Thermal Cycling Simulation and Sensitivity Analysis of Wafer Level Chip Scale Package with Integration of Metal-Insulator-Metal Capacitors

Wafer-level chip scale package (WLCSP) under temperature cycles could be subjected to reliability issues due to thermal mismatch of different materials. Various factors could contribute to WLCSP thermo-mechanical performance, such as bond pad and solder material, die dimensions, as well as the locations of packaging components. In this paper, a novel sensitivity analysis technique is integrated with ANSYS® APDL to improve the thermal cycling performance of a WLCSP package with metal-insulator-metal (MIM) capacitors. In the first example, a finite-element model with details of passivation and metal layers were constructed in ANSYS® APDL. In the model, a total of 30 different variables were defined. Then the model was imported into an optimization software optiSLang to perform sensitivity analysis under thermal cycling loadings. Two runs of sensitivity analyses were carried out, each with 100 designs of experiments generated by space-filling Latin hypercube sampling technique. During sensitivity analysis, optiSLang was able to run 4~8 designs in parallel for faster results. In the first run, the individual contribution of all the parameters to the passivation stresses was examined through correlation coefficients and coefficients of prognosis. The results of the first 100 designs revealed that the Young's modulus of isolation material in MIM structure is a dominating parameter for the high thermal stress. Specifically, dielectric material Si3N4 had a much higher modulus than SiO2 and was adopted as the isolation material in the MIM capacitor, causing higher stress. Due to the dominance of isolation modulus, the contribution of the other parameters was not visible. Therefore, the isolation modulus was fixed in the second run of sensitivity analysis. Then the three most important parameters were identified for both low and high isolation modulus. In the second example, a sensitivity analysis was performed by considering only the three critical parameters using a 3D parametric model to capture the stress patterns and magnitudes better. Through sensitivity analysis using the ANSYS® parametric model in optiSLang, we concluded that the methodology could quickly identify the key parameters for the high thermal stress in the WLCSP and could also provide potential solutions to mitigate the cracking issues.