Interfacial Interactions and Tribological Behaviour of Metal-Oxide/2D-Material Contacts


 This work combines experimental atomic force microscopy (AFM) and DFT simulations to study oxidized metal (oxidized copper & titanium) and 2D material (graphene & MoS2) interfaces. Combining AFM and DFT allowed identifying the interfacial interaction and established a correlation between tribological behavior, interfacial charge distribution, and variations in the potential energy profile with sliding along the metal/2D-materials interfaces. The TiO2 (rutile) and CuO (cupric oxide) metal oxides were mostly found to chemisorb along the interface with the 2D-materials. Both the metal-oxide counter-surfaces (TiO2 and CuO) exhibited higher friction force and adhesion on graphene than on MoS2. The CuO surface was inferred to be copper rich based on comparison with DFT simulations. The interfacial electronic charge distribution and relative energy change were identified to strongly influence sliding and adhesive behavior between oxidized-metal/2D-material contacts when considering only electronic effects in the DFT simulations. More homogenous interfacial charge distribution/sharing and lower surface energy variation, as found on the MoS2 surfaces, were identified to lower friction and adhesion. Non-electronic effects not captured by simulations were found to likely dominate interfacial shear strength measurements experimentally. Therefore, MoS2 should be used in interfacial applications involving TiO2 and copper rich CuO surfaces requiring lower adhesion and friction.


Introduction
Titanium and copper have been used extensively for decades to manufacture electrical and mechanical components. 1,2 At the macroscale, the tribological properties of metals can signi cantly improve when lubricated by lamellar-materials such as graphite and thick-MoS 2 primarily due to the formation of bene cial tribo lms. [2][3][4][5] The formation aids in the signi cant reduction of coe cient of friction (COF) and wear rate. 2,3,5 The structure of tribo lms can be complex, consisting of ultrathin-2D materials, unoxidized-metal and oxidized-metal species (e.g. CuO, TiO 2 ) when exposed to ambient conditions (e.g. air, water). 1 Identi cation of interfacial phenomenon for these individual species within tribo lms remains unknown mainly for macroscale studies since it is challenging to isolate them individually. 3 Indepth understanding of interfacial phenomena for 2D-materials and oxidized-metals has become vital as they are increasingly used together for designing nanoscale hybrid material systems.
Some of the applications of these nanoscale oxidized-metals/2D-material hybrid systems include advanced lubricants 2 , electronics 6-8 , and energy storage devices. 7,9 They exhibit superior properties including enhanced lubriciousness such as through a 50% reduction in coe cient of friction for copper/graphene vs. pure copper, hardness with an 80% increase for copper/graphene vs. pure copper, Young's modulus (e.g. 3-fold increase for copper/graphene vs. pure copper), wear resistance, thermal/electrical conductivity and fatigue resistance (5-6x enhanced fatigue resistance for copper/graphene vs. pure copper). [10][11][12][13][14][15][16][17] The improved properties are often dominated by the interfacial interaction between the metal/2D-material interface. 18,19 For instance, weakly interacting interfaces such as copper/graphene can minimize crack-formation/crack-propagation resulting in improved fatigue life within nanoscale composite lms. 16,20 Since emerging technologies such as exible electronics are being designed using metal-and oxidized-metal/2D-materials, there is the need for studying and understanding the mechanisms in play at the interfaces which can affect processes such as delamination. Controlled experimental studies of oxidized-metal/2D-material interfaces remain a considerable gap in the current body of literature as existing atomic/nanoscale studies are primarily limited to atomistic simulations such as density functional theory (DFT). 21 DFT studies are bene cial for gaining insight into atomic scale interfacial mechanisms by predicting the interaction between two surfaces, [21][22][23][24] the in uence of changing interfacial energy pro les on sliding, 25 and identifying the correlation between interfacial charge distribution and friction/adhesion. 26,27 During sliding, interfacial interaction can vary depending on the relative position of two contacting surfaces and dictate adhesion at the atomic scale. 26 The interfacial energy corrugation acquired using DFT allows identifying the preferential sliding path with the lowest energy resistance for a given interface. 25 Wolloch et al. 26 recently reported a correlation between surface charge redistribution and surface energy corrugation with adhesion for contacting solid surfaces. Higher interfacial charge redistribution and greater change of interfacial energy were both reported to increase interfacial adhesion. 26 Interaction between pure metals and 2D-materials was predicted using DFT, where copper was reported to have a weaker physical interaction as compared with titanium, which exhibited a stronger chemical interaction with graphene. [21][22][23][24] In real applications, exposure of these metals to operational ambient conditions can quickly change the surface by forming oxide layers. Their in uence on the interfacial interaction and tribological behavior with 2D materials remains unknown.
