ISIJ International, Vol. 53 (2013), No. 12, pp.

2275–2277

Martensite decomposition (i.e., diffusive reversion and

Note cementite precipitation) occurs easily during reheating
toward As because diffusion is easy in low-alloyed steel at
elevated temperatures around As. This difference in the
reversion mechanism between high Ni or Mn steel (e.g. 18%
Ni steel) and low-alloyed steel is schematically shown in
Fig. 1 in the form of a continuous heating reversion dia-
Transition from Diffusive to Displacive Austenite gram. In these diagrams, the austenite-start and -finish tem-
Reversion in Low-Alloy Steel peratures for diffusive reversion during heating are defined
as Ac1 and Ac3, respectively, for distinguishing them from
Nobuo NAKADA,1,2)* Toshihiro TSUCHIYAMA,1,2) the austenite-start and -finish temperatures for martensitic
Setsuo TAKAKI,1,2) Dirk PONGE3) and Dierk RAABE3) reversion, As and Af. Although diffusive reversion is prefer-
ential, upon observing Fig. 1(b), one can expect that the
1) Department of Materials Science and Engineering, Graduate reversion mechanism is changeable depending on the heat-
School of Engineering, Kyushu University, 744 Moto-oka, Nishi- ing rate and that martensitic reversion should occur even in
ku, Fukuoka, 819-0395 Japan. 2) International Institute for low-alloyed steels if the heating rate is sufficient for avoid-
Carbon Neutral Energy Research (WPI-I2CNER), Kyushu ing the diffusive reversion. In this study, the effect of heat-
University, 744 Moto-oka, Nishi-ku, Fukuoka, 819-0395 Japan. ing rate on reversion behavior was investigated in 0.15%C–
3) Max-Planck-Institute für Eisenforschung, Max-Planck-Str. 1, 5%Mn steel for understanding the reversion transition
Düsseldorf, 40237 Germany. mechanism in steel.
0.15%–5%Mn steel (0.14%C–4.97%Mn; % = wt.%), was
(Received on June 24, 2013; accepted on July 23, 2013) used in this study as a low-carbon and low-alloyed steel†.
Mn addition lowers T0 (T0: temperature where the Gibbs
free energies of fcc and bcc are equal) as well as atomic dif-
For understanding the transition from diffusive to displacive fusivity in steel, which is thought to promote the occurrence
austenite reversion mechanism in steel, the effect of heating rate of martensitic reversion. The material was initially solution
on austenite reversion behavior was investigated in 0.15%C– treated at 1 273 K in the austenite single-phase region for
5%Mn steel. Austenite reversion temperature first increased grad- 1.8 ks, followed by water quenching to obtain an initial lath
ually with the heating rate owing to the superheating effect and martensitic structure. Subsequently, this solution-treated
then remained at a constant temperature above a critical heating material was cut into cylindrical rods of length 10 mm and
rate. In response, the austenite formed by rapid heating exhibited diameter 2 mm; these rods were re-austenitized at different
a coarse prior austenite grain structure, indicating the occurrence heating rates from 20 to 450 K/s using induction heating
of displacive reversion even in low-alloy steel. apparatus. The transformation temperature during re-
