Carbon Diffusion Measurement in Austenite in the Temperature

Range 500
C to 900
C

P. THIBAUX, A. MÉTENIER, and C. XHOFFER

Carbon diffusion in austenite plays a critical role in phase transformation in steel. However, it can
only be estimated in the fully austenitic range and has then to be extrapolated to the temperature
range of the phase transformation. Therefore, published data are limited to temperatures above
750
C. In this study, new experiments are carried out to determine the carbon diffusion coefficient
in austenite at temperatures as low as 500
C. Carburization experiments are performed in the
austenitic range for a Fe-1.5 pct Mn 0.13 pct C and a Fe-31 pct Ni alloy (wt pct). Composition
profile measurements, which are done using glow discharge optical emission spectrometry
(GDOES), show that the surface composition is not constant with time. A methodology has been
developed to assess the diffusion coefficient of carbon in austenite combining the measured carbon
profiles and a numerical method to compute the diffusion profile taking into account the time
evolution of the boundary condition. This method is first validated on the Fe-C-Mn steel. Car-
burization experiments are carried out on a Fe-31 pct Ni alloy at 900
C, 800
C, 700
C, 600
C,
and 500
C. The carbon diffusion coefficient is assessed using the method described above and
15;050
fitted with the following expression (T in Kelvin): D ¼ 1:23
106
e TðKÞ ðm2 /sÞ. The new
expression is compared with previous experimental results measured for comparable nickel
content at higher temperatures, and it shows a reasonable agreement. The model proposed by
Ågren for carbon diffusion has been modified to take into account the thermodynamic contri-
bution of nickel. This model also shows good agreement with the present experimental results,
even if it was fitted to experiments performed at higher temperatures.

DOI: 10.1007/s11661-007-9150-5
 The Minerals, Metals & Materials Society and ASM International 2007

I. INTRODUCTION lower than 800
C, it is of great interest to obtain

experimental data for lower temperatures.
THE carbon diffusion coefficient is one of the most The main difficulties encountered in measuring the
important parameters in the modeling of phase trans- diffusion coefficient of austenite below this temperature
formation in steel. For instance, the kinetics of the range consist of avoiding the austenite/ferrite phase
austenite/ferrite transformation is assumed to be mainly transformation and measuring the very shallow profiles
a diffusion-controlled process.[1] There are also diverg- induced by diffusion at these lower temperatures. We
ing opinions about the bainitic transformation. Some propose to overcome this first difficulty by using a steel
authors assume that it is a reconstructive transformation grade that is austenitic at room temperature. The second
where carbon diffusion determines the phase transfor- difficulty is resolved by analyzing diffusion profiles using
mation kinetics;[2,3] other authors describe this transfor- the glow discharge optical emission spectrometry
mation as a displacive transformation where nucleation (GDOES). In order to create diffusion profiles, we
is the rate determinant step.[4] Recent studies try to perform carburization experiments. Diffusion coeffi-
combine both modes of growth.[5,6] In all cases, the cients are computed from these diffusion profiles.
carbon diffusion coefficient in austenite is used at
temperatures where austenite is no longer stable.
There exists nowadays several models for the carbon
diffusion in austenite.[7,8] Most of them are fitted to the II. METHODOLOGY
experiments of Wells et al.[9] These experiments have A. Selection of Steel Grades
been carried out at temperatures between 800
C and
1200
C. Because the austenite/ferrite and the austenite/ Two steel grades are investigated in this study. A Fe-
bainite transformations occur at temperatures much C-Mn steel grade is used as reference, because its
composition is close to the ones used for assessing
diffusion coefficients in the literature. The second steel,
P. THIBAUX, Research Engineer, Physical Modelling, Substrate,
Metallic Coating Development and C. XHOFFER, Research Engi- Fe-31 pct Ni, is fully austenitic in the temperature range
neer, Materials Characterisation are with Arcelor Research Industry 500
C to 900
C.
Gent/OCAS N.V, B-9060 Zelzate, Belgium. Contact e-mail: philippe. The Fe-C-Mn steel is slightly alloyed with Mn, Cr,
thibaux@ arcelor.com A. MÉTENIER, formerly Graduate Student, and C (Table I). The Ac3 and Ar3 temperatures are
Materials Science Department, Ecole Polytechnique Universitaire de
Lille, Lille, France, is External Consultant, CEV, Aulnoye, France.
estimated to be around 850
C and 760
C, respectively.
Manuscript submitted June 20, 2005. This steel was industrially continuously cast, hot rolled,
Article published online June 13, 2007. cold rolled, and annealed. The initial state consists of a

