Study of bonding and doping properties of SP² carbon nanostructures numerical simulations and development of empirical interaction potentials.

by Kolmogorov, Aleksey N.

The main topics of this dissertation are the development of empirical interaction potentials and the study of electronic and mechanical properties of sp2-bonded carbon nanostructures. First, we investigate the weak interlayer binding in graphitic structures. We argue that although the absolute cohesion is not properly described in the local DFT approximations, the variation in the binding energy under interlayer shifts appears to be much more sound than previously suspected. We combine this result with experimental data to introduce a new empirical potential, which describes the variation in the interlayer binding with the relative alignment (registry) of the layers. Lacking a registry dependence, the commonly used Lennard-Jones potential significantly underestimates the variation in energy. We use our potential to study interwall sliding in nested nanotubes. We find that the well-defined geometry and extreme structural anisotropy of a multiwalled carbon nanotube can bring qualitatively new features to its nanometer-scale tribology. Efficient cancellation of registration-dependent interactions in incommensurate tubes (and also, surprisingly, certain axial commensurate tubes) can induce extremely small and nonextensive shear strengths. This result suggests the use of multiwalled carbon nanotubes as nanoscale bearings. We also apply our potential to look at the alignment of nanotubes on a graphitic substrate. We discover that the interaction of a one-dimensional tube with a two-dimensional substrate then leads to an unusual registry phenomenon not visible in standard layer-on-layer growth: the system develops favorable orientations which clearly are incommensurate. This effect could be used for nanotube separation by their helical angle. Next, we study the effect of doping in carbon nanostructures. Using a self-consistent tight-binding model we examine the radial distribution of charge in a bromine-doped double-walled carbon nanotube system. Our results confirm recent Raman measurements that most of the charge resides on the outer wall, even when the outer nanotube is semiconducting and the inner nanotube is metallic. According to experimental data nanoporous carbon also exhibits interesting behavior under doping with alkali atoms: it undergoes graphitization at relatively low temperatures. We propose a representative model of a fully sp2-connected defect (a wormhole) in a carbon network, which could be used to study the graphitization phenomenon. We investigate structural properties of iii the wormhole and consider possible mechanisms of its annihilation. Finally, we explore the possibility to describe interatomic interaction in solids with neural networks methods. We find that with a proper choice of the network and input variables the forces on atoms and the total energy of the system around an equilibrium can be described with ab initio accuracy. We formulate an algorithm how to build and train the network for modeling solid states systems. iv