The effect of over-based calcium sulfonate detergent additives on white etching crack (WEC) formation in rolling contact fatigue tested 100Cr6 steel

Over-based calcium sulfonate (OBCaSul) additives in oils are considered potentially critical in driving WEC formations, this ‘driving’ effect being unclear. Rolling contact fatigue testing of 100Cr6 steel using FE8 and PCS Micro-Pitting-Rig lubricated with oils containing varying OBCaSul concentrations have been conducted, tested samples being analysed using serial sectioning, Thermal desorption analysis (TDA) to measure hydrogen diffusion, and SEM/EDX of the tribofilms formed. Results show that OBCaSul concentration appears to affect WEC formation propensity, a reversal relationship of formations being shown between test rigs. Evidence shows oils containing OBCaSul form thick Ca dominated tribofilms, potentially promoting hydrogen diffusion and WEC formations in the FE8 rollers, where thinner Zn-S tribofilms on the FE8 raceways may demote hydrogen diffusion and WEC formations.


Introduction
Premature failures of wind turbine gearbox bearings (WTGBs) cause significant downtime and high repair costs causing substantial impact to the wind energy industry. One of the most noted failure modes of WTGBs is that of White Structure Flaking (WSF) due to the formation of White Etching Cracks (WECs) typically ∼ 1 mm below the contact surface, WECs being associated with a microstructural alteration known as White Etching Area (WEA).
Driving mechanisms for WSF and WECs have been debated for a number of years and a consensus into what the major influencing factor is has yet to be decided. One specific factor is the effect of additive chemistry and in addition, the effect additive chemistry has on the generation and diffusion of hydrogen, as a prominent driver for WSF and WECs. Additives found in lubricants that have been shown to reduce RCF life and promote WSF include; greases and lubricants containing sulphur-phosphorous extreme-pressure/anti-wear (EP/AW) additives, such as zinc dithiophosphates (ZDDP/ZnDTP/ZnDDP) and detergent/rust preventatives such as calcium sulfonates [1][2][3][4][5][6][7][8][9][10]. Hydrogen poisoning sulphur can aid in hydrogen diffusion by inhibiting molecular hydrogen recombination [11], where decomposition of EP/ AW additives such as ZDDPs at nascent catalytic surfaces can generate hydrogen [12,13].
A number of studies conducted by the authors of this manuscript and others have used an automotive gear oil known to be 'bad' and rapidly promote WECs [3,6,[14][15][16][17][18][19], this oil containing Over-based calcium sulfonate (OBCaSul) and AW ZDDP additives. This oil causes bearing failure on the FAG-FE8 test rig between 16 and 40 h [14,18,20] in contrast to the 1000 h test under the same conditions without WSF, where a pure base oil was used [20]. FAG-FE8 testing conducted by the authors using this oil has shown that the diffusible hydrogen measured in the test rollers increased with RCF test duration, where this was coupled with an increase in WEC formations [21]. FAG-FE8 testing has shown that oils containing ZDDP, without the introduction of metal or non-metal additives formed WECs, addition of Ca/Na sulfonates also formed WECs as well as Ca/Na in the absence of ZDDP [5]. It is also noted that FAG-FE8 tests have created cracks using a mineral oil without any 'considerable' oil additives [22].
Calcium sulfonate detergents and more specifically OBCaSul are typically used in automotive gear oils; therefore, most of the literature focuses on their impact on AW performance of engine components for the automotive industry. Calcium sulfonate detergents are used to keep metal surfaces clean by forming a protective layer as well possessing good AW/EP, friction reducing and anti-scuffing properties. ZDDP decomposition rates have been found to be retarded by the OBCaSul detergents since the reaction between ZDDP and OBCaSul is proposed to be antagonistic to the performance of the AW tribofilm [23]. OBCaSul are suggested to have an antagonistic effect toward sulphur containing species in film formation, OBCaSul inhibiting ZDDP tribofilm formation [24][25][26] [26][27][28][29][30] and potentially aiding in hydrogen diffusion processes by promoting and prolonging nascent surface exposure [31].
A number of suggestions explaining the antagonistic effect of OBCaSul are as follows; OBCaSul inhibits the formation of a ZDDP film due to the formation of a colloid dispersion with ZDDP [32]. OBCaSul competition with ZDDP at the contact surface in the formation of a tribofilm, the OBCaSul breaks down to form carbonate, which is deposited on the metal surface [33]. It has been shown from multiple surface techniques that Ca displaces the Zn polyphosphates in the AW film [34,35], OBCaSul may inhibit polyphosphate chain formation by formation of Ca phosphate [36]. X-ray absorption near-edge structure spectroscopy (XANES) has shown that mix of ZDDP and OBCaSul resulted in the formation of considerable Ca phosphate in the film [37]. The over-basing component is suggested to retard the rate-determining step of AW action (decomposition stage) of ZDDP [38], however, others suggest that it can improve AW performance by forming a 'paste-like' structure in the contact [39,40].
Based on past literature, it is considered that the OBCaSul in the 'bad' oil mentioned above is the critical additive driving WEC formations, this 'driving' effect being unclear, where this study aims to investigate this. To accomplish this, rolling contact fatigue (RCF) testing of 100Cr6 steel using FAG-FE8 and PCS-MPR test rigs, lubricated with the 'bad' oil and purposely-formulated oils containing varying wt% of OBCaSul additive have been conducted. The tested samples have been examined and analysed using serial sectioning techniques, Thermal desorption analysis (TDA), and SEM/EDX of the tribofilms formed (RCF testing, serial sectioning, and TDA for the FE8 tests lubricated with the 'bad' oil are from previous studies conducted by the authors [18,21], and are used for comparison in this study). By using these techniques, the effect that oils containing OBCaSul additives have on the tribofilm formed in the contact, diffusion of hydrogen into the steel during operation, and WEC formations has been investigated, where the effect that these additive have on tribofilm and hydrogen diffusion influencing WEC formations have been explored.

