Austenitization has attracted much attention owing to its significant roles in controlling the final microstructure and mechanical properties of advanced high-strength steels (AHSS) [[1], [2], [3], [4]]. The initial microstructure of AHSSs prior to austenitization is usually a mixture of pearlite and ferrite (e.g. cementite and ferrite) with a heterogeneous distribution of elements (e.g. C, Mn, etc.). The inhomogeneity in both structure and composition lead to the complicated austenitization behavior upon subsequent heating [5,6].

Speich et al. [7] studied austenite formation in Fe-C-Mn steels with an initial microstructure of pearlite and proeutectoid ferrite. Compared with ferrite, pearlite is thermodynamically and kinetically favorable to transform into austenite due to enrichment of austenite stabilizers (C and Mn). Three kinetic stages during austenite formation were identified: (i) rapid growth of austenite into pearlite; (ii) slower growth of austenite into the remained proeutectoid ferrite; (iii) Mn equilibrium between austenite grains. Nevertheless, due to the structural complexity of pearlite, the role of cementite in austenite formation was not discussed in details.

Different from interval arranged cementite plates in pearlite, the coupled diffusion effects [8] are weakened through spheroidization of cementite during austenite formation, which simplified the related study on its dissolution behavior. Lenel et al. [9] suggested that austenite preferentially nucleates at the interface between ferrite and cementite particles, and it would encircle the cementite particles instantaneously upon nucleation and then extend to both ferrite and cementite. Austenite growth would then proceed via the cooperative migration of the γ/α and γ/θ interfaces, which is closely related to the partitioning behavior of carbon and alloying elements [10].

Miyamoto et al. [11] systematically investigated the effects of Si, Mn and Cr addition on kinetics of isothermal austenite formation from tempered martensite consisting of ferrite and spheroidized cementite in Fe-0.6C steels. It was suggested the alloying elements could partition at the γ/α and γ/θ interfaces and retard austenite growth. The kinetics of isothermal austenite formation from cementite and ferrite have also been simulated using local equilibrium (LE) model [[12], [13], [14], [15]]. The complex kinetic transitions between non-partitioning local equilibrium (NPLE) and partitioning local equilibrium (PLE) were predicted to occur during the migration of γ/α and γ/θ interfaces. In the NPLE mode, the migration of γ/α and γ/θ interfaces is carbon diffusion-controlled and proceeds quite fast, while in the PLE mode, interface migration is sluggish owing to the partitioning of substitutional elements. Nevertheless, austenite formation from cementite and ferrite upon continuous heating was relatively less investigated, while it is of practical interest to steel production. Upon continuous heating, kinetic transitions and alloying element partitioning behaviors at the γ/α and γ/θ interfaces are expected to be thermodynamically and kinetically complex, which should be heating rate dependent. Heating rate in the conventional annealing line is usually near 5 °C/s, and austenite formation upon continuous heating is expected not to be significantly different from isothermal austenite formation.

Recently, the fast-heating (100-300 °C/s) technology was proposed to be a promising technology for producing the strip steels [16,17]. It was found that fast heating not only improves energy efficiency and enhances productivity, it could also lead to unique ultrafine microstructures and significant improvement of the mechanical properties due to the non-equilibrium austenite formation upon fast heating [[18], [19], [20], [21]]. Despite fast heating technology has shown its great potential in the strip steels production, the non-equilibrium austenite formation during fast heating is still not well understood.

In this contribution, austenite formation upon continuous heating from a designed spheroidized cementite structure in Fe-C-Mn alloy was investigated both experimentally and theoretically, focused on clarifying the role of heating rate in kinetic transitions and element partitioning behavior. Combining nano-auger electron spectroscopy equipped with electron backscatter diffraction (AES-EBSD) analysis and local equilibrium simulations, the effects of C and Mn partitioning on the synergetic migration of γ/α and γ/θ interfaces during thermal cycle with slow heating rate (0.1 °C/s) and fast heating rate (100 °C/s) were investigated comparatively. The formation mechanism of the heterogeneous microstructure was discussed in details. In a view of controlled diffusion of elements, we suggested tuning heating rate may open new routes for microstructural design in steels.

