Robust structural superlubric interfaces under high current density, from understandings to applications

Abstract

Structural superlubricity (SSL) exhibits significant potential for applications in micro/nanoelectromechanical systems, switches, and sensors, owing to its characteristics of near-zero friction and zero wear. However, the deployment of the more universal 2D/3D SSL interfaces in efficient electronics is severely limited by an insufficient understanding of their electrical stability and failure mechanisms under high current density. In contrast to the Joule-heating-induced interlayer bonding that limits the critical current density in 2D/2D SSL interfaces, we report a fundamentally different failure mechanism for 2D/3D SSL interfaces under high current density: the interfacial Au layer melts and transitions from single-crystalline to polycrystalline state, which drastically increases surface roughness and thus causes a friction surge, severe graphite wear, and the eventual collapse of the SSL state. Notably, under this new failure mechanism, the critical current density of the interface increases by nearly an order of magnitude compared to that of 2D/2D SSL interfaces, exceeding 110 GA/m². Furthermore, we developed a high-power switch prototype that significantly outperforms conventional ones, demonstrating a critical step in translating SSL from principle to practice and paving the way for addressing contact issues in high-power, long-life electronics.

Introduction

Microsystems, particularly MEMS (Micro-Electro-Mechanical Systems)¹, integrate sensing, actuation, and processing on a single chip, making them foundational to the Fourth Industrial Revolution and critical across consumer electronics, automotive and healthcare applications^2,3. High-performance MEMS switches and sensors^4,5 are essential for enabling advanced capabilities in 5 G/6 G telecommunications^6–8, autonomous vehicles^9,10 and next-generation robotics¹¹, where they provide compact, efficient, and reliable functionality. However, traditional MEMS face challenges^12–19 in miniaturization, power density and longevity under extreme operational demands in emerging fields like humanoid robotics^20–22, 6 G communications²³ and unmanned aerial vehicles^24,25. A critical factor in overcoming these limitations is the development of electrical transport interfaces that exhibit high current density, excellent thermal conductivity, and superior wear resistance, which are vital for significant improvements in power handling and device miniaturization^26,27.

Structural superlubricity (SSL) has emerged as a promising solution to the above challenges^28,29. In the SSL state, friction is reduced to near-zero levels³⁰, and wear is virtually eliminated³¹. This phenomenon, which occurs at certain interfaces—such as between graphite and other materials^31–33 – could address the fundamental issues of friction and wear in sliding contacts. More importantly, SSL interfaces exhibit exceptional electrical contact properties, including ultrahigh critical current density (CCD) and ultra-low contact resistivity. Recently, an ultrahigh CCD of J_c = 17.5 GA/m² has been observed at graphite/graphite interfaces during sliding, while maintaining the SSL state³⁴. Beyond that, ultralow contact resistivity and friction have been simultaneously achieved at wear-free Au/single crystal graphite (SCG) and graphite/diamond like carbon SSL interfaces ^35,36.

Existing research findings demonstrate the promising potential of SSL interfaces for SEC. However, our understanding of their failure mechanisms under practical conditions remains limited. To date, studies have reported on the failure of 2D/2D SSL interfaces under extreme normal loads³⁷. Theoretical work has also predicted failure mechanisms in 2D/2D graphite interfaces under high current densities³⁴, suggesting that failure may be related to Joule-heating-induced interlayer bonding. Despite these insights, the failure mechanisms of SSL interfaces—particularly those involving metal/graphite interfaces, which offer better compatibility with micro-fabrication processes—are still not well understood, lacking direct experimental characterization and first-principles analysis. Specifically, it remains unclear how electrical loading leads to structural damage at the interface: whether it promotes bond formation, causes bond breaking and defect generation due to Joule heating, or alters the state of the metal interface. These questions are critical for the practical application of SSL technology application in high-efficiency electronic devices, yet they remain inadequately addressed in current research.

