Vous êtes sur la page 1sur 14

Materials Science & Engineering A 560 (2013) 653–666

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A


journal homepage: www.elsevier.com/locate/msea

The effect of training on two-way shape memory effect of binary NiTi


and NiTi based ternary high temperature shape memory alloys
K.C. Atli a, I. Karaman a,b,n, R.D. Noebe c, D. Gaydosh c
a
Department of Mechanical Engineering, Texas A&M University, College Station, TX 77843, USA
b
Materials Science and Engineering Program, Texas A&M University, College Station, TX 77843, USA
c
NASA Glenn Research Center, Structures & Materials Division, Cleveland, OH 44135, USA

a r t i c l e i n f o abstract

Article history: The propensity for various high-temperature shape memory alloys (HTSMA), i.e., Ni28.5Ti50.5Pt21,
Received 3 August 2012 Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5, to exhibit two-way shape memory effect (TWSME) was
Received in revised form compared to that of a conventional binary Ni49.9Ti50.1 shape memory alloy (SMA). Thermomechanical
2 October 2012
training in the form of thermal cycling under constant stress levels was employed to induce two-way
Accepted 4 October 2012
shape memory behavior in the various materials. The resulting TWSME was characterized for its
Available online 12 October 2012
magnitude and stability under stress-free conditions, while parameters such as training stress and
Keywords: upper cycle temperature during training were investigated for their influence on this phenomenon. For
Shape memory alloys Ni49.9Ti50.1, a negative correlation was found between an increasing training stress, from 80 MPa to
Martensitic transformation
200 MPa, and the magnitude of the resulting TWSM strain, while a positive correlation was observed
Thermomechanical training
for Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5. The stability of the TWSME for the Ni49.9Ti50.1, measured by
Two-way shape memory effect
Thermal stability the strain evolution of the cold (martensitic) and hot (austenitic) shapes of the samples upon stress-free
thermal cycling, was found to depend on the stress and temperature interval during training.
Conversely, the stability of the NiTiPd based HTSMAs was much greater and less sensitive to these
parameters over the stress and temperature intervals investigated. No TWSME was seen in
Ni28.5Ti50.5Pt21 due to the higher upper cycle temperatures required during thermal cycling, which
resulted in the recovery of any favorable dislocation structures generated during training.
& 2012 Elsevier B.V. All rights reserved.

1. Introduction actuator to remember both its low-temperature and high-


temperature shapes without the need for a re-biasing force.
NiTi based shape memory alloys (SMAs) have been exploited Unlike OWSME, TWSME is not an inherent characteristic of SMAs,
in a wide range of product forms such as eyeglass frames, but rather it is obtained after thermomechanical treatments
orthodontic wires, valves, pipe couplings, and switches due to (training), such as stress or temperature cycling.
their unique functional properties [1]. In addition, the fact that Recently, a need for compact, high-temperature actuation
the shape memory effect can be used to do work against a load sources has emerged in several aerospace, oil exploration, and
has led to the development of SMAs as compact, solid-state automotive applications. One drawback of NiTi based binary SMAs
actuators. Compared to D.C. motors or their pneumatic counter- is the relatively low transformation temperatures (o100 1C), limit-
parts, these actuators have several advantages such as light ing their use when high temperature actuation is required.
weight, reduction in total part count, ease of inspection and To overcome this restriction, NiTi can be alloyed with Pd, Pt, Au,
higher energy densities [2]. SMA actuators mostly operate based Hf and Zr, to raise transformation temperatures [3]. Among these
on the one-way shape memory effect (OWSME) combined with a ternary NiTi based high-temperature SMAs (HTSMAs), the NiTiPd
biasing force to reset the SMA after each actuation cycle. How- system has recently attracted considerable attention due to its
ever, it would be advantageous in many designs to eliminate the adequate workability, low thermal hysteresis, and good strain
need to mechanically reset the actuator. In this respect, the two- recovery under constrained and stress-free conditions [4–11].
way shape memory effect (TWSME) renders it possible for an Early studies on the TWSME mostly concentrated on conventional
Cu-based [12–20] and binary NiTi SMAs [21–38]. These studies
focused on the explanation of the TWSME mechanisms, generation
n
Corresponding author at: Department of Mechanical Engineering, Texas A&M
of TWSME using different training procedures, and the effects of
University, College Station, TX 77843, USA. Fax: þ1 979 862 2418. training parameters on the magnitude and stability of the TWSME.
E-mail address: ikaraman@tamu.edu (I. Karaman). Similar studies were also conducted on ternary NiTi based SMAs, such

0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.10.009
654 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

as NiTiCu [38–41], NiTiFe [35], and NiTiNb [42,43]. However, there is been shown that excessive residual strain in the form of generalized
very limited data on the TWSME characterization of HTSMAs. The plastic deformation and the subsequent formation of retained
only studies in this area for NiTi based HTSMAs were performed on martensite have undesirable effects on the TWSME [14]. Thus, the
NiTiHf alloys [44,45]. In the first study [44], Ni49Ti36Hf15 (at%) plates approach for obtaining maximum TWSM strain is to induce just
were trained via bending in martensite followed by unconstrained enough localized plasticity that will yield the highest magnitude of
recovery for up to 30 cycles at different temperatures. The highest stable oriented stress fields for nucleation of single variant marten-
two-way shape memory (TWSM) strain of only 0.88% was achieved site [28].
after a bending strain of 7.1% at room temperature. But even at this
limited strain level, the stability of the TWSME was poor. This was 1.2. Thermomechanical training procedures to obtain TWSME
attributed to the low strength of the martensite, which permitted the
introduction of dislocations during the TWSME cycles, relaxing the Different training techniques have successfully been imple-
oriented stress fields. Furthermore, the already limited TWSM strain mented to obtain TWSME in various SMA systems. The most
decreased by 50% in just 10 stress-free cycles. In an effort to improve common training techniques are: deformation in martensite
the stability of the TWSME, the same training procedure was applied followed by constrained or unconstrained recovery (OWSME
to a precipitation-hardened Ni50.6Ti29.4Hf20 (at%) HTSMA [45]. While cycling) [24,26,37,40,42,44]; stress cycling above the Af tempera-
the stability of the TWSME improved, the alloy still exhibited a 20% ture (superelastic cycling) [12]; temperature cycling through the
decrease in the TWSM strain after 30 stress-free cycles and the martensitic transformation under a constant stress level (isobaric
magnitude of the initial TWSM strain was only slightly increased thermal cycling) [13,14,20,23–25,27,28,32,36] or a combination
compared to the solutionized material. of the latter two methods [29]. Common to all these techniques is
To the authors’ knowledge, there has not been a systematic the repeated growth and shrinkage of particular martensitic
study on the TWSME of promising HTSMA systems such as NiTiPd variants, which is responsible for the generation of stable dis-
and NiTiPt. In this study, a series of NiTi based HTSMAs are location arrays [47–49]. Another technique, which is different in
compared to a conventional binary Ni49.9Ti50.1 SMA in terms of principle than the aforementioned techniques, is aging under
the magnitude and stability of the TWSME as a result of a 100- constraint. This method has been used to obtain TWSME in
cycle thermomechanical training procedure. The effects of Sc Ni-rich NiTi SMAs. With this method, coherent precipitates are
microalloying to NiTiPd are also investigated in an effort to formed in preferred orientations under applied stress, resulting in
enhance the TWSME. The effects of training parameters on the oriented internal stress fields that bias the formation of single-
resulting TWSME, such as training stress and the upper cycle variant martensite [30].
temperature (UCT) are also considered. Since most of the emer- Among the training procedures mentioned above, thermal
ging actuator applications require the SMA to do work against a cycling through the martensitic transformation under a constant
load, it is also useful to characterize the work output capability of stress level has been shown to yield satisfactory results in terms
the TWSME and its stability during actuation, which initially has of the magnitude and stability of the TWSME generated. This
been addressed for a Ni24.5Ti50.5Pd25 HTSMA [46]. However, this method has been proven to be an efficient procedure, and
broader study is focused on optimization of the TWSME and thus depending on how it is implemented, can involve relatively low
characterization of this phenomenon under stress-free conditions stresses, which can result in minimal plastic deformation com-
is sufficient and will provide direction for the selection, proces- pared to other training procedures [28]. Consequently, this
sing, and design of HTSMA actuators exploiting the TWSME. approach was adopted in the current study.

