**1. Introduction**

The performance of advanced gas turbine engines is currently limited by degradation of the mechanical components, in particular, rolling bearing elements, such as the raceway [1]. This is because aerospace engine bearings are subject to extreme operating conditions, including elevated temperatures, high speeds, vibratory stresses, rolling contact fatigue, and complex lubricant and environment interactions [2]. Accordingly, both high hardness and high toughness are critical requirements for aerospace bearing materials, ye<sup>t</sup> achieving both in a single material is challenging. M50, a through-hardened carbon steel, was developed for aircraft engine bearing applications and has become the standard bearing steel used in the United States due to its ability to perform well at high temperatures while maintaining relatively high fracture toughness compared to earlier generation

carbon steels, such as AISI 52100 (UNS G52986) [1,3,4]. In the case of sea-based or coastal aircraft operations however, open turbine engine systems can limit the ability of ester-based lubricants to provide wear and corrosion protection, as the surrounding environment introduces water and marine aerosols into the engine during both storage and operation [5]. The presence of water in the lubricant can then serve to initiate aqueous corrosion during engine cycling and downtime [5]. Consequently, current aero-engine performance is limited by corrosion-enhanced wear of the metallic bearings and drive components, which leads to increased maintenance and premature failure [1,6–8]. Thus, there has been significant research effort to develop alternative bearing steels to M50 that exhibit enhanced corrosion resistance to support increased engine performance [3,4,7–10].

Martensitic stainless steels (MSSs) were developed for use in applications where high wear resistance and toughness is required whilst maintaining high corrosion resistance. These properties, combined with their potential for high hardness upon heat treatment [1,11–15], have led to MSSs being implemented in many demanding applications, including bearings, molds, nuclear reactors, hydroelectric engines, and petrochemical steam and gas turbines and buckets [1,11–20]. To improve surface wear resistance while maintaining the corrosion resistance of the core, MSSs can instead be surface treated (carburized), with carbon incorporated into the sample surface at elevated temperatures to form hard carbides with alloying elements such as chromium or vanadium [1,21–23].

Highly corrosion-resistant MSSs (e.g., Cronidur 30 or XD15NW) include additions of alloying elements (and/or nitrogen) and can have poor adhesive and wear performance [24]. While not as corrosion resistant, UNS S42670 or AMS 59030B (referred to herein as P675) are relatively cost-efficient MSSs with high corrosion resistance (equivalent to 440C steel) and bulk fracture toughness (higher than M50) [25]. P675 was specifically engineered for aerospace bearing applications in advanced gas-turbine engines, where conventional bearing steels (e.g., M50 and 440C) are adversely affected by corrosion in aggressive environments and/or do not have sufficient high temperature wear performance [8]. Although P675 shows improvement in corrosion resistance relative to conventional bearing steels, higher surface hardness would lead to a longer wear lifetime in-service. Accordingly, secondary surface processing has been targeted as a way to increase the hardness and wear resistance of P675 [7,9,10,26]. Such surface treatments impart a graded microstructure that extends ~1000 μm below the metal surface. Optimized wear properties are obtained by balancing the surface hardness and core ductility of composite microstructures across the gradient region. However, the increased surface hardness typically comes at the expense of corrosion resistance, as the formation of carbides on the surface locally depletes corrosion-resistant elements (e.g., chromium, vanadium, molybdenum) from the surrounding matrix [7,20,22,23,27,28].

The corrosion performance of various P675 surface treatments has been previously assessed through accelerated DC and AC electrochemical testing in aqueous solutions [7,9,10]. These investigations provided a ranking of corrosion performance, showing that the final tempering temperature and processing atmosphere had a considerable influence on both the overall corrosion rate and damage morphology. Compared to M50, surface hardened P675 can be significantly more corrosion-resistant, and higher processing temperatures typically increased susceptibility to general corrosion damage, while lower temperatures exhibited more localized corrosion relative to untreated P675 [7]. The influence of processing on P675 wear performance for the same steels in non-corrosive wear testing has also been reported, where higher processing temperatures (HTT) yielded longer bearing lifetimes compared to low-temperature temper (LTT) [29,30]. However, there remains a need for research into the interdependency between simultaneously balancing corrosion resistance and surface hardness for bearing applications, since wear resistance (i.e., bearing performance) in corrosive environments is ultimately limited by corrosion [11].