FAG-FE8
Testing was conducted on a standard FAG-FE8 rig, comprehensive details of which can be seen in Ref. [18]. The test conditions (with the exception of the oils used) can be found in Ref. [18]. The tested bearing raceway are pre-roughened before testing to R q value of 0.5 μm. As illustrated in Ref. [18], the centre of the raceway track and rollers experiences pure rolling with rising slip of up to ± 12.5% slide-roll-ratio (SRR) along the contact major axis towards the edges. The rollers experience both -ve and + ve directional slip due to being sandwiched between two raceways.
Four tests were conducted in total using an oil specifically formulated for this study containing Group III ISO VG 68 with ZDDP AW additive, where the OBCaSul detergent wt% in the oil for each of the four tests was altered (ZDDP only (0% OBCaSul), 1.4%, 2.8% and 5.6%), base oil and ZDDP staying constant. Tests were shut down manually at 18 h to keep duration constant between tests. The calculated maximum Hertzian contact pressure Pmax is in the range of ∼1.5-1.9 GPa.
Initial minimum oil film thickness (h min ) between rollers and raceways is calculated using the Hamrock and Dowson equation [41,42]. Lambda ratio (λ) is calculated based upon h min and the roughness (R q ) values given in Ref. [18], the bearing has been run under boundary lubrication. Oil temperature in the contact is controlled at 100°C.

Micro-Pitting Rig (MPR)
Testing was also conducted on a PCS-Micro-Pitting Rig (MPR) (Fig. 1) so that a comparative assessment between the MPR and FAG-FE8 rigs on WEC formations could be made. The MPR is a computer controlled RCF rig in which three equal diameter counterface rings are in contact with a smaller diameter roller positioned in the middle of the rings. The three rings have a diameter of 54 mm and the test roller is 12 mm with a 1 mm wide wear track ( Fig. 1 (c)). Rings and rollers are AISI 52100 martensitically through hardened steel. Chemical composition of the roller and rings can be found in Ref. [3]. Rollers and rings have a Rockwell C hardness of 60 (746 HV) and 63 (867 HV) respectively [3]. Servo controlled motors control the rings and roller separately to allow for testing at different SRR. Lubricant is supplied to the contact through a dip lubrication system. A motorised ball screw applies load, which acts through an arm. A piezoelectric accelerometer is used to measure vibration in the contact during operation, once a threshold is reached the test is stopped automatically.
Five tests were conducted using the same four OBCaSul oils tested on the FAG-FE8 as detailed above, and a fully formulated semi-synthetic gear oil (Oil X) that contains ZDDP and OBCaSul 2.8% which is known to be 'bad' and promote WEC formations (full details can be found in Ref. [18]). The water content of these oils was also measured pre-RCF testing using Karl Fischer titration (see Table 1). All tests were run to 18 h before being shutdown manually. The calculated Hertzian contact pressure P max is 1.9 GPa. Subsurface shear stresses have been calculated using Hertizian contact theory. The SRR during operation is set at -30%. The test conditions of the five tests are given in Table 2.
Initial minimum oil film thickness (h min ) between the rings and roller has been calculated using the Hamrock and Dowson visco-elastic equation [41,42]. Lambda ratio (λ) has been calculated based upon h min and the roughness (R q ) values given in Table 2. Oil temperature is kept at 100°C during operation as is measured by a probe inserted into the test chamber.

Contact surface inspection and subsurface examination using metallographic analysis
All tested sample surfaces were initially inspected using optical macroscopy. For the FAG-FE8 test samples, images were taken at 60°i ntervals around the circumference of the rollers. For the MPR samples, images were taken at 120°intervals around the circumference of the rollers. Example images of the contact surface for each test can be seen in Fig. 3.
Metallographic analysis was then conducted to record individual WECs and associated damage features in all the RCF tested samples  from both FAG-FE8 and MPR testing using the same procedures as detailed in Ref. [18]. For the FAG-FE8 tests, rollers were prepared in the same way as in Ref. [18]. The MPR test roller were prepared in a similar way, the single test roller being analysed. Rollers were mounted such that they were sectioned in the axial direction from the outer roller edge (see Fig. 1 (c)). Details of the sectioning analysis for FAG-FE8 and MPR samples are given below.

