Effect of Hot Rolling on Microstructural Evolution and Wear Behaviors of G20CrNi2MoA Bearing Steel

Hot rolling can improve the mechanical properties after heat treatment by improving the microstructure. The effect of hot rolling (HR) deformation on the microstructural transformation of G20CrNi2MoA bearing steel in the subsequent CQT (carburizing-quenching and tempering) and RQT (reheating-quenching and tempering) processes was studied. The results indicate that the austenite grain size decreased by 20% after 45% hot rolling reduction, and the number of large-angle grain boundaries increased due to the recovery and recrystallization induced by hot deformation. The refinement effect of hot deformation on austenite grains was retained after dual austenitizing, and the large-angle grain boundaries and massive dislocation in the grains caused by hot deformation promoted the diffusion of carbon atoms during carburization, resulting in a higher surface carbon concentration. The refined grains and higher carbon concentration affected the volume fraction and size of undissolved carbides in RQT specimens. When the initial hot rolling reduction reached 45%, the average particle size of carbides decreased by 40%, and the area volume fraction increased by 37%. The Vickers hardness increased, but the friction coefficient and wear rate were significantly reduced with the increase in the initial hot rolling reduction. The main reasons for the improved wear resistance were fine grains, superior carbide distribution and high hardness.


Introduction
Bearings are an important part of high-speed railway trains, and thus, their quality and stability receive considerable attention [1,2]. The fatigue performance and damage evaluation of the high-speed railway bearing steel are also taken seriously [3,4]. To bear the thermal-mechanical load during operation, the service performance requirements of high-speed railway bearings are mainly focused on two characteristics: high wear resistance of the surface and high impact toughness of the core. Carburized G20CrNi2MoA steel can fully meet these requirements after proper forming and heat treatment. In particular, the main manufacturing processes are hot-rolled ring forming, carburizingquenching; reheating-quenching; and additional low-temperature tempering after each quenching [5,6]. 2 of 14 To meet the increasing performance requirements, numerous studies have been conducted on the microstructural evolution of G20CrNi2MoA steel during the manufacturing process. From the forming process perspective, high-efficiency hot-rolled ring technology is characterized not only by material saving, energy efficiency and lower pollution but also by high forming accuracy. The team of Hua et al. [7] carried out a thorough systematic analysis of the rolling forming process from the aspects of design, theory [8,9] and experimental research [10,11]. From the perspective of microstructural evolution, hot rolling can not only refine the microstructure [12,13] but also yield a more uniform texture distribution [14,15] and provide a high-quality initial microstructure for subsequent heat treatment.
The carburizing-quenching technique is an important part of the heat treatment routes of G20CrNi2MoA steel. During this process, the part acquires a carbon concentration gradient, which is the decisive step in realizing the high wear resistance of the surface and high toughness of the core [16]. Arif Sugianto et al. [17] studied the carburization process of low-carbon steel in detail using finite elements. Their results showed that the carbon content of the surface hardening layer was related to the carbon potential of the carburizing gas, the initial carbon content of the steel and the carburization time and temperature. However, their analysis did not consider the deformation and heat treatment history of the material before carburization. Sougata Roy [18] studied the effect of the carburization process on the carbon concentration gradient and analyzed the wear behavior under dry sliding conditions. Their results indicated that abrasive wear resistance was enhanced due to the high surface carbon content, which was characterized by high hardness and retained austenite content. Early studies confirmed the positive effect of double quenching and tempering heat treatment on the fatigue resistance of steel [19,20]. Over nearly a decade, scholars further studied the microstructures and mechanical properties of steels with different carbon content after double quenching and tempering [21,22]. The findings of these studies suggested that the substantial refinement of austenite grains was the reason for the remarkable improvement of impact toughness, and the yield strength and tensile strength were also improved. However, it is worth noting that these studies did not focus on the microstructural evolution and mechanical properties after secondary quenching when the prior microstructure was different.
