The Evolution of White Etching Cracks (WECs) in Rolling Contact Fatigue-Tested 100Cr6 Steel

Serial sectioning analysis of 18-h rollers revealed a large number of small near-surface (< 25 µm zone) WEC/WEAs (see Fig. 7). These were rarely seen with connection to the contact surface, limited interaction with inclusions, and it was also found in cases that there was no apparent presence of a crack; thus, evidence shows that a crack is unnecessary for WEAs to form in this case. These near-surface features were predominantly found in the outer half of the rollers with a decrease in their respective numbers being found across the inner half. The occurrence of these near-surface features has been documented previously by the authors in FAG-FE8 tests [24], hydrogen charged TE-74 two-roller specimens during RCF testing [6] and actual WTGBs from service [1], where rare connection to the contact surface has been found with limited cases of inclusion interaction [6].

It is hypothesised that the occurrence of these features is a result of either, or a combination of; (1) an increase in traction between contacts at later stages of RCF operation, resulting in a rise of the maximum subsurface shear stress zone towards the surface. This would be significantly increased in the case of insufficient lubrication and higher surface roughness; (2) microindentations at the contact surface causing local regions of mixed/boundary lubrication regimes increasing the traction coefficient; (3) areas of localised increase in contact pressure, and (4) increase in the concentration of mobile diffusible hydrogen over longer RCF operation [44], where locally at the near-surface, higher localised penetration and concentrations may exist and aid in the acceleration of these features. Figure 15 illustrates the hypothesised mechanisms for the formation of these features.

Analysis has shown that the propensity and average size (see Fig. 13) of WECs recorded in the rollers increase with RCF test duration (see Table 3, and Fig. 12, 13) no WECs being found in the raceway. Two factors to explain this could be: (1) steel cleanliness has shown that the raceway is ‘cleaner’ than the roller (see Sect. 2.3.1); (2) a lack of hydrogen being available to accelerate WEC formations [44]; and (3) the raceway is ~ 23% softer (590 HV) than the rollers (765 HV); therefore, the raceway is less prone to cracking due to an increased toughness. The importance of high toughness steel has been highlighted in the reduction in intergranular subsurface cracking and the subsequent movement of crack faces in generating WEA [41]. It is well recognised that hydrogen affects high strength steels, where hydrogen in its atomic mobile form is able to retain its mobility [52]. Hydrogen has, however, been shown to have little influence on toughness and no effect on the hardness of 100Cr6 bearings steel [53]. To confirm the non-existence of WECs in the raceway washer, at a later date 9 randomly selected individual sliver sections of raceway were mounted such that the sections were examined in the opposite axial direction (side on instead of top down contact surface direction). Two sections at 100-um intervals were taken and examined through optical microscopy. No evidence of WECs was found.

No WECs were recorded at 2 h; this could be thought logical due to the short RCF test time. Results show that the number, size and severity of WECs do not increase linearly (see Fig. 12, 13), a ramped increase seen at the later stages of RCF operation (12–18 h). This could be due to WECs coalescing to form larger crack networks resulting in a ‘weakening’ of the surrounding steel accelerating WEC growth, this being heightened in the event of a sufficient threshold concentration of diffusible hydrogen being reached [44]. Variance in severity is also observed between rollers. This highlights the importance of analysing a representative volume of steel. Differences in the severity between outer and inner roller halves at 18 h are also shown, the outer half having a greater severity than the inner.

Condition monitoring on FAG-FE8 tests [29] has suggested a rapid release of subsurface WECs occurring at ~ 20% outstanding RCF test time before WSF failure. This is proposed to be due to the steel experiencing a failure-free period (no WECs detected), in which energy is absorbed (explained by Barkhausen noise (BN) measurements [54]), a limit being reached with a sudden release of WECs. This failure-free period could be local subsurface transformations that have been observed as ‘crack-free’ dark etching regions suggested to lead to the formation of WEA and subsequently WEC [39]. This investigation shows that WECs do exist during this period before a sudden rupture occurs, ~ 20% outstanding RCF time corresponding to 14.4 h, WECs being recorded between 4 and 12 h.

