Severe Plastic Deformed Zones and White Etching Layers Formed During Service of Railway Wheels

The near-surface regions of rail wheels experience a complex thermo-mechanical loading, which varies along the location of the tread and causes severe changes to the microstructure. Occasionally, brittle white etching layers (WEL) are formed, representing a high risk of wheel damage. Therefore, we studied the depth-evolution of the surface-near microstructure along different regions of the tread of a wheel being ~ 200,000 km in service. The microstructural variations point toward a higher thermal loading history in the middle of the tread and predominant mechanical loading in the neighboring regions, where a severe plastic deformed microstructure with rolling contact fatigue cracks is present. Varieties of WEL-like microstructures were investigated, the presented analysis contributes to the knowledge of their microstructural characteristics and outpoints these regions in terms of potential crack initiation sites on railway wheels.


Introduction
Medium carbon steels are commonly used for railway wheels due to a good balance of hardness and fracture toughness, as well as solid wear properties. Among rail wheel steels, the grade ER7 is one of the most used grades on freight trains in Europe. During service, the contact surface is affected by a complex mechanical loading situation in terms of vertical, tangential, and longitudinal loads [1]. High shear stresses can cause rolling contact fatigue (RCF) damage [2][3][4][5], and cracks propagating into the material can lead to catastrophic failure [6]. Besides the mechanical loads, the wheel tread surface is additionally affected by high thermal loads due to frictional heating during curving, braking, short slippage events when acceleration, or even occasional full slippage events. Temperatures can exceed 500 °C in slip zones on wheel tread surfaces [7] leading to a massive influence on microstructure and mechanical properties [8,9]. Even though modern trains are equipped with wheel slide protection systems [10,11] thermally induced damage is still an issue [10,12,13]. Additionally important are mechanical loads, which in combination with temperature heavily affect the near-surface microstructure [9,[14][15][16]. Besides the loading history, the microstructural evolution also varies on the location along the wheel tread surface [17,18].
The formation of a white etching layer (WEL) can initiate cracks and promote their propagation [19][20][21][22][23][24][25][26][27], since the evolved microstructure of the WEL comes up with higher hardness and lower fracture toughness. Most studies are performed for railheads, but the comparable materials and rolling-sliding contact situation may allow transferring the gained know-how about the relevant mechanisms to railway wheels as well. The formation process is still under discussion, some studies predict heating above austenitization temperature followed by rapid cooling [21,22,25,28] is the major cause, while others claim heavy plastic deformation [23,24,29], and further ones their combination [30,31]. However, WEL can form in several variations-a common and consistent classification is still missing-on railway wheel tread surfaces [13,[32][33][34]. Recently, the terms brown etching layer (BEL) and stratified surface layer (SSL) are introduced [35][36][37][38], outpointing the diversity of the evolved microstructures.

3
The different variations are further summarized as WELlike microstructures within this work.
The aim of this study is to study the evolved microstructure on railway wheels in greater detail, since this is the initial microstructural state before a possible formation of WEL-like microstructures (WEL, BEL, SSL). For this, the three main zones along a railway wheel tread after service are investigated. We concentrated on the microstructural evolution in transversal and rolling directions-in different depths from the surface-using light optical microscopy (LOM) as well as high magnification scanning electron microscopy (SEM) and electron backscatter diffraction (EBSD) measurements. Our studies also reveal different variations of WEL-like microstructures, characterized by different microstructures and hardness.

Experimental
A railway wheel of the wheel steel grade ER7 (EN13262 [39]) with a diameter of 0.95 m and after nearly 200,000 km in service was provided by the Austrian Federal Railways. Table 1 summarizes the chemical composition of this material, which typically shows an upper yield strength of R eH ≥ 520 MPa, a tensile strength R m between 820 and 940 MPa, and an elongation at break A 5 of ≥ 14%. Brinell hardness is above 235 HB and a fracture toughness between 40 and 80 MPa m 1/2 is standardized [40], when treated and tested according to EN13262.
The samples for our investigations were prepared with a laboratory cutting device (Struers Secotom-50; Struers ApS, Denmark) out of three different zones of the railway wheel. These zones are named according to [41] as zone 1 (being taken from the peripheral tread region, at the field side), zone 2 (taken from the tread surface close to the wheel flange), and the middle zone 3 (major wheel tread) (Fig. 1a).
For each zone, multiple cross sections in transversal direction (CS-TD) and in rolling direction (CS-RD) are analyzed (Fig. 1b).
