Sei sulla pagina 1di 8

100

IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 1, FEBRUARY 2000

Characterization of Interfacial Adhesion Damage Induced by Accelerated Life Testing


Jongwoo Park, D. Gary Harlow, and Herman F. Nied
AbstractThe purpose of this effort was to characterize statistically and experimentally the interfacial adhesion damage of generic dual in-line packages (DIPs) exposed to accelerated stress conditions. DIPs, consisting of three parallel gold conductors in a meander pattern on a 24 pin Al2O3 ceramic substrate spin-coated with Pyralin PI-2555 polyimide with and without an adhesion promoter (APS), were chosen as the test vehicles. Leakage current measurements and optical microscopy were employed to monitor the degradation process at the interface between the polyimide and the substrate. In order to characterize the damage adequately, the entire surface of each DIP was analyzed. A two-parameter Frechet cumulative distribution function (cdf) was found to be excellent for characterizing the statistical scatter in the geometrical features, such as, the cross-sectional area of the damage located over the interface. It was observed that the adhesion damage tends to be larger and clustered for more severe testing conditions. It was demonstrated that APS statistically improves long-term reliability. Furthermore, interfacial adhesion damage was not initiated solely by contamination at the interface. Index TermsInterfacial damage, adhesion, leakage current, acceleration test, optical microscopy, image analysis, Frechet cumulative distribution function, and statistical analysis.

apply electrical, mechanical, or environmental conditions more severe than typical, to induce failure mechanisms more quickly than would occur under normal operation. A variety of accelerated tests have been reported in the literature [7], [8]. One of the standard test conditions for reliability testing of plastic packages is 121 C with either saturated or unsaturated steam, using a conventional pressure cooker test (PCT) or a highly accelerated stress test (HAST). Commercial test standards for plastic-encapsulated packages that describe procedures and define test requirements for qualification, screening, and lot acceptance of high reliability plastic packages are available [9]. The goal of this study was to characterize statistically and experimentally the interfacial adhesion damage over the entire surface area of dual in-line packages (DIPs) exposed to HAST and PCT. The test conditions were 121 C and 100% relative humidity (RH) for both HAST and PCT. Observations on DIPs with and without DuPont APS and PI-2555 polyimide were obtained by optical microscopy and analyzed statistically.

I. INTRODUCTION PIN-COATED polyimide (PI) is widely used in microelectronic applications as an interlayer dielectric material. PI provides high resistivity, thermal stability, chemical inertness, good adhesion, and planarization, and it is suitable for very large scale integration processes [1], [2]. However, it is well known that in the presence of moisture, ionic impurities, and electrical fields, the delicate integrated circuitry is vulnerable to failure [3], [4]. Since polymers are permeable to moisture, it is impossible to exclude moisture completely from packaged devices. Interfacial adhesion damage caused by moisture is a precursor to an electrical or a mechanical failure, and it strongly affects the reliability of the polymer-encapsulated microelectronics. Measurement of leakage currents is one of the widely accepted methods for evaluating encapsulant performance [5], [6]. Exposure to a deleterious environment generally causes an increase in leakage current, which may be attributed to water or ionic contamination at the surface. The effect of humidity in microelectronic devices can increase ionic mobility and can induce surface leakage current. Accelerated life tests are designed to
Manuscript received February 9, 1999; revised November 2. This work was supported by the Semiconductor Research Corporation under Contract 96-PJ-428. J. Park is with the Optoelectronics Center, Lucent Technologies, Breinigsville, PA 18031 USA. D. G. Harlow and H. F. Nied are with the Department of Mechanical Engineering and Mechanics, Lehigh University, Bethlehem, PA 18015-3085 USA. Publisher Item Identifier S 1521-3323(00)01388-5.

