Scripta Materialia 140 (2017) 82–86

Contents lists available at ScienceDirect

Scripta Materialia
journal homepage: www.elsevier.com/locate/scriptamat

Regular article

Bainite formation kinetics in steels and the dynamic nature of the

autocatalytic nucleation process
Ashwath M. Ravi* , Jilt Sietsma, Maria J. Santofimia
Dept. of Materials Science and Engineering, Delft University of Technology, Mekelweg 2, Delft 2628 CD, The Netherlands

A R T I C L E I N F O A B S T R A C T

Article history: Over the years, a quantitative theory to explain bainite formation kinetics has been proposed based on the
Received 27 February 2017 nucleation kinetics of bainitic sub-units. Although the theory shows acceptable correlation with experimen-
Received in revised form 2 June 2017 tal results, it is observed that the kinetic models show a certain degree of discrepancy with actual kinetics.
Accepted 29 June 2017 It is identified that these mainly arise due to the inadequate estimation of autocatalytic nucleation, espe-
Available online 17 July 2017
cially as a function of progress of bainite formation. With the help of this observation, the kinetic model is
modified and a better insight into the process of autocatalytic nucleation, essential in bainite formation, is
Keywords: obtained.
Bainite © 2017 Acta Materialia Inc. Published by Elsevier Ltd.This is an open access article under the CC BY-NC-ND
Kinetics
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
Nucleation
Isothermal heat treatments

Among the wide variety of products formed by the decomposition nucleation at bainite/austenite interfaces (autocatalytic nucleation)
of austenite in steels, formation of bainite is one of the least under- and can be expressed as
stood phenomena [1–8]. According to the displacive theory of bainite
 
formation, the rate of bainite formation, df /dt, is proposed as DQ ∗
k = exp (2)
kT
df /dt = (1 − f )(1 + kf )jf (1)  
where DQ ∗ is the difference QG∗ − QA∗ in the activation energy for
 
where f is the bainite fraction, k is the autocatalytic parameter grain-boundary nucleation (QG∗ ) and autocatalytic nucleation QA∗ , k
and jf is the rate parameter which accounts for the thermally acti- is Boltzmann’s constant and T is the bainite formation temperature.
vated nature of the bainite nucleation process [9–15]. This equation Studies claim that nucleation based models using the functional
was derived based on the displacive mechanism of bainite forma- form described in Eq. (1) accurately predict the bainite formation
tion [9,13]. Bainite formation in steels begins with nucleation of kinetics in steels [9,11-13,15]. However, a close examination of the
bainitic ferrite at austenite grain boundaries. Subsequently, nucle- published results suggests that these models still show a certain
ation continues further through autocatalytic nucleation of bainitic degree of miscalculation of kinetics [10-12,14]. Santofimia et al. [13]
ferrite at the newly created bainitic ferrite/austenite interfaces [16]. evaluated the applicability of several kinetic models which are based
The difference in nucleation rate of bainite formed by autocatalysis on the displacive theory of bainite formation. They also observed
compared to the nucleation rate of bainite due to grain-boundary that the models imprecisely estimate the nucleation rate and called
nucleation is accounted for by the term kf in Eq. (1) [9–13]. Most for a better treatment of autocatalytic nucleation [13]. Although
studies applying Eq. (1) treat k as an empirical dimensionless fitting these discrepancies are rarely investigated in detail, they have been
constant [13]. The values obtained for k are however not satisfacto- attributed to improper estimation of final volume fraction of bainite
rily analysed in the literature [6]. Recently, the present authors [15] or to unaccounted carbide precipitation [13]. This implies that these
proposed that k is determined by the difference in activation energy discrepancies are generally considered to be due to the improper
for bainite nucleation at austenite grain boundaries and bainite estimation of the degree of carbon enrichment of austenite, since the
carbon content in austenite determines final volume fraction of bai-
nite [10,14] and the degree of carbide precipitation during bainite
* Corresponding author.
formation [11].
E-mail addresses: [email protected] (A.M. Ravi), [email protected] In the current work, it is shown that an adequate estimation of
(J. Sietsma), M.J.Santofi[email protected] (M.J. Santofimia). carbon enrichment alone is not sufficient to accurately simulate the

http://dx.doi.org/10.1016/j.scriptamat.2017.06.051
1359-6462/ © 2017 Acta Materialia Inc. Published by Elsevier Ltd.This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/
4.0/).
 A. Ravi et al. / Scripta Materialia 140 (2017) 82–86 83

