Advancing the Computational Exploration for Thermoelectric Materials
出版項
2019
說明
1 online resource (203 pages)
文字
text
無媒介
computer
成冊
online resource
附註
Source: Dissertations Abstracts International, Volume: 80-12, Section: B
Publisher info.: Dissertation/Thesis
Advisor: Stevanovic, Vladan;Toberer, Eric
Thesis (Ph.D.)--Colorado School of Mines, 2019
Includes bibliographical references
In the computational search for new thermoelectric materials, high-throughput, semi- empirical models have proven to be one of the more fruitful approaches. The best models take into account both electronic and thermal properties since good thermoelectric materials must maintain a difficult balance between the two to achieve high efficiency. In this vein, one route which has proven particularly insightful is to rank materials by their intrinsic thermoelectric quality, which takes into account the density of states, the carrier mobility, and the lattice thermal conductivity, all of which are parameters that either do not change or change in a predictable manner when adjusting temperature or Fermi level. Semi-empirical models of the thermoelectric quality factor (β) have proven successful in suggesting new materials as candidates for good thermoelectric performance.The semi-empirical model of β developed by the Stevanovic/Toberer research groups relies on purely isotropic models of the density of states effective mass, carrier mobility, and the lattice thermal conductivity. One aspect which is missing from this approach, however, is any account of the electronic component of thermal conductivity (κe). In thermoelectric materials with very low lattice thermal conductivity, the electronic component can often contribute an equivalent amount to the total thermal conduction. To investigate the potential importance of the electronic component of κ, I developed a novel approach to search for materials with potentially low Lorenz number, which is the coefficient that relates κe to the electrical conductivity, σ. Although the Lorenz number is typically calculated assuming a single parabolic band, I showed that by theoretically driving a material’s energy dispersion away from parabolic, specifically by applying the equivalent of a low pass filter to the energy transport, the Lorenz number can be drastically reduced, leading to a significant enhancement in zT over the single parabolic band approximation. Among the mechanisms by which such an affect could occur are the existence of offset multiple bands in conjunction with intervalley phonon scattering. Based on this plausibility argument, I developed a high-throughput search metric, the density of states shape factor, to provide insight in the search for materials with potentially low Lorenz number. By using this metric as a secondary screening tool in conjunction with the existing semi-empirical β, the vast majority of known thermoelectric materials were found to fall within the search parameters. By extension, new materials within the same bounds were identified for further investigation.In addition to enhancing the search for new isotropic thermoelectric materials, the bulk of my research has been devoted to the development of models to screen materials based on anisotropic transport properties. A computational investigation of anisotropic transport within thermoelectric materials had yet to occur. This absence would have neglected materials which have average isotropic performance, but potentially promising properties along one direction. Well-known thermoelectric materials, such as SnSe, Bi2Te3, and Mg3Sb2, are layered materials in which the intrinsic transport would be inherently anisotropic in single crystals, warranting an investigation into anisotropic transport for single-crystal thermoelectric applications.Before thorough investigations into anisotropic transport began, I conducted a survey of materials which we would expect to demonstrate anisotropic single crystal transport. Layered systems, due to their inherent quasi-2D structures, were obvious candidates. Layered thermoelectrics such as SnSe and Bi2Te3 belong to a class of materials which are bound by loose van der Waals (vdW) bonds. A large set of these vdW layered materials had previously been identified through the use of a slab-cutting algorithm. In addition to the vdW layered materials, there exists another set of layered compounds which is of interest to the thermoelectrics community. Compounds such as Mg3Sb2 and the A1B2C2 Zintl thermoelectrics belong to class of layered materials which are more tightly bound than vdW layered materials, due to the existence of an ionically-bound “spacer” elements between the layers. Using a modified version of the slab-cutting algorithm, that I redesigned specifically for the purpose of finding systems belonging to this class, I identified over 1500 of the so-called “ionic” layered compounds, which are often clays. These compounds exhibit inherent structural anisotropy, yet they are a distinct class from the vdW layered compounds because the bonding between layers can range from very weak, near vdW bonding, to very tight, nearly covalent bonding. Additionally, I was able to show that these materials can be structurally linked to the vdW layered materials from which they can be derived by the addition of a spacer element. By conducting an in-depth analysis of the similarities and differences between these two classes of layered systems and assessing the elastic anisotropy, I revealed a rich diversity of anisotropic behavior within this set, which laid the groundwork for further studies of anisotropic transport on this class of materials.Using both the vdW and ionic layered compounds as a material test set, I began the work of building new semi-empirical models for anisotropic transport. The first step was to create a model of anisotropic thermal conductivity. In addition to thermoelectric applications, a model of anisotropic κL is important for any application in which single crystal thermal conductivity is of interest. I created a new anisotropic model for thermal conductivity which achieved an accuracy within a factor of 2 across 5 orders of magnitude. Applying this model to vdW and ionic layered compounds revealed that anisotropy within κL wanes as the minimum value approaches the amorphous limit. Additionally, by examining the high end and low end of thermal conductivity, new potential materials were identified for thermoelectric or power electronic applications based on their predicted κL(θ,φ).The last part of my investigation into anisotropic transport was to build a new model for the prediction of anisotropic carrier mobility. Building upon intuition from the existing semi-empirical model of mobility, I created the new model for directional mobility by using isotropic and anisotropic elastic parameters along with the conductivity effective mass tensor. By fitting the new model to a set of experimental values gathered for over 60 compounds, the new model achieved an accuracy within a factor of 3 across 4 orders of magnitude, which was a significant improvement over the previous isotropic model. Combining the anisotropic mobility and anisotropic lattice thermal conductivity, I was able to create three different metrics by which to rank materials and screen the vdW and ionic layered compounds to search specifically for materials with ideal anisotropic transport
Electronic reproduction. Ann Arbor, Mich. : ProQuest, 2020
