙𝑒𝑎𝑘 (1)
𝑑𝑡 𝑑𝑥 𝑑𝑥 𝑑𝑥
Where 𝐴𝑐 is the cross section of the regenerator housing. 𝑇𝑓 is the fluid temperature. The conduction term
through fluid is given by the 𝜅𝑑𝑦𝑛 [4]. The heat transfer term (𝑄ℎ ) between fluid and solid is given by non-
equilibrium heat transfer equations [5]. The heat transfer coefficient ℎ𝑓 is calculated from is calculated from
the from published correlations [4], [9]. For the dead volume a degrading factor calculated by Engelbrecht is
included as the sphere size of the glass spheres reduce the known correlations [8]. The heat leak (𝑄𝑙𝑒𝑎𝑘 ) is
given by the conduction (𝑘) through the housing (G10) and air, radius of the housing and magnet (ID housing,
𝑟1 , OD Housing, 𝑟2 , magnet, 𝑟3 ) and the perimeter (𝑃𝑐 ) of the housing [5].
𝑃𝑐 (𝑇𝑎𝑚𝑏 − 𝑇𝑓 )
𝑄𝑙𝑒𝑎𝑘 = 𝑟 𝑟 𝑟 𝑟 (2)
1
ln (𝑟2 ) + 1 ln (𝑟3 )
𝑘𝑔10 1 𝑘𝑎𝑖𝑟 2
The solid phase energy equation is:
𝑑𝑇𝑠 𝑑 𝑑𝑇𝑠
(1 − 𝜖)𝐴𝑐 𝜌𝑠 𝑐𝑝,𝑠 + ((1 − 𝜖)𝐴𝑐 κstat ) = 𝑄ℎ + 𝑄𝑀𝐶𝐸 (3)
𝑑𝑡 𝑑𝑥 𝑑𝑥
Where 𝑇𝑠 is the solid temperature, 𝜌𝑠 is the density of Gd, 𝑐𝑝,𝑠 is the specific heat of Gd as a function of internal
field and temperature. The conduction term through solid is given by the 𝜅𝑠𝑡𝑎𝑡 [4]. For the magnetocaloric
76
effect (MCE) the adiabatic temperature change is used. MFT theory is used to calculate the entropy curves for
Gd [6]. The resulting entropy curves are imported into COMSOL and an interpolation function is used to find
the adiabatic temperature change as a function of starting field, ending field position and temperature. The
total increase in heat over the simulation time step is given by:
Δ𝑇𝑎𝑑 (𝐻𝑠𝑡𝑎𝑟𝑡 , 𝐻𝑒𝑛𝑑 , 𝑇𝑠 )

𝑄𝑀𝐶𝐸 = (1 − 𝜖)𝐴𝑐 𝜌𝑠 𝑐𝑝,𝑠 (𝐻, 𝑇) (4)
Δ𝑡
The internal field is reduced to the demagnetizing field in the regenerator, as follows:
𝐻𝑖𝑛𝑡 = 𝐻𝑎𝑝𝑙 − 𝑁𝑑 ⋅ 𝑀𝑔𝑑 (𝐻𝑎𝑝𝑙 , 𝑇𝑠 ) (5)
where the 𝐻𝑎𝑝𝑙 is taken from Fig. 2, the 𝑁𝑑 is the demagnetizing coefficient of the regenerator [7], 𝑀𝑔𝑑 is
calculated with MFT theory [6]. To increase computational time the 𝐻𝑎𝑝𝑙 field is directly used to find the
magnetization.
4. RESULTS
4.1 Experimental Results
For each regenerator location a no load (governed by heat

leaks) and 2.5W experiment is done. The load is applied to
the cold heat exchanger. The hot heat exchanger is kept at
20C by a chiller. The cold heat exchanger is floating. Heat
is added to the cold heat exchanger by an omega heater. In
the zero load experiment this heater is off and in the 2.5W,
2.5W is supplied. When the mean cold side temperature is
at rest the steady state temperature is taken. The
temperature span is determined from the temperature
measured at the end of regenerator housing. From Fig. 4
we see that the center placed beds are outperforming the
beds that are placed at the edges. Figure 4: Experimental results
4.2 COMSOL Results
0W 0W 0W
2.5W 2.5W 2.5W

a) Cold side b) Center c) Hot side
Figure 5: COMSOL results, Blue lines indicate the temperature of the fluid phase after hot and cold blows. The
orange lines indicate the temperature of the solid after the hot and cold blow.
77
A series of COMSOL simulations are done where the temperature span, location and total content of
the regenerator housing is taken into consideration.
5. CONCLUSIONS
From the experimental results it is very clear that the applied field has a large impact on the measured
temperature span of the bed however, the COMSOL simulations show that the dead volume acts as an extra
resistance between the regenerator and the Heat Exchangers. The temperature span is much larger across the
regenerator than the measured. Park et. al showed similar results by introducing a dead volume on the hot side
of the regenerator [3]. This increase of temperature span across the regenerator causes the heat pumped to be
much lower due to the increased conduction losses. The modeling of the glass spheres is an important to
understand the functioning of the regenerators.
ACKNOWLEDGMENTS
We would like to thank the Natural Sciences and Engineering Research Council of Canada and BASF New
Business for their financial support of this research.
REFERENCES
[1] A. Tura and A. Rowe, “Permanent magnet magnetic refrigerator design and experimental
characterization,” Int. J. Refrig., vol. 34, no. 3, pp. 628–639, 2011.
[2] P. V. Trevizoli, J. R. Barbosa, P. a. Oliveira, F. C. Canesin, and R. T. S. Ferreira, “Assessment of
demagnetization phenomena in the performance of an active magnetic regenerator,” Int. J. Refrig.,
vol. 35, no. 4, pp. 1043–1054, 2012.
[3] I. Park, Y. Kim, and S. Jeong, “Development of the tandem reciprocating magnetic regenerative
refrigerator and numerical simulation for the dead volume effect,” Int. J. Refrig., vol. 36, no. 6, pp.
1741–1749, 2013.
[4] K. K. Nielsen, G. F. Nellis, and S. A. Klein, “Numerical modeling of the impact of regenerator
housing on the determination of Nusselt numbers,” Int. J. Heat Mass Transf., vol. 65, pp. 552–560,
2013.
[5] P. V. Trevizoli, “DEVELOPMENT OF THERMAL REGENERATORS FOR MAGNETIC
COOLING APPLICATIONS,” Federal University of Santa Catarina, 2015.
[6] J. M. D. Coey, Magnetism and Magnetic Materials. New York: Cambridge University Press, 2006.
[7] D. X. Chen, J. a. Brug, and R. B. Goldfarb, “Demagnetizing factors for cylinders,” IEEE Trans.
Magn., vol. 27, no. 4, pp. 3601–3619, 1991.
[8] K. L. Engelbrecht, G. F. Nellis, and S. a. Klein, “The Effect of Internal Temperature Gradients on
Regenerator Matrix Performance,” J. Heat Transfer, vol. 128, no. 10, p. 1060, 2006.
[9] D. a Nield and a Bejan, Convection in Porous Media, vol. 165. 2006.
78
EFFECT OF PET INSULATOR LAYERS ON THE CONDUCTION LOSSES
OF PLATE ACTIVE MAGNETIC REGENERATORS
I. Mugica(a)*, S. Poncet(a) J. Bouchard(b)
Université de Sherbrooke, 2500 boul.de l’Université, J1K 2R1, Sherbrooke, Canada
(a)
(b)
Laboratoire des technologies de l’énergie, 600 av. de la Montagne, G9N 7N5, Shawinigan, Canada
*
Corresponding author. E-mail: i.mugica@usherbrooke.ca
ABSTRACT
The following paper proposes a feasible solution to an idea presented by Nielsen and Engelbrecht [1]. They
explain that higher performances are linked with lower longitudinal thermal conductivity of the magneto-
caloric material (MCM) in magnetic regenerators. The concept presented here; involves the insertion of
insulator layers along the length of the magnetic regenerator. This idea is investigated by means of a 1D
numerical model. It solves not only the energy equations of fluid and solid domains, but also the magnetic
circuit that conforms the experimental set-up of reference. In conclusion, to intercalate insulator layers within
the MCM increments the adiabatic temperature change, and thus the maximum cooling power of the machine.
Keywords: Magnetic, Refrigeration, Gadolinium, Plates, Insulator, Regenerator.

DOI: 10.18462/iir.thermag.2016.0100
1. INTRODUCTION
Many past studies have presented magnetic refrigeration systems with greater efficiencies than conventional
vapor-compression cycles. On top of that, magnetic refrigeration uses a solid refrigerant, which means that no
ozone depleting gases can be released by the use of this technology. Indeed, the principle of such promising
technology is founded on the Magneto-Caloric Effect (MCE) of materials. The MCE is a physical phenomenon
for which Magneto-Caloric Materials (MCM) change their entropy level when they are subjected to a varying
magnetic field. The most popular refrigeration cycle using MCE is the Active Magnetic Regeneration (AMR).
In the AMR cycle, magnetization and demagnetization processes are induced between oscillating fluid surges.
The oscillating flow transports the heat created during magnetization to a hot sink (Cold blow), and it also
takes heat from the cold source to warm up the regenerator again after a demagnetization step (Hot blow).
During its operation, the magnetic regenerator develops a temperature gradient inside the MCM that starts at
the cold end (next to the cold source) and finishes at the hot end, (next to the hot sink). This temperature
gradient is caused by the oscillating flow. In fact, similar to a heat exchanger, the Heat Transfer Fluid (HTF)
loses or gains temperature as it travels through the regenerator, leaving the MCM with a similar temperature
distribution. As a consequence, this temperature gradient causes regions of the solid domain to work further
form the Curie temperature, involving less work in the cycle. The temperature gradient is also the origin of
heat conduction within the MCM. As shown by [1] longitudinal thermal conductivity of the MCM should be
as low as possible, so that the temperature gradient is easier to maintain. In, [1] it is shown how increasing the
thermal conductivity of the solid leads to losses on the cooling load. Thus, they proposed that the ideal
Magneto-Caloric Material (MCM) should be highly anisotropic, having a high thermal conductivity towards
the HTF, but a low one in the longitudinal direction. This kind of material does not exist, so an investigation
was conducted to explore whether it is possible to block the conduction heat transfer within the MCM by
inserting insulator layers in the regenerator (Fig 1). In the subsequent sections, the following hypotheses are
investigated by means of a 1D numerical model:
I. Inserting insulator layers in the AMR regenerator helps reduce the conduction losses in the MCM during
steady state operation. Therefore, the maximum temperature span at zero cooling power will increase.
II. As a result of the lower temperature gradient between the insulator and the HTF, the heat transfer towards
the HTF will be damaged, and thus the optimal utilization rate will be higher.
Figure1. Disposition of the insulator layers in the regenerator
79
2. Numerical experiment
2.1 Numerical model
Like all the 1D AMR codes published so far, this mathematical model will solve the energy conservation
principle on the solid and fluid domains.
∂Ts ∂2 T ∂Tf ∂T ∂2 T
ρs c s =λs 2s +Q transfer +Q MCE ,and ρf cf ( +u f) =λf 2f +Q transfer (1)
∂t ∂x ∂t ∂x ∂x
Technically, for the MCM, 𝑐𝑠 also depends on H and T, and it should be included in the temporal derivative.
Anyway, for the sake of simplicity, 𝑐𝑠 values are updated at each time step and assumed constant. These “Eq.
(1)” will be solved at all times: during (de)magnetization, cold and hot blows, and at idle timeframes.
The mathematical model will be representative of the experimental setup of AMR “A” described in [2]. This
regenerator is made of Gd plates and the applied magnetic field (1.15T) is parallel to the plates. In [2] an
overall idle time of 0.9s during each cycle is specified. The influence of the idle time is shared in the code
between the regular processes of an AMR. Thus, the cycle develops as follows: 1.Magentization, 2.Idle time,
3.Cold blow, 4.Idle time 5.Demagnetization 6.Idle time 7. Hot blow 8.Idle time. The heat sources of “Eq. (1)”
will include the heat produced by the MCE in (de)magnetization steps, and the heat transport between solid
and fluid domains at all times.
∂T(H,T) ∂T(H,T) T ∂M(H,T) ∂H
Q MCE =ρMCM cMCM (H,T) ∂t | ,where ∂t
| =- c(H,T) ∂T | ∂t (2)
s s H
Heat capacity values were interpolated from [3], and magnetization data from [4] was fitted to an expression
presented in [5] (See Appendix). Heat transport between solid and fluid domains will happen by convection if
there is fluid flow, or by conduction if there is not.
hAHT (Tis -Tif ) λf AHT (Tsi -Tfi )
Q convection = ,or Qconduction = (3)
V 𝑉 ef /2
λf Nu
The Nusselt number used to evaluate the heat transfer coefficient h= Dh
is taken from [6] for a constant heat
exchange in rectangular ducts: 𝑁u=8.235(1-2.0421r+3.0853r -2.4765r +1.0578r4 -0.1861r5 ), where r is the
2 3
aspect ratio. In heat conduction, because λs ≫λf (See Appendix) the fluid domain alone is considered to
accurately represent Qconduction .
Finally, for an accurate representation of the experimental set-up, the effective magnetic field needs to be
calculated. In order to do that, two effects are taken into consideration: the demagnetization field and the air
gap. According to [7] and the geometry of the regenerator, D=0.0403 for a magnetic field parallel to the plates.
Thereby, the demagnetization field was deemed negligible. In regards to the air gap effect, [8] shows how the
magnetic circuit of the experimental set-up is solved for each point of the regenerator, and at each time step of
the (de)magnetization processes.
Figure 2. Sketch of the magnetic circuit.

In order to solve the magnetic circuit it is necessary to know the reluctance of the MCM, which is dependent
on the effective magnetic field (HMCM):
Lplate M(HMCM ,T)

RMCM = ,where μr (HMCM ,T)=1+ (4)
μ0 μr (HMCM ,T)eMCM dx HMCM
Happ Lgap RMCM (HMCM ,T)Rwater

HMCM = (5)
Rtot (HMCM ,T)Lplate RMCM (HMCM ,T)+Rwater
80
So, “Eq. (5)” is solved numerically with the Newton-Raphson method. The relative magnetic permeability of
HTF, air and PET casing are assumed to be μr ≈1. The numerical discretization of “Eq. (1)” was done by 2nd
order (in space and time) finite difference schemes (Crank-Nicholson, implicit, unconditionally stable). The
resulting algebraic system is solved with the TDMA algorithm. The rest of the physical properties of Gd, HTF
and PET (used for the insulator layers) can be found in the appendix.
2.2 Grid sensitivity and Validation

As is the case for all numerical models, one needs to choose a discretization size that is not too far from the
analytical solution of the mathematical model. Therefore, the spatial grid sensitivity is checked for the first
cycle of the AMR in Fig 3 below.
20
No insulator layers 1,85 1mm layers
1,8 18
1,8
1,6 16
∆T [K]
∆T [K]
1,75
∆T [K]
1,4 14
1,2 1,7 12
1 1,65 10
15 46 74 105 135 0 200 400 600 0 0,2 0,4 0,6 0,8 1 1,2
Number of nodes [-] Number of nodes [-] V [-]
Figure 3. Temperature span in the regenerator as a function of the spatial nodes after the first cycle (first two graphs).
The curves on the middle graph represent different number of insulator layers ( 5, 20). On the right, temperature
𝑚̇ 𝐻𝑇𝐹𝑐𝐻𝑇𝐹 𝜏𝑏𝑙𝑜𝑤
span in the regenerator for different utilization ratios (𝑉 = ) is shown: the line corresponds to the
𝑚𝑀𝐶𝑀 𝑐̅ 𝑀𝐶𝑀
numerical solution and X symbols correspond to the experimental data from [2].
Using the criteria from [9], when the GCI is sufficiently low and similar to its prediction, the solution is
assumed to be in the asymptotic region. These discretizations are 125 nodes for the regenerator without layers,
and 500 nodes for a regenerator with 5 and 20 layers. The predicted temperature span (shown in the last graph
of Fig 3) is close to the experimental values of [2]. The biggest discrepancies appear at low utilization rates.
In fact in [2], when operating at low mass flow rates, the author had unintended higher values of temperature
at the hot sink, making the experimental temperature span slightly higher than what was planned.
2.3 Results and discussion

26 No layers 24 No layers
24 5 22 1mm
22 10
20 2mm
20 15
∆T [K]
18 5mm
∆T [K]
18 20
16 10mm
16
14 14
12 12
10 10
0 0,5 1 1,5 0 0,5 1 1,5
V [-] V [-]
Figure 4. On the left, maximum temperature span for different number of evenly distributed1mm layers. On the right
maximum temperature span for 5 layers with different thicknesses. V corresponds to the utilization ratio.
Fig 4 shows how incrementing the number of insulator layers increases the maximum temperature span, at low
utilization ratios. If the length of the passive layers is increased though, the temperature span at higher
utilization augments. The graph on the left supports the first hypothesis, and it confirms how the continuous
heat exchange between thermodynamic cycles that normally undergo all the points of the regenerator is
detrimental. Indeed, to form AMR cascades seems to be beneficial not only for the application of graded
alloys, but for single MCM regenerators as well. However, looking at the graph on the right, it is thought that
the second hypothesis cannot be confirmed. In fact, the optimal utilization rate moves very little to the right
(from 0.3 with 1mm layers, to 0.35 with 10mm layers). So, because of the additional length that the insulator
layers add to the regenerator, the HTF might just need to travel a little bit more to have the same effect on the
81
cycle. Anyway the maximum temperature span is still visibly incremented by the insulator layer length at
higher utilization rates. This is thought to come from the role of the insulator layers in the cycle as passive
regenerators.
3. CONCLUSIONS
This paper constitutes the first attempt to propose a viable solution to the idea presented in [1]. This attempt
consists on reducing the longitudinal thermal conduction within the MCM, by the insertion of insulator layers.
Results show higher temperature spans at zero cooling power with the same amount of MCM. This increment
in maximum temperature span does not come without a cost though. Apart from incrementing the pumping
loses due to a longer regenerator, the spatial distribution of relative magnetic permeability is continuously
changing: from μr ≈1 for the PET layers, to μr =1~6 for the Gd ([8]). This means an increment of the force
needed to move the magnet (dF=(M∙∇)Hdν, [10]) and thus, of the total work. Thereby, it would be of interest
for future work to optimize the number of layers needed to maximize the COP.
REFERENCES
[1] K.K. Nielsen, K. Engelbrecht, “The influence of the solid thermal conductivity on active magnetic
regenerators”, J. Phys. D: Appl. Phys. 45, 7 (2012).
[2] J. Tušek et al., “A comprehensive experimental analysis of gadolinium active magnetic regenerators”,
Applied Thermal Engineering 53, 57-66, (2013).
[3] M. Risser et al., “Construction of consistent magnetocaloric materials data for modelling magnetic
refrigerators”, International Journal of refrigeration 35, 459-467, (2012).
[4] J. S. Lee, “Evaluation of the magnetocaloric effect from magnetization and heat capacity data”, Phys. Stat.
Sol. (b) 241 No.7, 1765-1768, (2004).
[5] W. F. Brown, “Theory of the Approach to Magnetic Saturation”, Physical Review 58, 736-743, (1940).
[6] W.M. Rohsenow et al., “Handbook of heat transfer”, McGraw-Hill (1998).
[7] A. Aharoni, “Demagnetizing factors for rectangular ferromagnetic prisms”, Journal of Applied Physics 83
No. 6, 3432-3434, (1998).
[8] D. Vuarnoz, T. Kawanami, “Experimental validation of a coupled magnetic-thermal model for a flat-
parallel-plate active magnetic regenerator”, Applied Thermal Engineering 54, 433-439, (2013).
[9] P.J. Roache, “Perspective: A Method for Uniform Reporting of Grid Refinement Studies”, Journal of Fluid
Engineering 116, 405-413, (1994).
[10] F. Allab, A. Kedous-Lebouc, J.P. Yonnet, J.M. Fournier, “A magnetic field source system for magnetic
refrigeration and its interaction with magnetocaloric material”, International Journal of refrigeration 29, 1340-
1347, (2006)
Appendix
a(T) b(T) c(T) d(T)

M(T,H)=Ms (T) [ + + + ] +e(T)H
H1/2 H1 H3/2 H2
-0.04583T2 +27.08T-3954 -1.613

Ms(T)= -0.00245T2 +0.438 a(T)= +0.5643
1+0.732 0.4213(T-291) 1+0.7320.4213(T-291)
0.7575 -0.1173
b(T)= -0.3996 c(T)= +0.06826
1+0.7320.4213(T-291) 1+0.7320.4213(T-291)
0.005889 0.07188T-21.99
d(T)= -0.003573 e(T)= +0.03931T-4.578
1+0.7320.4213(T-291) 1+0.7320.4213(T-291)
Table 1. Fitted function of magnetization to data from [4]. The ratio between the variance of the experimental data, and
the fitted function is R2 =0.9981.
Gd PET HTF (67%water+33% ethylene glycol)

𝑐 [J/(kg K)] See [3] 1425 36628
𝜌 [kg/m3 ] 7900 920 1042
𝜆 [W/(m K)] 10.5 0.3 0.4534
Table 2. Physical properties of substances in the regenerator.
82
INFLUENCE OF VOID VOLUME AND INLET FLOW MALDISTRIBUTION
ON THE PERFORMANCE OF THERMAL REGENERATORS
P.V. Trevizoli(a,b)*, M.S. Capovilla(a), G.F. Peixer(a), A.T. Nakashima(a), J.A. Lozano(a),
J.R. Barbosa Jr.(a)
(a)
Polo ‐ Research Lab. for Emerging Technologies in Cooling and Thermophysics. Department of
Mechanical Engineering, Federal University of Santa Catarina, Florianópolis, SC, Brazil.
(b)
Institute for Integrated Energy Systems, Department of Mechanical Engineering, University of
Victoria, Victoria, BC, Canada
*
Corresponding author. E-mail: paulot@uvic.ca
ABSTRACT
Void (dead) volumes represent a penalty to the thermal performance of Active Magnetic Regenerators
(AMR). While much effort has been put in designing AMRs with small void volumes, in most designs,
penalties associated with inlet flow maldistribution and pressure loss actually increase with the reduction of
the void volumes. In this work, an experimental evaluation of the combined effect of void volume and inlet
flow maldistribution is carried out. The analysis is performed based on the regenerator thermal effectiveness
determined experimentally in a passive regenerator test apparatus. The results show that the inlet flow
maldistribution has a more significant negative impact on the regenerator effectiveness than the void
volume.
Keywords: Void volumes, inlet flow maldistribution, passive regenerator apparatus, thermal effectiveness.
DOI: 10.18462/iir.thermag.2016.0102
1. INTRODUCTION
The performance of active magnetic regenerators (AMR) is penalized by magnetic, thermal and flow losses
[1]. Magnetic and thermal losses have been studied in several works published recently [1‐4]. On the other
hand, void volumes and inlet flow maldistribution are, in general, not considered in AMR performance
analyses or in numerical models. This work presents an experimental evaluation of the combined effect of
void volume and inlet flow maldistribution. Three different regenerator‐to‐tube (R‐T) adapters are
evaluated, as shown in Fig. 1(a). R‐T adapter #1 has a large void volume ( and a large opening angle ( ),
while R‐T adapter #3 has a small void volume and a small opening angle. The analysis is based on the
regenerator thermal effectiveness determined experimentally in a passive regenerator test apparatus. For a
more in‐depth analysis, three different packed bed matrices composed of stainless steel spheres, with the
same cross sectional area but different aspect ratios were tested. Another regenerator geometry with a
rectangular cross sectional area and a negligible void volume (Fig. 1(b)), proposed in Ref. [5], was also
experimentally tested. The results show that the inlet flow maldistribution has a more significant impact on
the regenerator effectiveness than the void volume, directly reducing the AMR performance.
Figure 1. (a) R-T adapters with different void volumes [4]; (b) Regenerator housing designed with rectangular cross
sectional area and zero void volume [5].
2. EXPERIMENTAL
The experimental evaluation of the pressure drop and thermal effectiveness was carried out in a passive
regenerator test apparatus at Federal University of Santa Catarina. More information can be found in Ref.
[4, 6]. In the tests with different R-T adapters, Fig. 1(a), three regenerative matrices with the same diameter
(22.22 mm), but different lengths (Casings #1 to #3), have been used. Therefore, different void volumes
fractions
83
were obtained. The matrices were composed of 1-mm stainless steel spheres. The porosity was equal to
0.36. The void volume fraction ( ∗) is defined as the ratio between the dead volume ( ) (internal volume of
the R-T adapter) and the fluid volume inside the porous media ( ), where is the porosity and the
regenerator total volume.
In the tests with the zero dead volume regenerator ( ∗ 0 ), Fig. 1(b), the matrix (Casing #7) has 27.85-
mm width, 10-mm height and 80-mm length. The matrices were composed of 0.5-mm and 1-mm stainless
steel spheres. The porosity was about 0.37. Table 1 summarizes the properties of all the matrices tested.
Table 1. Dimensional properties and void volumes fractions for different regenerator matrices.
∗
Casing Length [mm]
R-T #1 - 67.4° R-T #2 - 53.1° R-T #3 - 31.0°
#1 50 14.5% 9.67% 7.25%
#2 75 11.2% 7.44% 5.58%
#3 100 7.66% 5.1% 3.82%
#7 80 0
The tests were performed under different utilization factors (0.2 to 1.4) and operating frequencies (0.25 to 2
Hz). The temperature span imposed on the passive regenerator test was 40 K; the hot reservoir was kept at
320 K and the cold source was at 280 K. The ambient temperature was kept at 300 K. Distilled water was
used as the thermal fluid [4, 6].
3. RESULTS AND DISCUSSIONS
3.1 Results for different R-T adapters
As discussed in Ref. [4, 6], the presence of a void volume imposes changes on the boundary conditions for
the thermal balance in the regenerator matrix that are responsible for decreasing the heat transfer
effectiveness. Hence, it becomes intuitive that the regenerator performance improves as the void volume
becomes smaller and, in a limiting situation, a regenerator without void volumes will have the best
performance. The experimental results in Fig. 2(a) (Casing #1), on the other hand, exhibit an opposite
behavior, i.e., by decreasing the void volume size, the regenerator effectiveness is reduced. The same
behavior was observed in Casing #2. In Casing #3 (Fig. 2(b)), the effectiveness reduction was smaller, but
no improvement of the effectiveness was observed.
The reduction of the regenerator effectiveness can be explained by a maldistribution of the inlet flow
through the regenerator. The void volume is reduced as the diverging angle ( ) of the inlet R-T adapter
becomes smaller (see Table 1), approaching the shape of a sudden expansion. If an abrupt variation in cross
sectional area exists, especially for R-T Adapter #3, an abrupt change in average flow velocity is
experienced. Thus, a flow recirculation region is created inside the matrix, near the inlet. The fluid in this
recirculation region does not interact effectively with the inlet stream, which acts as an internal void space
that reduces the total heat transfer area and the effective thermal mass of the matrix. As a result, the heat
transfer effectiveness decreases. Also, the void region creates internal gradients in the transverse direction.
Therefore, the inlet flow maldistribution imposes heavier penalties on shorter matrices, such as Casing #1,
as the length fraction of the recirculation region is larger. In longer matrices, such as Casing #3, the
influence of the flow maldistribution is smaller. However, this was not sufficient to improve the
performance in comparison with a regenerator with a larger void volume.
3.2 Results for the regenerator with no void volume
Lozano [5] proposed a regenerator housing with a round rectangular cross sectional area, in which the inlet
and outlet streams are different, as can be seen in Fig. 1(b). This geometry guarantees a zero void volume
but the inlet flow tubing is similar to R-T Adapter #3. Fig. 3 compares the experimental results for the 1-
mm spheres with a validated passive regenerator model presented in [4]. The tests with 0.5-mm spheres
presented similar results.
84
(a)
(b)
Figure 2. Results for the effectiveness as a function of the NTU for different utilization factors (ϕ) and
different R-T adapters: (a) Casing #1; (b) Casing #3.
As can be seen in Fig. 3, there are considerable differences, as high as 7%, between the numerical and
experimental effectiveness results. These discrepancies in the thermal effectiveness can also be attributed to
the inlet flow maldistribution. The design of Casing #7 gave rise to recirculation regions near the flow inlet,
creating a large thermally inactive region, which reduced the effective heat transfer area, also creating
internal temperature gradients transversal to the flow direction. In addition, in this case the 1-D model
approach is not capable of predicting these losses because they involve two-dimensional effects related to
the fluid flow and heat transfer. In summary, the design geometry chosen for Casing #7, albeit having a zero
void volume, deteriorated the regenerator thermal performance compared with a regenerator with a round or
square cross sectional area, in which the inlet flow maldistributions are reduced. Also, as in Ref. [4] the
Casing #7 also increased the viscous losses due to the two-dimensional nature of the fluid flow.
85
Figure 3. Results for the effectiveness as a function of the NTU for different utilization factors (ϕ): Casing #7.
4. CONCLUSIONS
The present paper showed how the void volumes and inlet flow maldistribution can impact negatively on
the performance of thermal regenerators. Also, in regenerators with two-dimensional effects, their thermal-
hydraulic performance is not accurately predicted by 1-D models. A forthcoming paper will discuss how
these effects can influence the thermal performance of an AMR.
ACKNOWLEDGMENTS
Financial support from Embraco and CNPq through Grant No. 573581/2008-8 (National Institute of Science
and Technology in Cooling and Thermophysics) is duly acknowledged.
REFERENCES
[1] P.V. Trevizoli et al. "Magnetic heat pumps: an overview of design principles and challenges", Sci.
and Techn. for the Built Environment, in press (2016).
[2] K.K Nilesen et al. “Review on numerical modeling of active magnetic regenerators for room
temperature applications”. Int. J. of Refrig. 34, 603-616 (2011).
[3] T. Burdyny et al. “Performance modelling of AMR refrigerators”. Int. J. of Refrig. 37, 51-62 (2014).
[4] P.V. Trevizoli. Development of thermal regenerators for magnetic cooling applications. PhD
Thesis (2015).
[5] J.A. Lozano. Designing a rotary magnetic refrigerator. PhD Thesis (2015).
[6] P.V. Trevizoli. et al. “Thermal–hydraulic evaluation of oscillating-flow regenerators using
water: Experimental analysis of packed beds of spheres”, Int. J. Heat and Mass Transfer 99, 918-930
(2016).
86
MODELING AND SIMULATION OF A THERMOMAGNETIC
RECIPROCATING ENGINE BY GRAVITY
A. C. Souza(a), E. Evaristo(a), F. C. Colman(a), W. A. S. Conceição(a), G. L. Foleiss(a), W.
Szpak(b), A. M. G. Carvalho(c), C. S. Alves(a)
(a)
Universidade Estadual de Maringá – UEM, Bloco 104, Avenida Colombo, 5790 – Zona 07
(Maringá/Brazil)
(b)
Universidade Tecnológica Federal do Paraná – UTFPR, Rua Marcílio Dias, 365 – Jardim Paraíso
(Apucarana/Brazil)
(c)
Laboratório Nacional de Luz Síncrotron - LNLS, CNPEM, Rua Giuseppe Máximo Scolfaro, 10000 - Guará
(Campinas/Brazil)
Alisson Cocci de Souza. E-mail: pg48958@uem.br
ABSTRACT
Worldwide concern over global warming and the search for less harmful energy sources to the environment
has driven the scientific community to develop means of sustainability. In this sense, we have developed a
gravitational thermomagnetic engine, based on Tesla’s engine operating principle. A magnetic plate
containing MnAs was used as working piece. An electric resistor was used as energy source to heat the plate.
We present in this paper our mathematical model and calculations, which simulate the engine’s dynamics.
Using an interactive software, we studied the heat transfer process on the plate using a one-dimensional finite
element model associated with the thermomagnetic properties of the material. Some performance parameters
were obtained and the engine applications are also discussed.
Keywords: magnetocaloric effect, thermomagnetic engine, finite element method.

DOI: 10.18462/iir.thermag.2016.0104
1. INTRODUCTION
Demand for new and renewable energy sources are growing as fossil fuels become scarce and environmental
concerns are gaining space in our society.
In this context, the magnetocaloric effect (MCE) presented by the magnetic materials is an alternative for
energy. Upon an adiabatic process, the MCE is a thermal response of a material when subjected to a
magnetic field variation [1].
Some MCE applications have been studied by the scientific community, such as the thermomagnetic engine,
where a temperature gradient is used to provide power. Also in 1889, the Austrian engineer naturalized
American, Nikola Tesla, proposed simple arrangements of an alternated thermomagnetic engine. [2].
The interest in these devices has increased recently with the discovery of the giant and the colossal
magnetocaloric effect [3,4]. Since then, several studies have been developed approaching the modeling,
simulation or construction of another thermomagnetic engine arrangement, called Curie wheel, which the
power is provided by a movement of a circular anchor [5,6,7]. However, there are few studies about the
Tesla engine.
In this context, the purpose of this paper is to develop the modeling and simulation of the operation of a
thermomagnetic engine, in which the power is provided by a reciprocating motion, similar to the
arrangement proposed by Tesla, where the restitution force is due to gravity. Through this simulation, we
intended to obtain some performance parameters, as torque and frequency, and examine possible engine
applications.
2. METODOLOGY
The engine arrangement consists basically in a plate containing a MnAs compound and a Nd-Fe-B
permanent magnet. The plate, initially at rest, at room temperature (298 K), which is below the Curie
temperature (Tc) of MnAs, is attracted by the magnetic field generated by the permanent magnet. Thus, a
resultant force arises on the plate, moving it vertically towards to the magnet gap (Fig. 1a). Upon arriving at
the end of its course, an electrical resistor located inside the plate is activated, heating the plate from the
center to the boards. With the increase of the plate temperature, eventually the material reaches the transition
87
temperature Tc and it passes from the ferromagnetic phase to the paramagnetic phase. Consequently, the
material magnetization reduces and the weight force increases its influence on the plate, causing it a
descendent movement (Fig 1b). With the start of the descendent movement, the electrical resistor is turned
off. Reaching start position, the plate is cooled by a water flux at room temperature (298 K), bringing the
plate back to initial conditions, where the material returns to ferromagnetic phase and restarts the cycle. In
the modeling, the cooling process of the plate was considered instantaneous.
Figure 1. Schematic representation of engine operation
The plate has a length of 0.1 m, height of 0.1 m and a thickness of 0.01 m. During modeling, temperature and
magnetic analysis, the finite elements method was used. The plate was divided in discrete elements with both
length and high of 0.01 m. As the resistor is along the centerline of the plate, it was considered that the
heating takes place symmetrically relative to this line, in a one-dimensional x-direction (Fig. 2).
Figure 2. Board nodes analyzed during simulation
The nodes temperature has been determined by applying an energy balance, obtaining the equations (1), (2)
and (3) [8]:
𝒑+𝟏 𝒑 𝒑
Node 1: 𝑻𝒎 = 𝟐𝑭𝒐(𝑻𝒎 + 𝑩𝒊𝑻∞ ) + (𝟏 − 𝟐𝑩𝒊𝑭𝒐 − 𝟐𝑭𝒐)𝑻𝒎 (1)
𝒑+𝟏 𝒑 𝒑 𝒑
Node 2 to 5: 𝑻𝒎 = (𝟏 − 𝟐𝑭𝒐)𝑻𝒎 + (𝑻𝒎−𝟏 + 𝑻𝒎+𝟏 )𝑭𝒐 (2)
𝒑+𝟏 𝒑 𝒑
Node 6: 𝑻𝒎 = 𝟐𝑭𝒐𝑻𝒎−𝟏 + (𝟏 − 𝟐𝑭𝒐)𝑻𝒎 , (3)
where the indices p and m, indicate the instant and the position of the node analyzed and Fo e Bi are the
Fourier and Biot number, respectively.
The magnetic circuit was analyzed based on Furlani’s procedure [9], from which magnetic field and
magnetic flux density in the magnet gap are determined. Magnetic energy difference between two nodes was
used to calculate the magnetic force acting on the plate, given by eq. (4) [7]:
𝒑 𝒑
𝑾𝒈,𝜟𝒙 −𝑾𝒈
𝑭𝒎𝒂𝒈 = | 𝜟𝒙
|, (4)
where Fmag is the magnetic force, W is the magnetic energy and Δx is the dimension of one discrete element.
88
A connecting rod-crank system linked to the plate was considered for torque developed in the system. For
this analysis, the dimensions of 100 mm and 23.5 mm were adopted for the rod length and the crank radius,
respectively and applied to the eq. (5) [10]:
𝒓
𝑴𝒕 = 𝑭𝒓 [𝟐𝑳 𝒔𝒆𝒏(𝟐𝜽) + 𝒔𝒆𝒏𝜽], (5)
where Mt is the torque developed, F is the resultant force, r is the crank radius, L is the rod length and θ is the
turning angle of the crank.
The temperature variation in the six analyzed nodes is shown on Fig. 3. The periods where the temperature
remains constant corresponds to the rising and fall time of the plate. The abrupt decrease in temperature
corresponds to the instant cooling of the plate, promoted by cooling water. Through Fig. 3, it is possible to
note the period of each cycle fulfilled by the plate, corresponding to 23.22 seconds. Thus, the engine
frequency is 4.307 x 10-2 Hz.
Figure 3. Temperature of the nodes during operation of the engine
The resultant force and the variation in plate position are shown on Fig. 4. One can see that the initial force is
7.88 N, where the plate is attracted to the magnet gap. It is possible to note a maximum force of 12.95 N,
achieved during the upward trajectory. Periods where the force is zero are the ones where the plate remains
still, inside the gap. The constant force of 6.48 N acting on the plate during its descending movement stands
for its weight force, and in this interval, it does not have an expressive magnetic force.
Figure 4. Net force acting on the plate
The torque behavior developed by the rod-crank mechanism is shown on Fig. 5. As one can expect, in a
similar way to force, when the plate finds itself inside the gap torque is null. In the periods where the plate is
moving, it is possible to note that the system develops bigger torque during the ascending movement,
89
showing a maximum value of 30.5 N.cm. Also, during the descending movement, the maximum value is of
12.5 N.cm. During the cycle, in the maximum torque moment, (30.5 N;cm), the power of the engine is 8.25 x
10-2 W. This value doesn’t permit practical application to the motor. In addition to this, the low frequency
contributes to this low performance.
Figure 5. Torque obtained during engine operation
4. CONCLUSION
As initially proposed, this paper exposed the behavior of some performance parameters of an
alternated movement thermomagnetic engine through the modeling and computer simulation.
Analyzing the obtained data, it is legit to conclude that, for the proposed arrangement, the engine
still does not have practical applications due the low values of torque and frequency. This leads us
to think of new approaches for future papers, such as changing the plate’s magnetic material, adopt
a permanent magnet with a higher remanence or modifying system’s arrangement, in order to
analyze performance parameters as a function of those variables.
ACKNOWLEDGMENTS
The authors are grateful to CAPES and CNPq for providing financial support for our research.
REFERENCES
[1] P. Weiss, A. Piccard, Le phénomène magnetocalorique, Journal. de Physique, 5th serie, 7, p.103 (1917).
[2] N. Tesla, Thermo-Magnetic Motor. US 396.121 Patent (1889).
[3] V. K. Pecharsky, K. A. Gschneidner Jr., Giant magnetocaloric effect in Gd5(Si2Ge2). Physical Review
Letters, 78, p.4494 (1997).
[4] S. Gama, A. Coelho, A. Campos, A. M. G. Carvalho, F. C. G. Gandra, P. J. von Ranke, N. A. de Oliveira,
Pressure-induced colossal magnetocaloric effect in MnAs, Physical Review Letters, 93, p. (2004).
[5] Y. Takahashi, H. Yamamoto, M. Nishikawa, Fundamental performance of triple magnetic circuit type
cylindrical thermomagnetic engine, Eletrical Engineergin in Japan, 154, p.68 (2006).
[6] D. A. Gabielyan, V. V. Semenov, D. S. Martirosov, Analysis of nonstationary heating and cooling of a
thermomagnetic engine gadolinium working element, Russian Aeronautic, 56, p.266 (2013).
[7] C. S. Alves, F. C. Colman, G. L. Foleiss, W. Szpak, G. T. F. Vieira, A. C. Bento, Simulation of solar
Curie wheel using NiFe alloy and Gd, International Journal of Refrigeration, 37, p.215 (2014).
[8] F. Incropera, D. Dewitt, T. Bergman, A. Lavine, Fundamentos de Transferência de Calor e de Massa.
6th ed., LTC (2008).
[9] E. Furlani, Permanent Magnet and Electromechanical Devices: Materials, Analysis and Applications,
first ed., Academic Press, p. 518 (2001).
[10] F. Brunetti; Motores de combustão interna: volume 2, first ed., Bucher, p.302 (2012).
90
INVESTIGATION OF AN ACTIVE MAGNETIC REGENERATOR SUBJECT
TO DIFFERENT FLUID FLOW WAVEFORMS
R. Teyber*, P.V. Trevizoli, T.V. Christiaanse, P. Govindappa, I. Niknia, A. Rowe
Institute for Integrated Energy Systems, Department of Mechanical Engineering, University of Victoria,
Victoria, BC, Canada
*Corresponding author. E-mail: rteyber@uvic.ca
ABSTRACT
A flow control device based on cam actuated valves is designed and implemented on the PM II AMR test
apparatus. The objective is to overcome the brief low field period of the nested concentric Halbach array by
decreasing the fluid blow width; displacing fluid only when the magnetic field is close to the minimum and
maximum values. Flow waveforms are simulated to evaluate varying blow durations with the same displaced
volume. AMR experiments are performed which demonstrate an 11.7% increase in exergetic cooling power
with a blow fraction of 59%.
Keywords: Active magnetic regenerator, magnetic refrigeration, Halbach array, flow control, velocity
waveform
DOI: 10.18462/iir.thermag.2016.0105
1. INTRODUCTION
A large number of AMR configurations can be created from the available combinations of magnetic field
generators, fluid flow systems, regenerator geometries and magnetocaloric materials [1, 2]. The AMR
prototypes developed at the University of Victoria [3, 4] use nested concentric Halbach cylinders as a magnetic
circuit and a piston to sinusoidally drive fluid in the regenerator matrix. Although nested concentric Halbach
arrays generate high magnetic fields, with constant angular velocity the magnetic field waveform has a brief
low field period which reduces AMR performance [5]. In the present work, the sinusoidal fluid velocity
waveform is modified so that fluid flow occurs when the magnetic field is near the maximum and minimum
values. A new flow control system is implemented using a cam-actuated valvetrain which selectively diverts
fluid from the regenerators to shape a user-defined fluid blow. The cam actuated flow system is modeled and
simulations are used to design waveforms of varying blow widths with equal displaced volumes. AMR
experiments are conducted with three blow widths of constant displaced volume.
2. VALVING SYSTEM
A schematic of the PM II device with the bypass loop is shown on the left of Fig. 1 and the camshaft prototype
is shown on the right of Fig. 1. A total of four one-way, normally closed valves are used (Clippard MJV-
2 Valve and MBA-1 Actuator); two of which modulate flow to the PM II device (1A, 2A) while the other two
modulate each direction of the bypass blow (1B, 2B). Flow is diverted when the bypass valves open, creating
a static period of no fluid flow in the regenerators. The fluid blow begins when the PM II valves open and the
bypass valves close over a transition period of 22.5 degrees, limited by the tangential loading of the cam on
the valve stem.
The flow in this valve transition period is simulated to insure waveforms are evaluated at constant displaced
volume, as this strongly affects AMR operation. The bypassed PM II schematic is modeled as two fluid loops
in parallel; a bypass loop (B) containing a single control valve, and a PM II loop (A) treated as a control valve
in series with a packed bed of spherical particles. The porous media is modeled with the Ergun equation and
∆𝑃
each valve is treated as a minor head loss, 𝐾𝑣𝑎𝑙𝑣𝑒 = 1 2 , which varies with the valve stem position. The
𝜌𝑣𝑓𝑙𝑢𝑖𝑑
2
minor head loss is evaluated by measuring the pressure drop and the fluid velocity through the valve for ten
stem depressions ranging from fully closed to fully open. In general, the pressure drop across the valve is very
low relative to the packed bed (~0.1 bar) until the valve abruptly closes. Assuming an incompressible fluid, the
pressure drop expression across each branch is shown in Eq. 1 and Eq. 2. In a pipe network, the pressure drop
across parallel branches is balanced, ∆𝑃𝐴 = ∆𝑃𝐵 , and mass is conserved at a junction, shown in Eq. 3.
91
Figure 1. (Left) PM II schematic with bypass loop to control blow width, (Right) Camshaft assembly.
ρ 𝑄𝐴2 𝜇(1 − 𝜖)2 Q A (1 − 𝜖)𝜌 Q2A (1)

∆𝑃A = K valve (𝑥A ) ( 2 ) + 𝐿𝑏𝑝 [150 ( ) + 1.75 ( 2 )]
2 𝐴valve 𝐷𝑝2 ϵ3 𝐴Reg 𝐷𝑝 𝜖 3 AReg
ρ Q2B (2)
∆𝑃𝐵 = K valve (𝑥𝐵 ) ( 2 )
2 𝐴valve
𝐴𝑝𝑖𝑠𝑡𝑜𝑛 𝑟𝑐𝑟𝑎𝑛𝑘 𝜔 sin(𝜔𝑡) = 𝑄𝐴 + 𝑄B (3)
where 𝑄 is the volumetric flow rate, 𝜌, 𝜇, 𝜖, 𝐿𝑝𝑏 and 𝐷𝑃 are the fluid density, fluid viscosity, bed porosity, bed
length and particle diameter, respectively, and 𝐴 is the flow area of the piston, regenerator or valve. Solving
the system of equations at each time step, with the parameters in Table 1, yields the simulated flow rate
waveform. The solution is repeated in a binary search to identify cam profiles (𝑥𝑎 (𝑡), 𝑥𝑏 (𝑡)) with equal
displaced fluid volumes. The simulated waveforms are shown in Fig. 2 for the existing sinusoidal waveform
and two decreasing blow widths. The simulated magnetic field waveform is shown as a reference. The bypass
loop is preferential until the bypass valves are nearly closed; causing an abrupt transition between the simulated
no blow and blow states.
Table 1 : Simulation parameters
Parameter Value Parameter Value

𝑓 0.8 (ℎ𝑧) mtotal (Gd) 0.25 (𝑘𝑔)
𝑇𝐻 298 (𝐾) Aspect ratio (D/L) 2.9 (𝐷/𝐿)
∆𝑉𝐷 13.9 (𝑐𝑚3 ) DP (μm) 300 (𝜇𝑚)
92
Figure 2: Simulated fluid waveforms of constant displaced volume with magnetic field waveform.
3. EXPERIMENTAL RESULTS
𝑇𝑠𝑝𝑎𝑛
Initial experimental results for the temperature span and exergetic cooling power, 𝐸𝑥𝑄 = 𝑄𝐶 ( 𝑇𝐶
), as a
function of the applied load for the three fluid waveforms are shown in Fig. 3. The 59% bypassed waveform
develops both the largest no-load temperature span (16.4 C vs. 16.1 C) and the largest exergetic cooling power
(1.62 W vs. 1.45 W), obtained at an applied load of 60 W. Although the results suggest that more waveforms
should be tested between the 59-100% blow fraction, the PM II’s stroke lengths are fixed in increments of 5
mm. Further decreasing the blow width to 46% does not yield performance improvements.
Figure 3: Experimental temperature span and exergetic cooling power as a function of applied load.
93
4. DISCUSSIONS
Decreasing the blow width increases the flow-averaged magnetic field change, which increases the cooling
power and temperature span. The flow rate, however, increases as the blow width decreases (100%, 59%, 46%)
to displace the same fluid volume, as in Fig. 2. Elevated flow velocities increase dispersion and decrease the
NTU. Additionally, thermal losses may penalize the AMR operating with a significant no-blow period. As a
result, an optimal blow width exists which balances these effects; obtained with a 59% blow width for the
presented data.
Although the experimental results demonstrate performance improvements with the bypassed fluid waveforms,
an unbalanced displaced fluid volume is present. Unbalanced flow is when the displacement of fluid in one
blow direction exceeds the displacement of fluid in the returning blow, causing a net circulation of flow. The
unbalanced displaced volume is caused by the large torques between the PM II Halbach cylinders, where a
brief change in rotational velocity coincides with the valve transition period. This brings warm fluid from the
hot reservoir into the regenerator [6], creating a difference in cold side temperature measurements. At the peak
exergetic cooling power, the difference in cold side temperature measurements is 0.6 C for the 46% blow
fraction, compared to 0.1 C for the 59% and 100% blow fractions. Several device alterations are in progress
to address the unbalance and improve the experimental procedure and the results will be published in a future
paper.
5. CONCLUSION
The PM II AMR test apparatus uses nested concentric Halbach cylinders which have a brief low field period.
In the present work, the sinusoidal fluid velocity waveform is modified with a no-blow period to displace fluid
when the magnetic field variation is increased. To accomplish this, a cam-actuated valvetrain is developed
which diverts fluid from the regenerators to shape a user-defined fluid blow. The fluid waveforms are simulated
to evaluate waveforms of varying blow widths with equal displaced volumes. AMR experiments are conducted
with three blow widths of constant displaced volume, where the 59% blow width develops both the largest no-
load temperature span (16.4 C) and exergetic cooling power (1.62 W). The experimental results show that the
performance of an AMR with nested Halbach arrays can be improved by overcoming the brief low magnetic
field period, however further work is required to balance the displaced fluid volumes and explore a range of
displaced volumes.
ACKNOWLEDGMENTS
REFERENCES
[1] A. Kitanovksi et al., “Magnetocaloric Energy Conversion: From Theory to Applications”, Springer
International Publishing (2015).
[2] P.V. Trevizoli et al., “Magnetic heat pumps: an overview of design principles and challenges”, Accepted
for publication in Science and Technology for the Built Environment.
[3] A. Tura, A. Rowe, “Permanent magnet magnetic refrigerator design and experimental characterization”,
I.J. Refrigeration 34, 628 (2011).
[4] D. Arnold et al., “Design improvements of a permanent magnet active magnetic refrigerator”, I.J.
Refrigeration 37, 99 (2014).
[5] P.V. Trevizoli et al., “Modeling of thermo-magnetic phenomena in active magnetic regenerators”, ASME
Proceedings, HT 2013-17108 (2013).
[6] D. Eriksen et al., “Effects of flow balancing on active magnetic regenerator performance”, A. Thermal
Engineering 103, 1 (2016).
94
EXPERIMENTAL EVALUATION OF TWO-MATERIAL ACTIVE
MAGNETIC REGENERATORS
R. Teyber*, P.V. Trevizoli, T.V. Christiaanse, P. Govindappa, I. Niknia, A. Rowe
*Corresponding author. E-mail: rteyber@uvic.ca
ABSTRACT
Four regenerator compositions comprised of two-layers are experimentally characterized in PM II with an

interface temperature sensor. In all the regenerators the warm layer is packed with Gd, while the cold layer
consists of different Gd1-xYx alloys. The performance is analyzed assuming a linear temperature profile
between the hot, interface and cold side temperature measurements. The linear temperature distribution is used
to evaluate the adiabatic temperature change, specific heat and entropy change distributions in each
multilayered regenerator. The regenerator with the largest refrigerant capacity (RC’) is found to develop the
largest temperature span at each operating condition.
Keywords: Active magnetic regenerator, magnetic refrigeration, magnetocaloric effect, layered regenerator,
interface temperature measurement
DOI: 10.18462/iir.thermag.2016.0106
1. INTRODUCTION
The layered concept in the active magnetic regenerator (AMR) has been studied in several numerical and
experimental works [1, 2], however further analysis of regenerator layer operation is required. Regenerators
consisting of two layers are experimentally characterized in PM II [3] with a temperature sensor assembled
between layers [4]. The warm layer of all tested regenerators is Gd, while the cold layer is made of four
Gd1-xYx alloys with 𝑥 varying from 0 to 0.075. In the present work, the experimental temperature, adiabatic
temperature change (∆𝑇𝑎𝑑 ), specific heat and entropy change distributions of the four regenerators are
evaluated assuming a linear layer temperature profile between experimental hot, interface and cold side
temperature measurements. Trends between regenerator performance and property distributions are
investigated.
2. METHODS
The presented experiments are conducted at a constant hot side temperature, displaced volume and frequency
of 298 K, 13.9 cm3 and 0.5 Hz, respectively. A 70/30 (%vol) water glycol solution is used as the heat transfer
fluid and the device operates in an ambient temperature of 293 K. The regenerator housing, made of Ultem
polyetherimide, is divided into hot and cold layers. The length of each layer is 65mm, yielding a total
regenerator length of 130mm. The magnetocaloric matrix consists of packed spheres with a particle diameter
range of 300-600 𝜇m (450 𝜇m average) and a porosity of 0.36. The total mass of magnetocaloric material
ranges from 120 to 130 g per layer, yielding a total mass of 500-520 grams for the two regenerators in PM II.
The total mass varies between regenerator compositions as the Gd1-xYx alloys have a slightly lower density
than Gd. Temperature measurements are performed at the hot, middle and cold ends of both regenerators to
characterize the layer and regenerator temperature spans. The presented temperature spans are the cycle-
averaged measurements of both regenerators. The four regenerator configurations (R1-R4) are shown in Table
1 below.
Table 1 : Tested regenerator compositions
Regenerator Composition and 𝑇𝑐𝑢𝑟𝑖𝑒 (K)

R1 R2 R3 R4
Hot layer material 𝐺𝑑/290 𝐺𝑑/290 𝐺𝑑/290 𝐺𝑑/290
Cold layer material 𝐺𝑑/290 𝐺𝑑0.975 𝑌0.025 /286 𝐺𝑑0.95 𝑌0.05 /283 𝐺𝑑0.925 𝑌0.075 /277
95
The spatial distribution of the adiabatic temperature change, specific heat and entropy change are used to
analyze performance trends, assuming a linear temperature profile between the interface and regenerators ends.
The high and low applied magnetic fields are averaged over the hot and cold blows (0.35-1.13 Tesla). The
applied field is corrected for demagnetization and the MFT model is used to evaluate the adiabatic temperature
change and magnetic properties. The Gd1−x Yx adiabatic temperature change and specific heat are treated as
Gd, shifted to the respective Curie temperature [5].
3. RESULTS
The total and layer temperature spans as a function of the non-dimensional position are shown on the left side
of Fig. 1 for the four regenerator compositions at 0, 20 and 40 watts. On the right side of Fig. 1 is the
corresponding adiabatic temperature change distribution along the regenerator dimensionless length.
Figure 1 : Assumed temperature distribution for R1-R4 (left) and corresponding ∆𝑻𝒂𝒅 distribution (right).
Even with identical warm layer compositions, for experiments with the same thermal load, the warm layer
span decreases as the total span increases; a result of the increased cold layer magnetic work and parasitic load.
96
Although the largest spans are obtained with the largest ∆Tad distribution, the specific heat and entropy change
should also be considered. The cycle-average specific heat is shown on the left side of Fig. 2 and the entropy
change is shown on the right side of Fig. 2.
Figure 2 : Average specific heat of R1-R4 (left) and corresponding entropy change distributions (right).
4. DISCUSSION
For each cooling load, the highest temperature spans are observed when the entropy change is largest along
the regenerator. For example at 0W, R3 has the highest entropy change distribution and produces the largest
temperature span. At 20 W, the warmer cold side temperature favors a reduced Curie temperature spacing
where R2 develops the largest temperature span. At 40 W, the single material R1 has the maximum entropy
change and produces the largest temperature span. Still at 40 W, the Curie temperature spacing of R3-R4 is
too large to sustain the 40 W load and the layer temperature span inverts. The refrigerant capacity
97
T
(RC ′ = ∫T hot |∆S(T, H)|dT) represents the distribution of entropy change with a single value [6], and is shown
cold
in Fig. 3 as a function of temperature span. The regenerator composition with the largest refrigerant capacity
(RC’) develops the highest temperature span at each load.
Figure 3 : RC’ versus experimental temperature span.
5. CONCLUSION
To gain understanding of multilayered regenerators, four regenerator compositions comprised of two-layers

with increasing Curie temperature spacing are experimentally characterized in PM II with an interface
temperature sensor. The temperature distribution is obtained by assuming a constant layer temperature gradient
between the hot, interface and cold side temperature measurements. The temperature distribution is used to
evaluate the adiabatic temperature change, specific heat and entropy change distributions in each multilayered
regenerator. The regenerator with the largest refrigerant capacity (RC’) develops the largest temperature span
at each operating condition.
ACKNOWLEDGMENTS
REFERENCES
[1] K. Engelbrecht et al., “Experimental results for a magnetic refrigerator using three different types of
magnetocaloric material regenerators”, I.J. Refrigeration 34, 1132 (2011)
[2] S. Jacobs et al., “The performance of a large scale rotary magnetic refrigerator”, I.J. Refrigeration 37, 84
(2014)
[3] D.S. Arnold et al., “Design improvements of a permanent magnet active magnetic refrigerator”, I.J.
Refrigeration 37, 99 (2014)
[4] R. Teyber et al., “Performance evaluation of two-layer active magnetic regenerators with second-order
magnetocaloric materials”, Applied Thermal Engineering 106, 405 (2016)
[5] B. Monfared, B. Palm, “Optimization of layered regenerator of a magnetic refrigeration device”, I.J.
Refrigeration 57, 103 (2015)
[6] A. Smith et al., “Materials challenges for high performance magnetocaloric refrigeration devices”, Adv.
Energy Mater. 2.11 1288-1318
98
MAGNETOCALORIC EFFECT IN SEVERE PLASTIC DEFORMED
Gd-X (X=In, Ga, B, Y, Zr)
S. Taskaev (a,c,d)*
, V. Khovaylo(a,c,e), K. Skokov(a,b), A. Pellenen(d), D. Karpenkov(a),
M. Ulyanov(a) and O.Gutfleisch(b)
(a)
Chelyabinsk State University (Chelyabinsk, Russia)
(b)
TU Darmstadt (Darmstadt, Germany)
(c)
National University of Science and Technology “MISiS” (Moscow, Russia)
(d)
NRU South Ural State University (Chelyabinsk, Russia)
(e)
ITMO University (Saint Petersburg, Russia)
*
Corresponding author. E-mail: tsv@csu.ru
ABSTRACT
Nowadays, a bottleneck in designing an effective heat exchanger for magnetic refrigeration devices is the
development of magnetocaloric materials exhibiting a good magnetocaloric effect in a wide temperature
range suitable for utilization in active magnetic refrigeration (AMR) cycle. One of the best candidates is Gd
solid solutions in which one can tune Curie temperature Tc by doping solvent element with a subsequent
decreasing magnetocaloric effect. Solid solutions of Gd treated with the help of cold rolling in severe regime
(relative deformation is about 60 times) show unusual magnetic properties initialized by induced magnetic
anisotropy. Although magnetization, heat capacity, magnetocaloric effect are depressed, they can be
completely restored by a proper thermal treatment of the rolled specimens. It was shown that Gd-Y, Gd-In
and Gd-Zr solid solutions have comparable to pure Gd magnetocaloric effect in a wide temperature window
up to 37 K, 35 K, and 16 K, respectively, and therefore they can be useful for designing an effective AMR
heat exchanger.
Keywords: magnetocaloric effect, rare earth alloys, magnetic refrigeration, ferromagnets, gadolinium.
DOI: 10.18462/iir.thermag.2016.0109
1. INTRODUCTION
Beginning from the discovery of giant magnetocaloric effect (MCE) in intermetallic compound Gd5Si2Ge2
[1], there has been an active search for other intermetallics suitable for the application in novel energy
efficient technology of room temperature magnetic refrigeration. Since that time, a large number of
compounds with impressive MCE characteristics have been found [2]. Theoretical limit for MCE is 18 K/T
and nowadays it is an elusive quantity, even for the materials with a first order magnetostructural phase
transition [3]. Despite the impressive progress in the search for new MCE materials, rare earth element Gd it
is still one of the best materials for the use in magnetic cooling devices, judging by the fact that Gd and its
alloys still have been widely used in prototype magnetic refrigerators [4].
Development of an effective heat exchanger for magnetic cooling device is the key factor in the entire
process of designing the device. For the development of the heat exchanger operating as the AMR system it
is necessary to develop and explore the physical properties of a series of materials which have different
magnetic phase transition temperatures. In the case of Gd, its solid solutions can be used for this aim,
however, this possibility is limited because only a few elements form solid solutions with Gd. Because the
exchange interactions between the atoms of rare earth elements occurs via the RKKY indirect exchange, a
dilution of the system of magnetic atoms by dia- or paramagnetic 3d metals or metalloids should lead to a
noticeable change in the exchange interactions, and hence to a decrease in the Curie temperature. However,
if the concentration of impurity atoms were small, the effective magnetic moment per atom of the solid
solution will be close to the magnetic moment of pure Gd and therefore MCE of such solid solution can be
expected to be close to MCE of Gd.
To create an effective heat exchanger is not sufficient to have only a set of MCE materials. Theoretical
estimates [5] showed that the optimal thickness of MCE material (e.g., Gd) is of about 200 µm or less. In this
case, it is possible to construct a magnetic refrigerator operating at frequencies much higher than 10 Hz.
However, production of such thin materials by conventional methods (electroerosion cutting, grinding, etc.)
faces the problem of cracking at the grain boundaries during machining. One way out of this situation is to
modify the microstructure of the material aimed at reduction of crystallite grain size. This can be done, for
99
example, by severe plastic deformation (SPD) methods which allow one to obtain thin metal strips with good
mechanical properties. Besides SPD methods, melt spinning technique can also be used for this aim.
Despite the intensive study of the rare earth metals and alloys over the past 50 years, the issue of the physical
characteristics of rare earth metals subjected to severe plastic deformation has remained virtually untouched.
As a result of such deformation, the microcrystalline structure of the material is changed which is
accompanied by modification of mechanical, magnetic, transport, and thermal properties. Indeed, resent
studies have shown [6] that properties of plastically deformed materials considerably differ from the
properties of the raw alloys, but these changes can be completely reversible [7].
2. EXPERIMENTAL DETAILS AND DISUSSION
The starting materials for producing Gd-R (R = Ga, B, In, Zr, Y) alloys was 99.98% pure elements. The
investigated compounds (see Table 1) were prepared with the help of arc-melting in a protective argon
atmosphere. For a better homogeneity the ingots were remelted at least 3 times. Plastic deformation was
carried out on two four-high rolling mills. The rollers were made from tungsten carbide. The first one had
100 mm work rolls and was used down to a thickness of 0.4 mm. Thinner strips were produced on a second
mill, whose work rolls were 40 mm in diameter. The thickness reduction was between 10% and 20% per run.
The speed of rolling was 0.05 m/s and maximal mechanical stress was 50 kN. The rolling was performed at
room temperature. No heat treatment was applied for improving ductility.
Direct measurements of adiabatic temperature change Tad were carried out in an apparatus described in
details elsewhere [8]. The variable magnetic field was generated by a permanent magnet assembly and
limited to 1.9T. The field was applied in the plane of the rolled strips, to avoid demagnetization. The brief
overview of some physical properties of the samples is shown in Table 1.
Table 1. The studied compounds and their physical properties.

Concen- Width of MCE reduction
Solid Solution Deformation Tc,
tration stripe (in compare to RC,J/kg
Solution type hinitial/hfinish K
range (x) (hfinish), mkm pure Gd), %
Gd100-xYx 0..15 substitution 52-63 ~100 0..25 ~37 149.2
Gd100-xInx 0..3 substitution 49-54 ~100 0..32 ~36 163.6
Gd100-xZrx 0..3 substitution 50-60 ~100 0..17 ~16 115.2
Gd100-xGax 0..10 substitution 42-50 ~130 0..40 ~5 171.4
Gd100-xBx 0..15 iterstitial 49-61 ~140 0..23 ~5 196.3
The presented compositions of solid solutions show the same trend, namely the dilution of the magnetic
subsystem of gadolinium by paramagnetic or diamagnetic impurity atoms results in a decrease of the
effective magnetic moment per atom (which is reflected in the fall of the saturation magnetization) and in a
decrease of the exchange interactions (which is reflected in the reduction of the Curie temperature of the
solid solutions). These trends lead to a reduction in the magnitude of the magnetocaloric effect and to a shift
of the position of its peak which are proportional to the concentration of impurity atoms in the solid
solutions.
The tendencies observed in the as-cast samples of solid solutions are also reproduced in the plastically
deformed samples, but the magnetization of the plastically deformed sample in low and moderate magnetic
fields is lower than in the corresponding as-cast samples of solid solutions [6, 7]. Note that in high fields (of
the order of tens Tesla) the saturation magnetization of the as-cast and plastically deformed samples is the
same. This anomalous behavior is a consequence of magnetic anisotropy arising in the course of the plastic
deformation whose energy density is of about ~10 MJ/m3 for plastically deformed pure Gd [6], ~1.7 MJ/m3
for Gd90Ga10 in the field of 2.85 T, and ~2.44 MJ/m3 for Gd85В10 in the field of 3 T. For the comparison,
magnetocrystalline anisotropy of polycrystalline pure Gd is of about 0.17 MJ/m3.
As has shown theoretically by the Green's functions method [9], in metals containing static impurities the
RKKY interaction falls off with distance as r-3, but the impurities make the phase of the RKKY oscillation
random. This effect increases the tendency to form a spin-glass-type ordering in metals with a random
arrangement of the localized magnetic moments. Note that even in the case of regular arrangement of the
magnetic moments, the presence of non-magnetic impurities leads to the random nature of the RKKY
interaction, and as a consequence, the possibility of implementing a spin-glass ordering. Nevertheless, the
100
spin-glass state can not explain the experimentally observed changes in the magnetic properties of plastically
deformed samples (particularly, appearance of the magnetic anisotropy). Besides, no relaxation of the
magnetization characteristic of spin glasses was observed in the plastically deformed samples.
In our opinion, the observed behavior can be explained qualitatively by the random magnetic anisotropy
induced by plastic deformation which leads to a random distribution of easy axes in the deformed alloy.
However, Gd has not to be sensitive to a change of the crystal field, because it has a half-filled 4f shell with
zero orbital angular momentum. Therefore, we can conclude that, most likely, the observed for the plastically
deformed samples changes in the magnetic properties are not determined by the main part of the material
concentrated in the crystallites, but the atoms of "grain boundary phase" and the atoms located around the
defects. As is shown in [7], annealing completely restores the properties of the deformed material to the
original properties of the alloy.Based on the experimental data one can construct the following overview
diagrams (Figs. 1, 2) modern magnetocaloric materials and the materials studied in this work.
Figure 1. Magnetic entropy change for Gd solid solutions and some perspective MCE alloys.
Figure 2. Adiabatic temperature change in Gd solid solutions and in some perspective MCE alloys [2].
101
As can be seen from the figures, classification of MCE materials with respect to the magnitude of the
magnetic entropy change is only qualitative, because the measurements of the magnetic entropy change are
only possible by indirect methods, and do not provide a comprehensive information about the
magnetocaloric effect. In contrast, measurements of MCE by the direct measurement shows that solid
solutions of Gd have the best characteristics among MCE materials with second order magnetic phase
transition (see Fig. 2). Only La(Fe,Si)13 alloys with first-order metamagnetic phase transition, Heusler alloys
and Gd5Si2Ge2-based alloys with coupled magnetostructural phase transitions can compete with the solid
solutions of Gd.
3. CONCLUSIONS
Summarizing the data, it can be stated the following.

1. Magnetocaloric effect was experimentally investigated in severe plastic deformed solid solutions of Gd-R
(R = Ga, B, In, Zr, Y). It is shown that Gd100-xYx (x = 0…15) system is the best candidates for utilizing in
AMR heat exchangers because it has comparable to Gd MCE in 37 K temperature window. The Gd-based
solid solutions are ductile and technologically friendly for large scale production.
2. It is found that a significant depression of magnetization and MCE occurs in Gd-R (R = Ga, B, In, Zr, Y)
after SPD. Origin of this phenomenon lies presumably in a random anisotropy which is induced after the
plastic deformation.
3. The degradation of magnetic and thermodynamic properties of SPD solid solutions of Gd-R (R = Ga, B,
In, Zr, Y) is fully reversible and these properties can be restored after annealing at temperatures which
depend on the deformation degree.
ACKNOWLEDGMENTS
This work was supported by Russian Science Foundation (Project No. 15-12-10008).
REFERENCES
[1] V.K. Pecharsky, K.A. Gschneidner, Jr, Phys. Rev. Lett. 78, 4494 (1997).
[2] J. Liu, T. Gottschall, K. P. Skokov, J.D. Moore, O. Gutfleisch, Nat. Mater. 11, 620 (2012).
[3] A.M. Tishin, Y.I. Spichkin, V.I. Zverev, P.W. Egolf, Int. J. Refrig. (2016),
http://dx.doi.org/doi:10.1016/j.ijrefrig.2016.04.020.
[4] B.Yu, M. Liu, P. W. Egolf, A. Kitanovski, Int. J. Refrig. 33, 1029 (2010).
[5] M.D. Kuz’min, Appl. Phys. Lett. 90, 251916 (2007).
[6] S.V. Taskaev, M.D. Kuz`min, K.P. Skokov, D.Yu. Karpenkov, A.P. Pellenen, V.D. Buchelnikov, O.
Gutfleisch, J. Magn. Magn. Mater. 331, 33 (2013).
[7] S. Taskaev, K. Skokov, V. Khovaylo, V. Buchelnikov, A. Pellenen, D. Karpenkov, M. Ulyanov, D.
Bataev, A. Usenko, M. Lyange, O. Gutfleisch, J. Appl. Phys. 117, 123914 (2015).
[8] K.P. Skokov, V.V. Khovaylo, K.-H. Muller, J.D. Moore, J.Liu, O.Gutfleisch, J. Appl. Phys. 111, 07A910
(2012).
[9] L.N. Bulaevskii, S.V. Panyukov, JETP Lett. 43, 240 (1986).
102
GENERATING THE OPTIMAL MAGNETIC FIELD FOR MAGNETIC
REFRIGERATION
R. Bjørk*, A. R. Insinga, A. Smith, C. R. H. Bahl
Department of Energy Conversion and Storage, Frederiksborgvej 399, 4000 Roskilde, Denmark
*
Corresponding author. E-mail: rabj@dtu.dk
ABSTRACT
In a magnetic refrigeration device the magnet is the single most expensive component, and therefore it is
crucially important to ensure that an effective magnetic field as possible is generated using the least amount of
permanent magnets. Here we present a method for calculating the optimal remanence distribution for any
desired magnetic field. The method is based on the reciprocity theorem, which through the use of virtual
magnets can be used to calculate the optimal remanence distribution. Furthermore, we present a method for
segmenting a given magnet design that always results in the optimal segmentation, for any number of segments
specified. These two methods are used to determine the optimal magnet design of a 12-piece, two-pole
concentric cylindrical magnet for use in a continuously rotating magnetic refrigeration device.
Keywords: Permanent magnet, Design, Remanence, Optimization, Segmentation, Virtual

DOI: 10.18462/iir.thermag.2016.0110
1. INTRODUCTION
Generating a strong magnetic field is of great importance in magnetic refrigeration. Most magnetic
refrigeration devices use permanent magnets to accomplish this task, as these do not require any energy input
to generate a magnetic field. However, permanent magnets are expensive and thus it is important to utilize
them most efficiently. This means that the permanent magnet structure must generate the high magnetic field
using the least amount of magnet material possible; indeed the magnet is the most expensive component in a
magnetic refrigeration device [1,2].
Previously, design of permanent magnet systems for magnetic refrigeration have relied on adapting existing
well-known geometries, such as the Halbach cylinder or the “C”-shaped magnet, to a given regenerator
geometry [3,4]. An optimization method exists that can be used to optimize a given magnet design [5], but it
requires an existing geometry and remanence distribution. The large spread in efficiency of published magnet
designs [6] indicates the very diverse methods used to design magnets for magnetic refrigeration, but also the
potential for improving performance.
Here we present a method that can determine the optimal distribution of remanence and the border of the
magnet for any desired magnetic field. Furthermore, we present a method for segmenting a given magnet
design that always results in the optimal segmentation, for any number of segments specified.
2. THE RECIPROCITY THEOREM AND THE OPTIMAL MAGNET
The optimization method presented here is not applied to a predefined magnet design but is based solely on
the desired magnetic field. The foundation for determining the optimal distribution of permanent magnet
material is the so-called reciprocity theorem in magnetostatics, which can be expressed as [7]:
B r ,1 ( x)  H 2 ( x) dV   Br , 2 ( x)  H1 ( x) dV (1)
This equation is an energy equivalence between two magnetic systems, 1 and 2. The equation states that the
magnetic energy in system 1 in virtue of its remanence, Br ,1 , when placed in the field generated by system 2,
H 2 , is equal to the energy in system 2 with remanence Br , 2 , when placed in the field generated by system 1,
H 1 . The integration is performed over all space. However, since Br ,1 is only non-zero in places where
permanent magnet material is present in system 1, and likewise for system 2, we can limit the integrals in Eq.
(1) to these regions. In deriving Eq. (1) we have assumed that there are no free currents in the system and that
103
the materials in the system obeys a linear B  H relation, B  H  Br , where the permeability,  , must
be the same for both systems [8].
The theorem in Eq. (1) can be used to determine the optimal remanence, Br ,1 , that produces a given desired
magnetic field, H 1 . We consider a virtual magnet system that has a virtual remanence, Br , 2 , parallel to the
desired magnetic field, H 1 , in every point in the air gap [9,10]. In this case the right hand side of equation Eq.
(1) will be maximized as Br , 2 || H1 . The left hand side will thus also be maximized if the real remanence is
aligned to the field produced by the virtual remanence everywhere. An illustration of this concept is shown in
Fig. 1. Here we consider a system where we want to generate a magnetic field as shown on the figure to the
left, i.e. a field that is radial in two air gaps and zero between these. Firstly, virtual magnets are ‘placed’ in the
air gap, with a remanence identical to the desired magnetic field. These virtual magnets generate a magnetic
field, shown in the right hand side of Fig. 1. The optimal remanence in the design area is then aligned
everywhere to this virtual magnetic field, to maximize the averaged projection of the produced field onto the
desired virtual remanence.
Figure 1. An illustration of the reciprocity theorem. The left-hand figure show the design area as well as the desired
magnetic field in the air gap. The right-hand figure show the corresponding virtual system, with virtual magnets
with a remanence identical to the desired magnetic field of the real system and the field produced by these.
It is also seen from Eq. (1) that once the real remanence is aligned to the virtual field H 2 in a given point x ,
the contribution of that site to the integral in Eq. (1) is proportional to the norm H 2 . This means that if the
magnet is surrounded by air (  r  1 ) the optimal border between magnet and air will be a contour level of
H 2 , as points inside a contour level of H 2 will contribute more to the integral in Eq. (1) than points outside.
If the magnet design area is surrounded by a high permeability material, e.g. iron, the optimal border can also
be calculated. Assuming an infinite permeability for iron, the virtual field H 2 will be normal to the border of
the areas with iron present. Now note that the magnetic field can also be written as the gradient of a magnetic
scalar potential, H 2   2 . Here the magnetic field is also normal to contours of  2 . Thus, there is an
analogy between these two cases. By choosing a contour of  2 as the border of the magnet, we can ensure
that the energy efficiency of the system is maximized [11]. If the relative permeability of the surrounding
matter is neither 1 nor infinite, an iterative approach can be used to determine the border between magnet and
the surroundings.
3. OPTIMAL SEGMENTATION
Once a remanence vector field has been selected, be it optimal or not, the magnetic structure must be segmented
before it can be realized. When segmenting a magnet design, the system is split into uniformly magnetized
segments. Previously segmentation of magnet designs was done by numerically determining the optimization
104
direction of the remanence on a predefined segmented geometry. However, it has recently been shown that it
is possible to determine the globally optimal segmentation of a 2D magnetic system [8].
Because of superposition, the field generated by each point in the magnet is independent from the other points.
One can show that this implies that it is never optimal to split a region over which the direction of the virtual
field H 2 is uniform [8], which for 2D systems can be shown to be equivalent with the fact that the optimal
 
border between two adjacent segments must lie on a contour of   arctan H 2, y / H 2, x . We are thus left with
choosing the optimal contours that gives the desired number of segments. This problem can be shown to be
equivalent with maximizing the perimeter of a piecewise linear approximation of a continuous curve  int  2 
defined as the integral of H 2 between two contours  1 and  2 [8]. It is possible to determine the globally
optimal solution to the curve approximation problem up to the desired precision by dynamic programming
using numerical techniques described elsewhere [8].
4. EXAMPLE
We will consider an example to illustrate the applicability of the techniques presented above for use in
designing magnets for magnetic refrigeration. We consider a rotating magnet design with an outer cylindrical
magnet, an inner iron cylinder and two high field regions and two low field regions in the air gap between the
cylinders. This is a geometry well known from literature [12]. We assume an inner radius of the air gap of
0.125 m, and an outer radius of 0.165 m. The total high field cross-sectional area is thus 182 cm2. The magnet
is assumed to have a maximum outer radius of 0.33 m. We consider a design where the part of the outer
cylinder that is not permanent magnet is iron, which here is assumed to have a permeability of  r  1000 .
The remanence of the permanent magnets is 1.4 T.
We desire a magnetic field in the high field air gaps similar to the field illustrated in Fig. 1. In order to design
the magnet, we follow the procedure described above. First, virtual magnets with a remanence equal to the
desired magnetic field are ‘placed’ in the high field air gaps. The remanence in the design area is then aligned
to the field produced by these virtual magnets. Finally a contour of the magnetic scalar potential of the virtual
field,  2 , is chosen as the border between the magnet and the surrounding iron. The choice of contour of  2
is a choice of the area of the permanent magnet. Following this, segmentation into 12 pieces is chosen,
following the optimal segmentation technique described above.
The difference in average field between the high and the low field regions as function of the fraction of the
outer cylinder that is permanent magnet material is shown in Fig. 2a. Also shown in this figure is the  cool
figure of merit parameter for the magnet design [6]. As can be seen from the figure, at a difference in average
Figure 2: a) The  cool and the difference in averaged field between the high and low field regions as function of the
amount of permanent magnet material and b) the optimal magnet design for the case of Amag / Aouter cylinder  36 % .
105
2/3
field of 1 T, a value of  cool  0.35 T can be obtained, a value significantly higher than previously reported
values [6]. An illustration of the resulting design is shown in Fig. 2 below for the case of a cross-sectional
magnet area of 930 cm2, i.e. Amag / Aouter cylinder  36 % for this design. The average field in the high field
region is 1.25 T, while the average field in the low field region is 0.13 T.
The methods described above always produce the magnet with optimum remanence and optimum
segmentation for the desired field distribution in the air gap. However, as can also be seen from Fig. 2, the
shape of the individual magnet segments are not polyhedral as is usually required for cheap manufacturing.
Thus, a further simplification of the determined magnet design might be needed.
5. CONCLUSIONS
We have presented two methods for designing the optimum magnet. One method can be used to determine the
optimal distribution of remanence that produces a desired magnetic field. The other method can segment a
given magnet design into optimally shaped segments. The methods were used to segment a magnet design
typically used for rotating magnetic refrigeration devices, and the field produced was characterized. The  cool
figure of merit was found to be significantly higher than previously reported values in literature.
ACKNOWLEDGMENTS
This work was financed by the ENOVHEAT project, which is funded by Innovation Fund Denmark (contract
no 12-132673).
REFERENCES
[1] A. Tura, A. Rowe, “Concentric Halbach cylinder magnetic refrigerator cost optimization”, International
[2] R. Bjørk, C. R. H. Bahl, K. K. Nielsen, “The lifetime cost of a magnetic refrigerator”, International Journal
of Refrigeration 63, 48 (2016).
[3] P. V. Trevizoli, J. A. Lozano, G. F. Peixer, J. R. Barbosa Jr., “Design of nested Halbach cylinder arrays for
magnetic refrigeration applications”, Journal of Magnetism and Magnetic Materials 395, 109 (2015).
[4] J. Tušek, S. Zupan, A. Šarlah, I. Prebil, A. Poredoš, “Development of a rotary magnetic refrigerator”,
[5] R. Bjørk, C. R. H. Bahl, A. Smith, N. Pryds, “Improving magnet designs with high and low field regions”,
IEEE Transactions on Magnetics 47, 1687 (2011).
[6] R. Bjørk, C. R. H. Bahl, A. Smith N. Pryds, “Review and comparison of magnet designs for magnetic
refrigeration”, International Journal of Refrigeration 33, 437 (2010)
[7] W.F. Brown, “Magnetostatic principles in ferromagnetism”, North-Holland Publishing Company,
Amsterdam (1962).
[8] A. R. Insinga, R. Bjørk, A. Smith, C. R. H. Bahl, “Globally optimal segmentation of permanent magnet
systems”, Physical Review Applied 5, 064014 (2016).
[9] N.I. Klevets, “Synthesis of magnetic systems producing field with maximal scalar characteristics”, Journal
of Magnetism and Magnetic Materials. 285, 401 (2005).
[10] N.I. Klevets, “Optimal design of magnetic systems”, Journal of Magnetism and Magnetic Materials 306,
281 (2006).
[11] J. H. Jensen, M. G. Abele, “Maximally efficient permanent magnet structures”, Journal of Applied Physics
79, 1157 (1996).
[12] D. Eriksen, K. Engelbrecht, C. R. H. Bahl, R. Bjørk, K. K. Nielsen, A. R. Insinga, N. Pryds, “Design and
experimental tests of a rotary active magnetic regenerator prototype”, International Journal of Refrigeration
58, 14 (2015).
106
THERMAL LOSS ESTIMATION FOR ACTIVE MAGNETIC
REGENERATOR AIMING AT RAILWAY AIR-CONDITIONER
APPLICATION
Y. Miyazaki, K. Ikeda, K. Waki(a)*, T. Kawanami(b)
(a)
Cryogenic Systems, Maglev Systems Technology Division, Railway Technical Research Institute Hikari-
cho, Kokubunji-shi, Tokyo, Japan
(b)
Department of Mechanical Engineering, Kobe University Rokkodai-cho, Nada-ku, Kobe, Japan
*
Corresponding author. E-mail: miyazaki.yoshiki.23@rtri.or.jp
ABSTRACT
We carried out a basic experiment and numerical analysis for estimating the losses quantitatively. A heat
transfer fluid back and forth moving in an active magnetic regenerator (AMR) mixes with the fluid in the
plumbing connected to the AMR at both ends of the AMR. Mixing of those fluids that have different
temperatures causes a decreasing of the cooling capacity. It was found out that displacers placed at the both
ends of the AMR could prevent those fluids from mixing. The result of the numerical analysis that considered
the heat leak from the ambient temperature had a good agreement to the result of the experiment of the AMR
with displacers placed at the both ends of the AMR.
Keywords: Thermal loss, magnetic heat pump, active magnetic regenerator, railway air-conditioner
DOI: 10.18462/iir.thermag.2016.0112
1. INTRODUCTION
In 1976, Brown first applied the magnetic refrigeration in a room temperature range using gadolinium (Gd) as
a magnetic material in a 7 T magnetic field by a superconducting magnet [1]. Operation near room temperature
recently became feasible with availability of large MCE materials and high field permanent magnets.
Since the magnetic heat pump uses magnetic materials as a solid state refrigerant, it has no vapor pressure and
zero global worming potential. Furthermore, the magnetic heat pump technology has a potential for high
efficiency compared to the conventional vapor technology [2].
Railway Technical Research Institute has been studying a magnetic heat pump for trains’ air conditioners. A
typical trains’ air conditioner in Japan has a cooling capacity of 25 kW, an electric consumption of 10 kW and
a coefficient of performance of 2.5. A prototype MHP has been developed, which consists of fixed active
magnetic regenerator (AMR) beds and Halbach arrayed NdFeB magnets with a peak field of 1.5 Tesla rotating
over the beds. Each bed was packed with sphere Gd or Gd-based compounds. The system has a maximum
cooling capacity of 1.4 kW at a zero span of temperature [3]. However, it is possible that the MHP has several
thermal losses so that the cooling capacity of the MHP is decreased. Therefore, we worked on investigation of
those losses and carried out a basic experiment and numerical analysis for estimating the losses quantitatively.
2. THERMODYNAMIC CYCLE of MAGNETIC HEAT-PUMP
Since the heat capacity of magnetic materials is larger at room temperature than at low temperature, the
temperature change of a magnetic material is also smaller. The refrigeration method using AMR (Active
Temperature
Temperature profile
Position
Fluid flow
AMR AMR AMR AMR
Magnet
(1) Adiabatic magnetization (2) Heat exchange (3) Adiabatic demagnetization (4) Heat exchange
Fig. 1 Schematic of the Active Magnetic Regenerator.
107
Magnetic Regenerator) has been developed in order to obtain large temperature spans even if the magnetic
field change is relatively small [4]. The AMR is thought to be the most effective method for the magnetic
refrigerator at room temperature. Magnetic Brayton cycle is the most basic one used in AMR cycle, which
consists of two adiabatic processes and two constant magnetic field processes. Figure 1 illustrates the principle
of the AMR cycle. The temperature profile of the magnetic material in the AMR is shown in the figure.
3. EXPERIMENTAL SET UP
Some experiments were carried out in order to estimate the mixing loss that a heat transfer fluid at both ends
of the AMR mixes with the fluid in the dead volume. The experimental set up is illustrated in Fig. 2. A fixed
AMR is magnetized and demagnetized by a permanent magnet reciprocating. The heat transfer fluid
reciprocates by a pump synchronized with the magnet motion, exchanging the heat with the AMR magnetic
material. Gadolinium (Gd) was used as the magnetic material (sphere particle size: 0.6 ~ 0.85 mm). Surface
size of the magnetic pole is 64 cm2 and magnetic flux density at the surface of the magnetic pole is 1 T. A
container of the AMR is made from acrylic resin that has the inner diameter of 14 mm and outer diameter of
18 mm. A total length of acrylic pipe is 180 mm including the length of 60 mm for packing the magnetic
material in the AMR. The both sides of the AMR are dead volumes. The total amount of Gd was 46 g and
packing factor was 63%. The temperature of the AMR was measured by thermocouples inserted in the both
ends of the AMR. To minimize the dead volume, displacers were placed in the dead volumes. The displacer
can separate the heat transfer fluid in the AMR from the other fluid in the dead volume, moving reciprocatingly
by the fluid pumped to the AMR. Thus, it is expected to prevent the heat transfer fluid in the AMR from mixing
with the other fluid in the dead volume.
4. NUMERICAL ANALYSIS
Figure 3 shows a model for a numerical analysis. Equations (1) and (2) shows the heat balance of the magnetic
material and the heat transfer fluid [5]. The left-hand side of the equation (1) indicates the time derivative of
the heat capacity, the first term in the right-hand side is the thermal conduction, and the second term is the
convection. In the left-hand side of the equation (2), first term shows the time derivative of the heat capacity
and the advection of the heat transfer fluid, the first term in the right-hand side is the thermal conduction, the
second term is the convection. The magnetocaloric effect was given as an adiabatic temperature change of Gd
at 1 T before and after the heat transfer. In the both equations, the last term of the right-hand side indicates the
heat leak to the surrounding at the AMR and the dead volumes. The heat balance was calculated based on
boundary condition that both ends of the dead volume were at room temperature. 𝐾𝑠 and 𝐾𝑓 were estimated
from the result of the temperature measurement at the AMR and dead volumes.
Magnetic material Permanent magnet
(moving)
Displacer AMR Displacer AMR

(Active magnetic
regenerator) (fixed)
Flow line
Permanent magnet Data logger
Pump
Pump
14 mm 60 mm
Thermocouples
Fig. 2 Schematic of experimental set up for AMR. Fig. 3 Photograph of experimental set up.
Dead volume AMR Dead volume
Fig. 4 Model for numerical analysis of AMR.
108
𝜕 𝑇𝑠 𝜕 𝜕 𝑇𝑠
𝜌𝑠 𝑐𝑠 𝑣𝑠 𝜕𝑡
= (1 − 𝜖) 𝜕 𝑥 (𝜆𝑠 𝐴𝑠 𝜕𝑥
) Δ𝑥 + ℎ 𝐴(𝑇𝑓 − 𝑇𝑠 ) + 𝐾𝑠 𝐴𝑙𝑜𝑠𝑠 (𝑇𝑟 − 𝑇𝑠 ) (1)
𝜕 𝑇𝑓 𝑢𝑓 𝜕 𝑇𝑓 𝜕 𝜕 𝑇𝑓
𝜌𝑓 𝑐𝑓 𝑣𝑓 ( 𝜕𝑡
+ 𝜖 𝜕𝑥
) = 𝜖 𝜕 𝑥 (𝜆𝑓 𝐴𝑓 𝜕𝑥
) Δ𝑥 + ℎ 𝐴(𝑇𝑠 − 𝑇𝑓 ) + 𝐾𝑓 𝐴𝑙𝑜𝑠𝑠 (𝑇𝑟 − 𝑇𝑓 ) (2)
𝐾 overall heat transfer coefficient related with heat

Nomenclature
loss
𝑇 temperature 𝑢 velocity
𝑡 time ℎ heat transfer coefficient
𝑥 position 𝐴′ surface area for heat transfar
𝜖 porosity 𝐴 cross-sectional area of magnetic material and fluid
𝜌 density Subscript
𝑣 volume s solid (magnetic material)
𝑐 specific heat 𝑓 fluid
𝜆 thermal conductivity 𝑙𝑜𝑠𝑠 loss parameter
5. RESULT AND DISCUSSION
5.1 Temperature spans of experimental result

Figure 5 shows the results of the temperature span of the ends of the AMR as a function of the flow rate when
the fluid flow time was 0.1, 0.2 second and the moving speed of the permanent magnet was constant. In fig.
5, the results of the temperature span using the displacers was compared with the results of not using the
displacers. When using the displacers, the temperature span of both ends of the AMR became larger than not
using the displacers. The results in Fig. 5 indicate that the displacers placed at both ends of the AMR could
prevent those fluids from mixing.
20 20
18 HEX
Fluid time 0.1 0.1
flow time [sec]
sec 18 HEX flow
Fluid timetime
0.20.2
[sec]
sec
16 16
span [℃]
[C]
[C]
span[℃]
14 14
Temperature span
Temperature span
12 12
10 10
Temperature
Temperature
8 8
6 6
4 Without DP
With Displacer
4 Without DP
With Displacer
Without With
Displacer
DP WithoutWith DP
2 2 Displacer
0 0
0 0.5 1 1.5 0 0.5 1 1.5
Flow rate [liter/min] Flow rate [liter/min]
Fig. 5 temperature spans of experimental result.

