Challenges in Molecular Dynamics of Amorphous ZIFs Using Reactive Force Fields

Abstract

While amorphous metal–organic frameworks form an emerging class of materials of growing interest, their structural characterization remains experimentally and computationally challenging. Out of the many molecular simulation methods that exist to model these disordered materials, one strategy consists in simulating the phase transition from a crystalline MOF to the amorphous state using molecular dynamics. ReaxFF reactive force fields have been proposed for this purpose in several studies to generate models of zeolitic imidazolate framework glasses by melt quenching. In this work, we investigate the accuracy and reliability of this approach by reproducing the published procedures and comparing the structure of the resulting glasses to other data, including ab initio modeling, in detail. We find that the in silico melt-quench procedure is extremely sensitive to the choice of methodology and parameters and suggest adaptations to improve the scheme. We also show that the glass models generated with ReaxFF are markedly different from their ab initio counterparts, as well as known experimental characteristics, and feature an unphysical description of the local coordination environment, which in turn affects the medium-range and bulk properties.

Computation of Finite Temperature Mechanical Properties of ZIF Glasses by Molecular Dynamics

“…Able to explore larger spatial and time scales, they could in principle be used to obtain the bulk modulus with several methods while reducing the finite size effects inherent to AIMD. However, we showed in a previous work the atypical structural properties of the glasses obtained with this approach, 40 and there is still no in-depth validation of the mechanical properties obtained with this force field in the existing literature. 15 In order to offer a meaningful comparison to other MD schemes, we consider several glass models: a ReaxFF glass obtained in our previous work, 40 the same three ab initio glass models studied in the previous section 43 and a glass obtained by Reverse Monte Carlo (RMC) modeling.…”

“…Due to the difficulty of performing (N, P, T ) ab initio simulations of soft porous crystals 40,68 and to the reduced computational cost of (N, V, T ) MD runs, we employed the finite strain difference method. All ab initio simulations were performed in the (N, V, T ) ensemble to enforce a volume V and compute the resulting pressure P (V ) of the equilibrated system.…”

“…We employed a parametrization of ReaxFF, a flavor of reactive force fields, 48 developed for ZIF materials, 49 and used the same parameters as in our previous work. 40 ReaxFF simulations were performed using LAMMPS. 50,51 A timestep δt of 0.25 fs was used in the MD runs and, the temperature (and pressure when applicable) were controlled using a Nosé-Hoover thermostat (and barostat).…”

“…Structural analyses were performed using the parameters used and further detailed in a previous study, 40 using the Python library aMOF.…”

“…43 All these models have been generated and equilibrated in the (N, V, T ) ensemble, and thus are all at the same density (taken to be the crystallographic density 1.22 g cm −3 ). Although this difficulty in capturing changes in density has been identified as an important limitation of the ab initio methodology, 25,40 such glass models still represent the most chemically accurate atomistic description of ZIF-4 glasses published to date in the literature. 15 Here, we have computed the bulk modulus K for four of these systems: the crystal and three melt-quenched glasses, chosen to preserve the diversity found in the coordination environments.…”