Materials Science & Engineering A 561 (2013) 321328

Contents lists available at SciVerse ScienceDirect

Materials Science & Engineering A

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

Inuence of short heat pulses on properties of martensite in medium

carbon steels
n, Birger Karlsson 2
Krste Cvetkovski 1, Johan Ahlstrom
Chalmers University of Technology, Department of Materials and Manufacturing Technology, SE-41296 Gothenburg, Sweden

a r t i c l e i n f o

abstract

Article history:
Received 10 August 2012
Received in revised form
20 September 2012
Accepted 3 October 2012
Available online 9 October 2012

The process of tempering a martensitic medium carbon steel was investigated with the aim to study
resulting material properties. The experimental results were used to model residual stresses caused by
local heating, with the nite element method. Tempering was followed for times from 0.1 s up to 1 h by
using laser heating and conventional salt bath furnace treatments within the temperature interval
500700 1C. In addition, the thermal expansion was evaluated using dilatometry. Experiments showed
that the initial stages of martensite decomposition, associated with loss of crystal tetragonality, proceed
almost instantly. An initial large decrease of hardness within the rst tenth of a second of the tempering
process was measured, followed by only limited further softening with increased tempering time. Thus
for the current material the tempering time had limited inuence on hardness, governed primarily by
the peak temperature during the heating process. Finite element modelling of rapid local heating and
cooling showed that the tempering behaviour and associated dilatation effects yield a peak temperature dependent residual stress eld with a broad tensile stress distribution for the case of un-tempered
martensite. However, for tempered martensite the residual stress eld depends primarily on the
heating rate and peak temperature and shows large gradients with tensile stresses in the surface and
compressive below. Thereby, for both cases, residual stresses were obtained but with completely
different residual stress gradients.
& 2012 Elsevier B.V. All rights reserved.

Keywords:
Martensite tempering
FE
Flash heating
Dilatometry
Laser heat treatment
Residual Stresses

1. Introduction
Pearlitic steels are commonly used for a wide range of
applications due to their strengthwear properties and cost
efciency. In industrial processes like laser or electron beam
welding, laser ash hardening, adiabatic shearing and in operating environments such as wheelrail contact in railway trafc, the
material is sometimes subjected to rapid heating and cooling
within a localised volume [16]. Such local heating often has a
short duration and is normally followed by self-cooling with large
cooling rates. If the heating pulses cause peak temperatures above
some 400600 1C, softening of the pearlitic microstructure takes
place resulting in easier plastic deformation and distortion under
loading [7]. Even more important are heating cycles extending
over the austenitisation temperature (TA1) which may lead to
martensite formation during rapid cooling. The volumetric

Corresponding author. Tel.: 46 31 772 1532; fax: 46 31 772 1313.

E-mail addresses: krste.cvetkovski@chalmers.se (K. Cvetkovski),

johan.ahlstrom@chalmers.se (J. Ahlstrom),

birger.karlsson@chalmers.se (B. Karlsson).
1
Tel.: 46 31 772 1244.
2
Tel.: 46 31 772 1242.
0921-5093/$ - see front matter & 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.msea.2012.10.005

increase caused by martensite formation leads to stress elds

and may cause local cracking. Subsequent re-heating in the
affected area above TA1 allows for new formation of martensite,
but, for lower temperatures tempering occurs with associated
decrease of volume. A key issue is the speed of martensite
softening and how internal stresses develop during tempering
in these shock heating and cooling sequences. As the kinetics of
martensite tempering is not well known for such short heating
and cooling histories the present study was undertaken. Methods
used to predict the material properties after tempering, for
example the HollomonJaffe parameter are suitable for long
tempering times when appropriate model calibration is performed [812]. However for tempering of seconds or less, it
may not be valid to extrapolate material properties from long
time tempering experimental results [13].
In this study, laser- and salt bath heating have been used to
study martensite tempering and resulting mechanical properties,
for time periods from a tenth of a second up to 1 h. Dilatometric
experiments have also been made to determine associated phase
transformations. Finally the effects of rapid heating and cooling of
martensite and tempered martensite have been analysed with
nite element (FE) simulation to evaluate residual stresses within
a component, for the case of local heating.