FAG-FE8
Macro sectioning (at 50 μm per slice material removal rate) was first conducted on rollers from all the tests from the outer edge of the rollers until the point where WECs were first found. This was then followed by coarse serial sectioning (at average material removal rate of ∼30, 35 and 38 μm per slice depending on the test). This is deemed a suitable sectioning removal rate to catch the majority of WECs including smaller networks; however, very small WECs may have been missed due to the slice removal rates. Imaging of WECs was taken at every five slices. Inclusions were recorded during sectioning. Inclusions may have been missed due to the sectioning removal rates conducted, where inclusions observed in the FAG-FE8 rollers are typically small in size (typically ∼ 2-15 μm) [18]. The individual sectioning ranges (sectioning start and stop range) and removal rates (μm/sectioning slice) for each test are listed in Table 3. Oil X data is from Ref. [18], data for Oil X rollers given in Table 3 is for an 18 h test for comparison. Fig. 2 details the sectioning analysis procedure.

MPR
Careful material removal was first conducted on all tests from the outer edge of the roller up until the edge of the wear scar at ∼3.5 mm (no contact between roller and rings before this point). Once completed, fine serial sectioning was started (∼4.3-7.4 μm per slice depending on the test). Imaging of WECs was taken at every five slices. Inclusion-WEC interactions were recorded during sectioning. The individual sectioning interval ranges and removal rates for each test are listed in Table 3. Fig. 2 details the sectioning analysis procedure.

Thermal desorption analysis (TDA)
To measure the hydrogen content in the tested samples, that may have diffused into and been trapped in the steel during RCF testing, TDA was performed for the samples from the FAG-FE8 tests using two different experimental set-ups under BS ISO 3690 standards [43] as detailed in Ref. [21].

Tribofilm analysis
The rollers and raceways from FAG-FE8 testing have been analysed using SEM/EDX to check the tribofilms formed on the contact surface during operation. The MPR samples were not analysed in this study. Tribofilms formed by ZDDP only (0%), OBCaSul (1.4%-5.6%) and Oil X have been examined, where for Oil X the tribofilms formed at 2, 6 and 18 h RCF test durations have been examined (Oil X RCF testing is from Ref. [18]). Prior to the analysis, samples are rinsed in heptane, isopropyl alcohol and blow-dried. 5 kV accelerating voltage is used for the analysis. The estimated tribofilm thickness is calculated based upon an empirical power fit of 100+ MTM-Slim analyses spanning ashless and ZDDP AW tribofilms. The fit is developed between a normalisation   factor for the 100 + tests and the tribofilm thickness when measured by the MTM-Slim. For the FAG-FE8 analysed tribofilms, no corresponding MTM-Slim data was available; therefore, the normalisation factor is used to get an estimate.

FAG-FE8
Wear across the rollers is observed on both sides of the central pure rolling region corresponding to outer and inner zones (Fig. 3). No visible signs of spalling were seen on any of the rollers except that from Oil X test. Mild wear is observed on the ZDDP only (OBCaSul 0%) and OBCaSul 1.4% tests with a slight increase seen for OBCaSul 2.8% and 5.6%. Oil X [18] in comparison showed visible signs of damage on the surface in the form of dents/impressions, no signs of spalling were found on the rollers sectioned, however, spalling was observed on other rollers from the same bearing.

MPR
Only Oil X showed visible signs of spalling damage ( Fig. 3 (j)). Wear and micropiting are observed on all rollers from the OBCaSul oil tests.  Optical macroscope images of rollers from each test for the FE8-FAG and MPR test rigs. Images (a)-(e) show one of the 60°interval zones around the circumference of the roller for the FE8-FAG tests, for Oil X see Ref. [18]. Images (f)-(j) show one of the 120°interval zones around the circumference of the roller for the MPR tests.

Metallographic analysis
An overview of the FAG-FE8 and MPR tests results, including numbers of WECs, NMIs interacted with WECs and whether the WECs have surface connections, is summarised in Table 4. Inclusions were evaluated using a ranking system described in Appendix A.
Examples of NMIs found to have interacted with WECs are shown in Fig. 4, where the corresponding information of the inclusion type, size, location and test oil type is provided. As it can be seen, from the MPR testing, inclusions observed in the OBCaSul 0%, 1.4% and Oil X are small (average size at ∼3-11 μm, ∼2-8 μm and ∼3-13 μm respectively) and are either globular sulfide-oxides (D Dup ), globular oxide (D) of manganese sulphide or manganese duplex (A). For the FAG-FE8 tests, the inclusions have been previously observed in the rollers as either globular sulfide-oxides (D Dup ) and globular oxide (D) based upon extensive collection of inclusion results [18]. Due to the size of the serial sectioning rates (15-38 μm) conducted on the rollers from the ZDDP only (OBCaSul 0%) and OBCaSul (1.4%-5.6%) oils tests in this study, it was not possible to gather clear data on the length of inclusions. However, since the same type of bearings are used; it is assumed similar type and size of inclusions present in the rollers from the ZDDP only (OBCaSul 0%) and OBCaSul (1.4%-5.6%) oil tests.
The spatial distributions and depth of inclusion-WEC interactions for the FAG-FE8 and MPR tests are shown in Fig. 5.
During the serial sectioning processes, where possible, the dimension and location of individual WECs was observed and have been recorded, the results from the FAG-FE8 and MPR tests are summarised in Fig. 6.

