Hybrid molecular-continuum simulations of water flow through carbon nanotube membranes of realistic thickness

Ritos, Konstantinos and Borg, Matthew Karl and Lockerby, Duncan A. and Emerson, David and Reese, Jason (2015) Hybrid molecular-continuum simulations of water flow through carbon nanotube membranes of realistic thickness. Microfluidics and Nanofluidics, 19 (5). pp. 997-1010. ISSN 1613-4982 (https://doi.org/10.1007/s10404-015-1617-x)

[thumbnail of Ritos-etal-Microfluid-Nanofluid-2015-Hybrid-molecular-continuum-simulations-water-flow-through-carbon-nanotube-membranes-realistic-thickness]
Preview
Text. Filename: Ritos_etal_Microfluid_Nanofluid_2015_Hybrid_molecular_continuum_simulations_water_flow_through_carbon_nanotube_membranes_realistic_thickness.pdf
Final Published Version
License: Creative Commons Attribution 4.0 logo

Download (1MB)| Preview

Abstract

We present new hybrid molecular-continuum simulations of water flow through filtration membranes. The membranes consist of aligned carbon nanotubes (CNTs) of high aspect ratio, where the tube diameters are ~1–2 nm and the tube lengths (i.e. the membrane thicknesses) are 2–6 orders of magnitude larger than this. The flow in the CNTs is subcontinuum, meaning standard continuum fluid equations cannot adequately model the flow; also, full molecular dynamics (MD) simulations are too computationally expensive for modelling these membrane thicknesses. However, various degrees of scale separation in both time and space in this problem can be exploited by a multiscale method: we use the serial-network internal-flow multiscale method (SeN-IMM). Our results from this hybrid method compare very well with full MD simulations of flow cases up to a membrane thickness of 150 nm, beyond which any full MD simulation is computationally intractable. We proceed to use the SeN-IMM to predict the flow in membranes of thicknesses 150 nm–2 μm, and compare these results with both a modified Hagen–Poiseuille flow equation and experimental results for the same membrane configuration. We also find good agreement between experimental and our numerical results for a 1-mm-thick membrane made of CNTs with diameters around 1.1 nm. In this case, the hybrid simulation is orders of magnitude quicker than a full MD simulation would be.