**3. Results**

#### *3.1. Surface Composition*

The carburizing and carbo-nitriding heat treatment processes performed on MSS P675 resulted in the development of well distributed metal-carbon precipitates (carbides) ranging in size from approximately 10 nm to 2 μm in diameter (Figure 2a), surrounded by the martensitic matrix at the sample surface. In addition to the surface, the carbides are present diminishingly, approximately 1000 μm radially inward into each of the samples (data not shown). Sample surfaces were analyzed via EDS (Figure 2b) to resolve carbide chemistry and determine alloying elements that segregated from the matrix to form these carbides during heat treatment. Carbides resulting from all three surface treatments were found to be predominantly carbon- and chromium-rich with lesser amounts of vanadium, molybdenum and/or manganese, while the surrounding matrix showed primarily iron, cobalt, and nickel. In previous work done on P675, X-ray diffraction (XRD) and electron beam backscattered diffraction (EBSD) determined M7C3 (orthorhombic) and M23 C6 (face-centered cubic) to be the primary carbides formed in P675 (M represents the metal in the carbide), with M23 C6 precipitating after M7C3, and chromium being the primary metal constituent present in the carbides [27,53]. HTT contains a greater population of M23 C6 carbides than LTT and CN due to its higher tempering temperature (i.e., increased kinetics). By stoichiometry, the HTT carbides contain more chromium than the carbides of the other two surface treated steels despite all having the same bulk composition before heat treating. The large amount of chromium present in the bulk (pre-heat treatment) P675 alloy (Table 1), coupled with the presence of molybdenum, should yield a magnetic MSS [54–56]. However, EDS analysis (Figure 2a,b), performed on the bulk surface of each steel, showed that the chromium and molybdenum primarily segregated within the carbides following heat treatment (thereby increasing the likelihood of magnetic carbides). EDS was performed on the bulk steel and not on the individual carbides due to inconsistent results obtained since large interaction volumes (by the EDS) penetrated both the carbide and surrounding matrix. In contrast, nickel, in the presence of iron and carbon acts as an austenite stabilizer and thus promotes a non-magnetic austenitic (fcc) structure [57]. MFM was therefore utilized to observe how the secondary processing performed on these steels affected the magnetic properties of the surface.

**Figure 2.** (**a**) Grayscale backscattered electron (BSE) images (left column) of the three different P675 surface-treated samples (carbides appear darker than surrounding matrix) with corresponding colored energy-dispersive X-ray spectrometer (EDS) compositional maps highlighting the principal components of the carbides (middle columns) and bulk matrix (right columns) for the HTT, low-temperature tempered (LTT), and carbo-nitrided (CN) samples (images for each row share the same micron bar). (**b**) Elemental composition in wt.% (determined via EDS) for the surface of each steel (not individual carbides).

#### *3.2. Scanning Probe Microscopy*

#### 3.2.1. Magnetic Force Microscopy (MFM)

MFM was utilized to map variations in the magnetic moment projections (surface normal direction) on the surface of the steels (Figure 3). In Figure 3, purple regions are identified as carbides since these coincide with regions that are raised in topography and visually similar to carbides seen in SEM/EDS analysis (see Figure 2). Topographical relief of the carbides was expected due to differential polishing rates during sample prep, resulting in the harder carbides slightly protruding above the surrounding matrix. MFM results indicated that the carbides and the matrix both exhibit out-of-plane magnetic domains (i.e., positive magnetic direction, non-parallel to surface), but with varying magnitude, carbides being noticeably larger than the matrix as expected from the enhanced chromium concentration (see Figure 2). Within the matrix, nanoscale variations in magnetic domain were also evident. In HTT these were larger and more elongated those on either HTT or CN. CN had the finest distribution of different magnetic domain regions able to be resolved within the matrix.

**Figure 3.** 3D magnetic response maps with changes in height representative of differences in magnetism. Color scale ranges are 7 degrees (0◦ = yellow, +7◦ = blue) for magnetic response.

