Surface and Coatings Technology 120–121 (1999) 358–365

www.elsevier.nl/locate/surfcoat

Mathematical modeling of a carburizing process

of a SAE 8620H steel
H. Jiménez a, M.H. Staia b, E.S. Puchi b, *
a Siderúrgica del Orinoco S.A., Matanzas, Puerto Ordaz, Edo. Bolı́var, Venezuela
b School of Metallurgical Engineering and Materials Science, Faculty of Engineering, Central University of Venezuela, Apartado Postal 47885,
Los Chaguaramos, Caracas 1045, Venezuela

Abstract

Pack carburizing experiments have been conducted employing samples of an SAE 8620H steel. Three different carburizing
mixtures have been used, each with a different content of carbonate, added to a metallurgical coke as catalyst. The change in
carbon concentration along the case depth has been determined quantitatively by means of chemical analysis conducted at every
0.1 mm from the outer surface of the workpiece. It has been shown that the carbon concentration profile can be satisfactorily
modeled employing the classical solution derived from Fick’s second law. For this purpose, a two-step optimization procedure
has been developed in order to compute, as a first step, the optimum values of the activation parameters (activation energy for
diffusion and the diffusivity coefficient) of carbon in austenite, as well as the carbon potential that is established at the surface of
the carburizing specimen under different processing conditions. The second step of the algorithm involves the re-computation of
the carbon potential at the surface, once the activation parameters are assumed to be constant. In general, it has been observed
that the addition of carbonates (BaCO and NaCO ) to the metallurgical coke gives rise to an increase in the carburization rate
3 3
and case depth which allows the achievement of the required carbon concentration profiles more efficiently. High carburization
temperatures are observed to give rise to undesired acicular structures in the workpiece at low temperatures, due to the austenite
grain growth during the carburizing processing. © 1999 Published by Elsevier Science S.A. All rights reserved.

Keywords: Carburizing process; SAE 8620H steel

1. Introduction produced. The carbon-rich layer or case can be subse-

Steel parts and components have been case-hardened
at industrial scale by carburizing for many years and
therefore it has become an important commercial pro-
cess, which is particularly true for both gas and liquid
carburizing treatments [1,2]. The latest advances in gas
carburizing and nitriding technology including develop-
ments in processes, process modeling, alloys, equipment
and control have been reported recently [3]. Also,
plasma processing is currently being employed in com-
mercial carburizing practice and the principles involved
have been discussed by Grube and Verhoff [1].
In general, during a carburizing process a low carbon
steel is heated at elevated temperatures, usually in the
range of 1173–1323 K, in the presence of a carbon-rich
medium. Such a process gives rise to the diffusion of
carbon into the steel and therefore a carburized layer is
* Corresponding author. Tel./fax: +58-2-6628927.
E-mail address: epuchi@reacium.ve ( E.S. Puchi)
quently hardened either by reheating and quenching or
by quenching directly from the carburizing temperature.
The final desired hardness can then be achieved by
tempering. In contrast to gas and liquid carburizing,
pack carburizing is a minor commercial process which
can make use of different furnaces since it produces its
own contained environment.
Furthermore, it allows the slow cooling of the work-
piece from the carburizing temperature and offers a
selection of stop-off techniques for selective carburizing.
However, this process is not well suited to the production
of shallow case depths which might require strict toler-
ances and, obviously, it is not well suited for direct
quenching. Additionally, it requires more processing
time. Pack carburizing is typically applied for the pro-
duction of different case depths in a wide variety of
components including different types of gears, crane-
cable drums, drive pinions, motor-brake and crane
wheels, leveler and processor pinch rolls and blooming-
mill screws.