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 38A, JUNE 2007—1169

 Table I. Composition of the Steels (Weight Percent)

Steel C Mn Si Cr Mo Ni W Fe
Fe-C-Mn 0.136 1.485 0.132 0.372 0.201 — — balance
Fe-Ni 0.004 — — — — 31.3 0.04 balance

microstructure of ferrite (85 pct) and martensite converted into a concentration vs depth profile. Intensity
(15 pct). The ferrite grain size is in the range of 8 to to concentration conversion is done using calibration
15 lm. According to Reference 10, the concentrations of standards having different C concentrations. The conver-
alloying elements are too low to influence significantly sion of time to depth is based on the sputtering rate
the carbon diffusion coefficient. (erosion rate). The analysis is performed with an anode
The second steel is alloyed with 31 pct Ni and has a diameter of 2.5 mm, a tension of 1000 V, and a current of
very low content of any other alloying element (Table I). 15 mA. The sputtering rate is estimated to be 15 nm/s.
Such a high Ni content should make the steel fully The sensitivity limit for the measurement of C is around
austenitic at temperatures as low as 500
C. The steel 1 pct. Therefore, no measurement of carbon concentra-
was cast at the University of Ghent in a vacuum furnace tion is significant for concentrations above 1 pct.
and hot rolled by laboratory rolling equipment. It has
been verified by X-ray diffraction that this steel is fully
D. Numerical Simulation of Diffusion
austenitic at room temperature.
Because there is no analytical solution of the diffusion
Eq. [1] when the diffusion coefficient is not constant or
B. Carburization Experiments when the boundary conditions are arbitrarily a function
The determination of the C diffusion coefficient is of the time, a numerical method is required to solve the
carried out by measuring and analyzing diffusion diffusion problem. Two packages are used for the
profiles. In order to create carbon diffusion profiles, a numerical simulations: DICTRA and a specific diffusion
gas-solid carburization reaction is carried out at the program. The simulations with DICTRA are carried out
surface of the samples. The gas compositions are with version 22 of the package, with the second version
indicated in Table II. of the solid solution database (SSOL2) and the standard
The experiments are carried out in a reactive annealing mobility database (MOB).[11,12] A specific diffusion
furnace. Due to the specific design of this furnace, the program has been written to solve efficiently the
maximum sample thickness is 1.5 mm, with a diameter of diffusion Eq. [1].
10 mm. The temperature is measured by two thermo-

couples in direct contact with the sample. The heating is @C @ @C
¼ D
½1
provided by two halogen lamps of 150 W. A high vacuum @t @x @x
(10–7 bar) is first achieved; the gas is then introduced,
followed by heating. The internal pressure of 1 bar is kept In Eq. [1], C is the carbon concentration and D the
constant by an automated gas flow injection. diffusion coefficient. A classical method to take into
The sample can be cooled to 450
C either in vacuum account nonhomogeneous mesh and material properties
or by using a nitrogen flow directed at the sample. (as the diffusion coefficient) is the finite element method,
Standard sample preparation has been used: grinding to as applied in Reference 13. However, implicit time
decrease the thickness to the maximum one allowed by integration of the diffusion equation requires a mini-
the furnace, followed by mechanical polishing (final step mum time-step, which is function of the size of the
is 1-lm diamond suspension). element.[14] The time-step of finite element simulation
has to be large enough such that the diffusion length
pffiffiffiffiffiffiffiffiffi
C. Carbon Profile Measurements during a time-step ( DDt, Dt is the length of a time-
step) is larger than the typical length of one element. If
The carbon profiles are measured using GDOES. the diffusion length is smaller than the length of one
Carbon profiles in the range of 5 to 100 lm are investi- element, the numerical solution is perturbed by oscilla-
gated using a Spectruma GDA 750 device (Spectruma tions, while too large a time-step is detrimental to the
Analytik GmbH, Fabrikzeile 21, D-95028, Hof). Succes- accuracy. Due to these drawbacks of the finite element
sive layers of the surface are eroded (sputtered) by an Ar- method, an original method based on a finite difference
plasma. The light emitted from the atoms of the sample is scheme has been developed. This method is original for
analyzed using an optical emission spectrometer. The a finite difference scheme, because it is implicit and can
GDOES delivers an intensity vs time profile that has to be take into account irregular grid and nonconstant diffu-
sion coefficients. It is based on the computations of
Table II. Gas Composition (Mole Percent) flows and mass balance in an irregular grid. Consider
Figure 1. Around point i, where the concentration is
Gas CO H2 N2 computed, the points i – 1 and i + 1 are at different
1 4 pct 4 pct 92 pct
distances. The carbon flows on the left- and right-hand
2 10 pct 4 pct 86 pct sides of i (Eq. [2]) allow computation of the mass
balance in the element centered on i (Eq. [3]).

