(197ag) Computationally Exploring Structure-Property Relationships of Thermal Transport in Metal-Organic Frameworks Using High-Throughput Screening and Machine Learning

Metal-Organic Frameworks (MOFs) are emerging as a highly promising class of materials for applications such as gas storage, separation, and catalysis due to their large surface area, tunable pore geometry, and high porosity. However, the practical utilization of MOFs is often limited by the thermal energy generated during the exothermic adsorption process. Unfortunately, the thermal transport properties of MOFs have not been studied extensively, resulting in a lack of fundamental knowledge about the structure-thermal conductivity relationships necessary for designing MOFs with specific thermal properties. To fill this knowledge gap, we conducted the first computational high-throughput screening of hypothetical MOFs using classical molecular dynamics simulations and the Green-Kubo method. We screened over 10,000 hypothetical MOFs generated by the ToBaCCo-3.0 code, accounting for significant structural and compositional features, including pore size, density, surface area, node-linker mass mismatch, and metal node connectivity. Our analysis revealed that small pores, high density, small node-linker mass mismatch, and four-connected metal clusters are associated with high thermal conductivity.

Furthermore, we trained a graph convolutional neural network on the thermal conductivities of nearly 10,000 hypothetical MOFs. This machine learning model can predict MOF thermal conductivity with a mean absolute error of 0.05 W m^-1 K^-1, which provides an efficient means of exploring the vast design space of MOFs. Additionally, we examined the effect of randomly distributed missing linker and missing cluster defects on the thermal conductivity of UiO-66 and HKUST-1, two well-known MOFs. We also investigated the thermal conductivity of three experimentally determined correlated defect nanodomains of UiO-66 with underlying topologies of bcu, reo, and scu nets. Our findings suggest that both randomly introduced missing linker and missing cluster defects reduce thermal conductivity, while the correlated missing linker defect nanodomain (bcu topology) displayed increased thermal conductivity than pristine UiO-66. Harmonic lattice dynamics calculations also supported these results, indicating an increase in phonon group velocity. In conclusion, our study provides crucial new insights into the design principles of MOFs with tailored thermal properties and highlights the significance of considering structural characteristics and defects when designing high-performance MOFs for various applications. Our application of classical molecular dynamics simulations, machine learning, and defect analysis offers a comprehensive framework for understanding and controlling the thermal transport properties of MOFs.