Macroscale superlubricity enabled by graphene nanoscroll formation

See allHide authors and affiliations

Science  05 Jun 2015:
Vol. 348, Issue 6239, pp. 1118-1122
DOI: 10.1126/science.1262024

Slip sliding away

Many applications would benefit from ultralow friction conditions to minimize wear on the moving parts such as in hard drives or engines. On the very small scale, ultralow friction has been observed with graphite as a lubricant. Berman et al. achieved superlubricity using graphene in combination with crystalline diamond nanoparticles and diamondlike carbon (see the Perspective by Hone and Carpick). Simulations showed that sliding of the graphene patches around the tiny nanodiamond particles led to nanoscrolls with reduced contact area that slide easily against the amorphous diamondlike carbon surface.

Science, this issue p. 1118; see also p. 1087


Friction and wear remain as the primary modes of mechanical energy dissipation in moving mechanical assemblies; thus, it is desirable to minimize friction in a number of applications. We demonstrate that superlubricity can be realized at engineering scale when graphene is used in combination with nanodiamond particles and diamondlike carbon (DLC). Macroscopic superlubricity originates because graphene patches at a sliding interface wrap around nanodiamonds to form nanoscrolls with reduced contact area that slide against the DLC surface, achieving an incommensurate contact and substantially reduced coefficient of friction (~0.004). Atomistic simulations elucidate the overall mechanism and mesoscopic link bridging the nanoscale mechanics and macroscopic experimental observations.

Macroscopic friction and wear remain the primary modes of mechanical energy dissipation in moving mechanical assemblies such as pumps, compressors, and turbines, leading to unwanted material loss and wasted energy. It is estimated that nearly one third of the fuel used in automobiles is spent to overcome friction, while wear limits mechanical component life. Even a modest 20% reduction in friction can substantially affect cost economics in terms of energy savings and environmental benefits (1). In that context, superlubricity is desirable for various applications and therefore is an active area of research. To date, superlubricity has been primarily realized in a limited number of experiments involving atomically smooth and perfectly crystalline materials (25) and supported by theoretical studies (6, 7). Superlubricity has been demonstrated for highly oriented pyrolytic graphite (HOPG) surfaces (8), as well as for multiwalled carbon nanotubes (MWCNTs), when the conditions for incommensurate contacts are met in a dry environment (9). Because these conditions are due to the incommensurability of lattice planes sliding against each other, they are referred to as structural lubricity and restricted to material interactions at the nanoscale. At the macroscale, this structural effect (hence, superlubricity) is lost because of the structural imperfections and disorder caused by many defects and deformations.

Low friction has recently been observed in centimeter-long double-walled carbon nanotubes with perfect atomic structures and long periodicity (10). Ultralow friction in disordered solid interfaces, such as self-mated DLC films (1114) and in fullerenelike nanoparticles such as molybdenum disulfide (MoS2) (15), has been observed under specific environmental and sliding conditions. However, the exact superlubricity mechanism in the above cases is still debatable and is not realized for industrial applications. In recent studies at the nano- and macroscale, graphene has shown a potential to substantially lower friction (1618) and wear (1921) under specific conditions. However, sustained macroscale superlubricity, particularly at engineering scales, has yet to be demonstrated.

We demonstrate our observation of stable macroscale superlubricity while sliding a graphene-coated surface against a DLC-coated counterface. Our initial assumption was that the random network of mixed sp3/sp2 bonded carbon in DLC might provide the perfect incommensurate surface needed for the ordered graphene flakes to slide against DLC with least resistance. This was indeed proved to be true; however, the coefficient of friction (COF) values for graphene sliding against DLC in a dry environment were not in the superlubric regime (COF ~ 0.04, as shown in Fig. 1B). Initial observation of the wear debris revealed formation of graphene nanoscrolls in the wear track. This prompted us to use nanodiamond as an additive, which may act as nano ball bearings when covered by graphene, providing extra mechanical strength. We saw a dramatic reduction in friction, reaching the superlubric state [in Fig. 1B and inset, the COF dropped to near zero (0.004)] in a dry environment, when we introduced nanodiamond in combination with few-layer (three to four layers) graphene flakes on the silicon dioxide (SiO2) substrate by means of a solution process method (figs. S1 and S2), providing a partial coverage on the SiO2 surface (22). The observed wear marks on the flat (Fig. 1C) and ball sides were minimal and produced primarily by the contact pressure during sliding tests (fig. S3). Raman analysis of the wear track (Fig. 1C, inset) showed modification of graphene inside the wear track as observed in a decreased 2D peak (at ~2660 cm−1) and an increased D peak (at ~1330 cm−1) in comparison with the initial graphene’s Raman signature, indicating a gradual loss of crystallinity of few-layer graphene and an increase in defects possibly due to tearing of graphene under constant sliding at high contact pressure (0.3 GPa). The Raman spectrum indicates no DLC transfer in the wear track during sliding, thus confirming that the superlubricity regime is not connected with the previously observed low-friction performance of DLC against DLC (fig. S14) (12).