Thermal desorption analysis
Both rollers and the raceways from the FAG-FE8 tests have been analysed to measure the content of diffusible hydrogen, and the TDA results are summarised below and in Table 5 and Figs. 8 and 9 (TDA results for Oil X are from Ref. [21]). TDA results from set-up 1 (Fig. 8) have shown that.
• Elevated concentrations of diffusible hydrogen were measured in the rollers.
• For both TDA setups, diffusible hydrogen contents in raceway sections are found to be negligible across all oils (close to the detection limit of the instrument) and are similar to that of the untested (0-h) control rollers.
• Higher water content (ppm) coupled with higher concentrations of diffusible hydrogen measured.
• For Oil X rollers and raceways, hydrogen desorbs out from the start of the analysis, this being later (250-300 min) for the OBCaSul 0% and 5.6% oils at a temperature range of ∼125-150°C.
• A single desorption peak is seen at ∼250-300°C for the ZDDP only oil (OBCaSul 0%) and OBCaSul 5.6% oil roller and raceways, this being much shallower for the raceways.

Tribofilm analysis
Tribofilm analysis results using SEM/EDX on rollers and raceways from FAG-FE8 testing are summarised in Figs. 10-14 and Table B1-B3 in Appendix B. EDX results of the film formed across the axial contact length of rollers and width of the wear track on the raceways for Oil X (2, 6 and 18 h) are shown in Fig. 10, and the corresponding results for the ZDDP only (0%) and OBCaSul (1.4%-5.6%) oils are shown in Fig. 11. Estimated tribofilm thicknesses are presented in Fig. 12. Figs. 13 and 14 present EDX phase maps for Oil X (2, 6 and 18 h) rollers and raceways.

WEC formation propensity
A significant difference in propensity and size of WEC formations is found for Oil X compared to the ZDDP only (OBCaSul 0%) and OBCaSul (1.4%-5.6%) oils (see Table 4 and Fig. 6 and 7), thus the different oil chemistry of Oil X promotes WEC formations more so than OBCaSul oils. ZDDP only oil (OBCaSul 0%) when tested on the FE8 showed no WECs (see Table 4 and Fig. 6), conversely, when tested on the MPR rig, this is the most detrimental oil to WEC size and propensity (see Table 4 and Figs. 6 and 7).
For the FE8 rollers, an increase in OBCaSul wt% from 1.4% to 2.8% and 5.6% resulted in a jump from zero WECs recorded to a number of WECs being recorded (17 and 12 WECs recorded for the 2.8% and 5.6% oils respectively) (see Table 4 and Fig. 6). Further testing with different wt% of OBCaSul between 1.4% and 2.8% should be conducted to    Fig. 7 for details). YZ-projection dots represent the position of each independent WEC in the Y-axis and Z-axis. For FAG-FE8 testing, note that Oil X data is taken from Ref. [18] and presented here for comparison.  Table 4 and Fig. 6), the 5.6% oil also showing no WECs. Testing of OBCaSul wt % below 1.4% but more than 0 wt% should be conducted to see if there is any trend for decrease in WEC propensity. It is suggested that the opposite behaviour between the FAG-FE8 and MPR tests may be due to differences in the rigs respective contact dynamics influencing the behaviour of OBCaSul, where differences may include.
(i) Slip -On the FAG-FE8 at the centre of bearing contact exists a pure rolling region, with rising slip of up to ± 12.5% SRR along the contact major axis, corresponding to slip regions on the roller/raceway. The MPR tests on the other hand run at −30% SRR, calculated 'frictional energy accumulation' [16] is thus far greater for the MPR (300%) compared to the FAG-FE8 (100%) [44]. It is noted, however, that lubricant factors are not accounted for in the energy accumulation model [44]. Frictional energy accumulation in areas of high slip [16] and resultant higher hydrogen adsorption are proposed to lead to higher concentrations below the surface resulting in WEC failures. (ii) Contact conditions -smaller area of contact for the MPR in comparison to the FAG-FE8, where for the MPR case it is proposed through modelling that the smaller contact area allows for less hydrogen diffusion/penetration and a faster escape rate due to the surrounding 'out of contact' area [44]. (iii) Contact cycles and 'regeneration time' -(the time between subsequent contact cycles, rollers being slightly longer than the raceways in the FE8 case), where longer regeneration times lead to longer WEC lives [16]. This is considered to be due to lower frictional energy accumulation and influence on surface film formation, which in turn may affect hydrogen diffusion.
It was also suspected that the differences in lubrication chemistry between the tested oils could affect the tribofilm formed during RCF operation (this is discussed in 4.4 of this paper), where contact dynamics/conditions could effect this, in turn potentially playing a role in hydrogen diffusion, and WEC formations. WECs recorded in the FAG-FE8 rollers across all oils fall within the zones (around ∼ 2 mm zone) of high slip energy (PVmax, see Fig. 6). It has been identified previously by the authors on serial sectioning analysis of the FE8 Oil X rollers [18], that this zone is where the highest propensity for WEC initiations exists (as shown in Fig. 6).