The investigated ternary alloy has a chemical composition of Fe-0.23C-1.54 Mn (wt. %). To obtain an initial microstructure with large ferrite grains and uniformly distributed cementite particles, the heat treatment process is designed as illustrated in Fig. 1. The hot-rolled plate with a thickness of 7.5 mm was re-homogenized at 1200 °C for 24 h, and water quenched to room temperature afterwards. The plate was then tempered at 300 °C for appropriate period to make the martensite deformable, and it was subsequently cold rolled to 2.5 mm with a reduction of 67 %. High-temperature (650 °C) tempering treatment for the cold-rolled plate was carried out at the Ar-filled tube furnace for 120 h including heating and cooling process.

Samples 2 × 4× 10 mm³ in size were cut from the middle layer of pre-treated plate to avoid the influence of the possible decarburization. Phase transformations during heating and quenching were measured using a DIL-805 A/D type dilatometer. To avoid the possible microstructure changes during heating from room temperature to the start temperature of austenite formation, samples were firstly heated to 650 °C with 100 °C/s. The specimens were then heated at heating rates of 0.1 °C/s and 100 °C /s to various temperatures and quenched to room temperature with a measured cooling rate of 250 °C/s.

Microstructure of the specimens were firstly examined by scanning electron microscopy (SEM) after the mechanical polishment etched with 4 % Nital solution. The further microstructural characterizations were carried out by a SEM equipped with a PHI-710 type nano-auger electron spectroscopy and electron backscatter diffraction (AES-EBSD). Samples were electrolytically polished in a mixed solution of 20 % perchloric acid and 80 % ethanol at 15 V-1.2A for 20 s. EBSD measurements were focused on the thermocouple-welded area of each sample, using a step size of 30 nm, a tilt angle of 70°, an accelerating voltage of 20 Kv and a beam current of 10 nA. The phase constitutes and orientations were analyzed by the TSL-OIM software. A minimum confidence index of neighbor CI correlation was selected as 0.01 to clean up the noise points. At the same location, carbon and manganese profiles were measured through the line-scan mode of Nano-AES with a spatial resolution about 18 nm. The atomic percentage of Mn was quantified with the aid of a group of high-purity Fe-C-Mn standard samples with linearly varying Mn contents, more details can be found in Ref. [22]. It is noteworthy that the line scanning is conducted before the EBSD measurements in order to avoid the effects of electron beam on the element content at the surface.

The image quality phase color map of the investigated steel after spheroidization is shown in Fig. 2(a). Cementite and ferrite are in yellow and gray, respectively. The initial microstructure prior to austenization consists of a sparse distribution of spheroidized cementite in the coarse ferrite matrix. The substructures of the cold rolled martensite were almost eliminated by recrystallization during tempering. Fig. 2(b) exhibits orientation map of ferrite grain and cementite particles in Fig. 2(a). Cementite particles have different orientations even though they are in the same ferrite grain. Fig. 2(c, d) show a Mn profile across a cementite particle and ferrite measured using AES technique. It shows that Mn is highly enriched and uniformly distributed in cementite with a content close to the equilibrium value calculated by Thermo-calc. A statistical measurement for Mn content in cementite with various size and its surrounding ferrite is displayed in Fig. 2(e). These cementite particles are captured in a square field with 20 × 20 um² in a SEM image. It indicates that the prolonged tempering makes Mn content in most cementite particles close to equilibrium [23,24]. Few particles with a rather small size less than 100 nm were not taken into account.

Fig. 3(a) shows the dilatometric curves of the investigated steels subjected to continuous heating at different rates from ambient temperature to austenitization temperature. With an increase in heating rate from 0.1 °C /s to 100 °C /s, the Ac₁ increases from 720 °C to 776 °C, while Ac₃ increases from 828 °C to 878 °C. The shape of the two curves remains similar, and the abnormal deviation from linear expansion during heating before Ac₁ is considered to be caused by the switching of heating rate and the magnetic transition [25].