Fig. 1 Experimental demonstration of the superlubricity regime.

(A) Not-to-scale schematic of the superlubricity test. (B) The COF for DLC ball sliding in a dry nitrogen environment against (i) graphene-plus-nanodiamonds (superlubricity state with COF ~ 0.004 ± 0.002), (ii) graphene alone (COF ~ 0.04 ± 0.01), and (iii) nanodiamond alone (COF ~ 0.07 ± 0.01). (Inset) A plot for superlubricity. (C) In the case of superlubricity, the wear tracks on the flat side and on the ball side (fig. S3) are almost invisible. (Inset) A typical Raman signature of defective graphene. (D and E) For graphene-plus-nanodiamond sliding against a DLC ball in a humid environment, (D) the COF reveals a high value of ~0.27 ± 0.04, (E) the corresponding wear track on the flat side is wide, and the inset shows a Raman signature corresponding to graphitized carbon debris. The tests were performed at room temperature under 1 N load and with 3 cm/s linear speed.

The necessity of using graphene-plus-nanodiamonds in establishing the superlubricity is demonstrated in Fig. 1B. In particular, graphene or nanodiamond when used alone on a SiO2 substrate sliding against a DLC ball in a dry environment displays higher values of COF (0.04 and 0.07). Additionally, the erratic nature of COF indicates large wear debris formation.

Our experimental studies confirm that the stable superlubricity regime occurs over a wide range of test conditions; when the load was changed from 0.5 to 3 N, velocity was varied from 0.6 to 25 cm/s, temperature increased from 20°C to 50°C (fig. S15), and the substrate was changed to nickel or bare silicon (fig. S16). The temperature and velocity range for maintaining stable superlubricity is further backed by theoretical simulations (tables S2 and S3).

For graphene-plus-nanodiamonds in an ambient humid environment (relative humidity ~30%), both COF and wear were comparatively large (Fig. 1, D and E). A substantial amount of graphitized carbon debris was formed in the wear track, as shown by the optical images and Raman data (Fig. 1E and fig. S3), and the substrate itself suffered from substantial wear during sliding. The distinctiveness of the tribopair and dramatic dependence on the environmental conditions led us to further explore the underlyng mechanism for the observed superlubricity.

We carried out more detailed analysis of the wear track that formed during the superlubricity regime in dry nitrogen by sampling and examining the wear debris with transmission electron microscopy (TEM). As shown in the TEM images in Fig. 2, a large fraction of the nanodiamonds were wrapped by graphene nanoscrolls (more detailed scroll images are shown in fig. S4). Electron energy-loss spectra (EELS) confirmed the presence of diamond in the wear debris, as evident from the typical EELS signature for diamond. The π* peak (at ~285 eV) in the carbon K-edge represents a small fraction of sp2 bonded carbon owing to the presence of the few layers of graphene wrapped around the nanodiamond, which is similar to the disordered carbon shell observed previously in detonated nanodiamonds (23). For pure nanodiamonds, this π* peak should be absent (23, 24), whereas in case of pure graphene scrolls, this π* peak is higher (Fig. 2B, inset). Because of the random orientation of scrolls with diamond embedded inside, we had to focus the TEM differently in order to view clearly the diamond lattice and graphene layers; therefore, some of the scrolls in Fig. 2 do not show nanodiamonds inside.

Fig. 2 Graphene nanoscrolls formation.

(A and B) TEM images of the wear debris for DLC ball sliding against graphene-plus-nanodiamonds, demonstrating superlow friction in the dry environment. Graphene scroll formation is observed. (Insets) EELS for both diamond and graphene in the wear debris.