Subsurface or surface initiation of WECs
The vast majority of WECs recorded in the FAG-FE8 rollers across all oils were contained entirely within the subsurface and without connection to the contact surface, therefore surface initiation of these WECs is infeasible (Table 4). Any interaction/connections to the surface Fig. 8. Thermal desorption analysis (set-up 1 TDA @ 400°C) of rollers from FAG-FE8 ZDDP only (OBCaSul 0%) and OBCaSul (1.4%-5.6%) tested oils and Oil X. Oil X TDA data is from Ref. [21]. Diffusible hydrogen concentrations in ZDDP only (OBCaSul 0%) and OBCaSul (1.4%-5.6%) oil tested rollers (all run to 18 h), superimposed onto the measured concentrations for Oil X tested rollers on the FAG-FE8 rig. Fig. 9. Hydrogen desorption curves for FAG-FE8 ZDDP only (0%), 5.6% OBCaSul and Oil X (18 h only) RCF tested bearings using TDA set-up 2 (ramped TDA to 300°C including the total diffusible hydrogen concentration measured. (a) Hydrogen desorption rate (ppm/minute) vs. desorption temperature (°C). (b) Hydrogen desorption rate (ppm/minute) vs. analysis time (minutes). See Table 5 for individual TDA measurements. Note: Data for Oil X is from Ref. [21]. made were predominantly small/short in size relative to the size of the WEC network. Only five cases of interaction/connection to the contact surface were recorded for WECs across all oils tested on the MPR rig (Table 4), connections again being small/short in size. It must be noted that for any WECs that had not started and finished within the sectioning range, interactions/connections still may exist, and further  sectioning would be needed to confirm this. A number of inclusion-WEC interactions have been recorded in the OBCaSul oil rollers (see Table 4 and see Fig. 4 for examples). Inclusions recorded are small/short in size (FE8: ∼2-20 μm, and MPR: ∼2-13 μm) and being predominantly globular sulfide-oxides (D Dup ), globular oxide (D) or manganese sulfide/sulfide duplex (A) (A type only recorded in MPR rollers) in type. For the FAG-FE8, inclusion-WEC interactions recorded in the OBCaSul oil rollers are within a similar size range and type to the inclusion-WEC interactions recorded in the Oil X rollers [18]. It is considered that a number of inclusions may have been missed in the OBCaSul oils due to size of the sectioning removal rates (∼15-38 μm) conducted relative to the small size of the inclusions. A far greater number of inclusion-WEC interactions have been recorded in the FAG-FE8 Oil X rollers in comparison to the OBCaSul (1.4%-5.6%) oils (see Table 4). This is proposed to be due to (i) smaller removal rates (∼3-5 μm); therefore, more chance to 'catch' WEC-inclusion interactions; (ii) higher number of WECs recorded in the Oil X rollers, thus more inclusion interactions; (iii) specific lubricant chemistry of Oil X allowing inclusion initiation to occur more easily.
Given the evidence, one mechanism for WEC initiation and propagation for the FAG-FE8 and MPR tested oils is in the subsurface, where differences in lubrication chemistries between the tested oils has not affected this. This further verifies the author's previous findings on FAG-FE8 testing using Oil X [1,18] and other works by the authors [45][46][47] that WECs can initiate and propagate entirely within the subsurface. In contrast to the results found in this paper and previous works by the authors of this study [14,18], where WECs were recorded in the FE8 rollers only, in Ref. [22], as well as cracks also being recorded in FE8 raceways, cracks found in the FE8 raceways and rollers are proposed to originate from the surface in high slide regions, initiated by local high stresses at asperity contacts due to the rollers high surface roughness (it is noted that rollers were intentionally not ground to achieve a high surface roughness and intentionally embrittled through high temperature austenisation), cracks subsequently propagating into the subsurface toward the zone of maximum subsurface shear stress. WECs have been created and recorded previously through MPR testing [3,4] however, the authors of this manuscript point out that conclusive evidence on whether WECs initiated at the surface/ subsurface or interacted with inclusions cannot be made, since 3D mapping of WECs in their entirety was not conducted. This study shows conclusively that subsurface initiation and propagation can be one mechanism in the MPR tests.
In addition, evidence from this study further verifies the author's previous revelations that subsurface initiation is highly likely at small/ short non-metallic inclusions [1,18,[45][46][47]. Again, differences in lubrication chemistries between the tested oils has not affected this. For the FAG-FE8 case this is strongly supported, since inclusion-WEC interactions have been recorded at the infant stages of their evolution [18]; however, for the MPR case, this would require further sectioning analysis to catch inclusion-WEC at the early stages.