1.1. Origin of TWSME


2. Experimental procedures
Two mechanisms for the TWSME have been proposed in the
literature. The first mechanism attributes the TWSME to the oriented 2.1. Materials
residual stress fields of the dislocation arrays generated during
thermomechanical training [12]. These residual stress fields are able Four different material compositions were chosen for this study:
to induce the same variants of martensite in the absence of an conventional binary Ni49.9Ti50.1, Ni24.5Ti50.5Pd25, Ni24.5Ti50Pd25Sc0.5
external stress as the ones that are generated by the external training and Ni28.5Ti50.5Pt21 (all in at%). It should be noted that all composi-
stress, thus resulting in the TWSME [20]. The magnitude and stability tions fall on the Ti-rich side of stoichiometry so that precipitation is
of the TWSME depends to a great extent on the magnitude of these not a complicating factor in any of the materials in this study. Binary
stress fields and how they can be maintained through repeated Ni49.9Ti50.1 (at%) was acquired from Special Metals, New Hartford, NY.
thermal cycling. The same mechanism has also been explained from a in the form of 10 mm diameter rods in the hot-rolled and hot-
thermodynamics point of view [15]. According to this view, the straightened condition. This lot of material has been characterized in
dislocation arrays generated during thermomechanical training create some detail and properties as diverse as thermal expansion [50],
low energy configurations in the repeatedly induced martensite elastic moduli [51], basic shape memory behavior [52,53] and
variants, while a relatively higher energy configuration is induced in textural evolution during thermomechanical testing are available in
the less frequently induced variants. As a result, the growth of the literature [52,54].
martensite variants with the low energy configuration is favored Ingots of the high-temperature ternary and quaternary alloys
even when the external stress is removed, resulting in TWSME. were prepared by vacuum induction melting of high-purity elemental
The second mechanism is based on the local stabilization of constituents (99.95% Ti, 99.98% Ni, 99.995% Pd, 99.995% Pt and
martensite, retained above the austenite finish (Af) temperature. 99.95% Sc). Each ingot was homogenized at 1050 1C for 72 h under
Similar to the residual stress fields in the first mechanism, vacuum and allowed to furnace cool. Subsequent to the homogeniza-
retained martensite plates are attributed with influencing the tion, ingots were placed in mild-steel extrusion cans and extruded at
nucleation of certain preferred martensite variants during the 900 1C with an area reduction ratio of 7:1. Rectangular and cylindrical
transformation process, resulting in TWSME [12]. dog-bone shaped tension specimens with gage dimensions of
Common to these two mechanisms is the prerequisite for 8 mm  3 mm  1.5 mm and 3.81 mm Ø  16.4 mm, respectively,
some amount of plasticity to occur in the material during training, were cut from the hot-rolled and hot-extruded rods for thermome-
which manifests itself as residual strain [12]. However, it has also chanical training and TWSME characterization.
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 655

Binary Ni49.9Ti50.1 was chosen to represent the baseline of a 7.5 kW induction heater. Temperature was measured with a
behavior of a conventional SMA without complications from K-type thermocouple, which was spot-welded to the middle of
off-stoichiometry strengthening and precipitation, as occurs in the gage section. Samples were heated and cooled at rates of
Ni-rich compositions. Ni24.5Ti50.5Pd25 was selected as an HTSMA 30 72 1C/min and 2072 1C/min, respectively. Additional experi-
appealing to applications requiring intermediate transformation mental details related to this test setup can be found in [8].
temperatures of around 200 1C [3,5,7–10]. Ni24.5Ti50Pd25Sc0.5 has
transformation temperatures about 10 1C lower compared to the
ternary alloy. However, as shown in a previous study [4], this 3. Results and discussion
HTSMA displays slightly improved functional behavior in the
form of better dimensional stability and smaller thermal hyster- Fig. 1a illustrates a typical strain vs. temperature evolution
esis during isobaric thermal cycling compared to the ternary during the 100-cycle thermomechanical training procedure
alloy. One question to be answered is whether these improve- employed in this study. The results are for the binary Ni49.9Ti50.1
ments will also extend to the TWSME response. Ni28.5Ti50.5Pt21 trained under 150 MPa. Training is initiated by loading the
was chosen as a representative alloy for high-temperature appli- sample at room temperature in martensite and subsequently
cations, capable of actuation around 350 1C. Constant stress heating above the Af temperature to the upper cycle temperature
thermal cycling behavior of a similar composition, Ni30Ti50Pt20 (UCT), which in this example was 165 1C, to obtain austenite
has previously been studied by Noebe et al. [55,56] and this under the applied load. The first complete and all subsequent
material exhibited good work output and dimensional stability thermal cycles are comprised of cooling from the UCT to the lower
with transformation temperatures around 300 1C. cycle temperature (LCT) and heating of the sample back to the
UCT, e.g., through the forward and reverse transformations,
2.2. Thermomechanical training and TWSME characterization respectively. A common characteristic for all these materials is
that the 1st training cycle was completed with a relatively high
The choice of training procedure to induce the TWSME in the value of residual strain, eres , as compared to subsequent cycles.
currently studied materials was thermal cycling under constant We use the term eres for residual strain per cycle to denote that it
stress for 100 cycles. Training was carried out under tensile is strain that was not recovered during a given thermal cycle and
stresses of 80 MPa, 150 MPa and 200 MPa for all alloys. One- not to imply any given mechanism for the strain generation. With
hundred thermal cycles was selected, since it was expected to increasing number of cycles, the material obtains an almost stable
yield a reasonably stabilized material response with minimal shape memory response, demonstrating a subtly different strain
changes in shape and transformation temperatures upon further
thermal cycling [8,9].
For the Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5 HTSMAs, train-
ing was carried out on a custom-built constant-load testing frame
using the small rectangular tensile samples. Heating and cooling
was done at a rate of 571 1C/min. Samples were heated by
conduction from the grips, which were in turn heated by radiation
through the use of an environmental furnace equipped with four
1 kW halogen lamps. For cooling, water was circulated around the
grips flowing through copper tubing. Temperature was controlled
with a K-Type thermocouple attached to the middle of the sample
gage section. To minimize the radiation heat transfer on the
sample surface (which will lead to erroneous temperature
readings that are not representative of the bulk specimen), the
sample was shielded with a 1 mm thick reflective aluminum foil.
A capacitive displacement probe (Capacitecs HPC-75) with a
linear range of 0–1.5 mm was attached to the grips to measure
the displacements during the training process. Axial strain was
calculated by dividing the change in length to the initial gage
section length.
Following thermomechanical training, samples were unloaded
and TWSME was characterized using a separate MTS testing
frame for 10 stress-free cycles to assess its magnitude and
stability. Strain was measured using an MTS high-temperature
extensometer with a gage length of 12.7 mm and a  20/þ20%
strain range. Samples were heated through conduction from
the grips with heating bands. Cooling was achieved through
conduction by flowing liquid nitrogen in copper tubes wrapped
around the grips. Heating and cooling rate of the samples was
10 72 1C/min. Similar to the training procedure, temperature was
measured using a K-type thermocouple attached to the middle of
Fig. 1. (a) An example of the thermomechanical training procedure used and the
the sample gage section. resulting strain evolution due to repeated isobaric thermal cycling. Data is for the
For the Ni49.9Ti50.1 and Ni28.5Ti50.5Pt21 samples, both training Ni49.9Ti50.1 alloy at 150 MPa. The 1st and the 100th cycles are highlighted. Total
and TWSME characterization were performed on an MTS servo- residual strain,etotal
res , was measured as the cumulative residual strain per cycle, eres,

hydraulic load frame using the cylindrical dog-bone samples. after 100 cycles. (b) 10 stress-free TWSME cycles following the training regimen
shown in (a). TWSM strain, eTWSM, was calculated as the strain difference between
Strain was measured using an MTS high-temperature extens- the cold and hot shapes for a given thermal cycle. Note that the stress-free strain–
ometer with a gage-length of 12.7 mm and a strain range of temperature cycles were rezeroed for easier comparison between conditions and
10/ þ20%. Heating of the samples was achieved through the use various materials.
656 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