Samples taken from these zones are embedded in conductive compounds and prepared by coarse grinding, 1 μm diamond fine polishing, and etching with ethanolic nitric acid (3% HNO3, 97% ethanol). The microstructure is studied LOM (Axio Imager M2m, Carl Zeiss AG, Germany) and SEM (Jeol JIB 4700F, Jeol Ltd., Japan) equipped with a Schottky field emission gun, secondary and backscattered electron detectors, and an electron backscatter diffraction (EBSD) detector (Bruker e-Flash HR, USA). SEM investigations were performed at 15 kV acceleration voltage. To gain optimal results at the EBSD measurements, additional finish polishing for 10 min with colloidal Silica (< 0.25 μm) was carried out. EBSD measurements were performed in the SEM with optimized electron beam conditions at 15 kV acceleration voltage and a probe current of 3.6 nA. The working distance of the SEM was set to 16 mm, and the EBSD detector distance to 18 mm. Kikuchi patterns were acquired at 200 × 150 px with an exposure time of 20 ms. Setup and analysis were performed via the Bruker Esprit 2.2 software package, whereas the American Mineralogist Crystal Structure Database (AMCSD) database was used for phase identification. The inverse pole figure (IPF) maps, the grain average misorientation (GAM) maps, and the grain boundaries (GB) maps are presented in this work to describe the evolved microstructure in more detail. The IPF map is used to describe the grain structure in general, possible preferred orientations, and the grain size distribution. The GAM maps show the calculated misorientation between each neighboring pair of points within the grain and therefore give information about the orientation changes inside the grains. The GB maps are useful to evaluate the fraction of small angle grain boundaries, defined in this study with GB angles up to 10°, and large angle grain boundaries (> 10°).  Hardness was investigated by low load Vickers hardness depth profiles (load of 0.05 kp (0.49 N)) indenting the cross sections at increasing distances from the surface with a Future-Tech FM-700 hardness tester. The diagonals of the indents are measured with a LOM. Furthermore, nanoindentation measurements of individual regions within the samples were performed with a Bruker Hysitron Triboindenter TI980-Performech II (equipped with a Berkovich diamond tip) to characterize their indentation hardness as well as moduli. The load-displacement curves (with a peak load of 5 mN) were evaluated and the hardness was deduced via the load/area ratio, whereby the reduced Young's modulus was evaluated via the Oliver and Pharr method [42].

Results
Out of each zone, several cross sections in transversal direction and rolling direction were prepared and analyzed. The presented images indicate that there is no preferred rolling direction, this is because this study is based on a railway wheel from the field where the running direction was changing over the service.

Characteristic Microstructural Evolution in Different Wheel Tread Zones
In the peripheral tread region (field side) of the wheel, zone 1, the cross section CS-TD (Fig. 2a) shows numerous cracks with depths of ~ 100 µm propagating in a severely plastically deformed (SPD) microstructure, which expands down to ~ 300 µm from the surface. Some of the cracks are wide opened and/or changing crack growth direction leading to multiple spalling of the material, hence, resulting in a rough surface. In a depth of 300-600 µm from the surface the microstructure gradually changes from a finegrained SPD microstructure with random orientation to a ferritic-pearlitic microstructure (with elongated grains in rolling direction, aligned with about 30-45° to the surface). The elongation, alignment, and grain size decrease with increasing distance from the surface, featuring a ferriticpearlitic bulk microstructure without preferred orientation upon ~ 600 µm from surface.
For the cross section CS-RD in zone 1 ( Fig. 2b) a rough surface can be seen as described for CS-TD due to numerous spallation. But this cross section in rolling direction additionally shows long cracks propagating almost parallel to the tread surface in depths of about 100 µm. These cracks are branching or merging at certain stages, which leads to large spallation. The evolution of the sub-surface microstructure is in general comparable to that shown for the cross section CS-TD: A gradual change from a fine-grained and randomly orientated SPD microstructure at ~ 300 µm from surface to a ferritic-pearlitic microstructure. The latter is aligned parallel to the surface with increasing grain size from 300 to 600 µm with increasing distance from the surface. At ≥ 600 µm below the surface, the original bulk microstructure is present.
