Fe-C-Si: Modelling
Fe-C-Si: Modelling
714-722
Jacques LACAZE
and Val~rie GERVAL
Laboratoire des Mat6riaux. ENSCT,UPRESA-CNRS
5071 .
31 077 Toulouse cedex, France. E- maii: jlacaze@ensct.fr
Theso-called eutectoid reaction of spheroidai gra phite cast irons (SGl) proceeds by competitive nucleation
and growth of ferrite and pearlite, In the present study are first reviewed the physical models of the ferritic
reaction in SGI previously described in the literature. Then, a new model is presented that uses a recent
description of the conditions to be fulfilled for the ferritic and pearlitic reactions to start. This description
is based
upon the knowledge of the relevant phase diagram. Growth of the ferrite halos during the ferritic
reaction is described as controlled by carbon transfer from the austenite/ferrite interface to the graphite
nodu[es and by an interfacial reaction at the ferrite/graphite interface. Modelling of the pearlitic reaction
accounts for nucleation of pearlite colonies, and their growth law is expressed according to experimental
data available in the literature. It
appeared also necessary to describe the diffusion of carbon in austenite
before the beginning of the decomposition of this phase. Predictions are compared to experimental
transformation kinetics obtained by meansof differential thermal anaiysis on spheroidal graphite Fe-C-Si
alloys, and cou[d be easily extended to alloys with low [evel additions of pearlite promoter elements.
KEYWORDS: spheroidal graphite cast irons; Fe-C-Si alloys; eutectoid reaction; transformation kinetics;
physical modelling.
l. Introduction
The effects of nodule count, of the cooling rate (in the
range I to 20 K/min), and of low level addition of Mn
Spheroidal graphite cast irons (SGI) are made of and/or Cu, have been studied previously by differential
graphite nodules embeddedin an iron base matrix. The thermal analysis (DTA).10) The characteristic tempera-
most common types of these alloys are mainly Fe-C-Si tures at the beginning of the transformation, in the stable
ternary alloys, with low level of Mg added as a nodu- and metastable systems, as well as the kinetics of austenite
larizer. The matrix is austenitic at high temperature; decornposition, were evaluated from the DTAsignal. It
upon cooling to room temperature after solidification was observed that the decomposition of austenite starts
or heat treatment, austenite decomposeselther to ferrite at a temperature which is well below the upper limit of
and graphite (in the stable system), or to pearlite (in the triphasic austenite/ferrite/graphite or austenite/fer-
the metastable system), or else to both of them. The rite/cementite domain. For both the stable and metasta-
control of the actual microstructure of the material at ble transformations, the undercoolings observed
were of
room temperature is of someimportance as the number the same magnitude than those reported by Ekpoom
of nodules and the ratio of the ferrite and pearlite and Heinell) who used also DTA. In order to deepen
fractions determine the service properties of the cast the understanding of the solid-state transformation of
components.1) It seems established that higher nodule SGI, a second series of experiments was carried out
counts and lower cooling rates (in the temperature with samples which were held at Ih I
100'C so as to
range of the eutectoid transformation) both favor the smooth out microsegregations.12) Such a treatment did
decomposition of austenite in the stable system.2) On in fact eliminate the extrema of the solute distributions
the other hand, adding low level of specific elements rather than actually homogenizethe matrix. The trans-
such as Cu, Mn, Sn or Sb, is a practical meansto obtain a formation temperatures were slightly modified with
pearlitic matrix.3~5) The ferrite to pearlite ratio respect to the untreated material, but the effect of the
aiso depends, but to a lesser extent, on microsegrega- cooling rate on the transformation kinetics remained
tion, i.e. chemical heterogeneities at the level of the unchanged.
