直播时间：2025年4月29日（周二）20:00-21:30

直播平台：

科学网APP

https://weibo.com/l/wblive/p/show/1022:2321325160765513400423

（科学网微博直播间链接）

科学网微博

科学网视频号

北京时间4月29日晚八点，iCANX Youth Talks第95期邀请到了香港科技大学助理郑琪野、北京科技大学教授邱琳担任主讲，北京大学教授王玮、北京大学助理教授杨林担任研讨嘉宾，北京大学教授张海霞担任主持人，期待你一起加入这场知识盛宴。

【嘉宾介绍】

郑琪野

香港科技大学

材料研究中的传热测量技术创新：分析方法与机器学习解决方案

【Abstract】

Effective thermal management is a critical challenge across microelectronics, energy technologies, aerospace, and advanced manufacturing. Accurate measurement of thermal conductivity (λ) is essential for optimizing materials in these fields, particularly as the complexity and heterogeneity of new materials increase. Rapid and precise thermal characterization is vital for advancing additive manufacturing, composite development, and high-performance electronics.First, I will discuss our recent advancements in the transient plane source (TPS) method (ISO 22007-2:2022), a widely adopted technique for bulk solid thermal conductivity measurement. Our work addresses the substantial systematic errors—up to 97%—that arise when measuring high-λ materials (λ > 30 W/(m·K)), primarily due to the limitations of conventional analysis models that neglect sensor/sample interfacial resistance (Rc) and heat conduction within the sensor itself. We introduce two new analytical frameworks: the Green’s function-based realistic sensor model (RSM) and the thermal quadrupole-based multilayer model (MLM), both explicitly accounting for sensor heat transfer and Rc effects. Additionally, we propose a derivative-based nonlinear regression approach that significantly improves robustness and accuracy, reducing λ measurement errors from as high as 97% to below 10% for high-λ materials, even under 60 times initial Rc variations. Our sensitivity analysis, leveraging singular value decomposition (SVD), guides optimal time interval selection and parameter identification, while 3D finite element modeling and experiments across a wide range of materials validate the superior performance of our models and fitting strategies. For applications with known sample heat capacity, a one-parameter fitting further enhances computational efficiency by 30–80%. Collectively, these advances offer a comprehensive, practical solution for accurate TPS measurements, with particular benefits for high-λ materials in demanding real-world study of metallic and composite materails.Secondly, I will introduce our development of a physics-informed fully convolutional network (PIFCN) architecture for solving forward and inverse problems in thermal conduction. Unlike traditional PINNs, which suffer from high computational cost and limited accuracy due to their fully connected architecture, PIFCNs leverage the localized connectivity of convolutional networks. This enables efficient, node-level matching between inputs and outputs, allowing direct discretization of PDEs via finite difference methods. Our results show that PIFCNs flexibly implement Dirichlet and Neumann boundary conditions, accurately predict temperature distributions, and can estimate unknown thermal diffusivity with over 99% accuracy even with sparse data and incomplete boundaries—substantially outperforming PINNs. Furthermore, the PIFCNs can be integrated with our high-throughput photothermal metrology system (SI-TI), which uses structured illumination and thermal imaging for rapid, parallel characterization of multiple samples. This synergy represents a promising direction for adaptive, high-throughput thermal characterization of complex materials.

在微电子、能源技术、航空航天及先进制造等领域，有效的热管理始终是一项关键挑战。随着新型材料的复杂性和异质性不断提升，准确测量材料的热导率（λ）对于材料性能优化尤为重要。快速且精确的热特性表征对于增材制造、复合材料开发和高性能电子器件的推进至关重要。首先，本次演讲将介绍本课题组在瞬态平面热源（TPS）法（ISO 22007-2:2022）方面的最新进展。TPS法是一种广泛应用于块体固体热导率测量的技术。本课题组针对高热导率材料（λ > 30 W/(m·K)）测量中出现的高达97%的系统性误差，提出了改进方法。该误差主要源于传统分析模型未能充分考虑传感器/样品界面热阻（Rc）及传感器内部的热传导。本课题组创新性地提出了两种新的分析模型：基于格林函数的真实传感器模型（RSM）和基于热四极子的多层模型（MLM），二者均明确考虑了传感器热传递及Rc效应。此外，本课题组提出了一种基于温度导数的非线性回归方法，显著提升了高热导率材料测量的鲁棒性和准确性，将λ测量误差从最高97%降低至10%以内，即使在Rc变化达60倍的情况下亦保持优异表现。本课题组结合奇异值分解（SVD）进行敏感性分析，为最佳拟合时间区间和参数识别提供指导；同时，通过三维有限元建模和多种材料的实验，验证了新模型和拟合方法的优越性。对于已知样品热容的应用，本课题组的一参数拟合方法可进一步提升计算效率30–80%。上述进展为精确TPS测量提供了全面而实用的解决方案，尤其适用于对金属及复合材料等高热导率材料的实际应用研究。其次，本次演讲将介绍本课题组开发的物理信息全卷积网络（PIFCN）架构，应用于热传导正反问题的求解。与传统的物理信息神经网络（PINNs）因全连接结构导致的高计算成本和有限准确性不同，PIFCN充分利用卷积网络的局部连接特性，实现了输入输出节点级的高效匹配，并可通过有限差分方法直接离散偏微分方程。研究结果表明，PIFCN能够灵活实现Dirichlet与Neumann边界条件，准确预测温度分布，并可在边界信息不完整、采样数据有限的情况下，以超过99%的准确率识别未知热扩散率，显著优于PINNs。此外，PIFCN还可与本课题组开发的高通量光热表征系统（SI-TI）结合，利用结构化光照与热成像，实现多样品的快速并行热特性测量。这种协同方式为复杂材料的自适应、高通量热表征提供了全新方向。

