2019 Fiscal Year Final Research Report
Development of automatic potential construction for reaction molecular dynamics based on machine learning
| Project/Area Number |
18K18801
|
|---|
| Research Category |
Grant-in-Aid for Challenging Research (Exploratory)
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Medium-sized Section 18:Mechanics of materials, production engineering, design engineering, and related fields
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2018-06-29 – 2020-03-31
|
| Keywords | 原子間ポテンシャル / 反応力場 / 機械学習 / マルチフィジックス |
|---|
| Outline of Final Research Achievements |
In this study we have developed a new algorithm of automated potential fitting, which dramatically improves efficiency and easiness of ReaxFF, as ReaxFF force fields are essential for reaction molecular dynamics which is a powerful tool to describe chemical reaction within the framework of atomistic modeling, without requiring computationally demanding quantum mechanical calculations. The algorithm is designed to renew a set of atomistic structures of reference data by evaluating the score of an optimized potential function. We picked the problem of NiO reduction as a test case and proved the efficiency of the developed algorithm by successful reproduction of accurate melting point and reasonable reduction process, which were not realized by existing force fields.
|
| Free Research Field |
計算材料科学
|
| Academic Significance and Societal Importance of the Research Achievements |
燃料電池の反応プロセスやガス環境下での材料破壊など、化学反応過程を原子レベルで明らかにすることが求められる問題は多く、これをスパコンなどの超大規模計算機を必要とすることなく効率的にシミュレーションする技術が反応分子動力学法であるが、そこで必要となる反応力場（原子間ポテンシャル）の作成は関数の複雑性のため極めて難しく、その技術は世界でも一部のグループに独占されていた。我々は独自の機械学習アルゴリズムによって力場作成の工程を自動化することで、反応力場作成を大幅に効率化・簡単化した。これによって、研究室レベルの計算機資源で化学反応過程の原子シミュレーションを行う基盤が整備された。
|
