Sessions Index

Bio:
2011 Ph.D., University of Hamburg, Germany 2011-2017 Postdoctoral Fellow, RIKEN, Japan 2017-2021 Researcher, Nissan Arc, Japan 2021-now CEO and Founder, ANVOS Analytics Co. Ltd., Japan

Abstract:
The effect of an AlN heat spreader formed on bow-tie-shaped GaN-HEMT devices is characterized by measuring the temperature distribution under current flow. The temperature of the GaN epilayer was extracted via Raman thermometry using the anti-Stokes-to-Stokes ratio. The peak temperature of 179°C n the narrowest channel region (hot spot) was reduced to 126°C by the cap-AlN heat spreader at 0.9 A/mm. A gradual temperature decrease from the center was observed, and the distance to reach a constant temperature was shorter with the cap-AlN heat spreader. The temperature measurement metrology under current flow is useful for analyzing the effectiveness of capped heat spreader.

 

Bio:
Hello, I’m Youngmin Park, a master’s student at Jeonbuk National University. My research focuses on subcooled flow boiling heat transfer.

Abstract:
INTRODUCTION
Flow boiling is a representative two-phase heat transfer mechanism that provides high heat transfer performance under high heat flux conditions. Owing to this advantage, it has been widely applied to highly integrated thermal management systems, including power electronics cooling, battery thermal management systems, and compact heat exchangers. In flow boiling systems, surface wettability plays an important role in determining bubble nucleation, growth, coalescence, and detachment behavior, thereby significantly affecting the overall heat transfer characteristics. In this study, numerical simulations are performed by varying the wall contact angle to investigate the effect of surface wettability on flow boiling heat transfer.

Numerical Methods
This study numerically investigates the effect of surface wettability on bubble dynamics and heat transfer characteristics in subcooled flow boiling within a vertical rectangular channel. The computational domain consists of a two-dimensional fluid region bounded by two opposite heated copper walls. A uniform inlet condition is applied at the channel inlet, while a constant pressure condition corresponding to atmospheric pressure is imposed at the outlet. All walls, except for the heated surfaces, are treated as adiabatic. The simulations are conducted using ANSYS Fluent. The Volume of Fluid (VOF) multiphase model, coupled with the Continuum Surface Force (CSF) model, is employed to capture the liquid–vapor interface dynamics. The Lee phase-change model is used to calculate mass transfer due to evaporation and condensation. Turbulence effects are modeled using the standard k–ε turbulence model. The wall contact angle, θ_w, is defined as the equilibrium contact angle at the liquid–solid interface and is varied as 5°, 80°, and 155° to represent different surface wettability conditions. FC-72 and copper are used as the working fluid and solid wall material, respectively. The simulations are performed under fixed operating conditions: an inlet mass velocity of 445.75 kg m⁻² s⁻¹, an inlet temperature of 300.97 K, a gauge outlet pressure of 0 Pa, and a uniform wall heat flux of 146,301 W m⁻².

RESULTS
The numerical results showed that the bubble nucleation sites, bubble growth, and detachment behavior varied significantly with the wall contact angle. At higher contact angles, vapor accumulation near the heated wall became more pronounced, and bubbles tended to remain attached to the wall for a longer period. This behavior led to a local increase in wall temperature due to the formation of vapor-rich regions near the heating surface. In contrast, as the contact angle decreased, bubble detachment from the wall occurred more readily, which promoted liquid rewetting of the heated surface. As a result, the wall temperature was maintained at a relatively lower level, indicating improved heat transfer characteristics.

Conclusions
The present study confirms that surface wettability is a key factor governing bubble dynamics and heat transfer performance in subcooled flow boiling within a vertical channel. Variations in the wall contact angle strongly affect the interfacial behavior, including bubble nucleation, growth, attachment, and detachment, which in turn influences the thermal response of the heated wall. Enhanced bubble detachment at lower contact angles contributes to improved heat transfer performance by reducing vapor accumulation near the wall. These findings suggest that controlling surface wettability can be an effective approach for improving flow boiling heat transfer in confined vertical channels.

Acknowledgment
This research was partly supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2025-00557360), and by the BK21 FOUR (Fostering Outstanding Universities for Research, No.2120240815466) funded by the Ministry of Education (MOE, Korea) and National Research Foundation of Korea (NRF).

 

Bio:
Currently serving as a Senior Thermal Engineer at ASE Group, Chung-Li Branch. Responsible for conducting thermal simulations and thermal measurements, as well as providing comprehensive solutions to package-level thermal issues. Actively involved in optimizing thermal performance and reliability for advanced semiconductor packaging technologies.

Abstract:
As semiconductor devices continue to scale toward higher performance and finer dimensions, advanced packaging technologies face increasing challenges in thermal management and long‑term reliability. Flip‑chip bonding, as a key process step, is highly sensitive to thermal stress, and the conventional Mass Reflow (MR) process often induces excessive warpage due to global substrate heating, particularly for fine‑pitch bump structures. These thermal‑stress issues limit the applicability of MR in next‑generation packaging. Laser Assisted Bonding (LAB) has emerged as a promising alternative, in which laser irradiation generates localized heating near the chip and bump regions, enabling rapid solder melting while reducing overall temperature rise, process time, and the heat‑affected zone. However, LAB may still encounter uneven temperature distribution, local overheating, or insufficient solder melting if laser parameters are not properly controlled, highlighting the need for accurate prediction of transient thermal behavior.