KEY WORDS: displacive reversion; transition of reversion mech- austenitization was determined using a thermal dilatation
anism; low-alloy steel; heating rate.
tester equipped with a temperature control system. In this
system, temperature is measured by an extra-fine thermo-
couple having high-speed thermal response. Dislocation
density in the martensite was calculated by the X-ray
diffraction method using Co-Kα radiation according to
Martensitic transformation caused by diffusionless shear Hall-Williamson’s method.6,7) The microstructure and crys-
takes place athermally in steel when the decomposition of tallographic characteristics of the rods were analyzed by the
austenite to such products as ferrite, pearlite and bainite is electron back-scattering diffraction (EBSD) method using a
suppressed during cooling. In general, martensitic transfor- field-emission scanning electron microscope (JSM-6701F
mation depends primarily on the cooling rate and composi- developed by JEOL Ltd.).
tion of steel. A high cooling rate and/or the addition of For visualizing prior austenite grain boundaries in the lath
alloying elements to avoid diffusive reactions results in a martensitic structure, a specific technique was used in the
significant degree of super cooling. Eventually, the large crystallographic analysis. In general, each martensite variant
driving force necessary for martensitic transformation has a Kurdjumov–Sachs (K-S) orientation relationship with
(about 1 200 J/mol) can be obtained at the martensite-start prior austenite in low-carbon steels. Therefore, block and
temperature, Ms. In principle, martensitic transformation is packet boundaries should lie between different two of 24 K–S
reversible. It is known that the reversible transformation
from martensite to austenite (martensitic reversion) occurs
in maraging steel owing to its high Ni content.1–4) Previous-
ly,5) the authors of the current study directly observed an
austenitic structure formed by martensitic reversion (mar-
tensitically reversed austenite) in 18%Ni–C (% = wt.%)
steel. It was reported that after reheating the sample to a
temperature above the martensitic austenite-start tempera-
ture, As, all martensite laths originally formed via an
fcc→bcc martensitic transformation reverse into new austen-
ite during the bcc→fcc martensitic reversion, thus returning
to an original austenitic structure with the same crystallo-
graphic orientation. However, the mechanism of this two- Fig. 1. Schematic diagram of continuous heating transformation
directional transformation is still under debate because such explaining the difference in reversion mechanism between
martensitic reversion almost never occurs in low-alloyed high Ni or Mn steel (a) and low-alloy steel (b).
steels. The main reason that inhibits such reversibility is the
high As peculiar to athermal martensitic transformation. †
The definition to distinguish ‘low’ and ‘high’ alloy steels varies among
countries and among standard-setting organizations. As a general indica-
* Corresponding author: E-mail: [email protected] tion, low-alloy steel can be regarded as alloy steels containing 1–5% of
DOI: http://dx.doi.org/10.2355/isijinternational.53.2275 elements except carbon by the ISO definition.