#### 3.2.2. Inert Environment Scanning Kelvin Probe Microscopy (SKPFM)

Freshly polished, cleaned, and dried samples underwent ex situ SPM imaging in an inert atmosphere glovebox. Images were acquired using sequentially larger scan areas of 10 × 10 μm2, 20 × 20 μm2, and 90 × 90 μm2, with contrast between carbides and the surrounding matrix observed in both Volta potential and topography (Figure 4). Numerical VPD results were calculated per the method described earlier and compiled for comparison (see Figure 5, error bars are indicative of one standard deviation). The measured VPD of the carbides ranged from 60 to 200 mV greater than the steel matrix, depending on the scan size analyzed, with HTT possessing the highest difference and CN the lowest. The relative magnitudes of the carbide-matrix VPDs remained consistent regardless of scan size, suggesting even the smallest imaging areas chosen (10 × 10 μm2) were large enough to be representative of the sample while also providing the highest spatial resolution of VPD variations.

**Figure 4.** High-resolution atomic force microscopy (AFM) topography (dark brown to white color scale, 100 nm full scale) and SKPFM Volta potential (green to pink color scale, 600 mV full scale) images over different size scan areas showing the different sizes and shapes of carbides distributed throughout the three sample types.

**Figure 5.** Plot of measured VPDs (with standard deviation error bars) of carbide precipitates versus the surrounding matrix for the three P675 surface-treated steels as a function of scan area.

#### 3.2.3. Intermittent SKPFM

Intermittent ex situ SKPFM was performed to track the evolution of the surfaces resulting from sequential sustained exposure to corrosive conditions. Samples were placed in a corrosive salt solution and the VPD maps were obtained at intervals of 0, 1, 2, 10, and 15 cumulative minutes of exposure to 1 M NaCl solution (Figure 6). Qualitative differences in both appearance (surface topography and morphology) and carbide-matrix VPD over time were observed for the steels. The HTT sample showed the formation of particulates on the surface and degraded uniformly with time, leading to a progressively lower variation in surface VPD. In contrast, the CN sample showed little change in VPD or topography on the surface, indicating corrosion reaction kinetics were much slower despite the distinct VPD contrast between the carbides and matrix. LTT exhibited behavior somewhere in between the other two steels. Initially, salt deposits on the LTT surface obscured the steel topography and VPD variations. However, with increasing time LTT appeared similar to CN, as evidenced by the relatively large contrast in topography and VPD apparent by the 15 min mark (see Figure 6).

**Figure 6.** SKPFM Volta potential maps (green to pink color scale, 400 mV full scale) overlaid on the evolving 3D topography (30 nm full scale) of the three heat-treated MSSs as a function of immersion time in 1 M NaCl solution.

Figure 7 presents VPD maps (left column) and plots of Volta potential versus location (middle and right columns) for cross sections of different carbide/matrix interfaces as a function of exposure time. As can be seen in the top row of Figure 7, the VPD between the HTT carbides and the surrounding matrix decreased with exposure time, while the VPDs of the LTT (Figure 7(b1,b2)) and CN (Figure 7(c1,c2)) carbides remained relatively constant throughout testing. For HTT, corrosion proceeded simultaneously both along grain and carbide boundaries as well as within the matrix. Corrosion products evolved and settled on both the matrix and surface carbides, where cathodic activity was supporting anodic dissolution of the matrix. With this production and deposition of corrosion products, the VPD between carbides with a native oxide and matrix decreased on the HTT surface until there was very little difference observed between the two, as seen in Figure 7(a1,a2). Conversely, the LTT and CN samples underwent typical localized corrosion (see Figure 6), wherein highly localized attack adjacent to grain boundaries/carbides was seen, as evidenced by particulates settling on or near the carbide-matrix interface. As time in solution progressed, the VPD between the carbides and steel matrix remained essentially unchanged throughout the duration of testing, with matrix attack relatively shallow. Therefore, there are notable differences in the initiation of corrosion mechanisms between different heat treated samples.

