2020 Volume 60 Issue 8 Pages 1774-1783
The effect of retained austenite (γR) on the rolling contact fatigue (RCF) properties of carburized SAE4320 steel was carefully investigated. We prepared specimens comprising four volume fractions of γR from 6% to 39% by controlling the subzero heat treatment. The effect of γR on the RCF was investigated using hardness measurements, X-ray diffraction, scanning electron microscopy (SEM), and transmission electron microscopy. The RCF test revealed that the sub-surface initiated spalling life was prolonged as the volume fraction of γR increased. In the area at the depth z0 where the orthogonal shear stress was maximum, the majority of the γR was transformed to martensite, thus resulting in a significant increase in the Vickers hardness. The result of SEM observation showed that the region initially comprising γR exhibited a high resistance to RCF. Moreover, the TEM analysis revealed that the initial γR region changed into a mixture of very fine hard martensite and some unchanged γR during RCF. This suggests that the transformation of γR into fine hard martensite during RCF contributed to the improvement of the RCF life.
In recent years, awareness regarding the protection of the global environment has been increasing worldwide. In addition, a reduction in the size and weight of rolling bearings is required to contribute to energy and resource savings. Rolling bearings are mechanical parts that support loads via a point or line contact, and a high contact stress of several GPa acts locally in rolling bearings. Therefore, high-strength steels are used for fabrication of the bearing races and the rolling elements. As the fabricated bearings become smaller and lighter, the contact condition becomes more severe, and thus, the microstructural control of steel having superior rolling contact fatigue (RCF) properties is required to be realized.
The failure of bearings due to RCF can be categorized as sub-surface initiated spalling and surface initiated spalling depending on the difference in the process of spalling. Among these, surface initiated spalling is a mode that occurs due to the inclusion of hard foreign substances in the lubricant and also owing to poor lubrication. Many studies1,2,3,4) have highlighted the relationship between the microstructure of the steel and the dent initiated spalling behavior that is caused by hard foreign substances entering between the bearing race and the rolling element. These studies explained that the dent initiated spalling life was improved by utilizing soft retained austenite (γR) among the martensite and carbide (carbonitride), which dominated the microstructure of quenched and tempered steel. The proposed mechanism was based on the ideas of 1) controlling the dent shape in order to reduce the stress concentration1,2,3) and 2) suppressing the microstructural fatigue via the deformation-induced martensitic transformation of γR4) during rolling contact.
Sub-surface initiated spalling of the bearing is a mode that is initiated from non-metallic inclusions existing in steel. Therefore, improving the steel cleanliness by reducing the size and quantity of non-metallic inclusions is the main countermeasure for suppressing sub-surface initiated spalling,5) and thus, the required improvement are implemented in the steelmaking stage. However, in recent years, the cleanliness has reached a very high level due to advancements in steelmaking technology, and the improvement of the sub-surface initiated spalling life owing to this high cleanliness has plateaued.6,7) Therefore, the suppression of the nucleation and propagation of cracks by strengthening the microstructure around the non-metallic inclusions is considered to contribute to the improvement of sub-surface initiated spalling life. As the fatigue progresses and the local microstructure alternates through cyclic stresses in the fatigued area, which then becomes the initiation point of cracking under the rolling contact surface,8) it is necessary to suppress this microstructural alteration in order to improve the sub-surface initiated spalling life. For this purpose, it has been recognized that the adjustment of steel chemical compositions6,8,9) and optimization of its microstructure via heat treatment are effective. Regarding the adjustment of steel compositions, effort is focused on increasing the resistance to temper softening of the martensite and suppression of the microstructural alteration by adding Si, Cr, and Mo. However, measures for improving the bearing life without relying on these additions are required because these additions deteriorate the workability of steels, and in addition, Cr and Mo are rare alloying elements. Therefore, it is necessary to improve the fatigue strength by optimizing the microstructure via heat treatment.
γR is known to improve the strength and mechanical properties in various types of steels. A representative example of this is transformation-induced plasticity steel.10) In terms of the influence of γR on the RCF, the presence of γR is generally considered to be beneficial.