II. EXPERIMENTAL PROCEDURE A. Test Vehicle and Preparation A generic 24 pin DIP consisting of three parallel gold conductors in a meander triple track pattern on a 99.5% alumina substrate was used as a test vehicle. A schematic diagram of a DIP is shown in Fig. 1. The conductors are 53 m wide, 2.1 m high, and spaced 95 m apart. Each conductor is approximately 21.6 cm long. The total number of conductors in a DIP is 48. Leakage current is measured between the outer tracks and the center track, which are oppositely biased. Such DIPs have been used to monitor surface conduction and the resulting corrosion that accompanies it. Sbar et al., [10], [11] used this type of DIP on an alumina substrate to study the corrosion of Ti-Pd-Au metallization with and without a protective coating. Iannuzzi [12] used similar test chips to investigate the effect of chloride on the corrosion of aluminum. Mancke [13] also used a triple track pattern as a test vehicle to evaluate the quality of polymer coatings by monitoring changes in leakage currents. Thus, this type of DIP has been used successfully to investigate the fundamental failure process of encapsulated microelectronic devices. The PI used in these experiments, obtained from DuPont, is Pyralin PI-2555. PI-2555 was selected because of its processability, thermal stability, and chemical resistance [14], [15]. The physical properties of PI-2555 as received from the manufacturer are summarized in [16], and details on its characteristic behavior have been described elsewhere [17], [18]. Surface cleaning is the most important step in the fabrication of

1521-3323/00$10.00 2000 IEEE

PARK et al.: CHARACTERIZATION OF INTERFACIAL ADHESION DAMAGE

101

Fig. 1. Schematic diagram of the cross-sectional view of a DIP.

any semiconductor device, and its efficiency is critical to the reliability and performance of the device. Six different cleaning methods with various combinations, including hydrogen peroxide cleaning [19] and dry cleaning methods, were employed to reduce the initial leakage current. Initial leakage currents were lowest when the DIPs were cleaned by ultraviolet ozone (UV-O3), which is suitable for oxidative removal of adsorbed organics and also does not cause further contamination of the tester surface [20]. Thus, the adopted cleaning procedure consisted of the following: deionized (DI) water rinse, spin drying or drying at 120 C for 30 min, and UV-O3 treatment. In order to minimize the contamination on the freshly cleaned DIPs, spin coating was performed immediately after cleaning. Mixing an organosilane coupling agent (VM651), DI water and methanol produces amino propyltriethoxy silane (-APS), which was used as an adhesion promoter on the substrate. A 0.1% aqueous solution of VM651 was spun-dried at 5000 rpm for 30 s on the DIPs shortly after cleaning. Then, the PI-2555 was coated on the APS-treated DIPs within 30 min. Since PI-2555 is relatively viscous and difficult to spin into thin layers, it was diluted. The solution consists of eight parts PI-2555 and one part T-9039 thinner with respect to the total volumetric ratio. The diluted PI-2555 was spun for 60 s at either 5000 rpm or 3500 rpm. The PI-2555 was cured in nitrogen (N2) atmosphere at 100 C for 30 minutes, then at 150 C for another 30 min, and finally at 200 C for 1 h, to facilitate the escape of the solvent from the film. Further curing was conducted at 300 C for one hour and then at 400 C for 10 min, where the PI is most thermally stable and chemically resistant. The ramp at each step is less than 5 C/min. After curing the PI coating, the thicknesses were determined to be 1.5 0.5 m and 2.35 0.5 m for the 5000 rpm and 3500 rpm rates, respectively. The film thickness was measured using a Tencor P-2 long scan profiler. The resulting film after curing is a translucent brown color with a glossy surface. B. Pressure Cooker Test (PCT) To accelerate aging of some of the coatings, PCT stress conditions, which were 121 C with 100% RH at 2 bars atmospheric pressure, were used. In PCT, the temperature and humidity are maintained merely by boiling water, and thus, saturated steam is the only humidity condition possible. To minimize the surface water condensation on the DIPs the specimens were placed in covered Petri dishes. The DIPs were stressed for specified length of time, e.g., 1, 3, 5, 7, 9 or 24 days. At regular time intervals, they were removed for leakage current measurements

