Transistors are aggressively scaled down to meet the ever-increasing demand for improved performance. Such a scaling leads to increased power density and electrothermal issues leading to performance degradation in transistors. For reliable circuit design and thermally aware device design, there is a need to physically understand and estimate the rise in device temperature beyond the ambient condition. This work attempts to model the steady-state self-heating in three modern-day transistor structures: (a) silicon-on-insulator (SOI) FinFET, (b) multi-finger silicon germanium heterojunction bipolar transistor (SiGe HBT), and (c) gallium nitride high electron mobility transistor (GaN-HEMT). This work also proposes methods to extract thermal resistance (Rth), its components, and temperature coefficient in SiGe HBTs. The methods are validated on SPICE-generated synthetic data, detailed thermal simulation from TCAD and measured data.

In SOI FinFETs, the device is divided into two regions based on two separate heat flow paths. The model equations that can estimate the peak temperature in each region was developed. Later both the models are combined to obtain the peak temperature for the device. The temperature dependence of thermal conductivity of the semiconductor material is considered in the model. Modeling results are validated with thermal and electrothermal numerical simulations and with experimental data.

In multifinger HBTs, where multiple heat sources act together, each finger's temperature rises above its self-heating temperature due to coupling from the other fingers. We propose a model that intuitively incorporates the effect of thermal coupling among the neighboring fingers in the framework of self-heating, bringing down the overall model complexity. The model offers better speed of simulation as it requires a lesser number of nodes for circuit implementation. The model is validated with 3-D thermal simulations and measured data from STMicroelectronics B4T technology. The Verilog-A implemented model simulates 40% faster than the conventional model.

The presence of different materials in the FEOL heat flow path makes the thermal modeling problem in GaN HEMTs difficult. Two methods are proposed to estimate the temperature in such multi-layered structures. The first method uses spreading angle approximation to develop a compact analytical model to predict the temperature in the structure. The second method uses the infinite series solution of the heat equation to calculate temperature. Modeling results are verified against detailed three-dimensional TCAD thermal simulations of GaN-on-Substrate structures.