In this study, we have elucidated for the first time the failure mechanism of 2D/3D SSL interfaces under high current density. Our work reveals a coupled electro-thermo-mechanical pathway that leads to the breakdown of the SSL state. This new mechanistic understanding has enabled the discovery of an ultrahigh CCD of 117 GA/m² in the Au/SCG interface, which surpasses previous records by nearly an order of magnitude. These insights directly inform the design of robust RF MEMS switches incorporating SSL contacts capable of operating at exceptionally high transmission current densities. Our work establishes a unified framework for understanding failure limits in superlubric contacts and accelerates their integration into high-performance power and MEMS applications.

Results

We realized epitaxial growth of a single-crystal graphite (SCG) thin film (1 cm × 1 cm) with a thickness of 20 μm on a nickel foil³⁸. The SCG film was subsequently transferred onto a silicon substrate. For comparative analysis, we employed Electron Backscatter Diffraction (EBSD) to characterize both SCG and highly oriented pyrolytic graphite (HOPG), as shown in Fig. S1 in Supplementary Information (SI), demonstrating that the epitaxially grown graphite was single-crystalline, sharply contrasting with the polycrystalline nature of HOPG. Raman spectroscopy mappings of the SCG interface (Fig. S2 in SI) confirmed the absence of defects^39–41, underscoring the high quality of the SCG film. For subsequent fabrication, a 200 nm-thick Au layer was nondestructively deposited onto the SCG surface using thermal evaporation. Employing photolithography and additional microfabrication techniques, we prepared uniformly shaped Au mesas. Detailed methodologies were provided in Fig. S3 and Section 1 in SI.

Firstly, we developed a custom microscale electrical test system, as illustrated in Fig. 1a. To provide a comprehensive view of the measurement configuration, a detailed schematic of the experimental setup and an optical photograph of the actual testing apparatus are presented in Fig. S4 of the SI. A micromanipulator (MM3A, Kleindiek) was employed to precisely control a tungsten probe (tip radius between 1 and 2 μm), enabling the manipulation of the Au mesa and application of high current. The SCG film was mounted on a piezoelectric ceramic stage to provide periodic reciprocating displacement. A source meter (Kethley 2470) was used to supply the required DC current and record the measurements. Specifically, the tungsten probe was pressed onto the Au mesa (typical size 3 μm×3μm)with the normal load of 200 μN-400 $\mathrm{\mu }\mathrm{N}$, while the sample substrate was driven in a periodic reciprocating motion at 1 Hz by a piezoelectric positioning stage (PZS), producing a relative sliding displacement of ~1–2 $\mathrm{\mu }\mathrm{m}$ between the Au mesa and the SCG substrate. During sliding, a source meter applied a constant current across the Au/SCG interface for 1 to 2 s. By finite element method (FEM) and theoretical calculations (details in Section 2 of the SI), the relaxation time of the system was estimated to be on the order of 10⁻⁶ s, indicating that our experiments have reached thermal equilibrium.

Fig. 1.

Setup and comparison of friction force, differential friction coefficient, morphology and grain characterization of Au at high/low current densities for the Au/SCG interface. a The illustration of our experimental setup. b The relationship between the friction force f and the previously applied current density at the Au/SCG interface. These data were obtained from nearly 70 Au mesas of the same size (3 μm×3 μm) after passing through different current densities. c The relationship of the friction force and normal load at the Au/SCG interface as a function of the previously applied current density. J was the current density passing through the Au/SCG interface. μ was the differential friction coefficient. For J = 0 GA/m², μ = −2.3 × 10^-4. For J = 19.44 GA/m², μ = −3.1 × 10⁻⁴. For J = 36.11 GA/m², μ = 3.8 × 10⁻⁴. For J = 55.56 GA/m², μ = 4.52 × 10^-3. Error bars were smaller than the friction force symbol at the corresponding position and were therefore not shown in (c). The averaged friction of the four sets of data in (c) is highlighted in red squares in (b). d (i-ii) Topography of the lower surface (sliding interface) of Au mesa with low J (i) and high J (ii). Both figures of topography use the same color bar. Inset: height profile of the surface at the red line. The scale bar is 500 nm. (iii-iv) EBSD IPF-X maps of the lower surface of Au mesa with low J (iii) and high J (iv). The scale bar is 2 μm

After the current tests, we used the current-treated Au mesas to measure the friction by atomic force microscopy (AFM) (details such as friction loops could be reached in Fig. S5 in SI). The relationship between the friction force and the previously applied current density J for each sample was shown in Fig. 1b.