Investigation of surface electronic properties can provide information to aid in the prediction of corrosion initiation sites [31]. Recently, scanning Kelvin probe force microscopy (SKPFM) has been used to investigate the role of nano- and micro-scale surface features on corrosion behavior [19,32–42]. Additionally, magnetic force microscopy (MFM) [43–45] has been used to similarly provide insight into

the magnetic behavior of alloy surfaces. SKPFM permits measurement with nanoscale resolution of Volta potential differences (VPDs), which are related to the electronic work function (EWF), while MFM provides information regarding the magnitude and orientation of the magnetic moments of surface domains. Likewise, in situ atomic force microscopy (AFM) has been used to monitor morphological changes during corrosion in electrolyte solution and link them to the electrochemical behavior of the material [19,46–49]. The current work presents the first application of such techniques to investigate corrosion behavior of MSS P675 with various surface treatments. Since corrosion is the most common precursor to wear damage during aero-engine operation [8], the time to onset and rate of corrosion can directly control maintenance requirements and operational costs. Initiation and propagation are critical considerations because they determine both wear behavior as well as the lifetime of the part or engine [8,50,51]. The focus of this study is to understand the effects of heat treatment processing parameters on corrosion evolution in P675 by utilizing a combination of scanning probe microscopy (SPM) techniques and accelerated corrosion testing, thereby linking surface microstructural differences (on the nanoscale) with observed macroscale surface corrosion behavior and wear performance.

#### **2. Materials and Methods**

## *2.1. Materials*

The nominal bulk composition of P675 (UNS S42670, the MSS studied here) prior to heat treatment is shown in Table 1 [29]. To increase surface hardness, P675 samples were carburized, followed by quenching and tempering, to harden the outer layer or case. Samples were cylindrical (9.5 mm diameter × 12 mm height) with post-treatment case depths of 750–1250 μm radially inward [9]. Samples differed in the final tempering temperature and carburization atmosphere: high-temperature tempering (HTT) at 496 ◦C, low-temperature tempering (LTT) at 315 ◦C, and carbo-nitrided (CN) where the case was obtained through a carburizing cycle followed by nitriding cycle during heat treating. Further details on the processing routes are discussed in previous works [9,10,29,30]. Prior to SPM characterization, samples were mechanically ground with SiC paper (to 2000 grit) in deionized (DI) water, followed by sequential polishing to 0.02 μm with a colloidal silica aqueous slurry. After polishing, samples were rinsed with ethanol and sonicated for 1 min in ethanol to remove any polishing residue.



#### *2.2. Electron Microscopy*

A field emission scanning electron microscope (SEM, FEI Teneo, Hillsboro, USA) coupled to an energy-dispersive X-ray spectrometer (EDS, 80 mm<sup>2</sup> Energy+, Oxford Instruments, Abingdon, UK) was utilized to characterize the surface microstructure and corrosion morphology of all samples, as well as construct elemental composition maps of the heat-treated surfaces. SEM analyses were conducted in both secondary electron (SE) and backscattered electron (BSE) imaging modes using 10–20 keV accelerating voltages.

#### *2.3. Scanning Probe Microscopy*

#### 2.3.1. Ex situ Scanning Probe Microscopy (SPM)

Ex situ AFM, MFM, and SKPFM were performed under an inert argon atmosphere containing <0.1 ppm H2O and O2 using a Bruker Dimension Icon AFM housed in an MBraun glovebox (MBraun, Stratham, USA). Prior to imaging, previously polished and sonicated samples were cleaned with HPLC/spectrophotometric grade ethanol (Sigma-Aldrich, 200 proof, St. Louis, USA) using lint-free

wipes (Kimtech). Following ethanol cleaning, compressed ultra-high purity nitrogen gas (Norco UHP, 99.999%) was used to dry the surface of the steel and remove any remaining surface particulates before introducing the samples into the glovebox antechamber.

Both MFM and SKPFM were performed using a dual-pass lift mode implementation in which the first pass over each scan line acquires surface topography. Upon completing the first pass, the probe then lifts off the surface to a user-defined height above the surface. This height (i.e., tip-sample separation, 100 nm in this study) remains constant throughout the second pass as the electromagnetic property of interest (i.e., Volta potential difference in the case of SKPFM or magnetic moment in the case of MFM) is measured. Surface topography was mapped using either intermittent contact (tapping) mode in the case of MFM imaging or PeakForce tapping mode (Bruker Nano, Santa Barbara, USA), which employs rapid force curve acquisition with a user-defined force setpoint (typically 2 nN here), in the case of AFM and SKPFM. In MFM, the magnetic force gradient between a magnetized Co-Cr coated AFM probe (Bruker MESP, *k* = 2.8 N/m, *f0* = 75 kHz, μ = 1 × 10−<sup>13</sup> EMU, where 1 EMU = 1 erg <sup>G</sup>−1) and the surface of the material was observed during the lift mode pass. For consistency, all MFM imaging reported herein was performed with the same MESP probe, which was magnetized immediately prior to imaging with its magnetic axis perpendicular to the sample surface. In SKPFM, the Volta potential difference (VPD) between a conductive probe (Bruker PFQNE-AL, *k* = 0.8 N/m, *f0* = 300 kHz) and the surface was quantified by application of a DC bias to null the tip-sample electric force gradient arising from the difference in Volta potential between the probe and sample surface. VPD maps were acquired utilizing frequency modulation SKPFM [31], as described in detail elsewhere [37,38]. These VPD maps were used to predict the corrosion behavior of the samples by suggesting the cathodic and anodic sites and the relative driving force for galvanic corrosion.

