You are on page 1of 12

Computational Materials Science 113 (2016) 280291

Contents lists available at ScienceDirect

Computational Materials Science


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

Multiscale modeling of tempering of AISI H13 hot-work tool


steel Part 1: Prediction of microstructure evolution and
coupling with mechanical properties
A. Eser a, C. Broeckmann a, C. Simsir b,
a
b

Institute for Materials Applications in Mechanical Engineering (IWM), RWTH Aachen University, Augustinerbach 4, 52062 Aachen, Germany
_
Metal Forming Center of Excellence (MFGE), Atlm University, Kzlcasar Mah., 06836 Incek-Glbas
, Ankara, Turkey

a r t i c l e

i n f o

Article history:
Received 6 September 2015
Received in revised form 10 November 2015
Accepted 14 November 2015
Available online 2 December 2015
Keywords:
Tempering
Multiscale modeling
Precipitation simulation
Microstructureproperty relationships
AISI H13

a b s t r a c t
In the first part of this two part study, the mechanical properties necessary for the simulation of tempering of an AISI H13 (DIN 1.2344, X40CrMoV5-1) tool steel was derived using physically based precipitation
simulations and microstructureproperty relationships. For this purpose, the precipitation of fine
carbides were simulated using a thermo-kinetic software which allows prediction of the evolution of precipitation/dissolution reactions and the particle sizes. Then, those microstructural findings were coupled
with physically based microstructureproperty models to predict the yield stress, flow curve and creep
properties. The predicted mechanical properties were verified with corresponding experiments and a
good agreement was found. In the second part of this study, those properties were coupled with a
Finite Element (FE) model in order to predict the relaxation of internal stresses and the evolution of deformations at the macroscopic scale.
2015 Elsevier B.V. All rights reserved.

1. Introduction
Hot-work tool steel AISI H13 (DIN 1.2344, X40CrMoV5-1) is
mainly used as tools for extrusion, forging dies and especially as
die-casting dies for the processing of light metals. AISI H13 has distinguished mechanical properties like high hardness, good wear
resistance, thermo-cyclic stability and high strength at elevated
temperatures which are essential for high temperature tool applications. These properties can be generated only by a controlled
heat treatment process which consists of an austenitization treatment with a subsequent hardening and a multiple tempering procedure. The superior mechanical properties are due to the
nanometer size alloy carbides which precipitate during the tempering. In order to predict the mechanical properties of AISI H13
during and after tempering, the development of those carbides
should be well known. The chemical composition of the investigated steel is given in Table 1.
There are several attempts in the literature which concentrate
on predicting the macroscopic material properties from the
microstructural parameters. Most of the previous work in the liter-

Corresponding author.
E-mail address: caner.simsir@atilim.edu.tr (C. Simsir).
http://dx.doi.org/10.1016/j.commatsci.2015.11.020
0927-0256/ 2015 Elsevier B.V. All rights reserved.

ature focused on the prediction of yield stress [15]. On the


other side there exist several studies about predicting the high
temperature creep behavior of 912% CrMoV-Steels [6,7]. All of
those models require microstructural parameters which should
be determined from highly costly experiments. Furthermore, most
of the previous attempts concentrated on one particular heat treatment state and did not simulate the evolution of the mechanical
properties during the heat treatment. Therefore the prediction
capability of those models for different microstructures is limited.
Moreover there is the commercial software JMatPro which simulates the material properties for different heat treatment conditions [8,9]. However, at best of authors knowledge, prediction of
the yield stress, flow and creep properties with this software during the tempering of steels is not possible.
In the present work, the previously mentioned singular efforts
were coupled together to determine the mechanical behavior of
an industrial steel during a heat treatment process. The precipitation kinetics were determined using the software MatCalc (Version
5.51.1011) [1012]. The results of the simulations namely size, volume fraction of carbides and the concentration of alloying elements in the matrix are used as input parameters for the
physically based material models to predict the mechanical
properties.

281

A. Eser et al. / Computational Materials Science 113 (2016) 280291

with rfe = intrinsic lattice strength of Fe (PeierlsNabarro Stress),


rss = solid solution hardening of matrix, rP = precipitation hardening, rdis = dislocation strengthening rsgb = subgrain strengthening.

Table 1
Chemical composition of steel AISI H13 (wt%).
C

Cr

Ni

Mo

Mn

Si

0.40

5.37

1.34

1.22

0.3

0.97

2.3.1. PeierlsNabarro stress


The temperature dependent
lowing expression [22]:

rfe can be estimated using the fol-

2. Modeling

rfe 78  0:023  T

2.1. Stress relaxation during tempering

where T is the absolute temperature in Kelvin.

Although this subject is also covered in Part-II of this article, it is


worth summarizing the stress relaxation mechanisms before starting to calculate the required material properties to simulate this
effect.
During quenching of tool steels, residual stresses are induced
due to non-homogeneous inelastic deformation due to thermal
gradients in the part. Those stresses relax during tempering either
by time-independent (yielding/time-independent elasto-plastic
deformation) or by time-dependent (creep/visco-elasto-plastic
deformation) deformation. The deformation occurs so that the geometric mismatch between the neighboring regions in the parts
decrease resulting in a decrease (relaxation) in residual stress level.
Relaxation (deformation) by yielding occurs instantaneously
when the local yield strength of the material is below the yield
strength of the material due to the increasing temperature. This
usually occurs during heating up to the tempering temperature.
However, the bigger extent of the relaxation (deformation) usually
occurs during holding at the tempering temperature by creep
mechanism since creep occurs as time goes on at any non-zero
stress level without the need of exceeding a yield strength. Moroever, the regions having larger stresses creep (relax) faster than the
regions with low stresses; which also enhance the relaxation
process.
In order to simulate the stress relaxation and deformations during the tempering, both elasto-plastic properties (yield strength,
flow curve) and creep properties must be determined either experimentally or computationally.