The experimental data revealed a sharp increase in friction force (f) near a specific range of J (42–45 GA/m²), indicating that the Au/SCG SSL interface likely underwent structural failure in this region. This abrupt change suggested the occurrence of a first-order transition at these current densities⁴². Additionally, experiments about the dependence of friction force on normal load also showed that the friction coefficient of sample under current density larger than CCD (55.56 GA/m²) was much higher than those of samples under current density lower than CCD (0, 19.44, 36.11 GA/m²), as presented in Fig. 1c. These observations provided crucial insights into the threshold conditions for SSL interface failure and highlighted the interplay between electrical and mechanical factors in defining interface stability.

Additionally, we further characterized the topography of the lower surface of the Au mesa (sliding surface) without J and after high J by AFM (Cypher ES, Oxford Instruments). As shown in Fig. 1d(I, ii), the results revealed that the roughness of Au mesa at interface without J (0.363 nm) was significantly lower than that of the roughness after high J (1.887 nm). Furthermore, we employed EBSD characterization to characterize the lower surface of the Au mesa without J and after high J, as shown in Fig. 1d(iii, iv) and Fig. S6 in SI. The results revealed that the lower surface of the Au mesa was initially single-crystalline, with a (111) crystallographic orientation. However, after high J, poly-crystallization was observed on the lower surface of Au mesa. These observations strongly indicated that the failure of Au/SCG SSL interface originated from high-current-induced damage to the lower surface of the Au mesa.

Based on the characteristics of the first-order phase transition (sudden change in friction) and the characterization results, we hypothesized that the observed changes in friction might result from interface failure induced by localized temperature increased by the applied high current density. Specifically, this failure could arise from the melting of the lower surface of the Au mesa.

To investigate this, we estimated the current density required for the melting of Au at the interface by FEM simulation, which meant to calculate the current density when the highest temperature of the interfaces reached 1337 K (melting point of Au). Details of the simulation setup were provided in Section 2 of the SI. The simulation results, shown in Fig. S7 (also presented in Fig. 2b: the correspondence of two lateral axis), revealed that the predicted current density aligns closely with the CCD values obtained from experiments. This agreement strongly suggested that Au was melted at CCD and thus caused the failure of the interface.

Fig. 2.

Molecular dynamics simulation models and results. a Front view of the simulation model and set-ups. b Dependence of the static friction stress and the temperature T_h. The correspondence of the current density and T_h was shown in upper and lower x-axis. The blue-to-red background color indicating the transition from near-zero to finite friction. c Per-atom lateral force f_x at Au interface for systems annealed from T_h = 1200 K (upper) and 1500 K (lower). d The T=0 radial distribution function (rdf) of the Au mesa after annealing from T_h. Narrow peaks (blue) suggested good single crystallinity. The Au-Au distance corresponding to the first four peaks were marked in the figure

To evaluate the operational uncertainty of this conversion and verify the reliability of our model, we performed a comprehensive sensitivity analysis by varying both the interfacial contact resistivity and the interfacial thermal contact resistance within a ±20% range (Fig. S8 of SI). The calculated CCD exhibits a well-bounded dependence on these boundary parameters, confirming the robustness of our electro-thermal predictions. Furthermore, the FEM simulation reveals a highly uniform temperature distribution across the Au/SCG interface, with a maximum spatial temperature variance of less than 50 K (Fig. S9, SI). This uniform thermal profile demonstrates that instead of a minor, localized defect initiating the failure, a substantial fraction of the interfacial Au layer reaches the melting threshold almost simultaneously, driving a system-wide, irreversible structural reconfiguration. Furthermore, considering that the exceptionally high thermal stability of graphite⁴³, we suppose that the SSL interface failure is caused by the structural breakdown of the bottom surface of Au under CCD.

