Importance of surface oxide for the tribology of a Zr-based metallic glass

Thermally grown surface oxide layers dominate the single-asperity tribological behavior of a Zr60Cu30Al10 glass. Increase in oxidation time leads to an increased contribution of shearing and a corresponding decreased contribution of ploughing to friction. This change in the dominating friction and wear mechanism results in an overall minor decrease of the friction coefficient of oxidized surfaces compared to the metallic glass sample with native surface oxide. Our results demonstrate the importance of creating a stable oxide layer for practical applications of metallic glasses in micro-devices involving sliding contact.


Introduction
Metallic glasses are structurally amorphous metallic alloys whose compositions are designed to prevent crystallization upon cooling from the melt [1]. Owing to their low elastic moduli high hardness metallic glasses have been reported to provide low friction and low wear rates [2][3][4]. Studies comparing the macro-scale wear behavior of metallic glasses and their crystalline counter-parts or steel have reported ambiguous results. Kwon et al. [5] compared the wear behaviour of a Zr-based bulk metallic glass with hardened steel. While the friction coefficients of both materials were similar, the authors reported a significantly lower wear resistance for the Zr-based bulk metallic glass and discussed the difference of underlying wear mechanisms: brittle wear in the case of the Zr-based bulk metallic glass and adhesive wear in the case of the hardened steel. Tam and Shek reported a higher friction coefficient for metallic glasses than for their crystalline counterparts despite their higher mechanical strength and hardness [6], while Yoon et al. [7] observed an increase in hardness upon crystallization of a Ni-based metallic glass but also an increase in the friction coefficient and of the wear rate. No unique correlation between the hardness and the tribological behaviour of metallic glasses has been established so far. At the macro-scale the tribological mechanisms are complex and involve a combination of plasticity, structural relaxation and transformation [8], and material transfer [9], or mixing [10,11] with subsequent chemical reactions and structural changes.
Meanwhile, metallic glasses have been recognized as potential materials for micro-electromechanical applications, such as micro-gears and media-storage devices [12,13]. This potential owes to their excellent thermoplastic formability down to the nm-scale within their undercooled liquid region [14]. For such applications, the tribological performance of metallic glasses at the nanometer scale is of crucial importance. Recently, some of the authors have investigated the nanoscopic friction and wear mechanisms of a Pt-based metallic glass surface in UHV after its native oxide removal in comparison to a Pt(111) surface [15]. We found that the wear-less friction coefficient was barely measurable on both surfaces with an inert diamond counter-body, while it significantly increased with a SiO x counter-body. This observation demonstrates the effect of surface chemistry on tribological processes that may involve interfacial alloying. At the nanometer scale also, native and grown oxide surface films play an increasing role. A recent study has shown that the oxide layer on a Ni-based metallic glass enhances the wear resistance [16]. The tribological response of oxides is not unique and has been reported to differ depending on the bonding strength of oxygen with the metallic elements: a weaker bond was found to correspond to a higher friction coefficient [17].
In this work we use TEM, AFM, and XPS to characterize the structure and properties of oxide layers grown on amorphous Zr 60 Cu 30 Al 10 ribbons by annealing in air at 504 K for 30 min and 60 min and we report and discuss the load dependence of friction for oxide layers of different thicknesses.