When comparing the inclusion–WEC interactions recorded at 4–18 h, the inclusions are: (1) consistent in type and size, typically small/short ~ 2–15 μm D _Dup or D-type inclusions, (2) the inclusion–WEC interaction depths are within/close to the zones of maximum subsurface shear stress (τ _{0, max} = 92 μm, τ _{uni, max} = 145 μm), specifically at the early stages of RCF (4 and 6 h) where initiation is suggested to occur (see Fig. 8), and (3) when visually comparing the inclusion–WEC interactions recorded across 4–18 h, a number of similarities in regard to crack shape/angle and continuity, inclusion type and size, direction of crack propagation and location of inclusion within the WEC network can be observed (see Figs. 9, 10, 16). It is therefore proposed that the WECs recorded in this study were formed as a result of WECs that initiated at NMIs in the subsurface, individual WECs propagating and coalescing at later stages of RCF (12–18 h) to form larger WEC networks. It can also be said that the 19 WECs that did make a connection to the contact surface at 18 h are likely formed as a result of subsurface initiation at inclusions. A number of these surface connections/interactions had very small contact crack volumes and connection to the surface over short axial lengths (see Fig. 6a–c); it is proposed that these connections are not sufficient to drive such extensive WEC networks in the subsurface. Note that a number of these surface connections were in fact very near (< 1 μm) surface ‘interactions’ (see Fig. 6c, f), where under optical microscopy no apparent connection to the contact surface was observed. Further evidence for subsurface initiation comes from the fact that for the 9 fully mapped subsurface WECs at 4 and 6 h, each WEC had at least one or multiple inclusion–WEC interactions. Additional evidence to support subsurface initiation of WECs by NMIs is shown through the visual comparison of typical recorded WECs across the 4–18-h tests (see Fig. 16). As it can be seen the initial shape and propagation route of WECs at 4 and 6 h follow a close link to those WECs recorded at 12 and 18 h, this is in conjunction with the fact that the inclusion–WEC interactions are also closely linked by the similarities discussed above. At 4 and 6 h, WECs are found to initially resemble butterfly cracks, which propagate into ‘star-like’ cracks with forking of the butterfly crack tips. This ‘star-like’ crack shape can be seen to fit a number of the WECs found at 12 and 18 h as shown in Fig. 16. It is thus proposed that the WECs recorded at 4 and 6 h are the early initiation stages of WECs that subsequently propagate and evolve into the large WEC networks recorded at 18 h. The proposed evolutionary stages of WEC initiation are shown in Fig. 16g.

The outer and inner roller halves have shown to significantly influence the propensity and size of WEC formations, the outer half being dominant over the inner (see Table 3, Figs. 11, 12, 13 and Video 2). Slip has been shown to influence the formation of WECs in both FAG-FE8 and three ring roller micropitting rig (MPR) tests using the same ‘special’ oil known used in this study [29, 55, 56], where evidence for the influence of negative slip being more dominant in WSF over positive slip is provided [56]. More recently the influence of slip on WEC formations has also been shown in a two-disc test rig set-up, where again negative slip showed dominance in WEC production in contrast to positive [57]. This dominance has been attributed to higher material stressing, lowered fracture mechanic properties under alternating load and preferential surface crack propagation due to the traction force and surface motion vectors pointing in the same direction in negative slip as opposed to positive [57]. It is proposed that negative slip results in the compressive closure of cracks enhancing the crack rubbing mechanism for WEA formation [56]. The localisation of the WECs recorded across 4–18 h is more densely populated in the 2–3 mm (outer) and 8–9 mm (inner) zones across the roller (see Fig. 12, Video 2). These zones correspond to areas of high slip energy (PVmax, the product of contact pressure P and slip velocity V, MPa ms⁻¹, see Fig. 12), slip energy taking into account asperity contact (P_cV value which takes into account the asperity contact pressure P_c) and asperity friction accumulation e _{a,c max} which relates the regeneration time span between consecutive contact load cycles on tested WEC lives and the specific frictional energy input into the a surface during the contact load cycle [58]. Slip energy criteria have been linked to WSF, WSF occurring at areas of greatest PVmax. These areas have also been found to coincide with zones of highest concentrations of hydrogen [50, 59, 60]. Supporting evidence for the slip energy criterion has been shown on FAG-FE8 tests where WECs appeared firstly at areas of high frictional energy, this also being demonstrated in tests using angular contact ball bearings [29, 58]. A number of inclusion–WEC interactions were also recorded during fine serial sectioning corresponding to the 2–3-mm zone of high slip energy dissipation and asperity accumulation (see Fig. 10). Further progression of the slip energy criteria concept has been developed based upon information from different test rigs, using normal contact load and representing the slip energy criteria per film thickness sheared (N V/λ, N ms⁻¹) to determine a threshold for WEC formation in most roller bearing configurations [21, 61]. It is postulated that this threshold could exist due to the fact that sliding energy generates local flash temperatures influencing the tribochemical reactions taking place at nascent surfaces [62]. Limitations, however, do exist as this criterion does not take into account the lubricant formulation. Evidence for the degree of boundary lubrication (the range of λ) controlling the propensity for WEC formation is also suggested, more WECs forming for more severe boundary regimes (λ in the range of 0.06–0.7) [56]. No WECs were found in the raceway washers, with no evidence of WEC formations being observed in the zones corresponding to high slip energy dissipation or asperity friction accumulation. It is noted that the asperity friction energy accumulation is greater in the washers than the rollers, where energy dissipation is greater in the inner raceway than the outer (see Fig. 6, [58]). This is contradictory to the result seen in this study, this discrepancy not being understood.

