國立虎尾科技大學 |

Computational Methods for Materials Discovery.

紀錄類型:	書目-語言資料,手稿 : Monograph/item
正題名/作者:	Computational Methods for Materials Discovery./
作者:	Lee, Andrew S.
面頁冊數:	1 online resource (234 pages)
附註:	Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
Contained By:	Dissertations Abstracts International85-06B.
標題:	Materials science. -
電子資源:	click for full text (PQDT)
ISBN:	9798381175042

Computational Methods for Materials Discovery.
Lee, Andrew S.

Computational Methods for Materials Discovery. - 1 online resource (234 pages)

Source: Dissertations Abstracts International, Volume: 85-06, Section: B.

Thesis (Ph.D.)--Northwestern University, 2023.

Includes bibliographical references

The process of materials discovery has rapidly changed in recent decades, with advancements in computational methods being a main driving force for this change. The introduction of density functional theory (DFT) first enabled researchers to accurately calculate the thermodynamic energy of almost any conceivable material. Improvements in computing power and the emergence of new machine learning (ML) methods have further unlocked new methods and strategies for finding new materials. In this dissertation, we explore several studies that leverage computational methods to accelerate different stages of materials discovery.This dissertation's first chapter summarizes how materials discovery has evolved and introduces several key concepts. The second chapter details a study that combines DFT-calculated stabilities with ML to predict material synthesizability. By combining these methods, we make accurate predictions that DFT cannot make, while aligning ML predictions with first-principles calculated properties. Extending the concept of DFT-calculated stability as a descriptor for materials synthesizability, we later devised a novel descriptor for predicting organic molecule synthesizability in chapter three. This descriptor is inspired by DFT stability, which only directly applies to crystalline materials. Our new synthesizability descriptor works well on organic molecules, suggesting there are commonalities in synthesizability between crystalline and organic materials. In the fourth chapter, we pivot our attention toward a tool for processing experimentally-measured X-Ray diffraction histograms. Using features developed from expert intuition, we built an algorithm that leveraged unsupervised clustering to group visually similar histograms. The algorithm groups histograms accurately, where different groups tend to reflect physical differences in observed histograms. With this tool, we analyzed ≈ 18,000 histograms in a short timeframe. The fifth chapter returns our focus to a DFT-driven study, where we employed high-throughput calculations to identify solid solution forming mixtures. This work focused on the Half-Heusler class of materials and reported calculation results for over 800 DFT calculations on binary mixtures. We devised a categorization system to help analyze our results and identify chemical trends that can explain our calculated energies. The study concludes by identifying solid solution forming mixtures and discussing how energies from binary mixtures can be used to identify promising ternary mixtures. Finally, we touch upon some work on ML methods development, where we focused on predicting transfer learning performance. While transfer learning is a well-demonstrated concept, it is not well-understood. Therefore, we introduce several hypotheses that attempt to intuitively explain transfer learning performance. The chapter concludes by proposing procedures for testing these hypotheses and a discussion on obstacles that have been encountered during this study.

Electronic reproduction.
Ann Arbor, Mich. :
ProQuest,
2024

Mode of access: World Wide Web

ISBN: 9798381175042Subjects--Topical Terms:

557839
Materials science.
Subjects--Index Terms:

Density functional theoryIndex Terms--Genre/Form:

554714
Electronic books.

Computational Methods for Materials Discovery.
LDR:04392ntm a22003977 4500 001 1149576
005 20241022112135.5
006 m o d
007 cr bn ---uuuuu
008 250605s2023 xx obm 000 0 eng d
020 $a 9798381175042
035 $a (MiAaPQ)AAI30694173
035 $a AAI30694173
040 $a MiAaPQ $b eng $c MiAaPQ $d NTU
100 1 $a Lee, Andrew S. $3 1475870
245 1 0 $a Computational Methods for Materials Discovery.
264 0 $c 2023
300 $a 1 online resource (234 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 85-06, Section: B.
500 $a Advisor: Wolverton, Christopher.
502 $a Thesis (Ph.D.)--Northwestern University, 2023.
504 $a Includes bibliographical references
520 $a The process of materials discovery has rapidly changed in recent decades, with advancements in computational methods being a main driving force for this change. The introduction of density functional theory (DFT) first enabled researchers to accurately calculate the thermodynamic energy of almost any conceivable material. Improvements in computing power and the emergence of new machine learning (ML) methods have further unlocked new methods and strategies for finding new materials. In this dissertation, we explore several studies that leverage computational methods to accelerate different stages of materials discovery.This dissertation's first chapter summarizes how materials discovery has evolved and introduces several key concepts. The second chapter details a study that combines DFT-calculated stabilities with ML to predict material synthesizability. By combining these methods, we make accurate predictions that DFT cannot make, while aligning ML predictions with first-principles calculated properties. Extending the concept of DFT-calculated stability as a descriptor for materials synthesizability, we later devised a novel descriptor for predicting organic molecule synthesizability in chapter three. This descriptor is inspired by DFT stability, which only directly applies to crystalline materials. Our new synthesizability descriptor works well on organic molecules, suggesting there are commonalities in synthesizability between crystalline and organic materials. In the fourth chapter, we pivot our attention toward a tool for processing experimentally-measured X-Ray diffraction histograms. Using features developed from expert intuition, we built an algorithm that leveraged unsupervised clustering to group visually similar histograms. The algorithm groups histograms accurately, where different groups tend to reflect physical differences in observed histograms. With this tool, we analyzed ≈ 18,000 histograms in a short timeframe. The fifth chapter returns our focus to a DFT-driven study, where we employed high-throughput calculations to identify solid solution forming mixtures. This work focused on the Half-Heusler class of materials and reported calculation results for over 800 DFT calculations on binary mixtures. We devised a categorization system to help analyze our results and identify chemical trends that can explain our calculated energies. The study concludes by identifying solid solution forming mixtures and discussing how energies from binary mixtures can be used to identify promising ternary mixtures. Finally, we touch upon some work on ML methods development, where we focused on predicting transfer learning performance. While transfer learning is a well-demonstrated concept, it is not well-understood. Therefore, we introduce several hypotheses that attempt to intuitively explain transfer learning performance. The chapter concludes by proposing procedures for testing these hypotheses and a discussion on obstacles that have been encountered during this study.
533 $a Electronic reproduction. $b Ann Arbor, Mich. : $c ProQuest, $d 2024
538 $a Mode of access: World Wide Web
650 4 $a Materials science. $3 557839
650 4 $a Engineering. $3 561152
653 $a Density functional theory
653 $a Machine learning
653 $a Materials discovery
653 $a Materials informatics
653 $a Synthesizability
655 7 $a Electronic books. $2 local $3 554714
690 $a 0794
690 $a 0800
690 $a 0537
710 2 $a ProQuest Information and Learning Co. $3 1178819
710 2 $a Northwestern University. $b Materials Science and Engineering. $3 1181794
773 0 $t Dissertations Abstracts International $g 85-06B.
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=30694173 $z click for full text (PQDT)