Fast Mutual-Heating Prediction Method for Integrated Electronics and Photonics

In this work, a rapid method is developed for predicting device-level temperature shifts in electronic and photonic integrated circuits. The method models the most dominant thermal interactions through resistive star networks. The resistance values are obtained through interpolation of data acquired by a finite element method solver or experimental characterizations and can be used to predict all device temperatures. Simulations using COMSOL and measurements of a silicon IC have verified the key assumptions of the method. After calibrating the method with experimentally obtained thermal reference data, thermal estimations have been shown to be as close as 8.1% to measurements. Compared to a finite element based approach, the method can allow for a 1 to 2 orders of magnitude speed-up, which the user may choose to trade off for a higher estimation accuracy.