Influence of oil chemistry on hydrogen diffusion
Discussions are focused on FE8 testing only. Due to time considerations, TDA was not conducted on the MPR rollers. TDA for set-up 1 has shown that oils containing OBCaSul with ZDDP, and ZDDP only oils have allowed the diffusion of hydrogen into the FE8 steel rollers during RCF operation and not the raceways (negligible concentrations measured similar to untested '0 h control' samples) ( Table 5 and Fig. 8). Differences in oil chemistry, therefore, having  no effect on the diffusion of hydrogen into the raceway, this potentially being due to the tribofilm formed on the raceway (discussed in 4.4). For the OBCaSul (1.4%-5.6%) oils, there is a positive correlation between increase in OBCaSul wt% and increase in diffusible hydrogen concentration. Again, this could be due to the tribofilms formed (see 4.4 for further discussion), where it is considered that differences between oil chemistries (and respective tribofilms formed) could affect the flux/rate of hydrogen diffusion into the steel during operation. TDA results from set-up 2 showed that hydrogen diffused into the ZDDP only (OBCaSul 0%) and OBCaSul 5.6% oils (Table 5 and Fig. 9). TDA was not conducted on the OBCaSul 1.4% and 2.8% oils due to time considerations and so trends cannot be examined, further TDA on these oils should be conducted so that a comparative assessment can be made. Differences in measured hydrogen concentrations between roller and raceway may be due to.
(i) The differing dynamics experienced between roller and race, for example, number of contact cycles experienced and 'regeneration times' as mentioned above in 4.1.1. However, regeneration times are slightly longer for the rollers than the raceways, where longer times are said to lead to longer WEC lives, only the raceways failing during FAG-FE8 testing [16]. This was said to be due to the higher asperity energy accumulation experienced by the raceways. This is contradictory to results found by the authors of this manuscript, where only the rollers contained WECs [18]. (ii) It may be due to differences in steel cleanliness, the raceway being shown to be far 'cleaner' than the rollers [18], thus a lack of trapping sites for hydrogen being available. (iii) Differences in respective tribofilms formed on the roller and raceway (see 4.4 for further discussion).
Finally, differences in the water content measured between the tested oils (see Table 2) may play a role in the amount of diffusible hydrogen measured. Results show a correlation between increase in water content (ppm) measured in the oils pre-RCF testing, and an increase in diffusible hydrogen concentration measured in the rollers (see Fig. 8). The hygroscopic character of ZDDP, Na, Mg and Ca alkyl sulfonate additives may be a reflection on the increase in water content measured for the respective oils, where it is proposed that water dissociation and subsequent hydrogen generation due to high friction films formed by these additive components can occur [5]. Although water contents are low, small amounts of water as little as ∼100 ppm have been shown to have a detrimental effect on fatigue life [48]. It has been said that water desorption spectra should be taken into account when measuring the total amount of diffusible hydrogen in steels, where thermal desorption and XPS analysis showed that the permeation of hydrogen is strongly influenced by the formation of oxide and/or hydroxide film when water is present [13]. It is noted, however, that when water is present hydrogen is mainly sourced from the oil opposed to water [49].
The trapping behaviour of hydrogen is discussed in Fig. 15 below.

Hydrogen diffusion and its relation to WEC formations
Discussions are focused on FE8 testing only. While the diffusible hydrogen concentrations are comparable in the tests lubricated by the Oil X and OBCaSul oils, the number of WECs recorded in the rollers are significantly different, i.e. the concentration of diffusible hydrogen found in rollers lubricated by different oils may not have had the same effect on WEC formation, where minute differences or similar diffusible hydrogen concentration between rollers 'with' and 'without' WECs between the tested oils are found. Comparing for example TDA results for Oil X at 12 h and OBCaSul 1.4% at 18 h (Fig. 8). Although the average hydrogen concentrations are similar, no WECs were recorded in the 1.4% OBCaSul rollers while a number of WECs were found to have formed in the Oil X 12 h rollers (see Ref. [18] for Oil X 12 h WEC details). Again, if we compare diffusible hydrogen concentrations between Oil X and OBCaSul 5.6% (at 18 h), although their average hydrogen concentrations are similar, there is a vast difference in the number of WECs recorded (Table 4 and Fig. 6). This supports the suggestion made in a previous study by the authors, that hydrogen may not be an initiator but an accelerator to WEC formation Fig. 15. Trapping behaviour of hydrogen in FAG-FE8 tested rollers and raceways analysed through TDA set-up 2.
A.D. Richardson et al. Tribology International 133 (2019) 246-262 [21], and/or becomes instrumental in formation when in combination with other factors e.g. tribofilm formed. It is noted that, at this stage, it is not clear what the influence of local diffusible hydrogen, i.e. at crack tips or inclusions, may have due to the limitation of the TDA, which only measures diffusible hydrogen in the bulk of the specimens. It is also considered that differences in oil chemistries may affect the flux/ rate of hydrogen diffusion and penetration, this being directly related to the respective tribofilms formed (further discussion in 4.4). It could be that.
(i) Hydrogen diffused later towards the end of RCF operation, not allowing a sufficient amount of time for hydrogen to take effect. Further TDA to measure diffusible hydrogen at varying stages of RCF operation for the OBCaSul oils (test run to only 18 h for these oils) would be able to provide more detail on this. (ii) Hydrogen penetration may be outside the zones of maximum subsurface shear stresses, or where critical inclusions are located in the depth of the steel. (iii) Insufficient hydrogen diffusion in critical zones where greater propensity for WEC formations exists, the diffusion of hydrogen in these zones being governed by the presence of metal-containing additives in the film such as calcium sulfonates under high frictional energy accumulation [44].
A difference in hydrogen trapping behaviour is seen between Oil X, and the ZDDP only (OBCaSul 0%) and OBCaSul 5.6% oils (Fig. 15). This may be a result of the greater number of WECs formed in the Oil X lubricated rollers in comparison, where it is suggested that crack interfaces have acted as 'weak' 'reversible' hydrogen traps.
Based on the TDA and serial sectioning analysis, a positive correlation is shown between the OBCaSul concentration and the amount of diffusible hydrogen. In addition, a jump in diffusible hydrogen concentration is seen for a jump in WEC formations between OBCaSul 1.4% and 2.8%, further TDA and serial sectioning for OBCaSul wt% between 1.4% and 2.8% is recommended to explore this. An increase in hydrogen concentration is seen between OBCaSul 2.8% and 5.6%, however, no great difference in number of WECs was seen, although the length of WECs formations increased. TDA was not conducted on the MPR rollers due to time constraints and so it is not possible to derive this relationship for the MPR tests, further testing is thus required to make a comparative assessment. The vast majority of WECs recorded for the OBCaSul FE8 tested oils were contained entirely within the subsurface. This supports previous findings by the authors on FAG-FE8 testing lubricated with Oil X [21], that the suggested mechanism of hydrogen entry into the steel during operation is at wear induced nascent surfaces or areas of heterogeneous tribofilm, hydrogen being generated by decomposition of lubricant through catalytic reactions and/or tribo-chemical reactions of water [12,13].