To further substantiate this hypothesis and obtain direct atomic-level information, we conducted MD simulations, mimicking the experimental progress. The simulation model consisted of an Au (111) mesa and graphene substrate in incommensurate contact (Fig. 2a). Periodic boundary conditions were applied to the in-plane x and y directions to isolate the intrinsic response of the contact interior from edge effects³⁰. The rationale for this choice and the influence of finite edges was discussed in Section 3 of the SI. We followed the standard simulation protocols for simulating SSL systems²⁹: the graphene substrate was constrained by springs (stiffness k_z) to mimic the semi-infinite substrate, while the center-of-mass of the Au mesa’s top layer was connected to a slider via a pulling spring with stiffness K_p. The sliding of the Au mesa was achieved by moving the slider, mimicking the lateral stiffness of the AFM tip in experiments. The system temperature was firstly raised up from room temperature to T_h to simulate the thermal effects induced by the applied J; after which, the system was annealed for quasi-static sliding simulation⁴⁴ (detailed set-ups in Methods).

As shown in Fig. 2b, the simulated dependence between static friction (stress) and T_h exhibited a significant first-order transition, with a critical temperature T_c≈1300 K – below which the frictions were always near-zero, while above which the frictions were large, and the values varied in several independent simulations. This high temperature corresponded to a current density of $48\mathrm{GA}/{\mathrm{m}}^2$, which agreed well with the CCD observed in the experiments, $\approx 44\mathrm{GA}/{\mathrm{m}}^2$. The slight overestimation in the simulation might come from the higher temperature (and thus current) required to achieve the solid-liquid phase transition within the relatively short simulation time (∼ns) – bearing in mind the hysteresis in the transition. From structural analysis, we identified the fundamental reason of this sudden increase in friction—melting of the Au interface at high temperatures, which transformed to a polycrystalline state upon cooling. Consequently, the originally incommensurate contact interface with regular, well-canceling lateral forces became irregular, preventing force cancellation and leading to a significant friction (Fig. 2c), as well as energy dissipation during realistic sliding. By comparing the (T = 0) radial distribution functions (rdf) of systems treated at T_h = 1200 K and T_h = 1500 K (Fig. 2d), we revealed that the Au reconstructed after exposure to high currents (or temperatures). Above factors ultimately cause the failure of SSL. Note that the simulated static friction and experimental kinetic friction are not directly comparable in magnitude, particularly because of the large difference in sliding velocity. Nevertheless, the quasi-static and additional finite-velocity simulations consistently show a pronounced friction increase following interfacial reconstruction. Details are provided in Section 4 of the SI.

Building on the results from FEM and MD simulations, we concluded the following mechanism for the current-induced failure of the Au/SCG interface: when the current density exceeded the critical threshold, the Au layer began to melt. When the current was removed, the interface cools down and forms reconstructed polycrystalline structures. The newly formed grain boundaries introduced significant friction, disrupting the perfect cancellation of atomic-scale lateral forces that originally enabled the SSL state. This led to a sharp increase in overall friction. Moreover, the localized pinning sites at the interface, the surface corrugation and/or nanoclusters that may be formed during the cooling process, and the consequent stress concentration might further promote wear of the interface and lead to the failure of superlubricity ²⁹.

To further validate this theoretical prediction, we conducted comparative experiments as shown in Fig. 3a. We first measured the friction force of the Au mesa at Position 1 (P1) to confirm that the Au/SCG interface here was intact and in the SSL state. Subsequently, we applied a high current density (J = 55 GA/m²) exceeding the CCD value here on Au mesa and measured the friction of the interface after this test. The results showed a significant increase in friction (P1 after high J in Fig. 3b) compared to the initial state (P1 before high J in Fig. 3b). We then identified another defect-free region near P1, labeled P2, as shown in Fig. 3c(i) and Fig. 3d), using Raman spectroscopy. Next, we used an AFM tip to move the Au mesa to P2 and measured the friction force without any J on Au mesa. The results indicated that the friction at P2 was almost identical to that at P1 after the application of high J, and significantly higher than that of the Au/SCG interface in the SSL state (P1 before high J). The fluctuation of friction implied the wear was generated during sliding. Then we removed the Au mesa away and performed Raman spectroscopy on the sliding region at P2 and observed a prominent defect-related peak in Fig. 3c(ii), d.

