2023 年 63 巻 3 号 p. 559-568
The changes in the state of carbon in tempered martensite and retained austenite in carburized SAE4320 steel under the rolling contact fatigue (RCF) were investigated using atom probe tomography (APT). In the tempered martensite, the carbons in solid solution and in carbon cluster were readily transferred to the preexisting metastable (ε) carbide due to rolling contact, resulting in a localized change from tempered martensite to ferrite accompanied by the growth of carbides. This supports the recently proposed dislocation assisted carbon migration theory. On the other hand, retained austenite with uniformly distributed enriched solute carbon was partially transformed into the very fine deformation-induced martensite due to rolling contact. Furthermore, carbon seemed to be partitioned into retained austenite from the deformation-induced martensite during further rolling contact cycles. This is a new insight into the characteristics of deformation-induced martensite and retained austenite generated by rolling contact. The present study provides a plausible explanation to the phenomenon that the deformation-induced martensitic transformation improves the RCF life.
A rolling bearing is a mechanical part that supports rotational motion with little friction while receiving a large load. The bearing ring and rolling element support the load via point or line contacts, owing to the structure of the bearing; therefore, a high contact stress acts locally. This is repeated as the bearing rotates. A microstructural change unique to rolling contact fatigue (RCF) occurs immediately below the rolling contact surface, starting with the occurrence of a dark contrast structure (e.g., dark etching region (DER), dark etching constituent (DEC), and dark etching area (DEA)),1,2,3,4,5,6,7,8,9,10,11,12,13) followed by the formation of a white contrast banded structure (e.g., white band (WB) and white etching band (WEB)).5,6,7,8,9,10,11,12,13,14,15,16,17,18) DER is a structure that has been etched by Nital or Picral and is indicated by a dark contrast by optical microscopy (OM). It is the first microstructural change that appears in rolling contact; therefore, its morphology must be understood to elucidate the RCF phenomenon. There have been many reports on the observation of DER using OM, scanning electron microscopy (SEM), and transmission electron microscopy (TEM).1,2,3,4,5,6,7,8,9,10,11,12,13) In particular, the details of the dislocation structure and precipitation of carbides have been confirmed via TEM; and studies have shown that, in addition to the increase in dislocation density and formation of dislocation cell structure, tempered martensite changed to ferrite with locally rearranged dislocations and precipitated fine carbides.3,4,7,8) These observations indicate that carbon-migration behavior in tempered martensite during RCF is an important factor affecting the microstructural changes caused by rolling contact. Carbon migration and subsequent carbide precipitation in tempered martensite was initially believed to be caused by re-tempering during rolling contact.3,5,6) However, microstructural changes do not occur at the rolling contact surface, where the temperature rise is highest, but they occur at the depth where the orthogonal shear stress below the rolling contact surface becomes large. Therefore, microstructural changes cannot be sufficiently explained by re-tempering. In contrast, Swahn et al.7) and Voskamp9) predicted that dislocations induced by cyclic stress promote the migration of solute carbon atoms from tempered martensite. Recently, this has been theoretically explained via a dislocation-assisted tempering model19) and dislocation-assisted carbon-migration theory.20,21) These models consider carbon migration, in which carbon atoms form a Cottrell atmosphere with the dislocations introduced via rolling contact and are transferred to existing carbides via gliding dislocations for carbide growth. Fu et al.21) investigated this carbon-migration behavior experimentally by measuring hardness and analyzing carbon distribution using atom probe tomography (APT). DER exhibited a lower hardness than tempered martensite, accompanied by the precipitation of θ-Fe3C and the transition carbides such as η-Fe2C and ε-Fe2.4C. Consequently, DER changed its structure from tempered martensite to a low-carbon ferrite with high density of carbides. These results confirmed that DER consists of a re-tempered structure of tempered martensite and demonstrated the validity of the dislocation-assisted carbon-migration model.