Tsushima and Maeda11) suggested that γR and RCF are not directly related. Based on the idea of mitigating the stress concentration around the non-metallic inclusions as described subsequently, it was explained that γR had no effect on the improvement of the sub-surface initiated spalling life in the case of the high cleanliness steels produced in recent years. Furthermore, in the past, an observation of a decrease in RCF with γR in high-speed tool steel for bearing reported by Scott and Blackwell,12) and no specific explanatory factor was mentioned.
The general consensus that γR improves RCF has been well documented.7,8,13,14,15,16,17,18) Various mechanisms have been proposed to explain the improvement in the sub-surface initiated spalling life due to γR. Shiko et al.8) estimated that γR improves the sub-surface initiated spalling life because of the deformation-induced martensitic transformation that occurs prior to the decomposition process of martensite during the RCF and the retardation of the fatigue progress. Similarly, Zhu et al.13) and Dommarco et al.14) have described the positive effect of the deformation-induced martensitic transformation of γR. Shoji and Eguchi15) have mentioned the possibility that the plastic deformation of the rolling contact surface increases owing to the existence of soft γR, and the sub-surface initiated spalling life improves as the substantial contact stress decreases. However, this phenomenon has been disproved experimentally, which indicates that γR improves the sub-surface initiated spalling life via transformation to martensite during the rolling contact in their report.15) Yajima et al.16) have stated that the properties of martensite basically influence the sub-surface initiated spalling life, but γR also has the auxiliary effect of improving it, and it has been presumed as a mechanism for delaying the propagation of cracks and impeding the progress of martensite tempering. Muro et al.17) has explained that the sub-surface initiated spalling life is improved by relaxing the stress concentration due to the existence of soft γR around non-metallic inclusions. Several reports have indicated the influence of γR on the sub-surface initiated spalling life as described above. However, a unified view has not been obtained in contrast to the case of dent initiated spalling life. In addition, the majority of mechanisms for improving the sub-surface initiated spalling life due to γR have been speculative, and a satisfactory explanation based on experimental results is yet to be arrived at.
In this study, we focused on the sub-surface initiated spalling life and aimed to clarify the relationship between the amount of γR and the sub-surface initiated spalling life by preparing specimens with intentionally altered volume fractions of γR and performing RCF tests on them. Furthermore, another objective of this study was to investigate the effect of γR on the sub-surface initiated spalling life by clarifying the influence of the behavior of γR on the microstructural alteration of steel during rolling contact.
In order to investigate the influence of γR on the sub-surface initiated spalling life, SAE4320 (0.2%C steel), which contains a large amount of Ni—1.7 mass% (hereinafter mass% is abbreviated as%)—was used. Addition of Ni, as a γ-stabilizing element, makes it easy to control the amount of γR via a heat treatment. The chemical composition of the steel used is presented in Table 1. The hot-rolled steel bar (26 mm in diameter) of SAE4320 fabricated using the actual production equipment was cut and heat treated as described below and then finished to obtain a cylindrical shape of 20 mm in diameter and 36 mm in length after performing grinding and super finishing. The heat treatment of the specimens included carburizing, quenching, subzero treatment and subsequent tempering. The amount of γR was intentionally altered by performing a subzero treatment after the quenching and adjusting the treatment temperature. First, the above cylindrical specimens were carburized at 960°C in a carburizing atmosphere for 26 h in such a manner that the carbon concentration in the surface layer was approximately 1.1%. Following carburization, the cooled specimens were heated to 820°C, maintained for 70 min, and oil quenched at 80°C. Finally, tempering was performed by maintaining the specimen at 180°C for 2 h. The microstructure of specimen without subzero treatment prepared using above process was comprised of tempered martensite as the basic microstructure, γR and cementite. In the specimens with a reduced amount of γR, subzero treatments were performed for 1 h at −50°C, −80°C, or −196°C after the quenching before the tempering to transform some austenite into martensite. The specimens with different amount of the γR were subjected to RCF testing. In order to obtain an insight into the influence of the microstructure on the RCF, all the specimens were prepared from the same material to equalize the condition of non-metallic inclusions, which could become the site of crack initiation.
| C | Si | Mn | P | S | Cu | Ni | Cr | Mo | O |
|---|---|---|---|---|---|---|---|---|---|
| 0.20 | 0.19 | 0.55 | 0.018 | 0.006 | 0.10 | 1.70 | 0.53 | 0.21 | 0.0009 |
The schematic diagram of the machine used for the RCF test is shown in Fig. 1. This test machine was used to rotate one specimen (20 mm in diameter and 36 mm in width) against two JIS-SUJ2 steel balls (31.75 mm in diameter) supported by three guide rollers, and the lubricating oil was supplied to the specimen surface through a driving roll. The test conditions are listed in Table 2. The test load was set such that the Hertzian maximum pressure reached 5.8 GPa, and the depth z0 at which the orthogonal shear stress—which was thought to affect the sub-surface initiated spalling life—had a maximum value was 0.24 mm. The L10 life, which represents a 10% probability of failure, was calculated from the Weibull plot obtained from the RCF test results of the eight specimens with various amounts of γR.
