A Dynamic Time Evolution Method for Accelerating Self-Heating Simulations of Advanced Technologies

Accurate thermal evaluations based on the classical subcircuit method for advanced transistors (e.g., FinFETs and GAAFETs) have become more complex and time-consuming. This study proposes a dynamic time evolution method (DTEM) to evaluate the device temperature without introducing the circuit thermal nodes, which effectively accelerates the device-circuit level self-heating effect (SHE) simulation. The results show that the basic DTEM-0 method achieves an SHE simulation speedup ranging from 2.1 to 9.1 times when compared with the existing subcircuit approach while maintaining acceptable levels of accuracy. Moreover, a new concept of quasi-steady-state (QSS) temperature is proposed to describe the steady thermal profiles of devices under the operation conditions in digital circuits. In conjunction with the proposed effective power method, the QSS model can predict the steady temperature of circuits under complex power waveforms, and the experimental results show that the DTEM-QSS potentially offers ~1000 times speedup for the evaluation of circuit thermal conditions. This work provides a scalable acceleration framework for SHE evaluation of advanced technologies, such as the PPA-T evaluations of standard cells with FinFETs and GAAFETs.