vs. temperature curve at the end of the 100th cycle that has a extracted from the 80 MPa data along with 150 MPa and 200 MPa
nearly negligible eres per cycle. During the course of training, the training results and summarized in Table 1.
material accumulates a total residual strain, etotal
res , which is the The 80 MPa training results, shown in Fig. 2, reveal that all
total strain increase in the sample after 100 isobaric thermal materials exhibited a relatively small transient response during
cycles, measured in the austenite state at the UCT. the very first heating cycle. After this initial transient response, a
Following training, the sample is unloaded at room tempera- more consistent erec was observed that evolved much more subtly
ture and heated above the stress-free Af temperature before the during the course of training. The highest erec values were seen in
1st stress-free TWSME cycle, recovering a fraction of the strain
associated with the detwinned, post-trained microstructure
Table 1
(Fig. 1b). During TWSME cycling, degradation in stability of the A summary of recovered transformation strain, erec, during the 1st and 100th
TWSME is encountered, evidenced by a decrease in cold and hot- training cycles and total accumulated strain, etotal after 100 cycles, determined
res
shape strains (Fig. 1b). This is the opposite behavior of what is from thermomechanical training tests conducted under different stress levels and
observed in Fig. 1a during training, where both the cold and hot- upper cycle temperatures (UCTs).
shape strains increase with training cycles. In this study, the
Material (at%) Training Training Rec strain Rec strain Total
stability of the TWSME is also quantified by the change in the stress UCT (1C) (erec) 1st (erec) residual
two-way shape memory strain, eTWSM, which is calculated as the (MPa) cycle (%) 100th strain,
difference between the cold and hot-shape strains for a given cycle (%) (etotal
res ) (%)
stress-free cycle, as demonstrated in Fig. 1b.
Ni49.9Ti50.1 80 165 2.04 3.68 4.29
150 165 3.62 3.39 7.15
150 200 3.82 3.76 12.38
3.1. Thermomechanical training and the evolution of shape memory 200 165 3.76 3.44 10.11
behavior Ni24.5Ti50.5Pd25 80 280 1.68 2.40 1.10
150 280 2.28 2.55 2.26
150 320 2.43 2.73 2.36
Fig. 2 is a compilation of the strain vs. temperature response
200 280 2.71 2.69 3.00
for binary Ni49.9Ti50.1, Ni24.5Ti50.5Pd25, Ni24.5Ti50Pd25Sc0.5 and Ni28.5Ti50.5Pt21 80 500 0.32 0.55 1.17
Ni28.5Ti50.5Pt21 during the 100-cycle training process at 80 MPa. 150 500 1.34 2.24 4.53
To better understand the generation of TWSM behavior in each 200 500 2.09
trained material, pertinent information such as etotal and the Ni24.5Ti50Pd25Sc0.5 80 280 1.80 2.41 1.12
res
150 280 2.49 2.74 1.74
recovered transformation strain, erec for the 1st and 100th train- 200 280 2.73 2.75 2.80
ing cycles, measured during the reverse transformation, were

Fig. 2. 80 MPa, 100-cycle thermomechanical training results for (a) Ni49.9Ti50.1, (b) Ni24.5Ti50.5Pd25, (c) Ni24.5Ti50Pd25Sc0.5 and (e) Ni28.5Ti50.5Pt21.
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 657

Ni49.9Ti50.1 at all stress levels investigated, while Ni28.5Ti50.5Pt21 martensite variants during the martensitic transformation, while
had the lowest values, which was attributed to a higher stress actually inhibiting the formation of non-favorable variants [15].
needed to reorient martensitic variants in this material [55]. The oriented stress fields are usually imposed in the material
All samples accumulated strain during the thermomechanical through the presence of defects such as dislocations that are
training procedure. At the end of the 80 MPa, 100-cycle training, generated during thermomechanical training. Thus, some amount
the lowest etotal res levels occurred in the Ni24.5Ti50.5Pd25 and of residual strain is expected during training; yet overstressing
Ni24.5Ti50Pd25Sc0.5 HTSMAs, while Ni49.9Ti50.1 exhibited the high- the material might result in an increase in the plastic deformation
est values (Table 1). and facilitate the formation of retained martensite. This in turn,
As expected, an increasing training stress led to an increased would result in a decrease in the TWSME due to a decrease in the
value of etotal
res for all materials, though for a given stress level, the amount of transforming volume. Thus, it would be anticipated
values were always larger for binary Ni49.9Ti50.1 (Table 1 and that there is an ideal stress or range of stresses for optimizing
Fig. 3). It is also interesting to see that Ni28.5Ti50.5Pt21 accumu- TWSME through thermomechanical training.
lated a relatively higher value of etotal
res under 150 MPa compared to
Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5, while it had almost the
same etotal
res at 80 MPa. This is most likely the result of creep 3.2.1. Ni49.9Ti50.1
deformation due to the increased stress level and the need to For Ni49.9Ti50.1, training stress had a complex effect on the
cycle to a much higher UCT than the other materials, which was magnitude and stability of the TWSME. The magnitude of the
necessary to complete the reverse transformation. For this same 10th cycle eTWSM varied inversely proportional to the training
reason, Ni28.5Ti50.5Pt21 could not endure 100 thermal cycles under stress, with increasing training stress resulting in lower 10th
200 MPa and failed at the 60th cycle generating a etotal res of 8.3%. cycle eTWSM (Fig. 5). However, while the magnitude of the eTWSM
Fig. 4 illustrates the evolution of the TWSME during the subse- decreased with increasing training stress, the stability of the
quent 10 stress-free cycles after training at 80 MPa (Fig. 4a, c, e and g), eTWSM increased as evident from the smaller changes between
as well as the changes in eTWSM as a function of cycle count after successive cycles (Fig. 4b). For instance, upon unloading the
training at all three stress levels (Fig. 4b, d, f and h). Table 2 lists 1st 80 MPa-trained material and heating above the Af temperature,
and 10th cycle eTWSM values; efficiency factor for each trained more than 4% strain was recovered (Fig. 4a). Only 73% of this
material, defined as the ratio of the 1st cycle eTWSM and the erec of value, or about 3% strain, was carried over to the 1st TWSME cycle
the 100th training cycle; the amount of degradation in eTWSM as well and the eTWSM degraded by a further 10% in 10 thermal cycles
as the cold and hot-shape strain values during 10 TWSME cycles. (Table 2). The eTWSM for the 150 MPa-trained material decreased
With the exception of Ni28.5Ti50.5Pt21, efficiency factors for materials from 2.6% to 2.5% in 10 cycles, while the eTWSM for the 200 MPa-
trained under 80 MPa were found to be quite similar, around 0.8, trained material was found to be very stable during cycling, and
which indicated that the 1st cycle eTWSM values were only  20% less may have actually increased slightly by the end of the 10th cycle
than the erec values of the 100th training cycle (an efficiency factor of (Fig. 4b). At this point, it should be noted that a stable eTWSM is not
1 indicates that 1st cycle eTWSM is equal to the erec of the 100th necessarily a sign of a stable TWSME. eTWSM might be quite stable
training cycle). In terms of TWSME stability, Ni49.9Ti50.1 demonstrated while there is a significant change in overall dimensions of the
poor performance with large degradations in eTWSM (10%) and cold sample due to an almost equal amount of degradation in cold and
and hot-shape strains (a decrease in strain of 0.6% and 0.3%, hot-shape strains. On the other hand, minimal changes in cold
respectively) during TWSME cycling. Ni24.5Ti50.5Pd25 and Ni24.5Ti50 and hot-shape strains will by definition result in a stable eTWSM.
Pd25Sc0.5 HTSMAs exhibited superior stability with less degradation in NiTi was found to display poor TWSME stability by exhibiting
eTWSM and cold and hot-shape strains. large changes in cold and hot-shape strains upon TWSME cycling
even though the changes in eTWSM were sometimes very minor
(Table 2). For example, after 10 cycles, the hot shape of the
3.2. Effect of training stress on the TWSME
200 MPa-trained material changed by 0.7% strain even though the
magnitude of eTWSM changed by only 0.1% strain.
As mentioned previously, TWSME arises from the presence of
For the 150 MPa-trained Ni49.9Ti50.1, etotalres was around 7%
oriented internal stress fields, which bias the formation of specific
(Fig. 3). Stress-free thermal cycling altered the post-trained
dislocation structure in a way that each cycle caused a progres-
sive relaxation of the oriented internal stresses. This inevitably
caused a decrease in the cold-shape strain due to the formation of
more self-accommodated martensite instead of heterogeneously
nucleated single-variant martensite (Fig. 4a). At the end of each
cycle, it was noticed that the hot-shape strain did not match the
strain value of the previous cycle, i.e., there was also a decrease in
the hot-shape strain. The primary reason for the decrease in hot-
shape strain upon stress-free thermal cycling is that after a total
deformation of 7%, there is probably retained martensite in the
material at the UCT [54]. But upon stress-free thermal cycling and
gradual relaxation of the internal stresses some of this oriented
retained martensite is able to transform back to austenite result-
ing in a decrease in the hot-shape strain as well. It is also possible
that back stresses due to the dislocation structures developed
during training were relaxed, which would also result in a slight
strain recovery. TWSM behaviors of 80 MPa and 200 MPa-trained
NiTi can be described based on the same scenarios. The notable
difference between these conditions is the amount of etotal res
Fig. 3. Total residual strain (etotal
res ), measured in the austenite, after 100 thermo-
generation during training. This variation could result in different
mechanical cycles as a function of stress level for the different SMA systems. amounts of retained martensite and dislocation densities in the
658 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