Previous experiments have confirmed the significant effect of cold ring rolling on the microstructural evolution of GCr15 bearing steel during quenching and tempering [23]. The main reason for this effect is that the deformed initial microstructures lead to the refinement of austenite grains and the accelerated dissolution of carbides in the subsequent austenitization step [24], thus affecting the microstructure transformation during quenching and tempering [25,26]. However, the influence mechanism of hot deformation on the microstructural evolution of carburized steel during subsequent heat treatment may be more complex. In particular, the material undergoes multiple austenitization cycles after the initial hot deformation. Additionally, the material has low carbon content during the hot deformation and high carbon content after carburization. Therefore, the effects of hotrolled ring deformation on the evolution of austenite grains, the formation of the carbon concentration gradient and undissolved carbides during subsequent heat treatment are still unclear. Previous studies have mainly focused on the effect of process parameters on the microstructures and properties within a single process step. Authors in [27,28] studied the dry sliding wear behavior of martensite steel under different friction conditions. However, only conventional quenching and tempering was mentioned, and the carburizing process was not involved. Authors in [29] investigated the microstructure and wear resistance of carburized 20CrNi2MoV steel under cryogenic treatment conditions. However, the effect of reheating-quenching had not been considered. Authors in [30] analyzed the dynamic recrystallization behavior of 20CrNi2Mo steel and established the critical stress and strain model. Nevertheless, the effect of subsequent heat treatment on the microstructure and properties was not discussed. The lack of systematic research on the microstructural evolution throughout the whole forming process, especially the effect of hot rolling on subsequent steps, limits the ability to further optimize the mechanical properties of highspeed railway bearings.
The aim of this study was to analyze the effect of prior hot rolling on the microstructural evolution of carburized G20CrNi2MoA bearing steel throughout the whole forming process. The effects of different hot rolling reduction rates on the evolution of austenite grains after hot deformation, carburizing-quenching and reheating-quenching were studied. The carbon concentration distributions after carburization with different hot rolling reductions were measured, and the phase contents after carburizing-quenching and reheating-quenching were quantified. After the whole forming and heat treatment process, the surface friction properties and hardness of specimens were measured.

Material and Experiments
The raw material used in this work was a typical G20CrNi2MoA carburized bearing steel (same as SAE 4320), which was received as an annealed bar with a size of φ 120 mm × 120 mm with varying thicknesses. The chemical composition of the steel, which was measured by speck spectrograph, is listed in Table 1. Four sets of plate specimens with dimensions of 120 mm × 60 mm × 21.8 mm, 120 mm × 60 mm × 17.1 mm, 120 mm × 60 mm × 14.1 mm and 120 mm × 60 mm × 12 mm were prepared from the steel bar for further deformation and heat treatment. All plate specimens were heated to 1000 • C for hot rolling (HR). The finishing rolling temperature was controlled so that it remained above 900 • C, and air cooling was applied after rolling. The thickness reduction rates were 0 (raw material), 15, 30 and 45%, respectively, to ensure that the final thickness of the rolled plate specimens with different initial sizes was 12 mm. The specimens are named HR-1, HR-2, HR-3 and HR-4, respectively. Subsequently, the HR specimens were directly heated to 930 • C and held for 21 h in a carburizing furnace with a constant carbon potential of 1.28% wt%. After reaching the specified holding time, all specimens were slowly cooled to 870 • C, oil quenched to room temperature and then tempered at 190 • C for 4 h. Specimens that had been subjected to carburizing-quenching and low-temperature tempering are named CQT-1, CQT-2, CQT-3 and CQT-4, respectively. After the CQT process, a reheating-quenching and low-temperature tempering (RQT) process was implemented. All specimens were heated to 805 • C, held for 1 h for austenitization and then oil quenched to room temperature. The parameters of the second low-temperature tempering were the same as those of the first cycle. These specimens are named RQT-1, RQT-2, RQT-3 and RQT-4, respectively. A schematic diagram of the HR, CQT and RQT of the specimens is shown in Figure 1. The conditions and corresponding abbreviations are listed in Table 2, and the schematic diagram of the forming process of the specimens is shown in Figure 1.