1170—VOLUME 38A, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS A

 From the values at time t + hDt, the values at time
t + Dt are extrapolated (Eq. [8]):
Ci ðt þ h
DtÞ  Ci ðtÞ
Ci ðt þ DtÞ ¼ Ci ðtÞ þ ½8
h

This integration scheme is unconditionally stable if

0.5 £ h £ 1.
Fig. 1—Schematic representation of a grid used for the computation
of the diffusion profile using a finite difference scheme.

 III. METHODOLOGY VALIDATION

Ci þ Ci1 Ci  Ci1
Ji1;i ¼ D

Si1;i A. Profile Measurement before Carburization
2 xi  xi1

 A surface carbon contamination is always present due
Ci þ Ciþ1 Ciþ1  Ci
Ji;iþ1 ¼ D

Si;iþ1 ½2 to exposure of steel to ambient air. In order to measure
2 xiþ1  xi this effect, carbon profiles are obtained on both steels
prior to any treatment. Concentration profiles are
acquired for C, as well as for other elements such as
dCi 1  
Si, Cr, Mn, and Ni. The surface contamination is seen
¼
Ji1;i  Ji;iþ1 ½3
dt Vi by the rapid decay of the C profile up to 2 lm below the
surface. For the Fe-C-Mn steel, steady levels for Fe,
In Eqs. [2] and [3], Ji–1,i is the carbon flow between Mn, Cr, and C close to the nominal composition are
points i – 1 and i, and Si–1,i is the surface between the achieved for a depth greater than 2 lm. In the Fe-Ni
elements centered at i – 1 and i. The term Ci is the alloy, the carbon concentration reaches the bulk con-
concentration at point i, xi is the position of point i, and centration at a depth varying between 2.5 and 5 lm.
Vi is the volume associated with point i. By changing the Because the purpose of this study is to perform
subscripts to i + 1 (i – 1, respectively), the same con- carburization experiments, to measure the diffusion
cept is applied to the point or element at the right (left, profile, and then to simulate numerically the experi-
respectively) side. For a one-dimensional problem, Eqs. ments, the determination of an interface determined
[2] and [3] become equivalent to Eq. [4]. experimentally and its correspondence with the bound-


 ary conditions for the computations is of primary
dCi 2 Ci þ Ci1 Ci  Ci1 importance.
¼
D

dt xiþ1  xi1 2 xi  xi1 From these preliminary measurements, however, we

  conclude that the influence of a carburization treatment
Ci þ Ciþ1 Ciþ1  Ci

D
½4 will not be significant until a minimum depth is reached.
2 xiþ1  xi
In the further analysis, concentration profiles are ana-
lyzed for depths higher than 5 lm.
The system is then integrated with a Crank–Nichol-
son scheme.[15] Equation [4] can be written in the form
of Eq. [5]. B. Carburization of Fe-C-Mn
1. Experiments
dCi
¼ a
Ci1 þ b
Ci þ c
Ciþ1 ½5 Carburization experiments on the Fe-C-Mn alloy are
dt performed with the gas 1 given in Table II. The
experiments are carried out in the austenitic range at
By integration of Eq. [5] from the time t up to the time 900
C, well above the Ae3 point. Different holding
t + hDt, one finds Eq. [6]. times are used: 90, 120, 150, 300, and 600 seconds. It
takes about 20 seconds to heat from 700
C to 900
C.
Ci ðt þ h
DtÞ ¼Ci ðtÞ þ h
Dt The time spent between 900
C and 700
C during
Z
tþh
Dt cooling is 5 seconds. Heating and cooling times are kept