20 20 20
18 FluidHEX
flowtime
time0.1
0.1[sec]
sec 18 Fluid HEX
flow time 0.2 sec
time 0.2 [sec]
18 FluidHEX
flow time
time 0.3 [sec]
0.3 sec
[C]
[C]
span [℃]
[C]
span [℃]
16 16 16
span[℃]
Calculation
Calculation Calculation
Temperature span
Temperature span
14 14 14
Temperature span
12 12 12
Temperature
Temperature
Temperature
10 10 10
8 8 8
6 6 6
4
Experiment 4 Experiment 4 Experiment
2 2 2
0 0 0
0 0.5 1 1.5 0 0.5 1 1.5 0 0.5 1 1.5
Flow rate [liter/min] Flow rate [liter/min] Flow rate [liter/min]
Fig. 6 Temperature spans of numerical analysis.
109
5.2 Temperature spans calculated by numerical analysis
Figure 6 shows the results of the numerical analysis by the dashed line, together with the experimental results
using the displacers. Here it can be seen that the shorter the fluid flow time was, the larger the temperature
span became in the high flow rate region. The results of the numerical analysis had the trends same as the
results of experiment. The coefficients 𝐾 in the model were adjusted to the experiments in order to match the
heat losses to the ambient which are present in the AMR volume, while the mixing losses were reduced in the
experiment. Therefore, the results of the calculation in consideration of the heat leaks were in close agreement
with the experimental results.
5.3 Estimation or analysis of losses
From the comparison of the experiments and calculations, the heat losses ware estimated. There was not a heat
load in the experiments, the term of the heat load was placed at the low temperature side of the AMR in the
numerical analysis. Since the numerical analysis which considered the heat leak to teh ambient was in good
agreement with the experimental result using the displacers, the temperature difference between the calculation
and experimental result without the displacers was regarded as the mixing loss. The mixing loss was estimated
from the temperature rise by the heat load in the calculation.
Figure 7 shows the example of the estimation of the heat loss. If there is no heat leak, the calculated cooling
capacity is 24 W at the temperature span of zero. As can be seen from Fig. 7, the estimated heat leak loss was
8%, the mixing loss was 29%, and the real cooling capacity was decreased to 63%.
From the results, we found that it is important to decrease the mixing loss, which accounts for almost 30 percent
of all the cooling capacity.
Ideal cooling
capacity 24 W Heat leak
8%
2W
15 W Mixing loss
Cooling 29%
capacity
63% 7W
Fig. 7 Estimation of losses.
6. CONCLUSIONS
Some experiments were carried out in order to estimate the mixing loss that a heat transfer fluid at both ends
of the AMR mixes with the fluid in the dead volume. It was found that the displacers at both ends of the AMR
were able to prevent the heat transfer fluid from mixing with the fluid in the dead volume.
The numerical analysis in consideration of the heat loss was also carried out. The results of the calculation
were in close agreement with the experimental results with the displacer. Thus, it is found that the numerical
analysis in consideration of the heat loss can explain the characteristics of the experimental result of the
magnetic heat pump.
REFERENCES
[1] G. V. Brown, Magnetic heat pumping near room temperature, J Appl Phys 47 8 (1976) 3673–80.
[2] S.L. Russek, C.B. Zimm, International Journal of Refrigeration 29 (2006) 1366-1373.
[3] Y. Miyazaki, K. Waki, K. Mizuno, K. Ikeda, Quarterly Report of RTRI 56 2 (2012) 130-136.
[4] J. A. Barclay, W. A. Steyert, Active Magnetic Regenerator, US patent, 4,332,135, 1981.
[5] A. Rowe, Thermodynamics of active magnetic regenerators: Part I, Cryogenics 52 (2012) 111-118.
110
ANALYTICAL SOLUTIONS OF THE MAGNETIC FIELD GENERATED BY
TWO-POLE NESTED HALBACH CYLINDERS
F.P. Fortkamp*, J.A. Lozano, J.R. Barbosa Jr.
POLO — Research Laboratories for Emerging Technologies in Cooling and Thermophysics,
Department of Mechanical Engineering, Federal University of Santa Catarina,
Florianópolis, SC, 88040-900, Brazil
*Corresponding author. E-mail: fabio@polo.ufsc.br
ABSTRACT
One of the most common permanent magnet configurations employed in room-temperature magnetic
refrigeration applications is the Halbach cylinder, a structure whose magnetic remanence is defined as a
sinusoidal curve. We present an analytical solution of the magnetic field generated in the air gap between
two concentric Halbach cylinders (where the active magnetic regenerators are placed). The model has been
adapted from previous works and the solution has been extended to the case of two concentric cylinders with
two magnetic poles so as to simplify the manufacture of this permanent magnet configuration. The two
Halbach cylinders are assumed to be infinite in length and built from Nd-Fe-B permanent magnets with a
magnetic remanence of 1.47 T. In the present analysis, the radial dimensions of the magnets are varied to
quantify the influence of the air gap size on the magnet performance and identify optimal configurations.
Keywords: Magnetic refrigeration, permanent magnet, Halbach cylinder

DOI: 10.18462/iir.thermag.2016.0125
1. INTRODUCTION
Since the work of Brown [1], who developed the first device to produce room-temperature refrigeration
using the magnetocaloric effect (MCE), many other magnetic refrigerator design concepts have been
reported [2,3,4,5,6]. However, to this date, no magnetic cooling device has been capable of producing values
of cooling capacity (𝑄̇C ), system temperature span (Δ𝑇sys ) and coefficient of performance (COP), at a single
operating point, which are compatible with those encountered in vapor compression systems.
The magnetic circuit is a crucial component of the magnetic refrigerator for it is the most expensive single-
component of any device [7]. Most state-of-the-art devices reported to date make use of magnetic circuits
based on the permanent magnet Halbach cylinder [8], a structure in which the direction of the remanence
varies continuously along its circumference. A systematic study and optimization of this configuration for
applications in magnetic refrigeration was performed by Bjørk [7], who also reviewed magnet designs from
other sources [9] and introduced a parameter (Λ cool) to compare the performance of the different magnet
configurations. The optimized four-pole nested Halbach magnetic circuit developed in [7] was used in the
magnetic refrigerator reported in [6]. The latter device consisted of a stationary magnet with a rotary AMR
ring that was capable of lifting a thermal load of 100 W with a regenerator temperature span of 21 K. A
detailed thermodynamic performance analysis of the device was carried out in [10]. The main conclusion
from this study was that friction in the flow distribution system and thermal leakage to the ambient were the
major sources of irreversibility. A more recent design by the same group [4] achieved the best
thermodynamic efficiency reported so far by employing a novel flow distribution system, reducing the heat
transfer losses in the magnetic gap, and by using two-pole nested Halbach cylinders and an odd number of
beds in the AMR, to prevent conditions of magnetic equilibrium with the two-pole configuration.
In this paper, analytical solutions of two-pole nested Halbach cylinders are presented. The solutions are an
extension of the single cylinder solutions obtained in [11] with particular emphasis on the influence of the air
gap geometry on the performance of the magnet configuration. The analytical solutions developed in this
paper are original and are useful to identify strategies to improve the performance of new magnetic circuits
for AMRs.
111
The geometry of the present analytical model (Fig. 1) is described by a cylindrical coordinate system, with a
generic position (𝑟, 𝜙, 𝑧), but with infinite length along the 𝑧-axis (making the problem two-dimensional).
Figure 1. Geometry of the nested Halbach cylinders. Regions: (I) shaft, (II) Halbach magnet, (III) magnetic gap, (IV)
Halbach magnet, (V) external environment and (VI) hypothetical layer with infinite magnetic permeability.
The problem domain is divided into layers (regions) and the constitutive relation for each region 𝑘 (each
with constant relative permeability 𝜇r,𝑘 ) is given by:
𝑩𝑘 = 𝜇0 𝜇r,𝑘 𝑯𝑘 + 𝑩rem,𝑘 (1)
where 𝑩𝑘 and 𝑯𝑘 are respectively the magnetic flux density and magnetic field at region 𝑘 and the
permeability of free space is 𝜇0 = 4𝜋 × 10−7 N/A2. For 2D Halbach cylinder magnets, their remanence
𝑩rem,𝑘 can be expressed as [11]:
𝑩rem,𝑘 = 𝐵rem,𝑘 (cos(𝑝𝑘 𝜙) 𝒆̂𝑟 + sin(𝑝𝑘 𝜙) 𝒆̂𝜙 ) (2)
The design proposed in this work assumes 𝑝II = −1 and 𝑝IV = 1, which can be considered the simplest
possible configuration for two Halbach cylinders, resulting in two magnetic poles within the gap. The
external layer (region VI) is helpful to give mathematical closure, and solutions will assume 𝑅e → ∞.
The governing equations for the present problem are the Maxwell equations for the magnetostatic case due to
the low operating frequencies and the absence of conduction currents:
∇ × 𝑯𝑘 = 0 (3)
∇ ⋅ 𝑩𝑘 = 0 (4)
The solution can be carried out in terms of the magnetic vector potential (𝑨):
𝑩 𝑘 = ∇ × 𝑨𝑘 (5)
The complete solutions of Eqs. (3) and (4) in terms of 𝑨𝑘 were originally presented in [11]. It is possible to
adapt the solution methodology to the present geometry (two Halbach cylinders) by including new boundary
conditions, solving the expanded system of equations and finding new coefficients. Due to space restrictions,
the complete expressions are not shown here.
The energy product for the permanent magnet regions can be expressed by [7]:
̂ rem,𝑘 ‖‖𝑩 ⋅ 𝑩
Ψk = ‖𝑯𝑘 ⋅ 𝑩 ̂ rem,𝑘 ‖ (6)
112
where 𝑩
̂ rem,𝑘 is the unit vector in the direction of the remanence of region 𝑘.
3. RESULTS
The geometry parameters and the properties of each region assumed in the analytical calculations carried out
in this work are summarized in Table 1. The main parameter of interest is the air gap size ℎgap = 𝑅g − 𝑅o .
Table 1. Geometry parameters and region properties assumed in the calculations (unless stated otherwise).
𝑅i 25 mm 𝜇r,I 1.00
𝑅g 110 mm 𝜇r,II 1.05
𝑅s 150 mm 𝜇r,III 1.00
𝐵rem,II 1.47 T 𝜇r,IV 1.05
𝐵rem,IV 1.47 T 𝜇r,V 1.00
The maps of the magnetic flux density in the magnetic gap and of the energy product for both magnet
regions for a particular value of ℎgap are shown in Figs. 2(a) and 2(b), respectively.
(a) (b)
Figure 2. (a) Magnetic flux density in the air gap and (b) energy product in the magnet regions for ℎgap = 20 mm.
The maximum energy product for a linear magnet with a magnetic remanence of 1.47 T and a relative
permeability of 1.05 is 409.4 kJ/m3. It can be seen in Fig. 2(b) that the inner magnet is mostly working at low
energy products, thus delivering less magnetic flux to the air gap than it is capable. This can be explained by
the inner magnet opposing to the field generated by to the outer one, resulting in regions of the inner cylinder
where the magnetization is nullified. This issue deserves further investigation but indicates that the inner
magnet is a critical component to be optimized; therefore, a performance improvement is expected by
keeping the dimensions of the outer magnet constant and decreasing the value of 𝑅o (increasing ℎgap for 𝑅g
fixed), as shown in Fig. 3. The Λ cool parameter is defined in Eq. (7), where the 𝜏h∗ parameter represents the
fraction of the AMR cycle during which there is at least one bed in the high field region and is assumed
equal to unity. As the air gap size increases, the difference between the average flux density (to the power of
2/3) in the high and low field regions (the first term in parentheses in Eq. (7)) decreases, but the ratio of high
field volume ( 𝑉h ) to magnet volume (𝑉magnet ) increases, a conflict that results in an optimum point. It can
be concluded that using an inner magnet is always beneficial, but care should be taken to avoid having a
large mass of permanent magnet with a low energy product.
2 2
𝑉h
Λ cool = (⟨𝐵h3 ⟩ − ⟨𝐵l3 ⟩) 𝜏h∗ (7)
𝑉magnet
113
Figure 3. Dependence of the performance parameter on the air gap height and external radius of the outer magnet.
4. CONCLUSIONS
Analytical solutions for a magnet configuration of two-pole nested Halbach cylinders have been presented in
this work. Initial calculations have shown that there is an optimal value of the air gap size that improves the
permanent magnet usage in the configuration. Even though the values of the magnetic circuit performance
parameter are still low, it has been demonstrated that there is room for optimization, especially of the inner
magnet. The results can be considered preliminary, since the thermo-hydraulic phenomena in AMR beds are
not yet taken into account. Future steps of this research will allow the development of an optimization
procedure that maximizes the performance of the magnet-regenerator integrated assembly.
ACKNOWLEDGMENTS
Financial support from CNPq, Embraco and the EMBRAPII Unit Polo/UFSC is duly acknowledged.
REFERENCES
[1] G.V. Brown, “Magnetic heat pumping near room temperature”, Journal of Applied Physics 47, 3673
(1976).
[2] S. Jacobs et al., “The performance of a large-scale rotary magnetic refrigerator”, International Journal of
[3] A. Tura, A. Rowe, “Permanent magnet magnetic refrigerator design and experimental characterization”,
[4] D. Eriksen et al., “Design and experimental tests of a rotary active magnetic regenerator prototype”,
[5] J.A. Lozano et al., “Development of a novel rotary magnetic refrigerator”, International Journal of
Refrigeration, doi: 10.1016/j.ijrefrig.2016.04.005 (2016).
[6] K. Engelbrecht et al., “Experimental results for a novel rotary active magnetic regenerator”, International
[7] R. Bjørk, “Designing a magnet for magnetic refrigeration”, PhD Thesis, Technical University of
Denmark (2010).
[8] K. Halbach, “Design of permanent multipole magnets with oriented rare earth cobalt material”, Nuclear
instruments and methods 169, 1 (1980).
[9] R. Bjørk et al., “Review and comparison of magnet designs for magnetic refrigeration”, International
[10] J.A. Lozano et al., “Performance analysis of a rotary active magnetic refrigerator”, Applied Energy 111,
669 (2013).
[11] R. Bjørk et al., “Analysis of the magnetic field, force and torque for two-dimensional Halbach
cylinders”, Journal of Magnetism and Magnetic Materials 322, 133 (2010).
114
OPTIMIZATION OF ACTIVE MAGNETIC REGENERATORS WITH TWO
AND THREE LAYERS OF Gd AND Gd-ALLOYS
J.E. Cararo(a), J.A. Lozano(a), P.V. Trevizoli(b), R. Teyber(b), A. Rowe(b), J.R. Barbosa Jr.(a),*
(a)
Florianópolis, SC, 88040-900, Brazil.
(b)
IESVic — Institute for Integrated Energy Systems, Department of Mechanical Engineering,
University of Victoria, 3800 Finnerty Rd, Victoria, B.C. V8W 3P6, Canada.
*Corresponding author. E-mail: jrb@polo.ufsc.br
ABSTRACT
Since the magnetocaloric effect (MCE) of typical materials is large only over a narrow temperature range,
layered regenerators can improve the temperature span and thermal performance by using materials with
different Curie temperatures to enhance the MCE along the active magnetic regenerator (AMR). In this
work, an AMR numerical model was used to simulate layered AMRs. Experimental magnetocaloric
properties of Gadolinium (Gd) were employed to simulate different Gd1-xYx alloys. Layered AMRs with two
and three layers of Gd spheres were studied by varying the layer length. Single-layer AMRs were simulated
for comparison purposes. Two different temperature spans of 15 and 20 K were considered. The layered
AMRs were optimized for the largest cooling capacity. The cooling capacity was observed to increase with
the number of layers, especially for the 20-K temperature spans, where a 47.3% increase in the cooling
capacity was achieved by a three-layer regenerator in comparison with a single-layer regenerator.
Keywords: Magnetic refrigeration; magnetocaloric effect; active magnetic regenerator; multilayer.

DOI: 10.18462/iir.thermag.2016.0126
1. INTRODUCTION
After the cyclic steady state operation is achieved in an active magnetic regenerator (AMR), a temperature
profile is established along the magnetocaloric material (MCE) between the hot and cold sources. Since the
MCE of a magnetocaloric material is larger around its Curie temperature (TC), regenerators with layers of
materials with different TC may improve the AMR temperature span due to an effective use of the MCE
along the porous bed.
The performance of multilayer AMRs has been studied numerically and experimentally. Nielsen et al. [1]
modified a two-dimensional numerical model to investigate two-layer parallel-plate AMRs. The analysis was
carried out by varying TC and the fraction of the regenerator occupied by each layer. Gadolinium (Gd) was
assumed as the magnetocaloric material (MCM), and the MCE was modeled through the mean field model
(MFM). By grading different materials in the regenerator, higher cooling capacities and larger temperature
spans were attained when compared with a single-layer AMR. You et al. [2] employed a one-dimensional
AMR model to study the performance of a two-layer regenerator composed of Gd and Gd0.73Tb0.27 for
different system temperature spans ( ). They concluded that the two-layer AMR can effectively improve
the cooling capacity ( ̇ ) and the coefficient of performance (COP) at larger . A numerical evaluation
of the performance of multilayer parallel-plate and microchannel regenerators was carried out by Kamran et
al. [3] employing compounds of LaFe13-x-yCoxSiy and hypothetical compounds of Gd. They reported that
when the number of layers is increased, the cooling capacity is also increased for a certain , and for a
given length of the regenerator there is an optimum number of layers that yield the maximum performance.
The performance of a two-layer regenerator composed of Gd and three different Gd1-xYx alloys was
evaluated numerically and experimentally by Teyber et al. [4]. The results were compared with those of a
single-layer Gd regenerator. The experiments were performed in the AMR test apparatus (PM II) developed
by Arnold et al. [5]. The authors reported higher maximum temperature spans for the two-layer regenerator.
Jacobs et al. [6] designed and constructed a large-scale magnetic refrigeration system using six layers of
115
LaFeSiH of different Curie temperatures. This system attained an impressive cooling capacity of about 3 kW
at zero temperature span. Eriksen et al. [7] built a rotary AMR using 11 regenerator beds composed of
spheres of Gd and three Gd1-xYx compounds. They achieved the best values reported in the open literature for
the coefficient of performance (COP) for ranges of cooling capacities compatible with a domestic
refrigerator.
In the present work, a one-dimensional AMR numerical model developed by Trevizoli [8] has been modified
to evaluate multilayer packed-bed regenerators. Experimental magnetocaloric properties of Gd have been
employed to simulate two other Gd1-xYx alloys. AMRs with two and three layers with different layer lengths
have been optimized in terms of the maximum cooling capacity for a given .
2. METHODOLOGY
The AMR numerical model developed by Trevizoli [8] was adapted to allow the evaluation of multilayer
regenerators with up to five layers of MCM with different lengths. Since the model is based on the finite
volume method (FVM), the regenerator length is divided into a given number of volumes. When the volume
contains only one kind of MCM, the MCE of this volume is the MCE of that material. However, when the
volume contains a boundary between two layers of MCM, the MCE of the volume is given by the mass
weighted average of the MCE of each MCM inside the volume.
The MCM properties are based on experimentally determined properties of Gd (with 290 K). The Gd1-
Y
x x alloys [4] are assumed to have thermophysical properties identical to Gd, except for TC, which is shifted
to lower temperatures according to the stoichiometry of the alloy. In this work, in addition to Gd, the
following alloys were considered: Gd0.95Y0.05 with 283 K and Gd0.925Y0.075 with 277 K. The search
for the best AMR configuration was carried out by varying the length of each MCM layer in the regenerator
and evaluating its performance in terms of ̇ for a fixed .
The numerical simulations were performed based on the operating conditions of the Permanent Magnet
Magnetic Refrigerator (PMMR II) experimental apparatus developed in [5]. The operating AMR frequency
was 0.5 Hz and the average mass flow rate was 52 kg/h (sinusoidal flow profile). This corresponds to a
constant utilization factor of about 0.60. The hot reservoir temperature was maintained at 298 K while two
different cold reservoir temperatures were simulated, 283 and 278 K, resulting in two of 15 and 20 K,
respectively. The packed-bed regenerator was assumed to be composed of spheres with an average diameter
of 450 µm and a porosity of 0.36. The bed dimensions were 130 mm (length), 22.4 mm (internal diameter)
and 0.9 mm (casing thickness). The heat transfer fluid was a mixture of water and ethylene glycol (70/30
%vv.). The magnetic field profile was a rectified cosine function, with an intensity increasing from 0.06 T to
a peak field of 1.45 T. Heat transfer through the casing (considering an ambient temperature of 293 K) and
demagnetization losses were considered in the simulations.
3. RESULTS
Initially, simulations of single-layer regenerators considering each MCM were carried out. For a system
temperature span of 15 K, a maximum cooling capacity of 22.5 W was attained for Gd ( 290 K). For a
span of 20 K, a maximum cooling capacity of 13.1 W was obtained using Gd0.95Y0.05 ( 283 K).
For the two-layer AMR simulations, Gd was used as the first layer at the hot end of the regenerator, while
the other layer (at the cold end) was a Gd1-xYx alloy. The cooling capacity as a function of the fraction of
Gd0.925Y0.075 and Gd0.95Y0.05 in the regenerator length (varied in steps of 5% of the total length) are shown in
Figs. 1(a) and 1(b), respectively. The performance was evaluated for of 15 and 20 K. The numerical
results for two-layer regenerators in Fig. 1 have been compared and validated with the experimental results
obtained by Teyber et al. [4] for cases with 50% of Gd for both combinations. Here, it is worth to point out
that the extreme values of 0 and 100% in Fig. 1 represent the performance of the single layer regenerator of
each MCM at the given . The maximum cooling capacity for a temperature span of 15 K (27.3 W) was
attained by a regenerator composed of 40% of Gd0.95Y0.05 and 60% of Gd. This corresponds to a 21.3%
performance increase compared to a single-layer regenerator. On the other hand, the maximum cooling
116
capacity for a temperature span of 20 K (16.7 W) was achieved with a regenerator composed of 40% of
Gd0,925Y0,075 and 60% of Gd. This represented a 27.5% performance increase compared to a single-layer
regenerator. The improvement associated with a material with a lower at larger temperature spans is due
to a better use of its MCE at lower cold end temperatures.
(a) (b)
Figure 1. Cooling capacity as a function of the fraction of (a) Gd0.925Y0.075 and (b) Gd0.95Y0.05 in two-layer AMRs.
Three-layered configurations of AMRs were simulated assuming different fractions of Gd at the hot end of
the regenerator, fractions of Gd0,925Y0,075 at the cold end of the regenerator and Gd0.95Y0.05 in between. The
cooling capacity as a function of the different combination of length fractions in the three-layered
regenerator for system temperature spans of 15 and 20 K are shown in Figs. 2(a) and 2(b), respectively.
Here, the edge lines (for 0 and 100% fraction of each MCM) of the triangle in each figure represent the cases
of two-layer AMRs, while the vertices correspond to the single layer AMRs. It can be seen in Fig. 2 that the
performance of a three-layered regenerator with these MCM is always better than a two-layered regenerator
as the cooling capacity increases from the edges to the inner part.
(a) (b)
Figure 2. Cooling capacity as a function of the fraction of Gd0.925Y0.075, Gd0.95Y0.05 and Gd in three-layer AMRs for
system temperature spans of (a) 15 K and (b) 20 K.
The maximum cooling capacity for a temperature span of 15 K (28.4 W) was obtained with a three-layer
regenerator with 10% Gd0,925Y0,075, 30% Gd0.95Y0.05 and 60% Gd. This corresponds to a performance
improvement of 26.2% and 4.0%, in comparison with a single-layer and a two-layer regenerator,
117
respectively. For a temperature span of 20 K, the maximum cooling capacity (19.3 W) was attained with a
three-layer regenerator with 20% Gd0.925Y0.075, 30% Gd0.95Y0.05 and 50% Gd. This corresponded to a 47.3%
increase in performance compared to a single-layer regenerator, and a 15.5 % increase in performance
compared to a two-layer regenerator.
It is important to note that for both operating conditions, when a two or three-layer regenerator is used, the
maximum cooling capacity is obtained with at least half the length of the regenerator composed of Gd.
Further studies will be carried out to determine the reason why those configurations of materials attained the
largest cooling capacities.
4. CONCLUSIONS
The performance of multilayer AMRs composed of Gd and two Gd1-xYx alloys was evaluated in this work. A
numerical model was used to evaluate the configuration of materials that resulted in the largest cooling
capacity for a given operating point and a fixed temperature span. Results indicated that by increasing the
number of layers, the cooling capacity also increases, especially for larger temperature spans. Working with
a system temperature span of 15 K, the optimized configuration of a three-layer regenerator presented a
26.2% increase in performance compared to a single-layer regenerator. For a temperature span of 20 K, the
performance was approximately 47.3% higher than that of a single-layer regenerator. It also has been
concluded that, for the operating conditions and Curie temperatures used in this work, at least 50% of Gd is
necessary in the multilayer regenerator in order to achieve the largest cooling capacities. The optimized
multilayer AMR will be experimentally tested in the test apparatus developed in [5].
ACKNOWLEDGMENTS
The authors would like to acknowledge CNPq, Embraco and the EMBRAPII Unit Polo/UFSC for financial
support.
REFERENCES
[1] K. Nielsen et al., “Numerical modeling of graded active magnetic regenerators”, International
Conference on Magnetic Refrigeration at Room Temperature (2010).
[2] Y. You et al., “A numerical study on the unsteady heat transfer in active regenerator with multi-layer
refrigerants of rotary magnetic refrigerator near room temperature”, International Journal of Refrigeration
65, 238 (2016).
[3] M. Kamran et al., “Performance optimization of room temperature magnetic refrigerator with
layered/multi-material microchannel regenerators”, International Journal of Refrigeration, In press.
[4] R. Teyber et al., “Performance evaluation of two-layer active magnetic regenerators with different
second-order magnetocaloric materials”. Accepted for publication in Applied Thermal Engineering (2016).
[5] D. Arnold et al., “Design improvements of a permanent magnet active magnetic refrigerator”,
[8] P.V. Trevizoli, “Development of Thermal Regenerators for Magnetic Cooling Applications”, PhD thesis,
Federal University of Santa Catarina (2015). Available in: http://tede.ufsc.br/teses/PEMC1634-T.pdf
118
INFLUENCE OF PARTICLE SIZE ON THE MAGNETOCALORIC
PROPERTIES OF Mn1.30Fe0.65P0.5Si0.5 POWDERS
A. Bartok(a)*, M. Kuepferling(b), C. Curcio(b,c), V. Basso(b), A. Pasko(a), K. Zehani(d),

L. Bessais(d), F. Mazaleyrat(a), M. LoBue(a)
(a)
SATIE, ENS Cachan, CNRS, Université Paris-Saclay,
61 avenue de Président Wilson, 94235 Cachan, France
(b)
Instituto Nazionale de Ricerca Metrologica, Strada delle Cacce 91, 10135Torino, Italy
(c)
DISAT, Politecnico di Torino, Duca degli Abruzzi 24, 10129, Torino, Italy
(d)
CMTR/ICMPE, CNRS, 2-8 rue H. Dunant, 94230 Thiais, France
*
Corresponding author. E-mail: andras.bartok@satie.ens-cachan.fr
ABSTRACT
We present the results of an experimental investigation on the magnetocaloric properties of

Mn1.30Fe0.65P0.5Si0.5 powders with different average particle sizes. Specific heat capacity under magnetic field
cp(H,T) and magnetic field induced isothermal entropy change Δsiso(H,T) were measured by Peltier
calorimetry. Thermal hysteresis is found to decreases with the particle size, but the refrigerant performance,
evaluated by using the product Δsiso ΔTad of the isothermal entropy change Δsiso and the adiabatic temperature
change ΔTad , has only slight dependence on the particle size.
Keywords: magnetocaloric materials, differential scanning calorimetry, powder metallurgy, magnetic cooling
DOI: 10.18462/iir.thermag.2016.0127
1. INTRODUCTION
Mn-Fe-P-Si alloys are among the best candidates as active substance for room temperature magnetic
refrigeration [1]. Detailed research studies have been already devoted to the influence of the composition on
the magneto-elastic transition [2]. However, extrinsic properties, such as the role of particle size, have not
been the object of a systematic assessment yet. Complex refrigerator parts are often realized by starting from
powders which are already in the desired phase. It is therefore of interest to know if the particle size of the
Mn-Fe-P-Si powder has an influence on the magnetocaloric properties. In this study we focus on the
investigation of the compound Mn1.30Fe0.65P0.5Si0.5 that is characterized by a transition temperature around
273K and a low temperature hysteresis.
2. EXPERIMENTAL
Samples of Mn1.30Fe0.65P0.5Si0.5 have been prepared by high-energy ball milling under Ar atmosphere from
high purity Fe2P powder and P, Mn and Si chips (see details in [3]). A two step heat treatment was employed:
the powder was first annealed at 1373 K for 2 hours in 200 hPa Ar atmosphere, and then at 1123 K for 20
hours. From the resulting powder a classical sieve separation method was employed to obtain different
batches with particle sizes: a) <10 μm; b) 10-35 μm; c) 35-75 μm; d) 75-100 μm and e) >100 μm. Powders
were analysed by XRD at room temperature using Cu Kα radiation (Bruker D8 Advance diffractometer) in
order to assure the absence of secondary phases. Results shown that, beside the main hexagonal Fe 2P-type
phase (P62m space group), cubic (Mn,Fe) 3Si phase was detected, but of negligible amount. All the samples
are in the paramagnetic state at room temperature.
Prior to measurement each sample was subjected to a number of preliminary cycles through the transition at
zero magnetic field to eliminate the so-called virgin effect [3]. Experience shows that at least 5 cycles are
necessary to obtain a repeatable signal from differential scanning calorimetry (DSC). Calorimetric
characterization in magnetic field was carried out on a Peltier cell DSC calorimeter [4]. Measurements were
performed in Ar atmosphere. Data was collected from temperature heating and cooling scans under constant
magnetic field and from magnetic field scans under isothermal conditions. From the measured heat flux
signal we derived the specific heat cp(H,T) and the entropy s(H,T). The isothermal entropy change and the
adiabatic temperature change were derived from the s(H,T) diagram [4,5].
119
Figure 1. Specific heat measured on Mn1.30Fe0.65P0.5Si0.5 at different field (0, 0.5, 1, 1.5 T) for fine (left panel) and coarse
powder (right) upon cooling (dashed lines) and heating (solid lines).
Figure 2. Entropy difference measured on Mn1.30Fe0.65P0.5Si0.5 with respect to a low temperature reference state s(H, T) -
s0 for fine (left) and coarse (right) powders upon cooling (dashed lines) and heating (solid lines). s0 is a reference
entropy value.
3. RESULTS
Here we present the results of two particle sizes: <10 μm and 75-100 μm representative of small and large
particles. The specific heat cp (H,T) for the two different sizes is shown in Fig. 1 while the entropy s(H,T) is
shown in Fig.2. In both graphs both the measurement upon heating and cooling are shown. The relative
vertical position of the curves of Fig.2 is set by using the measurement of the isothermal field induced
entropy change. The magnetic field induced entropy change of Fig.3 is the difference -Δs(H,T)=-(s(H,T)-
s(0,T)) computed by subtracting heating with heating and cooling with cooling. The effective entropy change
Δsiso and the adiabatic temperature change ΔTad (Fig.4) which are effectively available for magnetic
refrigeration are computed by the scheme sketched in Fig.5.
By comparing the measurements of the two different particle sizes we notice, in Fig. 1, that the maximum
value of the specific heat is slightly smaller for the small particles than for the large ones, and the same effect
is seen for magnetic field induced entropy change of Fig. 3. This effect is opposite for what concerns the
temperature hysteresis ΔThyst whose values are reported in Table 1. The effect of the magnetic field is to
decrease the values of the specific heat. This effect is stronger for coarser particles. At 1.5 T the specific heat
properties for the two samples are similar.
To investigate the magneto-caloric refrigerant performance of these samples, the product Δsiso ΔTad with other
parameters (Table 1) were derived by using the scheme of Fig. 5. Temperature hysteresis has been derived
from Fig. 2 by taking the temperature differences at half of the transition entropy. We find that the entropy
change and the adiabatic temperature change in different fields are slightly dependent on the particles size.
The most important difference can be observed in the thermal hysteresis. In both cases thermal hysteresis
decreases with increasing magnetic field.
120
Figure 3 : Magnetic field induced entropy change measured on Mn1.30Fe0.65P0.5Si0.5 fine (left) and coarse (rightl) powders
upon cooling (dashed lines) and heating (solid lines).
Figure 4 : Adiabatic temperature change of Mn1.30Fe0.65P0.5Si0.5 . Fine (left) and coarse (right) powders.
Figure 5. Sketch of the entropy versus temperature curves. The isothermal entropy change Δsiso and the adiabatic
temperature change ΔTad obtained in a cooling thermodynamic cycle are indicated. The transition temperatures T h and
Tc are defined as the midpoint of the transition between the paramagnetic and the ferromagnetic phases.
121
However, coarse powder shows stronger dependence. Between 0 and 1.5 T this hysteresis change is 31%
instead of 75% observed in the case of fine powder. On the other hand, the product Δsiso ΔTad shows a larger
dependence on field in the case of coarser particles. At 1 T the products Δsiso ΔTad have almost the same
values, however the thermal hysteresis is decreased by a factor 2 for finer particle size.
A possible clue to interpret the observed differences, comes from XRD data and from SEM imaging
performed on the samples. These data show that the average grain (i.e. crystallite) size lays between 5 and 10
μm for both samples. Therefore the fine particles will be mainly single crystal grains, while the coarser
particles will be mainly polycrystalline. We can argue that mechanical stresses can develop during the phase
transition in polycrystalline particles in which each single crystal grain may have random orientation. The
elastic energy, which is necessarily present in polycrystalline particles in the low temperature phase, is then
playing a non negligible role. These random stresses give rise to two different effects. From one side they
introduce an additional extrinsic hysteresis which is observed in coarse particles with respect to the fine ones.
On the other side they force the single grains of a polycrystalline particle to change phase simultaneously,
then giving a sharper ΔTad and cp peaks observed in the coarse particles. Further efforts will be devoted to
understand this behaviour in a quantitative way.
particle size : d < 10 μm particle size : 75 < d < 100 μm
Table 1 Temperature hysteresis, entropy change, adiabatic temperature change and the product Δsiso ΔTad derived from
the scheme of Fig. 5.
4. CONCLUSIONS
In this paper the magnetocaloric effect of Mn-Fe-P-Si powders has been investigated as a function of the
particle size. Samples of Mn1.30Fe0.65P0.5Si0.5 were prepared by high energy ball milling, annealed, separated in
ranges of particle size and investigated by DSC in magnetic field. Results on fine versus coarse particle size
show that fine particles have a smaller temperature hysteresis by a slightly lower magnetocaloric effect with
respect to coarser particles. The two effects compensate each other at a certain magnetic field between 0.5
and 1T. Then the energy product Δsiso ΔTad is larger for the fine particles for 0.5T while it is larger for coarse
particles for 1T or above. This result may be helpful for the optimization of the material though the
microstructure for the application in magnetic refrigerating machines.
ACKNOWLEDGMENTS
The research leading to these results has received funding from the European Community 7th Framework
Programme under Grant agreement 310748 DRREAM.
REFERENCES
[1] E. Bruck, Developments in magnetocaloric refrigeration, J. Phys. D: Appl. Phys. 38 (23), R381 (2005).
[2] N. H. Dung et al., From first-order magneto-elastic to magneto-structural transition in (Mn,Fe) 1.95P0.50Si0.50
compounds, Appl. Phys. Lett. 99 , 092511 (2011).
[3] A. Bartok et al., Study of the first paramagnetic to ferromagnetic transition in as prepared samples of
Mn–Fe–P–Si magnetocaloric compounds prepared by different synthesis routes, J. Magn. Magn. 400, 333-
338 (2015).
[4] V. Basso et al., A Peltier cells differential calorimeter with kinetic corrections for the measurement of
cp(H,T) and Δs(H,T) of magneto-caloric materials, Rev. Sci. Instrum. 81, 113904 (2010).
[5] V. Basso et al., Specific heat and entropy change at the first order phase transition of La(Fe-Mn-Si)13-H
compounds, J. Appl. Phys. 118, 053907 (2015).
122
INFLUENCE OF THE HEAT EXCHANGERS THERMAL CONDUCTANCE
ON THE THERMODYNAMIC PERFORMANCE OF A MAGNETIC
REFRIGERATOR
R.S. Calomeno(a), J.A. Lozano(a),*, P.V. Trevizoli(b), J.R. Barbosa Jr.(a)
(a)
(b)
IESVic — Institute for Integrated Energy Systems, Department of Mechanical Engineering,
University of Victoria, 3800 Finnerty Rd, Victoria, B.C. V8W 3P6, Canada
*Corresponding author. E-mail: jaime@polo.ufsc.br
ABSTRACT
In the majority of the magnetic refrigeration devices developed so far, the thermal load is provided by an
electrical heater in direct contact with the working fluid in the cold heat exchanger. This is acceptable in
laboratory devices, but does not represent the real thermal interactions taking place in the heat exchanger in
contact with the refrigerated environment (e.g., the air inside the cabinet of the refrigerator). In this work, the
influence of the thermal conductances of the heat exchangers on the thermodynamic performance of a
magnetic refrigerator has been analyzed. A mathematical model based on the 𝜀-NTU method was developed
and implemented in an AMR numerical model. The results reveal a significant deterioration of the system
performance as the thermal conductances are decreased. The system performance is more sensitive to the
overall thermal conductance of the hot-side heat exchanger. This should be considered in the design of actual
AMR operating with real heat exchangers.
Keywords: Heat exchanger; thermal conductance; thermodynamic performance; magnetic refrigerator.

DOI: 10.18462/iir.thermag.2016.0129
1. INTRODUCTION
Most magnetic cooling prototypes developed to this date emulate the refrigerator thermal load via electrical
heaters in direct contact with the working fluid in the cold heat exchanger [1]. However, this does not
represent the actual thermal interaction between the cooling system and the refrigerated environment in real
cold-side heat exchangers. In a real magnetic refrigerator, for instance, the cold and hot heat exchangers
(CHEx and HHEx, respectively) have finite overall thermal conductances (UA) that affect the
thermodynamic performance of the AMR refrigeration system. The thermal design of the cabinet itself and
of the heat exchangers is restricted by space (volume) and cost limitations.
In this paper, we analyze the influence of the cabinet and heat exchanger thermal conductances on the
thermodynamic performance of a magnetic refrigerator. A mathematical model based on the effectiveness (𝜀)
and number of transfer units (NTU) method was developed and implemented in an active magnetic
regenerator (AMR) numerical model developed in [2]. Numerical analyses of the AMR thermal performance
were carried out considering a gadolinium (Gd) packed-sphere regenerator at specified operating conditions.
Different values of the overall thermal conductance were assigned to the cold and hot heat exchangers to
quantify their impact on the system thermodynamic performance.
2. METHODOLOGY
The various temperatures and other relevant variables of a magnetic refrigeration system are schematically
represented in Fig. 1. The system temperature span ( ) is defined as the difference between the hot and
cold environment temperatures, and , respectively. The regenerator temperature span ( ) is defined
as the difference between the temperature of the fluid exiting the hot and the cold ends of the regenerator,
and , respectively. Due to the finite overall thermal conductances of the hot and cold heat exchangers
( and , respectively) there are temperature differences associated with each heat exchanger
123
( and , respectively). Thus, the fluid temperatures entering the regenerator are, respectively,
and .
Figure 1. Schematic diagram of a refrigeration system operated by an AMR and its related variables.
and are usually output parameters in AMR models, and correspond to the temperatures entering the
heat exchangers. In this work, unidirectional flow of the heat transfer fluid is assumed in the heat
exchangers, so the 𝜀-NTU method [3,4] is employed to predict the outlet fluid temperature in each heat
exchanger for different values of overall thermal conductance. The NTU is the ratio of the heat exchanger
overall thermal conductance and the lowest thermal capacity rate between the two streams. Assuming that
the air in hot and cold environments has a much higher thermal capacity, the NTUs of the hot and cold heat
exchangers can be defined as:
(1)
̇
(2)
̇
where ̇ is the mass flow rate and is the specific heat capacity of the fluid.
The effectiveness of a heat exchanger is defined as the ratio of the actual heat transfer rate and the maximum
possible heat transfer rate. In terms of temperatures, the effectiveness of the hot and cold heat exchangers are
given, respectively, by:
𝜀 (3)
𝜀 (4)
For very large thermal capacity rates of the air in the hot and cold environments, their temperature changes
along the heat exchangers are expected to be small. In this situation, the effectivenesses of the hot and cold
heat exchangers are related to the numbers of transfer units by the following expressions:
𝜀 (5)
𝜀 (6)
The above 𝜀-NTU relationships were implemented coupled with the one-dimensional AMR numerical model
developed in [2]. After each blow, the temperatures of the fluid exiting the regenerator are updated and
124
become the inlet temperatures of the heat exchangers. Since and are treated as input data,
the NTU for each heat exchanger are also known so the temperatures of the fluid entering the regenerator are
updated until the convergence is obtained.
In this work, numerical simulations were carried out to evaluate the thermal performance of an AMR
connected to heat exchangers with different thermal conductances and subjected to different operating
conditions. The simulations assumed a Gd packed-sphere bed regenerator with a geometry based on those
analyzed in [2]. The heat transfer fluid was a water/ethylene-glycol mixture (20% vv.). A square-wave
(instantaneous) fluid pumping profile with equal blow periods was assumed. A summary of the regenerator
geometry and the simulation conditions are presented in Table 1.
Table 1. Regenerator geometry and simulation parameters.

Regenerator length 100 mm Porosity of AMR matrix 0.36
Regenerator diameter 22.22 mm Mass flow rate 100 kg/h
Magnetic field Profile Cosine wave form Operating frequency 1 Hz
Magnetic field (Min./Max.) 0 T / 1.5 T Hot reservoir temperature 300 K
Sphere diameter 550 µm
3. RESULTS
The first analysis consisted of verifying the performance of the AMR when the overall thermal conductance
of both heat exchangers is varied. The performance curves (cooling capacity as a function of the system
temperature span) for the AMR with ideal heat exchangers (infinite ) and for real heat exchangers with
25, 15 and W/K are shown in Fig. 1. In most AMRs developed so far, the heat
exchangers have been considered ideal. In small capacity refrigeration systems (e.g., a domestic refrigerator),
overall thermal conductances of 25, 15 and 10 W/K correspond to values that are typically twice those
obtained with static (i.e., natural convection) condensers and evaporators [5,6]. The results have
demonstrated that the performance of the regenerator deteriorates significantly as the thermal conductance of
the heat exchangers is decreased; that is, for the same temperature span the cooling capacity is lower as the
values are smaller. Therefore, it is very important to find optimal designs (in terms of cost and volume)
of the heat exchangers for an actual cabinet. Generally, the heat exchanger cost is directly proportional to the
, as this increases with the amount of solid material (i.e., metal) in the heat exchanger.
Figure 2. Performance curves for an AMR with ideal and real heat exchangers with different thermal conductances.
Usually, in an actual cabinet, the cold heat exchanger has less space available in comparison to the hot heat
exchanger (which is placed outside the cabinet). The thermal performance of an AMR operating with real
heat exchangers was evaluated through numerical simulations assuming fixed thermal conductances of the
hot heat exchanger and different thermal conductances of the cold heat exchanger. Two scenarios were
simulated, was maintained at 10 W/K and at 15 W/K while varying from 5 to 20 W/K. The
resulting performance curves for an AMR operating with these combinations are shown in Figs. 3(a) and
3(b), respectively. Here, represents two values of thermal conductance commonly used in real
125
domestic refrigerators. As can be seen in Figs. 3(a) and 3(b), the performance of the AMR is reduced as
is decreased. As is varied, the AMR can absorb the same cooling capacity but at different
system temperature spans. The difference between curves at the same cooling capacity increases as the
temperature span becomes smaller.
(a) (b)
Figure 3. Performance curves for an AMR with real cold heat exchangers with different thermal conductances and real
hot heat exchangers with (a) 10 W/K and (b) 15 W/K.
The increase of results in an improvement of the AMR performance, but this is very small at the
largest values of temperature span where the desired operating points of the refrigerator are located. A
similar behavior is found when is increased from 10 to 15 W/K. Nevertheless, having a large
does not compensate for a small , as a higher heat rate is transferred to the hot reservoir.
Therefore, a balance between the design of the heat exchangers and the performance of the AMR must be
found so that a competitive (low-priced) refrigerator can be developed.
4. CONCLUSIONS
The influence of the heat exchangers thermal conductance on the thermodynamic performance of a magnetic
refrigerator has been analyzed in this work. A mathematical model based on the 𝜀-NTU method was
developed and implemented in an AMR numerical model. Numerical simulations assuming an AMR with
hot and cold heat exchangers with different overall thermal conductances were carried out to evaluate the
influence on the thermodynamic performance. It has been demonstrated that special care has to be taken
when designing an AMR that will operate an actual cabinet with real heat exchangers. The design of heat
exchangers and the thermodynamic performance evaluation, in terms of the coefficient of performance and
the second law efficiency, for an actual AMR will be subjects of further studies.
ACKNOWLEDGMENTS
REFERENCES
[1] Kitanovski, A., Tusek, J., Tomc, U. et al., “Magnetocaloric Energy Conversion: From theory to
applications”, Springer (2015).
[2] Trevizoli, P.V., “Development of Thermal Regenerators for Magnetic Cooling Aplications”, Ph.D.
Thesis, Federal University of Santa Catarina (2015). Available in: http://tede.ufsc.br/teses/PEMC1634-T.pdf
[3] Kays, W. M. and London, A. L., “Compact Heat Exchangers”, Krieger Publishing Co. (1984)
[4] Incropera, F. P. and DeWitt, D.P., “Fundamentals of Heat and Mass Transfer”, John Wiley & Sons
(2006).
[5] Melo, C. and Hermes, C.J.L. A heat transfer correlation for natural draft wire-and-tube condensers,
International Journal of Refrigeration 32, 546–555 (2009).
[6] Hermes, C.J.L., Melo, C. and Negrão, C.O.R., A numerical simulation model for plate-type, roll-bond
evaporators, International Journal of Refrigeration 31, 335-347 (2008).
126
MODIFIED MEAN-FIELD THEORY WITH PHENOMENOLOGICAL
MODEL OF VOLUME EFFECTS FOR 1ST ORDER TRANSITION
MAGNETOCALORIC MATERIALS
M. Risser(a)*, V. Hardy(b), S. Lionte(a), C. Muller(a)
(a)
Cooltech Applications, 5 imp Antoine Imbs, 67810, Holtzheim, France
(b)
CRISMAT, CNRS UMR 6508, ENSICAEN, 6 Bvd Maréchal Juin, 14050, Caen, France
*
Corresponding author. E-mail: m.risser@cooltech-applications.com
ABSTRACT
The simulation of an Active Magnetic Regenerator (AMR) behavior in a numerical model is strongly
dependent on the Magnetocaloric Material (MCM) data. The Bean and Rodbell theory [1] is often used in
order to model 1st order transition MCM. However, its results can differ from experimental measurements with
a non-zero magnetic field. In this paper we present a phenomenological model introducing the effect of the
volume change of the MCM on its magnetic and calorimetric behavior. We use a shape function to adapt the
exchange interaction. The free parameters of the model are adjusted with respect to experimental
measurements of magnetization and heat capacity. Volume change is deduced to minimize the Gibbs free
energy. These more representative modeled data of strong 1st order MCM can be used for the simulation of the
magneto-thermodynamic cycles at work in the MCM with an improved accuracy.
Keywords: Magnetocaloric, Magnetoelastic, Data Modeling, Experimental Measurements.

DOI: 10.18462/iir.thermag.2016.0131
1. INTRODUCTION
The magnetocaloric (MC) systems constitute an emerging technology showing a good potential in the market
of refrigeration or energy conversion. The MC systems implement the Magnetocaloric Effect (MCE) in
magnetic materials subjected to applied magnetic field changes. For ferromagnetic materials, the MCE is
maximized close to their Curie temperature (Tc). Since the Magnetocaloric Materials (MCM) of first order
exhibit a sharp transition, their peak of MCE is higher than for a second order MCM, but it also has a smaller
width making it unable to directly cover the thermal span required for commercial applications. Thus, one
solution is to design a detailed 1st order MCM Tc cascade in an Active Magnetic Regenerator (AMR) in order
to reach the thermal span required for the applications.
Optimizing the design of these systems is a challenge. It requires the use of numerical multiphysics modeling
taking into account magnetic fields, thermodynamic behavior of the MCM, heat transfers, fluid mechanics,
etc. Three physical properties of the MCM are needed to properly model the magnetothermodynamic cycles
at work in the MCM and to calculate the cooling power and the magnetic work consumption with a good
accuracy: the magnetization, the heat capacity and the adiabatic temperature change or magnetic entropy
change [2]. These data must not only be well representative of the experimental measurements but must also
be consistent with each other from the thermodynamic point of view in order to avoid any artefact of energy
and to obtain a conservative model [3]. Over a thermodynamic cycle these data should lead to equality between
the magnetic work and the anergy generation. For this purpose, the magnetic data used in AMR numerical
models can be calculated using either the Weiss mean-field theory or the Bean-Rodbell theory [1]. Then the
magnetic contribution for the heat capacity can be derived from the magnetization.
In this paper we introduce a thermodynamic model which takes into account the assumed magnetoelastic effect
in the Weiss mean field theory. Bean and Rodbell [1] introduced a model using an explicit dependence of the
exchange interaction on the specific volume (V) through a Curie temperature Tc = To(1+βω) where β is a
dimensionless coefficient of magnetoelastic coupling, ω = (V-V0)/ V0 represents the lattice reduced volume and
T0 is the Curie temperature if the system had the volume V0. The exchange interaction is treated as in the mean-
field theory. Magnetization, magnetic entropy and magnetoelastic energy are thereby solved in order to
minimize the Gibbs free energy leading to a narrower transition than the data obtained with the original Weiss
model according to the value of β. M. Piazzi et al. [4] investigated the competition between the magnetoelastic
and magnetic entropies on the basis of the Bean-Rodbell theory. Two metastable magnetizations can be
obtained for the minimization of the Gibbs free energy. They showed that this theory is a good candidate to
127
model the hysterical transition at the first order. With values of β selected to obtain a first order transition with
a zero magnetic field, the transition changes from the first to the second order with the increase of the applied
magnetic field in adequacy to the experimental behaviors, as shown in [4]. The critical point at which the
transition becomes a second order is investigated by Basso et al. [5] on La(Fe-Mn-Si)13-H compounds.
Nevertheless, this modeled behavior can leads to high differences with respect to the experimental ones for
strong first order MCM as some GdSiGe, MnFePSi or LaFeSi compounds. For these materials, such a critical
point occurs at a higher magnetic field (>5 [T]) [6] than the predictions of Bean-Rodbell theory. As
consequences, MCM modeled data can be responsible for inappropriately modeled magnetothermodynamic
cycles and thereby, generate deviations between AMR simulations and the experimental behaviors for the
strong 1st order transition MCM.
2. MODEL PRESENTATION FOR MATERIAL DATA
The magnetization of the Weiss mean field theory is given by M  N g B J BJ  y   M 0 BJ  y  where g is the
Landé g-factor, µB is the Bohr magneton, J is the total angular momentum, N is the number of magnetons per
unit volume and M0 is the saturation magnetization. BJ(y) is the relative magnetization given by the Brillouin
function where y  0 g B J H  H m  k B T  is the ratio of Zeeman to thermal energies, µ0 is the vacuum
permeability, H is the applied magnetic field, Hm = λM is the Weiss mean field and kB is the Boltzmann
constant. The mean field coefficient (λ) is assumed to be independent of the temperature in the original Weiss
model. We assume that the interaction energy is impacted by the interatomic distances related to the volume.
The volume change at the 1st order magnetic transition is assumed to be similar as the one presented by Fujieda
et al. [7]. To obtain modeled MCM data with improved fitting to the experimental ones in the cases of narrow
first order magnetic transition, we consider the volume dependence of the mean field coefficient (λ) as:
3  k B  Tc   T 
 (1)
0  N  g 2  B 2  J  J  1
through a magnetic interaction shape function α(T), defined as  T   ek 1T Tc   having the slope –k x/Tc at the
x
transition. The parameters k and x are the coefficients of distortion of λ with respect to the temperature. They
are used to adjust respectively the amplitude and the width of the transition in order to obtain a good correlation
between modeled and experimental magnetic data. We assume the effect of the volume change on N is
negligible with respect to its counterpart on α(T), leading us to consider N = N0V0/V ≈ N0 in Eq. 1; here N0 is
the amount of magnetons per unit volume for the reference volume V0, taken at the Curie temperature and in
zero magnetic field. The thermal expansion  v   T H  1 V0 V T H will be derived from the shape
function α(T) in order to minimize the Gibbs free energy in the model. The second assumption deals with the
Curie temperature dependence on the applied magnetic field. Assuming that the total entropy at the transition
is conserved whatever the magnetic field, we define the recursive magnetic field dependent Curie temperature
H
as Tc H   T0   T S0  / H S dH where T S0  / H S is the adiabatic temperature change with respect to
0
the magnetic field (H) at constant entropy (S0) [3], this entropy being equal to the entropy at T0 and H = 0.
Because λ and Tc are now temperature and magnetic field dependent, the ideal shape function α(T) should
allow  H S  0 to match with the conventional exchange energy equation. Because this condition cannot
be verified at this step since the Curie temperature is recursively defined from the MCE, the exchange energy
has to be numerically solved. The exchange energy is derived from the first principle dUex = Q + Wex with
Wex the work resulting from the exchange, i.e. the mean field on the magnetons. We focus on a given quantity
of magnetons in a variable volume V, hence in [J/kg] we have lim Wex S   0d M HmS 2  , the ½ factor
M 0
avoiding the double counting of the exchange energy, and because in an adiabatic process Q = 0 we have
dU ex   0  M dM S 1 2 M 2 d S   and therefore, by integrating over the magnetization we obtain:
M
 0  H H
  M  M 2    
U ex   dU ex   
2

0
M H  
  M   dH    dH  (2)
0
  2 0
0  H  s 2  H  s 
128
If the condition of the isentropic invariance of λ is verified according to the magnetic field, we would have
U ex  0 2   M 2 . Since the isothermal compressibility K  1 V0 V p H ,T is assumed to be independent
of the magnetic field, the magnetoelastic energy is given by U el   2 2K  . The heat capacity at constant
magnetic field, CH  E T H , is deduced from the magnetic enthalpy E  U  p   0  H M with p0
such as no pressure work can be associated to the volume change. The total heat capacity has three components
CH  CHLat  CHmag  CHel which are the lattice contribution, the magnetic contribution and the magnetoelastic
contribution, respectively, the latter being equal to CHel  U el T H   T H  K    v K . The
magnetic heat capacity from the exchange and Zeeman energy with the assumption of an isentropic invariance
of λ is given in Eq. 3 where Hr = Hm + H is the total resulting magnetic field.
        M  M 2    
CHmag   U ex  0 H M     0 H 
 r T     (3)
 T   H    H 2  T  H 
Assuming at this stage that the volume change is acting at a second order on the lattice contribution, the lattice
heat capacity is deduced from the Debye model. The entropy and the magnetic entropy change are derived
from the heat capacity. Since the magnetization is obtained through α(T) and Tc(H), one solution for minimizing
the Gibbs free energy G  E  TS (setting V = V0 for T = Tc in such a way that ω(Tc) = 0) is given in Eq. 4.
This equation leads to the volume change that should be obtained to match the magnetic behavior of the MCM.
2 T
V  V0 3 J N k B Tc BJ K    J N k B Tc K  2    
J  1 Tc  J  T  H 
     3 B   dT (4)
V0 2 J  1 v  T  H
3. COMPARISON OF THE MODEL WITH EXPERIMENTAL MEASURMENTS
We refer to FeSiLaMnH room temperature 1st order magnetocaloric alloys. The experimental heat capacity is
derived from a method designed to catch the whole latent heat component which can be hidden with
conventional calorimetric analysis [8, 9]. The measurements of FCC and FCW magnetizations are performed
in sweep mode at 0.5 [K/min] around the Tc. The sample holders ensure a magnetic field direction parallel to
the long side of the sample in order to minimize the demagnetizing field for both heat capacity and
magnetization measurements. Experimental results are presented in Fig. 1. Magnetization measurements show
the presence of a parasitic ferromagnetic phase above the Tc which is not reproduced by the model. The
modeled data of the alloy for Tc = 33°C are presented in Fig. 2 and Fig. 3. According to the experimental heat
capacity, the magnetic transition is switching to the 2nd order with the magnetic field faster than the modeled
data because the model is designed to reproduce data with critical point at high magnetic field.
Figure 1. Experimental measurements of heat capacity for the Tc = 33 [°C] (left) and magnetization for the three Tc
of 0, 20 and 33[°C] (right) as a function of the temperature in cooling and heating for FeSiLaMnH alloys.
129
0
µ0 H [T]
2
µ0 H [T]
0 2
Figure 2. Heat capacity (left) and reduced magnetization (right) from 0 [T] (blue) to 2 [T] (purple) with a step of 0.5
[T], for the Tc = 33 [°C] alloy, calculated with J=2.
0
µ0 H [T]
2 2
µ0 H [T]
0.5
Figure 3. Total entropy (left) from 0 [T] (blue) to 2 [T] (purple) with a step of 0.5 [T], magnetic entropy change and
Curie temperature dependence on the magnetic field (right), for the Tc = 33 [°C] alloy, calculated with J=2.
4. CONCLUSIONS
In this paper we present a phenomenological model to introduce the volume effect in the mean field theory.
This model can be used as an alternative to the Bean-Rodbell theory for magnetic materials with low switch
from the first order to the second order in the presence of an external magnetic field. The experimental data
taken as reference are not the most suitable to validate the model. Nevertheless the modeled data are consistent
with strong first order MCM.
REFERENCES
[1] C.P. Bean, D.S. Rodbell, “Magnetic Disorder as a First-Order Phase Transformation”, Physical Review
126-1, 104 (1962).
[2] M. Risser et al., “Improvement and application of a numerical model for optimizing the design of magnetic
refrigerators”, Int. J. Ref. 36, 950 (2013).
[3] M. Risser et al., “Construction of consistent magnetocaloric materials data for modelling magnetic
refrigerators”. Int. J. Ref. 35, 459 (2012).
[4] M. Piazzi et al., “Modeling speciﬁc heat and entropy change in La(Fe-Mn-Si)13-H compounds”, J. of Mag.
and Mag. Mater. 400, 349 (2016).
[5] V. Basso et al., “Specific heat and entropy change at the first order phase transition of La(Fe-Mn-Si)13-H
compounds”, J. of App. Physics 118, 053907 (2015).
[6] W. GaoFeng et al., Magnetic and Calorimetric Study of the Magnetocaloric Effect in Intermetallics
Exhibiting First-order Magnetostructural Transitions, Thesis, Zaragoza university (2012)
[7] S. Fujieda et al., “Giant isotropic magnetostriction of itinerant-electron metamagnetic La(Fe0.88Si0.12)13Hy
Compounds”, Applied Physics Letters 79-5, 653 (2001).
[8] V. Hardy et al., “Derivation of the heat capacity anomaly at a first-order transition by using a semi-adiabatic
relaxation technique”, J. Phys. Cond. Matt. 21, 075403 (2009).
[9] F. Guillou et al., “Calorimetric investigation of the magnetocaloric effect in Ni45Co5Mn37.5In12.5”, J. Phys.
D:Appl. Phys. 45, 255001 (2012).
130
THE OPTIMIZATION OF THE ENERGY PERFORMANCES OF A PMRR
BY USING NEURAL NETWORKS

C. Aprea(a), A. Greco(b), A. Maiorino(a)*, C. Masselli(a)
(a)
Department of Industrial Engineering, University of Salerno (Fisciano (SA), Italy)
(b)
Department of Industrial Engineering, University of Naples Federico II (Napoli, Italy)
*
Corresponding author. E-mail: amaiorino@unisa.it

ABSTRACT
In recent years, a large number of experimental and numerical studies have highlighted the potential of the
permanent magnetic rotary refrigerators (PMRR) than those reciprocating. For a PMMR, it is well known it is
possible to obtain the desired performance by contemporary acting on two operational parameters: the mass
flow rate and the cycle frequency. Consequently, with the aim to improve the energy performances of an actual
PMRR, it is necessary to experience an innumerable amount of operating conditions regarding mass flow rate
and cycle frequency. The present work introduces ANNTEO (artificial neural networks technique for
optimization), a technique based on artificial neural networks and able to reduce the number of experiments
necessary to define an optimization map for an actual PMRR. The experimental setup and test procedure are
here reported to demonstrate the technical soundness of ANNTEO.
Keywords: magnetic refrigerator, experimental, neural network, optimization, energy save.

DOI: 10.18462/iir.thermag.2016.0132
1. INTRODUCTION
The interest of the scientific community regarding magnetic refrigeration at room temperature is constantly
growing. In recent years, there has been an increase in the experimental work relating to both new prototypes
and new magnetocaloric materials [1]. Among the different configurations realized, one can recognize that
rotary devices, having rotating magnets and static regenerators, are of particular interest because of their
excellent energy performances. One of the main characteristic, which distinguishes the rotary devices from
those reciprocating, is the capability to ensure higher cycle frequencies so as to increase the available cooling
power. Also, the rotary devices may operate at a variable mass flow rate of the regenerating fluid without
change the cycle frequency, offering further opportunity for improvement in the energy performance in terms
both of COP and of cooling capacity. To date, several works [2,3] have carried out useful suggestions for
designing a PMRR, and some experimental works [4,5,6] have shown the actual behavior of the PMRRs under
different operating conditions. However, no one has shown the way to optimize the working of a PMRR against
the change of the main boundary conditions, such as the thermal load and the heat rejection temperature. For
this reason, this paper introduces a technique (ANNTEO) useful to obtain an optimization tool, which may be
implemented as controller of an actual PMRR.
2. ANNTEO
ANNTEO is a technique based on artificial neural networks and able to reduce the number of experiments
necessary to define an optimization map for a real PMRR. The idea is to identify an empirical model of the
PMRR by employing the artificial neural networks (ANN) and designed to predict the energy performances of
the device. A little number of experiments can be used to train the ANN, so as to carry out an easy and robust
model of an actual device. After a validation process, the ANN can be employed to predict the energy
performance and then it can provide the basis to control the working of the PMRR.
131
2.1 The experimental setup and test procedure
All data employed to build the neural network have been collected by designing a test set to carry out the
energy performances of an actual magnetic refrigerator operating under different conditions. The device used
is 8Mag, a permanent magnetic rotary refrigerator built and designed at University of Salerno [7]. Its magnetic
system is based on a double U configuration of permanent magnets, and it performs a peak flux density of 1.25
T when the air gap is equal to 43 mm. Gadolinium is selected as magnetic refrigerant and demineralized water
is employed as regenerating fluid. The total mass of gadolinium (1.20 kg), shaped as packed bed spheres (of
400÷500 microns), is confined within 8 static regenerator enclosures, which are alternatively magnetized and
demagnetized with the rotation of the magnets. Each regenerator has a height of 20 mm, a length of 45 mm,
and a width of 35 mm. The aspect ratio is 1.5 and the volume available for the MCM is 31.5 cm3. A rotary
valve, mechanically coupled with the magnetic system, imparts the direction of the water through the
regenerators. A rotary valve mechanically coupled with the magnetic field generator imparts the direction of
heat transfer fluid through the regenerators. The cycle frequency (fAMR) is determined by rotating the magnets;
specifically, for each rotation of the magnets (𝑛"#$ ), each regenerator experiences two AMR cycles. A
hydraulic system obtained by the combination of a rotary valve and a vane pump ensures the proper distribution
of the regenerating fluid in each component of the apparatus in accordance with the AMR cycle phases. The
total fluid flow rate entering in the rotary valve is partitioned into two equal shares and then is transported to
the regenerators. At each instant, there are four regenerators hydraulically connected to each other that are
subject to the fluid flow: a couple is magnetized and another one is demagnetized. At the same time, the
remaining four regenerators are disconnected from the hydraulic circuit to experiment on their adiabatic
temperature change. Based on the design specifications, the ratio between the magnetization period and the
fluid flowing period is 1:1 and their duration is imposed by the rotational speed of the magnets. The drive
system consists of a brushless DC motor rotating the magnets at variable speed between 0.1-1 Hz. A digital
encoder and a programmable speed controller complete the drive system. This results in an available maximum
continuous torque of 70 Nm at 54 rpm. A photo of 8Mag is shown in Fig. 1.

Figure 1. Picture of the PMRR prototype named 8Mag
Using PT100 thermo-resistances placed between the input and output of each component, the temperature of
the regenerating fluid is measured. In particular:
§ the temperature span (∆Tspan) is the average time between the temperature of the regenerating fluid
exiting the hot side (THFo) and the temperature of the regenerating fluid exiting the cold side (TLFo);
§ the temperature of heat rejection (TH) is the average time between the temperature of the water exiting
the hot side of the magnetic refrigerator (THFo) and the water exiting the geothermal probe (THFi), both
measured under steady state condition.
By using a magnetic flow meter placed between the pump and the hot heat exchanger, the volumetric flow rate
of the regenerating fluid (V&,( ) is measured. To estimate the pressure drop of the entire device, two piezoelectric
pressure sensors placed between the input and the output of the pump are employed. The total pressure loss is
evaluated as the average time over the difference between the pressure at the outlet (Pout) and the pressure at
the inlet of the pump (Pin). The pressure measurements are also performed at the ends of the regenerators, with
the objective of measuring the blow duration and the pressure drop through the regenerators. Following the
procedure reported in [7], the measurement of the utilization factor (𝜑) is performed. Using the digital encoder
on the motor, the AMR cycle frequency is also measured. The test apparatus is equipped with 32 bit A/D
converter acquisition cards with a sampling rate of up to 10 kHz. To measure the refrigeration duty provided
by 8Mag, a cold heat exchanger has been realized by the combination of electric resistance with a thermally
132
insulated pressure vessel. A variable voltage supply feeds the electrical resistance to provide a thermal load
that is variable from 0 to 500 W. Water, the temperature of which can be adjusted by an electrical heater
managed by a PID controller, is used as a secondary fluid in the hot heat exchanger. In this way, it is possible
to consider the following equation to be valid:
Q (+& = W+.,/ (1)
where W+.,/ is the electrical power that is absorbed by the electric heater inserted into the exchanger and
measured using a watt transducer. Using a PT100, temperature measurements on the outer surface of the
insulation coat of the exchanger are performed to verify the quality of the insulation level. Moreover, using a
two-watt transducer, the electrical power absorbed by the pump (W+.,2 ) and the electrical power absorbed by
the electric motor used for magnet rotation (W+.,3 ) are evaluated. For the evaluation of the COP, the following
equation is used:
89:; <:=,A
𝐶𝑂𝑃 = = (2)
<:=,> ?<:=,@ <:=,> ?<:=,@
In Table 1, the characteristics of the sensors used and the accuracy of the measurements performed are reported.
Quantity Characteristic Accuracy

Temperature RTD PT100 ± 0.1 [K]
Pressure Piezoelectric Pressure gauge ± 75 [kPa]
Volumetric flow rate
Electromagnetic Flowmeter ± 0.5%
Rpm / Frequency Optical Encoder ± 0.01 [° s -‐‑1]
Mass Electronic Balance ± 0.2 [g]
Electrical power Electromagnetic W attmeter ± 0.2%
Torque Torque transducer ± 0.5%
𝜑 Procedure introduced in [7] ±5.3%
COP Eq. 2 & Propagation Error Analysis ±0.35%

Table 1. Characteristics of the sensors used and the accuracy of the measurements performed.
With the aim to define an optimization map for 8Mag, a set of experiments has been designed and carried out
by changing the following parameters: the TH, the W+.,/ , the V&,( , and the 𝑛"#$ . Three TH (16°C, 22°C, and 32
°C) were selected as hot thermal reference, and for each of them a set of experiments were carried out for
different values of the V&,( and the 𝑛"#$ . Because the vane pump is driven by an inverter, a set of three V&,(
values (5.0 l min-1, 6.0 l min-1, and 7.0 l min-1) were employed. Consequently, the 𝑛"#$ were selected so that,
for each V&,( value experimented, the utilization factor were equal to the following values: 2.70, 1.80, 1.30,
0.87, and 0.52. Finally, for each test condition different values of the thermal load were imposed: by starting
under the zero load condition, the thermal load was incremented until ∆Tspan was greater than zero. In this way,
224 experiments were collected in agreement with the operational limits both of the pump and of the electrical
motor.
2.2 The artificial neural networks

An ANN is a mathematical tool used for a wide tasks variety (such as classification, data mining, pattern
recognition, image compression, process modeling). ANN are algorithms implemented in a computing
program or electronic model based on the human brain functioning. The ANN basic components are neurons
which are simple processing elements interconnected and layered. Each connection among neurons is
weighted, and a training algorithm calculates these weights. Hence, each neuron computes a weighted sum
based on the input variables values. A transfer function allows determining the neurons output as a result of
the input weighted sums. The ANN realization requires defining inputs, type of network, topology, training
paradigm and transfer function. The ANN modeling allows for carrying out the required output starting from
corresponding input vectors without considering the assumption of any determinate relationship between the
input and output. There are many types of connection for the data transfer: the most used is the multi-layer
perceptron (MLP). MLP is a feed-forward ANN where data flow from the input layer to the output layer
133
through a different number of hidden layers without any feedback loop. The hidden layers represent the
network computation model core. MLP networks can learn complex relationships between input and output
patterns, and they show a better approximation of the constants allowing a quick link between constant and
non-constant input values. The most selected learning rule for an MLP is the error Back-Propagation (BP)
algorithm. It calculates the gradient of the network error related to its modifiable weights. The BP learning
approach can be implemented considering different topologies and training functions. This typical problem,
when the ANN is developing, is solved using the cross-validation which is a validation technique to estimate
how a model generalizes an independent data set. In particular, the k-fold cross-validation technique is applied
to calculate the mean squared error (MSE), which defines the fitting assessment through the following
equation:
J
FKL EF GHF I
𝑀𝑆𝐸 = . (3)
M
The MSE results can be used to select the best set of ANN parameters. Hence, the cross-validation is applied
to choose the best model between different plausible ANN alternatives. The main choices are related to the
training algorithm, the number of hidden layers and hidden neurons and the transfer functions. For the training
algorithm, the BP paradigm allows to compare the gradient descent with the momentum and weight decay to
Levenberg–Marquardt algorithm [8]. For the topology selection, it is possible to consider one or two hidden
layers, as assured by the universal approximation theorem. Regarding the number of the hidden neurons, an
empirical rule may be considered:
𝑁OM = 2𝑁Q + 1 . (4)
Finally, for the transfer functions, the best solution for the output layer is a linear function, while, because of
the non-linearity problem, the best transfer function for the hidden layer is nonlinear. The most nonlinear
selected functions are the sigmoid or hyperbolic tangent because their derivatives simplify the application of
the BP algorithm.
2.3 The design of the ANN to predict the energy performances

The design process started with the definition of the inputs and the outputs. Following the main objective of
ANNTEO, one has to select as inputs the operating parameters which can influence the energy performances
of a magnetic refrigerator; then, three inputs may be chosen: the V&,( , the 𝑛"#$ , and the TH. However, looking
at the conventional applications in the field of the refrigeration, one may have to keep a wanted difference
between the hot and the cold source; consequently, a fourth input has been considered: the ∆Tspan.
Figure 2. The ANN proposed in support to ANNTEO
134
At the same manner, one has to select as outputs the quantities which identify the energy performances of a
PMRR. The Q (+& and the COP are sure the most significant quantities, but not least is the electrical power
absorbed by the device; in particular, the knowledge of the 𝑊UV,$ and the 𝑊UV,# could help identify the main
causes of the reduction of the energy performances. Due to these considerations, ANNTEO leads to an ANN
having four inputs and four outputs. Regarding the topology of the wanted ANN, as mentioned in the
introduction, in the literature no references are available. However, one may expect to take in consideration
similar works, which refer to the conventional refrigeration systems. As suggested in [9] the best topology of
an ANN, devoted to tracking the energy performances of a refrigerator, is a feed-forward MLP having one
hidden layer. Regarding the input data, the same literature suggested to normalize them appropriately in the
range of values between -1 and 1. In this manner, in the hidden layer, the hyperbolic tangent can be chosen as
the activation function. To compose the training set, the 224 experiments, obtained with the campaign tests
reported in section 2.2, were considered. The Levenberg-Marquardt algorithm was selected as a training
process, applied along with methods of early stopping cross-validation to avoid the over-fitting problem and
to test the generalization ability of the ANN. In particular, about the latter, the set of experimental data has
been divided in the following way: 60% was included in the training set; 30% was used for the validation set;
10% constituted the test set.
The generation of the network and the learning algorithm implementation were performed by means of the
Neural Network Tool available in Matlab. In addition to the MSE of the network, for each output the coefficient
of determination (R2) was calculated as:
J
HF GE I
𝑅 X = FKL
J (4)
FKL EF GE I
With the aim to not make the network affected by the data used as training set, they have been chosen by means
of a random extraction routine. Several extractions were performed, and a new network was generated for each
of them. Among all the generated network both the MSE and the R2 showed slight changes. By considering
the Eq. 4, the suggested number of hidden neurons is 9. However, a trial error algorithm, finalizing to the
minimization of the MSE, has been employed to carry out the best number of hidden neurons. This procedure
has led to an ANN having 11 hidden neurons (Fig. 2), which was able to obtain an MSE equal to 4.91*10-3.
As shown in Fig. 3, the experimental data and the predicted ones are in good agreement each other: the worst
R2 is equal to 0.98.