322

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

Table 1
Chemical composition of material in wt%.
C

Mn

Si

Cr

Mo

Cu

0.50

0.75

0.37

0.12

o0.08

o0.06

o0.11

2. Experimental
Ferriticpearlitic steel with some 90 vol% ne pearlite and
10 vol% proeutectoid ferrite was used as a starting material. The
nominal chemical composition of the steel is presented in Table 1.
Martensitic material was obtained by austenitisation at 900 1C in
a tube furnace with protective argon atmosphere for 5 min and
thereafter quenched in brine. The Vickers hardness after this
treatment was 780710 HV30.
Samples for laser experiments were coin-like with a thickness
of 3.0 mm and diameter of 22.5 mm. To attain rapid heating and
cooling of the material, a 2.8 kW diode laser was used for heating
a spot on the sample surface. To have uniform laser radiation
absorption between samples, the radiated surface was ground
and then polished with 3 mm diamond suspension. The laser
beam was guided via an optical bre and defocused to a diameter
of 6 mm at the centre of the coin surface. The temperature within
the innermost 2 mm diameter of the laser exposed area was
measured with a pyrometer, using a sampling frequency of
60 kHz. The samples were mounted in a temperature controlled
copper sample holder within a vacuum chamber, evacuated down
to 10  8 bar during the experiments.
Tempering treatments for times longer than 10 s were made in
a salt bath furnace using sodium nitrate for tempering at 500 1C
and barium chloride salt for higher temperatures. Samples were
thin rectangular plates with 2 mm thickness. A type S thermocouple inserted within a reference sample was used to obtain the
actual sample temperature. Hardness indentation was made on a
cross-section to avoid inuence of thin decarburised layers on the
surface.
For laser tempered samples, hardness (HV5) was measured
along two orthogonal lines on the surface both crossing the centre
of the irradiated area. On a perpendicular section containing one
of the surface lines with surface hardness indents, hardness
mapping was made with a distance between indents of 0.5 mm.
Retained austenite in martensite was measured with X-ray
diffraction technique using an Xstress 3000 X-ray equipment
with vanadium Kb blocking lters.
Dilatometry was performed for martensite and pearlite by means
of a push-rod dilatometer, Netzsch Dil 402 C, with a tungsten
thermocouple in protective atmosphere. Thermal expansion and
solid state transformations of as-quenched and tempered material
were studied using single and multi-runs. Analyses were made for
different maximum temperatures up to 990 1C, all starting at 30 1C
(isotherm, 30 min before each run). Heating and cooling rates were
5 1C/min and 10 1C/min respectively. Alumina (Al2O3) was used as a
reference. Samples were cylindrical with diameter 6 mm and length
50 mm.

3. Finite element model

3.1. Software and denitions
Simulations were performed with the nite element (FE) method
in the commercially available software Abaqus 6.9 Standard by
Dassault Syste mes. The software has the ability to link user-dened
subroutines into the compilation. Three such subroutines were
used in the current work: UEXPAN for implementation of phase

Fig. 1. Mesh and denition of directions.

dependent thermal expansion, DFLUX to model the heat distribution

from the laser source and HETVAL to model martensite tempering
and to keep track of phase contents [14]. An axisymmetric mesh
with 3000 elements was used, see Fig. 1; the element size close to
the heated surface was 0.025 mm thick and 0.075 mm wide.
Coupled temperaturedisplacement analyses were run with fournode bilinear displacement and temperature elements (CAX4T).
Each of the 3111 nodes has three degrees of freedom: temperature
and displacement in radial and depth directions.