The grain morphology of HR specimens was analyzed by electron backscatter diffraction (EBSD). The grain morphology of CQT and RQT specimens, which were prepared with a supersaturated picric acid solution and Seagull detergent at 90 • C for 10 s, was observed with a Zeiss metallographic microscope (Gesellschaft mit beschrankter Haftung, Göttingen, Germany). The grain sizes were statistically analyzed by Image Pro Plus software (Version 6.0.0.260, Media Cybernetics Inc., Singapore), and over 400 grains were examined. The carbon concentrations along the thickness direction of the CQT and RQT specimens were measured by speck spectrograph. The amount of retained austenite was measured by X-ray diffraction (XRD) (D/max-2500, Rigaku, Tokyo, Japan) with Mo Kα and a scanning speed of 3 • min −1 . The morphology of carbides in CQT and RQT specimens was observed with a JEM-7500F field-emission scanning electron microscope (FESEM, Japan Electron Optics Laboratory, Beijing, China). The equivalent radius and distribution of carbide were statistically analyzed by Image Pro Plus software. The grain morphology of HR specimens was analyzed by electron backscatter dif fraction (EBSD). The grain morphology of CQT and RQT specimens, which were prepared with a supersaturated picric acid solution and Seagull detergent at 90 °C for 10 s, was observed with a Zeiss metallographic microscope (Gesellschaft mit beschrankter Haftung Göttingen, Germany). The grain sizes were statistically analyzed by Image Pro Plus soft ware (Version 6.0.0.260, Media Cybernetics Inc., Singapore), and over 400 grains were ex amined. The carbon concentrations along the thickness direction of the CQT and RQT specimens were measured by speck spectrograph. The amount of retained austenite was measured by X-ray diffraction (XRD) (D/max-2500, Rigaku, Tokyo, Japan) with Mo Kα and a scanning speed of 3° min −1 . The morphology of carbides in CQT and RQT specimens was observed with a JEM-7500F field-emission scanning electron microscope (FESEM, Ja pan Electron Optics Laboratory, Beijing, China). The equivalent radius and distribution o carbide were statistically analyzed by Image Pro Plus software.
Vickers hardness of the RQT specimens was measured with a HV-1000 Vickers hard ness tester (Laizhou Huayin Test Instrument Co., LTD, Laizhou, China). The loading force for the tests was 500 g, and the duration was 10 s. For each specimen, seven sites were tested, and then the values were averaged as the final result. The frictional properties o the RQT specimens were measured on an HT-1000 high-temperature friction and pin-on disk wear testing machine under dry sliding conditions. The wear test specimens ma chined from the RQT specimens were 25 mm in diameter and 15 mm in thickness. Wear tests were performed on the specimens against a ZrO2 ball with a diameter of 3 mm. The tests were conducted under an applied load of 20 N, with a fixed sliding velocity of 0.236 ms −1 . Each wear test was repeated three times to ensure the precision of the results. The friction coefficient was recorded in real time by the apparatus. Before and after the test the pin and disk were soaked in petroleum ether and cleaned by ultrasonic cleaners for 5 min, and then mass loss values of the pin and disk were determined from weight differ ences by the precise BSA224S electronic balance (Shanghai Xiren Scientific Instrument Co. LTD, Shanghai, China)with an accuracy of ±0.1 mg. The friction surface and cross-sec tional morphology were observed with a field-emission scanning electron microscope (FESEM) and atomic force microscope (AFM).