Figure 3. Regression line of the outputs of the ANN
135
2.4 Implementation of ANNTEO
It has been already evidenced the main idea which is the basis of ANNTEO is to obtain an optimization map
for an actual magnetic refrigerator without exceeding in the experiments. Once the ANN has been defined, one
can use it to find one or more parameters, which can optimize the energy performances of the device. Due to
its mathematical expression, an ANN can be easily adapted to a function Y=f(X) where the array of the
independent variables represents the array of the ANN’s inputs, while the array of dependent variables
represents the array of the ANN’s outputs. The inputs of the ANN may be distinguished in three group of
parameters: operating (the V&,( and the 𝑛"#$ ), which are controllable using the pump and the electrical motor;
characteristic (the ∆Tspan), which is usually desired; external (the TH), which is uncontrollable because it
depends on the external factors. A possible way to obtain an optimal working point for a PMRR is to bind the
TH and to set the ∆Tspan to the desired value. In this manner, the extra degrees of freedom are two, both of a
type operating. Consequently, one can research that couple of values concerning the V&,( and the 𝑛"#$ , which
optimize the function f, then the energy performances of the PMRR. In particular, the proposed ANN leads to
two possible operational logics: the refrigeration duty optimization, and the COP optimization. Another couple
of logics may be defined regarding the remaining ANN’s outputs (the 𝑊UV,$ and the 𝑊UV,# ), but they could
result similar to the COP optimization. In both cases, a rectangular domain for the independent variables was
defined; in particular, the limits for each quantity was defined in agreement with the limits of working of the
pump and the electrical motor: 5-7 l min-1 and 10.8 – 53.8 rpm. Successively, a script in Matlab was defined
with the aim to resolve both the optimization problems. As a result of the optimization under different
conditions concerning the TH and the ∆Tspan has been carried out the optimal couple ( V&,( ; 𝑛"#$ ) able to
maximize the refrigeration duty but not the COP, and that able to maximize the COP but not the refrigeration
duty. As a sample of the performed optimization in Fig. 4 are reported the surface respectively of the Q (+& and
the COP as a function of the operating parameters of the device. In addition to the optimization of the energy
performances, the Fig.4 suggest ANNTEO provides full operating maps of a PMRR. These maps could be
used to identify the values of the couple ( V&,( ; 𝑛"#$ ) which guarantee the desired refrigeration duty value or a
minimum COP value.
3. CONCLUSIONS
In the present paper a technique named ANNTEO has been introduced. It is a technique based on artificial
neural networks and able to reduce the number of experiments necessary to define an optimization map for an
actual PMRR. Considering the experimental analysis performed with a PMRR, an MLP neural network has
been developed to predict the energy performances by receiving as input four quantities: the V&,( ,the 𝑛"#$ , the
TH and the ∆Tspan. The data predicted by the ANN has shown a significant agreement with those experimental:
Figure 4. The Q̇ (+& and the COP as function of the operating parameters of the device (TH=22°C and ∆Tspan=8°C).
136
the worst R2 has been equal to 0.98, while the MSE has reached a value of 4.91*10-3. The ANN has been
proposed as the basis of ANNTEO and then has been considered as mathematical function Y=f(X). With a
particular focus on the operating parameters such as the V&,( and the 𝑛"#$ , the function f(X) has been optimized
by considering two different logics: the refrigeration duty optimization, and the COP optimization. The results
obtained have suggested ANNTEO provides full operating maps of a PMRR. These maps could be used to
identify the values of the couple (V&,( ; 𝑛"#$ ) which guarantee the desired refrigeration duty value or a minimum
COP value.
REFERENCES
[1] A. Kitanovski, J. Tušek, U. Tomc, U. Plaznik, M. Ozbolt, A. Poredos, 2015 Magnetocaloric energy
conversion, Springer International Publishing, 2015, 269-‐‑330
[2] C. Aprea, A. Maiorino, “A flexible numerical model to study an active magnetic refrigerator for near
room temperature applications”, Applied Energy, 87, 2690-‐‑2698 (2010)
[3] J. Tušek, A. Kitanovski, I. Prebil, A. Poredoš, “Dynamic Operation of an Active Magnetic Regenerator
(AMR): Numerical Optimization of a Packed-‐‑Bed AMR” . International Journal of Refrigeration, 34, 1507–
1517 (2011).
[4]D. Eriksen, K. Engelbrecht, C. R.H. Bahl, R. Bjørk, K.K. Nielsen, A.R. Insiga, N. Pryds, “Design and
experimental tests of a rotary active magnetic regenerator prototype”, International Journal of
Refrigeration 58, 14-21 (2015).
[5] C. Aprea, A. Greco, A. Maiorino, Masselli C., “The energy performances of a rotary permanent magnet
magnetic refrigerator”, International Journal of Refrigeration 61, 1-11 (2016).
[6] J. A. Lozano, M. S. Capovilla, P. V. Trevizoli, K. Engelbrecht, C. R.H. Bahl, J. R. Barbosa Jr.,
“Development of a novel rotary magnetic refrigerator”, International Journal of Refrigeration (2016),
http://dx.doi.org/doi: 10.1016/j.ijrefrig.2016.04.005.
[7] C. Aprea, A. Greco, A. Maiorino, Mastrullo R., Tura A., “Initial experimental results from a rotary
permanent magnet magnetic refrigerator”, International Journal of Refrigeration 43, 111-122 (2014).
[8] D. Marquardt, “An algorithm for least-squares estimation of non-linear parameters”, Journal of Applied
Mathematics 11, 431-444 (1963)
[9] J. Navarro-Esbrì, V. Berbegall, G. Verdu, R. Cabello, R. Llopis, “A low data requirement model of a
variable-speed vapour compression refrigeration system based on neural networks”, International Journal of
Refrigeration 30, 1452-1459 (2007).
137
HIGH-TEMPERATURE HYDROGENATION BEHAVIORS AND
MAGNETOCALORIC PROPERTIES OF LaFe11.6Si1.4 COMPOUND
Tang Yongbai(a)*, Chen Yungui(a), Zheng Hongyan(a), Wu Jianghong(b), Wang Huasheng(c)

(a)School of Materials Science and Engineering, Sichuan University, Chengdu 610065, China
(b) School of Mechanical ＆ Automobile, South China University of Technology, Guangzhou, 510640, China
(c) School of Engineering and Materials Science, Queen Mary, University of London E1 4NS, UK
*
Corresponding author. E-mail: tyb1973@163.com
ABSTRACT
LaFe11.6Si1.4 samples were hydrogenated at the initial hydrogen-filling pressure of 0.6 MPa and different
temperatures (from 393 K to 478 K). The hydrogen absorption kinetics performance, magnetocaloric effect and
Curie temperature (TC) of these alloys were investigated. The results show that all samples have excellent
hydrogen absorption kinetics performance and the hydrogen saturation time of these samples decreased from 45
min (at 393 K) to 4 min (at 478 K). The magnetic entropy changes were 16.27 J/kg.K (before hydrogen
absorption), 12.2 J/kg.K (at 393K), 10.48 J/kg.K (at 423K), 7.81 J/kg.K (at 453) and 10.64 J/kg.K （at 478K,
respectively. The Curie temperature varied from 325.7K for an absorption temperature of 393K to 276.8K for an
absorption temperature of 478K.
DOI:10.18462/iir.thermag.2016.0134
1. INTRODUCTION
LaFe13-xSix (1.1≤x≤1.6) and related compounds with cubic NaZn13-type structure are one of competitive
candidates for the magnetic refrigerants because of the low price of material and adjustable Curie temperature TC.
However, these compounds are not directly suitable for room temperature application due to their low Curie
temperature (TC, 190 K)[1]. In order to make LaFe13-xSix compounds work near ambient temperature, it is
necessary to adjust their TC to room temperature. A great deal of effort has been done to adjust the TC, such as
substituting Fe with Mn or Co [2-5], substituting La with Ce or Pr or Nd [5,6] and adding interstitial atoms H or C
[7-9]. It has been reported that the introducing of interstitial H atoms is the most efficient method to increase the
TC to near room temperature as well as to keep the large magnetocaloric effect [10,11]. A positive linear
relationship between the H content y and Tc for the La (Fe0.88Si0.12) 13Hy compounds was found by A.Fujita [12].
The hydrogen absorption effect of LaFe13-xSix compounds is influenced by many process parameters, such as the
treatment temperature,hydrogen gas pressure, hydrogenation time and particle size[13-16]. For a given particle
size, the hydrogen saturation time as well as the hydrogen absorption capacity decreases with the increase of the
treatment temperature [13]. The usual way to control H content is firstly to hydrogenate compounds until
saturation, and then to desorb some H by thermal desorption [17]. However, the concentration of H desorption
during this process is difficult to control. So it is necessary to find a method to precisely adjust the H content in
LaFe13-xSixHy compounds.
In this paper, the high-temperature hydrogenation method is used to prepare a set of fully hydrogenated
LaFe11.6Si1.4Hy compounds. The hydrogen absorption kinetics performance, magnetocaloric effect and Curie
138
temperature (TC) of these compounds were investigated. The high-temperature hydrogenation can significantly
reduce the hydrogen saturation time of LaFe11.6Si1.4 compounds and precisely adjust the TC.
2. EXPERIMENTAL
The investigated LaFe11.6Si1.4 compounds were prepared by arc melting under the protection of high-purity
argon atmosphere. The purities of raw materials La, Fe and Si were 99.4, 99.9 and 99.9999 wt. %, respectively.
Considering the loss of lanthanum, an excess of 4 % of La was systematically added. The alloy buttons were
re-melted five times to ensure homogeneity. To obtain NaZn13-type main phase, the as-cast LaFe11.6Si1.4
compound was annealed in a ZM-40-16 vacuum molybdenum wire furnace at 1473 K for 5 h [17-18].
The hydrogen absorption and the hydrogen absorption kinetics curves of the compounds were measured by the
PCTPro-2000 Siverts Gas Sorption Analyzer made by Hy-Energy Company. The powder samples were prepared
by mechanical crushing and the particle size measured by Standard sample sieve is between 0.15 and 0.45 mm.
Before the hydrogen absorption, the annealed material was activated twice as follows: after putting the powder
samples into the reactor, the reactor was firstly evacuated at 298K for 30 min and subsequently evacuated at 653
K for 60 min. When hydrogenating, the H2 with purity of 99.999 wt. % was introduced. the hydrogen-filling
pressure was 0.6 MPa, and the samples were kept at different high treatment temperature (393 K, 423 K, 433 K,
443 K, 453 K, 463 K, 468 K, 473 K and 478 K, respectively) for 1 h to saturate, and then the samples were cooled
slowly to room temperature when the hydrogen was taken out. Finally, a series of samples with different and
calculable H content were prepared.
Magnetic measurements were performed using vibrating sample magnetometer (VSM) from Quantum Design.
The Curie temperature (TC) was determined from the maxima of dM/dT of the M-T curves, which were measured
in an applied magnetic field of B=0.02 T. The magnetic entropy changes ΔSM (T, H) were calculated from
isothermal magnetization curves (M-H curves) under magnetic fields of 2T.
Fig. 1 shows the hydrogen absorption kinetics performance of LaFe11.6Si1.4 compounds. The hydrogen absorption
of all samples can approach to a saturation state within about 1 h at the same initial hydrogen-filling pressure of
0.6 MPa but varying treatment temperature. The hydrogen saturation time of these samples decreases from 45 min
Fig. 1. The hydrogen absorption kinetics curves of the LaFe11.6Si1.4 compounds at 0.6 MPa
hydrogen-filling pressure and different treatment temperatures
139
(at 393 K) to 4 min (at 478 K). Compared with the hydrogen saturation time of 670 min at initial hydrogen-filling
pressure of 2 MPa and room temperature [19], it can be seen that the hydrogen absorption kinetics performance of
samples in this work is much better, indicating that high-temperature hydrogenation is an efficient way to save the
hydrogenation time and improve the production efficiency. As shown in Fig.1, the hydrogen absorption capacity
of LaFe11.6Si1.4 compounds decreases with increasing temperature.
Fig. 2 displays the temperature dependence of the magnetization measured under a magnetic field of 0.02 T in the
heating and the cooling cycles for LaFe11.6Si1.4Hy compounds. The Curie temperature TC is found to decrease with
the treatment temperature (Ttr) from 325.7 K for Ttr =393 K to 276.8 K for Ttr =478 K. The increase of TC with
increasing H content y in LaFe11.6Si1.4Hy compounds can be attributed to the lattice expansion caused by the
introduction of interstitial hydrogen atoms which lead to a stronger Fe–Fe exchange interaction. An irreversible
behavior, accompanied by little thermal hysteresis for these compounds, between the heating and the cooling M-T
curves is observed, which is one of the main characteristic of first-order magnetic refrigerants.
Fig. 2 Temperature dependent magnetization for LaFe11.6Si1.4Hy compounds at heating and cooling processes under field of 0.02 T: (a)
before hydrogenation; (b) hydrogenated at 393 K; (c) hydrogenated at 423 K; (d) hydrogenated at 453 K; (e) hydrogenated at 478 K.
Fig.3 The magnetic entropy changes of LaFe11.6Si1.4Hy compounds
140
Fig.3 shows the magnetic entropy change (ΔSM) of the compounds calculated from isothermal magnetization
curves shown in fig.4. The magnetic entropy change was 16.27 J/kg.K (before hydrogen absorption), 12.2 J/kg.K
(at 393K), 10.48 J/kg.K (at 423K), 7.81 J/kg.K (at 453K) and 10.64 J/kg.K(at 478K), respectively. The ΔSM of the
hydrogenated compounds decreased slightly, but still maintained a high value.
CONCLUSIONS
The hydrogen absorption kinetics performance, magnetocaloric effect and Curie temperature of LaFe11.6Si1.4Hy
compounds have been investigated. All samples show excellent hydrogen absorption kinetics performance at high
temperature. The shortest hydrogen saturation time is only 4 min at 478 K. The Curie temperature varied from
325.7K at 393K to 276.8K at 478K.The magnetic entropy change was 16.27 J/kg.K (before hydrogen absorption),
12.2 J/kg.K (at 393K),10.48 J/kg.K (at 423K), 7.81 J/kg.K (at 453K) and 10.64 J/kg.K (at 478K), respectively.
ACKNOWLEDGMENTS
The work was supported by the Key Project of National Natural Science Foundation of China (grant number
51176065).
REFERENCES
[1] J. L. Wang, S. J. Campbell, et al. Hyperfine Interact. 226 (2014) 405-413.

[2] Jia L, Sun J R, Wang F W, et al. Appl Phys Lett, 2008, 92:101904
[3] Y. f. Chen, et al. J Phys. Condens. Matter. 15 (2003) 161-167
[4] Z. W. Li and A. H. Morrish, Phys. Rev. B 55, 3670 1997.
[5]E. Bruck, Handbook of Magnetic Materials, edited by K. H. J. Buschow North-Holland, Amsterdam, 2008, Vol.
17, p. 293.
[6] Jin-liang Zhao, et al. J. Appl. Phys 107, 113911 2010
[7] H. Zhang, B.G. Shen, et al. J. Appl. Phys. 111 (2012) 07A909.
[8] S. Fu, Y. Long, Y.Y. Sun, J. Hu, Intermetallics 39 (2013) 79-83.
[9] M. Balli, D. Fruchart, D. Gignoux, Appl. Phys. Lett. 92 (2008) 232505.
[10] A. Fujita, S. Fujieda, Y. Hasegawa, K. Fukamichi, Phys. Rev. B 67 (2003)104416e104418.
[11] J. Lyubina, K. Nenkov, L. Schultz, O. Gutfleisch, Phys. Rev. Lett. 101 (2008)177203.
[12] A. Fujita, S. Fujieda, Y. Hasegawa, K. Fukamichi, Phys. Rev. B 67 (2003) 104416.
[13] J.W. Wang, Y.G. Chen, Y.B. Tang, S.F. Xiao, T. Liu, E.Y. Zhang, J. Alloys Compds.485 (2009) 313-315.
[14] G.F. Wang, L.J. Mu, X.F. Zhang, Z.R. Zhao, J.H. Huang, J. Appl. Phys. 115 (2014)143903.
[15] F.X. Hu, L. Chen, J. Wang, L.F. Bao, J.R. Sun, B.G. Shen, Appl. Phys. Lett. 100(2012) 072403.
[16] H. Zhang, Y. Long, et al. J. Appl.Phys. 113 (2013) 17A911.
[17] M. Krautz, et al. J. Appl. Phys. 112 (2012) 083918.
[18] H.N. Beza, et al. J. Magn. Magn. Mater. 386 (2015)125-128.
[19] Luo Huihui, Chen Yungui, Tang Yongbo, Chen Xiang, Wu Jianghong, Liu Chaopeng, Rare Metal Materials
and Engineering.42 (2013) 2136-2138.
141
IMPROVEMENT OF THE AMR’S PERFORMANCE: MULTI-SEGMENT
REGENERATOR AND BY-PASS FLOW
S. Lionte(a)*, M. Risser(a), C. Vasile(b), L. Elouad(a), C. Muller(a),

(a)
Cooltech Applications, 5 Impasse Antoine Imbs, 67810, Holtzheim, France
(b)
LGeCo, National Institute of Applied Sciences (INSA), 24 Boulevard de la Victoire, 67084, Strasbourg,
France
*
Corresponding author. E-mail: s.lionte@cooltech-applications.com
ABSTRACT
This paper analyses the improvement of an Active Magnetic Regenerator performance by two methods. The
first one is the layering of several magnetocaloric materials in series by their Curie temperature in order to
create a higher temperature span. The second method is the use of a countercurrent heat exchanger by-pass
flow to increase the machine’s final Coefficient of Performance and cooling power. A time-dependent
numerical model was used to analyze multiple cases of magnetocaloric segment layering. The best layering
was identified and compared with the by-pass configuration. Some interesting results were obtained which
highlighted the point that the joint use of a by-pass on a magnetic refrigeration system with proper MCM
segment layering significantly changes the behavior of an AMR. These results were compared to
experimentally-obtained data from an existing rotary research prototype which had been specially adapted to
fit the requirements of these new operating methods.
Keywords: Magnetic Refrigeration, Active Magnetic Regenerator (AMR), by-pass, multi-segment

regenerator, rotary system.
DOI: 10.18462/iir.thermag.2016.0141
1. INTRODUCTION
Like every development program for new technologies, magnetic refrigeration today faces many challenges
before being widely accepted as a replacement technology for cooling. Since the first near-room temperature
magnetic refrigeration conference, Thermag held in 2005 in Montreux, Switzerland, most of these challenges
have been discussed, partially proven and validated. However, validation of the majority of these constraints
has mainly relied on numerical simulations and laboratory-scale experimentation. There has been much
research done in-silico on different multiphysics numerical models, less research done in-vitro on laboratory-
scale prototypes and even less experiments done in-vivo, reporting on prototypes ready to be deployed in actual
living conditions such as temperature variation and integration of thermal losses. Therefore, in order to obtain
good performance in terms of temperature span and cooling power, with a reasonable quantity of
magnetocaloric materials and a value of magnetic field intensity that could be generated by permanent
magnets, complex device engineering is needed. This could be oriented towards many research axes such as
system development, improvement in materials, adaptive hydraulics solutions and thermal loss reduction.
2. DESCRIPTION OF THE SYSTEM
In order for a commercial Magnetic Refrigeration System (MRS) to be used as an end-user device, a span of
at least 30 K has to be provided along with a sufficient value of cooling power depending on each application
type and a good value of Coefficient of Performance (COP). Cooltech has developed and tested a new
generation of rotary magnetic cooling systems that could now confront these technical requirements.
A multi-segment regenerator made from 8 different segments of up to 5 Curie temperatures (Tc) placed in
serial is used. The magnetocaloric materials consist of Gd-based alloys with Curie temperature values from 0
to 20 °C and a regenerator geometry based on thin parallel plates. The system uses an NdFeB magnet assembly
with a magnetic field intensity value of approximately 1 Tesla (measured in the air gap), which Cooltech
considers to be the optimal economic value. The machine uses several frequency values from 1 to 2 Hz and a
variable flow rate from 3 to 6 l/min. as operating parameters.
142
2.1 Multi-segment Regenerator
The layering of MCM segments with different Curie temperatures is one of the key elements in the proper
functioning of the MRS [1]. This arrangement must be done not only according to the cold and hot source
temperatures but also to the operating parameters such as the Rvol (dimensionless displaced volume) and
frequency that can modify the temperature gradient of the regenerator in the steady-state mode. In order to find
the best configuration of MCM segment layering, several configurations were numerically tested and then
compared to experimentally-obtained data. The numerical model used for the simulation of the AMR behavior
was a time-dependent finite differences-based, previously developed and validated model [2] which was
adapted to integrate the new multiple Curie temperature segment layering.
Five configurations were tested, based on the Curie temperatures of the 8 MCM segments. : 20-20-20-15-15-
10-5-0 °C (Batch 1), 20-20-15-15-10-10-5-5 °C (Batch 2), 20-15-15-10-10-5-5-0 °C (Batch 3), 20-20-15-15-
10-5-5-0 °C (Batch 4) and 20-20-15-15-10-5-0-0 °C (Batch 5). The configurations were chosen based on the
initial temperature gradient of the regenerator in steady-state mode with an ideal pseudo-material with linear
properties and then adapted by iterations to the regenerator temperature gradient given by the point with the
best performance, as a function of the operating parameters.
The best performance (maximum cooling power with a good COP value), in our case, was obtained with the
configuration given in Batch 1 (20°C / 20°C / 20°C / 15°C / 15°C / 10°C / 5°C / 0°C), and the layering of the
MCM segments according to the regenerator's temperature gradient as shown in the figure below:
Figure 1: The multi Tc layering according to regenerator temperature gradient and length
As one can see in Figure 1, this configuration is well adapted to the regenerator temperature gradient on the
cold side, from 0 up to 15°C, while on the hot side a material with a Curie temperature superior to 20 °C is
needed. As we are limited here by the maximum Curie temperature of Gadolinium at 20°C, an MCM serial
layering with different Curie temperatures covering the entire temperature gradient area of the regenerator up
to 30°C would increase the performance of the machine both in terms of temperature span and COP.
Several cases of operating parameter variation (frequency from 1 Hz to 2 Hz and Rvol from 0.3 to 1.5) were
numerically analyzed for five cases of Curie temperature material layering as in Figure 2.
Based on these points the supposed envelope curves were drawn for each Curie temperature layering
configuration. The most adapted configuration for the source temperatures (0 and 30°C) is the configuration
made from 8 segments with 5 Curie temperatures: 20°C / 20°C / 20°C / 15°C / 15°C / 10°C / 5°C / 0°C.
These five configurations were tested experimentally on a specific rotating research prototype and the results
showed the same pattern, with the best performance obtained with the same Curie temperature configuration.
143
Figure 2: Cloud of numerical simulations and envelope curves for 5 configurations of Tc layering
2.2 By-pass Configuration
The magnetic refrigeration heat pump is generally based on an operation with a relatively small temperature
difference between the inlet and outlet of the cold heat exchanger (around 1 or 2 K) and its principle is well
suited to a "dynamic temperature" refrigeration type.
Figure 3: Classic configuration and by-pass configuration applied to an AMR with several Curie temperature layering
For a cooling application of a continuous fluid flow [3], the magnetic heat pump operates with a large
temperature span (ΔT1-Figure 4) between the inlet secondary fluid temperature and the outlet of the heat
transfer fluid from the AMR. The outspread temperature profiles of the two fluids lead to increased
irreversibility and thus a reduction of the Coefficient of Performance.
In a configuration adapted to a cooling application of a continuous fluid flow in a magnetic heat pump with a
by-pass exchanger, a percentage of the heat transfer fluid (1-10%) from the cold side of the AMR passes the
by-pass exchanger [4] and is reinjected into the hot side of the AMR, as in Figure 3. Therefore the temperature
profiles of the fluids in the by-pass exchanger are closer (ΔT2-Figure 4) which lead to an increase in the
system's performance. However, if this value of the fluid flow is too high, the AMR regeneration is
incompletely realized which implies a modification of the thermal gradient of the regenerator with a possible
impact on the cooling power value.
Figure 4: Heat exchanger temperature diagram on a classic configuration (left) and a by-pass configuration (right)
On a classic configuration, the secondary fluid temperature difference is superior to the temperature difference
of the heat transfer fluid and the heat exchange effectiveness is given by Eq. 1:
T1in  T1out 32  2
left    0.91 (1)
T1in  T2in 32  0.95
with T1in = 32°C (outside ambient temperature), T1out = 2°C (conventionally chosen) and T2in = -0.95°C (from
simulations).
144
If a by-pass configuration is used, the secondary fluid temperature difference is inferior to the temperature
difference of the heat transfer fluid and the heat exchange effectiveness is given by Eq. 2:
T2out  T2in 30  0.95
right    0.94 (2)
T1in  T2in 32  0.95
with T1in = 32°C (outside ambient temperature), T2in = - 0.95°C (from simulations) and T2out= 30°C (hot source
temperature). As the effectiveness of the by-pass heat exchanger is higher, it leads to temperature values better
adapted to this type of application.
In order to determine the minimum percentage value of the heat transfer fluid flow through the by-pass heat
exchanger from which the configuration with by-pass gives the same cooling power as the classic
configuration, the following properties are used: heat transfer fluid with Cp = 4070 J/kg·K and ρ =1010 kg/m3,
mass flow rate = 0.021 kg/s and heat transfer temperature = - 0.95°C (both values taken from the point giving
the highest cooling power in simulations).
The heat transfer rate from the heat transfer fluid on a configuration without by-pass is given by Eq. 3:
  m 2  Cp2  (T2out  T2in )  0.021kg / s  4070J / kgK  0.95K  81.12W (3)
The minimum percentage value of the heat transfer fluid flow through the by-pass exchanger from which the
configuration with by-pass gives the same cooling power as the classic configuration is given by:
 81.12W
x   0.031  3.1% (4)
m 2  Cp2  (T2out  T2in ) 0.021kg / s  4070 J / kgK  30.95C
Under this percentage value of fluid flow through the by-pass heat exchanger, the by-pass solution is not
interesting from the point of view of cooling power. This value of percentage is confirmed by simulations of
a pair of two regenerators working dephased and giving a continuous by-pass fluid flow.
3. CONCLUSIONS
For the proper functioning of a magnetic heat pump the layering of the magnetocaloric materials of the
regenerator is one of the key elements. This layering, based on Curie temperatures, must be made according
to the temperature of the cold and hot sources and also in accordance with the operating parameters (such as
the Rvol and frequency) that can modify the temperature gradient of the regenerator in the steady-state mode.
On a continuous fluid flow cooling application the by-pass heat exchanger leads to reduced irreversibility due
to closer temperature profiles and is better adapted to this application’s temperature values. The percentage of
the by-pass fluid flow must be chosen in order not to modify the temperature gradient, which can lead to a
possible impact on the cooling power. Even better performance could be obtained if the layering is directly
adapted to the temperature gradient of the regenerator given by the by-pass configuration.
4. ACKNOWLEDGEMENTS
The authors would like to express their gratitude to the French National Research Agency (ANR), this study
being partially supported by ANR-10-STOCK-E project.
REFERENCES
[1] A. Rowe, A. Tura, “Experimental investigation of a three-material layered active magnetic regenerator”,
International Journal of Refrigeration 29, 1286-1293 (2006).
[2] M. Risser et al., “Numerical simulation of magnetocaloric system behavior for an industrial application”,
International Journal of Refrigeration 33, 973-981 (2010).
[3] J. Barclay et al., “Unique features of liquefaction of hydrogen and natural gas using magnetic refrigeration”,
Proc. of the 6th IIF-IIR Int. Conf. on Magnetic Refrig. Victoria, BC, 7-10 September 2014.
[4] B.T. Kulakowski, F.W. Schmidt, “Design of Packed Bed Regenerator with By-pass”, Heat Transfer
Engineering 4, 17-23 (1983).
145
NEW MAGNETIC REFRIGERATION PROTOTYPE WITH APPLICATION
IN HOUSEHOLD AND PROFESSIONAL REFRIGERATORS
B. Monfared*, B. Palm
KTH Royal Institute of Technology, Brinellvägen 68, 10044, Stockholm, Sweden
*
Corresponding author. E-mail: behzadam@kth.se
ABSTRACT
The number of magnetic refrigeration prototypes with high cooling capacity and large temperature span is
limited and there is ample room for new designs and improvements. In this paper a new prototype, designed
and built, aiming at 200 W cooling capacity and about 40 K temperature span is presented. Such a unit is
suitable for applications in household and professional refrigerators. In the current work, design of the
prototype is described, practical issues solved to make the prototype running are explained, and preliminary
test results are presented.
Keywords: magnetic refrigeration, prototype, high capacity, rotary.

DOI: 10.18462/iir.thermag.2016.0142
1. INTRODUCTION
In this article a magnetic refrigeration prototype designed and built for application in professional
refrigerators is presented. The professional refrigerator, for which the prototype is designed, keeps the
cabinet temperature at 2 °C for ambient temperatures as high as 43 °C. The estimated heat leak to the cabinet
for ambient temperature of 25 °C and 43 °C are about 90 and 165 W. Therefore, to replace the vapor-
compression system currently used in such units, the magnetic refrigeration prototype is designed to deliver
200 W cooling at 43 °C ambient temperature, which covers the capacity needed at moderate ambient
temperature of 25 °C.
2. THE PROTOTYPE, EXPERIMENTAL RESULTS AND PRACTICAL ISSUES
In this section the design of the prototype, the magnetic field, the test results, and the practical issues related
to the prototype are presented.
2.1 Design of the prototype
The designed prototype is of rotary type with rotating magnet and fixed regenerators. The regenerators are
packed beds of La-Fe-Mn-Si-H particles with irregular shapes. As shown in Fig. 1, the number of
regenerators is 12; however, the results presented in this article are obtained with only 2 of them filled with
the magnetocaloric materials. Each regenerator takes about 0.12 kg magnetocaloric materials
Figure 1. Regenerators of the prototype filled with magnetocaloric materials.
The regenerators are inside a soft-magnetic yoke, shown in grey in Fig. 2, needed to concentrate the
magnetic field. The magnet assembly, which creates two high-field regions and two low-field regions, is
placed inside the regenerators’ casing. Therefore, the regenerators at two ends of the casing’s diameters go
through similar cycles. As the magnet rotates continuously, the magnetized regenerators (red ones in Fig. 2)
remain in the high field region during about 30° rotation angle, which provides enough time for the heat
146
transfer fluid to flow from the cold end to the warm end. As the magnet continues its rotation the similar
flow goes through the next regenerators; therefore, the warm fluid leaving the regenerators and entering the
warm heat exchanger is unidirectional, allowing the heat exchanger works effectively. The flow in the
opposite direction for the demagnetized regenerators (blue in Fig. 2) and in the cold heat exchanger is
similar. Two mechanisms are used for distributing the heat transfer fluid between the regenerators: a number
of solenoid valves or a rotary valve. The solenoid valves give flexibility for laboratory testing; however, for
the final application the rotary valve coupled mechanically with the magnet shaft is used.
Figure 2. Continuous operation of the prototype shown for 60° rotation of the magnet assembly.
For laboratory tests a stepper motor is used for full control over the magnet position. In the actual application
the stepper motor can be replaced by an ordinary inverter-motor.
2.2 Magnetic field

The magnetic field in the air-gap between the magnet and the yoke is calculated to be 1.1 T for the high-field
region (Fig. 3) and less than 0.1 T for the low field region. The simulation is done using FEMM software.
Figure 3. High-field region (central 30°) and declination of the field out of it.
Since the permeability of the magnetocaloric materials are significantly higher than that of air, it is expected
that the external field becomes stronger in presence of magnetocaloric materials, although demagnetization
field lowers the internal field.
2.3 Test results

The temperature measurements shown in Fig. 4 are done for a test with only two, out of 12, regenerators
filled. The layers of the magnetocaloric materials chosen in this test have transition temperatures between -3
and 26 °C, not intended to create the maximum design temperature span. The test was done with water as
heat transfer fluid therefore it is stopped before the water leaving the cold end reaches 0 °C.
147
Figure 4. Temperature measurement results.
The cold end and the warm end temperatures in Fig. 4 are the average heat transfer fluid temperatures
leaving the two regenerators. Flow rate of the heat transfer fluid is 8.3×10-6 m3s-1, and the frequency is 0.5
Hz. Although no external cooling load is applied, the cooling capacity has been used on reducing the
temperature of the yoke and regenerators’ casing closer to the cold end. The parasitic loads such as heat
transfer from the ambient to the cold end and also the axial heat transfer from the warm end to the cold end
constitute other parts of the cooling load.
2.4 Practical issues

Limiting the heat transfer between the ambient and the regenerators is challenging since the high-field region
produced by magnet, due to the cost of magnet materials and the overall size of the refrigeration system,
cannot be designed so large that it leaves room for insulating materials. In this prototype, after initial heat
transfer tests, it was decided to reduce the outer diameter of the regenerator casing by 1 mm to leave an air
gap between the yoke and the casing, acting as insulation. The added thermal resistance is estimated to be
about 0.3 KW-1.
The main challenges slowing the experimental work down are related to the magnetocaloric materials. Due
to non-uniform properties of the materials in each package it was necessary to separate the materials
manually. The separation was done by changing the temperature of a bath containing particles in 1K steps
and separating the strongly magnetized particles at that temperature by a magnet. By using unseparated
original materials no or very little cooling was achievable. In addition, non-uniform shape and size of the
particles and also their pulverization during packing result in filling the void between larger particles by
smaller ones and increasing the pressure drop. A possible solution to fix the particle positions is bonding
them together using epoxy [1].
We successfully managed to find a procedure to recycle mixed and corroded materials. The corroded
materials were first cleaned in an ultrasonic bath to become shiny again. Afterwards separation was done as
explained above. Finally the particles were dried in a vacuum oven. However, during each process some of
the materials get pulverized and wasted.
3. CONCLUSIONS
A rotary magnetic refrigeration prototype is designed and built for application in professional refrigerators.
Its design and core parts of it are presented. The results with only one sixth of the regenerators filled and
rather low operation frequency shows cooling with the rate of about 1.8 Kmin-1 in transient operation with
initial thermal equilibrium with the ambient. The major part of the cooling load during the operation is
pulling down the temperature of the yoke and the regenerator casing close to the cold end.
148
Apart from the higher magnetocaloric effect through developing new materials or the processes, it is also
desirable to have improved quality of the materials, such as uniformity in size and properties, for reliable
operation of magnetic refrigeration prototypes.
The manual procedure reported for recycling materials can be developed and automatized to be applicable to
larger scales of used materials. Through recycling of the magnetocaloric materials the environmental profile
of magnetic refrigerators can be improved and their cost can be lowered.
As the future work, the prototype is going to be completed and tested systematically with different working
conditions and configurations.
ACKNOWLEDGMENTS
The authors would gratefully like to acknowledge that this study was financed by the Swedish Energy
Agency and Electrolux through the research program EFFSYS EXPAND.
REFERENCES
Refrigeration, 37, p.p. 84 (2014).
149
SIMULATION OF MAGNETIC REFRIGERATION SYSTEMS WITH
THERMAL DIODES AND AXIAL CONDUCTIVE HEAT TRANSFER
B. Monfared*
KTH Royal Institute of Technology, Brinellvägen 68, 10044, Stockholm, Sweden
*
Corresponding author. E-mail: behzadam@kth.se
ABSTRACT
In conventional magnetic refrigeration cycles with heat transfer fluid, the achievable cycle frequency, and
therefore, specific cooling capacity is limited. Furthermore, ineffective use of magnet in low frequency
devices makes them expensive. In this work, as an alternative technique, utilizing conductive heat transfer
controlled by two different types of thermal diodes, gas-liquid and contact-break diodes, is investigated. For
that purpose two software models are made to simulate the performance of a magnetic refrigerator with each
of the diodes. The results of simulations are presented and comparison is made between these results and
the results of older studies which used ideal properties. According to the results, due to the limited thermal
conductivity of the magnetocaloric materials, the increase in the capacity becomes small with too high
frequencies. Among the thermal diodes and materials studied, the liquid metal Galinstan as the conducting
fluid in an active diode gave the best results.
Keywords: magnetocaloric, refrigeration, solid-state, thermal diode, heat gate, heat switch.
DOI: 10.18462/iir.thermag.2016.0143
1. INTRODUCTION
Conventional magnetic refrigeration devices with a heat transfer fluid going through a porous media have
limited operation frequency. The low frequency necessitates large amount of magnetocaloric materials
(MCM), which implies larger magnet bores and magnet mass, be used to get the needed cooling capacity.
Other drawbacks of such systems are the considerable pressure drop in the porous media and complications
related to the hydraulic system. Kitanovski and Egolf [1] suggested thermal diodes facilitating heat transfer
from cold end to warm end and having higher thermal resistance in the opposite direction as a solution to
increase the frequency. The thermal diodes use different technologies summarized by Kitanovski et al. [2].
Tomc et al. [3] used Peltier elements, active components consuming power, as thermal diodes in a special
design with a long flat bar of magnetocaloric materials (MCM), above and below which series of Peltier
elements guide the heat from the cold to the hot heat exchanger. Tasaki et al. [4], Olsen et al. [5], and Silva
et al. [6] have modeled a solid-state magnetic cooler for mobile applications. No specific design is
considered for the passive thermal diodes in their works. Olsen et al. [5] and Silva et al. [6] assumed zero
conductivity for inactive thermal diodes and finite values for active diodes. Tasaki et al. [4] assumed infinite
conductivity for active diodes. The thermal conductivity values for Gd in the works of Tasaki et al. [4] and
Olsen et al. [5] are much larger than the actual values since these works investigate the performance limits,
not the actual performance, of a solid-state refrigeration system.
In this work two possible techniques for thermal diodes and their implications for the performance of the
refrigeration system are investigated. First, moving fluids in and out of a small gap between MCM layers
are considered for making thermal diodes. Possibility of putting together and taking apart the layers for
controlling the direction of conduction is also investigated. In addition, comparison with the older results
are made.
150
2. MODELING AND RESULTS
The working principles of both modeled systems are the same. First every other MCM layer is magnetized.
Then, thermal diodes on their side closer to the heat sink are activated to facilitate the heat transfer in heat
source to heat sink direction. After that, all the diodes become inactive and the other MCM layers become
magnetic and warm. Finally, the thermal diodes which have not been active so far become active. Then the
next cycle starts. For more detailed explanation of the working principle please see reference [5].
The differential equation describing the processes is Eq. (1), which is 1D heat equation with magnetocaloric
effect as a temperature dependent source term:
𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕

�𝑘𝑘 � − 𝜌𝜌𝜌𝜌 = 𝜌𝜌𝜌𝜌 (1)
𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕 𝜕𝜕𝜕𝜕
By defining ∂s/∂B as a function of space and setting it to zero for thermal diodes, a single differential
equation can describe the whole system consisting of MCM and thermal diodes. In Eq. 1 conductivity, k,
density, ρ, and heat capacity, c, are functions of both time and position. The rest of the parameters are T for
temperature, t for time, and x for position. The results are obtained by solving Eq. 1 using finite element
method for model 1 and finite difference method for model 2. The choice of the numerical methods are to
make modeling more straightforward considering the nature of each problem. The magnetic field variation
used in the models is a linear change from 0 to 1 T. Gadolinium with properties reported in [7] is used as
MCM; however, for multi-layering the Curie temperatures are shifted as explained in [8].
2.1 Gap filled with different fluids as thermal diode (model 1)

In model 1, the diodes are modeled as shown in Fig. 1. A liquid and a gas are encapsulated in small cells
between two layers of MCM or MCM and hot/cold heat exchangers (HE). The lower part is a flexible
membrane used to push the liquid into the gap while the upper part is rigid allowing to compress the gas. In
the model the fluid leaving the gap keeps its average temperature until it reenters the gap. That is the
temperature of the gap elements changes together with their thermophysical properties after switching the
fluids. The pressure drop due to moving the fluids is not considered since the speed of movement depends
on the detailed design of the machine, which is not in the scope of this study.
Figure 1. Liquid-gas thermal diode. a) inactive diode with low conductivity b) active diode with high conductivity
For active phase of the diodes, it is desirable to use a fluid with high thermal conductivity, such as
Galinstan, a liquid alloy with melting point of -19 °C. Since Galinstan reacts with air, an inert gas with low
conductivity such as Argon, Ar, the most common noble gas, is needed to be paired with it. Table 1
lists thermal conductivity of a number of fluids for comparison. The pairs tested in this work are Galinstan-
Ar, Galinstan-air (although not practical), and water-air.
air Ar NH3 CO2 Galinstan Hg water ethanol engine oil

k (Wm K ) 0.026 0.016 0.025 0.017
-1 -1
15.5 8.5 0.59 0.17 0.15
Table 1 A number of fluids with the highest and the lowest thermal conductivities at 25 °C and 100 kPa
151
Table 1 summarizes the simulated cold end temperature of arrays consisting of two 4 cm2 layers of MCM
separated with thermal diodes of the length 0.2 times that of each MCM layer. The simulations are done for
no-load condition on the cold end and convective heat transfer with a heat sink at 21 °C on the hot end with
heat transfer coefficient of 200 Wm-2K-1.
fluid pairs Galinstan-Ar Galinstan-air water-air

2 mm layer length, 5 Hz 19.72 19.71 21
0.5 mm layer length, 10 Hz 17.40 17.42 18.03
Table 2 Calculated cold end temperature for different fluid pairs used in model 1
As can be inferred from the results in Table 2, water-air pair, which failed to establish a temperature
span for larger MCM layers and slower cycle, has a limited potential for this purpose and use of liquid
metals such as Galinstan, although more expensive and difficult to work with, can pay off.
2.2 Mechanical contact breaking diodes (model 2)

Another possible mechanism for thermal diodes investigated in this work is separating the MCM layers
during inactive phase of thermal diodes and placing them together for activating the thermal diodes. With
this solution, even during active phase of thermal diodes the contact resistance between the MCM layers
should be taken into account. The contact resistances tested are 10-5, 10-4, and 10-3 m2KW-1. Inactive thermal
diodes are modeled as 1 mm air gap between the MCM layers. Table 3 summarizes the calculated cold end
temperatures for two layers of MCM with 4 cm2 cross section area. The simulations are done for no-load
condition on the cold end and convective heat transfer with heat transfer coefficient of 200 Wm-2K-1 with a
heat sink at 21 °C.
contact resistance for active diodes (m2KW-1) 10-5 10-4 10-3

2 mm layer length, 5 Hz >21 >21 >21
0.5 mm layer length, 10 Hz 19.34 19.65 >21
Table 3 Calculated cold end temperature (°C) for different contact resistance for active diodes in model 2
The two layered array of MCM and contact-break diodes did not create a temperature span for the 5Hz cycle
and 2mm length of MCM layers. Similarly, with 0.5 mm layers of MCM, 10 Hz cycle frequency and contact
resistance of 10-3 m2KW-1 for active diodes the cold end temperature did not go below the heat sink
temperature, 21 °C, which means there was a need to reject input work as heat from both ends.
2.3 Comparing with previous works

In this section comparison is made between the performance of a device with the two thermal diodes
described and the extreme conditions investigated by Olsen et al. [5] (Fig. 2). For that purpose, the
Figure 2. a) values reported by Olsen et al. [5] b) our results with assumptions made by Olsen et al. [5] c) ideal
diodes but actual conductivity of MCM d) contact-break thermal diode with 10-5 (m2KW-1) contact resistance for
active diode e) Galinstan-Ar diode f) water-air diode. (The figure to the right magnifies curves c to f).
152
magnetocaloric effect modeled by mean field theory, which overestimates the effect, is used for 1.37 T field
change. The results shown in Fig. 2 (given for unit volume of the whole array of MCM and diodes with 24
layers of MCM) are for a temperature span between 308 and 248 K. Each layer is 0.5 mm and the length of
the diodes is 0.2 time of the MCM layers. Each of the magnetization or demagnetization processes takes 5%
of the cycle period. The values reported by Olsen et al. [5], with ideal conductivities as high as 400 Wm-1K- 1
for both MCM and the thermal diode and 0 Wm-1K-1 for inactive diodes, are shown in Fig. 2. To show that
the results of our model and the one used by Olsen et al. [5] are in agreement, an extra curve showing the
outcome of our model with the conductivity values used by Olsen et al. [5] is also added.
3. CONCLUSIONS
In this work solid-state magnetic refrigeration is considered as a solution for increasing the specific cooling
power of magnetic refrigeration devices. Two numerical models simulating solid-state magnetic
refrigeration devices are developed to investigate the liquid-gas and contact-break thermal diodes, which
have simpler mechanisms, and probably are easier to build, among the proposed technologies for passive
thermal diodes. More details, including loss mechanisms, are taken into account in these models compared
to the previous works using rather ideal passive thermal diodes.
A limitation for the liquid-gas thermal diodes is the low conductivity of the ordinary fluids. Ordinary liquids
such as water give rather poor results, while liquid metals are more promising. The Galinstan-Ar pair gave
results even slightly better than the contact-break diodes. The liquid-gas diodes can be more feasible since
the position of the MCM layers does not vary during operation and axial force lowering contact resistance
is not needed.
As shown in Fig. 2, with actual materials the capacity cannot be increased effectively by increasing the
frequency over a limit, about 50 Hz in the studied conditions. The main hindrance is the thermal conductivity
of MCM (compare curves b and c in Fig. 2), not the diodes. However, the non-ideal diodes reduce the
performance even further. Apart from the limitations due to the thermophysical properties of the materials,
the relaxation time for phase transition in MCM and the need for changing magnetic field fast enough in
high frequencies can be limiting as well.
REFERENCES
[1] A. Kitanovski, P. Egolf, “Innovative ideas for future research on magnetocaloric technologies”,
International Journal of Refrigeration, 33, pp. 449 (2010).
[2] A. Kitanovski et al., Magnetocaloric Energy Conversion: From Theory to Applications, Springer (2015).
[3] U. Tomc et al., “A new magnetocaloric refrigeration principle with solid-state thermoelectric thermal
diodes”, Applied Thermal Engineering, 58, pp. 1 (2014).
[4] Y. Tasaki et al., “A study on the fundamental heat transport potential of an in-vehicle magnetic
refrigerator”, Proceedings of the 5th International Conference on Magnetic Refrigeration at Room
Temperature, IIR, pp. 445 (2012).
[5] U.L. Olsen et al., “Modeling of in-vehicle magnetic refrigeration”, International Journal of
Refrigeration, 37, pp. 194 (2014).
[6] D. J. Silva et al., “Maximizing the temperature span of a solid state active magnetic regenerative
refrigerator”, Applied Energy, 113, pp. 1149 (2014).
[7] J. A. Lozano et al., “Experimental and numerical results of a high frequency rotating active magnetic
refrigerator”, International Journal of Refrigeration, 37, pp. 92 (2014).
[8] B. Monfared, B. Palm, “Optimization of layered regenerator of a magnetic refrigeration device”,
International Journal of Refrigeration, 57, pp. 103 (2015).
153
OPTIMIZATION OF HEAT EXCHANGE IN MAGNETIC
REFRIGERATORS AND THERMOMAGNETIC MOTORS
C. V. X. Bessa(a)*, L. D. R. Ferreira(a), S. Gama(b), T. Santos(c), J. S. Amaral(c), V. A. F.
Costa(c), V. S. Amaral(c), O. Horikawa(a)
(a)
University of São Paulo, 2231 Av. Prof. Mello Morais, 05508-030, São Paulo, Brazil
(b)
Federal University of São Paulo, 1535 Manoel da Nóbrega, 09910-720, Diadema, Brazil
(c)
University of Aveiro, Campus Universitário de Santiago, 3810-193, Aveiro, Portugal
*
Corresponding author. E-mail: carlosviniciusxb@usp.br
ABSTRACT
The power and efficiency of magnetic refrigerators and thermomagnetic motors depend on their operating
frequencies. To improve the efficiency of these machines, it is necessary to design the regenerators/heat
exchangers so to present heat transfer as high as possible, allowing higher operational frequencies. In several
machines, a fluid flowing through internal channels promotes the heat transfer. In such cases, as the
temperature difference governing heat transfer decreases along the channels length, the heat transfer also
decreases. One solution for this problem is the use of non-linear heat transfer area increase along the
channels. That can be done by changing the cross section area or the profile of the channels. This study
presents an experimental analysis that shows the effectiveness of this technique by comparing heat
exchangers with cylindrical, conical and non-prismatic shaped internal channels using infrared
thermography.
Keywords: heat transfer, magnetocaloric refrigeration, thermomagnetic motors, thermography analysis.

DOI: 10.18462/iir.thermag.2016.0145
1. INTRODUCTION
The magnetocaloric effect can be used in refrigeration or heating as an alternative to the traditional air-
conditioning applications [1]. The cooling or heating capacity and the performance of these machines are
strongly related to the operation frequency [2, 3], and that defines the time of heat exchange between the
magnetocaloric regenerator and the working fluid. Therefore, the optimization of efficiency of these
machines can be done through the optimization of heat transfer. The thermomagnetic effect can be used to
convert thermal energy into mechanical energy in thermomagnetic motors [4, 5]. An important problem
verified in this technology is the low operation frequency, as consequence of the limitations in the heat
transfer through heat exchangers [6]. Magnetocaloric refrigerators and thermomagnetic motors can be
described by Brayton cycles operating in forward or reverse way. Many of these refrigerators and motors use
the flow of a working fluid in internal channels to promote the heat transfer in the magnetocaloric
regenerators [1, 7] or thermomagnetic heat exchangers [6, 8]. In both cases, the enhancement of the heat
transfer can improve the efficiency and power of these machines.
This paper presents an experimental study about transient heat transfer in solid plates containing internal
channels with different shapes. These plates emulate the heat exchangers of a thermomagnetic motor or the
regenerators of a magnetocaloric refrigerator.
2. HEAT EXCHANGERS STUDIED
The heat exchangers considered here are plates with internal channels shown in Fig.1a. These three designs
are inspired in the thermomagnetic motor proposed by Ferreira et al. [6]. All the plates have the same
external dimensions of 50.14 mm × 8 mm × 110 mm and volume of 23,907 mm3. The difference is only in
the geometry of internal channels. The first plate has cylindrical channels, therefore each channel has
constant perimeter and cross section area along the axial direction. The second plate has conical channels,
and with that both the perimeter and the cross section area of the channel increase along the plate length. The
channels of the third plate have different shapes in the plate inlet and outlet. In the inlet these channels are
circular and smoothly change along the channels length to become triangular in the outlet. These non-
154
prismatic channels have a practically constant cross section area but an increasing perimeter along the length
of the plate. This study focuses on the impact of channels geometry on heat exchange and temperature
distribution of heat exchangers. Therefore, for manufacturing simplicity, instead of thermomagnetic
materials, exchangers made of steel were used. In order to avoid rust in the channels, the stainless steel was
used, specifically the 403 stainless steel having a density of 8,0 kg/m3, thermal conductivity of 16.2 W/(m.K)
and specific heat of 500 J/(kg.K).
The heat transfer in the plates is a function of the heat transfer area (AT), and that can be calculated by
∫ , where n is the number of channels, p is a channel perimeter, l is the total length of the plate
and x is the axial direction of the channels. For a linear increase of the heat transfer area, as studied here in
cylindrical channels, the heat transfer decreases from the inlet to the outlet as the temperature difference
between the working fluid and the channels walls (temperature difference governing heat transfer) decreases
along the channels length. One alternative to equalize the heat transfer along the heat exchanger is to
increase p along x by changing the channels cross section area or profile, resulting in a non-linear AT
increase. Fig. 1b shows AT/n as a function of the channel length for each studied case. It must be noted that
the total heat transfer area is the same for cylindrical and conical channels.
Figure 1. Studied heat exchangers. (a) Heat exchangers profiles (dimensions in mm). (b) Accumulated heat transfer area
of one channel as a function of the heat exchanger length.
3. EXPERIMENTAL SETUP
The setup shown in Fig. 2a was used to compare the heat transfer aspect and the temperature distribution in
the heat exchangers plates during the heating and cooling transient thermal cycles. The experiment consisted
on measuring the temperature at the surface of each heat exchanger as a function of time when hot/cold
water flowed in its internal channels. By comparing results, it was possible to verify how the non-linear heat
transfer area increase affects the heat transfer. To do that, an infrared camera (IR FLIR SC 5650) measured
the temperature of the heat exchanger surface via IR thermography. Ti and To are thermocouples that
measured the temperature of the water in the inlet and outlet, respectively. Pi and Po are sensors that measure
the inlet and outlet pressure, respectively. V is a mass flow rate sensor and a gate valve assembly, which
permitted to adjust the water mass flow rate (the subscripts H and C indicate hot and cold water,
respectively). The setup has two water circuits with different temperatures. By switching these two circuits
with 2/3 way solenoids valves (S), it was possible to promote a cooling cycle or a heating cycle. Fig. 2b
shows a picture of the system.
In the experiments, first, the mass flow rates of the hot and of the cold water are adjusted. Then, by changing
the position of S, cold water flows until the exchanger reaches a thermal steady-state. Now, by changing the
position of S, hot water flows in the exchanger internal channels. From this instant, the IR camera measures
the temperature distribution as a function of time on the heat exchanger surface during this heating cycle
until a thermal steady-state was reached. After that, the position of S changes again, the cooling cycle starts.
The IR camera measures the temperature distribution on the exchanger surface again. We applied this
procedure for the three studied heat exchangers and for water mass flow rates of 0.02 kg/s, 0.03 kg/s and
0.04 kg/s, and compared the results.
155
Figure 2. Experimental setup. (a) Schematic view. (b) Picture of the assembled setup.
4. RESULTS
To compare the studied cases it is convenient to show the results in dimensionless form. The dimensionless
temperature (θ) is shown in Eq. (1). T is the instantaneous temperature, i.e., the temperature in a given time,
Tmin is the smallest value of temperature in the studied cycle, and Tmax is the highest value of temperature in
the studied cycle. The dimensionless temperature of water in the inlet (θi), measured by Ti (Fig. 2a) as a
function of the time is shown in Fig. 3a. It is possible to observe that θi follows the equation proposed by
Liang and Yang, given by Eq (2), t is the time and i is the characteristic time of the temperature of the water
in the inlet [9].
⁄ (1)
⁄
(2)
The average dimensionless temperature of the exchanger surface (θs) during a heating and a cooling cycles,
for a mass flow rate of 0.02 kg/s, is shown in Fig. 3b. The heat exchangers with cylindrical and conical
channels have the same total heat transfer area, however the exchanger with conical channels presented a
faster increase of its surface temperature than the exchanger with cylindrical channels. This is a consequence
of the non-linear increase of the heat transfer area in the conical channels. The heat transfer area distribution
compensates the decrease/increase of temperature in the water along of the channels length during the
heating/cooling cycle. The heat exchanger with non-prismatic channels shows the best response in terms of
heat transfer, as expected, since it has the largest total heat transfer area and a non-linear increase of the heat
transfer area. However the exchanger with conical channels and non-prismatic channels reached the same
value of θs when t/i is around 43, before the thermal steady-state. The mass flow rates of 0.03 kg/s and 0.04
kg/s resulted in the same conclusions and curve distributions.
The dimensionless temperature surface distributions of the three studied heat exchangers for t/i = 8 and a
mass flow rate of 0.02 kg/s during a heating cycle are shown in Fig. 3c. It is possible to verify using
thermography that the fluid velocity is higher in the central channels than in the lateral channels of the heat
exchangers by verifying the temperature surface profiles. This distribution of velocities was predicted by
computational fluid dynamics analysis (CFD). In thermomagnetic motors, a distribution of temperature as
homogeneous as possible in the heat exchangers is important to avoid motor locking. The heat exchanger
with non-prismatic channels has problems with its temperature distribution, and this is caused by the
asymmetry in the heat exchanger mass. This asymmetry can be observed in Fig 1a at the outlet side of this
heat exchanger. The heat exchanger with conical channels has the best temperature distribution.
The pressure drop is practically the same for the plates with cylindrical and non-prismatic channels. The
pressure drop for the plate with conical channels was around 15% smaller than the other two for the studied
mass flow rates range.
156
Figure 3. Results. (a) Dimensionless temperature of water in the inlet for a mass flow rate of 0.02kg/s. (b) Average
dimensionless temperature of the heat exchangers surface. (c) Temperature distribution on the heat exchangers surfaces.
5. CONCLUSIONS
The conducted experimental analysis demonstrates that, the non-linear increase of the heat transfer area
along the heat exchanger is an effective way to avoid the reduction in the heat transfer, consequence of the
reduction of the temperature difference governing heat transfer along the heat exchanger. The heat
exchangers with conical and non-prismatic (variable perimeter) channels present a better response than the
heat exchanger with cylindrical channels (constant perimeter). The heat exchangers with cylindrical and
conical channels, considered in this study, have the same total heat transfer area, however the conical one has
a better heat transfer response. This way, a magnetocaloric refrigerator with regenerators with channels that
have a non-linear increase of the heat transfer area can operate in frequencies higher than one with linear
increase of the heat transfer area. The same is expected with respect to thermomagnetic motors. This
technique can be used to improve the efficiency and power of these machines, however the challenge of
fabricating this type of plates using magnetocaloric materials still has to be addressed.
ACKNOWLEDGMENTS
The Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP) grant number 2012/09486-0
supported this work. The authors also thank the Conselho Nacional de Desenvolvimento Científico e
Tecnológico (CNPq). This work is dedicated to our deceased colleague Dr. I. da Silva.
REFERENCES
[1] L. Tagliafico et al., “Preliminary experimental results from a linear reciprocating magnetic refrigerator
prototype”, Applied Thermal Engineering, vol. 52, 492 (2013).
[2] A. Tura et al., “Experimental and modeling results of a parallel plate-based active magnetic regenerator”,
International Journal of Refrigeration, vol. 35, 1518 (2012).
[3] P. Tevizoli et al., “Modeling of thermomagnetic phenomena in active magnetocaloric regenerators”,
Journal of Thermal Science and Engineering Applications, vol. 6, 031016-1 (2014).
[4] C. Palmy, “A new thermo-magnetic wheel”, European Journal of Physics, vol. 27, 1289 (2006).
[5] M. Trapanese et al., “Optimization of a thermomagnetic motor”, Journal of Applied Physics, vol. 117,
17A750-1 (2015).
[6] L D. R. Ferreira et al., “A heat transfer study aiming optimization of magnetic heat exchangers of
thermomagnetic motors”, International Journal of Refrigeration, vol. 37, 209 (2014).
[7] M. Risser et al., “Improvement and application of a numerical model for optimizing the design of
magnetic refrigerators”, International Journal of Refrigeration, vol. 36, 950 (2013).
[8] L. D. R. Ferreira et al., “A linear reciprocating thermomagnetic motor powered by water heated using
solar energy”, Green Design, Materials and Manufacturing Processes, CRC Press (2013).
[9] C. Liang, W. Yang, “Modified single-blow technique for performance evaluation on heat transfer
surfaces”, Journal of Heat Transfer-Transactions of the Asme, vol. 97, 16 (1975).
157
EPOXY-BONDED La(Fe,Mn,Si)13Hz AS A MULTI LAYERED ACTIVE
MAGNETIC REGENERATOR
Henrique Neves Bez(a), Kristina Navickaitė (a), Tian Lei(a), Kurt Engelbrecht(a), Alexander
Barcza(b), Christian R. H. Bahl(a)*
(a)
Department of Energy Conversion and Storage – Technical University of Denmark, Frederiksborgvej 399,
Roskilde, Denmark
(b)
Vacuumschmelze GmbH & Co. KG, D-63450 Hanau, Germany
*
Corresponding author. E-mail: chrb@dtu.dk
ABSTRACT
The high magnetocaloric effect and tunability of the Curie temperature over a broad range makes
La(Fe,Mn,Si)13Hz a promising magnetocaloric material for applications. Due to a volume change across the
transition and the brittleness of the material as well as erosion due to fluid flow, the particles of the material
may break apart during operation. In this context, we studied epoxy-bonded La(Fe,Mn,Si)13Hz regenerators,
in a small versatile active magnetic regeneration (AMR) test device with a 1.1 T permanent magnet source.
The magnetocaloric material was in the form of packed irregular particles (250-500 µm), which were
mechanically held in place by an epoxy matrix connecting the particles, improving the mechanical integrity,
while allowing a continuous porosity for the fluid flow. Water with 2 wt% ENTEK FNE as anti-corrosion
additive was used as the heat transfer fluid for the epoxy-bonded regenerators. A series of AMRs was
evaluated by varying the epoxy content in the range 1-4 wt%.
Keywords: Magnetocaloric, refrigeration, epoxy-bonded, layered regenerator, AMR.

DOI: 10.18462/iir.thermag.2016.0147
1. INTRODUCTION
Since the discovery of a material with a giant magnetocaloric effect two decades ago[1] there has been a
desire to translate the very high entropy change observed into an increased performance in a magnetocaloric
device. Since the first room temperature device presented by Brown in 1976[2] most magnetocaloric devices
have used materials with a second order phase transition (SOPT). In these materials the transition between
the ferromagnetic and the paramagnetic phase results in a smooth entropy change across a broad temperature
range. Contrary to this, the discontinuous transition in materials with a first order phase transition (FOPT)
results in a very sharp and narrow entropy change. However, the peak values of the entropy change are in
general several times higher than those found in SOPT materials.
In order to improve performance, first order magnetic phase transition magnetocaloric materials may be used
in the AMR. One of the most promising FOPT materials’ series is La(Fe,Mn,Si)13Hz[3]. This material has the
Curie temperature, TC, near room temperature, while it can be finely tuned by the Fe/Mn/Si ratio. Still, using
FOPT materials instead of SOPT ones is not trivial and further considerations must be taken. Firstly, FOPT
materials exhibit hysteresis [4], although it has been shown that for the La(Fe,Mn,Si)13Hz series it is rather
small and in the range of 1 K [5]. Additionally, by definition FOPT materials exhibit a discontinuous phase
transition over temperature and field, which is the reason that they exhibit large magnetocaloric effect.
Nevertheless, the temperature range where there is a significant magnetocaloric effect is much narrower for
FOPT materials. For example, the full width at half maximum of the ∆s peak of Gd is about 30 K [6] for a
field change of 1 T, while for La(Fe,Mn,Si)13Hz it can be as low as 7 K [5]. This brings forth the necessity of
layering the regenerator with La(Fe,Mn,Si)13Hz with slightly different compositions along the direction of the
temperature span, in such a way that the working temperature of each layer is optimized. This engineering is
not straightforward and many complications have been shown elsewhere [7]. Furthermore, due to the
necessity of interstitial H in order to bring the TC to near room temperature without significant changes of the
magnetocaloric properties, the shapeability of the material is compromised, as the hydrogenation process
must be performed while the material is in the powder form and further sintering options remain limited [8].
Moreover, this material undergoes a volume change of ~0.9 vol% during the magnetic phase transition [8,9].
Given the brittleness of the material, microstructure and secondary phases, the volume change may lead to
158
cracking [10,11]. A method proposed to avoid cracking and erosion of the material under application is to
bond the particles of the magnetocaloric material with epoxy, increasing the mechanical stability of the
regenerator while still maintaining a continuous porosity [12].
In this context, we evaluate the performance of La(Fe,Mn,Si)13Hz epoxy-bonded double layer AMRs. The
factors investigated here are: the effect of epoxy content; the heat transfer fluid; the effect of layering; and
the utilization.
2. EXPERIMENTAL PROCEDURE
Fine particles of five slightly different compositions of La(Fe,Mn,Si)13H1.65 were used to make regenerators.
The materials were produced by the same processing route[3] and will in the following be identified by their
Curie temperature, measured by magnetometry.
Two types of regenerators were made: one with loose particles, held in place mechanically with a mesh at
either end and a second where the particles were bonded together by a small amount of epoxy. In both cases
the regenerators were cylindrical and made to fit into the versatile test device described previously[13,14].
The regenerator moves in and out of the 40 mm diameter bore of a 1.1 T permanent magnet Halbach
cylinder. The device is placed in a temperature controlled cabinet, and the hot end temperature, Thot, is fixed
by a heat exchanger to be slightly above the ambient temperature in the cabinet, Tamb.
A table listing the regenerators tested is given below. The particle sizes were in a range of 250-500 µm and
the regenerators had a porosity of 55%. The relatively high porosity is due to the relatively high irregularity
of the shape of the particles, leading to a low packing density. The epoxy bonded regenerators were all fixed
in 40 mm long pieces of Perspex tube, as shown in Fig. 1 all with masses of 95 g. The non-bonded
regenerators were held in a custom container allowing filling with different amounts of material.
Additionally, the viscosity and density of the transfer fluids were evaluated in an Anton Paar Lovis 2000
M/ME viscometer.
Name Description TC Mass

Single Single layer regenerator with no epoxy 23.3 ºC 48 g
Double Double layer regenerator with no epoxy 19.2 and 23.3 ºC 25 g each
Single4wt% Single layer regenerator with 4 wt.% epoxy 22 ºC 95 g
Double1wt% Double layer regenerator with 1 wt.% epoxy 23 and 26.6 ºC 95 g
Table 1: Description and properties of the regenerators for each given name.
Fig. 1 : Photo of an epoxy bonded regenerator, pen top for scale.
159
For the regenerator tests we used two different corrosion protection additives, which have previously been
shown to be effective for La(Fe,Mn,Si)13Hz type material, namely ethylene glycol based automotive
antifreeze (EG) and Entek FNE (NT) [15]. Viscosity and density of water with different amounts of these
additives were measured and compared to pure water, as shown in Fig. 2. The measurements were done at 20
ºC, which is close to the range the regenerators are tested in. Previous experience has shown that a
concentration of around 15% of EG is needed for good corrosion protection, while concentrations in the
order of 1-2% are sufficient when using NT. It is observed in Fig. 2 that the effect of the additive on the
density and viscosity is much less for NT than for EG, presumably due to the much smaller concentrations. It
can be noted that the change in viscosity per percent of additive is approximately the same while the effect
on density is higher per percent of EG than NT. Also, while there is a significant change in the absolute
viscosity of about 50% for EG and 20% for NT, the difference in density is only of the order of 3% for EG
and less than 1% for NT.
Fig. 2 : (a) The density and (b) the viscosity of the different fluids.
The apparent mechanical strength of the regenerators is generally good. However, during the regenerative
tests the Double1wt% regenerator with just 1 wt.% of epoxy could not withstand the forces during operation
and the regenerator broke apart.
We initiate the investigation of the regenerative performances by evaluating the influence of the utilization at
a Thot in the region of the maximum temperature span. The utilization, φ, is defined as
𝑚𝑐
𝜙 = 𝑚 f 𝑐f (1)
s s
where mf is the mass of the fluid pushed through the regenerator in one direction, cf is the specific heat of the
fluid, ms is the mass of the solid regenerator and cs is the specific heat of the regenerator material. The values
used for cs and cf are 480 [5] and 4200 Jkg-1K-1, respectively.
Since the utilization is linearly proportional to the blown fluid mass, mf, in Fig. 3 ∆Tspan is shown as a
function of the blown fluid mass. One can see that the trend of the span versus the mf is approximately the
same for the two regenerators without epoxy, independent of the number of layers and amount of material.
The epoxy bonded regenerators have the peak at the same blown fluid mass as the ones without epoxy.
Nonetheless, the mass of the solid regenerator is basically double of the ones without epoxy, and as we can
see in equation (1) the effect will be to decrease the utilization by half.
160
Fig. 3 : Temperature span as a function the blown fluid mass for three different regenerator at the region of maximum
temperature span.
The temperature span as a function of the hot end temperature is shown in Fig. 4, where the left graph shows
the results of the material without epoxy, while the right shows the results of the material with epoxy. It
should be noted that the results of the regenerators without epoxy were done with water and 15 vol.% EG as
the heat transfer fluid, while for the epoxy containing ones the fluid was changed to water with 2 vol.% NT.
As discussed above these additives do change the viscosity, but only have a weak impact on the density. Due
to the low frequency (0.13 Hz) the viscous dissipation is small when pushing the fluid through the
regenerator, so even for the EG this will not have a large influence on the results. The regenerators were
operated with the best resulting mf, 5.4 g, which due to the difference of mass in the regenerators lead to
different utilizations: 0.5 for the epoxy-bonded regenerators while those without epoxy were operated at a
utilization of 0.9, in accordance with the results found in Fig. 3.
A feature common in all the results is a steady increase of temperature span with increasing Thot up to a
maximum at a temperature about 3-5 K above the highest TC, then the span sharply decreases at higher hot
end temperatures[7]. Additionally, it is observed that increasing the number of layers, from one to two,
significantly increases the span, as expected. The measured temperature spans in the regenerators were not
affected much by the epoxy-bonding. In fact they increased a little, as the dashed lines in Fig. 4 suggest.
This, however, could be a combined effect of the different heat transfer fluids, differences in the mass of
magnetocaloric materials and effects related to the epoxy itself. Additionally, the TC spacing of the double
layer regenerator without epoxy and the ones epoxy-bonded is slightly different. As it is shown in the
literature[7], the temperature span can be heavily influenced by this spacing.
Although good performance was observed in the regenerators without epoxy, these regenerators were not
mechanically stable. After three weeks of measurements, the regenerators started to release some very fine
particles to the heat transfer fluid. These fine particles are believed to be the corners of the irregular particles,
which are more susceptible to cracking. We believe that erosion during the fluid blow periods, coupled with
the ~0.9% volume change that the material experiences at the transition [9, 11] result in breaking apart of the
particles. Additionally, once this breaking apart starts, the packing of the regenerators is lost, and the
particles are free to move inside the regenerator housing leading to further degradation of the regenerators.
Indeed, this is what was observed. Shortly after the first fine particles were observed the number of particles
in the fluid increased until the whole regenerators broke apart.
The reason for introducing the epoxy into the regenerator matrix was to enhance the mechanical stability and
prevent breaking apart of the regenerators. As expected, the mechanical stability of the regenerators was
increased significantly by epoxy bonding the particles. No degradation related to particles breaking was
observed, even after 5 weeks of testing. By comparing the temperature span of regenerators Double2wt% and
Double4wt% in Fig. 4 it is observed that increasing the amount of epoxy was detrimental to the span. So
when epoxy bonding it is important to find the minimum amount of epoxy that can be used while still
161
maintaining the structural integrity of the regenerator. As discussed above the regenerator Double1wt%
broke apart during initial testing, so the minimum amount of epoxy to avoid this seems to be 2 vol.%.
Fig. 4 : Temperature span as a function of the hot end temperature for different regenerators. The Dashed line is a guide
to the eye.
4. CONCLUSIONS
The performance of La(Fe,Mn,Si)13Hz as an active magnetic regenerator is evaluated in a custom-built

device. The results show that irregular particles of La(Fe,Mn,Si)13Hz deteriorate during application which
would limit the use of this material as an AMR. By epoxy bonding the particles, the mechanical stability of
the regenerators increased significantly and no deterioration due to particles breaking was observed. The
temperature span reached by the epoxy-bonded regenerators suggest that 2 wt.% of epoxy maximizes the
span, while still maintaining the mechanical integrity. The characterization of the transfer fluid showed that
the Entek solutions have a lower viscosity and density then the ethylene glycol ones. Although the effect of a
lower viscosity does not influence of the tests here done since the frequency is relatively low, this might play
a very important role under high frequency operations.
ACKNOWLEDGMENTS
This work was financed by the ENOVHEAT project which is funded by Innovation Fund Denmark (contract
no 12-132673). The authors are grateful to Vacuumschmelze GmbH for providing the samples.
REFERENCES
[1] V. K. Pecharsky and K. A. Gschneidner, Jr., “Giant Magnetocaloric Effect in Gd5Si2Ge2,” Phys. Rev.
Lett., vol. 78, no. 23, pp. 4494–4497, 1997.
[2] G. V. Brown, “Magnetic heat pumping near room temperature,” J. Appl. Phys., vol. 47, no. 8, p.
3673, 1976.
[3] A. Barcza, M. Katter, V. Zellmann, S. Russek, S. Jacobs, C. Zimm, “Stability and Magnetocaloric
Properties of Sintered La(Fe, Mn, Si)13Hz Alloys,” IEEE Trans. Magn., vol. 47, no. 10, pp. 3391–
3394, 2011.
[4] L. von Moos, C. R. H. Bahl, K. K. Nielsen, and K. Engelbrecht, “The influence of hysteresis on the
determination of the magnetocaloric effect in Gd5Si2Ge2,” J. Phys. D. Appl. Phys., vol. 48, no. 2, p.
025005, 2015.
[5] V. Basso, M. Küpferling, C. Curcio, C. Bennati, A. Barzca, M. Katter, M. Bratko, E. Lovell, J.
Turcaud, and L. F. Cohen, “Specific heat and entropy change at the first order phase transition of
La(Fe-Mn-Si)13-H compounds,” J. Appl. Phys., vol. 118, no. 5, p. 053907, 2015.
[6] K. K. Nielsen, H. N. Bez, L. von Moos, R. Bjørk, D. Eriksen, and C. R. H. Bahl, “Direct
measurements of the magnetic entropy change,” Rev. Sci. Instrum., vol. 86, no. 10, p. 103903, 2015.
[7] T. Lei, K. K. Nielsen, K. Engelbrecht, C. R. H. Bahl, H. Neves Bez, and C. T. Veje, “Sensitivity
study of multi-layer active magnetic regenerators using first order magnetocaloric material
162
La(Fe,Mn,Si)13Hy,” J. Appl. Phys., vol. 118, no. 1, p. 014903, 2015.
[8] S. Fujieda, a. Fujita, K. Fukamichi, Y. Yamazaki, and Y. Iijima, “Giant isotropic magnetostriction of
itinerant-electron metamagnetic La(Fe0.88Si0.12)13Hy compounds,” Appl. Phys. Lett., vol. 79, no. 5, p.
653, 2001.
[9] H. N. Bez, K. K. Nielsen, P. Norby, A. Smith, and C. R. H. Bahl, “Magneto-elastic coupling in
La(Fe, Mn, Si)13Hy within the Bean-Rodbell model,” AIP Adv., vol. 6, no. 5, p. 056217, 2016.
[10] J. Lyubina, R. Schäfer, N. Martin, L. Schultz, and O. Gutfleisch, “Novel design of La(Fe,Si)13 alloys
towards high magnetic refrigeration performance.,” Adv. Mater., vol. 22, no. 33, pp. 3735–9, 2010.
[11] A. Waske, L. Giebeler, B. Weise, A. Funk, M. Hinterstein, M. Herklotz, K. Skokov, S. Fähler, O.
Gutfleisch, and J. Eckert, “Asymmetric first-order transition and interlocked particle state in
magnetocaloric La(Fe,Si)13,” Phys. status solidi - Rapid Res. Lett., vol. 9, no. 2, pp. 136–140, 2015.
[12] J. Lanzarini, T. Barriere, M. Sahli, J. C. Gelin, A. Dubrez, C. Mayer, M. Pierronnet, and P. Vikner,
“Thermoplastic filled with magnetocaloric powder,” Mater. Des., vol. 87, pp. 1022–1029, 2015.
[13] K. Engelbrecht, K. K. Nielsen, and N. Pryds, “An experimental study of passive regenerator
geometries,” Int. J. Refrig., vol. 34, pp. 1817–1822, 2011.
[14] C. R. H. Bahl, T. F. Petersen, N. Pryds, and a Smith, “A versatile magnetic refrigeration test
device.,” Rev. Sci. Instrum., vol. 79, no. 9, p. 093906, 2008.
[15] D. Velázquez, M. Castro, E. Palacios, and R. Burriel, “Pratical Properties of LaFeCoSi Materials to
Be Used in Magnetic Cooling Applications,” in Proceedings of Thermag VI, 2014, pp. 133–134.
163
DESIGN AND FABRICATING OF A TWO-IN-ONE COMPACT MAGNET
ARRAY
Q. Jiang(1), H.B.Li(1), X.N.Xu(1),J.H.Huang(2),D.W.Lu*(1)
Physics Department of Nanjing University,(22 HanKou Road ,210093 ,Nanjing,P.R.China)
(1)
(2)Baotou research institute of rare earths(Baotau,014030, Inner Mongolia,P.R.China)
*Corresponding Author. E-mail:mrlab@nju.edu.cn
ABSTRACT
Usually two Halbach[1] magnet arrays are required in order to make a force balance for a whole rotary
magnetic refrigerator system. Except for a large volume of magnet array, a coupling gear is needed for the
two magnet arrays. It is found that a new magnet array may avoid the above disadvantages. The magnet
arrays are divided into inner two-hole Halbach-like magnet array and outer Halbach magnet array. The outer
magnet array may rotate around the inner magnet array. The inner Halbach-like magnet array contains two
rectangular holes parallel to cylindrical axis. The magnetic fields in the holes of inner magnet array are
uniform and equal with opposite directions. The field in two holes alternately varies with the rotation of
outer magnet array. It is shown that the field may be fairly uniform by calculation. The real two-in-one
magnet array is made successfully and the measuring data of magnetic field support the calculation results.
Keywords: Halbach magnet array, rotary magnetic refrigerator system, two-in-one magnet array, two parallel
rectangular holes
DOI: 10.18462/iir.thermag.2016.0149
1. INTRODUCTION
Halbach magnet array is usually used to obtain a high magnetic field for magnetic refrigeration. A rotatable
bi-layer Halbach magnet array[2] is suitable for a rotary magnetic refrigerator. The bi-layer structure contains
both inner and outer cylindrical magnet arrays (Halbach magnet array) : the outer magnet array may rotate
while the inner magnet array is fixed, and there is a cylindrical working space near the center where the
magnetic field is uniform around the axis. Considering the interaction between the inner and outer magnet
array two sets of bi-layer magnet systems are necessary for the torque balance while the outer array rotates.
The two sets of magnet arrays keep synchronous linkage through gear meshing and their magnet field’s
phases differ by semi-cycle, that is, when the two outer magnets are driven to rotate synchronously , the
strength of the magnetic fields in the two working space will change periodically and their phase will differ
by 180 °[3] .Based on a magnetic refrigeration system requires the two sets of well-coupling magnet array,
we propose a new design setting two sets of magnet array into one special set of magnet array. The array
contains two spaces where the magnetic field varies inversely when the outer Halbach magnet array rotates.
2. MAGNETIC CIRCUIT STRUCTURE AND DESCRIPTION

The basic structure and magnetic induction field of a bi-layer Halbach magnet array is shown in Fig. 1a. The
arrow represents the direction of magnetic induction in the cavum or the direction of magnetic remanence in
Fig. 1a.From left to right: the structure and magnetic induction Fig. 1b. The variation of the assemblage magnetic
vector field of the outer, inner and the assemblage magnet field, while t=0 represents the magnetic field of
array of a simplified bi-layer Halbach magnet. outer and inner array are parallel.
the magnet array. When the outer Halbach magnet array rotates around the inner magnet array the magnet
induction variation in the space with the time is shown in Fig. 1b. We may integrate the two sets of bi-layer
Halbach magnet array as long as the magnetic induction in the two holes keeps inverse change with the outer
164
magnet rotating. We simply use the outer magnet array as a rotary part, thus we mainly focus on the inner
magnet system. Since the outer magnet array rotates around the inner magnet array, they both need to take a
cylindrical shape. The two parallel cavum in the cylinder can take different shapes, in this article we only
discuss the situation of rectangular cavum. After setting several main magnet arrays right along the magnetic
induction of the holes the other arrays play an auxiliary role of homogenization. Roughly speaking, the
magnetic circuit is similar to the Halbach magnet array at the cross section of two holes.
3.SIMULATION AND RESULTS
We change the inner cylinder from a single cavity to double holes, and it will have two spaces where the
magnet fields are uniform and point oppositely. In order to reduce the volume of the system and avoid
deforming the cylinder we may get the rough structure of a compact two-in-one magnet array system. The
final design after multiple calculation and modification is shown in Fig. 2a, where the arrows represent the
direction of magnetic remanence, the slight color difference in magnet array represents different magnet
properties, and the jagged edges are for assembling convenience. Given a magnet system with structure and
size, the magnetic field can be calculated in details. It’s convenient to use special software Ansys-Maxwell of
magnetic calculation. As a simple approximation, we take the system as two dimensions. The calculation are
carried out according to the structure in Fig. 2a with measuring scale, and the arrow represents the direction
of magnetic remanence. The simulation of this magnet model is introduced as follows.
3.1 Simulation Settings
The design of the inner bi-hole magnet array is shown in Fig. 2b. In order to get a convergence result, a soft
magnet layer of 10mm is required to warp the cylinder of inner magnet array. We applied NdFeB magnet as
permanent magnet material. The two types are N45 and SH45 in Fig. 2b. As for the soft magnet material, we
use electrical pure iron DT4’s magnetization curve as its magnetic properties. The outer magnet array is a
common Halbach magnet.
Fig. 2a. Structure,size and magnetic remanence vector Fig. 2b. Different magnetic materials used in the
field direction of the inner bi-hole magnet design. manufacture of the inner magnet array.
3.2 Simulation Results
Firstly we calculate the distribution of magnetic induction in the space and compare both horizontally and
vertically. The results are shown in Fig. 3a and Fig. 3b. Figure 3a shows magnetic induction vs. vertical
coordinates in different x-axis position. As vividly depicted, besides some partial areas, most of the magnetic
induction in the space are about 0.6T. Figure 3b shows magnetic induction vs. horizontal coordinates in
different y-axis position within the hole, and the distribution is mostly around 0.6T. This result proves good
uniformity of the magnetic induction in working space. In order to study the magnetic induction in detail,
we adjust the measurement range according to the calculation result in Fig. 3a and Fig. 3b. Figure 4a and Fig.
4b show the simulation of magnetic induction after adjusting the measurement range. As can be seen from
the figures, stronger magnetic field exists in the left corner areas, and the area of low magnetic induction
near the center is still quite large. Adding the volume of the center magnet may amend the magnetic
induction distribution but it leads to smaller working space. By comparing the gain and loss, we decided not
to increase central magnet’s volume to maintain a bigger working area.
In addition to the magnetic induction’s value, its direction should also be considered. The external magnetic
field applied by outer magnet array is homogeneous, in order to make the magnetic field in the cavity be
superposed to a weaker strength, the magnetic field should be along the Y-axis. Fig. 5a is the simulation of
165
Fig. 3a. The comparison of the magnetic induction’s Fig. 3b. The comparison of the magnetic induction’s
strength along Y-axis in the inner magnet’s cavum with strength along X-axis in the inner magnet’s cavum with
different x coordinates. The unit of magnetic induction’s different y coordinates. The unit of magnetic induction’s
strength is MT. strength is MT.
Fig. 4a. The distribution
of magnetic induction in
the cavum of inner
magnet array when
measurement range is
0.54T-0.61T. The color
legend represents
magnetic induction’s
strength.
Fig. 4b. The distribution

of magnetic induction in
the cavum of inner
magnet array when
measurement range is
0.57T-0.61T. The color
legend represents
strength.
magnetic induction’s direction. As can be seen, direction deviation from Y-axis only exists near the right
side of the cavum. Fig. 5b shows the value of the magnetic induction’s component along X-axis. This
component of the magnetic induction is very small. In most of the cavum, the magnetic induction component
is below 0.04 T. In general, the horizontal deviation of the magnetic induction in the cavity is quite small.
Fig. 5a. The direction of magnetic induction in the left Fig. 5b. The strength of the magnetic induction’s
vavum of the inner magnet array. component along x-axis in the left vavum .
Fig. 6 shows the simulation after adding the outer cylinder. Fig. 1a depicts the outer cylinder, and the outer
Halbach magnet array generates magnetic induction and it directs along the positive Y-axis. Fig. 6 shows the
superposition of field between the inner magnet array and outer Halbach magnet array in the working area.
166
The magnetic induction is in agreement with the assumptions within most of the internal working area except
the fluctuation of magnetic induction in the lesser marginal parts of the two cavities.
Fig. 6. The superposition
of the inner Halbach-like
magnet array and outer
Halbach magnet array in
the working area. The
outer magnet array’s
magnetic field is along
the positive Y-axis. The
color legend represents
strength.
4.THE REAL MAGNET

The magnet system shown in Fig. 2a assembling with outer magnet array is manufactured. The top view of
the magnet is shown in Fig. 7. Prime measurement by a magnetometer shows the maximum and minimum
magnetic induction along Y-axis are 0.087T and 1.094T, which are in agreement with the calculation.
Fig. 7. Top view of the magnet entity. Fig. 8. The distribution of magnetic induction in the cavum
for the normal magnet shape without jagged edges.
5.DISCUSSION
The ideal result according to calculation is shown in Fig. 8. The real design shown in Fig. 2a is based on this
diagram after minor mending for practical reasons.The surprising result is that there still is not a smooth-
going rotating for the magnet system. Since the inner system is a mirror image reverse symmetry, the fact of
unbalanced torque should be related with procedure dependence effect. The Halbach-like magnet array is
made of many small magnet units which will subject to the outer magnetic field. The anisotropy of
magnetization properties will break the symmetry and result in an unbalance torque. The torque depends on
the angle between the inner and outer magnet array in experiment. The intrinsic unbalance effect may be
weakened by adjusting the shape of magnetic induction curve shown in Fig. 1b. The method is: along a
certain circle, to cut the outer Halbach magnet array into bi-layer Halbach magnet arrays[4] which rotate
inversely. The tri-layer magnet arrays may make the system rotary more smoothly.
REFERENCES
[1]Klaus Halbach, “Design of permanent multipole magnets with oriented rare earth cobalt material”,
Nuclear Instruments and Methods, Volume 169, Issue 1, 1-10(1980).
[2]A. TURA, A. ROWE, “DESIGN AND TESTING OF A PERMANENT MAGNET MAGNETIC
REFRIGERATOR”, Proceedings of the Second IIF-IIR International Conference on Magnetic Refrigeration
at Room temperature(2007).
[3]Jin Peiyu et al., “Design of Double Cylinder Style Room-temperature Magnetic Refrigerator”,Chinese
Rare Earths, 35(5):21-24(2014).
[4]D.S. Arnold, et al, “Design improvements of a permanent magnet active magnetic refrigerator”,
International Journal of Refrigeration, Volume 37, 99 -105(2014).
167
A NEW CONCEPT REFRIGERATION MODEL BASED ON THERMAL SWITCH
AND NUMERICAL CALCULATION
H.B. Li, Q. Jiang, X.N.Xu, D.W. Lu*
Department of physics, Nanjing University, 22 Hankou Road ,210093 ,Nanjing,P.R.China
*Corresponding author. Email:mrlab@nju.edu.cn
ABSTRACT
A new concept electrocaloric effect(ECE) refrigeration model based on a certain thermal switch(TS) is
presented in this paper. In the model tabular ECE materials and tabular TS alternately arrange parallelly and
tightly, forming many units of the same structure. The TS has two states, on and off according to external
control orders. The electric field is applied in every other ECE material sheet and let every other TS sheet be
on within the fore semi-period at the same time, and alternately going adversely within the rear semi-period.
The thermal conductivity of the TS in the off-state should be much lower than that in the on-state. The
calculation shows that the cooling power with 20 K temperature span may be more than 1 kW per kg ECE
material, and unit area cooling power may be more than 8 KW/m2 when the sheet is 0.15mm thin. In
addition the COP may be up to 12 at most.
Keywords: thermal switch, electrocaloric effect refrigeration, numerical simulation.
DOI: 10.18462/iir.thermag2016.0150
1.INTRODUCTION
ECE refrigeration is a new kind of refrigeration mode based on ECE of ferroelectric materials[1]. The ECE of
conventional materials is too weak to use in real life[2]. However, recently the research group of Q.M.Zhang
from Pennsylvania State University of United States discovered that the ECE of P[VDF-TrFE] materials is
giant. When applying an electric field the adiabatic temperature change (ΔTad) of P[VDF-TrFE] may
approach 20K, it may be also up to 10K even for the ceramics materials[3-6]. This discovery sets the
foundation of practical ECE refrigeration. However, similar to room temperature magnetic refrigeration, the
adiabatic temperature drop of ECE materials is less than the real requirement, thus those regular refrigeration
methods like Carnot cycle cannot be utilized for refrigeration. There are two practical methods to use the
small temperature drops realizing large temperature span refrigeration: one is in series and the other is
regenerative. When the drop is large enough the cooling mode in series is of advantage and adversely the
regenerative mode is necessary. The typical regenerative methods refer to active magnetic regenerative
(AMR) and passive regenerative(Brown type), but it showed low refrigeration power and cooling
efficiency(COP) when applied to prototypes[7]. In fact, the ECE and magnetocaloric effect (MCE) occur
within the solid materials and usually fluid is chosen to carry the heat out and in. The heat exchange of fluid
acts poorly in both frequency and power, the Peltier effect method improves in frequency but still lacks in
COP[8].
2. STRUCTURE DESCRIPTION
The new kind of refrigeration model is based on TS.

