Acta Materialia 149 (2018) 68e77

Contents lists available at ScienceDirect

Acta Materialia
journal homepage: www.elsevier.com/locate/actamat

Full length article

Carbon enrichment during ferrite transformation in Fe-Si-C alloys

H.-D. Wu a, *, G. Miyamoto b, Z.-G. Yang a, C. Zhang a, H. Chen a, T. Furuhara b
a
School of Materials Science and Engineering, Key Laboratory for Advanced Materials of Ministry of Education, Tsinghua University, Beijing, 100084, China
b
Institute for Materials Research, Tohoku University, Aoba-ku, Sendai, Miyagi, 980-8577, Japan

a r t i c l e i n f o a b s t r a c t

Article history: The microstructure evolution, transformation kinetics and especially the C enrichment during isothermal
Received 14 December 2017 ferrite transformation at 800, 750, 700
C for Fe-(1.5, 3.0) mass%Si-0.4mass%C alloys were investigated in
Accepted 14 February 2018 the present study. Both transformation kinetics and C enrichment in austenite (g) at grain boundary
Available online 2 March 2018
ferrite (GBF)/g interfaces suggest that negligible partitioning local equilibrium (NPLE) mode operates
during non-partitioned ferrite growth under the conditions investigated. Deviations of C enrichment in g
Keywords:
at Widmanstatten ferrite (WF)/g interfaces from NPLE limit at early stages were observed. These de-
Pro-eutectoid ferrite
viations are deduced to be resulted from intrinsic interface mobility rather than solute drag effect. In
Carbon enrichment
NPLE limit
addition, mobility of WF/g interfaces is found to be lower than that for incoherent GBF/g interfaces
Interface mobility reported in literature.
Si containing steels © 2018 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
EPMA

1. Introduction NPLE/PLE transition temperature (also called NPLE limit) while that
for PE model is para-Ae3 temperature in this study.
Austenite (g) to ferrite (a) transformation in steels has been one Both NPLE and PE models are idealized models and consider
of the most widely studied transformations due to its industrial and “equilibrium” condition at the interface. In the actual case, how-
theoretical importance [1]. In alloyed steels, diffusivities of substi- ever, some phenomenon related to interface reaction may cause
tutional elements (e.g. silicon (Si), manganese (Mn), denoted as X) additional energy dissipation and thus interfacial condition de-
and interstitial elements like carbon (C) usually differ by several viates from the “equilibrium” state. The two well acknowledged
orders of magnitudes. At relatively large undercooling from Ae3 phenomenon are intrinsic friction [5e7] and solute drag effect
temperature, non-partitioned a growth occurs, namely growth [8e10].
without macroscopic partitioning of X between a and g, and thus it Si is a common alloying element known as a a stabilizer and
is controlled by C diffusion in g. Time needed for growth of non- inhibiting cementite precipitation. Compared with Fe-Mn-C
partitioned a is usually comparable to the time scale for heat ternary system, investigations on a transformation in Fe-Si-C
treatment employed industrially and thus non-partitioned a ternary system is less sufficient. Guo and Enomoto [10] made a
transformation is of special interest in industry. Two models, i.e. comparison on a growth kinetics between NPLE, PE predictions and
negligible partitioning local equilibrium (NPLE) model [2,3] and the experimental ones by in Fe-0.12%C-1.47%Si [11] (mass%, here-
paraequilibrium (PE) model [4], have been proposed to describe the after) and in Fe-0.40%C-1.73%Si [9] and concluded that the
a/g interfacial condition during non-partitioned a growth. NPLE measured a parabolic constant agreed reasonably well with both
model assumes local equilibrium (LE) at the migrating interface NPLE and PE models' predictions. However, the difference between
where C and X have the same chemical potential in a and g. This two predictions was too small to be discriminated within experi-
would result in a thin spike of X in g at a/g interface during a mental error. Recent decarburization studies [12] on steels with Si
growth. PE model assumes that LE condition is held only for content ranging from 0.45%Si to 1.46%Si found that the a growth
interstitial elements while the ratio of X and Fe is a constant kinetics is slightly slower than PE and NPLE predictions and the
throughout the system. The critical temperature for NPLE model is deviations were interpreted to be resulted from solute drag effect.
On the other hand, van Landeghem et al. [13] observed no Si
segregation or even Si depletion at C enriched g side of a/g in-
* Corresponding author. Room 2516, Yifu Technology and Science Building,
terfaces for Fe-0.84%Si-0.76%C alloy decarburized at 775
C which
Tsinghua University, Haidian, Beijing, 100084, China. was not in agreement with suggestions from solute drag effect. The
E-mail address: [email protected] (H.-D. Wu). understanding on the role that Si addition plays during a

https://doi.org/10.1016/j.actamat.2018.02.040
1359-6454/© 2018 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77 69

transformation is far from satisfactory and further investigation is

necessary.
Recently, Liu et al. [14] and Xia et al. [15] measured the C content
in g at a/g interface with field emission electron probe micro-
analysis (FE-EPMA) technique in Fe-Mn-C based alloys during a
transformation and made insight discussions. FE-EPMA was also
utilized in present study to directly measure C enrichment in g with
Fe-(1.5, 3)%Si-0.4%C alloys at several temperatures in the aim of
clarifying the regime operating during non-partitioned a growth
stage as well as the effect of intrinsic friction and solute drag effect.
In addition, microstructural evolution and transformation kinetics
were also included for the purpose of comparison and verification.
Note that present authors has utilized similar techniques to study
incomplete baintie transformation phenomenon with Fe-(1.5, 3)%
Si-0.4%C alloys and highlighted the limitations of current theories
[16].