Fig. 4. The evolution of strain vs. temperature behavior during no-load thermomechanical cycling used to determine TWSME after training at 80 MPa for (a) Ni49.9Ti50.1,
(c) Ni24.5Ti50.5Pd25, (e) Ni24.5Ti50Pd25Sc0.5 and (g) Ni28.5Ti50.5Pt21. Note that the strain–temperature cycles were rezeroed after training for easy comparison. The variation of
the TWSM strain during stress-free thermal cycling for (b) Ni49.9Ti50.1, (d) Ni24.5Ti50.5Pd25, (f) Ni24.5Ti50Pd25Sc0.5 and (h) Ni28.5Ti50.5Pt21 after training at various stress levels.

trained materials, which alter the volume fraction of oriented stable eTWSM but unstable axial dimension upon cycling, can be
martensite variants that cause the TWSME. For instance, the attributed to a recovery of a large volume fraction of retained
peculiar TWSM behavior of the 200 MPa-trained material, i.e., martensite. For this material, the decrease in the hot-shape strain
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 659

Table 2
Magnitude and stability of the TWSME for the various SMA systems after thermomechanical training under different stress levels and upper cycle temperatures (UCTs).

Training Training TWSME 1st cycle TWSM 10th cycle TWSM Efficiency TWSM strain Hot-shape strain Cold-shape
stress UCT (1C) UCT (1C) strain (eTWSM) (%) strain (eTWSM) (%) factora degradationb degradationc strain
(MPa) degradationc

Ni49.9Ti50.1 80 165 165 3.06 2.75 0.83 0.10 0.33 0.60


150 165 165 2.60 2.53 0.77 0.03 0.68 0.64
200 165 2.95 2.82 0.78 0.04 0.46 0.49
200 2.43 2.09 0.65 0.14 0.64 0.91
200 165 165 2.34 2.36 0.68 -0.01 0.69 0.58
Ni24.5Ti50.5Pd25 80 280 280 2.12 1.99 0.88 0.06 0.10 0.24
150 280 280 2.41 2.28 0.95 0.05 0.17 0.31
320 280 2.34 2.20 0.86 0.06 0.13 0.27
200 280 280 2.58 2.46 0.96 0.05 0.08 0.23
Ni28.5Ti50.5Pt21 80 500 500 -0.04 -0.08 -0.08 -1.00 0.04 0.09
150 500 500 0.13 0.00 0.06 1.00 0.03 0.09
200 500
Ni24.5Ti50Pd25Sc0.5 80 280 280 2.11 2.04 0.88 0.03 0.20 0.26
150 280 280 2.46 2.38 0.90 0.03 0.13 0.22
200 280 280 2.56 2.46 0.93 0.04 0.10 0.22

a
Efficiency factor was calculated as the ratio of the two-way shape memory strain (eTWSM) of the 1st stress-free cycle to the recoverable transformation strain (erec)
of the 100th training cycle, i.e., (1st cycle eTWSM/100th cycle erec).
b
Degradation in eTWSM upon stress-free thermal cycling was calculated as the ratio of the difference between the 1st and 10th stress-free cycle eTWSM values to the
eTWSM of the 1st stress-free cycle, i.e., ((1st cycle eTWSM – 10th cycle eTWSM)/(1st cycle eTWSM)).
c
Degradations in cold and hot-shape strains were calculated as the difference between the percent shape strain values of the 1st and 10th stress-free cycles.

Fig. 5. (a) TWSM strain values after 10 stress-free cycles for the tested materials as a function of training stress. The 10th no-load thermal cycles for (b) Ni49.9Ti50.1
and (c) Ni24.5Ti50.5Pd25 after training under 80 MPa, 150 MPa and 200 MPa.

was very similar, if not larger, than the decrease in the cold-shape 3.2.2. Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5
strain, resulting in a very stable or even slightly increasing eTWSM, Compared to Ni49.9Ti50.1, much lower etotal res values were
while resulting in a significant change in the axial dimension of recorded at all training stress levels for both Ni24.5Ti50.5Pd25 and
the sample. Ni24.5Ti50Pd25Sc0.5 HTSMAs (Fig. 3). While etotalres was 1.1% for
660 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

training at 80 MPa for Ni24.5Ti50.5Pd25, it doubled to 2.3% when the stress used, the first stress-free cycle at 500 1C relaxed any
stress was increased to 150 MPa and at 200 MPa, etotal res was 3.0%. beneficial dislocation structure that might have developed during
Marginally smaller values for etotalres were observed at each stress training [56], eliminating the driving force for the TWSME (Fig. 4g
level for Ni24.5Ti50Pd25Sc0.5 (Table 1). These values were much less and h).
than that observed for Ni49.9Ti50.1. Thus the NiTiPd(Sc) HTMSAs
were much more resistant to the development of plastic deforma-
tion at the stress levels investigated than Ni49.9Ti50.1, even though 3.3. Effects of composition on the TWSME
the NiTiPd(Sc) HTSMAs were thermally cycled to much higher
temperatures. In addition, the transformation strain for the The most obvious change in the TWSME characteristics
NiTiPd(Sc) HTSMAs increased with training stress. Thus an between the various SMA systems is the basic effect of composi-
increasing transformation strain with increasing training stress tion on transformation temperatures and the role that cycling
naturally led to an increasing eTWSM with training stress. above the various transformation temperatures plays on subse-
Furthermore, eTWSM of the trained NiTiPd(Sc) HTSMAs were quent microstructural development. Fig. 6 shows the 10th
found to be very close to the maximum recoverable strain levels TWSME cycles for all materials previously trained under
under stress, resulting in high training efficiencies (Table 2), 150 MPa. The substitution of 25 at% Pd for Ni in NiTi results in
which actually increased with increasing training stress level an increase in the transformation temperatures of Ni24.5Ti50.5Pd25
(Fig. 5), while the opposite behavior was observed in binary to around 200 1C. The small quaternary alloying addition of Sc to
Ni49.9Ti50.1. For example, the strain recovered during the initial Ni24.5Ti50.5Pd25 subsequently decreased the resulting transforma-
heating of the 150 MPa-trained material above the Af temperature tion temperatures slightly, by about 10 1C, consistent with the
was very close to the 1st cycle eTWSM. Out of the 2.55% strain isobaric thermal cycling results and differential scanning calori-
recovered during the last training cycle, 95%, or 2.4% strain, was metry (DSC) analysis reported previously [4]. Pt has a greater
carried over to the subsequent TWSME cycle and only degraded effect per unit of alloying addition on transformation tempera-
by about 5% at the end of 10 stress-free thermal cycles for all tures than Pd, such that the addition of 21% Pt to NiTi raised the
training stress levels. transformation temperatures of Ni28.5Ti50.5Pt21 to between 350 1C
In terms of TWSME stability, cold and hot-shape strains of and 400 1C. Thus cycling above this temperature to ensure
Ni24.5Ti50.5Pd25 were very stable and changed much less upon complete transformation under stress, i.e., 500 1C, was high
stress-free thermal cycling than Ni49.9Ti50.1 (Fig. 4c and Table 2). enough to allow complete recovery of any dislocation structure
The hot-shape strain change was less than 30% of that observed in that may have developed during training, removing any driving
Ni49.9Ti50.1 for a given stress level while the change in cold-shape force for subsequent TWSME. Noebe et al. [56] have previously
dimensions was less than half of that observed in the binary alloy. shown that recovery processes begin at approximately 450 1C in a
This would indicate that the dislocation structures developed in Ni29.5Ti50.5Pt20 HTSMA. It thus remains a technological challenge
Ni24.5Ti50.5Pd25 during training were more resistant to relaxation to obtain stable TWSME response in HTSMAs when transforma-
and recovery of the internal stresses than in Ni49.9Ti50.1, in spite of tion temperatures are close to the recovery/recrystallization
cycling to higher temperatures. But there was also a consistent temperatures of the particular alloy.
anisotropy in the cold versus hot-shape strain change, which can be Substitution of Pd with Ni not only increases the transformation
seen in Fig. 4c and quantified in Table 2. In general, the hot-shape temperatures, but also acts as a solid-solution strengthener and
strain change was about half of that observed for the change in cold results in a change in the martensite structure from monoclinic B19’,
shape strain during the ten stress-free thermal cycles. As discussed to orthorhombic B19, resulting in improved compatibility [4]
above, the change in shape of the sample measured at the LCT in the between transforming phases (cubic, B2 austenite transforming to
martensite phase is due to a slight reversion of oriented martensite orthorhombic martensite instead of monoclinic martensite). This
back to self-accommodated martensite. The change in hot-shape improved crystallographic compatibility coupled with increased
dimensions measured at the UCT is presumably due to reversion of strength levels, results in a much more stable thermomechanical
retained martensite to austenite. This would indicate that in cycling response compared to that of Ni49.9Ti50.1. This is evidenced
Ni24.5Ti50.5Pd25 either very little retained martensite was generated by smaller etotal
res values during training under all stress levels (Fig. 3)
during training or the retained martensite that was formed was and smaller changes in cold and hot-shape strains during TWSME
very stable. Ni24.5Ti50Pd25Sc0.5 displayed very similar results to
Ni24.5Ti50.5Pd25, with similar or slightly improved eTWSM and cold
and hot-shape strain stability (Table 2).