Table 1 which grain-boundary nucleation sites can contribute to the over-

Chemical compositions of steels used for study (values in wt%). all nucleation rate. Numerically, (df /dt)v can be determined using
Steel C Mn Si Mo Al Reference experimentally obtained (df /dt) data and the corresponding bainite
S1 0.2 3 – – – Current work fraction, f .
S2 0.2 3.51 1.52 0.25 0.04 [17] The displacive theory of bainite formation suggests that the rate
of bainite formation is determined by the nucleation of bainitic sub-
units. Thus from Fig. 1 (c) and (d), the rate at which bainite nucleation
occurs within the available austenite can be interpreted. It can be
seen from these figures that the rate of austenite transformation into
rate of bainite formation. A strong argument for this is presented bainite constantly changes as bainite formation progresses.
here. A fundamental change in the assumptions for autocatalytic One of the well documented reasons for such a change in the
nucleation, especially as a function of increasing bainite fraction, is austenite transformation rate is due to the possible carbon enrich-
required to adequately estimate bainite formation kinetics. Based ment of austenite during bainite formation [20]. Since (df /dt)v is
on these new assumptions, the bainite kinetics is simulated and a measure of the rate of austenite to bainite transformation, the
compared with experimental results. effect of carbon enrichment on the rate of bainite formation can be
The compositions of the steels used for the present work are given understood by interpreting its effect on (df /dt)v . Using Eqs. (1) and
in Table 1. Studies in the current work have been carried out using (3), (df /dt)v can be given as
kinetic data obtained from isothermal bainite formation experiments
on Steel S1. The bainite formation experiments were carried out
(df /dt)v = (1 + kf )jf . (4)
in a Bähr DIL805A/D dilatometer. Specimens were first austenized
at 1273 K and then isothermally held at 653 K. The bainite fraction
formed as a function of time was determined based on the dilatome- With the help of Eq. (4) and the kinetic model proposed by the
ter data obtained. Results obtained from these studies were further authors [15], a physically based interpretation of (df /dt)v can be
validated using the kinetic data published in the literature on Steel derived since the carbon enrichment of residual austenite during
S2 (isothermal treatment at 603 K for 1 h after austenization at 1173 bainite formation is well accounted for in the proposed model. The
K) [17]. underlying principles used in Ref. [15] for calculation of bainite for-
Fig. 1 (a) gives the experimentally obtained bainite fraction, f vs mation kinetics are similar to other published models [12,13,18,19]
time for both steels, while Fig. 1 (b) gives the rate of bainite forma- that use displacive theory of bainite formation. In Ref. [15], the effect
tion, df /dt, as bainite formation progresses. Fig. 1 (c) and (d) shows of carbon enrichment on the rate of bainite formation is calculated
experimentally obtained (df /dt)v as a function of f in Steel S1 and using a fitting constant, Xb . Xb accounts for the carbon which does
Steel S2 respectively, where (df /dt)v is the rate of bainite forma- not participate in the carbon enrichment of austenite. Using the
tion per unit volume of untransformed austenite (volume fraction approach given in Ref. [15], jf in Eq. (1) can be given as
available for bainite formation). It is given as
 