Variation in Grain Size
The grain morphology of different HR specimens is shown in Figure 2. The grain size of HR-1 in Figure 2a was visibly larger than that of HR-4 in Figure 2d. For specimens HR

Process
Thickness Reduction Rates 0% 15% 30% 45% Vickers hardness of the RQT specimens was measured with a HV-1000 Vickers hardness tester (Laizhou Huayin Test Instrument Co., LTD, Laizhou, China). The loading force for the tests was 500 g, and the duration was 10 s. For each specimen, seven sites were tested, and then the values were averaged as the final result. The frictional properties of the RQT specimens were measured on an HT-1000 high-temperature friction and pin-on-disk wear testing machine under dry sliding conditions. The wear test specimens machined from the RQT specimens were 25 mm in diameter and 15 mm in thickness. Wear tests were performed on the specimens against a ZrO 2 ball with a diameter of 3 mm. The tests were conducted under an applied load of 20 N, with a fixed sliding velocity of 0.236 ms −1 . Each wear test was repeated three times to ensure the precision of the results. The friction coefficient was recorded in real time by the apparatus. Before and after the test, the pin and disk were soaked in petroleum ether and cleaned by ultrasonic cleaners for 5 min, and then mass loss values of the pin and disk were determined from weight differences by the precise BSA224S electronic balance (Shanghai Xiren Scientific Instrument Co., LTD, Shanghai, China)with an accuracy of ±0.1 mg. The friction surface and cross-sectional morphology were observed with a field-emission scanning electron microscope (FESEM) and atomic force microscope (AFM).

Variation in Grain Size
The grain morphology of different HR specimens is shown in Figure 2. The grain size of HR-1 in Figure 2a was visibly larger than that of HR-4 in Figure 2d. For specimens HR-2 to HR-4, the grain size showed a decreasing trend as the hot rolling reduction increased. Grain refinement phenomenon and mechanism by hot rolling deformation has been studied by many scholars. Guo et al. [31] found hot rolling can obviously refine the grain size, while grain refinement has limitations at a certain sufficient plastic deformation. Zhao et al. [32] pointed out that the grain size was refined through dynamic recrystallization. A large number of dislocations caused by plastic deformation increases the driving force for dynamic recrystallization, which will promote the sites for nucleation and growth of grains. al. [32] pointed out that the grain size was refined through dynamic recrystallization. A large number of dislocations caused by plastic deformation increases the driving force fo dynamic recrystallization, which will promote the sites for nucleation and growth o grains. The statistical results of EBSD diagrams are shown in Figure 3. As the hot rollin reduction increased, the average grain size of the raw material decreased from 13.0 μm (HR-1) to 11.1 μm (HR-2), 10.7 μm (HR-3) and 10.4 μm (HR-4). The proportion of grai sizes smaller than 10 μm was 34.51% in HR-1 and increased to 56.55% after 45% defor mation, which indicates that the fine grains markedly increased after hot rolling reduc tion. The major difference in the misorientation distribution is reflected by the frequenc of large-angle grain boundaries, as shown in Figure 3b. The statistical results for large angle grain boundaries (>15°) revealed that the proportion increased from 45.67 to 51.73% with the increase in hot rolling reduction. The statistical results of EBSD diagrams are shown in Figure 3. As the hot rolling reduction increased, the average grain size of the raw material decreased from 13.0 µm (HR-1) to 11.1 µm (HR-2), 10.7 µm (HR-3) and 10.4 µm (HR-4). The proportion of grain sizes smaller than 10 µm was 34.51% in HR-1 and increased to 56.55% after 45% deformation, which indicates that the fine grains markedly increased after hot rolling reduction. The major difference in the misorientation distribution is reflected by the frequency of largeangle grain boundaries, as shown in Figure 3b. The statistical results for large-angle grain boundaries (>15 • ) revealed that the proportion increased from 45.67 to 51.73% with the increase in hot rolling reduction.
The grain morphology of CQT and RQT specimens is presented in Figure 4 where d is the average grain size. The average grain size of CQT specimens increased to about 16 µm, while those of RQT specimens decreased to nearly 4 µm. With the increase in the initial hot rolling reduction, the average grain size of CQT and RQT samples decreased, with a narrow variation range.