As described previously, the microstructural changes associated with rolling contact are affected by the migration behavior of carbon dissolved in tempered martensite. However, retained austenite also exists in the microstructure of quenched and tempered steel, in addition to tempered martensite. Austenite has a higher solubility limit of carbon than martensite and can dissolve a large amount of carbon; however, there has been no research on the migration behavior of solute carbon in retained austenite. The aforementioned dislocation-assisted tempering model19) and dislocation-assisted carbon-migration theory20,21) apply to tempered martensite, and have been confirmed experimentally using quenched and tempered high-carbon–chromium bearing steel with low amounts of retained austenite. However, bearings often use steels that have been carburized or carbonitrided. Retained austenite exists in a volume fraction of approximately 10–50% in these materials; therefore, the migration behavior of carbon in retained austenite must be elucidated to understand the structural changes in RCF. An appropriate amount of retained austenite may improve the RCF life by transforming it into deformation-induced martensite during rolling contact.3,22,23,24,25,26,27,28,29) Various possibilities have been proposed regarding the role of deformation-induced martensite on RCF, such as the suppression of crack propagation via deformation-induced martensitic transformation at the stress-concentration part in the tip of the internal crack,23) delayed fatigue of tempered martensite via formation of new martensite,3,23) or introduction of compressive residual stress.24,26) However, the mechanism remains largely unclear. Clarifying the migration behavior of carbon in retained austenite at the rolling contact may elucidate the mechanism by which it enhances RCF life.
This study aims to investigate the migration behaviors of solute carbon in tempered martensite and retained austenite via rolling contact and discuss the mechanism of RCF-life improvement by retained austenite based on the difference between these carbon-migration behaviors.
We selected SAE4320 steel, which has a high Ni content (austenite stabilizing element) and from which retained austenite can be easily obtained via carburization, as a sample among the materials commonly used for bearings. Table 1 shows its chemical composition. A hot-rolled steel bar with a diameter of 26 mm manufactured using an actual facility was machined, heat-treated, and ground into a cylindrical shape with a diameter and width of 20 and 36 mm, respectively. The carburization treatment was as follows: In a carburizing atmosphere at 960°C, the carbon concentration of the surface layer was adjusted to approximately 1.1% over 26 h. The cooled sample was then heated to 820°C in the austenite region, held for 70 min, and then quenched with oil at 80°C. Finally, the sample was tempered at 180°C for 2 h. The as-prepared samples consisted of tempered martensite as the main structure, as well as retained austenite and cementite (θ-Fe3C). The retained-austenite amount (volume fraction) was 39%, as evaluated by X-ray diffraction (XRD) measurements.
| 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 |
Figure 1 schematically shows the machine used for the RCF test. The sample and two high-carbon–chromium bearing steel balls (diameter of 31.75 mm) supported by three guide rolls were brought into rolling contact under the conditions shown in Table 2. The depth of the maximum orthogonal shear stress owing to rolling contact (z0) was 0.24 mm below the rolling contact surface. As previously reported, rolling contact causes microstructural change (DER formation) around the depth z0.28) The test was terminated after 3.7 × 106, 10 × 106, and 44 × 106 cycles, and the samples were removed from the test machine and subjected to microstructural analysis.
Schematic of the radial type rolling contact fatigue (RCF) 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 | |
The following analysis was conducted to clarify the microstructural changes and carbon-migration behavior associated with rolling contact. All microstructural analyses were conducted at a depth of z0 from the rolling contact surface, where the RCF-induced microstructural changes were the most prominent.
2.3.1. Vickers Hardness and Retained-austenite AmountThe Vickers hardness was measured using a Vickers hardness tester at a test load of 2.94 N after cutting the sample in the direction parallel to the rolling direction, such that the cutting surface was just below the rolling contact surface, and then buffing the cross-section. The retained-austenite amount was measured via XRD after electropolishing the rolling contact surface of the sample to a depth of z0. A Rigaku SmartLab equipped with a Co tube was used for XRD, and the retained-austenite amount was calculated using the peak integrated intensity ratios of α(200), α(211), γ(200), and γ(220).30)
2.3.2. MicrostructureFor the microstructural analysis, the cross-section used for Vickers hardness measurement was mirror-polished using colloidal silica, etched using Nital (95% methanol and 5% nitric acid), and then subjected to field emission-SEM (FE-SEM) of JSM-7100F (JEOL Ltd.).
