CAREER: Local Correlation Approaches for High-Level Density Functional Theory Simulations of Large Systems

佛罗里达州立大学的陈黄获得了化学系化学理论、模型和计算方法项目的支持，以开发新的理论和计算方法来研究大而复杂的分子和材料的电子性质。在电子水平上为大型系统建模的能力对于清洁能源、高效电池和新型数据存储设备的分子和材料的工业和技术设计至关重要。然而，用最精确的电子结构方法来研究大型系统是具有挑战性的，因为计算成本随着系统的规模呈指数增长。黄博士和他的团队正在开发新的量子力学嵌入方法，以分而治之的方式确定大系统的电子性质。这些方法的重点是准确描述电子之间的微观相互作用-它们的量子关联--在尺度上。寻找解决这个“电子关联问题”的新方法被认为是21世纪科学的重大挑战之一。黄博士的新算法和方法将在最先进的、开源的量子化学代码中实现，并提供给更广泛的研究社区。结合他的研究，黄博士正在培训一批来自工程、化学、物理和材料科学的本科生和研究生，通过一门新的基于问题的密度泛函理论(DFT)课程，有效地利用电子结构方法。课程讲座和YouTube教程旨在降低学生将DFT模拟整合到他们的研究中的门槛。与佛罗里达州立大学的推广项目合作，黄博士正在让初中生和高中生参与他的研究，以提高他们追求STEM职业的兴趣。这个项目的目标是开发新的量子力学嵌入方法，以获得精确的反应能量和电子结构，这些反应能量和电子结构可以扩展到大型化学系统。嵌入方法的基本思想是利用感兴趣区域内的高级电子结构方法，并用低级别、计算高效的方法处理系统的其余部分(环境)。黄博士和他的团队正在向三个新方向推广嵌入方法：(I)使用类赝势方法将密度矩阵嵌入推广到金属系统；(Ii)扩展XCPP(交换相关势修补)方法以逐个原子的方式构造大系统中的RPA(随机相近似)关联能和势；以及(Iii)通过施加适当的矩约束来加速随机抽样方案计算的RPA能量差的收敛。这些新技术正被应用于具有重大基础和技术意义的挑战性问题，包括研究金属铜在甲醇合成中的作用；YBa2Cu3O7/La2/3Ca1/3MnO3(超导/铁磁)界面的电荷转移和磁耦合；以及用于多相催化的CeO2中氧的传输、存储和释放。这些方法在ABINIT、PSI4和CP2K量子化学包中以开放源代码的形式实现。该教育计划的重点是开发一门基于问题的DFT课程，包括YouTube教程，使学生能够发展这一广泛使用的当代分子和材料建模工具的专业知识。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Chen Huang of Florida State University is supported by an award from the Chemical Theory, Models and Computational Methods program in the Division of Chemistry to develop new theoretical and computational methods for investigating the electronic properties of large, complex molecules and materials. The ability to model large systems at the electronic level is essential for the industrial and technological design of molecules and materials for clean energy, efficient batteries, and novel data storage devices. However, it is challenging to investigate large systems with the most accurate electronic structure methods, since computational cost grows exponentially with system size. Dr. Huang and his group are developing new quantum mechanical embedding methods for determining the electronic properties of large systems in a divide-and-conquer manner. The methods are focused on accurately describing the microscopic interactions among electrons---their quantum correlations---at scale. Finding new ways to address this "electron correlation problem" is regarded as one of the grand challenges of 21st century science. Dr. Huang's new algorithms and methods will be implemented in state-of-the-art, open source quantum chemistry codes and made available to the broader research community. Integrated with his research, Dr. Huang is training a diverse group of undergraduate and graduate students from engineering, chemistry, physics, and materials science to effectively utilize electronic structure methods, through a new problems-based density functional theory (DFT) course. The course lectures and YouTube tutorials are aimed at lowering the barrier for students to integrate DFT simulations within their research. Working with outreach programs at FSU, Dr. Huang is involving middle school and high school students in his research to increase their interest in pursuing STEM careers.The goal of this project is to develop new quantum mechanical embedding methods to obtain accurate reaction energies and electronic structures that scale to large chemical systems. The basic idea of embedding methods is to utilize a high-level electronic structure method within a region of interest, and to treat the rest of the system ("the environment") with a low-level, computationally-efficient method. Dr. Huang and his group are extending embedding methods in three new directions: (i) generalizing density-matrix embedding to metallic systems using a pseudopotential-like approach; (ii) extending the XCPP (exchange-correlation potential patching) methodology to construct RPA (random phase approximation) correlation energies and potentials in large systems in an atom-by-atom manner; and (iii) accelerating the convergence of RPA energy differences calculated with a stochastic sampling scheme by imposing appropriate moment constraints. The new techniques are being applied to challenging problems of significant fundamental and technological interest, including investigating the role of metallic copper in methanol synthesis; charge transfer and magnetic coupling at YBa2Cu3O7/La2/3Ca1/3MnO3 (superconducting/ferromagnetic) interface; and transport, storage, and release of oxygen in ceria for heterogeneous catalysis. The methodologies are being implemented as open source code in the ABINIT, Psi4, and CP2K quantum chemistry packages. The educational plan is focused on developing a problems-based DFT course including YouTube tutorials, to enable students to develop expertise in this widely-used tool of contemporary molecular and materials modeling.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.