SKPFM was also used to observe corrosion initiation and propagation mechanisms by carrying out intermittent imaging at well-defined intervals throughout the corrosion process. While all such imaging was carried out within the controlled environment (<0.1 ppm H2O and O2) of the argon-filled glovebox, corrosion was initiated and allowed to proceed outside the glovebox, where samples were sequentially soaked for prescribed amounts of time in a 1 M NaCl solution prepared from reagen<sup>t</sup> grade NaCl (Sigma Aldrich, St. Louis, USA) and deionized (DI) water. After each time increment, samples were rinsed with DI water to remove any adhered salt, dried with UHP nitrogen, and cleaned with ultrapure ethanol using lint-free wipes. The samples were then reintroduced into the glovebox and imaged via dual-pass SKPFM. Repeated nanoscale imaging at specific recurrent locations with micron-scale positional accuracy was made possible by fiduciary marks created with a diamond tip indenter. Testing and imaging were performed ~500 μm away from the fiduciary mark to ensure results obtained were not influenced by the indent.

#### 2.3.2. In Situ SPM

To capture images of corrosion initiation and propagation in real time, in situ PeakForce tapping (topographical) AFM was also performed. In contrast to the ex situ (i.e., glovebox) SPM imaging, samples for in situ AFM imaging were mounted in a fluid cell and immersed in a 0.1 M NaCl solution under ambient atmosphere. The NaCl concentration was chosen such that it would initiate corrosion on samples at an appropriate timescale to reveal changes in topography concurrent with corrosion propagation and progression. Silicon nitride probes with a nominal tip radius of 20 nm (Bruker ScanAsyst-Fluid, *k* = 0.7 N/m, *f0* = 150 kHz) were used for repetitive imaging (0.5 Hz scan rate) of 10 × 10 μm<sup>2</sup> areas at 512 × 512 pixel resolution, corresponding to a refresh rate of ~8.5 min to capture each image. Due to differences in time between initial immersion of each sample and the initial image capturing (driven by optimization of imaging parameters), the specific timing of subsequent images is not exact between samples. The total amount of time each sample had been exposed to the corrosive salt solution was documented at both the start and end of captured images.

#### 2.3.3. Image Processing

SPM image processing and quantitative analysis were conducted using NanoScope Analysis 1.90 (Bruker). All topographical images were processed with a first order flatten filter to remove sample tip and tilt as well as any individual line-to-line offsets. The images for HTT at 116 and 135 min required a 2n<sup>d</sup> order flatten procedure to account for the deposited debris. To quantify the findings from SKPFM mapping, a threshold technique was implemented (see example image in Figure 1 below) that utilized a user-determined cut-off potential based on the distribution of Volta potentials observed in the corresponding data histogram (512 bins). From the resulting thresholded data, the average Volta potential (with corresponding standard deviation) was calculated for each of the two phases present on the surface (i.e., matrix and carbides, identities confirmed through SEM/EDS characterization) [52]. Figure 1a shows a representative SKPFM Volta potential map for HTT P675. Figure 1b shows the matrix in dark brown with the carbides (data in blue) excluded, while the light brown areas visible in Figure 1c correspond to the carbides (with the matrix excluded and indicated by the dark blue areas). Using this method, an average VPD between the matrix and carbides was calculated for each SKPFM image.

**Figure 1.** Representative 10 × 10 μm<sup>2</sup> scanning Kelvin probe microscopy (SKPFM) images of P675- high-temperature temper (HTT). Dark brown corresponds to the softer matrix phase, which is lower in height following polishing than the harder, lighter brown carbides. Images show (**a**) the original Volta potential image (600 mV full-scale range) and subsequent implementation of thresholding cutoffs (blue) to calculate average Volta potential differences (VPDs) for the (**b**) matrix and (**c**) carbides.

#### *2.4. Electrochemical Corrosion Testing*

Electrochemical cyclic polarization testing was used to characterize corrosion behavior for each type of heat-treated steel. Sample preparation details can be found in a previous publication, thus the sample testing area was defined by masking off the sample such that only a circular area (diameter ~6.6 mm) test area was in contact with the electrolyte solution [7]. Testing was conducted in 0.01 M NaCl electrolyte solution with a potentiostat (SP-300, Bio-Logic, Seyssinet-Pariset, France) used to control and monitor a three-electrode system in a modified flat cell. A saturated calomel electrode (SCE) served as the reference electrode and a platinum mesh as the counter electrode. Following sample immersion, open circuit potential (OCP) was monitored for 30 min. The sample was then polarized at a scan rate of 0.5 mV/s from 100 mV below OCP to 600 mV above OCP or when pitting had stabilized, followed by a reverse scan back to OCP.