Metallographic analysis has shown that the volume of WEA associated with cracks increases for longer RCF test operation. Through quantitative WEA analysis, the average WEA volume (μm³) and area (μm²) associated with cracks increased between 4 and 18 h, a ramped increase found between 12 and 18 h (see Fig. 14a). Analysis also reveals that for greater axial WEC lengths a significant increase in the associated WEA volume is found (Fig. 14b). It is proposed that the evidence found in this study supports the theory of crack rubbing/beating in the formation of WEAs [36]. As WECs grow and propagate during RCF operation, further crack rubbing/beating occurs at the newly formed crack faces, larger cracks having a greater amount of ‘free’ crack faces available for extended crack rubbing/beating to occur. This can also be exhibited in Fig. 16 where it can be seen that the amount of WEA associated with the cracks increases across 4–18 h, an increase being observed between 12 and 18 h. Further to this, by visually observing the mapped WECs at 4 and 6 h (see Fig. 8) a decrease in WEA volume is seen at the start and ends of the WEC, i.e. the extreme tips when visualised as a 3D network. Taking for example the 6-h WEC in Fig. 8, where it is proposed that the inclusion is the site of initiation; it can be seen that the volume of WEA is greater around the inclusion site (see Fig. 8g) than at the end ‘tips’ where branching/forking has occurred (final stages of propagation) where it is proposed that less time has been available for crack rubbing/beating. This is also exhibited in two videos through serial sectioning of an 18.5-h FAG-FE8 test previously conducted by the authors [24]. This point is strengthened through WEA analysis where it has shown that the average amount of WEA (μm²) is less at the tips (1st (start) and 5th (end) measurements) than at the centre (3rd (middle) measurement) (see Fig. 14a). As discussed, a large number of near-surface WEA/WECs were recorded at 18 h (see Table 3, Fig. 7). A number of these features were found not associated with a crack, leading to the conclusion that near-surface WEAs do not seem to require a crack to form WEA. However, note that in this study only optical microscopy has been used in the classification of WEAs associated with cracks, further analysis using SEM to confirm the non-existence of small cracks that may be present inside the WEA regions should be conducted.