and microscopy. Under these conditions the temperature distribution in the test chamber ranged from 121 C to 125 C, while the vapor pressure ranged between 15 psi and 19 psi. The pressure gauge error was approximately 3%. C. Highly Accelerated Stress Test (HAST) Also, HAST was used to accelerate the degradation process of several encapsulants. The DIPs were subjected to 121 C, 100% RH, and unbiased in a HAST chamber manufactured by the Tobai Espec Company. Thus, the specimens were free of water condensation. The dew point temperature and vapor pressure under the nominal HAST conditions were 121 C and 15 psi, respectively. For these tests, the temperature distribution in the test chamber was within 0.5 C in the center of the test chamber, and the RH error accuracy was estimated to be 3%. The accuracy was similar to that for conventional stress testing at 85 C and 85% RH. D. Image Analysis DIPs coated with PI-2555 were examined using optical microscopy (LECO 2001 Image Analyzer with a programmable stage pattern) at a magnification of 50. To scan the entire coating-substrate interface, 66 fields, 11 fields along the axis and six fields along the y axis, were required. It should be noted that the depth of field for the optical microscopy was adjusted so that primarily interfacial adhesion damage was measured and the effects of surface imperfections were minimal. Details of the applications and methodology of analysis may be found in [21]. The entire area of the DIPs measured by image analysis was 12 770 m 7398 m. The area, maximum chord, minimum chord, and coordinates of the centroid for each damage feature were recorded. The measurements were used to statistically characterize the geometrical properties of individual damage sites along the interface. It is anticipated that this classification will lead to an improved understanding of the critical failure mechanisms for interfacial adhesion damage. E. Leakage Current Measurements Leakage currents between the positively biased two outer conductors and the grounded center conductor were measured with a Keithley Instrument Model 617 digital electrometer before and after each test at different time intervals. Further details of the methodology used in measuring leakage currents, e.g., minimizing noise for reproducible measurements, may be found in [4]. The leakage current fluctuation was monitored as a function of the accelerated HAST or PCT exposure conditions.

102

IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 1, FEBRUARY 2000

Fig. 2. Example of a large damage site induced by PCT for 38 days; PIA47-3 with APS.

Fig. 4. Example of the interfacial adhesion damage induced by PCT for 38 days; PIA47-5 without APS.

Fig. 3. Largest damage site induced by PCT for 38 days; PIA47-7 with APS.

Fig. 5. Largest contiguous interfacial adhesion damage site induced by PCT for 38 days; PIA47-6 without APS.

Initial leakage current measurements were typically in the pico ampere (pA) range, but they increased considerably after the long term stress testing. F. Chemical Analyzes A Scienta Instruments ESCA-300 system X-ray photoelectron spectrometry (XPS) was used to analyze the chemical state of the elements in selected damage sites. Differential Scanning Calorimetry (DSC) was used for measuring the glass transition temperature Tg. III. EXAMPLES OF INTERFACIAL ADHESION DAMAGE Typical examples of interfacial adhesion damage on DIPs with and without APS subjected to PCT for 38 days are shown with various magnifications in Figs. 25. The black features between the gold conductors are indicative of substantial delamination between the coating and the substrate that evolved from undamaged DIPs. The features of lighter color are damage that were not developed fully. The DIPs with APS exhibited much smaller and more regularly shaped interfacial adhesion damage than that observed on the DIPs without APS. The damage was scattered uniformly over the entire surface of each DIP. Fig. 2, which depicts the gold (Au) conductor and the ceramic substrate, shows a common damage feature. The damage appears to have propagated from the conductor. This behavior is expected because the corner formed by the conductor on the substrate is more difficult to coat properly and is also a region of high stress

concentration. The single damage site of largest area, i.e., 3,900 m2, is shown in Fig. 3. Since the damage bridges the conductors, the leakage current of about 0.1 A for this DIP was higher than any other DIP with APS. Without APS, however, individual interfacial adhesion damage sites span considerably larger areas of the substrate, as seen in Fig. 4. Notice that the damage has bridged two large features by running along the conductor. Of the three DIPs without APS subjected to PCT, the largest single cross-sectional area of the damage sites is 6,000 m2 and the smallest is 150 m2. Fig. 5 shows the largest contiguous region of damage with a total area of 16 500 m that was observed. It spans several tracks and the intermittent substrate. There probably was water in this delaminated area, because a spectrum of colors was observed from this site during optical microscopy. As a result of the extensive damage, the average leakage current for this DIP was 1.86 A, which is the largest leakage current among all of the DIPs tested. Certainly, damage to this extent severely degrades the electrical integrity of the device. For the DIPs with APS tested by HAST, examples of the interfacial adhesion damage are shown in Figs. 68. Even the largest damage areas do not bridge the conductors, and consequently leakage currents were quite small, 0.1 to 10 nA. It is obvious that an electrical short is not imminent. In fact, the worst damage site from all of the DIPs with APS, tested by the HAST for up to 20 days, is shown in Fig. 6, and the leakage current was only 1 nA. For the DIPs with APS tested under HAST, the maximum and minimum cross-sectional areas were

PARK et al.: CHARACTERIZATION OF INTERFACIAL ADHESION DAMAGE

103

Fig. 6.