microstructure, which build up during the solidification From the experimental fact that growth of ferrite
stage.6~8) Finally, in case of heat treatment, the aus- during cooling of SGI is not associated with partitioning
tenitizing temperature affects the eutectoid reaction be- of substltutionnal elements (Si, Cu. Mn, ') between ' '
cause of the change of the solubility of carbon content ferrite and austenite, we proposed to follow the de-
in the parent austenite with temperature.9) composition of austenite in the stable system in an
@1998 ISIJ 71 4
ISIJ International, Vol. 38 (1998), No. 7
materialis below the lower temperature of the triphasic
~)
F,
c:,
domain, named T* in the following. Using the tem-
(,
perature T. as the reference temperature for the
~
~ Y
Q)
ee+?
ferritic reaction, it was found that the undercooling
H 7+g
increases with the cooling rate, as expected, and drops
Ta to zero whenthe data is extrapolated to a zero cooling
\
\ rate.12,13) This latter observation gives someconfidence
oe a+7+g \
Toc
\ to the above description of the growth conditions during
l \ the ferritic reaction.
oe+g \
1
Aw l
\
\ Anearly similar reasoning was applied to the for-
"~'~ mation of on the basis that no redistribution of
pearlite
I ll,
l \
substitutional elements between pearlite and austenite
l l l has been reported in the case of cast irons. This leads to
a/g ce/? Y/ce
wc ~c Carbon content wc the condition that pearlite cannot grow at temperatures
Fig. l. Schematic isopleth FeCsection of the relevant stable higher than the lower limit of the corresponding three-
phase diagram in the temperature range corresponding phase austenite/ferrite/cementite field.12) The reference
to the eutectoid reaction. Solid lines represent the limits
temperature for the beginning of the pearlitic reaction,
of the phase fields and broken lines the metastable
extrapolation of the ferrite/austenite (o(/^/) and aus-
named Tp, was selected as the intersection of the ylcc
tenite/ferrite (y/ce) boundaries. extrapolation with this iimit. In addition, it is expect-
ed that pearlite grows in SGI as it does in steels. Thus,
it is worth mentionlng that the model for isothermal
71 5 O 1998 ISIJ
ISIJ International, Vol. 38 (1998). No. 7
Venugopalanl9) summarized the main steps of the iso- austenlte in Eq. (2), Ieaving everything else unchanged.
thermal decomposition of austenite to ferrite and graph- The growth rate of the ferrite shell was then given by:
ite as follows:
(1) nucleation of occurs at the austenite/
ferrite (vv y/' w~/y) dr "
Dc
aw6
..........(5)
graphite interface. An incubation period is associated
c - dt ar ..
with this step which is attributed to the limited kinetics Equation (5) is similar to the equations previously derived
of nucleation. In his study, Venugopalanconsidered that in the works devoted to isothermal ferritization.18,21)
the incubation period actually corresponds to the time
The models developed by Chang et al.22) and by
at which a ferrite nucleus has formed and has grown to Svensson et al.23) are very much alike the mode] of
an observable volume fractlon, typically o/o. The author
I Venugopalan20) but the three following points:
pointed out that the development of this nucleus is (1) nucleation of
Instantaneous, it occurs
ferrite is
attended by diffusion of carbon in the parent austenite. at a temperature which has to be experimentally de-
However, for modelling purpose, Venugopalanassumed termined. Changet al. madeuse of records while DTA
that all the nodules are instantaneously surrounded by Svensson and Wess6nused the line defining the start of
a ferrite shell whenthe incubation time is reached. The the transformation on the CCT
diagram of their material.
thickness of this shell is such that the volume fraction These two methods appear to be very similar to each
of ferrite is equal to the initial value chosen. other, they are indeed a way to handle the very large
(2) growthofferritelsattendedbyrejectionofcarbon temperature difference betweenthe observed temperature
which diffuses to the graphite nodules and into un- at the beginning of the decomposition of austenite and
transformed austenite. The mass conservation condi- 7~ which was used as the reference temperature by these
tion for carbon at the austenite-ferrite interface is given authors. As described in the introduction, part of this
by: discrepancy is associated with the fact that the reference
temperature should be T* and not T~-
p"(w cy/' vt'~/v) dr " =p"D~ aw~
PyD~
aT,v~
...(2) (2) Changet al. replaced Eq. (3) glving the growth
- dt a,. ar
.. ** rate of the graphite nodule by an approximate mass
where prp the density of phase q) (either ferrite:
is o( or balance, while Svensson et al. simply did not account
austenite: D~is the diffusion coefficient of carbon in
y), for graphite growth. These differences are not significant
phase (p, w~/' and w~/y are the mass fraction of carbon because the growth of the graphite phase is in any case
respectively in austenite and in ferrite at the austenite/ very limited.