Experimental methods and materials preparation
A Zr 60 Cu 30 Al 10 master alloy was prepared by melting elemental Zr, Cu, and Al with a purity > 99.9 wt.% in the corresponding weight proportion by arc-melting in Zr-gettered Ar-atmosphere. The melting process was repeated three times to ensure homogeneous mixing of the elements. Subsequently, 0.2 mm thick Zr 60 Cu 30 Al 10 metallic glassy ribbons were prepared by melt-spinning on a water-cooled Cu-wheel, using Ar as an injection gas. The amorphous structure of the as-spun ribbons was confirmed by X-ray diffraction (XRD) with K a radiation and differential scanning calorimetry (DSC). Further samples from the same batch have been oxidized in air at T = 504 K for 30 min and 60 min. The temperature was selected well below the glass transition on-set temperature T g, on-set = 673 K in order to avoid structural relaxation of the metallic glass [18,19]. The surface composition and oxide layer thickness on as-spun and oxidized samples have been determined by X-ray photoemission spectroscopy (XPS) using a K-Alpha + XPS System, manufactured by ThermoFisher Scientific. In particular, the composition was determined from the peak area of the different elemental spectra and their bonding states were determined from the peak energy-shifts. The effect of oxidation treatment in air on the surface topography was determined by tapping-mode AFM using a NanoWizard 3 Ultra AFM manufactured by JPK Instruments, Germany and a single crystalline Si cantilever (PPP-NCLR, NanoSensors, Germany). AFM scratching tests on as-spun and oxidized samples were performed at room temperature and in ambient conditions with a diamond-coated silicon cantilever (CDT-NCLR, NanoSensors, Switzerland). The normal and lateral spring constant of the cantilever, N k and L k , were determined from the geometrical beam model according to , where E is Young's modulus, G is the shear modulus. The width w, the length l, and the tip height h were measured optically. The thickness of the cantilever was determined from the first bending resonance frequency of the cantilever 0 f according to  where  is the mass density. The sensitivity of the photodetector S was determined by recording forcedistance curves, in which the slope of the repulsive part corresponds to 1/S. The normal and lateral forces were calculated from the vertical and lateral photodiode voltages by assuming the same photodiode sensitivity in both directions according to N , where V N and V L are the vertical and lateral voltages of the photodiode, respectively. For the diamond-coated cantilever used in this work we found -1 48 Nm Single cycles of reciprocal scratch tests perpendicular to the cantilever axis were performed over a length l s = 300 nm at normal force F N = 500-7,000 nN, during which F L was recorded. Wear scratches were imaged by tapping mode AFM to determine their width. The friction force was calculated according to f where F L,trace/retrace are the lateral forces recorded during the forward and backward sliding motion of the cantilever during reciprocal scratching, respectively. For all measurements, the sliding velocity was 600 nm/s v  .
All results presented in this report are recorded with the same diamond-coated cantilever to directly compare the tribological performance of the different sample surfaces. All AFM-measurements were repeated twice with a different tip for each comparative series of measurements on the different samples. Given geometrical difference of the AFM-tips minor differences in the numerical values of the friction forces and the width of the scratches were observed but did not significantly affect the overall trend and the values of the friction coefficient and of the scratch hardness. This number of measurement repetition does not allow for statistical analysis. However, the authors would like to stress out the representative character of the measurements presented here. The significance of AFM-measurements relies on the idealization of a tribological contact between a smooth surface and a single asperity, i.e., the AFM-tip. For a given type of cantilevers, the geometry of the tips can be treated as equivalent, thanks to optimized production processes and controls of the manufacturers.

Samples characterization
The chemical composition of the oxide layers on the as-spun and the annealed samples is listed in Table 1.
The bonding state of the elements was determined from the peak energy-shift. Correspondingly, Table 1 also indicates the molecular fraction of the respective oxide compounds. We find that the oxide layer mostly consists of a ZrO 2 and Cu 2 O. After oxidation treatment, an increase in oxygen concentration was measured. This increase in oxygen concentration is attributed to a departure from stochiometry for ZrO 2 and Cu 2 O. Both these oxides have a wide concentration range for oxygen. Increasing the oxidation time resulted in an increase in the Cu 2 O molar fraction. Recently, Louzguine-Luzgin et al. have investigated the structure and growth rate of the native oxide on a Cu 47 Zr 45 Al 8 bulk metallic glass [22]. They found that after several days of exposition to ambient air the oxide layer mostly consisted of Zr 2 O and Cu 2 O, and a minor amount of Al 2 O 3 . The chemical composition of the metallic glass investigated in Ref. [22] is comparable to the one of the metallic glass investigated in this work and we expectedly find our results in good agreement with the results by Louzguine-Luzgin et al. Figure 1 shows high-resolution TEM images on the cross section of a sample oxidized for 90 min. Both high resolution TEM images show that the structure of the underlying metallic alloy was amorphous after the oxidation treatment. However, crystallized regions Zr-rich amorphous oxide matrix. There, the authors identified the nanocrystalline particles based on the spacing between the lattice fringes. For selected particles, they measured 213 pm, corresponding to the interplanar distance of {200}-planes in Cu 2 O. In this work though, the contrast of the TEM image is to shallow to measure the spacing of the lattice fringe with accuracy. This is partly due to the fact that the TEM sample was too thick. Thus, the lattice fringes of the crystallites are overlaid with the amorphous structure of the matrix. Figure 2 shows topography images recorded by tapping-mode AFM. The RMS roughness was between 50 pm and 80 pm. The results confirm the impressive flatness of melt-spun metallic glass ribbons, whose roughness is almost at the limit of the sensitivity of AFM experiments. It is found that the oxidation treatments did not affect the surface roughness of the samples.  Figure 3 shows topography images of the as-spun Zr 60 Cu 30 Al 10 sample after scratch tests with normal forces ranging from F N = 0. 9

