Strong electron correlations in quantum chemistry: new approaches from machine learning, quantum computing and time-dependent quantum control

量子化学中的强电子相关性：机器学习、量子计算和瞬态量子控制的新方法

• 批准号: RGPIN-2020-04306
• 负责人: DeBaerdemacker, Stijn
• 金额: $ 2.48万
• 依托单位: University of New Brunswick
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=747843
• 关键词: Strong; electron; correlations; quantum; chemistry

The ability to universally solve quantum many-body systems at high accuracy with low computational cost would have unprecedented consequences for human kind, allowing us to design and accurately compute sophisticated molecules with important applications in chemical, bio-medical, and material science on computer chips, rather than in expensive experimental laboratories. While we are not quite there yet, promising techniques are on the horizon to scale up current theoretical methods to more than a few atoms per molecule. In the past five years, the recently relocated quantum chemistry group at the University of New Brunswick (QuNB), and collaborators have proposed a computationally tractable Geminal theory: an electronic structure method for strongly correlated molecules based on the premises of Lewis electron pairs, rather than molecular orbitals. The Geminal project emerged from a fruitful cross fertilization between theoretical chemistry and mathematical/nuclear physics with ideas from integrability and the seniority scheme. Notwithstanding the strong quantum correlations encoded in the Geminal wavefunction ansatz, the method has a good computational scaling, so other research groups are currently investigating it for larger systems. While the mathematical formulation of Geminal theory is currently investigated in the framework of a Tier-2 Canada Research Chair, the present Discovery Grant proposal goes beyond traditional electronic structure methods by embracing newly emerged technologies with disrupting potential to the field. In three Work Packages (WP), we will explore, develop and establish WP1(GemQC): the power of Quantum Computing to design a fully correlated Geminal theory; WP2(ML-DMET): the connection between Machine Learning and Density Matrix Embedding Theory; WP3(e-FMD): Fermionic Molecular Dynamics as an intuitive classical picture of electron bonding and dynamics probed by attosecond laser fields. The proposed research program will provide excellent HQP training opportunities in both hard and soft skills for 3 PhD students, 2 MSc students and several undergraduate students. The hard skills that come with theoretical electronic structure theory are mathematical abstraction and numerical modeling, both skills that are in high demand in academia and industry. The embedding within a CRC research group will allow for high-quality training on important soft skills such as (international) collaboration and independence, with strong respect for Equity Diversity and Inclusion. Canada is a global leader in both the fundamentals and applications of Machine Learning and Quantum Computing, so this research program will enhance Canada's leading position, both by delivering new theoretical methods published in top-tier academic journals, open source (quantum) computer software packages that implement those methods, as the training of HQP in that area.

以较低的计算成本高精度地普遍求解量子多体系统的能力将对人类产生前所未有的影响，使我们能够在计算机芯片上设计和精确计算复杂的分子，在化学、生物医学和材料科学中具有重要的应用，而不是在昂贵的实验实验室中。虽然我们还没有完全做到这一点，但有前景的技术即将出现，将当前的理论方法扩展到每个分子多个原子。 在过去的五年里，新不伦瑞克大学（QuNB）最近搬迁的量子化学小组和合作者提出了一种计算上易于处理的Geminal理论：一种基于路易斯电子对前提的强相关分子的电子结构方法，而不是比分子轨道。 Geminal 项目是理论化学和数学/核物理之间富有成效的交叉融合的产物，其思想来自可积性和资历方案。尽管双子座波函数 ansatz 中编码了很强的量子相关性，但该方法具有良好的计算扩展性，因此其他研究小组目前正在针对更大的系统研究它。 虽然双子座理论的数学公式目前正在加拿大二级研究主席的框架内进行研究，但目前的发现资助提案超越了传统的电子结构方法，采用了对该领域具有颠覆潜力的新兴技术。在三个工作包（WP）中，我们将探索、开发和建立WP1（GemQC）：利用量子计算的力量来设计完全相关的Geminal理论； WP2(ML-DMET)：机器学习和密度矩阵嵌入理论之间的联系； WP3(e-FMD)：费米子分子动力学是阿秒激光场探测的电子键合和动力学的直观经典图像。 拟议的研究计划将为 3 名博士生、2 名硕士生和几名本科生提供优秀的 HQP 硬技能和软技能培训机会。理论电子结构理论带来的硬技能是数学抽象和数值建模，这两种技能在学术界和工业界都非常需要。嵌入 CRC 研究小组将允许对重要软技能（例如（国际）合作和独立性）进行高质量培训，同时高度尊重股权多元化和包容性。 加拿大在机器学习和量子计算的基础和应用方面均处于全球领先地位，因此该研究计划将通过在顶级学术期刊上发表的新理论方法、开源（量子）计算机软件包来增强加拿大的领先地位实施这些方法，如总部在该领域的培训。