Schematic diagram of radial-type rolling contact fatigue test machine.
| Contact condition | Hertzian maximum pressure | 5.8 GPa |
|---|---|---|
| Maximum orthogonal shear stress (depth z0) | 1.4 GPa (0.24 mm) | |
| Loading speed | 285 Hz | |
| Lubricant | Mineral oil (ISO-VG100) | |
| Operating temperature | 60±5°C | |
In order to obtain an insight into the influence of the amount of γR on the microstructural alteration during rolling contact fatigue, the RCF test was interrupted at a predetermined cycle before the spalling. Full width at half maximum (hereinafter referred to as FWHM) of the martensite X-ray diffraction (XRD), residual stress, volume fraction of γR, microstructure, and Vickers hardness of the specimen were investigated. The FWHM, residual stress, and volume fraction of γR of the electropolished rolling contact surface of the specimen were measured by irradiating X-rays in the rolling direction using the XRD method. The FWHM was measured from the α(211) peak, and the residual stress was evaluated from the same α(211) peak using the sin2ψ method. The volume fraction of γR was calculated from the peak integral intensity ratio of α(200), α(211), γ(200), and γ(220). The microstructure was observed using an optical microscope after cutting and mirror polishing in a plane normal to the rolling direction and then etching with nital (95% methanol and 5% nitric acid). The hardness was measured by Vickers hardness tester with 300 gf of the test load at the cross section where the microstructure was observed.
For some specimens, the microstructure in a plane parallel to the rolling direction was observed using a scanning electron microscope (SEM). The specimens for the observation were prepared by mirror polishing them using colloidal silica dispersed suspension and then etching with nital.
Furthermore, a detailed microstructure analysis was performed using a transmission electron microscope (TEM). The thin foil used for the TEM analysis was prepared using conventional electropolishing methods after collecting the specimen using a precision cutting machine at the depth z0 from the rolling contact surface so as to obtain a plane parallel to the rolling direction. Electropolishing was performed using a twin-jet electropolishing apparatus. 10% perchloric acid–10% methanol–glacial acetic acid was used as the electrolytic solution.
As shown in Table 3, the specimens with various volume fractions of γR were prepared by altering the subzero temperature. 39% of γR (S39) was prepared without subzero treatment. 21% of γR (S21) was obtained by subzero treatment at −50°C. Other subzero temperatures were −80°C for 12% of γR (S12) and −196°C for 6% of γR (S6). The remainder of the microstructure mainly comprised tempered martensite and cementite. In addition, the Vickers hardness decreased as volume fraction of γR increased, thus suggesting that γR was softer than the tempered martensite. Incidentally, the volume fraction of γR and Vickers hardness values in Table 3 are measured at the depth z0 (0.24 mm depth from the surface).
| Specimen | Properties at z0 | |
|---|---|---|
| Volume fraction of γR | Vickers Hardness | |
| S39 | 39% | 747 HV |
| S21 | 21% | 809 HV |
| S12 | 12% | 848 HV |
| S6 | 6% | 866 HV |
The Weibull plots of the obtained RCF life of specimens are presented in Fig. 2. All the failures of the specimens were judged as sub-surface initiated spalling based on their appearance. The L10 of the specimens was 1.9 × 106 cycles for S6, 13.8 × 106 cycles for S12, 33.3 × 106 cycles for S21, and 120.0 × 106 cycles for S39, and they improved as the volume fraction of γR increased.
Sub-surface initiated spalling life.
Figure 3 shows the result of the obtained L10 as a function of the initial volume fraction of γR at the depth z0. As can be seen from this result, the specimen with a large initial volume fraction of γR had a long sub-surface initiated spalling life. As it was inferred that the difference in the initial volume fraction of γR had an influence on the sub-surface initiated spalling life, an attempt was made to elucidate the influential factors via the following investigation.