2. Experimental

Two high purity Fe-(1.5, 3)%Si-0.4%C alloys were used in this

study. Hereafter they are referred to as 1.5Si and 3Si alloys. The
exact compositions together with the calculated ortho-Ae3, para-
Ae3, NPLE/PLE transition and ortho-Acm temperatures of two al-
loys are listed in Table 1. The temperatures as well as other ther-
modynamic calculations in present study were all performed by
Thermo-Calc with TCFE7 database. The alloys ingots were pre-
pared by vacuum melting and casting, followed by hot rolling to
~15 mm. The two alloys were encapsulated into Ar-filled silica
tubes and homogenized at 1150
C for 24 h, after which no banded
structure could be observed.
Specimens of ~10  5  3 mm were cut from homogenized al-
loys and austenitized in vacuum furnace at 1150
C for 10 min
producing nominal g grain size of 193, 195 mm for the 1.5Si and 3Si
alloys, respectively. After austenitization, the specimens were
immediately quenched into salt bath and isothermally held at 800,
750, 700
C for various time periods after which the specimens Fig. 1. Vertical phase diagram of (a) 1.5Si and (b) 3Si alloys together with the heat
were quenched in iced water. The temperatures utilized here for a treatment conditions shown by circles.
transformation was designed to locate in aþ g two phase regions, as
was shown in Fig. 1.
(EBSD) technique. It should be noted that untransformed g has
Transformed specimens were cut along the plane parallel to the
been transformed to martensite during quenching and thus the g
5  10 mm2 plane into two pieces. Then, cross sections close to the
orientation was reconstructed from the orientation of martensite
center of the specimens were observed to minimize the possible
[17]. EBSD measurements with step size of 0.5 mm on polished
loss of C during heat treatment or the effect of oxidization. Before
specimens was carried out by scanning electron microscopy (SEM;
etching, standard metallurgical polishing method was performed.
JEOL JSM-7001F) equipped with an EBSD (TSL OIM ver. 6) system.
The polished specimens were etched by 3% nital for optical mi-
croscopy (OM) observations. Volume fractions of a or transformed
g during transformation were measured by point counting method. 3. Results
For each specimen, 10 OM images with magnification of 200
(contain ~70 prior g grains in total) were counted and averaged. C 3.1. Effect of ferrite morphologies on C enrichment
and Si profiles across the a/g interfaces were measured by field-
emission electron probe micro-analyzer (FE-EPMA, JEOL JXA- Two morphologies of ferrite have been observed during this
8530F), the procedure of which is described in detail in Ref. [14]. investigation, namely grain boundary ferrite (GBF) formed along
For specimens of interests, the orientation relationship (OR) be- prior g grain boundaries and Widmanstatten ferrite (WF) mainly
tween a and g was characterized by electron backscatter diffraction nucleating at GBF/g interfaces while growing into matrix g grains.

Table 1
Chemical compositions and characteristic temperatures for the two alloys.

Alloy Compositions (mass%) Temperatures (
C)

Fe C Si Mn P S Ortho-Ae3 Para-Ae3 NPLE/PLE Ortho-Acm

1.5Si Bal. 0.375 1.48 <0.01 <0.005 0.0011 847 835 823 593
3Si Bal. 0.379 3.09 <0.01 <0.005 0.0016 935 902 879 631

Bond letters highlight the difference of two alloys.