Fig. 3(b) exhibits the dilatometric curves of the investigated steel heated at 0.1 °C /s to temperatures lower than Ac₃ and then followed by quenching to room temperature. About 50 % and 80 % of austenite was formed at 775 °C and 797 °C, calculated by the lever rule [26]. Upon quenching, a clear signal of martensite transformation was detected in both samples, and martensite starting temperature was found to increase with increasing intercritical annealing temperature. Except for martensite transformation, a slight expansion of the sample was also identified at the very beginning stage of cooling, which was deduced to be caused by ferrite formation. It will be discussed later.

Fig. 3(c) shows the dilatometric curves of the investigated steel heated at 100 °C /s to various temperatures and then quenched to ambient temperature. About 50 % and 90 % of austenite were estimated to be formed at intercritical temperatures of 827 °C and 857 °C, respectively. The phase transformation behaviors during cooling are quite complex for the fast-heated samples, which is strongly dependent on intercritical annealing temperature. At the beginning of cooling, no obvious phase transformation was observed in the sample quenched from 857 °C, while a non-linear contraction due to austenite formation were detected as the sample was quenched from 827 °C or 837 °C. The abnormal austenite formation upon cooling was named as inverse transformation in Ref. [27], which is considered caused by non-equilibrium conditions at the heating-cooling reversal temperatures. After a short transitory stage, an obvious non-linear expansion due to ferrite formation can be observed in the samples quenched from 827 °C or 837 °C.

Fig. 4(a) shows the microstructure of the sample heated at 0.1 °C/s to 775 °C. The austenite/martensite/cementite islands are found to be uniformly distributed in the ferrite matrix, which indicates that the ferrite/cementite interface served as the nucleation sites for austenite formation. A representative austenite/martensite/cementite island was characterized in detail using the Nano-AES-EBSD. The SEM image (Fig. 4(b)) clearly indicated that there are four layers of structures with different contrast distended from center to the edge. The image quality map taken from the same region obtained by EBSD is shown in Fig. 4(c). The undissolved cementite (in yellow) could be identified in the center, which is surrounded by austenite (in green). The dark gray region reveal at the outer shell of austenite is identified as martensite due to the poor image quality (IQ) value [28]. Ferrite can be distinguished at the outermost layer by higher IQ. The corresponding orientation map is shown in Fig. 4(d). It shows that the displayed blocky austenite belongs to the same austenite grain. The martensite/austenite interfaces with K-S orientation relationship (OR) was indicated by the black line. Using the reconstruction method of the parent austenite [29,30], a core-shell structure of a cementite particle and its enveloping austenite shell could be observed, as shown in the inset of Fig. 4(d).

C and Mn profiles along this multi-layer structure, indicated by a yellow line in Fig. 4(b), are investigated by Auger Nano Probe, as shown in Fig. 4(e). The corresponding interface positions between phases are highlighted by the yellow dash line. A carbon gradient is detected from the central cementite to the outer ferrite, where two sharp descents could be noted at the θ/γ interface and α’/α interface. The Mn distribution, indicated by the blue curve, suggests that Mn partitioning from cementite to austenite proceeded during the thermal cycle. Through the austenite and martensite, Mn content decreased from near 16 % in cementite to 1 % ~ 2 % in ferrite gradually. Fig. 4(f) shows the trace of epitaxial ferrite identified by its different contrast with surrounding ferrite due to a distinct compositional gradient [31], which is in accordance with the non-linear expansion at the beginning of quenching in Fig. 3(b).

Fig. 5(a) shows microstructure of the fast-heated (100 °C/s) samples quenched from 827 °C, where the austenite-martensite-cementite islands are also homogeneously distributed but with a smaller size compared with the slow-heated samples (see Fig. 4(a)). A representative island characterized by the Nano-AES-EBSD indicated that the fast-heating leads to a pronounced difference in multilayer structures. Only three layers of structure was found in the islands, as displayed in the SEM image with different contrasts (Fig. 5(b)). Image quality map with superposed phase color map in Fig. 5 (c) clearly highlights a core-shell structure consisting of cementite (yellow) and the surrounding austenite (green). In this island, fresh martensite was not observed, which could be confirmed by the orientation map (Fig. 5 (d)). Fig. 5(e) shows the distribution of C and Mn across the cementite, austenite and ferrite measured by Auger Nano Probe. It is seen that the carbon intensity displays a distinctive sharp gradient from cementite to austenite. Mn redistribution is marginal at the austenite/cementite interfaces, while a distinct Mn gradient was detected across the austenite/ferrite interface. The region where Mn gradient exists might be the original position of cementite/ferrite interface. Considering the very short annealing time, the Mn enriched austenite could mostly form via the migration of austenite/cementite interface into cementite. It is difficult to deduce from the experiments whether the austenite/ferrite interface has moved or not during the fast annealing.