0257-8972/99/$ – see front matter © 1999 Published by Elsevier Science S.A. All rights reserved.
PII: S0 2 5 7- 8 9 7 2 ( 9 9 ) 0 0 46 4 - 8
 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365
Table 1
Chemical analysis of the experimental alloy (wt%)
C Mn Si Cr
0.22 0.74 0.22 0.46
Depending on the particular component, the case
depth to 50 HRC can vary from 0.6 to 6.9 mm.
According to Foreman [2], during pack carburizing the
carbon monoxide derived from the carburizing com-
pound or mixture decomposes at the metal surface into
nascent carbon and carbon dioxide. The fresh carbon
formed in this process is absorbed into the metal,
whereas the carbon dioxide reacts with the carbonaceous
material present in the carburizing mixture giving rise
to more carbon monoxide. It has also been pointed out
that the formation of carbon monoxide is enhanced by
the different carbonates added as catalysts which facili-
tates the reduction of carbon dioxide with carbon to
form carbon monoxide.
The present investigation was conducted in order to
study the carburizing process of an SAE 8620H steel
currently employed in the manufacture of processor
pinch rolls, by means of pack carburizing, employing
three different carburizing mixtures based on metallurgi-
cal coke and different proportions of carbonates. The
study was conducted over the temperature range of
1123–1223 K and carburizing times ranging between 2
and 6 h. It involves the determination of the carbon
concentration profile along the case depth and its
description by means of the classical equation derived
from the solution of Fick’s second law, that is to say,
without taking into consideration any dependence of
the diffusion coefficient on carbon concentration.
Thus, a numerical procedure is proposed in order to
determine the optimum values of the diffusion coefficient
and the carbon potential at the surface of the workpiece
that give rise to the minimization of the quadratic mean
difference between the experimental data and their theo-
retical description.
2. Experimental procedure
The present investigation was conducted with speci-
mens of an SAE 8620H steel whose mean chemical
analysis (eight samples) is given in Table 1. C and S
were determined employing a Leco CS-46 analyzer,
whereas the remaining elements were obtained using a
Quantovac ARL analyzer. The material was produced
at SIDOR (Siderúrgica del Orinoco, S.A., Venezuela) in
an electric arc furnace with a capacity of 150 tons.
Ingots of 10 tons were cast, reheated and subsequently
hot rolled to billets of 230 mm×230 mm of cross-
section. After hot rolling, the stock was reheated again
and further rolled starting at 1513 K. Slabs of 186 mm
359
Ni Mo P S Al Fe
0.46 0.18 0.013 0.013 0.022 bal.
thickness were finally obtained and from one of them a
sample of 700 mm in length was cut in order to perform
the present study.
The hardenability of the material was evaluated by
means of the Jominy End Quench Test, employing three
samples machined with their axes parallel to the rolling
direction of the slab, according to the ASTM standard
A255. Brinell hardness measurements ( Wolpert) were
conducted along the cross-section of the samples
employing a load of 30 kg and an indenter of 1 mm
diameter. The metallographic analyses of the samples
were conducted (Leitz MM-6) in order to evaluate both
the inclusion level of the material, according to the
ASTM standard E45, and the austenitic grain size
following the ASTM standard E112 which employs a
comparative method. Both analyses were carried out on
the short transverse section of the slab.
Samples of 10×40×60 mm were machined for pack
carburizing, ensuring that the 60 mm length was parallel
to the rolling direction and the 40×60 mm face of the
specimens was parallel to the long transverse section of
the slab. The specimens were maintained carefully in
order to avoid any oxidation of the surfaces. Pack
carburizing boxes of 90×90×110 mm were made using
stainless steel sheets of 5 mm thickness. Three different
pack carburizing mixtures were prepared employing
metallurgical coke (84–86% C, 11–13% ashes, 1% max
S and 0.01% max P) with a particle size of 3–6 mm
diameter, BaCO of 99.5% purity and NaCO of 99%
3 3
purity.
The composition of the mixtures employed is pre-
sented in Table 2. The mixtures were prepared by pre-
heating the coke at 353 K and subsequently adding the
liquid solutions of carbonates, increasing the temper-
ature up to 383 K and keeping it for 90 min in order to
eliminate the humidity. The pack carburizing process
was performed by filling in the boxes with the corre-
sponding mixtures up to a height of 25 mm and subse-
quently placing two samples in a vertical position with
a separation of 25 mm. Afterwards, the boxes were
completely filled in with the mixture and sealed off.
Table 2
Composition of the pack carburizing mixtures used in the present
study (wt%)
Composition Mixture A Mixture B Mixture C
Coke 90 100 85
BaCO 5 – 10
NaCO
3
5 – 5
3
360 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365