The above analysis indicates that there were clear differences in grain size at different processing stages. The austenite grain sizes at different process steps decreased in the following order: CQT (~16 µm), HR (~11 µm) and RQT (~4 µm). The main factors affecting grain size varied between different stages. For the HR process, the deformation energy storage introduced by hot rolling provided the driving force for grain recrystallization [33] and thus led to a reduction in grain size. In the subsequent process, the long-time holding at 930 • C during carburization and the low austenitizing temperature (805 • C) in the RQ process were the main factors that determined the grain size of CQT and RQT specimens, respectively. The grain size was significantly reduced by 75% after the RQT process comparing with previous research [34,35] in which the result was 49% and 48%, respectively. The same phenomenon was also found in the literature [21,22]. There are a large number of defects in the microstructure after CQT process, which is beneficial to the nucleation of austenite grains during reheating-quenching process. Another interesting observation was that the difference in grain size among HR specimens markedly decreased as the deformation increased, while this trend was not evident in the other two stages. In the CQT stage, the long-time holding during carburization weakened the effect of the rolled microstructure on grain refinement. Therefore, the differences in the grain sizes of CQT specimens with different initial hot rolling reduction rates were very small, and thus, their differences after RQT treatment were small. In other words, the refinement effect of hot rolling on the grain size of RQT specimens was weakened after multiple austenitizing processes. The refined grains of HR specimens may further affect the subsequent microstructural evolution, especially the diffusion behavior of carbon during the carburizing process and the final properties, as discussed below. The grain morphology of CQT and RQT specimens is presented in Figure 4 where d is the average grain size. The average grain size of CQT specimens increased to about 16 μm, while those of RQT specimens decreased to nearly 4 μm. With the increase in the initial hot rolling reduction, the average grain size of CQT and RQT samples decreased, with a narrow variation range. The above analysis indicates that there were clear differences in grain size at different The grain morphology of CQT and RQT specimens is presented in Figure 4 where d is the average grain size. The average grain size of CQT specimens increased to about 16 μm, while those of RQT specimens decreased to nearly 4 μm. With the increase in the initial hot rolling reduction, the average grain size of CQT and RQT samples decreased, with a narrow variation range. The above analysis indicates that there were clear differences in grain size at different processing stages. The austenite grain sizes at different process steps decreased in the following order: CQT (~16 μm), HR (~11 μm) and RQT (~4 μm). The main factors affecting grain size varied between different stages. For the HR process, the deformation energy

Carbon Concentration Distribution Behavior
The carbon gradient of specimens after carburization as a function of distance from the surface is shown in Figure 5. The carbon concentration along the thickness direction of all CQT specimens presented a decreasing trend, as illustrated in Figure 5a. The RQT specimens generally inherited the carbon concentration of this distribution. In particular, the carbon concentration of the specimens at the same depth increased with the increase in the initial deformation, which indicates that hot rolling promoted the diffusion of carbon during carburization. Figure 5c,d represents the relationship between the surface carbon concentration of the specimens and the hot rolling reduction. The surface carbon concentration of the RQT specimens is notably lower than the CQT ones because of the decarburization phenomenon during the RQT process. Nevertheless, the increasing trend of the surface carbon concentration with the increase of hot rolling reduction did not change.
of the rolled microstructure on grain refinement. Therefore, the differences in the grain sizes of CQT specimens with different initial hot rolling reduction rates were very small, and thus, their differences after RQT treatment were small. In other words, the refinement effect of hot rolling on the grain size of RQT specimens was weakened after multiple austenitizing processes. The refined grains of HR specimens may further affect the subsequent microstructural evolution, especially the diffusion behavior of carbon during the carburizing process and the final properties, as discussed below.