Fig. 6(a) shows the IQ map of 5 islands in the sample fast-heated to 827 °C at 100 °C/s and then quenched. There is no trace of martensite at each island. The RA islands (No. 1, 2 and 5) suggest that cementite is totally dissolved. The corresponding crystal orientation imaging map is shown in Fig. 6(b). Assuming each RA island is originated from a cementite particle, we can clearly find several austenite grains could nucleate within the same cementite particle under fast-heating conditions, as shown by the RA island of No. 1, 3 and 5. The Mn profiles across islands 1-4 (indicated by the arrow line in the crystal orientation imaging map) shown in Fig. 6(c) are all quite symmetrical. A steep Mn gradient could be observed at all these austenite/ferrite interfaces. According to Fig. 6(c), the Mn contents in γ at the No. 1 and 2 RA islands (12.2 ± 1.1 wt. % and 11.8 ± 0.4 wt. %) are found to be lower than the value in islands No. 3 and 4 (14.5 ± 0.6 wt. % and 14.7 ± 1.0 wt. %) in which cementite was not dissolved completely. This indicates that the Mn content in cementite would also influence the kinetics of cementite dissolution during fast heating.

Fig. 7(a, b) show the IQ map of the fast-heated samples quenched from higher temperatures, 857 °C and 900 °C, respectively. Based on the dilatation curves, about 80 % of austenite is formed at 857 °C, while the sample is fully austenized at 900 °C. The matrix of the as quenched fast heated samples is martensite, containing a small amount of RA and ferrite. The morphology of RA in the fast-heated samples quenched from 857 °C is quite similar to that of initial cementite (see Fig. 2(a)). Fig. 7 (c, d) show the image quality map of the slow-heated samples quenched from 797 °C and 900 °C, respectively. In the slow heated sample quenched from 797 °C a small amount of austenite was retained. Nevertheless, no RA is identified in the sample quenched from 900 °C. The corresponding orientation maps of the slow heated and fast heated samples in Fig. 7(e-h) indicate that fast-heating could lead to a grain refinement of parent austenite grains derived from multiple orientations of RA (highlight with dark circles) and the substructures of martensite. Fig. 7(i-k) display the typical austenite grains by SEM and the corresponding Mn distributions among each phase. In the fast-heated sample quenched from 857 °C, a tiny undissolved cementite still exists inside the large austenite grain, as shown in Fig. 7(i). The sharp Mn gradient at the RA/martensite boundary indicates less Mn has partitioned during the fast heating and quenching, and RA is formed from the original cementite. The similar microstructure has demonstrated a great potential to reach spectacular mechanical properties [32,33]. However, in the slow heated sample quenched from 797 °C shown in Fig. 7(k), cementite has almost all dissolved, and the Mn gradient has diffused out.

During austenite formation from cementite and ferrite, it was experimentally found that austenite could grow with and without distinguished alloying element partitioning during continuous heating. Fig. 8 shows the schematic isothermal sections of Fe-C-Mn phase which indicate Mn diffusion-controlled growth of austenite (Fig. 8(a)), and C diffusion-controlled growth of austenite (Fig. 8(b)), which are actually both expected to occur during heating. The critical temperature at which the transformation mode switches was denoted as partition to non-partition transition temperature (PNTT) [34,35]. At a temperature below PNTT, substitutional alloying element diffusion is required to balance the carbon activity (denoted as a_c) difference at γ/α and γ/θ interfaces. At temperatures above PNTT, austenite could grow without partitioning of substitutional elements via the establishment of a positive carbon activity gradient of from γ/θ interface to γ/α interface (denoted as a_c^γ/θ and a_c^γ/α). Y_Mn in the Fig. 8 is the site fraction of Mn, defined mathematically as Y_Mn=X_Mn/(X_Fe+ X_Mn), where X_Fe and X_Mn are mole fractions of Fe and Mn. The gap between the two carbon isoactivity lines (Δα_c) in Fig. 8(b) enlarges with increasing temperature until one phase is totally dissolved.