Worst damage site induced by HAST for 20 days; PIA24-3 with APS.

While certain insights into the interfacial adhesion damage evolution are possible from these observations, further detailed experimental efforts are warranted. The interfacial adhesion damage resulting from PCT is worse than that caused by HAST. The major difference between the test conditions is that the steam for PCT is condensing while it is uncondensing for HAST. Thus, moisture condenses on the DIPs during PCT. Due to the dependence of humidity on temperature, fluctuations in temperature are greater in PCT than HAST. Finally, the time of testing was also different for the two tests. On the other hand, the thickness of the film made little difference on the interfacial adhesion damage or leakage current behavior for the DIPs. IV. STATISTICAL ANALYSIS OF GEOMETRICAL FEATURES From Figs. 28, the variability in the size, shape and location of the geometrical features of the damage is apparent. A combination of statistical estimation techniques was used to analyze the image analysis data. The first task was to estimate the cumulative distribution function (cdf) for the geometrical features of the interfacial adhesion damage. The two-parameter Frechet cdf given by: (1) where is the shape parameter and is the scale parameter, of was considered. The mean and coefficient of variation this cdf are given by and (2)

Fig. 7. Example of the clustering of small blisters induced by HAST for 20 days; PIA24-3 with APS.

Fig. 8. Typical damage that forms along a conductor induced by HAST for 20 days; PIA24-3 with APS.

only 2600 m and 130 m , respectively. Compared with PCT, the interfacial adhesion damage induced by HAST is quite different. The damage shown in Fig. 7 appears to be delamination blisters. Since these blisters are quite small, they are representative of an earlier stage of damage growth. Given sufficient time these blisters would continue to grow and coalesce into a single feature of substantial size. In fact, Fig. 8 may reflect the growth through the gray scale shading. The black region represents early damage and the ensuing stages propagate in the rings about it. The lighter area is damage that has coalesced recently with the larger damage. Notice that outside the large feature there are numerous very small blisters, which would coalesce into the large feature, if continued growth were to occur.

is the incomplete gamma function. The Frechet cdf where, was chosen because it is the cdf associated with the extreme value of the maxima of random variables that are concentrated on the nonnegative real numbers [22]. Such a cdf is preferred because the larger features are indicative of more severe damage. Fig. 9 is the graph for the data for the maximum chord, minimum chord, and cross sectional area of the damage observed from the DIP, denoted as PIA24-1, with APS and subjected to HAST for 20 days. The maximum chord is the longest dimension of the feature, and likewise the minimum chord is the shortest. Similarly, Fig. 10 is for the DIP, denoted by PI47-5, without APS and subjected to PCT for 38 days. The data are plotted on Frechet probability graph paper, for which linearity implies that the Frechet cdf is an appropriate fit to the data. The choice of the Frechet cdf to represent the data is confirmed by the excellent linear fit of the data. It is suggested that the geometrical features of interfacial adhesion damage obtained from all DIPs with and without APS tested by HAST or PCT can be represented by a two-parameter Frechet cdf. The parameters for the cdfs were estimated by using the maximum likelihood method [23]. Furthermore, the goodness-of-fit of the data to the estimated Frechet cdfs were validated with the Kolmogorov-Smirnov test [24] at a confidence level of 99%. Tables IIV summarize the primary statistical parameters for several different DIPs. Table I is for 3 DIPs without APS subjected to PCT for 38 days; Table II is for 4 DIPs with APS

104

IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 1, FEBRUARY 2000

Fig. 9. Frechet cdfs for the geometrical features induced by HAST for 20 days; PIA24-1 with APS.