4. Analysis
4.1. Temperature control and thermal properties
The initial sample temperature was set to 20 1C. On mounting
into the sample holder the sample was assumed to be heated to the
temperature of the holder, which in accordance with the experimental setup was set to 100 1C. The laser heating of the most central
part of the surface was done with different powers for different
cases, and adjusted to achieve a top-hat temperaturetime curve in
conformity with the experiments. It was found in the experiments
that the peak temperatures dominate the following material behaviour by controlling the degree of tempering. Initial cooling was
modelled by lm cooling to the sample holder; however, the hottest
parts of the specimen were mainly self-cooled. As the material
models are formulated as rate-independent, the exact cooling curve
is not required when temperature gradients are low. Mechanical
interaction between the sample and the sample holder was
neglected. Final cooling to 20 1C was accomplished by setting a
temperature boundary condition to the nodes. A nal temperature
of 20 1C was required to allow comparisons with experimental
hardness data.
The power from the laser was applied on the top surface at
0rr r3.0 mm with scaling of absorption coefcient to achieve a
narrow temperature range during the time of exposure within the
pyrometer sampling area, 0 rr r1.0 mm. Temperature dependent specic heat and thermal conductivity were implemented
for the actual chemical composition of the coin sample based on
data from Thermo-Calc (Thermo-Calc software), reported in earlier publications [15].
4.2. Mechanical properties and dilatation
The elastic response of the material was implemented with
slight temperature dependence of the Youngs modulus E,
decreasing from 210 GPa at 20 1C down to 150 GPa at 700 1C
according to data from JMatPro6.0 (Sentesoftware) calculated for
the actual chemical composition. The plastic behaviour was
described using a material model with isotropic hardening predened in Abaqus. This model has the capability to account for
both temperature and solution dependent variables, in this case a
tempering parameter explained below. Underlying data were
compiled from own experiments (compression tests at temperatures from 25 to 275 1C), extrapolated to higher temperatures

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

using JMatPro6.0, see ow stress curves in Fig. 2. To handle the

effect of tempering on ow stress, a tempering parameter with
values corresponding to the room temperature Vickers hardness
was dened, HV. Thus, for each position the peak temperature
(Tpeak, in 1C) so far gave the degree of tempering according to the
formula: HV810 0.67Tpeak (rationalisation of experimental
results is presented in Fig. 3). Flow stresses were interpolated
from data tables containing stressstrain curves at certain temperatures and tempering parameters. A local decrease of

323

tempering parameter value was assumed to give a proportional

decrease of local ow stress. An initial hardness of 780 HV was set
for the as-quenched (un-tempered) coins and 408 HV, corresponding to tempering at 600 1C, for the tempered coins. Initially,
the coins were assumed to be globally stress free.
The thermal expansion was modelled using the data from the
dilatometry experiments. The comparison between the actual
temperature and the peak temperature at each location gave
the microstructural dependence of the thermal expansion, i.e.
during the rst heating of un-tempered martensite the primary
heating curve was followed and during cooling and re-heating
below the peak temperature of that location, the secondary
heating curve representing tempered martensite was used. The
curves were implemented by three polynomials, valid in discrete
temperature ranges.
4.3. Computational strategy
The subdivision of the simulations in terms of computational
steps is outlined in Table 2 and the cases simulated are seen in
Table 3 below.

5. Results
5.1. Hardness and microstructural evaluation after tempering
Fig. 2. Flow stress curves for as-quenched martensite (QM) and tempered
martensite (TM) at different degrees of tempering, and at different temperatures.
Points connected with lines represent experimentally performed compression
tests. Room temperature (RT) ow stresses are assumed to be proportional to the
decrease in room temperature hardness. High temperature behaviour above
275 1C extrapolated with JMatPro6.0.

Fig. 3. Hardness after tempering. Minimum surface hardness measured on coin

samples is presented.