70 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77

At 800
C, GBF was the only product for the 1.5 Si alloy (Fig. 2(a)) from NPLE limit. On the other hand, C content at WF/g interfaces
while both GBF and WF were observed for the 3Si alloys (Fig. 2(b)). (Fig. 4(e)) were lower than the NPLE limit. At later transformation
As is generally acknowledged [18], formation of WF needs more stage when transformation kinetics slows down, Si partitioning
undercoolings from ortho-Ae3 temperature than GBF formation. could be detected by EPMA measurements. Examples of the 1.5Si
Since Si addition shifts ortho-Ae3 line in vertical phase diagram alloy transformed for 43.2 ks and the 3Si alloy transformed for
upward, the a morphology difference in two alloys at 800
C is due 18.6 ks at 800
C are shown in Fig. 3(d)~(f) and Fig. 4(g)~(i),
to different undercoolings for a formation resulted from different Si respectively. Si is depleted in g at a/g interface since Si is a ferrite
content. At early stage, transformation kinetics for the 3Si alloy is former. In addition, Si-C repulsive interaction results in increment
faster than that for the 1.5Si alloy at 800
C (Fig. 2(c)). Trans- of the C enrichment in g at a/g interface in order to maintain the
formation kinetics clearly slows down when a volume fraction uniformity of C activity in g.
reaches the value predicted by NPLE limit. Fig. 5(a) and (b) summarize the evolution of C content in g at a/g
The C and Si profiles across GBF/g interfaces and WF/g interfaces interfaces for the 1.5Si and the 3Si specimens isothermally held at
were measured. Fig. 3 and Fig. 4 shows typical results for the 1.5Si 800
C, respectively. Each point corresponds to the averaged
and 3Si alloys transformed at 800
C for different periods, respec- measured value at least from 6 a/g interfaces. For the 1.5Si alloy
tively. Carbon is rejected from a regions while no long range where only GBF was observed, the difference between predictions
redistribution of Si atoms occur (Figs. 3(c), Fig. 4(c) and (f)) when given by PE model and NPLE limit is comparable to the experi-
reaction time is short. Figs. 3(b) and Fig. 4(b) clearly show that C mental error (Fig. 5(a)). It is not easy to obtain a safe conclusion on
contents in g at most GBF/g interfaces are close to the predictions whether NPLE limit or PE model gives a better description on the
experimental data during non-partitioned a growth stage of this
condition. For non-partitioned a growth of the 3Si alloy at 800
C, C
contents in g at GBF/g interfaces could be well described by NPLE
limits at holding times less than 1.8 ks (Fig. 5(b)). On the other
hand, C content in g at WF/g interfaces of specimens isothermally
held for 30 s significantly deviates from NPLE limits and approaches
gradually to NPLE limit (Fig. 5(b)). Si redistribution was observed
when further increasing the holding time at this temperature and
was accompanied with increasing of C content in g at a/g interfaces.
The difference of interfacial C between GBF/g and WF/g interfaces
content becomes negligible during this stage.

3.2. Effect of reaction temperature on microstructures,

transformation kinetics and C enrichment

Besides Si content, reaction temperature also plays a vital role

on microstructures as well as transformation kinetics and C
enrichment. For the 1.5Si alloy, WF was observed when lowering
the reaction temperature to 750
C (Fig. 6(a)). Further lowering the
reaction temperature to 700
C results in the formation of more WF
(Fig. 6(b)). In addition, pearlite is formed at this temperature. For
the 3Si alloy at 750
C (Fig. 6(c)), GBF is reduced to a thin layer along
prior g grain boundaries and WF becomes more closely spaced
compared with 800
C (Fig. 2(b)). After prolonged holding, pearlite
was observed at 750
C. Pearlite transformation kinetics becomes
much faster at 700
C (Fig. 6(d)) and the untransformed g is soon
consumed by pearlite resulting in a smaller amount of a formed
from g than 750
C. The change of a microstructure is in accordance
with general understanding on the effect of Si addition and reaction
temperature while effects of Si on competitive transformation be-
tween ferrite and pearlite will be reported elsewhere.
a transformation kinetics of two alloys at 750
C and 700
C are
summarized in Fig. 7. Overall transformation kinetics including
pearlite fraction are together plotted in cases where pearlite is
formed. a transformation kinetics at 750 and 700
C becomes faster
by increasing Si content in accordance with transformation kinetics
at 800
C shown in Fig. 2. The overall transformation kinetics curve
for 3Si alloy exhibits two stages (red open circles in Fig. 7(a)). Only a
is growing in first stage and the second stage mainly involves
pearlite formation. Pearlite starts to form much earlier when
further lowering the reaction temperature to 700
C and two stages
could be no longer found (Fig. 7(b)). It should be noted that a
transformation kinetics for two alloys slows down significantly
when the volume fraction of a reaches the value predicted by NPLE
Fig. 2. OM images of (a) 1.5Si and (b) 3Si alloys transformed at 800
C for 10 min. (c):
limit in cases where interference from pearlite transformation is
ferrite transformation kinetics for two alloys at 800
C. M(A): martensite transformed not significant (Figs. 2(c), Fig. 7(a)).
from untransformed austenite (same meaning in following figures). The C and Si profiles across GBF/g interfaces and WF/g interfaces
 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77 71

Fig. 3. The measured C and Si profiles across a/g interface for 1.5Si alloy transformed at 800
C for (a)(b)(c) 600 s and (d)(e)(f) for 43200 s. (a)(d) are the OM images showing the
lines where EPMA measurements are conducted; (b)(e) and (c)(f) are the measured C and Si profiles, respectively.

