(482g) Large-Scale Free-Energy Calculations in Metal-Organic Framework Prototypes

The chemical and structural tunability of metal-organic frameworks (MOFs) can be exploited to alter how molecules adsorb, diffuse and react within their pore structure, making MOFs promising for applications in gas storage, chemical separations, and catalysis. However, MOF tunability also implies that there are millions of potential MOF variations, making it challenging to rapidly identify the best MOF for a given purpose. As a response to this challenge, computational high throughput screening emerged as a tool to rapidly identify promising MOF designs to be pursued by more focused (and slower) experimental efforts. This high throughput approach has already proven successful for the prediction of application-relevant adsorption properties [1] while efforts to calculate mechanical[2] and catalytic properties[3] are also starting to emerge. However, a significant obstacle that remains when progressing from the identification of a promising (in silico) MOF prototype to actually synthesizing it is the uncertainty whether the prototype is synthesizable.

Uncertainty in MOF prototype synthesizability arises due to an insufficient understanding of what makes a MOF thermodynamically and kinetically accessible. While the holistic view provided by high throughput screening of adsorption properties has provided valuable insights on what gives a MOF prototype high adsorption performance, similar insights on what makes a MOF prototype “synthesizable” have not been obtained due to the absence of large-scale efforts to calculate measures of thermodynamic stability, such as free energy, in MOF databases. This gap is perhaps due to the relative complexity of free energy calculations in porous materials. Here we present our efforts to automatize free energy calculations of MOF prototypes using the Frenkel-Ladd (FL) path method. We compare free energies calculated using the FL method to those calculated using the harmonic approximation (HA), which is a method often used to calculate crystal free energies due to the simplicity of its principles. However, the HA is a multistep process that includes the calculation of the phonon density of states (PDOS), making it a relatively involved approach not ideal for large-scale screening.

Based on our simulation automatization efforts, we find the FL method to be better suited for calculation of free-energies in large MOF databases. Furthermore, the FL method does capture the contribution of anharmonic vibrations to crystal free energy, which are important in some structures. Indeed, we show that the entropic contributions to free energy calculated using the HA method tend to match poorly with those calculated using the FL method, which could result in substantial inaccuracies when considering highly flexible MOFs. Finally, we compare the free energies of a set of MOFs calculated using three “off-the-shelf” force-fields (including the MOF-focused UFF4MOF) to gauge how much different force field parameterizations affect MOF stability predictions.

1. Gómez-Gualdrón, D. A.; Colón, Y. J.; Zhang, X.; Wang, T. C.; Chen, Y.-S.; Hupp, J. T.; Yildirim, T.; Farha, O. K.; Zhang, J.; Snurr, R. Q. Evaluating Topologically Diverse Metal–Organic Frameworks for Cryo-Adsorbed Hydrogen Storage. Energy Environ. Sci. 2016, 9 (10), 3279–3289.
2. Anderson, R.; Gómez-Gualdrón, D. A. Increasing Topological Diversity during Computational “Synthesis” of Porous Crystals: How and Why. CrystEngComm 2019, 21 (10), 1653–1665.
3. Rosen, A. S.; Notestein, J. M.; Snurr, R. Q. Structure–Activity Relationships That Identify Metal–Organic Framework Catalysts for Methane Activation. ACS Catal. 2019, 9 (4), 3576–3587.