Effect of Hydrogen on GaAs MMICs in Hermetic Packages. D. N. Goswami Picosecond Pulse Labs 2500 55th Street Boulder, CO 80301, USA Phone: (303) 209-8126, FAX: (303) 209-8203, e-mail: dgoswami@picosecond.com Abstract Presence of hydrogen in a hermetic package containing GaAs semiconductor devices can adversely affect performance and reliability of the GaAs devices. The problem is believed to be caused by conversion of molecular hydrogen to atomic hydrogen through the catalytic reaction with platinum or palladium in the gate structure of a GaAs device, which then diffuse into the active area of the device affecting its electrical performance. The primary source of hydrogen in the package is the trapped hydrogen in the ferrous metals used in package fabrication, and hydrogen that gets absorbed in the nickel plating during electroplating of the package. Past studies have shown that performance of GaAs device can degrade severely at hydrogen level as low as fraction of a percent. The device fabrication process and gate structure/metallurgy also impact the sensitivity to the hydrogen level. This paper presents the results of a study on the level of hydrogen in gold/nickel plated aluminum package as a function of different high temperature bake, and presents the Residual Gas Analysis (RGA) results for hydrogen in hermetic hybrids with GaAs MMICs that have successfully completed 1000 hour life test as a part of Telcordia Hybrid Qualification. The results show that hydrogen content in a hermetic package can be reduced to less than 1000 ppm by high temperature bake, and for telecommunication applications (operating ambient 60°C or less) hydrogen poisoning of GaAs MMIC in hermetic hybrid is not a concern so long as the hydrogen level is kept below 5000 ppm. Key Words: GaAs PHEMT, Aluminum Package, Hydrogen outgassing, Hydrogen poisoning, 1. INTRODUCTION Use of GaAs ICs in various high frequency applications has increased significantly over the last few years. This has created considerable interest in the reliability of these devices. One area of concern has been the effect of hydrogen gas present in hermetic packages containing GaAs devices. Early work in this area reported that hydrogen gas in quantities as low as 0.5% of ambient atmosphere can cause significant performance degradation at temperature of 125 C in a relatively short period (500 hours) [1]. The degradation has been reported for both metal semiconductor field effect transistor (MESFET) and pseudomorhic high electron mobility transistor (PHEMT) devices involving amplifier functions [2-12], but not for heterojunction bipolar transistor [HBT], PIN diodes, and phase shifter functions [12,13]. The problem manifests as a sudden drop in current and transconductance, and is believed to be caused by conversion of molecular hydrogen to atomic hydrogen through the catalytic reaction with platinum or palladium in the gate structure, which then diffuse into the active area of the device affecting its electrical performance [1,14]. No degradation has been observed when the gate did not contain any Platinum or palladium [8]. The primary source of hydrogen in the package is the trapped hydrogen in the ferrous metals used in package fabrication, and hydrogen that gets absorbed in the nickel plating during electroplating of the package. Studies have shown that hydrogen content of the package materials can be reduced, but not eliminated by thermal annealing [2]. Since aluminum does not absorb hydrogen, use of aluminum package is preferred. But hydrogen can still get adsorbed in the nickel under-layer during gold plating of the package that can later outgas creating a hydrogen rich environment inside the package cavity. There is also the concern that hydrogen may react with other residual gases and oxides inside a hermetic hybrid creating a moisture rich environment. Although the problem has been known for more than a decade, and various ways to mitigate the problem has been discussed in the literature [13,14], there is no clear guideline or limit as to what level of hydrogen might be considered acceptable in hermetic hybrids. The problem is somewhat involved since the device type, fabrication process, feature size, gate structure/metallurgy, and operating ambient temperature, - all can impact the sensitivity to the hydrogen level. This paper presents (1) the effect of thermal bake on hydrogen content in Ni/Au plated aluminum packages, and (2) 1000 hr life test data including RGA results for hydrogen in hermetic hybrids that used plated aluminum package with no thermal bake. 2. Background Figure 1. Test Package Several OEMs using Picosecond Pulse Labs (PSPL) hybrid amplifier devices in the design of their products required the hybrids to be qualified to full Telcordia requirements