Statistical Mechanical and Quantum Mechanical Modeling of Condensed Phase Systems

Understanding adsorption in nanoporous media such as carbon nanotubesare vital to improving fluid storage and separations processes. Onemajor objective of this research is to shed light on an on-goingcontroversy in literature over where gases adsorb on single walledcarbon nanotube bundles. Grand-canonical Monte Carlo simulations havebeen performed using models of carbon nanotube bundles composed oftubes of all the same diameter (homogeneous) and tubes of differentdiameters (heterogeneous). We use three metrics with which we compareour simulation results to those found in experiments on HiPconanotubes: the specific surface area, the isosteric heat ofadsorption, and adsorption capacity. Simulations of classicallybehaved fluids Ar, CH$_4$, and Xe indicate that nanotubes prepared bythe HiPco process are best described by a model consisting ofheterogeneous bundles with $sim11\%$ of the nanotubes opened. Nerequires additional considerations to describe the quantum effects atthe temperatures of interest. Simulation results from Ne simulationsare consistent with those from classical fluids. However, Nesimulations strongly indicate that the small interstitial channelsformed by exactly three nanotubes are closed. Combined with previousstudies on classically behaved fluids Ar, CH$_4$, and Xe, experimentaldata including Ne adsorption are best matched by hetergeneous bundleswith $sim11\%$ open nanotubes.The development of a heterogeneous Co/C/O reactive force field(ReaxFF) potential has also been a major objective of this research.ReaxFF provides a method to describe bond-breaking and bond-formingevents that can be applied to large-scale molecular dynamicssimulations. The many-bodied semi-empirical potential has been trainedfrom emph{ab initio} density functional theory calculations. Thetraining set originally included description of bulk and surfacecondensed phase cobalt systems, but was later expanded to includebinary (Co/C, Co/O) and tertiary (Co/C/O) heterogeneous interactions.We have tested these parameters against additional DFT calculationsnot included in the fitting data set. The parameter optimization hasproduced a force field capable of describing new configurations withsame accuracy as those used in the fitting procedure. The optimizedparameters have been used to predict the melting point and diffusioncoefficients of bulk fcc cobalt. Large-scale simulations of a Co$_2$Cphase nanoparticle show segregation on short time scales (less than300 ps), with all C atoms forming graphene precursors on the surfaceof a Co nanoparticle core. ReaxFF has also been used to showdiffusion of Co is more energetically favorable than oxygen through acobalt oxide crystal. This is consistent with experimentalobservations that oxidized cobalt nanoparticle form hollow cobaltoxide nanospheres. These two binary applications demonstrate thatReaxFF is transferable to heterogeneous systems and a computationallyinexpensive means by which transition metal surface reactions can beexplored