A Bi-Layer compact thermal model for uniform chip temperature control with non-uniform heat sources by genetic-algorithm optimized microchannel cooling

摘要

Due to the difference of local power in semiconductor chips, the chip temperature is usually non-uniform with hot spots, which may cause problems like thermal stress and interfacial delamination, thus deteriorating the reliability and decreasing the electrical performance. Therefore, uniform chip temperature is demanded and pursued both by both industry and academics. The most commonly used single-phase microchannel liquid cooling is efficient but suffered from severe temperature non-uniformity, which may be tackled by tuning local channel densities. However, the existing one-dimensional model is not enough for the channel density design of microchannel with non-uniform heat flux, for it's lack of the description of heat spreading effect. In this study, we presented a bi-layer compact thermal model with considering the three-dimensional heat spreading and thermal conduction for the junction temperature gradient prediction in microchannels with non-uniform heat source. This analytical model accurately predicts the junction temperature gradient with error within 4.8% compared with COMSOL simulation results. Combining the present model with genetic algorithm, we designed a microchannel with optimized local channel widths to achieve the temperature uniformity as high as ～90%. The maximum junction surface temperature difference is reduced from 45?°C to 13?°C with the existence of hot-spot with heat flux density of 200W/cm2. The present model and design strategy can be used to diminish the hot spot and achieve uniform temperature in high-power chips for extensive applications like integrated circuits, IGBT, HEMT, etc.