In this study, a comprehensive three‑dimensional transient thermal modeling is established to analyze the thermal behavior of LAB. By incorporating realistic bump distribution, the model improves geometric fidelity and reduces potential bias from oversimplification. The predicted temperature distribution throughout the package, including both global and local bump temperatures, shows excellent agreement with experimental measurements, confirming that the model can accurately reproduce thermal characteristics observed in practice. In addition, the model explicitly captures the full temporal evolution of bump temperature during both heating and cooling phases, providing a more complete representation of transient thermal behavior. These temperature data, at both the package level and individual bump level, can also be utilized in subsequent thermal stress analysis.

Overall, this study provides a more comprehensive framework for LAB evaluation by integrating realistic bump geometries and simulating the complete heating–cooling temperature evolution. The validated modeling approach not only improves the completeness and accuracy of LAB simulation, but also offers transient thermal results that can serve as a foundation for future stress and reliability studies.

 

Bio:
I am Ching-Heng Lin from the Department of Mechanical and Electro-mechanical Engineering at National Sun Yat-sen University. My current research focuses on thermal management and CFD analysis of two-phase transport in ultra-thin silicon-based vapor chambers. In this study, I use ANSYS Fluent to investigate how different filling volumes affect vapor transport, liquid return, pressure distribution, and vapor–liquid distribution in vapor chambers.

Abstract:
A three-dimensional CFD model was developed using ANSYS Fluent to investigate filling-volume effects on two-phase transport in an ultra-thin silicon-based vapor chamber. Based on previously reported experimental conditions, 2.50, 5.00, and 7.50 μL filling volumes were examined under 0.93–4.87 W. The 5.00 μL case showed the most balanced transport behavior, while insufficient or excessive filling may cause liquid starvation or vapor-core suppression. Since ΔT deviations remain at high power, the model is mainly used for internal mechanism analysis and relative trend evaluation.

 

Bio:
I'm currently pursuing a master’s degree in the Department of Materials and Optoelectronic Science at National Sun Yat-sen University, Taiwan. My research focuses on semiconductor packaging, power module reliability, and thermal management, with particular interests in high-thermal-conductivity epoxy molding compounds, package warpage behavior, and material characterization. His work involves both experimental analysis and simulation, including thermal, mechanical, and reliability evaluation for advanced power packaging applications.

Abstract:
With the rapid development of electric vehicles and high-power electronic systems, power modules are increasingly designed toward higher power density and higher operating temperature. As a result, the performance of encapsulation materials in thermal conduction, warpage control, and reliability has become increasingly important. In this study, epoxy molding compounds (EMCs) with different thermal conduction capabilities were evaluated for power module encapsulation applications, and encapsulated samples were fabricated using a transfer molding process. Material properties were first characterized by DSC, TMA, DMA, and LFA to obtain specific heat, coefficient of thermal expansion, elastic modulus, and thermal conductivity. Shadow Moiré was then used to measure the warpage behavior of the encapsulated modules, while SAT and thermal cycling tests were conducted to evaluate interfacial defects and reliability. In addition, a finite element model was established to analyze the influence of material properties on the thermomechanical behavior of power modules.

The results show that the high-thermal-conductivity EMC exhibited better thermal diffusion capability, reducing internal heat accumulation and improving the uniformity of temperature distribution within the encapsulated module. The LFA results indicated that its thermal conductivity was approximately 2.9–3.2 W/m·K, significantly higher than that of the conventional EMC at 0.85–0.9 W/m·K. This confirms that material composition and filler content strongly influence the thermal conduction capability of EMCs. However, package warpage was also affected by the coefficient of thermal expansion, elastic modulus, and glass transition behavior. The Shadow Moiré results showed that different EMCs exhibited distinct warpage variations during heating and cooling, mainly due to thermal expansion mismatch among the encapsulation material, substrate, and chips.

Furthermore, the SAT and thermal cycling test results showed that EMCs with better material properties could reduce the risk of interfacial defects and improve the long-term thermal cycling reliability of power modules. Overall, this study establishes the relationship among EMC thermal conduction capability, thermomechanical properties, package warpage, and reliability, and can serve as a reference for material selection and structural design of high-power module encapsulation.

 

Bio:
Muhammad Firman Friyadi is a first-year Ph.D. student at the Academy of Innovative Semiconductor and Sustainable Manufacturing (AISSM), National Cheng Kung University (NCKU), Tainan, Taiwan, under the supervision of Prof. Chi-Hua Yu. His research focuses on AI-driven thermal management for high-power and 3D-stacked electronic packages, combining computational fluid dynamics with deep learning and reinforcement learning to automate the design of microchannel liquid-cooling structures. His current work, conducted in collaboration with ASE Group, develops CFD-validated surrogate models and design-optimization frameworks for next-generation cooling solutions.