【BIOGRAPHY】

Dr. Qiye Zheng is current an Assistant Professor of Mechanical and Aerospace Engineering at The Hong Kong University of Science and Technology. He worked at UC Berkeley and Lawrence Berkeley National Lab from 2019-2022 after obtaining his PhD in 2017 and a 1-year postdoc in Materials Science and Engineering from the University of Illinois at Urbana-Champaign. His current research interests include nanoscale heat transfer, non-invasive thermal wave characterization, switchable thermal materials, dynamic thermal insulation, and thermal metrology. He has published 35 papers in prestigious journals such as Science, Applied Physics Review, Advanced Functional Materials, and Applied Energy with 2200+ citations and a h-index of 20. He is the recipient of NSFC Excellent Young Scientists Fund (Overseas, 2021). He serves as an independent reviewer for 62 distinct journals including Physical Review Letter, Advanced Functional Materials, and Nano Letter, a Topic Editor for Frontiers in Mechanical Engineering, and a Topical Advisory Panel member of the journal Crystals. He also served as session chair and poster judge for various conferences.

郑琪野博士于2022年起任香港科技大学机械与航空航天系助理教授，2012年于北京大学物理学院取得理学学士学位，2017 年获得美国伊利诺伊大学香槟分校材料科学工程博士学位并进行了一年的博士后研究，师从美国科学院院院士David Cahill教授，加入港科大前任加州大学伯克利分校博士后研究员，合作导师是美国工程院院士Ravi Prasher教授和UC Berkeley机械系主任Chris Dames教授。长期从事传热表征测量和功能材料相关的研究，在新型热测量技术、材料表征、和传热仿真方面积累了丰富的经验，近几年以通讯或第一作者在 Science, Applied Physics Review，International Journal of Heat and Mass Transfer， Advanced Functional Materials等期刊发表学术论文，相关 SCI 文章35篇，美国专利2项， Web of Science 引用2100余次， h 因子20。在新型光学高通量热测量技术，电子束纳米测温技术以及热功能材料方面取得了一定成果。其研究受EurekAlert!, Electronics Weekly, ScienceDaily, Physics World等媒体报道。目前担任美国ASME K9委员会成员，受邀为Physical Review Letter, Nano Letter, Advanced Functional Material ，Applied Energy等62种期刊审稿240余次，多次在香港、大陆和韩国和美国进行受邀报告。

邱琳

北京科技大学

“2.5D”全碳sp2/sp3杂化界面实现超高界面导热

【ABSTRACT】

Towards the development of next-generation integrated circuits, carbon-based thermal interface materials (TIMs) with excellent thermal conductivity play a crucial role, especially single-layer and multi-layer graphene, which has become a research hot spot due to its excellent high temperature resistance and heat dissipation characteristics. However, when carbon-based TIMs are transferred onto the chip, their performance is severely compromised due to the extremely low interface thermal conductivity (GI) between them and the chip, which hinders the heat dissipation efficiency of the chip. This interface is called a ex situ interface because TIMs are not directly grown on chip materials. For example, once graphene TIM is transferred to Si/SiO2 substrate, the carrier mobility will significantly decrease due to non-covalent interface interactions. Therefore, by directly growing carbon-based TIMs on the chip, this problem can be effectively avoided. For the widespread application of the fourth-generation semiconductor material diamond in integrated circuits, carbon materials grown directly on diamond, namely all carbon interfaces, are undoubtedly the most ideal heat dissipation solution in the future.This presentation will introduce that our research group has developed an ultrafast quenching process from the perspective of optimizing the all-carbon interface, and successfully prepared a 2.5D all-carbon interface with a large number of sp2/sp3 hybrid C-C bonds. This method utilizes direct current arc plasma jet for rapid heating, forming a high-temperature C/Ni solid solution on the surface of diamond. During the precipitation of C, a large number of metastable covalent sp2 or sp3 bonds are formed. In addition, ultra-fast quenching through water cooling can preserve a large number of metastable structures, allowing the graphene layer to covalently bond tightly to the diamond surface. The sp2/sp3 hybrid bond not only provides a high-speed interface heat transfer pathway, but also significantly increases the vibrational density of states (VDOS) of low-frequency phonons on both sides of the interface, leading to an increase in the heat carried by phonons. The research results are a breakthrough in efficient all-carbon TIM and have significant implications for the development of all-carbon devices and circuits.