2275 © 2013 ISIJ

 ISIJ International, Vol. 53 (2013), No. 12

Fig. 2. Distribution of misorientation angle in lath martensitic Fig. 3. Continuous heating reversion diagram in 0.15%C–5%Mn
structure and random orientation structure. steel.

variants and should have specific grain boundary character- 1 130 K under the heating rate of 400 K/s. However, the aus-
istics. The possible misorientation angle characterized by tenite-finish temperature was much lower in practice, and
two different K–S variants can be calculated geometrically, the austenite-start and -finish temperatures remained con-
as reported by Morito et al.8) Actual distribution of the mis- stant at 1 006±8 and 1 047±6 K, respectively, under further
orientation angle in the lath martensitic structure of rapid heating. Such discontinuous changes in the reversion
0.15%C–5%Mn steel (shown later in Fig. 4(a)) is shown in temperature correspond to Fig. 1(b), that is, this result
Fig. 2. Given that a block is a smaller subunit than a packet, implies that the transformation mechanism changes from
the density of block boundaries is considerably higher than diffusive to martensitic reversion as the heating rate is
that of packet boundaries. Therefore, the distribution profile increased. Some researchers investigated martensitic rever-
has clear peaks with deviations of several degrees at approx- sion in some types of high-alloyed steels,10,11) and the driv-
imately 10.5°, 49.5° and 60°, which are the theoretical mis- ing force necessary for martensitic reversion was calculated
orientation angles of block boundaries in a K–S relationship. to be 300–400 J/mol from its own As. Thermo-calc. results
By contrast, prior austenite grain boundaries should have a indicate that the Gibbs free energy gap between fcc and bcc
random misorientation distribution, i.e., a Mackenzie pro- in this material is approximately −350 J/mol at 1 000 K if
file,9) because the two martensite variants that form a prior carbide does not precipitate within the lath martensite
austenite grain boundary had been in different prior austen- matrix during heating, which allows the occurrence of mar-
ite grains and, therefore, no orientation relationship. That is, tensitic reversion in terms of thermodynamics. However, we
to observe prior austenite grain boundaries, we should focus cannot ignore carbon diffusion at such elevated tempera-
on the difference between the specific misorientation distri- tures. Assuming that a transformation interface migrates
bution in the lath martensitic structure and the Mackenzie 10 μ m, which corresponds to a conventional martensite
profile shown in Fig. 2. When we selected the misorienta- block size during heating from Ac1 to Ac3, the length of car-
tion selection window between 20 to 50° in the orientation bon diffusion field in the martensite matrix at the austenite/
analysis software, the fraction of the grain boundary in the martensite interface, L, can be estimated as being approxi-
selection window was 72% in the Mackenzie profile. This mately 1 μ m from the diffusion coefficient of carbon, D, and
means that the selection window enables us to visualize the the interface velocity, v, i.e., L~D/v.12) This estimate sug-
majority of prior austenite grain boundaries, excluding gests that even if martensite reverses to austenite via a dis-
block and packet boundaries in the lath martensitic struc- placive mechanism, the austenite reversion should be
ture. accompanied by carbon diffusion.
Figure 3 shows the continuous heating reversion diagram Figure 4 shows the bcc orientation imaging maps (Figs.
of 0.15%C–5%Mn steel, including changes in the austenite- 4(a-1)–4(d-1)) and the corresponding boundary maps with
start and -finish temperatures as a function of the heating misorientation angles over 15° (Figs. 4(a-2)–4(d-2)) and 20–
rate. Additionally, Ae1 (780 K), Ae3 (997 K) and T0 (883 K), 50° (Figs. 4(a-3)–4(d-3)) obtained by high-resolution EBSD
as calculated by Thermo-calc. (SSOL6 database), are also at a step size of 500 nm. The maps show the microstructure
shown in this diagram. Because Ae1 was too low for pro- of the solution-treated specimens and the re-austenitized
moting atomic diffusion of iron, there is a large gap between specimens reheated at heating rates of 200, 300 and 450 K/s.
Ae1 and the austenite-start temperature even at the slowest Each re-austenitized specimen was reheated at the respec-
heating rate. However, the austenite-finish temperature was tive heating rate and then held for 10 s at 1 073 K, which is
almost the same as Ae3 under the heating rate of 20 K/s, higher than the austenite-finish temperatures (see Fig. 3),
which suggests that diffusive reversion must have been followed by He gas quenching. All specimens had a typical
completed under near thermal equilibrium conditions in the lath martensitic structure consisting of packet and block
case of slow heating. The austenite-start and -finish temper- boundaries (Figs. 4(a-1)–4(d-1)), and X-ray diffraction anal-
atures increased gradually with the heating rate, and the aus- ysis proved there was no retained austenite. Boundary maps
tenite-finish temperature was higher than Ae3 by 50 K under with misorientation angles over 15° (Figs. 4(a-2)–4(d-2))
the heating rate of 300 K/s. The austenite-finish temperature indicate that the lath martensitic structure is characterized
increased along the dashed line, which is Ac3, as roughly by high angle block and packet boundaries and that there is
estimated using DICTRA (Diffusion Controlled Transforma- no significant difference in the density of these boundaries
tion, SSOL6 and Mob2 databases). This implies that the among specimens. To reveal the occurrence of the displa-
heating rate dependence of the reversion temperature was cive reversion, we must focus not on block and packet
due to the superheating effect, and diffusive reversion took boundaries but on the prior austenite grain boundaries; this
place at heating rates of up to 300 K/s. According to the esti- is because while the original austenite grain boundaries
mated Ac3, the reversion should have been completed at remained in displacive reversion,3,5) new austenite grains

© 2013 ISIJ 2276

 ISIJ International, Vol. 53 (2013), No. 12

Fig. 4. Orientation imaging map and grain boundary map of 0.15%C–5%Mn steel. Initial solution treated material (a) and
re-austenitized materials at heating rate of 200 (b), 300 (c) and 450 K/s (d).