a
Ci1 þ b
Ci þ c
Ciþ1 dt ½6 short compared to holding times in order to avoid any
diffusion from the surface during heating or cooling.
t
After treatment, the samples are analyzed by GDOES.
An example of a concentration profile for different
This allows us to write Eq. [7] when it is assumed that elements is shown in Figure 2. The different alloying
the coefficients are constant within the time interval, elements, except carbon, have a flat concentration profile
which defines a linear system to find the composition at for a depth of 5 lm below the surface. For carbon, the
time t + hDt. profile reaches the bulk value at a depth around 50 to
60 lm. The carbon profile has an unusual shape: starting
 h
Dt
a
Ci1 þ ð1  h
Dt
b
Ci1 Þ from the surface, it first increases, reaches a maximum,

Ci  h
Dt
c
Ciþ1 ¼ Ci ðtÞ ½7 and then decreases with two different slopes.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 38A, JUNE 2007—1171

 or equivalently that the gas/metal reaction is not fast
enough compared to bulk diffusion. As the interface
concentration is not equal to the constant value given by
a thermodynamic computation, a specific method has to
be developed. The experiments are numerically simu-
lated with DICTRA to validate the interpretation of the
diffusion profile in zone III. The boundary condition at
the center of the material is a zero-flow condition. The
boundary condition at the other boundary (external
surface) requires more attention: instead of imposing the
boundary condition at the surface of the material, we
focus on what is happening at a certain depth below the
surface of the material. For this purpose, the boundary
condition in the simulation will be the measured carbon
content in the material at a depth of 12 lm below the
surface. This boundary condition evolves in time,
starting from the bulk concentration at time 0 seconds,
Fig. 2—Concentration profile in the Fe-C-Mn steel after carburiza-
tion at 900
C during 150 s.
to the experimental concentrations at 12 lm after 90,
120, 150, and 300 seconds.
In Figure 3, the carbon profiles of all experiments are The results of the numerical simulation are shown in
represented. We distinguish three zones in the carbon Figure 4. The agreement between simulated and exper-
profiles. In the first one, the carbon concentration imental profiles is satisfactory. For the longest time,
increases (I), then it decreases with a high slope (II), and there is a deviation between the desired boundary
finally it keeps decreasing but with a lower slope (III). conditions and the one assumed by the program
It is not clear why the carbon profile has this shape. (difference of concentration between simulated and
The goal of these carburization experiments on the Fe- measured concentration at 0 lm). Using this time-
C-Mn steel is to validate the methodology and to check dependent boundary condition at a certain depth in
the modeling method. In our opinion, stages I and II of the sample, we are able to simulate the experimental
the diffusion profiles could be related to the segregation profiles. This methodology is thus valid to analyze the
profile of substitutional elements such as Mn and Si and carburization experiments on the Fe-Ni alloy, for which
to carbon diffusion occurring during the austenite-to- the diffusion coefficient is not known.
ferrite phase transformation during cooling. Stage III
should be more representative of the diffusion occurring
IV. EXPERIMENTAL RESULTS ON Fe-Ni
during the carburization treatment.
GRADE
2. Numerical simulation of the carburization A. Carburization of Fe-Ni
experiments
Although preliminary equilibrium computations with Experiments are performed at 900
C, 800
C, 700
C,
Thermocalc indicate that cementite should be present at 600
C, and 500
C with gas 2 (Table II). For each
the gas/metal interface, our experimental results did not temperature, three carburization times are used. The time
show that the surface was covered by cementite. This of the tests is increased as the temperature decreases.
implies that the gas/metal interface is not at equilibrium,

Fig. 3—Carbon concentration profiles in the Fe-C-Mn steel for the Fig. 4—Comparison between experimental and computed profiles for
different holding times. the Fe-C-Mn steels after carburization at 900
C.