Carbon Concentration Distribution Behavior
The carbon gradient of specimens after carburization as a function of distance from the surface is shown in Figure 5. The carbon concentration along the thickness direction of all CQT specimens presented a decreasing trend, as illustrated in Figure 5a. The RQT specimens generally inherited the carbon concentration of this distribution. In particular, the carbon concentration of the specimens at the same depth increased with the increase in the initial deformation, which indicates that hot rolling promoted the diffusion of carbon during carburization. Figure 5c,d represents the relationship between the surface carbon concentration of the specimens and the hot rolling reduction. The surface carbon concentration of the RQT specimens is notably lower than the CQT ones because of the decarburization phenomenon during the RQT process. Nevertheless, the increasing trend of the surface carbon concentration with the increase of hot rolling reduction did not change. During the carburizing process, the long-range diffusion of carbon atoms occurred simultaneously with austenite grain growth. The diffusion behavior of carbon atoms at During the carburizing process, the long-range diffusion of carbon atoms occurred simultaneously with austenite grain growth. The diffusion behavior of carbon atoms at the austenite grain boundary followed an interface diffusion mechanism, in which the diffusion rate was much greater than that within the crystal [36]. The dislocations and vacancies reduced diffusion energy to form a fast diffusion channel. The EBSD results for grain size and distribution revealed that the grain number and the proportion of high-angle grain boundaries increased with the increase in hot rolling reduction. It is reasonable to infer that the specimens subjected to intense thermal deformation had smaller grains and massive dislocations in the early stage of the carburizing process, which promoted the diffusion of carbon atoms during isothermal holding. Therefore, the high diffusion rate of carbon atoms favored an increase in the overall carbon concentration in the depth range of the carburized layer, as presented in Figure 5. However, the low austenitizing temperature in the RQT stage was not conducive to the long-range diffusion of carbon atoms through the depth of the material, resulting in a similar carbon gradient in RQT specimens. In addition, the difference in carbon content between CQT specimens induced by different initial hot rolling reduction rates would produce differences in the microstructures of RQT specimens, especially carbides.

Evolution of Carbides
The SEM morphology of CQT specimens is shown in Figure 6a-d. The microstructures mainly consisted of coarse martensite and large amounts of retained austenite. No carbides were observed at low magnification, but nano-sized carbides were observed in the matrix at high magnification, as presented in Figure 6e-h. Although the carbon content of the surface layer was as high as 0.84-0.92% wt.% (Figure 5c) after carburizing treatment, the carbon atoms were mainly in martensite and retained austenite in the form of solid solution after long holding at 930 • C. The nano-carbides were mainly precipitated during tempering. ls 2021, 11, x FOR PEER REVIEW 9 of Figure 6. SEM morphology in different hot rolling reduction after the CQT and RQT processes: (a-d) CQT specimens at the multiples of 5K; (e-h) CQT specimens at the multiples of 100K; (i-l) RQT specimens.
The SEM morphology of the RQT specimens is shown in Figure 6i-l. In addition tempered martensite and retained austenite, high carbide content was observed. Add tionally, the statistical results in the figures indicate that the average size (S) of carbides RQT specimens gradually decreased, while the area fraction (A) increased with the crease in the initial hot rolling reduction. Figure 7 shows the XRD data of the CQT, RQ and RQT specimens in different stat Based on the data in Figure 7a, the content of retained austenite in CQT-1 was 30.2% a increased to 35.1% in CQT-4. During carburization, the carbon concentration increas with the increase in hot rolling deformation (Figure 5a), resulting in an increase in retain The SEM morphology of the RQT specimens is shown in Figure 6i-l. In addition to tempered martensite and retained austenite, high carbide content was observed. Additionally, the statistical results in the figures indicate that the average size (S) of carbides in RQT specimens gradually decreased, while the area fraction (A) increased with the increase in the initial hot rolling reduction. Figure 7 shows the XRD data of the CQT, RQ and RQT specimens in different states. Based on the data in Figure 7a, the content of retained austenite in CQT-1 was 30.2% and increased to 35.1% in CQT-4. During carburization, the carbon concentration increased with the increase in hot rolling deformation (Figure 5a), resulting in an increase in retained austenite content after CQT treatment. Figure 7b shows the XRD diffraction data of RQ specimens. The retained austenite content was about 23%, and the martensite peak shifted to the left. After low-temperature tempering, the content of retained austenite did not significantly decrease, but the martensite peak moved to the right, as illustrated in Figure 7c. The c/a values of martensite in RQ and RQT specimens are presented in Figure 7d. With the increase in hot rolling deformation, the c/a values of RQ specimens increased, while those of RQT specimens substantially decreased, and the difference between RQ and RQT specimens increased. The carbides in carburized steel after the final heat treatment were from multiple sources. Before the RQ process, many fine dispersed carbides already existed in the matrix due to the martensite tempering in the CQT stage, as shown in Figure 6e-h. In addition, the decomposition of retained austenite in CQT specimens during the heating stage of the RQ process also produced carbides. The sources of the CQT carbides were the fine carbides precipitated from tempered martensite and the cementite produced by the decomposition of retained austenite. During the RQT stage, the carbides were mainly precipitated from martensite because the change in retained austenite content before and after the second low-temperature tempering was not significant. To differentiate them from other types, these carbides are defined as RQT carbides. Therefore, CQT carbides and RQT carbides constituted the complex carbide system of carburized bearing steel after the final heat treatment.