Q∗
jf ∝ (Th − T)(T0 − T) exp − G (5)
kT
(df /dt)v = (df /dt)/(1 − f ). (3)

where Th and T0 are the critical temperatures which define the ther-
(df /dt)v is an important parameter in understanding the bainite modynamic conditions for bainite formation [3]. The factor (Th − T)
formation kinetics. The overall rate of bainite nucleation in steels signifies the driving force available for bainite nucleation, while
mainly depends number of grain-boundary nucleation sites avail- the factor (T0 − T) signifies the driving force available for bainite
able and the potency of these nucleation sites to form bainitic ferrite growth [3]. Th , T0 and QG∗ are all functions of carbon concentration of
sub-units. It should be noted that the rate of autocatalytic nucle- austenite [3,15,20] and can be expressed in terms of f and Xb [15].
ation also depends on the rate at which grain-boundary nucleation Studies suggest that Th and T0 decrease linearly and QG∗ increases
occurs, since grain-boundary nucleation is a prerequisite for creating linearly with increasing carbon enrichment of austenite [15].
bainite/austenite interfaces and subsequent autocatalytic nucleation. In order to understand the effect of carbon enrichment of austen-
The potency of grain-boundary nucleation sites to transform into ite on the rate of bainite formation, (df /dt)v was calculated assuming
bainitic sub-units and facilitate autocatalytic nucleation is influenced varying degrees of the carbon enrichment (by numerically varying
by rate governing parameters such as bainite formation tempera- the value of Xb ) using Eqs. (2), (4), and (5). In Eq. (5), the proportion-
ture and carbon concentration in austenite [12,13,18,19]. (df /dt)v ality constant is a material dependent parameter and is calculated
gives a measure of this potency. Physically, it represents the rate at according to Ref. [15] (for Steel S1 = 1.92 s −1 K −2 ). According to the

-3 -3
1 8 x 10 25 x 10 x 10-3
5
Steel S1 (T = 653 K) (b) (c) Steel S1
(d) Steel S2
Steel S1 20 4
6 (T = 653 K) (T = 603 K)
(df/dt)v [s-1]
(df/dt)v [s-1]
df/dt [s-1]

15 3
f

0.5 4
Steel S2 10 2
(T = 603 K) 2 Steel S2
5 1
(a)
0 0 0 0
0 1000 t [s] 2000 0 0.5 1 0 0.5 1 0 0.5 1
f f f

Fig. 1. (a) Comparison of experimentally obtained kinetics (Steel S1 at T = 653 K, Steel S2 at T = 603 K). (b) Experimentally obtained df /dt vs f (c, d) Experimentally obtained
(df /dt)v vs f .
84 A. Ravi et al. / Scripta Materialia 140 (2017) 82–86