were formed through nucleation and growth in diffusive effect of the heating rate on austenite reversion behavior was
reversion. In other words, the austenite grain size does not investigated in 0.15%C–5%Mn steel. The results obtained
change during displacive transformations. Therefore, are summarized as follows:
boundary maps with misorientation angles of 20–50° were (1) Displacive reversion takes place even in low-alloy
shown in this figure for visualizing the prior austenite grain steels when the heating rate is above a critical value for
structure. The shape of the prior austenite grains in the solu- avoiding diffusive reversion.
tion-treated material can be observed roughly in Fig. 4(a-3), (2) In diffusive reversion, the reversion temperature
and it was found that the initial austenite grain size was increases gradually with the heating rate owing to the super-
approximately 100 μ m. The prior austenite grains were sig- heating effect, but remains constant in displacive reversion.
nificantly refined by re-austenitization at heating rates of (3) Specimens re-austenitized through displacive rever-
200 and 300 K/s, and the grain size decreased slightly up on sion had a lath martensitic structure with coarse prior aus-
increasing the heating rate (Figs. 4(b-3), 4(c-3)). This tenite grains, and higher hardness and dislocation density
implies that diffusive reversion occurred under these heating than the specimen before re-austenitization.
conditions and that the superheating effect should enhance
the austenite nucleation frequency. It should be noted that Acknowledgement
the austenite formed under the heating rate of 450 K/s had This study was supported under Grant-in-Aid for Young
coarse prior austenite grains, not the fine ones (Fig. 4(d-3)). Scientists (A) No. 22686065 (2010–2011) and Scientific
Indeed, the grain boundary length per unit area of each spec- Research (C) No. 24560852 (2012–2015) by the Japan Soci-
imen was evaluated at (a-3) 5.9×104, (b-3) 18.3×104, (c-3) ety for the Promotion of Science and the International Insti-
20.5×104 and (d-3) 13.2×104 m/m2. The presence of a tute for Carbon Neutral Energy Research (WPI-I2CNER),
coarse-grained prior austenitic structure demonstrates that sponsored by the Japanese Ministry of Education, Culture,
the reversion mechanism was changed discontinuously by Sports, Science and Technology. Additionally, the authors
rapid heating above a critical heating rate (between 300 and would like to thank Mr. Tatsuo Yokoi at Nippon Steel &
400 K/s). Furthermore, it was confirmed that there is little Sumitomo Metal Cop. for supply of the steel used in this
difference in the hardness and dislocation density among the study.
solution-treated specimens and the re-austenitized speci-
mens reheated under heating rates of 200 and 300 K/s REFERENCES
[4.00±0.03 GPa, (3.7±0.1)×1015 m–2], but the re-austenitized
specimen reheated at a heating rate of 450 K/s had higher 1) Y. E. A. Fakina, L. V. Smirnov and V. D. Sadovskiy: Phys. Met.
hardness and dislocation density (4.48 GPa, 4.6×1015 m–2). Metallogr., 31 (1971), 199.
2) K. Hosomi, Y. Ashida, R. Atagi and H. Hato: Trans. JIM, Suppl. 17
These results strongly suggest that the austenite formed by (1976), 87.
displacive reversion exhibits a high dislocation density, 3) T. Maki, H. Morimoto and I. Tamura: Tetsu-to-Hagané, 65 (1979),
which contributes to the strengthening and toughening of 1598.
martensite after quenching due to an ausforming-like 4) T. Tasuno, A. Koganei, K. Kuribayashi, T. Hasegawa and R.
Horiuchi: ISIJ Int., 36 (1996), 595.
effect.4,13,14) In contrast, a few fine austenite grains were 5) N. Nakada, R. Fukagawa, T. Tsuchiyama, S. Takaki, D. Ponge and
found to be dispersed within the coarse-grained matrix (Fig. D. Raabe: ISIJ Int., 53 (2013), 1288.
4(d-3)), as indicated by the arrows. It is known that marten- 6) G. K. Williams and R. E. Smallman: Philos. Mag., 8 (1956), 34.
sitically reversed austenite is easy to recrystallize at elevated 7) W. H. Hall: J. Inst. Met., 77 (1950), 1127.
temperatures owing to a high dislocation density.1–5,15) 8) S. Morito, H. Tanaka, R. Konishi, T. Furuhara and T. Maki: Acta
Mater., 51 (2003), 1789.
Therefore, it is thought that displacive reversed austenite 9) J. K. Mackenzie: Biometrika, 45 (1958), 229.
statically recrystallizes in a short time resulting in the for- 10) K. Tomimura, S. Takaki and Y. Tokunaga: ISIJ Int., 31 (1991), 1431.
mation of fine austenite grains. These experimental results 11) S. J. Lee, Y. M. Park and Y. K. Lee: Mater. Sci. Eng., A515 (2009),
demonstrate that displacive reversion accompanied by car- 32.
12) M. Enomoto: Phase Transformation in Metals, Uchida Rokakuho
bon diffusion takes place even in low-alloyed steels when Pub. Co., Ltd., Tokyo, (2000).
the heating rate is sufficient for suppressing diffusive rever- 13) G. Krauss: Acta Metall., 11 (1963) 499.
sion and also that the displacive-reversed austenite can eas- 14) C. A. Apple and G. Krauss: Acta Metall., 20 (1972), 849.
ily recrystallize. 15) N. K. Nagpaul and D. R. F. West: J. Iron Steel Inst., 208 (1970), 276.
For understanding the reversion mechanism in steel, the

2277 © 2013 ISIJ