Fig. 3.

Experiments results and analysis of the failure mechanism at Au/SCG SSL interface. a Schematic diagram of experimental steps. P1 and P2 means position 1 and 2, respectively. The dashed box means the sliding area of the position. b The friction force of Au mesa/SCG interface at P1 (before (pink line) and after (red line) passing through high current density) and at P2 (transferred from P1 after passing through high current density). All friction experiments used the same Au mesa. c (i) Raman spectrum mapping (I_D/I_G) for P2 of SCG before sliding (defect-free). (ii). Raman spectrum mapping (I_D/I_G) for P2 of SCG after sliding (defective). d Raman signals of dots (blue in (i)) and crosses (red in (ii), clear D-peak was observed) in (c) with the Raman shift from 1000 to 3000 cm⁻¹

Throughout the entire testing process at P2, no current was applied. Nevertheless, the originally intact graphite surface still exhibited clear damage (D peak in Fig. 3d). This indicated that the defects observed in the graphite interface were caused by the already degraded bottom surface of the Au mesa. The high current density induced poly-crystallization and roughening of the Au interface, which significantly enhanced the friction force at the Au/graphite interface, thus leading to the defects of the graphite interface and the breakdown of the SSL state. This comparative experiment strongly validated the failure mechanism process we previously derived from experiment and MD simulations.

In addition, we further evaluated the potential impact of the Au mesa edges (e.g., thermal desorption, oxidation, or structural reconstruction) prior to the phase transition. These progressive, edge-initiated phenomena are experimentally ruled out as the primary failure mechanism. Specifically, the sliding friction remains perfectly stable and near-zero right up to the critical current density (Fig. 1b). Meanwhile, independent external heating tests reveal that friction decreases monotonically with temperature (Fig. S10), directly contradicting any thermally activated friction surge. Moreover, the permanent post-cooling friction jump points to an irreversible bulk phase transition (melting) rather than transient edge modifications, which would otherwise relax upon cooling.

The experimental and simulation results presented above revealed a novel failure mechanism in the Au/SCG SSL interface under high current density. This newly identified mechanism suggested the potential for superior current-carrying capacity. Based on this finding, we further investigated the current density characteristics of the Au/SCG interface across different scales.

All results of the CCD values at different scales were shown in Fig. 4a. Representative results, including a current density of 117 GA/m² with Au mesa size of 920 nm × 920 nm, and no observed damage on the sliding region of SCG after the CCD test, were shown in Figs. S11 and S12.

Fig. 4.

Analysis and comparison of CCD characteristics in 2D/3D SSL interfaces. a CCD values for Au mesas of different sizes on SCG. The results were obtained in atmospheric environment, vacuum environment and COMSOL simulation. b Comparison of mechanical (wear rate (mm³/N⋅m) and the coefficient of friction (μ)) and electrical properties (J_c (GA/m²) and σ_c (S μm²) is the CCD and conductivity of the interface, respectively) among our work, graphite/graphite SSL contact, and traditional SEC

For small samples (<2 μm), CCD test was performed under the scanning electron microscope to facilitate manipulation and to compare results in vacuum and atmospheric conditions, as shown in Fig. S13. The experimental results revealed no significant differences between the two environments, indicating that the effects of atmosphere were negligible.

In addition, we conducted systematic FEM (details in Section 2 of the SI) simulations to determine the CCD values of the Au/SCG interface across various length scales, based on an established critical temperature criterion of T_c= 1337 K. The simulation results exhibited excellent agreement with our experimental results as shown in Fig. 4a.

Moreover, the experimental and simulated CCD results across different sample sizes were fitted to a logarithmic-linear model, yielding a slope of k = −0.6435 (R² = 0.98), demonstrating a power-law relationship between CCD and contact scale, as shown in Fig. S14 in SI.