3.2.3. Ni28.5Ti50.5Pt21
The high transformation temperatures of Ni28.5Ti50.5Pt21
necessitated thermal cycling to much higher temperatures during
the training process, i.e., 500 1C compared to just 280 1C for the
NiTiPd(Sc) HTSMAs. Heating of the 80 MPa-trained sample above
the Af temperature resulted in a strain recovery of 0.5% (Fig. 4g),
which is almost the same as the erec of the last training cycle
(Table 2). Further stress-free cycling, however, did not yield any
useful eTWSM. At the end of 10 stress-free thermal cycles, the
eTWSM of Ni28.5Ti50.5Pt21 was almost non-existent regardless of the
training stress.
Although Ni28.5Ti50.5Pt21 exhibited reasonable recoverable and
residual strains during training for an alloy with such high
Fig. 6. Effect of composition on the TWSME for materials previously trained at
transformation temperatures [57], it did not show significant 150 MPa for 100 thermal cycles. The strain–temperature curves represent the
evidence of TWSME. This can be attributed to the high UCT TWSM response for the 10th load-free thermal cycle. The right axis represents the
required for thermal cycling. Therefore, regardless of the training efficiency of the eTWSM compared to the erec during training.
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 661

cycles (Fig. 4c), even though the sample was cycled to a higher UCT. reaching a maximum erec at a UCT of around 250 1C under
For these same reasons, there was also a decrease in energy 200 MPa [60]. It is of interest to determine whether similar trends
dissipation during transformation, contributing to a significant will be seen after a relatively high number of thermal cycles and
reduction of the thermal hysteresis during both training and the how this may affect the resulting TWSME. Consequently,
TWSME cycles. Ni49.9Ti50.1 and Ni24.5Ti50.5Pd25 were trained using two different
Microalloying with Sc further improved the compatibility of UCT values under 150 MPa. Ni49.9Ti50.1 was cycled to UCTs of
the transforming phases in Ni24.5Ti50.5Pd25 and increased the 165 1C and 200 1C, while the Ni24.5Ti50.5Pd25 was cycled using
material strength due to solid-solution strengthening [4]. As a UCTs of 280 1C and 320 1C.
result, Ni24.5Ti50Pd25Sc0.5 had slightly smaller thermal hysteresis
compared to Ni24.5Ti50.5Pd25 during both training and TWSME
cycles. Recently, the second eigenvalue, l2, of the transformation 3.4.1. Ni49.9Ti50.1
stretch tensor that maps the austenite lattice to the martensite The strain vs. temperature response of the Ni49.9Ti50.1 during
lattice, has been related to the thermal hysteresis associated with thermomechanical training under 150 MPa is shown in Fig. 7a
the martensitic transformation [58]. It has been shown that as the and b with UCTs of 165 1C and 200 1C, respectively. At a first
l2 value gets closer to 1, compatibility between transforming glance, it is clear that etotal
res generated during training increased
phases increases, leading to a smaller thermal hysteresis. Current from 7.2% to 12.4% due to an increase of UCT from 165 1C to
results for the l2 of binary NiTi (0.9663) [59], Ni24.5Ti50.5Pd25 200 1C. In spite of the large difference in etotal res there are also
(1.0171) [4] and Ni24.5Ti50Pd25Sc0.5 (1.0158) [4] are consistent similarities in the strain vs. temperature evolution for these two
with this finding for the thermal hysteresis associated with the cases. For both materials, the 1st training cycle is characterized by
TWSME, as well. a relatively large eres and thermal hysteresis. However, during the
In terms of eTWSM, Ni49.9Ti50.1 exhibited a higher absolute value course of training, the materials started to show a more stable
at the end of 10 TWSME cycles compared to Ni24.5Ti50.5Pd25 and behavior characterized by smaller shifts in transformation tem-
Ni24.5Ti50Pd25Sc0.5 (Fig. 6). However, when the TWSM values were peratures during cycling and smaller values of eres per cycle. Also,
normalized with respect to the highest amount of recoverable a smaller thermal hysteresis was observed, indicative of slightly
strain that could be obtained from each material at a stress level less dissipation of elastic stored energy during later transforma-
of 150 MPa, it was observed that Ni24.5Ti50.5Pd25 and Ni24.5Ti50 tion cycles, which could be due to a reduced interaction between
Pd25Sc0.5 HTSMAs outperformed Ni49.9Ti50.1 with values very martensite variant pairs or to the development of low energy
close to 1 (Fig. 6). This indicates that for a specific thermomecha- dislocation structures with cycling.
nical treatment, the NiTiPd(Sc) HTMSAs responded to training There are also obvious differences between these two condi-
more efficiently and have nearly perfect TWSME relative to their tions in terms of erec . Compared to the 165 1C-UCT trained
load-biased behavior during training. A high value of training material, the 100th cycle of the 200 1C-UCT trained material had
efficiency indicates that dislocation structures and local stress a noticeably higher erec (3.8% vs. 3.4%) and higher transformation
fields generated by cycling are very effective in biasing the temperatures. For the 165 1C-UCT trained material erec decreased
same martensite variants that were formed through load-biased from 3.6% to 3.4% after 100 thermal cycles at 150 MPa. For the
thermal cycling, during the subsequent thermal cycling under 200 1C-UCT trained material, erec remained quite stable at around
zero stress [28]. On the other hand, Ni28.5Ti50.5Pt21 had a value 3.8% during the entire 100 cycle training regime. As plasticity was
close to 0, indicating that this material is not suitable for TWSME more prominent at the higher UCT value (Fig. 7a vs. b), one would
applications. anticipate a reduced transforming volume. Regardless, the higher
transformation strain of the 200 1C-UCT trained material can be
3.4. Effect of training upper cycle temperature (UCT) on the TWSME explained by a retained martensite argument. While a higher UCT
results in higher values of etotal res during training, the amount of
It is commonly accepted that for a chosen SMA system, retained martensite is probably diminished compared to the
transformation strains generated during thermomechanical train- lower UCT case, as evidenced by the increased erec . In addition,
ing will be representative of the resultant eTWSM. This is the case while plasticity was greater in the 200 1C-UCT trained material as
for the Ni49.9Ti50.1 SMA and NiTiPd(Sc) HTSMAs studied here. measured by etotal
res , it is possible that the transforming volume was
Thus, any training parameter that can potentially increase the actually relatively unchanged or even greater than that for the
erec of the last training cycle will most likely increase the eTWSM of 165 1C-UCT trained material, because excess dislocations could
the following stress-free cycles. While erec is a function of applied have been annihilated during cycling or rearranged into low
stress, it also can be affected by the UCT the material is exposed to energy networks, given the additional thermal energy introduced
during cycling [53]. While a high UCT is more likely to trigger by cycling to 200 1C.
global plasticity due to the decrease in critical shear stress (CSS) The resultant TWSME obtained from these two cases is also
for slip, it is also possible that higher UCT values may reduce the different in terms of stability and magnitude. While a UCT of
amount of retained martensite present in the alloy by heating 200 1C during thermomechanical training resulted in a higher
further above the Af temperature. value of etotal
res , it also yielded a higher initial eTWSM of around 3.0%
The initial choice of UCT for thermomechanical training was (compared to 2.6% when a UCT of 165 1C was used). The
made to ensure complete transformation of martensite to auste- degradation levels associated with cold and hot-shape strains
nite under the applied stress. Since all materials have different are also different. After 10 TWSME cycles, the material trained
transformation temperatures, different UCTs were used during with a UCT of 165 1C exhibited a eTWSM of around 2.5% (Fig. 7e)
training. Ni49.9Ti50.1 was thermally cycled to 165 1C, whereas with about 0.7% decrease in both cold and hot-shape strains
NiTiPd(Sc) HTSMAs were cycled to 280 1C and Ni28.5Ti50.5Pt21 (Fig. 7c). Fig. 7d illustrates the TWSM response of the same
was heated to 500 1C. Padula et al. [53] have already demon- material trained with a UCT of 200 1C. Compared to the material
strated that UCT affects the thermomechanical response of the trained with a UCT of 165 1C, the decrease in both cold and hot-
Ni49.9Ti50.1 SMA used in this study, at least for a relatively low shape strains are less by about 0.2% and the resulting eTWSM was
number of thermal cycles. They have shown that the material 2.8% at the end of 10 TWSME cycles. The larger decrease in the
cycled using a higher UCT, up to some maximum value depending cold-shape strain, in the material trained with a UCT of 165 1C, is
on the stress level, exhibited higher transformation strains probably due to a larger relaxation of internal stresses, and thus
662 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