In this work, we custom-fabricate cantilevers to take an experimental atomic force microscope (AFM) based approach to compare the interfacial behavior of copper and titanium with ultrathin 2D-materials (graphene and MoS 2 ) at the nanoscale. In our study, by combining experimental AFM and DFT simulations, we rst identify the surface chemistry of the metals (i.e. copper and titanium) and the interfacial interaction with ultrathin 2D-materials. Secondly, we study the effect of varying the interfacial interaction on sliding and adhesion. By comparing experimental AFM and DFT simulations, correlation is established between the tribological behavior, the interfacial electronic charge, and the changing interfacial energy pro le along the metal/2D-material interfaces.

Results And Discussion
Characterization of copper and titanium counter-surfaces Analytical high-resolution X-ray photoelectron spectroscopy (XPS) was used herein to characterize copper and titanium samples to understand the metal surface chemistry (Figure 1). It was found that both the metals had undergone surface oxidation. The titanium surface is composed primarily of TiO 2 (Figure 1a, b), where the Ti2p peak energy con rmed the TiO 2 to have a rutile structure. 28 Furthermore, traces of Ti metal, TiO and Ti 2 O 3 compounds were detected from the subsurface (Figure 1b). The thickness of the oxide was estimated by comparing the relative intensities of the Ti2p, Ti3p regions, where the difference in the mean free path due to the universal curve can be used to estimate the thickness of the overlayers (Supporting Information; Section S1). Herein the thickness of the titanium oxide layer was calculated to be ~1.1nm.
Analysis of the copper sample identi ed the presence of cupric oxide (CuO) as the outermost layer ( Figure  1c). Due to the small chemical shifts observed between Cu and its oxides, the copper oxides were analyzed by a chemical depth pro le approach using monotonic Ar + etching while analyzing the copper Auger peak. Mathematical integration of the area under the Auger peak allowed for Cu in its different oxidation states to be followed as a function of depth (Figure 1d and Supporting Information; Figure S1).
It was found that a thin layer of CuO existed as the outermost layer followed by ~6 times thicker Cu 2 O subsurface underneath (Figure 1d). The total thickness of the copper oxide layer is estimated to bẽ 3.3nm. 29,30 Oxidized copper and titanium beads were attached to tipless cantilevers to act as counter- Friction results were repeatable for three separate datasets. On graphene, overall higher friction and adhesion for TiO 2 /graphene than CuO/graphene was observed ( Figure 2a & Table 1). ISS was measured by tting the friction-normal load plot using the procedure proposed by Carpick et al. 31 (Supporting Information; Figure S2). Carpick et al. developed a parametrized approach based on Maugis-Dugdale model. 31 The procedure allows for understanding the range of surface forces at the contact, the contact and extract ISS by tting the FFM data. The interfacial shear strength (ISS) was measured to be ~21.3 ± 6.8 MPa for CuO/graphene and ~83.4 ± 18.1 MPa for TiO 2 /graphene interface (Table 1) Table 1). The more resistive sliding behavior on the TiO 2 counter-surface was accompanied by higher interfacial adhesion ( Table 1). As for CuO counter-surface, there was a greater relative change in friction force as a function of normal load compared to other interfaces and ISS between CuO/MoS 2 was measured to be 13.8 ± 4.9 MPa (CuO/MoS 2 ). It the later section using DFT, it was observed that the interfacial behavior along MoS 2 is highly sensitive to any presence of exposed oxygen on the oxidized-copper counter-surface. Furthermore, comparison of graphene and MoS 2 using the same counter-surface (e.g. CuO/graphene vs CuO/MoS 2 ) shows lower adhesion and friction at low normal loads on MoS 2 than on graphene (Table 1) for both the counter-surfaces, when the sliding behaviour is dominated by interfacial interaction. However, the CuO/MoS 2 has greater change in friction with normal load and exhibits higher friction during sliding at high loading regime. Adhesion on graphene was measured to be several folders higher using the same metal oxide counter-surface than that on MoS 2 (Table 1). To gain insight on the interfacial interaction/mechanism, indepth DFT simulations were conducted to simulate sliding interfaces, and are presented in the subsequent section.
