Microstructure defects are the source of many interesting physical phenomena in materials. The influence of microscopic imperfections are visible among a background of a crystalline material and can noticeably change macroscopic materialproperties. Ab initio computations are a wide range of computational methods used to study materials at such microscopic levels. Classical and quantum-level computations are often viewed as competing methods, but we instead take a complementary approach. We combine strengths of both methods to study two classes of materials—silicon and the two-dimensional (2D) transition metal dichalcogenide (TMD) molybdenum disulfide (MoS2). In MoS2, we interrogate the effects of the presence of dopants. Conversely in Si, we take amorphization as a disruption to the base crystalline structure. In both, we are interested in how these changes to microstructure affect energetic efficiencies. In a-Si, the photovoltaic efficiency is hampered by the presence of dangling bonds, while in MoS2, mechanical efficiencies of lubrication are enhanced when doped.
Both systems pose their unique challenges and deviations from standard workflows. We will review the overarching strategies used to study them in Ch. 2. Throughout our studies, we have developed several techniques to overcome specific challenges which are detailed in their respective sections. With a-Si, a classical potential Monte Carlo code is used to generate realistic, non-biased, fully amorphous coordinates faster than what could be achieved by currently available quantum-mechanical methods. Then, density functional theory is used to relax the structures. With MoS2, higher-accuracy DFT energy computations are used to parametrize a classical force-field for the computation of larger systems which would otherwise be difficult with DFT alone.
We cover a wide range of analyses between both materials. We have developed methodologies for analyzing the amorphous and 2D material systems using classical methods and density functional theory. We have computed a range of material properties for MoS2 and amorphous Si and hydrogenated a-Si:H. With Si, we modified the Wooten-Winer-Weaire algorithm to produce amorphous networks with included voids by application of initial strain. The size of the voids is somewhat controlled by the strain value. These voids emerge naturally as a part of the amorphization process.
We find that the amorphous networks generated by simple Keating springs when applied in the WWW method is retained when relaxed by DFT. Structure-scale approximations of the Keating potential yields the result that ∆θ explains a large portion of the structure’s energy, and this holds even in density functional theory.
With 2D materials, we outline a specific multi-step method to quantify the sliding of defected materials in Ch. 6. Successive increases in the degrees of freedom while sliding allow us to access different components of sliding—namely the potential barrier differences, low-energy sliding pathways, and slip planes. The sliding potential in MoS2, even while intercalated, is composed of pairwise interactions of the MoS2 interfaces. This means that computing arbitrarily sized systems can be theoretically computed by only considering interactions of their interfaces.
We find tetrahedrally intercalated Ni-doped MoS2 to be stable and thus more important than is considered in the literature. When intercalated, we find Ni can bind layers together, explaining the material’s increased resistance to wear, or material loss during sliding.
Re-doped MoS2 has shown an increase in friction with an increasing layer count as measured with atomic force microscopy. This is counter to typical 2D materials. We find that intercalated Re can explain this relationship as an alteration to outof-plane stiffness. For this material, in computing the vibrational spectroscopy we overcame a difficulty in computing its Raman spectra. It has a metallic character, thereby limiting our Raman intensity computation due to an infinite dielectric constant. We developed a method to circumvent this computation and approximate the spectra by substitution of the atomic Raman tensor.