Fig. 7. The evolution of strain vs. temperature response of Ni49.9Ti50.1 trained under 150 MPa using a UCT of (a) 165 1C and (b) 200 1C. The evolution of the TWSME during
the subsequent 10 stress-free cycles for the material trained with UCT of (c) 165 1C and (d) 200 1C. (e) A comparison of 10th TWSME cycles for the Ni49.9Ti50.1 alloy trained
with different UCTs.

the formation of more self-accommodated martensite at the with a UCT of 320 1C resulted in a etotal res value of 2.4% (Fig. 8b).
expense of the preferred oriented martensite. The larger etotal
res of When the 1st and 100th cycles are inspected for both cases, it is
12.4% in the 200 1C-UCT trained material is likely to have clear that the evolution in strain vs. temperature responses are
produced a more stable, lower energy dislocation network, which unaffected by the choice of UCT. Unlike Ni49.9Ti50.1, the 1st and
would be more resistant to relaxation of internal stresses in this 100th cycles for Ni24.5Ti50.5Pd25 are very comparable in terms of
material, leading to an improved TWSME stability. Finally, the erec and the shape of the hysteresis loops. These are all indications
smaller degradation in the hot-shape strain of the same material of a shape memory response that is not affected by the choice of
is possibly associated with smaller amounts of retained martensite UCT, at least over the temperature range investigated. In addition,
present in the microstructure that reverts back to austenite during cycling to either UCT resulted in an approximately 16 1C shift in
stress-free cycling. transformation temperatures during the course of training com-
pared to a 7 1C shift in Ni49.9Ti50.1. This is an indication that larger
residual stresses are generated in the Ni24.5Ti50.5Pd25 alloy,
3.4.2. Ni24.5Ti50.5Pd25 assisting the transformation.
In contrast to Ni49.9Ti50.1, a 40 1C increase in the UCT used The initial stress-free heating curves after the training procedure
during training of Ni24.5Ti50.5Pd25 resulted in no significant follow almost the same strain–temperature path for the two cases
difference in etotal
res levels (Table 1). The sample, which was (Fig. 8c and d) indicating similar values of internal stresses in both
thermally cycled under 150 MPa using a UCT of 280 1C had a materials regardless of the UCT used. Consequently, the changes in
etotal
res of 2.3% (Fig. 8a), while the same thermomechanical training cold and hot-shape strains are indistinguishable and are minimal
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 663

Fig. 8. The evolution of strain vs. temperature response of Ni24.5Ti50.5Pd25 trained under 150 MPa using a UCT of (a) 280 1C and (b) 320 1C. The evolution of the TWSME
during the subsequent 10 stress-free cycles for the material trained with UCT of (c) 280 1C and (d) 320 1C. (e) A comparison of 10th TWSME cycles for the Ni24.5Ti50.5Pd25
alloy trained with different UCTs.

compared to Ni49.9Ti50.1. The 10th TWSME cycle for each case is UCT2W on the stability and magnitude of TWSME, the binary
shown in Fig. 8e. The two strain vs. temperature curves are nearly Ni49.9Ti50.1 alloy, trained under 150 MPa using a UCT of 200 1C,
identical with similar transformation strain, hysteresis, and trans- was thermally cycled stress-free using a UCT2W of 165 1C and
formation temperatures. The 0.1% difference in eTWSM for the 10th 200 1C. As expected for this alloy, a higher UCT2W used during
stress-free cycle (Fig. 8e) is essentially within the uncertainty of the stress-free cycling severely affected both the magnitude and
strain measurements and is hardly significant. stability of TWSME. Out of the 3.8% erec for the last training cycle,
only 2.4% (vs. 3.0% with a UCT2W of 165 1C) could be carried over
3.5. Effect of TWSME cycling upper cycle temperature (UCT2W) on to the 1st TWSME cycle (Table 2). The eTWSM decreased by 0.3%
the TWSME (vs. 0.1% with a UCT2W of 165 1C) in 10 cycles and the largest
decrease in cold-shape strain was observed for this condition.
In the previous section, the significance of the UCT used for This increased degradation in stability is undoubtedly due to the
thermomechanical training cycles was demonstrated for the rapid relaxation of oriented internal stresses at this higher cycling
binary Ni49.9Ti50.1 alloy (while the Ni24.5Ti50.5Pd25 alloy was temperature.
insensitive to this effect over the narrow range of temperatures
investigated). Consequently, the selection of UCT for the TWSME 3.6. Factors that affect the magnitude and stability of TWSME
cycles also deserves attention. To avoid any confusion with the
UCT used during the training process, the UCT used during the For a review of the factors that can affect the magnitude and
TWSME cycles will be denoted as UCT2W. To show the effect of stability of TWSME, the schematic in Fig. 9 is introduced.
664 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

Fig. 9. Schematic showing the parameters involved in the generation of TWSME, its magnitude and stability. See text for details. (?: dislocations, s: applied stress during
training, UCT: upper cycle temperature during either training or TWSME cycles, TRIP: transformation-induced plasticity, RM: retained martensite, etotal res : total residual
strain measured in the austenite state at the UCT, OM: oriented martensite (responsible for the TWSME), SAM: self-accommodated martensite.).

The figure is divided into five main parts, each summarizing less generally throughout the material [62–65]. Finally, the effects
relevant microstructural parameters, mechanisms, or testing dislocation relaxation, recovery and even creep deformation
conditions that can directly or indirectly influence the magnitude cannot be ignored during thermal cycling under stress at even
and stability of the TWSME. moderate temperatures [66] as seen in the Ni28.5Ti50.5Pt21 case.
Part I illustrates the key parameters that are likely to affect the Consequently, thermomechanical cycling during the training
microstructural evolution during thermomechanical training: the process results in an accumulation of defects in the microstruc-
applied stress (s), the choice of UCT, and the number of cycles. ture. These defects are usually either dislocations or deformation
In this study, only s and UCT were chosen as variables. The most twins [47–49,61–65,67] generated due to the mechanisms sum-
evident effect of an increase in either s (Fig. 3) or UCT (Figs. 7 and 8) marized in Part II. In addition to their contribution to plasticity,
was an increase in etotal
res generated during training. The effect of the and potentially acting as a source for preferred nucleation of
number of cycles on the TWSME was not directly investigated in oriented martensite (OM), these defects also form a barrier
this study, with the number of cycles being kept constant at 100. But against reverse transformation, stabilizing some of the martensi-
it is clear from the data that etotal
res increases with cycle count, in a tic variants such that they cannot revert back to austenite during
manner that appears to saturate at some higher number of cycles. transformation.
Though the effect of cycle count on the TWSME was not directly Part III of Fig. 9 represents the formation of dislocations (?)
investigated, it can be postulated that as the cycle count increases and retained martensite (RM) as a result of training. The hor-
and the strain–temperature response saturates with cycling, a more izontal arrow is used between ? and RM to represent the inter-
stable TWSME should be developed. play between dislocation structures and martensite stabilization.
In addition to the martensitic phase transformation in SMAs it A consequence of both dislocation generation and to a much
is possible for a number of additional inelastic and predominantly lesser extent the formation of RM is an increase in the etotal
res and
non-recoverable deformation processes to occur concurrently, as change in the hot-shape strain of the sample, which can be
illustrated in Part II of Fig. 9. These are what give rise to the major quantified from the thermomechanical testing results (Fig. 3).
component of the etotal res generated during thermomechanical Some amount of etotalres is probably necessary for the generation of
cycling or training. Depending on the temperature and applied TWSME, yet excessive deformation leads to a decrease in the
stress under which the transformation occurs, several deforma- TWSME, as seen in binary Ni49.9Ti50.1 (Fig. 5b). Although there is
tion mechanisms can come into play. The most critical, because it no unique relationship between etotal res and the TWSME, dislocation
will occur regardless of the stress–temperature conditions, is structures and RM determine the amount of transforming volume
accommodation of the volume mismatch during the martensitic available after training. Since it is not exactly known how disloca-
transformation by dislocation generation at austenite/martensite tion structures contribute to the amount of transforming volume,
interfaces, a process also known as transformation-induced plas- the connection between ? and transforming volume has been
ticity (TRIP). Evidence for this behavior in NiTi, through detailed indicated with a dashed arrow. Due to the presence of oriented
TEM studies, has been presented by a number of investigators internal stresses induced during training, a fraction of the trans-
[47,49,61]. In addition, general plasticity is also likely as internal forming volume will be comprised of OM, which is responsible
stresses or stress concentrations in the material reach the critical for the TWSME and directly controls the magnitude of the eTWSM.
resolved shear stress for slip at first locally [52] and then more or The remainder will transform into self-accommodated martensite
K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666 665

