Atomistic understandings of reduced graphene oxide as an ultrathin-film nanoporous membrane for separations
Summary¶
Reduced graphene oxide (rGO) is often discussed as an ultrathin nanoporous membrane, but pore creation is tied to how oxygen leaves the graphene oxide (GO) lattice during thermal reduction. Lin and Grossman combine ReaxFF simulations in LAMMPS to model GO→rGO chemistry with explicit epoxy and hydroxyl functionality, then use classical (non-reactive) permeation MD—DREIDING with SPC/E water and TIP3P-style ions—to evaluate water desalination and CO₂/CH₄ separation through representative rGO pores. The two-stage workflow links atomic-scale defect chemistry to continuum-like transport metrics.
Methods¶
MD application — GO→rGO formation (ReaxFF in LAMMPS). GO sheets ~3.4 nm × 3.2 nm in-plane with random epoxy/hydroxyl decoration are placed in ~15 nm-tall 3D periodic cells so CO\(_2\) and H\(_2\)O desorbates can accumulate along ±z during reduction. ReaxFF structural optimization precedes thermal reduction: the cell is ramped from 10 K to the target reduction temperature at ~6 K/fs, then annealed ~400 ps with a Berendsen thermostat (damping = 100 MD steps) and temperature-dependent timesteps (0.1 / 0.08 / 0.05 / 0.05 fs for 1500 / 2000 / 2500 / 3000 K, respectively, as listed in Nature Communications Methods). By-products more than 5 nm away from the sheet along z are removed every 200 fs to mimic vacuum; a 0.175 nm bond cutoff culls leftover small molecules after reduction.
Parameter / morphology survey. Ten randomized GO samples per oxygen concentration (17 / 25 / 33%) × epoxy:hydroxyl ratio (2:1, 1:1, 1:2) combination are reduced at 1500–3000 K, yielding 360 rGO realizations for pore-statistics analysis as stated in the article.
Permeation MD (downstream classical FF). Desalination and CO\(_2\)/CH\(_4\) permeation use hybrid setups described in Methods: the rGO membrane is still treated with ReaxFF, while water uses SPC/E, ions use Joung–Cheatham-family parameters, and nonbonded rGO–water interactions use Lennard-Jones + Coulomb with DREIDING LJ parameters for C/O/H and Lorentz–Berthelot mixing—see the membrane/piston box description in the paper for geometry and driving-force details. Replica exchange / applied electric-field MD: N/A — not reported for these permeation workflows.
Ensembles and pressure. GO→rGO reduction: constant-volume cells with Berendsen thermostat (damping = 100 MD steps) during heating/annealing; NPT barostat is not reported for this stage. Permeation: a piston-style membrane cell with pressure control is described in Nature Communications Methods—read the article for GPa-level targets and box geometry rather than transcribing numbers from this note alone.
Findings¶
Pore dimensions correlate with total oxygen, the epoxy/hydroxyl balance, and the reduction temperature: higher oxygen load and epoxy-rich starting points tend to enable more carbon removal and larger pores under the simulated annealing windows. For transport, certain rGO realizations exhibit high permeance with selectivity favorable for desalination or gas separation in the classical MD tests; absolute numbers inherit force-field and pore statistics uncertainty, which the authors acknowledge when contrasting qualitative trends versus quantitative engineering predictions. The two-stage workflow—ReaxFF for pore genesis, classical FF for permeation—mirrors how membrane design often separates chemical defect formation from hydraulic performance estimation.
Limitations¶
Authors note qualitative structural predictions owing to ReaxFF accuracy and timescale limits; separation numbers depend on classical water/ion models. Membrane engineers should treat permeation metrics as illustrative unless validated against experimental pore distributions for specific rGO processing routes.
Relevance to group¶
Demonstrates ReaxFF-based GO→rGO modeling coupled to downstream membrane performance estimates.
Citations and evidence anchors¶
DOI 10.1038/ncomms9335.