The TS may be driven by mechanical system or
electronic system, the later being studied. The basic
TS and ECE materials should be shaped sheet in
the refrigerator for the sake of quick heat exchange.
A basic structure of ECE refrigerator consists of
ECE material sheets and TS sheets array as
ABABAB...ABABA, as showed in fig.1, where A
represents TS unit while B stands for ECE material
unit and the number of ECE material unit must be
even and the number of TS unit is one more than
that of ECE unit. Each TS has two faces and
Fig1. The four components of the structure of the whole connects to two ends of both the adjacent ECE
system when N equals 10.
Seventh IIF IIR International Conference on Magnetic Refrigeration at Room Temperature, Thermag VII
168
units respectively. The TS has two states, one is ON with high thermal conductivity and the other is OFF
with low thermal conductivity. The cooling cycle includes two procedures: in the first half cycle, applying
electric field to the ECE material units of odd numbers while removing electric field from units of even
numbers, put the odd TS units to ON and the even TS units to OFF resulting in the ECE-TS system
absorbing heat from the cold end and expelling heat to hot end. In the second half cycle, applying electric
field to the ECE material units of even numbers while removing electric field from units with odd numbers,
put the odd TS units to OFF and the even TS units to ON resulting in the ECE-TS system recovering the
inner temperature gradient field. And repeat the above cycle, the heat from the low temperature heat resource
will be absorbed and then expelled to the high temperature heat resource.
3.CALCULATION MODEL & RESULTS
Based on thermodynamic equations, a detailed numerical calculation is presented for the new concept
refrigeration mode as follows. The basic assumptions are: (I)the appliance and removal procedures of electric
field are adiabatic and instantaneous, (II)all the ECE materials properties keep constant in the working
temperature range, (III)all heat transfer through the TS and ECE material sheets takes place only in one
dimension, (IV)the surface heat transfer coefficientbetween the boundary TSes and hot/cold exchangers is
infinite.
The parameters in the calculations are listed below. The parameters of the TS are based on a new kind of
electronic driven thermal switch which is still under research. This kind of new thermal switch concludes
some material whose physical parameters are similar to ECE materials, thus the choice of parameters. As for
thermal conductivity of the TS, we gave a approximated assumption based on its components.
The constant temperature of the hot heat exchanger and cold heat exchanger are 300K and 280K, i.e. Tspan is
20K. For the ECE materials, we used two sets of physical properties parameters corresponding to ceramic
material and polymer material[9]. The TS’s parameters are well matched with ECE materials. The thickness
of ECE materials dF=0.15mm, the thickness of TS dS=0.30mm. For ECE ceramic materials, the density
ρF=6*106g/m3, the heat capacity CF=1J/(g*K), the thermal conductivity λF=10W/(K*m), the average ΔTad in
the working temperature range is 8K. For TS, the density ρS=1.2*106g/m3, the heat capacity CS= 1J/(g*K),
the thermal conductivity of state on λON=10W/(K*m) while for state off λOFF=1W/(K*m). For ECE polymer
materials, ρF=1.5*106g/m3, CF=1J/(g*K), λF=2W/(K*m), ΔTad in the working temperature range is 4K. For
TS, ρS=1*106g/m3, CS=2.1J/(g*K), λON=2W/(K*m) while λOFF=0.5W/(K*m). The characteriistic cycle period,
under comprehensive consideration of sufficient heat exchange and lower heat leakage may be derived as
Eq.1.
 C *d *d
  4* F F F S (1)

ON
Where ρF represents the density of ECE materials, CF represents the heat capacity of ECE materials, dF
represents the thickness of ECE materials, dS represents the thickness of TS, and λON represents the TS’s
thermal conductivity of state on.
The differential equations used in the calculation is Eq. 2.
T  2
 T (2)
 c
The boundary conditions is Eq. 3.
T1 s  T2 s
T1 T (3)
   2 2
1
n s n s
The performance of the refrigeration system depends on the number, size and properties of both ECE
material and TS units. To evaluate the performance of the refrigeration system, we introduce three indexes:
COP, unit area cooling power (UACP) and unit mass cooling power (UMCP).
When the system works stably, the temperature distributions of the ECE material unit are almost linear.
While only adjusting the number of ECE material unit N, the curves of all three indexes are single peak
despite of different position. For ceramics materials, the peaks position in fig.2 shows that the maximum
COP is 12.45 when N=20, the maximum UACP is 1.40*105 W/m2 when N=70, the maximum UMCP is
169
8.98W/g when N=10. The COP rises first then drop with the increase of number N. In order to explain such
variation, we presume that when the number N is low, the real Tspan is close to maximum Tspan which
increases with N, thus COP rises first. However, temperature gradient drops with the increase of N,
decreasing the rate of heat transfer, which causes the decrease of COP. As for the UACP, the increase of
number N increases the amount of ECE material per unit area, which causes the increase of UACP until N
reaches 70.
Fig.2. COP, UACP and UMCP corresponding curves of ceramics

material.
After that, the decrease of COP plays the main role, causing the decrease of UACP. The UMCP falls all the
time due to increase of the amount of ECE material.
Fig.3. COP, UACP and UMCP corresponding curves of polymer material.

The Tspan are 20K and 10K respectively , which makes the temperature of
cold junction 280K and 290K.
In the case of polymer materials, Tspan is taken as 10 and 20K respectively, the response curves are shown in
fig.3. It shows similar signal peak phenomenon. When Tspan equals 20K, the performance of the system is far
poorer than the situation when Tspan equals 10K. The main cause is that the limit temperature Tspan is very
close to 20K. (Tspan maximun is 16.13K, 23.80K and 31.22K when N equals 20, 30 and 40.) For the present
polymer material, the maximum Tspan is quite low due to lower ΔTad and λF when compared to ceramics
material, so the cycle will be long and the performance becomes bad. Even the performance is still not good
for Tspan=10K. A trial calculation shows that the performance of system obviously depends on how the
properties of ECE materials match that of the TS.
4. DISCUSSIONS
We propose a new concept refrigeration method for ECE materials and MCE materials and the new
refrigeration system is based on TS and may match a very high cycle frequency and is of large cooling
capacity and high COP. In physics, the cycle cannot be simply classified into refrigeration in series or
regenerative mode. In fact the present cycle using TS is something like a mode in series with N grades, but it
is also like AMR with solid as regenerative but heat exchange being limited between the adjacent sites.
Based on the method, a quantitative calculation shows that unit mass cooling power maximizes 8.99W/g
while COP is 11.58 and unit area cooling power is 87.3kW/m2 for ceramics material when N=10. When
170
N=20, COP maximizes 12.45 while unit area power is 118.7kW/m2 and unit mass power is 6.59W/g. When
N=70, unit area power maximizes 140.0kW/m2 while COP is 6.33 and unit mass power is 2.22W/g. For
polymer material, low heat conductivity limits performance. Roughly speaking, COP, UVCP and UACP are
all high with 20 grades of ceramic ECE sheets. The practical appliance should be determined by
consideration of all three indexes.
REFERENCE
[1]Lines M, Glass A. Principles and Applications of Ferroelectrics and Related Materials. Oxford: clarendon Press,
1977.
[2]A.S. Mischenko1, et al, Giant Electrocaloric Effect in Thin-Film PbZr0.95Ti0.05O3, Science, Vol. 311, Issue 5765,
1270-1271,2006.
[3]Bret.Neese, Baojin.Chu, Q.M.Zhang, et al, Large Electrocaloric Effect in ECE Polymers Near Room
Temperature, Science, 821-823, 2008.
[4]S.G.Lu, et al, Enhanced electrocaloric effect in ferroelectric poly(vinylidene-fluoride/trifluoroethylene)
55/45mol% copolymer at ferroelectric-paraelectric transition, Appl. Phys. Lett. 98, 122906 (2011).
[5]Haiming Gu, et al, A chip scale electrocaloric effect based cooling device, Appl. Phys. Lett. 102, 122904
(2013).
[6]Hui-Jian Ye, et al, Giant electrocaloric effect in BaZr0.2Ti0.8O3 thick film, Appl. Phys. Lett. 105, 152908 (2014).
[7]M.-A.Richard, A.M. Rowe, R. Chahine, Magnetic refrigeration: Single and multimaterial active magnetic
regenerator experiments, J. Appl. Phys. 95, 2146 (2004).
[8]P.W.Egolf, et al, High-frequency magnetocaloric modules with heat gates operating with the Peltier effect,
International Journal of Refrigeration, Volume 37, 2014.
[9]Defay E, et al. The electrocaloric efficiency of ceramic and polymer films. Adv. Mater. 25(24): 3337-3342,
2013.
171
IMPACT OF THE INTEGRATION WORK ON THE PERFORMANCE
MEASUREMENTS OF THE MAGNETOCALORIC COOLING SYSTEM
AND ITS TARGET REFRIGERATED CABINET
M. Hittinger(a)*, JB. Chaudron(a), L. Elouad(a), C. Muller(a), P. Haegel(a), C. Vasile(b)
(a)
Cooltech Applications, Impasse Antoine Imbs, 67810 Holtzheim, France
(b)
LGeCo, National Institute of Applied Sciences (INSA), 24 boulevard de la Victoire, 67084 Strasbourg
Cedex, France
*
Corresponding author. E-mail: m.hittinger@cooltech-applications.com
ABSTRACT
Cooltech Applications has completed the construction and tests of a new Scale-up rotary magnetic cooling
system. This specific prototype has been designed for research purposes. Its modular architecture has been
used to generate improvements on the results to be integrated into the more compact commercial devices.
This system uses NdFeB permanent magnets providing a peak field of 0.8 T in the air gap and 3.1 kg of
parallel plates of Gd alloys. On the technology roadmap focused on the integration inside a target
refrigerated cabinet, external liquid/air heat exchangers and airflow configurations have been identified as
key drivers. In this paper the performance of this latest interfaced system is reported for a liquid flow rate of
6 liters/min. Actual measurements demonstrate that it was achievable to reach the target specification of the
first application from industrial partners: to cool an air volume of 350 liters down to a temperature of +2°C.
Keywords: magnetic cooling, heat exchangers, integration, airflow.

DOI: 10.18462/iir.thermag.2016.0151
1. INTRODUCTION
The scientific community and market players are convinced that magnetocaloric cooling could be the best
technology for a green future. Nearly twenty years after the discovery of the giant magnetocaloric effect and
with the rising interest in this research field, functional devices meeting the requirements of a final
application have started appearing. They validate the potential of this energy-efficient magnetocaloric
cooling technology before a successful market penetration.
The task and engineering efforts are complex and take time to transfer a magnetocaloric system from
conditions of maximum temperature span with no thermal load towards integration inside a functional useful
end application with target specifications for the needed heat exchanges. All the irreversibilities in the
system need to be reduced to provide an efficient energy interface between the magnetic device and the cool
air stream inside the industrial refrigerated cabinet. Important items such as the design of external heat
exchangers (HEX), thermal insulation of the refrigerated cabinet and air stream distribution are elements that
also need attention.
Specifications, actual operating measurements with room temperature around +22°C and control strategies
on a final integration are detailed in this paper. The experimental measurements show the impact of different
integration configurations on the performance of the system, cold temperature and dynamics of a target
refrigerated cabinet keeping the same flow rates of liquid and air.
2. EXPERIMENTAL DEVICE AND RESULTS
2.1 System Description

The prototype described in this paper is a Scale-up rotary magnetic cooling system called “DPSI+”. This
specific machine DPSI+ has been designed as a modular architecture to explore different research
opportunities such as the layering of Tc for example. For this purpose the bulk volume reduction for
integration was not a target and it was more intended to be easily tuned to investigate different configurations
on an easy-to-access test rig.
The magnetocaloric system uses a rotating magnetic assembly made of NdFeB magnets. Based on an
extensive simulation campaign it has been designed while searching for an optimum on economic criteria for
capital cost [1]. The goal is to minimize the mass of magnets while ensuring a magnetic field high enough to
172
generate a sufficient magnetocaloric effect. This architecture with permanent magnets produces a peak field
of 0.8 T that is measured in the air in the middle of the air gap of 34 mm. Simulations indicate a magnetic
induction of 1.2 T in the solid core of the representative Active Magnetic Regenerator (AMR).
In the described configuration of the DPSI+, 3.1 kg of magnetocaloric alloys are used. Magnetocaloric
materials have been selected with 4 different Tc inside the layered regenerator bed to improve the
performance over the target temperature span of 25 K in exchange mode with the sources. The coldest Tc is
5°C with an alloy of Gadolinium and the warmest Tc is 20°C with Gadolinium. Cooltech is also working on
more powerful first-order transition magnetocaloric materials for future applications. However, at this stage,
Gadolinium alloys are less expensive and more convenient to master with a wider flexibility of operations.
Parallel plates of alloys have been selected with a porosity of 27% based on simulations made with the multi-
physics model developed by Cooltech. This AMR configuration with thin material blades and thin fluid
channels is a trade-off between high exchange coefficients, reduction of pressure drops, flow homogeneity
and industrial manufacturing [2]. Based on past experiences, packed spheres or porous matrix AMR are not
considered by Cooltech as a viable industrial path to explore. They present a lower mass of active
magnetocaloric material in the same volume. Moreover, in spite of its higher heat exchange coefficient, this
kind of AMR generates much higher pressure drops, additional heat, cooling power decrease and COP
reduction due to the increased pumping power.
The rotating magnetic assembly is set in motion with an external drive motor via a belt. The spatial
configuration of all magnetized beds has been optimized to reach a better magnetic force balancing, a steady
torque and a lower motor power consumption. The heat transfer fluid is a mixture of water and glycol. The
liquid flow in the thermal circuits is performed thanks to a circulation pump and the synchronization of the
cycle phases is done through solenoid valves.
Thermal insulation has been used on liquid pipes and at strategic locations on the devices to limit the heat
transfer from room temperature to the cold spots of the device. External liquid/air heat exchangers are used
to transfer energy between the cold side and the hot side using the AMR in the middle. The interfaced system
of the DPSI+ and the test refrigerated cabinet can handle liquid flow rates between 6 and 12 liters/min and
cool an air volume between 300 to 1000 liters down to a temperature of +2°C for fresh products. In this
paper the actual measurements report performances on the selected minimum possible parameters: a liquid
flow rate of 6 liters/min and a refrigerated cabinet of 350 liters, these conditions simulating the functioning
parameters of the smaller compact integrated commercial devices of Cooltech.
2.2 Measurement of Maximum Temperature Span with no Thermal Load

First the maximum temperature span of the device was measured with no thermal load. In this case the
external heat exchangers were not connected. The collected flow from the cold and hot outlets of the AMR
was sent back to the AMR via an insulated pipe. Figure 1 represents the liquid cold and hot temperatures at
the outlets of the AMR for a magnetocaloric frequency of 1.2 Hz.
Figure 1. Measurements of temperature evolution on the DPSI+ device at the hot outlet of the AMR (curve in red) and
at the cold outlet of the AMR (curve in blue), representing the maximum temperature span with no load.
173
This measurement was performed in the workshop at Cooltech with a room temperature of +22.5°C. The
coolest liquid temperature was -5.3°C, reached thanks to the lowest Tc on the cold side at +5°C. The cooling
dynamics of the liquid was fast, with a cooling rate of -8 K/min. The maximum temperature span with no
load reached with this specific device was 45.9 K. 90% of the cold span, for a cold liquid temperature of -
2.5°C, achieved in only 319 seconds.
This first test was relevant to validate the correct assembly of the device and check that the cascade of Tc
works in a cooperative way. The target specification for the needed heat exchanges is an air temperature in
the refrigerated cabinet of +2°C. It took several months of research efforts to understand and control the heat
exchanges in link with the magnetocaloric system capability.
2.3 Comparison of Two Types of Heat Exchangers on the Cooling Dynamics of the Target
Refrigerated Cabinet
The intention was not to characterize the isolated magnetocaloric cooling device with fixed return
temperatures back to the AMR; the control of this equilibrium permanent state is usually done with electrical
heaters [3]. The goal here was to demonstrate that once integrated inside a refrigerated cabinet the functional
device could actually cool down the air volume of a target application in real operating conditions over a
useful application temperature span. The collected liquid flow from the cold outlet of the AMR was sent into
the external cold liquid/air heat exchanger connected to the cabinet. More than maximum cooling power at
zero temperature span figures, the minimum cold air temperature reached inside the cabinet and the time to
attain it from room temperature are treated as Key Performance Indicators by Cooltech’s industrial partners.
The actual opening of doors of refrigerated cabinets makes important the fast cooling dynamics and
adaptation of the whole system. The experimental set-up is shown in a picture of the system in Figure 2.
Figure 2. Modular and easy-to-access test rig with the white 350-liter air refrigerated cabinet to be cooled down, the
pink insulated appendix containing the different versions of cold liquid/air heat exchangers to be tested and the
connected black magnetocaloric cooling device DPSI+.
The two references of tested external liquid/air heat exchangers are called “Baseline HEX” and “Adapted
HEX for magnetic cooling”. In these two types of heat exchangers, the liquid from the AMR stays liquid and
there is no phase change like there is in HEX from compressor systems (neither evaporator nor condenser).
On the first integrations in 2013 industrial partners provided the external liquid/air heat exchangers that were
dedicated to compressor systems. These units were not suitable at that time for the magnetocaloric devices
due to oversized internal fluid volumes, heat exchange surfaces and mass. They use copper tubes and
aluminum fins, are retested here and called “Baseline HEX”. Since then Cooltech has worked to develop
custom heat exchangers specifically for magnetocaloric cooling devices. These “Adapted HEX” have been
designed as modular cores to adapt the heat exchange surfaces. The idea was to target compact heat
exchangers with thin solid walls and mini or micro liquid channels. They should present a low exchanged
liquid ΔT that is lower than the magnetocaloric effect and a reduced temperature difference between liquid
and air (pinch of the heat exchanger). The developed product with brass tubes and 25 μm copper fins exhibits
174
high exchange coefficients and a great density of heat exchange surfaces in a limited volume (convenient for
integrations). This approach to decrease all the unnecessary ΔT on the heat chain also aims to reduce the
exergy losses in the system. The specifications of these two references are compared in Figure 3.
Figure 3. Design and specifications comparison of the two references of tested external liquid/air heat exchangers:
“Baseline HEX” on the left and “Adapted HEX for magnetic cooling” on the right.
The test procedure outlined here was to change the liquid/air heat exchangers only on the cold side of the
integration (in the pink insulated appendix shown in Figure 2) and to compare the impact of the two
references on the cold air temperature and the dynamics of the target refrigerated cabinet. The experimental
measurements were compared using the same flow rates of liquid (6 liters/min) and air (varying with control
of fan speed). The two references have totally different cross sections for air flow: 5.2 dm² vs 2.2 dm². To get
a homogeneous air flow rate over the whole exchange surface 4 fans were used to blow on the “Baseline
HEX” and 2 fans were used to blow on the “Adapted HEX”. An air speed sensor was set in the middle of the
suction pipe with a diameter of 120 mm to make sure that the same air flow rate was sent on the heat
exchanger surface. The air speed was tested between 2 and 10 m/s. This measurement provided the precise
maximum air speed in the middle of the suction pipe; the calculation of the resulting air flow rate is only an
approximation and just provides a rough idea for comparison (zero speed on the pipe walls, integration
needed on the air speed profile). The air temperature was measured 15 cm downstream of the HEX. At tested
air flow rates the air temperature was homogeneous inside the refrigerated cabinet. The mean difference
average between the upper part and the lower part was 0.5 K and the air temperature stratification was
measured at a maximum of 1.8 K for the minimum air flow rate. A large amount of test data was generated.
Only the most interesting runs are presented in this paper.
Figure 4. On the left: Measurements of temperature evolution at the outlets of the AMR on the DPSI+ device (the
coolest temperature out of the 2 sets of curves) and in the air inside the refrigerated cabinet with the same air speed of
5.4 m/s. Comparison of two references of liquid/air heat exchangers: “Baseline HEX” and “Adapted HEX”; On the
right: Impact of air speed (and air flow rate) of the minimum cold air temperature reached inside the cabinet and the
cooling dynamics of the whole system. Comparison of two references of liquid/air heat exchangers: “Baseline HEX”
and “Adapted HEX”.
175
The results are encouraging. In 2013 the “Baseline HEX” did not fit at all with magnetic cooling systems at
that time. Compared to that period the generations of devices have since been improved in multiple ways.
For example, while maintaining the same target temperature span, the DPSI+ device generates 2.7 times
more cold flow rate per circuit compared to the 2013 units. Combined with this new DPSI+ machine the
“Baseline HEX” meets the requirements of the target refrigerated cabinet thanks to its large exchange
surfaces. However the results are better with the “Adapted HEX” and major aspects should be noted. In the
post-treatment of the results, the different number of fans (4 or 2) running at different speeds to provide the
same resulting air flow rate has been considered. Electrical measurements were performed and the difference
and contribution of the additional heat from the 2 additional fans was found to be minor: only 5 W. As
shown in Figure 3 the bulk volume of the “Adapted HEX” is much more compact: this point is crucial in the
integration work where space is often limited and usually adapted to compressor systems. In spite of lower
heat exchange surfaces (both liquid and air), the measurements show that “Adapted HEX” provide better
performance results. The reduced volume of internal liquid and solid mass of these “Adapted HEX” units
allows a low thermal inertia and fast response to actual fluctuating operational environment such as the
opening of doors or additional product loads. It exhibits a faster cooling rate of air in K/min better than 30%
up to 53% depending on the conditions. These “Adapted HEX” present great potential of heat exchange in a
limited volume with a high effectiveness. It can be explained by their high exchange coefficients thanks to
their thinner fluid channels (both liquid and air). It also allows a low temperature difference between liquid
and air (limited on these units to 2 K). This permits us to limit the temperature span that the AMR should
generate with regeneration giving more opportunity to use this energy for cooling capacity. Future
developments for heat exchangers will be focused on further optimizations of these “Adapted HEX” and they
will be adjusted to handle even more powerful magnetic units.
2.4 Impact of the Air Flow Configuration on the Cool-down Time

Airflow configuration has also been identified as a key item on the technology roadmap focused on the
integration inside a target refrigerated cabinet. Multiple set-ups have been tested. The stake is to limit the
temperature difference of contact between the liquid and the air to minimize the exergy losses while ensuring
sufficient heat transfer rate by design. Here a specific test was performed to demonstrate the importance of
this consideration when this phenomenon is maximum for low air flow rate. With the “Adapted HEX” the
performance measurements were carried out with and without the suction pipe. Even if the air temperature
was quite homogeneous inside the refrigerated cabinet some air stratification still existed. The purpose of the
suction pipe is to suck up the coolest air at the bottom of the cabinet to blow it onto the cold heat exchanger.
Figure 5. Benefit of the suction pipe on the heat exchange conditions and the cool down time of air inside the
refrigerated cabinet with the “Adapted HEX” for an air speed of 2.2 m/s. The curves in blue are the configuration
without the suction pipe and in green with the suction pipe. The two sets of curves in the same color are measurements
of temperature evolution at the outlets of the AMR on the DPSI+ device (the coolest temperature) and in the air inside
the refrigerated cabinet.
176
From these measurements it can be noticed that the cold temperature reached for air is cooler in the end with
the suction pipe (in green): +2.4°C compared to +3.3°C. This appendix adjusts the inlet air temperature for
the heat exchange during the cooling transient phase for lower ΔT and better effectiveness of the heat
exchanger.
2.5 First Elements of Control Strategy

Based on the previous results from part 2.3 it is clear that the air speed and air flow rate blown onto the heat
exchanger also has an impact on the heat transfer rate and then on the return temperature back to the AMR.
This latest effect influences the regeneration process necessary to maintain the temperature span and the
capability of the AMR to cool down the source changing its operating point on the temperature span-cooling
power map. Several tests were performed to adapt the heat transfer rate with different air flow rates via fan
speeds according to the period in the transient phase. Additional dynamic tests were carried out to simulate
the operation of a practical system with fluctuating conditions such as the opening of doors. An example of
control strategy is shown in Figure 6.
Figure 6. Measurements of temperature evolution at the liquid outlet of the AMR on the DPSI+ device (in blue) and in
the air inside the refrigerated cabinet (in green) with an “Adapted HEX” and control of different air flow rates.
3. CONCLUSIONS
Cooltech has tested a new Scale-up rotary magnetic cooling system dedicated to research studies. The
measured results have met the requirements of a target refrigerated cabinet with a volume of air of 350 liters
that needed to be cooled down to a temperature of +2°C. External liquid/air heat exchangers and airflow
configurations have appeared to be key items to consider in the integration work. All the lessons learnt have
been implemented in the more compact industrial devices from Cooltech. The energy consumption figures
(kWh/24h) are being analyzed on this version using higher efficiency electrical components. The improved
efficiency compared to the COP of Carnot will be communicated in a future publication.
REFERENCES
[1] A. Rowe, “Performance metrics for active magnetic refrigerators”, Proceeding of Thermag III, 3rd IIF-IIR
International Conference on Magnetic Refrigeration at Room Temperature, Des Moines, Iowa (2009)
[2] A. Kitanovski et al., Magnetocaloric Energy Conversion – From Theory to Applications, Springer (2015)
[3] S. Jacobs et al., “The performance of a large-scale rotary magnetic refrigerator”, Proceedings of Thermag
V, 5th IIF-IIR International Conference on Magnetic Refrigeration at Room Temperature, Grenoble, France
(2012)
177
THEORETICAL DESCRIPTION OF THE MAGNETOCALORIC
EFFECT IN Mn-Fe-P-Si ALLOYS
M. Piazzi(a)*, C. Bennati(a,b), C. Curcio(a,b), M. Kuepferling(a), V. Basso(a)

(a)
Istituto Nazionale di Ricerca Metrologica, Strada delle Cacce 91, 10135 Torino, Italy
(b)
DISAT, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
*
Corresponding author. E-mail: m.piazzi@inrim.it
ABSTRACT
We apply a thermodynamic model describing magnetoelastic effects and based on mean field theory to
understand the physical origin of the first order ferro- to paramagnetic magnetostructural transition occurring
in Mn1.3Fe0.65P0.5Si0.5. The phenomenological parameters of the model are determined through comparison
with the entropy data measured by Peltier calorimetry under magnetic field. The values obtained support the
idea that: the phase transition is driven by the strong magnetoelastic coupling present along the a lattice axis;
the spin entropy plays a key role in the transition; the structural contribution has a minor counteracting role.
Moreover, the number of magnetic moments derived from the model agrees with the picture describing the
paramagnetic phase of the system as due to the disordered magnetic moments at the Mn sites only.
Keywords: magnetoelastic coupling, mean field theory, Mn-Fe-P-Si alloys, first order phase transitions.
DOI: 10.18462/iir.thermag.2016.0157
1. INTRODUCTION
La(Fe-Si)13-based and (Mn-Fe)2(P-X) (X=As, Si, Ge) alloys are currently regarded as the most promising
classes of magnetocaloric (MCE) materials for magnetic refrigeration applications around room temperature.
The Mn1.3Fe0.65P 0.5Si0.5 system has an hexagonal crystal structure and it undergoes a ferro- (FM) to
paramagnetic (PM) first order transition, at Tt ~ 270 K, characterized by the fact that magnetic moments on
Mn sites remain substantially unaltered, while those on Fe sites drop significantly [1]. The system shows a
strong coupling between structural and magnetic degrees of freedom, highlighted by the abrupt anisotropic
change in the unit cell lattice parameters occurring at the transition. Indeed, the c/a ratio between the out-of-
plane (c) and the in-plane (a) lattice parameters increases, upon heating, of ~ 5% at Tt, because of the cell
expansion in the c direction and the simultaneous contraction in the a direction [1, 2].
In this work, in order to investigate more deeply the phase transition of Mn1.3Fe0.65P0.5Si0.5 compounds and to
enlighten the nature of the magnetostructural coupling, we apply the same phenomenological thermodynamic
model that we have already used for the study of La(Fe-Mn-Si)13-H1.65 alloys [3]. In Sec. 2 we describe the
thermodynamic model and in Sec. 3 we present the results of the comparison between the entropy behaviour
theoretically expected and experimentally measured. Finally, in Sec. 4 we discuss the results obtained,
presenting our conclusions and suggesting possible directions for future works.
2. THERMODYNAMIC MODEL
Let us consider a FM crystal lattice with magnetization M, expressed as mass density (i.e. in units Am2/kg),
along the same direction of the applied magnetic field H. The Landau free energy of the system, neglecting
any pressure effect, is given by:
G L (M , T , W ) = FM (M , T , W ) + FS (T , W ) - m 0 HM . (1)
In Eq. (1), FM, FS are the magnetic and lattice contributions to the free energy, m0 is the permeability of
vacuum, W = da/a + db/b + dc/c = Wa + Wb + Wc (Wi = di/i) is the relative lattice volume variation and a, b, c
are the in-plane and out-of-plane lattice parameters of the system, respectively. The dependence of FM on W
encompasses the coupling between the structural and magnetic degrees of freedom. A viable way to describe
this coupling can be developed assuming that its effects result not only as an isotropic change of the lattice
volume at the FM-PM transition, but as a relation between the molecular field coefficient W, describing the
exchange interactions in the mean field theory (MFT) approximation, and Wi (i = a, b, c):
W = W0(1 + baWa + bbWb + bcWc) (2)
178
where W0 > 0 is the FM exchange interaction strength in absence of magnetoelastic coupling, i.e. at ba = bb =
bc = 0. Eq. (2) allows to describe both isotropic and anisotropic magnetostructural transitions. Indeed, the
case ba = bb = bc, originally proposed by Bean and Rodbell [4], reduces to an isotropic volume change. The
general case ba ≠ bb ≠ bc, applying to the Mn1.3Fe0.65P0.5Si0.5 system, will be instead discussed in Sec. 4.
For the formal development of the theory we introduce now the reduced dimensionless variables
w(W) = W/W0 = 1 + bW (where W can be either the lattice volume change or the linear variation of the lattice
parameters), m = M/M0 (with –1 ≤ m ≤ 1 and M0 being the spontaneous magnetization of the system),
h = H/(W0M0) and t = T/T0. The FM-PM transition temperature at H = 0, b = 0 is T0 = a J m 0W0 M 02 (nk B ) ,
with aJ = (J + 1)/(3 J) and n, kB, J being the number of magnetic moments per unit mass, the Boltzmann
constant and the total angular momentum quantum number, respectively. Eq. (1), using the MFT
approximation, can be written then in dimensionless form as:
GL (M , T , W ) 1
g L (m, t , W ) = = - w(W )m 2 - a J ts M (m ) + f S (t , W ) - hm. (3)
m 0W0 M 0 2
2
In Eq. (3), sM = SM/(n kB) is the spin contribution to the entropy that, in the Brillouin theory, can be expressed
as:
é sinh ((2 J + 1)x (2 J )) ù
s M ( x ) = ln ê ú - xBJ ( x ) (4)
ë sinh ( x (2 J )) û
where x = BJ-1 (m ) and B J ( x ) is the Brillouin function. In the present study we will assume J = S = 1/2,
valid for magnetism due to the electrons. The structural part of the energy, f S (t , W) = FS (t , W ) m 0W0 M 02 , ( )
can be finally obtained as a series expansion around T0 [5]:
2
W2 a éa p v0 ù 1æ a ö
f S (t , W ) = f 0 + 0 W + S 0 ú (t - 1) - çç J ÷÷ bv m 0W0 M 02 (t - 1) .
v 2
- J ê (5)
2k T m 0W0 M 0 nk B ë k T
2
û 2 è Bø
nk
In Eq. (5), f0 is a constant term, kT is the isothermal compressibility, ap is the thermal expansion coefficient,
S0 is a dimensionful entropy reference value, bv is the entropy coefficient at constant volume.
The equation of state describing the magnetostructural behaviour of the system is obtained from Eq. (3) as
¶g L (W = W ) ¶m t = 0 Þ h = - m[1 + z (t - 1)] - hm 3 3 + a J tB J-1 (m ) (6)
where b W = z (t - 1) + hm 2 3 is the volume minimum and h = 3 / 2 (nk B k T ) (a J v0 )T0 , z = a p b T0 are
phenomenological parameters. As demonstrated in Ref. [4] and shown in Fig. 1a, by increasing the value of
h the phase transition changes from second to first order. The z parameter is instead related to the entropy of
(
the system. Indeed, the latter is given by s = -(1 a J )¶g L W = W ¶t = s M (m ) + s w ( m) + s S (t ) , where sM ) m
is defined in Eq. (4), sS is a pure structural contribution depending only on the temperature and due to
phonons and electrons, while sw = 1/(2 aJ)zm2 is the contribution coming from the magnetoelastic coupling.
The dependence of sw on z implies that the magnetoelastic entropy at a given t and m can either add (z < 0)
or subtract (z > 0) from the magnetic one depending on the sign of z. This fact is shown in Fig. 1b, where the
magnetic field induced entropy change –Ds = s(h = 0,t) – s(h,t) is reported for different values of z.
1
m z=0
h = 1.5 z = -0.5 h=2 z = 0.5 h=2
0.6 0.6
h = 1.2
h=1
h = 0.8
0.4 0.4
1st
h = 0.5
FM PM
-Ds
0.5
h=0
h
FM PM FM PM
0.2 0.2
2nd
h
h h h h
0 0
0.8 0.9 1 1.1 1.2 1.3 1.4 0.8 0.9 1 1.1 1.2 1.3 1.4
0
0.8 0.9 1 1.1 1.2 t t
t
(a) (b)
Figure 1. (a) Evolution of m vs. t behaviour for increasing h and J = 1/2, z = 0, h = 0. (b) - Ds = s(h = 0, t ) - s(h, t )
for negative and positive z and J = 1/2, h = 2, 0 ≤ h ≤ 0.05.
179
3. RESULTS
The MFT thermodynamic model described in Sec. 2 has three independent phenomenological parameters, n,
h, z, whose values can be determined through comparison with experimental data. As discussed in a
previous work (see Ref. [3]), these parameters affect the
m H=0 T
peak height, the shape and the height to width ratio of –Ds. 3000
0
m H=0.5 T
In particular, in this work we have applied the same 0
m H=1 T
approach followed in Ref. [3] to determine the parameters 0
m H=1.5 T
values for the Mn1.3Fe0.65P0.5Si0.5 alloys. The samples have
cp (Jkg-1K-1)
0
been prepared by conventional sintering, pressing the

2000
powders obtained by high-energy ball milling into pellets

and exposing the pellets to thermal annealing [6]. The
specific heat cp of the system has been experimentally 1000
measured with a Peltier calorimeter under magnetic field

and the curves obtained for the heating process are shown Mn1.3Fe0.65P0.5Si0.5 - exp
in Fig. 2. The sharp peak shown by cp at H = 0, close to 0
260 270 280 290
Tt ≈ 270 K, clearly indicates the first order nature of the T (K)
transition. By increasing H the peak moves towards higher Figure 2. cp vs. T upon heating measurements in
temperatures and smooths out, meaning that the transition Peltier calorimeter at different magnetic fields.
becomes more and more second order type.
The entropy S(H,T) and the magnetic field induced entropy change –DS(H,T) of the system, evaluated from
cp, are shown in Fig. 3a and Fig. 3b, top panels. By comparing these experimental data with the predictions
of our model, evaluated numerically for J = S = 1/2 (electron spin) and T0 = 272.6 K (experimental value of
the transition temperature at H = 0), we have found good qualitative and quantitative agreement among their
behaviours by choosing the following parameters values: nkB = 135 J/(kg K), h = 1.2 and z = 0.3. The
resulting theoretical curves for S(H,T) and –DS(H,T) are depicted in Fig. 3a and Fig. 3b, bottom panels.
4. DISCUSSION AND CONCLUSIONS
The parameters values found in Sec. 3 for the Mn1.3Fe0.65P 0.5Si0.5 system allow to derive the following
physical picture. The transition occurring in the compound is driven by magnetoelastic effects, responsible
both for the stabilization of low temperature FM and high temperature PM states, encompassed in the z < 1
value, and for the its first order nature, encompassed in the h > 1 condition. Moreover, the fact that z > 0
implies that the spin entropy plays a major role in driving the system through the transition, since the lattice
one has a small counteracting role. Finally, the number of spins per formula unit evaluated from the model is
nf.u. = nPM/NA ≈ 2.2 (NA and PM being the Avogadro’s number and the molecular weight of the system). The
Mn1.3Fe0.65P0.5Si0.5 compound is known to have ~2.5 spins/f.u. at 3g lattice sites occupied by Mn ions only,
and ~1.2 spins/f.u. at 3f lattice sites, occupied either by Fe or Mn ions [1]. Hence, the nf.u. value is much less
than ~3.7 spins/f.u., i.e. the sum of the spins at 3f and 3g sites contributing to the spontaneous magnetization
M0 of the system, but it is very close to the ~2.5 spins/f.u. present at the 3g sites. This means that our model
has correctly evaluated the magnetic entropy sM as due to the spins at 3g sites only, in agreement with the
explanation of the FM-PM transition given in Ref. [1], according to which at T > Tt magnetic moments at 3g
sites remain unaltered in amplitude but become disordered, while those at 3f sites drop to zero.
It is worth noting that similar conclusions were found in the case of the La(Fe-Mn-Si)13-H1.65 alloys [3].
However, the major difference with respect to that case is that the change in the lattice structure of
Mn1.3Fe0.65P0.5Si0.5 is anisotropic, while La(Fe-Mn-Si)13-H1.65 undergoes a first order transition accompanied
by an isotropic shrink of the unit cell volume W ~ –1% [3]. We can obtain anyway a meaningful picture for
Mn1.3Fe0.65P0.5Si0.5 by remembering that W = Wa + Wb + Wc (Sec. 1) and knowing that for this system W ≈ 0
and a = b: this way we easily have Wc = –2Wa. By considering now the temperature dependence of the lattice
parameters, see Ref. [2], we can notice that a shows an unusual contraction close to Tt. It is possible to
suppose that this contraction shall be ascribed to the magnetoelastic coupling in the same way the W shrink
in La(Fe-Mn-Si)13-H1.65 is. Then, the abrupt expansion along c at the transition, superimposing to the thermal
one, is simply a consequence of the a behaviour. Therefore, the relevant lattice parameter to be used in our
model in order to give a meaning to the whole physical picture here described shall be Wa instead of W and,
consequently, the magnetoelastic coupling parameter b should be also related to it. A rough estimate of ba
180
Mn1.3Fe0.65P0.5Si0.5 - exp 15 Mn1.3Fe0.65P0.5Si0.5 - exp m0H=0.5 T
0
m0H=1 T
S-SPM (Jkg-1K-1)
m0H=1.5 T
-DS (Jkg-1K-1)
10
-10
m0H=0 T
-20 m0H=0.5 T 5
m0H=1 T
m0H=1.5 T
-30 0
Mn1.3Fe0.65P0.5Si0.5 - theory 15 Mn1.3Fe0.65P0.5Si0.5 - theory

0 m0H=0.5 T
z = 0.3 z = 0.3 m0H=1 T
S-SPM (Jkg-1K-1)
h = 1.2 h = 1.2
-DS (Jkg-1K-1)
m0H=1.5 T
-10 10
m0H=0 T
-20 m0H=0.5 T 5
m0H=1 T
m0H=1.5 T
-30 0
260 270 280 290 260 270 280 290
T (K) T (K)
(a) (b)
Figure 3. Experimental (top) and theoretical (bottom) behaviour of: (a) entropy S-SPM; (b) entropy change –DS =
S(H=0,T)–S(H,T). For experimental curves, the function SPM = cPMln(T) is evaluated by using the specific heat constant
value cPM = 600 J/(kg K). Theoretical curves obtained for J = 1/2, T0 = 272.6 K, z = 0.3, h = 1.2, nkB = 135 J/(kg K).
can be obtained from the values of T0, z that we have found for Mn1.3Fe0.65P0.5Si0.5 and by assuming that
ba = bb, bc = 0 and that the linear thermal expansion coefficient along the a axis far from Tt is 1/3 of the
volume thermal expansion coefficient. Since the latter, extrapolated from experimental data, is
ap ≈ 3×10-5/K, we conclude that ba ≈ 55.
It is important to mention that the role of hysteresis, although present in the system, has been completely
neglected here. Indeed, hysteresis is a phenomenon mainly due to defects and models more sophisticated
than the one we have applied here are better suited to describe it [7]. However, the free energy of a system,
that can be easily evaluated in MFT through Eq. (1) providing the phenomenological parameters are known,
is a key quantity for all those models aiming to describe hysteresis effects. In this sense, the present study
may provide an interesting starting point for future investigations concerning the description of hysteresis in
Mn1.3Fe0.65P0.5Si0.5 compounds. Finally, further developments of the model may be envisaged also for what
concerns the study of magnetocaloric insulating systems in which magnetism arises from magnetic atoms
with integer or half-integer spins S > 1/2, as for example manganites.
REFERENCES
[1] N. H. Dung et al., “High/low-moment phase transition in hexagonal Mn-Fe-P-Si compounds”, Phys. Rev.
B 86, 045134 (2012).
[2] A. Pasko et al., “X-ray diffraction analysis of the magnetoelastic phase transition in the Mn-Fe-P-Si
magnetocaloric alloy”, AIP Adv. 6, 056204 (2016).
[3] M. Piazzi et al., “Modeling specific heat and entropy change in La(Fe-Mn-Si)13-H compounds”, J. Magn.
Magn. Mater. 400, 349 (2016).
[4] C. P. Bean and D. S. Rodbell, “Magnetic Disorder as a First-Order Phase Transformation”, Phys. Rev.
126, 104 (1962).
[5] V. Basso, “The magnetocaloric effect at the first-order magneto-elastic phase transition”, J. Phys.:
Condens. Matter 23, 226004 (2011).
[6] A. Bartok et al., “Study of the first paramagnetic to ferromagnetic transition in as prepared samples of
Mn-Fe-P-Si magnetocaloric compunds prepared by different synthesis routes”, J. Magn. Magn. Mater. 400,
333 (2016).
[7] V. Basso et al., “Modeling Hysteresis of First-Order Magneto-Structural Phase Transformations”, IEEE
Trans. Magn. 44, 3177 (2008).
181
MODELING OF A NOVEL DIGITAL HYDRAULIC SYSTEM FOR A
MAGNETIC REFRIGERATOR
P.O. Cardoso(a), M.C. Destro(a), M.G. Alvares(a), J.A. Lozano(a),*, V.J. De Negri(b),
J.R. Barbosa Jr. (a)
(a)
Florianópolis, SC, 88040-900, Brazil.
(b)
LASHIP — Laboratory of Hydraulic and Pneumatic Systems, Department of Mechanical Engineering,
Federal University of Santa Catarina, Florianópolis, SC, 88040-900, Brazil.
*Corresponding author. E-mail: jaime@polo.ufsc.br
ABSTRACT
In rotary magnetic refrigerators, the flow distribution system can be the major energy-consuming component
of the overall transmission system. Most magnetic refrigerators developed to this date employ rotary valves
(face-to-face sealing) for the fluid flow management. In this work, a mathematical model of a novel digital
hydraulic system for a magnetic refrigerator was developed. The modeling parameters of the flow
distribution system were based on the operating conditions of the device developed at POLO/UFSC. By
varying the input parameters for the solenoid valves and the system operating conditions, the dynamic
response of the hydraulic system was evaluated in terms of the absolute pressure and the volumetric fluid
flow rate at any point of the circuit. The energy consumption of the digital hydraulic system was also
determined. Based on the solenoid valves power consumption, a qualitative analysis was carried out to
compare the new solenoid valve system to that of a face-to-face flow management system.
Keywords: Magnetic refrigeration, digital hydraulic system, solenoid valve.
DOI: 10.18462/iir.thermag.2016.0158
1. INTRODUCTION
There are two main types of hydraulic systems: power systems and fluid transport systems. While the aim of
the former is to perform mechanical work, the latter are responsible for transferring fluid from one location
to another to achieve a certain practical purpose which, in the case of magnetic refrigeration, is to perform
the heat transfer to and from the magnetocaloric material. Thus, the flow management system in a magnetic
refrigerator is essential to the execution of the thermodynamic cooling cycle. In a magnetic cooling device,
the hydraulic system contains valves that are responsible for controling the direction and timing of the fluid
flow in the regenerators.
Because of friction and other effects, operation of hydraulic systems involves some degree of energy
dissipation. In fact, friction is the one of the dominant loss mechanisms in magnetic cooling systems, which
makes the hydraulic system responsible for a significant share of the energy losses. Eventually, a fraction of
the heat generated by friction is transferred to the fluid in the regenerator, thus reducing the efficiency of the
device [1]. Among the magnetic refrigeration prototypes developed so far [2], only few have done systematic
evaluations of losses considering those in the hydraulic system.
In magnetic refrigerators, a synchronization is required between the magnetic and the hydraulic systems.
Most prototypes reported in the literature employ rotary valves with a face-to-face sealing as flow
management systems, taking advantage of the torque transmitted to the magnet (or to the regenerator) to
drive the hydraulic system. Designers recognize the simplicity of face-to-face sealing, where the opening
time is controlled by oblongs and orifices. However, it has been demonstrated that the friction generated by
these types of seals [3,4] and the redundant constraints commonly found in these projects compromise the
overall performance of magnetic refrigerators. Eriksen et al. [5] obtained the best performance to date of a
magnetic refrigerator using a camshaft-driven valve system. However, the camshaft system impedes the
adjustment of different pressure drops through the regenerator beds, requiring manual equalization of the
pressure drop according to the regenerator with the lowest fluid flow rate.
182
Due to the aforementioned challenges, a novel digital hydraulic system based on solenoid valves is proposed
in this work as a flow management system for a magnetic refrigerator. This solution appears to have low
power consumption and is more versatile, being controlled electronically by software. Through the
mathematical model presented in this paper, a selection map for solenoid valves with different flow rate
coefficients has been developed as a function of the supply pressure of the system (volumetric flow rate).
This aids the designer to define the pressure drop given by each valve. Also, a comparison of the flow
management system developed by Lozano et al. [6] with face-to-face rotary valves, and that of a novel
digital hydraulic system has been evaluated the use of solenoid valves already available in the market.
A non-linear model of a hydraulic digital system for a simplified magnetic refrigerator was developed in this
work, as shown schematically in Fig. 1. The system is composed of the following components: a hydraulic
power unit (HPU), four digital (solenoid) valves (V), two regenerators (R), two heat exchangers (Z), two
flowmeters (S), and hoses (H). The model assumes the simplest configuration of an AMR with continuous
pumping and unidirectional fluid flow in the heat exchangers. The HPU is responsible for supplying and
receiving the fluid to and from the hydraulic system. It acts simultaneously as a pump and as a reservoir with
constant pump and discharge pressures, respectively. These are considered as the input parameters of this
model. The reservoir absolute pressure ( ) was assumed at atmospheric pressure (1.0 bar). The flow rate
provided by the HPU is independent of the internal system flow rate, since a directional valve, which acts as
a by-pass, controls it to maintain a constant supply (pump) absolute pressure ( ). The latter is an
idealization of the proposed layout by Ebel et al. [7] to avoid water hammers and smooth the flow.
1Z2 1S2
1R1 1R2
1H1 1H2
1V1 1V2 1V3 1V4
pS
1Z1 1S1
pT
HPU
Figure 1. Schematic representation of a digital electro-hydraulic circuit for a magnetic refrigeration system.
The porous regenerators (1R1 and 1R2) are modeled as packed-sphere beds and the pressure drop is
determined by the correlation proposed by Ergun [8]. The spool displacement of the solenoid valves ( ) is
function of the dynamic behavior of the electrical voltage ( ) at the solenoid of the valves, which can be
approximated by a RL series circuit [9] with a first-order behavior given by ⁄ . The
mathematical modeling of the directional valves is composed of two parts: a dynamic opening (or closing)
behavior described by a first-order equation as ⁄ , where is the steady state gain
and is the time constant. The volumetric flow rate ( ̇ ) through the valves is function of the valve flow rate
coefficient ( ), the valve opening ratio (where is the nominal voltage) and the valve pressure
drop ( ) and described as follows:
̇ ( )√ (1)
The flowmeters (1S1 and 1S2) and the heat exchangers (1Z1 and 1Z2) are simulated by the pressure drop
equation for laminar fluid flow. The hoses (1H1 and 1H2) are modeled by the continuity equation in the lines
183
between the valves and the regenerators, considering to have an internal area ( ) and a pressure loss
coefficient ( ). Here, an apparent fluid mass was implemented to simulate the inertial behavior of the fluid
flow by using the acceleration equation proposed in [10], which is function of the fluid density ( ), the hose
internal volume ( ), the hose flow rate ( ̇ ) and the hose pressure drop ( ), and given by:
̇
̇ (2)
3. RESULTS
The steady state values of the volumetric flow rate and the valve pressure drop from the dynamic model
simulations are shown in Figs. 2(a) and (b), respectively, for several combinations of valve flow rate
coefficient and supply pressure . As seen in Fig. 2(a), the volumetric flow rate has a larger dependence
on the supply pressure than in the valve . This is because in most of the analyzed cases, the valve pressure
drop (Fig. 2(b)) represents a small proportion of the system pressure drop, especially for
√ , since the valve pressure drop has an inversely proportional relation with the volumetric flow rate.
Thus, the higher the , the lower the valve pressure drop. In a first approach, these two maps can aid the
designer of a magnetic refrigerator to select solenoid valves using the valve for certain system ,
resulting in a volumetric flow rate through the AMRs. These plots were built by the simplified model
presented in Fig. 1, but they can be extended to relate the volumetric flow rate in the case of multiple
regenerator beds operating with parallel blows.
(a) (b)
Figure 2. Steady-state dependence of the (a) volumetric flow rate and (b) valve pressure drop from the dynamic model
simulations for several combinations of flow rate coefficient ( ) and supply pressure ( ).
As a case study, consider a desired ̇ 175 L/h at a pump 5 bar. In principle, it would be possible,
using Fig. 2(a), to select valves with between and √ . Each of those valves at the
previous desired operating conditions, would have, by looking at Fig. 2(b), a pressure drop of around 0.3 to
0.4 bar. On the other hand, by having a valve with √ and the desire volumetric flow rate
of 250 L/h, according to Fig. 2(a), the device would have to operate with a supply pressure of about 7.3 bar.
In this case, by checking in Fig. 2(b), the valve pressure drop would be approximately 0.34 bar.
The total power consumption of a solenoid valve system ( ̇ ), as proposed in this work, can be calculated
as ̇ ̇ ̇ , where ̇ is the nominal power consumption of the valve, ̇ is the power
consumption of the relay (drive system of the valve), is the number of valves (in this setup, twice the
number of regenerator beds), and is the duty cycle, i.e. the time percentage of the blow over the total AMR
cycle period. Also, the magnetic refrigerator developed by Lozano et al. [6] is considered, which is
composed of 8 regenerator pairs operated by a face-to-face rotary valve system that would be equivalent to a
digital hydraulic system of 16 solenoid valves, each with . Typical values of the power
̇
consumption ( ) of commercial solenoid valve candidates to substitute the rotary valves, which can
184
operate with water and be supplied by a 24 V DC source, is about 4 W with a settling time of approximately
50 ms. Taking into account the drive power consumption ( ̇ ) for an 8-relay system, for which each relay
actuates over two solenoid valves, the average power consumption of each relay is about 0.36 W. A
comparison between the power consumption of the rotary valves evaluated experimentally by Capovilla et
al. [4] and the proposed solenoid valve system is shown in Fig. 3. A relevant conclusion is that due to the
cyclic behavior of the system and the fast time response of the solenoid valves, the power consumption of the
solenoid valve system is independent of the operating AMR frequency, and this is case would be
approximately 16.7 W.
Figure 3. Comparison of the power consumption of the hydraulic system operated by rotary and solenoid valves.
4. CONCLUSIONS
A mathematical model of a novel digital hydraulic system has been developed in this work. Two designer
maps has been generated as a function of the volumetric flow rate and valve pressure drop for a simplified
magnetic refrigerator in which solenoid valves can be selected for a certain magnetic refrigerator. The power
consumption of a solenoid valve system that would substitute the rotary valve system of an actual magnetic
refrigerator has demonstrated a frequency-independent behavior, in addition to a lower power consumption.
Therefore, improvements in energy efficiency and operation versatility are expected through the use of the
proposed digital hydraulic system.
ACKNOWLEDGMENTS
REFERENCES
[1] I.V. Linsingen, Fundamentals of hydraulic systems, UFSC Publishing (2008).
[2] A. Kitanovski et al., Magnetocaloric Energy Conversion: From theory to applications, Springer (2015).
669 (2013).
[4] M.S. Capovilla et al., “Performance evaluation of a magnetic refrigeration system”, Science and
Technology for the Built Environment 22, 531 (2016).
[7] T.R.V. Ebel et al., “Simulation of a hydraulic circuit for a magnetic refrigerator”, Proceedings of 7th
International Conference on Magnetic Refrigeration at Room Temperature (2016).
[8] S. Ergun, “Fluid flow through packed columns” Chem. Process Eng. London 48, 89 (1952).
[9] H. Joo et al., “Dynamic Analysis of Linear Electromagnetic Solenoid for Electric Vehicle Relay”, XIX
International Conference on Electrical Machines, (2010).
[10] R. Szpak, “Análise teórico-experimental do comportamento das pressões em posicionadores
hidráulicos”, Master Thesis, Federal University of Santa Catarina (2008). Available:
http://www.tede.ufsc.br/teses/PEMC1105-D.pdf
185
THERMAL ENERGY HARVESTING: THERMOMAGNETIC VERSUS
THERMOELECTRIC GENERATOR
M. Almanza(a)*, A. Pasko(a), A. Bartok(a), F. Mazaleyrat(a), M. LoBue(a)
(a)
SATIE, ENS Cachan, CNRS, Université Paris-Saclay, 94235 Cachan, France
*Corresponding author. E-mail: morgan.almanza@ens-cachan.fr
ABSTRACT
We compare the efficiency and the power density of thermoelectric and thermomagnetic generators at
maximum power. The performances of thermomagnetic generator are computed using an equation of state,
either extrapolated from experimental data for 2nd order transition or deduced using a phenomenological
Landau model on measured data for 1st order transition. The performances of thermoelectric generator are
computed using the Onsager model. Moreover, the heat exchange in finite time is estimated using a simple
model of thermal conductance. According to the results, thermomagnetic generator is more efficient and have
slightly higher power density than thermoelectric for temperature difference lower than 10 K. Therefore low
grade heat thermal energy harvesting could consider thermomagnetic generator.
Keywords: magnetocaloric materials, thermal energy harvesting, thermomagnetic cycle, thermoelectric

generator, thermomagnetic generator
DOI: 10.18462/iir.thermag.2016.0160
1. INTRODUCTION
As indicated by the US department of energy [1], high and low grade waste heat are widespread, for example
in the USA between 20% and 50% of the global industrial energy inputs, corresponding to to
quadrillion /yr, is lost as waste heat. This makes thermal energy harvesting at different scales a key
research area to improve the sustainability of our electricity supply. So far thermal energy harvesting is
mainly oriented towards thermoelectric generation (TEG). However, recent advances on magnetocaloric
materials (MCM), aimed to applications to room temperature magnetic refrigeration, could pave the way for
a new generation of thermomagnetic generators (TMG) with high power density and better efficiency.
TMG is based on the magnetization change in MCM, induced by temperature (i.e. cycling between thermal
contact with a hot and cold reservoirs) and by successive application and removal of an external magnetic
field. Work can be produced in mechanical [2] or electrical [3] forms. Although TEGs are very simple thanks
to a direct production of electricity, TMGs can potentially attain better performances due to their higher
relative efficiency.
2. FINITE TIME THERMODYNAMICS APPLIED TO THERMOGENERATOR
One of the pillars of thermodynamics is the limitation in the conversion of heat into work imposed by the
second law. This maximum is reached with a fully reversible cycle – the Carnot cycle where thermodynamic
transformations are quasi-static. From the practical viewpoint, heat exchange in finite time and its
irreversibility has to be considered, as proposed in [4] where a heat exchange model is introduced. Whatever
the thermogenerator used, the main consequence is the introduction of a tradeoff between efficiency and
power [4]. Depending on the thermogenerator, different sources of irreversibility have to be considered: heat
transfer rate between the heat reservoir and the engine, the heat leaks from the hot to the cold reservoir and
the internal dissipation, only considered in TEG. They are modeled with thermal conductance . and , the
thermal leakage conductance and the dissipated heat , respectively, as shown in Figure 1
186
Figure 2 Examples of two ideal cycles: Ⓐ isotemperature-adiabatic, Ⓑ
Figure 1. Model of irreversibility
in thermogeneration. isotemperature-isofield, and a generic cycle Ⓒ.
In order to estimate the power, the heat exchange is modeled with and
, where and are the temperatures of two reservoirs, T1 and T2 are the
material temperatures when in contact with cold or hot reservoir respectively, and t is time. Different
configurations of thermogenerators can be assessed as a function of the operating points, the thermodynamic
cycles (Figure 2), the temperature difference between the hot and the cold reservoir ( ), the active
material and its volume, each one with different relative efficiency and power. Considering the relative
efficiency at the maximum power (EMP) [5] using realistic values of heat transfer coefficient, we compare
thermomagnetic generator and thermoelectric generator in terms of relative efficiency and power density for
different . The relative efficiency is the efficiency divided by the Carnot efficiency.
3. THERMOMAGNETIC GENERATOR
Thermomagnetic generator use the coupling between the entropy and the magnetization of MCM, given by
the entropy and the magnetization functions of temperature and magnetic field , with an
example of ideal equation of state in Figure 3.
Figure 3. From the left to the right: at different fields, at different temperatures, at different fields
derived from the equation of state [6] with an example of thermomagnetic cycle corresponding to cycle Ⓑ (Figure 2).
The thermomagnetic cycle Ⓑ drawn in Figure 2 and Figure 3 has four steps:
1- Isotemperature transformation where the MCM is in thermal contact with the cold reservoir and the
applied field increases in order to maintain the temperature constant;
2- Isofield transformation where the MCM is in thermal contact with the hot reservoir at constant field
;
3- Isotemperature transformation where the MCM is in thermal contact with the hot reservoir and the
applied field decreases in order to maintain the temperature constant;
4- Isofield transformation where the MCM is in thermal contact with the cold reservoir at constant field
.
In our study the MCM is at thermodynamic equilibrium and the model does not take into account intrinsic
irreversibility of MCM, namely both 2nd and 1st order transitions are assumed to take place at equilibrium
(i.e. entropy production associated with the transition is zero, ). From standard thermodynamic
relations applied to an elementary volume of MCM, always assumed to be in quasi-static process, we write:
and (1)
187
with T the temperature, the thermal capacity at constant field, the internal energy per unit of volume,
the heat exchange per unit volume and the magnetization. Because thermodynamic convention is used,
work produced by the system is counted as negative.
The TMG need thermal switch that allows the MCM to be in thermal contact alternatively with the hot and
the cold reservoir. We considered the conductances and controlled between two values and
, i.e when the MCM is in thermal contact with the hot reservoir, , , conversely
when in contact with the cold reservoir and , with >> . The power
density and the efficiency depend on , , as in magnetic refrigeration [7]. Because the
conductance depends on the geometry and the volume, we introduce conductance per unit of volume. The
constant is expected to assume values of about using thickness sheet of
MCM with heat transfer coefficient, as commonly used in magnetic refrigeration with
forced flow. The constant , associated with the heat leakage, is assumed null.
The entropy is divided in two parts: the lattice contribution with its thermal capacity constant,
and the magnetocaloric contribution which is central to the conversion. The lattice entropy does
not contribute to the conversion because it does not produce change in magnetization as shown by the
Maxwell relation. In the thermomagnetic cycle Ⓑ (Figure 2), it induces an extra heat exchanged with the
reservoirs, i.e heat leakage, mathematically given by the integration of between and . This term
is like putting a non-active material alternatively in contact with the hot and the cold reservoir. Because the
term is smaller than with 1st order MCM, the performance of TMG with 1st order Figure 5 is better
than with 2nd order Figure 4. Numerical simulation of TMG from [4] gives:
Figure 4. Maximum power and relative efficiency for Figure 5. Maximum power and relative efficiency for
different for cycle Ⓑ with and 2nd different for cycle Ⓑwith and 1st order
order and [7]. [6].
4. THERMOELECTRIC GENERATOR
Assuming the system at local equilibrium, because the system is in quasi-static process, and stationary state
we obtained from the relation of Onsager [8] the fluxes (Figure 1):
(2)
with the Seebeck coefficient , N the number of junctions, the internal

electrical resistance. The current is deduced from with the load
resistance connected to the thermoelectric generator. The use of Onsager model allows a description of the
internal dissipation, , and of the heat leakage between reservoirs, , and thus puts forward the
irreversibility due to thermal exchanges. For different , the length of the TEG is optimized, giving a
realistic estimate of the efficiency at the maximum power as proposed in [8], [9]:
188
and (3)
Here is the electrical resistivity ( from [8]); the length of the thermoelement; the length of
the insulating ceramic layer ; and where and are the electrical and
thermal contact resistivity ( and from [8], [9]). Factor of merit is then 0.8. Same as
for TMG, the addition of conductances based on [10] gives an optimum when
and . As expected, the performance is lower when the heat
exchange is taken into account.
Figure 6. Maximum power for different and relative Figure 7. Maximum power for different and relative
efficiency for TEG without the heat exchanger. efficiency for TEG with the heat exchanger (solid lines)
and TMG with 1st order MCM (dashed lines).
5. CONCLUSIONS
As shown in Figure 7, for large the power density is higher for TEG, however for small , lower
than 10 K, the power density is equal while the efficiency is always much higher for TMG with first order
phase transition MCM. Although the weight of the magnetic circuit and the efficiency of conversion of the
magnetic energy into electricity are not included, improvement on the shape of the thermodynamic cycle
could be expected. Moreover, further studies are needed to have better estimation of the coefficient used in
the heat exchange conductance.
REFERENCES
[1] US Department of Energy, “Waste Heat Recovery: Technology and Opportunities in U.S. Industry.” prepared by
BCS, Incorporated, Mar-2008.
[2] M. Ujihara, G. P. Carman, and D. G. Lee, “Thermal energy harvesting device using ferromagnetic materials,”
Appl. Phys. Lett., vol. 91, no. 9, p. 93508, Aug. 2007.
[3] T. Christiaanse and E. Brück, “Proof-of-Concept Static Thermomagnetic Generator Experimental Device,”
Metall. Mater. Trans. E, vol. 1, no. 1, pp. 36–40, Mar. 2014.
[4] M. Almanza, A. Pasko, F. Mazaleyrat, and M. LoBue, “Numerical study of thermomagnetic cycle,” J. Magn.
Magn. Mater., vol. submitted, 2016.
[5] F. L. Curzon, “Efficiency of a Carnot engine at maximum power output,” Am. J. Phys., vol. 43, no. 1, p. 22, 1975.
[6] A. Pasko, A. Bartok, K. Zehani, L. Bessais, F. Mazaleyrat, and M. LoBue, “X-ray diffraction analysis of the
magnetoelastic phase transition in the Mn-Fe-P-Si magnetocaloric alloy,” AIP Adv., vol. 6, no. 5, p. 56204, May
2016.
[7] M. Almanza, A. Kedous-Lebouc, J.-P. Yonnet, U. Legait, and J. Roudaut, “Magnetic refrigeration: recent
developments and alternative configurations,” Eur. Phys. J. Appl. Phys., vol. 71, no. 1, p. 10903, Jul. 2015.
[8] D. M. Rowe, Ed., CRC handbook of thermoelectrics. Boca Raton, FL: CRC Press, 1995.
[9] D. M. Rowe and G. Min, “Evaluation of thermoelectric modules for power generation,” J. Power Sources, vol.
73, no. 2, pp. 193–198, Jun. 1998.
[10] C. Goupil, H. Ouerdane, and Y. Apertet, “Thermoélectricité Thermodynamique et applications,” Tech. Ing.
Vecteurs Énergétiques Hors Électricité Convert., vol. TIB206DUO, no. be8080, 2013.
189
DYNAMIC TEMPERATURE MEASUREMENT FOR MAGNETOCALORIC
MATERIALS
M. Almanza(a)(b)*, W. Cherief(a), A. Kedous-Lebouc(a)

a
Univ. Grenoble Alpes, G2Elab, F-38000 Grenoble, France
CNRS, G2Elab, F-38000 Grenoble, France
b
Univ. Paris Saclay, SATIE, 94230 Cachan, France CNRS, SATIE, 94230 Cachan, France
*Corresponding author. E-mail: morgan.almanza@ens-cachan.fr
ABSTRACT
A device to measure the dynamic response of magnetocaloric effect in terms of temperature change at
different speed and up to has been developed. The magnetic source is based on rotating magnets
and the temperature measurement on thermocouple. This prototype will be able to explore the dynamic
response of magnetocaloric effect under controlled temperature for different type of material: first order and
second order phase transition material.
Keywords: dynamic response, magnetocaloric effect, temperature measurement, thermocouple, magnetic

source.
DOI: 10.18462/iir.thermag.2016.0161
1. INTRODUCTION
Assuming that heat exchange can be improved, increasing the operating frequency is a key solution to
achieve high cooling power density. Because at each cycle the heat is taken at the cold reservoir,
when the number of cycle per second, i.e. the frequency, increases, the cold power linearly increases with the
frequency [1]. Currently, almost devices run around few Hertz, some prototypes [2] were designed to go up
to 10 Hz but higher frequency could be investigated in the future and therefore a better knowledge of the
magnetocaloric material (MCM) behavior under dynamic excitation will be required.
At G2Elab, a dedicated characterization device is under development to measure the adiabatic temperature
change as a function of the field level , its rate of time variation and temperature . It will
be also used to study the time relaxation and hysteresis of the transition [3], [6] and to cycle the material in
order to study its ageing. It is designed to explore dynamics up to and cycling up to 300 Hz. In
this paper, we describe the system and preliminary measurements obtained for second order phase transition
(MCM). Section 2 describes the experimental set up used to produce a rotating magnetic field. The
methodology for temperature measurement and the first results obtained are provided in section 3. While 4
presents the conclusions and future works.
2. MAGNETIC SOURCE
The magnetic source is based on spinning magnets where a tunable air gap controls the applied field applied
from 0.2 T to 1.2 T, as shown in Figure 1 and Figure 2. The rotational angle defines the magnetic field, as
illustrated in the Figure 1, therefore by controlling the rotational speed from 0 to 10000 rpm the
magnetic field sweep rate is controlled from 0 to . Systems as [4], [5] used linear
configuration but the field rising time is smaller compared to the 1 reached by this device. However
recent work [6] using superconductors reaches for a field of . To minimize the effect of the
demagnetizing field the samples are elongated, being 10 mm long and 1 mm wide. In this study we
considerer a 2 mm thick Gadolinium sample. This leads to a demagnetizing factor of around 0.05. In this
study the fixed air gap is 15 mm and the applied field is 0.9 T.
190
Sample
Magnets
Shaft
Motor
Figure 1 Experimental setup and magnetic field level Figure 2 Magnetic field source with the
evaluated with simulation in the air gap for different magnets, the motor, the sample holder
configurations as a function of the rotation angle
3. TEMPERATURE MEASUREMENT
The main difficulty is to measure the temperature at high field sweep rates of up to 10 kHz and sufficiently
high precision of . Conventional approaches involve gluing or welding the Copper/Constantan junction
[2] or the thermistor [3] on to the sample (MCM). Here, this is accomplished using a thermocouple junction
that closes via its mechanical contact with the MCM sample itself (Figure 3, Figure 4).
Figure 3 Connection of the thermocouple: Left) with thermal

Figure 4 Thermocouple used.
grease, Right) mechanical contact.
Even though this method is sensitive to oxidation and the quality of mechanical contact, it has fulfilled the
bandwidth needed. Because the Seebeck measured voltage comes from the temperature at successive
junctions, here especially the copper/MCM and MCM/constantan junctions, the thermal exchange through
the glue is no longer needed, as illustrated on Figure 3 and Figure 4. An example of measurement is shown
in Figure 5. At low speed, we measure the voltage of the thermocouple and using the thermocouple relation
of , we obtained as expected 1.25 K in the Gadolinium (Figure 5).
The high speed temperature change involves an adiabatic transformation, because at 400 rpm a complete
magnetization needs of time length of respectively seconds (Figure 1), while the diffusivity of Gd and
Plexiglas, material of the sample holder are respectively around and . Accordingly in the
Figure 5, we have a small temperature change during the magnetization and the demagnetization states
because of a heat exchange.
However, this method has low sensitivity because of the small detected signal of , the background
noise of and the induced voltage dues to a strong time variation of the magnetic field. Despite
geometry efforts to minimize the loop generated by the wires, the induced voltage is around for a
rotational speed of 2000 rpm, as illustrated in Figure 7.
To overcome the undesired effect of the induced voltage in the thermocouple, clockwise (CLW) and
anticlockwise (ACLW) measurements are investigated, rotating the system at the exact speed but in the two
opposite ways, e.g. +400 rpm vs -400 rpm. In fact, in that cases, as measurements are perfectly synchronized
191
with the rotational angle, the detected signals haves symmetries. Figure 6 illustrates this clearly for 400 and
2000 rpm rotation speeds. So by averaging the CLW and ACLW signals, we can extract the time evolution
of the EMC. Figure 8 shows that results are independent of the magnetic field direction which is
characterizing the EMC property.
MCE
Average
Voltage
Figure 5 Example of measurement at low speed.
It is surprising that the values of the first measurement (Figure 5) at low speed are different from the ones at
higher speed (Figure 7). But those experiment have been made at different time, therefore it could come from
the sample. But others explanations could be appropriate as: the heterogeneity of the field inside the sample
for different frequency, different placement of the thermal thermocouple junction or different average
temperature of the sample.
Figure 6 Thermocouple voltage for different Figure 7 Thermocouple voltage function of the
rotational speed. position of the rotor for different speeds with Gd.
The control of the average temperature, currently done with air flow, will be improved by Peltier cell, as
shown in the Figure 8.
192
Figure 8 Synoptic of the system with the control of temperature.
4. CONCLUSIONS
A prototype to measure high dynamic response of the magnetocaloric temperature change is proposed. The
magnetic source uses rotating magnets while the temperature measurement uses open junction thermocouple.
Using superposition of clockwise and anticlockwise measurement, the effect of the induced voltage in the
thermocouple is removed. However, experimental investigations have shown dependence respect to the
speed of rotation, contrary to the results expected on Gadolinium at our frequencies. Although possible
explanations have been raised as the heterogeneity of the field at high frequency, further measurement using
different MCM and different shapes are needed to conclude.
ACKNOWLEDGMENTS
The authors are grateful to the French ANR-MagCool.
REFERENCES
[1] M. Almanza, A. Kedous-Lebouc, J.-P. Yonnet, U. Legait, et J. Roudaut, « Magnetic
refrigeration: recent developments and alternative configurations », Eur. Phys. J. Appl. Phys.,
vol. 71, no 1, p. 10903, juill. 2015.
[2] A. Kitanovski, J. Tušek, U. Tomc, U. Plaznik, M. Ožbolt, et A. Poredoš, Magnetocaloric
Energy Conversion. Cham: Springer International Publishing, 2015.
[3] J. D. Moore, K. Morrison, K. G. Sandeman, M. Katter, et L. F. Cohen, « Reducing extrinsic
hysteresis in first-order La(Fe,Co,Si)13 magnetocaloric systems », Appl. Phys. Lett., vol. 95, no
25, p. 252504, déc. 2009.
[4] G. Porcari, M. Buzzi, F. Cugini, R. Pellicelli, C. Pernec ele, . aron, . r ck, et M. Solzi,
« Direct magnetocaloric characterization and simulation of thermomagnetic cycles », Rev. Sci.
Instrum., vol. 84, no 7, p. 73907, 2013.
[5] H. Yibole, F. Guillou, L. Zhang, N. H. van Dijk, et E. Brück, « Direct measurement of the
magnetocaloric effect in MnFe(P, X )( X = As, Ge, Si) materials », J. Phys. Appl. Phys., vol. 47,
no 7, p. 75002, févr. 2014.
[6] T. Gottschall, K. P. Skokov, F. Scheibel, M. Acet, M. G. Zavareh, Y. Skourski, J. Wosnitza, M.
Farle, et O. Gutfleisch, « Dynamical Effects of the Martensitic Transition in Magnetocaloric
Heusler Alloys from Direct Δ T ad Measurements under Different Magnetic-Field-Sweep
Rates », Phys. Rev. Appl., vol. 5, no 2, févr. 2016.
193
PECULIARITIES OF THE MAGNETOCALORIC EFFECT IN THE
VICINITY OF THE FERROMAGNETIC-ANTIFERROMAGNETIC
TRANSITION IN AN IRON-RHODIUM ALLOY
R.R. Gimaev(a),(b)*, V.I. Zverev(a), A.M. Saletsky(a), A.M. Tishin(a),(b), T. Miyanaga(c), J. B.
Staunton(d)
(a)
Faculty of Physics, M.V. Lomonosov Moscow State University, 119991, Moscow,
Russia
(b)
Advanced Magnetic Technologies and Consulting LLC, 142190, Troitsk, Russia
(c)
Department of Mathematics and Physics, Hirosaki University, Hirosaki, Aomori, 036-8561, Japan
(d)
Department of Physics, University of Warwick, Coventry CV4 7AL, U.K.
*
Corresponding author. E-mail: gimaev@physics.msu.ru
ABSTRACT
We have investigated the large magnetocaloric effect (MCE) in a Fe50.4Rh49.6 alloy subjected to a slowly
cycled magnetic field of up to 1.8 T in magnitude over a range of temperatures, 250 K < T < 350 K. Our
measurements showed that the MCE associated with the first order ferromagnetic/antiferromagnetic phase
transition in this material was asymmetric with respect to whether the transition took place by heating the
material from temperatures below the transition or by cooling it from above. We have explained these
peculiarities using ab-initio density functional theory-based disordered local moment theory calculations.
Keywords: Magnetocalorics; magnetic refrigeration, iron-rhodium alloys.

