In-Situ Ultrasonic Inspection of Thickness and Morphology of Thermal Interface Material in Multilayer Structure Using Modified Minimum Entropy Deconvolution

Power density of electronic devices inevitably increases as the feature size of integrated circuits (ICs) shrinks, leading to the overheating of chips. The heat dispersion performance and reliability of electronic package devices are significantly affected by the bond line thickness (BLT) and morphology of the thermal interface material (TIM). The conventional ultrasonic inspection with pulse-echo mode fails to accurately quantify the BLT and morphology of TIMs due to the overlapping of echo waves in the multilayer structure. In this work, a third-order nonlinear minimum entropy deconvolution (TON-MED) algorithm is proposed to decouple the overlapped high-frequency ultrasonic signals, and the best performance on sparsity of the theoretical and experimental signals is demonstrated by comparing the TON-MED algorithm with the commonly used deconvolution algorithms, Wiggins minimum entropy deconvolution (WMED), and optimal minimum entropy deconvolution adjusted (OMEDA). The TON-MED algorithm is consequently implemented to accurately determine the BLT of TIM and image the internal defects, in-situ acquired from the Si-TIM-Cu three-layer structure. The low measurement deviation (6%) and high-resolution image prove the reliability and capability of the proposed TON-MED algorithm for characterizing the BLT and morphology of the TIM simultaneously, which benefits to design and fabricate high-performance TIMs for heat dispersion of chip devices.