For every mixture, three boxes were introduced in an

electric furnace at three different temperatures: 1123,
1173 and 1223 K. The boxes were removed from the
furnace after 2, 4 and 6 h, respectively. The carbon
profile after carburizing was determined from chips
obtained at 0.1 mm intervals from the outer face of the
samples until the chemical analysis was found close to
the carbon composition of the base steel. The machining
process was carried out carefully without lubrication.
The determination of the carbon content of the case
layer was conducted in a Leco CS-46 analyzer, taking
the mean of three samples and the value obtained was
then reported in relation to the mid point of the
cemented case, according to the ASTM standard G79-C.
The carbon profile of the cemented case was also
evaluated by means of SEM techniques (Phillips 505)
employing a WDX analyzer (Microspec WDX-2A). In
this case, the change in carbon content was evaluated
linearly from the outer surface to the center of the
samples. The microhardness profile was determined
employing samples of 10×10 mm prepared metallo-
graphically according to the ASTM standard E3, etched
with nital 2%. This preparation allowed the microstruc- Fig. 2. Carbon concentration profile obtained with the carburizing
compound designated as mixture A, at 1173 K. The solid curve repre-
tural observation of the cemented case layer according sents the theoretical description obtained by optimization of the
to the ASTM standard G79-B. parameters C and D.
S
The microhardness profile was obtained employing a
Knoop indenter and conducting microhardness tests at
every 0.1 mm from the edge of the cemented samples to
the center, according to the ASTM standard E384.

3. Results and discussion

The photomicrograph of Fig. 1 illustrates the typical

microstructure of the experimental alloy employed in
the pack carburizing treatment. It consists of a mixture
of acicular ferrite and pearlite characteristic of a hot
rolled structure. The evaluation of the austenitic grain
size allowed to determine an ASTM value between 5

Fig. 3. Carbon concentration profile obtained with the carburizing

compound designated as mixture B, at 1173 K. The solid curve repre-
sents the theoretical description obtained by optimization of the
parameters C and D.
S