ferrite interface, w8 is the mass fraction of carbon in (3) Changet al. and Svensson et al. assumedthat
phase ip, and r" is the radius of the ferrite envelop. the composition of the austenite at the ferrite/austenite
The growth rate of the graphite nodule is given by the interface follows the metastable extrapolation of the ocly
carbon massbalance at the ferrite-graphite interface: boundary, i,e. that there is local equilibrium between
these phases.
w~/g)
d,' g aw~ Since their preliminary works, Svensson and Wess~n
= p"D~
g(1 •(3)
p - dt ar
• •• • •• • ••• • •
where w~/g is the massfraction of carbon in ferrite at the position of the austenite to ferrite and graphite.24) They
ferrite/graphite interface and rg is the radius of the proposed to consider three successive steps:
graphite nodule. (1) the formation of ferrite halos which is controlled
It is generally accepted that diffusion of carbon through by diffusion of carbon in the parent austenite as Venu-
the ferrite shell maybe expressed as a stationary process, gopalan stated but did not describe. The authors as-
i.e. that it is possible to
express the carbon profile in the
sumed that ferrite nucleates 20 below T"o and that K
the thickness of the halos which develop from these nuclei
ferrite halo as w~ a+b/r, where the constants a and
= b
are to be determined from the value of the carbon content
is constant
as long as the halos are not totally formed.
in the ferrite halo at the two lirniting interfaces. This (2) the growth of the ferrite shells controlled by an
gives for the gradient in carbon content in ferritel9)' interfacial reaction at the graphite/ferrite interface which
was found to depend on (rg)~ l
awc" r "A wc aw~ r gAl't'c (3) the growth of the ferrite shells controlled by
= and = (4)
diffusion of carbon through the ferrite phase.
. . . .
ar .*
,'g(,'"-rg) a,'
..
,'"(r"-1"g)
O 1998 ISIJ 71 6
ISIJ International, Vol. 38 (1998), No. 7
e)
f~
diffusion of carbon in austenite and that local equilibrium ~
holds at the austenite/graphite interface. The growth rate
1,
H ,b,
Il~l
- Toa- -
II'l
J~llt~ Toe-
of the graphite nodules Is given by: 800
l~ ~ ~ -
'~
'~
drg aw~ ferrite + graphite
'b Itb
pg( I - = PvD~
.(6)
T4'~/g) l~l
. .. . . .. . .