Tribofilm composition
Discussions below focus on SEM/EDX analysis of the tribofilm from the FE8 tests only.
SEM/EDX has shown that the OBCaSul (1.4%-5.6%) oils have formed thicker Ca dominated films across wear zones on the rollers in comparison to 'normal' ZneS films that are shown to form for the ZDDP only (OBCaSul 0%) rollers and raceways, and the Oil X and OBCaSul (1.4%-5.6%) raceways. For Oil X, an increase in RCF test duration has also resulted in thicker Ca films forming across wear zones on the rollers. ZDDP + OBCaSul combinations have shown that surfaces covering worn areas showed mainly Ca and O in the film [32], the Oil X and OBCaSul (1.4%-5.6%) lubricated rollers in this study exhibiting this behaviour. AES has shown that by using neutral calcium sulfonate with ZDDP, films contained Ca but were otherwise similar to ZDDP, the OBCaSul showing Ca and O with no traces of ZDDP in the film and having a greater effect on the AW properties [32]. The dominance of Ca over Zn shown on the rollers is suggested to be a reflection of OBCaSul additive being a 'strong' competitor to ZDDP in the oil. It is proposed that a replacement of Zn in the anti-wear tribofilm has resulted in thicker Ca films forming on the rollers. Ca replacement of Zn has been shown using multiple techniques on the analysis of automotive AW tribofilms on cam/tappet friction test specimens [35] and other engine parts [34].
As shown in Fig. 10, for Oil X at 2 h, the elements in the film are sporadically dispersed across the roller, while a well formed ZDDP tribofilm is shown to have developed on the raceway (indicated by the higher concentrations of Zn, S and P in the wear zones). In addition, the film thickness on the raceway is found to be thicker than that on the rollers at 2 h (see Fig. 11). Over time, while the film on the raceway does not change significantly, those on the rollers have developed and grow thicker, tribofilms on the raceway being on average thinner than the rollers. It is not clear why tribofilms form quicker on the raceway, but it may be an explanation for why no WECs were found to have formed in them. It is suspected that the differing contact dynamics between roller and raceway may affect film formation. For example, the rollers experience slightly longer 'regeneration times' (the time between subsequent contact load cycles) than the raceways, where longer regeneration times are said to improve resistance to WEC formation [16]. This is, however, contradictory to the results found in studies by the authors on FE8 tests using Oil X, where rollers contained multiple WECs, the raceway sections not showing any [18]. It is considered that longer regeneration times could influence film formation [16], which may be a reason as to why differences in tribofilm have been found in this study. More recently, it has been proposed that high surface shear forces in sliding areas due to the tribofilm formed (heterogeneous films composed of AW/EP and detergent/rust inhibiting additives) caused the initiation of surface cracks in FE8 RCF tested raceway samples [17].