2.3.3. Phase DistributionThe phase distribution was analyzed using electron backscattering diffraction (EBSD) after mirror-polishing the cross-section used for the Vickers hardness measurement finished by colloidal silica. The measurement was conducted using an EBSD data collection system (OIM, TSL) attached to the FE-SEM under the conditions of an acceleration voltage of 15 kV, working distance of 15 mm, and step size of 10 nm. For fine grains smaller than the step size, the Kikuchi pattern becomes unclear and the measurement accuracy deteriorates. Therefore, measurement points with lower accuracies were excluded from the analysis targets during data processing.
2.3.4. Carbon DistributionThe carbon distribution was measured using APT. The focused ion beam (FIB) method was used to prepare the needle sample that was used for the measurement. An FIB-SEM (Helios450, FEI) was used for sample preparation. At room temperature, gallium (Ga) ions were irradiated at an acceleration voltage of 30 kV for rough processing and 5 kV for finish processing. Tempered martensite and retained austenite were identified via the same microstructural analysis as in Section 2.3.2 using FIB-SEM, and the sampling positions of the needle samples of each phase were determined. We used a local electrode atom probe (LEAP4000X Si, AMETEK) for the APT measurements. The laser-pulse mode was used, and the laser wavelength, laser pulse power, pulse frequency, evaporation rate, and sample temperature range were 355 nm, 20 pJ, 500 kHz, 0.002 counts/pulse, and −253 to −243°C, respectively. The acquired data were three-dimensionally constructed using the AMETEK’s IVAS (ver.3.8.4) software. Depending on the material, the crystal phase of the needle sample can be determined by transmission-EBSD measurements or TEM observations. However, in this study, the transformation of retained austenite into martensite during needle preparation had been confirmed via FIB; therefore, these analyses were not conducted. The APT measurement was conducted assuming that the phase transformation during the preparation of the needle sample did not affect the state of carbon.
Additionally, for some samples, the solute carbon content (Cγ, mass%) of the retained austenite was measured via XRD using a Rigaku SmartLab equipped with a Co tube. The average lattice constant (aγ, pm) of the austenite was determined, and the solute carbon content in the retained austenite was calculated using the Roberts Eq. (1).31)
| (1) |
A large amount of retained austenite exists in the surface layer of carburized alloy steel; therefore, cyclic stress causes deformation-induced martensitic transformation of the retained austenite, resulting in a decrease in the retained-austenite amount and an increase in hardness.28) Figure 2 shows the changes in the retained-austenite content and Vickers hardness in this study with an increasing number of stress cycles. The measurement position is that of depth z0 from the rolling contact surface. The result with 0 cycles corresponds to the measurement result before RCF, with a retained-austenite amount of 39% and Vickers hardness of 747 HV. Most of the retained austenite underwent deformation-induced transformation to martensite in the samples up to 3.7 × 106 cycles, showing a large increase in hardness. After subsequent stress cycles, the decrease in the amount of retained austenite and the corresponding increase in hardness were small.
Changes in the volume fraction of retained austenite (γR) and Vickers hardness due to rolling contact at z0 depth.