at early stages for the 1.5Si alloy at 750
C and 700
C are shown in interfaces at the early non-partitioned a growth stage for the 3Si
Fig. 8. Clearly, Si long range partitioning between a and g did not alloy at 750
C (Fig. 9(b)) and the 1.5Si alloy at 700
C (Fig. 9(c)).
occur (Fig. 8(c)(f)(i)). C contents in g at two GBF/g interfaces at
750
C coincides with NPLE limit (Fig. 8(b)). On the other hand, at 4. Discussions
700
C (Fig. 8(h)), C content at one GBF/g interface is close to NPLE
limit while that at another GBF/g interface deviates from NPLE 4.1. Agreement between the C enrichment at incoherent GBF/g
limits (Fig. 8(h)). Similar with the case of the 3Si alloy transformed interfaces and NPLE limits
at 800
C (Fig. 4(e)), deviation from NPLE limits was observed for
WF/g interfaces at 750
C (Fig. 8(e)) and 700
C. As can be seen from Fig. 1, the temperature range employed for
Fig. 9 summarizes the evolutions of C enrichment in g at GBF/g studying proeutectoid a transformation in present investigation is
and WF/g interfaces for two alloys at 750
C and 700
C. Similar 700e800
C, lying well below PE and NPLE/PLE transition temper-
with the case for the 3Si alloy at 800
C (Fig. 5(b)), C content in g at atures of each alloy. Therefore, proeutectoid a growth would start
GBF/g interfaces for the 1.5Si alloy at 750
C (Fig. 9(a)) clearly with non-partitioned a growth in accordance with EPMA mea-
coincide with NPLE limit during non-partitioned a growth stage surements (see Figs. 3(c), 4(c)(f) and 8(c)(f)). The C content in g at
while that at WF/g interfaces deviates from NPLE limit at early non- GBF/g interface during non-partitioned growth stage could be well
partitioned a growth stage and the deviation gradually diminishes described by NPLE limit (Figs. 5 and 9). These agreements suggest
with increasing reaction time. Redistribution of Si again increases C that a Si spike has been fully developed for specimens when the
content in g at interfaces. For the 3Si alloy at 750
C and two alloys reaction time is as short as 30 s, the shortest isothermal holding
at 700
C, only limited data points at early stage were obtained. It time at three temperatures. In addition, that the experimental
appears that lowering reaction temperature and increasing Si transition of a transformation kinetics from fast to slow occurs
content would increase the difference between measured C when the proeutectoid a volume fractions coincide with pre-
enrichment in g at WF/g interfaces and NPLE limits at a given re- dictions from NPLE limits in several conditions where interference
action time. It should be noted that relatively small deviation from from pearlite transformation is not significant further validates the
NPLE limits is also observed for C enrichment in g at GBF/g operation of NPLE regime during non-partitioned a growth.
72 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77

Fig. 4. The measured C and Si profiles across (a)(b)(c) GBF/g interface, (d)(e)(f) WF/g interface for 3Si alloy transformed at 800
C for 10 min and (g)(h)(i) a/g interface for 3Si alloy
transformed at 800
C for 3 h. (a)(d)(g) are the OM images showing the lines where EPMA measurements are conducted; (b)(e)(h) and (c)(f)(g) are the measured C profiles and Si
profiles, respectively.

Fig. 5. Evolution of interfacial C content in g at GBF/g and WF/g interfaces during ferrite transformation for (a) 1.5Si and (b) 3Si alloys at 800
C.

The width of alloy spikes r could be expressed with r ¼ Dg =v utilized in experiments of each condition are used in order to avoid
[11], where Dg is the diffusion coefficient of the alloying element in soft impingement effect. Because GBF grows into both g matrixes in
g (values for Si diffusivity in g is obtained from Ref. [19]) and v is the general, the migrating distance is assumed to be equal to 0.5dGBF .
migrating rate of a/g interface. Migrating rate of GBF/g interfaces Average vGBF between t1 and t2 was evaluated with corresponding
(vGBF ) were estimated as follows. GBF thickness dGBF
1 and dGBF
2 by,
Conditions where no interference of pearlite transformation or
soft impingement effect is involved were considered. The volume 1 dGBF  dGBF
2 1
fraction of GBF (fGBF ) is firstly counted from the experimental OM vGBF ¼ (2)
2 t2  t1
images. According to Liu et al. [20], average thickness of GBF (dGBF )
could be calculated with following equation, The estimated vGBF as well as calculated r at the beginning
stages of different conditions when soft impingement effect is ab-
pffiffiffi
p sent are summarized in Table 2. It is found that r is smaller than g
dGBF ¼ f $dg (1) lattice constant (~0.36 nm) [21] in most cases. It seems that fully
4 GBF
building up of a Si spike in g matrix is not feasible at the early stage
where dg is prior g grain size. Two shortest reaction times (t1 ; t2 ) of non-partitioned a growth for the conditions employed in present
 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77 73

Fig. 6. OM images of (a)(b) 1.5Si alloy transformed (a) at 750
C for 60s, (b) at 700
C for 30s and (c)(d) 3Si alloy transformed at (c) 750
C for 30s, (d) 700
C for 30s. P: pearlite (same
meaning in following figures).