DOI: 10.18462/iir.thermag.2016.0163
1. INTRODUCTION
Conventional nearly equiatomic, stoichiometric Fe-Rh alloys are unusual in that they undergo a first order
ferromagnetic (FM)-antiferromagnetic (AFM) transition close to room temperature [1–5]. These alloys
exhibit a giant magnetocaloric effect (MCE) which has led to a proposal [6] for their use in magnetic
refrigerators. Further discussion about the creation of a magnetic refrigerator which exploits such a 1st order
FM/AFM transition has also been published [6,7]. Another promising area of application for these alloys is
in heat-assisted magnetic recording (HAMR) [8–10] in high density magnetic storage technology. There is
also some scope for their use in medical applications [11]. In this paper we report on some subtleties of the
MCE associated with the first-order transition in this intriguing material.
2. THEORETICAL CONSIDERATION
We used the ab-initio disordered local moment theory [12,13] to describe the MCE properties of Fe-Rh
alloys as reported in [13,14]. In the calculations we specified a Fe-Rh alloy for a composition close to the
stoichiometric, perfectly ordered CsCl-type (B2) phase as Fe(100-x)Rh(x) – Rh(100-y)Fe(y) where both x
and y are small percentages. The proportion of iron in the material c = (100-x+y)/2 and the long-range order
parameter S = 1-y/100. For the Fe50.4Rh49.6 sample, which we studied experimentally, we used values c=50.4
and y=1.5 which were consistent with the experimental characterization data. For a specific composition, c,
y, the free energy of the alloy is given by the expression [13]
𝐹(𝑐, 𝑦, 𝐻, 𝑇) = 𝑈(𝑐, 𝑦, 𝑚𝑓 , 𝑚𝑎 ) − 𝑇 (𝑆𝑚𝑎𝑔 (𝑚𝑓 , 𝑚𝑎 , 𝑇) + 𝑆𝑙𝑎𝑡𝑡 (𝑇)) − 𝐻 𝑚𝑓 (1)
where U is the internal magnetic energy, Smag the magnetic entropy, Slatt the lattice vibration entropy [14]
and H the magnitude of applied magnetic field. mf is the ferromagnetic order parameter (proportional to the
overall magnetization) and ma the anti-ferromagnetic order parameter. For complete AFM order ma = 1, mf =
0, for complete FM order, ma = 0, mf = 1 and in the high temperature paramagnetic state both are zero. From
ab-initio calculations for several c and y values [13] we found an explicit expression for the internal energy,
U(c,y,mf,ma), in terms of mf and ma as reported in [14]. The values of the ferro- and antiferromagnetic order
194
parameters, mf and ma, were taken from where the free energy of Eq.(1) was minimized. We used this model
to analyse the experimental results reported in this paper.
3. EXPERIMENTAL TECHNIQUE
We performed direct magnetocaloric effect (MCE) measurements using a MagEq MMS 801 (AMT&C LLC)
system which has been described elsewhere [14–16]. As with all magnetic materials which undergo a 1st
order phase transition, Fe-Rh alloys demonstrate temperature hysteresis in their magnetization [3,17,18].
Consequently, for further insight into the magnetocaloric properties of Fe-Rh, direct measurements of the
magnetization of our Fe50.4Rh49.6 sample were performed by both heating and cooling through the phase
transition. On heating there was a ‘direct’ AFM-FM transition and on cooling there was a ‘reverse’ FM-AFM
transition. The AFM-FM transition was observed after having first slowly cooled the sample to 240 K. This
value was chosen after several preliminary measurements and was checked to be a temperature at which the
sample was fully in the AFM state. In a similar vein the sample was initially warmed to 350 K to be in the
fully FM state for measurements of the ‘reverse’ FM-AFM transition. All measurements of field induced
adiabatic temperature changes were performed uniformly with a 2 K/minute heating or cooling rate to
exclude significant temperature fluctuations and to start the measurements from as stable a state as possible.
Fig.1(a) shows the results of direct adiabatic temperature change, ∆T(T), measurements in the bulk
Fe50.4Rh49.6 sample taken on both routes through the phase transition. Those taken on heating the sample (red
open circles) have already been presented in Fig.3 of [14]. It is noticeable that the route through the phase
transition had a significant influence on both the temperature of the maximum MCE value and the MCE
value itself. In the present work we have defined the magnetic phase transition temperature as the
temperature at which the MCE reached its maximum value [19,20]. We found different phase transition
temperature values, Ttr, of 323.5 K and 319 K for the AFM-FM and FM-AFM transformations respectively.
The maximum MCE value, obtained by cooling from the FM state, was -5K, whereas the comparable value
obtained on heating from the AFM state, was -7.5 K. Also it is worth noting that the ∆T(T) associated with
‘reverse’ transition had significant values over a temperature range which had half the extent of that
associated with the ‘direct’ one. Evidently the hysteresis of the magnetocaloric properties accompanying the
1st order phase transition in the Fe-Rh sample appeared to be influenced directly by the overall
magnetization.
In our recent paper [14] the effect of irreversibility of the sample’s temperature to its initial value, when the
sample was subjected to a cycled external magnetic field at a sequence of increasing temperatures through
the AFM-FM transition, was shown (Fig. 3 in [14]). We explained this effect in the framework of the ab-
initio disordered local moment density functional theory (DLM-DFT) mentioned above [13]. Here we
present new results of ∆T(H) measurements for the FM-AFM transition taken while cooling the same sample
from the fully FM state. Figs. 1(b)-(d) show the measured ∆T(H) dependences: (c) at T tr (319 K) where the
MCE reached its maximum, (b) below Ttr (312.1 K) and (d) above Ttr (321.1 K), at which a MCE was still
observed. We present the results at two different temperatures in the vicinity of the phase transition in order
to reveal the comparable behavior of the MCE field dependences in both dominating AFM and FM phases
below and above the phase transition temperature, respectively. A similar hysteresis to what was observed on
heating through the AFM-FM transition was found as the field was varied (FWHM of about 1 T).
There was a noticeable absence of the ‘irreversibility’ effect of the type found in the earlier ∆T(H)
measurements which were obtained on heating the same sample from the AFM state [14]. In the case of
cooling from the FM state after a full magnetic field cycle we found the difference between the initial
temperature and that after a complete cycle to be zero, TFIN - TINIT = 0.
195
Fig. 1. (a) Experimental ΔT(T) values for Fe50.4Rh49.6 in a magnetic field 1.8T measured through the AFM-
FM transition on heating (red open circles) and through the FM-AFM transition on cooling (black filled
diamonds). Theoretical (blue dashed line) ΔT(T) values calculated for the system cooled through the FM-
AFM transition are also shown. The experimental ΔT(H) values for one full cycle of the magnetic field while
cooling the sample from the FM phase to specific temperatures are shown for (b) 312.1 K, (c) 318.7 K and
(d) 321.1 K. The theoretical ΔT(H) calculations for one full cycle of the magnetic field while cooling from
the FM phase to specific temperatures are shown for (e) 306 K, (f) 312 K and (g) 314. K. The black and red
arrows in (a) indicate the direction in which the temperature was changed at a rate of 1 T/s.
In order to model the irreversibility in the AFM/FM heating ∆T(H) measurements in [14] we had invoked a
simple mechanism. We assumed that for these measurements there was a weak, long-ranged, dynamic,
magnetic response effect in play with a slow time variation so that equilibrium thermodynamics could be
applied at every time step, ti. The extents of ferromagnetic or anti-ferromagnetic order, mf(ti) and ma(ti), were
averaged over compositional variations and ΔT for each value of the cycling magnetic field H(ti) was found.
This dynamic effect led to an extra effective magnetic field, Heff proportional to mf(ti-1), being added to the
external field H(ti). This extra field incorporated into our model successfully reproduced the irreversibility
feature. We have used the same model here to simulate the cooling results but with the one key difference
that for each temperature we constructed the extra effective magnetic field at the first time step, t 0, from
initial values of mf, mf(t0), taken from the FM state from where the sample was cooled (350K) instead of
mf(t0) = 0 (suited for the AFM state), which was the case for the heating calculations. In both sets of
calculations 20 time steps were allowed to ensue before the field-cycling was started.
The results are shown in Figs. 1(a), (e)-(g). Fig. 1(a) (blue dashed line) shows the theoretical ∆T(T) values
for cooling. ∆T(T) for heating was presented in [14]. As in experiment the non-symmetric character
stemming from the 1st order transition in Fe-Rh was evident and the cooling results had a smaller maximum
than that of the results obtained for the heating scenario. In line with experiment we found TFIN - TINIT = 0 for
this case. Figs. 1(e)-(g) show the cooling ∆T(H) values for one full cycle at three temperatures: (b) below
the temperature where ∆T(H) had its maximum value (306K), (c) at this point (312K) and (d) above it
(314.5K) where it is clear that TFIN - TINIT = 0.
5. CONCLUSIONS
Direct MCE temperature and field measurements in Fe50,4Rh49,6 alloy have been carried out and analysed in
terms of a theoretical model. We have found a heating/cooling asymmetry of its magnetocaloric properties
in the vicinity of the FM/AFM first order transition. This was demonstrated by the presence of an
‘irreversibility’ of the sample’s temperature to its initial value during a full cycle of the magnetic field on
heating which was absent in the analogous cooling measurement.
196
REFERENCES
[1] S.A. Nikitin, G. Myalikgulyev, A.M. Tishin, M.P. Annaorazov, K.A. Asatryan, A.L. Tyurin, The
magnetocaloric effect in Fe49Rh51 compound, Phys. Lett. A. 148 (1990) 363–366. doi:10.1016/0375-
9601(90)90819-A.
[2] M.P. Annaorazov, K.A. Asatryan, S.A. Nikitin, A.M. Tishin, A.L. Tyurin, Pisma Zh. Tekh. Phys. 12
(1991).
[3] M.P. Annaorazov, K.A. Asatryan, G. Myalikgulyev, S.A. Nikitin, A.M. Tishin, A.L. Tyurin, Alloys of
the Fe-Rh system as a new class of working material for magnetic refrigerators, Cryogenics. 32 (1992)
867–872. doi:10.1016/0011-2275(92)90352-B.
[4] M.P. Annaorazov, H.M. Güven, K. Bärner, COP of cooling cycles around the AF–F transition in FeRh
based on experimental data, J. Alloys Compd. 397 (2005) 26–30. doi:10.1016/j.jallcom.2005.01.016.
[5] J.-U. Thiele, T. Hauet, O. Hellwig, Design of Co∕Pd multilayer system with antiferromagnetic-to-
ferromagnetic phase transition, Appl. Phys. Lett. 92 (2008) 242502. doi:10.1063/1.2946654.
[6] M.P. Annaorazov, M. Ünal, S.A. Nikitin, A.L. Tyurin, K.A. Asatryan, Magnetocaloric heat-pump
cycles based on the AF–F transition in Fe–Rh alloys, J. Magn. Magn. Mater. 251 (2002) 61–73.
doi:10.1016/S0304-8853(02)00477-8.
[7] M.P. Annaorazov, S.A. Nikitin, A.L. Tyurin, S.A. Akopyan, R.W. Myndyev, Heat pump cycles based
on the AF–F transition in Fe–Rh alloys induced by tensile stress, Int. J. Refrig. 25 (2002) 1034–1042.
doi:10.1016/S0140-7007(02)00028-2.
[8] C.W. Barton, L. Saharan, G. Hrkac, T. Thomson, Effect of Fe under layer in ultrathin FeRh films, in:
2015 IEEE Magn. Conf. INTERMAG, 2015: pp. 1–1. doi:10.1109/INTMAG.2015.7157518.
[9] F. Garcia-Sanchez, O. Chubykalo-Fesenko, O.N. Mryasov, R.W. Chantrell, Multiscale models of hard-
soft composite media, J. Magn. Magn. Mater. 303 (2006) 282–286. doi:10.1016/j.jmmm.2006.01.135.
[10] T.J. Zhou, K. Cher, J.F. Hu, Z.M. Yuan, B. Liu, The concept and fabrication of exchange switchable
trilayer of FePt/FeRh/FeCo with reduced switching field, J. Appl. Phys. 111 (2012) 07C116.
doi:10.1063/1.3677838.
[11] A.M. Tishin, Y.I. Spichkin, V.I. Zverev, P.W. Egolf, A review and new perspectives for the
magnetocaloric effect: new materials and local heating and cooling inside the human body, Int. J.
Refrig. (2016). doi:10.1016/j.ijrefrig.2016.04.020.
[12] B.L. Gyorffy, A.J. Pindor, J. Staunton, G.M. Stocks, H. Winter, A first-principles theory of
ferromagnetic phase transitions in metals, J. Phys. F Met. Phys. 15 (1985) 1337. doi:10.1088/0305-
4608/15/6/018.
[13] J.B. Staunton, R. Banerjee, M. dos S. Dias, A. Deak, L. Szunyogh, Fluctuating local moments, itinerant
electrons, and the magnetocaloric effect: Compositional hypersensitivity of FeRh, Phys. Rev. B. 89
(2014) 54427. doi:10.1103/PhysRevB.89.054427.
[14] V.I. Zverev, A.M. Saletsky, R.R. Gimaev, A.M. Tishin, T. Miyanaga, J.B. Staunton, Influence of
structural defects on the magnetocaloric effect in the vicinity of the first order magnetic transition in
Fe50.4Rh49.6, Appl. Phys. Lett. 108 (2016) 192405. doi:10.1063/1.4949355.
[15] Y.I. Spichkin, A.M. Tishin, D.B. Kopeliovich, A.Y. Malyshev, Direct measurements of the
magnetocaloric effect: realization and results, in: Third IIF-IIR International Conference on Magnetic
Refrigeration at Room Temperature (Thermag III), Des Moines, Iowa, USA, 2009: pp. 175–180.
[16] Y.I. Spichkin, R.R. Gimaev, Experimental methods of the magnetocaloric effect studies, Int. J. Refrig.
37 (2014) 230–236. doi:10.1016/j.ijrefrig.2013.09.018.
[17] M.P. Annaorazov, S.A. Nikitin, A.L. Tyurin, K.A. Asatryan, A.K. Dovletov, Anomalously high
entropy change in FeRh alloy, J. Appl. Phys. 79 (1996) 1689–1695. doi:10.1063/1.360955.
[18] E. Kren, L. Pal, P. Szabo, Neutron diffraction investigation of the antiferromagnetic-ferromagnetic
transformation in the FeRh alloy, Phys. Lett. 9 (1964) 297–298. doi:10.1016/0031-9163(64)90369-5.
[19] V.K. Pecharsky, K.A. Gschneidner, A.O. Pecharsky, A.M. Tishin, Thermodynamics of the
magnetocaloric effect, Phys. Rev. B. 64 (2001) 144406. doi:10.1103/PhysRevB.64.144406.
[20] A.M. Tishin, K.A. Gschneidner, V.K. Pecharsky, Magnetocaloric effect and heat capacity in the phase-
transition region, Phys. Rev. B. 59 (1999) 503–511. doi:10.1103/PhysRevB.59.503.
197
ELECTROCALORIC POLYMERS: DIFFERENT ELECTRIC FIELD
CYCLES TO INCREASE THEIR EFFICIENCY
Sébastien Pruvost(a)*, Vittorio Basso(b), Lauric Garbuio(c), Afef Kedous-Lebouc(c),

Morgan Almanza(d) and Jean-François Gérard(a)
(a)
Univ. Lyon, INSA Lyon, UMR CNRS 5223, IMP Ingénierie des Matériaux Polymères,
F-69621 VILLEURBANNE, France
(b)
Istituto Nazionale di Ricerca Metrologica INRIM, Strada delle Cacce 91, 10135 Torino, Italy
(c)
G2Elab, Univ. Grenoble Alpes /CNRS, F-38000 Grenoble
(d)
Univ. Paris Saclay, SATIE, F-94230 Cachan, France
*Corresponding author. E-mail: sebastien.pruvost@insa-lyon.fr
ABSTRACT
The evaluation of electrocaloric properties of a material can be achieved by two methods: either by the
indirect method which uses Maxwell relations to link the electrocaloric effect to the polarization versus
temperature or by the direct measurement of the heat flux response to an electric field variation. In this paper
we show how the heat flux method is able to evaluate the efficiency of different types of electric field cycles
applied to a ferroelectric material with hysteresis. We present experiments on the polymer P(VDF-TrFE)
70/30 and discuss how to improve the cooling power in connection with respect to the non zero electrical
conductivity of the material.
Keywords: Electrocaloric effect, ferroelectric polymers, electric field cycles, cooling efficiency.
DOI: 10.18462/iir.thermag.2016.0173
1. INTRODUCTION
The electrocaloric effect corresponds to the adiabatic temperature change of a material or the isothermal
entropy change induced by application of an electric field. This effect is small in most dielectric materials.
Recently, ferroelectric ceramics [1] and fluorinated polymers [2] have been found to exhibit a giant
electrocaloric effect, then opening the door for applications in the field of refrigeration.
Many studies, dealing with the electrocaloric effect, calculate the entropy by integrating the Maxwell relation
of equilibrium thermodynamics ∂s/∂E =∂P/∂T using the polarization versus electric field cycles P(E)
measured as a function of temperature T. If this method is used when the system is out of thermodynamic
equilibrium, as for example when the relationship P(E) is hysteretic, it may provide unreliable results,
therefore alternative methods are required. Among the different methods that have been developed, many of
them directly evaluate the electrocaloric effect as the electric field induced temperature change under
adiabatic conditions. The temperature of the sample can be measured by thermistors [3], thermocouples [4]
or by infrared cameras [5, 6]. However, the adiabatic conditions are always problematic to realize and, as an
alternative, differential scanning calorimeter DSC under electric field have been successfully tested [7-9].
In particular the calorimetry with Peltier cells, originally developed to characterize magnetocaloric materials,
has been proved to be very effective for the characterization of the electrocaloric effect of dielectric
polymers [10,11]. Peltier cell calorimetry also allows to separate reversible contributions (electrocaloric
effect) and irreversible ones (hysteresis losses). As the efficiency of an electrocaloric material is the key
point in the future development of cooling systems [12], then the evaluation of the losses is a particularly
important aspect. In the present study we address the problem of finding appropriate electric field cycles able
to improve the performance of electrocaloric materials with hysteresis without increasing the maximum
applied electric field.
2. DIRECT MEASUREMENT OF THE ELECTROCALORIC EFFECT
Calorimetry with Peltier cells, that has been originally developed to characterize magnetocaloric materials,
has been recently adapted to characterize dielectric polymers [10]. This basic scheme of the measurement
system is reported in Fig. 1. The system works in isothermal conditions and permits to measure
198
simultaneously the electric current density passing through the sample and the heat flux, qs, exchanged by the
sample (S) with the thermal bath (B). The entropy change Δs is computed by subtracting the Joule heating
contribution, qJh, from the measured heat current, qs, and by making the integration
1
Δs =
vT
∫ (q s − qJh )dt (1)
where v is the sample volume. With the knowledge of the electrical conductivity σ, the Joule heating
contribution is evaluated as qJh = - v σ E2. The result of the measurement is the entropy change Δs of the
material upon a given variation€of the electric field. The nice feature of the system is the possibility to apply
it to ferroelectric materials with hysteresis, where the cyclic variation of the electric field between two
extrema, correspond to traversing a minor hysteresis loops of the material.
Figure 1. Scheme of the system measuring the electrocaloric effect. (A) Peltier cell controlling the temperature, (B)
copper block, (C) thermocouple, (D) Miniaturized Peltier heat flux sensors, (S) sample. The arrow shows the positive
direction of the heat flux qs.
3. ELECTRIC FIELD CYCLES
For this study, the P(VDF-TrFE) copolymer (70/30) was selected as a test ferroelectric material.
Measurements have been performed below its Curie temperature, where the materials display ferroelectric
hysteresis [11]. Measurements have been performed by starting from both a polarized and a non-polarized
initial state. We found that electric field cycles around the remanence of the polarized material represent an
advantage, because the corresponding entropy change has the same sign of the electric field variation in time
dE/dt, provided that the material will not depolarize. Fig. 2 shows the results of the measurement of the
entropy change corresponding to different electric field cycles performed around the remanent polarization
of the material using as negative extremum an electric field Eneg which is smaller, in absolute value, than the
coercivity. Experiments on the P(VDF-TrFE) copolymer (70/30) show that with Eneg > -40 MVm-1 the
depolarization of the material is avoided. For the maximum field there is also a limitation. The results of
Fig.2 show that, for an electric field variation |ΔE|= Epos - Eneg, larger than 125 MVm-1, quite large
differences, between the entropy changes of the two branches of the loop, are obtained. This is interpreted as
the result of the presence of hysteresis and therefore of hysteresis losses. For large fields, the hysteresis
losses become significant and detrimental for the efficiency of an electrocaloric cooling system. To exploit
the behavior at relatively small fields, one may then compare different electric field cycles with the same
variation ΔE. By comparing the behavior of an electric field variation of 80 MVm-1, one finds the two
corresponding cycles (-40/40) and (0/80) lead to the same entropy variation under cyclic behavior. However
the smaller is the absolute value of the applied electric field, the smaller will be the energy loss by the Joule
heating generated by the electric conduction of the material. Then for the same electric field variation ΔE,
the symmetric cycle is more efficient.
199
Figure 2. Top left. Scheme of a generic asymmetric P(E) cycle between electric fields Eneg and Epos, starting from the
positive remanence state. Right. Entropy variation measured on P(VDF-TrFE) (70/30) copolymer around its positive
remanent polarization for different asymmetric cycles. Results are plotted as a function of |ΔE|= Epos - Eneg, with (Eneg,
Epos): (1) (0,25); (2) (-25, 25); (3) (-40, 40); (4) (-25, 100); (5) (-40, 100); (6) (-40,150); (7) (-40, 200) (x 106Vm-1).
Bottom left. Example of different electric field cycles close to the remanence with the same |ΔE|: symmetric (-40,40)
and asymmetric (0,80).
4. CONCLUSIONS
We have employed a Peltier cell calorimeter to perform direct measurements of the electrocaloric effect. The
device allows to detect reversible contributions (electrocaloric effect) and irreversible ones (hysteresis
losses). The method was employed to test a ferroelectric copolymer P(VDF-TrFE) (70/30). Asymmetric
cycles close to a remanence of a ferroelectric material with hysteresis, appear to be valuable for improving
the performance of electrocaloric materials. We have tested different electric field cycles with the same
electric field variation. We found that symmetric cycles have the advantage to limit the impact of the Joule
heating due to the non zero electric conductivity. Therefore electric field cycles performed around a
remanence of the material can be efficiently used for eletrocaloric refrigeration, provided that the coercivity
is large enough to avoid the depolarization. This criterion can guide the design of electrocaloric cooling
systems working with ferroelectric material in the hysteretic state.
ACKNOWLEDGMENTS
The authors gratefully acknowledge INSA Lyon for giving a visiting professor fellowship to Vittorio Basso
and the Rhône-Alpes region for its financial support by ARC ENERGIE.
REFERENCES
[1] A.S. Mischenko, Q. Zhang, J.F. Scott, R.W. Whatmore, and N.D. Mathur, "Giant electrocaloric effect in
thin-film PbZr0.95Ti0.05O3", Science 311, 270 (2006)
[2] B. Neese, B. Chu, S.G. Lu, Y. Wang, E. Furman, and Q.M. Zhang, "Large electrocaloric effect in
ferroelectric polymers near room temperature", Science 321, 821 (2008)
[3] S.G. Lu, B. Rozic, Q.M. Zhang, Z. Kutnjak, R. Pirc, M. Lin, X. Li, and L. Gorny, "Comparison of
directly and indirectly measured electrocaloric effect in relaxor ferroelectric polymers", Appl. Phys. Lett. 97,
202901 (2010).
[4] S. Kar-Narayan and N.D. Mathur, "Direct and indirect electrocaloric measurements using multilayer
capacitors", J. Phys. D: Appl. Phys 43, 032002 (2010).
200
[5] S.G. Lu, B. Rozic, Q.M. Zhang, Z. Kutnjak, X. Li, E. Furman, L.J. Gorny, M. Lin, B. Malic, M. Kosec,
R. Blinc, and R. Pirc, "Organic and inorganic relaxor ferroelectrics with giant electrocaloric effect" Appl.
Phys. Lett. 97, 162904 (2010).
[6] G. Sebald, L. Seveyrat, J.-F. Capsal, P.-J. Cottinet, and D. Guyomar, "Differential scanning calorimeter
and infrared imaging for electrocaloric characterization of poly(vinylidene fluoride-trifluoroethylene-
chlorofluoroethylene) terpolymer", Appl. Phys. Lett. 101, 022907 (2012).
[7] G. Sebald, S. Pruvost, L. Seveyrat, L. Lebrun, D. Guyomar, and B. Guiffard, "Electrocaloric properties of
high dielectric constant ferroelectric ceramics", J. Eur. Ceram. Soc., 27, 4021 (2007).
[8] G. Sebald, L. Seveyrat, D. Guyomar, L. Lebrun, B. Guiffard, and S. Pruvost, "Electrocaloric and
pyroelectric properties of 0.75Pb(Mg1/3Nb2/3)O3-0.25PbTiO3 single crystals", J. Appl. Phys. 100, 124112
(2006).
[9] Y. Bai, G.-P. Zheng, and S.-Q. Shi, "Kinetic electrocaloric effect and giant net cooling of lead-free
ferroelectric refrigerants", J. Appl. Phys. 108, 104102 (2010).
[10] V. Basso, F. Russo, J.-F. Gerard, and S. Pruvost, "Direct measurement of the electrocaloric effect in
poly(vinylidene fluoride-trifluoroethylene-chlorotrifluoroethylene) terpolymer films", Appl. Phys. Lett. 103,
202904 (2013).
[11] V. Basso, J.-F. Gerard, and S. Pruvost, "Doubling the electrocaloric cooling of poled ferroelectric
materials by bipolar cycling", Appl. Phys. Lett., 105, 052907 (2014).
[12] E. Defay, S. Crossley, S. Kar-Narayan, X. Moya, and N.D. Mathur, "The electrocaloric efficiency of
ceramic and polymer films", Adv. Mater. 25, 3337 (2013).
201
THEORETICAL MODELING OF HEAT TRANSFER PROCESSES IN
Ni-Co-Mn-In MAGNETIC WIRES
Oksana Pavlukhina(a)*, Vladimir Sokolovskiy(a,b), Vasiliy Buchelnikov(a)
(a)
Chelyabinsk State University, Bratiev Kashirinykh st., 454001 Chelyabinsk, Russia
(b)
National University of Science and Technology “MIS&S”, Leninskiy prospect, 119049 Moscow, Russia
E-mail: pavluhinaoo@mail.ru
ABSTRACT
Evidently, the choice of geometry for a regenerator or magnetic cooling cell is an important
problem in the magnetic cooling technology. In our work, we study a heat transfer in the three-
dimensional magnetic cooling cell involving microwires of Ni45Co5Mn37In13 Heusler alloy. In the
present study of heat transfer processes in the system with microwires, we used the system of
differential equations including the Navier–Stokes and heat equations for fluid and solid, while the
incompressible fluid is passed. The relaxation times of the temperature of wires, the temperature
distribution in the magnetic wires and coolant in the cell were calculated. Our data clearly point to
the fact that the relaxation time increases practically linearly with the increasing microwire
diameter.
Keywords: Heusler alloys, magnetic wires, heat transfer, simulation.

DOI: 10.18462/iir.thermag.2016.0174
1. INTRODUCTION
At the present time, a large number of the intermetallic alloys and compounds including Gd–Si–Ge, Fe–Rh,
La–Fe– (Si,Al), Mn–Ga–(C,N), Mn–Fe–(P,As), and Ni–Mn–(Ga, In, Sn, Sb) are also promising candidates
in magnetic cooling technology [1–6]. Co -doped Heusler alloys Ni-Mn-Z (Z= Ga, In, Sn) have attracted a
lot of attention due to the observation of shape memory effect, the giant magnetoresistance, magnetocaloric
effect (MCE) etc. [5-9]. Thus, the Co addition can greatly change both the martensitic transformation
temperature and Curie temperatures of austenite and martensite as well as it can change magnetization curves
across the structural transition resulting in an enhancement of the inverse MCE. For instance, Liu et al. [1]
directly measured the adiabatic temperature change (ΔTad) across the magnetostructural transformation for
Ni1.8Co0.2Mn1.48In0.52 alloy upon variation of magnetic field from 0 to 1.9 T. The authors have found the
largest value of ΔTad (≈ - 6 K) obtained for Heusler alloys. Recent experimental investigations of magnetic
microwires have shown that the microwires of Ni–Mn-based Heusler alloys can be also promising candidates
for magnetic cooling applications near room temperature [5–9]. Thus, Zhukova et al. [11] fabricated Ni–Mn–
In and Co-doped Ni–Mn–In Heusler-type glass-coated microwires and demonstrated the temperature
dependences of magnetization of the as-prepared and annealed samples in small applied magnetic fields.
Evidently, the choice of geometry for a regenerator or magnetic cooling cell is an important problem in the
magnetic cooling technology. There have been reports about engineering design and numerical investigations
the active magnetic regenerators consisting of meshes of channels with square cross-sections, packed
particles and stacks of parallel plates of magnetocaloric materials [12-14]. Sarlah et al. [12] reported that the
parallel plates offer considerable promise for the regenerator design compared with other regenerator
geometries such as packed particles or spheres due to the lowest pressure drop as evaluation criteria. To our
knowledge, the theoretical investigation of thermophysical properties of Heusler Ni-Co-Mn-In microwires
has not been shown previously. Therefore, our interest in this paper was to study the distribution of a heat
transfer and thermophysical properties in Ni45Co5Mn37In13 Heusler microwires.
2. DETAILS OF MODELING
In order to realize a mathematical model of heat transfer, it is necessary to consider two mechanisms of heat
transfer involving the mechanism of heat conduction and convection as well as the movement of the coolant.
202
As the result, we need to solve the Navier–Stokes equations for an incompressible fluid, continuity equation,
and thermal conductivity.
(1)
Here, T = T (x, y, z) is the temperature at a given point in time t, Cp(l) and Cp(s) are the heat capacities of
liquid (l) and solid (s), kl and ks are the corresponding thermal conductivities, ρ is the density, η is the
viscosity, u = u(x, y, z, t) is the velocity vector of the fluid at the point (x, y, z) in space at time t. Molecular
adhesion forces always exist between the solid surface and any viscous liquid. These forces result in that the
solid wall adjacent to the liquid layer is delayed, like adhering thereto. Accordingly, there is a boundary
condition to the equations of a viscous liquid movement. The boundary condition is treatment demand zero
fluid velocity on fixed solid surfaces. The boundary conditions for the heat equation - heat flow coming from
one environment should be equal to the flow entering the second medium. The above system of equations
with regard to the boundary conditions was solved by using the finite element method. The magnetic cooling
cell is shown in Fig. 1. With respect to the geometry of a microwire, the wire thickness was varied from 10
to 50 µm, the wire length was fixed to 1 mm. Concerning the geometry of the magnetic cooling cell, the
height, width, and length of the cooling cell were taken as 0.2, 0.6, and 2.5 mm, respectively.
Figure 1. The three-dimensional model cooling cell involving five microwires.
It is significant that a choice of model parameters is the important initial step before the design of optimal
cooling cell in magnetic refrigeration devices. Table 1 shows the model parameters for both the
Ni45Co5Mn37In13 microwire and coolant (water). The parameters for the water and Ni45Co5Mn37In13 are taken
from Refs. [1, 8–11]. We chose the constant temperature defined as TC - ΔTad (314 K) of wires as the initial
temperature for both the coolant and microwires. On the left in the cell is take coolant flow at a constant rate.
Moreover, the coolant inlet velocity U (Fig. 2) was taken at a theoretical Curie temperature (320 K) remind
that all results of the heat transfer simulations are presented for two values of the coolant velocity, 0.7 and
2.5ms−1.
H2O Ni45Co5Mn37In13
density (с), kgm−3 1000 8020
heat capacity (Cp), J kgK−1 4183 1279
Curie temperature (TC), K 320
MCE (∆Tad), K 6
thermal conductivity (k), WmK−1 0.582 13
dynamic viscosity (з), Pa×s 0.001
Table 1. Model parameters for coolant and microwire.
Before proceeding to discuss the obtained results, we need to determine the optimal control points located on
the microwires for the discrete measurements of a temperature. These points are shown in Fig. 2. It is worth
noting that to determine the relaxation time we need to know the measurement temperature. The other
locations of thermocouples are similar. To do this, the following criteria were used: the microwires are fully
heated when the temperature difference between the liquid and wires is reduced to e times. As the result, the
measurement temperature is 317.79 K.
203
Figure 2. The set of control points (thermocouples) located on the microwires. We note that the indicated positions of
thermocouples are optimal for the determination of characteristic relaxation times, which are important for the
estimation of cell cooling capacity. The direction of coolant flow is indicated by the arrow.
In Fig. 3, we illustrate the time dependences of the temperature difference between the liquid and wires,
which are calculated for different thermocouples located on the microwires with different diameter. As it was
mentioned above, the relaxation time is estimated at the temperature of 317.79K for different control points.
Figure 3. The difference temperature between the liquid and microwires with different diameters as a function of time.
The inlet coolant velocity is 0.7 ms−1. The diameters of microwires are (a) 10 µm and (b) 50 µm. Temperature curves
are presented for different thermocouples (see points in Fig. 2).
It can be also observed from both Figs. 3 that the optimal temperature control point (thermocouple) located
on the microwire is represented by the triangle symbol. We summarized the obtained results of the relaxation
time calculation in Table 2.
U = 0.7 ms−1 U = 2.5 ms−1
(d = 10 µm) relaxation time, ms 2.2 0.6
Table 2. Results of the relaxation time calculation in microwires with different diameter (d) at the fixed inlet coolant
velocity.
Figure 4. Results of the distribution of a velocity field (m s−1) in the magnetic cooling cell. The inlet coolant velocity is
(a) 0.7 ms−1 and (b) 2.5 ms−1.
Fig. 4 presents the velocity field distribution in the magnetic cooling cell involving the microwires with fixed
diameter of 10 µm. It is shown that the velocity of the fluid in the system is heterogeneous and different in
size significantly. This is due to different cross sections of the coolant channels of the current and plays an
important role in the formation of a total picture of the heat transfer process. Next, we can observe that the
204
coolant flow increases due to the fact that the microwires change the channel section. The velocity
distribution shows that the coolant flow velocity increases significantly at a distance from the wire, which is
approximately equal to the wire diameter. This is provided by changes in the channel diameter. In the study,
we examined the distribution of the temperature gradient in Km-1. Note that a thermal gradient is observed at
a greater distance from the surface when the fluid velocity is 0.7 ms-1. We also investigated the thermal
energy in Wm-2. We observe that the thermal energy transfer occurs at a distance, which is no more than the
microwire thickness. The thermal energy transfer cannot propagate further. It is related with the fact that the
coolant flow velocity becomes sufficient at the same distance but from the microwire surface. As a result, the
inlet flow carries away the thermal energy in the direction of flow.
4. CONCLUSIONS
In summary, we have investigated the heat transfer in the three-dimensional magnetic cooling cell involving
five microwires. The calculations have shown that in the case of the cooling cell with the microwire diameter
10 (50) µm, the relaxation time was found to be 4.1 (2.2) ms at the constant coolant velocity of 0.7 ms−1,
respectively. The increase in the flow velocity up to 2.5 ms−1 has led to a decrease of the relaxation time to
0.6 (2.3) ms for the fixed microwire diameter 10 (50) µm, respectively. Our calculations have shown that the
cooling power depends on the magnetic wire thickness and the coolant flow velocity. For the considered
three-dimensional model, we have found the heat transfer process propagates at a distance from the wire
surface, which is close to the wire diameter.
ACKNOWLEDGMENTS
This work was supported by the RFBR (Grant No. 14-02-01085), Ministry of Education and Science RF No.
3.2021.2014/K, and RSF No. 14-12-00570\14 and the Creation and Development Program of NUST
“MIS&S.”, President RF Grant -8480.2016.2
REFERENCES
[1] J. Liu et al., “Giant magnetocaloric effect driven by structural transitions”, Nature Mater., 11, 620
(2012).
[2] O. Pavlukhina et al., “Monte Carlo study of the magnetic and magnetocaloric properties of La1-xCaxMnO3
(x = 0.33 and 0.5)”, Solid State Phenomena, 190, 347 (2012).
[3] R. Das et al., “Effect of Co and Cu substitution on the magnetic entropy change in Ni46Mn43Sn11 alloy”,
J. Appl. Phys., 109, 07A901 (2011).
[4] O. Pavlukhina, V. Buchelnikov , V. Sokolovskiy, “Modeling of the magnetic and magnetocaloric
properties of La0.7Ba0.3MnO3 manganites by Monte Carlo method”, Functional Materials, 19, 97 (2012).
[5] V. Buchelnikov et al., “Monte Carlo study of the polycrystalline Ni2MnGa Heusler alloy”, Solid State
Phenomena, 233-234, 251 (2015).
[6] D.Y. Cong, S. Roth, L. Schultz, “Magnetic properties and structural transformations in NiCoMnSn
multifunctional alloys”, Acta Mater., 60, 5335 (2012).
[7] S. Fabbrici et al., “Co and In doped Ni-Mn-Ga magnetic shape memory alloys: a thorough structural,
magnetic and magnetocaloric study”, Entropy, 16, 2204 (2014).
[8] O. Pavlukhina et al., “First principles study of the structural and magnetic properties of Cr-doped
Ni1.75Co0.25Mn1.5In0.5 Heusler alloys”, Materials Science Forum, 845, 138 (2016).
[9] V. Sokolovskiy et al., “Predictions of a large magnetocaloric effect in Co- and Cr substituted Heusler
alloys using first-principles and Monte Carlo approaches”, Physics Procedia, 75, 1381 (2015).
[11] V. Zhukova et al., “Magnetic properties of Ni-Mn-In-Co Heusler-type glass-coated microwires”, J.
Appl. Phys., 115, 17A939 (2014).
[12] A. Sarlah et al., “Comparison of thermo-hydraulic properties of heat regenerators applicable to active
magnetic refrigerators”, J. Mech. Eng., 58, 16 (2012).
[13] L. Kuhn et al., “Magnetic refrigeration at room temperature – from magnetocaloric materials to a
prototype”, J. Phys: Conf. Ser., 303, 012082 (2011).
[14] K. Nielsen et al., “Review on numerical modeling of active magnetic regenerators for room temperature
applications”, Int. J. Refrig., 34, 603 (2011).
205
DIRECT IN SITU STUDY OF MAGNETOCALORIC EFFECT AND
MARTENSITIC TWINS STRUCTURE OF Ni-Mn-Ga HEUSLER ALLOY IN
HIGH MAGNETIC FIELDS UP TO 14 T
Yu.S. Koshkid’ko(a)*, E.T. Dilmieva(a,b,d), J. Ćwik (a), A. P. Kamantsev(b), V.V. Khovaylo(c),
V.V. Koledov(a,b), V.S. Kraposhin(d), A.V. Mashirov(b), V.G. Shavrov(b)
(a)
International Laboratory of High Magnetic Fields and Low Temperatures, PAS,
53-421 Wroclaw; Poland
(b)
Kotelnikov Institute of Radio-engineering and Electronics of RAS, Moscow 125009, Russia
(с)
Department of Functional Nanosystems and High-temperature Materials, National University of Science
and Technology "MISIS," Moscow 119049, Russia
(d)
Bauman Moscow State Technical University, Moscow, 105005 Russia
*
Corresponding author. E-mail: yuri.koshkidko@ml.pan.wroc.pl
ABSTRACT
This work reports the results of the experimental studies of magnetocaloric effect (MCE) in
Ni2.16Mn0.84Ga Heusler alloy with separate magnetic and structural phase transitions in adiabatic (∆T) and quasi
isothermal (∆Q) regimes by direct methods in magnetic fields up to 10 T. The data obtained allow for
comparing ∆T and ∆Q values in the vicinity of martensitic transition and Curie point. It is found maximal
values of ∆T = 4,7 K at T = 311 K and ∆Q = 4,6 kJ/kg at T = 313 K in the fields H = 10 T. Optical observations
of martensitic twins structure origin and evolution under the influence of high magnetic field confirm the
reversible magnetic-field-induced martensitic transition at constant temperature T = 313 K in the field 14 T.
Keywords: magnetocaloric effect, Heusler alloy, direct method, in-situ mode, the latent heat of transition
DOI: 10.18462/iir.thermag.2016.0175
1. INTRODUCTION
The study of the materials with high magnetocaloric effect (MCE) is of a great interest in view of
perspectives of the development of the new solid state technology of magnetic refrigeration. In addition, study
of MCE is very attractive for obtaining a new important information about fundamental aspects of magnetic
phase transitions and subsystems interaction in magnetically ordered solids. Recentely it was proved, that at
the first order magnetic phase transition the MCE values measured by indirect methods can differ substantially
from the values obtaind by the direct method [1]. Therefore study of MCE by reliable direct method is very
actual.
A lot of works are devoted to investigation of the magnetic and functional properties of Ni-Mn-Ga
ferromagnetic Heusler alloys [1-5]. The remarkable properties of these alloys are due to the magnetic field
influence on thermoelastic martensitic structural transition, with the magnetic-field-controlled shape memory
effect [4] and high values of MCE [5], being the most attractive. For Ni2+xMn1-xGa Heusler alloys at x≤0.16
the martensitic transformation takes place in a ferromagnetic state. The alloys with x = 0,18..0,27 demonstrate
merged magnetic and martensitic structural transition. For x≥0.3 the structural transition happens in
paramagnetic state [2].
MCE studies of Ni-Mn-Ga Heusler alloys by direct method had been done for the first time in the
work [1]. These experiments were conducted in comparatively low magnetic fields (~ 1.8 T) and only for the
compositions of the alloy with merged magnetic and structural phase transitions (Ni2,19Mn0.81Ga). The adiabatic
MCE maximal value in the vicinity of merged magnetostructure transition was found to be about 1,2 K. [1].
In the work [3] a new vacuum calorimeter was used to simultaneously measure adiabatic MCE (∆T) and quasi
isothermal MCE (∆Q) of magnetocaloric materials using Bitter coil magnet in the fields up to 140 kOe.
Measurements were performed of the adiabatic temperature change (ΔT) and the specific heat transfer (ΔQ)
in quasi isothermal regime in Ni2.18Mn0.82Ga Heusler alloy with merged martensitic transition and Curie point.
The maximal values of MCE in H=140 kOe were found to be ΔT=8.4K at initial temperature 340 K and
ΔQ=4,9 kJ/kg at 343 K. The magnetostructural phase transition is found to be not completely reversible even
206
at 14 T due to comparatively wide thermal hysteresis of the martensitic transition in this alloy and its low
sensitivity to the magnetic.
The purpose of the present work is to study experimentally MCE in adiabatic (∆T) and quasi isothermal
(∆Q) regimes of Ni2.16Mn0.84Ga Heusler alloy with separate magnetic and structural phase transitions by the
direct methods in high magnetic fields of Bitter coil and to compare ∆T and ∆Q values in the vicinity of the
martensitic transition and near Curie point of this alloy. Another purpose of the work is to perform the direct
observation of martensitic twin structure formation under the influence of magnetic fields up to 14 T.
Previously the observations of martensitic twins structure evolution in Ni-Mn-Ga Heusler alloys were carried
out only in comparatively small stationary magnetic fields (about 1,1 T) or impulse magnetic fields (up to 31
T) and in Heusler alloys samples with the inverse MCE [6-8]. The purpose of the present work is also to
improve the qualitative knowledge about the MCE nature by in situ optical observations of the origin and
annihilation of martensitic twins during reversible magnetic-field-induced martensitic transition.
2. EXPERIMENTAL
2.1 Methodology of the experiment

2.1.1 MCE measurement in adiabatic regime (∆T) by the direct method
The experimental device has been constructed, targeted to allow for direct MCE measurements in high
magnetic fields in the temperature range of 4,2-350 K, using the extraction method [9]. The Bitter coil is used
to generate the cuasi stationary magnetic fields up to 14 T. The temperature of the sample was measured using
a thermocouple. The sample was encased in vacuum within the thermostat with thermal screen to minimize
the heat exchange with the environment. The sample was moved from/to the center of the Bitter coil magnet
on the distance of 0.35 m during 1 second by linear actuator. A Hall sensor placed in the sample holder was
used to measure the magnetic field. It should be noted that the magnetic field value in the point of minimal
field was less than 2 % of the maximum field in the center of the Bitter coil.
2.1.2 MCE measurement in quasi isothermal regime (∆Q) by the direct method
The new method of MCE direct measurement in quasi isothermal regime was applied in these experiments
(for more details see [3,10]). For this purpose the copper bloc was placed in good thermal contact with the
sample as shown on the inset in Figure 1a. The system of "block+sample" (inset in Figure 1a) was placed in
thermally insulated sample holder for MCE measurement. The thermocouple was fixed between the two parts
of the copper bloc (see Figure 1a). The quasi isothermal heat transfer from the sample was calculated by the
formula:
∆Q ≈ (M/m)•C•ΔTb, where M » m. (1)
M – the mass of the block, C – it’s specific heat, m – the mass of the sample, ΔTb - the temperature
change of the block measured by thermocouple. The weight ratio between copper block and the sample was
not less then 7:1.
2.1.3 Observation of martensitic domain structure
For visualization of austenite-martensite phase transformation the original device was constructed. It consists
of the optical microscope and vacuum thermostat with the temperature range from 77 K to 420 K exposed to
a magnetic field. The observation of martensitic twins structure was carried out on the sample’s surface
polished in austenitic state. The temperature of the sample was measured in situ with magnetic field cycling.
The video records were synchronized with the magnetic field change and the sample temperature measurement.
2.1.4 The samples
The samples of Ni2+xMn1-xGa Heusler alloy with different stoichiometric compositions were produced
by arc melting method[2]. According to the results of measurement of magnetization in low magnetic field
(100 Oe), the characteristic temperatures of phase transition in the samples of Ni2.16Mn0.84Ga alloy were found
to be AS=307 K and AF=315 K, TC =339 K (see Figure 1a).
2.2 The experimental results and discussion
Fig. 2 a illustrates the temperature dependences of MCE measured in the adiabatic regime (∆T) in
magnetic fields of 5 and 10 T. As it can be seen, near TC = 339 K the temperature changes ∆T = 2 K and 4 K
were observed for magnetic fields of 5 and 10 T, respectively. In the vicinity of martensitic structural transition
the narrow peaks of MCE, which have 3.3 K and 4,7 K maximal values in the magnetic field of 5 and 10 T,
respectively were observed. The increase of magnetic field up to 10 T widens the MCE pick in the vicinity of
martensitic transition and shifts MСE maximum near Curie temperature to higher temperatures region.
207
Worthwhile mentioning that temperature dependences of MCE measured both in adiabatic and
isothermal regimes demonstrate the irreversible character close to martensitic transition point and difference
of the MCE values at first switching on and at consequent cycling of the magnetic field. The ∆T-effect value
in the cooling mode of a sample at a temperature of 311 K and in case of the first switching on of magnetic
field of 10 T are equal to 4,7 K. At the second and further switchings on of the magnetic field of 10 T the ∆T-
effect equals to 4,2 K. (see figure 1b).
copper block
Sample
Figure 1. MCE measurements in quasi isothermal (∆Q) and adiabatic (∆T) regimes in Ni2.16Mn0.84Ga. ( a) The
dependence magnetic moment versus temperature in 100 Oe magnetic field at cooling. The inset shows the photo of the
copper block with a sample for ∆Q measurements. ( b) The temporal dependences of the magnetic field change (red
line) and the resulting temperature change (circles) of the sample during the adiabatic MCE measurements.
Figure 2. Temperature dependence of MCE: a) in adiabatic regime (∆T) under 5 and 10 T magnetic field, b) in quasi
isothermal regime (∆Q) under 2, 5 and 10 T magnetic field.
The results of the measurement of ∆Q values versus temperature at heating and at cooling are given in
Fig. 2b. It is seen that ∆Q in the vicinity of martensitic transition considerably exceeds ∆Q measured near
Curie point. ∆Q maximal value in the field of 10 T is equal to 4.6 kJ/kg, and near Curie point it is equal only
to 1,6 kJ/kg. Higher ∆Q values in the vicinity of martensitic transition are caused by the impact of the structural
subsystem. It can be compared with the latent heat of martensitic structural phase transition of the first-order.
The latent heat of the transition is measured on zero field by differential scanning calorimetry (DSC) and is
found to be near 7 kJ/kg. Therefore, ∆Q value determined by direct method in 10 T magnetic field is less by
34% than the latent heat of the martensitic transition measured at zero field. The presented measurements were
done in the field up to 10 T. The analysis of the curves on Figures 2 a and b reveals that there is still not
complete structural transition even at optimal initial temperature near 311 K. This difference can be related
to the fact that magnetic field diminishes difference between internal energies of the martensitic and austenitic
structural phases of the solid state ferromagnetic magnetic material with first order structural transition. In
order to clarify the relation of MCE and change of the structural state under the influence of magnetic field the
208
direct in situ study is undertaken of the origin and evolution of the martensitic twin structure under the influence
of magnetic field from 0 to 14 T at 308 K temperature. It is illustrated by micro photographs given on Fig. 3.
It is seen, that formation of martensitic domain structure at 308K temperature begins in a magnetic field of 4
T. With the further increase of the magnetic field the proportion of martensitic twins and austenitic uniform
phase grows and the martensitic formation is almost entirely completed in 12 T field. At 14 T magnetic field
only the minor changes of martensitic twins structure in the lower left corner of a frame are observed. This
qualitative confirms full completion of martensitic transition at 14 T.
0T 4T 8T 12 T 14 T
Figure 3. In situ video registration of magnetic-field-induced martensitic twin structure origin and evolution in
Ni2.16Mn0.84Ga alloy at the temperature of 308 K (size 2x2 mm2).
3. CONCLUSIONS
MCE was studied for Ni2.16Mn0.84Ga Heusler alloy in adiabatic (∆T) and quasi isothermal conditions
(∆Q) in magnetic fields up to 10T. ∆Q value in the vicinity of martensitic transition considerably exceeds ∆Q
value near Curie point. It is found maximal values of ∆T = 4,7 K at T = 311 K and ∆Q = 4,6 kJ/kg at T = 313
K ar H = 10 T. The magnetic-filed-induced magnetostructural transition is not completely reversible even at
10 T. Optical observations of martensitic twins evolution under high magnetic field confirm qualitatively the
reversible magnetic-field-induced martensitic transition at constant temperature T = 313 K in the field 14 T.
ACKNOWLEDGMENTS
The research was supported by the project No.146-MAGNES of the ERA.Net RUS Plus initiative of the EU
7th Framework Programme (Yu.S.K, E.D., J.Ć.) and RFBR project No 16-32-00905. The direct MCE
measurements in isothermal regime are done under support of RSF project No 14-22-00279.
REFERENCES
[1] Khovaylo, V.V., Skokov, K.P., Koshkid'ko, Yu.S. et al. “Adiabatic temperature change at first-order
magnetic phase transitions: Ni2.19Mn0.81Ga as a case study”, Phys. Rev. B 78, 060403 (2008).
[2] Khovaylo V.V., Buchelnikov V.D., Kainuma R., at al. “Phase transitions in Ni2+xMn1−xGa with a high Ni
excess”, Phys. Rev. B. 72, 224408 (2005)
[3] Kamantsev A. P., Koshkid'ko Yu. S., Khovaylo V.V., et al.“Magnetocaloric and thermomagnetic properties
of Ni2.18Mn0.82Ga Heusler alloy in high magnetic fields up to 140 kOe”, J. Appl. Phys. 117, 163903 (2015).
[4] Cherechukin A.A.; Dikshtein I.E.; Ermakov D.I.; et al “Shape memory effect due to magnetic field-induced
thermoelastic martensitic transformation in polycrystalline Ni-Mn-Fe-Ga alloy” Phys.Lett.A 291, 175 (2001).
[5] Aliev A.; Batdalov A.; Bosko S.; et al “Magnetocaloric effect and magnetization in a Ni-Mn-Ga Heusler
alloy in the vicinity of magnetostructural transition.” JMMM 272, 2040 (2004).
[6] T. Kihara, I.Katakura, et al. “Optical imaging and magnetocaloric effect measurements in pulsed high
magnetic fields and their application to Ni–Co–Mn–In Heusler alloy”, J.All Comp. 577S, S722 (2013)
[7] T. Gottschall, K. P. Skokov, R. Burriel, O. Gutfleisch “On the S(T) diagram of magnetocaloric materials
with first-order transition: Kinetic and cyclic effects of Heusler alloys”, Acta Materialia 107, 1 (2016)
[8] Korpusov, OM; Grechishkin, RM; Koledov, VV et al. “Simultaneous magnetooptic observation and
thermomagnetic analysis of phase transitions in shape-memory Ni-Mn-Ga alloys.” JMMM 272, 2035 (2004).
[9] I.D. Rodionov, Yu S. Koshkid’ko, J. Cwik, et al., “Magnetocaloric effect in Ni50Mn35In15 Heusler alloy in
low and high magnetic fields”JETP Lett. 101, 385 (2015).
[10] A. Kamantsev, V. Koledov, et al. “Thermomagnetic and magnetocaloric properties of metamagnetic Ni-
Mn-In-Co Heusler alloy in magnetic fields up to 140 kOe”. EPJ Web of Conferences, 75, 04008 (2014).
209
MAGNETOCALORIC PROPERTIES OF THE STACKED
Ni50Mn18.75-xCu6.25+xGa25 (x = 0÷1) POLYCRYSTALLINE ALLOYS
R. Wroblewski(a)*, K. Sielicki(a) , M. Leonowicz(a)
(a)
Faculty of Materials Science and Engineering, Warsaw Universtity of Technology, Woloska 141 St,
02-507, Warsaw,
*
Corresponding author. E-mail: r.wroblewski@inmat.pw.edu.pl
ABSTRACT
Nowadays, much effort is devoted worldwide to the development of refrigerators utilizing the magnetocaloric
effect (MCE). The MCE is observed in the course of heating and cooling of the material during the successive
cycles of magnetization and demagnetization of materials exhibiting magnetic transitions, i.e. ferromagnetic -
paramagnetic transition at the Curie temperature. To this group belong Heusler phase based Ni2MnGa alloys.
In our paper the effect of stacking of NiMnCuGa alloys, having various compositions, was studied towards
achieving a broader temperature range of the relatively high magnetocaloric properties. Three stacked alloys,
with 6.25, 6.75 and 7.25 at. % of Cu content exhibit a broad spectrum, over 25°C, of the magnetic entropy
change (ΔS), calculated on a basis of thermomagnetic curves recorded with 200 kA/m step over the
0-1200 kA/m range. A single alloy, having 6.25 at. % Cu, exhibits high value of the magnetic entropy change,
approx. 12 J/kgK, however, the change occurs within a narrow temperature range of a few centigrades.
Calculations of the relative cooling power (RCP) revealed the value of 62.5 J/kg for the stacked alloys, over
a broad temperature range of 25°C, which exceeds the respective RCP for the single alloy – 46.5 J/kg. Although
the maximum value of the ΔS is three times lower than for the single alloy the direct measurement of the
adiabatic temperature change (ΔTad) revealed the constant value of the ΔTad equal to 0.4 °C in the temperature
range from 35 to 65 °C.
NiMnCuGa alloys package seems to be an interesting alternative to expensive alloys based on gadolinium or
lanthanum. The temperature change can be relatively easily controlled with the chemical composition and/or
the heat treatment of the alloys thereby increasing the operating temperature range of the devices made of
them. The alloys described are also simple to prepare by casting and heat treatment.
Keywords: magnetocaloric effect, adiabatic temperature change, Heusler alloys, polycrystalline.

DOI: 10.18462/iir.thermag.2016.0176
1. INTRODUCTION
A refrigerator can be found in every household. In addition, various types of refrigerators are widely spread in
the industry, not just food related. A pioneer of modern refrigeration was William Cullen, who in 1755 built
a device capable of producing a very small amount of ice utilizing the boiling of diethyl ether phenomenon [1].
Less than three years later, Benjamin Frankilin and John Hadley showed that the phenomenon of evaporation
can used to reach a temperature below the point of water freezing [2]. The breakthrough was an invention of
the compressor device for the production of ice described in 1805 by Oliver Evans [3]. The first, fully
operational devices, based on compression-decompression cycle, were developed by Carl von Linde in the
years 1874-1901 and led to condensation of air and to obtain liquid oxygen [4]. For over 100 years refrigerating
devices have been continuously improved in order to optimize efficiency and energy consumption, and reduce
negative environmental impact.
In the last few decades other, already known, physical phenomena were introduced to cooling. The most
famous are thermoelectric devices based on the Peltier effect [5]. Currently, scientists and designers devote
a lot of effort to the development of refrigerators utilizing the magnetocaloric effect, i.e. heating and cooling
of the material during successive cycles of magnetization and demagnetization of the materials exhibiting
magnetic transformation, eg. a ferromagnetic - paramagnetic Curie temperature. In the laboratory conditions,
as a benchmark material with known properties, gadolinium is used, however, because of its high cost and
potential problems of accessibility, alternative materials are being sought. Among them a wide group is
represented by the Ni2XY alloys. Those Heusler phase-based alloys are characterized by the occurrence of
magnetostructural and magnetic transformations. Depending on the chemical composition it is possible to
210
control the position of both transition temperatures. The reference [6] describes the substitution of lead for
copper in NiMnCuGa alloy, which resulted in a slight decrease of the Curie temperature and a significant shift
of the magnetostructural transition toward lower temperatures inflicting the separation of these
transformations. Thus, it is possible to control the position and magnitude of the magnetocaloric effect.
However, the main problem with the use of Heusler alloys is narrow operating temperature range of just few
degrees. In order to broaden it we have carried out an experiment with stacked plates of alloys having similar
chemical composition and slightly different magnetostructural transition temperatures.
2. EXPERIMENTAL
A series of alloys, having chemical compositions described by the Ni50Mn18.75 - xCu6.25 + xGa25 formula, was
prepared by induction melting of pure elements in a protective atmosphere of argon. The ingots were annealed
at a temperature of 810°C for 48 hours, and slowly cooled with the furnace.
The chemical composition was investigated with the X-ray microanalysis (EDS) technique in the scanning
electron microscope Hitachi S-3500.
Magnetic properties and transition temperatures were determined using the LakeShore vibrating sample
magnetometer (VSM), equipped with a cryostat. The field of 4 kA/m (5 mT) was applied in the temperature
range of 0 to 130°C.
The magnetic entropy change was calculated using Maxwell realtion:
 M 
H
S M T , H    0    dH (1)
0
T  H
where: SM – magnetic entropy, μ0 – magnetic permeability of vacuum, M – magnetization, T – temperature and
H – magnetic field.
The adiabatic temperature change (ΔTad) was recorded in a magnetic field of 1600 kA/m (2 T) with a GaAlAs
diode. The sensor has got a 1.9 x 1.9 mm footprint. The examined samples were square in shape with 5 x 5 mm
long edges and a 0.8 mm thickness. They were stacked parallel onto the sensor plane, one on top of each other.
Thermal contact between the sensor and plates was established with a stock CPU thermal paste, containing no
silver particles, having a thermal conductivity of approx. 4 W/mK. The frequency of the magnetic field changes
was 1/30 Hz and the
Figure 1 shows the results of magnetization versus temperature measurement for the stacked
Ni50Mn18.75 - xCu6.25 + xGa25 alloys, with 6.25 to 7.25 at%. Cu content, registered in the field of 1 200 kA/m
(1.5 T). The characteristic deflection points of the magnetization curves, indicated by arrows, correspond to
ferromagnetic - paramagnetic transitions of the alloys with different copper content. For the alloy having
6.25 at%. Cu the transition is at a temperature of 41°C, for 6.75 at%. Cu - 50°C, and for 7.25 at%. Cu - 59°C.
The temperature hysteresis between heating and cooling curves is 11°C. These results are consistent with those
published by Stadler et al [7].
Figure 1. Magnetization versus temperature for stacked NiMnCuGa alloys having 6.25, 6.75 and 7.25 at. % Cu;
H = 4 kA/m (vertical arrows indicate ferro-paramagnetic transformation temperature for each alloy).
211
In the same work [7] the calculations of the magnetic entropy change for the Cu NiMnGa doped alloys are
presented. Magnetic field of 1 600 kA/m (2 T) induce a ΔSM from -20 to -30 J/kgK for the alloys described
therein, however, the entropy peak is very narrow and its width does not exceed 2°C . The way to extend the
working range of the alloys proposed by the authors seems to be very interesting, however, it is not described.
NiMnGa alloys, containing Cu and Co, with a total content of these elements higher than 5% at., were patented
in 2008 [8] for magnetocaloric applications. Magnetic entropy changes, for them, reach -20 J/kgK in the field
of 1 600 kA/m. The patent even mentioned a possibility of the application of the stacked materials, exhibiting
different transformation temperatures, in the refrigerating device, however, no figures indicating the validity
of such a solution were published. Therefore, in our paper we attempt to determine the magnetocaloric
performance of three stacked alloys with similar transition temperatures. Figure 2 shows the relation between
the SM change and the temperature for the fields 200, 400, 800, and 1 200 kA/m (0.25, 0.5, 1 and 1.5 T).
Fig. 2. Magnetic entropy changes for stacked NiMnCuGa alloys having 6.25, 6.75 and 7.25 at. % Cu; inset shows
magnetic entropy change for the Ni50Mn18.75Cu6.25Ga25 alloy only.
The three minima, corresponding to the transformations occurring in the studied alloys, are clearly visible.
The ΔSM magnitude is not high, and reaches -2 to -3 J/kgK, but its width is greater than 25°C. In comparison,
the magnetic entropy change for an alloy containing 6.25 at%. Cu is -12 J/kgK but it is only 5°C wide (Fig. 2,
bottom right). The RCP parameter (relative cooling power) was also calculated, using the formula:
RCP   S Mmax  T FWHM (2)
where: ΔSMmax – maximum value of SM and δTFWHM – temperature full width at half maximum.
For an alloy with 6.25 at% Cu the RCP is 46.5 J/kg, and for a stack RCP = 62.5 J/kg, but the width of the
changes are 2.75°C and 25°C for former and latter, respectively. We are aware that the RCP is often taken as
a measure of a potential performance and it should be applied for a materials with a single phase transition but
we do believe it may be a first approximation of available temperature span even in case of multiple peaks.
Liu et al. proposed a similar solution to overcome the narrow range of the magnetocaloric effect in Heusler
alloys in the work of dedicated to NiMnInCo alloys [9]. Their alloy containing 13 at%. indium and
5 at%. cobalt showed reproducible temperature change in the magnetic field equal to ΔTad = 1°C, and the
authors suggested the possibility of stacking of several alloys, which transition temperatures position can be
controlled by annealing at different times.
212
Fig. 3. The adiabatic temperature changes (∆T) measured for stacked NiMnCuGa alloys having 6.25, 6.75 and
7.25 at. % Cu; H = 1 600 kA/m.
Stacked alloys, described in this work, are characterized by a reproducible ∆Tad = 0.4°C in the range from 35
to 65°C. Figure 3 shows the cycles of magnetization - demagnetization of the tested stack of NiMnCuGa alloys.
As one can see the magnetic field induced changes are repeatable and occur quickly with application and
removal of the magnetic field (the increase and decrease of temperature, respectively), and it proves the
stability of the magnetocaloric effect in NiMnCuGa alloys. The directly measured temperature span of
repeatable MCE is in good agreement with its approximation based on the Eq. (2), however, we are not
postulating this method (RCP calculation) as an appropriate way to examine the potential performance of
magnetocaloric composites.
4. CONCLUSIONS
Presented stack of NiMnCuGa alloys seems to be an interesting alternative to the alloys based on expensive
gadolinium or lanthanum. Transformation temperatures can be relatively easily controlled with the chemical
composition as well as the heat treatment increasing the operating range of a device made of them. The
materials described are also simple to prepare by casting and heat treatment.
ACKNOWLEDGMENTS
This work was supported by the project No. UDA-POIG.01.03.01-00-058/08-00 granted by the Polish Ministry
of Science and Higher Education within the Operation Programme Innovative Economy
REFERENCES
[1] A.R. Chandra, “Refrigeration and air conditioning” PHI Learning Pvt. Ltd. (2010).
[2] B. Franklin, “Cooling by evaporation” (Letter to John Lining), Londyn, (June 17th 1758).
[3] D. Banks, “An introduction to thermogeology: ground source heating and cooling”, Oxford, UK: Blackwell
p. 120 (2008).
[4] http://www.britannica.com/biography/Carl-Paul-Gottfried-von-Linde, (acces June 1st 2015).
[5] J.F. DiSalvo, “Thermoelectric cooling and power generation”, Science 30, 285, no 5428, p. 703 (1999).
[6] K. Sielicki et al., “A Comparative studies of magnetocaloric effect in Ni-Mn-Cu-Ga and Ni-Mn-Pb-Ga
alloys”, Acta Physica Polonica A 127, no 2 p. 644 (2015).
[7] S. Stadler et al., “Magnetocaloric properties of Ni2Mn1−xCuxGa”, Applied Physics Letters 88 p.192511 1-3
(2006).
[8] N. Ali et al., patent US20080276623 A1, US 12/119,485 (2008).
[9] J. Liu et al., “Giant magnetocaloric effect driven by structural transitions”, Nature Materials 11 p. 620
(2012).
213
PROPERTIES OF GADOLINIUM WIRE FOR MAGNETIC
REFRIGERATION
R. Nomura(a), T. Kizaki(a), K. Ueno(a)*, M. Kondo(a), K. Takeuchi(a)
(a) Fujikura Ltd./1440,Mutsuzaki,Sakura-shi (Chiba,Japan)

*
Corresponding author. E-mail: ryujiro.nomura@jp.fujikura.com
ABSTRACT
In an active magnetic regenerator (AMR) cycle, the shape of the magnetocaloric material (MCM) has a large
influence on the pressure losses of the refrigerant fluid. The results of a simulation suggest that a wire-
shaped MCM lined up in the regenerator lead to more reductions in the pressure losses compared to packed
particles of MCM [1]. However, no experimental study verified the simulation results.
We have developed a gadolinium (Gd) wire with a diameter of 0.25 mm to increase the power of an AMR
based on the simulation results. This paper reports mechanic and magnetic properties of the wire. The tensile
strength of the wire was about 400 MPa and the elongation was about 3%. These properties were almost the
same as those of a copper wire. This suggests that Gd wires can easily be handled like common metal wires.
For example, this wire can be lined up side by side in the regenerator. The maximum magnetic entropy
change, ΔSm, of the wire was about 2.6 J/Kg*K. This value was lower than that of the particles (‐2.8
J/Kg*K). The Curie temperature, TC, of the wire was slightly lower than that of the particles.
More detailed data will be reported in the poster session.
Keywords: Gadolinium, material, shape, wire, AMR, ΔSm

DOI: 10.18462/iir.thermag.2016.0177
1. INTRODUCTION
In a magnetocaloric refrigeration (MC) system, speeding up an active magnetic regenerator (AMR) cycle is
one of the solutions to increase the refrigeration power of a magnetic heat pump. The bottleneck is
considerable pressure losses of the refrigerant fluid due to the particles of magnetocaloric material (MCM).
A large contact area is required of an MCM to transfer the heat between the MCM and the refrigerant fluid.
However, the small particles make the flow channel geometry more complex and cause large pressure losses.
For these reasons, the particles of 0.3 mm are commonly used because they satisfy the balance between the
contact area and the pressure losses.
In previous studies, a plate and a wire were considered candidates for use in a regenerator since these shapes
were assumed to reduce pressure losses. Stacking these materials on top of each other in layers can make the
flow channel straight and allow large contact areas that are almost the same as those of the particles of an
equal volume, depending on the thickness or the diameter.
In some studies, plate-shaped MCMs were examined [2][3], but no study has actually examined a wire-
shaped MCM. It has been considered difficult to make a wire-shaped MCM because of its low workability.
The wire with a diameter of 0.25 mm has almost the same contact area as that of the particles with a 0.3 mm
diameter of an equal volume, which is suitable to verify the simulation results. Based on the assumption, we
have developed a sufficient amount of the gadolinium (Gd) wire with a diameter of 0.25 mm by a wire
drawing technology. This paper describes the mechanic and magnetic properties of the Gd wire with a
diameter of 0.25 mm.
214
2. METHODS
We chose pure Gd as wire material for the AMR because of its high workability as metal. The Gd wire with
a diameter between 2 mm and 0.25 mm was developed by applying a wire drawing technique (Figure. 1).
We determined the properties of the Gd wire with a diameter of 0.25 mm.
The tensile strength and the elongation of the wire were measured using a tensile testing machine. A copper
wire with a diameter of 0.25 mm was also tested for comparison. the ΔSm and the TC of the wire was
measured between 240 K and 330 K and a magnetic field of 1 Tesla with a magnetic property measurement
system. The properties of Gd particles with a diameter of 0.3 mm were also determined for comparison
purposes.
3. RESULTS AND CONCLUSIONS
Figure 2 shows the tensile strength and the elongation of the three materials. The properties of Gadolinium
wire were almost the same as those of the copper wire. Furthermore, the Gd wire can easily be assembled
like common metal wires. For example, the wires can be lined up side by side in the case of the regenerator
with a winder.
The ΔSm of the Gd wire at the TC (-2.6 J/kg*K) was lower than that of the particles (-2.8 J/kg*K) (Figure 3).
The TC of the wire was lower than that of the particles. In the range of temperatures lower than the TC, the
ΔSm of the wire was -0.2 ~ -0.6 J/kg*K higher than that of the particles (Figure 4). We have set up some
hypotheses to explain the difference of magnetic properties between the wire and the particles. The
demagnetizing coefficient depends on the shape of the MCM. The shape of the Gd wire may have exerted an
influence on a demagnetizing field. Another possibility is that dislocations of the gadolinium crystal may
have been caused by the wire drawing and this may have affected the magnetic properties of the Gd wire.
500
450 Gd
Tensile strength (MPa)
400 Cu
350
300
250
200
150
100
50
0
0 1 2 3 4 5
Strain (%)
Figure 1 0.25 mm dia. Gd wire Figure 2 Stress-strain curve of 0.25 mm

dia. Gd wire and copper wire
-3.5 -3
(a) Particle (b) -2.9 Particle
-3
Wire -2.8 Wire
-2.5 -2.7
ΔSm (J/kg*K)
ΔSm (J/kg*K)
-2 -2.6
-2.5
-1.5 -2.4
-1 -2.3
-2.2
-0.5
-2.1
0 -2
240 260 280 300 320 340 280 285 290 295 300
Figure 3 (a) ΔSm of 0.25 mm dia. Gd wire and 0.3 mm dia. Gd particles in 1Tesla
(b) Enlarged graph of the vicinity of TC
215
-0.8
-0.6
-0.4
ΔSm (J/Kg)
-0.2
0.2
0.4
240 260 280 300 320 340
Temperature (K)
Figure 4 difference of ΔSm of 0.25 mm dia. Gd wire and 0.3 mm dia. Gd Particle in 1Tesla
ACKNOWLEDGMENTS
The authors thank Prof. Matsumoto of Kanazawa University for his support in the measurement of the
magnetic properties of the Gd wires and particles.
REFERENCES
[1] D. Vuarnoz, T. Kawanami," Numerical analysis of a reciprocating active magnetic regenerator made of
gadolinium wires", Applied Thermal Engineering 37, 388 (2012).
[2] B.R. Hansen, M. Katter, "Characterization study of a plate of the magnetocaloric material
La(Fe,Co,Si)13" , The 3rd IIF-IIR International Conference on Magnetic Refrigeration at Room Temperature,
Des Moines, (2009).
[3] Barbara Pulko, Jaka Tušek, "Epoxy-bonded La–Fe–Co–Si magnetocaloric plates", Journal of Magnetism
and Magnetic Materials, 375, 65 (2015).
216
ON THE ROTATING MAGNETOCALORIC EFFECT IN MULTIFERROIC
RMn2O5 COMPOUNDS
M. Balli(a)*, S. Mansouri(a), B. Roberge(a), S. Jandl(a), P. Fournier(a,b), D. Z. Dimitrov(c,d)
(a)
Département de Physique ,Université de Sherbrooke,J1K 2R1, QC, Canada
(b)
Canadian Institute For Advanced Research, Ontario M5G 1Z8, Canada
(c)
Institute of Solid state Physics, Bulgarian Academy of Science, Sofia 1184, Bulgaria
(c)
Institute of Optical Materials and Technologies, Bulgarian Academy of Sciences, 1113 Sofia, Bulgaria
*
Corresponding author. E-mail: mohamed.balli@usherbrooke.ca
ABSTRACT
Thanks to the strong magnetic anisotropy shown by RMn2O5 compounds, a large magnetocaloric effect can
be induced (around 10 K) by rotating them in constant magnetic fields instead the standard magnetization-
demagnetization method. Particularly, the TbMn2O5 single crystal reveals a giant rotating magnetocaloric
effect (RMCE) under relatively low constant magnetic fields reachable by permanent magnets. On the other
hand, it was found that the RMCE in RMn2O5 crystals strongly depends on the R3+ ion nature. Under a
constant magnetic field of 2 T, the maximum rotating adiabatic temperature change exhibited by TbMn2O5 is
more than 5 times larger than that presented by HoMn2O5. The rare earth ion size could play a crucial role in
the determination of the magnetic and magnetocaloric properties of RMn2O5 compounds through the
modulation of exchange interactions via lattice distortions, a scenario that seems to be supported by Raman
scattering.
Keywords: RMn2O5, Multiferroics, Anisotropy, Single crystals, Magnetocaloric effect.
DOI: 10.18462/iir.thermag.2016.0178
1. INTRODUCTION
The search for materials with excellent magnetocaloric properties in the temperature range from about 2 to
30 K is of great interest from fundamental, practical and economical points of view. This is mainly due to
their potential use as refrigerants in several low temperature applications such as space industry, scientific
instruments and gas liquefaction [1]. On the other hand, the development of new designs which are able to
render the magnetic cooling more competitive is also a key parameter for the commercialization of this
emergent technology. In this context, the RMn2O5 (R = magnetic rare earth) multiferroics seem to be
promising candidates for magnetocaloric tasks around 10 K [2-8].Usually, the competition between different
magnetic exchange interactions in the orthorhombic RMn2O5 compounds results in strongly frustrated
systems. Consequently, a large magnetocaloric effect (MCE) could be obtained in RMn2O5 crystals by
rotating them between their easy and hard axes in constant magnetic fields, instead the conventional
magnetization-demagnetization process (via field variation). At room temperature, this kind of materials
crystallizes in the orthorhombic structure of space group Pbam. As shown in Fig.1-a, their unit cell consists
of Mn3+O5 pyramids and Mn4+O6 octahedra which are connected each to other through oxygen atoms [2, 3].
The octahedra are aligned along the c-axis with sharing their edges. The formed ribbons are linked by pairs
of corner-shared Mn3+O5 pyramids within the ab plane. The rare earth R3+ ions are located in the empty
interstitial sites surrounded by octahedra and pyramids. In this paper, we discuss the magnetic and
anisotropic magnetocaloric properties of RMn2O5 compounds, recently studied in Sherbrooke [6, 7].
Figure 1-b shows the Raman spectra of RMn2O5 (R = Tb and Ho) at 5 K obtained with an incident light
(632.8 nm) polarized in the xy-plane. The analysis of different Raman excitations confirms that the crystals
under study form in a high quality orthorhombic crystallographic structure with Pbam space group. It is
worth noting that the competition between different magnetic exchange interactions makes RMn2O5 systems
highly frustrated. Consequently, consecutive magnetic and ferroelectric phase transitions occur at around 45,
38, 20 and 10 K [2-5]. Usually, the Mn3+/Mn4+ spins order in an incommensurate antiferromagnetic (AFM)
217
state at TN1 ̴ 45 K becoming commensurate with decreasing temperature at a lock-in transition point ( TL = 33
K). A second magnetic phase transition at which AFM ordering of Mn moments becomes incommensurate
takes place at TN2 ̴ 20 K. The onset of ferroelectric order was observed slightly below TN1, at TC ̴ 38 K, while
(a) (b)
Figure 1. (a) Crystal structure of RMn2O5 compounds [2]. (b) Micro Raman spectra at 5 K for the
orthorhombic single crystals HoMn2O5 and TbMn2O5.
the rare earth moments usually order bellow 15 K [2-5]. The temperature dependence of the magnetization
for a typical RMn2O5 (R = magnetic rare earth) multiferroic (HoMn2O5) under a low magnetic field of 0.15 T
applied along its easy axis b is reported in Fig.2-a (for example). As shown, only the magnetic transition
related to the Ho3+ spins ordering is clearly visible at low temperatures around 10 K. For TbMn2O5, the
ordering point of Tb3+ moments (not shown in Fig.2) was observed around 5 K, which is in good agreement
with early works [4, 5]. However, the phase transitions involving the manganese sublattice and occurring at
TN1, TC and TN2 cannot be clearly seen in the thermomagnetic curves shown in Fig.2-a but their presence can
be easily identified from specific heat measurements as reported in Ref. 5. This mainly arises from the
(a)
(b)
Figure 2. (a) Temperature dependence of magnetization under a magnetic field of 0.15 T applied along the b
axis for HoMn2O5. (b) Isothermal magnetization curves of RMn2O5 (R = Ho, Tb) measured at 2 K under
magnetic fields applied along their easy and hard axes.
complex arrangement of the Mn moments in RMn2O5 [2-5]. The magnetic and magnetocaloric properties
(particularly RMCE) are very sensitive to the nature of the R3+ ions (Figs. 2b to 4). The data in Fig.2-b
indicate that the magnetic easy direction of TbMn2O5 is along the a axis while that of HoMn2O5 is along the
b axis. From the linear fit of the inverse magnetic susceptibility (not shown here), the paramagnetic Curie-
Weiss temperatures along the easy axes were found to be about 0.9 K for HoMn2O5 and 20 K for TbMn2O5.
The weak value of Tθ in the case of HoMn2O5 reflects a paramagnetic behavior and/or a weak
antiferromagnetic order of Ho3+ ions. In contrast, the relatively large positive value of Tθ as in the case of
TbMn2O5 suggests a dominant ferromagnetic ordering of Tb3+ moments. This leads to a marked difference in
the behavior of the field dependence of magnetization along the easy axes of HoMn2O5and TbMn2O5, as
shown in Fig. 2-b. With increasing field, the HoMn2O5 magnetization increases slightly with a weak
tendency to saturate even under high magnetic fields (127 Am2/kg under 7 T). For TbMn2O5, the
218
magnetization easily reaches the saturation state under relatively low magnetic fields of about 2 T. The
magnetization saturation is found to be about 140 Am2/kg (8.75 µB/f.u) being closer to the Tb3+ magnetic
moment (9 µB). This indicates that the Tb3+ magnetic moments in TbMn2O5 can be completely aligned using
magnetic fields higher than 2 T, since the contribution of the Mn sublattice to the full magnetization is
negligible [2-7]. On the other hand, an enhancement of the magnetocrystalline anisotropy is observed in
TbMn2O5 (Fig. 2-b). When changing the magnetic field direction from the easy axis to the hard axis, the
magnetization under a magnetic field of 7 T is reduced by 94 % in the case of TbMn2O5 and 70 % for
HoMn2O5. The observed difference in the magnetic behavior of both HoMn2O5 and TbMn2O5 single crystals,
leads to a significant deviation between their MCEs represented by the isothermal entropy change as shown
in Fig.3-a,b. Since the hysteresis effect is negligible in HoMn2O5 and TbMn2O5 compounds, their isothermal
entropy changes ∆S can be accurately determined by using the well-known Maxwell relation [9].
Figure 3. Isothermal entropy changes of RMn2O5 (R = Ho, Tb) as a function of temperature along their
easy (a) and hard (b) axes for a magnetic field changing from 0 to 2 T.
In Fig.4-a, we report the temperature dependence of the rotating entropy change (∆SR), associated with the
rotation by an angle of 90 ° of HoMn2O5 (in the cb plane) and TbMn2O5 (in the ca plane) between their easy
and hard axes. ∆SR can be written as ∆SR = ∆S (H//easy axis) - ∆S (H//hard axis) [6, 7], where ∆S (H//easy
axis) and ∆S (H//hard axis) are the entropy changes resulting from the application of the magnetic field along
the easy and hard axes, respectively. As can be seen in Fig. 4-a, TbMn2O5 unveils a rotating entropy change
that is about 2 times larger than that shown by HoMn2O5. Under a constant magnetic field of 2 T, ∆SR is
found to be 6.36 J/kg K for TbMn2O5 and only 3 J/kg K for HoMn2O5. The improvement of ∆SR in the
(a)
(b)
2T 2T
Figure 4. (a) Temperature dependence of the rotating isothermal entropy change in RMn2O5 (R = Ho,
Tb) under 2 T. (b) Associated adiabatic temperature change under 2 T. Both compounds show a
3+
maximum RMCE at different temperatures, corresponding to the ordering points of Ho (̴ 10 K) and
3+
Tb (̴ 5 K) spins.
TbMn2O5 compound is mainly attributed to the reinforcement of the magnetocrystalline anisotropy as well as
the enhancement of the isothermal entropy change (arising from Tb3+ions) along the easy axis (Fig.3-a).
More interestingly, TbMn2O5 presents a rotating adiabatic temperature change [6, 7] which is about 5 times
219
larger than that obtained with HoMn2O5 under 2 T (Fig. 4-b). Considering initially the magnetic field parallel
to the hard axis, the rotation motion around the intermediate axis by an angle of 90 ° induces a maximum
temperature change larger than 8 K for TbMn2O5 and only 1.6 K for HoMn2O5 under a constant magnetic
field of 2 T. On the other hand, the ∆Tad, R shown by TbMn2O5 is much larger than that exhibited by the
garnets DGAG (̴ 1 to 2 K/T), usually used in low temperature applications such in the hydrogen liquefaction
[1]. The giant ∆Tad, R shown by TbMn2O5 is particularly due to its low specific heat and large rotating
isothermal change. Around the ordering point of the rare earth moments, TbMn2O5 has a specific heat of
about 6.8 J/kg K being 3 times lower than that exhibited by HoMn2O5 [5].
It is worth noting that the fundamental mechanisms behind the nature of the coupling between the
magnetic ordering, crystal structure and magnetocaloric properties in RMn2O5 are still unclear. However,
according to available data [2, 3], we first speculate that the R3+ spins ordering, magnetic anisotropy and
consequently the strength of the rotating magnetocaloric effect in multiferroic RMn2O5 could be strongly
controlled by the atomic radius of the rare earth ions. This can be well understood when taking into account
the interplay between lattice distortions and exchange interactions. Looking at the RMn2O5 crystallographic
structure reported in Fig. 1-a, the Mn3+ (S = 2) and Mn4+ (S = 3/2) spins are ordered within the ab plane in
loops of five Mn following the arrangement Mn4+-Mn3+ -Mn3+-Mn4+-Mn3+ . Based on the crystal structure,
the magnetic exchange interactions are mainly driven by five nearest-neighbors of Mn lattice identified as
Mn4+-O2-Mn4+ (J1), Mn4+-O3-Mn4+ (J2), Mn4+-O4-Mn3+ (J3), Mn4+-O3-Mn3+ (J4) and Mn3+-O1-Mn3+ (J5) [2,
3]. According to the Goodenough-Kanamori-
Anderson rules [10-12] and the Mn-O-Mn bond
angles associated with the exchange interactions in
HoMn2O5 and TbMn2O5 [3], it seems that J3 and J4
reinforce the ferromagnetic interactions of the Mn4+
magnetic moments located in adjacent edge-shared
octahedra, either side of the R3+ layer (J1), as
shown in Fig. 1-a. This could explain the marked
difference in the magnetic and magnetocaloric
behaviours of RMn2O5 since the interaction
between nearest Mn4+ spins are strongly modulated
by the radius of the rare earth. As reported by Blake
et al [3], the Mn4+-O2-Mn4+ bond angle increases
with increasing the R3+ size. At 60 K, it was found
Figure 5. Temperature dependence of the to increase from 97.10 ° in the case of R = Ho to
-1 97.45 ° for Tb, increasing consequently the
Raman-active phonon ~ 630 cm Ag for
HoMn2O5 and TbMn2O5. corresponding interatomic distance Mn4+-Mn4+
from 2.887 to 2.902 Å. The resulting interactions
may play a role in the determination of the magnetic arrangement of R3+ via the local magnetic field
produced by Mn4+ ions, and consequently the rotating magnetocaloric effect in RMn2O5. This scenario seems
to be supported by Raman scattering investigations. The temperature dependences of ~ 630 cm-1 Ag mode of
RMn2O5 (R = Tb and Ho) are reported in Fig.5. For both, the frequency of this phonon (Mn-O stretching
mode) deviates from the regular anharmonic behavior and hardens below T* ~ 65 K. According to early
studies [13-15], T* is a characteristic temperature attributed to the short magnetic correlations often observed
just above the Néel transition temperature [13-15]. This frequency hardening retraces the reduction of the
unit cell volume below T* and TN previously observed in RMn2O5 (R = Tb, Ho and Bi) ( ∆ω ≈ −γ .ω0 .∆V V )
[13-15]. The volume contraction was attributed to the Mn-Mn exchange-striction [13-15]. The frequency
hardening of the considered mode in TbMn2O5 (~1 cm-1) is 2 times larger than its equivalent in HoMn2O5 (~
0.5 cm-1. This result underlines the important lattice effect (R3+ size) on the Mn exchange interactions and
therefore the ordering of the R3+ magnetic moments. However, the crystalline field effect on the 4f magnetic
moments ground state manifold may also play a role in the magnetic configuration of the R3+ spins. This
scenario is currently explored by our group (Mansouri et al).
3. CONCLUSIONS
In summary, we have investigated the magnetic and magnetocaloric properties of HoMn2O5 and TbMn2O5
single crystals. Around the ordering temperature of the R3+ moments (̴ 5 and 10 K), the two compounds show
220
a gigantic magnetic anisotropy leading to a large rotating magnetocaloric effect. However, the reported
values of the RMCE in TbMn2O5 were found to exceed largely those shown by HoMn2O5. The giant RMCE
shown by TbMn2O5 can be mainly attributed to the “colossal” anisotropy of the isothermal entropy change,
the enhancement of the Tb3+ magnetization under relatively low magnetic fields (enhancement of ∆S) along
the easy axis and the low specific heat. The R3+ size seems to have a role in the control of the magnetocaloric
properties via the configuration of the magnetic exchange interactions in RMn2O5. On the other hand, the
large RMCE shown by RMn2O5 opens new avenues for the design of new cryomagnetocaloric devices.
ACKNOWLEDGMENTS
The authors thank M. Castonguay and S. Pelletier for technical support. We acknowledge the financial
support from NSERC (Canada), FQRNT (Québec), CFI , CIFAR and the Université de Sherbrooke.
REFERENCES
[1] K. Matsumoto, T. Kondo, S. Yoshioka, K. Kamiya and T. Numazawa, J. Phys.: Conference Series. 150,
012028 (2009).
[2] Y. Noda, H. Kimura, M. Fukunaga, S. Kobayashi, I. Kagomiya and K. Kohn, J. Phys.: Condens. Matter
20, 434206 (2008).
[3] G. R. Blake, L. C. Chapon, P. G. Radaelli, S. Park, N. Hur, S-W. Cheong and J. Rodríguez-Carvajal,
Phys. Rev. B 71, 214402 (2005).
[4] N. Hur, S. Park, P.A. Sharma, J. S. Ahn, S. Guha and S-W. Nature. 429, 392 (2004).
[5] N. Hur, S. Park, P. A. Sharma, S. Guha, and S-W. Cheong, Phys. Rev.Lett. 93, 107207 (2004).
[6] M. Balli, S. Jandl, P. Fournier, M. M. Gospodinov, Appl. Phys. Lett. 104, 232402 (2014).
[7] M. Balli, S. Jandl, P. Fournier, M. M. Gospodinov, Appl. Phys. Lett. 108, 102401 (2016)
[8] B. Mihailova, M. M. Gospodinov, B. Güttler, F. Yen, A. P. Litvinchuk, and M. N. Iliev, Phys. Rev. B.
71, 172301 (2005).
[9] M. Balli, D. Fruchart, D. Gignoux and R. Zach, Appl. Phys. Lett. 95, 072509 (2009).
[10] J. B. Goodenough, Phys. Rev. 100, 564 (1955).
[11] J. Kanamori, J. Phys. Chem. Solids 10, 87 (1959).
[12] P. W. Anderson, Solid State Phys. 14, 99 (1963).
[13] A. F. Garcia-Flores, E. Granado, H. Martinho, R. R. Urbano, C. Rettori, E. I. Golovenchits, V. A.
Sanina, S. B. Oseroff, S. Park and S-W. Cheong, Phys. Rev. B 73, 104411 (2006).
[14] L. C. Chapon, G. R. Blake, M. J. Gutmann, S. Park, N. Hur, P. G. Radaelli, and S.-W.Cheong, Phys.
Rev. Lett. 93, 177402 (2004).
[15] E. Granado, M. S. Eleotério, A. F. Garcia-Flores, J. A. Souza, E. I. Golovenchits, and V. A. Sanina,
Phys. Rev. B 77, 134101 (2008).
221
ACTIVE CALORIC REGENERATOR CYCLES: AN ANALYTIC ELEMENT
MODEL
A. Rowe*, T. V. Christiaanse, P. Govindappa, I. Niknia, R. Teyber, P. V. Trevizoli
Institute for Integrated Energy Systems (IESVic)

Department of Mechanical Engineering
*
Corresponding author. E-mail: arowe@uvic.ca
ABSTRACT
We describe an analytic model for active caloric regenerator (ACR) cycles using any type of caloric
material. An ACR element is defined which provides expressions for cooling power and work input directly
without solving the governing partial differential equations for fluid and solid temperature. The ACR
performance is determined by fundamental material properties such as specific h e a t a n d a d i a b a t i c
t e m p e r a t u r e c h a n g e i n addition to the properties that define an effective passive regenerator. The
ACR element includes losses due to conduction, convection, and depolarization. We demonstrate the
predictive ability of the element model with comparisons of loss mechanisms in an ACR using a first order
magnetocaloric material. We then show how the ACR element is used with regenerators made up of layers
of different materials with a two-layer ACR example. In addition to providing a clear functional relationship
for performance, the element model allows for fast screening of materials, design, and operating conditions.
Keywords: active regenerator, caloric, modeling,
performance. DOI: 10.18462/iir.thermag.2016.0179
1. INTRODUCTION
Caloric materials are solid substances with one, or more, reversible work modes [1]. They have potential as
working materials in high efficiency systems for refrigeration, heat pumping and power generation. One of
the challenges to using caloric materials in systems that are to compete with incumbent devices is the
magnitude of the caloric properties and the temperature range over which adiabatic temperature change is
significant. Ideally, a single material would have a large entropy change and the temperature range of the
transition would be broad [3]. The active regenerator cycle has been used with magnetocaloric materials to
overcome problems of relatively small adiabatic temperature change and narrow windows of activity. Other
caloric systems suffer from similar problems with amplitude and narrowness of entropy change with respect
to temperature [2], [4]. Modeling tools are needed to facilitate material selection and system design using
caloric materials. This paper describes analytic expressions for active regenerator performance using caloric
materials.
2. THEORY
An active caloric heat pump (or refrigerator) uses an active regenerator cycle to convert work into heat
transfer. The working material is assumed to be a solid, with a single work mode, formed into a porous
matrix (the regenerator.) A heat transfer fluid oscillates through the pores of the matrix, exchanging heat
with the solid. The solid caloric material is subjected to a periodic variation in generalized force (magnetic,
strain, electric, or stress) and, due to changes in generalized displacement (magnetization, lattice structure,
polarization), exhibits a reversible temperature change. These characteristics define a generic active caloric
regenerator (ACR) cycle.
A generic heat pump based on an ACR cycle is shown in Fig. 1. On the left we recognize the system has net
work input, W, which results in heat transfer from a low temperature reservoir at TCHX to a high
temperature reservoir at THHX. Due to thermal imperfections in the system a parasitic heat leak between the
warm and cold heat exchangers exists, QHC. The system schematic on the right side of Fig. 1 defines
additional thermal imperfections using a resistor symbol with conductance, K. Here we see the ACR must
work over a larger temperature span, TC to TH, due to finite heat exchanger size. In addition, the heat pumped
through the cold side of the ACR, QC, is greater than the applied load in the cold heat exchanger, QNET, due
to heat leaks from the environment and the warm side. Unwanted thermal resistances b e t w e e n t h e
regenerator and the heat
222
Figure 1. (left) A heat pump operating between reservoirs TCHX and THHX. Imperfect thermal islotation results in a
heat leak from the hot reservoir to the cold. (right) A more detailed schematic showing various heat leak paths related
to the operating span of the ACR. TC and TH are not the same as TCHX and THHX due to heat exchanger limits.
exchangers can arise due to dead volumes [5] and, in the case of heat pump operation (heat engine mode
will not be considered here), the supply of higher temperature heat may be reduced by heat leaks to the
environment at TA. Heat leaks into the ACR, Qp, combine with interal irreversibilities causing the efficiency
to be reduced.
A basic ACR element can be thought of as a fraction of the regenerator composed of a single caloric
material. An example of the energy balance on an ACR element with external heat leaks to the cold side is
shown in the left of Fig. 2. The centre schematic shows a basic ACR element, while the main functional
characteristics are shown on the right. A force field cycles between low, FL, and high, FH, magnitude in
some defined waveform while fluid oscillates through the pores. Integrating over the blow period, τB,
results in a defined mass displaced through the ACR, md, and, when normalized by the blow period, is
equivalent to an average flow rate. The regenerator is assumed to have a constant cross-sectional area
normal to the flow, A, the fluid specific heat, cp, is constant, and the perimeter is sufficiently well insulated
so that parasitic transverse heat leaks, Qp, can be neglected. We will assume that effects of viscous
dissipation have a small impact on the fluid temperature.
With a defined fluid and waveform, caloric material matrix structure and regenerator shape, the
thermohydraulic properties such as porosity, α, surface area, convective heat transfer coefficient, h, and
thermal conductivity, k, can be determined. One of the main design parameters influencing thermal
effectiveness of a regenerator is the thermal mass. This is determined by the thermal capacity of fluid in the
pores of the matrix and the thermal mass of the caloric material. The thermal mass of the solid is a function
of the mass, ms, and the average specific heat. Caloric materials tend to exhibit useful properties near phase
transitions where specific heat is a strong function of temperat u r e a n d f i e l d . I n t h e o r y , f i r s t
o r d e r p h a s e transformations result in latent heat; in practice, the specific heat can vary rapidly and be
very large, but is finite and continuous. We will assume that hysteresis in the caloric materials is negligible.
Figure 2. (left) A single ACR element with heat leak paths to TC from ambiant, TA, and the hot side TH. (centre) A
basic ACR element in heat pump mode. (right) The ACR is a porous structure with a constant cross-sectional area A.
Synchronized field, F(t), and flow waveforms, m(t), transform temperature oscillations in the caloric material into net
heat transfer from TC to TH across the length of the regenerator, L.
223
The average specific heat of the solid, cs is based on the values at low, FL, and high field, FH,
TH
1 1
cs 
TH  TC  2 c
TC
F (T , FL )  cF (T , FH )  dT . (1)
A parameter, R, relating the ratio of total thermal mass to the thermal mass of the solid only can be defined,
m f cp
R 1 (2)
ms cs
where mf is the total mass of fluid in the pores of the matrix. In an ACR, ideally, the pore thermal mass
should be small relative to the solid so that irreversibilities due to heat transfer are minimized.
The adiabatic temperature change, ΔT, due to change in field from FL to FH is a function of temperature.
The reduced adiabatic temperature change is ξ = ΔT/T. Based on the prescribed temperature span across the
ACR element, the minimum reduced adiabatic temperature change, ξmin, for the caloric material can be
determined,
 T 
min    . (3)
 T min
∆T and ξ are the average values of ΔT and ξ between TC and TH.
The two key energy transfers describing the ACR energy balance are heat and work. Pumping work is input
elsewhere in the system so that the work term on the ACR element can be considered only caloric work, WF.
Heat transfer at the hot side is QH = QC + WF. The analytic expression for cooling power is,
md c p 1 1 md c p  TH   keff A
QC   minTC 1    1   TH  TC  . (4)
C R  2 R ms cs  TC  L
The caloric work is,
ms cs 1 md c p 
WF     R  1 T  TH  TC   . (5)
C R ms cs 
In both Eq. (4) and (5) we see an operating parameter called the utilization which is the ratio of the
thermal mass of displaced fluid to the thermal mass of the solid, Φ = mdcp/mscs. Additional details describing
closure relations, effective conductivity, keff, (including convection), operation in heat engine mode and the
significance of reduced adiabatic temperature change are described elsewhere [6]. The derivation and
validation of analytic expressions for heat and work transfer can be found in [7][8].
3. METHODS
The cooling power, Eq. (4), is described by two terms; the first is the active component which is a function
of reduced adiabatic temperature change. The second term is the passive component which reduces the
cooling power due to the effective conductivity. In passive regenerator theory, the second term is
described by the effectiveness, εp. Using the passive effectiveness, we can rewrite Eq. (4) as,
1
md c p 1 md c p  TH  md c p
QC   min TC
 1     1   1   p  TH  TC  . (6)
C R  2 R ms cs  TC  C
  
