Conference Agenda

This study presents an induction heating process for transient liquid phase (TLP) bonding of Cu-Sn layers at high frequencies, providing a rapid and localized heating solution for MEMS packaging applications. In order to model the coupled electromagnetic and thermal fields, with particular emphasis on capturing the frequency-dependent skin effect, FE simulations are conducted alongside experimental validation. Two modeling strategies are proposed: Strategy I facilitates direct visualization of the skin depth, while Strategy II offers a computationally efficient alternative with comparable accuracy. Both models demonstrate a deviation below 5% relative to experimental results, highlighting their reliability. With a low current of 12 A, the results indicate a minimum operating frequency of 27.12 MHz within the ISM band as essential to achieve the Cu-Sn bonding temperature of 250 °C and a maximum heating rate of 35 K/s at 55 MHz.

The antenna is an essential device in modern communication systems, antenna designs are imperative to provide a high-performance option. As the antenna is trending towards multifunction and systematization, antenna designs are often involved with a large scale of optimization variables and high dimensional objectives. A decomposition algorithm based on dynamic random grouping and aggregation trees is proposed to solve the antenna designs with a large scale of variables and objectives. To search the variable space efficiently, variables are divided into several groups, which are checked and updated during the evolutionary process. To enhance the evolutionary pressure in high dimensional objective comparison, the aggregation trees method is used to remove the non-conflict or weak-conflict objectives. A 30-element linear antenna array and a filter antenna are used to test the effectiveness and efficiency of the proposed method. The numerical results demonstrate that the proposed decomposition method can obtain qualified and satisfied options for complex antenna designs with fewer computational costs.

This paper demonstrates the far-field response of magnetically-biased graphene plasmonic resonators at THz frequencies by employing quasi-normal mode schemes. First, the eigenstates are normalized using a non-trivial procedure, accounting for the anisotropic nature of graphene’s surface conductivity. Next, the scattered field is computed by means of the modal coefficients of the quasi-normal modes for a general incident plane-wave. Finally, the far-field reconstruction, derived analytically from the scattered fields, is successfully validated against results from a full-wave solver, exhibiting exceptional accuracy.

To design efficient power converter, the electromagnetic behavior of the printed circuit board (PCB) at high switching frequencies should be well-known. Then, numerical approaches can be investigated to analyse the voltage and current waveforms as well as the current density distribution associated with a PCB layout. In this communication, considering a half-bridge converter connected to a passive electrical load, we propose a co-simulation based on the Cauer Ladder Network (CLN) approach providing an equivalent circuit of the PCB combined with models of GaN transistor and passive components.

This study proposes a scalable framework for uncertainty quantification (UQ) in large-scale high-frequency electromagnetic wave problems by integrating Non-Intrusive Polynomial Chaos methods with a parallel finite element solver. The approach addresses the critical need to model uncertainties in material properties, geometric configurations, and boundary conditions, which significantly impact performance in advanced electronic and communication systems. Key innovations include the use of Karhunen-Loeve Expansion to reduce the dimensionality of random fields and regression-based NIPC for efficient probabilistic analysis. Computational optimizations such as coarse-mesh eigenvalue analysis and hierarchical domain decomposition further enhance scalability and efficiency, enabling the framework to handle large-scale simulations effectively. Numerical experiments on a monopole antenna model with a dielectric validated the method's ability to accurately quantify uncertainties while significantly reducing computational costs compared to traditional Monte Carlo approaches. The results demonstrated precise modeling of spatially correlated uncertainties and their influence on electromagnetic field distributions. This work bridges the gap between deterministic solvers and advanced UQ methods, providing a robust and computationally efficient tool for real-world applications in electromagnetic compatibility analysis and the design of next-generation electronic devices.

Planar magnetics such as wireless power transfer (WPT) devices are being developed and applied to various kinds of applications. Thanks to the advantages such as high-quality factor and transmission efficiency, various printed-circuit-board (PCB) resonators are designed for different practical needs. Usually, these planar PCB devices are composed of thin dielectric layers and even thinner copper traces printed on the planar dielectric board. Furthermore, since the WPT devices can work at frequencies as high as several MHz, there is strong skin, proximity and distributed capacitive coupling effects between conductors. For planar magnetic having thin solid foil conductors, the stray capacitance between conductor segments has to be taken also to address the resonant phenomena. In the design of such kind of WPT structures, for example to design the spiral pattern of the winding copper traces, there is a strong need to resort to numerical methods without introducing simplification or equivalent methods. In this work, the layered meth technique is proposed to remedy the tough meshing process for thin conductors. A numerical example will be given to showcase the efficiency of the method

CONTROLLING phased antenna arrays is crucial and challenging in satellite communications. Control can be continuous or quantized [1],[2],[3],[4], where quantized control limits the phase and amplitude of each antenna element to discrete levels, introducing noise which can degrade array performance. Therefore, a robust method is essential to ensure high performance despite these quantization effects.