DFT simulations of metal-oxide/2D-material interface In this section, the interfacial interaction between metal-oxide counter-surfaces (CuO and TiO 2 ) and 2Dmaterials (graphene and MoS 2 ) were investigated using DFT (Figure 3a,b). In the case of CuO countersurface, two possible surface terminations (i.e. oxygen-rich and copper-rich) arise due to alternating layers of Cu and O atoms (Supporting Information; Figure S4). Additionally, the CuO counter-surface was also simulated with the less oxidized Cu 2 O subsurface layer (i.e. Cu 2 O + CuO) earlier identi ed in Figure 1.
Unlike CuO, the TiO 2 structure has no such distinguishable layers, resulting in a single surface termination ( Figure 3a). A TiO 2 /bilayer-MoS 2 system was also investigated to gain insight into dthe effect of 2D material thickness on interfacial interaction (Table 1).
For all the material systems, the change in relative energy (with regards to initial step) was tracked as the metal-oxide counter-surfaces translated in incremental sliding steps (ΔԀ) along the 2D-material basal plane. The interface was allowed to relax at each sliding step, in the direciton normal to each surface, and the change in relative energy ( ) was used to evaluate the ISS (τ), where τ = /ΔԀ (Supporting Information; Figure S3). In general, the sliding of metal-oxide counter-surfaces along the graphene basal plane was observed to have higher ISS than MoS 2 for the same metal-oxide (Tables 1 & S2). The metal-oxides had to overcome higher energy barriers along the graphene basal except for the CuO (O-surface) countersurface, where sliding along MoS 2 basal plane was found to be more resistive than graphene (Supporting Information; Figure S3 & S4). The highest average ISS of all systems occured for CuO (Cu-surface) with graphene, while the lowest occured for the CuO (O-surface) with graphene (Supporting Information; Table  S2). Table 1: Comparison of theoretical DFT predictions and experimental measurements for oxidizedmetal/2D-material interfaces. Max energy difference (ΔMax Energy) refers to the difference between relative energies of the least and most stable positions along sliding path. Adsorption distance is for the most stable (lowest energy) position along the path. Average ISS is reported with standard deviation. F POF is the pull off force to measure adhesion between the interface. F f is the friction force with 0nN normal load. ISS Exp is the experimental interfacial shear strength measured from friction force. All experimental interfaces composed of ultrathin-2D materials (i.e. not mono/bilayer 2D material) The interfacial adsorption distances (d ad ) between the oxidized-metal counter-surfaces and 2D-material basal planes were calculated to establish the interaction regimes (chemical or physical interaction) ( Table  1). The interfacial adsorption distances (d ad ) were acquired for the lowest energy depicting the most stable con guration (Supporting Information; Table S1). Previous studies have denoted adsorption distance between 2.0 < d ad < 2.5 Å as chemical interactions. 21 According to the above distance criteria, all of the systems display chemisorption behavioiur (Table 1) with the exception of the CuO-(Osurface)/graphene system (Supporting Information; Table S1). While the majority of interfaces are interacting chemically, the metal-oxide/2D-material interfaces exhibit varying ISS values suggesting additional interfacial mechanisms contributing to the ISS. Along the sliding path, larger electron density and bonding variations result in larger changes in local chemical environment as atoms pass through. These variations are tyically caused by increased electron localization for surface atoms, whereas a more uniform electron distribution is caused by a metal like spread out of the electron distribution. This is in agreement with the correlation that higher relative energy change results in higher ISS (Tables 1 & S2). Overall, if the metal oxide surface termination has metal atoms (copper or titanium), then graphene will present higher