Zigzag pore based molecular simulation on the separation of CO₂/CH₄ mixture by carbon membrane

Abstract

A zigzag-type pore structure was proposed for the separation of a CO 2 /CH 4 binary mixture by carbon molecular sieve membrane (CMSM). A defect-free pore and three kinds of defect pores, i.e., random, uniform, and partial defect, were considered to improve the separation performance. Adsorption and diffusion behaviours of pure gases and a gas mixture were simulated by the grand canonical Monte Carlo (GCMC) method and non-equilibrium molecular dynamics (NEMD) method, respectively. The pressure ranged from 10-100 kPa and the temperature was between 273-348 K during simulation. The calculated isotherms of CO 2 and CH 4 showed an acceptable agreement with the experimental data for the defect-free pores of 0.67 nm in size at 298 K. Examination on the effect of pore sizes showed that 0.67 nm was the appropriate option for separation. The total selectivity was 20.1 at 298 K and 100 kPa with the pore size of 0.67 nm, which is consistent with the experimental value. The adsorption was determined as the dominant separation mechanism between adsorption and diffusion. Compared with the defect-free pores, the introduction to the random and uniform defect pores can improve the total selectivities and the random defect provided superior separation performance. An appropriately low temperature and small pore size were beneficial to the separation.

Analysis of Influencing Factors on the Gas Separation Performance of Carbon Molecular Sieve Membrane Using Machine Learning Technique

“…Exploring the influence of factors, such as the microstructural characteristics and gas properties, on the gas separation performance, or the so-called permselectivity, would be helpful to the adjustment of the microstructure and the optimization to the membrane preparation. However, gas permeability of CMS membrane is hard to correlate functionally with the plural factors which are not directly interrelated [9][10][11]. The machine learning technique, which is the core of artificial intelligence, makes it possible to realize the correlation [12,13].…”