definition of Xb [15], the degree of carbon enrichment of austen- a certain bainite fraction is formed, the carbon enrichment of the
ite increases for decreasing values of Xb (Fig. 2 (a)). The varying austenite becomes too high to sustain an increasing rate of bainite
trends of (df /dt)v as a function of Xb shed light on the effect of formation. This results in the maximum seen in Fig. 2 (a). This expla-
the carbon enrichment of austenite on bainite formation kinetics. nation of the maximum is consistent with the observation that for
(df /dt)v can be calculated for various Xb values assuming the initial Xb = X̄, which implies that carbon enrichment of austenite does not
QG∗ value (at f = 0, referred as Q ∗ ) to be 155 kJ/mol and DQ ∗ to be take place, no maximum occurs. Based on these trends, it can be con-
GX̄
20 kJ/mol. These assumed values are based on previously published cluded that the maximum is due to carbon enrichment of austenite
data for the activation energy for bainite nucleation [11,12,15,21] in during bainite formation.
the literature. However, by further comparing Fig. 2 (a) to Fig. 1 (c) and (d), it
Fig. 2 (a) shows the (df /dt)v trends as a function of bainite frac- can be seen that the carbon enrichment alone is insufficient to accu-
tion by assuming four different Xb values in Steel S1 for isothermal rately simulate the experimentally obtained trends. The non-linear
bainite formation at 653 K. The assumed Xb values can be related to increase in rate of bainite formation before the maximum (clearly
the overall carbon concentration of the steel, X̄. For instance, in Fig. 2 seen in Fig. 1 (c)) is not predicted by the current description (Fig. 2
(a), Xb = 0.8X̄ suggests that Xb is assumed to be equal to 80% of the (a)). This calls for revisiting the assumptions used for simulating
overall carbon concentration of steel. In Fig. 2 (a), a maximum is seen bainite formation kinetics.
to occur for Xb < X̄. Furthermore, the bainite fraction at which the As mentioned previously, most studies use a constant k to
maximum occurs shifts to a lower value when the degree of carbon account for autocatalytic nucleation during bainite formation. As
enrichment of austenite is higher (i.e., when Xb is lower). It must given in Eq. (2), k is dependent on the difference in the activa-
be noted that although (df /dt)v is only shown for Steel S1, similar tion energy for grain-boundary bainite nucleation and autocatalytic
trends apply for Steel S2 (or any other steel composition) since Eqs. bainite nucleation. This difference is assumed to be constant [15].
(2), (4), and (5) are applicable for all steels. This suggests that the activation energy for grain-boundary nucle-
Fig. 2 (a) can be compared with the experimental (df /dt)v curves ation and autocatalytic nucleation increase at the same rate with
in Fig. 1 (c) and (d) in order to understand the effect of carbon enrich- increasing f .
ment of austenite on bainite kinetics. It can be seen that the carbon However, it can be argued that k will not remain constant
enrichment of austenite during bainite formation in Steel S2 is much throughout the bainite formation process. Bainite nucleation is an
greater than for Steel S1 in agreement with Fig. 2 (a); a clear max- interfacial process and thus depends on the chemistry and the mor-
imum is visible in Fig. 1 (d) at a low value for bainite fraction (f ≈ phological characteristics of the interface at which bainite nucleates.
0.4) while the maximum in Fig. 1 (c) appears close to the end of These aspects heavily depend on the type of the interface [22,23].
the bainite formation process (f ≈ 0.9) and is less pronounced. This This suggests that the activation energy required for grain boundary
can be attributed to the chemical composition of the steels. Steel S1 nucleation can be expected to be different compared to the activation
is a silicon-free steel and exhibits carbide precipitation during bai- energy required for autocatalytic nucleation. Furthermore, bainitic
nite formation, while in Steel S2, carbide precipitation is suppressed, growth is a displacive process and leads to plastic deformation of
leading to greater carbon enrichment of austenite [17]. surrounding austenite matrix [2,16,24]. This suggests that the dislo-
The carbon enrichment of austenite (whose degree increases cation densities around the bainite/austenite interfaces may vary as
as bainite fraction increases) should ideally lead to a decrease in the bainitic growth continues to form sheaves. Therefore, the acti-
rate of bainite formation as bainite fraction increases. Neverthe- vation energies of autocatalytic nucleation and of grain-boundary
less, Fig. 2 (a) shows an increase in rate of bainite formation until nucleation will increase at different rates. To account for these
a maximum is reached at a relatively high bainite fraction. This effects, it is assumed that the difference in the activation energy is a
is due to the competition between increasing autocatalytic nucle- function of bainite fraction f can be expressed as,
ation and increasing degree of carbon enrichment as bainite fraction
increases. The carbon enrichment of austenite affects both auto-
DQ ∗ = QG∗ − QA∗ = DQX̄∗ + hf (6)
catalytic and grain-boundary nucleation. As the bainite formation
progresses, both the activation energy for autocatalytic nucleation
and the activation energy for grain-boundary nucleation decrease where h is the proportionality constant between DQ ∗ and f . DQ ∗ is
X̄
with increasing degree of carbon enrichment. On the other hand, DQ ∗ at f = 0.
the rate of autocatalytic nucleation increases initially as the bainite ∗
Using the above assumption of DQ being linearly related to f , the
formation progresses as a result of increased number density of auto- trends for (df /dt)v were recalculated for various h values in Steel S1
catalytic nucleation sites due to increasing bainite fraction. When for isothermal bainite formation at 673 K (Fig. 2 (b)), using the same
values as for Fig. 2 (a). It must be noted that for Fig. 2 (a), DQ ∗ was
assumed to be constant. For Fig. 2 (b), DQ ∗ is assumed to depend on f,
according to Eq. (6), and DQ ∗ is assumed to be constant. Furthermore,
X̄
x10-3 x10-3 in the calculations, in order to account for the effect of a slight carbon
30 30
Xb=X =0 enrichment, Xb is considered to be equal to 0.99X̄.
Xb=0.99X = 2 It can be seen in Fig. 2 (b) that the calculated trends are now much
Xb=0.9X = 5 closer to the experimentally obtained (df /dt)v curve in Fig. 1 (c). The
20 Xb=0.8X 20 = 8
(df/dt)v [ s-1]