Figure 3 shows the microstructure determined via FE-SEM before RCF, after 3.7 × 106 and 44 × 106 cycles. The microstructure at 10 × 106 cycles, which did not differ significantly from the results after 3.7 × 106 and 44 × 106 cycles in Fig. 2, has been omitted. Before RCF (Fig. 3(a)), there were a representative globular cementite (indicated by the arrow), a rough structure with large unevenness that was formed by Nital etching (indicated by the white dashed line), and a flat structure with small unevenness (indicated by the solid orange line). The rough structures with large unevenness corresponded to tempered martensite, and the flat structures corresponded to retained austenite.28) As the number of stress cycles increased, the tempered martensite was deformed significantly, and striped elongated gains were formed, as indicated by the white solid lines in Figs. 3(b) and 3(c). Meanwhile, most of the retained austenite was presumed to undergo a deformation-induced transformation to martensite; however, as indicated by the solid orange lines in Figs. 3(b) and 3(c), the rather flat internal structure was maintained, suggesting that the structure was less deformable than tempered martensite.
SEM images showing microstructure at z0 depth. The number of cycles were (a) 0 (before RCF), (b) 3.7 × 106 and (c) 44 × 106. (Online version in color.)
Next, Fig. 4 shows the changes in the phase distribution observed via SEM-EBSD. Figures 4(a)–4(c) show α-phases corresponding to representative tempered martensite or elongated grains, and Figs. 4(d)–4(f) show those of representative retained-austenite grains. As in a previous report,32) tempered martensite was refined over 3.7 × 106 cycles, and upon increasing the stress cycles to 44 × 106, some of the fine grains exhibited growth, resulting in a unidirectionally elongated WB nucleus. Meanwhile, as representatively shown by arrows, the retained austenite resulted in the generation of fine deformation-induced martensite over 3.7 × 106 cycles. Analysis using automated crystal orientation mapping by TEM (ACOM-TEM), which has a higher spatial resolution than SEM-EBSD, confirmed that the smallest deformation-induced martensite was a few nanometers in size.29) From 3.7 × 106 cycles onwards, at least up to 44 × 106 cycles, there was no significant progression of deformation-induced martensitic transformation, as can be also inferred from the results in Fig. 2. This was presumably owing to an increase in the mechanical stability of the retained austenite, which was attributable to the work hardening of the retained austenite as a result of rolling contact stress and refinement of the retained austenite as a result of deformation-induced martensitic transformation.33)
Phase maps overlaid on IQ maps obtained using EBSD focused on (a)–(c) α phase and (d)–(f) γ phase. The number of cycles were (a)(d) 0 (before RCF), (b)(e) 3.7 × 106 and (c)(f) 44 × 106. Arrows in (e) and (f) show representative deformation-induced martensites. (Online version in color.)
We conducted APT measurements to elucidate changes in the carbon distribution for tempered martensite and retained austenite, which exhibit the aforementioned microstructural changes owing to rolling contact.
3.3.1. Tempered MartensiteFigure 5 shows the results for tempered martensite before RCF. Tempered martensite exhibited a characteristic microstructure with concentration fluctuations. Concentration fluctuations of carbon atoms in tempered martensite are caused by segregation of carbon to dislocations, lath interfaces, and the formation of carbon clusters and carbides;34,35) and similar results were obtained in the present study. The high carbon concentration region, defined as the region where the carbon concentration is 15 at% or more, existed locally shown in Fig. 5(b) and was plate-like in three dimensions. Figures 5(c) and 5(d) show the concentration profiles measured in the plate-thickness direction for the representative high carbon concentration regions (X1, X2). The thickness of the high carbon concentration region was 2–3 nm, and the maximum carbon concentration was approximately 20 at%. Additionally, no enrichment of other alloying elements (Mn, Ni, Cr, and Mo) was observed. The total length of each high carbon concentration region is unknown because the needle sample was prepared by cutting the plate surface. However, a length of at least 40 nm has been confirmed, which was the same for Y1, Y2, Z1, and Z2 in Figs. 6 and 7. Cementite (θ-Fe3C) and metastable carbides such as ε carbides are typically precipitated during low-temperature tempering.36) The carbon concentration of cementite, a stable carbide, has been reported to be approximately 25 at% via APT measurements,37) which is close to the concentration predicted in the equilibrium state. Meanwhile, the carbon concentration of ε carbide, a metastable carbide, differs with the measurement method but has been reported to be 18.9–22.6 at% via APT measurements.38) Therefore, although the high carbon concentration region confirmed in this study cannot be identified because electron diffraction was not conducted, the region certainly has a lesser carbon content than cementite. Therefore, the high carbon concentration region is speculated to be ε carbides which precipitated via self-tempering or low-temperature tempering after quenching. Furthermore, the fluctuated regions in the carbon concentration in the matrix of tempered martensite was confirmed in addition to ε carbide. These did not show a clear plate shape, the interface was diffuse, and the carbon concentration was low; therefore, they were inferred to be carbon clusters with the same crystal structure as martensite, which was the parent phase.