This scaling behavior is fundamentally governed by the competitive balance between localized electro-thermal Joule heat generation at the Au/SCG interface and multi-directional heat dissipation through both the highly conductive SCG substrate and the measurement probe. As the contact size decreases, the relative efficiency of heat dissipation increases significantly due to the localized diffusion profile of current and heat in the SCG, thereby allowing smaller interfaces to sustain a much higher CCD before reaching the critical failure temperature. To safely extrapolate this scaling relationship to other dimensions, three key physical assumptions are established: (i) the interfacial Au layer maintains its bulk melting point of 1337 $\text{K}$, which holds well for lateral dimensions above ∼20 $\text{nm}$ where size-dependent melting point depression is negligible⁴⁵; (ii) the contact area between the tungsten (W) probe and the Au mesa scales proportionally with the mesa’s lateral dimensions; and (iii) the bottom and outer edges of the SCG substrate are maintained at a constant temperature of T = 293$\text{K}$ to serve as an ideal heat sink.

The edge and in-plane inhomogeneous transport revealed here was mainly due to the diffusion transfer of current in SCG³⁵. A detailed discussion of the scale-dependent electrical conduction and heat transfer mechanisms underlying this relationship was provided in Section 5 of the SI, enabling CCD predictions for Au/SCG interfaces across all scales.

Here, the tribological and electrical properties of this Au/SCG SSL contacts were compared with those of the graphite/graphite SSL contacts and nearly the best parameters screened from various traditional sliding electrical contacts (SECs), as shown in Fig. 4b.^{34–36,46–52}. In that four-dimensional plot evaluating friction-wear performance, interface electrical contact quality, and CCD, the Au/SCG contacts broke the recent record set by the graphite/graphite SSL contacts and completely surpassed all traditional SECs.

To exhibit the excellent mechanical (wear free for longevity and robust low coefficient of friction) and electrical properties (ultralow contact resistivity³⁵ and ultrahigh CCD) of this Au/SCG contact, we designed a contact mode MEMS switch based on SSL technology. Figure 5a showed the SEM image of one fabricated MEMS switch. The switch was controlled by applying a driving voltage that caused the adsorption and release of the cantilever, enabling the opening and closing of the circuit. A fixed graphite flake (Fig. 5b), serving as the upper contact, was integrated on the cantilever, while a fixed Au hemisphere (Fig. 5c, d), serving as the lower contact, was located on the substrate. When the cantilever was adsorbed under the application of driving voltage, the upper contact and lower contact came into connection, and the circuit was “on”. During the switch’s closing process, there would be a slight sliding between the graphite and Au contacts, causing the wear failure and unstable connection of device⁵³. The use of the SSL contact prevented failure due to frictional wear, significantly extending the device’s operational lifespan.

Fig. 5.

MEMS switches with SSL contacts. a SEM image of one fabricated MEMS switch. b Optical microscope image of the graphite contact after test (Contact pair 1, and pair 2 in Fig. S15 of SI). c Optical microscope image of the Au contact after test. d Topography of Au contacts. Inset image: Height profile of the Au contact. e Cold switch of the device. Current signal: blue line, right axis. Driving voltage of the switch: red line, left axis. f Voltage and current curves of the MEMS switch in the “on” state when the probe was pressed on cantilever

Here, we controlled the switching operation by applying and removing a 95 V driving voltage (controlling section), while (functional section, signal or power) passing through the contacts during closure. The “on” and “off” of current controlled by driving voltage (cold switch) was shown in Fig. 5e. The normal force between the contacts while being absorbed was estimated to be approximately 172 μN (details in Section 6 of the SI), corresponding to a contact area of ~1.04 μm² by Hertzian contact theory in Section 7 of SI.

Therefore, we could calculate the current density of this MEMS switch was about 67.3 GA/m², although the ideal CCD value for Au/SCG contact of this scale was approximately 75 GA/m² according to Fig. 4a. This was due to the roughness of Au contacts, about 3.155 nm (the topography of Au contact could be reached in Fig. 5d, causing the real contact area was a little less than calculated. We also characterized the graphite surfaces by Raman spectroscopy (Figs. 5b and S156 in SI), and the results demonstrated that both the graphite and Au contacts (optical observation) exhibited no observable damage even after sustaining ultra-high currents.

