Oral: Prediction of Thermal Conductivity in CALF-20 with First-Principles Accuracy via Machine Learning Interatomic Potentials

We report the thermal transport study of CALF-20, a recent addition to the Metal-Organic Framework (MOF) family known for high capacity of CO2 capture capability, using artificial neural network-based machine learning potentials (MLP). We applied a renormalization technique to the raw heat flux to eliminate non-contributing components and reduce noise in the calculated thermal conductivity (κ) via Green-Kubo formalism, marking the first implementation of this approach in MLP-based molecular dynamics (MD) simulation generated heat flux. The κ predicted by our MLP-based MD study reveals the anisotropic thermal transport properties of CALF-20, yielding a low value below 1 W/m.K, making it suitable for thermoelectric energy conversion applications. Our analysis of the temperature (T) dependence of κ in CALF-20 reveals a weak T dependence (κ ∽1/T^0.56), which stands out from the typical trend observed in crystalline materials (κ ∽1/T). Surprisingly, the κ of CALF-20 remains nearly constant with a slight overall decrease in response to rising pressure, whereas in typical crystalline solids, κ may either increase or decrease depending on the mass ratio of the constituent atoms. The outcome of the study, leveraging advanced computational techniques for predictive modeling, offers valuable insights into more suitable applications of CALF-20 with tailored thermal properties.