and 6. Such a grain size as well as the level of non

Fig. 1. Optical photomicrograph illustrating the initial microstructure metallic inclusions and hardenability behavior deter-
of the 8620H steel employed in the present investigation. mined from the Jominy test were found to fulfill the
 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365
Fig. 4. Carbon concentration profile obtained with the carburizing
compound designated as mixture C, at 1173 K. The solid curve repre-
sents the theoretical description obtained by optimization of the
parameters C and D.
S
requirements of the SAE standard for this material.
Figs. 2–4 illustrate, as an example, the carbon profiles
determined by means of chemical analysis of the chips
machined from the specimens, after carburizing at
1173 K for 4 h with the three mixtures employed.
The carbon profiles determined at 1123 and 1223 K
have been omitted due to space confinement. These
profiles were qualitatively corroborated by means of
SEM and WDX measurements. As expected, as the
carburizing time increases the case depth also increases
up to values that range between approximately 1.5–
2 mm. Furthermore, at a constant carburizing temper-
ature and time, the case layer is deeper as the carbonate
content of the carburizing compound increases. This
fact corroborates the catalytic action of these additives
reported in the literature [2], regarding the reduction of
carbon dioxide to form carbon monoxide.
The carbon concentration is observed to decrease
smoothly from the surface to the interior of the work-
piece, with the trend to achieve asymptotically the
carbon content of the steel. The solid lines shown in
these figures correspond to the theoretical description
of the experimental data by means of the classical
expression determined from the solution of Fick’s second
law, of the form:
x/(2앀Dt)
G P
2 the evaluation of the parameters C and D at some
Ĉ(x, t)=C +(C −C ) 1− exp−y2 dy
0 S 0
앀p (1)
0 The iterative procedure for determining the optimum
where Ĉ(x, t) represents the carbon content at any
361
distance from the outer surface of the carburizing
sample, C the carbon concentration or carbon potential
S
at the surface, C the carbon concentration of the steel
0
and D the diffusion coefficient of carbon in austenite
which in turn is given by:
D=D exp(−Q/RT ). (2)
0
Here, D represents the diffusivity coefficient, Q an
0
experimental activation energy for the diffusion of
carbon in austenite and R and T their usual meaning.
This theoretical description does not take into
account the dependence of the diffusion coefficient of
carbon on concentration, as it has been reported in the
earlier work of Batz and co-workers [4]. According to
this study, at a temperature of 1400 K, for carbon
concentrations above approximately 1 wt%, the diffu-
sion coefficient increases very rapidly, and for concen-
trations in the order of 1.5–1.6 wt%, this parameter
becomes more than twice the value achieved below
1 wt%.
Regarding this issue, Romig and co-workers [5] have
dealt with the numerical modeling of several phenomena
controlled by volume diffusion in which it is often
necessary to solve the diffusion equation under initial
and boundary conditions for which an analytical solu-
tion does not exist, including cases where the initial
composition profile is neither constant nor a simple
error function. Among the numerical algorithms that
have been employed in this context are implicit and
explicit finite differences, lines and finite element meth-
ods. In the present investigation, the fit of Eq. (1) to
the experimental data has been conducted by determin-
ing the value of the constants C and D that minimize
S
the function:
N
Q= ∑ {C(x, t) −Ĉ(x, t)}2 (3)
i
i=1
which represents the quadratic mean difference between
the experimental values of the carbon concentration
profiles and those predicted from the solution of Fick’s
second law. The optimization of the parameters C and
S
D has been conducted by means of the Newton–Gauss
method, which involves the expansion of Eq. (1) in
terms of a series of two variables, of the form:
C D
∂Ĉ(x, t)
Ĉ(x, t)=[Ĉ(x, t)] + (C −C )
0 ∂C S S0
S 0
C D
∂Ĉ(x, t)
+ (D−D ) (4)
∂D 0
0
where the square brackets and the subscript zero imply
H
S
initial value reasonably close to the optimum solution.
values of C and D involves the computation of the
S
362 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365

Table 3
Optimized values of carbon diffusivity and carbon potential at the
surface computed for different carburizing temperatures and expo-
sures times

Diffusivity×10−6 Carbon Temperature Time

(mm2s−1) potential (wt%) (K) (s)

1.70 1.00 1123 7200

7.26 0.97 1123 14 400
9.50 1.12 1123 21 600
8.33 0.84 1173 7200
8.37 1.01 1173 14 400
14.9 1.12 1173 21 600
14.5 0.80 1223 7200
10.5 0.98 1223 14 400
14.5 1.16 1223 21 600
0.92 0.57 1123 7200
2.17 0.65 1123 14 400
3.48 0.89 1123 21 600
8.54 0.71 1173 7200
10.6 0.79 1173 14 400
7.11 0.95 1173 2 1600
8.99 0.73 1223 7200
5.20 0.94 1223 14 400 Fig. 5. Comparison between the experimental values of carbon concen-
12.4 0.97 1223 21 600 tration and the predicted ones obtained from the theoretical curves
2.09 0.91 1123 7200 after optimization of the parameters C and D.
S
8.14 0.97 1123 14 400
9.53 1.08 1123 21 600
8.45 0.89 1173 7200 atures for the different mixtures, it is possible to compute
15.7 0.98 1173 14 400 both the diffusivity coefficient and the activation energy
27.1 1.12 1173 21 600 for the diffusion of carbon in austenite, by means of
17.2 0.85 1223 7200
23.1 0.97 1223 14 400
Eq. (2).
34.9 1.26 1223 21 600 Fig. 6 illustrates an Arrhenius-type plot showing the
change in the diffusion coefficient with the inverse of