ISS. However, if there is an oxygen termination, the MoS 2 will have higher ISS. The presence of an additional layer of 2D material, at least for the TiO 2 /MoS 2 case, did not signi cantly alter the sliding behavior of the interface. This is likely affected by the fact that interlayer spacing for bulk solids of the 2D materials (~3 Å for MoS 2 33 and 3.3 Å or higher for graphite 34 ) is higher than the adsorption distances for the interfaces. The CuO surface with Cu 2 O subsurface, implying less oxygen content as one moves away from the surface into the bulk, results in a reduction of ISS for both graphene and MoS 2 interfaces, with the latter seeing a much greater reduction.

Comparison of Experimental and Theoretical Trends
It should be noted that simulation results are only useful for obtaining qualitative relative trends between surfaces and their quantitative numerical predictions cannot be compared to experimentally measured values as complex real world surfaces are not well captured by the atomic scale DFT models. The DFT simulations are capturing nuclear and electronic interaction which can represent chemical bonding and van der Waals attraction. The insight provided by DFT is that it can isolate these factors and their contribution to friction behaviour (qualitatively), among a number of other factors affecting total friction force measurements.
A comparison between experimental and DFT simulation reveals similar general trends. Experimentally graphene was found to consistently exhibit higher friction and adhesion as compared to MoS 2 against both TiO 2 and CuO counter-surfaces. Higher friction trend on graphene as compared to MoS 2 was also observed using DFT simulations. In the case of CuO counter-surface, experimentally, it is di cult to precisely characterize and control the termination of a surface layer, as was achieved in simulations (Osurface vs Cu-surface). Hence, experiments reported only a single CuO system for each 2D material.
Experimentally, CuO/graphene was reported to have higher ISS than CuO/MoS 2 , while this was found to be the case for CuO (Cu-surface) and Cu 2 O+CuO (Cu-surface) for the simulations ( against oxidized-copper counter-surface (Table 1). This also suggests that interfacial adhesion is more sensitive to the change in surface energy corrugation than friction.
It should be noted that differences between ISS values for the simulations are far more pronounced than for experiments. The simulations utilize idealized surfaces and are mainly capturing the electronic effect of surface chemical species on the sliding behavior in static equilibrium snapshots. The dynamics of kinetic friction are not captured by the simulation. The experimental values are in effect measuring the average of various contributions to sliding force over an area orders of magnitude larger than the simulation cell. There will also be other factors such as contact size, surface morphology (fragments, holes, islands), defects, hetergenous layering, sliding oreintation, thermal effects, contaminants, environmental conditions (including humidity which was not zero for the experiments) and internal mechanical processes involving dissipiation of stress, internal layer movement or structural recon guration which may add to or contrast with these electronic effects in real-world applications and experiments and as such may be dwar ng the purely electronic effect differences captured by DFT simulations. The ISS experimental results thus suggest that chemical interacitons are not the dominant contributor to friction during interfacial shearing.
The metal oxides are less likely to be well described by atomic scale simulations than the 2D materials as the latter are actually atomically thin and have simpler structures. One of the experimental metal oxide surfaces might differ more from its DFT representation than the other. Therefore, comparing trends for the same metal oxide surface should be more reliable than across the two different metal surfaces and the former do follow the experimental trends exactly.