The growth of austenite under local equilibrium condition was simulated by DICTRA software, using the TCFE7 thermodynamic database and MOB2 mobility database. In DICTRA simulations, a core-shell structure consist of cementite and ferrite is set as initial structure, where it is assumed that cementite particles have a fraction of about 3.4 % according to thermodynamic calculations with an average radius of 400 nm. The size of the simulation cell is assumed to be R_cell = 2140 nm. Mn content of the initial cementite is set as 16 wt. pct according to the AES measurements, and the composition of ferrite was set according to mass balance. The PNTT in this model was calculated near 762 °C thermodynamically. As illustrated in the inset figure of Fig. 9(a), austenite is set as “inactive” phase which will form at the interface of cementite and ferrite during heating when the driving force exceeds 10^-5 J/mol [15].

Fig. 9(a) shows the simulated kinetics of austenite formation from a mixture of cementite and ferrite (dash line), which is in a qualitative agreement with the dilatometry measurements (solid line). It is worth noting that the predicted kinetic transitions are defined based on the predicted migration modes of austenite/ferrite and austenite/cementite interfaces, which can be derived from the predicted C/Mn profiles. As shown in Fig. 9(a), the simulations show that the switch of interface migration mode indeed change the kinetics of austenite formation to some extent, while it is not significant enough to be detected by dilatometry. The deviation between the simulation and experimental curves are presumably result from heterogeneous distribution of cementite particles which may influence the austenization kinetics on aspects of nucleation and hard impingement during growing. The start and finish temperatures of austenite formation are predicted to increase with heating rate increasing, which is in good agreement with experiments. Three distinct kinetic transitions during the growth of austenite were also predicted during the slow and fast heating, as marked by the blue dots in Fig. 9(a). For the slow-heated sample, austenite grows slowly from 700 °C to near 765 °C (stage I), then grows rapidly in a very short temperature interval from 765 °C to ~771 °C (stage II), and slows down thereafter until the fraction of austenite reaches 100 % at 825 °C (stage III). For the fast-heated sample, austenitization is mainly accomplished by the fast growth stage from ~770 °C to ~837 °C (stage II). Before 770 °C, it is seen that the growth of austenite is extremely restrained (stage I). It is noteworthy that the measured Ac3 (878 °C) at fast-heating is a little bit higher than the predicted missive transformation transition temperature (870 °C) for ferrite with 1 wt. % Mn [36]. However, it is quite challenging to directly prove the presence of massive transformation via experiments as the austenite/ferrite and austenite/cementite interfaces migrate cooperatively.

Fig. 9(b) displays the position of the γ/α and γ/θ interface during continuous heating. The Y axis represents the radius direction of the cell. Austenite starts to nucleate at the position of 0.4 um, then grows into cementite and into ferrite via the γ/α and γ/θ interfaces. It is clearly shown that there exists a critical temperature (PNTT) for the migration of both γ/α and γ/θ interfaces during continuous heating, above which the kinetics of both interfaces remarkably speeds up. For the slow-heated case, in the stage I, it is predicted that the γ/α interface starts to move into ferrite, while the γ/θ interface is almost immobile. Both γ/θ and γ/α interfaces migrate in the stage II. Cementite dissolves completely at the end of stage II, while the γ/α interface continues to migrate sluggishly in the stage III. For the fast-heated case, the migrations of the γ/θ and γ/α interfaces are negligible below ~770 °C, but then migrate synchronously towards the opposite direction as the temperature increases. Different from the slow heating case, it is predicted that cementite could not fully dissolve even after the full dissolution of ferrite, which results in the third stage for the dissolution of the remaining cementite. It is also in good agreement with the experimental observation in Fig. 7.