Hardness after laser and salt bath temperings is presented in

Fig. 3. It can be observed that small differences in hardness
between the two methods were measured after the tempering.
The difference is believed to originate from problems to assess
and calibrate an accurate temperature pyrometer reading. Irrespective of the uncertainties of absolute temperature in the laser
trials, the result clearly indicates that initial martensite tempering
occurs fast. Tempering at 600 1C for 0.110 s by both laser and
salt bath treatments gave close to 50% decrease of hardness. After
1 h tempering at the same temperature, the hardness decreased
to about 40% of the original.
Hardness proles made on the cross-section of test samples
after rapid laser tempering are presented in Fig. 4. Surface
hardness values are seen as downward pointing triangles on the
upper surfaces. Increased heating time or temperature gave a
larger tempered volume. The tempered volume had a typical half
elliptic form until about 2 s. For longer heating time heat losses to
the sample holder give a signicant inuence on the hardness
eld. Surface hardness from tests at the same temperature, but
different time, was rather similar. This indicates that the time
factor during rapid tempering has a limited inuence on the
hardness and primarily affects penetration depth.
In the rapidly tempered martensite, lath-like carbide agglomerates were observed; Fig. 5, cf. [1618]. Microstructural differences
resulting from secondary tempering effects such as coarsening of
laths or spheroidised cementite were not observed, and neither
were any microcracks visible in the coin surfaces. X-ray diffraction

Table 2
Subdivision of the computation into different computational steps (NBnot time steps, each step contains several time steps).
Computational
step

Boundary conditions and loads

What is simulated?

Step 1

Nodes along symmetry y axis xed in radial direction, bottom node also in y
direction. Retained for all following steps. Temperature of all nodes
increased to 100 1C
Heat absorption on central part of top surface; otherwise insulated surfaces
Heat conduction to sample holder on bottom surface
Temperature of all nodes decreased to 20 1C

Insertion of sample into sample holder and temperature stabilisation at

100 1C. Thermal expansion of entire coin from 20 to 100 1C (no stresses
generated)
Laser heating, thermal expansion and arising stress elds
Cooling, thermal contraction and arising stress elds
Thermal contraction and arising stress elds on cooling down to 20 1C

Step 2
Step 3
Step 4

324

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

crystallographic measurement made after tempering veried lattice

transformation from bct to bcc which has taken place.

5.2. Dilatometry
Two main tempering stages were observed for the martensite
(see Fig. 6), visible as dilatation effects on the linear thermal
expansion curve (DL/L0). The linear thermal expansion differentiated with respect to temperature (al al(T)) is also included. The
negative dilatation effect (starting with a negative slope of al)
with a peak value around 160 1C is associated with carbon
segregation and precipitation of transition carbides, presumably
e carbide. A second tempering stage ranging from 220 to 400 1C
was observed in the form of second negative dilatation effect,
indicating further decomposition into ferrite and cementite
[16,1921]. The small positive dilatation effect around 230 1C
may be related to transformation of retained austenite [20,22].
Yet, this effect is minor since present XRD measurements indicated a retained austenite content of less than 1.5%.
The behaviour for re-tempering of martensite was also studied
to obtain an understanding of how tempering and associated
volumetric shrinkage proceed as a function of the peak temperature for multiple heat cycles. One such double run is presented in
Fig. 6b, where the total thermal cycle consisted of a rst heat run
to 200 1C with subsequent cooling to 30 1C, and a second run to
570 1C with subsequent cooling. After the initial heat run no
dilatation effects were apparent in the subsequent run-up to the
peak temperature of the initial run, while above the peak
temperature of the initial run, tempering advanced accordingly.
For a multi-run with tempering to 570 1C in both cycles, the
length decrease was 0.22% after the rst run and practically none
for the consecutive.
Table 3
Different cases simulated. Approximate peak temperature (Tpeak) and time spent
within some 30 1C from Tpeak are given.
Case

Heating rate

Tpeak (1C)

Time @ Tpeak (s)

Starting microstructure

M1
M2
M3
M4
M5
TM3
TM4
TM5

High
High
High
Low
Very high
High
Low
Very high

225
400
575
575
575
575
575
575

2
2
2
4
0.1
2
4
0.1

Un-tempered M, 780 HV
Un-tempered M, 780 HV
Un-tempered M, 780 HV
Un-tempered M, 780 HV
Un-tempered M, 780 HV
Tempered (600 1C), 408 HV
Tempered (600 1C), 408 HV
Tempered (600 1C), 408 HV