In the middle of the wheel tread, designated as zone 3, the cross section CS-TD (Fig. 2c) shows that the finegrained microstructure reaches deeper into the material than for zone 1, although, the depth is varying widely between 300 and 700 µm from the surface. Below that region, the microstructure changes gradually to an aligned SPD microstructure with decreasing deformation degree with increasing depth. Within zone 3, the thickness of the influenced microstructure varies widely within a maximum of ~ 1,000 µm. Also in rolling direction, the microstructure is comparable (Fig. 2d). Here, a formation of a thin WEL can be detected with a thickness of about 30 µm.
The cross-sectional LOM image of the region close to the wheel flange (zone 2, Fig. 2e) clearly shows massive cracks, which mainly originate from regions where a WEL is present. These cracks typically propagate down to depths of ~ 600 µm with an angle of about 45° from surface. Some cracks branch within the WEL and/or grow along the interface between WEL and underlying material, leading to spallation. A fine-grained microstructure is present down to almost 900 µm from surface. Below this uniform microstructure, there is immediately the original microstructure.
Microstructural investigations of the material in rolling direction (Fig. 2f) exhibit similar features as perpendicular to it. Below the fine-grained microstructure, there is directly the original microstructure, without the presence of an aligned SPD microstructure as within the other two zones.
Detailed investigations of zone 1 (Fig. 2a) by higher magnification SEM and EBSD measurements on cross sections perpendicular to the rolling direction in different depths from the surface are presented in Fig. 3. At a depth of 50 µm (Fig. 3a) a fragmentation (and possible dissolution) of the cementite is evident and the spherical cementite particles are randomly distributed. At a depth of 100 µm (Fig. 3b) and 150 µm (Fig. 3c) the microstructural features are comparable, with a slightly less dense arrangement. For the regions at distances of ~ 150 µm (Fig. 3c), 300 µm (Fig. 3d), 600 µm (Fig. 3e), and 1500 µm (Fig. 3f) below the surface detailed EBSD measurements were conducted and evaluated for their IPFs, GAM, GBs. The IPF map for the region at ~ 150 µm from the surface shows a fine-grained microstructure with slightly elongated grains and no preferred orientation. The fraction of large angle grain boundaries (with misorientation angles between adjacent grains of ≥ 10°) is relatively high (Fig. 3c). At a depth of 300 µm, the SEM investigations suggest for larger grains (Fig. 3d), bent and broken cementite lamellae (next to spheroidized particles), and proeutectoid ferrite. EBSD investigations of this region prove the larger grains, which seem to be elongated along the rolling direction (Fig. 3h). The GAM map analysis shows a wider GAM angle distribution than the region closer to the surface.
At a depth of 600 µm from the surface, the microstructure is comparable to the original condition with clearly present pearlite and ferrite grains (Fig. 3e). The original microstructure is essentially present at a distance of 1500 µm from the surface (Fig. 3f). The GAM analysis of the EBSD measurements (Fig. 3i) show that the highest misorientation angles are at the lamellar structures of pearlite, and the respective GB analysis yield only a fraction of 25-18% of small angle GB (angle below 10°).
Detailed analysis of zone 3 (Fig. 2c) by SEM (Fig. 4) show spheroidized cementite (Fig. 4a) down to 100 µm (Fig. 4b) and even 150 µm from the surface (Fig. 4c). With increasing distance from the surface, the overall microstructure gradually changes toward the original one (Fig. 4d, e, and f). EBSD analyzes of the region 150 µm below the surface (Fig. 4g) show a fine-grained microstructure with slightly elongated grains. The GAM angle map pictures low misorientation angles and the fraction of large angle grain boundaries (angles > 10°) is high with ~ 86%. At a distance of 300 µm from the surface (Fig. 4h), the microstructure is still very similar with slightly more pronounced elongated grains (represented by the IPF map) and a smaller fraction of large angle grain boundaries.
With further distance from the surface, (300 µm, Fig. 4d) the fraction of spheroidized cementite decreases and that of cementite lamellae increases, suggesting less broken and annealed cementite lamellae upon this depth. However, even at 600 µm from the surface (Fig. 4e), spheroidized particles are present. The corresponding IPF map (Fig. 4i) shows a mixture of elongated grains (orientated in rolling direction) and polygonal grains. At this distance from the surface there are more areas with larger GAM angles than closer to the surface and the fraction of small angle grain boundaries (< 10° misorientation) increased to almost 17%. Contrary to zone 1, here, even at a depth of 1500 µm (Fig. 4f), broken and bent cementite lamellae are present within the pearlite grains. Further, the EBSD investigations do not represent an unaffected bulk microstructure like in zone 1 at this depth. Clearly, a sub-grain microstructure is present.
