DFT-Quality Adsorption Simulations in Metal-Organic Frameworks Enabled by Machine Learning Potentials

Reaching Quantum Accuracy in Predicting Adsorption Properties for Ethane/Ethene in Zeolitic Imidazolate Framework-8 at Low Pressure Regime

Nanoporous materials in the form of metal-organic frameworks such as zeolitic imidazolate framework-8 (ZIF-8) are promising membrane materials for the separation of hydrocarbon mixtures. To compute the adsorption isotherms in such adsorbents, grand canonical Monte Carlo simulations have proven to be very useful. The quality of these isotherms depends on the accuracy of adsorbate-adsorbent interactions, which are mostly described using force fields owing to their low computational cost. However, force field predictions of adsorption uptake often show discrepancies from experiments at low pressures, providing the need for methods that are more accurate. Hence, in this work, we propose and validate two novel methodologies for the ZIF-8/ethane and ethene systems; a benchmarking methodology to evaluate the performance of any given force field in describing adsorption in the low-pressure regime and a refinement procedure to rescale the parameters of a force field to better describe the host-guest interactions and provide for simulation isotherms with close agreement to experimental isotherms. Both methodologies were developed based on a reference Henry coefficient, computed with the PBE-MBD functional using the importance sampling technique. The force field rankings predicted by the benchmarking methodology involve the comparison of force field derived Henry coefficients with the reference Henry coefficients and ranking the force fields based on the disparities between these Henry coefficients. The ranking from this methodology matches the rankings made based on uptake disparities by comparing force field derived simulation isotherms to experimental isotherms in the low-pressure regime. The force field rescaling methodology was proven to refine even the worst performing force field in UFF/TraPPE. The uptake disparities of UFF/TraPPE improved from 197% and 194% to 11% and 21% for ethane and ethene, respectively. The proposed methodology is applicable to predict adsorption across nanoporous materials and allows for rescaled force fields to reach quantum accuracy without the need for experimental input.

Molecular Recognition between Carbon Dioxide and Biodegradable Hydrogel Models: A Density Functional Theory (DFT) Investigation

In this research, we explore the potential of employing density functional theory (DFT) for the design of biodegradable hydrogels aimed at capturing carbon dioxide (CO2) and mitigating greenhouse gas emissions. We employed biodegradable hydrogel models, including polyethylene glycol, polyvinylpyrrolidone, chitosan, and poly-2-hydroxymethacrylate. The complexation process between the hydrogel and CO2 was thoroughly investigated at the ωB97X-D/6-311G(2d,p) theoretical level. Our findings reveal a strong affinity between the hydrogel models and CO2, with binding energies ranging from -4.5 to -6.5 kcal/mol, indicative of physisorption processes. The absorption order observed was as follows: chitosan > PVP > HEAC > PEG. Additionally, thermodynamic parameters substantiated this sequence and even suggested that these complexes remain stable up to 160 °C. Consequently, these polymers present a promising avenue for crafting novel materials for CO2 capture applications. Nonetheless, further research is warranted to optimize the design of these materials and assess their performance across various environmental conditions.

Transferable Water Potentials Using Equivariant Neural Networks

Machine learning interatomic potentials (MLIPs) have emerged as a technique that promises quantum theory accuracy for reduced cost. It has been proposed [J. Chem. Phys. 2023, 158, 084111] that MLIPs trained on solely liquid water data cannot accurately transfer to the vapor-liquid equilibrium while recovering the many-body decomposition (MBD) analysis of gas-phase water clusters. This suggests that MLIPs do not directly learn the physically correct interactions of water molecules, limiting transferability. In this work, we show that MLIPs using equivariant architecture and trained on 3200 liquid water structures reproduces liquid-phase water properties (e.g., density within 0.003 g/cm3 between 230 and 365 K), vapor-liquid equilibrium properties up to 550 K, the MBD analysis of gas-phase water cluster up to six-body interactions, and the relative energy and the vibrational density of states of ice phases. We show that potentials developed using equivariant MLIPs allow transferability for arbitrary phases of water that remain stable in nanosecond long simulations.