5.3. FE simulations
The inuence of peak temperature on the residual stresses
of un-tempered martensite was studied via cases M1M3
(cf. Table 3), and the inuence of heating rate and time at peak
temperature was examined in M4M5. For tempered martensite,
only the inuence of heating rate and time at peak temperature
(combined) was studied in TM3TM5, as heating at lower
temperatures did not generate residual stresses, as will be
discussed later. Temperature curves for the centre node on the
top surface for all cases are shown in Fig. 7.
The radial stress for case M5, the case with the highest
temperature gradients is shown in Fig. 8. The stress distribution
just before cooling (Tpeak E575 1C) is shown in Fig. 8a; the peak
stress is approximately  550 MPa. The von Mises stress has
approximately the same magnitude and is not sufcient to
plastically deform the material; this is valid in every location
and every time step for all of the cases M1M5. The radial stress
distribution for M3 after nal cooling to room temperature is
shown in Fig. 8b.
The radial stresses for case TM3 are shown in Fig. 9a. The
stress at the centre reaches 240 MPa, and rapidly declines;
predominantly the volume which was plastically deformed during heating shows large tensile stresses (Fig. 9b).
Depth proles of radial stresses for all cases along the centre
line after nal cooling are shown in Fig. 10. For case M1 stresses
are small, below 25 MPa. Case M2 shows a maximum radial stress
of around 100 MPa, decreasing towards the bottom surface, while
case M3 has a more even distribution of residual stresses, as most
of the dilatation differences have been exploited through the
entire sample thickness (the peak temperature at the centre line
towards the bottom of the sample is around 370 1C). Case M4 is
more extreme in this respect and stress levels are lower as a
larger part of the coin has decreased its volume (heating the
entire coin to the same temperature would give only shrinkage,
no residual stresses). M5, on the other hand, shows that the larger
temperature gradients achieved with rapid heating give correspondingly larger gradients in residual stresses. The fact that all
cases M1M5 show somewhat lower stresses at the surface is
believed to be due to the increasing degree of plane stress
condition naturally prevailing on a free surface. As seen from
the stress elds, the stresses were below the yield stress for all
locations in M1M5 (see Fig. 2); thus the strains are caused by
volume decrease upon tempering of martensite and to some
degree of warping of the sample. The most obvious difference
for the tempered materials (TM3TM5) is the much shallower

Fig. 4. Hardness maps for cross sections of tempered samples. Downward facing triangles represent surface hardness as measured before sectioning, while each diamondshaped mark represents an indent on the cross-section. Hardness scale bar applicable to all graphs. (a) 600 1C; 0.1 s, (b) 600 1C; 0.4 s, (c) 600 1C; 4.9 s, (d) 700 1C; 0.4 s and
(e) 500 1C; 0.4 s.

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

325

Fig. 5. Scanning electron micrographs (in-lens secondary electron detector). (a) Comparison of initial martensitic microstructure and (b) tempered at 700 1C for 0.1 s.
Carbide precipitates are brightly imaged in the tempered martensite.

Fig. 7. Temperaturetime curves for nodes on the heated surface. Time for heating
from room temperature to 100 1C (temperature of sample holder) and the nal
abrupt cooling to room temperature is arbitrary (needed for computational
reasons).

thermal strain gradient is needed, while lower heating rates will

render less or no residual stresses.

6. Discussion
6.1. Martensite tempering

Fig. 6. Thermal expansion for multi-pass dilatometer runs (cooling in-between

runs not showed in graphs). (a) Two runs both to 570 1C; after the initial run no
further dilatation effects are measured. (b) First run to 250 1C and second run to
570 1C, in the second run tempering progress as normal above the peak temperature of the rst run.

residual stress prole. For these cases, no dilatation effects affect

the solution. Thus, all residual stresses developing are due to the
plastic deformations caused by the thermal strain with respect to
the ow stresses. Plastic deformation in compression occurs in
the heated volume at high temperature, giving residual tensile
stresses after cooling. However, to reach the yield stress a certain