Through metallographic analysis, it is indicated that the angle of crack propagation, crack width and zone of maximum subsurface shear stresses can influence the degree of WEA generated. WEA analysis has shown that the amount of WEA associated with a crack increases the nearer to 0° or perpendicular to the axis of compression a crack propagates (see Fig. 14c). A vertical crack ± 90° parallel to the axis of compression is found to have very little WEA associated (see Fig. 14c). This is also exhibited clearly in the 18-h crack shown in Fig. 17. As suggested by others [36], it is proposed that a vertically branching crack will be subjected to a much lesser amount of crack rubbing. The localisation of strain has been found to be strongly reliant on crack orientation in relation to stress [63]. High strain rate compressive tests have shown regions of WEA [64], this being in comparison with equivalent tensile tests; as a result it is proposed that crack rubbing/beating under RCF shear stresses or compressive loading results in WEA formations and thus adiabatic shearing is an unlikely cause [65]. The crack width also appears to influence WEA formations. For example, in Fig. 18 it is seen that very little or no WEAs are associated with sections (Area 1) of the WEC that have large crack width when compared to crack faces that are close together (Area 2). It is proposed that for adjacent crack faces that are further apart, less action is available for crack rubbing/beating. This is not to say, however, that these particular areas of the WEC network would have not been associated with WEAs at some point during operation. The proposed mechanism (see Fig. 18) to explain this is as follows: (1) inclusion initiation of butterfly and ‘star-like’ WECs with subsequent generation of WEAs due to crack rubbing/beating. (2) WECs propagate and WEAs continue to develop. (3) Short crack growth from inclusions/butterflies by Mode I loading [66] stops and further growth is governed by Mode II/III shear loading if the Mode II/III stress intensity factor threshold is surpassed [66], WECs may propagate and coalesce to form larger networks. Until a critical length is reached, crack growth rate may be slow, where once exceeded rapid propagation results under applied stress. (4) Due to the rapid growth of the crack and crack volume, the time and action available for WEA development is alleviated and thus a reduction or non-existent presence of WEA is seen. In the event of hydrogen diffusion, hydrogen acts to decrease the Mode I/II stress limits for crack growth and propagation [66, 67], it may be reasoned that this step increase in WEC formations is due to a threshold concentration of hydrogen being reached for a decrease in Mode II crack growth [44].

Analysis also indicates that the zone of maximum subsurface shear stresses influences WEA generation. This is most clearly exhibited in WECs recorded at 18 h. For example, WEC-2 R1 in Fig. 17 shows that a greater amount of WEA exists within and in the regions around the zone of maximum subsurface shear stresses (τ _{0, max} = 92 μm, τ _{uni, max} = 145 μm).

Eighty-two NMIs were recorded during the metallographic analysis; 62 were ranked with a high likelihood of WEC initiation (rank 1 or 2). Most inclusions were found to be small sized (~ 2–15 μm (diameter) and ~ 4–21 μm in axial length) globular duplex inclusions (globular manganese and/or calcium sulphide surrounding aluminate) (D _Dup) and globular oxide inclusions (D). At the early infant stages of WEC formation (4 and 6 h), D_Dup and D-type inclusions were found to interact with the WECs, EDX analysis of inclusions at these stages being either Al₂O₃ or MnS surrounding Al₂O_3. Inclusion types found to interact with the large WEC networks found at the later stages of RCF duration (18 h) are consistent with those found at the early infant stages. This agrees with the findings found from the 18.5-h RCF test in [24] that found 49 NMIs and 41 rank 1 or 2 NMIs in 5 fully mapped WECs, these predominantly being small sized (~ 2–15 μm) D _Dup, D _DupTi(C,N) and D-type inclusions. The oxide encapsulations are responsible for hardness discrepancy with the martensite matrix, induced tensile residual stresses due to differing coefficients of thermal expansion and weak coherence/de-bonding of the oxide and matrix [9, 68, 69]. The majority of the inclusion–WEC interactions were recorded at a depth of ~ 50–200 μm, this being consistent with the depth of high subsurface shear stresses (τ _{0, max} = 92 μm, τ_{uni, max} = 145 μm) (see Fig. 10).

Steel cleanliness analysis indicates that the raceway is ‘cleaner’ than the rollers (C _i  = 1.5 (standard) and C _i  = 344.1 (non-standard) for the rollers and (C _i  = 0.2 (standard) and C _i  = 35.6 (non-standard) for the raceway). The ‘cleaner’ raceway would therefore have fewer inclusions available to initiate cracks, which could explain why no WECs were recorded. The lower cleanliness and therefore greater density of inclusions in the rollers would also lead to an increased propensity for WECs to initiate and coalesce to form more extensive networks, this being elevated if inclusions lie in critical locations. It is important to note the significant increase in C _i when only counting inclusions recorded under the thresholds set by the ISO 4967-B standard [48] and when counting inclusions recorded outside of these domains; however, the ratio difference in cleanliness between the roller and raceway remains similar. Small/short inclusions have been found to be dominant in initiating and interacting with WECs. Thresholds set by the standard do not factor in these small/short inclusions, comparisons of the C _i highlighting the potential limitations of the standard when considering steels used in WTGBs. These limitations are currently being investigated and will be presented in a future study by the authors.