merged statistically with confidence levels that exceed 90%. In other words, the individual sets of data represented by Table II statistically may be considered to be from one population which, in turn, may be characterized by a single Frechet cdf. The same is true for the data represented by Table III. The maximum likelihood estimations for the geometrical features based on the merged data represented in Tables IIII are given in Table IV, with the caution that merging the data from the DIPs without APS tested under PCT needs special consideration. It may be that one of the specimens listed in Table I was subjected to conditions that were somewhat inconsistent with the others. Unfortunately, this cannot be determined after the fact. A more likely possibility is that the severe testing conditions from PCT applied to specimens without APS leads to more statistical variability than the other test conditions. In PCT moisture condenses on the DIPs, which may cause erroneous measurements from water stains. In any case, further experimentation and analysis is needed to sort through this dilemma. Thus, the reasonable approach seems to be that HAST testing, with varying RH levels of 90, 95, and 100%, is preferred, along with image analysis of the surface before and after each test. V. RESULTS AND DISCUSSION In solution, PI-2555 is susceptible to trapped air during the spin coating process. If the spin coating is not done carefully, spiral clusters of trapped air bubbles can form. While the initial leakage current seems to be unaffected by these bubbles, they are sites in which delamination of the coating and substrate readily ensue. Consequently, all of the DIPs considered herein were free of bubbles, as determined by microscopy prior to stress testing. The leakage current behavior is accelerated by PCT and can be classified qualitatively by three regions. Small values for the leakage current prior to stress testing reflects good adhesion at the interface. The first region consists of leakage currents in the pA to nA range. This reflects the gradual penetration of moisture into the PI. In the second region, the leakage current fluctuates between 1 and 10 nA. This behavior may be due to ionic mobility and the charging and discharging of ionic traps. Mobile ions change the shape of the current transient, but steady state current is not strongly dependent on ionic concentration. To correlate these measurements with interfacial adhesion damage, each DIP was observed with optical microscopy. The observed interfacial adhesion damage is sufficiently small that an electrical failure is not imminent. Electrical failure commonly results in DIPs that have poor adhesion at the interface, but such is not the case with PI-2555. The third region covers dramatic increases in leakage currents extending from 0.1 to 1.0 A. Typically the corresponding damage consists of either a single delamination that spans the gap between the gold conductors or a large cluster of damage. Continuous degradation at the interface from stress testing, induces an increase in leakage currents, ultimately leading to an electrical failure. Because of the different test conditions for HAST and PCT, it is difficult to compare directly each of the test results. Comparisons were made on the same day of each test by microscopy

Fig. 10. Frechet cdfs for the geometrical features induced by PCT for 38 days; PIA47-5 without APS.

subjected to PCT for 38 days; and Table III is for 4 DIPs with APS subjected to HAST for 20 days. The maximum chord and minimum chord have dimensions of m, and the area has dimensions of m2. The carat over the parameters in these tables indicates that they have been estimated from the available data. The maximum likelihood method was used to obtain and in the Frechet cdf, whereas and were computed. The sample and size, i.e., the number of observed damaged sites, is and are the sample average and sample standard deviation, respectively. It is apparent from Table I that and for the geometrical features are not that similar for these three specimens without APS tested by PCT. In fact, the likelihood ratio test [25] confirmed that these data could be merged only for a confidence level that exceeds 99.9%. This is somewhat surprising, because the preparation and testing of the DIPs were nominally identical. Statistically, the reason that the three populations should be merged with caution is that the differences between the values for are quite large. In contrast, the geometrical features of the DIPs with APS tested by HAST or PCT can be

PARK et al.: CHARACTERIZATION OF INTERFACIAL ADHESION DAMAGE

105

TABLE I ESTIMATED PARAMETER VALUES FOR THE FRECHET CDF FOR THE GEOMETRICAL FEATURES OF DAMAGE FOR THREE DIPs WITHOUT APS AND SUBJECTED TO 38 DAYS OF PCT; 121C/SATURATED STEAM/UNBIASED. THE UNIT OF LENGTH FOR THE MAXIMUM AND MINIMUM CHORDS IS m, AND IT IS m2 FOR THE AREA

