Abstract
Nowadays, electronic packaging has developed to achieve high power, high I/O, good reliability performance and small form factor characteristics. Among the various packages have been adopted by industry, Wafer Level Chip Size Packaging (WLCSP) fulfills above demands. Unlike flip chip packaging, WLCSP does not need underfill protection since it possesses a soft stress buffer layer between silicon chip and solder bump, the buffer layer can reduce the stress/strain in solder joint and prolong its lifetime. However, no underfill protection induced the high stress/strain which is caused by coefficient of thermal expansion mismatch (CTE) in thermal loading. This effect allows WLCSP failure more easily. In general, using experiments to assess reliability of WLCSP will take much time and money, so the finite element method and design-on-simulation technology have been widely used for predicting the thermal fatigue life of packaging under thermal cycling loading condition. If the simulation results are exact to match the experiments, this simulation can substitute the experiments to progress some analysis such as parameter designs, fatigue life prediction, etc. Using finite element method would spend less time than conduct some experiments, so this research gives finite element method of WLCSP to analyze creep effect and discuss various thermal cycling loading conditions. In order to reduce the development time and ensure the reliability quality of electronic packaging, the Accelerated Thermal Cycling Test (ATC) is a standard method which is currently used to characterize the reliability performance of electronic packaging. The most commonly used temperature range for commercial electronic products is from -40°C to 125°C, and within a predefined temperature range the creep effect is more significant in the solder material because the homologous temperature is excessed to 0.33 Tm (in K) in the thermal cycling loading. According to the simulation result, it can be found the creep effect is the main factor that caused the solder to failure. The increase in the ramp rate of the thermal cycling loading is often used to reduce test duration, but increase in the ramp rate causes the material to change its stress/strain properties. In addition, creep effect becomes smaller in the ramp section of thermal cycling because it allows lesser time in the ramp section of thermal cycling. Due to increasing ramp rate, it causes a less creep effect, so the stress relaxation effect also becomes less. Because of the less stress relaxation effect, it makes the materials have more stress at the end of ramp section that allows the materials to produce plastic strain more easily. To sum up, changing the ramp rate in the thermal cycling loading influences the reliability of electronic packaging. On the other hand, the dwell time of thermal cycling also affects the reliability of materials because increases in the dwell time that causes more creep strain and has less fatigue cycles in the solder material. According to the above discussion, it is very important to use an exact simulation method to assess the reliability of electronic packaging. In this study, the Anand Model and the Garofalo Hyperbolic Sine Model with Chaboche Kinematic Hardening Model are used to simulate creep and plasticity behavior in the simulation process. The results of simulation indicates that both of them have the same tendency during thermal cycling loading. There are two standard methods to calculate the predict fatigue cycles such as Coffin-Mason strain based model and Darveaux energy based model. The results of simulation demonstrate that the incremental inelastic strain is changed insignificantly, so the fatigue cycle can’t coincide with the experimental results which are calculated by using Coffin-Mason strain based model. In addition, the incremental energy density is increased when the ramp rate is increased. Substituting this results to Darveaux energy based model would get the life prediction cycles which is agreed with the experiment results. Moreover, in the case of fixed ramp rate and varied the dwell time, the life prediction cycles which are calculated by both models and have the same tendency of the life predicted cycles. In Norris-Landzberg Acceleration Factor Model, it doesn’t consider creep effect in this formula, so it will predict a violation result when using the experimental data which are progressed on fixed dwell time and various ramp rate in thermal cycling test. Finally, the modified AF formula which considers the creep effect will be proposed. As a result, the life prediction cycle which is calculated by the modified AF formula is corresponded with all of experimental results.