We demonstrate how machine-learning approaches can significantly speed up the way materials are characterized and designed at their molecular scale. Using a multi-level computational approach, we delineate key structural features in metalorganic frameworks (MOFs) that influence their mechanical properties. Importantly, we highlight the strength of artificial neural networks in producing MOFs with mechanical properties in a matter of seconds without the need for complex and time-consuming calculations or experiments. The results guide MOF researchers to assess and design structures with improved mechanical stability.
Nanoporous single-layer graphene is promising as an ideal membrane because of its extreme thinness, chemical resistance, and mechanical strength, provided that selective nanopores are successfully incorporated. However, screening and understanding the transport characteristics of the large number of possible pores in graphene are limited by the high computational requirements of molecular dynamics (MD) simulations and the difficulty in experimentally characterizing pores of known structures. MD simulations cannot readily simulate the large number of pores that are encountered in actual membranes to predict transport, and given the huge variety of possible pores, it is hard to narrow down which pores to simulate. Here, we report alternative routes to rapidly screen molecules and nanopores with negligible computational requirement to shortlist selective nanopore candidates. Through the 3D representation and visualization of the pores’ and molecules’ atoms with their van der Waals radii using open-source software, we could identify suitable C-passivated nanopores for both gas- and liquid-phase separation while accounting for the pore and molecule shapes. The method was validated by simulations reported in the literature and was applied to study the mass transport behavior across a given distribution of nanopores. We also designed a second method that accounts for Lennard-Jones and electrostatic interactions between atoms to screen selective non-C-passivated nanopores for gas separations. Overall, these visualization methods can reduce the computational requirements for pore screening and speed up selective pore identification for subsequent detailed MD simulations and guide the experimental design and interpretation of transport measurements in nanoporous atomically thin membranes.
The regulation of mass transfer across membranes is central to a wide spectrum of applications. Despite numerous examples of stimuli-responsive membranes for liquid-phase species, this goal remains elusive for gaseous molecules. We describe a previously unexplored gas gating mechanism driven by reversible electrochemical metal deposition/dissolution on a conductive membrane, which can continuously modulate the interfacial gas permeability over two orders of magnitude with high efficiency and short response time. The gating mechanism involves neither moving parts nor dead volume and can therefore enable various engineering processes. An electrochemically mediated carbon dioxide concentrator demonstrates proof of concept by integrating the gating membranes with redox-active sorbents, where gating effectively prevented the cross-talk between feed and product gas streams for high-efficiency, directional carbon dioxide pumping. We anticipate our concept of dynamically regulating transport at gas-liquid interfaces to broadly inspire systems in fields of gas separation, miniaturized devices, multiphase reactors, and beyond.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.