National Institute of Fuel-Cell Technology
Department of Mechanical & Aerospace Engineering
West Virginia University
A common drawback of molecular dynamics (MD) simulations is their inability to describe bond formation, which is essential in modeling of chemical reactions This drawback can be remedied if an MD simulation could incorporate not only the intermolecular force potentials, but also allow for the possibility of different states of molecular excitation and bonding. These effects can in principle be modeled in a probabilistic sense, if the MD simulation takes advantage of the relevant probability data from ab-initio quantum mechanical simulations or experimental results. In this study a molecular dynamics code was developed for computing reactive gas mixtures, including catalytic reactions at active surfaces (electrochemistry) based on the Collision Theory.
In simulations of a gaseous phase and gas-solid interactions, each molecule is advanced following the classical laws of conservation of its linear and angular momenta. At the same time a reaction possibility is considered, which determines the transition event for the two species to enter into a reaction with the production of new species or forming a molecular bond. The reaction event occurs when the impact energy of the molecules plus their internal energy exceeds the known activation energy for the reaction. Each molecule of a newly formed specie is again treated as a rigid body subjected to Newton's laws of motion. Energy distribution among the molecules which result from the collision are calculated in accordance with the molecular degrees of freedom determined by molecule's specific heat constants.
To be able to trace the motion and interactions of millions of molecules efficiently, a domain segmentation algorithm was implemented, which enabled to reduce the time of processing interactions from N2 to a near-linear dependence. The code was implemented in C++ programming language. The figure below shows a snapshot of a simulation of hydrogen interaction with the reactive surface with the production of H2O. The simulation inside one micron pore under atmospheric conditions performed on a 2GHz CPU, 4GB RAM workstation took about one week. The program enables one to simulate up to 10 million molecules per each Gigabyte of RAM.