(a) Three dimensional (3D) C atom map and (b) iso-concentration surfaces with 15 at%C of (a) in tempered martensite before RCF. (c) and (d) were C, Mn, Ni, Cr, Mo concentration profiles in the thickness direction of X1 and X2, respectively. (Online version in color.)
(a) 3D C atom map and (b) iso-concentration surfaces with 15 at%C of (a) in elongated grain of tempered martensite at 3.7 × 106 cycles. (c) and (d) were C, Mn, Ni, Cr, Mo concentration profiles in the thickness direction of Y1 and Y2, respectively. (Online version in color.)
(a) 3D C atom map and (b) iso-concentration surfaces with 15 at%C of (a) in elongated grain of tempered martensite at 44 × 106 cycles. (c) and (d) were C, Mn, Ni, Cr, Mo concentration profiles in the thickness direction of Z1 and Z2, respectively. (Online version in color.)
Figures 6 and 7 show the results of the tempered martensite after RCF at 3.7 × 106 and 44 × 106 cycles. These results indicate that the carbon distribution of the tempered martensite was changed via rolling contact. First, after 3.7 × 106 cycles, it was inferred from the 15 at% C iso-concentration surface shown in Fig. 6(b) that the plate-like high carbon concentration region was enlarged. Figures 6(c) and 6(d) are concentration profiles measured in the plate-thickness direction for Y1 and Y2 in Fig. 6(b). Compared with the pre-RCF results (Figs. 5(c) and 5(d)), the maximum carbon concentration did not change significantly, but the thickness of the high carbon concentration region expanded to 5–7 nm, which was attributed to the growth of the ε carbides that existed before RCF. After 44 × 106 cycles, as shown in Figs. 7(a) and 7(b), relatively large plate-like high carbon concentration regions (e.g., Z1, Z2) also existed, and it can be seen from the concentration profiles in the plate-thickness direction shown in Figs. 7(c) and 7(d) that the thickness had further expanded to 11–15 nm. Additionally, as indicated by ROI3, there was a region where the carbon concentration locally decreased, which was not observed before RCF (Fig. 5) or after 3.7 × 106 cycles (Fig. 6).
3.3.2. Retained AusteniteFigure 8(a) shows the carbon atom map of the retained austenite before RCF. Clearly different from the tempered martensite, the retained austenite exhibited no concentration fluctuations, indicating homogenously dissolved carbon atoms. Figures 8(b) and 8(c) show the post-RCF carbon atom maps corresponding to 3.7 × 106 and 44 × 106 cycles, respectively. The concentration fluctuated slightly with increasing number of stress repetitions; however, the growth of ε carbides and significant decreases in carbon concentration, as observed in the tempered martensite, were not observed.
3D C atom maps of flat structure. The number of cycles were (a) 0 (before RCF), (b) 3.7 × 106 and (c) 44 × 106. (Online version in color.)
The change in carbon distribution differed significantly between the tempered martensite and retained austenite, owing to rolling contact. The changes were prominent in the former and slightly occurred in the latter. These changes, and their differences, are discussed in the following subsections.