**Figure 7.** SKPFM Volta potential maps ((**<sup>a</sup>**,**b**,**<sup>c</sup>**) 600 mV full scale, exposure time given below each image) for each of the three heat-treated MSSs with time-dependent Volta potential profiles (**a1**–**c2**) across two representative carbides plotted as a function of duration of exposure to 1 M NaCl solution. The location of the carbide represented by each profile is indicated by the corresponding dotted box in the exemplary SKPFM maps at left.

#### 3.2.4. In Situ Atomic Force Microscopy (AFM)

To observe the progression of corrosion in real time while samples were immersed in 0.1 M NaCl solution, in situ AFM was employed to monitor topographical changes over time. Figure 8 shows the results for the three heat-treated P675 steels with no applied bias voltage. (Variations in exposure time across samples are due to differences in corrosion rate and the time necessary to implement optimized imaging parameters.) For HTT, corrosion activity rapidly progressed and large surface deposits (~1–2 μm wide) appeared on the surface after ~107 min (Figure 8). EDS analysis indicated these large features to be iron-rich corrosion products with NaCl (analysis not shown). Despite the deposited particles, distinct localized corrosion was not seen on the HTT sample. As testing progressed, corrosion reactions proceeded, depositing corrosion product particulates on the surface (see Figure 8–HTT 116 & 134 min). In comparison, highly localized corrosion was evident at the carbide-matrix interfaces in both the CN and LTT samples. CN showed the greatest segregation of corrosion between matrix attack and the unaffected carbides, as indicated by near complete but shallow etching attack along carbide boundaries (Figure 8). LTT appeared to behave somewhere in the middle of these two extremes, with particle build-up similar to HTT seen initially, but eventually, these particles cleared to reveal evidence of localized corrosion propagation in the matrix adjacent to some of the carbides, similar to CN.

**Figure 8.** Time-lapse in situ AFM topography maps (160 nm full scale) for each of the heat-treated MSSs in 0.1 M NaCl solution, with approximate exposure time at the end of each scan indicated below the corresponding map (image time was ~8.5 min).

Time-dependent line profile analysis of selected carbide particles was conducted on each of the samples (Figure 9), confirming the qualitative observations arising from the images presented in Figure 8. HTT showed an increase in surface contrast of the carbides, up to 50 nm, with corresponding slight, uniform changes in the height of the surrounding matrix. For LTT, height changes across the carbide/matrix interface initially (44 min) showed ~100 nm deep attack immediately adjacent to the carbides (Figure 9). Then at longer times (112 min), the height of the carbides increased, accompanied by shallower apparent depth of attack in the adjacent matrix area. These changes are likely associated with the production and deposition of insoluble corrosion products. CN exhibited the sharpest contrast in topography by the end of exposure to salt solution, with the carbide surface height increasing by ~25 nm relative to the adjacent bulk matrix, with matrix attack limited to ~75 nm deep and only extending approximately 0.5 μm away from the carbide interface. The depth of attack also decreased from 103 min to 112 min, indicating slight corrosion product deposition within the highly localized area of matrix attack.

**Figure 9.** Topography maps ((**<sup>a</sup>**,**b**,**c)**, 160 nm full scale, exposure time indicated below corresponding map) for each of the three heat-treated MSSs with height profiles across selected carbide-matrix interfaces shown as a function of exposure time to 0.1M NaCl solution (**a1**–**c2**). Location of each profile is indicated by the corresponding box in the exemplary topography maps presented at left for each of the three heat-treated steels.

Post-testing SEM imaging was conducted on the same sample surfaces (Figure 10) to record surface morphological differences following the in situ AFM testing. HTT exhibited a distinctively different surface morphology compared to LTT and CN, characterized by the presence of large, fluffy appearing salt-laden corrosion deposits. Beneath these deposits and surrounding the carbides, the entire matrix surface area was uniformly corroded with no indication of matrix passivity. In contrast, both the LTT and CN carbide boundaries were attacked, with NaCl particles present along the grain boundaries and carbide-matrix separation and subsequent grain separation (Figure 10). LTT showed some attack along carbide boundaries as well as some generalized attack as indicated by roughening of the entire surface due to corrosion product deposition. CN displayed much more localized attack at the carbide boundaries than LTT (dotted oval in the right panel of Figure 10), and narrow "valleys" on the order of ~0.5 μm wide were observed around the CN carbides, confirming observations in Figure 8. Furthermore, unlike LTT or HTT, CN did not show evidence of adhered or deposited corrosion products. Tracing the representative "line of attack" for the CN sample in Figure 10 reveals a grain undergoing intergranular attack, indicative of microgalvanic corrosion between the noble carbides and the active matrix.