Tribofilm formation influence on H-diffusion and WEC formations
TDA showed a positive correlation between OBCaSul wt% and diffusible hydrogen concentration for the FE8 rollers (see Table 5 and Fig. 9), where OBCaSul in oils containing ZDDP have formed thick Ca films. Therefore, it is suggested that the thick Ca films formed on the FE8 rollers have promoted hydrogen diffusion into the bearing steel. For Oil X an increase in RCF test duration also resulted in thicker Ca films forming, this being correlated with an increase in diffusible hydrogen concentration and WEC formations in the FE8 rollers. Based on TDA and serial sectioning analysis of the FE8 raceways, negligible hydrogen concentrations were detected and no evidence of WECs found, therefore, it is considered that the thinner ZneS containing films formed on the raceways has demoted hydrogen diffusion and WEC formations.
Literature reports that an oil containing OBCaSul resulted in WEC detection (using ultrasound) in FE8 raceways, XPS/SNMS analysis showing increased film thickness (reaching 100 nm) over test duration and comprising of CaeP on the raceway slip zones [2]. In contrast, removing OBCaSul resulted in no WEC detection, the developed film being thinner (20 nm) and comprising of ZneS in the same regions. It must be noted that these tests used ceramic rollers to 'push' WEC failure to the raceways. ZDDPs are shown to reduce hydrogen permeation into steel, where it is suggested that these additives can form protective film preventing hydrogen permeation [50]. Studies in literature also propose that OBCaSul inhibit film formation and create 'weakened' heterogeneous ZDDP films [24][25][26][27][28][29][30], OBCaSul being suggested to promote and prolong nascent metal exposure [31].
In literature, STEM analysis shows heterogeneous tribofilms containing Ca, S, Zn, P, Na and O with varying thickness (5-100 nm) to form on FE8 raceways that also showed WECs [17]. These elements were also observed on the roller films, where the study notes that the morphology and composition are different to the raceways, however, limited results are presented. It is worth mentioning that the rollers were not examined for damages since the majority of damage was found on the raceway surfaces, this contradicts what the authors of this study have seen on FE8 bearings using Oil X, where damage is only seen on the rollers [1,18].
Finally, it has been shown that for higher water contents (ppm) measured in the tested oils pre-RCF testing, higher diffusible hydrogen concentrations were measured (see Fig. 8). Water dissolved in the oil is said to influence the functionality of ZDDP AW additives and enhance wear [51][52][53][54]. A feature of ZDDP, and Ca alkyl sulfonate is that they are hygroscopic, where these additives may carry water molecules to rubbing contact surfaces [5]. MTM-Slim studies have shown oils containing ZDDP, Na, Mg and Ca alkyl sulfonates form thick patchy tribofilm, which are proposed to cause high friction in an initial incubation period (∼20 h for FE8) resulting in water dissociation at fresh nascent surfaces and subsequent generation of hydrogen that diffuses into the steel promoting WEC formations [5].
Mechanisms for why Ca tribofilms are thicker than the ZneS tribofilms and why these films may promote/demote hydrogen diffusion is still unclear based on the SEM/EDX analysis. Further tribofilm analysis using XPS and STEM has been conducted by the authors (to be published in a later study) and may be able to provide more information on the properties of the tribofilms and reveal physical mechanisms further. One answer may be that sulphonates can act as hydrogen poisoners, meaning that they can inhibit the recombination of hydrogen atoms (or radicals) to molecular hydrogen, hydrogen ions that then exist at the contact surface can recombine with electrons at fresh nascent metal surfaces forming atomic hydrogen that can diffuse into the steel [11]. Adding to this effect, it is also proposed that due to OBCaSul wanting to keep surfaces clean and prevent deposits at the contacting surface, the exposure of fresh nascent metal surfaces is proposed to be heightened [31].

Conclusions
RCF testing has been conducted under non-hydrogen charged conditions on FAG-FE8 and MPR rigs using 100Cr6 steel samples lubricated with oils containing varying wt% of OBCaSul detergent additive. The effects that oils containing OBCaSul have on tribofilm formation, diffusion of hydrogen into the steel, and the formation of WECs has been investigated. The findings of this study are: • Serial sectioning has shown that WECs can initiate and propagate entirely within the subsurface and are frequently found to interact with small/short non-metallic inclusions. This has confirmed and added further verification to the author's previous findings.
• The number of WECs formed in the FE8 rollers appears to be increasing with the increase of OBCaSul wt% from 1.4% at 0 WEC, to 2.8% and 5.6% at 17 and 12 WECs respectively, suggesting that OBCaSul has promoted WEC formation. However, in the MPR tested rollers, only the OBCaSul 1.4% lubricated roller showed WECs, higher OBCaSul wt% (2.8% and 5.6%) not creating any potentially due to their differences in contact dynamics.
• Similar to the results obtained by FAG-FE8 testing using the fully formulated 'bad' oil (Oil X), TDA has also shown hydrogen to have diffused into the FE8 rollers and not the raceways during RCF operation for the OBCaSul containing and ZDDP only oils. TDA has shown a positive correlation between OBCaSul wt% and concentration of diffusible hydrogen. Further TDA should be conducted on the MPR rollers to confirm this relationship.
• Detailed tribofilm analysis has shown that oils containing OBCaSul additive form thicker Ca dominated tribofilms on the FE8 rollers as opposed to the normal ZneS dominated tribofilm. This may have promoted hydrogen diffusion and WEC formations in the FE8 rollers. Thinner ZneS containing films have been found to form on the FE8 raceways however, and these may have demoted hydrogen diffusion and WEC formations. 71

Declarations of interest
None.

Funding
This work is supported by a UK EPSRC PhD studentship; and Afton Chemical.