Atomistic-scale simulations on graphene bending near a copper surface
Summary¶
ReaxFF reactive MD is used to study graphene bending in vacuum and on Cu surfaces, comparing two published ReaxFF parameter sets for graphene/Cu (including the Zhu 2020 set with Srinivasan 2015 carbon parameters cited in the paper). The authors report bending stiffness, binding of H and Cu adatoms, and the draping angle of graphene over Cu step edges; simulated draping angles agree with scanning-tunneling-microscopy-based experiments, supporting the newer parameterization for collector–anode interface modeling. The motivation is battery-relevant: current collectors and anodes often place graphene or carbon coatings next to Cu, where mechanical draping and adatom binding influence interfacial stability during cycling.
Methods¶
MD application (ReaxFF, ADF). The Catalysts article (Section 2) reports all main ReaxFF trajectories in ADF (not LAMMPS): velocity Verlet integration, time step 0.1 fs, and 3D periodic boundary conditions. Thermostat / barostat: Berendsen thermostat with a 100 fs coupling constant in NVT; in NPT the Berendsen barostat keeps the system at zero pressure with a 100 ps pressure coupling constant.
System size, protocol, and ensembles (from the same section). Free-standing graphene: ~8.3 × 5.8 nm with ~2000 C atoms—energy minimization, NPT at 5 K to show thermal rippling, then NPT annealing from 1500 K to 77 K at 1.2 K/ps and a 100 ps hold at 77 K (Section 2.1). Long ribbons: 20, 40, and 80 nm lengths, NPT relaxation at 300 K, heating to 1300 K or 2000 K at 10 K/ps with a 100 ps plateau, then NVT at 77 K (Section 2.2). Graphene on Cu (Zhu 2020 / Srinivasan 2015 C): rippled free-standing structure placed on a Cu surface; annealing compared before/after in Figure 3 (Section 2.3). Step edges (draping angle): a 40 nm ribbon on Cu with 6- vs 12-layer-high steps; NVT at 300 K for up to 500 ps to establish one- or two-sided contact with the terraces; angles extracted from the last 10 structures with ImageJ and compared to STM linecuts (Sections 2.4–2.5).
ReaxFF parameter lines compared include CHON-2019 (Srinivasan et al. 2015 C parameters in the CHO-2016 / CHON-2019 lineage) versus legacy CHO-2008 (Chenoweth et al.) for mechanical and hydrogenation tests, and the Zhu 2020 Cu–C field with Srinivasan C for Cu interfaces.
Blueprint line items. Boundaries / periodicity: 3D PBC in all cases as stated. Ensemble: NPT for isotropic or zero-pressure relaxations/anneals; NVT for confined 77 K stages and 300 K step–edge production. Timestep 0.1 fs; durations as in the subsections above. Temperature: 5, 77, 300, 1300, 1500, and 2000 K in the different protocols. Barostat — only in NPT steps (0 target pressure in those segments). NVT step-edge runs: N/A separate barostat. Pressure — N/A as an independent control in NVT step-edge segments. Electric field — N/A. Shock / shear / replica / metadynamics — N/A. Electrostatics / ReaxFF QEq — default ReaxFF/ADF treatment; standard ReaxFF charge training limits are discussed vs eReaxFF in the Introduction, not a separate numerical Ewald block.
Force-field training — N/A (uses published parameter sets; no de novo fit in this work).
Static QM / DFT — N/A for production MD; DFT from Yi et al. is comparison literature for hydrogenation (Section 2.5), not a separate pipeline here.
Findings¶
Outcomes and mechanisms. CHON-2019 reproduces the reference potential energy per atom for annealed ribbons better than CHO-2008 (Section 2.2). Placing rippled graphene on Cu with the Zhu 2020 set alters the annealed conformation relative to the free sheet because of strong Cu–C contact at bend regions (Figure 3). Draping angles from the 40 nm step-edge systems are about 28° ± 4° (6 Cu layers) and 30° ± 5° (12 layers); STM linecuts on CVD graphene on Cu give 32° ± 3°—taken in the paper as support for the ReaxFF Cu–C interface (Section 2.4–2.5). Hydrogenation and Cu-atom binding (Section 2.5, Tables 1–3): CHON-2019 aligns better with DFT (Yi et al.) for one-sided hydrogenation and with 4Hc/4Hb ordering; CHO-2008 is weaker on those distinctions.
Comparisons vs experiment / reference QM. ReaxFF vs DFT tables for H chemisorption; Zhu 2020+ReaxFF draping vs STM for step angles.
Sensitivity and levers. Ribbon length and heating path (1300 vs 2000 K) change ripple multiplicity; step height switches one- vs two-sided draping within the 500 ps NVT@300 K window; parameter set (CHON-2019 vs CHO-2008) shifts mechanical and H-binding metrics.
Limitations (as stated). The authors flag that ReaxFF is not trained for high-fidelity charge distributions and point to eReaxFF as a way to add explicit electronic response (Introduction).
Limitations¶
Reactive MD captures mechanics and binding trends but not quantitative electronic structure or charge transport across the interface. Step-edge geometries in experiment include substrate miscut and oxide contamination that are simplified in ideal Cu terraces; draping comparisons should therefore emphasize trends and relative parameter-set performance rather than sub-Å absolute agreement without surface reconstruction studies. Hydrogen adatom tests probe local binding; electrochemical environments may introduce additional adlayers not represented in vacuum benchmarks.
Relevance to group¶
Direct van Duin-group application of ReaxFF to battery-relevant graphene–Cu interfaces.
Citations and evidence anchors¶
- Catalysts 11(2), 208 (2021); DOI 10.3390/catal11020208 — Section 2 (simulation techniques) and results for bending/draping.