Machine Learning Surrogates for Simulating Quantum Materials

量子材料具有不同寻常的磁性和电学特性，如果能够被理解和控制，将给技术领域带来革命性的变化，例如，能够实现高能效的电气系统或更快的电子设备。该项目有助于开发计算效率高、功能强大的“虚拟实验室”，在虚拟实验室中，量子材料系统可以比实验更便宜、更有针对性地进行研究。这将使候选量子系统的高通量筛选成为可能，从而加速科学发现和技术转让。实现这一目标的障碍在于，尽管研究人员可用的计算资源持续快速增加，但对具有科学和技术意义的量子材料系统的高保真模拟仍然遥不可及。关键因素是这样的量子系统的规模和需要准确地处理强相关效应来描述真正的量子现象的出现。该联盟-量子-催化剂项目的目标是通过将现代机器学习方法与对量子材料系统的化学和物理学的深刻见解相结合，显着扩大可以可靠和准确模拟的量子系统的范围。

Quantum materials have unusual magnetic and electrical properties that, if understood and controlled, will revolutionize the technology sector, e.g. enable highly energy-efficient electrical systems or faster electronic devices.This project contributes to the development of computationally efficient and robust "virtual laboratories" in which quantum materials systems can be studied far more cheaply and in a more targeted way than in experiments. This will enable high through-put screening of candidate quantum systems and thus accelerate scientific discovery and technology transfer. The barrier to this goal is that, despite the continuing rapid increase in computational resources available to researchers, high-fidelity simulation of quantum materials systems of scientific and technological interest remains out of reach. The critical factors are the scale of such quantum systems and the need to accurately treat strong correlation effects to describe the emergence of genuine quantum phenomena. The objective of this Alliance-Quantum-Catalyst project is to significantly expand the range of quantum systems that can be simulated reliably and accurately by merging modern machine-learning methodology with deep insights into the chemistry and physics of quantum materials systems.