Structurally Driven Environmental Degradation of Friction in MoS₂ Films

MoS₂ thin films were prepared via direct current (DC) magnetron sputtering under 10 mTorr pressure of argon gas. High purity MoS₂ and Ti targets (Angstrom Sciences) were used to produce 650 nm thick MoS₂ films on top of a 75 nm Ti adhesion layer at 200 W and 300 W bias voltages on 1 cm² Si wafers. Separate samples were prepared for PEEM and KPFM analysis.

Tribological testing was conducted on a custom built linear-reciprocating microtribometer [41‐43]. Unless otherwise noted, experiments were performed in a dry N₂-filled glovebox (< 10 ppm O₂ and < 10 ppm H₂O), using a chilled mirror hygrometer to measure humidity. Three test series were run on each sample, the first at a 200 mN load for 250 cycles in two different environments, one track in humidified (50% RH) N₂ and the other in dry N₂, the second at a constant 200 mN load in dry N₂ with variable cycle count (1, 10, and 250 cycles) and the third in dry N₂ for 250 cycles with varying load (10, 50, and 200 mN). All experiments used a 1 mm diameter ruby ball with a sliding speed of 1 mm/s. Wear scars were spaced in close proximity to each other to fit multiple tracks into the 2 mm diameter analysis window available to PEEM. This allowed for direct comparisons between tracks of a given set without re-adjusting the PEEM system.

Molecular dynamics (MD) simulations with a reactive force field (ReaxFF) [44] were used to probe the interactions between molecular oxygen, water and triangular MoS₂ lamellae to investigate how adsorbed species induce structural changes and influence shear strength (and thus friction coefficient). Simulations consisted of three layers of randomly rotated lamellae (triangles ~ 3 nm on a side) of MoS₂ sandwiched between four monolayers of MoS₂. The outer layers were held rigid to apply normal load and shear. After initial shearing, the top layers of MoS₂ were removed, and the lamellae were exposed to oxygen or water at a pressure of 100 atm. These systems were equilibrated for 100 ps before atmospheric molecules were removed and the upper MoS₂ layers replaced for further shear. Simulation methods and schematic of the simulation structure can be found in a previous publication [45].

Work function of different size lamellae, monolayer and bulk MoS₂, were calculated with density functional theory (DFT). The Vienna Ab initio Simulation Package (VASP) v5.4.4 [46, 47] was used with the standard PAW_PBE pseudopotentials packaged with VASP (Mo: 08Apr2002 and S: 06Sep2000). The energy cutoff was 520 eV, equal to 1.3 times the largest ENMAX value for any of the pseudopotentials. The exchange correlation was treated with the PBE functional [48], and dispersion corrections were added using the DFT-D3 method from Grimme [49, 50] using Becke-Johnson damping. Gaussian smearing was employed with a sigma value of 0.05, and the k-points were treated using a 4 × 4 × 1 Monkhorst–Pack grid [51]. Dipole corrections were used for the Z- direction (normal to the plane of MoS₂). The calculation was not spin polarized. The SCF convergence criterion was 10⁻⁴, and structures were geometrically relaxed to a maximum force criterion of 0.02 eV/Å. VASP calculations were set up and managed using the Atomic Simulation Environment [52]. Integration was done with the “accurate” flag. The work function was computed usingwhere E_vac is the energy of the vacuum level and E_Fermi is the Fermi level from the VASP output. The vacuum energy is computed from a 1D averaged electrostatic potential based on the 3D potential output from VASP using the postprocessing utility from Walsh and Catlow [53]. The bulk layer calculations used 5 × 5 unit cells in the X and Y directions with at least 20 Å of vacuum space in the Z direction to avoid self-interaction.

KPFM measurements across wear scars were performed on a Bruker Dimension Icon atomic force microscope (AFM). The MoS₂ sample was mounted on a stainless-steel AFM puck and electrically grounded to the puck using silver paint. A dual-pass lift off was utilized on each line scan to obtain both surface height z and contact potential difference V_CPD data. On the first pass, z was measured via PeakForce Tapping, while on the second pass, V_CPD was measured by retracing the topography at a user-defined height and varying the DC sample voltage V_DC to null the cantilever oscillations (V_DC = V_CPD when the tip-sample electrostatic force gradient was minimized). The work function of the sample ϕ_s was calculated from the V_CPD data through the relationshipwhere ϕ_t is the work function of the tip and e is the elementary charge [54]. In this study, a single Bruker PFQNE-AL probe (nominal tip radius R = 5 nm, spring constant k = 0.8 N/m) was used to collect the z and V_CPD data for all of the wear scars, such that tip-to-tip variations in R and ϕ_t were mitigated. For this cantilever, k was taken to be the nominal value and ϕ_t was taken to be 4.88 eV from earlier work [55]. A Bruker PFKPFM-SMPL standard was imaged before and after the KPFM measurements; minimal changes in z and V_CPD across selected regions of the grating suggested that R and ϕ_t remained relatively constant throughout the experiments. The scan parameters were largely set by the ScanAsyst optimization technique; notable exceptions included a PeakForce setpoint of 10 nN and a lift-off height of 50 nm. A plane fit correction was applied to all z maps to account for sample tilt, while the V_CPD and ϕ_s maps were unprocessed. Cross-sectional profiles of the z and ϕ_s maps were taken across the widths of the wear scars, and the results presented herein were typical of data along the lengths of the wear scars.