Abstract:
The relentless scaling of high-power and 3D-stacked processors for AI workloads has made onchip thermal management a primary limiter of performance and reliability. Embedded
microchannel liquid cooling offers a promising path, but designing channel geometries that
suppress localized hotspots remains slow and largely manual, as each candidate design must be
evaluated through computationally expensive computational fluid dynamics (CFD) simulation.
This work presents an integrated, AI-driven framework that couples deep-learning surrogate
prediction with reinforcement-learning (RL) optimization to automate thermal-aware
microchannel design, directly advancing the conference theme of energy-efficient AI from
computing to system-level application.
A central enabler of the framework is a large, consistently generated simulation database. We
have completed the generation of nearly 10,000 CFD cases spanning a wide parametric space of
microchannel unit-cell designs, including variations in channel width, channel spacing, angle, and
density, with each case paired to its resulting temperature and thermal-gradient field. To our
knowledge this is among the largest dedicated datasets for microchannel cooling design, and it
provides the statistical foundation required for reliable data-driven modeling.
Building on this dataset, the framework contributes two coupled AI capabilities. First, a
conditional generative adversarial network (Pix2Pix) surrogate predicts the full thermal-gradient
field, together with average, minimum, and maximum temperatures, directly from a channelpattern geometry. Once trained, the surrogate replaces full CFD evaluation during design
exploration, reducing per-design assessment from minutes of simulation to near-instant inference
and enabling rapid screening of thousands of candidates. Second, a reinforcement-learning agent
uses this surrogate as its evaluation engine to generate optimized designs: given a user-defined
thermal target, the agent iteratively adjusts unit-cell parameters and proposes the best design
candidate, steering coolant toward hotspot regions to improve temperature uniformity.
A key strength of the approach is that design quality is assessed directly against CFD-based flow
and thermal behavior, ensuring that the optimized geometries reflect realistic coolant flow and
heat-removal performance rather than simplified estimates. By learning from high-fidelity
simulation data, the framework captures the coupling between channel geometry, coolant mass
flow, and the resulting temperature distribution that governs effective cooling.

 

Bio:
Dr. Yonhua Tzeng is a Distinguished Professor of the College of Semiconductor Research of National Tsing Hua University in Hsinchu, Taiwan. Before joining NTHU, Prof. Tzeng was a University Chair and Distinguished Professor of Microelectronics at NCKU, where he retired on August 1, 2025. At NCKU, Professor Tzeng served as VP for Research and Dean of College of EECS. Professor Tzeng has more than 30 years of experiences in CVD diamond production and applications. His accomplishments were published in peer reviewed journals and awarded with more than 50 invention patents. Prof. Tzeng was elected Life Fellow of IEEE and served as the President of IEEE Nanotechnology Council and Chair of the Fellow Evaluation Committee. Prof. Tzeng was elected fellow of US National Academy of Inventors.

Abstract:
Polycrystalline silicon carbide (poly SiC) was used for the fabrication of microchannel heat sinks (MCHS). The MCHS is compared with single crystalline semi-insulating SiC (SI-SiC) based MCHS. The polycrystalline SiC was grown by physical vapor transport, which is like the growth process for single crystalline SiC except that a polycrystalline SiC seed substrate is used to grow a transparent thick polycrystalline SiC for slicing into the desired thickness. No intentional dopant was added. The size is restricted by the size of the seed SiC substrate and the PVT dimensions making it promising to produce square or round substrate of 30 cm by 30 cm or 30 cm diameter or larger. The square shaped SiC is desirable for applications, which use square SiC substrates because more pieces of usable square SiC substrates can be cut from a 30 cm x 30 cm square substrate than a piece of 30 cm diameter SiC. The process window for growing polycrystalline SiC is wider than that for growing single crystalline SiC. Besides, the yield of producing high-quality and transparent polycrystallne SiC is higher than that of single-crystalline SiC while the production cost is about 30% to 50% lower than that for crystalline SiC. The polycrystalline SiC is provided by Global Wafers Co,. Ltd. (Hsinchu, Taiwan) with a trade name of “Spoly-SiC”.

 

Bio:
Yinglei Ren is a principle engineer in Intel cooperation focusing on signal integrity, power integrity and PCB technology. She received her M.S.E.E. and B.S.E.E. from Shanghai Jiao Tong university in 2005 and 2002 correspondingly.

Abstract:
Over the past years, artificial intelligence (AI) has experienced rapid and transformative growth. Printed circuit boards (PCBs) play a critical role in enabling the performance and reliability of AI systems. As signal speeds continue to increase and thermal design power (TDP) rises in AI designs, certain PCB manufacturing details that were previously overlooked are now directly impacting system performance, making precise design control and manufacturing quality more important than ever.
This paper introduces two examples of how PCB manufacturing details impact signal integrity (SI) / power integrity (PI) performance in AI designs.