The rapid tempering experiments showed that extensive

martensite decomposition occurs within very short time for the
type of steel studied, leading to instantaneous softening within
parts of a second (Fig. 3). Further softening proceeds at a much
lower rate. It is evident that extrapolation from long tempering
times in the HVlog t diagram (Fig. 3) gives rather accurate
predictions of hardness down to times as low as at least 0.1 s
for each of the temperatures investigated. The deviation in the
extrapolation of the hardness at the highest temperature 700 1C is
judged to emanate from a slight underestimate of the temperature in the pyrometer temperature recordings. The slight parabolic shape of the curves based on the more precise data from the
salt bath experiments together with some experimental uncertainties in the temperature recordings in the laser experiments
would indicate that the tempering time has a very minute
inuence on softening for times below parts of a minute or so.
The softening rate for times smaller than 0.1 s is, however, not
possible to evaluate from the present data.
Dilatation effects on the al curve for re-tempered martensite were
not detectable. Comparison of tempered martensite (Tpeak 575 1C)

326

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

Fig. 8. Radial stresses overlaid on the deformed mesh; deformations magnied 50  . (a) Case M5 before cooling. The peak stress is not sufcient to plastically deform the
material. (b) Case M3 after cooling to room temperature. Shrinkage due to local martensite tempering is clearly seen.

Fig. 9. (a) Radial stresses (MPa), and (b) accumulated plastic strain (  ) of case TM3 after cooling to room temperature. Deformations magnied 50  .

Fig. 10. Radial stresses after cooling to 20 1C for all cases in Table 3.

with the pearlitic (non-quenched) base material, Fig. 11, showed

the difference in al to be insignicant at lower temperatures. Above
ca. 400 1C a slight deviation is visible, which may be related to
spheroidisation of pearlite [7]. Similar thermal expansion during
heating for pearlite and tempered martensite would be expected
since the phase constituents are similar and morphology effects are
small. Consequently, to estimate material properties for martensite

Fig. 11. Comparison of thermal expansion for tempered martensite and pearlite;
the behaviour is similar.

tempering at short heating times, accurate prediction of the peak

temperature is more crucial than time history.
By SEM examination it is possible to resolve only larger changes in
the microstructure and thus martensite lath structure appeared to be
unaffected after rapid tempering. For rapid tempering processes,

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

substitution elements are anticipated to have small impact on the

nal hardness due to limited diffusion time, i.e. low atomic mobility
[18,23]. Malik and Lund [23] for a similar medium carbon steel
presented the difference in carbide spacing, increasing from 610 A at
250 1C to 1010 A at 550 1C after 1 h of tempering. Mean distance
between carbides principally determines their hardness contribution
to the material [23,24]. The carbide coarsening rate is probably very
low compared to the initial rapid softening during tempering. Based
on the current result, carbon diffusion rate out from the bct lattice is
anticipated to be high. Dislocation densities in martensite may be as
high as 1017 m  2 [23,25,26], enabling pipe diffusion of atoms in
addition to volume diffusion. Iron diffusion can be a rate limiting
factor for carbide precipitation particularly during growth of ecarbide. Motion of iron atoms has been suggested to take place by
pipe diffusion as volume diffusion solely would not be sufcient at
low temperature [2730]; this would help to explain why carbides
are able to form rapidly upon tempering. An important implication
when pulse heating a small volume is the creation of high temperature gradients with resulting strains, which can accelerate the
diffusion rate of matrix atoms and carbides [31].
Nevertheless, the large and similar hardness decrease for long and
short time temperings above 500 1C shows that extensive cementite
precipitation is present nearly instantly when heating, as a function
of temperature. As the main contribution to hardness decrease during
the tempering process for long times in low alloyed martensitic steels
can be attributed to carbide coarsening, the initial tempering stage is
assumed to be related to carbon rejection from the supersaturated
martensite lattice taking place within fraction of a second.

327

an even larger gradient in thermal strain due to the low thermal

expansion. Thus a combination of the two, principally different,
processes can occur for extremely rapid heating to high temperatures.
An isotropic plasticity model is not optimal for medium carbon
steel, but proved sufcient for evaluating the principal behaviour
for the cases studied. In the as-quenched martensite, no plastic
deformation occurred and thus only the elastic part of the
mechanical material model was utilised. For the tempered martensite, a lower ow stress prevailed already from the start, and it
was further lowered by the temperature dependence. For the
current cases, the plastic deformation was limited, leading to
rather low stresses after cooling compared to the ow stress. In
cases where the ow stress in tension (after deformation in
compression) is exceeded, an accurate prediction of residual
stresses would require a kinematic hardening model to be used.