2.3.2. Solid Solution Hardening


According to Lacy rss , the strength contribution due to the
alloying elements can calculated as [23]:

2.2. Simulation of precipitations


The thermo-kinetic software package MatCalc is based on the
numerical KampmannWagner model [13,14] describing the precipitates in terms of size classes. The nucleation approach is based
on the classical nucleation theory [15,16]. The thermo-dynamic
calculations are considered by CALPHAD methods and CALPHAD
databases [17,18]. The program calculates chemical potential and
driving forces of all phases as well as the diffusion coefficient by
using chemical composition and temperature dependent mobility
database for all the elements. Based on these data, the development of the precipitate size and the composition of the precipitates
can be calculated. There are several studies in the literature for the
simulation of carbide development during the heat treatment of
hot-work tool steels and good agreement between the experimental findings and simulation results were reported [1921].
2.3. Prediction of yield stress
Several strengthening mechanisms contribute to the overall
strength of the steel AISI H13. The overall yield strength of a complex tempered martensite (ryield ) can be assumed to be a linear
superposition of each individual strengthening mechanism [13]:

ryield rfe rss rP rdis rsgb

rss

f
X
n
K iXi i

i1

with X i being the weight percent of alloying atoms dissolved in the


matrix, K i and ni are alloying element dependent constants
(Table 2). The temperature dependency of the solid solution hardening is assumed to be similar to the temperature dependency of
the shear modulus (Eq. (7)).
2.3.3. Precipitation hardening
In order to calculate the strengthening of the steel due to the
precipitations, the MatCalc internal model was used. The model
is based on the work of Holzer et al. [4,5]. and based on the
assumption that the coherent precipitates are shearable on the
contrary incoherent ones are not shearable by dislocations. Then
the strengthening due to precipitation hardening can be expressed
as:

rP

Ncoh
Ntotal

1=2 q 
1=2
N
r2chem r2coh r2mod incoh
ro
Ntotal

where Ncoh ; N incoh ; N total are the number densities of the coherent,
the incoherent and the sum of the number densities of precipitates,
respectively. The terms in the second square root are the strength
contributions due to chemical hardening effect (rchem ), the coherency strain effect (Misfit strengthening) (rcoh and the modulus
hardening effect (rmod ). The final term (ro ) in Eq. (4) is the strength
contribution of the impenetrable precipitates.
The chemical hardening term (rchem ) is described by the general
FriedelBrownHam equation [26]:

rchem

2M
bkC1=2

 cb

3=2

where M is the Taylor factor, C is the line tension of the dislocation,


b is the magnitude of the Burgers vector and k is the mean particle
spacing in the slip plane of the dislocation.
The so called misfit strengthening (rcoh ) due to the lattice
coherency of precipitates with the ferritic matrix is described
according to Brown and Ham [26]:

rcoh 8:4  MG  e3=2 

 1=2
N
 r2
b

where r represents the mean precipitate radius, G is the shear modulus, e is the linear misfit strain, and N is the corresponding number
density.
The modulus strengthening effect is due to the intersection of
dislocations with a spherical precipitate having a shear modulus
lower than the shear modulus of the matrix. The formulation based
on the work of Russell et al. [27]:

282

A. Eser et al. / Computational Materials Science 113 (2016) 280291

Table 2
Parameters K and n from Eq. (3) (for carbon from [3,24,4,25] and other alloying elements from [23]).

K i (MPa/wt%)
ni ()

rmod M

Al

Cr

Mn

Mo

Nb

Ni

Si

1171:3
1=3

4000
0:75

1400
0:75

7000
0:75

9600
0:75

4000
0:75

6100
0:75

11000
0:75

4000
0:75

11000
0:75

"
 2 #3=4
Gb
Up
1
k
Um

where U p and U m are the line energies of the dislocation in the precipitate (U p ) and in the iron matrix (U m ), respectively.
Finally the strengthening effect of impenetrable particles (ro ) is
described by the Orowan Mechanism [28]:

ro CM

 
Gb
ro
ln
k
ri

where C is a constant. ri and ro represent the inner and outer cut-off


radius of the dislocation stress field.
2.3.4. Dislocation strengthening
The contribution of the dislocations for the strengthening of the
steel can be defined by the well-known equation of Taylor [29,30]:

rdis a  G  b  qdis

where qdis is the dislocation density. The temperature dependency


of the strengthening due to dislocations is taken into account with
Eq. (10) [1]:

a 0:4  4:42  104  T

10

concerned. Thus the strengthening due to subgrains rsgb [MPa]


can be calculated for the martensitic steel by Eq. (14) [33,34,37]:

rsgb

115
2w

14

where w in lm. The martensite lath width is mainly a function of


carbon content of the steel. For a carbon content of 0.4% for the steel
of interest, the lath width is calculated as 0.1 lm [38]. This value
agrees well with the microstructural measurements from previous
studies for similar steels [31,39]. The temperature dependency of
the subgrain strengthening is correlated with the temperature
dependency of the shear modulus [40]. In this work, the lath width
was assumed to be constant for the tempering temperatures of
520 C and 600 C which are rather lower than the temperatures
at which static recovery of the martensitic lath structure can take
place. For the tempering temperature of 650 C, the static recovery
of the martensitic laths (coarsening of the martensite laths) is
known from the literature [41]. The evolution of the martensitic
lath width is predicted with an Arrhenius type equation similar to
Eq. (13).



dw
Q
C w wexp  m
dt
RT

15

The shear modulus of a martensitic steel was measured in the


work of Straub within the temperature range of 298973 K [6]:

where C w is a constant with a value of 1.4  103. After double tempering at 650 C for two hours the martensite lath width was calculated to be around 0.17 lm from Eq. 15.

G 0:039431  T 98:46456

2.4. Flow curve

11

where G is in GPa.
Bhadeshia et al. [3] proposed a general equation to determine
qdis of martensite at room temperature (RT) after quenching
depending on the martensite start temperature (M s in Kelvin)

logqdis 9:2840

6880:73 1780360

Ms
M 2s

12

M s was determined as 300 C from dilatometer experiments in the


previous work of the authors [35]. Finally the dislocation density of
the martensitic structure of AISI H13 is calculated as 7.41  1015
[1/m2] at RT from Eq. (12). The evolution of dislocation density
qdis during tempering can be estimated with an Arrhenius type
equation [21,32]:



dqdis
Q
C dis qdis exp  v
dt
RT

13

where C dis is a constant, Q v m is the activation energy for the motion


of vacancies (134 kJ/mol, [1]) and R is the gas constant. The
dislocation density of a double tempered tool steel (at 600 C for
two hours) of X38CrMoV5-1 which has a very similar chemical
composition to the steel of interest in this work was measured to
be 6  1014 [1/m2] [31]. Using the initial and final dislocation
density for a tempering process of two hours at 600 C, constant
C can be calculated from Eq. (13) as 3.002  104.
2.3.5. Subgrain strengthening
The contribution of the subgrains on the strength of the martensitic steel rsgb depends inversely on the mean value of the larger
diameter of slip planes [4,33,34,36]. The mean slip plane diameter
depends mainly on the thickness of the martensite lath w

Concerning the stress relaxation during tempering, not only the


evolution of the yield stress during the tempering but also a complete temperature dependent flow curve is necessary to describe
this phenomenon quantitatively.
In this work, the development of the work hardening was modeled by KocksMecking type equation [42]. Assuming that the
mobile dislocations move a mean free path L before they are
immobilized or annihilated, the evolution of the dislocation density qdis can be described by [43]:

dqdis M e_ pl

dt
bL

16

where e_ pl is the plastic strain rate, M is the Taylor factor. The mean
free path L is limited by the average spacing between dislocations:

A
L p

17

qdis

where A is a material constant.