Machine learning potential for modelling H2 adsorption/diffusion in MOFs with open metal sites

Metal-organic frameworks (MOFs) incorporating open metal sites (OMS) have been identified as promising sorbents for many societally relevant-adsorption applications including CO2 capture, natural gas purification and H2 storage. This has been ascribed to strong specific interactions between OMS and the guest molecules that enable the MOF to achieve an effective capture even under low gas pressure conditions. In particular, the presence of OMS in MOFs was demonstrated to substantially boost the H2 binding energy for achieving high adsorbed hydrogen densities and large usable hydrogen capacities. So far, there is a critical bottleneck to computationally attain a full understanding of the thermodynamics and dynamics of H2 in this sub-class of MOFs since the generic classical force fields (FFs) are known to fail to accurately describe the interactions between OMS and any guest molecules, in particular H2. This clearly hampers the computational-assisted identification of MOFs containing OMS for a target adsorption-related application since the standard high-throughput screening approach based on these generic FFs is not applicable. Therefore, there is a need to derive novel FFs to achieve accurate and effective evaluation of MOFs for H2 adsorption. On this path, as a proof-of-concept, the soc-MOF-1d containing OMS, previously envisaged as a potential platform for H2 adsorption, was selected as a benchmark material and a machine learning potential (MLP) was derived for the Al-soc-MOF-1d from a dataset initially generated by ab initio molecular dynamics (AIMD) simulations. This MLP was further implemented in MD simulations to explore the H2 binding modes as well as the temperature dependence distribution of H2 in the MOF pores from 10 K to 80 K. MLP-Grand Canonical Monte Carlo (GCMC) simulations were then performed to predict the H2 sorption isotherm of Al-soc-MOF-1d at 77 K that was further confirmed using sorption data we collected on this sample. As a further step, MLP-based molecular dynamics (MD) simulations were conducted to anticipate the kinetics of H2 in this MOF. This work delivers the first MLP able to describe accurately the interactions between the challenging H2 guest molecule and MOFs containing OMS. This innovative strategy applied to one of the most complex molecules owing to its highly polarizable nature, paves the way towards a more systematic accurate and efficient in silico assessment of MOFs containing OMS for H2 adsorption and beyond to the low-pressure capture of diverse molecules.

Predicting Spin-Dependent Phonon Band Structures of HKUST-1 Using Density Functional Theory and Machine-Learned Interatomic Potentials

The present study focuses on the spin-dependent vibrational properties of HKUST-1, a metal-organic framework with potential applications in gas storage and separation. Employing density functional theory (DFT), we explore the consequences of spin couplings in the copper paddle wheels (as the secondary building units of HKUST-1) on the material's vibrational properties. By systematically screening the impact of the spin state on the phonon bands and densities of states in the various frequency regions, we identify asymmetric -COO- stretching vibrations as being most affected by different types of magnetic couplings. Notably, we also show that the DFT-derived insights can be quantitatively reproduced employing suitably parametrized, state-of-the-art machine-learned classical potentials with root-mean-square deviations from the DFT results between 3 cm-1 and 7 cm-1. This demonstrates the potential of machine-learned classical force fields for predicting the spin-dependent properties of complex materials, even when explicitly considering spins only for the generation of the reference data used in the force-field parametrization process.

Screening Cu-Zeolites for Methane Activation Using Curriculum-Based Training

Machine learning (ML), when used synergistically with atomistic simulations, has recently emerged as a powerful tool for accelerated catalyst discovery. However, the application of these techniques has been limited by the lack of interpretable and transferable ML models. In this work, we propose a curriculum-based training (CBT) philosophy to systematically develop reactive machine learning potentials (rMLPs) for high-throughput screening of zeolite catalysts. Our CBT approach combines several different types of calculations to gradually teach the ML model about the relevant regions of the reactive potential energy surface. The resulting rMLPs are accurate, transferable, and interpretable. We further demonstrate the effectiveness of this approach by exhaustively screening thousands of [CuOCu]2+ sites across hundreds of Cu-zeolites for the industrially relevant methane activation reaction. Specifically, this large-scale analysis of the entire International Zeolite Association (IZA) database identifies a set of previously unexplored zeolites (i.e., MEI, ATN, EWO, and CAS) that show the highest ensemble-averaged rates for [CuOCu]2+-catalyzed methane activation. We believe that this CBT philosophy can be generally applied to other zeolite-catalyzed reactions and, subsequently, to other types of heterogeneous catalysts. Thus, this represents an important step toward overcoming the long-standing barriers within the computational heterogeneous catalysis community.