TABLE II ESTIMATED PARAMETER VALUES FOR THE FRECHET CDF FOR THE GEOMETRICAL FEATURES OF DAMAGE FOR THREE DIPs WITH APS AND SUBJECTED TO 38 DAYS OF PCT; 121C/SATURATED STEAM/UNBIASED. THE UNIT OF LENGTH FOR THE MAXIMUM AND MINIMUM CHORDS IS m, AND IT IS m2 FOR THE AREA

TABLE III ESTIMATED PARAMETER VALUES FOR THE FRECHET CDF FOR THE GEOMETRICAL FEATURES OF DAMAGE FOR THREE DIPs WITH APS AND SUBJECTED TO 20 DAYS OF HAST; 121C/100% RH/UNBIASED. THE UNIT OF LENGTH FOR THE MAXIMUM AND MINIMUM CHORDS IS m, AND IT IS m2 FOR THE AREA

TABLE IV ESTIMATED PARAMETER VALUES FOR THE FRECHET CDF FOR THE GEOMETRICAL FEATURES OF DAMAGE FOR THE MERGED DATA. THE UNIT OF LENGTH FOR THE MAXIMUM AND MINIMUM CHORDS IS m, AND IT IS m2 FOR THE AREA

and corresponding changes in leakage currents. The evolution of interfacial adhesion damage induced by PCT was more severe than that caused by HAST during the same time intervals. Leakage currents for PCT are 1 to 2 orders of magnitude higher than those for HAST after 20 days of testing. Again, PCT is harsher because of saturated moisture and the temperature variations. Therefore, the use of HAST for accelerated aging is preferred.

The effect of APS based on the leakage current for DIPs tested under PCT is not as significant as would be expected for an increase in adhesion strength at the interface between the PI and the substrate. The reasons may be due to the following: i) ionic impurities in the adhesion promoter [26] or ii) decomposition of chemical bonds of APS at high temperature [27]. The glass transition temperature Tg for PI-2555 was measured by DSC and was found to be 312 C. To estimate the effect of

106

IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 1, FEBRUARY 2000

moisture reaction on thin PI films, a 1 mg water drop in an aluminum pan was added, and the Tg diminished to 128 C. This may imply that a small amount of water at the interface can have drastic consequences. It has been reported that the decrease in Tg during PCT exposure suggests that a hydrolytic degradation may have taken place in the PI coating [28]. It also is interesting that the time required for a significant decrease in Tg under PCT is nearly the same as the time required for a decrease in adhesion [28]. The hydrolysis of PI-2555 at the interface reflected by the leakage current and the reduction in Tg is similar to previous studies of PI-2555 on DIPs [14], [15]. Alternatively, whether or not APS is added, it is known that PI-2555 has good adhesion strength compared to other PIs, even if other mechanical properties, e.g., high residual stresses and a high coefficient of thermal expansion, are not as optimal [29], [30]. Thus, for PI-2555 the addition of APS may be of little importance. To examine the interface of a delaminated region, XPS was used. The atomic percentages of the chemical elements were found to be the following: 50% C, 25% Al, 4% Au, with traces of O, N, and Cu. Neither Na nor Cl was detected. Note that the initial content of Cu in PI-2555 is less than 0.5 ppm. Thus, the growth of interfacial adhesion damage is not directly associated with contaminants. The XPS results suggest that the increase in leakage current may be related primarily to the penetration of moisture through the PI-2555 during the humidity test conditions. VI. CONCLUSION DIPs coated with PI-2555, with and without APS, and exposed to HAST and PCT were examined by optical microscopy and leakage current measurements to monitor the degradation at the interface. The maximum chord, minimum chord and area of each interfacial adhesion damage site are represented well by a two-parameter Frechet cdf for each test condition. PCT is a more severe environmental test than HAST, and this is reflected by higher and more rapid changes in leakage current and sizes of the damage. There is more statistical variability in the observed geometrical features of the damage after PCT than after HAST, which reflects the uncertainties that lead to the low reproducibility of the results obtained for PCT. It has been shown that the leakage current is qualitatively related to the size and extent of the interfacial adhesion damage. Interfacial adhesion damage is spread uniformly over most of the interface, but there seems to be more clustering for PCT conditions. Furthermore, interfacial adhesion damage growth is not associated directly with contamination at the interface. A future goal of the present research is to quantify the growth kinetics of interfacial adhesion damage and the primary failure mechanisms by using an appropriate polymeric coating under unsaturated HAST conditions. ACKNOWLEDGMENT The authors would like to thank A. Benscoter, Lehigh University, for his help with the optical microscopy, P. Zeigler for information obtained, and Dr. R. Pearson, Lehigh University, for several discussions.