   
Active Passive
The passive effectiveness is a function of conduction, convection, and utilization and can be between 0.95 and
0.99 for a very good regenerator.
224
3.2 R → 1, Effectiveness → 1
We can now examine limiting conditions of ACR performance. Let us assume the solid thermal mass is much
larger than the pore fluid so that R = 1, and pumping work is small. The second law efficiency is then,
QC  TH  TC 
  
WF  TC 
(7)
   1 md c p 1   p   C T T 
 min 1      H C 
   2 ms cs  min  B  TC


 
If the regenerator effectiveness is very high so that the passive loss term is negligible, the efficiency of the
ACR element becomes,
 min  1 md c p TH  TC 
 1  . (8)
  2 ms cs TC 
As shown elsewhere [3], an ideal caloric material has a constant reduced adiabatic temperature change which
would make min /   1. In this case, for small utilizations the efficiency of an ACR element approaches the
ideal limit of 1.
The exergetic cooling power is,

QC
ExQ  TH  TC 
TC
(9)
md c p  1 m c 1   p   C  T 
 TH  TC   min 1    d p    H  1 
  TC
C  2 ms cs  min  B  
  
Using the expression for efficiency, Eq. (8), the exergetic power is,
md c p
ExQ    TH  TC  . (10)
C
3.2 Material Performance

We will use a magnetocaloric material with properties shown in Fig. 3 to demonstrate the use of the
analytic ACR expressions. The material’s response to field shows first order phase transformation
characteristics.
Specific Heat [J/g-K]
Entropy [J/kg-K]
Figure 3. (left) Entropy at zero field and 1.1 T. (right) Specific heat for 0 T and 1.1 T.
225
Table 1. System, regenerator and operating parameters.
Parameter Symbol Value
Solid mass ms 0.1 kg
Cycle Period τC 1s
Field Change FL →FH 0 – 1.1 T
Displaced Fluid Thermal Mass mdcp 50 J/K
Blow Fraction τB/ τC 0.5
Pore Thermal Mass R 1
External Loss Coefficients KAC, K0C 1, 0.5 W/K
Ambient Temperature TA 298 K
With the conditions specified in Table 1, the net cooling power of a single ACR with properties given in
Fig. 3 is determined using Eq. (6). The net cooling power represents the external load, QNET, and is less
than the load on the ACR due to parasitic external heat leaks shown in the system represented on the left of
Fig. 2.
QNET  QC  QHC  QAC . (11)
4. RESULTS
We consider three cases: (1) no internal (εp = 1) or external losses, (2) εp = 0.96 and no external losses, and (3)
εp = 0.96 and external losses. In reality, the passive effectiveness is a function of solid properties, but we
will neglect this for simplicity. The net cooling power for the three cases is shown as a function of rejection
temperature, TH, and temperature span, Tspan, in Fig. 4.
In the first case with no losses either internal or external to the ACR (Fig. 4 (a)), the cooling power is
determined by the active term in Eq. (6). Two effects impact the power – the shape of the reduced adiabatic
temperature change curve as a function of temperature, and the finite thermal mass of the solid. This
material has a narrow region where adiabatic temperature change is large, resulting in strong sensitivity of ξ
with temperature. Because of the strong dependence of average specific heat with temperature, utilization
varies with span. However, this is not the main reason for decreasing cooling power; instead, it is the
increasing energy for regeneration which leads to decreasing power. The regeneration energy is determined
by the enthalpy change in the fluid which increases with temperature span.
In the second case, internal losses are considered by reducing the effectiveness to 0.96 (Fig. 4 (b)). The main
impact of finite thermal effectiveness is to introduce larger parasitic load at higher temperature spans. Thus,
small-span cooling power is unchanged from the previous case. As seen in experimental results, the zero
span cooling power is largely determined by the total mass of solid and the adiabatic temperature change.
The effects of external heat leaks to the cold side are shown in Fig. 4 (c). The dominant conductance, KAC,
links the cold side to the ambient and, because the operating range straddles ambient temperature, can result
in a either an increase or decrease in load seen by the ACR depending on the operating temperatures.
Figure 4. Net cooling power with no external losses, (a) εp = 1 , (b) εp = 0.96. (c) εp = 0.96 and external losses.
226
The rapid decrease in performance with rejection temperature is a problem with single caloric materials
having a narrow operating band. This is often addressed by using more than one material in a layered
regenerator. The ACR element can be used to create a cascade of materials operating as a layered structure.
A layered regenerator structure comprised of two ACR elements is shown in Fig. 5.
Figure 5. A two layer regenerator represented by a series of two ACR elements.
The analytic description of each individual ACR element still holds; however, now additional nodes exist
which can be linked via thermal paths. Because the ACR elements are in series, the steady-state solution
of the system must satisfy individual energy balances. For example, the heat rejection at node 1 in Fig. 5 is
Q1 =Q0 + W1. Thus, the load on ACR 2 is determined by stage 1. If each ACR is comprised of material with
properties like those in Fig. 3, but with different transition temperatures, then the energy balance can only
occur when operating temperatures satisfy the individual ACR performance curves. For the two stage ACR,
the interface condition that must be satisfied is,
Q1  QH 1 (T0 , T1 )  Q C 2 T1 , T2  . (12)
Once material properties and thermal imperfections are considered, one can see how a layered structure
consisting of a large number of ACR elements can be difficult to design. The ACR element provides a
tractable way to rapidly investigate preferred combinations of properties, designs, and operating conditions.
5. CONCLUSIONS
A caloric material can be used to pump heat using an active caloric regenerator (ACR) cycle. The analytic
description of an ACR element is presented and the impacts of internal and external losses are quantified.
While the ACR element is not a replacement for more detailed numerical modelling, it is another tool for
understanding the behaviour of caloric materials and systems subject to a vast range of design choices.
ACKNOWLEDGMENTS
REFERENCES
[1] X. Moya, et al., “Too cool to work,” Nat. Phys., 11, 3, 2015.
[2] E. Stern-Taulats, et al., “Tailoring barocaloric and magnetocaloric properties in low-hysteresis
magnetic shape memory alloys,” Acta Mater., 96, 2015.
[3] A. M. Rowe and J. A. Barclay, “Ideal magnetocaloric effect for active magnetic regenerators,” J.
Appl. Phys., 93, 3, 2003.
[4] L. Mañosa, et al., “Advanced materials for solid-state refrigeration,” J. Mater. Chem. A, 1, 16, 2013.
[5] P. V. Trevizoli, et al., “Magnetic Heat Pumps: an overview of design principles and challenges,” Sci.
Technol. Built Environ., 2016.
[6] A. Rowe, et al., “Thermal effectiveness of active caloric regenerators,” in review, 2016.
[7] T. Burdyny, et al., “AMR thermodynamics: Semi-analytic modeling,” Cryogenics, 62, 2014.
[8] A. Rowe, “Thermodynamics of active magnetic regenerators: Part II,” Cryogenics, 52, 2–3, 2012.
227
MAGNETOCALORIC PROPERTIES OF Fe-Rh-(Z) (Z = Pd, Ni) ALLOYS
FROM AB INITIO AND MONTE CARLO CALCULATIONS
V. Sokolovskiy(a,b)*, O. Pavlukhina(b), V. Buchelnikov(b), P. Entel(c)

(a) Chelyabinsk State University, Br. Kashirinykh str. 129, 454001, Chelyabinsk, Russia
(b) NUST “MIS&S”, Leninsky pr. 4, 119991, Moscow, Russia
(c) University of Duisburg-Essen, Lothar str. 1, D-47057, Duisburg, Germany
*Corresponding author. E-mail: vsokolovsky84@mail.ru
ABSTRACT
In this work, we theoretically investigated the magnetic and magnetocaloric properties of Fe-Rh-(Z) (Z = Pd,
Ni) alloys by using the ab initio and Monte Carlo approaches. The equilibrium structural and magnetic
reference states as well as the magnetic exchange coupling constants for both low- and high-temperature
phases were determined by means of first-principles calculations. In order to obtain the temperature
dependencies of magnetic and magnetocaloric properties, we used the Potts-Blume-Emery-Griffiths lattice-
spin model with parameters obtained from ab initio calculations as input data. As the result, the temperature
behaviors of magnetization, specific heat, magnetic entropy as well as magnetocaloric effect (ΔS and ΔTad)
were obtained at the magnetic field change of 2 T by using the classical Monte Carlo routine. The obtained
results are found in a good agreement with available experimental data.
Keywords: ab initio and Monte Carlo methods, magnetocaloric effect, ferro and antiferromagnetic orders.
DOI: 10.18462/iir.thermag.2016.0180
1. INTRODUCTION
Today the tasks aimed at improving the effectiveness of magnetic cooling are of special interest among the
scientific world community. First of all, the interest is determined by the creation of energy-efficient, highly
reliable cooling device operating on a magnetocaloric effect (MCE) [1]. Among the main MCE
characteristics are isothermal magnetic entropy change (ΔS) and the adiabatic temperature change (ΔTad),
which arise in any magnetic material under the influence of applied external magnetic fields [1]. Generally
the change of the external magnetic field applied to the magnetically ordered materials, leads to MCE, where
upon the maximum MCE value is observed in the magnetic phase transition.
In particular, one of the most common examples of such materials is gadolinium (Gd), which up until
now has been considered as the main material in the prototype magnetic cooling technology at room
temperatures [1, 2]. Due to the Gd’s high price and business unprofitableness for the last fifteen years the
interest of the researchers has been associated with the search of ferromagnetic materials with similar or
better MCE as compared to Gd, manifested at room temperatures. Such candidates include intermetallic
alloys with metamagnetic transition (Fe-Rh, Mn3-Ga-(C, N), Ni-(Y)-Mn-Z (Y = Co, Z = In, Sn, Sb)) [1, 2]. It
should be noted that these alloys can be characterized by the presence of two MCE types (direct and inverse),
leading to respectively heating and cooling of the material when the magnetic field is changed.
As for today, Fe-Rh alloy with almost equiatomic composition is one of the best MCE materials
working at room temperatures. In this alloy, a sharp change in the magnetization when transiting from
antiferromagnetic (AFM) into ferromagnetic (FM) phase at the point 310 K is observed upon heating. In
addition, the reverse transition is accompanied by a hysteresis of about 10-20 K. It is commonly known that
the AFM-FM transition is a phase transition of the first order, accompanied by a giant volume
magnetostriction (1% increase in cell volume) and significant resistance decrease [3], which leads to a giant
magnetoresistance effect (ΔR/R ≈ - 50% in the field of 15 T) [4] and to larger MCE [5]. Direct MCE
measurements in Fe51Rh49 alloy [13] showed that the value of the adiabatic temperature change (ΔTad) was
found to be ≈ -13 K at the point 307 K with the magnetic field change ΔH = 2 T. The calorimetric studies of
Fe49Rh51 alloy by external pressure appliance (up to 2.5 kbar) and magnetic field (up to 6 T) have been
presented in [6]. The authors have shown that the barocaloric and magnetocaloric effects are reproducible
under cyclic application of pressure and magnetic field. In the recent paper [7] the calorimetric studies of
Fe48Rh52 alloy were carried out with the reverse MCE in strong magnetic fields up to 14 T. The authors
reported that with the magnetic field increase the latent heat transfer is reduced both in the forward and the
228
reverse transformations.
In this work, we present theoretical investigations the magnetic and magnetocaloric properties
of Fe-Rh-(Z) (Z = Pd, Ni) alloys by using the ab initio and Monte Carlo approaches.
2. CALCULATION DETAILS
The optimized lattice parameters of low- and high temperature phases as well as the initial spin
configurations of of Fe-Rh-based alloys were determined by using the VASP package [8] in combination
with the generalized gradient approximation (GGA) for the exchange correlation functional. The performed
calculations were semirelativistic and the spin polarization was taken into account for all the cases. The
VASP calculations were performed for the 16-atom L21-type supercell (Fe8Rh8) assuming the following spin
configurations (AFM-1, AFM-2, AFM-3, and FM) shown in Figure 1. This supercell was created from the
unit cell of the L21 structure, which can be represented as four interpenetrated fcc sublattices with two Fe and
Rh atoms located at next sites: (0, 0, 0), (0.5, 0.5, 0.5) and (0.25, 0.25, 0.25), (0.75, 0.75, 0.75), respectively.
Here, we considered that in the case of FM order the Rh atoms have small magnetic moment (< 1 µB) while it
takes zero value for an AFM order. To create the ternary Fe8Rh8Z8-x compounds, we replaced a Rh atom by Z
one according to the nominal composition.
Figure 1. Different spin configurations taken into account in ab initio calculations.
The equilibrium lattice parameters and spin configurations were further used to determine the magnetic
exchange coupling parameters for AFM and FM phases by means of the SPR-KKR package [9] with the
GGA approximation. In this case the chemical disorder in the off-stoichiometric FeRhx(Ni, Pd)1-x systems
were treated with the coherent potential approximation (CPA).
Finally, within the framework of microscopic approach at finite-temperatures, we performed Monte
Carlo (MC) calculations by using the Potts-Blume-Emery-Griffiths (PBEG) lattice-spin model, where
equilibrium spin configurations, magnetic moments and exchange coupling constants for low-temperature
AFM phase and high-temperature FM phase according the Fig. 1 were taken from ab initio calculations. The
total Hamiltonian includes the magnetic and elastic as well as magnetoelastic interaction parts. We note that
the extensive information about this Hamiltonian can be found in Refs. [10,11]. The model lattice with a real
unit cell consisted of 3925 atoms. As time unit, we used one MC step consisting of N attempts to change the
spin variables in magnetic and structural subsystems. For each temperature the properties of the system
(internal energy and order parameters) were analyzed allowing 106 MC steps and 104 thermalization steps.
In order to find the optimized lattice parameters of Fe-Rh-based alloys in dependence on the different
magnetic spin ordering and concentration of a doped atom, we performed calculations the total energy as a
function of lattice parameter. These results are presented in Figs. 2. It can be seen from Fig. 2a that for the
parent Fe-Rh compound, the AFM-1 spin configuration in a cubic cell is energetically favorable compared to
other AFM and FM configurations. In this casе, a total magnetic moment (μtot) is found to be of 0 μB/f.u. We
note that the calculated optimized lattice parameter for Fe-Rh alloy is in a good agreement with experimental
and other and theoretical values [1,3-5,12]. Besides, the stable AFM-1 order was found for both types of
alloys doping by the Ni and Pd atoms. In order to show the effect of the third atom in Fe-Rh system on
equilibrium lattice parameter, in Fig. 2b we present the composition dependence of the optimized lattice
parameter of cubic cell with the AFM-1 ordering for Fe8Rh8-x(Ni, Pd)x (x = 0 – 3) compositions. We can see
that the addition of Ni (Pd) atoms results in decrease (slightly increase) of the equilibrium lattice parameter
because of smaller (slightly larger) atomic radius of Ni (Pd) compared to that of Rh, respectively. We note
that it is crude assumption since the electronic structure and chemical bonding are also responsible for this.
229
Figure 2. (a) The total energy for Fe-Rh with different spin configurations as a function of lattice parameter. (b) The
equilibrium lattice parameter of Fe-Rh-based alloys with AFM-1 order as a function of concentration of a doped atom.
In order to investigate the possibility of martensitic transformation in these alloys in dependence on the
third element, we performed total energy calculations for tetragonal distortion of the cubic structure along z
axis. To accomplish this, we fixed the volume of a supercell as V0 = a03 ≈ a2c. The total energy differences
between the tetragonal distorted and cubic phases for Fe8Rh8-x(Ni, Pd)x (x = 0 – 3) compositions with AFM-1
ordering as functions of c/a ratio are shown in Fig. 3. In this case, the zero value of ∆E corresponds to the
austenitic phase for each compound. It is clearly seen that the addition of the third atom stimulates the
martensitic transformation with increasing doping concentration. Moreover, in the case Fe-Rh-Ni alloys, we
found larger difference between the cubic structure and tetragonally distorted one compared with the case of
Fe-Rh-Pd. It is indicated on a higher martensitic transition temperature in Ni-doped Fe-Rh alloys [14].
Figure 3. The variation of the total energy for (a) Ni- and (b) Pd-doped Fe-Rh alloys with AFM-1 as a function of the
tetragonal distortion.
By using the PBEG Hamiltonian with the Metropolis algorithm, we simulated temperature
dependences of magnetic and magnetocaloric properties. Here, MC simulations were performed for both
cooling and heating protocols. We note that for the low-temperature AFM phase we took the AFM-1 spin
configuration. In Figures 4a and 4b we show the temperature dependences of magnetization curves for Fe-Rh
alloy in magnetic fields of 5 mT and 2 T calculated in the vicinity of metamagnetic phase transition and in a
wide temperature interval. Here, we observe two phase transitions upon cooling, which are corresponded to
the para-ferromagnetic transition at the Curie point of ≈ 630 K and the metamagnetic transition from FM to
AFM phase at the point of Tm ≈ 310 K. In addition, large changes in the magnetization around 310 K as well
as the negative shift of Tm in the external magnetic field can be seen. The temperature dependences of
specific heat and magnetic entropy curves across the metamagnetic transition in magnetic fields of 5 mT and
2 T are shown in Fig. 4c. We can observe that the transition temperature is shifted towards low temperatures
at applying the magnetic field. The adiabatic temperature change upon variation of magnetic field of 2 T is
presented in Fig. 4d. It is seen that the large values of the inverse MCE in the vicinity of metamagnetic phase
transition. The same trends in M(T) and ∆Tad(T) behaviors for Fe-Rh have been found experimentally [3,13].
230
Figure 4. (a-c) Temperature dependences of magnetization, specific heat and magnetic entropy (See inset) for Fe-Rh in
magnetic fields of 5 mT and 2 T. (d) Adiabatic temperature change for Fe-Rh alloy as a function of temperature at the
magnetic field change of 2 T. Here, open (filled) symbols note cooling (heating) protocols, respectively.
4. CONCLUSIONS
In summary, we have combined ab initio with Monte Carlo approaches to investigate the Ni- and Pd-doped
Fe-Rh alloys. The ab initio calculations have shown that the AFM-1 spin configuration is more energetically
favorable in contrast to other configurations. The addition of third element into Fe-Rh system slightly
changes the optimized lattice parameter and stimulates the martensitic phase transformation. By using the
Hamiltonian model with ab initio exchange constants as input data, the temperature behaviors of
magnetization and magnetocaloric effect for Fe-Rh system have been modeled by using the
classical Monte Carlo routine. The results obtained are found in a good agreement with available
experimental data.
ACKNOWLEDGMENTS
This work is supported by RSF-Russian Science Foundation No. 14-12-00570\14, Ministry of Education and
Science RF No. 3.2021.2014/K, President RF Grant -8480.2016.2 and RFBR No. 14-02-01085.
REFERENCES
[1] K. A. Gschneidner Jr et al. Rep. Prog. Phys. 68, 1479 (2005)
[2] K.A. Gschneidner Jr., V.K. Pecharsky, Int. J. Refrig. 31, 945 (2008).
[3] J.S. Kouvel et al., J. Appl. Phys. 33, 1343 (1962).
[4] P.A. Algarabel et al., Phys. Lett. 66, 3061 (1994).
[5] M.P. Annaorazov, K.A. Asatryan, G. Myalikgulyev et al. Cryog. 32, 868 (1992).
[6] E. Stern-Taulats et al., Phys. Rev. B 89, 214105 (2014).
[7] A. Kamantsev et al., Bulletin RAS Phys. 79, 1086 (2015).
[8] G. Kresse and J. Furthmuller, Phys. Rev. B. 54, 11169 (1996).
[9] H. Ebert. SPR-KKR package Version 6.3. http://ebert.cup.uni-muenchen.de.
[10] V.D. Buchelnikov et al., J. Phys. D.: Appl. Phys. 44, 064012 (2011).
[11] V.V. Sokolovskiy et al. Entropy, 16,4992 (2014).
[12] J. Kudrnovsky et al., Phys. Rev. B 91, 014435 (2015).
[13] A. Chirkova, K.P. Skokov, L. Schultz et al. Acta Mater. 106, 15 (2016).
[14] M. Siewert, M.E. Gruner, A. Dannenberg et al. Appl. Phys. Lett. 99, 191904 (2011).
231
FUNDAMENTAL STUDY ABOUT AN AIR CONDITIONER
WHICH COMBINED THE MAGNETIC REFRIGERATION
WITH HUMIDITY REGULATION SYSTEM
N. Hirano(a)*, K. Tanaka(b), M. Takahashi(b), R. Nagamine(c) and T. Okamura(c)
(a)
Chubu Electric Power Co., Inc., 20-1, Kitasekiyama, Ohdaka-cho, Midori-ku (Nagoya, Japan)
(b)
Takenana Corporation, 1-5-1, Ohtsuka (Inzai, Japan)
(c)
Tokyo Institute of Technology, 4259-G3-39, Nagatsuta-cho, Midori-ku (Yokohama, Japan)
*
Corresponding author. E-mail: Hirano.Naoki@chuden.co.jp
ABSTRACT
We have proposed a new humidity regulation system applying a magneto-caloric effect to the desiccant type
humidity regulator. When magneto-caloric materials coated with the sorbent are heated by magnetizing, the
moisture in the sorbent is released into the air. During demagnetization, the moisture in the air is adsorbed on
the sorbent. One-dimensional numerical analysis codes which simulate the experiment were developed and
verified the validity.The analysis which was considered the effect of high magnetic field was also done and it
was clear that the performance might be raised remarkably by using high magnetic field. In addition, the
performance calculation for the magnetic refrigerator using superconducting magnets was already carried out,
and we have obtained the result that COP of 5-8 is predicted when ΔH is 10 T. The possibility to apply the
large scale air conditioner using the magnetic refrigeration technologies with superconducting magnets and
humidity regulation system was confirmed.
Keywords: humidity regulator, magnetic refrigerator, high magnetic field, large scale, numerical analysis,
experimental result
DOI:10.18462/iir.thermag.2016.0182
1. INTRODUCTION
The fundamental study about a highly effective and large scale magnetic refrigeration system using a
superconducting magnet and humidity regulation system has been started. We have conducted R&D to convert
magnetic refrigeration technology, which uses magneto-caloric effect and is a highly-efficient, energy-
conserving/environment-friendly technology, into practical application. The performance calculation for the
magnetic refrigerator using superconducting magnets was already carried out, and we have obtained the result
that COP of 5-8 is predicted when ΔH is 10 T. This value is as good as that of conventional air-conditioners.
The possibility to apply the large scale air conditioner using the magnetic refrigeration technologies with
superconducting magnets was confirmed. Recently, we propose a new humidity regulation system applying a
magneto-caloric effect to the desiccant type humidity regulator. In conventional dehumidifiers, it is generally
necessary to warm the desiccant materials with a heater at approximately 80 °C to regenerate it and release
adsorbed moisture. On the other hand, sodium polyacrylate polymer sorbent that can be regenerated at
approximately 30–60 °C has been developed. At this temperature range, the magneto-caloric effect with 10T
level of high magnetic field is applicable for heating and cooling the sorbent. When magneto-caloric materials
coated with the sorbent are heated by magnetizing, the moisture in the sorbent is released into the air. During
demagnetization, the moisture in the air is adsorbed on the sorbent. In this poster, possibility to apply the highly
effective and the large scale air conditioner which combined the room temperature magnetic refrigeration
technologies using superconducting magnets and humidity regulation system applying a magneto-caloric
effect is presented.
2. EXPERIMENT
2.1 Experimental Apparatus

We used permanent magnets with a magnetic field of 1 T. For the experimental and numerical analysis of this
system, we measured the adiabatic temperature change of the magnetic material (Gd) caused by the magneto-
caloric effect. We coated the surface of Gd with the sodium polyacrylate polymer sorbent and analyzed the
adsorption and desorption of the sorbent caused by the adiabatic temperature change. Fig. 1 shows the
232
experimental apparatus, which regulates the flow rate and the relative humidity of the air flowing into the test
section. First, the compressor supplied air to the apparatus, and the air was filtered and dehumidified to
approximately 5% relative humidity. Part of the air was humidified in the humidifier, mixed with dry air in a
mixing vessel to control the relative humidity, and passed into the test section. The flow rate was controlled
with valves and the air temperature was controlled with an air conditioner. In the test section, there were two
ducts that contained a line of five test pieces as shown in Fig. 2. Air flowed along the surface of the test pieces
in the ducts. A magnetic field was applied to the test section with moving permanent magnets. During
magnetization, Gd generated heat and the moisture was desorbed from the polymer sorbent, and during
demagnetization Gd absorbed heat and the polymer sorbent adsorbed moisture. In this apparatus, inlet and
outlet temperature of the air and inlet and outlet relative humidity of the air of the test section were measured.
Figure 3 shows a test piece. Gd which is shaped a plate is coated with a polymer sorbent. Table 1 shows the
experimental conditions. In this study, the following parameters are defined as the basic conditions; 50%
relative humidity, flow rate 1 L/min, and temperature 20 °C. The conditions were changed from basic
conditions to 25 and 75% relative humidity, flow rate to 0.5 and 10 L/min, and air temperature to 15 °C and
25°C. The relative humidity and air temperature were measured at 1 s intervals.
Fig. 1 The block diagram of the experimental apparatus. Fig. 2 The schematic diagram of the test duct.
Table 1 Experimental Conditions

Flow Rate (L/min.) 0.5, 1,2,3,6,10
@50%RH, 20℃
Relative Humidity (%) 25,50,75
@1 L/min., 20℃
Temperature (do) 15,20,22,25
@1 L/min., 50%RH
Magnetic Field (T) 1
Fig 3 The schematic diagram of the test piece
2.2 Experimental Results

At first, we carried out the check to confirm whether the
measurements of the hygrometer showed a true value. The
typical result when we gave a step change of the humidity
is shown in Fig. 4. It is confirmed that it takes nearly 100
seconds for step response from the figure. Because of the
temperature change caused by magneto-caloric effect is
completed in approximately one second, we cannot ignore
the reply delay of the hygrometer.
Therefore we calculated the time constant (τ) and revised
measurement results from the following equation.
Fig. 4 Time constant confirmation test
(1) result for the hygrometer.
φt(t) is true value of relative humidity, φm(t) is measured value of relative humidity and Δt is duration,
respectively. Fig. 5 shows the experimental results for the inlet and outlet relative humidity and air
temperature under basic conditions. The relative humidity decreased during adsorption. The outlet
233
temperature of the air was not changed by adsorption. As for the outlet humidity, it is confirmed that
the corrected results is greatly different from the measurements. It is important for the measured outlet
relative humidity to revise with the time constant. Absolute humidity is calculated from the
experimentally obtained relation between relative humidity and air temperature by using Eqs. (2)–(4).
Ps  1000  10 7.203 11735 .74 /(T  234 ) (2)
 (3)
Pw  Ps 
100
Pw
X  0.622  (4)
P  Pw
Equation (1) is the Antoine equation, used to calculate the saturated vapor pressure (Ps) from the air
temperature (T). Equation (2) is used to calculate vapor pressure (Pw) from Ps and relative humidity (ϕ).
Equation (3) is used to obtain the absolute humidity (X) from Pw and total pressure (P). Fig. 6 shows the
absolute humidity change over time caused by the magneto-caloric effect, obtained with Eqs. (2)–(4).
The amount of adsorbed moisture (M) is calculated from the absolute humidity and mass flow rate (Q) as
follows. As shown in Fig. 5, we defined the time before the humidity changing from initial state and recovering
to an initial value again as the response time of adsorption.
M   X  t  Q (5)
Fig. 5 Relative humidity and air temperature over time. Fig. 6 Absolute humidity change over time.
3. DISCUSSION
3.1 Comparison between experimental results and numerical analysis

The Gas-Side Resistance model, which is a one-dimensional numerical analysis, was used to simulate the
dehumidification performance of the desiccant [1]. The model considers only the boundary film diffusive
resistance between the air and sorbent. We carried out the calculation in consideration of the balance of the
energy and the balance of the amount of vapor in a duct as a
function of the moisture conduction and moisture transmission [2].
The experimental and numerical analysis results were compared
for the basic conditions (50% relative humidity, flow rate 1 L/min,
20 °C). Fig. 7 shows the absolute humidity during adsorption. It is
confirmed that the analytical result is in agreement with the
experimental result corrected by the time constant. The
comparison between experiment and calculation results by
changing the experiment conditions such as air temperature is
shown in Fig. 8.The numerical analysis results agree very well
with the experiments.
3.2 The possibility to apply the large scale air conditioner

The performance calculation for the magnetic refrigerator using Fig.7 Results of experiment and numerical
superconducting magnets was already carriedout, and we have analysis for absolute humidity during
adsorption
234
obtained the calculation result that COP of 5-8 is predicted when
magnetic flux density was changed between 0 and 10 T. COP is
defined the value that divided cooling capacity by power
consumption. The power of cryocooler for cooling
superconducting magnets is taken into account as the power
consumption. The electrical power needed to produce the
magnetic field is not considered because of using
superconducting magnets. Both cooling capacity and power
consumption are calculation results. We also calculated the
adsorption properties in the high magnetic field using the
developed numerical analysis code. The calculation results of
the amount of adsorbed moisture and the response time of Fig.8 Results of experiment and numerical
adsorption by changing the magnetic field at the basic experiment analysis by changing air temperature
condition are shown in Fig. 9. When the magnetic field is
increased, both of the amount of moisture and the response time are increased. We consider the reason for this
is the adiabatic temperature change increased by increasing magnetic field, and the adsorption phenomenon is
promoted.Finally, dehumidification capacity (defined 90% amount of sorption divided by 90% sorption
completion time) per volume and fan power (calculated from pressure loss multiplied by flow rate) were
compared with the value of conventional dehumidifier as shown in Fig. 10. At nearly 6L/min, our experimental
apparatus using Gd plate is required fan power which is equivalent to value of conventional dehumidifier.
From this figure, in case of using magnetic flux density change (0 to 5T), our apparatus shows dehumidification
capacity multiplied conventional value by approximately 1.5. The possibility to apply the large scale air
conditioner using the magnetic refrigeration technologies in high magnetic field was confirmed.
Fig.9 Analysis result in the high magnetic Fig.10 Dehumidification capacity of experimental and analysis
field. results and a value of conventional dehumidifier.
4. CONCLUSIONS
We confirmed the adsorption and desorption with the magneto-caloric effect from the basic experiment.
Numerical analysis code considered the magneto-caloric effect was also constructed. Our numerical analysis
results agreed very well with the experiments. The possibility to apply the large scale air conditioner using the
magnetic refrigeration technologies with superconducting magnets and humidity regulation system was
confirmed. It will be necessary to study the structure to combine magnetic refrigeration system using a
superconducting magnet and a humidity regulation system.
REFERENCES
[1] R. Yoshie, F. Yamaguchi, et al., “Development of Numerical Simulation Method for Predicting
Dehumidification Performance of Desiccant Rotors”, Journal of Environ. Eng., AIJ. 78(682), 341 (2013).
[2] R. Nagamine, T. Okamura, N. Hirano et al., “Basic Study On Humidity Regulation System By Using
Magneto-caloric Effect”, Proceedings of the 24th International Congress of Refrigeration, August 16-22,
Yokohama, International Institute of Refrigeration, ID:232 (2015).
235
FEATURES OF MAGNETOCALORIC EFFECT IN RARE-EARTH BASED
R(Co-Fe)2 LAVES PHASES, WITH R = Ho, Er
M. Anikin(a)*, E. Tarasov(a), N. Kudrevatykh(a), M. Semkin(a),
A. Volegov(a), A. Inishev(b), A. Zinin(a)
(a)
Institute of Natural Sciences, Ural Federal University,Mira st. 19, 620002, Ekaterinburg, Russia
(b)
Institute of Metal Physics, S. Kovalevskoy st. 18, 620137, Ekaterinburg, Russia
*
Corresponding author. E-mail: maksim.anikin@urfu.ru
ABSTRACT
In this paper measurement results of heat capacity (CP) and magnetocaloric effect (MCE) in Ho(Co1-хFeх)2
and Er(Co1-хFeх)2 systems over the whole range of Co substitution by Fe are presented. Phase composition
was controlled by X-ray diffraction analysis. Heat capacity was measured in the temperature range 77-320 K.
MCE has been studied within the temperature range 5-670 K in magnetic fields up to 70 kOe. It was found
that Fe concentration increase causes the following: the CP maxima disappearing at Curie temperature point
and emergence of magnetic contribution to CP in a wide temperature range lower this point; a table-like
(plateau) MCE temperature dependence for both magnetic entropy change date (ΔS) and direct ∆T-effect
measurements independent on Fe concentration. The possible reasons of such behavior are discussed.
Keywords: Magnetic properties, magnetocaloric effect, magnetic entropy change, Laves phase, heat
capacity.
DOI: 10.18462/iir.thermag.2016.0183
1. INTRODUCTION
RCo2 type binary compounds (R are the heavy rare earth elements) possess high magnetocaloric effect
(MCE) or ∆T-effect at their magnetic phase transition temperatures [1]. In magnetic properties and MCE
studies of quasibinary R(M1-xFex)2 (M = Ni, Co) ferrimagnets, it was found that a partial Co or Ni
replacement by Fe causes a significant MCE in a wide temperature range below Curie point (ТС) [2]. Such
ability is very attractive for magnetic refrigeration.
Our recent MCE measurements for some Ho(Co1–xFex)2 [2] and Dy(Co1–xFex)2 [3] compounds with Co
substitution by Fe in the relatively narrow range ( 0 ≤ x ≤ 0.16) confirmed that results and allowed to suggest
the reasons of MCE peak widening in the temperature range below their ТС.
For deeper understanding of magnetic and MCE mechanisms formation in R(Co-Fe)2 compounds we studied
the magnetocaloric properties of such systems with the Co substitution by Fe in the whole concentration
range (0 ≤ x ≤1.0) for the system containing holmium and 0 ≤ x ≤0.8 for the system containing erbium.
2. EXPERIMENTAL DETAILS
R(Co1-хFeх)2 alloys were melted in induction furnace with argon protective atmosphere or in electric arc
furnace under a pure helium protection. An excess of rare earths metal (~2 wt. %) was added to the starting
compositions to prevent the formation of Co-rich phases. A homogenizing annealing of alloys was made in a
vacuum furnace at 1220 K during 7-32 hours. The structure of the samples was determined by X-ray
diffraction (D8 Advance, Bruker) with Cu Kα radiation source. Diffraction patterns were analyzed by
Rietveld method using the “Fullprof” software [4].
Magnetic properties were measured using both SQUID–magnetometer (MPMS-XL-7, Quantum Design) in
the magnetic field up to H = 70 kOe and vibrating sample magnetometer (7407, Lake Shore Cryotronics) in
the temperature range from 450 K to 660 K under a magnetic field up to 10 kOe. Heat capacity was
measured at zero magnetic field using adiabatic calorimeter with the relative error of ± 0.6 %. Direct MCE
measurements (∆Т-effect) were carried out using MagEq MMS SV3 experimental apparatus in the magnetic
field 17.5 kOe.
236
3. RESULTS
Analysis of the X-ray diffraction data at room temperatures showed that all samples contain mainly the 1:2
phase. The crystal lattice calculation parameter (a) presented in Figure 1. The temperatures of magnetic
transitions (Curie temperature - TC) for the studied samples (see Figure 1) were determined from the
positions of the dM/dT peaks on the temperature axis, taken from the specific magnetization temperature
dependencies (M(T)) in the magnetic field of 0.1 kOe.
Figure 1. Crystal lattice parameter (a) and Curie point (TC) dependences on iron concentration (x) in the studied
systems. For ErCo2 and ErFe2 data are taken from [5-7].
Taking in the consideration the results of [7-9], it is possible to infer that the non-linear TC(x) dependences
correlate with the mean magnetic moment of the 3d-ions subsystem µd(x) in these compounds. We can say
that non-monotonic dependence TC(x) correlates with the dependence of the magnetic moment of
the d-sublattice on concentration of iron μd(x). Thus we can conclude that the d-d- exchange interaction
dominates in these system for the whole Fe-concentration range, except only for the region with x ~ 0.
2.1 Heat capacity

In Figure 2 the experimental data on heat capacity temperature dependences CP(T) are given for some of the
studied samples. The algorithm of electron-lattice (Cel+Clatt) and magnetic (Cmag) contributions calculations is
presented in paper [2]. For Ho(Co0.96Fe0.04)2 sample in the vicinity of TC point a rather narrow maximum is
observed, which is a characteristic feature of the magnetic phase transition. In samples with higher Fe
concentration, the maximum on CP(T) dependence is observed in a wider temperatures interval.
Figure 2. On the left axis: temperature dependencies of the heat capacity for the studied compounds. Experimental data
– open symbols, calculated lattice and electronic contributions (Cel + Clatt) – dashed line, magnetic contribution (Cmag)
– closed symbols. On the right axis: specific magnetization temperature dependences at 5 kOe.
237
Such spread of Cmag(T) maximum reflects the emergence and existence of specific magnetic disorder in
a wide temperature region which is also reflected on the specific temperature dependences of magnetization -
M(T). It is worth to note, that M(T) dependences of Er(Co1-хFeх)2 samples are differ from the Weiss type and
all have the deflection with temperature rise which correlates with Cmag(T) data.
2.2 Magnetocaloric effect

The existence of magnetic disorder in studied compounds inferred from the analysis of temperature
dependences of heat capacity should be considered as a magnetic entropy change (∆S) in a wide temperature
range. In Figure 3, temperature dependences of magnetic entropy change ∆S(T) are presented. The ∆S(T)
value was calculated using the formula [10].
Figure 3. Temperature dependences of magnetic entropy change ∆S(T) in magnetic fields (0-10) kOe. Inserts – ∆S(T)
in magnetic fields (0-70) kOe. Arrows indicate the T C point. The dash line – extrapolation.
It is seen that the occurrence of the magnetic disorder causes the table-like (plateau) MCE to appear at the
T < TC. For compounds with x ≤ 0.20 this plateau is merged with the MCE peak at TC –point, so only the one
wide common peak is presented. At the higher Fe concentration, the MCE peak caused by the magnetic
phase transition at TC -point is detached from the plateau area which is connected with some magnetic
disorder in R-sublattice. The same picture of plateau-like ∆S(T) dependence has been observed in paper [11]
for ErFe2. Besides, for the Er(Co1-хFeх)2 samples with x ≥ 0.40 the magnetization compensation point was
found in [6] which accompanied by the reversed MCE.
Figure 4. Temperature dependencies of ∆T-effect. Arrows indicate the TC point.
238
Our ∆S(T) experimental dependences for the magnetic field change of 10 kOe are correlated with the direct
∆T-effect measurements data for the adiabatic external magnetic field change of 17.5 kOe (Figure 4).
4. CONCLUSIONS
We found that the increase of Fe concentration causes the following magnetothermal and magnetocaloric
properties changes in the studied systems:
 A heat capacity maxima disappearing at TC point and emergence of magnetic contribution to a heat
capacity in a wide temperature range below this point;
 The plateau-like MCE temperature dependence for both magnetic entropy change date and direct
∆T-effect measurements independently on Fe concentration.
To our mind, the mentioned MCE features in R(Co1-xFex)2 intermetallics originate from the specific magnetic
state of R-ions sublattice which belongs according of Belov classification [12] to a “weak” type. Due to that
reason the R-sublattices are partially magnetically disordered in the range 0 K – TC (state similar for
paramagnet), but able to give a great response to the external magnetic field. Another possible reason is the
sperimagnetic structure formation in R-sublattices due to the local variation of electric crystal field acting on
R-ion from the Fe-ion neighbors. In other words – the deflection of R-ions magnetic moments out from the
global easy axis. In this case the external magnetic field aligns them, which produces the specific
contribution to entropy. The obtained experimental data are very important for understanding of the origin of
MCE mechanism and for design of novel and potential magnetic refrigerant materials working at room
temperature.
ACKNOWLEDGMENTS
The authors are grateful to Dr. N.V. Selezneva for the help with X-Ray measurements. This work has been
supported by the State contracts No. 1362 between UrFU and the Ministry of Education and Science of
Russian Federation and by the Fund of assistance to development of small forms enterprises in scientific-
technical sphere No. 6576GU/2015.
REFERENCES
[1] K. A. Gschneidner Jr. et al., “Recent developments in magnetocaloric materials”, Rep. Prog. Phys. 68,
1479 (2005).
[2] M.S. Anikin et al., “Magnetic structure and magnetocaloric properties of Ho(Co1-хFeх)2 quasibinary
intermetallic compounds”, Phys. Procedia 75С, 1198 (2015).
[3] M.S. Anikin et al., “Untypical temperature dependence of the magnetocaloric effect in the Dy(Co 1-xFex)2
(x = 0.10; 0.15) compounds”, Solid State Phenom. 233-234, 247 (2015).
[4] J. Rodriguez-Corvajal, “Recent advances in magnetic structure determination by neutron powder
diffraction FullProf”, Physica B 192, 55 (1993).
[5] M.I. Bartashevich et al., “Collapse of the itinerant Co moment in Er1−xLuxCo2 by the application of high
magnetic fields”, Physica B 229, 315 (1997).
[6] A.V. Andreev et al., “Influence of 3d-metal on the magnetic properties of quasi-binary intermetallic
compounds of rare-earth Er(Fe1-xCox)2”, Phys. Metals Metallogr. 59, 481 (1985) (In Russian).
[7] X. B. Liu et al., “Moment variation in Er(Co1−xFex)2 Laves phase: magnetic measurements and
Mössbauer spectroscopy study”, J App. Phys., 105, 07E119 (2009).
[8] E. Burzo, “Magnetic properties and resonance studies of some pseudobinary gadolinium or yttrium
compounds”, J. Phys. Colloque, 40, C5-184 (1979).
[9] F. Pourarian et al., “Magnetic behavior of Laves phase RCo2-xRex (R = Ho, Er) compounds and their
hydrides”, J. Magn. Magn. Matter. 25, 299 (1982).
[10] K.A. Gschneidner. Jr, V.K. Pecharsky, “Magnetocaloric materials”, Annu. Rev. Mater. Sci. 30, 387–429
(2000).
[11] E. Burzo et al., “Magnetic and magnetocaloric properties of some ferrimagnetic compounds”, Journal
of optoelectronics and advanced materials 12, 1105 (2010).
[12] K. P. Belov, “Ferromagnets with "weak" magnetic sublattice”, Adv. in Phys. Sci. 166, 669 (1996) (In
Russian).
239
EXPERIMENTAL PERFORMANCE EVALUATION OF SINTERED Gd
SPHERES PACKED BEDS
A. Tura*(a), K.K. Nielsen(b), Ngo Van Nong(b), Nini Pryds(b), P.V. Trevizoli(a), T.V.
Christiaanse(a), R. Teyber(a), A. Rowe(a)
(a)
(b)
DTU Energy, Technical University of Denmark, Roskilde, Denmark
*Corresponding author. E-mail: atura@uvic.ca
ABSTRACT
ABSTRACT — Research in magnetic refrigeration heavily relies on the use of packed spheres in
regenerators, however little investigation to verify that such non-monolithic arrangements guarantee a
sufficiently constrained structure has yet been performed. This work presents a preliminary comparison of
the performance of AMRs consisting of Gd spheres with diameters ranging from 450-550 microns partially
sintered by Spark Plasma Sintering (SPS) to similar spheres, sorted in the same size range and from the same
batch, but merely packed. Pressure drop is compared at uniform temperature and at a range of heat rejection
temperatures and temperature spans. Performance is compared in terms of temperature span at a range of
heat rejection temperatures (295 – 308 K) and 0 and 10 W cooling loads. Results show a moderate increase
of pressure drop with the sintered spheres, while temperature spans were consistently 2.5 - 5 K smaller.
These results are coherent with previously presented results [1].
Keywords: Active magnetic regenerator, spark plasma sintering, partially sintered spheres regenerator, Gd.
DOI: 10.18462/iir.thermag.2016.0184
1. INTRODUCTION
While research is in progress for the development of optimal regenerator matrix structures for commercial
applications [2], beds of packed particles, preferably of spheroidal shape, have been extensively used by
researchers for prototyping regenerators [3]. Flow channeling induced by poor packing techniques,
regenerator housing geometry, and broad mesh size distribution when coupled with magnetic forces and flow
pressure can significantly degrade the performance of a regenerator. If the particles are not fully constrained
the fluid and magnetic forces can also cause a regenerator breakdown, especially if the refrigerant is made of
a brittle material. While such issues are known, there is no experimental study available in literature
quantifying the particular relevance of their impact and if they need to be accounted for in numerical models.
Generally, detectable regenerator performance degradation (change in performance and pressure drop) over
time can be an index of matrix structure deterioration. However even a stable regenerator may have flow
channeling issues which would go largely undetected unless displaying an abnormal pressure drop. To
overcome such issues the most common technique used to create a monolithic structure is epoxy
impregnation of packed beds [4], where a thin epoxy film is used to bond the particles and constrain them in
space. The objective of this work is twofold: experimenting with the novel concept of partial sintering as an
alternative to epoxy impregnation, specifically for second order rare earth alloys, and quantifying the impact
of flow channeling in packed regenerators.
2. EXPERIMENTAL
2.1 Spark Plasma Sintering (SPS)
In this section the manufacturing process of the partially sintered spheres is presented. This was already
presented in our previous paper results [1], however it is reiterated here for the reader’s convenience.
240
Plasma Spark Sintering Process
SPS is an established technological process for sintering electrically conductive powders, granulates or
spheres. It consists in driving a large current across a powder batch pressurized and contained in a graphite
mold. Point contacts of the particles are quickly affected by Joule heating causing melting. The device can
monitor pressure, temperature, and variation in porosity during the process.
Regenerator Fabrication
The regenerator pucks were made in an SPS device as
described above. The loose spheres were kept in a graphite
die with a diameter of 16.5 mm and maintained under a
pressure of 16 MPa via two graphite punches. Before
installing the upper punch, the spheres were shaken and
pressed together in order to reach the nominal porosity for a
random close pack (0.36). The SPS profile was set to ramp
from room temperature to 525 deg. C with 50 K/min and stay
at 525 deg. C for five minutes. Ramping down in temperature
was done by turning off the SPS current and let the sample
cool. The ramp profile for this setup would typically result in
an SPS current of the order 400 A. The path of the current
was not determined. In order to maintain good electrical
contact and to protect the punches, graphite paper (0.1 mm
thick) was placed between the punch and the top sphere layer. Fig. 1. Partial sintering through SPS of Gd spheres.
We do not claim the SPS profile to be optimal, however, we did test at both lower and higher temperatures
and pressures and a good compromise between regenerator stability and integrity of the spheres seems to be
approximately realized by the profile just described. The atmosphere during the fabrication was a vacuum of
approximately 10-3 mbar.
The whole process takes about 2 hours from preparing the graphite die with spheres to the point when the
regenerator is cooled back down to room temperature. The resulting regenerator cylinders were weighted and
their dimensions measured. The bulk porosity determined from these measurements was in the range 0.35 to
0.37, which indicates that the SPS process only deforms the spheres slightly. An experiment at a larger
temperature (550 deg. C) and pressure (50 MPa) was done and here the spheres transformed into an ingot
with essentially zero porosity. Scanning electron microscope (SEM) pictures were taken on several of the
regenerators (Fig. 1). These indicate of course the random packing but, more interestingly, they also show
the necking between the spheres.
2.2 Sintered bed regenerator assembly
The objective is to compare the performance of regenerators of exactly the same and size and composition
operating under the same conditions. As discussed in Ref. [1], the sintered beds may be susceptible to
preferential flow channeling in the gap between the sintered puck and the regenerator housing. In this new
study, to avoid preferential flow channelings, we propose a new assembly of the sintered beds. The
procedure is described in Fig. 2. First the sintered pucks (Fig. 2(a)) are assembled in a shrink tube and heat is
applied, resulting in a sealed form in Fig. 2(b). Later, epoxy is used and the sealed bed is assembled in the
Ultem housing (Fig. 2(c)). After the final epoxy cure, the sintered bed regenerator has the form presented in
Fig. 2(d). This way, the shrink tube is used to guarantee no preferential flow inside the AMR matrix, while
the epoxy avoids flow channeling in the gap between the sintered bed and the regenerator housing.
2.3 AMR Tests
The AMR tests are carried out in PM I test apparatus at University of Victoria [5]. No load (0 W) and 10 W
load tests are performed at a fixed frequency of 1 Hz and displaced volume of 15.5 cm3 (about 0.8 utilization
factor), while the rejection temperature is changed between 294 and 308 K. The regenerator matrices are
composed by spheres with an average particle size of 500 μm, with a total Gd mass of about 55 g. The
porosity of the sintered spheres regenerator is 0.36, with not significant increase form the packed spheres
bed. The heat transfer fluid used was a water ethylene-glycol mixture (80/20%vol).
241
(a) (b) (c) (d)
Figure 2. (a) Sintered Gd beds; (b) sintered bed sealed in the shrink tube; (c) Ultem PEI housing; (d) Final assembly of
the sintered beds after the epoxy procedure.
3. RESULTS AND DISCUSSIONS
3.1 Pressure Drop
Pressure drop tests are performed in two ways: Fig. 3(a) maintaining the regenerators at constant ambient
temperature and constant utilization (fixed piston displacement), while the operating frequency is varied; Fig.
3(b) pressure drop is monitored during temperature span performance tests at no-load conditions, frequency
of 1 Hz and varying the heat rejection temperature from 294 to 308 K.
The differences in the pressure results can be explained by the fact that going from spheres point contact to
soldered point contact (Fig. 1) might increase the tortuosity of the flow path, and small amount of flow
channeling in the packed regenerator spheres may occur.
105
Sintered beds - 0W
200 100
Loose spheres - 0W
Sintered beds 95
Loose spheres 90
150
85
DP [kPa]
DP [kPa]
80
100 75
70
65
50
60
55
(a) (b)
0 50
0 0.5 1 1.5 2 2.5 294 296 298 300 302 304 306 308
frequency [Hz] TH [K]
Figure 3. (a) Pressure drop test as a function of the frequency; (b) Pressure drop as a function of the rejection
temperature.
3.2 AMR Performance
Fig. 4 summarizes the performance test. The results validate our original tests [1], showing that the packed
regenerator beds consistently outperform by a considerable margin the sintered sphere beds. Such results are
surprisingly unforeseen, and do not seem to relate to experimental reading errors. If there is a problem with
the test set up it is systematic and currently undetected. The regenerators are structurally and composition
wise equivalent, with the sintered regenerators holding the advantage of a monolithic structure. Expectations
were that sintered regenerator performance should be the same, if the packed sphere beds have no flow
channeling, or better. Second order material, such as gadolinium is not expected to change properties in the
SPS process. Increased axial thermal conduction induced by the more intimate sphere contact does not seem
to be plausible because the regenerator thermal gradient does not seem to affect the difference in
performance. Assuming that the tests were carried out exactly in the same manner few and minor variable
242
seem affecting the results: (a) different regenerator shell material might affect thermal leaks to the
environment, however this should also be influenced by the thermal gradient, (b) SPS did indeed affect the
material properties of gadolinium, (c) some unforeseen flow channeling in the regenerator housing, or in the
regenerator structure, however pressure drop should reflect that, (d) some flow effect or dead volumes in the
sintered pucks interfaces. Further investigation is required at this point, including testing regenerator samples
for material properties.
22.5
Sintered beds - 0W
Sintered beds - 10 W
20
Loose spheres - 0W
Loose spheres - 10W
17.5
15
Tspan [K]
12.5
10
7.5
2.5
0
294 296 298 300 302 304 306 308
TH [K]
Figure 4. Temperature span and function of heat rejection temperature for 0 W and 10 W cooling load for both sintered
and packed spheres.
4. CONCLUSIONS
New performance test comparison was performed comparing Gd sintered monolithic spheres to simply
packed spheres. Earlier results showed that packed spheres outperformed sintered spheres, however the
experiments lacked confidence in the way the sintered beds were constructed. The new beds seem to address
the issue and new packed spheres tests were also performed to match at best the design and operational
variables of the two scenarios. Results validated the previous tests showing that the sintered spheres display
larger pressure drop and inferior performance. The reason of the lower performance is still unanswered and
is currently under investigation.
REFERENCES
[1] A. Tura and K.K. Nielsen. Experimental fabrication and performance testing of partially sintered
gadolinium spheres regenerator beds. In: Proceedings of 6th IIF-IIR Int. Conf. on Magnetic Refrigeration
(Thermag). Victoria, BC, Canada (2014).
[2] K.K. Nielsen, C.R.H. Bahl, and K. Engelbrecht “The effect of flow maldistribution in heterogeneous
paralle-plate active magnetic regenerators’ J. Phys. D: Appl. Phys. 46, 2013.
[3] B. Yu, M. Liu, P.W. Egolf, A. Kitanovski, “A review of magnetic refrigerator and heat pump prototypes
built before 2010”, Int. J. Refrigeration 33 (6), 2010, pp. 1029 - 1060.
[4] A. Rowe “Active Magnetic Regenerators: performance in the vicinity of Para-Ferromagnetic second
order phase transitions”, PhD thesis, 2002, University of Victoria, Victoria BC.
[5] A. Tura, A. Rowe “Permanent magnet magnetic refrigerator design and experimental characterization.”
Int. J. Refrigeration 34 (3), 2011, pp 628-639.
243
INTEGRAL OPTIMISATION OF A MAGNETOCALORIC AIR-
CONDITIONING SYSTEM FOR AN ELECTRIC VEHICLE
B. Torregrosa-Jaime*, J. Payá, J.M. Corberán
I.U.I. Ingeniería Energética, Universitat Politècnica de València, Camino de Vera s/n, 46022, Valencia,
Spain
*
Corresponding author. E-mail: bartorja@iie.upv.es
ABSTRACT
The application of a permanent-magnet active magnetic regenerator refrigerator (AMRR) in the air-
conditioning (AC) system of an electric vehicle is analysed. A comprehensive optimisation study is performed
in order to determine the design and working parameters of the AMRR that fulfil the thermal requirements of
the vehicle with minimum combined electric consumption and total mass. With the obtained designs, the
AMRR would weigh between 20 and 50 kg and would work with an EER from 2 to 4 and a COP from 2 to
3.5. The efficiency of the AMRR depends on the considered design point, which is determined by the settings
of the electrical auxiliaries. As the efficiency of the AMRR increases, the consumption of the auxiliaries
becomes more relevant. In order to determine the design that minimizes the total energy consumption of the
AC system, all the components must be taken into account.
Keywords: magnetic refrigerator, active magnetic regenerator, air conditioning, automotive, optimisation.
DOI: 10.18462/iir.thermag.2016.0191
1. INTRODUCTION
Magnetic refrigeration technology presents several advantages with respect to vapour-compression heat
pumps, as are the absence of a gaseous refrigerant, the lower maintenance and a theoretically higher efficiency.
These features make magnetic refrigeration a potentially attractive alternative for the air-conditioning (AC)
systems of automobiles, whose performance is key to reduce their energy consumption. This is especially
relevant in the case of fully electric vehicles, in which the waste heat from the motor is not enough to cover
the heating needs of the cabin and hence an additional energy consumption is needed. Installed cooling and
heating capacities in electric vehicles range between 3 and 5 kW, while typical temperature spans can be found
between 30 and 40 K. For the same ambient conditions, the comfort temperature can be maintained inside the
vehicle with different cooling or heating capacities and different temperature settings, which depend on the air
flow pushed to the cabin. Consequently, the energy consumption of the electrical blowers will be different.
Given the current trend of increasing the efficiency of the refrigerators, the consumption of the auxiliaries may
become relevant in the global energy balance of the AC system.
Mobile AC is a challenging application for magnetic refrigeration at room temperature. No published AMRR
prototype has demonstrated the mentioned performances up to date [1]. According to a previous study by the
authors [2], a permanent magnet AMRR based on Gd-like magnetocaloric materials (MCMs) could optimally
meet a target cooling demand of 3.0 kW at a 29.4 K temperature span working with an electric efficiency
(EER) of 2 to 4. The device would weigh between 20 and 50 kg and would work with a magnetic field of 1.4
T and a cycle frequency up to 10 Hz. In order to achieve such performances with a parallel-plates regenerator,
plates between 0.03 and 0.04 mm thick and channels between 0.02 and 0.04 mm high should be employed.
In this paper, an AMRR-based AC system for an electric vehicle is optimised from an integral point of view.
Firstly, the cooling and heating requirements of an electric minibus are determined. Two design points are
selected for each working mode, with different settings of the electrical auxiliaries. The cited AMRR
optimisation study is then repeated with the new cooling design point, with the aim of finding the AMRR
features that optimally meet these requirements in terms of efficiency and total device mass. The results of the
previous and the new cases are compared in terms of AMRR efficiency and also in terms of global energy
consumption, that is, including the electric consumption of the blowers. Subsequently, the heating performance
of the obtained AMRR designs is analysed. The main novelties with respect to previous AMRR optimisation
studies [2,3,4,5] are the covered parameter space, the consideration of the equipment external to the AMRR
and the optimisation of the heating performance of the AMRR.
244
2. METHODOLOGY
2.1 Mobile air-conditioning system definition and requirements

The target mobile air-conditioning (MAC) system (Fig. 1) consists of an AMRR connected to two hydraulic
loops containing the water-to-air heat exchangers (HEX) and the fluid transport auxiliaries (pumps and
blowers). The loops fulfil the heating or cooling requirements of the cabin of the vehicle and absorb or reject
heat to the ambient. The working mode is changed by means of a valve system. The proposed layout
corresponds to the MAC system developed during the European Project ICE [6] for the Iveco Daily Electric
commercial minibus and the characteristics of the conventional components were reported in previous
publications [2].
Figure 1. AMRR-based MAC system layout. a) Heating mode. b) Cooling mode.
The thermal requirements of the minibus have been determined by means of a global model of the vehicle as
detailed in [2]. Table 1 summarizes the design conditions and the obtained cooling and heating requirements,
excluding the possible heating power needed to demist the windows in winter. Two working points defined by
different settings of the blowers inside the cabin have been selected for each mode. When the air flow rate
pushed through the HEXs increases, the temperature span decreases but the thermal demand increases due to
the auxiliaries and ventilation loads. A higher air flow rate implies a larger electric consumption of the blowers.
Design conditions Summer Winter

Comfort target 25ºC at 50% RH 20ºC
External conditions 35ºC at 60% RH 0ºC at 80% RH
Requirements S1 S2 W1 W2
Capacity (W) 3027 2803 2706 2307
Cabin side temperature (ºC) 12.8 6.8 28.6 31.5
Ambient side temperature (ºC) 42.1 41.6 -2.4 -2.2
Temperature span, ΔTspan (K) 29.4 34.8 31.0 33.8
Air flow rate to cabin (m3 h-1) 811 428 646 560
Blowers’ consumption (W) 640 366 522 398
Table 1. Design conditions and requirements of the AMRR MAC system.
2.2 AMRR definition and simulation setup

A permanent-magnet parallel-plates AMRR with rotary architecture and axial fluid flow is addressed [2]. The
MCMs in the bed have been approached by shifting the Curie point of Gd. The average demagnetizing factor
has been estimated as 0.25. The working fluid is a 50% ethylene glycol solution for automotive applications.
The electric power consumption of the AMRR (Ṗ) is given by the magnetic work (Ẇmag) and the pumping
work (Ẇpump), Ṗ= Ẇmag/ηmotor + Ẇpump/ ηpump, where η is the efficiency of electric energy conversion into
mechanical work (ηpump=0.60 and ηmotor=0.85). Ẇmag is related to the AMR cooling (Q̇ c) and heating (Q̇ h)
capacities by the conservation equation, Ẇmag= Q̇ h-Q̇ c-Ẇpump. Finally, Ẇpump depends on the pressure drop (Δp)
of the fluid flow rate through the plates (V̇ ), Ẇpump= V̇ · Δp.
The mass of the AMRR is approached as the sum of the mass of the magnet assembly and the mass of the
regenerator, which are directly related to the cost of the device. The mass of regenerator needed to meet the
cooling or heating demand of the vehicle with a given AMR configuration is inversely proportional to its
specific cooling or heating capacity, respectively. The mass of the magnet assembly has been related to the
volume of the regenerator and to the magnitude of applied magnetic field by following the approach in [3].
245
The AMRR performance and hence its resulting mass depend on multiple design and operating parameters.
An optimisation study was carried out in [2] with the purpose of determining the values of these parameters
that define an AMRR with minimum combined electric consumption and device mass, under S1 conditions.
The composition of the regenerator bed was adapted to S1 working temperatures. A first run of simulations
defined a range of near-optimal values of the studied parameters (Table 2). In all cases the optimal applied
magnetic field was 1.4 T, the maximum considered for a practical permanent magnet. A second parametric run
with the parameters in Table 2 determined their optimal values for S1 conditions, as summarized in section 1.
Parameter Values
Channel height (Hf) 0.02, 0.03, 0.04, 0.05 mm
Porosity (ε) 0.33, 0.36, 0.40, 0.44, 0.48, 0.52, 0.55
Regenerator length (L) 30, 25, 20, 15, 10, 7.5 mm
Cycle frequency (f) 2, 4, 6, 8, 9, 10 Hz
Thermal utilisation (φ) 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1
Table 2. Variable AMRR model parameters.
The optimisation study of the AMRR parameters in Table 2 has been repeated under S2 conditions, adapting
the composition of the bed to S2 working temperatures. A relevant result is that despite the 5.5 K difference
in ΔTspan between S1 and S2 conditions, the AMRR cooling performance does not increase by adding more
than 4 layers of MCMs in both cases. The optimal cooling performance results in cases S1 and S2 have been
compared. Each optimal combination of parameters defines a specific AMRR device. The heating performance
of each of these devices has been calculated under W1 and W2 conditions with the aim of determining whether
the AMRRs designed for cooling mode would also cope with the heating demand of the vehicle.
3. RESULTS
Fig. 2a summarizes the results of the optimisation studies for cooling mode. The trends obtained for S2
conditions are very similar to the ones reported for case S1 [2] and the optimal values of the parameters are
practically the same. The optimal Hf is found between 0.02 and 0.04 depending on mass or efficiency priority
respectively. Thinner channels work better with shorter regenerator lengths due to the increase in the pressure
drop and the optimal L ranges from 10 to 20 mm. The optimal ε increases with Hf so there is an optimal range
of plate thicknesses for all cases between 0.03 and 0.04 mm. The optimal f was generally 10 Hz (the maximum)
and the optimal φ is in between 0.4 and 0.7, slightly lower than for case S1 (0.5 to 0.8).
Figure 2. a) AMRR and b) total electric consumption as function of AMRR mass for summer design points.
If cases S1 and S2 are compared in terms of AMRR performance (Fig. 2a), the decrease of the cooling demand
by 7.4% achieved with the reduction of the auxiliaries’ load in case S2 does not compensate for the increase
of ΔTspan by 5.5 K caused by the corresponding reduction of the air flow rate (if a device with the same mass
is considered). However, achieving S1 working conditions requires a larger power demand by the blowers.
When the consumption of the blowers is added to the one of the AMRR, case S2 results globally more efficient
(Fig. 2b). As the specific cooling power of the AMRR increases, it becomes relatively more inefficient and
from a certain point on the total energy consumption of case S2 is larger than the one of case S1.
246
Fig. 3 shows the optimal heating performance of the devices defined by the former study. For each AMRR, f
and φ were varied in order to find the working point that meets W1 and W2 requirements with minimum
electric consumption. In the studied range, f should be maximized while φ is regulated. The devices optimised
for S1 conditions could not reach the winter requirements in most cases considering a maximum f=10 Hz (Fig.
3a). The performances obtained in case S2 are significantly better (Fig 3b), accordingly to the wider range of
Curie temperatures included in the regenerator defined for the S2 case. The performance of a current vapour-
compression system has been also been represented for reference.
Figure 3. AMRR electric consumption and mass. a) Optimal S1 devices. b) Optimal S2 devices.
On the contrary to the summer cases, in case W2 the 14.7% savings in the heating demand achieved with the
reduction in the ventilation load did compensate for the 2.8 K increase in ΔTspan in all cases, despite the MCMs
were not adapted to the winter working temperatures. Since W2 working conditions were realized with less
power demand by the blowers, case W2 results even more efficient in global terms than case W1 if the devices
with the same composition are compared.
4. CONCLUSIONS
A 1.4 T permanent-magnet AMRR optimised for cooling mode could meet the cooling and also the heating
requirements of a vehicle with a regenerator made of plates between 0.03 and 0.04 mm thick and channels
from 0.02 to 0.04 mm width. The device would weigh between 20 and 50 kg and would work with an EER
between 2 and 4 and a COP (coefficient of performance in heating mode) between 2 and 3.5, similar to the
current vapour-compression technology. As the efficiency of the refrigerator increases, the energy
consumption of the electrical auxiliaries becomes more relevant. In order to minimize the total energy
consumption of a MAC system, all the components must be taken into account.
ACKNOWLEDGMENTS
B. Torregrosa-Jaime acknowledges the Spanish Ministry of Education, Culture and Sport (Ministerio de
Educación, Cultura y Deporte) for receiving the Research Fellowship FPU ref. AP2010-2160.
REFERENCES
[1] A. Kitanovski et al., Magnetocaloric Energy Conversion, Springer (2015).

[2] B. Torregrosa-Jaime, J. Payá, J.M. Corberán, “Application of magnetic cooling in electric vehicles”,
Science and Technology for the Built Environment In press, (2016).
[3] R. Bjørk, A. Smith, C.R.H. Bahl, N. Pryds, “Determining the minimum mass and cost of a magnetic
refrigerator”, International Journal of Refrigeration 34, 1805 (2011).
[4] A. Tura, A. Rowe, “Concentric Halbach cylinder magnetic refrigerator cost optimization”, International
[5] R. Bjørk, C.R.H. Bahl, K.K. Nielsen, “The lifetime cost of a magnetic refrigerator”, International Journal
of Refrigeration 63, 48 (2016).
[6] ICE Project FP7, Grant Agreement no. 265434, www.ice-mac-ev.eu (2014).
247
A CASCADING MODEL OF AN ACTIVE MAGNETIC REGENERATOR
SYSTEM
M. Tahavori(a), K. Filonenko(a)*, C. T. Veje(a), T. Lei(b), K. Engelbrecht(b), C. R. H. Bahl(b)

(a)
The Maersk Mc-Kinney Moller Institute, University of Southern Denmark, Campusvej 55, DK – 5230
Odense, Denmark
(b)
Department of Energy Conversion and Storage, Technical University of Denmark, Frederiksborgvej 39,
DK – 4000 Roskilde, Denmark
*
Corresponding author. E-mail: kfi@mmmi.sdu.dk
ABSTRACT
In recent years, significant amounts of studies have been done on modeling and analysis of active magnetic
regenerators (AMRs). Depending on the AMR geometry and the magnetocaloric material being modeled, the
AMR may not be able to provide the required performance demanded by practical applications. Some AMR
models in the literature predict high performance but with relatively low temperature spans at either end of
the AMR. Therefore, they may not be sufficient for practical applications, such as providing the heat
exchanger temperature spans required for residential and commercial space air conditioning. To remedy this,
one solution is cascading of multiple single layer AMRs. In this work, a cascading AMR model is presented
and studied. In a cascade configuration, N number of single layer AMRs are connected. The results show that
higher hot and cold side temperature differences may be achieved compared to the ones obtained with a
single AMR rendering the solution more suitable for use in residential and commercial space conditioning.
Keywords: cascading model, active magnetic regenerator, magnetocaloric material, temperature span,
practical applications.
DOI: 10.18462/iir.thermag.2016.0194
1. INTRODUCTION
Modeling and analysis of AMRs have received a lot of attention recently [1-6]. By reversing the refrigeration
cycle, the resulting device can be operated as a heat pump with improved residential and commercial space-
heating capabilities. Our primary focus is a particular implementation of the latter device depicted in Fig. 1a,
however, the results can be useful for refrigeration applications also.
The AMR itself, as the main component of the heat pump, requires a detailed mathematical model. Here we
adopt the 1D transient model developed in [2], where it is applied to several AMR geometries. Generally
such models predict relatively low temperature differences on both hot and cold sides of AMR made of Gd
spheres and utilizing water with 25 % ethylene glycol as a heat transfer fluid, Figs. 1b and 1c. In building
applications such low temperature differences will require unrealisticly high massflows to provide the
required capacity. This will hamper the system performance through increased viscous dissipation and
pressure loss. Depending on the geometry of the magnetocaloric material, the AMR may not be able to
provide sufficient coefficient of performance (COP), heating power, temperature spans, etc. required for
building heating. One way to increase temperature differences and enhance AMR performance would be to
use first order materials with higher magnetocaloric effect [5].
In this paper, we propose a different solution: cascading of multiple single-layer AMRs. We show that the
requirements for the conventional building heating can be met in cascading geometry even without layering
of the magnetocaloric material. Magnetocaloric cascading systems is a subject of active research, but
concerns mainly AMRs connected in series [6-8]. Such configuration leads to a larger temperature span
across the device. Our goal, on the other hand, is to demonstrate an increase in temperature difference on
either side of the system. For this purpose, we present a modified configuration detailed in the next section.
2. CASCADING AMR MODEL
A single-layer AMR 1D model developed in [2], is modified such that it can be used in a cascade system
248
configuration. Fig. 1b shows a modified 1D AMR with TCi and TCo ( THi and THo ) as inlet and outlet
! = ρ f V! as the mass flow rate
temperatures of the cold (hot) side, respectively, f as the AMR frequency and m
defined through volumetric flow rate V! and fluid density ρ f . As opposed to the temperature span across the
regenerator, THi − TCi , we will call the difference between the input and the output temperatures at cold (hot)
end, ΔTC = TCi − TCo (t) (ΔTH = THo (t) − THi ) , a cold (hot) end span. In the case of multiple regenerators, the
temperatures are enumerated by index n, e.g. ΔTH = THo(n) (t) − THi(n).
The cascading system shown in Fig. 2a consists of N number of single-layer modified 1D AMRs.
Neighbouring regenerators are connected through their hot ends and separately through their cold ends in
contrast to hot-to-cold and cold-to-hot connections for AMRs in series [8]. The total COP at the hot side of
the cascading AMR model is defined as COP = QH / (QH − QC ) with QX = QX,1 + QX,2 +... + QX,N where
1
QX,n = ! c
∫ m(t) f ΔTX dt have the meaning of cooling load and heating power of the n'th regenerator for
τ
X ≡ C and X ≡ H , respectively. Here, τ is AMR cycle time and c f is fluid heat capacity.
Figure 1. a) Magnetocaloric heat pump: the cold and hot ends are connected to the ambient and residential/commercial
building, respectively; b) geometry of AMR made of Gd spheres; c) modified 1D AMR model to be implemented in a
cascading configuration.
3. TEMPERATURE SPAN IN AMR CASCADING
In order to demonstrate the possibility of the substantial increase of the temperature span in a cascading
configuration in comparison to the single regenerator model, a preliminary study is described in this section.
We fix volumetric flow rate in all regenerators at V! = 2.23 m3/hr corresponding to a heating power of
approximately 2 kW for a single regenerator. By adding consecutively identical regenerators to the system,
we follow in Fig. 2b how the cold and hot end temperature differences change with number of regenerators.
The calculations are made for two regenerator masses: 33.6 kg and 50.2 kg. Obviously, the total mass values
are considerably larger compared to those used in conventional heat pumps. In subsequent research the
required mass can be reduced through optimization with respect to both material properties, e.g. by layering
each regenerator, and to device specifications, e.g. by varying operating frequency and regenerator aspect
ratio. Here we are interested in illustrating the concept of cascading and therefore not concerned with
249
keeping the regenerator size small. From Fig.2b the expected growth of temperature spans is observed: for a
six-regenerator cascading system, ΔTH is in the range of 4-5 K, whereas ΔTC has a value between 3 K and
4 K. This comes together with high enough heating power QH =1.6 kW and COP=3-4, which makes the six-
regenerator cascade a viable device for building heating.
Figure 2. Cascading AMR model: a) schematic drawing of a cascading system, b) temperature span vs. number of
AMRs which are all identical and have masses 33,6 kg or 50,2 kg each with V! fixed at 2.23 m3/hr in all regenerators.
4. MATCHING RESIDENTIAL HEATING REQUIREMENTS
With respect to in-house heating conditions we assumed that:

1. The amount of heating power which is required in a residential/commercial building is about 2 kW;
2. We aim for COP = 5;
3. The amount of temperature difference which is required on the hot side of AMR is about 4-5 K.
In this section, we limit ourselves to three regenerators and to moderate temperature spans below 3 K
keeping in mind that the analysis can be generalized by increasing the number of regenerators. This allows
us to match the system parameters for the first two heating requirements, while keeping computation time
relatively small.
Figure 3. Optimization of AMR cascading for N = 1, 3: a) COP , b) volumetric flow rate and b) temperature spans vs.
total cascading system mass.
Parameters of the cascading system with one and three AMRs are analyzed in Fig. 3 as functions of total
system mass mtot under condition that the overall heating power has a fixed value QH = 2 kW in all
calculations. All regenerators in a cascading system are assumed to be identical and have the same aspect
ratio AR = 6.6 independent of the number of regenerators and mtot . Constant AR allows to relate the
regenerator length L and cross-section area Ac to those of an arbitrary regenerator chosen as a reference. We
250
choose the reference regenerator with mass mreg = 2.8 kg, length L0 = 100 mm, cross-section area A0 =
229.25 mm2 and geometry of regenerators inside the cascading system is given by Lreg = L0 (M / N )1/3 and
Areg = A0 (M / N )2/3 , where M = mtot / m0 . Red and blue curves in Fig. 3 correspond to N = 1 and N = 3,
respectively, and the dashed line in Fig. 3a represents the requirement of COP = 5. It can be concluded that
the COP gets and stays larger for N>1, which is also supported by the calculations made for N > 3 (not
shown here). Beyond total mass values approximately equal to 25 kg COP gets larger than 5 thereby
fulfilling the second requirement for building heating. As is expected, the flow rate decreases and
temperature difference increases with increasing total mass and number of regenerators.
5. CONCLUSIONS
Through numerical analysis of AMR cascades we showed that cascading increases both temperature
difference and COP showing the possibility for fulfilling the requirements on heating power, COP and
temperature span imposed by the residential/commercial heating. Unlike in non-optimized case discussed in
Sec. 3, the 3-regenerator cascade discussed in Sec. 4 needs much less total cascade mass compared to the
single-regenerator case to achieve the same (and even higher) COP. Although we have fulfilled first two
requirements separately from the third, we have shown that the temperature span increases together with
number of regenerators. Therefore, it is possible to fulfil all three requirements by performing time-
consuming calculations.
Authors realize that the regenerator masses used in the paper are considerably larger than the masses of
currently used AMR prototypes and produced heat pumps. However, the goal in the present manuscript was
to explore methodology and check the hypothesis concerning the influence of cascading on AMR
performance and temperature span. Therefore, the single-layer Gd was used to simplify consideration. In
future research, this issue will be addressed and the required mass can be sufficiently reduced by layering the
AMRs in the cascading geometry. The cascading technique described in this paper will allow a further
increase of efficiency and temperature span of AMR and in combination with other techniques will bring
closer the industrial realization of commercial AMR- based devices.
ACKNOWLEDGMENTS
We are grateful for financial support provided by the ENOVHEAT project funded by Innovation Fund
Denmark (contract no 12-132673).
REFERENCES
[1] A. Kitanovski, P. W. Egolf, “Thermodynamics of magnetic refrigeration”, International Journal of

[2] K. Engelbrecht, “A Numerical Model of an Active Magnetic Regenerator Refrigerator with Experimental
Validation”, Ph.D. thesis, University of Wisconsin (2008).
[3] K. K. Nielsen et al., “Review on Numerical Modeling of Active Magnetic Regenerators for Room
Temperature Applications”, International Journal of Refrigeration 34, 603 (2011).
[4] A. Rowe, “Thermodynamics of active magnetic regenerators”, Cryogenics 52, 111 (2012).
[5] T. Lei et al., “Sensitivity study of multi-layer active magnetic regenerators using first order
magnetocaloric material La(Fe,Mn,Si)13Hy”, Journal of Applied Physics 118, 014903 (2015).
[6] P. J. Shirron, “Applications of the Magnetocaloric effect in Single-Stage, Multi-Stage and Continuous
Adiabatic Demagnetization Refrigerators”, Cryogenics 62, 130 (2014).
[7] P. Shirron, “Optimization strategies for single-stage, multi-stage and continuous ADRs”, Cryogenics 62,
140 (2014).
[8] P.W. Egolf, A. Kitanovski, D. Vuarnoz, M. Diebold, C. Besson, “An introduction to magnetic
refrigeration” [Online]. Available: http://www.centrogalileo.it/nuovapa/Articoli%20tecnici/INGLESE
CONVEGNO/XII%20Convegno%20English/II%20SESSIONE/Egolf%201_eng.pdf, (2008).
251
GIANT MAGNETOCALORIC EFFECT OF COMPRESSIBLE ISING AND
HEISENBERG LATTICES
J. S. Amaral(a,b)*, N. M. Fortunato(a), C. O. Amorim(a), J. N. Gonçalves(a), V. S. Amaral(a)
(a)
CICECO - Aveiro Institute of Materials, Department of Physics, University of Aveiro, 3810-193 Aveiro,
Portugal
(b)
IFIMUP and IN-Institute of Nanoscience and Nanotechnology, 4169-007 Porto, Portugal)
*
Corresponding author. E-mail: jamaral@ua.pt
ABSTRACT
Since the discovery of giant magnetocaloric materials, it has become clear that magnetovolume coupling can
lead to a large increase to a material's magnetocaloric effect, and so, to high-performance magnetic
refrigerants. A simple and computationally inexpensive approach to gauge magnetovolume effects is via the
Bean-Rodbell model, where a Curie temperature dependence on volume is included in the Weiss mean-field
model. The thermodynamic properties of giant magnetocaloric materials are described, namely discontinuous
magnetization curves and large entropy change values. Moving to a microscopic model is a natural step to
better characterize giant magnetocaloric materials. In this work we show how an Ising/Heisenberg-type
interaction together with elastic/magnetovolume coupling can be solved by a Monte-Carlo method,
reproducing the thermodynamic properties of giant magnetocaloric systems. The low computational cost of
this approach and the possibilities of simulating systems using Density Functional Theory calculated
magnetic interaction parameters allow the computational study and design of (giant) magnetocaloric
materials.
Keywords: Entropy change, magnetovolume coupling, Mean-Field, Monte-Carlo simulations.