During the softening due to spontaneous annihilation, the
reduction of the dislocation density can be defined by Eq. (18) [44]:

dqdis
dann
q Me_ pl
B2
dt
b dis

18

in which B is a constant depending on the different number of activated slip systems. The critical distance dann is described by [45]:
4

dann

Gb

2p1  mQ mf

19

283

A. Eser et al. / Computational Materials Science 113 (2016) 280291

where G is the shear modulus of the matrix material,

m is the

Poissons ratio and Q mf is the vacancy formation energy.


The modeling of recovery by dislocation climb is based on the
model of Lindgren et al. [43], with the assumption that the thermally activated climb is controlled by the self-diffusion coefficient
along dislocations Dd (pipe-diffusion) [46]:

dqd
Gb
C2Dd
q2  q2deq 
dt
kB T dis
3

20

where kB is the Boltzmann constant, T Temperature in Kelvin, qdeq is


the equilibrium dislocation density and C is a constant, which takes
solute trapping into account.
Finally the dislocation core diffusivity Dd is coupled to the
bulk diffusivity Db by Eq. (21) in MatCalc via a temperature
dependent factor ad :

Dd ad Db

21

For the calculation of the flow curve, three material dependent


parameter should be defined. First of all the parameter A which
depends on the average distance between dislocations and dislocation density should be known. In the work of Polcik [7] for a
martensitic chromium steel, the distance between the free dislocations was reported as 0.117 lm and for the dislocations in subgrains as 0.01 lm respectively. Hence the material parameter A
can be calculated from Eq. (17) between 0.17 and 4.59. This parameter was determined as 0.4 from iterative fitting of the calculated
and experimentally determined flow curve and is between the possible value of 0.17 and 4.59 [7]. The other two parameters namely B
from Eq. (18) and C from Eq. (20) were taken from the work of
Sherstnev et al. [47] and have values 5 and 0.001.

Especially for tool steels which are normally tempered for long
soaking times at elevated temperatures above 0:4T m , where T m is
the melting temperature of the steel, creep plays a major role on
the amount of stress relaxation.
The physically based creep model used in this work is based on
the composite model of Mughrabi et al. [48,49]. Blum et al. implemented this model to predict the high temperature creep behavior
of the martensitic steels [6,7,50,51]. The basic idea of the model is
that the heterogeneous dislocation distribution causes a heterogeneous stress distribution. Correspondingly the material is treated
as a composite of a hard (subscript h) phase with high dislocation
density and a soft (subscript s) phase with low dislocation density. It is assumed that the total strain etot is identical in both
regions:

rs
E

es

rh
E

eh

22

where etot is the sum of elastic and the plastic strain, E is the elastic
modulus, rs and rh are the local stress acting in soft and hard
region. The elastic strain is calculated by the terms rs =E; rh =E for
the soft and hard region respectively. es and eh are representing
the plastic strains in the corresponding regions. Due to the heterogeneous distribution of the dislocations, internal forward stresses
rf are formed at the (hard) subgrain boundaries corresponding to
internal back stresses rb in the subgrain interior. The local stresses
rs and rh can be defined as:

rs rA  rb ;

23

rh rA rf ;

24

where

r
etot 1  f h es f h eh

rA is the externally applied stress and can be calculated as:

rA f s rs f h rh :

25

26

 
where the first term rE is the elastic strain and the plastic strain
epl is defined by Eq. (27) as follows:

epl 1  f h es f h eh :

27

Differentiating Eq. (27) for the case of creep (r = constant) gives the
plastic deformation rate:

e_ pl 1  f h e_ s f h e_ h f_ h eh  es

28

where e_ s and e_ h are the deformation rates in the soft and hard
region of subgrains respectively. The local deformation rates are
formulated as:

e_ s

b
q ms ;
M dis

29

e_ h

b
q mh ;
M dis

30

with ms and mh the velocity of the dislocations in soft and hard region
and defined by Eqs. (31) and (32) respectively.

ms Bs  rs  expb  rn


s
b  s  rh
M  kb  T
2

mh v h;0  sinh

2.5. Creep

etot

In Eq. (25), the terms f h and f s is the volume fraction of the hard
and soft regions. The volume fraction of hard region is defined as
f h a  2=w (a = width of hard region, 2=w: subgrain boundary area
per volume) and for the soft region as f s 1  f h , so that the total
strain can be calculated from Eqs. (22) and (25) as:

31
!
32

The term s represents the distance between the dislocations in


the subgrain boundaries. b and n are parameters and Bs and v h;0
describes the temperature dependency with Arrhenius type
equations.



Q
B As  exp  s
RT

v h;0 Ah  exp

33



Q
 h
RT

34

Q s and Q h are the activation energies in soft and hard


regions and As ; Ah are constants. The effective stresses rs ; rh
defined in Eqs. (29) and (30) in soft and hard regions are defined
as follows:

rs rs  rdis  rp;s ;

35

rh rh  rp;h :

36

The effective stress in a soft region r is described as the difference between the local stress in soft regions rs and the stress components resulting from the interaction with other free dislocations
rdis and precipitates rp;s . On the other side, the effective stress in
hard regions rh is defined as the difference between the local stress
in the hard region rh and the stress component resulting from the
interaction with other free precipitates rp;h . The athermal stress
component rdis which results from the interaction of the free dislocations, is calculated using Eq. (9). The stress component rp;s
resulting from particle hardening in soft regions can be formulated
with particle size dp and the average volume fraction f p of the
precipitates.

s

rp;s Crs  rOrowan:s

37

284

A. Eser et al. / Computational Materials Science 113 (2016) 280291

rOrowan:s 3:32  G  b 

q
Np
f p;s;k
X
k1

38

dp;k

where N p is the number of precipitates. The phenomenological relationship between the parameter C p;s and the local stress rs is
described by Eq. (35).


rs
1  exp 
C1

C p;s

C 2 !!1=C2

39

with C 1 C 1 rOrowan:s ; C 2 C 2 rOrowan:s .


