Collaborative Research: Understanding Subatomic-Scale Quantum Matter Data Using Machine Learning Tools

现代量子物理学的一个中心目标是寻找新的系统和技术范例，利用物质的量子力学方面，而不是受到它们的限制。特别是，人们正在积极寻找新材料，这些材料表现出令人惊讶的物理性质，因为单个电子之间的强烈相互作用导致电子运动中的强关联，从而导致强关联量子物质。强关联量子物质(SCQM)的研究已经达到了一个临界点，经过过去十年的密集努力，产生了大量的实验数据。该领域的下一个突破将来自使用数据科学工具将这些实验数据与理论模型联系起来。然而，SCQM中数据驱动的挑战需要一种全新的数据科学方法，原因有两个：第一，量子力学成像是概率的；第二，从数据进行推理应该遵循基本的物理定律。因此，SCQM领域由数据驱动的新挑战要求NSF的十大想法中的两个：“不断增长的融合研究”和“驾驭数据革命”。该项目的目标是开发和传播机器学习(ML)工具，这些工具可以作为连接亚原子尺度SCQM实验中的数据革命和对SCQM的基本理论理解的双向高速公路。具体目标是：(1)开发用于位置空间图像数据的可解释的ML工具；(2)开发用于动量空间散射数据的无监督ML工具；(3)以从ML获得的洞察力为指导设计新的成像模式；以及(4)将ML工具与康奈尔高能同步加速器源(CHESS)光束线的操作中人机界面相结合。目标(1)和(2)触手可及，而(3)和(4)是更雄心勃勃的愿景，希望扩大到一个未来的研究所，涉及更多的学术机构和分散在全国各地的实验设施。该项目是国家科学基金会利用数据革命(HDR)大创意活动的一部分。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。

A central goal of modern quantum physics is to search for new systems and technological paradigms that utilize quantum mechanical aspects of matter rather than being limited by them. In particular, there is an active search for new materials that exhibit surprising physical properties because of strong interaction between individual electrons that leads to strong correlations in the motion of electrons and as a result, to strongly correlated quantum matter. The study of Strongly Correlated Quantum Matter (SCQM) has reached a tipping point through intense efforts over the last decade that have led to vast quantities of experimental data. The next breakthrough in the field will come from relating these experimental data to theoretical models using tools of data science. However, data-driven challenges in SCQM require a fundamentally new data science approaches for two reasons: first, quantum mechanical imaging is probabilistic; and second, inference from data should be subject to fundamental laws of physics. Hence the new data-driven challenges in the field of SCQM requires "Growing Convergent Research" and "Harnessing the Data Revolution", two of NSF's Ten Big Ideas. The objective of the project is to develop and disseminate machine learning (ML) tools that can serve as a two-way highway connecting the data revolution in SCQM experiments at sub-atomic scale to a fundamental theoretical understanding of SCQM. The specific goals are: (1) Develop interpretable ML tools for position space image data; (2) Develop unsupervised ML tools for momentum space scattering data; (3) Design new imaging modality guided by the insight gained from ML; and (4) Integrate ML tools with in-operando human interface to the Cornell High Energy Synchrotron Source (CHESS) beamline. Goals (1) and (2) are within reach, while (3) and (4) are more ambitious visions for scaling up to a future institute that can involve more academic institutions and scattering experiment facilities nationwide. This project is part of the National Science Foundation's Harnessing the Data Revolution (HDR) Big Idea activity.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.