NpjComput.Mater.:機器學習原子間勢:兼具高效能和高效率

海歸學者發起的公益學術平臺
分享資訊,整合資源
交流學術,偶爾風月

精準、高效地預測原子間勢能是推動材料設計與模擬的關鍵。然而,現有深度學習方法雖然在加速計算方面展現出潛力,但往往難以兼顧預測精度和計算效率,特別是在大規模體系模擬中仍存在顯著挑戰。要實現既精準又高效的勢能預測,關鍵在於兩方面:首先,為確保物理合理性,模型預測的能量需對剛體變換(平移與旋轉)保持不變,而預測的力場需對平移不變、對旋轉等變,從而保證物理一致性和預測準確性;其次,模型設計需在滿足等變性的同時保持足夠的計算效率,以適應大規模模擬的需求。
圖神經網路的基本原理以及等變與不變概念的解釋
針對這一挑戰,新加坡國立大學沈雷課題組與北京大學陳語謙教授團隊合作,提出了一種兼具效能與效率優勢的等變圖神經網路(E2GNN)。不同於傳統等變模型依賴複雜的高階張量表示,E2GNN 採用標量向量的簡潔表示對等變特徵進行編碼,並結合全域性訊息傳遞機制,精準建模長程原子間相互作用,實現了高精度與高效率的勢能預測。此外,全域性訊息傳遞與門控訊息傳遞機制的結合進一步增強了模型的整體表現。
2 E2GNN的原理圖
實驗結果表明,在催化劑、分子體系和有機異構體等多個數據集中,E2GNN均優於當前代表性基線模型,同時在計算效率上展現出明顯優勢。此外,基於 E2GNN 力場進行的固態、液態和氣態體系的分子動力學模擬,其預測精度可達到第一性原理級別。
對固體、液體和氣體的分子動力學模擬
本研究提出的高精度、高效率原子間勢能預測模型,有望加速新材料發現、最佳化現有材料效能,併為材料設計提供更高效的資料驅動工具,助力材料科學計算邁向更精準、更高效的新階段。該文近期發表於npj ComputationaMaterials11,:49 (2025)英文標題與摘要如下,點選左下角“閱讀原文”可以自由獲取論文PDF。
Efficient equivariant model for machine learning interatomic potentials 
Ziduo Yang, Xian Wang, Yifan Li, Qiujie Lv, Calvin Yu-Chian Chen & Lei Shen
In modern computational materials, machine learning has shown the capability to predict interatomic potentials, thereby supporting and accelerating conventional molecular dynamics (MD) simulations. However, existing models typically sacrifice either accuracy or efficiency. Moreover, efficient models are highly demanded for offering simulating systems on a considerably larger scale at reduced computational costs. Here, we introduce an efficient equivariant graph neural network (E2GNN) that can enable accurate and efficient interatomic potential and force predictions for molecules and crystals. Rather than relying on higher-order representations, E2GNN employs a scalar-vector dual representation to encode equivariant features. By learning geometric symmetry information, our model remains efficient while ensuring prediction accuracy and robustness through the equivariance. Our results show that E2GNN consistently outperforms the prediction performance of the representative baselines and achieves significant efficiency across diverse datasets, which include catalysts, molecules, and organic isomers. Furthermore, we conduct MD simulations using the E2GNN force field across solid, liquid, and gas systems. It is found that E2GNN can achieve the accuracy of ab initio MD across all examined systems.
媒體轉載聯絡授權請看下方

相關文章