7. Conclusions
Tempering characteristics of a medium carbon steel in martensitic condition were examined for tempering times ranging
from 0.1 s to 1 h. Resulting material properties were studied
primarily by the indentation technique, and characterisation of
thermal expansion and solid state transformation were made by
dilatometry. Using the experimental results, residual stresses
arising on rapid local heating with subsequent cooling to room
temperature have been evaluated using the nite element analysis for as-quenched and tempered martensite. The following
conclusions were drawn:

6.2. Residual stresses

The FE analysis showed that residual stresses are attainable by
local heating of both as-quenched and tempered martensite, but
governed by different mechanisms. The development of residual
stresses depends on heating rate, peak temperature and mainly
two material properties, thermal expansion/contraction and yield
strength. For the case of as-quenched martensite (cases M1M5),
the low thermal expansion and high yield strength caused no
plastic deformation to take place during the heating stage for all
studied cases. However, residual stresses were instead obtained
by local volume decrease. The amount of shrinkage is a function
of the peak temperature, which is analogous to reaching higher on
the thermal expansion curve in Fig. 6. Heating rate and time at
peak temperature have no profound effects; these factors affected
only the temperature eld and affected volume, in accordance
with the experimental data (Fig. 4). As long as the heated volume
is surrounded by non-affected material, i.e. with a xed boundary,
essentially tensile residual stresses are present. Therefore, after
local heating in a larger component, residual stresses may always
be expected for as-quenched martensite. In contrast, the magnitude of residual stresses in pre-tempered martensite (cases TM3
TM5) shows a strong dependence on heating rate and peak
temperature. While there is no shrinkage effect after heating,
the thermal expansion is larger for the tempered martensite and
yield strength is lower. For high heating rates, plastic deformation
in compression of the heated clamped volume results in residual
stresses after cooling, with large stress gradients. Tensile stresses
are present in a thin surface layer and compressive below (see
Fig. 10). There is a distinct limit in strain gradient below which no
residual stresses will form. The reasoning regarding residual stress
formation in tempered martensite is valid also for local heating of
pearlitic material. With its generally lower ow strength compared
to tempered martensite and almost identical thermal expansion,
pearlite is even more prone to residual stress formation by plastic
deformation of the surface. Also for as-quenched martensite, plastic
deformation of the locally heated volume may occur, but it requires

1) Experiments with short time laser tempering of martensite

showed that the initial stages of martensite decomposition,
associated with loss of crystal tetragonality, proceed almost
instantly. An initial large drop in hardness within the rst
tenth of a second of the tempering process is followed by
limited further softening for the times and temperature
studied. Thus for the current steel the tempering time has
limited inuence on hardness, governed primarily by the peak
temperature during the heating process.
2) Partial tempering to an intermediate temperature, cooling, and
tempering to a higher temperature showed that the dilatation
effects occur only the rst time at each temperature. Below the
peak temperature of a previous tempering run, the material
exhibits similar thermal expansion as that of the material in its
pearlitic state.
3) Finite element modelling showed residual stresses to develop
in as-quenched martensite solely by the effect of volume
decrease associated with tempering after local heating, independent of heating rate. Thus, any local temperature increase
in the material gives residual stress after cooling. For the case
of tempered martensite, the residual stress eld instead
depends on the heating rate and peak temperature, giving
sufciently large thermal strain gradients to cause local plastic
deformation. This results in a large stress gradient, with tensile
stress in the surface and compressive below. This is in contrast
to as-quenched martensite which exhibits a broad tensile residual
stress distribution around the whole tempered volume.

Acknowledgement
This investigation has been performed within the National
Centre of Excellence CHARMEC (Chalmers Railway Mechanics).
Authors are grateful for cooperation with Materials Center Leoben
regarding laser tempering experiments. Peter Krahl and Alexandre Gorski are acknowledged for help with experimental work.