Conclusions
This study investigated the tribological behavior of metal counter-surfaces (i.e. titanium and copper) with 2D-materials (i.e. graphene and MoS 2 ). The metal surfaces underwent native oxidation with the formation of few nanometers thick TiO 2 and CuO layers. These oxides played an important role in the interfacial properties and exhibited chemical like interactions for most of the interfaces studied herein. Experimentally, higher friction force and adhesion were measured on graphene than on MoS 2 for both the metal-oxide counter-surfaces (i.e. TiO 2 and CuO).
In the case of TiO 2 counter-surface, DFT simulations also predicted lower friction on MoS 2 . Comparison of monolayer vs bilayer MoS 2 showed no signi cant change in relative energy and ISS. This indicates that the thickness of 2D materials will have minimal in uence on the chemical interaction of sliding metal-oxide counter-surfaces. In the case of CuO counter-surface, two possible exposed terminations were identi ed (O-vs Cu-surfaces), and both were found to in uence the overall interaction in simulations (the two cannot be differentiated experimentally). For CuO (O-surface), sliding on graphene was found to have a lower ISS than MoS 2 while for CuO (Cu-surface) the opposite behavior was observed. This is suggestive that the Cu-surface dominates the interfacial interaction experimentally.
Experimental and DFT comparisons also identi ed the dependence between the tribological behavior, the interfacial electronic charge, and the changing interfacial energy pro le along the metal-oxide/2Dmaterials interfaces. While the majority of interfaces are interacting chemically, the metal-oxide/2Dmaterial interfaces exhibit varying interfacial behavior suggesting other dominating interfacial mechanisms. The small differences in ISS for the same meta-oxide countersurface experimentally in comparison to DFT results likely means that non chemcial interaction factors overwhelm electronic effects for shear behaviour. More homogenous interfacial charge distribution/sharing and lower surface energy variation, such as that observed for the TiO 2 /MoS 2 and CuO (Cu-surface)/MoS 2 interfaces, were identi ed to lower friction and adhesion. The nding from this study is suggestive to combine MoS 2 with metal-oxide counter-surface if lower adhesion and easier shearing at the interface is preferred when designing nanoscale oxidized-metals/2D-material hybrid systems. However, graphene should be preferred over MoS 2 if the metal-oxide counter-surface is highly oxygen terminated.

Methodology
Experimental Sample Preparation, AFM Cantilever and Characterization. Graphene and MoS 2 samples were prepared by mechanical exfoliation using scotch tape method. Highly oriented pyrolytic graphite (SPI supplies) and a large MoS 2 crystal (Graphene supermarket) were used to peel and deposit ultrathin sheets onto n-doped silicon wafer substrate (Graphene supermarket). The surface of silicon substrate was clean using ethanol and methanol in an ultrasonic bath prior to exfoliation. AFM measurements show the thickness for 2D materials as 6.6 nm for MoS 2 and 4.2 nm for graphene. The custom fabricated AFM cantilevers were prepared by attaching copper and titanium (Alfa Aesar) beads of 99.9% and 99.5% purity onto tipless cantilevers (APP-Nano). A custom built microscope-micromanipulator setup was used to apply the PC-Super epoxy and attached the metal beads (bead radius of 2.5 μm for titanium and 3.5 μm for copper).
The same beaded cantilevers were used to avoid the in uence of bead shape or size when compare the 2D materials. A Hitachi scanning electron microscope (SU3500) was used for image the cantilevers. X-ray photoelectron spectroscopy (XPS) was performed using an ESCALAB 250Xi (Thermo Fisher Scienti c -East Grinstead, UK) to identity and measure the thickness of oxide layers on the copper and titanium beads. The ESCALAB 250Xi system is equipped with monochromatic Al K α X-ray source and was used to collect the survey spectra with pass energy (PE) of 100 eV and nominal spot size of 400 µm. Surface characerization was followed by the spectral regions of interest at higher resolution (PE -20 eV) from which the composition (Rel. At.%) was obtained.