1172—VOLUME 38A, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS A

 The carbon profiles are shown from Figures 5, 6, 7.
Every carbon profile for temperatures above 500
C
shows two slopes: a very steep decay close to the surface,
which is also found in a nontreated sample and a second
one much more gradual over a longer depth. The switch
between these two segments is always at a depth less
than 5 lm. As expected, the carbon concentration
profile takes in most cases higher values for longer
annealing times. However, some experiments at inter-
mediate times lead to profiles that do not fit with the
profiles for the shortest time and for the longest time
(700
C, 1800 seconds and 600
C, 5400 seconds). In
these cases, these profiles at intermediate times have not
been taken into account in this study as they are
considered as unsuccessful experiments. We note that
one diffusion profile is sufficient to fit the diffusion
coefficient if the boundary conditions are constant. The Fig. 7—Carbon concentration profile after carburization treatment
experiments at intermediate times are mainly used to at 600
C for 2700 and 6200 s and at 500
C for 3600 and 7200 s.
assess the evolution of the boundary condition as a The crosses correspond to the numerical simulation results after
optimization of the diffusion coefficient.

function of time; it is expected that one point at

intermediate time does not have a major influence on
the final results. The carbon concentration at the
interface increases when the temperature is decreased.
At constant time, the diffusion length is decreased when
the temperature decreases.
During the experiments at 500
C, graphite is formed
at the surface. The presence of a growing layer of
graphite does not allow us to consider the concentration
at a depth of 5 lm as a boundary condition for the
simulations. The concentration of interest for the
simulation is the concentration at 5 lm below the steel
surface, which is different from 5-lm depth. The steel
surface is set at a depth where the nickel signal reaches
90 pct of its maximum value (Figure 8).

Fig. 5—Carbon concentration profiles after carburization treatment

at 900
C for 90, 300, and 600 s. The crosses correspond to the B. Simulation and Optimization of the Diffusion
numerical simulation results after optimization of the diffusion coef- Coefficient
ficient.
The carburization process is numerically simulated
in order to compute the diffusion coefficient. The last

Fig. 6—Carbon concentration profiles after carburization treatment

at 800
C for 300, 600, and 900 s and at 700
C for 1200 and 2400 s. Fig. 8—Profiles of Fe, Ni, and C after carburization treatment at
The crosses correspond to the numerical simulation results after 500
C for 3600 s. The concept of interface determined by the level
optimization of the diffusion coefficient. of the Fe signal is illustrated.

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 38A, JUNE 2007—1173

one is computed by minimizing the difference between
the experimental and computed diffusion profiles. The
boundary conditions are a function of time and are
interpolated between the experimental points for
different times at the same distance of the interface
of the material. For each temperature, one optimized
diffusion coefficient is obtained. Because the carbon
concentration remains in a narrow range [0 to 0.4 wt
pct C], the carbon diffusion coefficient is supposed to
be constant. The results of the simulation are given in
Figures 5 through 7 and show good agreement with
the experimental results.

V. DISCUSSION
A. Assessment of the Diffusion Coefficient
for the Fe-C-Mn Steel at 900
C
The same procedure used to assess the diffusion
coefficient of C in the Fe-Ni alloy is used to determine Fig. 9—Comparison of the value estimated from this study with the
the diffusion coefficient of C in the Fe-C-Mn steel. experimental value of Bose[16] and estimates from the formulas pro-
posed by Ågren.[7]
The results are summarized in Table III. Next to the
average measured value, the equation proposed by
Ågren[7] has been used with two concentrations,
15;050
D ¼ 1:23
106
e T ðm2 /sÞ ½9
namely, the bulk concentration and the concentration
at the boundary condition after the longest time. As In the last expression, the temperature is expressed in
these two concentrations are the lowest and the Kelvin. Using a gas constant R of 8.3144 J mol–1 K–1,
highest values, they give a range of values expected an activation energy of 125.132 kJ mol–1 is found.
for the diffusion coefficient. The new experimental Considering the limited concentration range over
value is lower than the expected one from the model which the diffusion data have been collected, it is rea-
of Ågren, but the agreement remains very good. sonable to assume that this formula is valid over the
Considering this comparison, the relative accuracy interval [0 to 0.2 wt pct C].
of the diffusion coefficient value should be within Values extrapolated from Eq. [9] are lower than the
50 pct. results of Bose and Grabke.[16] The ratio between the
experimental values of Reference 16 and Eq. [9] is
around 2.5. The reason for this discrepancy is not clear
at present.
B. Assessment of the Carbon Diffusion Coefficient
The diffusion coefficient of carbon in austenite is
for the Fe-Ni Alloy
mainly of critical importance in low-alloyed steels,
The values of the diffusion coefficient as a function of where different phase transformations and precipitation
the temperature are given in Table IV. The results phenomena occur. Therefore, the extrapolation of the
follow an Arrhenius law, as seen from Figure 9. A very present results to lower substitutional element content is
good fit of the different experimental points is obtained now discussed. In Reference 7, the diffusion coefficient is
by the following equation: given as