1.73%Si alloy at temperatures between 725 and 825
C.

Hillert [22] proposed that when r decreases to below atomic
dimensions, the spike will move inside the interface and thus
diffusivity inside the interface should be considered. He later
pointed out [23] that one could use diffusivity of alloying element
in a/g interface, DInt , to replace Dg when using the same expression
to estimate the width of spike in cases where the spike moves in-
side the interface. Due to lack of experimental data on DInt , one may
apply the diffusivity of Fe along grain boundaries [24] to approxi-
mate the magnitude with reasonably assuming the width of
interface d to be 0.5 nm. The spike width r 0 was then estimated
between 1:2  104  2:1  106 nm (Table 2) which is much larger
than d. However, this indicates that it is feasible for the develop-
ment of a Si spike if diffusion in both interface and in austenite are
involved. Similar conclusions have also been reached in Fe-C-Mn
system [10] as well as in Fe-C-Mn-Mo system [15]. Further direct
observation (e.g. by atom probe [13]) on the development of spike
is thus necessary to solid these conjectures. Van Landeghem et al.
[13] has used atom probe to observe segregation of alloying ele-
ments. Si depletion at a/g interfaces was observed in one specimen
of Fe-0.76%C-0.84%Si alloy when being decarburized at 775
C for
16 min while no depletion of Si was found in another specimen
which extended the isothermal transformation for 240 min. How-
ever, their results is less definitive due to reasons such as possible
interface motion during quenching which may change the local
interfacial condition, as suggested by Van Landeghem et al. [13].
Further systematic investigation on this topic is necessary.

4.2. Dissipations by intrinsic interface mobility at WF/g interfaces

The C contents in g at broad faces of WF/g interfaces boundary

are shown to be much lower than that at GBF/g interfaces at the
early stage for non-partitioned a growth (Fig. 4(d)~(f), Figs. 5(b),
Fig. 8(d)~(i) and Fig. 9). Deviations of interfacial C content from
Fig. 7. a transformation kinetics for two alloys at (a) 750
C and (b) 700
C. Overall NPLE limits at broad faces of WF/g interfaces are resulted from
transformation kinetics are also included in some cases due to pearlite formation. energy dissipations which could be intrinsic friction and/or solute
drag effect. Migrating rate of WF broad faces (vWF ) is firstly esti-
mated in order to clarify the energy dissipations during
study. Similar conclusion has also been obtained by Bradley and
transformation.
Aaronson [9] who studied GBF growth kinetics with a Fe-0.40%C-
vWF is also estimated with Eq. (3) which revised from Eq. (2)
74 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77

Fig. 8. The measured C and Si profiles of 1.5Si alloy across (a)(b)(c) GBF/g interface, (d)(e)(f) WF/g interface transformed at 750
C for 30s and (g)(h)(i) across GBF/g interface at
700
C for 30s. (a)(d)(g) are the OM images showing the lines where EPMA measurements are conducted; (b)(e)(h) and (c)(f)(g) are the measured C and Si profiles, respectively.

Fig. 9. Evolution of interfacial C in g at GBF/g and WF/g interfaces during ferrite transformation for (a)(c)1.5Si and (b)(d) 3Si alloys at (a)(b) 750
C and (c)(d) 700
C.

where GBF thickness was replaced by true thickness of WF plates

(dWF WF
1 ; d2 ).
1 dWF WF
2  d1
vWF ¼ (3)
2 t2  t1
 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77 75

Table 2
Summary of migrating rates of grain boundary ferrite (GBF)/g interfaces (vGBF ) which is estimated from experimentally measured a volume fractions, calculated spike width
based on Si diffusivity in austenite (r) and at grain boundary (r 0 ) as well as migrating rates of Widmanstatten ferrite (WF)/g interfaces (vWF ) estimated from experimentally
measured WF thickness.

Alloy 1.5Si 3Si

Temperature, C 800 750 700 800 750
Reaction time range, s 600e1800 30e60 30e60 30e60 30e60
vGBF for GBF/g interfaces, mm/s 1:4  103 5:6  102 4:1  102 5:5  102 7:2  102
r, nm 1:2 6:3  103 1:5  103 3:0  102 4:9  103
r 0 , nm 2:1  106 2:3  104 1:2  104 5:5  104 5:5  104
vWF for WF/g interfaces, mm/s e 8:6  103 1:8  102 2:8  103 1:2  102