non-linear increase in rate of bainite formation before the maximum

is predicted by the calculated trends. Based on these results, it can
be said that a combination of Xb and h values is required to simulate
10 10 the experimental trends observed. With the help of these results, the
Steel S1 Steel S1 kinetic model proposed in Ref. [15] was modified and h was intro-
(T = 653 K) (a) (T = 653 K) (b) duced as a new fitting parameter. The calculated kinetics from the
0 0 model was compared with kinetics data obtained by dilatometry
0 0.5 f 1 0 0.5 f 1
experiments in the current work as well as with published kinetic
Fig. 2. Calculated (df /dt)v vs f in Steel S1 (Bainite formation temperature = 653 K)
data in the literature [17].
for (a) varying degrees of carbon enrichment (varying Xb ) and (b) for varying h values Fig. 3 (a), (b) shows the comparison between simulated and
(in kJ/mol), with Xb = 0.99X̄. experimentally derived (df /dt)v as a function of bainite fraction. It
 A. Ravi et al. / Scripta Materialia 140 (2017) 82–86 85

x 10-3 x 10-3 3 x 10
4
3 x 10
4
25 5

Change in Activation Energy

(a) Steel S1 (b) Steel S2 (c) (d)
20 (T = 653 K) 4 (T = 603 K)
(df/dt)v [s-1] 2
*
QG - Q*AX¯ 2 *
QG - Q*AX¯
15 3

[J/mol]
Steel S1 Steel S2
10 2 Q*
1 1 Q*
5 1 QA* - Q*AX¯
QA* - Q*AX¯
0 0 0 0
0 0.5 f 1 0 0.5 f 1 0 0.5 f 1 0 0.5 f 1

Fig. 3. (a, b) Experimental (markers) and calculated (lines) (df /dt)v vs f (dashed line: h = 0 kJ/mol, solid line: h = −4.8 kJ/mol (Steel S1) and −6.1 kJ/mol (Steel S2)). (c, d) Change
in activation energy for nucleation (dashed line: autocatalytic nucleation, solid line: grain-boundary nucleation). The bainite formation temperature for Steel S1 is 653 K and Steel
S2 is 603 K respectively. Q ∗ is the initial QA∗ at f = 0.
AX̄