(SAM). The difference between the observed eTWSM and the alloys had eTWSM values around 2.4%, which corresponded to
theoretical maximum transformation strain for the polycrystal- almost 90% of the recoverable strain that could be obtained
line aggregate is then due to this SAM and any RM, which does from these materials during 150 MPa thermal cycling. Due to
not participate in the transformation process. the inherently high transformation temperatures of the
Obviously, microstructural features such as crystallographic Ni28.5Ti50.5Pt21, stress-free thermal cycling was performed at
compatibility between transforming phases, texture, grain size temperatures above the recovery range for the alloy, relaxing
and the initial dislocation density also affect microstructural or recovering any dislocation structures in the material. Con-
evolution during training (Part IV of Fig. 9). Higher strength levels sequently, no significant TWSME was observed.
through finer grain size, higher initial dislocation density, pre- 3. Both the stress level and UCT used during thermo-mechanical
cipitate strengthening, or preferred textures/orientations where training have significant but varied effects on the magnitude
slip is severely restricted, as well as improved compatibility and the stability of the resulting TWSME. UCT2W, the upper
between the transforming phases through compositional changes cycle temperature selected for stress-free thermal cycling
should decrease defect generation during training, resulting in a during measurement of the TWSM response, was equally
relatively lower dislocation density and possibly less RM. Other important for the stability of the TWSME. A higher UCT2W
than improved compatibility between transforming phases as was found to result in a faster degradation of the TWSME in
exemplified by comparing the response of the binary NiTi and Ni49.9Ti50.1.
NiTiPd(Sc) systems [4], the effects of these microstructural para- 4. NiTiPd(Sc) HTSMAs have been shown to be attractive candi-
meters on the TWSME were beyond the scope of this study, but dates for TWSME applications and are the only currently viable
are included in Part IV of Fig. 9 for completeness. option for use at higher temperatures (to about 200 1C). These
The actual stress-free cycling of the trained material, used to alloys generated relatively small amounts of residual strain
investigate the stability of the TWSME, can cause additional during thermomechanical training, over the range of UCTs and
microstructural evolution. First, a higher UCT2W will tend to cause stress levels investigated. The resultant TWSME was shown to
rearrangement and annihilation of the dislocation structures have excellent stability, including relatively stable cold and
generated during training, relaxing the internal stress fields and hot-shape strains, and adequate eTWSM.
affecting the proportion of OM that is formed. Furthermore, it
may assist in destabilization of any RM in the structure.
The magnitude of the eTWSM and the stability of the TWSME Acknowledgments
depend directly on the relative amounts of all types of martensite
structures (SAM, OM and RM) during stress-free cycling. Thus The authors gratefully acknowledge insightful discussions
relaxation of the dislocation structures that affect the volume with Ji Ma (Graduate research assistant at Texas A&M University)
fractions of OM and the fraction of SAM retained during the and the Shape Memory Alloy and Active Structures Group at
transformation process will result in a change in eTWSM and NASA Glenn Research Center. Particular thanks to Brian E. Franco
instability of the TWSME due to changes in cold and hot-shape (Graduate research assistant at Texas A&M University) for his
strains of the sample (Fig. 7c and d). Hence the selection of UCT help with the construction of the thermomechanical test setup
during the TWSME cycling, as well as the other microstructural and training of samples. This study has been supported by the
parameters mentioned above (Part IV of Fig. 9), will have a strong NASA Fundamental Aeronautics Program, Subsonic Fixed Wing
influence on the magnitude and stability of the TWSME. Project through Cooperative Agreement No. NNX07AB56A, with
additional support from the Aeronautical Sciences Project. IK also
acknowledges the support from the US Air Force Office of
4. Summary and conclusions Scientific Research, Grant No. FA9550-12-1-0218.

In this study, a binary Ni49.9Ti50.1 SMA and Ni28.5Ti50.5Pt21, References


Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5 HTSMAs were character-
ized in terms of the stability and magnitude of the two-way shape [1] K.N. Melton, in: T.W. Duerig, K.N. Melton, D. Stockel, C.M. Wayman (Eds.),
Engineering Aspects of Shape Memory Alloys, Butterworth-Heinemann, New
memory effect (TWSME). The TWSME was induced in these
York, 1990, pp. 21–35.
materials through a thermo-mechanical training procedure con- [2] C. Mavroidis, Res. Nondestr. Eval. 14 (2002) 1–32.
sisting of 100 thermal cycles under various stress levels with [3] J. Ma, I. Karaman, R.D. Noebe, Int. Mater. Rev. 55 (2010) 257–315.
different upper cycle temperatures (UCTs). In the process of [4] K.C. Atli, I. Karaman, R.D. Noebe, A. Garg, Y. Chumlyakov, I. Kireeva, Metall.
Mater. Trans. A 41 (2010) 2485–2497.
training, the evolutionary behavior of the strain–temperature [5] K.C. Atli, I. Karaman, R.D. Noebe, H.J. Maier, Scr. Mater. 64 (2011) 315–318.
response of the materials as a function of stress was also [6] K.C. Atli, I. Karaman, R.D. Noebe, A. Garg, Y. Chumlyakov, I. Kireeva, Acta
determined. Subsequently, the stability and magnitude of the Mater. 59 (2011) 4747–4760.
[7] G. Bigelow, R.D. Noebe, S.A. Padula, A. Garg, in: B. Berg, M.R. Mitchell, J. Proft
TWSME was assessed by running 10 stress-free thermal cycles (Eds.), Proceedings of the International Conference on Shape Memory and
following the training cycles. Superelastic Technologies, 2006, pp. 113–132.
A summary of the results and conclusions that could be [8] G. Bigelow, S.A. Padula, A. Garg, D. Gaydosh, R.D. Noebe, Metall. Mater. Trans.
A 41 (2010) 3065–3079.
derived from this study are: [9] G. Bigelow, D. Gaydosh, A. Garg, S.A. Padula, R.D. Noebe, in: S. Miyazaki (Ed.),
Proceedings of the International Conference on Shape Memory and Super-
1. In general Ni49.9Ti50.1 exhibited poor TWSME stability evi- elastic Technologies, 2007, pp. 83–92.
[10] G. Bigelow, S.A. Padula, A. Garg, R.D. Noebe, in: M.J. Dapino (Ed.), Proceedings
denced by large degradations in cold (martensite) and hot of SPIE: Behavior and Mechanics of Multifunctional and Composite Materials,
(austenite) shape strains upon stress-free thermal cycling. 2007, pp. 2B1-2B12.
Ni24.5Ti50.5Pd25 and Ni24.5Ti50Pd25Sc0.5 had superior stability [11] B. Kockar, K.C. Atli, J. Ma, M. Haouaoui, I. Karaman, M. Nagasako, R. Kainuma,
Acta Mater. 58 (2010) 6411–6420.
characterized by minimal shape changes.
[12] J. Perkins, R.O. Sponholz, Metall. Trans. A 15 (1984) 313–321.
2. After training at 150 MPa, Ni49.9Ti50.1 exhibited a eTWSM of 2.5% [13] R. Stalmans, J.V. Humbeeck, L. Delaey, Acta Metall. Mater. 40 (1992) 501–511.
at the end of 10 stress-free cycles. This value was about 60% of [14] R. Stalmans, J.V. Humbeeck, L. Delaey, J. Phys. IV 1 (1991) 403–408.
the recoverable strain level for the material under biased [15] R. Stalmans, J.V. Humbeeck, L. Delaey, Acta Metall. Mater. 40 (1992) 2921–2931.
[16] R. Stalmans, J.V. Humbeeck, L. Delaey, in: R. Yamamoto, E. Furubayashi, Y.
conditions at the end of training. On the other hand, NiTiPd(Sc) Doi, R. Fang, B. Liu (Eds.), Advanced Materials ‘93, Vol. V: Pt. A: Ecomaterials;
HTSMAs had much more efficient responses to training. These Pt. B: Shape Memory Materials and Hydrides, 1994, pp. 927–930.
666 K.C. Atli et al. / Materials Science & Engineering A 560 (2013) 653–666

