I. Introduction
With the ever and rapid increase of data rate, circuit complexity, and device density in electronic integration and packaging, the signal integrity (SI) and power integrity (PI) of high-speed heterogeneous integrated circuits (ICs) become an imperative and challenging issue especially at the design stage [1]. Numerous methods have been explored in the past decades to address the large-scale electromagnetic (EM) modeling problems, which can be mainly categorized into two groups, the differential equation methods [2], [3] and the integral equation methods [4]. The integral equation methods exhibit superior efficiency for modeling the problems involving conductors in free space as compared to the differential equation methods. Because unknowns from the background space are not required in the integral equation methods, the number of unknowns is thus significantly reduced. However, for heterogeneous problems involving finite-sized dielectrics, additional equivalent sources are needed to describe the dielectrics in the conventional integral equation methods, resulting in exponential increase of the computation effort for electrically large problems. The denser impedance matrix will incur a dramatic hike in the computational overhead. An integral equation method that is capable of modeling EM problems involving piecewise homogenous dielectrics, meanwhile, has less unknowns is highly desirable.