The volume fraction of different particles in the soft region f p;s;k
can be calculated as:

f p;s;k

1  Up;SGB
f p;k 
1  f p;SGB

40

where Up;SGB the fraction of particles located at subgrain boundaries


and f p;SGB is the volume fraction of the subgrain boundary region
where particles attached to subgrain boundaries are lying. The
value of f p;SGB is related to subgrain sizew and particle size dp;k :



p dp;k 3
f p;SGB 1  1  p 
2 6 w

41

The stress component resulting from the interaction with other


free precipitates rp;h in the hard region is calculated from

rp;h

2
f p;h;k
Eo X

b 1 dp;k

42

where Eo is a constant having a value of 14.4  109 J/m when b and


dp are expressed in meters.
The volume fraction of particles in the hard region is equal to
the average volume fraction f p;k of different particles times the
local relative particle density in the hard region:

f p;h;k f p;k 

Up;SGB

43

f p;SGB

The evolution of the microstructure during creep can be formulated from the equations above. The parameters of the dislocation
structure are the mean subgrain size w, the spacing between free
dislocations in the subgrain interior q0:5
, the width of the hard
f
region a, and the spacing s between the dislocations in the subgrain
boundaries. Their steady state values (after creep) designated by
the subscript 1 are given by the equations:

w1 10 

bG
jrj

q0:5
dis;1 3:9 

bG
jrj


s1 1:5  b 

a1

44

45
0:5
46

jrj

0:05  w1
2

47

The evolution of the structure parameters w; q0:5


dis ; a and s is
assumed to develop with strain e along a path relating the initial
value to the steady states values:


log f log f1 logfo =f1  exp 

e
kf


48

where f stands for w; q0:5


dis ; a and s with appropriate values for kf .

Under constant loading (creep), the rate of the local stress in the
soft region can be calculated by differentiation of Eq. (23):

r_ s f h Ee_ h  e_ s f_ h

rs  r
fh


49

Eq. (49) was integrated over the plastic strain, so that the development of the plastic strain rate during creep can be achieved. The
necessary material and microstructure parameters for the modeling
of creep is summed up in Table 3.
3. Results and discussion
3.1. Precipitation
For realistic simulation of the precipitation evolution during
annealing, the initial precipitation state should be known. In this
work, the initial microstructure (carbide size, carbide content)
after austenitization and quenching was determined from the
scanning electron microscope (SEM) investigations [52]. The
undissolved carbides after austenitization and quenching found
to be MC carbides with a volume fraction of 0.65 vol% and average
radius of 57.01 nm. In order to define the initial state, a so called
Virtual Pre Treatment in MatCalc was conducted. The parameters
of the Virtual Pre-Treatment can be arbitrarily selected so that
the phase fraction and the size of the carbides in the initial structure are consistent with the experimentally determined values.
After several iterative Virtual Pre Treatment simulations, the calculated carbide phase fraction and the average radius of the MC
carbides matched with the experimental findings, which describes
the initial state before tempering.
The carbide types and the possible nucleation sites considered
in the simulation were selected to be the same as in the previous
work of Sonderegger [53] where a similar hot-work tool steel
was investigated (Table 4). Furthermore dislocation density and
subgrain size were given as input for the simulation of precipitation during tempering, whose values were discussed already in
the previous section.
The simulation of the precipitation development during tempering was conducted for three different tempering temperatures;
520, 600 and 650 C (Figs. 13). 520 C is the tempering temperature, where the maximum hardness can be achieved [56]. 600 C
is a common tempering temperature for this kind of steels and
650 C is towards the highest temperature, that the hot-work tool
steel can be tempered. The heating and cooling times were selected
as 0.5 h and the soaking time at the corresponding tempering temperature as 2 h. The diagrams show only the secondary hardening
carbides, which precipitate during tempering. Phase fraction and
size of MC carbides, which do not dissolve during austenitization,
change only slightly during tempering.
For all of the tempering temperatures, cementite was formed
during the heating to the tempering temperature in the first tempering step and dissolves rapidly at the tempering temperature.
The M7C3 carbides were formed rapidly upon reaching the tempering temperature and the phase fraction was higher than all the
other carbides except for the tempering temperature of 650. However, its fraction decreases with further soaking time. In contrast,
the fraction of the second Cr rich carbide namely M23C6 increases
gradually during the tempering. For the tempering temperature
of 650 C, M23C6 carbide has the highest volume fraction in comparison with the other carbides. The calculated phase fraction of
M6C carbide has its maximum value during tempering at 600 C.
The phase content of this carbide also rises gradually with the
increasing tempering time, except during tempering at 650 C, like
the M23C6 carbide. However, the calculated phase portion of this
carbide was lower than that of the other carbides except for the

285

A. Eser et al. / Computational Materials Science 113 (2016) 280291


Table 3
Material and microstructure parameters for the creep model.
Definition

Value

Source

kw
ka
kqdis;0

Initial martensite lath width


Initial dislocation density
Initial spacing between the dislocations in the subgrain
the fraction of particles located at subgrain boundaries
Eq. (48)
Eq. (48)
Eq. (48)

0.1 (lm) for 600 C Eq. (15) for 650 C


Eqs. (8) and (9)
0.01 (lm)
0.87
0.1243
0.003
0.0005

[38]
[2]
[7]
[7]
[7]
[7]
[7]

ks
C 1 rOrowan;s
C 2 rOrowan;s

Eq. (48)
Eq. (39)
Eq. (39)

0.0005
116:37 53:54  ln0:25  rOrowan;s 21:65

24:55 0:31  rOrowan;s  5:08  105  r2Orowan;s 6:8026  107  r3Orowan;s

[7]
[7]
[7]

Eq. (10)

a 0:4  4:42  104  T

[1]

b
G
Qs
Qh
As
Ah
Bs
b
n

Burgers vector
Shear modulus
Activation energy (soft region) Eq. (33)
Activation energy (hard region) Eq. (34)
Eq. (33)
Eq. (34)
Eq. (31)
Eq. (31)
Eq. (31)

0.248 (nm)
Eq. (11)
562 (kJ/mol)
562 (kJ/mol)
3.59  1015 (m/s Pa)
4.39  1020 (m/s)
9.0  1019 (m/Pa s)
0.59
0.045