The evolution of the C and Mn profiles for the slow-heated and fast-heated sample during heating are presented in Fig. 10. For the slow heated case as shown in Fig. 10(a) and (b), at 750 °C (stage I), Mn in both cementite and ferrite diffused into the newly formed austenite, indicating that the migration of both γ/θ and γ/α interfaces are controlled by Mn partitioning, e.g. PLE mode. A carbon gradient is predicted to exist in austenite at 750 °C due to the existence of Mn gradient as carbon activity is Mn content dependent. At 770 °C (stage II), there are very sharp Mn spikes at both γ/θ and γ/α interfaces, and thus the migrations of both interfaces are controlled by carbon diffusion. A smooth distribution of Mn is left in the austenite which verified a long period PLE controlled austenite formation at the slow heating rate. Therefore, the migration mode for both interfaces switches from PLE into NPLE mode during heating from 750 °C to 770 °C. At 780 °C (stage III), cementite has totally dissolved. The Mn profile exhibits a zigzag shape at the γ/α interface, which means that the migration of γ/α interface is controlled by Mn diffusion, e.g. PLE mode.

The kinetic transitions of the γ/θ and γ/α interfaces for the fast-heated case are different from those for the slow-heated case. As shown in Fig. 10(c) and (d), at 760 °C, both γ/θ and γ/α interfaces migrate under PLE mode, and the migration of the interfaces is very sluggish. At 800 °C and 820 °C, positive C gradients are built up in austenite, demonstrating the migration mode is NPLE. Upon heating, a decrease of C content at γ/α interface but an increase at γ/θ interface is predicted, corresponding to the thermodynamic conditions as explained in section 4.1. Upon fast-heating, it is predicted that there is almost no Mn redistribution at the migrating γ/α and γ/θ interfaces, and the steep concentration gradients of Mn at θ/α interface is fully inherited by the austenite and yields a chemical boundary for both Mn and C within austenite.

For the austenite growth under NPLE mode during heating, the migrating velocity of austenite/cementite and austenite/ferrite interfaces based on mass balance can be described respectively by:

where x_θ/γ (≈25 at. %) is the carbon content in cementite at the cementite/austenite interface which is not temperature dependent. It is clearly shown that the competition between cementite dissolution (v_γ/θ) and ferrite dissolution (v_γ/α) is mainly dependent on the evolution of x_γ/θ and xγ/α during heating. With increasing temperature, x_γ/θ increases while x_γ/α decreases, which results in a higher dissolution kinetics of cementite and ferrite. The DICTRA simulation results further indicate that the ratio of v_γ/θ to vγ/α decreases with increasing heating rate. As a result, the cementite is remained to a higher annealing temperature when heating rate was increased from 0.1 °C /s to 100 °C /s. In addition to heating rate, the dependence of x_γ/θ and x_γ/α on temperature are also affected by Mn redistribution between cementite and ferrite, which is expected to affect the interface migration during heating. The influence of Mn redistribution on the austenite reversion from ferrite and cementite during fast heating should also be investigated in the future.

Fig. 11 presents simulation results of the austenite formation kinetics as a function of temperature during the intercritical thermal cycle. The intercritical temperatures are selected to be near the experimental intercritical temperature in section 3.3, with a similar volume fraction of austenite. Fig. 11(a) indicates the predicted kinetics of the γ/θ and γ/α interface migration during heating and cooling for the slow heated cases. The γ/α interface is predicted to be stagnant at the beginning of cooling and then start to migrate backward into austenite, which leads to the formation of epitaxial ferrite as experimentally observed in Fig. 4(f). Based on the evolution of C and Mn profiles in Fig. 11(b) and (c), there is a kinetic transition for the γ/α interface from PLE to NPLE during cooling, and the stagnant stage was because interface migration is in PLE mode. Similar kinetic transition for the γ/α interface was also found during the cyclic phase transformations [27,37]. As shown in Fig. 11(d), for the fast-heated case, the γ/α and γ/θ interfaces respectively migrate into ferrite and cementite at the beginning of cooling, which leads to the inverse austenite formation. The inverse transformation was also observed by dilatometer experiments as shown in Fig. 3(c). Similar as the slow heated case, the γ/α interface was also predicted to migrate backward into austenite. In both slow and fast heating cases, γ/θ interfaces are predicted to be nearly immobile during cooling, which means that the γ/θ interface migration is always controlled by Mn diffusion, as indicated in the inset of Fig. 11(c, f).

