DMREF: Collaborative Research: Machine learning exploration of atomic heterostructures towards perfect light absorber and giant piezoelectricity

摘要非技术描述：最近发现的原子薄二维（2D）材料，如石墨烯、过渡金属二硫族化合物、黑磷等，为通过垂直堆叠（即二维异质结构）控制原子水平的材料开辟了新的机会。二维异质结构材料设计的艺术还处于起步阶段，我们可以通过实验获得的二维单层的潜在库大约有1000个，仅仅是N个单层的最低能量堆叠就可以产生1000N个可能的异质结构。我们的项目旨在通过新颖的机器学习（ML）引导的密度泛函理论（DFT），对有用的2D异质结构的发现过程产生变革性影响，例如可见光谱中的完美光吸收和巨大的压电性。我们组建了一个由数据科学和机器学习应用、二维材料及其光学特性建模、二维异质结构的分子束外延生长以及材料和器件表征方面的专家组成的团队。这些新的二维异质结构的成功展示将开创一个高效和有目的的材料设计方法论的新时代。为所有参与该提案的群体制定了一个三管齐下的扩大参与计划，包括本科研究、社区外展和暑期项目。技术描述：对二维原子晶体的研究最近集中在它们的异质结构上，这一新兴领域的进步已经导致了诸如超导性和磁性等令人着迷的发现。然而，成千上万种不同的二维层状材料及其排列构成了几乎无限的异质结构组合。本研究将开发一种新的ml引导的DFT框架，结合物理动机原子描述符，将数据科学应用于寻找具有目标属性的设计异质结构。作为概念验证，我们将通过优化填充带和空带之间的带嵌套来展示具有完美光吸收的异质结构，以及通过设计电负性偶极矩来展示巨大的压电性。这些具有目标性质的异质结构将通过超清洁的最先进的MBE方法生长，并对其吸收和压电系数进行表征。实验和理论之间的佐证将指导对所提出的ML和DFT模型和总体策略的可能改进。这些新的二维异质结构的成功展示将开创一个高效和有目的的材料设计方法论的新时代。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

AbstractNontechnical Description: Recent discovery of atomically thin two-dimensional (2D) materials, such as graphene, transition metal dichalcogenides, black phosphorus, among many others, has opened new opportunities to atomic level materials control through vertical stacking, i.e. 2D heterostructure. The art of 2D heterostructure materials design is at its nascent stage, and with the potential library of 2D monolayers that we can access experimentally amounting to about 1000, just the mere lowest energy stacking of N monolayers would then lead to 1000N possible heterostructures. Our program seeks a transformational impact on the discovery process of useful 2D heterostructures, such as perfect light absorption in the visible spectrum and giant piezoelectricity, through novel machine-learning (ML)-guided density functional theory (DFT). We have assembled a team comprising of experts in data science and the application of ML, modeling of 2D materials and their optical properties, molecular beam epitaxial growth of 2D heterostructures, and materials and device characterization. The successful demonstration of these new designer 2D heterostructures would usher in a new era of efficient and purposeful materials design methodology. A three-pronged broadening participation program for all participating groups in this proposal is planned, which includes undergraduate research, community outreach, and summer programs.Technical Description: Research in two-dimensional atomic crystals has recently focused on their heterostructures, and the advancements in this emerging field has already led to fascinating discoveries such as superconductivity and magnetism. However, thousands of different 2D layered materials and their permutations amount to almost infinite heterostructure combinations. This research will develop a novel ML-guided DFT framework, in conjunction with physically motivated atomistic descriptors, which applies data science in the search for designer heterostructures with targeted properties. As a proof-of-concept, we will demonstrate heterostructures with perfect light absorption through optimizing the band nesting between the filled and empty bands as well as giant piezoelectricity through engineering the electronegativity dipole moments. These heterostructures identified with the targeted properties will be grown with ultra-clean state-of-the-art MBE approaches, and their absorption and piezoelectric coefficients characterized. Corroboration between experiments and theory will then instruct on possible improvements to the proposed ML and DFT models and overall strategy. The successful demonstration of these new designer 2D heterostructures would usher in a new era of efficient and purposeful materials design methodology.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.