표제지
목차
Abstract 8
1. 서론 10
2. 연구 방법 15
2.1. 암모니아 분해반응 메커니즘 15
2.2. 암모니아 분해 촉매 18
2.3. 기술자 선정 20
2.4. 기술자 검증 22
2.5. 데이터베이스 수집 방법 23
2.6. 흡착에너지 예측용 머신러닝 모델 27
3. 연구 결과 30
3.1. 머신러닝 모델 학습 30
3.2. 머신러닝 모델 학습 결과 33
3.3. 촉매 탐색 결과 35
4. 토의 37
5. 결론 39
참고 문헌 40
Table 1. Physical properties of ammonia. 12
Table 2. Measured TOF for ammonia decomposition over several supported metal catalysts. 19
Table 3. Comparison of catalyst activity experimental result and descriptor used in this study. 22
Table 4. Adsorption energy database. 26
Table 5. Dataset separation for machine learning model training. 31
Table 6. Hyperparameter list. 32
Table 7. Hyperparameter tunning result. 33
Table 8. Alloy catalyst candidates. 36
Fig 1. Global carbon dioxide emissions 2022. 11
Fig 2. Ammonia decomposition mechanism. 17
Fig 3. Ammonia decomposition mechanism over Ni(110). 17
Fig 4. Price per kilogram of metal catalysts. 19
Fig 5. A correlation between the rate of ammonia decomposition on several metals and the relative rate of N-H bond scission, and N-N recombination 21
Fig 6. Used elements for database from periodic table. 24
Fig 7. Structures of (a) single metal and (b) binary alloy. 25
Fig 8. Adsorption site of nitrogen. 25
Fig 9. Slab graph convolutional neural network model architecture. 29
Fig 10. K-fold cross validation (K=5) 31
Fig 11. SGCNN model training result. 34