However, the driving voltage imposed a limit on the normal force of the contact, as excessively high voltages might lead to breakdown of the driving electrodes, thereby restricting the real contact area of graphite/Au interface. To explore the limits of the MEMS switch with SSL contacts, we accomplished an on-state current of 210 mA across the two pairs of contacts by applying pressure to the cantilever of the MEMS switch with a probe, as shown in Figs. 5f and S17 in SI. Here, we compared our MEMS switch with Menlo Micro’s flagship product, the MM5130. In terms of single pair of contact current capacity, the MEMS switch with SSL contacts (∼105 mA) outperformed the MM5130 (∼62.5 mA) by 50%⁵⁴. Even with the application of probe pressure on the cantilever, which increased the contact area to S = 2.91 μm² (with a normal load of 800 μN), the corresponding current density was 36 GA/m². This value still fell short of the ideal critical current density (∼60 GA/m²) at the SCG/Au interface for this scale, which was also limited by the surface roughness of Au contact. Therefore, this was also a challenge which is vital to further advancing the performance of SSL-based MEMS devices.

Discussion and outlook

We introduce a novel failure mechanism for 2D/3D SSL interfaces, which enables a nearly order-of-magnitude improvement in the CCD values. Here, we aim to discuss what characteristics an SSL interface with high CCD values should possess by comparing the Au/SCG and graphite/graphite interfaces.

The critical factor was contact resistivity disparity. Graphite/graphite SSL interfaces require to be incommensurate to achieve superlubricity, but this introduces substantial electronic transport barriers, yielding contact resistivity far exceeding AB-stacked graphite^46,55. In contrast, Au/SCG interfaces exhibit ultra-low contact resistivity approximately two orders of magnitude lower than graphite/graphite SSL interfaces^35,46, generating nearly two orders of magnitude less Joule heat at equivalent current densities due to superior electrical transport.This indicates that SSL interfaces with high CCD values must first and foremost possess low contact resistivity; beyond that, we have further discovered that the interface failure mechanism significantly governs the CCD performance of SSL interfaces.

During preparation, graphite flake edges consistently form a 30 nm amorphous carbon layer. Most failures (∼2/3) in graphite/graphite SSL interfaces occur at these edges (Fig. S18a). We attribute this to bonding between the edge amorphous carbon and underlying graphite. MD simulations show interlayer bonding initiates at these interfaces above 700 K (Fig. S18b). Subsequent sliding causes bond breaking, damaging the pristine graphite interface and leading to SSL failure. The graphite/graphite SSL failure temperature (T_c1 = 700 K) is significantly lower than the Au/SCG interface (T_c2 = 1337 K), indicating edge amorphous carbon severely limits the CCD of graphite/graphite SSL interfaces.

In contrast, the Au/SCG interface, with its high-quality, pristine, edge-free structure and superior thermal stability—where interface failure occurs only upon Au melting—achieved an ultrahigh CCD, demonstrating that SSL interface failure under ultrahigh current densities is predominantly driven by thermal degradation-induced structural collapse. Moreover, the critical thermal failure temperature is intrinsically linked to the interface failure mechanism. Based on this analysis, we summarize two key characteristics of high-CCD SSL interfaces: (1) low contact resistivity and (2) high failure temperature. These enable CCD prediction/evaluation for other SSL systems: for typical 2D/3D interfaces (e.g., metal/graphite), CCD can be assessed through contact resistivity measurement, metal melting point referencing, and finite element model simulations, providing a practical methodology for rapid screening of high CCD SSL SEC interfaces.

Beyond its exceptional energy transport capacity, the Au/SCG interface offered immense potential for device integration due to its 2D/3D hetero-structure characteristics. Unlike traditional 2D/2D contacts, which often required complex transfer processes, the Au/SCG interface could be seamlessly integrated into Complementary Metal Oxide Semiconductor (CMOS) processes, paving the way for large-scale, high-volume manufacturing of advanced devices.