REFERENCES

[1] R. Ginsburg and J. R. Susko, High temperature stability of a polyimide film, IBM J. Res. Develop., vol. 28, no. 6, pp. 735740, 1984. [2] L. M. Baker, P. J. Brown, and J. L. Markham, Polyimide for dielectric layers, Amer. Chem. Soc., pp. 127139, 1989. [3] D. Y. Guan, T. F. Gukelberger, and E. C. Cahoon, New failure mechanisms in temperature and humidity tests, in Proc. IEEE 36th Electron. Comp. Conf., 1986, pp. 107110. [4] R. Kodnani, D. Jaffe, and R. Jaccodine, Evaluation of polymeric coatings for MCM application, IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 18, pp. 658665, Dec. 1995. [5] P. R. Troyk, D. Conroy, and J. E. Anderson, Modeling of triple-track and comb-pattern leakage current measurements, in Polymeric Materials for Electronics Packaging and Interconnection. Washington, DC: American Chemical Society, 1989, ch. ch. 22, pp. 249267. [6] J. E. Anderson, K. M. Adams, P. R. Troyk, and R. Frankovic, Polymercoated microelectronics: Comparison of bulk, surface, and interphase conductivities, IEEE Trans. Comp,. Hybrids, Manufact. Technol., vol. 14, no. 2, pp. 420427, June 1991. [7] J. E. Gunn, S. K. Malik, and P. M. Mazumdar, Highly accelerated temperature and humidity stress test techniques (HAST), in Proc. IEEE Int. Rel. Phys. Symp., 1981, pp. 4851. [8] J. E. Gunn, R. E. Camenga, and S. K. Malik, Rapid assessment of the humidity dependence of IC failure modes by use of HAST, in Proc. IEEE Int. Reliability Phys. Symp., 1983, pp. 6672. [9] L. T. Nguyen, R. H. Lo, A. S. Chen, and J. G. Belani, Molding compound trends in a denser packaging words: Qualification test and reliability concerns, IEEE Trans. Rel., vol. 42, pp. 518535, Dec. 1993. [10] N. Sbar and L. G. Feinstein, Performance of new based metallization systems in an 85C 80% RH Cl contaminated environment, IEEE Trans. Parts. Hybrids. Packag., vol. 13, pp. 208218, Sept. 1977. [11] N. L. Sbar and R. P. Kozakiewicz, New acceleration factors for temperature, humidity, bias testing, IEEE Trans. Electron Devices, vol. 26, pp. 5671, Jan. 1979. [12] M. Iannuzzi, Reliability and failure mechanisms of nonhermetic aluminum SICS: Literature review and bias humidity performance, IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 6, pp. 181190, June 1983. [13] R. G. Mancke, A moisture protection screening test for hybrid circuit encapsulation, IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 4, pp. 492498, Dec. 1981. [14] A. W. Lin, Evaluation of polyimides as dielectric materials of multichip packages with multilevel interconnection structure, IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 13, pp. 207213, Mar. 1990. [15] J. J. Burack, J. D. Legrange, and A. W. Lin, Enhanced moisture protection of electronic devices by ultra-thin polyimide films, IEEE Trans. Comp., Hybrids, Manufact. Technol., vol. 13, pp. 214218, Mar. 1990. [16] DuPont guideline and data sheet for PI-2555 as interlayer dielectric,. [17] H. J. Nehaus and S. D. Senturia, Conduction transients in polyimides, in Polymeric Materials for Electronic Packaging and Interconnection. Washington, DC: American Chemical Society, 1989, pp. 177189. [18] J. McHardy, D. I. Basiulis, G. Angsten, L. T. Higley, and R. N. Leyden, Accelerated testing of polyimide coating for neural prosthesis, in Polymeric Materials for Electronic Packaging and Interconnection. Washington, DC, 1989, pp. 165175. [19] W. Kern and D. A. Puotinen, Cleaning solutions based on hydrogen peroxide for use in silicon semiconductor technology, RCA Rev., vol. 31, pp. 187206, June 1970. [20] J. Ruzyllo, G. T. Duranko, and A. M. Hoff, Preoxidation UV treatment of silicon wafers, J. Electrochem. Soc., vol. 134, no. 8, pp. 20522055, Aug. 1987. [21] J. Park and D. G. Harlow, Statistical modeling of interfacial damage of polymer encapsulated microelectronic devices, Int. J. Microc. Electron. Packag., vol. 21, no. 2, pp. 171176, 1998. [22] R. E. Barlow and F. Proschan, Statistical Theory of Reliability and Life Testing: Probability Models. Silver Spring, MD: To Begin With, 1981. [23] L. M. Leemis, Reliability: Probabilistic models and statistical methods. Englewood Cliffs, NJ: Prentice-Hall, 1995. [24] R. B. DAgstino and M. A. Stephens, Goodness of fit techniques. New York: Marcel-Dekker, 1986, pp. 387393. [25] J. F. Lawless, Statistical Models and Methods for Life Time Data. New York: Wiley, 1982. [26] R. Kodnani, SRC Annu. Report. Bethlehem, PA: Lehigh Univ., 1994.