To further explore the superlubricity mechanism, we performed molecular dynamics (MD) simulations (table S1) (22), and our simulations suggest that nanodiamonds can activate, guide, and stabilize the scrolling of initially planar graphene patches (fig. S5). During sliding in a dry environment, nanodiamonds facilitate scroll formation via two mechanisms: (i) Graphene platelets are highly reactive and easily attach to the dangling bonds present on the surface of nanodiamonds, initiating the scroll formation; and (ii) the sliding graphene patches encounter the three-dimensional (3D) structure of nanodiamonds, which act as obstructions (fig. S9). Additionally, the presence of topological defects (such as double vacancies or Stone-Wales) in graphene is expected to promote the scrolling behavior (fig. S12). On the basis of relative binding energetics between graphene-DLC and graphene-nanodiamond, we found that graphene prefers to wrap around the nanodiamond to promote higher surface contact (fig. S10). Once in a scrolled state, the final structures of graphene on diamond are well coordinated and stabilized by van der Waals forces.

Scroll formation and evolution of the COF are shown in Fig. 3, A and B, respectively, for a single graphene patch in a dry environment. At time t < 1.0 ns, COF values are high, ~0.2 to 0.4, because the graphene patch is in an extended or unscrolled state. The wrapping of a graphene sheet over the nanodiamonds begins at ~1.5 ns; this coincides with COF values dropping substantially, leading to a superlubric state that is maintained until the end of the simulation. Once formed, these scrolls slide against randomly arranged DLC atoms, which provide an incommensurate contact. This constant out-of-registry sliding translates into a superlubric regime. The COF also depends on the contact area between formed graphene scrolls and DLC. The superlubricity is thus attributed to (i) reduction in the interfacial contact area (>65%) and (ii) incommensurability between DLC and graphene scrolls.

Fig. 3 Simulations of the single-scroll formation.

(A to D) Temporal evolution of nanoscale friction for DLC ball sliding against graphene-plus-nanodiamonds in [(A) and (B)] dry and [(C) and (D)] humid environments extracted based on the MD simulation trajectories. Graphene scroll formation over nanodiamonds is observed in a dry environment (the steady-state COF is 0.005 ± 0.004), whereas ordered water layers above the graphene flakes prevent scroll formation in the humid environment (COF is 0.12 ± 0.04).

At the molecular level, the observed superlubricity has its origin in graphene’s nanoscopic anisotropic crystal structure, which consists of strong covalent intralayer bonding and weaker dispersive interlayer interactions. The structural contact between an incommensurate DLC ball and the graphene scrolls allows DLC to slide on top of the underlying graphene sheets by overcoming relatively small energetic barriers. In recent experiments, Dienwiebel et al. (8) observed friction reduction to vanishingly small values, depending on the degree of commensurability between the graphene flakes and the extended graphite surface. In an incommensurate state, the unit-cells in contact have to overcome much smaller barriers at any point in time, leading to considerably reduced resistance toward sliding. Consistent with the prediction of Mo et al. (25), we found that the friction force depends linearly on the number of atoms that chemically interact across the contact. The effective contact area between the graphene sheets and DLC decreases with time upon scrolling. Because friction is controlled by the short-range interactions even in the presence of dispersive forces, scrolling-induced reduction in nanoscopic contact is substantial enough to lead to a superlubric state.

Our experiments suggest that the humid environment increases the friction and wear of the ball side because graphene layers remain strongly attached on the surface. We therefore performed MD simulations of the DLC-nanodiamond-graphene system in a humid environment (fig. S6). MD trajectories suggest formation of quasi-2D ordered water layers between the nanoscopic contacts, the DLC and graphene sheets (Fig. 3, C and D). These water layers prevent the scrolling of the graphene during sliding (fig. S7), and the ordered 2D water layers [based on calculated translational and tetrahedral orders (supplementary text) (22)] present a constant energy barrier for the DLC to overcome. These two effects result in little or no friction variation over time (Fig. 3D), and a nearly constant high-friction condition is maintained (COF ~ 0.1). We have simulated the effects of surface chemistry and considered the role of defects (supplementary text) (22). We found that the presence of defects greatly facilitates the adsorption of water from the ambient atmosphere (fig. S13). Water preferentially adsorbs and stabilizes defective sites, which further prevents the formation of scrolls.