Random lines were plotted on OM images and average inter- smaller than that of incoherent interfaces due to fewer sites for
cepting length between random line and WF were obtained atom attachment in diffusional process compared with the inco-
(dintercept ) with more than 30 measurements for a given specimen. herent ones. Fig. 10(a)(b) summarize the C content at g side of GBF/
The true thickness of WF plate (dWF ) can be calculated with g and WF/g interfaces for the 1.5Si and the 3Si alloys transformed at
following equation [25], different temperatures when the isothermal holding time is 30 s
(exception: 10 min, the shortest isothermal holding time, is used for
. the 1.5Si alloy at 800
C) and Fig. 10(c), (d) convert these experi-
dWF ¼ dintercept 2 (4)
mentally measured interfacial C concentrations into undercoolings
The estimated migrating rates of WF broad faces are also (DG) from PE predictions. NPLE limits could be regarded as com-
included in Table 2. It is seen that estimate migrating rate for WF bination of PE predictions with extra dissipations caused by the full
broad faces is smaller than that for GBF/g interfaces for each con- development of spike (DGspike ) and are also plotted in Fig. 10(c), (d).
dition. This agrees with the fact that WF broad faces are usually It is seen that deviation of C content in g from NPLE limit is much
semi-coherent [26] which is more difficult to migrate during more obvious at semi-coherent WF/g interfaces while that at most
diffusional transformation. GBF/g interfaces is negligible. This indicates that the observed de-
Guo and Enomoto [10] have showed that solute drag effect for viation at WF/g interfaces should be mainly attributed to intrinsic
Fe-0.12%C-1.47%Si, Fe-0.40%C-1.73%Si alloys during 700e900
C friction rather than solute drag.
reached maximum at velocities in range of 0.1 and 1000 mm/s The energy dissipated by intrinsic friction, DGm , could be
which is faster than any interfacial migrating rate met in present expressed as
study (Table 2). Therefore, solute drag effect would be less signifi-
cant to interface with a smaller migrating rate. Besides, less DGm ¼ v  Vm =M (5)
segregation is expected at interfaces of good coherency. On the
other hand, intrinsic mobility for semi-coherent interfaces are where v is the interface migrating rate, Vm is molar volume and M

Fig. 10. The interface C content in g at GBF/g interfaces and WF/g interfaces for (a) 1.5Si and (b) 3Si alloy specimens isothermal reacted for 30 s (exception is 1.5Si alloy at 800
C and
the reaction time is 10 min). (c) and (d) are calculated extra energy dissipations estimated from PE prediction based on the measured C content in (a) and (b), respectively.
76 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77

represents interfacial intrinsic mobility. Because spike develop- development of spike where DG is fully dissipated both by spike
ment at WF/g interfaces is not clarified, DGm was calculated in two development and interface migration. With its estimated migrating
ways, DGm ¼ DG for no development of spike where DG is fully rate (vWF , in Table 2), mobility for WF/g interfaces could be
dissipated by interface migration, DGm ¼ DG  DGspike for full computed. The results are summarized in Fig. 11 and a comparison
with mobility proposed for incoherent a/g interfaces [5e7] is made
(majority of the exact values are also presented in Table 3). The
estimated mobility for semi-coherent GBF/g interface is found to be
much smaller than the values reported in literature [5e7] and
nearly independent of temperature and Si content.

4.3. Effect of orientation relationship between GBF and g on

interfacial C enrichment

As discussed in previous section, the difference of C content in g

at GBF/g and WF/g interfaces is mainly attributed to intrinsic fric-
tions which may be closely related to interface properties. As
shown in Fig. 8(h), it is observed that C content in g at interfaces
adjacent to the same GBF are different. Fig. 12 shows another
example of 1.5Si alloy transformed at 700
C for 30 s where EBSD
measurement was carried out in the same area. The deviated angle
between GBF and adjacent g grains from the well-known Kurdju-
mov-Sachs (K-S) OR [27] was calculated and denoted as DK-S.
Straight GBF/g interface deviates only 0.8
from K-S OR and could
be regarded as (semi-)coherent interface. The other interface which
could be regarded as incoherent interface deviates significantly
from K-S OR and the a preferentially grows into this direction. The C
Fig. 11. Estimated mobility for WF/g interfaces at several conditions and for semi- content at the (semi-)coherent interface side is lower than the
coherent GBF/g interfaces of the 1.5Si alloy transformed at 700
C. Note that
mobility for coherent GBF/g interfaces are estimated when assuming DGm ¼ DG while
value predicted by NPLE limit while the C content of the other
mobility for WF/g interfaces are estimated both when DGm ¼ DG and when DGm ¼ interface coincides well with the prediction. This result strongly
DG  DGspike . suggests the effect of interface coherency on carbon enrichment

Table 3
Estimated mobility values (in m4/(Js), hereafter) in present study when assuming DGm ¼ DG  DGspike as well as values calculated from literature.