**Figure 10.** SE SEM images of the sample surfaces following in situ AFM testing. Red squares in the images in the top panels indicate areas of magnified images below. Dotted red oval area in magnified CN image indicates the "line of attack" (see discussion).

#### *3.3. Electrochemical Corrosion Testing*

To elucidate the corrosion pitting and repassivation behavior of the samples, cyclic potentiodynamic polarization (CPP) scans were conducted on each of the samples to explore the effects of the different heat treatments. Figure 11 shows the resultant polarization curves, along with macro images of the sample surfaces post-electrochemical testing. Testing indicated that HTT had the lowest OCP (−400 mV), followed by LTT (−200 mV) and CN (−80 mV), respectively. This ranking is in agreemen<sup>t</sup> with previous studies that ranked corrosion resistance for these same steels (i.e., corrosion rate determined via electrochemical methods) [7,9]. The LTT and CN samples exhibited a rapid change in potential over a minimal increase in current density (Figure 11a, green boxed areas), indicative of typical passive behavior. The breakdown potential of the LTT and CN samples occurred at 40 mV and 95 mV, respectively. Conversely, the HTT sample showed active corrosion behavior as demonstrated by linear growth of the current density over the potential sweep. However, pits were initially observed on the HTT surface (−200 mV), but did not grow and as the anodic overpotential continued to increase. The post-corrosion images in Figure 11b show the difference in corrosion morphology for each sample following CPP testing. For HTT, the entire test area darkened due to corrosion product formation (Figure 11b), engulfing the initially isolated areas of pitting. Arrows in Figure 11b indicate the four pits that first formed on the HTT sample before the entire test area underwent generalized corrosion. As expected from previous work [7], LTT and CN showed a distinctly different morphology of corrosion attack, with corrosion limited to only several dispersed pits on the surface of the sample. Compared to HTT, LTT showed limited regions of depassivation emanating from corrosion pits, evidenced by regions of minor surface darkening. In contrast, corrosion attack on CN displayed only highly localized, isolated pits (Figure 11b) with no visual evidence of any other associated areas of depassivation.

**Figure 11.** (**a**) Cyclic potentiodynamic polarization (CPP) scans (0.01 M NaCl electrolyte) for all three surface treatment samples. Passive regions for LTT and CN are indicated by green squares. (**b**) Images of the samples post-testing (after the area masking tape was removed) with dotted red circles indicating the test location on each sample surface. All samples display some isolated pitting. However, due to the difficulty in clearly seeing the pits on the HTT sample (which, in contrast to the other samples, underwent generalized corrosion attack), yellow arrows indicate the location of the pits present on the HTT sample.