To bridge the gap between the nanoscale mechanics and macroscopic contacts evident in our experiments, we performed a large-scale MD simulation for an ensemble of graphene-plus-nanodiamonds present between DLC and the underlying multilayered graphene substrate (fig. S8). The mesoscopic link is crucial to explain how the formation of nanoscrolls translates from a nanosystem with a single graphene patch (square-nanometer area of sliding interface) into the observed superlubricity at the macroscale (square-millimeter area). We evaluated the collective scrolling and tribological behavior of many individual graphene patches and created a density distribution of their tribological state in order to assess their contribution to the observed friction. During the initial sliding period at t = 0, the unscrolled graphene patches are in close contact with the interface. The contact area normalized with respect to the initial value at t = 0 is ~1 (22), as shown in Fig. 4C. The density distribution of COF values (Fig. 4B) shows a narrow distribution with a peak at ~0.6 to 0.7, suggesting that the system is in a high-friction state. With time (200 to 300 ps), the graphene patches increasingly scroll over nanodiamonds, and we observe a corresponding reduction in this peak intensity. The density profile shows a broader distribution and shifts prominently toward lower COFs (<0.2). The contact area, which is proportional to the number of interacting atoms, reduces by 40 to 50% during (26) this period. During the latter stages (~500 ps), most of the graphene patches are scrolled. The density profile shows a shift in the distribution to COF values <<0.01. The effective contact area in the present case is reduced significantly, by ~65 to 70%, and the mesoscopic system has reached a superlubric state.

Fig. 4 Mesoscale MD simulations of superlubricity.

Mesoscale MD simulations demonstrating the time evolution in the distribution of COF values. (A) Snapshot showing the scroll formation on nanodiamonds for an ensemble of graphene patches when subjected to sliding. (B) Temporal evolution of COF distribution averaged over an ensemble of graphene patches. (C) Evolution of the corresponding contact area. Initially at t = 0 ps, the patches are mostly sheetlike and in close contact with DLC, leading to an average COF of ~0.6 to 0.7. Sliding of DLC increases the probability of scroll formation by graphene patches, leading to a decrease in the average contact area, which manifests in the form of macroscopic superlubricity. The ensemble-averaged COF shifts to superlubric values at t = 500 ps, when most of the graphene patches are in a scrolled state.

The tribological evolution of a single graphene patch at the nanoscale resembles that of a single asperity contact, whereas the mesoscopic behavior resembles a multiple asperity contact. The friction mechanism at the mesoscale for an ensemble of graphene patches is not different from nanoscale (single patch). The initial tribological state of the patches, as well as the configuration of the patches versus nanodiamonds, dictates the dynamics of scroll formation, which in turn affects the dynamical evolution of COF for the mesoscopic system. The macroscopic contact in our experiments can be envisioned as comprising a much larger number of such smaller contacts or asperities, which explains the difference in time for the onset of the superlubric state in the experiments versus simulated systems.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S16

Tables S1 to S3

References (2745)

Movie S1

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: The help in the TEM data collection by Y. Liu is greatly appreciated. Use of the Center for Nanoscale Materials was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under contract DE-AC02-06CH11357. This research used resources of the National Energy Research Scientific Computing Center, which is supported by the Office of Science of the U.S. Department of Energy under contract DE-AC02-05CH11231. This research used tribological test facilities of the Energy Systems Division supported by the Vehicle Technologies Program of the Office of Energy Efficiency and Renewable Energy of the U.S. Department of Energy under contract DE-AC02-06CH11357. An award of computer time was provided by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) program. This research used resources of the Argonne Leadership Computing Facility at Argonne National Laboratory, which is supported by the Office of Science of the U.S. Department of Energy under contract DE-AC02-06CH11357. Experimental data and simulations are archived on servers at Argonne National Laboratory. Part of the experimental results are covered by a patent (US20140023864A1). Both D.B. and S.D. contributed equally in this work. D.B. performed the experiments and analyzed the data. S.D. and S.K.R.S. devised and performed the molecular dynamics simulations and performed all the related data analysis. A.V.S. conceived the idea, helped in the data analysis of experimental results, and directed the project. A.E. codirected the project and helped in the data analysis of tribological tests. S.K.R.S. guided the simulation effort. D.B., S.D., S.K.R.S., A.E., and A.V.S equally contributed to discussing the results and composing the manuscript.
View Abstract

Stay Connected to Science

Navigate This Article