[7]
[6]
[7]
[7]
[7]
[7]
[7]

w0

qdis;0
s0

Up;SGB

Table 4
Carbide types and their possible nucleation sites which were considered in the
precipitation simulations.
Carbide type

Possible nucleation sites

MC
M23C6
M7C3
M6C
Cementite (M3C)

Grain boundary,
Grain boundary,
Grain boundary,
Grain boundary,
Dislocations

Subgrain
Subgrain
Subgrain
Subgrain

boundary, Dislocations
boundary
boundary
boundary

tempering temperature of 600 C. The phase fraction of MC carbide


shows its maximum value after annealing at 650 C. In particular
the phase fraction of MC carbide increases significantly towards
the end of the second tempering step at 650 C (Fig. 3).
Another important parameter with regards to the precipitation
strengthening is the carbide size. The simulation results show that
the M23C6 carbides were larger than all other carbides at each tempering temperature. The coarsening rate of this carbide, especially
at 650 C was observed to be very high compared to the other

carbides (Fig. 3). Although the M7C3 carbide shows coarsening with
increasing tempering temperatures, the coarsening rate, when
compared with the M23C6 carbide was very low. However, the
coarsening of the M7C3 carbide is faster than the MC and M6C carbides. The M6C carbide has a higher resistance to coarsening, especially at elevated temperatures. The size of the carbide was higher
at the tempering temperatures of 520 C and 600 C than that of
MC carbide, however, the size of the carbide after tempering at
650 C was almost equal to be the size of the MC carbide. Finally,
the mean radius of the MC carbide after each tempering temperature was calculated as the smallest compared to the other carbides.
Fig. 4 shows the results of the calculated phase fraction and the
carbide size of the secondary hardening carbides after two times
tempering at the corresponding tempering temperatures. The total
fraction of the secondary carbides corresponds to about 5.6 vol.% at
520 C, 5.8 vol.% at 600 C and 5.3 vol.% at 650 C. In the work of
Ebner et al. [54] the phase fraction of secondary carbides was
determined for the steel X38CrMoV5-3 as 4.5 vol.% for the austenitization temperature of 1050 C. Although the exact tempering
temperature was not mentioned in the work, the tempering

Fig. 1. Kinetic simulation of the precipitate evolution of steel AISI H13 during tempering at 520 C: (a) tempering temperaturetime profile; (b) phase fraction; (c) mean
precipitate radius; and (d) number density.

286

A. Eser et al. / Computational Materials Science 113 (2016) 280291

Fig. 2. Kinetic simulation of the precipitate evolution of steel AISI H13 during tempering at 600 C: (a) tempering temperaturetime profile; (b) phase fraction; (c) mean
precipitate radius; and (d) number density.

Fig. 3. Kinetic simulation of the precipitate evolution of steel AISI H13 during tempering at 650 C: (a) tempering temperaturetime profile; (b) phase fraction; (c) mean
precipitate radius; and (d) number density.

Fig. 4. Simulated phase fraction and mean radius after two times tempering at the tempering temperatures of 520 C, 600 C, and 650 C.

A. Eser et al. / Computational Materials Science 113 (2016) 280291

process was carried out so that hardness values between 50 and 52


HRC were achieved, which corresponds to a tempering temperature of 600 C. In a further study [53], the carbides of the steel
X38CrMoV5-3 were examined with SANS (small-angle neutron
scattering). The samples were austenitized at 1060 C for 50 min
and tempered twice at 550 C for one hour and at 610 C for two
hours. The phase fraction of Cr-rich secondary hardening carbides
corresponded to 2.3  0.2 vol.% and the fraction of MC and Mo-rich
secondary hardening carbides 1.9  0.2 vol.%. The sum of the phase
fraction of the MC and M6C carbides secondary hardness in this
work was calculated as 0.97 vol.%, 1.58 vol.% and 1.60 vol.% for
the tempering temperatures of 520, 600 and 650 C correspondingly. On the other side, total phase fraction of M7C3 and M23C6
Cr-rich corresponds to 4.6 vol.%, 4.2 vol.% and 3.7 vol.% for tempering temperatures each for 520, 600 and 650 C. In this work, the
calculated phase fraction of the Cr-rich carbides was higher than
that of the experimentally measured values from the literature.
However, the phase fractions of the MC and M6C carbides were
more consistent with the literature data.
The above-mentioned discrepancy between the simulated and
experimentally observed phase fractions can be explained with
two aspects. First of all, the steel X38CrMoV5-3 contains less carbon (0.38 wt% instead of 0.40 wt%), chromium (4.8 wt% instead of
5.37 wt%) and vanadium (0.62 wt% instead of 1.22 wt%) and more
molybdenum (2.8 wt% instead of 1.34 wt%) than the steel AISI
H13. The higher estimated proportion of Cr-rich carbides in this
study can be explained by the higher percentage of chromium.
The lower phase fraction of M6C carbides MC and in this work compared to the literature [53] should be connected to the lower content of molybdenum. The higher phase fraction of the total
simulated secondary carbides in this work compared with the literature can be explained due to the higher carbon content of the
investigated steel. The second important aspect that can lead to
above mentioned deviations can be due to different heat treatment
processes. In Ref. [53], tempering time and tempering temperatures varies from the present work. (1st tempering step = 550 C
(1 h) and 2nd tempering stage = 610 C (2 h)). This can affect not
only the carbide size, but also the phase fraction. The coarsening
of Cr-rich carbides for the steel X38CrMoV5-1 was also
investigated in the work of Hu et al. with transmission electron
microscopy (TEM) measurements [55]. For the relative fraction of
Cr-rich carbides to the total carbides was reported to be 65.4%
which was very close to the calculated value of 66% in this work
(tempering temperature of 600 C). In contrast, the relative fractions of the M6C and MC carbides were identified in the work
[55] to be 23% and 11% respectively, in the present study they
are calculated to 16% and 18% (without considering the MC carbides, which do not dissolve during austenitization). At this point
it should be mentioned that the vanadium content of the steel in
the work of Hu et al. [55] was 1.02 wt% (instead of 1.22 wt% in
the present study) and the molybdenum content was 1.5 wt%
(instead of 1.34 wt% in the present work). The calculated higher
fraction of MC carbides and lower fraction of M6C carbides in this
work compared to reference [55] is due to the higher vanadium
and lower molybdenum content of the investigated steel in this
work.
The secondary carbides show a gradual coarsening with the
elevated tempering temperatures (Fig. 4). The mean radius of the
Cr-rich carbide after tempering was measured to be around
44 nm for a similar steel [55] for the tempering temperature of
610 C. This value lies between the calculated mean radius of
M7C3 and M23C6 at 600 (about 25 nm for M7C3 and M23C6) and
650 C (about 50 nm for M7C3 and M23C6). Furthermore in the same
study, the rod-shaped carbides of M6C reported to have the dimension of 20  55 nm (approximately 17.5 nm average radius if it was
assumed to be spherical carbides). This was between the calculated

