Session: 17-15-01: Society-Wide Micro/Nano Poster Forum

Paper Number: 100225

100225 - Lattice Thermal Transport Properties of Methane Hydrates From Deep Neural Network Interatomic Potentials 

Clathrate hydrates are non-stoichiometric crystalline compounds in which the lattice structure is composed of water molecules as the host and functions as a cage capable of accommodating guest atoms or molecules of various sizes and shapes. The equilibrium between the attractive and repulsive forces of guest molecules ensures the stability of clathrate hydrate, and the effective overlap or hybridization between the electron clouds of guests and water molecules contributes to the global stability. Methane molecules entrapped inside the voids of clathrate hydrates constitute the most common clathrate hydrate. The solid-state phase of methane hydrates formed at low temperatures and high pressures plays a vital role in both environmental and sustainable resource constraints. Clathrates possess three primary crystal structures, referred to as I, II, and H, each of which contains cages of varying sizes. Computer-based ab-initio molecular dynamics simulations (AIMD) have proven successful in understanding the properties and structural stability of clathrates varying with pressure and temperature and the presence of guest molecules. Amid this, classical molecular dynamics (MD) simulations have supplanted the high computational cost of AIMD. In contrast, the conventional empirical potentials, such as the Lennard-Jones (LJ), embedded atom method (EAM), Stillinger-Weber (SW), Tersoff, and charge-optimized many body (COMB) potentials, have been extensively utilized to handle thousands of atoms, but their transferability is a severe limitation. Additionally, the aforementioned potentials do not adequately account for chemical reaction events such as bond breaking or formation. In this work, we used a deep learning approach based on artificial neural networks to generate highly accurate interatomic potentials from extensive AIMD data of methane hydrates at different thermodynamic conditions. Using these deep neural network interatomic potentials, we performed large-scale molecular dynamics simulations and spectral phonon analyses to investigate the phonon properties of methane hydrates at various temperatures and pressures. In particular, we found that the largely unrestricted vibration and rotation of guest methane molecules in clathrate cages can significantly scatter and even localize phonons, greatly hindering thermal transport and thus leading to low thermal conductivity of methane hydrates. Then, polycrystalline methane hydrates with dense grain boundaries were simulated for thermal transport properties, revealing moderate to significant effect on thermal conductivity depending on the average grain size. Moreover, the effect of structural disorder on phonon transport in methane hydrates was investigated and phonon localization behaviors were analyzed rigorously. This work provides a detailed understanding of phonon transport properties in the largely elusive methane hydrate structures.

Presenting Author: Tengfei Ma University of Nevada Reno

Presenting Author Biography: Tengfei Ma hold Ph.D. degree in the Mechanical Engineering department. He received his B.Eng. degree in Nuclear Engineering from Wuhan University (WHU), Wuhan, China in 2014. Prior to his Ph.D. he earned a M.Eng. degree in Power Engineering from North China Electric Power University (NCEPU), Beijing, China in 2017. In the spare time he loves traveling and playing ping-pong.

Authors:

Iyyappa Rajan Panneerselvam University of Nevada Reno
Haoran Cui University of Nevada Reno
Tengfei Ma University of Nevada Reno
Yan Wang University of Nevada Reno