AI 中文总结
针对芯片热管理瓶颈，本文综述其热设计的模拟与测量技术，分析现存局限，探讨AI加速模拟等新兴方向，以推动热管理高效化、智能化。
AI 中文摘要
随着集成电路向更高功率密度、三维集成及异构封装方向发展，芯片热管理已成为限制器件性能、可靠性和寿命的关键瓶颈。本文系统综述了芯片热设计的数值模拟方法与实验测量技术，重点关注多尺度与多物理场耦合、热边界电阻测量、高热通量冷却相关的技术挑战。首先介绍了宏观及器件尺度的热模拟方法，包括等效热电路模型、有限元法、计算流体动力学，并讨论了微观尺度下声子输运理论与分子动力学的应用。随后分析了红外热成像、热反射法、拉曼测温法及嵌入式传感器的优势与局限性。当前的局限包括巨大的计算成本、不准确的多尺度耦合、昂贵的实验设施，以及传统冷却技术的物理极限。最后，探讨了新兴方向，包括AI加速热模拟、嵌入式微通道液冷、两相冷却、先进高导热材料及多物理场协同设计，旨在推动芯片热管理向更高效率与智能化发展。
英文摘要
As integrated circuits advance toward higher power densities, three-dimensional integration, and heterogeneous packaging, chip thermal management has become a key bottleneck limiting device performance, reliability, and lifetime. This article systematically reviews numerical simulation methods and experimental measurement techniques for chip thermal design, with particular emphasis on the technical challenges associated with multiscale and multiphysics coupling, thermal boundary resistance measurement, and high-heat-flux cooling. We first introduce macro- and device-scale thermal simulation methods, including equivalent thermal-circuit models, the finite element method, and computational fluid dynamics, and discuss the application of phonon transport theory and molecular dynamics at microscopic scales. We then examine the advantages and limitations of infrared thermography, thermoreflectance, Raman thermometry, and embedded sensors. Current limitations include the enormous computational cost, inaccurate multiscale coupling, expensive experimental facilities, and the physical limits of conventional cooling technologies. Finally, we discuss emerging directions, including AI-accelerated thermal simulation, embedded microchannel liquid cooling, two-phase cooling, advanced high-thermal-conductivity materials, and multiphysics co-design, with the aim of advancing chip thermal management toward greater efficiency and intelligence.
Journal refZHOU Junnian, ZHOU Feng, YUE Shengying. A review of simulation, measurement techniques, and development in chip thermal design. Acta Physica Sinica, 2026, 75(7): 070806