[17] R. Stalmans, J.V. Humbeeck, L. Delaey, Scr. Metall. Mater. 31 (1994) 1573–1576. [48] R. Delville, B. Malard, J. Pilch, P. Sittner, D. Schryvers, Int. J. Plast. 27 (2011)
[18] E. Cingolani, M. Ahlers, Mater. Sci. Eng., A 273 (1999) 595–599. 282–297.
[19] R. Rapacioli, V. Torra, E. Cesari, J.M. Guilemany, J.R. Miguel, Scr. Metall. 22 [49] T. Simon, A. Kröger, C. Somsen, A. Dlouhy, G. Eggeler, Acta Mater. 58 (2010)
(1988) 261–264. 1850–1860.
[20] L. Contardo, G. Guénin, Acta Metall. Mater. 38 (1990) 1267–1272. [50] S. Qiu, V.B. Krishnan, S.A. Padula, R.D. Noebe, D.W. Brown, B. Clausen,
[21] J. Perkins, Scr. Metall. 8 (1974) 1469–1476. R. Vaidyanathan, Appl. Phys. Lett. 95 (2009) 141906–141906-3.
[22] K. Escher, Metal. Wiss. Tech. 44 (1990) 23–28. [51] S. Qiu, B. Clausen, S.A. Padula, R.D. Noebe, R. Vaidyanathan, Acta Mater. 59
[23] D.A. Hebda, S.R. White, Smart Mater. Struct. 4 (1995) 298–304. (2011) 5055–5066.
[24] R. Lahoz, L. Gracia-Villa, J.A. Puertolas, J. Eng. Mater. Technol.-ASME. 124 [52] S. Manchiraju, D. Gaydosh, O. Benafan, R.D. Noebe, R. Vaidyanathan,
(2002) 397–401. P.M. Anderson, Acta Mater. 59 (2011) 5238–5249.
[25] R. Lahoz, J.A. Puértolas, J. Alloys Compd. 381 (2004) 130–136. [53] S.A. Padula, D. Gaydosh, R.D. Noebe, G. Bigelow, A. Garg, D.C. Lagoudas,
[26] Y. Liu, Y. Liu, J. Van Humbeeck, Acta Mater. 47 (1998) 199–209. I. Karaman, K.C. Atli, in: M.J. Dapino, Z. Ounaies (Eds.), Proceedings of SPIE:
[27] Y. Liu, P.G. McCormick, Scr. Metall. 22 (1988) 1327–1330. Behavior and Mechanics of Multifunctional and Composite Materials, 2008,
[28] Y. Liu, P.G. McCormick, Acta Metall. Mater. 38 (1990) 1321–1326. pp. 692912–692912-11.
[29] H.Y. Luo, E.W. Abel, Smart Mater. Struct. 16 (2007) 2543–2549. [54] S. Qiu, PhD Thesis, University of Central Florida, 2011.
[30] M. Nishida, T. Honma, Scr. Metall. 18 (1984) 1293–1298. [55] R.D. Noebe, D. Gaydosh, S.A. Padula, A. Garg, T. Biles, M. Nathal, in:
[31] E. Quandt, C. Halene, H. Holleck, K. Feit, M. Kohl, P. Schlomacher, A. Skokan, W.D. Armstrong (Ed.), Proceedings of SPIE: Behavior and Mechanics of
K.D. Skrobanck, Sens. Actuators, A 53 (1996) 434–439. Multifunctional and Composite Materials, 2005, pp. 364–375.
[32] H. Scherngell, A.C. Kneissl, Scr. Mater. 39 (1998) 205–212. [56] R.D. Noebe, S. Draper, D. Gaydosh, A. Garg, B. Lerch, N. Penney, G. Bigelow,
[33] H. Scherngell, A.C. Kneissl, Acta Mater. 50 (2002) 327–341. S.A. Padula, in: B. Berg, M.R. Mitchell, J. Proft (Eds.), Proceedings of the
[34] K. Wada, Y. Liu, J. Alloys Compd. 449 (2008) 125–128. International Conference on Shape Memory and Superelastic Technologies,
[35] J.J. Wang, T. Omori, Y. Sutou, R. Kainuma, K. Ishida, Scr. Mater. 52 (2005) 2006, pp. 409–426.
311–316. [57] S.A. Padula, G. Bigelow, R.D. Noebe, D. Gaydosh, A. Garg, in: B. Berg,
[36] L.M. Wang, Y.F. Zheng, W. Cai, X.L. Meng, L.C. Zhao, J. Mater. Sci. Technol. 17 M.R. Mitchell, J. Proft (Eds.), Proceedings of the International Conference on
(2001) 263–266. Shape Memory and Superelastic Technologies, 2007, pp. 787–802.
[37] Z. Wang, X. Zu, X. Feng, J. Dai, Mater. Lett. 54 (2002) 55–61. [58] J. Cui, Y.S. Chu, O.O. Famodu, Y. Furuya, J. Hattrick-Simpers, R.D. James,
[38] Z.G. Wang, X.T. Zu, J.Y. Dai, P. Fu, X.D. Feng, Mater. Lett. 57 (2003) 1501–1507. A. Ludwig, S. Thienhaus, M. Wuttig, Z.Y. Zhang, I. Takeuchi, Nat. Mater. 5
[39] L.P. Chen, N.C. Si, J. Alloys Compd. 448 (2008) 219–222. (2006) 286–290.
[40] A.V. Shelyakov, Y.A. Bykovsky, N.M. Matveeva, Y.K. Kovneristy, J. Phys. IV 5 [59] K.F. Hane, T.W. Shield, Acta Mater. 47 (1999) 2603–2617.
(1995) 713–716. [60] S.A. Padula, R. Vaidyanathan, D. Gaydosh, S. Qiu, R.D. Noebe, G. Bigelow,
[41] Z.G. Wang, X.T. Zu, X.D. Feng, S. Zhu, J.Y. Dai, L.B. Lin, L.M. Wang, Mater. Lett. A. Garg, Metall. Mater. Trans. 43A (2012) 4610–4621.
56 (2002) 284–288. [61] A. Pelton, J. Mater. Eng. Perform. 20 (2011) 613–617.
[42] X.L. Meng, F. Chen, W. Cai, L.M. Wang, L.C. Zhao, Mater. Trans. 47 (2006) [62] K. Gall, M.L. Dunn, Y. Liu, P. Labossiere, H. Sehitoglu, Y. Chumlyakov, J. Eng.
724–727. Mater. Technol.- ASME. 124 (2002) 238–245.
[43] L.M. Wang, X.L. Meng, W. Cai, L.C. Zhao, J. Mater. Sci. Technol. 17 (2001) [63] G. Kang, Q. Kan, L. Qian, Y. Liu, Mech. Mater. 41 (2009) 139–153.
13–14. [64] H. Sehitoglu, R. Anderson, I. Karaman, K. Gall, Y. Chumlyakov, Mater. Sci. Eng.,
[44] X.L. Meng, Y.F. Zheng, W. Cai, L.C. Zhao, J. Alloys Compd. 372 (2004) 180–186. A 314 (2001) 67–74.
[45] X.L. Meng, W. Cai, Y.D. Fu, Q.F. Li, J.X. Zhang, L.C. Zhao, Intermetallics 16 [65] X. Wang, B. Xu, Z. Yue, J. Alloys Compd. 463 (2008) 417–422.
(2008) 698–705. [66] P.K. Kumar, U. Desai, J.A. Monroe, D.C. Lagoudas, I. Karaman, G. Bigelow,
[46] K.C. Atli, I. Karaman, R.D. Noebe, Scr. Mater. 65 (2011) 903–906. R.D. Noebe, Mater. Sci. Eng., A 530 (2011) 117–127.
[47] D.M. Norfleet, P.M. Sarosi, S. Manchiraju, M.F.X. Wagner, M.D. Uchic, [67] W.J. Moberly, J.L. Proft, T.W. Duerig, R. Sinclair, Acta Metall. Mater. 38 (1990)
P.M. Anderson, M.J. Mills, Acta Mater. 57 (2009) 3549–3561. 2601–2612.

Vous aimerez peut-être aussi