It is predicted that heating rate plays an important role in C and Mn distribution in the newly formed austenite, which is expected to affect phase transformation upon the subsequent cooling. For the slow heated case, a smooth Mn gradient from nearly 16 to 1% was predicted in austenite near the γ/θ interface (as shown in the insert of Fig. 11(c)), which is accompanied with a carbon gradient. Given that the inhomogeneous C and Mn content in austenite, the outer ring of austenite with the lowest Mn and C enrichment would transform into ferrite, while the inner ring with a significant enrichment of Mn and C can be stabilized to ambient temperature. An intermediate enrichment of C and Mn in the middle ring of austenite could suppress ferrite formation upon cooling, while it is not enough to suppress martensite transformation. Therefore, a mixed microstructure consisting of cementite/austenite/martensite/ferrite is predicted to form, which is in good agreement with experiments (see, Fig. 4). In contrast to the slow-heated case, Mn concentration in cementite and ferrite were almost freezed and inherited by austenite for the fast-heated case. During cooling, the austenite/ferrite interface migrates into austenite quickly until it was blocked by the high Mn content inherited from previous cementite. Austenite was retained to ambient temperature due to the significant C and Mn enrichment, which results in the microstructure consisting of cementite/austenite/ferrite or austenite/ferrite (see Fig. 5, Fig. 6).

Fig. 12 summarized the evolution of microstructure and Mn distribution during thermal cycles. Both fast-heating and slow-heating could lead to chemical gradients within austenite due to the kinetic mismatch between elements diffusion and austenite formation, while the sharpness of chemical gradient is strongly dependent on heating rate and heating temperatures. The sharpness of chemical gradient within austenite would then significantly affect austenite decomposition during cooling. Therefore, heating rate and heating temperature can be used to adjust chemical heterogeneity within austenite and further tailor the final microstructure.

In this study, austenite formation from a mixture of ferrite and spheroidized cementite with a significant Mn partitioning in an Fe-0.23C-1.54 Mn alloy were systematically investigated, and the effect of heating rate on kinetic transition and elements partitioning was discussed in detail. Austenite growth was found to proceed via PLE (γ/α, γ/θ), NPLE (γ/α, γ/θ) and PLE (γ/α) controlled interfaces migration during slow-heating, while NPLE (γ/α, γ/θ) mode predominately controlled the austenitization upon fast-heating through a synchronous dissolution of ferrite and cementite. It was both experimentally and theoretically found that after austenite growth Mn distributions within austenite grains of slow and fast heated samples are inhomogeneous. For the fast-heated sample, the significant enrichment of Mn in original cementite particles was fully inherited by newly formed austenite, which leads to sharp Mn gradients within austenite grains. However, it diffuses out in the slow-heated sample due to the long-range Mn partitioning during austenite growth. The different Mn distribution within austenite grains of the slow and fast heated samples was found to play a significant role in phase transformations upon the subsequent cooling. The localized enrichment of Mn within austenite grains could suppress ferrite or martensite formation upon cooling, which could help stabilize the ultrafine austenite to ambient temperature. Therefore, it is expected that heating rate can be used as an effective parameter to tune the magnitude of local Mn enrichment and further tailor the microstructure of steels.

H. Chen acknowledges financial support from the National Natural Science Foundation of China (Grant U1860109, 51922054, U1808208 and U1764252) and Beijing Natural Science Foundation (2182024). Z. G. Yang acknowledges financial support from the National Natural Science Foundation of China (Grant 51771100). C. Zhang acknowledges financial support from the National Natural Science Foundation of China (Grant 51771097) and the Science Challenge Project (Grant TZ2018004). G. Liu acknowledges financial support from China postdoctoral science foundation (2018M631459).