Detailed SEM investigations of zone 2 (Fig. 2e), 50 µm below the surface perpendicular to the rolling direction, show fine globular particles (Fig. 5a). Although generally, the overall microstructure is similar to that of zone 1 and zone 3 at this depth (compare with Fig. 3 and Fig. 4), still some cementite lamellae are present here. At distances of 100 µm (Fig. 5b) and 150 µm (Fig. 5c) from the surface, the overall microstructure is very similar to that at 50 µm, with tendencies of an increased fraction of cementite lamellar. This tendency continues with increasing distance from the surface (300 µm, Fig. 5d), and at a distance of 600 µm (Fig. 5e) the grains are already more globular with cementite lamellae fragments and proeutectoid ferrite. At a distance of 1500 µm the original pearlitic microstructure is present (Fig. 5f).
EBSD measurements clearly show elongated grains aligned parallel to the rolling direction at a distance of 150 µm from the surface (Fig. 5g). Compared with the other zones, the grain size is larger. With increasing distance ( Fig. 5h and i), the grain size further increases while 1 3   still being aligned in rolling direction. Areas with larger GAM angles increase and the fraction of large angle grain boundaries decreases. Even at a distance of 1500 µm from the surface, the EBSD investigations (Fig. 5j) hint toward a modified microstructure with elongated ferrite and pearlite grains aligned in rolling direction.
Microhardness is determined by Vickers hardness measurements, the results show an increase within 200 µm from surface for zone 1 and zone 2, in contrast to no significant change in zone 3 (Fig. 6a). Below ~ 300 µm from surface the hardness values are constant in each zone, differences are supposed to be evident due to the thermal treatment when manufacturing the wheel. An attempt to evaluate nanoindentation hardness (Fig. 6b) is done at various depths from surface, referring to the depths of the detailed SEM investigations. An increase in hardness can be seen within the outermost 100 µm from surface in all zones, however, scattering is very pronounced.

Various WEL-Like Microstructures on the Wheels
Various WEL-like microstructures are identified by etching the cross-sectional samples, with variations in appearance (bright shining to brownish), thicknesses (20-600 µm) as well as microstructure (Fig. 7).
The white appearing layer pictured in Fig. 7a is present on the tread surface in the region between zone 2 and 3. It comes up with a thickness of ~ 30 µm, formed top of an SPD aligned microstructure and numerous cracks origsnate within this layer (defined as WEL1 in this work). Figure 7b shows different WEL-like microstructures, more precisely a stratification of different layers. A large brownish appearing region extends down to 500 µm below the surface of zone 3. On top of this brownish region (defined as BEL) a brighter, more homogenous and thinner layer is present, which is defined as WEL2. The combination can be designated as SSL. Some cracks stop at the interface of WEL2 and BEL, but at the margin of the SSL a crack propagates through both layers, arresting at the interface to the underlying microstructure. In some areas, the underlying material (below the SSL) is an SPD microstructure and in others an almost original coarse ferritic-pearlitic microstructure.
Detailed high-magnification SEM investigations of selected regions highlight the typical microstructure of the formed WEL-like microstructures ( Fig. 7a and b). The thin and homogenous WEL1 has a uniform microstructure (Fig. 7c). Further increasing the magnification highlights a fine mesh-like structure with some evenly distributed submicron globular particles < 100 nm (Fig. 7d). The underlying SPD region exhibits a coarser and more aligned microstructure ( Fig. 7e and f), corresponding to that presented earlier for a distance of ~ 50 µm below the wheel tread surface.
In general, WEL2 exhibits a comparable microstructure to WEL1 but the mesh-like structure tends to be demolished, compare Fig. 7c and d (WEL2) with Fig. 7g and h. Moreover, differences can be seen regarding hardness, where WEL1 reveals the highest values of 6.09 ± 0.43 GPa followed by WEL2 with 4.61 ± 0.46 GPa. The high-magnification SEM micrographs of the BEL (Fig. 7i and j) show a dissolved microstructure with spheroidized cementite and the hardness significantly decreased to 3.57 ± 0.50 GPa.

