Covalent organic frameworks (COFs) are penetrated with uniform and ordered nanopores, implying their great potential in molecular/ion separations. As an imine-linked, stable COF, TpPa-1 is receiving tremendous interest for molecular sieving membranes. Theoretically, atomically thin TpPa-1 monolayers exhibit extremely high water permeance but unfortunately no rejection to ions because of its large pore size (∼1.58 nm). The COF monolayers tend to stack to form laminated multilayers, but how this stacking influences water transport and ion rejections remains unknown. Herein, we investigate the transport behavior of water and salt ions through multilayered TpPa-1 COFs by nonequilibrium molecular dynamics simulations. By analyzing both the interfacial and interior resistance for water transport, we reveal that with rising stacking number of COF multilayers exhibit increasing ion rejections at the expense of water permeance. More importantly, stacking in the offset eclipsed fashion significantly reduces the equivalent pore size of COF multilayers to 0.89 nm, and ion rejection is correspondingly increased. Remarkably, 25 COF monolayers stacked in this fashion give 100% MgCl2 rejection, whereas water permeance remains 1 to 2 orders of magnitude higher than that of commercial nanofiltration membranes. This work demonstrates the rational design of fast membranes for desalination by tailoring stacking number and fashion of the COF monolayers.

It is generally considered that ion rejection of a desalination membrane is independent of the operation pressure drops (ΔPs), which is typically not higher than 10 MPa. However, this may not be true for pressures as high as hundreds of megapascals usually used in simulations. Therefore, simulation results of high ΔPs cannot be directly used to predict real-world ion rejections, which is often overlooked. Herein, we investigate the ion rejection of carbon nanotube membranes in a large scale of ΔPs via nonequilibrium molecular dynamics simulations. With effective pressure drops (ΔP e 's) increased from 2.85 to 996 MPa, the ion rejection drops from 100% to nearly zero. Rather than directly investigating the rejection, the relationships of ion and water fluxes with ΔPs are separately investigated. With rising ΔP e s, the water flux increases linearly, while the ion flux undergoes a two-stage increase: first, an exponential increase at ΔP e ≤ 53.4 MPa and then a linear increase. An equation describing the ΔP e -dependent ion rejection is then developed based on these observations. Moreover, the rejection mechanism is also discovered, which indicates that the enhanced input energy makes ions easier to overcome the energy barrier rather than the molecular-configurational reasons. These findings are expected to fill the big gaps between simulations and experiments and may also be helpful for the rational design of the next-generation desalination membranes.

Carbon nanotube (CNT) membranes have long been considered as next-generation membranes due to superfast water transport inside tubes. However, a large pressure loss occurs at the pore mouth, and consequently water transport through the whole tubes is significantly retarded. To find out the reason behind this, we conduct systematic non-equilibrium molecular dynamics (NEMD) simulations on water transport through CNT membranes with various tube diameters and lengths. The whole transport resistance is contributed by the interfacial and interior parts, and the interfacial contribution plays a dominating role in short tubes and only can be ignored when the tube length reaches a scale of several micrometers. With regard to the origin of the interfacial resistance, the hydrogen bonding rearrangement (HBR) effect accounts for at least 45%, and the rest is attributed to the geometrical or steric crowding of water molecules near the pore mouth. To reduce the dominant interfacial resistance, we change the shape of the pore mouth from plate to hourglass by mimicking the aquaporin water channels. The interfacial resistance is thus decreased by >27%. It is also found that the reduction is originated from the optimized HBR rather than the subdued steric crowding of water molecules near the pore mouth.