Computational discovery, design and optimization of high-performance thermoelectric materials

Qu, Jiaxing

Permalink

https://hdl.handle.net/2142/124553 Copy

Description

Title

Computational discovery, design and optimization of high-performance thermoelectric materials

Author(s)

Qu, Jiaxing

Issue Date

2024-04-26

Director of Research (if dissertation) or Advisor (if thesis)

Ertekin, Elif

Doctoral Committee Chair(s)

Ertekin, Elif

Committee Member(s)

• Johnson, Harley
• Perry, Nicola
• Cai, Lili

Department of Study

Mechanical Sci & Engineering

Discipline

Mechanical Engineering

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

Ph.D.

Degree Level

Dissertation

Keyword(s)

thermoelectrics, computational material discovery

Abstract

The Materials Genome Initiative (MGI) offers a promising route to revolutionize the realization and discovery of a diverse set of functional materials. A key to the success of the MGI is the synergistic effect of computational and experimental material science, wherein computation serves as a guide for experimentation. The field of thermoelectrics (TE) has historically been dominated by experimental work motivated by chemical intuition. However, sampling the search space by trial-and-error experiments is not efficient and the discovery of top TE materials is largely from serendipity. Therefore, it is critical to achieve computation-driven discovery, design and optimization of TEs, ideally in a high-throughput manner. The challenge comes from the stringent requirements for an effective thermoelectric material, which requires both favorable intrinsic properties relevant for charge carrier and heat transport, as well as defect characteristics to achieve optimal carrier concentrations for maximized TE performance. The objective of this work is to develop a first-principles materials discovery, optimization and design framework for TE candidates considering both transport and defect properties. Our work mainly examined two material classes computationally in detail --diamond-like semiconductors (DLS) and Zintl phases which exhibit reciprocal behaviors in terms of electron and thermal transport. In the first part of my work, I applied different optimization techniques in these two representative of diverse material systems to investigate the fundamental physics of electron transport and defect chemistry. The ternary and quaternary DLS phases exhibit unusually high electronic mobility and abnormally low lattice thermal conductivity ideal for thermoelectrics. For DLS materials with intrinsic favorable electrical and thermal transport, my work primarily focuses on defect engineering by controlling native defects and introducing extrinsic doping to tune the carrier concentrations to maximize their TE performance. The other family is the Zintl phases which have recently been recognized as promising TE candidates owing to their complex crystal structures, intricate chemical bonding, and thus intrinsically low thermal conductivity. For Zintl phases with favorable poor thermal transport, by contrast, my optimization approach concentrates on promoting electron transport via band engineering. The second part deals with computational discovery of high-performance TE materials. To avoid the computational extensive cost for transport and defect calculations, I designed a funnel-based framework to achieve high-throughput search of high-performance TE materials using first-principles simulations. This framework employs computationally tractable descriptors and semi-empirical transport models to assess the TE performance of large chemical spaces, thus can effectively select candidates with promising intrinsic transport properties for subsequent more computationally extensive defect analysis. We applied this framework to the search of n-type Zintl phases and successfully identified several n-type dopable Zintl phases, adding to the list of rare n-type Zintl phases. This framework also enables discovery of I-III5-Te8 (I = Cu, Ag; III = Ga, In) ordered-vacancy chalcogenides (OVCs), a special class of DLS materials, which closely resemble state-of-the-art TE materials I-III-Te2 chalcopyrites. Leveraging another computational tool - modern deep learning models, we designed a recommendation engine for material discovery empowered by natural language representation and demonstrated its application on TE materials. With the contextual information obtained from language representations, we demonstrated diversified recommendations of prototype structures and identify under-studied high-performance material spaces. In the third part, we focus on generalized material design framework and rules for high-performance TEs. For alloy Zintl systems, we designed a computational framework that combines first-principles calculations with alloy and point defect modeling to identify optimal alloy compositions. This design enables multi-task optimization of electronic, thermal, and defect properties in alloy systems. Band unfolding is performed to sketch the effective band structures of alloys and identify compositions to facilitate band convergence and minimize alloy scattering of electrons. Additionally, we further contributed to the design of low thermal conductivity bulk Zintl phases via local motif analysis. Leveraging a structural fingerprint, Zintl phases are classified into structure prototypes by coordination motifs and connectivity.This work reveals several design strategies for achieving low thermal conductivity Zintl phases from the motif-level perspective and identifies linear-chains and trigonal planar structure prototypes as promising TE candidates with ultra-low thermal conductivity. Together, our efforts ultimately advance the discipline of TEs in the direction of the MGI's vision. We have provided understandings for optimization strategies for TE performance, identified potential candidates for TE applications, and built-upon existing computational techniques to accelerate material discovery and design. These efforts have begun to reveal the complex structure-property dependency that advances TE performance.

Graduation Semester

2024-05

Type of Resource

Thesis

Copyright and License Information

Copyright 2024 Jiaxing Qu

Owning Collections