can be observed that (df /dt)v is accurately calculated with the mod- envisaged that as the bainite formation progresses, the austenite
ified model. This suggests that the nucleation rate and consequently, matrix in which the bainite formation occurs undergoes several
the activation energy for bainite nucleation is calculated precisely. changes, thereby affecting the bainite formation kinetics.
The values for the fitting parameters obtained are given in Table 2. In summary, even though a kinetic theory for bainite formation
It is observed that the h parameter shows a negative value. This sug- based on the displacive mechanism of bainite growth has been well
gests that the activation energy for autocatalytic nucleation increases established, literature evidence calls for a more detailed treatment
faster than the activation energy for grain-boundary nucleation as since deviations larger than the experimental uncertainty are found.
a function of bainite fraction, as seen in Fig. 3 (c) and (d). These The reasons behind such discrepancies are investigated in detail in
figures give the change of activation energy for bainite nucleation the current work. The proposed kinetic theory suggests that carbon
as a function of f for both autocatalytic and grain-boundary nucle- enrichment of the austenite during bainite formation has a different
ation. The change in activation energy is calculated with respect to influence on the activation energies for grain-boundary and auto-
the initial activation energy for autocatalytic nucleation, Q ∗ (i.e. catalytic nucleation. Furthermore, factors such as the instantaneous
AX̄
QA∗ at f = 0). These figures also show that the activation energy for deformation state of the austenite and the instantaneous volume of
grain-boundary nucleation in Steel S1 only increases significantly as the austenite grain can be expected to affect the bainite kinetics as
bainite fraction approaches unity. This
 is due to the limited carbon well. A new fitting parameter, h, is introduced to the kinetic model
enrichment in Steel S1 Xb = 0.99X̄ . given in Ref. [15] to account for these factors. h is a physical entity
In a physical sense, the rate at which bainite formation progresses which expresses the difference between the activation energy for
can be estimated by calculating activation energies for both grain- autocatalytic bainite nucleation and grain-boundary nucleation as a
boundary nucleation and autocatalytic nucleation. As discussed in function of bainite fraction. This interpretation improves the quanti-
already published literature, carbon enrichment of austenite leads tative theory and provides excellent correlation between experimen-
to increase in the activation energy for bainite nucleation, which tal and calculated bainite kinetics.
leads to a decreasing rate of bainite formation as the bainite frac- The research leading to these results has received funding from
tion increases. However, the rate at which the activation energies for the European Research Council under the European Union’s Seventh
grain-boundary nucleation and for autocatalytic nucleation increase Framework Programme (FP/2007–2013)/ERC Grant Agreement n.
is different. This suggests that the impact of carbon enrichment on [306292].
autocatalytic nucleation and grain-boundary nucleation might be
different. Also, other factors can lead to an increase in the activation References
energy for bainite formation, since the ability of austenite to trans-
[1] L.C.D. Fielding, Mater. Sci. Technol. 29 (2013) 383–399. http://dx.doi.org/10.
form into bainite is not just affected by its carbon enrichment during 1179/1743284712Y.0000000157.
bainite formation. One other factor can be the plastic deformation [2] H.K.D.H. Bhadeshia, D.V. Edmonds, Acta Metall. 28 (1980) 1265–1273. http://
of austenite, that is associated with bainite formation [1,2,25]. These dx.doi.org/10.1016/0001-6160(80)90082-6.
[3] H.K.D.H. Bhadeshia, Acta Metall. 29 (1981) 1117–1130. http://dx.doi.org/10.
factors are accounted for by the parameter h. Ongoing studies by
1016/0001-6160(81)90063-8.
the present authors on h also show that h can be affected by the [4] M. Hillert, ISIJ Int. 35 (1995) 1134–1140. http://dx.doi.org/10.2355/
grain size of the austenite in which the bainite formation occurs. isijinternational.35.1134.
[5] M. Hillert, Metall. Mater. Trans. A 25 (1994) 1957–1966. http://dx.doi.org/10.
Such behaviour can be due to the reducing volume of the austenite
1007/BF02649044.
grains within which bainite formation can occur as the bainite frac- [6] H.K.D.H. Bhadeshia, Mater. Sci. Eng. A 273–275 (1999) 58–66. http://dx.doi.org/
tion increases and to the stress state associated with this residual 10.1016/S0921-5093(99)00289-0.
austenite volume (due to surrounding bainite). It can be, therefore, [7] A. Borgenstam, M. Hillert, J. Ågren, Acta Mater. 57 (2009) 3242–3252. http://
dx.doi.org/10.1016/j.actamat.2009.03.026.
[8] F. Caballero, M. Miller, C. Garcia-Mateo, Acta Mater. 58 (2010) 2338–2343.
http://dx.doi.org/10.1016/j.actamat.2009.12.020.
[9] H.K.D.H. Bhadeshia, J. Phys. 43 (1982) 443–448.
Table 2
[10] N.A. Chester, H.K.D.H. Bhadeshia, J. Phys. IV France 07 (1997) 41–46. http://dx.
Values of fitting parameters obtained (95% Confidence Interval provided in doi.org/10.1051/jp4:1997506.
parentheses). [11] S.M.C. van Bohemen, J. Sietsma, Int. J. Mater. Res. 99 (2008) 739–747. http://
Steel T X̄ Xb Q∗ DQ ∗ h dx.doi.org/10.3139/146.101695.
GX̄ X̄ [12] S.M.C. van Bohemen, D.N. Hanlon, Int. J. Mater. Res. 103 (2012) 987–991.
(K) (wt%) (wt%) (kJ/mol) (kJ/mol) (kJ/mol)
http://dx.doi.org/10.3139/146.110744.
S1 653 0.2 0.199 164.5 23.6 −4.8 [13] M.J. Santofimia, F.G. Caballero, C. Capdevila, C. Garcia-Mateo, C.G. de Andres,
(±0.001) (±0.1) (±0.3) (±0.2) Mater. Trans. 47 (2006) 1492–1500. http://dx.doi.org/10.2320/matertrans.47.
1492.
S2 603 0.2 0.1 165.2 21.0 −6.1
[14] G. Sidhu, S. Bhole, D. Chen, E. Essadiqi, Scr. Mater. 64 (2011) 73–76. http://dx.
(±0.01) (±0.1) (±0.1) (±0.1)
doi.org/10.1016/j.scriptamat.2010.09.009.
86 A. Ravi et al. / Scripta Materialia 140 (2017) 82–86

