Application of Thermo-mechanical Measurements of Plastic Packages for Reliability Evaluation of Pems

Thermo-mechanical analysis (TMA) is typically employed for measurements of the glass transition temperature (Tg) and coefficients of thermal expansion (CTE) in molding compounds used in plastic encapsulated microcircuits (PEMs). Application of TMA measurements directly to PEMs allows anomalies to be revealed in deformation of packages with temperature, and thus indicates possible reliability concerns related to thermo-mechanical integrity and stability of the devices. In this work, temperature dependencies of package deformation were measured in several types of PEMs that failed environmental stress testing including temperature cycling, highly accelerated stress testing (HAST) in humid environments, and burn-in (BI) testing. Comparison of thermo-mechanical characteristics of packages and molding compounds in the failed parts allowed for explanation of the observed failures. The results indicate that TMA of plastic packages might be used for quality evaluation of PEMs intended for high-reliability applications. Introduction Excessive mechanical stresses in dice and fracture of wire bonds due to deformation of plastic packages, which are caused by mismatch of CTE between the molding compound (MC) and the lead frame (LF), are considered as one of the major failure mechanisms specific to plastic encapsulated microcircuits. To reduce mechanical stresses in the dice, a buffering layer of silicone die coating (usually drop-on coating) is used in many linear devices sensitive to mechanical stresses. Extensive investigations of the effectiveness of the silicone die coating have shown that in some cases they might adversely affect reliability of the parts [1, 2]. In particular, it has been shown that relatively thick silicone coatings might increase the probability of failures during temperature cycling [3]. This requires a thorough analysis and a method to evaluate risks associated with usage of die coatings in PEMs. Another possible mechanism of failure is due to excessive deformations of plastic packages as a result of swelling of the MC in humid environments [4]. Moisture sorption in plastic packages degrades mechanical characteristics of MCs, resulting in decrease of the glass transition temperature and reduction of adhesion between the MC and LF. This, together with expansion of the MC due to moisture-induced swelling, might result in delaminations between the LF and MC during HAST, thus providing a path for moisture and corrosive contaminations to penetrate to the die surface. Excessive delaminations were often observed on plastic devices failing HAST; however, this effect has not been properly analyzed and explained. Application of TMA for measurements of thermo-mechanical characteristics of MC directly on PEMs has been analyzed in [5]. It has been shown that anomalies in deformations of plastic packages might be due to the presence of the LF, sorption and desorption of moisture, and the presence of internal mechanical stresses, which result in warpage of packages. Elimination of these errors allows for accurate measurements of Tg and CTE to characterize molding compounds used in PEMs. In this work, anomalies in the TMA characteristics of plastic packages are used to reveal excessive deformations, which resulted in formation of rejectable defects and/or failures in the parts. Three case histories are described in which employment of TMA has helped in understanding the mechanism of failures during reliability testing of PEMs. Experiment In this study, the temperature dependence of the deformation of plastic packages and epoxy molding compounds was measured using a thermal mechanical analyzer, TMA2940, manufactured by TA Instruments. To avoid warpage-related errors, the characteristics of molding compounds were measured on small pieces of the devices cut from the packages. Experiments have shown that the best reproducibility is achieved when testing is performed at a rate of 3 C/min. during cooling from 220 C, followed by heating of the sample in the analyzer at the same rate. This allows for monitoring of the stress-relief effect in samples and assures elimination of possible errors related to the presence of moisture and built-in mechanical stresses. Measurements of package deformations were carried out directly on plastic parts with a probe centered above the die on the surface of the package. Case History I: HAST-Induced Delaminations Multiple failures due to delaminations were observed after HAST at 130 C/85% RH for 96 hours in comparator microcircuits encapsulated in plastic SOT-23-5 packages. The parts had normal electrical characteristics; however, delaminations, which were observed mostly at the lead finger-tips (see Figure 1), were considered critical due to the wire bond integrity concerns. a) b) Fig. 1. Example of delaminations at the finger-tips on the top (a) and the bottom (b) of the SOT-23-5 packages after HAST. To evaluate the risk related to the observed defects, three groups of the devices were tested: 15 parts after HAST, 15 parts from the same lot date code (DC0018) that failed C-mode scanning acoustic microscopy (C-SAM) examinations during screening, and 5 parts from another lot (DC 0020) that were used for comparison purposes. It was assumed that due to close date codes, these lots had similar characteristics. Three samples of each group were characterized for Tg and package deformations using a thermo-mechanical analyzer. Results of these measurements are shown in Figure 2 and Table 1. a) b) Fig. 2. Temperature dependence of package deformations for three samples: a) after HAST, b) after screening. Note that all samples after HAST had a significant hysteresis, with the increasing branch of the TMA curves being much higher than the decreasing one. The value of this hysteresis was evaluated by a relative deformation (dL/L) calculated at 50 C. Results of these calculations are also shown in Table 1. Lot Tg, C CTE1, ppm/C CTE2, ppm/C dL/L, % After HAST, DC 0018 170.1 (1.1) 9.2 (1.7) 37.3 (3.7) 0.21 (0.04) After Screen, DC 0018 171.1 (0.3) 9.1 (1.8) 42.6 (7.2) 0.12 (0.05) Initial, DC 0020 164.3 (0.7) 10.25 (0.8) 65.5 (4.0) Tab. 1. Glass transition temperatures (Tg), coefficients of thermal expansions (CTE), and hysteresis (dL/L) of the three groups of the comparators. Average values and standard deviations are shown in parentheses. The results show that samples with DC 0018 after HAST and after screening had similar Tg and CTE values; however, the hysteresis was much larger for the HAST parts, indicating swelling of the molding compound. This swelling is most likely the major reason of excessive delaminations observed after HAST. The TMA testing was performed in ~1,000 hours after HAST. With the thickness of parts, h = 1.15 mm, the characteristic time of moisture diffusion, which also indicates the time necessary to dry out the package, can be estimated as τ = h/4D, where D is the diffusion coefficient. Calculations yield τ ≈ 520 hours at room temperature. This means that most of the moisture absorbed during HAST should have been released from the package by the time of TMA measurements. The observed excessive hysteresis suggests that the exposure to hightemperature/high-humidity conditions during HAST has resulted in irreversible deformations of the molding compound and formation of delaminations. It is interesting to note that microcircuits with DC 0020 and DC 0018 had different molding compounds, thus suggesting that commercial parts with even close date codes might be manufactured using different materials, and that the TMA measurements are capable of discriminating among different encapsulating materials. Delaminations after HAST testing were observed on parts that passed screening. This indicates that similar defects potentially can develop with time due to moisture-induced swelling and creep in molding compound, even in samples that initially had no delaminations. To evaluate the risk related to the presence of delaminations in these devices, all samples in the three groups were subjected to preconditioning according to JEDEC standard JESD22-A113 (three runs through the solder reflow chamber) and temperature cycling from -55 C to 125 C. The parts were tested after 100, 300, and 1,000 cycles. No failures were observed during this testing, and the parts manifested only minor changes in their electrical characteristics, as shown in Figure 3.