## **4. Discussion**

#### *4.1. Nanoscale Origins of Corrosion Initiation*

Determining the nanoscale contributions to a material's bulk corrosion rate is inherently difficult due to the complexity and multitude of variables that influence its behavior in a corrosive environment. Corrosion is a spontaneous process driven by thermodynamics [58,59]. In a microgalvanic couple, the difference in electrode potential of the anode and cathode regions on the surface correlates with the magnitude of negative free energy change (thermodynamic propensity) for local corrosion to occur. SKPFM Volta potential (VPD) mapping is the highest spatial resolution method available to directly measure the relative thermodynamic propensity for corrosion between nanoscale heterogeneities in a material. For the MSSs considered in this study, the relatively high Cr composition of the carbides suggests they are likely noble in comparison to the matrix based on the galvanic series [60]. Hence, a larger VPD between carbides and the matrix will lead to a greater drive (i.e., increased microgalvanic full-cell potential) for corrosion of the matrix. Among the steels studied, HTT consistently exhibited the largest VPD between the carbides and the matrix (200 mV), while LTT (150 mV) and CN (90 mV) were considerably less (Figure 5). The relative magnitudes of these VPDs can likely be attributed to carbide chemistry, as HTT carbide composition is predominantly M(Cr)23C6 compared to predominantly M(Cr)7C3 compounds for LTT and CN. An interesting finding of this study is that for each of the surface treatments considered, the bulk OCP values measured inversely corresponded with the magnitude of the VPD between the carbide and matrix phases (Figure 5). HTT had the greatest carbide/matrix VPD and least noble OCP (−400 mV), CN had the lowest VPD and most noble OCP (−80 mV), and LTT was intermediate with a carbide/matrix OCP of −200mV. This observation demonstrates how local SKPFM measurements of the relative microgalvanic couple potential contribute to the bulk OCP observed on each of the different surface-treated MSSs investigated. In addition, variations in chromium enrichment of the carbides subsequently influenced both the VPD and degree of passivity of the surrounding chromium-depleted matrix. The steepest VPD gradients measured were across the carbide/matrix interface (Figure 7), and so SKPFM measurements also provided a technique to predict and locate expected points of microgalvanic corrosion initiation on the surface.

#### *4.2. Corrosion Propagation*

SKPFM measures VPDs on the surface, which are influenced by the presence of oxide layers. With MSSs, passivating chromium oxide layers are readily formed and act as a kinetic barrier to corrosion, which complicates any correlation of thermodynamic propensity derived from SKPFM measurements. However, for the steels considered herein, since the bulk composition is the same, data obtained from SKPFM also provided information on the spatial variations in surface properties that influence corrosion propagation. Intermittent SKPFM testing was conducted to monitor shifts in microgalvanic couples' VPD over time due to corrosion activity. For HTT, the VPD between the carbides and the matrix decreased with time (Figure 7). As a result, as the duration of corrosion propagation increased, the VPD between carbides and the matrix approached 0 mV for HTT, resulting in a more thermodynamically homogenous surface. In contrast, for LTT and CN, the initial VPD between the carbides and matrix phase was smaller, but remained nearly constant throughout testing, with only minor evidence of the corrosion activity apparent on the surface (Figures 6 and 7). This behavioral difference can be attributed to differences in the passive oxide layer performance, and is also reflected in the VPD measurements, which are highly influenced by the presence of surface oxides. Previous work by Schmutz and Frankel showed similar behavior on aluminum alloys and indicates that the shift in VPD observed on HTT following active corrosion was caused by oxide growth at cathodic sites and the generation and deposition of corrosion products at active sites creating a more homogenous surface [51]. For carburized MSSs, the magnitude of VPD surface variation measured by SKPFM pre-corrosion provided an indication of the how the VPD evolved as a result of exposure to corrosion conditions: smaller initial VPD between the carbides and matrix phase indicated more robust passivity during corrosion, as seen in CN and LTT steels. For HTT, the higher initial VPD between the carbides and matrix indicated a greater susceptibility to depassivation and more uniform corrosion activity during propagation. These findings were validated with bulk electrochemical testing (Figure 11), where CPP testing showed that LTT and CN had a more protective oxide layer as indicated by the presence of a passive region in the CPP scan. Moreover, during intermittent SKPFM testing the VPD on HTT evolved rapidly and HTT exhibited active corrosion behavior throughout CPP testing.

While the bulk amount of chromium present at the surface is the same for all steels considered, the spatial distribution is different among the three surface treatments, leading to distinctly different corrosion properties and behavior. Relative to LTT and CN, HTT tended to corrode more uniformly and had a higher VPD between carbides and matrix. HTT was more prone to depassivation compared to LTT despite both having identical bulk chemical composition and same carburization cycle (carburized in single furnace load). The different carbide-matrix VPDs among the samples influences or indicates how local solution chemistry likely evolves during active corrosion on MSSs. This suggests that for HTT, as pitting progressed, the local solution chemistry, most likely due to higher sensitization during tempering cycle, was sufficiently aggressive to cause widespread depassivation. Conversely, with LTT and CN samples, the VPD between carbides was smaller and pitting was unable to transition to more widespread corrosion, suggesting local solution chemistry evolution did not support auto-catalytic depassivation as corrosion propagated. Here the lower VPD observed for LTT and CN indicated the matrix phases exhibited more robust passivity than the matrix of HTT. The in situ SKPFM VPD measurements correlate with the observed corrosion morphology of the steels. That is, the measured carbide-matrix VPD for each steel is inversely proportional to the extent of general (uniform) corrosion resistance of the steel. The efforts in this paper show that SKPFM is able to effectively predict bulk corrosion behavior of different surface treatments by observing and measuring nanoscale surface VPD differences between carbides and the underlying matrix.