Methods

Sample fabrication

Initially, a 200 nm thick layer of Au was deposited onto an SCG film utilizing thermal evaporation technology. Subsequently, a layer of photoresist was applied to the Au coating through spin-coating, followed by exposure and development of the photoresist. The processed sample was then immersed in an Au removal solution, facilitating the etching away of the excess Au layer. The final step involved thoroughly washing away the photoresist from the surface.

Friction measurements using AFM system

The sliding friction force at the Au/SCG interface was meticulously measured using AFM technology, specifically with NT-MDT equipment. For the experiments, an AFM tip was precisely applied against the Au mesa to initiate a reciprocating motion. This motion was conducted at a frequency of 1 Hz and an amplitude of 1 μm. The measurement of the friction force was achieved by calculating the difference between the lateral forces observed during the reciprocating motion. Following the initial running-in phase, which encompasses approximately 500 cycles, the applied normal pressure was methodically adjusted every 20 cycles. This adjustment aimed to establish a detailed correlation curve between the friction force and the applied normal pressure, enhancing our understanding of the interface’s mechanical behavior.

Raman spectroscopy mapping characterization

Upon finalizing the contact resistance measurements at the Au/SCG interface, the Au mesa was laterally displaced to revealed the underlying SCG surface at the point of contact. To assess surface defects on the SCG after conduction and sliding tests, a Raman characterization (WITec Raman spectra, Ulm, Germany equipped with an alpha 300RA microscope using ZEISS 100x/0.85 objective, a 532 nm laser, a 300 gr/mm grating and a charge-coupled device cooled down to −60 °C. The laser power is set 6 mW and the Raman image of the sample was measured with a 300 nm step size and an acquisition time of 1–5 s. The acquired images were analyzed using a WITec Control/Project 5.3.) was performed on the surface. We specifically focused on the observation of the D peak at 1350 cm⁻¹ as a definitive marker for the presence of defects on the SCG surface. This approach allowed for a precise and detailed analysis of the surface integrity following electrical conduction.

Morphology characterization using AFM system

We used the semi-contact mode of atomic force microscope (AFM, Cypher S-Oxford Instruments) to conduct roughness characterization on the Au contact of the MEMS switches.

Other characterization

SEM images were obtained using an FEI Thermo Scientific Quattro S scanning electron microscope. EBSD characterizations were carried out using an Oxford NordlysMax3 instrument.

Molecular dynamics

The edge-free MD simulation model consists of the graphene substrate and a Au(111) slider aligned with the substrate. The in-plane size of the system is L_x = 4.9045 nm and L_y = 5.9964 nm. The resulting mismatch strain between slider and substrate, ε_xx and ε_yy, are both smaller than 0.5%. Interatomic interaction of Au-Au and Au-C were described by Foiles EAM⁵⁶ and registry-dependent SAIP force fields⁵⁷. The Interlayer and intralayer interaction for graphene substrate were described by Kolmogorov–Crespi and REBO force field respectively^58,59. Stiffnesses of the spring used in simulations were k_z = 2.896 N/m and K_p = 100 N/m. All simulations were performed with open-source code LAMMPS ^60,61.

Prior to sliding simulation, the system was firstly raised up to a high temperature T_h by Nosé-Hoover thermostat to mimic the thermal effect from current. According to the multi-physics FEM simulation, the current density J from 0 to 51 GA/m² corresponds to temperature from room temperature to ≈1500 K. A series of simulations for different T_h were performed starting from the same initial Au/Gra heterostructure. For each case, the system was firstly heated from room temperature to T_h within 0.1 ns. After maintaining the temperature at T_h for thermal relaxation, the annealing was performed within another 0.1 ns. The obtained structures were used for following sliding simulations. Direct MD simulations of kinetic friction generally require sliding velocities on the order of m/s because of the limited simulation timescale, several orders of magnitude higher than the μm/s velocities used experimentally. We therefore adopted a quasi-static sliding protocol to determine the static depinning stress and assess whether the interface remains structurally lubric. The simulation protocol follows our previous studies ^29,62.

Conflict of interest

The authors declare no conflict of interest.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41378-026-01455-2.