Table III. Diffusion Coefficient Measured for the Fe-C-Mn Steel at 900 ˚ C and Comparison with Estimates from the Literature for
the Concentration at the Interface and in the Bulk of the Material[7]

Experimentally Determined Computed Diffusion Coefficient Computed Diffusion Coefficient

Diffusion Coefficient Using the Interface Concentration Using the Bulk Concentration
6.2Æ10–12 m2/s 9Æ10–12 m2/s 7.2Æ10–12 m2/s

Table IV. Diffusion Coefficients Measured for the Fe-Ni Steel and Comparison with the Trivedi’s Average Based on the Concen-
tration at Interface and in the Bulk

Temperature (
C) 900
C 800
C 700
C 600
C 500
C

–12 –13 –13 –14
2
This study (m /s) 3.5Æ10 9.2Æ10 2.2Æ10 4.6Æ10 4.1Æ10–15
Agren (average) (m2 /s)[7] 6.6Æ10–12 1.8Æ10–12 3.5Æ10–13 4.7Æ10–14 2.8Æ10–15
Bhadeshia (average) (m2 /s)[8] 5.4Æ10–12 1.4Æ10–12 2.7Æ10–13 3.4Æ10–14 2.4Æ10–15

1174—VOLUME 38A, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS A

 concentration [bulk, boundary condition] using Eq. [14]
Dcc ðm2 /sÞ has been reported in Figures 9 and 5.


7 8339:9
¼ 4:53
10 1 þ yc ð1  yc Þ ZCBC
T 1

  Dcc ¼  BC 
Dcc ðcÞdc ½14
1 4 C  Cbulk
 exp   2:221
10 ð17; 767  yc 26; 436Þ Cbulk
T
½10 The agreement between Eq. [13] and the new exper-
imental data is fair. The ratio between the experimental
In this expression, the first bracket describes the data and the computed values ranges within [0.5, 2].
influence of thermodynamics on the diffusion coefficient, However, the experimental data are lower than values
while the other components describe the mobility of predicted by Eq. [13] at 900
C and 800
C, while the
carbon atoms in an austenitic lattice, as shown by the inverse relationship occurs at 500
C and 600
C.
equivalent formulation Comparing Eqs. [9] and [13], one notes that the

 activation energy for low carbon concentration is higher
2
LFe in Eq. [13] than in Eq. [9]. Considering that Eq. [13] is
Dcc ¼ Vm
R
T
MCVa
1 þ yc ð1  yc Þ CVa
½11
R
T based on extrapolation from a higher temperature
range, the agreement between the new assessment and
It is possible to take into account the thermodynamic existing modeling is more than satisfactory.
influence of the fraction of substitutional elements on For computations of phase transformations of low-
the diffusion coefficient. Using a sublattice model,[17] an alloyed steels, we thus recommend the use of the
equivalent mixture term is written as expression proposed by Ågren for the following reasons.
(1) Even though it is fitted to high-temperature data,
LCVa ¼ xNi
LNi Fe
CVa þ xFe
LCVa ½12 this expression gives a reasonable agreement with
the present new experimental results.
In the article of Hillert and Qiu,[18] no value is (2) This expression is fitted to binary steels and takes
reported for the LNi
CVa parameter. Therefore, this param- into account the dependence of carbon on the dif-
eter is expected to be zero. fusion coefficient, while the expression based on
In the article of Bose and Grabke, there is an the present experimental study is essentially carbon
estimation of the influence of Ni on the mobility of concentration independent.
carbon in the fcc phase at higher temperatures (950
C
to 1050
C). According to their results, the mobility of
carbon changes only slightly as a function of the
amount of nickel. They found that the mobility VI. CONCLUSIONS
changes by 50 pct in the composition range 20 to Carburization experiments are performed in the
100 pct Ni. If one considers a narrower Ni-content austenitic range for Fe-1.5 pct Mn 0.13 pct C and Fe-
range (20 to 40 pct Ni), the variation of the mobility 31 pct Ni alloys. Carbon composition profile measure-
is smaller than 10 pct. Therefore, it is expected that ments using GDOES show that the surface composition
results achieved with the 31 pct Ni alloy in this is not constant with time. A methodology is proposed to
study can be extrapolated with good confidence to assess the diffusion coefficient of carbon in austenite
lower Ni contents. It is of great importance to extend using the carbon profiles measured experimentally. This
the present findings to low Ni content steels, and methodology uses a numerical method to compute the
more specifically to steels undergoing a phase trans- diffusion profile and takes into account the time-
formation at lower temperatures. This gives a new dependent boundary condition. This methodology is
equation: first validated on the Fe-C-Mn steel, where it gives a
diffusion coefficient of 6.2Æ10–12 m2 /s at 900
C, com-
Dcc ðm2 /sÞ


