Learning from data has led to paradigm shifts in a multitude of disciplines, including web, text, and image search, speech recognition, as well as bioinformatics. Can machine learning enable similar breakthroughs in understanding quantum many-body systems? Here we develop an efficient deep learning approach that enables spatially and chemically resolved insights into quantum-mechanical observables of molecular systems. We unify concepts from many-body Hamiltonians with purpose-designed deep tensor neural networks (DTNN), which leads to size-extensive and uniformly accurate (1 kcal/mol) predictions in compositional and configurational chemical space for molecules of intermediate size. As an example of chemical relevance, the DTNN model reveals a classification of aromatic rings with respect to their stability -- a useful property that is not contained as such in the training dataset. Further applications of DTNN for predicting atomic energies and local chemical potentials in molecules, reliable isomer energies, and molecules with peculiar electronic structure demonstrate the high potential of machine learning for revealing novel insights into complex quantum-chemical systems.
Ring Polymer Molecular Dynamics and Active Learning of Moment Tensor Potential for Gas-Phase Barrierless Reactions: Application to S + H2
环状聚合物分子动力学(RPMD)已被证明是计算各种化学反应热速率系数的一种精确方法。为了更广泛地应用这种方法,需要有效的方法来生成底层的全维势能面(PES)和相应的能量梯度。最近,我们提出了一个完全自动化的程序,该程序基于将原始RPMDrate代码与使用矩张量势的PES动态主动学习相结合,并成功地将其应用于两个代表性的热活化化学反应[I]。S. Novikov,Y. V. Suleimanov,A. V. Shapeev,物理化学物理,29503-29512(2018)]。在这项工作中,使用原型插入化学反应S + H $_2$,我们表明,这一过程同样适用于另一类化学反应。我们发现,相应的PES可以通过拟合到不到1500自动生成的结构,而RPMD速率系数显示偏离参考值内的RPMDrate的典型收敛误差。我们注意到,更多的结构积累在实时传播的动态因子(再交叉因子),而不是以前的研究。我们还观察到,在进入络合物形成井之前,沿着反应坐标的相对平坦的自由能分布可能导致定位不太收敛的PES的自由能表面的最大值的问题。然而,最终的RPMD速率系数与分割表面的位置无关,这使得它不会受到这个问题的影响,从而将必要结构的总数保持在几千以内。我们的工作的结论是,在未来,所提出的方法可以应用到现实的复杂的化学反应与各种能源配置文件。
Obtaining transferable chemical insight from solving machine-learning classification problems: Thermodynamical properties prediction, atomic composition as good as Coulomb matrix
University of Warwick(华威大学);University of Cambridge(剑桥大学);ENS Paris-Saclay(巴黎萨克雷高等师范学院);Université Paris-Saclay(巴黎萨克雷大学);Aalto University(阿尔托大学);Brigham Young University(杨百翰大学);University of British Columbia(不列颠哥伦比亚大学)
Mechanistic Insights into Chemical Exchange during the Signal Amplification by Reversible Exchange Sensitization of Pyruvate
University Hospital Schleswig-Holstein, Kiel University(基尔大学施莱斯维希-霍尔斯坦医院);University of Oulu(奥卢大学);University of Bath(巴斯大学);Max-Planck-Institut für Kohlenforschung(马克斯·普朗克煤炭研究所);University of York(约克大学)
Exploring Non-Steady-State Charge Transport Dynamics in Information Processing: Insights from Reservoir Computing
Key Laboratory of Organic Integrated Circuit, Ministry of Education & Tianjin Key Laboratory of Molecular Optoelectronic Sciences, Department of Chemistry, School of Science, Tianjin University(天津大学理学院化学系有机集成电路教育部重点实验室与分子光电子科学天津市重点实验室);Tianjin University(天津大学);Collaborative Innovation Center of Chemical Science and Engineering (Tianjin)(化工过程开发与分子工程协同创新中心(天津))
Formation of phosphorus monoxide (PO) in the interstellar medium: insights from quantum-chemical and kinetic calculations
近年来,在恒星形成区和彗星67 P/Churyumov-Gerasimenko中发现了一氧化磷(PO)-生命前化学的一种重要分子。这些研究表明,在星际介质中,PO是系统地最丰富的含磷物质,其丰度是$\sim$的1-3倍,高于第二丰富的含磷分子氮化磷(PN)。PO比PN更丰富的原因仍然不清楚。目前还没有对气相中的磷进行实验研究,可能是因为处理其化合物有困难。因此,原子磷的反应性需要使用可靠的计算工具进行研究。为此,国家的最先进的量子化学计算已被用来评估准确的反应速率和分支比的P + OH $\rightarrow$ PO + H和P + H $_2$ O $\rightarrow$ PO + H $_2$反应的框架内的主方程方法的基础上从头过渡态理论。OH和H ${_2}$ O可以是原子磷的潜在氧化剂的假设是基于H ${_2}$ O在ISM中的普遍存在。它的破坏然后产生OH,这是另一个非常丰富的物种。虽然基态磷原子与水的反应由于出现能垒而不是PO的相关来源,但P + OH反应代表了PO在星际介质中的重要形成途径。动力学结果表明,该反应符合Arrhenius动力学行为,其速率系数α =2.28 $\times$ 10 $^{-10}$ cm ${^3}$分子$^{-1}$ s $^{-1}$,β =0.16,γ =0.37 K随温度升高而增大.