4.1.1. Tempered MartensiteTo reveal the changes in the carbon distribution in the tempered martensite, we selected cylindrical regions (ROI1, ROI2, and ROI3 in Figs. 5(b), 6(b), and 7(b)) with diameters and lengths of 10 and 50 nm, respectively, for analysis from the regions that excluded the ε carbide which had already revealed a state change. Figure 9 shows the carbon concentration profiles obtained in the direction of the arrows in the cylindrical regions. Before RCF, carbon clusters with a carbon concentration of approximately 7 at% existed at intervals of 30–40 nm (Fig. 9(a)). The carbon concentration of the matrix, excluding these carbon clusters, was presumed to be 1–2 at% on an average. Such tempered martensite showed no significant changes up to 3.7 × 106 cycles (Fig. 9(b)), but the carbon clusters locally disappeared and the carbon concentration in the matrix decreased after 44 × 106 stress cycles (Fig. 9(c)). Meanwhile, the growth of the ε carbide that existed before RCF was also confirmed (Fig. 7).
C concentration profiles of (a) ROI1 in Fig. 5, (b) ROI2 in Fig. 6 and (c) ROI3 in Fig. 7.
Thus, the carbon clusters and the solute carbon in the matrix were inferred to migrate, owing to cyclic stress caused by rolling contact, causing the growth of the ε carbide. Kang et al.19) and Fu et al.20,21) suggested that carbon in tempered martensite interacts with dislocations to form a Cottrell atmosphere and is transferred to existing carbides via gliding dislocations associated with rolling contact, resulting in the growth of carbides. The results of the present study also support this mechanism. Dislocations that migrate within α-Fe shear carbon clusters via a cutting mechanism have been experimentally confirmed39) and predicted via molecular dynamics simulations.40) Carbon clusters are particles that interact weakly with dislocations, so the repeated passage of dislocations causes the clusters to decompose owing to shear, resulting in the re-dissolution of the carbon in the matrix. Not only the redissolved carbon but also carbon that was initially dissolved in the matrix are believed to migrate easily via dislocations that form the Cottrell atmosphere; therefore, as shown in Fig. 9(c), the carbon concentration approached 0 and the tempered martensite was presumed to decay to ferrite. Meanwhile, carbides are particles that strongly interact with dislocations; therefore, they are less likely to be sheared by dislocations. Furthermore, when a dislocation passes through a carbide, it leaves an Orowan loop around it.41) Therefore, the carbon transferred via dislocation migration is believed to stagnate upon reaching the existing ε carbides, causing carbide growth.
Fu et al.21) used the index DER%, which corresponds to the ferrite area ratio in DER, to predict the decaying behavior of tempered martensite under cyclic stress. Despite varying with conditions such as temperature, contact stress, and stress cycles, DER% (i.e., ferrite area ratio) increases rapidly after approximately 1 × 107 cycles. Figure 9 also suggests a significant formation of ferrite in the present study, between 3.7 × 106 and 44 × 106 cycles, which is consistent with the prediction by Fu et al.
4.1.2. Retained AusteniteFigure 10 shows the carbon concentration profile measured in the axial direction in the cylindrical region with a diameter of 10 nm from the needle tip to the needle base in Figs. 8(a)–8(c). The carbon concentration before RCF was approximately 4 at%, and the distribution of carbon was relatively uniform. However, the fluctuations in the carbon concentration in retained austenite gradually increased with the number of stress cycles. The concentration of carbon after 44 × 106 cycles existed at intervals of approximately 20–40 nm, with a maximum and minimum concentration of approximately 6 and 1 at%, respectively.
Phase-distribution analysis via SEM-EBSD (Fig. 4) showed that deformation-induced martensite was formed in the retained-austenite grains via rolling contact. In our previous paper using ACOM-TEM,29) we investigated the morphology of the deformation-induced martensite, which had not been previously elucidated, and confirmed that fine martensite, with a minimum size of several nanometers, was dispersed in the retained austenite. Therefore, the state of the carbon in the deformation-induced martensite was investigated using APT together with the results of ACOM-TEM in the previous study.29) First, the average interparticle spacing of the deformation-induced martensite in the retained-austenite grains was evaluated for the specimen after 3.7 × 106 cycles based on the results of ACOM-TEM.29) As schematically shown in Fig. 11, circular deformation-induced martensite grains of equal sizes were assumed to be aligned at regular intervals within an initially (pre-RCF) square retained-austenite grain. Assuming that the initial retained-austenite grain area was S nm2, the number of deformation-induced martensite grains in a retained-austenite grain after RCF was N, and the interparticle spacing of the deformation-induced martensite was L nm, Eq. (2) was obtained.
