The experimental steel is a laboratory developed low carbon Si-Mn TRIP steel, and chemical compositions are shown in Table 1. The experimental steel was smelted in a 150 kg vacuum induction furnace, and forged into slab billet of 60 mm in height, and 100 mm in width. Several critical temperatures including the austenite non-recrystallization temperature (T _nr) were measured using Gleeble-3500 thermomechanical simulator, and listed in Table 2. On the basis of these temperature parameters, by varying the deformation temperature in austenite non-recrystallization region, two processes of A and B were designed, as shown in Fig. 1. The slab billets were austenitized at 1150 ℃ for 2 h in a resistance-heated furnace, and hot-rolled into 4.5 mm thick plates in seven passes by a pilot rolling mill with twin rolls of 450 mm in diameter. After hot rolling, the steel plates were cooled to 410 ℃ by a multi-step cooling process, and held for 50 min in a salt-bath furnace.

The specimens for microstructure observation were machined from the steel plates. The examined planes were parallel to the rolling direction. After mechanically polished, the samples were etched with 4% nital solution, and observed using a Leica DMIRM image analyzer. The contents of microstructural constituents (proeutectoid ferrite and bainite) were measured by Image Pro Plus.

Transmission electron microscopy (TEM) analysis was conducted to investigate more detailed microstructural characteristics. The thin foils for TEM were prepared by mechanical thinning from 300 μm to 80 μm along with twin-jet electropolishing in an electrolyte of 8% perchloric acid and 92% ethanol at the temperature of -30 ℃. TEM examination was performed on an FEI Tecnai G2 F20 TEM at an accelerating voltage of 200 kv.

RA phase characteristics were analyzed by electron back scattered diffraction (EBSD) and X-ray diffraction (XRD). The specimens for EBSD and XRD were obtained using a mechanical polishing followed by electrochemically polishing with a solution of 80% ethanol, 12% distilled water and 8% perchloric acid (volume percent) at a polishing voltage of 35 V. The specimens were examined at 0.02 μm scan step on an FEI Quanta 600 SEM equipped with EBSD system to evaluate RA particle size. The carbon content and volume fraction of RA were measured using a D/max 2400 diffractometer equipped with a CuK _α radiation. The specimens were scanned at a step size of 0.04 deg. in the 2θ range of 40 to 130 deg. under 55 kV and 180 mA. The integrated intensities of (200)_γ, (220) _γ and (311)_γ austenite peaks, and (200)_α along with (211)_α ferrite peaks were obtained by Jade version 6.5 software, and used to calculate the volume fraction of RA (V _γ) using Eq. (1) (Liu et al., 2018; Dong et al., 2021):

where I _γ and I _α are the average integrated intensities of above-mentioned austenite and ferrite diffraction peaks, respectively.

The carbon concentration in the RA was estimated by RA lattice parameter based on the (220)_γ peak using Eq. (2) (Li et al., 2018):

where a γ is the lattice parameter of RA in Å, and C γ is the carbon content of RA in wt%.

Rectangular tensile specimens with a gage length of 50 mm and a width of 12.5 mm were cut along the rolling direction, and tested using an Instron-8500 digital control tensile testing machine. Tensile properties were determined by the average values of tested results of five tensile specimens.

Figure 2 shows the metallographic microstructures of experimental steels subjected to different deformation temperatures. The contents of proeutectoid ferrite and bainite are listed in Table 3. It can be seen that the fractions of proeutectoid ferrite and bainite for two processes are obviously different. For Process A, the microstructure contains more amount of bainite (65.4%) but less proportion of proeutectoid ferrite (34.6%), while opposite results (i.e., 59.7% proeutectoid ferrite and 40.3% bainite) are observed in the case of Process B.

It is difficult to discern RA under optical microscope. Therefore, TEM analyses for RA were performed. In the case of Process A, as indicated in Fig. 3a, RA particles are present in the form of thin interlayer films along bainitic ferrite lath boundaries. Moreover, from Fig. 3b, it is clearly visible that RA has the Kurdjumov-Sachs (K-S) orientation relationship of {011}_Ferrite//{111}_Austenite and <111>_Ferrite//<011>_Austenite with adjacent bainitic ferrite.

TEM analysis results of RA for Process B were illustrated in Fig. 4. Similar to Process A, lamellar RA is also found to appear in the bainitic structure, and there exists definite orientation, i.e., Nishiyama-Wassermann (N-W) relationship of {110}_Ferrite//{111}_Austenite and <100>_Ferrite//<110>_Austenite, between RA and surrounding bainitic ferrite, as shown in Figs. 4a and 4b. However, in comparison to Process A, RA displays some differences from the view of morphology and distribution under the condition of Process B. Some blocky RA particles with different size are trapped in proeutectoid ferrite grains (Figs. 4c and 4d). Moreover, long RA strips as well as isolated RA islands are scattered in granular bainitic ferrite matrix (Figs. 4e and 4f).

