Quantifying the Uncertainty of Force Field Selection on Adsorption Predictions in MOFs.
J Chem Theory Comput
; 20(11): 4869-4884, 2024 Jun 11.
Article
en En
| MEDLINE
| ID: mdl-38818701
ABSTRACT
Comparisons between simulated and experimental adsorption isotherms in MOFs are fraught with challenges. On the experimental side, there is significant variation between isotherms measured on the same system, with a significant percentage (â¼20%) of published data being considered outliers. On the simulation side, force fields are often chosen "off-the-shelf" with little or no validation. The effect of this choice on the reliability of simulated adsorption predictions has not yet been rigorously quantified. In this work, we fill this gap by systematically quantifying the uncertainty arising from force field selection on adsorption isotherm predictions. We choose methane adsorption, where electrostatic interactions are negligible, to independently study the effect of the framework Lennard-Jones parameters on a series of prototypical materials that represent the most widely studied MOF "families". Using this information, we compute an adsorption "consensus isotherm" from simulations, including a quantification of uncertainty, and compare it against a manually curated set of experimental data from the literature. By considering many experimental isotherms measured by different groups and eliminating outliers in the data using statistical analysis, we conduct a rigorous comparison that avoids the pitfalls of the standard approach of comparing simulation predictions to a single experimental data set. Our results show that (1) the uncertainty in simulated isotherms can be as large as 15% and (2) standard force fields can provide reliable predictions for some systems but can fail dramatically for others, highlighting systematic shortcomings in those models. Based on this, we offer recommendations for future simulation studies of adsorption, including high-throughput computational screening of MOFs.
Texto completo:
1
Colección:
01-internacional
Banco de datos:
MEDLINE
Idioma:
En
Revista:
J Chem Theory Comput
Año:
2024
Tipo del documento:
Article
País de afiliación:
Reino Unido