Many factors affect the size of carbides during heat treatment. Generally speaking, the dissolution rate of carbide during austenitization is positively related to its initial size and also promoted by fine grains [37], providing more diffusion paths. During tempering, the carbide size is affected by the carbon content of the martensite matrix and the original austenite grains. From the perspective of kinetics, the former promotes carbide precipitation [38], and the latter affects the number of defects and provides more effective sites for carbide nucleation [39]. The fine CQT carbides were easier to dissolve during re-austenitization, which refined the residual undissolved carbides and increased the carbon content in the matrix after the RQT process, which is evidenced by the slight left shift of the The carbides in carburized steel after the final heat treatment were from multiple sources. Before the RQ process, many fine dispersed carbides already existed in the matrix due to the martensite tempering in the CQT stage, as shown in Figure 6e-h. In addition, the decomposition of retained austenite in CQT specimens during the heating stage of the RQ process also produced carbides. The sources of the CQT carbides were the fine carbides precipitated from tempered martensite and the cementite produced by the decomposition of retained austenite. During the RQT stage, the carbides were mainly precipitated from martensite because the change in retained austenite content before and after the second low-temperature tempering was not significant. To differentiate them from other types, these carbides are defined as RQT carbides. Therefore, CQT carbides and RQT carbides constituted the complex carbide system of carburized bearing steel after the final heat treatment.
Many factors affect the size of carbides during heat treatment. Generally speaking, the dissolution rate of carbide during austenitization is positively related to its initial size and also promoted by fine grains [37], providing more diffusion paths. During tempering, the carbide size is affected by the carbon content of the martensite matrix and the original austenite grains. From the perspective of kinetics, the former promotes carbide precipitation [38], and the latter affects the number of defects and provides more effective sites for carbide nucleation [39]. The fine CQT carbides were easier to dissolve during reaustenitization, which refined the residual undissolved carbides and increased the carbon content in the matrix after the RQT process, which is evidenced by the slight left shift of the martensite peak {001}α' in Figure 7b. The martensite peak was shifted to the right after the second low-temperature tempering, indicating that more carbon precipitated from the supersaturated martensite to form carbides. The data show that the RQT carbide content increased gradually with the increase in the initial hot rolling deformation. Therefore, the high initial thermal deformation and the cumulative effect of multiple heat treatments resulted in smaller particle sizes and a more concentrated distribution in RQT-4. Thus, the high hot rolling deformation increased the carbon concentration of carburizing quenched specimens and ultimately increased the carbide volume fraction of specimen RQT-4.

Hardness Analysis
The Vickers hardness of the RQT specimens is depicted in Figure 8. These results show that the Vickers hardness gradually increased with the increase in the initial hot rolling reduction. Previous studies have discussed in detail the factors that affect the hardness of tempered martensitic steel [40]. A strong, linear relationship between the hardness and the volume fraction of carbide has been reported [41]. Previous studies have also suggested that numerous fine dispersed carbides substantially help to improve the hardness [42,43]. In this work, the factors that significantly affected the Vickers hardness were the volume fraction and size of carbides. Compared with RQT-1, the carbides in RQT-4 had a higher volume fraction and smaller average size, which contributed to the improvement of the hardness. The Vickers hardness of the RQT specimens is depicted in Figure 8. These results show that the Vickers hardness gradually increased with the increase in the initial hot rolling reduction. Previous studies have discussed in detail the factors that affect the hardness of tempered martensitic steel [40]. A strong, linear relationship between the hardness and the volume fraction of carbide has been reported [41]. Previous studies have also suggested that numerous fine dispersed carbides substantially help to improve the hardness [42,43]. In this work, the factors that significantly affected the Vickers hardness were the volume fraction and size of carbides. Compared with RQT-1, the carbides in RQT-4 had a higher volume fraction and smaller average size, which contributed to the improvement of the hardness.

