Molecular Dynamics and Machine-Learning Studies on Structure-Property Relationship in Solid-State Electrolytes and Combustion
분자동역학과 머신러닝을 활용한 고체 전해질 및 연소 반응에서의 구조-물성 관계 연구
- 주제어 (키워드) structure-property relationship , machine-learning molecular dynamics , inorganic glass solid-state electrolytes , ion conductivity , graph neural network , combustion properties , explainable artificial intelligence , 구조-물성 관계 , 머신 러닝 분자동역학 , 유리 무기 고체 전해질 , 그래프 기반 인공신경망 , 연소 특성
- 발행기관 서강대학교 일반대학원
- 지도교수 성봉준
- 발행년도 2025
- 학위수여년월 2025. 8
- 학위명 석사
- 학과 및 전공 일반대학원 화학과
- 실제 URI http://www.dcollection.net/handler/sogang/000000081756
- UCI I804:11029-000000081756
- 본문언어 영어
- 저작권 서강대학교 논문은 저작권 보호를 받습니다.
목차
Abstract xi
1 Cascading Hopping as Ion Conduction Mechanism of Inorganic Glass Solid-
State Electrolytes 1
1.1 Introduction 1
1.2 Results 5
1.2.1 The composition dependence of ion conductivity and dynamic het-
erogeneity 5
1.2.2 The transport mechanism of lithium ions 10
1.2.3 The correlation between lithium hopping and AlCl –4 rotation 15
1.3 Discussions 16
1.4 Methods 18
1.4.1 The construction of the machine-learning potential for LixAlCl3+x . 18
1.4.2 The large-scale machine learning molecular dynamics (MLMD)
simulations 21
1.4.3 The hop function analysis 22
1.5 References 24
2 Chemomile: Explainable Multi-Level GNN Model for Combustion Property
Prediction 32
2.1 Introduction 32
2.2 Methods 36
2.2.1 Fragmentation and preprocessing 36
2.2.2 Chemomile Network 38
2.2.3 AttentiveFP 39
2.2.4 Particle Swarm Optimization (PSO) 41
2.2.5 Dataset 43
2.2.6 Training and Metrics 46
2.3 Results and discussion 46
2.3.1 Performance and Benchmark of Chemomile 46
2.3.2 Explainability of Chemomile 50
2.3.3 A Test Case: Combustion Enthalpy of Organosilcons 53
2.4 Conclusion 56
2.5 References 57