PEEM measurements were conducted using a LEEM-III system (Elmitec Elektronenmikroskopie GmbH) coupled to a continuous-wave (CW), tunable wavelength light source based on a Xe lamp. The tunable wavelength light source was comprised of a pressurized Xe lamp (Energetiq, EQ-77 LDLS), a Czerny–Turner monochromator (Acton research, SP2150), and reflective refocusing optics, and illuminated the sample at an incident angle of 73° relative to the surface normal. Maps of the relative work function were determined by recording the photoelectron intensity as a function of the electron kinetic energy using monochromatic light illumination (185 nm wavelength, 6.70 eV) [56]. Similar to conventional analysis of X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS) spectra, lines were fit to the low electron kinetic energy edge and the intersects of the fitted lines with the baseline of the spectra were determined. These intersects specified the locations of the vacuum level. The energy resolution of our electron energy filter was set to ~ 0.1 eV, which resulted in a total broadening of the spectral width by ~ 0.15 eV. This broadening due to the measurement resolution was corrected for post facto by shifting the edges of the spectra, similar to the procedure outlined in Yi et al. [57]. Maps of the vacuum level were obtained by fitting the spectra at each pixel. Local variation in the vacuum level across the field of view yielded the work function variation.

Slices from MD simulations through the plane of lamellae sheared between periodic MoS₂ lamellae are shown in Fig. 1a–d. Lamellae with edge sites that are fully passivated with sulfur atoms are defect-free but non-stoichiometric; these are referred to as defect-free throughout the manuscript. Those where stoichiometry is preserved by removal of selective sulfur atoms from the edges of the lamellae are not fully passivated, and therefore defective and referred to as such.

Stoichiometric (defective) lamellae sheared in vacuum (i.e., no water or oxygen present, Fig. 1b) exhibited the highest friction coefficients, with \(\mu\) ~ 0.15 (Fig. 1e). The simulations show that lamellae tend to anchor themselves at random to neighboring lamellae because of the highly reactive exposed edges. This is likely the atomic scale process that is operative at the interface early in sliding during macroscale experiments when friction is observed to be high. Results from simulations where lamellae were exposed to water and molecular oxygen (Fig. 1c, d) showed lower friction than the defective lamellae, \(\mu\) ~ 0.12 and 0.07, respectively, likely because of passivation of exposed Mo atoms at the edges of lamellae by adsorbed species. This effect is visible in Fig. 1d. These simulations demonstrate that water tends to agglomerate at edges more than molecular oxygen.

To isolate and independently determine the effects of water and oxygen on the friction (i.e., without the changes in passivation of edge sites), we compare friction coefficients between defective and defect-free systems (i.e., with fully passivated, sulfur-terminated lamellae). As shown in Fig. 1e, the defect-free systems exhibited the lowest friction coefficients. The effects of environmental species on microstructure evolution can be probed by counting the number of inter-lamellar bonds in all systems. While individual lamellae are held rigid so that no chemical reactions can occur, the lamellae are able to physisorb and form larger lamellar structures; we refer to these inter-lamellar interactions as bonds and quantify them through the atomic separation between different lamellae. The defect-free systems showed the highest number of inter-lamellar bonds, followed by the defective systems, and then defective systems exposed to water and molecular oxygen. This again indicates passivation from environmental species. Visualization of slices through the simulation box corroborate these results and show that under shear, lamellae order and begin to form larger lamellar structures.

Friction experiments in different environments (dry N₂ vs 50% RH N₂) were used to validate simulation results (Fig. 2a). Additional experimental parameters, including load and cycle count, were systematically varied to determine their impact on the evolution of the surface microstructure by producing surfaces that could be characterized with PEEM and KPFM.

Baseline experiments in dry N₂ using contact forces of 10, 50 and 200 mN exhibited friction behavior typical of MoS₂ films with high initial (few cycle) friction coefficients, \(\mu\) ~ 0.1–0.2, that gradually decreased throughout the first few hundred sliding cycles to values of \(\mu\) ~ 0.05. As in prior literature [58], these MoS₂ films obeyed Hertzian contact mechanics, with friction coefficients that were inversely proportional to contact force. Experiments in humid N₂ consistently led to higher friction than those in dry N₂ (Fig. 2a).

The work function was measured by PEEM and KPFM on separate wear tracks for which representative friction data are shown in Fig. 2. Results of these analyses are shown in Figs. 3 and 4, respectively, where PEEM data are presented in terms of the relative change in work function from the as-deposited surface and KPFM data are presented as absolute values. The work function of these films was independently determined to be 4.9 eV from separate ultraviolet photoelectron spectroscopy (UPS) measurements, matching the values found via KPFM and corroborating the assumed value for the work function of the KPFM tip. PEEM and KPFM results were in good agreement and showed that the shear modified surfaces exhibit a higher work function than the as-deposited films by about 0.10–0.25 eV. The debris surrounding the wear scars, conversely, shows a decrease in work function by a similar magnitude (0.10–0.25 eV) relative to the as-deposited film. Work functions increased with sliding cycles and were higher for friction experiments in dry N₂ than in humid N₂. Changes in work function as a function of normal force were less distinct, but local increases in work function appear to correlate with regions of increased peak contact pressure and were overall higher for the highest contact force. It should be noted that PEEM analysis was carried out in ultrahigh vacuum (UHV) while KPFM was conducted in laboratory air, but these differing analysis environments do not appear to have had an appreciable impact on results.

DFT was used to calculate changes in the work function of MoS₂ as a function of lamellae size and layer thickness, and we compare these results to prior literature as well as our PEEM and KPFM results. The work function for stochiometric lamellae (Mo₂₁S₄₂, Mo₂₈S₅₆ and Mo₃₆S₇₂) was found to increase with increasing lamella size from 4.79 eV to 5.02 eV. These values are smaller than that for a continuous monolayer of MoS₂ (5.69 eV), and the work function further increases to 5.94 eV for both 3- and 4-layer systems (Fig. 5), suggesting convergence to the bulk value. These results are in reasonable agreement with other DFT-calculated values for MoS₂ [59, 60].