Results
, where A f is the projected area between the submerged part of the indenter and the scratch front [23] and  s is the friction coefficient associated to  effect is more pronounced for the as-spun sample and less pronounced with increasing oxidation time. The friction force measured on the as-spun sample was slightly higher than on the oxidized samples over the whole range of normal forces, while it was very similar on both oxidized surfaces. The respective friction coefficient for shearing and ploughing, s  and p  , were determined as described above and are shown in Fig. 4(b). The values of the shearing friction coefficient s  increase from 0.177 for the as-spun  increasing oxidation time, the contribution of ploughing to friction decreases. This observation can be explained by considering that in the case of ploughing friction forces arise from the resistance of the surface material to displacement ahead and aside of the indenter. As the hardness of a material describes its resistance to plastic flow, the effect of ploughing on friction becomes smaller for a softer surface. Moreover, the increased contribution of shearing to friction with oxidation time indicates an increased adhesion and possible material transfer.

Discussion
The XPS and TEM results show that thickness of the oxide layer on Zr 60 Cu 30 Al 10 increased up to 10 nm upon annealing in air at 504 K for 90 min. For all samples, the surface oxide was found to mostly consist in a mixture of Zr and Cu oxides. From the topography images shown in Fig. 2 with roughness lower than 1 nm we conclude that the effect of the oxidation treatment on the geometrical surface parameters can be neglected. While the effect of the surface oxidation treatment on the friction coefficient was small, our results demonstrate that the presence of a thermally grown surface oxide layer changes the governing mechanism for friction and wear. After growth of a surface oxide the contribution of ploughing to the friction coefficient decreased from 58% for the as-spun sample to 25% and 16% for the samples oxidized for 30 min and 60 min, respectively. In the later cases wear was dominated by shearing. It is interesting to note that the change in the governing mechanism for friction and wear was accompanied by a surface softening as illustrated in Fig. 5(d) by the decrease in H s . Based on these observations we attribute the ploughing contribution to friction and wear to the plastic deformation of the metallic glass below the native or thermally grown surface oxide layer. On the other hand, shearing is attributed to adhesion of the surface oxide to the indenter. This is in line with the observation that the pile-ups on the sides and the ends of the scratches shown in Figs. 3(c) and 3(d) are much less prominent than those observed in our previous study on metallic glass surfaces in UHV conditions after oxide removal [16]. In a recent work, Louzguine-Luzgin et al. characterized the nano-scale wear of three different metallic glasses with a native surface oxide layer [24]. The authors estimated the wear rate from the scratched volume normalized by the length of the scratch performed with a sharper single crystalline diamond tip than the polycrystalline diamond coated tip used in this work. Owing to the sharpness of the tip used in Ref. [24] wear set at lower load values than in this work and the ploughing contribution is expected to be higher. Nevertheless, it is interesting to note that the authors observed a change of wear regime when the scratch depth became larger than the thickness of the native surface oxide layer. The specific wear rate became much larger when scratching deeper than the oxide layer. Though not explicitly mentioned in their work, we suggest that this corresponds to a transition from adhesive to abrasive wear. Classically, the unit less abrasive wear constant where the friction coefficients in the range of μ = 0.3-0.7 have been reported. The discrepancy can be explained by the absence of major plastic deformation in friction of oxide layers, which is however inherent to macro-scale friction experiments. The tribology of ZrO 2 has been a topic of intensive research (see for example Refs. [26−28]). Recently, the low friction and wear of ZrO 2 thin layers on steel have been connected to the stress-induced martensitic transformation from tetragonal to monoclinic zirconia [26]. In our experiments, the surface oxide layer was found to be partially crystalline. The structure of the nanocrystals was not identified so that a pressure or shear induced martensitic transformation of oxide nano-grains can only be suggested as a hypothesis. Our results on the favorable tribological response of mixed Zr and Cu oxides may shed light on previous results obtained on CuO doped Y-stabilized ZrO 2 by pin-on-disc and AFM experiments, where an interfacial layer with low shear strength generated in the early stage of sliding friction tests has been invoked to explain reduced wear and friction [28].

Conclusions
The contribution of surface oxide layers to the singleasperity tribological response of a Zr 60 Cu 30 Al 10 glass increases significantly upon annealing. The thermal growth of the oxide layer leads to a decrease of scratch hardness. Shearing mediated by adhesion becomes the governing mechanism of friction and wear while ploughing related to plastic deformation of the underlying metallic glass becomes less important. The results clearly demonstrate the dominant role of the oxide layer in frictional microscopic contacts with metallic glasses and the importance of creating a stable oxide layer for practical applications in micro-devices involving sliding contact.