Dielectric waveguides have been attracting attention as a transmission line that allows for highly flexible circuit configuration in the millimeter-wave band. The objective of this study is to develop a dielectric antenna array integrated with a dielectric waveguide. By configuring the radiating element of the antenna from a dielectric material, it is expected that the loss between the dielectric waveguide and the antenna can be suppressed. In this study, the analysis model of the antenna array integrated with the dielectric waveguide consists of a T-shaped branch and two radiation elements, and beam steering is performed by the structural design of the branch. Topology optimization using the normalized Gaussian function network (NGnet), which has a high degree of design freedom, is introduced into the structural design of the branch. The structural design is performed by setting the beam tilt angle in 76 GHz band. Although the radiation loss is still large, the optimization results show that beam steering is possible even with low relative permittivity of εr = 2.6.

This paper discusses an open-source simulation scheme for calculating the induced currents on antenna surfaces and the distribution of surrounding electric fields under the influence of various electromagnetic pulses. By designing a series of preliminary numerical experiments, the results from the open-source computations are compared with those from commercial software to verify the feasibility of the scheme. Subsequently, the open-source tools are employed for project experiments to provide references for the electromagnetic protection design of antennas. The study demonstrates that open-source electromagnetic simulation offers new possibilities for research in electromagnetic computation.

Modern communication systems for highest frequency range and bandwidths are compact solutions relaying on hybrid photonic-electronic integrated circuits. One of important problems of this technology are substrate resonances at mm-wave and sub-terahertz frequencies producing highly nonlinear frequency characteristics of RF components fabricated on top of the substrate, such as antennas and electro-optical modulators. This paper presents a method for simplified modeling of anti-reflective coatings in large and detailed 3-D electromagnetic simulations by representing complicated and periodical geometrical coating features via frequency dependent permittivity of the homogenous coating layer. The accuracy of the proposed method is verified by comparison against detailed high fidelity models.

The development of precise wireless propagation models is critical for the architectural planning and operational deployment of fifth-generation (5G) network infrastructures and base stations. To satisfy the growing demands and the staged deployment of 5G technologies, highly accurate and effective path loss (PL) predictive models are essential for optimizing network layouts and enhancing

deployment strategies. Although traditional propagation models have shown utility, their applicability is frequently constrained by their lack of adaptability to diverse propagation environments and the significant computational investments required by deterministic methodologies. Contemporary advances in data-centric artificial intelligence (AI) have substantially propelled the progress of adaptable and intelligent propagation models, markedly improving the accuracy of PL predictions. This detailed investigation uses an extensive dataset from NYUSIM software, which includes a comprehensive spectrum of atmospheric variables during various seasons in urban microcell configurations in South Asia, enriched with broad empirical data across multiple environmental contexts. The primary objective of the study is to formulate, train, and evaluate four different machine learning (ML) models tailored to predict PL at crucial high frequencies vital to the effective performance of 5G networks. The analysis rigorously scrutinizes each model’s ability to accurately predict PL values, focusing on their precision and efficiency in processing large datasets under varying atmospheric conditions.

Batteries are the essential elements in energy storage devices. Thermal control of batteries is imperative to the safe operation of energy storage devices. To decrease the maximum temperature and reduce the temperature difference inside the lithium-ion batteries, a cooling system combined with air cooling and plate-fin cooling is proposed and optimized. A novel structure of plate-fin is designed to promote cooling efficiency. To speed up the optimal design of the combined cooling system, global and local searching algorithms based on the random forest method and kriging surrogate model are introduced. A battery pack containing eight battery cells is used to test the proposed cooling strategy. The numerical results demonstrate that the optimized cooling method can decrease the maximum temperate and temperate difference of the battery cells and the global and local searching algorithm is an efficient tool for the design of battery cooling systems.