DOI: 10.18462/iir.thermag.2016.0200
1. INTRODUCTION
The performance of magnetic refrigeration devices depends heavily on the magnetocaloric effect (MCE)
intensity of the active magnetic refrigerant. The discovery of the giant magnetocaloric effect (GMCE) [1] led
to a renewed interest in this technology, now in its starting steps of commercialization following intense
research in new high performance materials and devices [2]. A strong coupling between magnetic and lattice
degrees of freedom is the underlying mechanism behind the GMCE, and the need for a fundamental
understanding of this effect led to the use of various theoretical models. The Bean-Rodbell model [3], in
particular, was successfully used in the study of several families of GMCE materials, including Gd-Si-Ge
[4], Mn-Fe-P-As-Si [5,6], La-Fe-Si [7] and La-Ca-Mn-O based manganites [8], and recently on general
properties of GMCE materials [9,10]. The Bean-Rodbell model consists on considering the Weiss molecular
mean-field model, together with an explicit linear dependence of the Curie temperature TC on volume:
  v  v0 
T C  T 0 1   



, (1)
  v0 
where β is a positive and constant factor of proportionality, v volume, v0 equilibrium volume in the absence
of magnetic interactions and T0 the Curie temperature for v = v0. For a given magnetic system with K
compressibility, the η parameter is defined as:
S S  1 
2
  40 Nk B
KT 0 
2
, (2)
2 S  1  1
4
where S is spin, N the spin density and kB the Boltzmann constant. For 0 < η ≤ 1 the paramagnetic (PM) to
ferromagnetic (FM) transition is of the second-order, while for η > 1 it is of the first-order, increasing the
isothermal magnetic entropy change ΔSM, particularly near TC.
252
The popularity and usefulness of the Bean-Rodbell model is due to its semi-quantitative validation of the
overall thermodynamic properties of a GMCE material, and also to its simplicity and low computational cost.
For reference, Bean-Rodbell model simulations of thousands of M(H,T) data points of a GMCE system
typically take minutes to run in a standard personal computer using available software packages, such as the
Mean Field Simulation Suite (MFSS) [11]. Still, the correlation between the microscopic properties of a
given MCE material (atomic types, structure) is difficult to gauge in a mean-field approximation. Typically,
an average spin value is used, and the remaining model parameters values, T0 and η, are chosen to replicate
experimental results. In short, it is possible to interpret experimental magnetic and magnetocaloric properties
of a given GMCE material, but difficult to predict them using a mean-field model.
A microscopic approach would allow a more realistic description of GMCE materials, provided the
thermodynamics of the first-order phase transition are adequately described. A predictive approach to
magnetocaloric performance could also be envisaged, as microscopic parameters can be estimated by well-
established Density Functional Theory (DFT) methods. To our knowledge, there are few quantitative
combined DFT + Monte-Carlo (MC) approaches to study MCE materials. A mixed-valence rigid Heisenberg
lattice study of La-Ca-Mn-O manganites [12], with previously calculated J values [13] has led to comparable
results with experiment, albeit considering no magnetovolume coupling, which is known to be relevant in
this system [14]. Co and In doped Ni-Mn-Ga shape memory alloys were studied via the use of a Potts
Hamiltonian with 2S + 1 possible spin projections, in accordance with the total spin of each magnetic atom,
and the three-state Blume-Emery-Griffiths model, which allows for a structural transformation from the
cubic (austenitic) phase to the tetragonal (martensitic) phase [15]. The magnetocoupling interaction was
taken into account via two parameters with values chosen in order to reproduce the experimental martensitic
transformation temperature and magnetization behavior in different magnetic fields. Simulated results
compared qualitatively well with experiment.
A more widespread use of microscopic models in the study of MCE materials would be achieved by
overcoming two main hurdles: the complexity of the models usually employed and the computational costs
of calculations. In this work we propose an Hamiltonian, following the ideas of Domb [16], which follows
essentially the same assumptions as the Bean-Rodbell model (section 2). Calculations of relevant
thermodynamic properties are performed via the use of a MC method (section 3), that drastically reduces
computational costs of simulating M(H,T) data, when compared to the more widely used Metropolis [17]
algorithm.
2. MAGNETOVOLUME COUPLING MODEL
For a simplified description of magnetovolume interactions, we start with a typical Ising/Heisenberg

interaction, where the magnetic interaction between neighboring spins depends on their interatomic distance,
a volume energy potential and interaction with external field:
1 1
 J ( v ) S  , (3)
2
   i
S j
Kv  MH
2 i, j 2
where J is the magnetic exchange parameter between Si and Sj spins, K compressibility, M magnetization and
H applied magnetic field. For a direct comparison with Bean-Rodbell model calculations, we here assume
that the J(v) dependence is linear, with positive slope. Explicitly, J(v)=J0+J'(v-v0)/v0, where J0 is the
magnetic exchange parameter for v = v0 and J' a positive constant. Fig 1 illustrates the energy potentials in
play:
253
Figure 1. Energy potentials of eq. 3: magnetic exchange parameter dependence on volume (solid line), volume energy
potential (dashed line) and total volume and magnetic energy potential (dotted line).
In the PM state, the system will tend to be near its structural energy minimum (v ~ v0), as magnetic energy
will be low. In contrast, the FM state will energetically favor a volume v > v0, as the magnetic energy
contribution is strong and it becomes favorable for the system to have a higher J value, and consequently a
larger v value. In analogy to the Bean-Rodbell model, for a sufficiently high magnetovolume coupling, the
system should discontinuously change M and v at TC. While solving the Ising/Heisenberg model Hamiltonian
with standard MC methods such as the Metropolis algorithm is now relatively straightforward, its
computational cost is considerable. Each M(H,T) data point requires a converged calculation. Also, data for T
~ TC, which is in the area of interest for MCE materials, requires longer convergence times, or the use of
more complex approaches (e. g. cluster models [18-20]) to avoid noise due to critical slowing down.
By introducing the magnetovolume term, the use of Metropolis method becomes more computationally
expensive. Both v and M parameters will require adequate convergence, which will not happen without the
introduction of an extra pressure term [21]. To avoid the use of this term, and also to reduce the overall
computational needs for the required data simulations, we propose the use of MC methods that determine the
Joint Density of States (JDOS), such as the Wang-Landau [22], transition matrix [23] and Random Path
Sampling (RPS) [24] methods. From (J,H,T)-independent JDOS calculations, which are done only once for
each lattice type, all relevant (J,H,T)-dependent thermodynamic quantities are obtained from either the
partition function or free energy minima. In terms of computational cost, this is a much more favorable
approach compared to individual M(H,T,J) converged calculations as required by the Metropolis method.
Due to its simple implementation and parallelization, we here employ the RPS method for JDOS
calculations. The methodology and implementation is briefly discussed in the next section.
3. MONTE-CARLO METHOD
For obtaining a reliable JDOS estimate of a system with a large number of possible Energy-Magnetization
states, an efficient statistical sampling of these states is required. A simple approach is the RPS method
which, by construction, results in unbiased, flat M sampling by sweeping from the -1 total M to the +1 total
M configuration via locally flipping spins from the -1 to the +1 state, from a random sequence. Energy
values are evaluated locally, much like in the Metropolis method, ensuring a fast update of total energy
value. Each complete RPS sweep corresponds to flipping all spins in the lattice. Depending on model type
and system size, for a number of sweeps that is orders of magnitude below the total number of possible
states, converged estimates of JDOS (and consequently of the partition function and free energy) are
achieved. For discrete spin models, such as the Ising model, the implementation is straightforward, while for
continuous spin model (Heisenberg) we here approximate the continuous spin direction degree of freedom to
14 total vector directions, which is larger than the minimum of 12 shown to be able to describe the critical
behavior of the model [25]. Convergence of JDOS calculations is evaluated by the overall smoothness of free
energy F(H,T) at zero field, particularly for T ~ TC. Data points shown in this work are taken from the
absolute F minimum.
254
4. RESULTS
We compare the mean-field Bean-Rodbell model simulations of second- and first-order FM-PM transitions,
from changing η values from 0 up to 2, while in the case of Ising and Heisenberg model, the J dependence on
volume slope changes from 0 to 3.5, considering a K value of 50. Fig. 2 (left) shows the zero-field M vs T
behavior of the Bean-Rodbell model and a 128 spin compressible BCC Ising lattice.
Figure 2. (left) Spontaneous magnetization and (right) isothermal magnetization calculations of Bean-Rodbell (top) and
microscopic (bottom) models, for rigid and compressible systems with increasing magnetovolume coupling. Dashed
lines are eye-guides.
Both the Bean-Rodbell and compressible Ising systems show qualitatively the same behavior, in terms of the
transition becoming first-order for sufficiently high magnetovolume coupling, as seen through the increase in
TC and the discontinuity of the M vs. T curve. In terms of isothermal M versus H behavior, for sufficiently
high magnetovolume coupling, the Bean-Rodbell model shows metamagnetic-like behavior (inflexions in
M(H)) and for stronger magnetovolume coupling, discontinuities up to a certain critical field. Fig. 2 (right)
shows the M(H) dependence of a first-order Bean-Rodbell system, together with a compressible HCP
Heisenberg system.
The metamagnetic-like isothermal M vs. H behavior of the Bean-Rodbell model is also qualitatively
reproduced by the compressible Heisenberg model, following the Hamiltonian of eq. 3. Taking into account
the two previous results, we should also see a large increase in the MCE when comparing rigid and
compressible lattices. Fig. 3 compares the isothermal entropy change of Bean-Rodbell and compressible
HCP Heisenberg systems.
Figure 3. Isothermal magnetic entropy change ΔSM vs T of Bean-Rodbell (left) and HCP Heisenberg (right) models, for
rigid and compressible systems with increasing magnetovolume coupling. Dashed lines are eye-guides.
255
The same overall thermodynamic behavior is observed between the two models, with similar increases in
entropy change and changes in the curve shape with increasing magnetovolume coupling. The necessary
ingredients to study GMCE materials are present in the Hamiltonian of Eq. 3.
5. COMPUTATIONAL COSTS
As stated previously, a more widespread use of combined DFT and MC approaches for MCE materials
studies will benefit from low computational cost of simulations. By using JDOS-based methodologies, such
as the RPS method here employed, the bulk of intensive calculations are done for obtaining JDOS itself. In
this work, we used previously calculated JDOS for all Ising and Heisenberg lattices here under study. The
data shown on Figs. 2 and 3 took minutes to simulate using a standard personal computer. Other (H,T)-
dependent thermodynamic properties, such as free energy and volume were also calculated. The Bean-
Rodbell simulations here shown, performed using the freely available MFSS software, took a comparable
amount of time to be performed. For a combined DFT and MC approach, the computational time required for
DFT calculations of magnetic interaction J parameters is naturally relevant, and will depend on the DFT
package used. Nonetheless, both the SPR-KRR [26] and openmx [27] packages, which employ the
Liechtenstein method for J value calculations [28], are sufficiently fast for systematic calculations in various
structures and compositions, as recently explored [15].
6. CONCLUSIONS
We have here proposed a simple Hamiltonian, that includes a magnetic interaction between spins together
with a lattice potential and volume-dependent magnetic coupling. Through Monte-Carlo computations of the
joint density of states of Ising and Heisenberg lattices, magnetic and magnetocaloric properties were
simulated with increasing magnetovolume coupling strengths and compared to simulations of the widely
used Bean-Rodbell mean-field model. We have found that the overall thermodynamics of first-order phase
transitions are represented in the model here proposed, making it applicable to the study of giant
magnetocaloric materials. The low computational cost of the calculations here performed is relevant for
widespread use, while a proposed combination with density functional theory can provide the basis for
predictive computational materials design of high-performance magnetic refrigerants.
ACKNOWLEDGMENTS
This work was developed within the scope of the projects CICECO-Aveiro Institute of Materials, POCI-01-
0145-FEDER-007679 (FCT Ref. UID/CTM/50011/2013) and FCT CERN/FIS-NUC/0004/2015, financed by
national funds through FCT/MEC(PIDDAC) and when appropriate co-financed by FEDER under the
PT2020 Partnership Agreement. FCT grants SFRH/BPD/111270/2015 (J. S. Amaral),
SFRH/BPD/82059/2011 (J. N. Gonçalves) and SFRH/BD/93336/2013 (C. O. Amorim) are acknowledged.
REFERENCES
[1] V. K. Pecharsky, K. A. Gschneidner Jr., Giant magnetocaloric effect in Gd5Si2Ge2, Phys. Rev. Lett. 78
(1997) 4494.
[2] K. A. Gschneidner Jr., V. K. Pecharsky, Thirty years of near room temperature magnetic cooling: Where
we are today and future prospects, Int. J. Refrig. 31 (2008) 945961.
[3] C. P. Bean, D. S. Rodbell, Magnetic disorder as a first-order phase transition, Phys. Rev. 126 (1962) 104.
[4] P. J. von Ranke, N. A. de Oliveira, S. Gama, Understanding the influence of the first-order magnetic
phase transition on the magnetocaloric effect: application to Gd5(SixGe1−x)4, J. Magn. Magn. Mat. 277 (2004)
78–83.
[5] G. F. Wang, L. Song, Z. Q. Ou, Z. R. Zhao, O. Tegus, Calculation of the magnetization and
magnetocaloric effect in the MnFeP0.45As0.55 compound, Acta Metallurgica Sinica (English Letters) 20
(2007) 265269.
256
[6] M. Balli, D. Fruchart, D. Gignoux, S. Miraglia, E. K. Hlil, P. Wolfers, Modelling of the magnetocaloric
effect in Gd1−xTbx and MnAs compounds, J. Magn. Magn. Mat. 316 (2007) E558–E56.
[7] J. D. Zou, B. G. Shen, J. R. Sun, Role of lattice contraction in the magnetocaloric effect in LaFe11.5Si1.5, J.
Phys.: Condens. Matter 19 (2007) 196220.
[8] J. S. Amaral, N. J. O. Silva, V. S. Amaral, A mean-field scaling method for first- and second-order phase
transition ferromagnets and its application in magnetocaloric studies, Appl. Phys. Lett. 91 (2007) 172503.
[9] J. S. Amaral, V. S. Amaral, Disorder effects in giant magnetocaloric materials, Phys. Stat. Sol. A 211
(2014) 971–974. [10] A. Smith, K. K. Nielsen, C. R. H. Bahl, Scaling and universality in magnetocaloric
materials, Phys. Rev. B 90 (2014) 104422.
[11] J. S. Amaral, Mean-Field Simulation Suite. URL http://sweet.ua.pt/jamaral/
[12] O. Pavlukhina, V. Buchelnikov, V. Sokolovskiy, M. Zagrebin, Monte carlo study of the magnetic and
magnetocaloric properties of La1−xCaxMnO3 (x = 0.33 and 0.5), Sol. St. Comm. 190 (2012) 347.
[13] E. Restrepo-Parra, C. D. Salazar-Enríquez, J. Londoño Navarro, J. F. Jurado, J. Restrepo, Magnetic
phase diagram simulation of La1−xCaxMnO3 system by using monte carlo, metropolis algorithm and
heisenberg model, J. Magn. Magn. Mat. 323 (2011) 1477.
[14] J. A. Turcaud, A. M. Pereira, K. G. Sandeman, J. S. Amaral, K. Morrison, A. Berenov, A. Daoud-
Aladine, L. F. Cohen, Spontaneous magnetization above TC in polycrystalline La0.7Ca0.3MnO3 and
La0.7Ba0.3MnO3, Phys. Rev. B 90 (2014) 024410.
[15] V. Sokolovskiy, A. Grünebohm, V. Buchelnikov, P. Entel, Ab initio and monte carlo approaches for the
magnetocaloric effect in Co- and In-doped Ni-Mn-Ga heusler alloys, Entropy 16 (2014)
4992–5019.
[16] C. Domb, Specific heats of compressible lattices and the theory of melting, J. Chem. Phys. 25 (1956)
783.
[17] N. Metropolis, A. W. Rosenbluth, M. N. Rosenbluth, A. H. Teller, E. Teller, Equation of state
calculations by fast computing machines, J. Chem. Phys. 21 (1953) 1087.
[18] R. H. Swendsen, J. Wang, Nonuniversal critical dynamics in monte carlo simulations, Phys. Rev. Lett.
58 (1987) 86–88.
[19] U. Wolff, Collective monte carlo updating for spin systems, Phys. Rev. Lett. 62 (1989) 361.
[20] J. R. Heringa, H. W. J. Blöte, Geometric cluster monte carlo simulation, Phys. Rev. E. 57 (1998) 4976.
[21] M. P. Aleen, D. J. Tildesley, Computer Simulation of Liquids, Oxford University Press, 1987.
[22] F. Wang, D. P. Landau, Efficient, multiple-range random walk algorithm to calculate the density of
states, Phys. Rev. Lett 86 (2001) 2050.
[23] A. L. C. Ferreira, R. Toral, Projected single-spin-flip dynamics in the ising model, Phys. Rev. E 76
(2007) 011117.
[24] J. S. Amaral, J. N. Gonçalves, V. S. Amaral, Thermodynamics of the 2-D Ising model from a random
path sampling method, IEEE Trans. Magn. 50 (2014) 1002204.
[25] A. Margaritis, G. Odor, A. Patkós, Sequence of discrete spin models approximating the classical
Heisenberg ferromagnet, J. Phys. A: Math. Gen. 20 (1987) 1917.
[26] H. Ebert et al., The Munich SPR-KKR package. URL http://ebert.cup.uni-muenchen.de/SPRKKR
[27] T. Ozaki et al., Openmx (open source package for material explorer). URL http://www.openmx-
square.org/
[28] A. I. Liechtenstein, M. I. Katsnelson, V. P. Antropov, V. A. Gubanov, Local spin density functional
approach to the theory of exchange interactions in ferromagnetic metals and alloys, J. Magn. Magn. Mat. 67
(1987) 65–74.
257
REVERSIBLE AND NON REVERSIBLE MAGNETOCALORIC EFFECT IN
PHASE SEPARATED MANGANITES
M. Quintero(a,b)*, D. Garcia Reparaz(a), D.Goijman(a), I. Irurzun(a), S. Passanante(a), A. G.

Leyva(a,b) J. Sacanell(a)
(a)
Departamento de Física de la Materia Condensada, GIyA, GAIANN, Comisión Nacional de Energía
Atómica, Av. Gral. Paz 1429 (1650), Buenos Aires, Argentina.
(b)
Escuela de Ciencia y Tecnología, Universidad Nacional de General San Martín, Martín de Irigoyen No.
3100 (1650), Buenos Aires, Argentina
*
Corresponding author. E-mail: mquinter@cnea.gov.ar
ABSTRACT
In this work we present a study of the magnetocaloric effect in phase separated manganites, which
display the intrinsic coexistence of ferromagnetic and non-ferromagnetic phases. In particular, we
focus on the influence of phase separation on the magnitude of the effect, analysing the adiabatic
change of temperature and the isothermal change of enthalpy, obtained by direct and indirect
methods, respectively. Our results show a superposition of reversible and non-reversible processes
during magnetic field induced transitions. We further present a model which is in well agreement
with the experimental data.
Keywords: Magnetocaloric, virgin, phase separation, manganites

DOI: 10.18462/iir.thermag.2016.0205
1. INTRODUCTION
The magnetocaloric effect (MCE) is characterized by the isothermal change of the magnetic entropy and the
adiabatic change of temperature, induced by the application of an external magnetic field [1]. Due to the
possibility of performing cycles of heat absorption and release just by varying a magnetic field, materials
exhibiting MCE are considered as an alternative to develop solid state refrigerators[2].
Since the discovery of giant MCE in Gd based compounds [3] and due to the high cost of production of these
materials, significant effort has been oriented to the search of alternative systems with large MCE and low
cost of production. Among those systems, the compounds known as manganites [4] are highly promising[5-
9]. As a consequence of a strong coupling between their magnetic, electronic and structural degrees of
freedom, some manganites display the phenomenon of phase separation (PS), characterized by an intrinsic
coexistence of ferromagnetic (FM) and non-ferromagnetic (nFM) phases. Thus, the analysis of MCE in
manganites is more complex than the typical one, just based on the use of the thermodynamic Maxwell
relations [1]. In fact, it has shown that the use of Maxwell relations to analyze MCE in PS manganites, can
either overestimate or underestimate their actual entropy change [8]. To overcome this drawback, we have
previously developed a phenomenological model which allows us to calculate the MCE of manganites with
PS [11].
The complexity of the system, obviously influences its magnetocaloric response. As, for example, has been
recently observed in Fe49Rh51, in which a different behavior is obtained in successive magnetic loops [12].
In this work we observed a similar behavior in La5/8-yNdyCa3/8MnO3 with y = 0.3 (LNCMO), a prototypical
manganite exhibiting PS [13].
We have performed magnetization and specific heat measurements to calculate MCE, and a differential
thermal analysis (DTA) for its direct measurement. We present direct measurements of the temperature
change and compare it with the results obtained from the enthalpy change, considering the phase separated
state as a non-interactive coexistence of FM and nFM phases.
2. MAIN SECTION
2.1 Experimental procedure
Polycrystalline samples of LNCMO were synthesized following a citrate/nitrate decomposition methods
using high purity reactants. The crystalline structure was obtained by X-ray powder diffraction which was
also used to verify the absence of undesired phases.
258
Magnetization (VSM mode) and specific heat (Cp) measurements were performed in a Quantum Design
VersaLab. We performed Differential Thermal Analysis (DTA) measurements with a homemade system,
formed by two Cernox CX-1080-SD thermometers (manufactured by Lake Shore Cryotronics) on a Teflon
piece to ensure thermal insulation between the sample and the reference thermometers [14]. We used a piece
of alumina as reference. The whole system was mounted in a VersaLab transport puck, in order to measure
the electrical resistance of the thermometers.
Thus, we performed magnetization, Cp and DTA measurements in the same range of magnetic field and
temperature.
2.2 Results and discussion

In figure 1 we present the magnetic phase diagram of the LNCMO sample. At high temperature the sample is
in a paramagnetic insulating (PI) state. Around 200K occurs a transition to a charge ordered phase (CO),
characterized by a periodic array of Mn +3 and Mn +4 ions. On further cooling, the ferromagnetic metallic
(FM) phase develops and coexist with the CO phase. This coexistence represent a significant portion of the
phase diagram (at low magnetic fields and intermediate temperatures) and an homogeneous FM state is
obtained at low temperatures or applying an external magnetic field of around 1T.
3.0
2.5
FM
Magnetic Field (T)
2.0
FM
1.5
1.0
CO PI
0.5
FM + CO
0.0
50 100 150 200 250
Temperature (K)
Figure 1: magnetic phase diagram of the sample LaNd, indicating the presence of the Ferromagnetic Metallic (FM),
Charge Ordered (CO), and Paramagnetic Insulating (PI) phases.
The localization of charges provokes a deformation in the oxygen octahedral that surrounds the Mn+3 ion due
to the Jahn Teller effect. Thus, the crystalline structure of the CO and the FM phases slightly differs, but not
their volume [15].
In figure 2 we present the temperature dependence of the adiabatic changes of temperature, measured by
DTA, induced by the application of magnetic field. We show the adiabatic change of temperature
corresponding to the first application of an increasing magnetic field from 0 to 3 Teslas (TADvir, where "vir"
stands for virgin state), the reduction of the magnetic field from 3 T to 0 (TADDOWN) and the subsequent
increase of the magnetic field from 0 to 3 Teslas (TADUP).
A clear difference can be observed between the TADvir with respect to that obtained during the next
application of the field (TADUP). It is also interesting to note the symmetry between TADUP and TADDOWN.
In the inset of figure 2 we show that the difference between TADvir and TADDOWN displays two clear peaks.
One around 90K, that is associated with transition to the low temperature FM homogeneous phase on
cooling. The other, around 170 K, which coincides with the temperature region where the CO phase forms in
the PI matrix.
The difference presented in the inset of figure 2, means that a significant MCE (as compared to that observed
in subsequent cycles) is observed for the virgin state.
We propose the following image to explain the "virgin effect" observed while applying the magnetic cycles
to our samples, that resembles an analogue effect observed on the thermal history of (Mn,Fe)2(P,Si) [16].
After cooling the sample under zero field, it presents a particular metastable state characterized by the
coexistence of amounts of CO and FM phases dependent on the thermal and magnetic history of the sample.
That state disappears when the first application of the magnetic field transforms the entire sample in FM and
leaves the sample in a stable state in which the amounts of the CO and FM phases are different than those
corresponding to the initial state.
To analyze the MCE during the application of the magnetic field, we considered the enthalpy change of the
sample during the process. We assume that the enthalpy of the sample is the result of the contribution of the
259
enthalpy of both phases with no interaction between them. Then, the total enthalpy can be written as
E  xE FM  (1  x ) E NF , being EFM and ENF the enthalpy of the magnetic (FM) and non-magnetic (CO)
phases respectively, and x the ferromagnetic fraction.
UP
0,8
-TAD
0,6 2,0
0,4
VIR
TAD
0,2 1,5
0,0
50 100 150 200
1,0
T (K)
TAD(K)
0,5
0,0
VIR
TAD -0,5
UP
TAD
-1,0
DOWN
TAD
-1,5
40 60 80 100 120 140 160 180 200
Temperature(K)
Figure 2: adiabatic temperature change extracted during the first application of the magnetic field from 0 to 3 T
(ΔTADUP), decreasing the field from 3T to 0 (ΔTADDOWN) and increasing for a second time (ΔTADUP). In the inset, the
difference between ΔTADUP and ΔTADUP.
Assuming that the magnetic field do not affect the CO phase and only affects the FM by a linear term related
with the magnetic work, the enthalpy change while applying a magnetic field H is
 x M S 
dE   ( E 0  M S H )  xH dH (1)
 H H 
being E0 the entropy difference between the samples at zero field, and Ms the magnetization of the
homogeneous FM phase, which depends on T and H. Those values were extracted from Cp and
magnetization measurements. We extracted the FM fraction from magnetization measurements.
The total enthalpy was obtained by the integration of equation (1) during the application of an increasing H,
both in the first and in the subsequent cycle. The difference between the first and the following cycle is
presented in figure 3. A peak around 80 K is observed, near the range where the transition between the phase
separated state and the homogeneous FM state occurs.
To gain further insight, we compared the results obtained from the enthalpy change (indirect, as were
obtained from magnetization measurements) with those extracted from DTA measurements (direct
measurement of temperature). We used that information to calculate the heat released during the transition.
Under isobaric conditions, the released heat (Q) coincides with the enthalpy change of the complete
transition (from zero to the maximum value). So, with our results we can directly obtain the released heat in
the virgin and the subsequent cycle (Qvir and QUP, respectively). By subtracting Qvir and QUP we
estimated the "excess" of heat released during the first cycle defined as Qexc = Qvir - QUP (figure 3).
The same quantity can be obtained from DTA measurements, assuming that the entire process was done
under adiabatic conditions and that the adiabatic temperature change was produced by the heat released.
Then dQ=C.dTAD (C is the heat capacity of the sample) to calculate the excess of heat. We also present Qexc
obtained from DTA measurements in figure 3.
We observe a good agreement between both results in the low temperature region indicating that the
proposed enthalpy description is adequate to study the PS to FM homogeneous transition. However, in the
high temperature transition the released heat obtained by the direct measurement, is markedly different from
that described by the (indirect) enthalpy approach. The reason behind this difference, is related with the fact
that the high temperature state is characterized by a mixture of the CO and PI phases [6, 17].
In similar systems, the PI to CO phases present a behavior resembling to the martensitic transitions in
metalic alloys. Due to the structural mismatch between the CO and the PI phases a new metastable phase
appears between them. This phase was described as an insulator with an incomplete charge ordering, named
charge disordered insulator (CDI) [6, 17]. We did not include that feature in our enthalpy description,
because it is not possible to extract the relative fraction of the coexisting phases from magnetization
measurements (as both the CO and the PI display similar H dependence and thus cannot be separated).
It is worth to note that our approach thus serves to fully describe the MCE effect related to the heat released
(or absorbed) during a metamagnetic transition between an FM and non-FM phases, in contrast with the
260
classical approach using the Maxwell relations [1]. On the other hand, it fails to describe the heat released in
a transition between two non-magnetic phases.
form TAD
form Enthalpy
50
40
Q (J/mol)
30
20
10
0
60 80 100 120 140 160 180 200
Temperature (K)
Figure 3: excess of heat liberated during the first application of the magnetic field extracted from adiabatic temperature
change (circles) and from the enthalpy change (triangles).
We propose that the structural difference between the CO and the FM phases is responsible for the formation
of a metastable phase that transforms irreversibly to the stable state during the first magnetic cycle. The
difference in the released heat during the first and the subsequent magnetic cycles is then due to the
additional energy required for that transformation.
3. CONCLUSIONS
Summarizing, in this work we present a study about the particular behavior observed in the magnetic
properties of a system characterized by the coexistence of different magnetic phases. The magnetization and
temperature change induced by the first magnetic field application at a given temperature is different from
that obtained in following cycles.
We compared the previous result with the enthalpy change of the system, considering that the coexisting
phases do not interact between them (except from the change in their relative fractions).
Comparing the heat released during the first and the following cycles we observed good agreement between
the direct and the indirect methods in the low temperature transition. A difference in the high temperature
transitions related with the CO formation can be related with the presence of a charge disorder insulator
phase formed during the nucleation of the CO in the PI phase.
ACKNOWLEDGMENTS
We are grateful with Leticia Granja for fruitful discussion. This work has been done with support of
ANPCyT PICT 1506/2012 and PICT 2116/2014. M.Q. and J.S. are also members of CIC-CONICET.
REFERENCES
[1] A. M. Tishin and Y. I. Spichkin, The magnetocaloric effect and its application, IOP Bristol and
Philadelphia (2003).
[2] O. Tegus et al., “Transition-metal-based magnetic refrigerants for room-temperature applications”,
Nature 415, 6868 P150-152 (2002); Vitalij K. Pecharsky, Karl A. Gschneidner Jr , “ Magnetocaloric effect
and magnetic refrigeration”, Journal of Magnetism and Magnetic Materials 200, 44(1999)
[3] V. K. Pecharsky and K. A. Gschneidner, Jr., “Giant Magnetocaloric Effect in Gd5(Si2Ge2)”,
Phys. Rev. Lett. 78, 4494 (1997).
[4] E. Dagotto et al., Phys. Rep. 344, 1 (2001); E. Dagotto, The Physics of Manganites and Related
Compounds, Springer – Verlag, New York (2003).
[5] H. Terashita et al., Phys. Rev. B 70, 094403 (2004).
[6] A. M. Gomes et al., Appl. Phys. Lett. 85, 4974 (2004).
[7] M. S. Reis et al., Phys. Rev. B 71, 144413 (2005).
[8] G. J. Liu et al., Appl. Phys. Lett. 89, 222503 (2006).
[9] A. Rebello and R. H. Mahendiran, Appl. Phys. Lett. 93, 232501 (2008).
261
[10] A. Biswas et al., Appl. Phys. Lett. 92, 215502 (2008); A. Biswas et al., Appl. Phys. Lett. 92, 12502
(2008); A. Biswas et al., Appl. Phys. Lett. 94, 233109 (2009)
[11] M. Quintero et al., “Magnetocaloric effect in manganites: Metamagnetic transitions for magnetic
refrigeration”, Appl. Phys. Lett. 97, 121916 (2010)
[12] Enric Stern-Taulats et al., “Reversible adiabatic temperature changes at the magnetocaloric and
barocaloric effects in Fe49Rh51” Appl. Phys. Lett. 107, 152409 (2015)
[13] J Sacanell et al., “Phase diagram of La5/8−yNdyCa3/8MnO3 manganites”, J. Phys.: Condens. Matter 19
186226 (2007)
[14] Y. Rotstein Habarnau et al., “Direct observation of magnetocaloric effect by differential thermal
analysis: Influence of experimental parameters” Physica B 407 (16), 3305–3307 (2012).
[15] M. Quintero et al., “Phase coexistence regimes in La0.325Pr0.300Ca0.375MnO3”, Physica B 354 63–66
(2004)
[16] X.F. Miao et al., “Thermal-history dependent magnetoelastic transition in (Mn,Fe)2(P,Si)”, Applied
Physics Letters 107, 042403 (2015)
[17] M. Uehara et al., “Percolative phase separation underlies colossal magnetoresistance in mixed-valent
manganites”, Nature 399 ,560 (1999); V. Podzorov et al., “Martensitic accommodation strain and the metal-
insulator transition in manganites”, S.-W. Cheong et al., Phys. Rev. B 64, 140406 (2001).
262
INVESTIGATION OF MAGNETOCALORIC EFFECT BY THE FAST
RESPONSE INFRARED OPTICAL TEMPERATURE PROBE
A.P. Kamantsev(a)*, V.V. Koledov(a), A.V. Mashirov(a), V.G. Shavrov(a),
L.N. Butvina(b), I.S. Tereshina(c), A.S. Los(d), A. Gilewski(d),
N.H. Yen(e), P.T. Thanh(e), N.H. Dan(e)
(a)
Kotelnikov Institute of Radio-engineering and Electronics of RAS, 11/7 Mokhovaya st., 125009, Moscow,
Russia
(b)
Fiber Optics Research Center of RAS, 38 Vavilova st., 119333, Moscow, Russia
(c)
Lomonosov Moscow State University, 1 Leninskie Gory, 119991, Moscow, Russia
(d)
International Laboratory of High Magnetic Fields and Low Temperatures, 95 Gajowicka st.,
53-421, Wroclaw, Poland
(e)
Institute of Materials Science of VAST, 18 Hoang Quoc Viet, Hanoi, Vietnam
*
Corresponding author. E-mail: kaman4@gmail.com
ABSTRACT
We present a new technique for experimental study of kinetics of phase transitions (PTs) and direct
measurement of the magnetocaloric effect (MCE) in pulsed magnetic fields by using the fast response
temperature probe with infrared fiber optical (IRFO) sensor. As a preliminary demonstration of the new
technique, the results are presented of MCE measurements for Gd near Curie point: ΔTad = 21.3 K under
pulsed magnetic field µ0H = 12.7 T, with 12.5 ms pulse duration and 1 MHz frequency response of the
electronic measurement system.
Keywords: magnetocaloric effect, kinetics of phase transition, infrared fiber optical probe, fast response
temperature measurements, gadolinium.
DOI: 10.18462/iir.thermag.2016.0207
1. INTRODUCTION
In recent years, all around the world, a large number of new magnetically ordered compounds are created
and studied with magnetic, metamagnetic and magnetostructural phase transitions (PTs) of the 1st and 2nd
orders, which are accompanied by the strong anomalies of magnetic, thermal and mechanical properties [1].
But, despite the fact that the PTs in magnetic substances are studied for a long time both theoretically and
experimentally, at the moment there is no sufficiently deep understanding of the kinetic phenomena,
accompanying PTs. By conventional theoretical approach, the growth of the new phase at PT of the 1st order
is described by the kinetic equation of the Fokker-Planck. The relaxation processes near PT of the 2nd order
is described by the Landau-Khalatnikov equation [2]. However, the applicability of these equations for the
magnetic PTs has not been tested experimentally still yet.
The problem of the rate of PTs requires immediate solution also because it is crucial for the creation of the
new technologies based on “giant” effects in the vicinity of PTs in solid state magnetic functional materials.
For example, the magnetocaloric effect (MCE) reaches maximum near the PTs in magnetically ordered
solids [1]. So, study of PT’s rate is necessary for creation of a new technology of magnetic refrigeration at
room temperature with high cooling power of a sold state working body [3]. The rate of PT limits the
frequency of thermodynamic cycles. Accordingly, the power of refrigeration will depend on the frequency of
cycles, and it is difficult to judge the competitiveness of this machine without determining the fundamental
restrictions on the parameter of specific cooling power of the prospective MCE materials [4]. The
comparative study of the kinetics of PTs of the prospective magnetic functional materials such as Heusler
alloys [5-7], Fe-Rh [8] and MnFe-based [9] alloys is very important for creation of the novel devices based
on MCE [10].
The purpose of the present work is to present a new technique for experimental study of kinetics of PTs and
direct measurement of MCE in pulsed magnetic fields by using the fast response temperature probe with
infra red fiber optical (IRFO) sensor comprising the optical fiber which is transparent in IR wavelength range
5-15 μm, and the photoresistor which is sensitive in the same range.
263
2. EXPERIMENTAL
2.1 Experimental technique

For direct measurements of MCE in pulsed magnetic field the new device was designed using pyrometric
principal. The device consists of optical fiber made of AgClxB1-x (0<х<1) compound by vacuum extrusion
through a die [11-15], and a photoresistor made of narrow-gap semiconductor Cd-Hg-Te, which placed into
cryostat with liquid nitrogen. The optical fiber is transparent in IR wavelength range 5-15 μm, the
photoresistor is sensitive in the same range. One optical fiber end is connected to the photoresistor, the other
optical fiber end is connected with a surface of the MCE sample (see Fig.1). The sample with MCE under
investigation is placed into a pulsed coil magnet with magnetic field up to 13 T and pulse duration of 12.5 ms
at room temperature. The electronic part includes analog amplifier of a signal from the photoresistor, and
analog-to-digital converter, connected to PC. The initial temperature of the sample was controlled by small
thermostat with conventional Pt thermoresistor temperature sensor.
The sensor was calibrated with the help of at heating of thin semiconductor film by short electrical impulses.
The sensitivity of IRFO sensor is 0.7 mV/K, the accuracy is about ± 0.1 К. The response time of the
electronic circuit is 1 µsec. It was found that the emissivity of the samples affects the measurement
insignificantly. The sample holder was made of textolite and does not affect the adiabatic conditions of the
measurement due to short magnetic pulse time.
2.2 The experimental results
For the first tests the new technique the pure gadolinium was chosen [16-18]. The samples had a disk shape
with a diameter of 5 mm and a thickness of 1 mm, with the mass of 136.5 mg. As a preliminary
demonstration of the possibility of the new technique for MCE measurements, the measured temporal
dependences of magnetic field and change of temperature are plotted on Fig. 2a. It is found for gadolinium
at initial temperature 300 K: ΔTad = 21.3 K under pulsed magnetic field µ0H = 12.7 T.
Our data are in good correlation with results of the work [19]: ΔTad = 15.4 K under pulsed magnetic field
µ0H = 7.5 T. However, the value of MCE, measured by the IRFO sensor, exceeds nearly by 10 % the value
obtained by a semiconductor diode temperature sensor in Bitter coil magnet for the Gd sample of the same
series [17]. The 10 % difference can be explained by the fact that the conditions are closer to adiabatic in the
pulse experiment, and the use of non-contact temperature measurement method by IRFO sensor excludes
additional heat loss provided by conventional sensor. Fig. 2b shows the MCE dependence on the magnetic
field. It has a minor hysteretic character, due to thermal relaxation of the Gd sample. This effect is difficult to
detect using conventional temperature measurement techniques – thermocouples or thermoresistors sensors
with significant relaxation time.
(a) (b)
Figure 1. (a) View of infrared fiber optical temperature probe. (b) Scheme of infrared fiber optical temperature probe.
3. CONCLUSIONS
We report the first experimental results of direct measurement of the MCE in pulsed magnetic fields by using
the fast response IRFO sensor. The value of MCE for gadolinium near Curie point (300 K) is found to be
264
322
322 13
Temperature 320 Temperature
320 12
Magnetic field Gd
318
318 Gd 11
10 316
316
Temperature (K)
9 314
Magnetic field (T)

Temperature (K)
314
8 312
312 7 310
310 6
308
308 5
306
4
306
304
3
304
2 302
302 300
1
300 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Magnetic field (T)
Time (ms)
(a) (b)
Figure 2. Measurements of MCE by the IRFO temperature probe. (a) Time dependence of temperature of Gd sample
(red) and magnetic field (black). (b) Temperature vs. magnetic field at one impulse, arrows show direction of magnetic
field change.
ΔTad = 21.3 K under pulsed magnetic field µ0H = 12.7 T. Signal-to-noise ratio is not less then 10:1. The new
system demonstrates higher noise immunity than existing systems based on micro-thermocouples [20-23]
and thin film thermoresistors [24-26].
We may conclude, that in order to achieve the breakthroughs in science and technology, it is necessary to
create the new functional materials and to study their properties in extreme conditions. One of the problems
is the lack of reliable methods for measuring and monitoring the temperature of the materials in varying and
pulsed high external fields. The presented IRFO sensor provides temperature measurements with high
accuracy and fast time response (presumably less than 1 µsec) in the conditions of very strong
electromagnetic noise.
ACKNOWLEDGMENTS
The work was supported by Russian Science Foundation, grant No. 14-22-00279, and Institute of Materials
Science, VAST, grant No. CSCL05.16.
REFERENCES
[1] A. Planes, L. Manosa, M. Acet, “Magnetocaloric effect and its relation to shape-memory properties
inferromagnetic Heusler alloys”, J.Phys.: Condens. Matter 21, 233201 (2009).
[2] E.M. Lifshitz, L.P. Pitaevskii, Physical Kinetics, Pergamon Press (1981).
[3] K.A. Gschneidner Jr., V.K. Pecharsky, “Thirty years of near room temperature magnetic cooling:Where
we are today and future prospects” Journal of Refrigeration 31, 945 (2008).
[4] A. Kamantsev, V. Koledov, A. Mashirov, E. Dilmieva, V. Shavrov, J. Cwik, I. Tereshina, “Fundamental
physical restrictions on power of magnetocaloric refrigeration based on gadolinium working body”,
Proceedings of the 6th IIF-IIR International Conference on Magnetic Refrigeration (THERMAG VI), 89
(2014).
[5] A. Kamantsev, V. Koledov, E. Dilmieva, A. Mashirov, V. Shavrov, J. Cwik, I. Tereshina, V. Khovaylo,
M. Lyange, L. Gonzalez-Legarreta, B. Hernando and P. Ari-Gur, “Thermomagnetic and magnetocaloric
properties of metamagnetic Ni-Mn-In-Co Heusler alloy in magnetic fields up to 140 kOe”, EPJ Web of
Conferences 75, 04008 (2014).
[6] A.P. Kamantsev, V.V. Koledov, A.V. Mashirov, E.T. Dilmieva, V.G. Shavrov, J. Cwik, and I.S.
Tereshina, “Direct measurement of magnetocaloric effect in metamagnetic Ni43Mn37.9In12.1Co7 Heusler
alloy”, Bulletin of the Russian Academy of Sciences. Physics 78, 936 (2014).
[7] A.P. Kamantsev, V.V. Koledov, A.V. Mashirov, E.T. Dilmieva, V.G. Shavrov, J. Cwik, A. S. Los, V.I.
Nizhankovskii, K. Rogacki, I.S. Tereshina, Yu.S. Koshkid’ko, M.V. Lyange, V.V. Khovaylo, and P. Ari-
265
Gur, “Magnetocaloric and thermomagnetic properties of Ni2.18Mn0.82Ga Heusler alloy in high magnetic
fields up to 140 kOe”, Journal of Applied Physics 117, 163903 (2015).
[8] A.P. Kamantsev, V.V. Koledov, A.V. Mashirov, E.T. Dilmieva, V.G. Shavrov, J. Cwik, I.S. Tereshina,
M.V. Lyange, V.V. Khovaylo, J. Porcari, and M. Topic, “Properties of metamagnetic alloy Fe48Rh52 in
high magnetic fields”, Bulletin of the Russian Academy of Sciences. Physics 79, 1086 (2015).
[9] F. Guillou, H. Yibole, A. Kamantsev, G. Porcari, J. Cwik, V. Koledov, N.H. van Dijk and E. Brück,
“Field dependence of the magnetocaloric effect in MnFe(P,Si) materials”, IEEE Transactions on Magnetics
51, 2503904 (2015).
[10] D.A. Kuzmin, I.V. Bychkov, I.Yu. Biryukov, A.P. Kamantsev, V.V. Koledov, V.G. Shavrov,
“Modeling of phase transitions kinetics in systems with two interacting order parameters”, Materials Science
Forum 845, 166 (2016).
[11] L.N. Butvina, E.M. Dianov, D.V. Drobot, Yu.G. Kolesnikov, N.V. Lichkova, “New optical materials
and fibers for mid IR region”, Proc. SPIE 1048, Infrared Fiber Optics 17 (1989).
[12] L.N. Butvina, E.M. Dianov, N.V. Lichkova, V.N. Zavgorodnev, L. Kuepper, “Crystalline silver halide
fibers with optical losses lower than 50 dB/km in broad IR region", Proc. SPIE 3849, Infrared Optical Fibers
and Their Applications 181 (1999).
[13] L. Küpper, H.M. Heise, L.N. Butvina, “Novel developments in mid-IR fiber-optic spectroscopy for
analytical applications”, Journal of Molecular Structure 563-564, 173 (2001).
[14] L.N. Butvina, O.V. Sereda, E.M. Dianov, N.V. Lichkova, V.N. Zagorodnev, “Single-mode
microstructured optical fiber for the middle infrared”, Optics letters 32, 334 (2007).
[15] L.N. Butvina, “Mid-infrared microstructured fibers and channel waveguides, based on halides crystals
and their applications”, Proceedings of the 2011 IEEE Photonics Society Summer Topical Meeting Series
(2011).
[16] A.P. Kamantsev, V.V. Koledov, V.G. Shavrov, I.S. Tereshina, “Thermodynamic and relaxation
processes near Curie point in gadolinium”, Solid State Phenomena 215, 113 (2014).
[17] A.P. Kamantsev, V.V. Koledov, A.V. Mashirov, E.T. Dilmieva, V.G. Shavrov, J. Cwik, I.S. Tereshina,
“Magnetocaloric effect of gadolinium at adiabatic and quasi-isothermal conditions in high magnetic fields”,
Solid State Phenomena 233-234, 216 (2015).
[18] E.T. Dilmieva, A.P. Kamantsev, V.V. Koledov, A.V. Mashirov, V.G. Shavrov, J. Cwik, I.S. Tereshina,
“Experimental simulation of a magnetic refrigeration cycle in high magnetic fields”, Physics of the Solid
State 58, 81 (2016).
[19] S.Yu. Dan’kov, A. M. Tishin, V. K. Pecharsky, K. A. Gschneidner Jr., “Magnetic phase transitions and
the magnetothermal properties of gadolinium”, Phys. Rev. B 57, 3478 (1998).
[20] A. K. Nayak, C. S. Mejia, S.W. D'Souza, S. Chadov, Y. Skourski, C. Felser, M. Nicklas, “Large field-
induced irreversibility in Ni-Mn based Heusler shape-memory alloys: A pulsed magnetic field study”,
Physical Review B 90, 220408 (2014).
[21] M.G. Zavareh, C.S. Mejia, A.K. Nayak, Y. Skourski, J. Wosnitza, C. Felser, M. Nicklas, “Direct
measurements of the magnetocaloric effect in pulsed magnetic fields: The example of the Heusler alloy
Ni50Mn35In15” Applied Physics Letters 106, 071904 (2015).
[22] C. S. Mejia, M.G. Zavareh, A. K. Nayak, Y. Skourski, J. Wosnitza, C. Felser, M. Nicklas, “Pulsed high-
magnetic-field experiments: New insights into the magnetocaloric effect in Ni-Mn-In Heusler alloys”,
Journal of Applied Physics 117, 17E710 (2015).
[23] T. Gottschall, K.P. Skokov, F. Scheibel, M. Acet, M.G. Zavareh, Y. Skourski, J. Wosnitza, M. Farle,
and O. Gutfleisch, “Dynamical Effects of the Martensitic Transition in Magnetocaloric Heusler Alloys from
Direct ΔTad Measurements under Different Magnetic-Field-Sweep Rates”, Physical Review Applied 5,
024013 (2016).
[24] T. Kihara, I. Katakura, M. Tokunaga, A. Matsuo, K. Kawaguchi, A. Kondo, K. Kindo, W. Ito, X. Xu, R.
Kainuma, “Optical imaging and magnetocaloric effect measurements in pulsed high magnetic fields and their
application to Ni–Co–Mn–In Heusler alloy”, Journal of Alloys and Compounds 577, S722 (2013).
[25] T. Kihara, Y. Kohama, Y. Hashimoto, S. Katsumoto, M. Tokunaga, “Adiabatic measurements of
magneto-caloric effects in pulsed high magnetic fields up to 55 T”, Review of Scientific Instruments 84,
074901 (2013).
[26] T. Kihara, X. Xu, W. Ito, R. Kainuma, M. Tokunaga, “Direct measurements of inverse magnetocaloric
effects in metamagnetic shape-memory alloy NiCoMnIn” Physical Review B 90, 214409 (2014).
266
MAGNETOCALORIC EFFECT SUPPRESSION INDUCED BY
MECHANICAL MILLING ON INTERMETALLIC DyCo2
V. G. de Paula(a)*, M. Gomes Silva(b), L. M. da Silva(b), A. O. dos Santos(b), R. Lang(c), L.
Otubo(d), A. A. Coelho(a)and L. P. Cardoso(a)
(a)
Instituto de Física Gleb Wataghin,Universidade Estadual de Campinas-UNICAMP, 13083-859 Campinas,
São Paulo, Brazil
(b)
Centro de Ciências Sociais, Saúde e Tecnologia-(CCSST,Universidade Federal do Maranhão-UFMA,
65900-000 Imperatriz, Maranhão, Brazil
(c)
Instituto de Ciência e Tecnologia-ICT, UNIFESP,12231-280 São José dos Campos, São Paulo, Brazil
(d)
Centro de Ciência e Tecnologia de Materiais, Instituto de Pesquisas Energéticas e Nucleares-IPEN,
05508-000 São Paulo, São Paulo, Brazil
*
Corresponding author. E-mail: vinicius.gomes.paula@gmail.com
ABSTRACT
Structural and magnetic properties of mechanically milled samples of intermetallic DyCo2 were investigated
by x-ray diffraction and magnetization measurements. The results show that the average crystallite size is
reduced down to 35 nm for the sample milled for 8 h. The milling process induces structural disorder as
evidenced by x-ray diffraction data, leading to a weakening of Itinerant Electron Metamagnetism (IEM)
mechanism and suppression of the magnetocaloric effect. The RCP values of milled samples were not
significantly reduced in comparison with the bulk sample due to the broadening of magnetic entropy change
peak.
Keywords: mechanical milling, magnetocaloric effect, magnetization.

DOI: 10.18462/iir.thermag.2016.0208
1. INTRODUCTION
Research and development of eco-friendly technologies are suitable efforts in a world with increasing
demand for environmental safety. In this context, the magnetocaloric effect (MCE) shows up as a key
research topic, once it is the basis for new magnetic refrigeration technique. Since 1917[1] when the MCE
was discovered, their theoretical and experimental aspects have been studied. Particularly in the last decades,
this subject attracts great attention with the discovery of materials with giant MCE near room-temperature[2-
4]
. Synthesis of new compounds is a fundamental step in search for materials with improved magnetocaloric
properties, leading to remarkable results in different families of intermetallic compounds[5-7]. In this way, the
search for materials with high MCE in low magnetic field is the main goal. However, despite the importance
of high values for isothermal magnetic entropy change (-ΔSM), a narrow -ΔSM peak is a limiting cooling
efficiency-factor, since a real magnetic refrigerator must be able to operate in a wide temperature range.
Recently, magnetocaloric measurements in GdAl2 milled samples has showed broad -ΔSM peak profile
attributed to a distribution of Curie Temperature (TC) values, occasioned by the structural disorder induced
by milling process[8]. This result demonstrates that mechanical milling is a viable alternative to synthetize
samples with broadened -ΔSM peak.
The RM2 (R=Rare earth, M=Transition metal) is one of the most extensively studied family of alloys[9]. High
angular magnetic moments J values associated to the R ions, presence of first or second order magnetic
phase transitions and low hysteresis values are essential pre-requisites to the interesting magnetocaloric
properties observed for this family of compounds. The Cobalt-based (RCo2) alloys are particularly
interesting, since instability of the Co sub-lattice, which leads to a metamagnetic first order transition for
some compounds of this family[5,10,11]. The field dependent character of phase transition is originated by 3d
itinerant electrons band, which under magnetic fields larger than a critical value induce changes in the
magnetic behavior of Co ions, resulting in non-zero net magnetic moment. The phenomena is called Itinerant
Electron Metamagnetism (IEM)[12] and is the origin for first order transitions with high ΔSM values in a
narrow temperature range in some RCo2 alloys. Although the magnetic and magnetocaloric properties of
bulk DyCo2 have been extensively studied by theoretical and experimental aspects [5,10] the complete
267
understanding of changes provided by reduction of particle size and its impacts on MCE are still not settled.
Motivated by this subject, here we report the synthesis, structural and magnetic characterization of DyCo2
particles obtained by mechanical milling technique in lower milling time (up to 8 h).
2. EXPERIMENTAL DETAILS
DyCo2 ingot with approximately 6 g was prepared by arc melting of the high purity elements (Dy=99.9%,
Co=99.95% Sigma Aldrich Elements) in an argon (Ar) atmosphere with subsequently annealed at 800 °C for
five days. The annealed alloy was then crushed and milled in a planetary ball mill (Fritsch Pulverisette7) at a
rotation speed of 200 rpm using a hardened steel container, also under Ar atmosphere. The milling was
carried out in a sequence of 30 min milling and 10 min resting; after each 2 h the container was opened to
uncompress the powder. The structural characterization was performed by using powder x-ray diffraction
(XRD) with Cu-K1 radiation (Rigaku Miniflex II) and pyrolytic graphite monochromator. A
superconducting quantum interference device magnetometer (MPMS SQUID magnetometer from Quantum
Design Inc.) was employed to measure the magnetization in the temperature interval between 2 and 300 K
under applied magnetic fields up to 50 kOe.
3. RESULTS
Fig. 1(a) shows the XRD measurements of the DyCo 2 unmilled sample (referred to as 0 h) as well as
of the samples submitted to 4 and 8 h of milling. It is possible to discriminate all the peaks corresponding
to the cubic MgCu2 FCC structure, no evidence of contamination from the milling materials or
oxidation was observed. The milling process induces changes in the sample structure i.e., significant
reduction in peak intensity and a consequent broadening. XRD Rietveld refinement analysis was
carried out for bulk and milled samples[13] and from results were extracted the lattice parameter,
average crystallite size, and microstrain values. Fig. 1(b) depicts in particular the refinement for the 8
h milled sample. The obtained lattice parameter for bulk sample is 7.194(1) Å in agreement with
literature[10] and no significant changes were observed for milled samples. The average crystallite size
reduces from 156 nm (bulk sample) to 35 nm for the 8 h milled sample, while the microstrain increases
reaching 1.2% regarding to the unmilled sample. Reduction in the crystallite size and large microstrain
values are common microstructural changes induced by milling process[14,15].
Figure 1. a) X-ray patterns for bulk and milled DyCo2 samples and b) Rietveld refinement for the 8 h milled sample.
Temperature dependence of magnetization, following ZFC and FC protocols, are exhibited in Fig. 2(a)
DyCo2 bulk sample presents ferromagnetic (FM) phase transition with TC = 140 K. The milling process
induces a small broadening in the FM transition that can be attributed to structural disorder due to the
mechanical impact. TC values remain unchanged for milling time up to 8 h, as shown in Fig. 2(b). On the
other hand, the remanent magnetization (Mr) at T=2 K shown in Fig. 2(c) increases from 1.68 μB/u.f. (0 h
sample) to 3.45 μB/u.f. (8 h sample), as expected for small particles. However, for higher temperatures as
T=137 K, Mr is strongly reduced, as shown in Fig. 2(d) (8 h sample), which is important for magnetic
refrigeration applications, once hysteretic losses lowers eficiency in the thermomagnetic cycles. In addition,
the reduction of saturation magnetization (MS) occurs due to the weakening of long-range magnetic ordering,
which is caused by the enhanced magnetic ions surface/volume ratio.
268
Figure 2. (a) Temperature dependence of magnetization for 0h (bulk) and 8 h milled DyCo2 samples following the ZFC
and FC protocols and (b) magnetization derivative for both samples. (c) Hysteresis loops for 0 and 8 h sample measured
at 2 K and (d) Magnetization as a function of magnetic field (increasing and decreasing) for 8 h sample at 137 K.
Magnetization isothermals curves M (H,T) as a function of applied magnetic field (0 to 5 T), measured
around TC for 0 h and 8 h samples, are shown in Fig. 3(a) and 3(b), respectively. The isothermals variation
obtained for 8 h sample is less pronounced compared to the 0 h sample ones. Also, the correspondent Arrot
plots shown in Fig. 3(c) and 3(d), suggests changes in the nature of the FM transition. According to Banerjee
criterion[16], a negative or positive slop of Arrot plots are evidence of first or second-order magnetic phase
transition, respectively. Negative slop on Arrot plot observed for unmilled DyCo2 sample indicates first order
ferromagnetic transition, in agreement with literature[5,10]. On the other hand, the milled samples show
positive slop consistent with a second order transition. These results indicate that the milling process
weakens the IEM mechanism responsible for the first order ferromagnetic transition on DyCo2.
Figure 3. Isothermal curves for (a) bulk and (b) 8 h milled samples. The corresponding Arrot plots are shown in (c) and
(d), respectively.
The magnetic entropy change -SM for DyCo2 bulk and milled samples was calculated from the isothermal
curves by numerical integration of Maxwell’ relation, h
.
 
f
M S M  dH
T H
hi
Figure 4. Magnetic entropy change (-SM) for DyCo2 bulk and milled by 4h and 8h in a magnetic field change of 5 T.
269
Fig. 4 shows the temperature dependence of -SM for 4 and 8 h samples calculed for different magnetic field
change of 5 T. The mechanical milling reduces the maximum values of magnetic entropy change (SMmax)
from 14.2 J/kgK to 5.1 J/kgK. On the other hand, a significant broadening in the -SM peak profile is
observed as the milling time increases. Both reduction of SMmax and the broadening of -SM peak are
indicative of second order FM transition for DyCo2 milled samples, in agreement with Arrot plots result.
Although milling process significantly reduces the SMmax, the relative cooling power (RCP=SMmaxx
δTFHWM) are not strongly affected. RCP values for H=5 T are 335, 300 and 249 J/kg for 0, 4 and 8 h
samples, respectively.
4. CONCLUSIONS
Mechanically milled DyCo2 samples were studied by means of x-ray diffraction and magnetic measurements.
The milling process used up to 8h has induced a strong reduction (~35 nm) in the DyCo2 average crystallite
size. The reduction of saturation magnetization at T = 2 K are strongly correlated with size effects, which
among other factors decrease the long-range magnetic ordering. The ball milling structural disorder
disturbs the R-Co coupling (responsible for IEM mechanism). This provokes the weakening of the
IEM mechanism which plays a central role on suppressing the MCE. The broadening of -SM peak positively
contributes by keeping the RCP value of the milled samples close to the DyCo2 bulk sample one.
ACKNOWLEDGMENTS
The authors would like to thank the Brazilian funding agencies CNPQ, FAPEMA and FINEP for partial
financial support.
REFERENCES
[1] A. Smith, “Who discovered the magnetocaloric effect ?” Eur. Phys. 38, 507 (2013)
[2] J. Lian, et al., “Giant magnetocaloric effect driven by structural transitions”, Nat. Mat. 11, 620 (2012)
[3] V. K. Pecharsky, K. A. Gschneidner, Jr, “Giant Magnetocaloric Effect in Gd 5(Si2Ge2)”, Phys. Rev. Lett.
78, 4494 (1997).
[4] V. Franco, et al., “The magnetocaloric effect and magnetic refrigeration near room temperature: materials
and models”, Annu. Rev. Mat. Res. 42, 305 (2012)
[5] Jr. Gschneidner, et al., "Recent developments in magnetocaloric materials", Rep. Prog. Phy. 68, 1479
(2005).
[6] V. Basso, et al., “Hysteresis and magnetocaloric effect at the magnetostructural phase transition of Ni-
Mn-Ga and Ni-Co-Mn-Sn Heusler Alloys”, Phys. Rev. B 85, 014430 (2012)
[7] S. Gupta, K. G. Suresh. "Review on magnetic and related properties of RTX compounds", J. Alloys
Comp. 618, 562 (2015).
[8] V. G. de Paula, et al., “Magnetocaloric effect and evidence of superparamagnetism in GdAl 2
nanocrystallites: A magnetic-structural correlation”, Phys. Rev. B 93, 094427 (2016).
[9] A. M. Tishin, Y. I. Spichkin, The magnetocaloric effect and its applications, CRC Press. 179 (2003)
[10] N. K. Singh, et al., “Itinerant Electron Metamagnetism and Magnetocaloric Effect in RCo2-based laves
phase compounds”, J. Mag. Magn. Mat. , 317 (2007) 68-79.
[11] J. Herrero, F. Bartolomé, L. M. García, “Nature and entropy content of the ordering transitions in
RCo2”, Phys. Rev. B 73, 134410 (2006).
[12] R. Z. Levitin, A. S. Markosyan, “Itinerant Metamagnetism”, Usp. Fiz Nauk.155, 623 (1988).
[13] A. C. Larson, R. B. Von Dreele, General Structure Analysis System (GSAS), Los Alamos National
Laboratory Report LAUR 86-748, Los Alamos, NM, USA, 2004.
[14] E. Yuzuak, et al., “The magnetic and magnetocaloric properties of nanostructured CoMnGe0.95Ga0.05”, J.
All. Comp. 541, 256 (2012)
[15] M. Khajepour, S. Sharafi, “Characterization of nanostructured Fe-Co-Si powder alloy”, Powd. Tech.
232 124 (2012).
[16] S. K. Banerjee, “On a generalized approach to first and second order magnetic transitions”, Phys. Lett.
12, 16 (1964).
270
CFD SIMULATIONS OF THE MAGNETOCALORIC HEAT TRANSFERS
BETWEEN A GADOLINIUM PLATE AND AN OSCILLATING LIQUID
FLOW
J.C. Roy*, T. De Larochelambert, P. Nika, Y. Bailly, R. Glises, A. Meunier
Institut FEMTO-ST, CNRS, University Bourgogne - Franche-Comté, 2, avenue Jean Moulin, 90000 Belfort,
France
*
Corresponding author. E-mail: jean-claude.roy@univ-fcomte.fr
ABSTRACT
This study deals with simulations of heat transfers between a 1 mm thick gadolinium plate submitted to a
variable magnetic field and an oscillating liquid flow. The channel is connected to a cold thermal source (Tc
= 290 K) and to a hot thermal source (Th = 295 K). The volume source term for the energy equation in the
gadolinium plate is computed for each time step. Simulations were performed with the Fluent CFD code for
a frequency f = 1Hz. The influence of the channel volume displacement was investigated. The evolution of
the temperature distribution in the plate for different volume rates shows that the gradient reaches a
maximum value in the plate for a volume rate displacement of 40%. This is confirmed by the determination
of the enthalpy variation of the liquid located in the thermal source. Simulations for f = 2Hz clearly show the
diminution of the enthalpy variations and consequently a sharp diminution of the possibility of heat transfer
between the sources.
Keywords: active magnetic regenerator; temperature, heat source; heat pump; enthalpy
DOI: 10.18462/iir.thermag2016.0220
1. INTRODUCTION
The magnetic heat pumping near room temperature constitutes a promising alternative to the conventional
heat pump or refrigerator using compressor and refrigerant gas, because of the polluting potential of these
gases [1,2]. In active magnetic regenerator (AMR) systems, heat is alternately exchanged between the
magnetic material and a channel fluid flow connected to both cold and hot heat sources. The forecasting of
the performance of such systems requires the accurate modeling of the oscillating flow which induces
specific heat transfer rates specially when annular flow occurs in the channels. Some authors have developed
models applied to flow between parallel plates: (i) 1D models [3,4] using a finite difference method (FDM)
or (ii) 2D models [5] using a finite element method (FEM) or a finite volume method (FVM) [6,7]. The
drawback of 1D models lies in the use of approximate correlations for the heat transfer coefficient between
the plate and the flow. The solving of the equation of energy in the flow for 2D models permits to avoid this
drawback and to simulate the heat transfer between the fluid flowing out of the channels and the heat sources
located on both sides of the AMR. A more realistic 3D model is developed which considers gadolinium
plates submitted to a variable magnetic field. A liquid flow oscillates through the channels between the plates
with a controlled frequency in relation with the magnetic field and perfect heat exchangers are considered on
both sides of the plates. Simulations are performed with the Fluent CFD code in order to highlight (i) the
influence of the volume displacement rate in the channel on the maximal temperature gradient along the
plate and (ii) the enthalpy variations of the liquid located in the heat source zones during each half period.
The influence of the frequency of the oscillating flow on enthalpy variations is also studied.
2. MODEL
2.1 Materials and domain
The three-dimensional geometry of the domain (Fig.1) consists of a gadolinium plate surrounded by two
channels connected at both ends to the liquid zones in contact with the cold and hot sources. The density of
gadolinium is ρg = 7900 kg m-3; its thermal conductivity is λg = 10.6 W m-1 K-1 and its specific heat,
dependent of temperature and magnetic induction, is determined from the Dan’kov et al. curves [8]. The
liquid is a specific antioxidant coolant, with physical properties slightly different from that of pure water
(density: ρf = 1089 kg m-3; thermal conductivity: λf = 0.54 W m-1 K-1; specific heat: Cpf = 3630 J kg-1 K-1;
viscosity: µ f = 0.00172 kg m-1 s-1). Because of the periodic arrangement of the gadolinium plates inside the
271
heat pump, symmetry planes are considered on the upper, lower and front sides of the domain while a solid
wall is considered on its rear side.
45 45 45
2 Cold zone Hot zone
0.75 Inlet 0.5 Gadolinium plate

0.25 Channel Outlet
Figure 1: View of the domain, dimensions in mm, scale not respected
2.2 Periodic conditions

A periodic trapezoid velocity-inlet condition is assigned to the left side. The evolution of the magnetic field
must be synchronized with the displacements of the fluid in order to optimize the heat transfer from the cold
to the hot source. The time-courses for one period τ (adimensional time t/τ) of both the velocity in the inlet
section and the magnetic field applied to the gadolinium plate are presented on Fig.2 and are similar to the
conditions chosen by Roudaut et al. [9].
V H
0.002 0.5
0.2 0.4 Hmax
Vmax
0.502 1
0.6 1 Hmin
0 t/τ t/τ
0.1 0.5 0
-Vmax 0.7 0.9

a b
Figure2. Time courses for (a) the velocity in the inlet section and (b) the magnetic field
2.3 Thermal conditions

The temperature of the fluid flowing in the domain is set to the cold source temperature Tc = 290 K while the
temperature of the backflow occurring in the outlet section when the inlet velocity is negative is set to the hot
source temperature Th = 295 K. The heat exchange between the liquid located in the cold (resp. hot) zone and
the cold (resp. cold) source is considered as perfect. For that purpose, temperature in every point of these
zones is numerically reset to the value of the concerned thermal source at the end of each half-period. The
magnetocaloric effect (MCE) in the gadolinium plate is taken into account by a source term q (W m-3) in the
solid energy equation and defined as:
 ∂M  ∂H 
q = − µ 0T    (1)
 ∂T  ∂t 
where µ0 is the vacuum permeability, T the temperature, M the magnetization of the solid, H the magnetic
field and t the time. M is a function of H and T and was determined from calorimetric measurements with
linear interpolations.
2.4 Parameters of the solver

The 3D domain is meshed with a 216.000 elements hexahedral grid. Continuity, momentum and energy
equations are solved for a transient laminar flow and the time step is set to τ/100. The source term q is
determined for each node and at each time step using a user defined function (UDF), which performs the
interpolation for the determination of M and its derivative.
272
3. RESULTS
3.1 Temperature distribution in the plate
The first series of simulations with no thermal load (adiabatic working) has permitted to investigate the
influence of the displacement rate D (in %) i.e. the ratio of the volume of fluid displaced in the channel at
each half-period, to the channel volume. A magnetic field difference Hmax – Hmin is considered corresponding
to a magnetic induction difference of 2 T. The temperature distribution along the plate at f = 1 Hz is
presented in Fig.3 for different displacement rates after 50 periods (steady-state working is reached). The
temperature span noticeably decreases for rate values upper than 40%. Consequently, this value is retained
for further simulations
297
296
295
Temperature ,K
294
293
292 20%
291 40%
60%
290
80%
289 100%
288
0 5 10 15 20 25 30 35 40 45
Position, mm
Figure 3. Temperature distribution along the plate for different displacement rates D; f = 1Hz; adiabatic working
Simulations allow to determine the intensity of the heat transfer between the fluid in the channel and the cold
and hot zones. The enthalpy variations of the liquid are presented on Fig.4 for f = 1 Hz as a function of the
adimensional time t/τ. The enthalpy of the liquid increases in the hot zone during the first half-period and
decreases in the cold zone during the second half-period. Simulation for f = 2 Hz is presented on Fig.5 and
shows an important decrease of the enthalpy variation and even a diminution of the enthalpy in the hot zone
at the end of the first half-period. Consequently, the efficiency of the heat pump is almost null and, for higher
values of f , heat is exchanged from the hot source to the cold source.
Hot zone Cold zone

0.03 0.01
Enthalpy variation, J
0.02
0
0.01
-0.01
0
-0.01 -0.02
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Adimensional time Adimensional time
a b
Figure 4. Enthalpy variation of the liquid for a period at f = 1 Hz; D = 40% (a) with the hot source at Th = 295 K;
(b) with the cold source at Tc = 290 K
273
Hot zone Cold zone
0.03 0.01
0.02
0
0.01
-0.01
0
-0.01 -0.02
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Adimensional time Adimensional time
a b
Figure 5. Enthalpy variation of the liquid for a period at f = 2 Hz; D = 40%; (a) with the hot source at Th = 295 K;
(b) with the cold source at Tc = 290 K
4. CONCLUSION
3D simulations of the heat transfer between a gadolinium plate and a liquid flow oscillating between two
heat sources were performed with the Fluent CFD code. This numerical tool permits to simulate the complex
heat transfer between the plate and the flow, because of the velocity profile in the channel, which is affected
by the oscillation of the flow. The great interest of this tool lies in its ability to determine the heat quantities
exchanged between the channel flow and both the hot and cold zones. Simulations highlight the great
sensitivity of the temperature span in the plate to the displacement rate in the channel. The frequency f of the
flow plays also a key role in the heat quantity transferred from the cold source to the hot source. These
results must be now validated with experimental results before being used to investigate a large range of
values for the frequency, the temperatures of the heat sources and the displacement rate. The shape and
dimensions of the hot and cold zones can also be studied in order to optimize the mixing of the liquid
flowing out from the channel.
REFERENCES
[1] B.F. Yu, Q. Gao, B. Zhang, X.Z. Meng, Z. Chen, “Review on research of room temperature magnetic
refrigeration” International Journal of Refrigeration 26, 622-636 (2003).
[2] K.A. Gschneidner, V.K. Pecharsky, “Thirty years of near temperature magnetic cooling: Where we are
today and future prospects” International Journal of Refrigeration 31, 945-961 (2008).
[3] F. Allab, A. Kedous-Lebouc, J. Fournier, J.P. Yonnet, “Numerical modeling for active magnetic
regenerative refrigeration” IEEE Trans. Magn. 41(10), 3757-3759 (2005).
[4] M. Risser, C. Vasile, T. Engel, B. Keith, C. Muller, “Numerical simulation of magnetocaloric system
behaviour for an industrial application” International Journal of Refrigeration 33(5), 973-981 (2010).
[5] T.F. Petersen, N. Pryds, N. Smith, J. Hattel, H. Schmidt, H.J.H. Knudsen, “Two-dimensional
mathematical model of a reciprocating room-temperature Active Magnetic Regenerator” International
Journal of Refrigeration 31(3), 432-443 (2008).
[6] K.K. Nielsen, C.R.H. Bahl, A. Smith, R. Bjork, N. Pryds, J. Hattel, “Detailed numerical modeling of a
linear parallel-plate Active Magnetic Regenerator” International Journal of Refrigeration 32(6), 1478-1486
(2009).
[7] U. Legait, “Caractérisation et modélisation magnétothermique appliquées à la réfrigération magnétique”
Ph. D. Thesis. Université de Grenoble, France (2011).
[8] S.Y. Dan’kov, A.M. Tishin, V.K. Pecharsky, K.A. Gschneider Jr, “Magnetic phase transitions and the
magnetothermal properties of gadolinium” Physical Review B 57(6), 3478-3490 (1998).
[9] J. Roudaut, A. Kedous-Lebouc, J.P. Yonnet, C. Muller, “Numerical analysis of an active magnetic
regenerator “ International Journal of Refrigeration 34, 1797-1804 (2011).
274
Ti SUBSTITUTION IN MnBi RARE-EARTH-FREE MAGNETOCALORIC
COMPOUND
C. Curcio(a, b)*, E. S. Olivetti(b), L. Martino(b), M. Küpferling(b), A. Bartok(c), M. LoBue(c),
V. Basso(b)
(a)
DISAT, Politecnico di Torino, Duca degli Abruzzi 24, 10129, Torino, Italy
(b)
Istituto Nazionale di Ricerca Metrologica, Strada delle cacce 91, 10135, Torino, Italy
(c)
SATIE, ENS Cachan, CNRS, Universitè Paris-Saclay,
61 avenue de Président Wilson, 94235, Cachan, France
*
Corresponding author. E-mail: c.curcio@inrim.it
ABSTRACT
The MnBi compound exhibits a first order magneto-structural phase transition around 630 K. This material
could be suitable as a rare-earth-free magnetocaloric material provided that one can shift the transition
temperature towards room temperature. In this work the effect of a partial substitution of Mn with Ti is
investigated. In particular we study the transformation kinetics of the magneto-structural transition, in samples
prepared with different techniques: powder metallurgy and rapid quenching from the melt. The samples show
a logarithmic dependence of the transition temperature on the temperature scan rate upon cooling. This
suggests that the transition occurs with the help of thermal fluctuations. The linear size of the activation volume
v0 of these thermally activated processes is found to pass from 2.5 nm in MnBi to around 1.8 nm in Ti
substituted samples.
Keywords: MnBi, kinetic effects, first order phase transition.

DOI: 10.18462/iir.thermag.2016.0222
1. INTRODUCTION
The intermetallic compound MnBi shows a first order magneto-structural phase transition around 630 K [1]
from a ferromagnetic low temperature α-phase to a paramagnetic high temperature β-phase, with an entropy
change of about 20 Jkg-1K-1 [2]. The α-MnBi has a hexagonal NiAs-type structure, which undergoes an a-axis
dilatation (1.2%) and a c-axis contraction (2.5%) when transforming into β-MnBi. The α-β transition is also
accompanied by the diffusion of ~ 15% of the Mn atoms that leave the initial octahedral sites to fill the trigonal
bipyramidal holes of the Bi sublattice [3]. The MnBi compound could be suitable as a rare-earth-free
magnetocaloric material provided that the transition temperature can be shifted towards room temperature.
Partial substitutions of Mn with other transition metals were already attempted in thin film samples [4-5],
however little is known on the properties of ternary bulk materials.
The MnBi compound exhibits a temperature scan rate (dT/dt) dependence of the α-β transition temperature
only upon cooling, already observed in the behavior of the cell parameter a [3] and in the magnetic and thermal
behavior of the transition [2]. The transition temperature changes approximately linearly with the logarithm of
the rate (dT/dt). This suggests that the thermal fluctuations are helping to overcome the energy barrier between
the two minima corresponding to the ferromagnetic and the paramagnetic states. These thermally activated
processes are accompanied by a correlated volume v0 in which the transition takes place with the help of the
thermal energy kBT [7]. In this work Mn50-xTixBi50 (x = 0 and x = 10) bulk samples have been prepared by
different techniques: powder metallurgy (PM) and rapid quenching from the melt (RQ). The aim of this work
is to study the role of Ti substitution in the MnBi microstructure. In particular, we investigate the effect of Ti
on the transformation kinetics of the first order magneto-structural transition by evaluating the thermal
activation volume v0 for Mn50-xTixBi50 samples, with x = 0 and x = 10.
2. SAMPLE PREPARATION AND EXPERIMENTAL
Polycrystalline samples are prepared with nominal composition Mn50-xTixBi50 (x = 0 and x = 10). Powder
metallurgy (PM) pellets are produced by mixing in a planetary ball mill the elemental starting powders for 2
hours (in zirconia jar sealed in nitrogen atmosphere, revolution speed of 450 rpm); then the mixed powders
are compacted in a green pellet (pressure 470 MPa, diameter 5 mm and thickness 0.5-1 mm). Finally, green
275
pellets are annealed in argon atmosphere for 1 hour at 580 K. Rapidly quenched (RQ) bulk plates (thickness
1.5 mm) are prepared by arc melting and suction casting of a master alloy in a water-cooled copper mould, in
argon atmosphere. The results presented in this paper concern samples labeled as PM-Ti0 with x = 0, PM-Ti10
with x = 10 and RQ-Ti10 with x = 10. The magnetic properties are investigated through Physical Property
Measurement System (PPMS) in the temperature range 10-290 K and by Vibrating Sample Magnetometer
(VSM) in the range 300-700 K, in both cases with an applied field of 1 T. The thermal behavior of the samples
is analyzed through Differential Scanning Calorimetry (DSC). The MnBi phase fraction, evaluated by
calorimetric measurements, is: 90 wt% for PM-Ti0 pellets, 10 wt% for PM-Ti10 and 40 wt% for RQ-Ti10
samples.
3. RESULTS
Fig. 1 shows the thermomagnetic curves of the studied samples. The curves displayed in Fig. 1a show the
typical thermomagnetic behavior of α-MnBi, with a spin reorientation transition around 110 K. The RQ Ti-
substituted sample of Fig. 1b exhibits a negligible magnetic moment at room temperature and a second order
phase transition around 180 K. The non-equilibrium conditions of the rapid quenching technique allow the Ti
to be incorporated into MnBi lattice [6], producing a ternary Mn(Ti)Bi phase. When this ternary metastable
phase is at T > 480 K the incorporated Ti atoms go out from the MnBi lattice and above 580 K the Mn(Ti)Bi
phase is fully transformed into MnBi phase plus secondary Ti-rich phases. In the subsequent part of the study
we try to clarify the role of Ti on the kinetics of the first order magneto-structural transition of MnBi.
(a) (b)
Figure 1. Thermomagnetic curves of the PM-Ti0 pellet (a) and the RQ-Ti10 sample (b), with µ0Happ = 1T.
The temperature scan rate (dT/dt) dependence of the α-β transition is investigated by DSC measurements at
rates from 0.3 to 30 K/min. Fig. 2 shows the DSC measurements with dT/dt = 10 K/min. On the heating curves
are displayed the α-β transition of MnBi at 631 K (PM-Ti0), 622 K (PM-Ti10) and 625 K (RQ-Ti10). On the
cooling curves are observed the β-α transition at 606 K (PM-Ti0), 587 K (PM-Ti10) and 590 K (RQ-Ti10).The
thermal hysteresis ΔT of the transition increases with Ti substitution. At 10 K/min the ΔT is 25 K (PM-Ti0),
36 K (PM-Ti10) and 34 K (RQ-Ti10). The measurement conditions may induce a spurious dT/dt dependence
related to the non-ideal thermal contact between the sample and the heat flux sensor [7]. Therefore we have
carefully evaluated the thermal contact resistance R, which depends on the contact area and on the quality of
the contact. The value of the thermal contact resistance is estimated by the coincidence of the melting point of
Bi-rich phase in DSC scans performed at different dT/dt. The R values are: 135 K/W for the PM-Ti0 pellets,
200 K/W for the PM-Ti10 pellets and 250 K/W for the RQ-Ti10 samples. The sample temperature Ts is then
estimated by the relation Ts = T – Rqs, where qs is the heat flux and T the measured temperature (Fig. 2).
276
Figure 2. DSC scans at 10 K/min.
After this correction we find that the rate dependence of the α-β transition upon heating is eliminated, while
the β-α transition temperature upon cooling Tc still depends on the temperature scan rate (dT/dt). This
dependence is linear with respect to the logarithm of the dT/dt (Fig. 3), and can be described by the relation:
𝑑𝑇
𝑇𝑐 = 𝑇𝑖 − 𝑇 ∗ |ln ( )| (1)
𝑑𝑡
where Ti is a constant and T* is a parameter. The slope of the linear dependence (T*) is correlated with the
typical volume v0, which transforms spontaneously and coherently from the β-MnBi to the α-MnBi phase by
the help of the thermal energy kBT [2]:
𝑘𝐵 𝑇
𝑇∗ = (2)
𝛥𝑠 𝑣0
where Δs is the entropy change at the transition in volume density [2].
Figure 3. Rate dependence dT/dt of the transition temperature Tc upon cooling.
277
The slopes T* and the thermal activation volumes v0 of the investigated samples were derived by fitting the
measured Tc (dT/dt) with the Eq. 1. The resulting values are reported in Table 1. The Ti substitution (x = 10)
in both PM and RQ samples decreases the size v01/3 of the region which transforms coherently by thermal
activation (Table 1). This can be interpreted as the effect of interstitial Ti which is contributing to decrease the
typical coherence region of MnBi transformation.
Mn50-xTixBi50 T* (K) v0 1/3(nm)

PM-Ti0 3.2 2.5
PM-Ti10 7.7 1.87
RQ-Ti10 7.6 1.88
Table 1. The slope of the linear dependence (T* ) and the linear size of the activation volume (v0 1/3).
4. CONCLUSIONS
In this paper we have studied the transition kinetics upon cooling of MnBi and Ti-substituted samples (PM
and RQ). The kinetics is found to be of thermal activation type and is described by a thermal activation volume
v0. The estimation of the activation volume shows that the Ti substitution (for both the PM-10 and the RQ-10
samples) slightly decreases the size of the region which transforms coherently from β-MnBi to the α-MnBi
structure. The linear size of the activation volume (v01/3) passes from 2.5 nm to around 1.8 nm with the Ti
substitution (x = 10).
Future investigation will be dedicated to clarify the role of the Ti substitution in the MnBi microstructure and
its effect on the magnetic properties.
ACKNOWLEDGMENTS
The research leading to these results has received funding from the EC 7th Framework Programme under Grant
Agreement No. 310748 (Project DRREAM).
REFERENCES
[1] T. Chen, E. Stutius, “The phase transformation and physical properties of the MnBi and Mn1.08Bi
compounds”, IEEE Trans. Magn. 10, 581 (1974).
[2] V. Basso et al., “Entropy change and kinetic effects at the magnetostructural phase transition of MnBi”,
[3] A. F. Andresen et al., “The magnetic and crystallographic properties of MnBi studied by neutron
diffraction”, Acta. Chem. Scand. 21, 1543 (1967).
[4] H. Göbel et al., “Properties of MnBi compounds partially substituted with Cu, Zn, Ti, Sb, and Te. I.
Formation of mixed phases and crystal structures”, Phys.Status Solidi 34, 553 (1976).
[5] H. Göbel et al., “Properties of MnBi compounds partially substituted with Cu, Zn, Ti, Sb, and Te. II.
Stability and magnetooptic properties of thin films”, Phys.Status Solidi 35, 89 (1976).
[6] E. S. Olivetti et al., ”Effect of Ti substitution on α and β phase formation and properties in Mn50-xTixBi50”,
Journal of Alloy and Compound 643, 270 (2015).
[7] M. Kuepferling et al., ”Rate dependence of the magnetocaloric effect in La-Fe-Si compounds”, EPJ Conf.
40, 06010 (2013).
278
STUDY ON COMPOSITE OF THERMOPLASTIC LOADED BY
MAGNETOCALORIC POWDERS
J. Lanzarini (a)*, J.C. Gelin (a), T. Barriere(a), M. Sahli(a), A. Dubrez(b)
(a)
Institute Femto-ST/Department of Applied Mechanics (Besançon, France
(b)
Erasteel SAS (Paris, France)
*
Corresponding author. E-mail: julien.lanzarini@femto-st.fr
ABSTRACT
Magnetic refrigeration is an innovative and promising alternative to traditional refrigeration systems.
However, the production of magnetocaloric components used as regenerators in magnetic refrigeration
systems is a significant challenge. This work consists of developing an innovative process to obtain
magnetocaloric components by powder extrusion moulding. A component containing up to 91.5 wt%
magnetocaloric powder was obtained using a gas atomised magnetocaloric powder La-Fe-Si alloy. The
powder is first mixed with a thermoplastic polymer to obtain a “ready to shape” feedstock. Then, the
-
feedstock is either extruded to obtain the magnetocaloric component. Characterisations of the thermo
physical properties are carried out during all stages of processing, from pure powder to injected parts. A
study on the modification of the Curie temperature as a result of dehydrogenation was performed during each
processing stages. Good processability was demonstrated for the highly loaded feedstock and conservation of
the magnetocaloric properties during the process.
Keywords: Composites, Functional polymer, Magnetocaloric properties, Rheology, Extrusion process
DOI: 10.18462/iir.thermag.2016.0227
1 INTRODUCTION
The gases used in conventional refrigeration are harmful to the environment because of their contribution to
the greenhouse gas effect. The aim of magnetocaloric materials and magnetic refrigeration at room
temperature is to replace the current refrigeration technologies in use today. Magnetic refrigeration is based
on cycles of magnetisation/demagnetisation around a magnetocaloric material using a heat transfer fluid.
This technology used the magnetocaloric effect of some materials. This effect is maximised around a
temperature known as critical order-disorder Tc (i.e., the Curie temperature) [1]. La(Fe,Si)13 alloys are
magnetocaloric materials which are among the most promising for magnetic refrigeration applications.
Conventional methods, such as ablation processes by milling or electro discharge manufacturing, do not
permit the production of components on an industrial scale. The solution is to shape the magnetocaloric
components via a composite material using a replication process (e.g., mixing and extrusion) [2]. In the
present study, La(Fe,Si)13 powders were used in the production of micro-composite mixtures. A previous
aging study of the powders highlights the impact of the processing temperature on hydrogen content [3]. In
this case, dehydrogenation can result in a shift towards lower temperatures of Curie temperature (Tc.) which
is the identified sensitive parameter. The purpose of this study is to develop functional composite
magnetocaloric mixtures. A twin-screw mixer was used to make develop the explorative mixtures according
to different formulations of the thermoplastic polymers and magnetocaloric powders. The study was
extended by performing a rheological analysis on the formulated composite magnetocaloric mixtures. The
evolution of the shear viscosity versus the shear rate and temperature is analysed. Understanding the flow
behaviour is essential for shaping this micro-composite material by extrusion.
2 MATERIALS
2.1 Powder and binder characteristics
The magnetocaloric powders are studied using the formulation La1-x (Ce, Pr)x ((Fe1-z-v Mnz Co v )1- y Si y ) w H n .
These magnetocaloric powders are produced from atomised powders of La(Fe,Si)13 alloys and provided by
ERASTEEL. The particle sizes are determined by laser diffraction, with d10, d50 and d90 being the powder
particle size of the cumulative volume percentages equal to 13 µm, 52 µm and 143 µm, respectively. The
binders are two thermoplastic polymers, polypropylene (PP) and low density polyethylene (LDPE). These
polymers are frequently used in several applications including auto manufacturing, construction, electronic
279
devices, etc. To preserve the magnetocaloric properties of the powders, the polymer melting temperature
must be low. This criterion is met by the PP grade and the LDPE grade used where the melting temperature
range are equal to 150-160°C and 100-110°C, respectively.
2.2 Mixing characterization