[15] A.M. Ravi, J. Sietsma, M.J. Santofimia, Acta Mater. 105 (2016) 155–164. http:// [22] M. Herbig, D. Raabe, Y.J. Li, P. Choi, S. Zaefferer, S. Goto, Phys. Rev. Lett. 112
dx.doi.org/10.1016/j.actamat.2015.11.044. (2014) 126103. http://dx.doi.org/10.1103/PhysRevLett.112.126103.
[16] H.K.D.H. Bhadeshia, Bainite in Steels: Transformations, Microstructure and [23] Y. Li, D. Ponge, P. Choi, D. Raabe, Ultramicroscopy 159, Part 2 (2015) 240–247.
Properties, Matsci Series IOM Communications. 2001. 1st International Conference on Atom Probe Tomography & Microscopy.
[17] A. Navarro-López, J. Sietsma, M.J. Santofimia, Metall. Mater. Trans. A 47 (2016) [24] S. Singh, in: E. Pereloma, D.V. Edmonds (Eds.), Phase Transformations in
1028–1039. http://dx.doi.org/10.1007/s11661-015-3285-6. Steels, Woodhead Publishing Series in Metals and Surface Engineering, vol.
[18] H. Matsuda, H.K.D.H. Bhadeshia, Proc. R. Soc. Lond. A 460 (2004) 1707–1722. 1, Woodhead Publishing. 2012, pp. 385–416. http://dx.doi.org/10.1533/
http://dx.doi.org/10.1098/rspa.2003.1225. 9780857096104.3.385.
[19] G.I. Rees, H.K.D.H. Bhadeshia, Mater. Sci. Technol. 8 (1992) 985–993. http://dx. [25] E. Swallow, H.K.D.H. Bhadeshia, Mater. Sci. Technol. 12 (1996) 121–125. http://
doi.org/10.1179/mst.1992.8.11.985. dx.doi.org/10.1179/mst.1996.12.2.121.
[20] H.K.D.H. Bhadeshia, J.W. Christian, Metall. Trans. A. 21 (1990) 767–797. http://
dx.doi.org/10.1007/BF02656561.
[21] D. San Martin, K. Aarts, P. Rivera-Diaz-del-Castillo, N. van Dijk, E. Brück, S.
van der Zwaag, J. Magn. Magn. Mater. 320 (2008) 1722–1728. http://dx.doi.org/
10.1016/j.jmmm.2008.02.002.