287

values of about 16 nm at 600 C and about 21 nm at 650 C. The


size of MC carbides values was reported between 20 and 40 nm
in Ref. [55]. These values were slightly higher than the calculated
values of approximately 8 nm at 600 C and 16 nm at 650 C.
Despite the existing deviations of the measured phase fractions
and carbide sizes in the literature and simulated values in this
work, the discrepancy is within the deviation range of the experiments and simulations.
3.2. Yield stress
The calculated precipitation strengthening during tempering
from Eq. (4) is represented in Fig. 5. The overall precipitation
strengthening decreases with increasing tempering temperatures,
due to the coarsening of the secondary carbides as expected. However the drop of the precipitation strengthening of the Cr-rich carbides (M7C3 and M23C6) compared to the V-rich MC carbides and
Mo-rich M6C carbides was much larger due to the higher coarsening rate of the Cr-rich carbides at elevated temperatures
(Figs. 2 and 3). Nevertheless, M7C3carbides has the highest precipitation strengthening for the tempering temperature of 520 C. In
contrast, the second Cr-rich carbide M23C6, despite higher phase
fraction has lower precipitation strengthening due to its large size.
For tempering temperatures of 600 C and 650 C provide the secondary MC carbides the largest contribution to the overall yield
strength. MC carbides show higher precipitation strengthening
due to their high resistance to coarsening, especially at higher
tempering temperatures and longer soaking times (2nd tempering
step). On the other hand the coarser MC carbides which were not
dissolved during the austenitization have minor effect on the
overall precipitation strengthening.
The The simulated total yield stress, each strengthening component and the experimental yield stresses for three different tempering temperatures are summarized in Fig. 6. The experimental
yield stresses were measured by compression tests in the previous
work [52]. Experiments were conducted with the same temperature profile applied in the simulations (0.5 heating/cooling time, 2
h soaking time at tempering temperature). By comparing the calculated and experimental values, very good agreement can be
observed in the error range of experiments. It should be noted that
during the tempering, the microstructure was not stable and the
experiments had larger error bars especially for the first tempering
step.
Subgrain boundaries of the martensitic steel are the major
obstacle against gliding of mobile dislocations [41]. The simulation
results capture this phenomena quite well and the subgrain
strengthening was the highest strength component for all tempering temperatures. For the tempering temperatures of 520 and
600 C, it was assumed that there is no change of the lath width
during tempering. For the tempering temperature of 650 C, coarsening of the martensitic lath was calculated from Eq. (15). The precipitation strengthening has its maximum contribution to the
overall strength for the tempering temperature of 520 C, where
the maximum hardness of the steel AISI H13 is also expected
[56]. The effect of the precipitation strengthening gradually
decreases with increasing temperature due to the coarsening of
the carbides.
3.3. Flow curve
One of the major stress mechanisms during tempering is due to
the yielding under the influence of quenching induced residual
stresses due to decreasing yield strength. After determining the
yield stress during tempering, the presence of the stress relaxation
due to this mechanism can now be estimated. However, the magnitude of this relaxation can only be calculated if the complete flow

288

A. Eser et al. / Computational Materials Science 113 (2016) 280291

Fig. 5. Simulated precipitation strengthening rp during tempering compared with the experimental findings: (a) Tempering temperature 520 C. (b) Tempering
temperature 600 C. (c) Tempering temperature 650 C.

Fig. 6. Simulated yield stresses and the corresponding strengthening terms during tempering compared with the experimental findings: (a) Tempering temperature 520 C.
(b) Tempering temperature 600 C. (c) Tempering temperature 650 C.

curve is known aside the yield strength. For this reason, simulations of the flow curves during tempering are required. The critical
point during tempering where the stress relaxation (time independent) due to the drop of yield stress can occur should be where the
tempering temperature is just reached. After reaching the tempering temperature, the time dependent relaxation due to creep

would be the significant mechanism. Consequently, the simulation


of the flow curve was carried out immediately after the tempering
temperature reached.
The calculated and measured flow curves at 520 C, 600 C and
650 C for the first and second tempering step are shown in Fig. 7.
The measured and simulated flow curves match quite well with

A. Eser et al. / Computational Materials Science 113 (2016) 280291

289

Fig. 7. Simulated flow curves at the corresponding tempering temperatures (immediately after reaching the tempering temperature) compared with the experimental
findings: (a) Tempering temperature 520 C. (b) Tempering temperature 600 C. (c) Tempering temperature 650 C.

each other. At this point it should be once more pointed out, that
especially for the first tempering due to the higher instability of
the microstructure at elevated temperatures, the measured values
has higher uncertainty [52]. Considering this, a better agreement of
the measured and calculated flow curves for the second tempering
cycle can be observed from Fig. 7 as expected.

3.4. Creep
Especially for steels with higher tempering temperatures, creep
(time-dependent/viscous deformation) is the major stress relaxation mechanism during tempering. The experimental observation
showed that short term creep effect for the steel AISI H13 becomes
significant for the tempering temperatures of 600 C and 650 C
[52]. In contrary to the instant compression tests conducted during
the tempering to determine yield stress and the corresponding
flow curves, the creep experiments were conducted on three times
tempered specimen, so that a stable microstructure before the
experiments is assured. During tempering, the microstructure is
instable. The compression tests to determine flow curves were relatively short, the effect of microstructure evolution during the test
can be neglected. However during a short term creep test (10 h) of
a not tempered martensite, the tempering process takes place during the test itself which will affect the measurements remarkably.
Consequently creep experiments were simulated only for the tempering temperatures of 600 C and 650 C for three times tempered
martensite under the loads of 400, 500 and 600 MPa.
The mathematical model of creep was programmed in Matlab
and solves the differential equation (Eq. (49)) with the Matlab