these parameters by means of linear regression analysis

applied to the above expanded equation and to employ
the values obtained as the initial approximation for the
next iteration until convergence is achieved.
Table 3 presents the results of the computation of
both D and C for the different conditions of carburizing
S
temperature and exposure time. For the set of experi-
mental data shown in Figs. 2–4, the only condition for
which major deviations are observed between the experi-
mental points and the optimized curve are those corre-
sponding to mixture C at 6 h of carburizing time. The
remaining conditions are described quite accurately by
Eq. (1). Fig. 5 illustrates a general comparison between
the predicted values of carbon concentration as function
of the depth case and those obtained experimentally
under all the experimental conditions explored in this
work. As shown in these figures, the overall fit of the
experimental data is satisfactory which demonstrates
both the ability of Eq. (1) for the description of such
data and the suitability of the optimization technique
that has been employed, in spite of the fact that it has
been assumed that the diffusion coefficient of carbon
does not depend on carbon concentration. Thus, once
the values of D have been determined from the experi- Fig. 6. Arrhenius plot of the inverse of the logarithm of the diffusion
mental data at different carburizing times and temper- coefficient versus inverse of the absolute carburizing temperature.
 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365 363

Table 4
Re-computed values for the carbon potential at the surface for different
conditions of carburizing temperature and exposure time. Q=
151 kJmol−1 and D =44.7 mm2s−1
0
Carbon potential (wt%) Temperature ( K ) Time (s)

0.82 1123 7200

1.07 1123 14 400
1.28 1123 21 600
0.84 1173 7200
1.01 1173 14 400
1.24 1173 21 600
0.78 1223 7200
0.90 1223 14 400
1.14 1223 21 600
0.43 1123 7200
0.58 1123 14 400
0.86 1123 21 600
0.71 1173 7200
0.82 1173 14 400
0.92 1173 21 600
0.66 1223 7200
0.74 1223 14 400
0.92 1223 21 600
0.78 1123 7200
1.09 1123 14 400
1.24 1123 21 600 Fig. 7. Carbon concentration profile obtained with the carburizing
0.89 1173 7200 compound designated as mixture A, at 1173 K. The solid curve repre-
1.09 1173 14 400 sents the theoretical description obtained by optimization of the
1.34 1173 21 600 parameter C .
S
0.86 1223 7200
1.03 1223 14 400
1.44 1223 21 600