Discussion
When comparing individual studies about wear and fatigue phenomena of railway wheels, it is of utmost importance to know the loading history and whether there is a preferred Fig. 6 Low-load Vickers hardness measurements (a) and nanoindentation hardness (b) graphs for the different zones rolling direction, or this changed periodically throughout the service like for the wheel studied in the current work. Whereas most of the studies focus on a certain zone along the railway wheel tread [43][44][45], we compared the microstructural evolution in the near-surface regions for different locations of the railway wheel tread. A corresponding study was conducted by Molyneux-Berry et al. [17] for a railway wheel being in service for ~ 280,000 km. For the region corresponding to our zone 1 (named region E in [17]) they report that there are few contacts (with the rail) but high tangential forces toward the middle of the tread, causing rolling contact fatigue (RCF) cracking. Our microstructural results are comparable ( Fig. 2a and b, Fig. 3) exhibiting an about 300-500 µm deep deformed region with numerous small RCF cracks along the surface and some long but shallow cracks propagating toward the middle of the railway wheel tread. For this middle region, moderate normal stresses and circumferential tangential forces are present [17]. Due to various characteristics (e.g., traction, braking), the peak loading tends to be in the near sub-surface region. Corresponding to the results reported in [17], our microstructural studies indicate a weaker line structure in transversal direction of this middle region (Fig. 2d) than for the cross section in rolling direction (Fig. 2c, Fig. 4). This-low shear strain values with minor deformed microstructure and practically featureless appearance for the middle sector of the railway wheel treat-is also reported by Cvetkovski and Ahlstrom [18]. Important to mention is also that for this middle region-being the main contact zone between wheel and rail-traction and breaking will lead to significant temperature influence on the railway wheel [17]. This thermal loading is sufficient to initiate spheroidization of broken cementite lamellae (due to the mechanical loading) [13,16,46], (Fig. 4a and b). Nikas et al. [9] studied different railway wheel materials and found comparable microstructures in this middle region of the railway wheel treat.
For the region of our zone 2, Molyneux-Berry et al. [17] proposed that significant normal and tangential stresses lead to the formation of fine surface cracks, which tend to propagate not deep into the material. While in general, our microstructural observations are similar for this region, we do observe also cracks that propagate almost 500 µm deep into the material (Fig. 2e). This might be affected by the formation of a different WEL-like microstructures, which easily results in a different fracture toughness [1,2,4,6,7,19,20,47,48].
Corresponding detailed EBSD studies at different regions of the railway wheel to ours are so far mainly conducted for wheel materials loaded via a twin-disc setup [33,[49][50][51][52][53][54][55][56][57][58][59][60]. These lab-tests do have their advantages as the loading conditions are controlled, but a direct comparison of the thereby obtained microstructural changes with those present in railway wheels after service is difficult. Nevertheless, the deformation of ferrite and cementite lamella, their fragmentation and the subsequent formation of a sub-grain structure and fine-grained microstructure-as reported for materials tested with the twin disc setup [51] -was also observed for our wheel after service (Figs. 3, 4 and 5). Differences for our ex-service wheel are that in zone 1 an SPD microstructure with high GAM angles is present, resulting from high tangential forces [17] and that the fraction of small angle GB is higher for the measurement depth of 150 µm (Fig. 3g). Also for the other zones, 2 (Figs3, 4g-j) and 3 (Fig. 5 g-j), the fraction of large angle GB is highest directly below the surface. The GAM is lowest in the middle region of the railway wheel treat (zone 3), where the highest thermal loading is present as discussed above, promoting dynamic recrystallization [52] and spheroidization of fragmented cementite lamellae. This leads to no significant hardness increase within zone 3 (Fig. 6a), in contrast to an increase for the other zones. Some studies measured an increase in hardness in the middle zone of an used rail wheel with values up to 300-400 HV1 [17,61], however, the wheel grade and the loading history are different. In general, plastic deformed regions and fine-grained layers near the surface come up with hardness increases [51,57] but due to local inhomogeneities of rail wheels from service the variation is large. This is also the issue for the presented nanoindentation measurement results (Fig. 6b), which were performed in the same depths as the detailed SEM investigations. A trend toward a hardness increase up to 5 GPa can be seen for all zones, but especially below 300 µm the coarse microstructure led to a large scattering of the nanoindentation results.
