[Paper Review] Poiseuille flow of liquid methane in nanoscopic graphite channels by molecular dynamics simulation

Martin Horsch, Jadran Vrabec+2 more|ArXiv.org|Nov 29, 2009

NMR spectroscopy and applications9 references3 citations

TL;DR

This study uses large-scale molecular dynamics simulations with the LS1/Mardyn code to investigate Poiseuille flow of liquid methane in nanoscopic graphite channels, demonstrating that Darcy’s law holds across the transition from nanoscale boundary-dominated flow to continuum behavior. The simulations reveal a sharp transition between slip-dominated and continuum regimes between 20 and 50 nm channel widths, with accurate scalability achieved on supercomputers for systems up to 4.8 million interaction sites.

ABSTRACT

MD simulations of methane confined between graphite walls with up to 4,800,000 interaction sites, i.e., carbon atoms and methane molecules, are conducted, where the channel width is varied to include both the boundary-dominated regime and the transition to the continuum regime. This proves that MD can be used today to cover the entire range of characteristic lengths for which continuum methods fail.

Motivation & Objective

• To investigate the hydrodynamic behavior of liquid methane in nanoscopic graphite channels using molecular dynamics.
• To assess the scalability of the LS1/Mardyn MD simulator for large-scale simulations of fluid–solid systems.
• To determine the transition regime between nanoscale slip-dominated flow and continuum-like flow in confined geometries.
• To validate the applicability of Darcy’s law across the nanoscale-to-continuum transition in graphite-confined methane.
• To enable large-scale MD simulations of fluid transport in carbon nanostructures using optimized parallel computing techniques.

Proposed method

• Employed the truncated and shifted Lennard-Jones (LJ/TS) potential to model methane–methane and methane–carbon interactions with parameters validated for methane thermodynamics.
• Used the Lorentz-Berthelot mixing rule to derive unlike interaction parameters between methane and graphite carbon atoms.
• Applied a rescaled Tersoff potential for graphite walls to correct bond length deviations from experimental values (1.421 Å vs. 1.461 Å).
• Implemented isotropic and channel geometry-based spatial domain decomposition in LS1/Mardyn to optimize load balancing and communication efficiency in heterogeneous systems.
• Used a PI controller to regulate external acceleration and maintain target flow velocity in the z-direction, simulating Poiseuille and Couette flow.
• Conducted simulations in the canonical ensemble with system sizes up to 4.8 million interaction sites, using supercomputers (cacau and HP XC6000).

Experimental results

Research questions

• RQ1 How does the flow behavior of liquid methane in graphite nanochannels transition from slip-dominated nanoscale flow to continuum-like behavior?
• RQ2 To what extent does Darcy’s law describe the pressure drop–velocity relationship in confined methane flow across varying channel widths?
• RQ3 What is the scalability of the LS1/Mardyn MD simulator for large-scale fluid–solid systems involving carbon nanostructures?
• RQ4 How do different spatial domain decomposition strategies (isotropic vs. channel-based) affect performance in heterogeneous MD simulations?
• RQ5 What is the critical channel width range where the transition between nanoscale and continuum regimes occurs for methane in graphite channels?

Key findings

• The transition from slip-dominated to continuum-like flow for liquid methane in graphite channels occurs in a narrow width range between 20 and 50 nm.
• Darcy’s law accurately describes the pressure drop–velocity relationship for methane flow in graphite channels across both the nanoscale and continuum regimes.
• Boundary slip velocities exceeded 99% of the mean flow velocity in channels below 2 nm, confirming strong non-continuum effects at the smallest scales.
• The LS1/Mardyn MD simulator demonstrated strong scalability on supercomputers, handling systems with up to 4.8 million interaction sites.
• Channel geometry-based domain decomposition outperformed isotropic decomposition for heterogeneous fluid–solid systems, improving load balancing and communication efficiency.
• The rescaled Tersoff potential for graphite walls improved structural accuracy by correcting the bond length from 1.461 Å to 1.421 Å, matching experimental values.