在下一代集成电路的发展过程中，具有卓越导热性能的碳基热界面材料（Thermal interface materials, TIMs）扮演着至关重要的角色，尤其是单层和多层石墨烯因其出色的耐高温和散热特性而成为研究的焦点。然而，当碳基TIMs被转移到芯片上时，由于其与芯片之间的界面热导（GI）极低，其性能会遭受严重损害，从而妨碍了芯片的散热效率。这种界面，由于TIMs并非直接在芯片材料上生长获得，因此被称为非原位界面。例如，石墨烯TIM一旦转移到Si/SiO2基底上，由于非共价界面作用，载流子迁移率会大幅下降。因此，通过在芯片上直接生长碳基TIMs，可以有效避免这一问题。对于集成电路中第四代半导体材料金刚石的广泛应用而言，金刚石上直接生长的碳材料，即全碳界面，无疑是将来最理想的散热解决方案。本次演讲将介绍我们课题组从全碳界面优化的角度出发，开发了一种超快淬火工艺，成功制备出了具有大量sp2/sp3杂化C-C键的2.5D界面。该方法利用直流电弧等离子体射流快速加热，在金刚石表面形成高温C/Ni固溶体，在析出C的过程中形成大量的亚稳态共价sp2或sp3键。此外，通过水冷的超快速淬火可以保留大量的亚稳态结构，从而使石墨烯层通过共价紧密地结合在金刚石表面。sp2/sp3杂化键不仅提供了高速的界面热传递途径，而且使界面两侧低频声子的振动态密度（Vibrational density of states, VDOS）显著增加，声子携带的热量增加。该项研究结果是高效全碳TIM的突破，对全碳器件和电路的发展具有重要意义。

【BIOGRAPHY】

Qiu Lin, Professor at Beijing University of Science and Technology, recipient of the National Excellent Youth Science Fund (2022), Beijing Nova Program (2020), and a distinguished scholar at University of Science and Technology Beijing. Her research focuses on advanced material thermophysical property measurement methods and heat transport mechanisms, wearable sensors and devices for physiological monitoring, high-throughput design calculations and preparation of low-dimensional carbon materials. She has hosted four national Natural Science Foundation projects, National Key R&D Program project, National Foreign Expert projects, Beijing Natural Science Foundation, Beijing Nova Program, Beijing Nova Program cross-collaboration project, and multiple enterprise projects. She has edited one English monograph and two textbooks and published 96 SCI-indexed papers on international high-level journals such as Phys. Rep., Angew. Chem. Int. Ed. (3), SmartMat (2) as the first/corresponding author, with an other-citation of over 3300 and H-Index of 33. 4 papers have been selected as ESI highly cited papers (2 have been selected as hot papers), 2 papers have been selected as journal covers/frontispiece. She has got 14 authorized invention patents, 2 software copyrights and 1 provincial and ministerial level award. She also holds positions as Associate Editor of Rev Sci. Instrum., Editor of Thermochimica Acta, Editorial Board of Carbon, Appl. Therm. Eng., Sci. Rep., J. Therm. Sci., ACS Appl. Engin. Mater.

邱琳，北京科技大学教授，北京市科技新星(2020)，北科鼎新学者。研究方向为先进材料热物性评价方法及热输运机理、生理监测用可穿戴传感器及设备、低维碳材料高通量设计计算及制备。主持国家自然科学基金4项，国家重点研发计划课题、国家级外专项目、北京市自然科学基金、北京市科技新星计划、科技新星交叉合作课题等省部级以上纵向项目和多项企业课题。主编英文专著1部、教材2部，以第一/通讯作者在Phys. Rep.、Angew. Chem. Int. Ed.(3)、SmartMat(2)等国际高水平期刊上发表SCI论文96篇，总他引逾3300次，H-Index为33，4篇曾入选ESI高被引论文(2篇曾入选热点论文)，2篇文章入选期刊封面/卷首插画，获授权发明专利14项、软件著作权2项、省部级奖励1项。担任Rev. Sci. Instrum.副主编，Thermochimica Acta编辑，Carbon、Appl. Therm. Eng.、Sci. Rep.、J. Therm. Sci.、ACS Appl. Engin. Mater.编委。

【主持人】

张海霞

北京大学

【研讨嘉宾】

王玮

北京大学

杨林

北京大学