Wear Resistance Properties
The wear morphology of the specimens is shown in Figure 9. In Figure 9a, significant plastic flow and furrows are clearly observed on the surface of the RQT-1 specimen. With the increase in the initial hot rolling deformation, the furrows on the sample surface were replaced by folds, shallow grooves and peeling debris, as shown in Figure 9c,d. These phenomena are consistent with abrasive wear characteristics. The corresponding wear tracks in Figure 9e-h suggest that the groove depth on the worn surface decreased with the increase in the initial hot rolling deformation, as well. The groove depth of the RQT-4 specimen decreased to about 40% of that of RQT-1 (from 377.1 to 145.9 nm). The crosssection of the wear track (Figure 9i-l) showed visible traces of carbide flaking in RQT-1.
However, this phenomenon gradually disappeared with the increase in the initial hot rolling deformation, and no coarse carbides were observed in the cross-section.
The wear morphology of the specimens is shown in Figure 9. In Figure 9a, significant plastic flow and furrows are clearly observed on the surface of the RQT-1 specimen. With the increase in the initial hot rolling deformation, the furrows on the sample surface were replaced by folds, shallow grooves and peeling debris, as shown in Figure 9c,d. These phenomena are consistent with abrasive wear characteristics. The corresponding wear tracks in Figure 9e-h suggest that the groove depth on the worn surface decreased with the increase in the initial hot rolling deformation, as well. The groove depth of the RQT-4 specimen decreased to about 40% of that of RQT-1 (from 377.1 to 145.9 nm). The crosssection of the wear track (Figure 9i-l) showed visible traces of carbide flaking in RQT-1. However, this phenomenon gradually disappeared with the increase in the initial hot rolling deformation, and no coarse carbides were observed in the cross-section. Figure 9. Morphology of RQT specimens in different hot rolling reduction after friction experiments: (a-d) worn surface; (e-h) AFM maps; (i-l) worn cross-section. Figure 10 shows the variation in the wear rate and friction coefficient of RQT specimens with different initial hot rolling reduction rates. Figure 10a reveals that with the increase in hot rolling deformation, the wear rate followed a significant declining trend. Figure 10b shows that, in the whole friction process, after a short period of initial wear, the friction behavior entered a stable wear stage, and the change in the friction coefficient was relatively stable over time. The results indicate that the friction coefficient had a trend of reduction with the increase in hot rolling deformation. Compared with the raw material without deformation, the friction coefficient decreased by about 50% after the 45% hot rolling reduction.  Figure 10 shows the variation in the wear rate and friction coefficient of RQT specimens with different initial hot rolling reduction rates. Figure 10a reveals that with the increase in hot rolling deformation, the wear rate followed a significant declining trend. Figure 10b shows that, in the whole friction process, after a short period of initial wear, the friction behavior entered a stable wear stage, and the change in the friction coefficient was relatively stable over time. The results indicate that the friction coefficient had a trend of reduction with the increase in hot rolling deformation. Compared with the raw material without deformation, the friction coefficient decreased by about 50% after the 45% hot rolling reduction. During dry friction, the coarse carbides, especially those with sharp edges and corners, had a high probability of falling off and leaving deep wear furrows on the friction surface, which would increase the wear rate, even at low deformation. Furthermore, the existence of deep furrows reduced the effective contact area between the friction pairs, leading to a high friction coefficient. On the other hand, when the number and distribution