Our results clearly show the variety of WEL-like microstructures, whereby its formation depends on the different loading conditions present at the different zones of the railway wheel. Further, the initial local microstructure along the wheel tread potentially influences the formation [28]. Most characteristically the WEL-like microstructures lead to different colors during LOM investigations (Fig. 7a,  b) and exhibit a different microstructure (during SEM investigations), as well as mechanical properties. The brightest WEL (WEL1 Fig. 7a, c, d) is thin and formed on top of an otherwise heavily deformed material, suggesting that short but intense slipping events are primarily responsible for their formation. A comparable microstructure and general appearance are reported by Handa et al. [11] for a wheel from service, and by Zhu et al. [50] for a wheel tested via a twin-disc setup with a 25% slip ratio. The large brownish (during LOM) appearing area (named here BEL, Fig. 7h) suggests a massive thermal influence (like present during heavy braking) because microstructural investigations after brake dynameter experiments yield a comparable result [11]. The microstructure is almost featureless with larger spheroidized cementite particles, which can be ascribed as BEL or tempered WEL, discussed in detail by Kumar et al. [35] for rails. Corresponding variations of WELlike microstructures at the same location along the tread wheel surface are also reported by others [13,[32][33][34], but a different classification is still missing and most likely not meaningful due to the complex loading situation during service.

Conclusions
The near-surface regions of a railway wheel after almost 200,000 km in service exhibit huge variations in their microstructural evolution from the field side of the wheel treat to its flange, due to variations in loading situations.
The field side (named here zone 1) reveals numerous RCF cracks, indicative of high tangential forces toward the middle of the tread. A severe plastic deformed microstructure extends down to a depth of ~ 300 µm from the surface, with cementite fragmentation and even spheroidization. EBSD investigations show that the microstructure exhibits signs of deformation even down to 600 µm and a relatively high grain average misorientation. The corresponding hardness variation-which increases from the core value to the surface by only ~ 8%-is less pronounced than for the other zones along the surface tread. The microstructural changes in the region next to the wheel flange (named here zone 2) extend deeper into the material than for the field side. The microstructural features-like more pronounced elongation in rolling direction as well as alignment ~ 45° to the tread surface-indicate a comparably lower thermal loading history. The hardness at the surface is ~ 15% above the core value. The region between these, being the major tread surface (named here as zone 3), exhibits a microstructure with spheroidized cementite down to deeper regions as for the field side. This suggests that the thermal loading history was more severe. Furthermore, the region is characteristic of its fine-grained microstructure with minor preferred alignment, low misorientation, and a higher fraction of large angle GBs near the surface.
In addition to these different microstructural evolutions and characteristics, variations of WEL-like microstructures are present, acting as potential crack initiation sites. The thin (~ 3 µm) and brightest appearing (after metallographic sample preparation) WEL (classified as WEL1) reveals a fine mesh-like structure and the highest hardness of 6.09 ± 0.43 GPa. The overall morphology of this WEL suggests a minor loading since formation, hence being relatively "young". This is in contrast to the other variation identified (classified as WEL2), which is up to 50 µm thick and exhibits a more flawed mesh-like microstructure suggesting severe loading since formation. This WEL appears more brownish-like during light optical microscopy and has a lower hardness with 4.61 ± 0.46 GPa. The third majorly different variation (classified as BEL) is softest (3.57 ± 0.50 GPa) and 500 µm thick. Its microstructure shows small spheroidized cementite particles and suggests even for their dissolution; hence, the ferrite would be supersaturated in C. Several reports indicate that such a BEL resulted from severe thermal loads generated during hard wheel braking and massive slippage events. The presented analysis shows the diversity of WEL-like microstructures in the field and can contribute to the knowledge of potential fatigue crack initiation sites. The highest tendency for crack-formation was found for the WEL1 and in its vicinity. Therefore, we classify WEL1 as more dangerous than WEL2 or BEL in terms of cracking risk. Knowledge of microstructural characteristics and mechanical properties, especially crack initiation and fracture behavior, is crucial for further improvement of railroad safety.
Acknowledgements This project has received funding from the Shift2Rail Joint Undertaking (JU) under grant agreement No 826255 and No 101012456. The JU receives support from the European Union's Horizon 2020 research and innovation programme and the Shift2Rail JU members other than the Union. In addition, this work was funded by the "Austrian COMET-Program" (project InTribology1, no. 872176) via the Austrian Research Promotion Agency (FFG) and the federal states of Niederösterreich and Vorarlberg and was carried out within the "Excellence Center of Tribology" (AC2T research GmbH). The authors acknowledge the TU Wien Bibliothek for financial support through its Open Access Funding Program. We thank DI Roman Schmid and Andreas Schachner from the Austrian Federal Railways Holding Stock Company for providing ex-service material and their help in this work. We are also thankful to Iñigo Sarries (CAF, Construcciones y Auxiliar de Ferrocarriles) for useful discussion and the supply of rail wheel materials.
Funding Open access funding provided by TU Wien (TUW).

Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.