dt ar **
750
where w~/g js the massfraction of carbon in austenite at
0.5 0.55 0.6 0.65 0.7 0.75 o8 0.85 0.9
of the austenite phase is shifted to higher carbon contents 3.2. The Ferritic Reaction
with respect to the austenite/graphite boundary, although As previously proposed,24) it is considered that the
this shift decreases, as expected, when the cooling rate
is decreased. Thus, the transformation of the bulk aus-
growth of graphite inside the ferrite shell
maybe partly
controlled by an interfacial reaction. On other hand,
the
tenlte to ferrlte cannot proceed before the path fol- 10cal equilibrium is assumed at the ferrite/austenite
lowed by I~~ has reached the metastable extrapolation interface, so that the carbon content in each phase at
of the austenite/ferrite boundary. The corresponding this interface is given by the extrapolation of the rele-
temperature will be noted Th in the following. For the vant boundary in the Fe-C section. The value of Awc
particular composition consldered and for the two which drives the transfer of carbon atoms from the fer-
selected cooling rates, it is noted that Th is above T.. It rite/austenite to the ferrite/graphite interface is thus
is worth to emphasizethat the formation of the halos as
Awc=w~/y _ w~, where w~ rs the actual carbon content
described by Wess~nand Svensson24) could not begin in ferrite at the interface with graphite. The growth rate
before the temperature of the alloy is below Th; the reason of ferrite is obtained by applying Eq. (2) below T.. Actual
is that it needs diffusion of carbon in austenite which growth of ferrite is however subjected to the condition
cannot proceed as long as w~/' is lower than I~)~. Also, that the growth rate given by Eq. (2) is positive. The
according to the present calculation, the temperature composition gradients in Eq. (2) are expressed by:
difference T;~- Th for the nucleation of ferrite depends
w~/'
on the coo]ing rate and cannot be considered as a
aw~ r gAT4'c
and aw~ ,~~
-
constant. Moreover, the thickness of the halos should ar _ r"(r" 1'g) ar
.* - ..
.(8)
increase during their formation, particularly when the
71 7 C 1998 ISIJ
ISIJ International, Vol. 38 (1 998), No. 7
Io'
drg 640
c= -pg(1 -w~)
dt """
"~"""(9) o,1 Io
growth rate (um/s)
c= - K"p"(~4'c w~/g)2 for the mterfaclal reactron and grow. Both steps have been considered. Nucleation
.(10) was assumedto be described by the following law:
where K" is a constant which characterizes the interfacial dN~= Ap(N~ - N~)( I - gt)(A Tp)"'d(A Tp)
al4'~
dNvP=0 forATp
c= -p"D~ ar
for the diffusion process ....(ll)
O 1998 ISIJ 71 8
ISIJ International, Vol. 38 (1998), No. 7
(g"*
+ gP ) 1
9~ (1 g)
K/minL.
Further indications on the procedure followed to make 0.8
Ho 06
LH
eo
F::
a~
4. Results l:~
o
~ 04
Thecalculations described below were intended at c:'
w
FL1
were heated to I
100'C for one hour and then cooled at
a controlled rate down to 600'C, then reheated to 900'C
for 10 min before being cooled downto 600'C at another o
680 700 720 740 760 780 800
rate and reheated again to 900'C for 10min and cooled Temperature ('C)
at a different rate. Twosamples will be considered here
Fig. 5. Comparison of calculated (dotted lines) and experi-
(the notation used is sample Cwas
that of Ref. 12)): (i)
mental (solid curves of transformed austenite in
lines)
cooled successively at 10, 5 and I K/min, and had a D
the case of sample with a nodule count of 330 ~ 2,
mm
nodule count of 190mm~2;(ii) sample D was cooled cooled from 900'C at 2 and lOK/min. Calculations
successively at 20, 10 and 2K/min, and had a nodule were madewith the selected value of D~.
count equal to 330mm~2.
The heat transfer in the DTAcell was simulated by Th Wasnot correctly predicted, i.e. that the change of
meansof a simp]e model described in the appendix. A the carbon content in austenite was not conveniently
numberof calculations were performed to investigate the described.
effect
on the kinetics of the ferritic reaction of the two In fact, Hillert27) showed in the case of white cast
input data, namely the kinetics coefficient K", and the irons that the graphitization process is not simply con-
initial thickness of the ferrite halo, 8r". This was done trolled by volume diffusion of carbon in austenite; as
by cornparing the calculated curves with the experimental the size of the graphite nodules increases, the transfer of
ones. The value of K" affects the shape of the curves carbon is successively controlled by diffusion of iron,
during the whole transformation while the value of ~r" creep of the matrix and finally diffusion of carbon. The
acts only at the beginning of the ferritic reaction. It first two steps
are related to transformation rates much
was thus possible to find a good estimate of K" by lower than the one corresponding to the third step, by
comparing the slope of the transformation curve at a factor in between five to ten.28) It was thus consider-
intermediate values of gt. The value finally selected was ed that the transfer of carbon could be estimated by
4.0 x l0~2 Km~1.