Figure 5 shows XRD analysis results of RA under different deformation temperatures. Based on the XRD patterns, V _γ values are calculated to be 7.59% for Process A, and 12.1% for Process B, respectively. Obviously, RA content for Process B is higher than that in the case of Process A.

As shown above, deformation temperature in austenite non-recrystallization region has remarkable effect on the metallographic microstructures, especially RA characteristics. This is mainly due to the state change of prior austenite caused by deformation. The literature documented that deformation in the austenite non-recrystallization region can lead to strain-hardening of austenite, and lower deformation temperature results in more significant strain-hardening effect in prior austenite (Zarei Hanzaki et al., 1995).

In the strain-hardened austenite, some crystal defects caused by accumulated strain in non-recrystallisation region, such as dislocation structure and deformed band, are formed (Feng et al., 2014; Hosseini et al., 2014; Lan et al., 2019). Evidently, compared with higher deformation temperature in non-recrystallisation region (i.e., Process A), the density of the lattice defects is higher in the austenite deformed at lower temperature (Process B). After hot rolling, fast cooling (water cooling) to the given austenite+ferrite intercritical region temperature (i.e., 760 ℃) can efficiently inhibit restoration behaviour of strain-hardened austenite. Therefore, the crystal defect structures can be remained in prior austenite, and usually act as the preferred intragranular nucleation sites for proeutectoid ferrite (Hosseini et al., 2014). During slow cooling in intercritical region from 760 ℃ to 690 ℃, in the case of Process B, a large amount of intragranular proeutectoid ferrite grains are formed due to high density of lattice defects. Consequently, prior austenite grains are geometrically partitioned severely by intragranularly formed ferrite. This results in striking decrease in particle size of remaining austenite after intercritical region transformation. Moreover, the formation of a large amount of proeutectoid ferrite can cause noticeable carbon enrichment and decrease in remaining austenite amount surrounded by proeutectoid ferrite grains. For Process A, at the end of intercritical region, the microstructure has opposite characteristics from the perspectives of the size, amount and carbon content of remaining austenite as well as amount of proeutectoid ferrite.

In the course of the holding at temperature of 410 ℃, austenite to bainite transformation starts. As mentioned above, in the case of Process B, the remaining austenite has small size and low volume fraction, but contains high density of dislocations and high average carbon content. Carbon enrichment in austenite decreases the driving force of bainitic transformation (Ranjan et al., 2015; Nakagaito et al., 2017). In addtion, the growth of bainitic ferrite by a displacive mechanism is hindered by tangled dislocations in austenite (Lin et al., 2022). These factors lead to chemical and mechanical stabilization of remaining austenite against bainite transformation. Moreover, small size of remaining austenite prior to transformation limits the development of bainitic ferrite plates due to the impingement of a large amount of intragranular ferrite. As a result, remaining austenite to bainite transformation does not developed fully. Therefore, lower bainitie volume fraction can be observed for Process B.

Generally, carbon enrichment is more remarkable in small austenite particles during the austenite to ferrite transformation (Park et al., 2015; Liu et al., 2016). Moreover, small austenite particles have high mechanical stability against bainitic and martensitic transformations and thus, contribute to the austenite retention (Hanamura et al., 2013). In the case of Process B, among different-sized remaining austenite particles after the completion of proeutectoid ferrite reaction, those with much small size could be stabilized during 410 ℃ holding and subsequent air cooling, and become blocky RA distributed in proeutectoid ferrite or at the grain boundaries of ferrite.

In contrast to Process B, austenite to bainite transformation develops extensively in the case of Process A because of less carbon enrichment and low density of crystal defect structures in remaining austenite. In addition, large amount and big particle size of remaining austenite are also the factors causing high volume fraction of transformation product of bainite in the microstructure. The formation of a large amount of bainite consumes dramatically remaining austenite accompanied with carbon enrichment in film-like retained austenite distributed between bainite platelets. Consequently, although high content of remaining austenite, the resulting volume fractions of RA is low.

The deformation temperatures in non-recrystallization region not only cause aforementioned differences in microstructural characteristics, on the other hand, but also have a strong influence on mechanical properties of experimental steels. Therefore, the tensile properties of experimental steels were tested, and representative engineering stress-strain curves are given in Fig. 6a. From these curves, yield strength (YS), ultimate tensile strength (UTS), total elongation (TEL) and the product of ultimate tensile strength and total elongation (PSE) are obtained, and listed in Table 4.

The engineering stress-strain curves obtained from the tensile tests were converted to true stress-strain ones using Eqs. (3) and (4) (Grajcar et al., 2022).

where σ e and ε e are the engineering stress and strain, respectively. Similarly, σ and ε are the true stress and strain, respectively.