PARK et al.: CHARACTERIZATION OF INTERFACIAL ADHESION DAMAGE

107

[27] H. Ishida and K. Nakata, The role of aminosilanes on the adhesion promotion at the E/Glass/Poly butylene terephthalate interface, SAMPE Quart., pp. 2128, 1987. [28] P. Hoontrakul, Interfacial adhesion of polymeric coatings for microelectronic encapsulation, M.S. thesis, Lehigh Univ., Bethlehem, PA, 1992. [29] T. G. Tessier, G. M. Adema, and I. Turlik, Polymer dielectric options for thin film packaging application, in Proc. IEEE 39th Electron. Comp. Conf., 1989, pp. 127134. [30] T. C. Hodge, N. R. Grove, B. Sinno, S. A. Bidsrup-Allen, and P. A. Kohl, Thermo-mechanical stresses in spin-coated films on silicon substrate, Int. J. Microcirc. Electron. Packag., vol. 20, no. 1, pp. 1220, 1997.

D. Gary Harlow received the B.A. degree from Western Kentucky University, Bowling Green, in 1973, and the M.S. and Ph.D. degrees from Cornell University, Ithaca, NY, in 1976 and 1977, respectively. He is a Professor of mechanical engineering and mechanics at Lehigh University, Bethlehem, PA. He has over 20 years of mechanistically based reliability modeling and analysis experience. His research has been sponsored throughout his career by the NSF, FAA, and DoD. Most recently, his primary research has been focused on aging aircraft issues and reliability of electronic packages.

Jongwoo Park received the B.S. and M.S. degrees in mechanical engineering from Hong-Ik University, Seoul, Korea, in 1989 and 1991, respectively, and the Ph.D. degree from Lehigh University, Bethlehem, PA, in 1998. From 1991 to 1993, he worked at ASIA Motor Company as a Research Engineer for the CAE team. He joined Lucent Technologies, Murray Hill, NJ, in 1999, as Member of Technical Staff. His research interests include mechanical and environmental degradation of electronic devices and the subsequent reduction in reliability.

Herman F. Nied received the B.S. degree in mechanical engineering from the University of Rochester, Rochester, NY, in 1976 and the M.S. and Ph.D. degrees in applied mechanics from Lehigh University, Bethlehem, PA, in 1978 and 1981, respectively. He has been a Professor of mechanical engineering and mechanics at Lehigh University since 1995. Prior to joining the Lehigh faculty, he was Member of Technical Staff, General Electrics Corporate Research and Development Center, Schenectady, NY. His research interests include electronic packaging, fracture mechanics, finite element simulation, and polymer processing. He holds 11 U.S. and three European patents. Dr. Nied received the CAREER Award from the National Science Foundation in 1997.

Potrebbero piacerti anche