Relationship between volume fraction of γR and L10 life.
It is generally known that the FWHM obtained using XRD is related to the dislocation density and increases with an increase in the number of cycles.19) However, it has been reported that the FWHM of martensite becomes narrower during the RCF process of quenched steel because carbides are precipitated from high-carbon martensite and the high dislocation density of martensite gradually decreases.16,20,21) Based on this phenomenon, the FWHM of martensite is often used as a parameter of fatigue analysis in RCF tests.22,23,24,25,26,27,28) In this study as well, we attempted to use the FWHM of martensite as the index of the degree of fatigue for consideration of the influence of the microstructural difference on the sub-surface initiated spalling life.
As an example, the results of S39 are shown in Fig. 4. The distribution of the FWHM of martensite was measured from the area directly below the rolling contact surface to the inside. The both results are shown before the occurrence of RCF (before RCF) and the cycles at 3.7 × 106, which is near the L10 life of S6 (after RCF). Incidentally, unless otherwise specified, the subsequent microstructural analysis after the RCF was also performed using specimens after 3.7 × 106 cycles. The FWHM of the martensite became narrow near the depth z0 owing to the rolling contact, thus indicating that the martensite fatigue was progressing. The specimens comprising various amounts of γR were measured in the same manner, and the results obtained at the depth z0 are summarized in Fig. 5. Prior to the rolling contact, the volume fraction of the martensite varied depending on the subzero treatment temperature, but each specimen had an equivalent FWHM of martensite. Furthermore, subsequent to the occurrence of the RCF, the FWHM of martensite became narrower by approximately 2.5° in each specimen, thus indicating that there was no clear difference. As described above, we attempted to explain that the improvement in the sub-surface initiated spalling life due to the increase in γR was related to the degree of fatigue as evaluated based on the FWHM of martensite. However, it became clear that the difference in the life could not be explained by the difference in the FWHM.
Change in the full width at half maximum of α(211) with depth from surface of S39 owing to the RCF (N = 3.7 × 106).
Relationship between initial volume fraction of γR and change in the full width at half maximum of α(211) at z0 (N = 3.7 × 106).
It is expected that γR also undergoes some microstructural alteration owing to the RCF. However, the FWHM of γR was not evaluated because the XRD intensity of γR was weak, and its FWHM could not be accurately measured in the specimens including a small amount of γR. Even in the case of previous fatigue analyses,16,20,21,22,23,24,25,26,27,28) the main study has been focused on quenched and tempered SUJ2 in which the volume fraction of γR is usually as small as 10% or less. Therefore, there is no example of an analysis of the degree of fatigue of γR based on the FWHM.
3.4. Changes in Residual Stress due to RCFThe residual stress before and after the RCF measured for each specimen is shown in Fig. 6. Figure 6(a) shows that the same level of compressive residual stress was imparted regardless of the microstructural condition before the occurrence of RCF. Subsequent to the RCF, the compressive residual stress due to plastic deformation was imparted around a certain depth as shown in Fig. 6(b). The depth at which the compressive residual stress had a maximum value was greater than the depth z0 at which the orthogonal shear stress was maximum. This depth is known to be the depth at which the principal shear stress is maximum.27,28) As is clear from Fig. 6(b), the residual stress distributions of the specimens after the occurrence of RCF demonstrate a slight difference. However, there is no clear correlation between the residual stress distribution and sub-surface initiated spalling life. Moreover, on taking into consideration the fact that there is an error of approximately 100 MPa in the measured value of the residual stress as reported in the quenched high-strength steel,29) the aforementioned difference is insignificant.
Depth distribution of the residual stress (a) before RCF and (b) after RCF (N = 3.7 × 106).