#### *4.3. SPM Characterization and Implications on Wear*

MFM provides a method to characterize local variations in magnetic properties that contribute to the bulk magnetic properties. For all steels studied, the carbides showed variable shades of purple/blue in the MFM maps (~1–3◦ phase shift), indicating slightly different magnetic properties within the individual phases (Figure 3), likely due to different carbide compositions in terms of the relative amounts of chromium and molybdenum, which influence the magnetic properties of phases [61–64]. Sample CN had a much less homogenous matrix that showed considerable variation in magnetic properties and is likely an effect of the relatively high surface retained austenite (18–22%) found within the matrix phases compared to LTT (10–13%) and HTT (1–2%) [9,30]. The bulk magnetism of the steels will change with tempering temperature and heat treatment process, following changes to the microstructural phases formed [9,30]. Further work is currently underway to investigate the implications of local magnetism and magnetic domains on resulting wear and corrosion mechanisms.

Similarly, the ability to resolve nanoscale variations in the resistance to deformation (elastic modulus) on a material's surface could help improve prediction of the wear behavior. The PeakForce tapping mode employed here measured differences in the elastic modulus distribution, as determined via the Derjaguin-Muller-Toporov (DMT) model [65], for the carbide and the matrices of the steels simultaneously with topography (see exemplary Figure 12). As seen in the CN image presented in Figure 12, carbides had a higher relative modulus than the matrix, suggesting potential sites for development of micro-cracking and fracture would likely lie at the interface between carbides and matrix where local modulus variation was greatest. Further work is underway to determine how these local differences in recorded modulus correlate to a material's ability to handle loads/stress in bearing applications.

**Figure 12.** (**a**) CN topography (160 nm full scale), and (**b**) DMT Modulus (1.5 GPa full scale). Images are representative of 103–112 min submersion in 0.1M NaCl solution.

In service, the uniform degradation seen on HTT could be effectively monitored conventionally via visual inspection, detection of wear debris, or thickness monitors installed on bearing raceways. For CN and LTT, current methods of monitoring engine health are less effective since significantly lower amounts of reaction products are generated from highly localized corrosion. Localized corrosion may not be detected until it has led to significant wear damage. Bearing steel developers should, therefore, be cautious with heat treatments that yield a surface similar to CN which, although highly corrosion resistant, the passive surface will inevitably be compromised in wear applications. Small areas of highly localized corrosion pits lead to surface crater development which can potentially lead to highly undesirable and unpredictable failure via spalling. LTT behavior was intermediate between the two other surface treatments, with some localized attack on grain/carbide boundaries as well as some evidence of wider depassivation. In corrosive environments, the overall wear lifetime may be controlled by resistance to corrosion initiation, in which case LTT and CN could provide greater benefit than HTT. Previously conducted wear studies are in agreemen<sup>t</sup> with the recommendations given, and the results of this study provide nanoscale insight to help understand why HTT outperformed both CN and LTT during rolling contact fatigue testing even though it had significantly lower corrosion resistance [29,30]. Based on this work, P675 HTT would be recommended over the other two tempering procedures for use in aerospace bearings where corrosion is not a primary concern. However, when the bearing assembly is prone to corrosion attack, HTT is not recommended due to its overall low corrosion resistance [7] which would lead to premature failure via degradation of the material. In this case, CN is recommended for bearing use due to its high resistance to both corrosion onset and propagation [7].
