Disorder at the nanoscale: A computational study

Disorder is an inherent component of real materials, with significant implications for their application in functional devices. Despite this, the theoretical modelling of disorder remains restricted, primarily due to the large simulation cells required to adequately represent disordered systems, and the associated computational costs. This has been remedied in part by the increased availability of resources for high performance computing. In this thesis, using a combination of computational techniques, primarily density functional theory and ab initio as well as classical molecular dynamics, we investigate disorder in two broad categories – physical and chemical disorder, in three distinct classes of materials: palladium nanoparticles, the negative thermal expansion cuprite Ag2O and the complex quaternary chalcogenide Cu2ZnSnS4, known commonly as kesterite. The ‘physical’ disorder discussed in the thesis constitutes shape- and adsorption-induced mechanical softening on the surface of Palladium nanocrystals used for nanocatalysis. This includes one study on the the adsorption of organic capping agents, and another on the adsorption of oxygen molecules and the subsequent oxidation of Pd. In the former, it was observed that the strain effect due to adsorption-induced surface disorder is significantly greater than that due to variations in surface termination, i.e. nanoparticle shape. Moreover in the latter case, different crystallographic facets with different degrees of disorder were found to affect the spin-flip induced activation of oxygen atoms, relevant to the catalytic oxygen reduction reaction in hydrogen fuel cells. In each case, the computational results were combined with a sophisticated, phenomenological whole powder pattern modeling of X-ray diffraction data primarily from synchrotron radiation, leading to an accurate characterization of the Debye-Waller coefficient, which was established as a reliable metric for disorder in crystalline systems. In the case of Ag2O instead, we demonstrated that the large experimental Debye-Waller coefficient was due to thermal diffuse scattering arising from the strong distortion of the Ag4O coordination tetrahedra. The second form of disorder which was investigated is ‘chemical’ disorder, which refers to cation disorder in the quaternary chalcogenide Cu2ZnSnS4 studied for its performance as a thermoelectric material. Similar to the studies on palladium, the disorder was quantified through the Debye-Waller coefficient using molecular dynamics simulations, this time from ab initio methods, and compared with X-ray diffraction data from a synchrotron source. The ordered phase of CZTS is known to crystallize in a tetragonal phase, with alternating Cu-Zn and Cu-Sn cation layers sandwiched between sulfur layers. Two forms of cation disorder were studied: disorder only in the Cu-Zn layer, leading to a disordered tetragonal phase, and full cation site randomization, leading to a disordered cubic polymorph. In the former case, it was found that the higher symmetry of the disordered tetragonal structure led to an average symmetrization of the nearest neighborhood of each individual cation, as a result of which there was a convergence of bands at the valence band maximum, leading to an experimentally observed increase in p-type carrier concentration. In the case of CZTS with full cation disorder, inhomogenous bond led to favorable modifications of the electronic and phonon properties, allowing for a simultaneous improvement of the experimentally measured electrical and thermal conductivities as well as the Seebeck coefficients. Finally, by studying the atypical electronic band structure of this cubic polymorph, we were able to identify topologically non-trivial behavior evidence of bulk band inversion, robust surface states, and an adiabatically continuous connection to a known TI phase. As such, we were able predict disordered cubic CZTS to be the first disorder-induced topological Anderson insulator in a real material system.