Solver ode45. In Table 1, the required microstructure and material parameters for the simulation of creep are summarized. In
addition to those parameters, the model requires for the description of the Orowan stress (rOrowan:s , Eq. (38)) the volume fraction
and the size of the carbides. These two parameters were calculated
by MatCalc and the creep model is taking those parameters from
the precipitation simulations. Two other parameters b and n (Eq.
(31)) used for the description of the dislocation velocity in the soft
region (Subgrain interior) were required to be estimated as in the
reference work [7]. In this work, these parameters were determined in Matlab by the optimization function fmincon. The
calculated parameters b and n are given in Table 1.
The simulation results and their comparison with the experimentally determined strain rate during loading is shown in
Fig. 8. Although the overall shape of the calculated curves particularly for smaller strains up to 0.004 at the tempering temperature
of 600 C, deviates from the experimental curves, the model predicts the minimum creep rate e_ min very well. It should be noted
that the initial microstructure for the creep tests at 600 C and
650 C are completely different, while the specimens were tempered for three times at the corresponding tempering temperatures. This effect was taken into account in the model with the
consideration of the dislocation density, subgrain size, volume
fraction and size of carbides.
One of the simplest material models to describe creep in the
macroscopic scale is Nortons creep law, where the stress dependency of the steady state creep rate e_ c is described in a power
law:

e_ c Ac rnc

50

Fig. 8. Simulated creep curves at the corresponding tempering temperatures compared with the experimental findings: (a) Tempering temperature 600 C. (b) Tempering
temperature 650 C.

290

A. Eser et al. / Computational Materials Science 113 (2016) 280291

Table 5
Parameter Ac and nc for the power law at temperatures 600 C and 650C.
Ac (s1 MPanc )

600 C
650 C

University Vienna) for his kind support in MATCALC precipitation


simulations.

nc ()

Exp.

Sim.

Exp.

Sim.

6.72393  1013
1.59987  1023

3.47308  1013
2.80594  1022

2.06132
6.57970

2.17178
6.12114

Applying this equation for the measured and simulated creep


curves, the Ac and nc values can be calculated (Table 5). The effect
of the discrepancy between the values of Ac and nc for the experimental and simulated creep parameters on the stress relaxation
will be examined in the second part of this study.
4. Conclusion
In this work, the dominant mechanical material properties
(yield stress, work hardening, creep) which are necessary to model
the stress relaxation during tempering of a tool steel were calculated using physically based approaches. Suitable modeling
approaches for the simulation of yield stress, flow curve and creep
were selected from the literature. Coupling those models with precipitation simulations was the most important achievement of this
study. Utilizing this methodology, different tempering conditions
(soaking time, tempering temperature) can be considered in the
FEM-Model to predict the stress relaxation without going through
a tedious, time consuming and expensive experimental campaign.
The MatCalc program was chosen for the simulation of the precipitation. Especially for similar steels, the prediction power of the
program was already demonstrated in previous work [1921,53].
In this work the simulated phase fraction and carbide size was
compared with the previous experimental results taken from the
literature. Although simulation results were in good agreement
with the experiments from the literature, a detailed experimental
investigation should be conducted in the future, to verify the prediction of carbide size and volume fraction over the whole range of
relevant tempering temperatures.
Considering the calculated and experimentally measured yield
stresses during tempering a very good prediction of the simple
model used in this study can be observed. The evolution of the dislocation density and martensite lath width was calculated using
new equations compared to previous work [52], which results in
better estimation of the measured yield stress values. The work
hardening model which is integrated in MatCalc also gives very
good results, although for the first tempering step, the calculated
flow curves deviates more from the measured ones. Finally the
creep model delivers quite good results especially for the steady
state strain rate e_ c .
The material models in use still need a big number of
microstructure and material parameters. The parameters used in
this study were taken from similar steels from the literature. The
experimental investigation of those parameters will definitely help
to increase the precision of the models. Beside the precipitations,
the evolution of the dislocations and the martensite lath width
during tempering are the key factors describing the tempering process. In this work the evolution of those parameters was taken into
account with simple equations, although these phenomena are
highly complex. The precipitation simulations should be coupled
in the future with physical models to simulate the evolution of
the dislocations and the recovery process.
Acknowledgement
The authors gratefully acknowledge Prof. Dr. Ernst Kozeschnik
(Institute of Materials Science and Technology, Technical

