Plenary Talks

The rapid growth of artificial intelligence (AI) has accelerated the demand for advanced packaging technologies that provide higher bandwidth, greater integration density, and reduced signal delay. Beyond chip-level scaling, recent advances in 2.5D/3D integration, chiplet architectures, heterogeneous stacking, and emerging substrate materials such as glass have become essential for next-generation AI semiconductor systems. Historically, back-end-of-line (BEOL) interconnect technologies have evolved to reduce RC delay through material and structural innovations, while continuously addressing key reliability issues such as electromigration, stress-induced failure, and dielectric degradation.

Advanced packaging can be viewed as an extension of these interconnect technologies into complex package-level geometries. As interconnect pitches continue to shrink toward the sub-micron scale and package integration becomes more heterogeneous, reliability assessment must go beyond conventional electrical testing. Thermomechanical stress, interfacial delamination, warpage-related degradation, oxidation, and moisture absorption are becoming increasingly important reliability factors in advanced packaging architectures.

This talk presents reliability challenges in advanced BEOL and packaging. Ensuring the reliability of chips and advanced packages requires a comprehensive evaluation framework that integrates electrical, environmental, and mechanical reliability perspectives, together with a fundamental understanding of materials properties. Such an approach is essential for predicting potential risks as advanced packaging technologies move toward smaller pitch, higher integration density, and more complex geometries. Reliability concerns and design–process–materials co-optimization strategies for advanced integration schemes such as 2.5D/3D stacking, chiplet architectures, and hybrid bonding will be discussed. By linking failure physics, modeling, and experimental data, this talk provides a basis for identifying reliability challenges and guiding future research directions for robust advanced packaging technologies in AI-era semiconductor systems.

In the last 60 years of Moore’s Law Scaling, the digital IT industry has pushed from micron (µm), nano (nm), Angstrom (Å), and to now AI era. The influence of Moore’s Law has expanded from advanced logic products (CPU/GPU) to other digital products like DRAM, Flash, CIS (CMOS Image Sensors), Flat Panel Displays and Hard Disk Drives (HDD). The economic scale of semiconductor products has grown 30 folds to close to 1 trillion USD from 40 years ago. Moore’s Law is not a law of Physics. It resulted from the integration of numerous engineering innovations.

Material innovation has been one of the most important pillars to enable continuous Moore’s Law scaling. The number of material species involved in semiconductor manufacturing has grown from a dozen to more than 80% of the periodic table. The material complexity has increased from single elements, binary compounds, and ternary compounds to multi-component compounds. The geometry of materials supporting modern semiconductor processes has transformed from 3-D, 2-D, and 1-D structures to volume-less dimensions. Without these innovations, Moore’s Law could have been terminated some time ago.

In this talk, we will reflect on the influences and contributions of key material innovations that have enabled engineers and scientists to overcome critical technical challenges throughout the 60-year history of Moore’s Law.

Long Live the Material Innovations!

Photonics and energy sciences are yielding advances that can open paths for conceptually new energy conversion technologies.  I will discuss three such “grand challenge” pathways. A first is challenging conventional wisdom and laws about the properties of the apparently incoherent emission of thermal radiation.  A second challenge is that of generating chemical fuels from sunlight, suitable for aviation and heavy transport use, from advances in photoelectrochemical water splitting and carbon dioxide reduction.  This requires new approaches to solar fuels systems design with unprecedented efficiency, durability, and chemical selectivity. The third challenge is space solar power innovation, powered by ultralight, inexpensive photovoltaics, flexible structures, and phased arrays. Space solar power technology can potentially enable generation of dispatchable, baseload solar electricity that doesn’t exhibit the intermittency of conventional solar photovoltaic systems, representing an approach to baseload solar power generation anywhere on earth.

A prevailing hypothesis suggests that the exceptional computational efficiency of brains arises because neurons operate in the 'edge of chaos' (EOC), a critical regime between order and disorder where systems achieve maximal computational capability and adaptability. This concept has been given a concrete mathematical description in a proof by Leon Chua within the discipline of nonlinear dynamics. Parallel research by Brian Ridley focused on nonequilibrium thermodynamics of systems with negative differential resistance (NDR) and revealed that entropy production minimization (EPM) could drive them to a state of optimum thermodynamic efficiency. These two lines of research are now converging to reveal that the same aspect within nonlinear dynamics, a local non-monotonic relationship between current and voltage, optimizes both the ability to perform computation and the power efficiency of the computation. This has important implications for understanding how brains compute so efficiently and for the design of future neuromorphic systems for AI.

Over millions of years, natural selection has produced surface architectures that achieve remarkable functionality using minimal material. This talk follows how such examples translate into practical technologies. First, I discuss our extensive work on gecko-inspired adhesive surfaces. Geckos adhere to walls primarily through van der Waals forces, enabled by dense arrays of microscopic, hair-like fibrils. Designing synthetic analogues requires careful optimization of geometry, stiffness, and surface energy — guided by theory and simulation as much as experiment. After more than a decade of development, these bioinspired adhesives are now used in robotic pick-and-place systems, particularly for handling challenging components such as micro-LEDs and thin battery foils. We have also developed controlled release mechanisms for ultra-light micro-objects, integrated optical monitoring for process reliability, and implemented machine-learning approaches to accelerate industrial adoption. More recently, we adapted these structures for adhesion to human skin, resulting in a clinically validated prototype for ear surgery.

Second, we turn to biological suction systems found in aquatic animals. Despite their widespread occurrence, their governing failure mechanisms are not fully understood. Using elastomeric micro–suction cups equipped with pressure sensing and simultaneous optical imaging, we identified the instability mechanisms that ultimately limit attachment strength. This provides a quantitative framework for designing more reliable suction-based gripping systems.