the absolute temperature. At each carburizing temper-

ature there are nine values of the diffusivity coefficient
which correspond to three exposures times for the three
mixtures employed. Therefore, the activation parameters
have been calculated from a total of 27 data points by
means of least square analysis and it has been determined
that Q=151 kJ mol−1 and D =44.7 mm2 s−1. The value
0
of Q computed in the present work is somewhat higher
than that reported by Verhoeven [4] of approximately
134 kJ mol−1 and agrees reasonably well with that
reported by Shewmon [5] of approximately
145 kJ mol−1. The value of D presently calculated is
0
also somewhat higher, though of the same order of
magnitude of those reported by the above cited authors
of approximately 12 and 27 mm2 s−1, respectively.
Fig. 6 clearly illustrates that the diffusion coefficient
at each carburizing temperature varies significantly,
which could be attributed both to the uncertainties Fig. 8. Carbon concentration profile obtained with the carburizing
involved in its computation from the experimental data compound designated as mixture B, at 1173 K. The solid curve repre-
and also to an actual dependence on carbon concen- sents the theoretical description obtained by optimization of the
tration thus far ignored. If the values thus determined parameter C .
S
of Q and D are re-applied to Eq. (1), it is possible to
0
calculate again the value of the parameter C for every carbon potential at the surface under different carburiz-
S
carburizing treatment and for the three different mix- ing conditions. A procedure similar to the one described
tures. Table 4 reports the re-computed values of the above can be employed, taking into consideration that
364 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365
Fig. 9. Carbon concentration profile obtained with the carburizing
compound designated as mixture C, at 1173 K. The solid curve repre-
sents the theoretical description obtained by optimization of the
parameter C .
S
in this second step of the algorithm, the expansion of
Eq. (1) will involve just one variable.
Figs. 7–9 illustrate the new fit of Eq. (1) to the
experimental data maintaining constant the diffusivity
parameters D and Q. Thus, it can be observed that
0
although most of the carburizing conditions are still
described satisfactorily, some of them particularly for
carburizing times of 6 h, present a significant difference
between the experimental points and the theoretical
prediction. Fig. 10 summarizes the final comparison of
the experimental values of carbon concentration with
those expected from the theoretical description provided
by Eq. (1) and the activation constants calculated under
the present conditions.
In general, it has been determined that the carbon
potential at the surface varied between approximately
0.57 and 1.26. The smallest value of the carbon potential
was obtained for mixture B, at 1123 K and 2 h of
carburizing time. This carburizing compound does not
contain any addition of catalysts, whereas the largest
value of such potential was computed for the data
corresponding to mixture C at 1223 K and 6 h, which
contains the maximum amount of catalysts (15 wt%).
Therefore, the carbon potential at the surface is
closely related to both carburizing temperature and to
the amount of carbon monoxide available for the carbu-
rizing process whose formation, as pointed out earlier,
is enhanced by the presence of BaCO and NaCO
3 3
employed as catalysts. Thus, the addition of carbonates
Fig. 10. Comparison between the experimental values of carbon con-
centration and the predicted ones obtained from the theoretical curves
after optimization of the parameter C .
S
to the metallurgical coke as energizers together with
elevated carburizing temperatures and exposure times
guarantee the achievement of the expected hardening
level for those applications in service where both high
surface hardness and wear resistance are required.
The resulting microstructure of the case layer was
observed to consist of a hypereutectoid structure at the
surface constituted by a mixture of pearlite within a
cementite net, which progressively transforms into an
eutectoid structure followed by a hypoeutectoid struc-
ture of ferrite and pearlite, as the case depth increases.
At 1223 K and 6 h of exposure time, the acicular struc-
ture of the workpiece was accentuated as a result of the
significant increase in the austenitic grain size that occurs
at this temperature.
4. Conclusions
The carbon profiles along the case depth of the pack
carburized samples has been satisfactorily modeled
employing the classical solution derived from Fick’s
second law. A two-step optimization procedure has been
developed in order to compute the optimum values of
the activation parameters (activation energy for diffusion
and diffusivity coefficient) of carbon in austenite as well
as carbon potential that is established at the surface of
the carburizing specimen under different conditions of
temperature, time and coke-catalysts mixtures. An acti-
vation energy of 151 kJ mol−1 was determined for the
diffusion of carbon in austenite, which agreed well with
the values reported in the literature. The carbon concen-
 H. Jiménez et al. / Surface and Coatings Technology 120–121 (1999) 358–365
tration at the surface or carbon potential determined
from the model varies between 0.57 and 1.26. The
addition of carbonates (BaCO and NaCO ) to the
3 3
metallurgical coke as catalysts as well as elevated carbu-
rizing temperatures and exposure times were shown to
be very effective in obtaining the required carbon con-
centration profiles that provide an appropriate surface
hardening of the workpiece. However, a careful control
of the austenitic grain size must be exercised in order to
avoid the formation of acicular structures that would
impart poor mechanical properties to the carburized
material.
Acknowledgements
Dr. E.S. Puchi acknowledges the financial support of
the Venezuelan National Council for Scientific and
365
Technological Research (CONICIT ) through the project
LAB-97000644.
References
[1] W.L. Grube, S. Verhoff, Plasma (ion) processes for case hardening
of metals, in: R. Kossowsky (Ed.), Surface Modification Engineer-
ing, Vol. II: Technological Aspects, CRC Press, Boca Raton, FL,
USA, 1989, p. 107.
[2] R.W. Foreman, Pack carburizing, in: J.R. Davis et al. (Eds.), ASM
Handbook Vol. 4: Heat Treating ASM International, ASM
International, Materials Park, OH, 1991, p. 326.
[3] J. Grosch et al. ( Eds.), Proceedings of the International Conference
on Carburizing and Nitriding with Atmospheres, ASM Interna-
tional, Materials Park, OH, 1995.
[4] W. Batz, R.F. Mehl, C. Wells, Trans. A.I.M.E. 188 (1950) 553.
[5] A.D. Romig Jr, N.Y. Pehlivanturk, O.T. Inal, in: A.D. Romig Jr,
M.A. Dayanada (Eds.), Diffusion Analysis and Applications, The
Minerals, Metals and Materials Society, Warrendale, PA, 1989,
p. 45.