Collaborative Research: CDS&E Decision Framework for Predictive Simulation of Highly Non-Equilibrium Thermal Transport in Nanomaterials

murthy（德克萨斯大学奥斯汀分校），Mahadevan（范德比尔特大学），Strachan（普渡大学）在过去的几年里，在纳米尺度上实验探测物理现象的能力有了显著的提高。实验技术正在产生关于石墨烯和硅等材料中热传输的详细纳米级数据，但这些数据是否被正确解释存在重大问题。一个问题是，用于解释这些数据的理论对于所涉及的高度非均衡制度来说过于简单。另一个问题是，由于涉及的长度和时间尺度极小，纳米尺度测量存在显著的可变性。为了使用实验数据来改进理论，人们必须充分考虑测量的不确定性、纳米尺度制造技术的统计变异性和材料特性的变异性，并开发一种系统的方法来利用这些不确定数据识别当前模型中的知识差距。在这个项目中，我们建议合并两个迄今为止截然不同的领域，决策科学和声子输运模拟，为纳米尺度热输运理论的系统发展和改进创建第一个决策框架。这项工作将影响各种各样的消费应用，包括微电子、能量转换和能量存储。该项目开发的研究和模拟工具将通过普渡大学的nanoHUB传播给研究界以及德克萨斯大学奥斯汀分校、普渡大学和范德比尔特大学的研究生和本科生项目，同时还将提供教育模块和教程，以帮助扩大使用范围。这三所学校都将积极参与现有的高效项目，招募女性和未被充分代表的少数族裔加入他们的研究项目。德克萨斯大学奥斯汀分校将从这项工作中提取本科生研究项目，将其纳入其创新的35-in-5 ME女性计划，该计划旨在到2018年将机械工程专业大一新生的女性比例提高到35%。本提案的总体目标是对纳米材料中的高度非平衡声子输运有更深入的了解。在过去的几年里，随着我们探测纳米尺度热和电子输运的能力的提高，人们已经认识到，非平衡输运在许多新兴纳米技术和测量系统的性能中占据主导地位。微拉曼和微布里渊光散射等实验技术正在产生详细的波矢量解析声子输运数据，如果要释放它们的真正潜力，必须正确解释这些数据。虽然理论和计算预测也在同时发展，但在这种粒度上很少对测量和理论进行直接比较，而且对现有理论是否足够缺乏信心。该项目将结合石墨烯、块体和薄膜硅以及其他材料的光学激发声子输运的详细模型和实验，并结合贝叶斯决策框架，以发展更好的理论，正确解释新兴实验，设计更好的实验和模拟，并量化我们预测中的不确定性。该项目的一个独特之处在于使用经典分子动力学（MD）模拟来评估基于半经典声子玻尔兹曼输运方程（BTE）的声子输运模拟中的模型形式不确定性。此外，通过利用UT Austin在一个平行的NSF项目中进行的独特的微拉曼和微布里渊光散射测量，我们将有一个独一无二的机会获得空间和模式分辨声子输运数据，这可以显著提高我们模型的质量。研究计划包括使用贝叶斯框架(i)通过分子动力学校准声子散射建模中的小扰动假设来量化模型形式的不确定性（ii）将原子间电位校准为空间和光谱-

CBET 1404991/1404823/1404919Murthy (U Texas at Austin), Mahadevan (Vanderbilt), Strachan (Purdue)During the last few years, the ability to experimentally probe physical phenomena at the nanoscale has improved dramatically. Experimental techniques are producing detailed nanoscale data on heat transport in materials such as graphene and silicon, but there are significant questions about whether these data are being interpreted correctly. One issue is that the theory used to interpret these data is too simplistic for the highly non-equilibrium regimes involved. Another issue is that there is significant variability in nanoscale measurements because of the extremely small length and time scales involved. In order to use experimental data to improve theory, one must fully account for measurement uncertainty, statistical variability in nanoscale fabrication techniques, and variability in material properties, and develop a systematic way to identify knowledge gaps in current models using these uncertain data. In this project, we propose to merge two hitherto distinct fields, decision science and phonon transport simulation, to create the first-ever decision framework for the systematic development and improvement of nanoscale thermal transport theory. The work will impact a wide variety of consumer applications including microelectronics, energy conversion and energy storage. The research and simulation tools developed in the project will be disseminated to the research community and to the graduate and undergraduate programs at UT Austin, Purdue and Vanderbilt through Purdue's nanoHUB, along with educational modules and tutorials to help broaden use. All three schools will actively engage their existing and highly-effective programs to recruit women and underrepresented minorities into their research programs. UT Austin will draw undergraduate research projects from this work to integrate into their innovative 35-in-5 Women in ME initiative which aims to increase the percentage of women in their freshman Mechanical Engineering batch to 35% by 2018.The overall objective of this proposal is to develop a deeper understanding of highly non-equilibrium phonon transport in nanomaterials. During the last few years, as our ability to probe nanoscale thermal and electronic transport has improved, it has come to be recognized that non-equilibrium transport dominates the performance of many emerging nanotechnologies and measurement systems. Experimental techniques such as micro-Raman and micro Brillouin Light Scattering are producing detailed wave-vector resolved phonon transport data which must be interpreted correctly if their true potential is to be unleashed. Though theory and computational predictions are also being developed simultaneously, few direct comparisons of measurements and theory have been made at this granularity and there is little confidence that existing theories are adequate. The project will combine detailed models and experiments for optically-excited phonon transport in graphene, bulk and thin film silicon and other materials with a Bayesian decision framework to develop better theories, interpret emerging experiments correctly, design better experiments and simulations and to quantify the uncertainty in our predictions. A unique feature of the project is the use of classical molecular dynamics (MD) simulations to evaluate model form uncertainty in phonon transport simulations based on the semi-classical phonon Boltzmann transport equation (BTE). Furthermore, by exploiting unique micro-Raman and micro Brillouin Light Scattering measurements being performed UT Austin in a parallel NSF project, we will have a one-of-a-kind opportunity to obtain spatially and mode-resolved phonon transport data that can significantly improve the quality of our models.The research plan includes the use of a Bayesian framework to (i) quantify model form uncertainties due small-perturbation assumptions in the modeling of phonon scattering through calibration with molecular dynamics (ii) calibrate interatomic potentials to spatially and spectrally-