The mixing tests are performed using a twin screw mixer which is equipped with “Z-blade” screws. This
screw shape leads to homogeneous mixing of the loaded polymer. The screw rotation (measured in rpm) and
mixing temperature are monitored. The critical powder volume loading for the magnetocaloric powders is
determined using an incremental test approach [4]. This starts with a half-full tank of a mixture of PP and
magnetocaloric powders. An initial powder volume loading is set first and increased for each iteration by
adding 4 vol% of the powder during the required mixing time of 80 min. Each time powder was added, the
peak mixing torque is recorded. Then, the screw torque decreases from the peak values to reach the
equilibrium state as homogenisation of the mixing process takes place. As the loading volume increases, the
mixing torque peaks increase. Instability in the mixing torque means that the mixing homogeneity becomes
more difficult to achieve, as shown in Figure 1 (a). This phenomenon occurs when the critical powder
volume loading is reached [4]. This instability corresponds to an excessive powder volume loading and
results in an inhomogeneous micro-composite. The experimental temperature increases higher than the
melting temperature, 180°C, and the screw speed was 50 rpm. The initial mixing volumetric load rate is
70%. After successive incremental tests, volumetric loading rates of 74 vol%, 78 vol% and 80 vol% were
achieved. At 80 vol% powder loading, the mixing torque increased drastically, from 5 Nm to 15 Nm [5].
Therefore, the critical powder volume for the magnetocaloric powder is equal to 80 vol%, as shown in Figure
1 a. The same study is performed with the LDPE, only the melting temperature changes where it is fixed at
140°C. In this case, the critical powder volume is between 77 vol% and 79 vol%, as shown in Figure 1 (b).
Figure 1. Evolution of the mixing torque associated with a continuous increase in the powder loading rate (a) PP, screw
speed: 50 rpm, mixing temperature: 180°C and (b) LDPE, screw speed: 50 rpm, mixing temperature: 140°C
3 RHEOLOGICAL ANALYSIS
The rheological measurements are performed using For capillary rheology measurements, a Rosand RH2000
double piston capillary rheometer with a capillary of 1 mm in diameter and 16 mm in length. The
measurements are carried out between 100 s-1 and 1000 s-1, which corresponds to the shear rate range for
extrusion processes [6]. The tests temperatures are between 120°C and 160°C and between 160°C and 200°C
for the composite using LDPE and PP, respectively, which are in accordance with the melting temperature
range for each polymer. Here, the rheological constraint of powder loading is investigated, which
corresponds to the maximum powder loading that will permit the mixture to flow. Four mixtures of PP are
formulated with different powder volume loadings as follows: 65 vol%, 70 vol%, 73 vol% and 75 vol%. The
tests show that mixture flow becomes impossible to push through the die for a powder volume loading of
75%, as shown in Figure 2 (b)). Thus, the rheological limit of the powder loading rate is between 73% and
75%. The same limit of the powder loading rate is obtained for the composite using the LDPE. Notably,
increases in the powder volume loading lead to increases in the shear viscosity of the micro-composite
mixtures (Figure 2 (a)). The shear viscosity of the mixtures decreases when the shear rate increases (Figure 2
(a)). Therefore, this magnetocaloric composite is a pseudo-plastic material [7].
280
Figure 2. (a) Comparison of the mixture shear viscosity based on the powder loading rate at 180°C (logarithmic scale)
and (b) evolution the shear viscosity as a function of the powder loading rate, test parameters: 180°C, shear rate: 1000/s
In our case, similar work is carried out to study the influence of temperature on the flow of the mixtures
made using magnetocaloric powders. The rheological tests are performed at three temperatures: 160°C,
180°C and 200°C. The results (Figure 3 b) show a variation in the shear viscosity versus temperature. When
the temperatures increase, the shear viscosity decreases [8]. In any case, the shear viscosity is 100 Pa.s to
1000 Pa.s when the shear rate ranges from 100 s-1 to 1000 s-1. This range of shear viscosities is characteristic
of this type of viscous material.
Figure 3. Comparison of mixtures shear viscosity based on the test temperature (powder loading rate fixed at 65 vol%)
4 THERMO-PHYSICAL ANALYSIS
The Curie temperature depends upon the composition of the magnetocaloric powder and may depend on the
proportion of hydrogen in the material [9]. Studies have been conducted on aging the hydrogenated powders
[3]. The results showed a dehydrogenation of the magnetocaloric powders after several months of storage as
measured by differential scanning calorimetry (DSC). This effect results in a shift in the Curie point towards
the left in relation to the Tc value at time t0 [3]. In additional work conducted on these magnetocaloric
composites, we studied the influence of the mixing temperature on the magnetocaloric properties of the
composite. The DSC measurements were carried out using a 92 Setaram DSC system.
Figure 4. Evolution of the Curie point (Tc) as a function of the mixing temperature for a mixing time of 20 minutes
281
The results were obtained over a temperature range from -60°C to 20°C at a heating rate of 10°C/min. The
composite mixture samples of magnetocaloric powders and polypropylene (PP) were prepared at three
temperatures, namely 160°C, 180°C and 200°C, with a mixing time of 20 minutes. The Curie point of the
mixture carried out at 200°C is offset from the mixtures formulated and tested at 160°C and 180°C, as shown
in Figure 4. The magnetocaloric properties of the micro-composite mixed at 200°C are shown to have been
modified. This indicates the first constraint application applications containing hydrogenated powder.
5 CONCLUSIONS
In this study, rheological and thermo-physical of a magnetocaloric composite were performed. The results
validate the ability of thermoplastic polymers to be mixed and loaded with magnetocaloric powders. The
rheological study of the magnetocaloric micro-composite highlighted the pseudo-plastic behaviour of the
material. The introduction of powders into the thermoplastic matrix increased the shear viscosity
proportionally to the powder volume loading. The shear viscosity over a certain range of shear rates is
compatible with the replication process for mixing and shaping by extrusion for the formulated composite. A
thermodynamic analysis using DSC shows the importance that the formulation mixture temperature has on
the final magnetocaloric properties. The DSC measurements carried out on mixtures at 200°C indicate a shift
in the Curie temperature. The Curie temperature of the initial magnetocaloric powder is partly defined by the
amount of hydrogen that is introduced in the alloy. Therefore, a small amount of hydrogen was released into
the atmosphere during the process of formulating the mixture. The work carried out here has led to the
development of composite mixtures that are highly loaded with powder, and the achieved powder volume
loading was on the order of 65 vol%. The obtained mixtures have interesting magnetocaloric properties that
may be used in the operation of a system based on the AMR cycle. However, the composite formulation
must be optimised to provide better performance. The formulation phase of the composite coupled to an
extrusion shaping process seems to be an efficient solution for manufacturing the magnetocaloric
components. The developed process can be adapted to industrial applications, which takes into account the
temperature constraints of the hydrogenated magnetocaloric powder.
ACKNOWLEDGMENTS
The authors wish to acknowledge the French OSEO – ISI program for their financial support.
REFERENCES
[1] A.M. Tishin, Y.I. Spichkin, The magnetocaloric effect and its applications, Institute of Physics Pub.,
Bristol; Philadelphia, 2003.
[2] J. Ma, M. Qin, L. Zhang, L. Tian, R. Li, P. Chen, X. Qu, Effect of ball milling on the rheology and
particle characteristics of Fe–50%Ni powder injection molding feedstock, J. Alloys Compd. 590 (2014)
41–45.
[3] C. Mayer, A. Dubrez, M. Pierronnet, P. Vikner, Towards the large scale production of
(La1-z Cez)(Fe1-x-yMn y Si x ) 13 H n products for room temperature refrigeration: Towards the large scale
production of (La1-zCez )(Fe1-x-yMnySix)13Hn products for room t emperature refrigeration, Phys. Status
Solidi C. 11 (2014) 1059–1063.
[4] X. Kong, T. Barriere, J.C. Gelin, Determination of critical and optimal powder loadings for 316L fine
stainless steel feedstocks for micro-powder injection molding, J. Mater. Process. Technol. 212 (2012)
2173–2182.
[5] C. Quinard, J. Song, T. Barriere, J.C. Gelin, Elaboration of PIM feedstocks with 316L fine stainless steel
powders for the processing of micro-components, Powder Technol. 208 (2011) 383–389.
[6] D.E. De Laney, J.F. Reilly, Rheology Conference: A Practical Approach to Quality Control: a Two-day
Conference, iSmithers Rapra Publishing, 1998.
[7] J.-F. Agassant, P. Avenas, J.-P. Sergent, B. Vergnes, M. Vincent, La mise en forme des matières
plastiques, 3ème édition, Technique & Documentation - Lavoisier, 1996.
[8] F. Thiébaud, J.C. Gelin, Characterization of rheological behaviors of polypropylene/carbon nanotubes
composites and modeling their flow in a twin-screw mixer, Compos. Sci. Technol. 70 (2010) 647–656.
[9] J.L. Zhao, J. Shen, H. Zhang, Z.Y. Xu, J.F. Wu, F.X. Hu, J.R. Sun, B.G. Shen, Hydrogenating process
and magnetocaloric effect in La0.7Pr0.3Fe11.5Si1.5C0.2Hx hydrides, J. Alloys Compd. 520 (2012)
277–280.
282
MULTIDIMENSIONAL NUMERICAL METHOD FOR PERFORMANCE
PREDICTIONS OF AN AMR CYCLE USING ANSYS FLUENT SOFTWARE
K. Hamdani, A. Smaili*
Laboratoire de Génie Mécanique et Développement
Ecole Nationale polytechnique
P.B. 182 El-Harrach Algiers, Algeria
*
Corresponding author. E-mail: arezki.smaili @g.enp.edu.dz
ABSTRACT
This work deals with the development of numerical simulation method of an active magnetic regenerator
(AMR) cycle by using advanced programming tools of ANSYS-Fluent code. For this purpose, the following
tools: scheme commands, RP-variables (RP refers to user-specifiable model macros in Fluent), and user
defined functions (UDFs) have been considered and used. These tools are expected to enhance the ability of
the software and to raise its degree of automation significantly, and thus leading to powerful numerical
simulation approach. The validity of the proposed numerical approach has been investigated through 2D
parallel plate regenerator model. Preliminary results that include regenerator temperature span and
temperature gradient and fluid velocity profile are presented and discussed.
Keywords: Magnetic refrigeration, Active magnetic regenerator, Numerical simulation, Thermal analysis.
DOI: 10.18462/iir.thermag.2016.0243
1. INTRODUCTION
Magnetic refrigeration (MR) motivated by environmental and economical considerations is a potential

candidate to replace conventional refrigeration systems. The most promising MR devices are based on active
magnetic regenerator (AMR) cycle. The numerical modeling of AMR cycle plays a key role in the design
and thermal optimization as well as developing new cooling devices. Numerical models for MR generally
deal with several tasks. Note here the time dependency (steady versus transient regime),space dependency
(1D, 2D, 3D), heat transfer fluid flow and its properties, MCE implantation (built-in, discrete), magnetic
field profile (discrete on/off, continue), solid regenerator thermal properties (constant, temperature and
magnetic field dependent), axial conduction and pressure drop[1].To overcome the points cited above,
simplifications and assumptions are considered in many models to reduce the complexity of the problem. 1D,
steady state, constant fluid and solid properties, constant and instantaneous MCE, discrete on-of magnetic
field profile, the average weighted temperature of entering flow at hot and cold sides are not evaluated with
respect to mass flow, neglected pressure drop and viscous heating. The approach developed in the present
work aims to propose a complete and detailed model able to reproduce the real behavior of the AMR and
make the model a powerful tool for predicting AMR cycle performance by using advanced tools of ANSYS-
Fluent software.
2. NUMERICAL MODEL
AMR cycle consists of four processes or steps; namely, magnetization and demagnetization steps, by
application and removal of a magnetic field (through adiabatic process), as well as cold and hot blows (i.e.
heating and cooling the circulating fluid). These processes are repeated several times to amplify temperature
span within a solid refrigerant according to AMR principal introduced by Barclay in which the solid plays
both roles of refrigerant and regenerator [2], [3]. The proposed approach consists in modeling the AMR
cycle (i.e. four processes) in scheme language files to be linked dynamically within ANSYS Fluent solver
when it is running [4]. Note that UDFs are used only to initialize the problem, evaluate the outlet fluid
temperature and define MCE as source term [5]. While the transition aspect and cycle simulation are assured
bay RP-variables defined within scheme files; this therefore constitutes the main contribution of the
presented work. Previously AMR cycle simulation method using Fluent code has been presented in Thermag
III [6]; in which the UDFs are the warhorse, the MCE is implanted directly and instantaneously as constant
temperature difference, and the fluid outlet temperature is introduced by taking the area averaged values at
283
the end of hot/cold blow steps. Also the cycle simulations cannot operate in an automatic way; it requires the
intervention of the user many times during the calculation.
2.1 Mathematical Model

In an AMR device the fluid flow is governed by continuity and Navier-Stokes equations given by
respectively Eq. (1) and Eq. (2)for incompressible and laminar flow conditions. The heat transfer between
fluid and solid mediums are governed by energy equations for fluid and solid. Eq. (3) and Eq. (4) represent
the energy conservation for fluid and solid mediums respectively
∇ · = 0 (1)
1
+ ( · ∇) = ( − ) (2)
· , + ( · ∇) = ∇· ∇ + Φ (3)
· , = ∇ · ( ∇ ) + ̇ (4)
Where , , , ,∇, , , , , , , ̇ , , , , Φ, ( · ∇) refer respectively to time, fluid velocity, fluid
temperature, solid temperature, divergence operator, fluid density, solid density, fluid viscosity, pressure,
fluid thermal conductivity, solid thermal conductivity, the solid source term, the fluid specific heat, the solid
specific heat, dissipation function (viscous heat) and velocity advection operator.
The source term for modeling the heating and cooling effect resulting from magnetization and
demagnetization processes respectively is given by Eq. (5).It takes positive or negative sign depending
weather the AMR step is magnetization or demagnetization.
. , ( , ) . ∆ ( , )
̇ = (5)
In Eq. (5), ∆ is the adiabatic temperature change, is the magnetization / demagnetization time duration
and B is the applied magnetic field.
2.2 Numerical Method and AMR Cycle Simulation

As mentioned in the introduction AMR cycle requires an intermittence of fluid flow, to perform that a
scheme program has been considered to attribute the appropriate boundary type for the fluid domain (inlet
and outlet). During cold blow process, the cold side will have inlet-velocity type and hot side will have
pressure-outlet type and for hot blow this configuration is reversed. This manipulation is done by a scheme
file linked dynamically to ANSYS Fluent solver. To accomplish and produce regeneration effect on the
magnetic solid the temperature of the entering fluid for each fluid flow (cold/hot blow) steps is the mass-time
weighted average temperature of the exiting fluid (bulk temperature of the total amount of leaving fluid) of
previous fluid flow step. This is done by a UDF with RP-variable linked to the solver via a scheme command
to be called at the end of each time step during transient calculation. During magnetization/demagnetization
steps both cold and hot fluid sides will have a wall type (adiabatic) and velocity of fluid domain is fixed to
zero ( stationary fluid), with this configuration the stagnated fluid will still exchange the heat with the solid
while all the system is insulated. The source term is enabled during magnetization/demagnetization and
disabled for cold/hot blow processes using RP-variable and UDF tools. The solid will have all walls
adiabatic. A coupled wall type is attributed to Fluid-solid interface shared between them which physically
consist on equal heat flux for both sides and one identical temperature value as shown by Eq. (6) and Eq.
(7).Finally the sequencing of the different AMR cycle steps and repeating them is performed by scheme file.
− = − (6)
= (7)
Where denotes for y axis and is the interface y coordinate in Cartesian coordinate system.
2.3 Preliminary Results

The validity of the proposed numerical approach has been investigated through the 2D parallel plates AMR
model (see Fig. 1), by considering zero cooling and heating loads. The Gadolinium (Gd) is used as magnetic
regenerator, assuming constant thermo-physical properties ( , = 270 , = 7900 / and =
·
284
10.6 ). The water is considered as heat transfer fluidwithout viscous heat,initiallythe system is at 298K
·
(fluid & solid). The cycle parameters are given in Table 1
Figure 1.ParallelAMR 2Dgeometry (not to scale).
The computational domain and grid topology of AMR system is shown by Fig. 2with a structured mesh of
100 000 cells (60 000 for fluid and 40 000 for solid) refined at inlet/outlet and solid-fluid interface to capture
high gradients. A time step of 0.01s is used for temporal discretization. Pressure-based solver, SIMPLE
method with second order upwind spatial discretization and second order implicit transient formulation
(unconditionally stable) have been chosen. Notice that because of disparate length scales of the AMR system
(very long and thin), single-precision calculation may be not adequate, so the use of double-precision will
lead to more accurate results.
Step Time (s) MCE (K) Velocity (m/s)
Magnetization τ1= 0.5 2 0
Cold blow τ2= 1.0 0 0.05
Demagnetization τ1= 0.5 -2 0
Hot blow τ2= 1.0 0 0.05
Cycle c= 3.0 Source term intermittence
Table1.AMR cycle parameters.
The simulations have been set to run for 100 cycles to attain the steady state conditions of AMR cycle.
Temperature span obtained between the hot and cold regenerator sides is about 16 K as show in Fig. 3. As it
can be seen the steady state conditions seems to be attained around 200s.Fig. 4 shows the evolution of the
temperature gradient within the regenerator which is establishing gradually from zero at 0 cycles until its
maximum value of 16 K after70 cycles. The effect of the entrance region is examined in Fig. 5to check
velocity profile for hydrodynamic developing and fully developed regions.
Figure 2.Computational domain and grid topology.
Figure 3.Regenerator temperature span.
285
Figure 4.Regenerator temperature gradient.
Figure 5.Velocity profile for entrance region.
3. CONCLUSIONS
The preliminary tests of the proposed numerical simulation have reproduced successfully the operating mode
of an AMR system and the degree of automation is raised to 100%. The method can be a powerful tool in
developing MR new devices. However, it is important to note that simplifications such as assuming a
constant adiabatic temperature change and constant specific heat are not thermodynamically consistent.
[3]Further development on more detailed numerical model that take into account most of physical and design
problems of AMR devices is in progress. Also, with such numerical approach 3D AMR systems can be
tackled and properly investigated.
REFERENCES
[1] K. Nielsen et al., «Review on numerical modeling of active magnetic regenerators for room temperature
applications», International journal of refrigeration, vol. 34, pp. 603-616, 2011.
[2] J. A. Barclay, «The theory of an active magnetic regenerative refrigerator», Los Alamos National
Labratory-LA-UR-83-1251; CONF-821237-1, New mexico, 1982.
[3] A. Smaili and R. Chahine, «Thermodynamic investigations of optimum active magnetic regenerators»,
Cryogenics, vol. 38, pp. 247-252, 1998.
[4] R. Kent Dybvig, The Scheme Programming Language, MIT Press, 1997.
[5] ANSYS, Inc., ANSYS FLUENT UDF Manual, 2011.
[6] U. Legait, A. Kedous-Lebouc et L. Rondot, «Numerical simulation and analysis of the refrigerant bed
behavior using Fluent software», Third International Conference on Magnetic Refrigeration at Room
Temperature, Des Moines,United States, 2009.
286
SIMULATION OF A HYDRAULIC CIRCUIT FOR A MAGNETIC
REFRIGERATOR
T.R.V. Ebel, J.A. Lozano, P.O. Cardoso, J.R. Barbosa Jr.*
*Corresponding author. E-mail: jrb@polo.ufsc.br
ABSTRACT
The flow distribution system plays a key role in the operation and the overall performance of a magnetic
refrigerator, but only a few works in the literature are dedicated to the design and evaluation of it. Therefore,
the focus of this work was to evaluate the transient behavior of a flow management system by means of
numerical simulations of a simplified hydraulic circuit of the magnetic refrigerator developed at
POLO/UFSC using a commercial software package. Water hammers were predicted to occur after switching
the fluid flow direction, and their magnitude was proportional to the volumetric flow rate. The performance
of the hydraulic system showed a dependence on the valve opening ramp; while fast ramps increased the
magnitude of the water hammer, slow ramps led to fluid flow bypassing the AMR. The use of a proportional
returning valve upstream of the high-pressure valves is recommended to absorb the impact of the water
hammers and make the flow through the regenerator beds smoother.
Keywords: Magnetic refrigeration, hydraulic circuit, numerical simulation.

DOI: 10.18462/iir.thermag.2016.0247
1. INTRODUCTION
Engineering design can be a valuable tool in the development of more efficient magnetic refrigeration
systems. From an engineering system perspective, a magnetic refrigerator can be divided into three main
sub-systems, namely the fluid flow management system, the magnetic field generation system and the active
magnetic regenerator (AMR). In the open literature, the flow management (or hydraulic) sub-system has
been the least investigated of the three. Nevertheless, it can be responsible for a significant fraction of the
total losses in magnetic cooling devices [1,2,3]. The present study advances a computational simulation in a
commercial software package of the time-dependent behavior of the hydraulic sub-system of the prototype
developed by Lozano et al. [4]. The focus is to understand the most important characteristics of the hydraulic
sub-system that should be considered in the design of new prototypes.
2. COMPUTATIONAL SIMULATION
A simplified hydraulic circuit model of the magnetic refrigerator prototype developed by Lozano et al. [4]
was carried out in this work. The model was implemented using the Computer-Aided-Engineering (CAE)
software package Flowmaster [5].
The prototype developed by Lozano [6] has 16 regenerator beds; however, only 2 regenerator beds are
sufficient to describe the basic system dynamics. A schematic diagram of the simplified flow management
system with 2 regenerators is shown in Fig. 1. The system layout implemented in Flowmaster is shown in
Fig. 2, with the appropriate measurement (reference) points. The most important discussions on the flow
behavior are related to Node 1 (upstream the high pressure rotary valve) and Node 5 (downstream the low
pressure rotary valve). The numbers makers in Figs. 1 and 2 regard to the same hydraulic nodes.
The purpose of the rotary valves is to control the direction of the fluid flow. While one side (coming from the
pump) has a rotating oblong, the other (going to the AMR) has orifices. The rotary valve and the magnet are
connected to the same shaft and have the same operating frequency. The opening (or closing) ramp is related
to the duration of the opening (or closing) of the rotary valve orifices, and it is expressed as degrees out of
one revolution (360º) to describe it independently of the operating frequency. The fluid switching time used
287
in [6] corresponded to 8º out of 180º. In this time interval, one of the orifices is closing while the other is
opening, thus giving the system its opening ramp characteristic. In the computational simulation, the valves
were modeled as gate valves in each orifice. In order to model the rotary valves in the software, each one is
modeled as a two gate valves with the opening times synchronized.
Figure 1. Schematic diagram of the simplified flow management system of a magnetic refrigerator [6].
Figure 2. System layout implemented in the software Flowmaster [5].
Assuming water as the working fluid, the connecting hoses have been modeled as flexible, with an internal
diameter of 6 mm and a wall thickness of 2.4 mm. The regenerators were modeled as packed-sphere beds,
for which the pressure drop can be determined by the Ergun correlation [7]. For practical reasons, the system
has not been considered a closed circuit. Rather, a fixed inlet mass flow rate was assumed. This approach
facilitates focusing on the dynamic response of the fluid switching device. On the other hand, a fixed
pressure boundary condition was used as the exit, which resembles the use of a reservoir open to atmospheric
pressure.
3. RESULTS
The model has been evaluated at different operating conditions in order to better understand the behavior of
the hydraulic circuit and its key characteristics. The analysis has been carried out for different mass flow
rates and opening ramps.
Figure 3 shows the behavior of the time-dependent absolute pressure behavior at Node 1 for an AMR
frequency of 1.0 Hz (which corresponds to the valves operating with blow durations of 0.5 second), an
288
opening ramp of 8º and volumetric flow rates of 50, 120, 170 and 200 L/h. A sharp decrease in pressure is
observed at Node 1, followed by severe oscillations when the rotary valves switch direction – when the blow
changes its direction, switching the orifices of the rotary valves. It can be inferred that the absolute pressures
at Nodes 1 and 5 tend to equalize while switching the direction of the blow. In addition, the fluid finds less
resistance to bypass the regenerators when both valves on the same line are opened for flow. During this
period, inside the regenerators, the fluid decelerates and is rapidly forced to flow in the opposite direction.
Once one of the high pressure valves is totally closed for flow (Node 1), a water hammer occurs within the
circuit. The simulated maximum pressure is more than twice the operating system high pressure for different
flow rates. This causes the many pressure oscillations which may propagate and harm the hydraulic system
or the device structure. As can be seen, the higher the mass flow rate the greater the intensity of the water
hammer and its effects.
Figure 3. Transient absolute pressure at Node 1 for different volumetric flow rates.
The duration of the opening ramp has been demonstrated to be a key parameter in the behavior of the
hydraulic system. Therefore, different scenarios have been evaluated. For a constant volumetric flow rate of
120 L/h and an operating frequency of 1.0 Hz, the pressure behavior at Node 1 has been quantified for
different ramp durations of 2º, 4º, 8º, 15º and 45º. Due to space restrictions, the results are not shown here.
However, it was inferred that the shortest opening ramps result in a quick drop in pressure and a severe water
hammer, whereas for the longer opening ramps the pressure in the high-pressure line drops during longer
times and the water hammer is less strong. Opening ramps shorter than 8º (roughly 2% of an AMR cycle)
tend to have greater influence on the severe water hammers on the circuit. On the other hand, for opening
ramps larger than 15º (roughly 4.2% of an AMR cycle) the fluid finds less resistance to bypass the
regenerators during the switching time, which reduces the water hammer effect but acts toward reducing the
system cooling capacity because it is not flowing through the regenerators. Based on the different scenarios
evaluated, the best overall performance was found for opening ramps between 8º and 15º.
Severe water hammering and oscillations in the hydraulic circuit are undesirable not only from a structural
reliability point of view, but also because they may reduce the thermal effectiveness and increase the fluid
pumping losses through the AMR. Therefore, to improve the hydraulic performance of the system, the use of
a proportional return valve with a low trigger pressure (4.0 bar, based on the working pressures for different
flow rates) is proposed. Additional simulations indicated that this device modifies the dynamic behavior of
the circuit, absorbing transient effects. Figure 4 shows the comparison between the transient absolute
pressures associated with the original and newly proposed layout. One benefit of the new layout is that it
makes the flow through the regenerator beds more stable, having a more well-behaved fluid flow profile. The
drawback is that the new layout demands a higher pumping power because it requires a higher flow rate at
the pump to maintain the same flow rate through the regenerator beds.
289
Figure 4. Comparison between the pressure at Node 1 for the original circuit and the proposed new layout.
4. CONCLUSIONS
A commercial CAE software package was used to simulate different operating scenarios of the fluid flow
management system of the prototype developed by Lozano [6] in order to better understand its dynamic
behavior. Different flow rates and valve opening ramps were evaluated. The magnitude of the water hammer
effects was proportional to the mass flow rate. In addition, the duration of the opening ramps played a key
role in the hydraulic circuit behavior: for opening ramps shorter than 8º, the circuit tends to have great
influence of the water hammer effect. On the other hand, for opening ramps larger than 15º the fluid finds
less resistance to bypass the regenerators when switching the fluid flow direction. The best overall
performance was found for opening ramps between 8º and 15º. A new layout of the hydraulic circuit was
proposed to improve the performance of the refrigeration system. Based on the obtained results, a
proportional returning valve before the high pressure valves and after the low pressure valves is proposed.
Such a device could absorb the impact of the water hammer effects and make the flow through the
regenerators beds smoother. However, since it increases the pumping power, the overall impact of this
modification on the system performance needs to be evaluated in further studies.
ACKNOWLEDGMENTS
Financial support from CNPq and Embraco is duly acknowledged.
REFERENCES
669 (2013).
[3] M.S. Capovilla et al., “Performance evaluation of a magnetic refrigeration system”, Science and
Technology for the Built Environment 22, 531 (2016).
[5] Flowmaster V7, Mentor Graphics, Inc. Wilsonville, OR, USA (2015).
[6] J.A. Lozano Cadena, “Designing a Rotary Magnetic Refrigerator”, PhD thesis, Federal University of
Santa Catarina (2015). Available: http://tede.ufsc.br/teses/PEMC1635-T.pdf
[7] S. Ergun, “Fluid flow through packed columns”. Chem. Process Eng. London 48, 89 (1952).
290
FUTURE DEVELOPMENTS IN MAGNETOCALORIC
REFRIGERATION AND HEAT PUMPING
A. Kitanovski*, U. Tomc, A. Poredos
University of Ljubljana, Faculty of Mechanical Engineering/Askerceva 6, 1000 Ljubljana,

Slovenia
*Corresponding author. E-mail: andrej.kitanovski@fs.uni-lj.si
Since the earliest days of engineering magnetocaloric devices for applications near or above room
temperature, a total of about 70 prototypes have been developed in different parts of the world. Despite
the fact that we have seen an exponential growth in the number of publications in this field, it is not easy
to relate these activities to the S-curve of technology development. A multitude of articles based on
minor modifications to already-know compositions and a large number of repetitions of earlier scientific
or engineering achievements tend to blur the transition from science to technology, with future directions
being unclear. There is strong competition among involved institutes and industries, but with a low level
of open, trust-based international collaborations. Simultaneously, important technical problems remain
unsolved or, in some cases, even untouched. It is therefore not a coincidence that we tend to see relatively
small steps towards the commercialization of magnetocaloric technology. However, there are different
engineering solutions that can really foster developments towards the first real market applications.
This contribution does not represent a scientific paper, but rather a critical review of activities and
achievements in the development of magnetocaloric refrigeration and heat pumping. It addresses some
of the very important scientific, technical and non-technical factors. These can provide a basis for future
improvements. Moreover, the most important technical obstacles are discussed and potential solutions
are outlined by pointing out their weaknesses, advantages, limits, as well as their future applicability.
Keywords: Magnetocaloric, Refrigeration, Heat pump, Cooling, Energy efficiency

DOI: 10.18462/iir.thermag.2016.0250
1. INTRODUCTION
Magnetocaloric refrigeration and heat pumping can theoretically, and with the absence of heat-transfer
and viscous irreversible losses, lead to close to Carnot efficiency. Compared to the irreversible
polytropic compression of a gas refrigerant and its isenthalpic expansion in vapour compression,
magnetocaloric cooling offers an opportunity for a substantial increase in the energy efficiency of future
refrigeration and heat-pump devices. However, by taking into account the irreversible losses due to the
heat transfer between the magnetocaloric material and the working fluid, as well as the viscous losses
of the fluid, the efficiency of existing magnetocaloric devices becomes very low. Moreover, the
mechanical transmission losses, the motor's and the pump's inefficiency, additional heat exchangers, and
the losses in auxiliaries, represent a further obstacle that prevents magnetocalorics from reaching the
market. This is particularly true for its heavy and expensive parts, which mostly relate to the rare-earth
permanent magnets (PMs) [1].
A number of existing publications in magnetocalorics characterize vapour compression as a mature and
highly inefficient technology, with an exergy efficiency of about 10–15 %. Of course, this is not true
and such statements should be avoided. Vapour compression is a very efficient technology, which for
larger devices exceeds an exergy efficiency of 50%. Furthermore, with the additional manufacturing
costs of compressor-based devices, even further improvements are possible. On the other hand, one can
keep in mind an important advantage of magnetocalorics over compressor cooling, which is related to
very small units (scale of household refrigeration), for their potential silent operation and as a solution
with environmentally friendly refrigerants. Theoretically, magnetic refrigeration can actually lead to
291
more compact devices than those using vapour compression. But there is a lot of work waiting for
researchers to bring this technology to the market.
The number of publications (including those of the authors of this article) in magnetocalorics
characterizes the technology as one with exponentially growing research activities. Such statements are
mostly based on the number of publications per year. However, a closer look at such publications reveals
that there is only an exponential growth of publications in material science. There are a surprisingly
small number of engineering activities, which are not seeing exponential growth (Fig. 1).
500
450
Number of papers per year
Material Science Engineering

400
350
300
250
200
150
100
50
0
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015
Year
Figure 1. Number of articles found on the ScienceDirect (keyword: magnetocaloric).
What we actually observe is a plethora of articles published in material science, which are mostly related
to the tuning of already known compounds. In certain cases the repetition of earlier achievements can
be observed. Moreover, research is performed on classes of materials that we already know will not lead
to useful applications [2, 3]. Therefore, the breakthroughs that were made by [4], and also by [5, 6, 7]
and others, are certainly missing. The community should become more careful about publishing research
in materials science.
Some groups are being too optimistic or being economical with the truth, which can be very dangerous
when it comes to losing confidence in the technology, not only with industry, but also with policy makers
and the community in general.
Most of the European engineering groups have experienced a large number of failures with respect to
calls offering funds from the European Commission. Engineering is not well supported by the H2020
programme, nor was it in FP7. This is despite the fact that with basic science one cannot develop new
market technologies, and that some very basic engineering problems need to be solved in
magnetocalorics (see also Table 1). With a lack of financial resources the development of
magnetocalorics is slow, and most of the EU’s engineering groups have to rely on national funds or
industrial support.
We can also observe a large gap between material science and engineering. In certain cases materials
scientists are trying to engineer devices, and engineers are trying to process or develop magnetocaloric
materials. But both should actually try to “walk in their own shoes” and collaborate with the exchange
of knowledge.
Materials scientists do not actually receive a clear message from engineers about which material
properties are the most crucial for successful future developments. On the other hand, they do not focus
on the solutions that engineers indicate as being the most important.
292
Many of the existing technical problems have not been solved yet. This is not just the case for the
materials and their processing. In engineering, the main obstacle, i.e., efficient heat transfer, has not
been solved yet. And the compactness (cost and mass) of devices is strongly related to efficient heat
transfer.
Today, the magnetocaloric devices are not market ready, at least not by the opinion of the authors of this
paper, and despite of the fact, that such statements can be found published in different media. Namely,
magnetocaloric devices, as they are today, cannot compete with the existing technologies. They simply
represent too expensive, too heavy, and in many cases even not so energy efficient devices, as this is the
case with the existing technologies on the market. By the opinion of the authors we still deal with the
Technology Readiness Levels TRL 3-4, what is actually far from the market. On the other hand there is
a large room for improvements. But surprisingly small and slow steps of the scientific community to
tackle those challenges.
What actually are the technical problems associated with magnetocalorics? Are there any potential
solutions that can make magnetocalorics one of the important future refrigeration and heat-pump
applications? Which directions or actions should be taken?
2. IDENTIFICATION OF THE TECHNICAL PROBLEMS AND SOME SOLUTIONS
Based on research activities [1, 8] we can indicate the most crucial problems that need to be solved in
both materials science and engineering. Table 1 shows a selection of some of the most important issues
that need to be addressed for future developments.
Part/Feature
Requirement/Solution Comment
of MC device
AMR parallel plate – the ideal
AMR parallel plate – even if this is solved one
thickness of the MC material is
should solve the maldistribution in channels.
about 50 microns, and this also
AMR packed bed – is strongly limited by the
holds approximately for the
friction losses.
voids.
In both cases the frequency of operation (no. of
AMR packed bed – particles
thermodynamic cycles per unit of time) with water
should be fixed (no moving in the
as the heat-transfer fluid will not exceed 5 Hz – if
bed) with a homogenous
the temperature span of device and a high
AMR diameter of 75–100 microns.
efficiency are the target
Do not rotate the AMR!
Additional improvements in heat transfer can be
made by reducing the thermal boundary layer.
Layering is a must!
According to our knowledge, nobody has touched
this issue yet.
Sintered materials, especially
for long-term operation, must
No evidence on cyclic stability of sintered
be bonded with epoxies or
materials (e.g. 100 million cycles).
similar.
Table 1: Most important requirements in magnetocalorics.
293
Part/Feature
of MC device
PM only up to 1–1.3 Tesla. Otherwise, it is too
Tad>3-4 K/Tesla is a must and heavy and too expensive.
by far the most important
feature! Research in materials with Tad< 3-4 K/Tesla
Magnetocaloric
effect A second and less important should be stopped!
feature is the specific heat or
isothermal s of the MC Materials scientists should focus on theoretical
material. investigations of the best material groups and not
only on experimental trial-and-error methods.
Should not provide dissipation of The best option is to avoid oscillation, but this is at
heat into the fluid or should be present only possible with thermal
installed correspondingly – close switches/diodes.
to the hot heat exchanger. The
Pump
market-available pumps do not Pumps should be custom made and this task
provide good efficiency for the demands a relation to engineering groups in such
oscillatory flow of the working domains. According to our knowledge, nobody
fluid. seriously touched this issue yet.
A smart design of magnetizer- Knowledge of PM motors required, demands a
motor as one part will drastically relation to engineering groups in such domains.
Motor
improve the compactness and According to our knowledge, nobody has touched
even the efficiency. this issue yet.
Magnetic field Multipole-rotating and torque No evidence on prototypes build with Alnico.
source balanced (see Motor).
Water with inhibitors is presently the best-known
Impact on the environment!
working fluid for packed-bed AMRs.
High thermal conductivity and
For parallel plates, liquid metals (e.g., Galinstan)
small viscosity (water with
perform better and enable a higher frequency of
inhibitors, liquid metals based on
operation up to 10 Hz. Despite the fact that they are
In–Ga - if in very small amounts)
expensive, the compactness can be substantially
Working fluid improved. According to our knowledge, only in
There exists a possibility of
our group have we touched this issue.
merging the conventional
A combination with the heat-pipe principle
refrigerants with the MC.
(condensation, evaporation) can boost the heat
According to our knowledge,
transfer from the thermal switch/diode to the
nobody has touched this issue
working fluid. According to our knowledge,
yet.
nobody has touched this issue yet.
Parallel plates today require rather long
Temporary, because the
AMR packed regenerators. This of course is also related to the
community did not find a good
bed VS AMR costs of a device. With packed beds for the same
way of processing, packed beds
parallel plate conditions, the AMR can be substantially smaller –
are having an advantage.
smaller required magnetic field source.
The thermodynamic cycle depends on the choice-
quality of the AMR (packed bed or parallel plate)
NEW
Can improve the efficiency and and the working fluids. According to our
thermodynamic
reduce the mass of magnets. knowledge, only in our group have we touched
cycles
this issue with comprehensive numerical and
experimental investigations.
Table 1: Most important requirements in magnetocalorics, …continued
294
Part/Feature
of MC device
Reduce as much as possible. The reduction of mass is of course possible also
Rare earths in Magnets represent much bigger with a higher frequency. But an AMR's upper limit
AMR and problem than MC materials. is about 5 Hz. Going beyond requires new
magnets Did you try or combine Alnico? approaches – presently known are thermal switches
See also the Motor. /diodes
There are many unbelievers in
It requires a new approach in the design of micro-
the community. Very important
systems (not nano!), which can be merged into a
to start immediately with the
macroscopic scale device. Solid-state rectification
integration of this domain. These
is inefficient unless nanostructures are being
Thermal approaches enable us to go far
developed. Solid mechanical contacts represent too
switches/diodes beyond the AMR’s 5 Hz and the
large heat resistance, unless solved on the micro-
targeted frequency is about 20–
scale. Apply micro-fluidics, e.g., electrowetting,
50 Hz. Only a few groups
magnetowetting, MHD, etc.
including ours have touched
this issue.
Table 1: Most important requirements in magnetocalorics, …continued
3. CONCLUSIONS
The magnetocaloric community has a lot of work to do. However, in order not to produce too much
entropy we need to improve our efficiency of research, especially because of the lack of financial and
human resources. Table 1 gives only a small example of numerous problems that need to be solved in
the future. And these numerous solutions can represent a basis for magnetocaloric market-available
applications. Because the groups involved in the research are rather small, our belief is that we should
perform the research with our collective intelligence – which of course requires fair, open, trust-based
international collaboration. And we have to involve many other engineering groups in order to fulfil the
competences in different domains. A number of industries involved in magnetocalorics actually do not
represent a competition, especially because of their different market products. An international business
cluster can offer a good way to foster R&D and substantially reduce research costs. Everybody can win,
or everybody can be a loser. It is up to us – we actually hold the future of magnetocalorics. And this can
only be done by innovation.
Lastly, we would like to add a quote from K.A.Gschneidner Jr.. The message he gave is certainly in line
with the above suggestions. Not only his achievements, but also his curiosity, optimism and positive
energy filled our hearts and minds.
To those that have been working in the field: “The limitations of magnetic refrigeration are only in the
minds of the individual scientists and engineers”
K.A.Gschneidner Jr., the father of magnetocalorics
REFERENCES
[1] A. Kitanovski et al., Magnetocaloric energy conversion: From theory to applications. Series: Green
energy and technology. Springer Publications, (2015).
295
[2] A.M. Tishin et al., “A review and new perspectives for the magnetocaloric effect: new materials and
local heating and cooling inside the human body”, International Journal of Refrigeration, Available
online (2016).
[3] K.G. Sandeman, “Magnetocaloric materials: The search for new systems”, Scripta Materialia Vol.
67, Iss. 6, 566–571 (2012).
[4] V.K. Pecharsky, K.A. Gschneidner Jr., “Giant magnetocaloric effect in Gd5(Si2Ge2)”, Physical
Review Letters Vol. 78, 4494-4496 (1997).
[5] H. Wada, Y. Tanabe, “Giant magnetocaloric effect of MnAs1-xSbx”, Applied Physics Letters Vol.
79, 3302-3304 (2001).
[6] A. Fujita et al., “Giant Magnetovolume and Magentocaloric Effects in Itinerant-Electron
Metamagnetic La(FeXSil_X)l3 Compounds”, Materia Japan Vol. 41, 269-275 (2002).
[7] O. Tegus et al., “Transition-metal-based magnetic refrigerants for room-temperature applications,”
Nature Vol. 415, 150-152 (2003).
[8] A. Kitanovski et al., “Present and future caloric refrigeration and heat-pump technologies”,
International Journal of Refrigeration Vol. 57, 288-298 (2015).
296
ROOM-TEMPERATURE MAGNETIC REFRIGERATION: FROM BASIC
RESEARCH TO DEVELOPMENT FOR APPLICATION
A. T. Saito(a)*, H. Nakagome(b)
(a)
Corporate R&D Center, Toshiba Corporation, 1, Komukai-Toshiba, Saiwai, Kawasaki, 212-8582, Japan
(b)
Department of Urban Environment Systems, Chiba University, 1-33, Yayoi, Inage, Chiba, 263-8522,
Japan
*
Corresponding author. E-mail: aki.saito@toshiba.co.jp
ABSTRACT
Cooling properties of active magnetic regenerative (AMR) refrigeration using spherical magnetic refrigerant
particles of Gd alloys and La(Fe,Si)13-based compounds were investigated by a primitive test apparatus for
the AMR-cycle refrigeration. The maximum temperature difference of 46°C and the lowest cold-end
temperature of −11°C were obtained by AMR-cycle operation using a single refrigerant material, GdY alloy,
under a magnetic field of 1.1 tesla. La(Fe,Si)13-based refrigerants showed high tolerance for heat load due to
their large heat capacity although the temperature difference (T) using La(Fe,Si)13-based refrigerants was
only half that using GdY-alloy refrigerants. Model calculation of AMR cycle indicates that layered
La(Fe,Si)13-based refrigerants having different TC would be effective to enlarge T and obtain good heat-
load properties at the same time. In the AMR cycle, the potential coefficient of performance is expected to
reach 5 under the condition of heat load of tens of watts with the T ~25°C and f ~1 Hz using 1 kg of
magnetic refrigerant.
Keywords: active magnetic regenerative (AMR) refrigeration, Gd-alloys, La(Fe,Si)13-based compounds,

pump-less system.
DOI: 10.18462/iir.thermag.2016.0257
1. INTRODUCTION
Demonstration of continuous operation for 18 months of the room-temperature magnetic refrigeration by

Gschneidner, Pecharsky and Zimm [1] raised great hopes of applying the magnetic refrigeration technique to
general-use refrigerating equipment. It was based on active magnetic regenerative (AMR) refrigeration [2],
which is considered a useful concept for generating a large temperature difference near room temperature,
and used a superconducting magnet to apply a magnetic field to a magnetocaloric material of Gd. In the field
of vapor compression refrigeration technology, the impacts of energy consumption and refrigerant gas
emission on the global environment have become prominent concerns. This is the context in which both
magnetocaloric materials and magnetic refrigerators, which are viewed as an alternative to vapor
compression technology, have been actively investigated in the last two decades. High-performance
magnetocaloric materials [3-6] enable magnetic refrigerators to operate with low magnetic fields of
permanent magnets. The rotary magnetic refrigerator [7] suggested the possibility for downsizing and cycle
operation at higher frequency. Refinement of magnetic circuits and heat exchangers has led to the
development of some prototypes [7-10]. However, since the prototypes were designed with different
applications and target performances in mind, their specifications differ. Therefore, phenomena fundamental
in the AMR cycle have not been elucidated sufficiently. An understanding of these phenomena is important
for improving the refrigeration performance and design in order to attain suitable properties. Thus, we started
with basic research using a primitive test apparatus of the AMR cycle in which various conditions were
varied in order to obtain guidelines for designing AMR refrigerators. The differing effects on refrigeration
properties arising from the difference of physical properties of materials, namely Gd alloys and La(Fe,Si)13-
based compounds, were also clarified using the test apparatus. Moreover, the potential coefficient of
performance (COP) was evaluated and feasible ways to improve the COP were investigated. In this paper,
basic aspects of refrigeration performance and guidelines for designing AMR refrigerators are described.
2. MAGNETOCALORIC MATERIALS
In order to understand the fundamental phenomena in the AMR-cycle, two different types of magnetocaloric
materials were chosen. Materials of the first type, comprising Gd and its alloys, have been thoroughly
297
studied for the application to the AMR refrigerator, and are easy to fabricate and handle. Those of the other
type, namely, La(Fe,Si)13-based compounds, are environmentally benign and expected to have larger
magnetocaloric effect than Gd, but are difficult to fabricate and their compositional homogeneity is difficult
to control. Evaluation of physical properties and preparation of spherical particles of these two types of
materials were carried out for application to the AMR-cycle test.
2.1 Properties of Magnetocaloric Materials

Gadolinium is a typical ferromagnetic material with the Curie temperature TC = 21°C, and shows a large
entropy change around TC. For several kinds of Gd alloys prepared by arc melting, magnetization
measurements were performed. The Curie temperatures of the Gd1-xRx alloys (R = Y, Tb, Dy, Ho, Er)
decrease linearly with increasing x in the range of x < 0.15. Magnetic entropy change Sm was estimated
from magnetization curves using Eq. 1 derived from the integration of the Maxwell relations,
H2
 M 
Sm 

H1
  dH
 T  H
(1)
where Sm and M are the magnetic entropy and magnetization of the magnetic material, T is the temperature
and H is the magnetic field. Figure 1(a) shows the Sm of Gd1-xYx alloys while changing external magnetic
field from 0 to 1 tesla. The peak value of Sm is somewhat enhanced around Y content x =0.02. As for the
La(Fe,Si)13-based compounds, the magnetic nature varies drastically from antiferromagnetic to ferromagnetic
according to the composition, and metamagnetic transition occurs just above TC accompanied by large
entropy change Sm in a certain composition region [6]. Curie temperature can be varied by partial
substitution for La- or Fe-sites, and also by hydrogen absorption to the interstitial site [6]. Figure 1(b) shows
the magnetic entropy changes of La(Fe,Co,Si)13 compounds.
-10 0 10 20 30 (oC) -10 0 10 20 30 (oC)

4 6
Entropy change ⊿Sm (J/kg･K)
Entropy change ⊿Sm (J/kg･K)
Gd1-xYx-alloy x =0.02 La(Fe,Co,Si)13 (b)

(a)
x =0.04 5
3
x =0.10 4
Gd
2 3 Gd
2
1
1
H =1 T
ext H =1 T
ext
0 0
260 270 280 290 300 310 260 270 280 290 300 310
Figure 1. Temperature dependences of magnetic entropy change of (a) GdY alloys, and (b) La(Fe,Co,Si)13
compounds.
2.2 Spherical Particle Preparation

One of the suitable shapes of magnetic refrigerant for AMR refrigeration is considered to be a small sphere
less than 1 mm in diameter. Thus, several Gd alloys and LaFeSi-based compounds were shaped into
spherical particles by means of the rotating electrode process (REP) shown in Figures 2 and 3. Both Gd and
LaFeSi-based particles having smooth surface with metallic sheen and diameters of 0.2 to 1.2 mm were
obtained. Regarding the LaFeSi-based compounds, phase separation and formation of the dendritic
microstructure consisted of Fe-rich and La-rich phases observed in the as-REP particles as shown in Figure
3(b) and 3(c), which are the cross-sectional backscattered electron (BSE) images in the scanning electron
microscopy. It is noted that the microscopic texture of the mother alloy affects microstructure in the REPed
particle significantly, that is, the mother alloy with coarse-grained texture leads to considerable elemental
segregation as shown in Fig. 3(b) in contrast to Fig. 3(c), which is the case of the mother alloy with fine
texture. The REPed particles were subjected to homogenizing anneal to obtain the objective La(Fe,Si)13
phase. After that, La(Fe,Si)13-based particles were coated with Cu, shown in Figure 2(c), to prevent corrosion
for the AMR-cycle experiments.
298
(a) Rotation
(a) (b)
Electrode
(c) (b) (c)
200μm
Figure 2. Photographs of spherical particles of Figure 3. (a) Rough sketch of REP, and the BSE
(a) Gd, (b) LaFeSi-based compound without images of the cross-section of REPed particles
coating and (c) La(Fe,Si)13-based compound made from LaFeSi mother alloy with (b) coarse-
coated with Cu. grained texture and (c) fine texture.
3. COOLING PROPERTIES OF AMR-CYCLE
3.1 Experimental Apparatus

To elucidate fundamental phenomena in the AMR cycle, a simple structure was employed for a primitive test
apparatus of the AMR-cycle refrigeration. Figure 4(a) and 4(b) show a photograph of the apparatus and a
sketch of the system layout. Magnetic refrigerant particles were packed into a cylindrical container (the
AMR unit), which has the capacity of 20 cm3, with a filling rate of about 63%, and the container was filled
with heat transfer fluid. The fluid was water or 20% ethylene glycol solution. The AMR unit can reciprocate
up and down by electric cylinder motors out and into the bore of a Halbach magnet. The strength of the
magnetic field at the center of the bore was approximately 1.1 T. The heat transfer fluid flowed reciprocally
through the packed magnetic refrigerant particles by moving pistons installed at the top and bottom parts of
the AMR unit. The AMR cycle was operated by the following procedure: (I) the AMR unit moves into the
bore of the magnet to apply a magnetic field to the magnetic refrigerant, (II) the heat transfer fluid flows
from the cold to the hot end through the magnetic refrigerant to transport heat to the hot side, (III) the AMR
unit moves out of the bore of the magnet to remove a magnetic field from the magnetic refrigerant, (IV) the
heat transfer fluid flows from the hot to the cold end (the opposite direction to step II) through the magnetic
refrigerant to transport cold heat to the cold side. After repeated cycles, a temperature gradient is generated
in the AMR unit. The cycle frequency and flow rate can be controlled by the speed of the cylinders and
(b)
(a)
up hot end
AMR-unit
down
Magnetic
refrigerant
particles
Halbach
Magnet
cold end
Figure 4. (a) Photograph of a primitive test apparatus of AMR cycle, and (b) sketch of the system layout.
299
pistons movement. Additionally, heat load was applied to the cold end of the AMR unit using an electrical
heater attached to the tip of the bottom piston. Thermocouples of type K were installed at 11 points in the
AMR unit, and the temperature at each point was monitored while operating the AMR cycle. The AMR unit
and a Halbach magnet were set in a temperature-controlled room to keep the ambient temperature constant
during AMR-cycle operation.
3.2 Basic Properties of AMR cycle

The temperature span Tspan defined as the temperature difference between the hot end and the cold end of
the AMR unit at a steady state in AMR-cycle operation was investigated under different conditions. Figure
5(a) shows the cycle frequency dependence of the Tspan with the use of spherical Gd particles 780 m in
diameter at ambient temperature Ti = 19°C, where the inset demonstrates the temporal changes in
temperature at both ends of the AMR unit during AMR-cycle operation. The Tspan increases approximately
linearly with frequency in the range of 0.1< f < 0.4. Figure 5(b) shows the particle size dependence of Tspan
as a parameter of piston displacement for the case of f = 0.4 Hz, Ti = 15°C and using Gd98.5Y1.5 spherical
particles. With increasing piston displacement, heat mixture between hot side and cold side becomes larger
and leads to decrease of Tspan. As the particle size becomes smaller, the Tspan becomes larger and saturates
in each displacement, because of the trade-off between an increases of heat exchange area and pressure loss.
Figure 6(a) shows the ambient temperature dependences of the Tspan for the different magnetic refrigerants
and they show broad peaks around the TC of each magnetocaloric material. It should be noted that the value
of the Tspan with the use of La(Fe,Si)13-based refrigerant was about half that in the case of using Gd-based
refrigerant. It is thought to be attributable to the heat capacity of La(Fe,Si)13-based materials being twice that
of Gd-based refrigerant. Furthermore, the AMR-cycle properties with heat load were investigated. The
decrease in Tspan with respect to heat load for Gd-alloy refrigerant was considerably larger than that for
50
Temperature span Tspan ( C)
(a)
40
(b)
40
30
35
最大熱分離
Gd+Water, 5mm, VR40000VR1000, Loop400, Delay0, Tmeas.19℃
353534
33
Hot end
32
30
( C)
31
30
20
29
28
Temperature
27
25 26
Temperature
25
Tspan
24
( C)
23
22
20 21
20
d = 5 mm
19
18
17
15 16
10
15
Cold end 30 d = 7 mm
14
13
10
12
11
10
d = 10 mm
9
8
7
0 1000 2000
0 500 1000 1500 2000
Time ( sec. )
Time (sec.)
0
0.1 0.2 0.3 0.4 0.5 400 500 600 700 800
Cycle frequency f (Hz) Particle size （μm ）
Figure 5. (a) Cycle frequency dependence of the Tspan defined as the inset shows and (b) particle size dependences
of the Tspan as a parameter by the piston displacement in the AMR-cycle operation.
40
50
Temperature span, Tspan ( C)
Temperature span, Tspan ( C)
35
(a) (b)
30 40
Gd-alloy
25
30
20
La(Fe,Co,Si)13
15 20
Gd
10 Gd0.985Yi0.015 10
La(Fe0.86Si0.14)13H1.2
5
La(Fe0.85Co0.07Si0.08)13
0 0
0 5 10 15 20 25 30 35 0 1 2 3 4 5
Ambient temperature Ti ( C) Heat load (W)
Figure 6. (a) Ambient temperature dependences and (b) heat-load dependences of the Tspan for different types of
magnetic refrigerants in the AMR-cycle operation.
300
La(Fe,Si)13-based refrigerant as shown in Figure 6(b). These results indicate that the large heat capacity of
La(Fe,Si)13-based compounds contributes to high tolerance for heat load although it is ineffective for
generating large Tspan.
3.3 Improvement of Cooling Performance for Application

In the AMR cycle, scale-up of the amount of refrigerant and higher-frequency operation are effective to
increase cooling power, and use of layered refrigerant materials having different TC would be effective to
enlarge the working temperature range. Figure 7(a) shows experimental results of the effect of layered
materials on the Tspan with respect to ambient temperature. Further, improvement of heat exchange owing to
turbulent flow by stepwise control enlarges Tspan to 46°C as shown in Figure 7(b). In terms of heat load,
model calculation was performed in a one-dimensional AMR model. Figure 7(c) shows the calculation
results of heat-load properties for the case of using 1 kg of La(Fe,Si)13-type refrigerant in view of the layered
number and cycle frequency. While the details of the calculations are described in other reports, the results
indicate that the AMR refrigerator using 1 kg of La(Fe,Si)13-based refrigerant in the several-layered structure
would be able to attain a cooling power of 100 W at Tspan ~25°C in the case of the cycle frequency of 1 Hz.
50 40 40
(a) (b) La(Fe,Si)13-type (c)
Temperature (oC )
45 30 Hot end
Tspan ( C )
30 9-layered
40 20
Tspan 3-layered
10 20
35
Temperature
f =0.4 Hz
0
30 GdY1.5 10
Cold end single
GdY1.5/GdY5/GdHo10 -10 f = 0.4 Hz
Time
25
0 5 10 15 20 25 0 1000 2000 00 20 40 60 80 100
Ambient temperature (oC) Time ( sec. ) Heat load (W)
Figure 7. (a) Experimental results of the Tspan with respect to ambient temperature for the case of using a single Gd
alloy or three kinds of Gd-alloy refrigerants. (b) Temporal changes in temperature at both ends of the AMR unit
during AMR-cycle operation using a GdY1.5-alloy refrigerant. (c) Heat-load dependences of the Tspan calculated by
the one-dimensional AMR model with La(Fe,Si)13-type refrigerants where the open triangle represents the case of a
single refrigerant with f = 0.4 Hz, the open square represents the case of three kinds of refrigerants with f = 1 Hz,
and the open circle represents the case of nine kinds of refrigerants with f = 1 Hz.
3.4 Design for Energy-saving System

The AMR cycle requires operating mechanisms of both magnetic field and fluid. With a view to realizing
highly efficient refrigerators, magnetic circuit design and a pump-less fluid flow mechanism were developed
to reduce power consumption for driving the AMR cycle. Figure 8(a) shows a schematic view of a magnetic
circuit designed for simultaneous low-torque operation and strong-magnetic-field generation. Strong
magnetic field of more than 1 T was achieved employing a Halbach structure of NdFeB magnets, and the
magnetic components and magnetic refrigerant units were aligned at the torque-canceled geometry during
relative rotation. Consequently, on/off cycles of the magnetic field of 1.39 T applied to/removed from 1 kg
of Gd refrigerant with frequency of 1 Hz by rotating operation of the magnetic circuit could be accomplished
(b)
rotation
Spring
Magnet
Magnetic
refrigerant
particles
Figure 8. Concept for energy-saving system of AMR refrigerator; (a) schematic view of a torque-canceled magnetic
circuit design and (b) photograph of a pump-less fluid flow mechanism.
301
with the input power of only 12 W. Figure 8(b) shows a photograph of a pump-less fluid flow mechanism
where the AMR unit moves automatically in the fluid-filled space by a balance between magnetic force and
elastic force during magnetic circuit rotation. Considering these matters concerning power consumption
together with the cooling capacity described above, the potential of the coefficient of performance (COP) of
the AMR cycle is expected to be larger than 5 in the case of a cooling power of tens of watts at Tspan ~25°C,
f ~1 Hz with 1 kg of magnetic refrigerant.
4. CONCLUSIONS
Basic research on the cooling properties of the AMR cycle using spherical magnetic refrigerant particles of
different types of magnetocaloric materials was performed. The temperature span Tspan depends strongly on
not only the operating conditions, such as the cycle frequency, fluid flow rate, and particle size, but also on
the physical properties of magnetic refrigerant materials such as magnetocaloric effect and heat capacity.
More than 40°C of Tspan was obtained by using a single refrigerant material of Gd alloy that is suitable for
generating large Tspan. On the other hand, La(Fe,Si)13-based compounds showed a great advantage over Gd-
alloys in terms of heat-load tolerance but were unsuitable for large Tspan generation. Cooling properties of
the AMR cycle were improved by the use of layered refrigerant materials having different TC, in addition to
scale-up and higher cycle frequency. Moreover, feasible techniques to reduce power consumption for the
AMR cycle were developed. In conclusion, the AMR cycle has a potential COP double that of current
household refrigerators. Further breakthroughs are required, such as efficient heat exchange and decrease of
the amount of magneto in the AMR refrigerator, but these will be achieved and that practical application of
room-temperature magnetic refrigeration will be realized in the near future.
ACKNOWLEDGMENTS
This work was partially supported financially by the Japan Science Technology (JST) and the New Energy
and Industrial Technology Development Organization (NEDO). We would like to express our gratitude to
Professor Karl A. Gschneidner, who has long been a source of encouragement and inspiration to researchers
in this field, and our a sorrow at his passing.
REFERENCES
[1] K. A. Gschneidner, Jr. et al., “Magnetic cooling for appliances”, Proceedings of the 50th Annual
International Appliance Technical Conference, (International Appliance Technical Conference, Inc., Largo
Florida), 144 (1999).
[2] J. A. Barclay, W. A. Steyert, “Active Magnetic Regenerator”, U.S. Patent 4,332,135 (1982).
[3] V. K. Pecharsky, K. A. Gschneidner, Jr., “Giant magnetocaloric effect in Gd5(Si2Ge2)”, Phys. Rev. Lett.
78, 4494 (1997).
[4] H. Wada, Y. Tanabe, “Giant magnetocaloric effect of MnAs1-xSbx”, Appl. Phys. Lett. 79, 3302 (2001).
[5] O. Tegus et al., “Transition-metal-based magnetic refrigerants for room-temperature applications” Nature
415, 150 (2002).
[6] A. Fujita et al., “Itinerant-electron metamagnetic transition and large magnetocaloric effects in La(FexSi1-
x)13 compounds and their hydrides”, Phys. Rev. B 67, 104416 (2003).
[7] C. Zimm et al., “Design and performance of a permanent magnet rotary refrigerator”, Proceedings of the
1st IIF-IIR International Conference on Magnetic Refrigeration at Room Temperature, International
Institution of Refrigeration, 367 (2005).
[8] A. Tura et al., “Generative Magnetocaloric systems: An overview of research and development activities
at the university of Victoria” Proceedings of the 5th IIF-IIR International Conference on Magnetic
Refrigeration at Room Temperature, International Institution of Refrigeration, 73 (2012).
[9] C. R. H. Bahl et al., “Development and experimental results from a 1 kW prototype AMR”, Proceedings
of the 5th IIF-IIR International Conference on Magnetic Refrigeration at Room Temperature, International
Institution of Refrigeration, 53 (2012).
[10] J. M. Gatti et al., “Magnetic heat pumps - configurable hydraulic distribution for a magnetic cooling
systems”, Proceedings of the 5th IIF-IIR Int. Conference on Magnetic Refrigeration at Room Temperature,
International Institution of Refrigeration, 341 (2012).
302
ROTARY MAGNETIC REGENERATOR DESIGN AND ASSEMBLY
F. Scarpa(a)*, L.A. Tagliafico(a), M. Gigante(a)

(a)
DIME/TEC, University of Genoa (Via all’Opera Pia 15a, 16145, Genoa, Italy)
*
Corresponding author. E-mail: fscarpa@ditec.unige.it
ABSTRACT
Magnetic refrigeration is an emerging technology that exploits a particular feature of some materials, called
“magneto caloric effect” to obtain a cooling effect. In recent years, several research groups have deepened
studies on this innovative cooling method, showing its potential mainly through the development of various
prototypes. The Augere Research Team (ART) at University of Genoa started the design that will lead to the
realization of a magnetic refrigeration machine, based on the AMR cycle, with fixed magnetic structure and
continuous rotating motion of the magnetocaloric material. The salient feature of this new refrigerator is the
total absence of valves and dedicated ducts; it is the structure itself, extremely compact, of rotor and stator to
provide a flow-guide to the heat transfer fluid. In the present work the functional plan for the construction
and assembly of this promising configuration is presented, along with the simulation tests performed during
the design stage.
Keywords: valve-less magnetic refrigerator, room temperature, rotating AMR, functional design.
DOI: 10.18462/iir.thermag.2016.0258
1. INTRODUCTION
Magnetic refrigeration is an innovative and alternative way of cooling at room temperature, that promises to
be competitive with actual refrigeration techniques by yielding a noticeable primary energy saving.
Currently, part of the residential sector in the European Community is one of those with the highest energy
consumption. Therefore the energy consumption decrease in this sector is a primary objective. Despite the
interest of the scientific community for magnetic refrigeration at room temperature has grown exponentially
in recent years, an experimental magnetic refrigerator with market ready characteristics does not yet exist. In
recent years, several research groups have deepened studies on this innovative cooling method, showing its
potential near room temperature mainly through the development of various prototypes. Exhaustive reviews
on this topic can be found in Kitanovski et al. [1].
In summary, a fully functional and marketable magnetic cooler has not yet been achieved. Therefore, it is
necessary to develop new experimental prototypes that guarantee a suitable cooling power providing at the
same time values of COP able to compete with the vapor compression systems. The presented design is
aimed to fulfill this gap. The AugERE Research Team (AUGmentation techniques for Energy, Refrigeration
and Environment) at University of Genoa started the design of a new magnetic refrigeration machine, based
on the well-known AMR (Active Magnetic Regenerative) cycle of the MCM (Magneto Caloric Material).
Starting from the early work of Egolf and coworkers [2], the concept is based on a fixed magnetic structure
and a rotating internal cylinder supporting a series of MCM sheets. The main feature of this new refrigerator
is the total absence of valves and dedicated ducts; it is the structure itself, extremely compact, of rotor and
stator, to provide a guide to the heat transfer fluid. This solution has been studied in such a way to have a
light rotating element (to reduce dynamic inertial losses) and to use materials able to avoid eddy currents in
the external iron structure.
The present work offers a detailed plan for the construction and assembly of this promising configuration, in
order to widespread the technology among research and SME communities. The ultimate goal of the study is
to make the magnetic refrigeration technology a competitive one in the refrigeration market, from the point
of view of the energy performance and of the installation costs.
2. DESIGN CONCEPT AND IMPLEMENTATION
During the design phase of a magnetic refrigerator, one of the first choices to be made concerns “how” the
active magnetic material enters and leaves the area affected by the magnetic field.
303
The decision concerns the type of motion, and there are basically two options: linear reciprocating motion
and rotary motion. In each of the two cases, the variation of the magnetic field sensed by the magneto-caloric
solid is however due to the relative motion between the latter and the permanent magnet.
There are other alternatives, for instance the use of a programmed electromagnet (replacing the permanent
magnet) can provide variations of the magnetic field, maintaining at rest all system components, including
the magnet. However this solution is not adopted because of the considerable expenditure of energy needed
to power the electromagnet and the resulting loss of efficiency which would arise. There is also a
configuration which combines the motion of two magnets so that the magnetic field is added to or subtract,
for example by using two coaxial cylindrical Halbach.
After a deep literature review and based on experiences with the construction of two linear reciprocating
prototypes [3], a rotating configuration was chosen since it is inherently simpler. In fact, the linear motion
configuration presents balancing issues originating from the force between the magnet and gadolinium in
order to move the material in and out of the magnetic field. Furthermore, it is difficult to manage the forces
due to motion inversion and often a portion of energy is wasted.
The continuous rotating concept overcomes both these drawbacks so we deeply analyzed the four rotary
configurations schematically reported in Fig. 1 which refers to topology analysis made in [4].
Figure 1. Various rotary configuration depending on the particular axes and fluid flow directions. The fourth
configuration (d) has been selected for the project.
In particular, the last configuration has been implemented in US patent US7481064 B2 by Kitanovski et al.
[2] and it presents some advantages in respect to the other embodiments. It can work both in continuous or
discontinuous, also reciprocating, motion. It does not need a bidirectional pump, usually implemented by
using three-way valves, and it is less prone to fluid leakage than the Steyert (c) solution [5].
2.1 The magnetic structure

Regarding the magnetic structure, a modified version of the Okamura [6] assembly was selected, but the
magnet set was kept at rest since, otherwise, eddy currents are likely to arise. As a consequence, it is the
magnetocaloric material (MCM) that rotates in the air gap formed between the fixed internal magnets (a
different configuration in respect to that used in the above patent, where the magnetic polar expansion are
conceived externally to the rotating MCM structure) and the static external iron bushing which closes the
magnetic circuit.
Figure 2. Selected magnetic configuration and cinematics. The magnetic induction is concentrated by proper iron
expansions (in green). The rotating sup the rotating support of the gadolinium is set in motion by four plastic spokes
304
Figure 3. Magnetic performance. Magnetic induction (magnetic flux density) reaches 0.9 T in air alone. In presence of
MCM sheets, this value increments to around 1.1 T in the sheets and reduces to 0.55 T in the air gaps.
Figs. 3 show the resulting magnetic induction. In fig 3a the strength of the induction increase up to 1.6 T in
the iron joke but a value of 1.1 T is granted in the MCM material as can be viewed in fig 3b and c. The iron
has been lead deliberately to saturation. In this way the reluctance of the circuit slightly increases but the
value of the flux density in the gap region results very stable and independent on small change in the
geometry, changes which often occur during the design phase.
In the gap four main sectors can be recognized, two with high and two with low inductance values. Figure
3(c) depicts the magnitude of the flux density norm in the MCM sheets and in the air gaps.
2.2 Thermodynamic analysis

The device size is undersized; an external radius of about 7 cm with a magnetic core with a 4 cm radius.
Obviously a small size negatively affects the performance. On the other hand the device has been conceived
as a direct substitute of a typical household refrigerator compressor and the low, but sufficient power (about
30 W), is compensated by the reduced weight and dimensions.
The refrigeration performance of the magnetic device is visible in fig. 4. The optimal rotation frequency is
around 0.25 Hz, with a utilization factor U=0.9. In these conditions the refrigeration capacity is about 30 W
with an internal COP of 6. These values have been obtained by adapting the procedure and the numerical
method detailed in [7] since the present rotary machine can be view as an ensemble of reciprocating devices.
Figure 4. Thermodynamic analysis. Refrigeration power Qref and performance COP as a function of the rotation
frequency, F. Utilization factor U varied. The considered MCM is gadolinium.
2.3 Stress analysis

In the high field zones, the magnetic flux density outward decreases so that the MCM sheets are attracted to
the magnets with a consequent stress to the supporting ring. This produces two important effects; first of all,
a possible creep of the ring with associated block of the device. Secondarily, due to the rotation of the ring,
the onset of a fatigue stress which would lead to the fracture of the material. So the need of an accurate
choice for the material of the ring. Fig.5 shows a detail of the rotating rim and results from a FEM analysis.
305
Figure 5. Von Mises stress and deformation analysis.
Once the magnetic and rotary configurations was selected, it was necessary to identify the right mechanical
embodiment that would allow an easy construction of the prototype. After the evaluation of various ways to
put into motion the device, the choice fell on an axial shaft handler as depicted in Fig.6. The same figure
shows an exploded view of the refrigerator core that feature a total weight of 15kg (1.5 kg of Gd).
Figure 6. Two different CAD views of the conceived device. The total weight is about kg 15.
3. CONCLUSIONS
The present descriptive work, which will be soon deepened in a proper publishing venue, proposes an
interesting magnetic refrigerator device. It is characterized by very small size and power but it fits the current
needs of household refrigeration. In particular, the moving parts are very simple, there are no valves or
dedicated ducts, and the structure is extremely compact. All these features translate in very low expected
production costs. While waiting for a fully mature perovskite-type oxides technology, we hope the proposed
assembly will stimulate an easy route to market for “room temperature” magnetic refrigeration technology.
REFERENCES
[1] A. Kitanovski, J. Tušek, U. Tomc, U., U. Plaznik, M. Ozbolt, M., A. Poredoš, Magnetocaloric Energy
Conversion – From Theory to Applications , Springer (2015).
[2] A. Kitanovski, P.W. Egolf, O. Sari, “Method and device for continuous generation of cold and heat by
means of the magneto-calorific effect”, US Patent No. US7481064 B2 (2009).
[3] L.A. Tagliafico, F. Scarpa, F, Valsuani, G. Tagliafico, “Preliminary experimental results from a linear
reciprocating magnetic refrigerator prototype”, Applied Thermal Engineering 52 (2), 492 (2013).
[4] F. Scarpa, G. Tagliafico, L.A. Tagliafico, “Classification proposal for room temperature magnetic
refrigerators”, International Journal of Refrigeration 35 (2), 453 (2012).
[5] W.A. Steyert, “Stirling-cycle rotating magnetic refrigerators and heat engines for use near room
temperature”, Journal of Applied Physics, 49, 1216 (1978).
[6] T. Okamura al., “Improvement of 100 W class room temperature magnetic refrigerator”, Proceedings of
the Second international conference on magnetic refrigeration at room temperature, IIF-IIR, 377 (2007).
[7] G. Tagliafico, F. Scarpa, F. Canepa, “A dynamic 1-D model for a reciprocating active magnetic
regenerator; influence of the main working parameters”, International Journal of Refrigeration 33 (2), 286
(2010).
306
AUTHOR INDEX
Abdelaziz, O. 59 Bjørk, R. 103
Almanza, M. 186, 190 Bouchard, J. 79
Almanza, M. 198 Buchelnikov, V. 202, 228
Alvares, M. 182 Butvina, L. 263
Amaral, J. 154, 252 Campbell, O. 35, 39, 67
Amaral, V. 154, 252 Capovilla, M. S. 83
Amorim, C. 252 Cararo, J. 115
Anikin, M. 236 Cardoso, L. 267
Aprea, C. 131 Cardoso, P. 182, 287
Bahl, C. 103, 158, 248 Carroll, C. 28
Bailly, Y. 55, 271 Carvalho, A. 87
Balli, M. 217 Chaudron, J. 172
83, 111, 115, 123, 182, Chen, Y. 138

Barbosa, J.
287
Cheng, J. 24, 63
Barcza, A. 158
Cherief, W. 190
Barrera-Medrano, D. 28
Chrigui, M. 51
Barriere, T. 279
35, 39, 67, 71, 75, 91,
Christiaanse, T.
Bartholomé, K. 47 95, 222, 240
Bartok, A. 119, 186, 275 Clareth Colman, F. 87
Basso, V. 119, 178, 198, 275 Cocci de Souza, A. 87
Bchiri, D. 51 Coelho, A. 267
Ben Sidhom, R. 51 Corberán, J. M. 244
Bennati, C. 178 Costa, V. 154
Bessa, C. 11, 154 Curcio, C. 119, 178, 275
Bessais, L. 31 Cwik, J. 206
Bessais, L. 119 Dan, N. H. 263
Bez, H. 158 De Negri, V. 182
307
Destro, M. 182 Gonçalves, J. 252
Dilmieva, E. 206 35, 39, 67, 71, 75, 91,

Govindappa, P.
95, 222
Dimitrov, D. 217
Greco, A. 131
Dos Santos Conceição, W. A. 87
Gutfleisch, O. 99
Dubrez, A. 279
Gérard, J. 198
Dötz, F. 28
Haegel, P. 172
Ebel, T. 287
Hai, X. 1
Elouad, L. 142, 172
Hamdani, K. 283
Engelbrecht, K. 158, 248
Hardy, V. 127
Entel, P. 228
Hess, T. 47
Espanet, C. 20
Hirano, N. 232
Evaristo, E. 87
Hittinger, M. 172
Ferreira, L. 11, 154
Hongwei, Y. 63
Filonenko, K. 248
Horikawa, O. 11, 154
Foleiss, G. L. 87
Huang, J. 164
Fortkamp, F. 111
Huang, J. 24, 63
Fortunato, N. 252
Ikeda, K. 107
Fournier, P. 217
Inishev, A. 236
Gama, S. 11, 154
Insinga, A. R. 103
Garbuio, L. 198
Irurzun, I. 258
García Reparaz, D. 258
Jandl, S. 217
Gelin, J. 279
Jeday, A. 51
Gigante, M. 303
Jiang, Q. 164, 168
Gilewski, A. 263
Jin, P. 24, 63
Gimaev, R. 194
Kalck, U. 28
Giurgea, S. 20
Kamantsev, A. 206, 263
Glises, R. 55, 271
Karpenkov, D. 99
Goijman, D. 258
Kawanami, T. 107
Gomes da Silva, M. 267
308
Kedous-Lebouc, A. 190, 198 Mansouri, S. 217
Khovaylo, V. 99, 206 Martino, L. 275
Kitanovski, A. 291 Mashirov, A. 206, 263
Kizaki, T. 43, 214 Masselli, C. 131
Koledov, V. 206, 263 Mayer, C. 1
Kondo, M. 43, 214 Mazaleyrat, F. 119, 186
Koshkidko, Y. 206 Meddeb, Z. 51
Kraposhin, V. 206 Mehdizadeh Momen, A. 59
Kudrevatykh, N. 236 Meunier, A. 55, 271
Kuepferling, M. 119, 178, 275 Michalski, R. 15
König, J. 47 Miraglia, S. 1
Lang, R. 267 Misra, S. 28
Lanzarini, J. 279 Miyanaga, T. 194
Lei, T. 158, 248 Miyazaki, Y. 107
Leonowicz, M. 210 Monfared, B. 146, 150
Leyva, A. G. 258 Moreira da Silva, L. 267
Li, C. 24 Mugica, I. 79
Li, H. 164, 168 Muller, C. 127, 142, 172
Li, Z. 24, 63 Nagamine, R. 232
Lionte, S. 127, 142 Nakagome, H. 297
Liu, C. 24, 63 Nakashima, A. 83
Lo Bue, M. 119, 275 Nakouri, K. 31
Lo Bue, M. 186 Nassif, V. 1
Los, A. S. 263 Navickaite, K. 158
83, 111, 115, 123, 182, Nielsen, K. 240

Lozano, J.
287
Nika, P. 20, 55, 271
Lu, D. 164, 168
Niknia, I. 35, 39, 67, 91, 95, 222
Mahlke, A. 47
Niknia, I. 71, 75
Maiorino, A. 131
Nomura, R. 43, 214
309
Okamura, T. 232 Sacanell, J. 258
Oliveira dos Santos, A. 267 Sahli, M. 279
Olivetti, E. 275 Saito, A. T. 297
Otubo, L. 267 Saletsky, A. 194
Palm, B. 146 Santiago Alves, C. 87
Pasko, A. 119, 186 Santos, T. 154
Passanante, S. 258 Scarpa, F. 303
Patissier, A. 31 Scharf, F. 28
Paul-Boncour, V. 31 Schneider Calomeno, R. 123
Pavlukhina, O. 202, 228 Schwind, M. 28
Payá, J. 244 Seeler, F. 28
Peixer, G. 83 Semkin, M. 236
Pellenen, A. 99 Shavrov, V. 206, 263
Piazzi, M. 178 Sielicki, K. 210
Plait, A. 20 Skokov, K. 99
Poncet, S. 7, 79 Smaili, A. 283
Porcher, F. 1 Smith, A. 103
Poredoš, A. 291 Sokolovskiy, V. 202, 228
Pruvost, S. 198 Sorin, M. 7
Pryds, N. 240 Staunton, J. 194
Quintero, M. 258 Szpak, W. 87
Reesink, B. 28 Tagliafico, L. A. 303
Risser, M. 127, 142 Tahavori, M. 248
Roberge, B. 217 Takahashi, M. 232
Rogge, O. 28 Takeuchi, K. 43, 214
35, 39, 67, 71, 75, 91, Tanaka, K. 232

Rowe, A.
95, 115, 222, 240
Tang, Y. 138
Roy, J. 55, 271
Tarasov, E. 236
Roy, S. 7
Taskaev, S. 99
310
Tereshina, I. S. 263 Zheng, H. 138
Zinin, A. 236
35, 39, 67, 71, 75, 91,
Teyber, R. Zverev, V. 194
95, 115, 222, 240
Zygadlo, J. 15
Thanh, P. T. 263
de Larochelambert, T. 20, 55, 271
Tishin, A. 194
de Paula, V. G. 267
Tomc, U. 291
van Asten, D. 28
Torregrosa-Jaime, B. 244
35, 39, 67, 71, 75, 83,

Trevizoli, P. 91, 95, 115, 123, 222,
240
Tura, A. 240
Ueno, K. 43, 214
Ulyanov, M. 99
Van Nong, N. 240
Vasile, C. 142, 172
Veje, C. T. 248
Volegov, A. 236
Waki, K. 107
Wang, H. 138
Wroblewski, R. 210
Wu, J. 138
Xu, X. 164, 168
Yen, N. H. 263
Zehani, K. 119
Zeilinger, M. 28
Zhang, C. 24, 63
Zhang, L. 28
Zhang, M. 59
Zhang, Y. 24, 63
311

ThermagVII PDF

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

ThermagVII PDF

Transféré par

Droits d'auteur :

Formats disponibles

REFRIGERATION SCIENCE AND TECHNOLOGY

7th International Conference on

11-14 September, 2016, Torino, Italy

INTERNATIONAL INSTITUTE OF REFRIGERATION

7th International Conference on

11-14 September, 2016

Under the auspices of Sponsor

INTERNATIONAL INSTITUTE OF REFRIGERATION

Droit de reproduction et avertissement

INTERNATIONAL INSTITUTE OF REFRIGERATION

LOCAL ORGANIZING COMMITTEE (INRIM)

LOCAL ORGANISING SECRETARIAT (INRIM)

PUBLICITY AND WEBSITE (IIF/IIR)

Karl Gschneidner Jr. (November 16, 1930 - April 27, 2016)

Karl Albert Gschneidner, Jr. was born in Detroit on

057 :: Multiobjective optimization of a reciprocating magnetic refrigerator using a genetic algorithm

059 :: A test stand to measure the magnetization work of magnetocaloric materials

070 :: Magnetocaloric bench: analytical model for gadolinium characteristics

072 :: Performance comparison of La0.8Ce0.2Fe11.47Mn0.23Si1.3H1.8 magnetic refrigerant prepared by two

077 :: A material screening technique for optimum performance of an AMR

078 :: Effects of magnetocaloric wire on increase in magnetic refrigeration cycle

086 :: Study of heat transfer in magnetocaloric regenerator with oscillating flow

096 :: Experimental evaluation of a single layer MnFeP1-xSix AMR

098 :: Transient simulation of the Active Magnetic Regenerator process

104 :: Modeling and simulation of a thermomagnetic reciprocating engine by gravity

106 :: Experimental evaluation of two-material active magnetic regenerators

110 :: Generating the optimal magnetic field for magnetic refrigeration

127 :: Influence of particle size on the magnetocaloric properties of Mn1.30Fe0.65P0.5Si0.5 powders

134 :: High-temperature hydrogenation behaviors and magnetocaloric properties of LaFe11.6Si1.4 compound

145 :: Optimization of heat exchange in magnetic refrigerators and thermomagnetic motors

147 :: Epoxy-bonded La(Fe,Mn,Si)13Hz as a multi layered active magnetic regenerator

149 :: Design and fabricating of a two-in-one compact magnet array

157 :: Theoretical description of the magnetocaloric effect in Mn-Fe-P-Si alloys

158 :: Modeling of a novel digital hydraulic system for a magnetic refrigerator

160 :: Thermal energy harvesting: thermomagnetic versus thermoelectric generator

161 :: Dynamic temperature measurement for magnetocaloric materials

174 :: Theoretical modeling of heat transfer processes in Ni–Co–Mn–In magnetic wires

176 :: Magnetocaloric properties of the stacked Ni50Mn18.75-xCu6.25+xGa25 (x = 0÷1) polycrystalline alloys

177 :: Properties of Gadolinium wire for magnetic refrigeration

178 :: On the rotating magnetocaloric effect in multiferroic RMn2O5 compounds

179 :: Active Caloric Regenerator Cycles: an analytic element model

184 :: Experimental performance evaluation of sintered Gd spheres packed beds

191 :: Integral optimisation of a magnetocaloric air-conditioning system for an electric vehicle

194 :: A cascading model of an active magnetic regenerator system

200 :: Giant magnetocaloric effect of compressible Ising and Heisenberg lattices

208 :: Magnetocaloric effect supression induced by mechanical milling on intermetallic DyCo2

222 :: Ti substitution in MnBi rare-earth free magnetocaloric compound

227 :: Study on composite of thermoplastic loaded by magnetocaloric powders

247 :: Simulation of a hydraulic circuit for a magnetic refrigerator

250 :: Future developments in magnetocaloric refrigeration and heat pumping

258 :: Rotary magnetic regenerator design and assembly

AUTHOR INDEX 307

X. Hai(a)(b)*, C. Mayer(b), V. Nassif(a), F. Porcher(c), S. Miraglia(a)

Magnetocaloric materials (La,Ce)(Fe,Mn,Si)13Hx have great potential as commercializable candidate for

Keywords: magnetocaloric materials, hydrogen sorption kinetics, neutron diffraction

2.1 Experimental Conditions

0,2 Absorption 0,2

2.3 Effect of C Interstitial Insertion

11,54 LaFe11.44Si1.56 11,54

(a) 2,80 La0.7Ce0.3Fe11.44Si1.56 (b) 2,48 La0.7Ce0.3Fe11.44Si1.56

[1] F. Wang et al., J. Phys. D.: Appl. Phys., 36, 1 (2003).

S. Roy*, S. Poncet, M. Sorin

Keywords: Multiobjective Optimization, Exergy, Efficiency, Genetic Algorithm.