References
[1] Y. Wang, B. Appolaire, S. Denis, J. Microstruct. Mater. Prop. 1 (2) (2006) 197
207.
[2] Q. Li, Mater. Sci. Eng. A. 361 (2003) 385391.
[3] C.H. Young, H.K.D.H. Bhadeshia, Mater. Sci. Technol. 10 (3) (1994) 209214.
[4] I. Holzer, Modelling and Simulation of Strengthening in Complex Martensitic
912%Cr Steel and a Binary FeCu Alloy, PhD thesis, Graz University of
Technology, Graz, 2010.
[5] I. Holzer, E. Kozeschnik, Mater. Sci. Eng. A. 527 (2010) 35463551.
[6] S.Straub: VDI-Fortschritt Berichte, Dissertation Universitt Erlangen- Nrnberg
1994, VDI-Verlag, Reihe 5, Nr. 405, Dsseldorf, 1995.
[7] P. Polcick, Modellierung des Verformungsverhaltens der warmfesten 9-12%Chromsthle im Temperaturbereich von 550650 C. Diss., Der Technische
Fakultt der Universitt Erlangen-Nrnberg, 1998.
[8] N. Saunders, X. Li, A.P. Miodownik, J.P. Schille, Computer modelling of
materials properties. In: Materials Design Approaches and Experiences,
Minerals Metals & Materials Society, 2001.
[9] Z. Guo, N. Saunders, J.P. Schill, A.P. Miodownik, Mater. Sci. Eng. A. 499 (2009)
713.
[10] J. Svoboda, F.D. Fischer, P. Fratzl, E. Kozeschnik, Mater. Sci. Eng. A. 385 (2004)
166174.
[11] E. Kozeschnik, J. Svoboda, P. Fratzl, F.D. Fischer, Mater. Sci. Eng. A 385 (2004)
157165.
[12] E. Kozeschnik, J. Svoboda, F.D. Fischer, CALPHAD 28 (2005) 379382.
[13] R. Kampmann, R. Wagner, Decomposition of Alloys: The Early Stages,
Pergamon, Oxford, 1984.
[14] R. Wagner, R. Kampmann, Mater. Sci. Technol. 5 (1991) 213303.
[15] R. Becker, W. Dring, Ann. Phys. 5 (1935) 719751.
[16] J.B. Zeldovich, Acta Physicochim. 18 (1943) 122.
[17] H.K. Lukas, S.G. Fries, B. Sundman, Computational Thermodynamics: The
CALPHAD Method, Cambridge University Press, Cambridge, 2007.
[18] N. Saunders, A.P. Miodownik, CALPHAD Calculation of Phase Diagrams: A
Comprehensive Guide, Elsevier Science Ltd., Oxford, 1998.
[19] H. Leitner, P. Staron, M. Bischof, H. Clemens, E. Kozeschnik, J. Svoboda, F.D.
Fischer, Coarsening of secondary hardening carbides in a hot-work tool steel:
experiments and simulation, in: Proceedings of MS&T 2004, 26.-29.9.New
Orleans, 2004, pp. 623632.
[20] B. Sonderegger, E. Kozeschnik, M. Bischof, H. Leitner, H. Clemens, J. Svoboda, F.
D. Fischer, in: Proc. 7th Tooling Conference, TOOL06, 2.-5. May 2006, Torino,
Italy, Characterization and Simulation of Precipitation Kinetics During Heat
Treatment of the Hot-Work Tool Steel X 38 CrMoV 5-3, 2006, pp. 533540.
[21] B. Sonderegger, E. Kozeschnik, H. Leitner, H. Clemens, J. Svoboda, F.D. Fischer,
P. Staron, Steel Res. Int. 81 (2010) 6473.
[22] C. Liebaut, Rheologie de la deformation plastique dun acier FE-C durant sa
transformation de phase, Diss, Nancy, INPL Nancy, 1988.
[23] C.E. Lacy, M. Gensamer, Trans. ASM 32 (88) (1944).
[24] P.G. Winchell, M. Cohen, Trans. ASM 55 (1962) 347361.
[25] J. Hald, Steel Res. 67 (1996) 369374.
[26] L.M. Brown, R.K. Ham, in: A. Kelly, R.B. Nicholson (Eds.), Strengthening
Methods in Crystals, Applied Science Publishers, London, 1965, pp. 9135.
[27] K.C. Russell, L.M. Brown, Acta Metall. Mater. 20 (1972) 969974.
[28] M. Ashby, in: G.S. Ansell, T.D. Cooper, F.V. Lenel (Eds.), Metallurgical Society
Conference, vol. 47, Gordon and Breach, New York, 1968, pp. 143205.
[29] G.I. Taylor, in: Proceedings of the Royal Society. vol. 145. 1934, pp. 362387.
[30] G.I. Taylor, J. Inst. Met. 62 (1938) 307324.
[31] H. Wurmbauer, H. Leitner, M. Panzenboeck, C. Scheu, H. Clemens, Mat.-wiss. U.
Werkstofftechn. 41 (2010) 1828.
[32] Zengshou Hou, Guangxi Lu, Physical Metallurgy, Presse des Sci. & Tech., de
Shanghai, 1989.
[33] J.P. Naylor, Metall. Trans. 10A (1979) 861.
[34] J. Daigne, M. Guttmann, J.P. Naylor, Mater. Sci. Eng. 56 (1982) 110.
[35] A. Eser, A. Bezold, C. Broeckmann, I. Schruff, T. Greeb, HTM J. Heat Treat. Mater.
69 (2014) 127137.
[36] D.W. Smith, R.F. Hehemann, J. Iron Steel Inst. 209 (1971) 467481.
[37] H.K.D.H. Bhadeshia, Bainite in Steels, second ed., The Institute of Materials,
London, 2001, ISBN 1-86125-112-2.
[38] S. Cobo, O. Bouaziz, in: New Developments on Metallurgy and Applications of
High Strength Steels, Argentina, Buenos Aires, Materials Science and
Technology, 2008.
[39] N. Mebarki, D. Delagnes, P. Lamesle, F. Delmas, C. Levaillant, Mater. Sci. Eng. A
Struct. Mater. Prop. Microstruct. Process. 387 (2004) 171175.
[40] K. Maruyama, K. Sawada, J. Koike, ISIJ Int. 41 (2001) 641653.
[41] H. Ghassemi-Armaki, R.P. Chen, K. Maruyama, M. Yoshizawa, M. Igarashi,
Mater. Lett. 63 (2009) 24232425.
[42] U.F. Kocks, ASME J. Eng. Mater. Technol. 98 (1976) 7683.
[43] L.E. Lindgren, K. Domkin, S. Hansson, Mech. Mater. 40 (2008) 907919.
[44] F. Roters, D. Raabe, G. Gotstein, Acta Mater. 48 (2000) 41814189.
[45] S. Brinckmann, R. Sivanesapillai, A. Hartmaier, Int. J. Fatigue 33 (2011)
13691375.
[46] M.S. Soliman, E.A. El-Danaf, A.A. Almajid, Adv. Mater. Res. 83 (2009) 407414.

A. Eser et al. / Computational Materials Science 113 (2016) 280291


[47] P. Sherstnev, P. Lang, E.P. Kozeschnik, in: Tagungsband zur European Congress
on Computational Methods in Applied Sciences and Engineering. Vienna,
Austria, 1014 September 2012.
[48] H. Mughrabi, in: P. Haasen, V. Gerold, G. Kostorz (Eds.), Proc. of 5th Conf. on
the Strength of Metals and Alloys (IGSMA 5), Pergamon Press, 1980, pp. 1615
1635.
[49] H. Mughrabi, Acta Metall. (1983) 13671379.
[50] W. Blum, S. Vogler, M. Biberger, A.K. Mukherjee, Mater. Sci. Eng. A Struct.
Mater. Prop. Microstruct. Process. 112 (1989) 93106.
[51] W. Blum, A. Rosen, A. Cegielska, J.l. Martin, Acta Metall. 37 (9) (1989) 2439
2453.

291

[52] A. Eser, Skalenbergreifende Simulation des Anlassens von Werkzeugsthlen,


Diss., RWTH Aachen, 2014.
[53] B. Sonderegger, E. Kozeschnik, H. Leitner, H. Clemens, J. Svoboda, F.D. Fischer,
P. Staron, Steel Res. Int. 81 (2010) 6473.
[54] R. Ebner, H. Leitner, D. Caliskanoglu, S. Marsoner, F. Jeglitisch, Z. Metallkd. 92
(2001) 820829.
[55] X. Hu, M. Zhang, X. Wu, L. Li, J. Mater. Sci. Technol. 22 (2006) 153158.
[56] Bhler Edelstahl GmbH: Werkstoffdatenblatt Warmarbeitsstahl W302,
November 2006.

You might also like