Tribology measurements. The Asylum Research MFP 3D atomic force microscope was used to perform friction force microscopy (FFM) and adhesion experiments. FFM and adhesion experiments were done using the custom fabricated copper and titanium cantilevers. The normal and torsion stiffness of the cantilevers were measuring using Sader's method. 35,36 The normal stiffness was measured ∼5.4 N/m for copper and ∼2.6 N/m for titanium cantilevers. Test probe method 37 was used to acquire the lateral sensitivity against a cleaved potassium bromide block and the normal sensitivity was measured by de ecting against silicon wafer. Friction voltage was measured as half the differences between the lateral trace and retrace signals as the cantilever slides at 90° scan angle with scan speed of ∼5 µm/s. Using the FFM (friction vs. normal load) data, the experimental interfacial shear strength (ISS) was calculated by tting the generalized Maugis-Dugdale model proposed by Carpick et al. 31 The tting of the generalized Maugis-Dugdale model allows for the extraction of pull off forces and , where indicates the transition between JKR and DMT models ( corresponds to the JKR case and corresponds to the DMT case). Using these extracted parameters, the transition parameter ( ) and interfacial energy ( ) can be determined, allowing for the estimation of the contact radius ( and calculating ISS experimentally. Detailed ISS tting procedure is provided in the supporting information (Supporting Information; Section S2). Lastly, adhesion experiments were done with maximum normal load of 90nN and dwell time of 1s. Dry environment was acquired by purging the environment locally around the contact with 99.9% pure N 2 gas.

Simulations
Plane wave-based DFT was run through the Quantum Espresso software package. 38,39 Interactions between the valence electrons and the ionic core were represented by the projector augmented wave (PAW) 40 method with Perdew-Burke-Ernzerhof (PBE) formulation. Kinetic energy cutoffs of 748 eV (55 Ry) and 8163 eV (600 Ry) were used for the wave functions and the charge density, respectively. All calculations were non spin polarized and van der Waals corrections were applied through Grimme's DFT-D3 method with Becke-Jonson (BJ) damping. 41,42 A Monkhorst-Pack de ned mesh of 2 x 2 x 1 k-points was used to sample the Brillouin zone. Adsorption distance was based on taking average position of surface atoms for each surface of the interface and then nding the difference in these two average positions.
All material systems were modelled periodically, where the surfaces were brought to a separation of 2.3 Å and the interface was allowed to relax in the out-of-plane direction (normal to each surface) for the most stable interfacial separation distance. The (001) plane of each metal oxide surface (CuO & TiO 2 ) was tested against both graphene and MoS 2 . For the oxidized metal counter-surfaces, only the layers closest to the interface were allowed to relax while those further away were xed in order to simulate the bulk solid. The 2D sheet was then moved in the in-plane direction by a 0.33 Å step displacement (ΔԀ) and again relaxed in the out-of-plane direction for a total of 10 steps. This simulates a pseudo-sliding motion where the surfaces are allowed to reach equlibrium separation distances at discrete points along the sliding path. Note that this does not simulate a dynamic process such as found in kinetic friction measurements or those with applied loads (no load is applied in the simulations and the systems are relaxed at absolute zero in a vaccum).  a) DFT-based interaction of TiO2/2D-materials (graphene and MoS2) for tracking the relative energy change during sliding. Insets: DFT material system and ELF analysis for TiO2/2D-material interfaces. b)

Declarations
DFT-based interaction of Cu2O+CuO/2D-materials (graphene and MoS2) for tracking the relative energy change during sliding. Insets: DFT material system and ELF analysis for Cu2O+CuO/2D-material interfaces. Purple grey atoms are Mo or Ti, red atoms are O, bright yellow atoms are C and pale yellow atoms are S.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.