For each specimen, the microstructure after the RCF observed using an optical microscope is shown in Fig. 7. The rolling direction is perpendicular to the micrographs. Figure 7(a) is a representative of the result of S39 and the area enclosed by a dotted circle in it indicates the dark etching area (DEA). The area is easily etched using nital and locates in the center of the depth z0 position just under the rolling contact surface, as previously reported.8) It is known that the DEA is generated by the local plastic deformation of martensite as the RCF progresses. However, Fig. 7 shows that the DEA has already appeared at 3.7 × 106 cycles in all the specimens; furthermore, the sizes of DEA observed area are nearly the same and do not exhibit clear differences. Therefore, it is suggested that there is no clear difference in the progress of fatigue in the martensitic structures among the specimens. As the L10 of S6 is 1.9 × 106 cycles, it is expected that spalling occurs before or immediately after the DEA appears. However, S39 with an L10 of 120.0 × 106 cycles does not spall until approximately 30 times the number of cycles required for the DEA to appear. Therefore, it can be concluded that there is no correlation between the appearance of the DEA and the sub-surface initiated spalling life.
Optical micrographs showing dark etching areas (DEA) under the track owing to RCF (N = 3.7 × 106) of (a) S39, (b) S21, (c) S12, and (d) S6. The area enclosed by the white dotted line in (a) indicates the DEA.
The Vickers hardness and volume fraction of γR were measured for each specimen after the RCF to clarify the changes in those values from those before the RCF. Figure 8 shows the changes in Vickers hardness and volume fraction of γR as a function of distance from the rolling contact surface to the inside of the specimens S39 and S6 as representatives. The Vickers hardness and volume fraction of γR change predominantly in the vicinity of the depth z0, although the magnitude of their variation is different depending on the initial amount of γR. This phenomenon was also confirmed in S21 and S12 although not shown. In the case of all specimens, γR decreased as the Vickers hardness increased. Figure 9 presents the relationship between the change in the volume fraction of γR (ΔγR) and Vickers hardness (ΔHV) after the occurrence of RCF at the depth z0 of each specimen. It is evident that the increase in the Vickers hardness becomes greater as the volume fraction of γR decreases significantly, which is the same as the tendency highlighted in previous results.8,15) This tendency can be understood as a result of the deformation-induced martensitic transformation of γR during rolling contact. Figure 10 shows the relationship between the Vickers hardness and the volume fraction of γR at the depth z0 before and after the occurrence of RCF for each specimen. Prior to the RCF, each specimen had a different volume fraction of γR and therefore different hardness. However, it can be observed that the hardness and volume fraction of γR of each specimen reached almost the same level (882 to 891 HV and 1 to 7%) after the occurrence of RCF. As describe above, although the hardness and volume fraction of γR are approximately at the same level at 3.7 × 106 cycles for all the specimens, the number of cycles until the subsequent sub-surface initiated spalling increased as the initial volume fraction of γR increased. This experimental result suggests that the difference in the behavior of microstructural alteration up to 3.7 × 106 cycles may affect the sub-surface initiated spalling life although the microstructural difference was not confirmed via an optical microscopic observation. Thus, for a higher initial volume fraction of γR, the increase in the hardness up to 3.7 × 106 cycles becomes greater and fresh martensite grains are simultaneously formed through the deformation-induced transformation of γR, whereby the fatigue strength is improved.
Change in the Vickers hardness of (a) S39 and (b) S6 and γR volume fraction of (c) S39 and (d) S6 owing to the RCF as a function of the depth from the surface (N = 3.7 × 106).
Relationship between change in volume fraction of γR (ΔγR) and the Vickers hardness (ΔHV) owing to the RCF at z0 (N = 3.7 × 106).
Relationship between volume fraction of γR and the Vickers hardness before and after RCF at z0 (N = 3.7 × 106).
As specified previously, the deformation-induced martensitic transformation of γR during rolling contact may improve the sub-surface initiated spalling life. In order to obtain some insight into this hypothesis, detailed microstructural observations were performed using SEM. The observations were made on S39 and S6. These microstructures at depth z0 before and after the occurrence of RCF were compared. The specimens for the cross-sectional observation were prepared by cutting them out parallel to the rolling direction. The obtained results are shown in Fig. 11. The microstructure of S6 before RCF (Fig. 11(a)) mainly comprises tempered martensite and slight globular cementite. Although the volume fraction of γR is 6% based on the XRD measurement, it is not clearly confirmed in the SEM image. However, in the microstructure of S39 before RCF (Fig. 11(b)), the flat structure was confirmed in addition to the tempered martensite and globular cementite observed in S6. As this flat structure can be confirmed throughout the microstructure in addition to the region surrounded by the orange line in Fig. 11(b), it can be inferred that it corresponds to γR based on the difference in the microstructure of S6.