Condition Interface type Estimated mobility Mobility calculated from Ref. [5] Mobility calculated from Ref. [6] Mobility calculated from Ref. [7]


1.5Si-750 C-30s WF/g 6:2  1016 1:1  109 1:2  1012 2:8  1014
1.5Si-700
C-30s WF/g 6:8  1016 4:5  1010 5:2  1013 1:2  1014
Semi-coherent GBF/g 4:8  1015
3Si-800
C-30s WF/g 7:4  1016 2:4  109 2:6  1013 6:1  1014
3Si-750
C-30s WF/g 5:2  1016 1:1  109 1:2  1012 2:8  1014

Fig. 12. (a) OM image showing an GBF with semi-coherent and incoherent a/g interfaces for 1.5Si alloy reacted at 700
C for 30 s. (b) Inverse pole figure (IPF) with confidential index
larger than 0.2 and is superimposed with image quality (IQ) map at the same region as (a). (c) C profile across the line indicated in (a) and (b).
 H.-D. Wu et al. / Acta Materialia 149 (2018) 68e77 77

during a transformation and is in agreement with the results at WF/ Acknowledgement

g interfaces.
In present study, EBSD measurement was carried out at all areas H.-D. Wu gratefully acknowledge the financial support from the
where conducted EPMA measurement for specimen of the 1.5Si Joint Education Program between Tohoku University and Tsinghua
alloy transformed at 700
C for 30 s. Semi-coherent GBF/g in- University, and National Natural Science Foundation of China
terfaces are defined as ones where measured OR between GBF and (Grant No. 51471094).
g deviated from K-S OR by angle less than 5
while that for inco-
herent GBF/g interfaces is larger than 5
. Results of interfacial C
content in g at incoherent and semi-coherent GBF/g are plotted References
separately for this condition in Fig. 10(a). C enrichment at inco-
herent GBF/g interfaces agrees well with NPLE limit while that at [1] M. Goune
, F. Danoix, J. Ågren, Y. Bre