328

K. Cvetkovski et al. / Materials Science & Engineering A 561 (2013) 321328

References
[1] K.A. Ryttberg, Materials and Manufacturing Technology, Chalmers University

of Technology, Goteborg,
2008.
[2] S. Moeinifar, A.H. Kokabi, H.R. Madaah Hosseini, Mater. Des. 31 (2010)
29482955.
[3] A. Zambon, F. Bonollo, Mater. Sci. Eng. 178 (1994) 203207.
[4] C. Yao, B. Xu, J. Huang, P. Zhang, Y. Wu, Opt. Lasers Eng. 48 (2010) 2026.
[5] S. Niederhauser, B. Karlsson, P. Sotkovszki, Z. Metallkd. 96 (2005) 370376.

[6] J. Ahlstrom,
B. Karlsson, Wear 232 (1999) 114.

[7] K. Cvetkovski, J. Ahlstrom,

B. Karlsson, Mater. Sci. Technol. 27 (2011)
648654.
[8] M.J. Balart, C.L. Davis, M. Strangwood, Mater. Sci. Eng. A 328 (2002) 4857.
[9] S. Murphy, J.H. Woodhead, Metall. Trans. 3 (1972) 727735.
[10] Z. Janjusevic, Z. Gulisija, M. Mihailovic, A. Pataric, Chem. Ind. Chem. Eng. Q. 15
(2009) 131136.
[11] T. Reti, M. Gergely, P. Tardy, Mater. Sci. Technol. 3 (1987) 365371.
[12] J.W. Elmer, T.A. Palmer, W. Zhang, B. Wood, T. DebRoy, Acta. Mater. 51 (2003)
33333349.
[13] T. Furuhara, K. Kobayashi, T. Maki, ISIJ Int. 44 (2004) 19371944.
[14] Dassault Syste mes, ABAQUS/Standard Documentation Collection, Version 6.9,
2009.

[15] J. Ahlstrom,
B. Karlsson, Wear 253 (2002) 291300.

[16] D.A. Porter, K.E. Easterling, Phase transformations in metals and alloys,
second ed., Stanley Thornes Ltd, Cheltenham, 2001.
[17] V.H. Baltazar Hernandez, S.S. Nayak, Y. Zhou, Metall. Mater. Trans. A 42
(2011) 31153129.
[18] X. Sauvage, X. Quelennec, J.J. Malandain, P. Pareige, Scr. Mater. 54 (2006)
10991103.
[19] M. Jung, S.J. Lee, Y.K. Lee, Metall. Mater. Trans. A 40 (2009) 551559.
[20] J. Pacyna, J. Achiev. Mater. Manuf. Eng. 46 (2011).
[21] G.R. Speich, W.C. Leslie, Metall. Trans. 3 (1972) 10431054.
[22] G.E. Totten, M.A.H. Howes, in: Marcel Dekker (Ed.), Steel Heat Treatment
Handbook, second ed., 1997.
[23] L. Malik, J. Lund, Metall. Mater. Trans. B 3 (1972) 14031406.
[24] R. Grange, C. Hribal, L. Porter, Metall. Mater. Trans. A 8 (1977) 17751785.
[25] G. Krauss, Mater. Sci. Eng. 273275 (1999) 4057.
[26] G. Krauss, Metall. Mater. Trans. B 32 (2001) 205221.

[27] P. Morra, A. Bottger,

E. Mittemeijer, J. Therm. Anal. Calorim. 64 (2001)
905914.
[28] E. Mittemeher, L. Cheng, P. van der Schaaf, C. Brakman, B. Korevaar, Metall.
Mater. Trans. A 19 (1988) 925932.
[29] L. Cheng, C. Brakman, B. Korevaar, E. Mittemeijer, Metall. Mater. Trans. A 19
(1988) 24152426.
[30] A. Sherman, G. Eldis, M. Cohen, Metall. Mater. Trans. A 14 (1983) 9951005.
[31] G. Eggeler, Acta Metall. 37 (1989) 32253234.