SEM images showing microstructure at z0 of (a) S6 before RCF, (b) S39 before RCF, (c) S6 after RCF, and (d) S39 after RCF (N = 3.7 × 106). (Online version in color.)
By comparing the microstructures before and after RCF, the microstructural alteration can be confirmed. First, it is evident that a striped structure is formed throughout the microstructure of S6 after the RCF (Fig. 11(c)). This striped structure becomes apparent by the nital etching, and it is considered to be the elongated grain that has been reported by Šmeļova et al.30) The presence of this elongated grain structure is also confirmed in the microstructure of S39 after RCF (Fig. 11(d)), but it is only partially visible, and occurrence degree of it is different from that in S6. Furthermore, on focusing on the region wherein the elongated grain is not manifested, it is found that the main microstructure is similar to the flat structure observed before RCF. The study of the elongated grain is insufficient, but it is presumed to be formed by the shear deformation of martensite during RCF, thus resulting in its softening owing to rearrangement of dislocation and carbide precipitation.8,31,32) As shown in Fig. 8(c), the volume fraction of γR decreases to approximately 7% in S39 and a considerable part of γR is transformed into fresh martensite by deformation-induced transformation during RCF. As this fresh martensite formation significantly increases the steel hardness as shown in Fig. 8(a), it is suggested that the resistance to shear deformation of fresh martensite is relatively high, and shear deformation thus occurs preferentially in the region of tempered martensite. Based on the above discussion, the flat structure existing as γR before the RCF is presumably transformed into martensite with a high resistance to shear deformation under rolling contact. The plastic deformation of the entire microstructure is suppressed by the dispersion of fresh martensite among the tempered martensite, and the specimen with a large amount of γR has a long sub-surface initiated spalling life. Furthermore, it is speculated that the mechanism by which excellent fatigue properties of dual phase (DP) steel consisting of ferrite and martensite may arise in steel containing a large amount of γR, which transforms into hard fresh martensite during rolling contact. Thus, it is expected that crack propagation will experience a stagnation/detour effect owing to the hard microstructures. The details of the mechanism of fatigue improvement due to γR are to be considered in future study. Here, it is predicted that the flat structure existing before RCF reaches one of the following states after RCF: ① it remains as γR, ② it changes to fresh martensite owing to deformation-induced transformation, or ③ it forms mixture of these. However, as these cannot be identified using SEM, the microstructure analysis using TEM was performed as presented in the next section. The results of the microstructure analysis obtained using SEM electron backscatter diffraction are described in the next report along with the mechanism for improving the fatigue characteristics of γR.
4.2. TEM Analysis of Flat Structure in DEA Originating Owing to γRIn order to clarify the characteristics of the flat structure that exists after RCF and originates from γR, the microstructure observed in S39 after RCF was analyzed using TEM. The thin foil used for the TEM observations was fabricated using electropolishing on the plane parallel to the rolling direction at the depth z0. The observation results are shown in Fig. 12. Similar to the microstructure obtained via SEM observation, both the striped structure, which seems to correspond to an elongated grain, and the flat structure of approximately several micrometer are observed. First, in order to analyze the striped structure, it was observed under high magnification as shown in Fig. 13. The selected area electron diffraction (SAED) was performed at two aperture sizes, 150 nm diameter (X) and 500 nm diameter (Y), as shown in Fig. 13(a). As a result, we found that the selected area of (X) exhibits the bcc structure shown in Figs. 13(b) and 13(c). Furthermore, we analyzed the selected area of (Y). The electron diffraction spots shown in Fig. 13(d) were spread concentrically, and it was found that the selected area consisted of an extremely fine polycrystalline bcc structure. Here, as observed using SAED in the area of (X), the striped structure presumably comprises tempered martensite as it has the bcc structure. The microstructure of the tempered martensite after the RCF has already been observed in detail using TEM.8,31) Shiko et al.8) revealed that phenomena similar to the tempering of martensite and precipitation of fine carbide occur during rolling contact. However, the white acicular structure confirmed at the same region has not been identified because of the limitation in the resolution of the TEM at that time. Sugino et al.31) revealed that martensite changed into rod-like or polygon-like fine grains with the precipitation of ε carbide during rolling contact, and they speculated that this grain was ferrite as its hardness was lower than that of the initial tempered martensite. Recently, Fu et al.32) reported the results of a microstructural analysis conducted using atom probe tomography (APT). According to the report, it was revealed that the microstructure after RCF has a lower carbon content than in the tempered martensite before RCF due to the decomposition and growth of carbides. Even in this study, there is a possibility that the same phenomenon has occurred because the precipitation of fine carbides has been confirmed by the dark field image of the spot and indicated by the arrow in the diffraction pattern of (Y) as shown in Figs. 13(d), 13(e). However, in order to clarify the behavior of carbon during RCF, it is necessary to separately conduct a detailed examination using TEM or APT.