| (2) |
Transforming Eq. (2) yields Eq. (3).
| (3) |
Schematic diagram to calculate the average interparticle spacing L in the deformation-induced martensite grains. S: Initial retained austenite grain area, L: Average interparticle spacing in deformation-induced martensite grains, N: Number of deformation-induced martensite grains.
Table 3 shows the initial retained-austenite grain area S, number of deformation-induced martensite grains N and their average grain area S’’, and the average interparticle spacing L of the deformation-induced martensite grains calculated from Eq. (3), which were obtained from the ACOM-TEM results of the specimen after 3.7 × 106 cycles. Here, ACOM-TEM was measured using rectangular pixels with a step size of 2 nm, and deformation-induced martensite grains with 5 or more pixels (area of 20 nm2 or more) were considered highly reliable data for calculation. Deformation-induced martensite grains were inferred to have sizes of approximately 76 nm2 and to be distributed at intervals of 22 nm, on an average, in the retained-austenite grains. This average interparticle spacing was small enough compared to the needle sample size (height: 140 nm; base diameter: 80 nm) measured via APT in Fig. 8(b). Therefore, it is implied that multiple grains of deformation-induced martensite with sizes of 20 nm2 or more may have existed in the region measured in Fig. 8(b).
| Initial retained austenite grain area, S | 176072 nm2 |
| Number of α’ grains over 20 nm2, N | 339 |
| Average area of α’ grains over 20 nm2, S’ | 76 nm2 |
| Average interparticle spacing of α’ grains over 20 nm2, L | 22 nm |
The solute carbon content of such deformation-induced martensite is of interest. Retained austenite has a high carbon-solubility limit; therefore, a large amount of carbon may continue to dissolve in the RCF. Meanwhile, deformation-induced martensite is formed by inheriting solute carbon from retained austenite; therefore, it is believed that martensite with a large amount of supersaturated solute carbon was formed at the initial stage of transformation (initial stage of RCF). However, carbon is gradually removed from deformation-induced martensite with increasing number of stress cycles. The removed carbon may behave in two ways: precipitate as carbides or partition into the surrounding retained austenite. ACOM-TEM29) indicated no precipitation of carbides in the retained austenite or deformation-induced martensite, and Figs. 10(b) and 10(c) by APT indicate no high carbon concentration region exceeding 6 at%. Therefore, it is possible that the carbides were not precipitated and that the carbon was partitioned from the deformation-induced martensite to retained austenite during RCF. XRD confirmed the change in the average solute carbon content in the retained austenite, with values of 4.6 at% (1.0 mass%) before RCF and 5.8 at% (1.3 mass%) after 44 × 106 cycles of rolling contact. Therefore, it was indicated that the solute carbon content in the retained austenite increased after rolling contact. Here, the solute carbon content in the martensite was not measured, because tempered martensite and deformation-induced martensite cannot be distinguished in the XRD diffraction peak, and tempered martensite exhibits a pronounced crystal orientation owing to rolling contact.32) There have been no studies on the partitioning of carbon from the α phase to the γ phase during fatigue, and no conclusive statements can be made in the present study as well. However, the phenomenon similar to the quenching and partitioning (Q&P) process, which causes the partitioning of carbon from martensite to austenite via soaking at an appropriate temperature above the Ms point42) may have occurred during RCF. Figure 10(c) clearly shows fluctuations of 1–6% in the carbon concentration at intervals of 20–40 nm after 44 × 106 cycles, which is close to the average calculated interparticle spacing of the deformation-induced martensite grains (22 nm) shown in Table 3. Here, it is assumed from Figs. 2 and 4 that the formation of new deformation-induced martensite between 3.7 × 106 and 44 × 106 cycles is negligible. Furthermore, the carbon concentration in the retained austenite after 44 × 106 cycles determined via XRD (5.8 at%) and the maximum carbon concentration measured via APT (approximately 6 at%) were consistent. Thus, deformation-induced martensite, which existed at intervals of approximately 20–40 nm, removed supersaturated carbon with increasing number of stress cycles and partitioned it to the surrounding retained austenite, resulting in fluctuations of carbon content.