Hollomon equation is usually used to describe the dependence of true stress and strain (Karimi et al., 2017). According to Hollomon equation, instantaneous strain-hardening exponent is calculated using Eq. (5) (Kong et al., 2018).

where n _in is instantaneous strain-hardening exponent, which reflects the uniformly strained ability.

n _in against ε curves are shown in Fig. 6b. When n _in is equal to ε, necking is initiated (Kong et al., 2018). Therefore, a designed n _in = ε curve is also ploted in green dot dash line in Fig. 6b. The points of intersection of two kinds of curves are indicative of the values of maximum uniform true strain (ε _umax.). Uniform elongation values (UEL) are calculated based on ε _umax. data using Eq. (4), and are listed in Table 4.

Tensile experimental results indicate that, compared with Process A, the experimental steel for Process B exhibits lower strength level, but higher elongation (including TEL and UEL) and PSE, i.e., TEL, UEL and PSE are increased by 7.9%, 9.5% and 5 GPa·%, respectively. This is attributed to the differences in RA characteristics along with the volume fraction of microstructural constituents in the microstructures for two processes. For Process B, TRIP effect of a large amount of different-sized RA particles remarkably improves the ductility during tensile deformation. Besides, the higher proportion of proeutectoid ferrite (correspondingly, lower bainite amount) also enhances elongation to some extent, but decreases strength.

The contribution of TRIP effect for RA to the ductility can be analyzed well from strain-hardening behaviour shown in Fig. 6b. In the initial stage of deformation, n _in values for two experimental steels all decrease quickly to minimum level because of dynamic restoration of soft strained ferrite (Rastegari et al., 2015; Zou et al., 2017). For Process A steel, the minimum n _in value is higher than that of Process B steel because of larger amount of bainite with high density of dislocations. After that, two experimental steels show different strain dependence with increase in strain. In Process B steel, n _in value is gradually increased with strain, and maintained at higher level over a considerably wide strain range before the onset of necking. On the contrary, for Process A steel, the n _in value keeps lower and nearly constant over a narrower strain scope, without increasing in n _in value over the entire strain range.

As mentioned above, small-sized RA particle has high stability against deformation induced transformation to martensite due to higher carbon enrichment and size effect. Moreover, the kind of neighboring phase around RA also plays an important role in RA stability (Tan et al., 2020; Lavakumar et al., 2023). Soft proeutectoid ferrite matrix around RA can absorb the volume expansion resulted from martensite transformation, and is disadvantageous to RA retention during strain. Conversely, bainitic ferrite can act as a shield for enclosed RA. Hence, stress or strain is effectively prevented from imposing on RA, which suppresses the austenite to martensite transformation during deformation. Therefore, film-like RA existing along bainite lath boundary or long strip RA lying in granular bainitic ferrite possess higher stability than those embedded in proeutectoid ferrite or presented at ferrite boundary under the same carbon content and particle size.

For Process B steel, a large amount of big blocky RA with low stability easily transform to martensite in the early stage of deformation. This results in significant strain-hardening effect, causing a fast and marked increase in n _in value. With increase in strain, smaller-sized blocky RA and lamellar along with long strip RA begin to gradually and continually transform to martensite owing to higher stability, keeping higher n _in value over a wide strain region. This gradual and sustained strain-hardening effect can effectively inhibit the necking initiation. As a consequence, the uniform deformation is extended to high strain level, and the ductility is improved dramatically, i. e., TRIP effect. During further straining, due to the consumption of a large amount of RA, strain-hardening effect gets weakened accompanied with the occurrence of necking.

For Process A steel without different-sized blocky RA, small amount of film-like RA in bainitic structure provide weak and transient TRIP effect, resulting in lower elongation and poorer combination of ductility and strength compared with Process B steel.

Figure 7 shows the XRD data of RA before and after tensile fracture. The calculation results of V _γ and C _γ were shown in Table 5. V _γ after fracture is decreased strikingly from 12.1% to 2.79%, indicating that a large amount of RA transform to martensite during tensile deformation. In the meantime, C _γ is increased from 1.21% to 1.54%, which illustrates that un-transformed RA after fracture has higher carbon concentration compared with that before fracture.

Figure 8 indicates EBSD results for RA, which intuitively demonstrates RA particle size change before and after tensile fracture. It is obvious that after fracture, particle size of un-transformed RA becomes much smaller in comparison to that before fracture. Statistics results on RA particle size are shown in Fig. 9, and the values of RA particles average diameter (d _mean) are listed in Table 5. Before fracture, particle diameter of RA varies continually from small to big, and mainly distributes between 0.2 μm and 0.6 μm. d _mean is about 0.47 μm. After fracture, RA particles with diameter more than 0.4 μm almost disappear, and particles less than 0.2 μm in diameter have very high count frequency. d _mean is decreased to 0.12 μm. Above results reveal again that RA with small size and high carbon content has high stability.