ABSTRACT
Metal-organic frameworks (MOFs) are highly tuneable, extended-network, crystalline, nanoporous materials with applications in gas storage, separations, and sensing. We review how molecular models and simulations of gas adsorption in MOFs have informed the discovery of performant MOFs for methane, hydrogen, and oxygen storage, xenon, carbon dioxide, and chemical warfare agent capture, and xylene enrichment. Particularly, we highlight how large, open databases of MOF crystal structures, post-processed to enable molecular simulations, are a platform for computational materials discovery. We discuss how to orient research efforts to routinise the computational discovery of MOFs for adsorption-based engineering applications.
Acknowledgements
A. S. and C. M. S. thank the School of Chemical, Biological, and Environment Engineering (CBEE) at Oregon State University for start-up funds. A. R. K. and C. E. W. gratefully acknowledge support from the National Science Foundation (NSF CAREER award CBET-1653375).
Disclosure statement
The authors declare the following competing financial interest(s): Christopher E. Wilmer is a co-founder of, and has a financial interest in, the start-up company NuMat Technologies, which is seeking to commercialize metal-organic frameworks. No potential conflict of interest was reported by the other authors.
Notes
* Official contribution of the National Institute of Standards and Technology; not subject to copyright in the United States. Certain commercially available items may be identified in this paper. This identification does not imply recommendation by NIST, nor does it imply that it is the best available for the purposes described.
1 In addition to thermal and chemical stability and the cost of the MOF, several other material properties influence the performance of a MOF in adsorbed natural gas storage, such as heat of adsorption, specific heat, thermal conductivity, diffusion coefficients, and adsorption of impurities that can ‘poison’ adsorption sites (note we approximated natural gas as methane).
2 We specify the MOF to be existing, as opposed to hypothetical or conceived, according to whether synthesis protocols have already been reported in the literature. Knowledge of these protocols, as well as activation procedures and information about stability, can expedite deployment.
3 That said, under the application of transition state theory, kinetic MC algorithms can simulate the diffusion of gases in MOFs [Citation74].
4 Typically, as illustrated in our survey in Section 6, some human judgment on e.g. ease of synthesis is also exercised in further shortlisting materials.
5 Consider a MOF immersed in a bath of gas at chemical potential μ and temperature T. Absolute adsorption is the number of adsorbate molecules in the MOF. Excess adsorption is the absolute adsorption minus the number of adsorbate molecules present in a volume of the bulk gas phase at chemical potential μ and temperature T, where
is the accessible pore volume offered by the MOF. While absolute adsorption is directly obtained in molecular simulation, excess adsorption is more directly obtained in experimental gas adsorption measurements [Citation167,Citation168].
6 The term ‘CoRE MOF’ was coined by Prof. David Sholl, who wrote a number of different combinations of words on a napkin during the Nanoporous Materials Genome Center Meeting (2013, Berkeley, CA).
10 Though, notably, DFT calculations were used as an energetic description of small gas molecules in MOF-74 to compute Henry coefficients via Widom insertions [Citation338] by biasing the samples towards low-energy regions. Similarly, Fetisov et al. [Citation339] conducted the first principle Monte Carlo simulations of CO, N
, and H
O adsorption in Mg-MOF-74. To avoid wastefully devoting DFT calculations to high-energy trial configurations, the authors employed (i) a configurational-bias Monte Carlo algorithm and (ii) a cheaper, approximate potential for pre-sampling configurations. Chen et al. [Citation340] simulated CH
adsorption in CuBTC using ab initio calculations on a grid to characterise the CuBTC-CH
interaction.
11 The following thought experimental clarifies why more data is needed to fit a neural network to the PES surface than to fit a traditional force field with an interatomic potential imposed. Assume that the Lennard-Jones potential is the ground truth for an interaction between two atoms of A. Then, two independent data points, i.e. the potential energy at two distances, is enough to determine the 12-6 Lennard-Jones σ and ε. In contrast, the neural network would need many more data points to learn the 12-6 scaling with interatomic distance.