 pared to the 7.2 to 9Æ10–12 m2 /s values reported in the
7 8339
yFe literature. Using the preceding methodology, experi-
¼ 4:53
10 1 þ yc ð1  yc Þ
T ments are carried out on a Fe-31 pct Ni alloy at 900
C,

  800
C, 700
C, 600
C, and 500
C. The following
1
 exp   2:221
104 ð17767  yc 26436Þ equation is able to describe correctly the experimental
T results (T in Kelvin):
½13 15;050
D ¼ 1:2296
106
e T ðm2 =sÞ
Values computed from Eq. [5] are compared with the
experimental results of this study (Figure 5 and 9). The The difference between this expression and measure-
diffusion coefficient in Eq. [5] is dependent of the carbon ments previously published for alloys with comparable
concentration. As the concentration varies over the alloying content is within a factor of 2. The discrepancy
concentration gradient, an average over the interval of between the present experimental results and the mod-

METALLURGICAL AND MATERIALS TRANSACTIONS A VOLUME 38A, JUNE 2007—1175

ified Ågren model remains within a factor of 2, which is 6. D. VanDooren, P. Thibaux, and B. De Cooman: in Modeling
very good agreement. Therefore, we recommend using Control and Optimization in Ferrous and Nonferrous Industry, F.
Kongoli, B.G. Thomas, and K. Sawamiphakdi, eds., TMS,
the expression proposed by Ågren in diffusion problems Warrendale, PA, 2003, pp. 481–93.
in the temperature range 500
C to 900
C, because it 7. J. Ågren: Sripta Metall., 1986, vol. 20, pp. 1507–10.
takes into account the influence of carbon concentration 8. H.K.D.H. Bhadeshia: Met. Sci., 1981, vol. 15, pp. 477–79.
on the diffusion coefficient. 9. C. Wells, W. Batz, and R.F. Mehl: Trans. AIME J. Met., 1950,
vol. 188, pp. 553–60.
10. S.S. Babu and H.K.D.H. Bhadeshia: J. Mater. Sci. Lett., 1995,
vol. 14, pp. 314–16.
11. S. Crusius, G. Inden, U. Knoop, L. Höglund, and J. Agren: Z.
Metalloved., 1992, vol. 83, pp. 673–78.
12. A. Dinsdale: Calphad, 1991, vol. 15, pp. 317–425.
REFERENCES 13. J. Agren: J. Phys. Chem. Solids, 1982, vol. 43, pp. 385–91.
14. Aliaga, Ph.D. Thesis, Ecole des Mines de Paris, Paris, 2000. .
1. A. Van der Ven and L. Delaey: Progr. Mater. Sci., 1996, vol. 40, 15. J. Crank: Mathematics of Diffusion, Oxford Science Publications,
pp. 181–264. Oxford, United Kingdom, 1975.
2. D. Quidort and Y.J.M. Brechet: Acta Mater., 2001, vol. 49 (20), 16. S.K. Bose and H.J. Grabke: Z. Metallkd., 1978, vol. 69, pp. 8–15.
pp. 4161–70. 17. M. Hillert and L.-I. Staffansson: Acta Chem. Scand., 1970, vol. 24,
3. M. Hillert: Scripta Mater., 2002, vol. 47 (3), pp. 137–38. pp. 3618–26.
4. H.K.D.H. Bhadeshia and D.V. Edmond: Acta Metall, 1980, vol. 18. M. Hillert and C. Qiu: Metall. Trans. A., 1991, vol. 22A, pp. 2187–
28, pp. 1265–73. 98.
5. O. Bouaziz, D. Quidort, and P. Maugis: Rev. Metal.-Cahiers
D’Inform. Techn., 2003, vol. 100 (1), pp. 103–08.

1176—VOLUME 38A, JUNE 2007 METALLURGICAL AND MATERIALS TRANSACTIONS A