chet, C. Hutchinson, M. Militzer, G. Purdy,
semi-coherent GBF/g interfaces significantly deviate from NPLE S. van der Zwaag, H. Zurob, Overview of the current issues in austenite to
ferrite transformation and the role of migrating interfaces therein for low
limit. Migrating rate of semi-coherent GBF/g interfaces is assumed alloyed steels, Mater. Sci. Eng: R 92 (2015) 1e38.
to be the same vGBF (Table 2) and mobility for semi-coherent GBF/g [2] D. Coates, Diffusion-controlled precipitate growth in ternary systems I, Metall.
interfaces was computed similarly to the estimation of interface Trans. 3 (1972) 1203e1212.
[3] D. Coates, Diffusion controlled precipitate growth in ternary systems: II,
mobility for WF/g interfaces (Fig. 11, Table 3). The estimated Metall. Trans. 4 (1973) 1077e1086.
mobility for semi-coherent GBF/g interface is found to be slightly [4] M. Hillert, Paraequilibrium, Internal Report, Swedish Inst. Met. Res., Stock-
smaller than the values reported by Krielaart [7] but larger than holm, Sweden, 1953.
[5] M. Hillert, Diffusion and interface control of reactions in alloys, Metall. Trans. 6
that for WF/g interfaces. (1975) 5e19.
The actual mobility for semi-coherent GBF/g interface, however, [6] J. Wits, T. Kop, Y. van Leeuwen, J. Sietsma, S. van der Zwaag, A study on the
should be much smaller than the estimated values due to following austenite-to-ferrite phase transformation in binary substitutional iron alloys,
Mater. Sci. Eng. A 283 (2000) 234e241.
reason. Semi-coherent GBF/g interfaces are less mobile than inco- [7] G. Krielaart, S. van der Zwaag, Kinetics of g / a phase transformation in Fe-
herent GBF/g interfaces and its true migration rate should be much Mn alloys containing low manganese, Mater. Sci. Technol. 14 (1998) 10e18.
smaller than vGBF . Therefore, interface mobility for semi-coherent [8] G.R. Purdy, Y.J.M. Brechet, A solute drag treatment of the effects of alloying
elements on the rate of the proeutectoid ferrite transformation in steels, Acta
GBF/g interfaces may be also significantly smaller than those re- Metall. 43 (1995) 3763e3774.
ported in literature [5e7]. [9] J. Bradley, H. Aaronson, Growth kinetics of grain boundary ferrite allo-
It has been previously reported [16] that C enrichment in g triomorphs in Fe-C-X alloys, Metall. Trans. A 12 (1981) 1729e1741.
[10] H. Guo, M. Enomoto, Effects of substitutional solute accumulation at a/g
during incomplete bainite transformation deviates significantly
boundaries on the growth of ferrite in low carbon steels, Metall. Mater. Trans.
from NPLE limits and PE predictions. Solute drag effect and spike A 38 (2007) 1152e1161.
development were found to hardly account the large deviations at [11] K. Kinsman, H. Aaronson, Influence of Al, Co, and Si upon the kinetics of the
such conditions in point view of that bainite growth is controlled by proeutectoid ferrite reaction, Metall. Trans. 4 (1973) 959e967.
[12] C. Qiu, H. Zurob, D. Panahi, Y. Brechet, G. Purdy, C. Hutchinson, Quantifying the
C diffusion in g. Since broad faces of BF/g interfaces are usually solute drag effect on ferrite growth in Fe-C-X alloys using controlled decar-
semi-coherent [28], the resulted small interface mobility may ac- burization experiments, Metall. Mater. Trans. A 44 (2013) 3472e3483.
count certain amount of C deviation from NPLE limits and PE pre- [13] H. Landeghem, B. Langelier, D. Panahi, G. Purdy, C. Hutchinson, G. Botton,
H. Zurob, Solute segregation during ferrite growth: solute/interphase and
dictions which worthy further examination. substitutional/interstitial interactions, JOM (J. Occup. Med.) 68 (2016)
1329e1334.
5. Conclusion [14] Z. Liu, G. Miyamoto, Z. Yang, T. Furuhara, Direct measurement of carbon
enrichment during austenite to ferrite transformation in hypoeutectoid
Fee2MneC alloys, Acta Mater. 61 (2013) 3120e3129.
The present study investigated microstructure evolution, [15] Y. Xia, G. Miyamoto, Z. Yang, C. Zhang, T. Furuhara, Effects of Mo on carbon
transformation kinetics and especially the C enrichment during enrichment during proeutectoid ferrite transformation in hypoeutectoid Fe-C-
Mn alloys, Metall. Mater. Trans. A 46 (2015) 2347e2351.
isothermal ferrite transformation at between 800 and 700
C for
[16] H. Wu, G. Miyamoto, Z. Yang, C. Zhang, H. Chen, T. Furuhara, Incomplete
Fe-0.4%C-(1.5, 3)%Si alloys. The following conclusions were bainite transformation in Fe-Si-C alloys, Acta Mater. 133 (2017) 1e9.
obtained: [17] G. Miyamoto, N. Iwata, N. Takayama, T. Furuhara, Mapping the parent
austenite orientation reconstructed from the orientation of martensite by
EBSD and its application to ausformed martensite, Acta Mater. 58 (2010)
1. Experimental transition of a transformation kinetics from fast to 6393e6403.
slow transformation occurs when the volume fractions of pro- [18] M. Enomoto, Thermodynamics and kinetics of the formation of widmansta €t-
eutectoid a coincide with predictions given by NPLE limits for ten ferrite plates in ferrous alloys, Metall. Mater. Trans. A 25 (1994)
1947e1955.
the 1.5Si and 3Si alloys in conditions where interference from [19] P. Franke, G. Inden, An assessment of the Si mobility and the application to
pearlite transformation is not significant. C enrichment in g at phase transformations in silicon steels, Z. Metall 88 (1997) 795e799.
GBF/g interfaces during non-partitioned a growth stage ob- [20] Z. Liu, G. Miyamoto, Z. Yang, T. Furuhara, Volume fractions of proeutectoid
ferrite/pearlite and their dependence on prior austenite grain size in hypo-
tained from EPMA measurements agrees well with NPLE limits. eutectoid Fe-Mn-C alloys, Metall. Mater. Trans. A 44 (2013) 5456e5467.
They together suggest that NPLE regime operates during non- [21] I. Seki, I.K. Nagata, Lattice constant of iron and austenite including its super-
partitioned a growth for the conditions investigated. saturation phase of carbon, ISIJ Inter 45 (2005) 1789e1794.
[22] M. Hillert, Diffusion in growth of bainite, Metall. Mater. Trans. A 25 (1994)
2. C enrichment in g at WF broad faces deviates from NPLE limits 1957e1966.
during non-partitioned a growth stage. The deviations gradually [23] M. Hillert, Nature of local equilibrium at the interface in the growth of ferrite
diminish with increasing isothermal holding time. Besides, such from alloyed austenite, Scripta Mater. 46 (2002) 447e453.
[24] J. Fridberg, L. Torndahl, M. Hillert, Diffusion in iron, Jernk. Ann. 153 (1969)
deviation was also observed at (semi-)coherent GBF/g interfaces
263e276.
in the 1.5Si alloy transformed at 700
C for 30 s. Intrinsic friction [25] R.T. Dehoff, F.N. Rhines, Quantitative Microscopy, McGraw-Hill Publishing,
rather than solute drag effect was concluded to be responsible 1968.
[26] W. Soffa, D. Laughlin, Diffusional phase transformation in the solid state, in:
for the deviations.
D.E. Laughlin, K. Hono (Eds.), Physical Metallurgy, fifth ed., 2014, p. 903.
3. Compared with incoherent a/g interface in literature [5e7], the Oxford.
estimated mobility of WF broad faces in different conditions and [27] G. Kurdjumow, G. Sachs, Über den mechanismus der stahlha €rtung, Z. Phys. 64
of semi-coherent GBF/g interface for 1.5Si alloy transformed at (1930) 325e343.
[28] T. Moritani, N. Miyajima, T. Furuhara, T. Maki, Comparison of interphase
700C for 30 s are found to be smaller and the mobility of WF/g boundary structure between bainite and martensite in steel, Scr. Mater 47
interfaces is less temperature dependent. (2002) 193e199.