Then, it was found convenient to set considering a corrected diffusion coefficient of carbon in
~r" equal to 0.1 um. In the calculations to be presented austenite, and it was found that Dyc should be divided
below, the value used for the constants of the nucleation by lO. The path followed by 'i)~ during cooling at and I
law of pear]ite were estimated from the experiments at 10 K/min thus calculated is shownwith solid lines in Fig.
the highest cooling rates:
m= 3and Ap= lO~ I m~3 K~4 3. It is seen that the sensitivity to the cooling rate is
increased and that the temperature Th is lower than in
but they cannot be considered to be firmly established.
Figure 5 compares the evolution of the fraction of the former calculations. At the lOK/min cooling rate,
transformed austenite as calculated (dotted lines) and as the curve ends before the y/oc boundary because the
experimentally estimated (full lines) upon both coolings pearlitic reaction begins.
of sample D
from 900'C. The calculated curves show a Figures and 6 7
present the calculations of the
shift towards lower temperature when the cooling rate transformation kinetics madefor samples (Fig. 5) and C
is increased. Moreover,
we can see that the effect of the D (Fig. 6) with the corrected value of the diffusion
cooling rate on the kinetics of phase transformation is coefficient of carbon in austenite. In both figures, we
satisfactorily described. However, the curves in Fig. 5 notice that there is a shift to higher temperatures of most
show that the effect of the cooling rate on the onset of of the calculated curves with respect to the experimental
the austenite decompositlon is not properly described. ones. This shift increases with the cooling rate, it is related
At first, one thought that this could be due to the to the simplicity of the heat transfer model of DTA
nucleation step of ferrite which was not described. A described in the appendix. In fact, there is a delay be-
model accounting for it was developed in which the size tween the beginning of the reaction in the sample and the
of the eutectoid spheres was calculated for all the size recording of the corresponding thermal arrest. This delay
classes defined by the successive nucleation events. It is of the order of
O
(the time constant characteristic of
was found that when the start of the reaction was the transfer function between the sample and the ther-
appropriately described, the bulk transformation was mocouple) at the beginning of the transformation
not. In other words, the discrepancy related to the onset and then decreases. The value of was found of the O
of the transformation could not be due to the nucleation order of 60 slo) for the DTAapparatus used in this study.
kinetics of ferrlte. It becameevident that the temperature Using this value, the predicted shift of the curves for a
71 9 C 1998 ISIJ
ISIJ International, Vo!. 38 (1998). No. 7
5. Conclusron
0.8
2K/min
1:'
(900 'C)
A model of the eutectoid reaction during cooling of
,L)
~
,4~~o
10 K/min Fe-C-Si alloys with spheroidal graphite has been
0.6
eo
(900 'C) presented. It is based on a description of the growth
f::
ce conditions of ferrite in the stable system and of pearlite
F::
~~
physical laws which take into account an interfacial re-
0.2 action at the ferrite/graphite interface as well as volume
diffusion of carbon through the ferrite halos. Thepearlitic
reaction accounts for nucleation and growth of pearlite
o colonies. The characteristics of the nucleation law of
680 700 720 740 760 780 800
pearlite have been determined so as to fit the experimen-
Temperature ('C)
tal data while the growth rate of pearlite
was expressed
Fig. 7. Comparison of calculated (dotted lines) and experi-
according to experimental data available in the litera-
mental (solid curves of transformed austenite in
lines)
ture. To reproduce the effect of the cooling rate at the
D
the case of sample with a nodule count of 330 2.
