表征恒电位过程自由能的量子化学方法及应用研究

通过恒电位条件下的自由能计算实现定量模拟反应动力学，是量子化学应用于电化学反应的核心、难点科学问题。对电化学催化应用于能源存储和转化与利用新技术上的强烈期待, 推动上述问题成为研究热点。为挑战这一理论难题，本项目在前期工作基础上进一步提出系统的解决方案：（1）结合总结归纳与机器学习，对DFT能量施加必要的电子相关能校正，改善其精度。（2）基于原子神经网络，以机器学习实现复杂电化学体系的势能面和表面功函的从头算精度快速计算；运用CPU-GPU异构计算技术加速计算；使机器学习自主化以节省人工成本。（3）实现大规模体系的长时间尺度增强采样分子动力学模拟计算，获得全反应路径的自由能势能面。（4）采用晶胞外推方法获得高精度的恒电位自由能。上述研究方案有望为恒电位自由能的量子化学计算提供标准的解决方案。通过项目的实施，还有望建立一套完整的、基于量子化学计算的高效率电化学反应理论模拟计算平台。

Accurately computing the free energies at constant electric-potentials is a fundamental and challenging open question, which is essential for the quantitative kinetics simulation of the electrochemical reactions. With the strong and urgent expectation for the application of electrocatalysis in energy transformation and storage, the issue is pushed forward as a hotspot in recent years. To overcome this difficult problem, we suggest in this proposal a systematical scheme of solution, which includes, (1) Based on the indepth exploration of energy difference between high-level quantum chemical methods and DFT theories, extract the rules for improving the accuracy of the DFT energy by introducing corrections either through explicit and analytical formula or machine learning. (2) To achieve the rapid evaluation of the potential energy, the energy gradients, as well as the surface work function for complex electrochemical systems, through the atomic neural network-based machine learning. To further accelerate the speed of calculation by employing the CPU-GPU heterogeneous platform. To reduce the labor work in refining the neural network by designing an Autonomous Learning Environment. (3) To realize the ultra-long term enhanced sampling molecular dynamics simulations on ultra-large scale models, for the reliable and simultaneous characterization of the whole free energy potential surfaces characterization along the main and side reaction routes. (4) To obtain potential related free energies for the critical structures in the reaction pathways through the “cell extrapolation method”. Hopefully, our research scheme could provide a standard solution to the calculation of the constant potential free energies, as well as provide a reliable, efficient, and complete toolkit for simulating the electrochemical reaction kinetics at the quantum chemical accuracies.