Third, we examine sharp natural penetrators, such as stingers and spines. Across length scales—from micrometers to millimeters—these structures resolve the mechanical conflict between penetration and buckling. They do so by adopting a near-parabolic geometry with an exponent close to 2, a shape that appears to be universal. This geometry minimizes insertion force while maintaining structural stability.

Together, these examples demonstrate a central message: Nature offers highly optimized, resource-efficient design strategies. By understanding the underlying mechanics rather than merely copying form, we can translate biological principles into robust and sustainable engineering technologies.

The rapid growth of artificial intelligence (AI) has accelerated the demand for advanced packaging technologies that provide higher bandwidth, greater integration density, and reduced signal delay. Beyond chip-level scaling, recent advances in 2.5D/3D integration, chiplet architectures, heterogeneous stacking, and emerging substrate materials such as glass have become essential for next-generation AI semiconductor systems. Historically, back-end-of-line (BEOL) interconnect technologies have evolved to reduce RC delay through material and structural innovations, while continuously addressing key reliability issues such as electromigration, stress-induced failure, and dielectric degradation.

Advanced packaging can be viewed as an extension of these interconnect technologies into complex package-level geometries. As interconnect pitches continue to shrink toward the sub-micron scale and package integration becomes more heterogeneous, reliability assessment must go beyond conventional electrical testing. Thermomechanical stress, interfacial delamination, warpage-related degradation, oxidation, and moisture absorption are becoming increasingly important reliability factors in advanced packaging architectures.

This talk presents reliability challenges in advanced BEOL and packaging. Ensuring the reliability of chips and advanced packages requires a comprehensive evaluation framework that integrates electrical, environmental, and mechanical reliability perspectives, together with a fundamental understanding of materials properties. Such an approach is essential for predicting potential risks as advanced packaging technologies move toward smaller pitch, higher integration density, and more complex geometries. Reliability concerns and design–process–materials co-optimization strategies for advanced integration schemes such as 2.5D/3D stacking, chiplet architectures, and hybrid bonding will be discussed. By linking failure physics, modeling, and experimental data, this talk provides a basis for identifying reliability challenges and guiding future research directions for robust advanced packaging technologies in AI-era semiconductor systems.

Synchrotron radiation (SR) has been widely applied in various fields of materials science. Especially, the temporal and spatial observation of various reactions is the biggest advantage of using SR. In the presentation, in situ observations with various ranges of time and space are presented:

(a) Temporal: iron-ore sintering (~min.), Fe-C phase transition (~msec.), and metal fracture (~nano sec);

(b) Spatial: iron-ore sintering (~μm) and iron-oxide reduction (~nm).

Time-resolved XRD and XAFS successfully revealed the reaction schemes of iron ore sintering, resulting in a new concept: a CCT diagram for sintering to optimize the sintering process. The combination of XRD and DXAFS (dispersive X-ray Absorption Fine Structures) allowed us to observe changes in short- and long-range order simultaneously during reactions such as phase transition and fracture at a timescale of nano sec~msec. The metastable states with middle-range order appear during the reactions, which is crucial information for understanding and controlling the reactions.

X-ray micro-spectroscopy, which involves X-CT measurements at different energies, can visualize 3D distributions of chemical states as well as microstructures. This technique provides us with information about the heterogeneity of reactions, which is essential for controlling reactions in real materials.

Temporal and spatial observation of reactions produces multi-dimensional big data. Therefore, it is rather difficult to “see” the data and find the location and timing of the formation of “trigger sites” which initiate and control reactions. We performed topological data analysis (TDA) to find “trigger sites” automatically, which can break the limitation of the empirical approach based on previous experience and knowledge. Cracks are formed within the sinter during the reduction of iron ore sinter, which is of great interest for industrial applications. We could non-empirically identify “trigger sites” of cracking from X-ray micro-spectroscopy data using persistent homology, a TDA approach widely applied to various systems.

Finally, I would like to discuss future directions in temporal and spatial observation of various reactions using synchrotron radiation, and the application of mathematics and/or informatics to analyze multidimensional big data.

We have known for many decades that grain boundaries (GBs) not only affect, but often dominate, the properties of polycrystals. And yet, our understanding of structure-chemistry-property connections for GBs has largely focused on individual cases, specific alloys, and certain high-symmetry boundaries. Through that focused effort, GB science has come a long way in its ability to explain specific observations. The premise of this talk, however, is that we are now entering the age of GB design, in which the tools of computation and data science allow us, for the first time, to rigorously address the “inverse problem”: designing polycrystalline materials from a blank slate to achieve GB networks with desired structures and properties.

The key to this paradigm shift is the ability to treat all the diverse atomic environments in the GB network as a quantifiable spectrum, producing a so-called “spectral model” connecting GB structure, chemistry, and properties. This talk will review what is now a complete first draft of the GB spectral model for GB chemistry in metal alloys, including all the thermodynamic elements needed to make predictions of GB configuration and energetics across the full span of an alloy phase diagram. Applications to experiments, materials design, and the commercialization of new alloys will also be discussed.

From the perspective of battery energy research, selecting appropriate materials and arranging them effectively are essential for developing high-performance batteries. In this talk, we introduce our research on battery materials and the process of manufacturing lithium-ion batteries. We then present a battery system designed to evaluate degradation states, which is crucial for determining the end of a battery’s lifespan.

In particular, we explore the application of electrochemical impedance spectroscopy as a technique for assessing degradation in energy storage systems. Finally, we introduce our Smart Energy System Innovation Center and its role in advancing these technologies.