mm~ beglnning of the austenite decomposition, it appeared of
Calculations were madewith the corrected value of
D~. The relevant cooling rate is indicated on each curve, first importance to describe the redistribution of carbon
with the starting temperature between brackets. between austenite and graphite before the beginning of
the eutectoid reaction. Calculations showed that this
10K/min cooling rate is at most lOK at the beginning stage should be controlled by stress effects and it was
of the transformation, in good agreement with the dis- found necessary to use an apparent diffusion coefficient
crepancy observed in Figs. 5 and 6, Nevertheless, ow- of carbon in austenite. In this line, work is going on in
ing to the various simplifications introduced in the order to characterize the possible effect of stresses and
physical model of the eutectoid transformation, as well strains on the ferritic reaction.
as in the heat transfer model, Figs. 6 and 7 show that Acknowledgement
the effect of the cooling rate on the onset of the
This work has received a financial support from
transformation as well as on the overall kinetics is
P6chiney-Electrom~tallurgie.
conveniently reproduced by the model proposed.
The present model has been developed and illustrated
with Fe-CSi alloys with spherical graphite, but could REFERENCES
be easily extended to alloys with low level additions of l) R. Elliott: Cast lron Technology. Butterworths et Co., London,
(1988).
promoter elements. There are however two main
pearlite
2) D. R. Askeland and S. S. Gupta: AFST,'a,Is., 83 (1975). 217.
drawbacks that should be emphasized. First, it is ex- 3) E. N. Pan, M. S. Lou and C. R. Loper: AFSTrans., 87 (1979),
pected that microsegregation of substitutional solutes 819.
25) M. Castro: PhDthesis, INPL, Nancy, France, (1991). selected by Liu and Agren33) and Agren34) respectively:
26) J. Lesoult. M. Castro and J. Lacaze: Acta Metal!. Mater., 46
l7767
27)
(1998), 983.
M. Hillert: Je,'nkonto,'ets Ann., 141 (1957), 67.
D~= 2343 10 5 exp - T
28) C. E. Birchenall and H. W. Mead: J. Met., (1956), 1004.
29) V. Gerval and J. Lacaze: Proc, of the Conf. SP97, ed. by J. Beech 10115
D~=2.0'10~6exp
and H, Jones, 506 - T
30) R. C. Voigt and R. C. Loper: AFST,'ans., 97 (1989), 595.
31) B. Sundman.B, Jansson and J. O. Andersson: Calphad, 9(1985),
O5898 l+-2 arctan 15Tc
629 15309 )J}
150 exp
T
. •
7c
32) B. Uhrenius: Hardenability Concepts with Application to Steel,
The Met. Soc. of AIME, (1977), 28.
33) Z. K. Liu and J. Agren: Acta Meta!!., 37 (1989), 3157.
where Tc is the Curie temperature, which is here calcu-
34) J. Agren: Acta Metall., 30 (1982), 841.
lated according to the following approximate formulae
35) I. Barin: Thermochemical Data of Pure Substances, VCH, which was evaluated from the binary Fe-Si diagram:
Weinheim, (1993). Tc (K) = I 043 I OOO ws~'
36) A. T. Dinsdale: NPLReport DMA(A)195, (1989). -
Heat Transfer Calculation
DTAexperiments were performed on a SETARAM
2000
Appendix. Thermophysical Data set up with alumina as reference body. The sample and
Description of the Phase Diagram the reference were placed in alumina crucibles high 7mm
The following expressions, where the Si content in and 4mm of Inner diameter. For modelling heat transfer,
austenite, wsi, is given in mass fraction, and the
it
was assumedthat the temperature of the alloy sample
temperature T in Celsius, have been obtained by fitting and of the reference are homogeneous.The heat balance
to calculatlons madewith the Thermocalc software31)
equation for the samplewaswritten in the following way:
using the data of Uhrenius.32)
pmVC
dTs ~Vdg
T2+0 125ws,
*
P dt = -
Ah(Ts~ Tf) + p AH (A-1)
w~/'
O1876-4.ll2•l0~4T+226 10 7 dt .....
m .
equal to 58 kJ kg~ I for both the stable and metastable
.
I
as 185 +0.1617 > (T- 500) J . kg~ I K- I which fits rea-
' transformations.
sonably well the assessed data35) In the temperature