TEM image showing microstructure in DEA of S39 after RCF (N = 3.7 × 106). (Online version in color.)
(a) TEM bright field image of elongated grains (high magnification of Fig. 12) in DEA of S39 after RCF (N = 3.7 × 106), (b) selected area diffraction pattern (SADP) of region (X), (c) indexed diffraction patterns of (b), (d) SADP of region (Y) and (e) dark field image produced from diffraction spot indicated by the arrow in (d). (Online version in color.)
Figure 14 shows the observation results of the flat structure after RCF at a high magnification of TEM. On performing SAED of 500 nm diameter as shown in Fig. 14(a), it was found that after RCF, the flat structure comprises a mixed crystallographic microstructures of fcc (corresponding to γR) and bcc structures (Figs. 14(b), 14(c) and 14(d)). Furthermore, it was confirmed that polycrystallization occurred due to severe plastic deformation in the striped structure as shown in Fig. 13, but in the flat structure, the bcc structure is not polycrystalline (although double diffraction is observed). It is suggested that the microstructure does not undergo a large plastic deformation in contrast to the striped structure.
(a) TEM bright field image of flat structure (high magnification of Fig. 12) in DEA of S39 after RCF (N = 3.7 × 106), (b) SADP, (c), (d) indexed diffraction patterns of (b) indicating α phase and γ phase, respectively. (Online version in color.)
From the above results, it can be inferred that the direct microstructure observation revealed the occurrence of the deformation-induced transformation of γR during rolling contact, which was suggested in the previous study.8,13,14,15) In the previous study, the deformation-induced transformation during rolling contact has been predicted mainly based on the results of the decrease in the volume fraction of γR and the increase in the hardness of the steel, as shown in Fig. 9. However, in the present study, the existence of finely dispersed martensite formed by deformation-induced transformation was clarified via a detailed analysis performed using SEM and TEM. It is assumed that the fresh martensite formed during rolling contact disperses in the fatigued area of the matrix, thereby suppressing the plastic deformation of the entire microstructure and improving the sub-surface initiated spalling life. The roles of the deformation-induced transformation of γR should be discussed in a separate paper after further analyzing the microstructural alteration during rolling contact.
Specimens were fabricated using SAE4320 containing various γR contents from 39% to 6% via a controlled heat treatment and were subjected to an RCF test. The influence of γR on the sub-surface initiated spalling life was investigated, and the following conclusions were drawn.
(1) The higher the initial volume fraction of γR, the greater the improvement in the sub-surface initiated spalling life.
(2) γR decreases and the hardness increases with a cycle of RCF at the depth z0, where the orthogonal shear stress becomes the maximum owing to rolling contact. As the volume fraction of γR and Vickers hardness after the same number of cycles (3.7 × 106 cycles) are approximately of the same degree irrespective of the sub-surface initiated spalling life, it is postulated that the difference of these alteration behaviors until 3.7 × 106 cycles affect the future life characteristics.
(3) The tempered martensitic structure becomes a striped very fine polycrystalline structure owing to the severe plastic deformation that occurs during RCF. However, γR maintains a flat structure even after the deformation-induced transformation into martensite, and no fine polycrystallization is confirmed. It is inferred that plastic deformation occurs preferentially in the tempered martensite.
(4) It is presumed that the fresh hard martensite produced by the deformation-induced martensitic transformation improves the sub-surface initiated spalling life by being finely dispersed in the microstructure.
In this paper, we summarize the results of a study that was focused on the microstructural alteration in steel during RCF. We have not conducted detailed experiments or presented discussions on the relation between the microstructural alteration and crack nucleation/propagation behaviors, and we intend to continue the research on these issues in a future work.
