Machine learning approaches for screening of materials in flexible electronic devices
In the rapidly evolving field of flexible electronics, the development of advanced materials is a cornerstone for innovation. Both active and substrate materials are key to the functionality and performance of flexible devices, which are becoming increasingly integral across various applications. Th...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis-Doctor of Philosophy |
Language: | English |
Published: |
Nanyang Technological University
2024
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/177761 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-177761 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1777612024-06-03T06:51:20Z Machine learning approaches for screening of materials in flexible electronic devices Deng, Siyan Li Shuzhou School of Materials Science and Engineering LISZ@ntu.edu.sg Engineering Machine learning Materials screening In the rapidly evolving field of flexible electronics, the development of advanced materials is a cornerstone for innovation. Both active and substrate materials are key to the functionality and performance of flexible devices, which are becoming increasingly integral across various applications. The integration of machine learning into the field of flexible electronics represents a groundbreaking shift in materials science, allowing for accelerated discovery and optimization of materials. This thesis presents two pivotal projects that leverage machine learning to promote advancements in flexible electronics. The first project employs machine learning-assisted high-throughput screening (HTS) to identify elastomers with desired mechanical properties, utilizing innovative structure-based multilevel (SM) descriptors derived solely from the molecular structure of the materials. Existing elastomer descriptors necessitate both experimental and simulation data for precise prediction of elastomer properties, which may not be available for all candidates of interest. This impedes the discovery of new elastomers through HTS. Our SM descriptors are derived solely from the universally accessible molecular structure. These SM descriptors are hierarchically organized to capture both local and global structures of elastomers. With the SM-Morgan Fingerprint (SM-MF) descriptors, one of our SM descriptors, a machine learning model accurately predicts elastomer toughness with a remarkable accuracy of 0.91. Furthermore, an HTS pipeline is established to swiftly screen elastomers with targeted toughness. We also demonstrate the generality and applicability of SM descriptors by constructing HTS pipelines for screening elastomers with targeted critical strain or Young’s modulus. The second project applies deep learning-assisted HTS, enhanced by transfer learning, to identify conjugated oligomers suitable for photovoltaic materials from a vast pool of candidates. The study employs transfer learning techniques to overcome the challenge of limited data, particularly prevalent in the study of conjugated oligomers. By transferring knowledge from a extensive dataset, the models accurately predicted essential optoelectronic properties of conjugated oligomers, including Highest Occupied Molecular Orbital (HOMO), Lowest Unoccupied Molecular Orbital (LUMO), and HOMO-LUMO gap, reducing mean absolute errors (MAE) significantly and outperforming direct learning models. Throughout this thesis, the efficacy of machine learning technologies in addressing the challenges of data scarcity and complexity in material prediction is demonstrated. The SM descriptors and transfer learning models developed here not only streamline the process of material selection for flexible electronics but also set a precedent for the use of these advanced computational methods in broader materials science applications. The success of these projects underscores the potential of machine learning to revolutionize the discovery and design of new materials, promising a new era of innovation in flexible electronic devices. Doctor of Philosophy 2024-05-30T05:56:42Z 2024-05-30T05:56:42Z 2024 Thesis-Doctor of Philosophy Deng, S. (2024). Machine learning approaches for screening of materials in flexible electronic devices. Doctoral thesis, Nanyang Technological University, Singapore. https://hdl.handle.net/10356/177761 https://hdl.handle.net/10356/177761 10.32657/10356/177761 en This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). application/pdf Nanyang Technological University |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering Machine learning Materials screening |
spellingShingle |
Engineering Machine learning Materials screening Deng, Siyan Machine learning approaches for screening of materials in flexible electronic devices |
description |
In the rapidly evolving field of flexible electronics, the development of advanced materials is a cornerstone for innovation. Both active and substrate materials are key to the functionality and performance of flexible devices, which are becoming increasingly integral across various applications. The integration of machine learning into the field of flexible electronics represents a groundbreaking shift in materials science, allowing for accelerated discovery and optimization of materials. This thesis presents two pivotal projects that leverage machine learning to promote advancements in flexible electronics.
The first project employs machine learning-assisted high-throughput screening (HTS) to identify elastomers with desired mechanical properties, utilizing innovative structure-based multilevel (SM) descriptors derived solely from the molecular structure of the materials. Existing elastomer descriptors necessitate both experimental and simulation data for precise prediction of elastomer properties, which may not be available for all candidates of interest. This impedes the discovery of new elastomers through HTS. Our SM descriptors are derived solely from the universally accessible molecular structure. These SM descriptors are hierarchically organized to capture both local and global structures of elastomers. With the SM-Morgan Fingerprint (SM-MF) descriptors, one of our SM descriptors, a machine learning model accurately predicts elastomer toughness with a remarkable accuracy of 0.91. Furthermore, an HTS pipeline is established to swiftly screen elastomers with targeted toughness. We also demonstrate the generality and applicability of SM descriptors by constructing HTS pipelines for screening elastomers with targeted critical strain or Young’s modulus.
The second project applies deep learning-assisted HTS, enhanced by transfer learning, to identify conjugated oligomers suitable for photovoltaic materials from a vast pool of candidates. The study employs transfer learning techniques to overcome the challenge of limited data, particularly prevalent in the study of conjugated oligomers. By transferring knowledge from a extensive dataset, the models accurately predicted essential optoelectronic properties of conjugated oligomers, including Highest Occupied Molecular Orbital (HOMO), Lowest Unoccupied Molecular Orbital (LUMO), and HOMO-LUMO gap, reducing mean absolute errors (MAE) significantly and outperforming direct learning models.
Throughout this thesis, the efficacy of machine learning technologies in addressing the challenges of data scarcity and complexity in material prediction is demonstrated. The SM descriptors and transfer learning models developed here not only streamline the process of material selection for flexible electronics but also set a precedent for the use of these advanced computational methods in broader materials science applications. The success of these projects underscores the potential of machine learning to revolutionize the discovery and design of new materials, promising a new era of innovation in flexible electronic devices. |
author2 |
Li Shuzhou |
author_facet |
Li Shuzhou Deng, Siyan |
format |
Thesis-Doctor of Philosophy |
author |
Deng, Siyan |
author_sort |
Deng, Siyan |
title |
Machine learning approaches for screening of materials in flexible electronic devices |
title_short |
Machine learning approaches for screening of materials in flexible electronic devices |
title_full |
Machine learning approaches for screening of materials in flexible electronic devices |
title_fullStr |
Machine learning approaches for screening of materials in flexible electronic devices |
title_full_unstemmed |
Machine learning approaches for screening of materials in flexible electronic devices |
title_sort |
machine learning approaches for screening of materials in flexible electronic devices |
publisher |
Nanyang Technological University |
publishDate |
2024 |
url |
https://hdl.handle.net/10356/177761 |
_version_ |
1800916255294619648 |