We have previously reported28) RCF life improvement via deformation-induced martensitic transformation of retained austenite during rolling contact. Here, we comprehensively discuss the mechanism of RCF life improvement via deformation-induced martensite, adding new findings on carbon-migration behavior in tempered martensite and retained austenite during rolling contact.
Tempered martensite shows a large change in carbon distribution due to rolling contact. Carbon clusters and solute carbon in the matrix migrate due to cyclic stress-induced dislocations, causing growth of existing ε carbides. Alongside this is the local decrease in the concentration of the carbon in the matrix, which is believed to decay tempered martensite to ferrite. Meanwhile, in retained austenite, in which a large amount of carbon is dissolved, fine deformation-induced martensite with a minimum size of several nanometers is generated via rolling contact. Newly transformed fine deformation-induced martensite contains a large amount of carbon as a solid solution; therefore, unlike tempered martensite, which is obtained via heat treatment, its microstructure is strengthened via grain refining, carbon-based-solid-solution strengthening, and local compressive stress due to volume expansion that accompanies the transformation. Retained austenite presumably improves RCF life by strengthening the microstructure caused by such deformation-induced martensitic transformation. Mechanisms for life improvement include the crack detour effect due to the formation of hard phases and the crack stagnation effect due to compressive stress. However, in the present study, we did not directly determine the initiation and propagation behavior of internal cracks during rolling contact. The effect of deformation-induced martensite on crack initiation and propagation may be studied in future. After transformation, deformation-induced martensite may have a gradually decreasing solute carbon content and partitioning of the carbon to the retained austenite with increasing number of stress cycles. Retained austenite, which did not undergo deformation-induced martensitic transformation, became mechanically stable, owing to an increase in the amount of solute carbon content due to carbon partitioning and refining and work hardening; the deformation-induced martensitic transformation may eventually cease even if the number of stress cycles was increased further, as shown in Fig. 2.
The microstructural changes and the changes in the state of carbon in tempered martensite and retained austenite due to rolling contact were investigated by ATP using carburized SAE4320 steel containing retained austenite. The following conclusions were obtained:
(1) Rolling contact generated fine deformation-induced martensite in retained-austenite grains. Additionally, the retained-austenite amount decreased, and the Vickers hardness increased.
(2) Carbon clusters and plate-like ε carbides existed in tempered martensite before RCF. Meanwhile, in retained austenite, carbon was homogenously dissolved.
(3) Tempered martensite decayed to ferrite due to a local decrease in the solute carbon in matrix together with the decomposition of carbon-clusters, and ε carbides simultaneously grew. It is presumed that the carbon clusters are sheared by dislocations, and carbon atoms are easily transferred to ε carbides during rolling contact, which cause the growth of pre-existing carbides.
(4) Fine deformation-induced martensite was dispersed in retained austenite via rolling contact, and the carbon concentration fluctuated with stress cycles. Deformation-induced martensite inherited the supersaturated carbon from the retained austenite; therefore, it is presumed that, although deformation-induced martensite has a large solute carbon at the initial stage of rolling contact, the carbon is partitioned to untransformed retained austenite with stress cycles, causing concentration fluctuations.
(5) The deformation-induced martensite was finer and had a higher solute carbon content than tempered martensite; hence, it is believed to have contributed to improved RCF life by strengthening the microstructure via grain-refining strengthening, solid-solution strengthening, and local internal compressive-stress introduced.