before using them in production systems. They also wanted assurance that there would be no hydrogen related device degradation over the operating life. Key points that had to be addressed were: (1) What level of hydrogen can be expected in a gold plated aluminum package used for PSPL hybrids, and does it pose any reliability risk? (2) If the hydrogen level is high, what thermal treatment may be used for the package to reduce the hydrogen level to an acceptable level? There was not enough time to fully address the above issues before starting the qualification life test if we were to meet customer’s deadline. However, there had been no evidence of any hydrogen related performance degradation of PSPL devices that had been out in the field and hydrogen outgassing did not appear to be a serious issue with the Ni/Au plated aluminum packages used in PSPL hybrids. Hybrid qualification testing was, therefore, started using as-plated aluminum package, but a parallel experiment was also carried out to study hydrogen outgassing from Ni/Au plated aluminum packages. 3. Hydrogen Outgassing plated Aluminum Package Test samples consisted of rectangular aluminum housings (1"X1"X0.375") with 0.125" thick sidewalls (Figure 1) which were representative of production packages used in PSPL hybrids. The test housings were plated with 50 microinch of gold over 100-300 microinch of nickel. An air circulating oven was used to bake the packages. Packages were baked, laser welded, conditioned at 125 C for 168 hour, and then sent out for RGA test. 3.1 Post Nickel Plating Bake It seemed logical that thermal bake of the nickel layer prior to gold plating would be the most efficient way to get rid of the hydrogen from the nickel plating. RGA results for hydrogen in samples subjected to different post nickel plating bake are shown in Table 1. Samples 1-3 with no thermal bake showed 16500 2000 ppm of hydrogen. Such high level of hydrogen resulting from the relatively thin (100300µ”) layer of nickel was somewhat of a surprise. Sample 3 was plated by a different vendor who claimed to have a nickel plating process that minimized hydrogen absorption; however, we did not see any evidence of it. Samples 4 and 5 with 190 C/3hr bake showed a sharp decrease in hydrogen to less than 5000 ppm. So, when the bake temperature/time was raised from 190 C/3 hr to 250 C/48 hr for samples 6,7, a very low level of hydrogen was expected inside the package. However, the samples still showed around 2000 ppm of hydrogen. A possible explanation was absorption of additional hydrogen by the nickel layer during subsequent gold plating. This implied that to reduce hydrogen level below 2000 ppm the packages would have to be baked after gold plating. Table 1. RGA Result for H2 (Post Ni plating Bake) Sample Post Ni Post AuCavity # Plating bake Plating bake Content H2 (ppm) 1,2 None None Empty 16,800, 14,500 3* None None Empty 18,200 4,5 3 hr / 190 C None Empty 4,986, 4,316 6,7 48 hr / 250 C None Empty 2,014, 2,147 * Package plated by a different vendor from the rest. 3.2 Post gold-plating bake Samples were prepared by performing postgold plating bake on packages with and without postnickel plating bake (see Table 2). One of the samples (sample11) contained a Duroid substrate laminated to Ni/Au plated Cu-W carrier which could be also a possible sources for hydrogen, while another (sample12) contained a RF absorbing silicone rubber which is used under the package lid in some rf hybrids. RGA results for hydrogen are shown in Table 2. Table 2. RGA Result for H2 (Post Au plating Bake) Sample Post Ni Post AuCavity Content H2 # Plating bake Plating bake (ppm) 8 3 hr / 190 C 168 hr / 125 C Empty 1305 9 3 hr / 190 C 130 hr / 215 C Empty 121 Empty 721 10 11 12 None 168 hr / 125 C 48 hr / 250 C 168 hr / 125 C Duroid Subs. Laminted to Cu-W Carrier * 48 hr / 250 C 168 hr / 125 C RF absorbing Siicone rubber* 777 511 * Pre-Baked at 125 C for 168 hr Comparing the results for samples 8, 9 with those of samples 4, 5 in Table-1, it is clear that postgold-plating thermal bake is quite effective to reduce the level of hydrogen. Sample 9 which was baked at a higher temperature of 215ºC showed only 121 ppm of hydrogen. Sample 10 did not have any post nickel plating bake, but it showed lower hydrogen content than sample 8 that was baked after both nickel and gold nickel plating. The difference is well within statistical error, but it indicates that thermal bake of the nickel layer prior to gold plating (which often creates considerable logistical problem at the Plater’s facility) is not essential. Gold Plating does not appear to impede the escape of hydrogen from the nickel layer and thermal bake of the package after final gold plating is quite effective in removing hydrogen from the nickel layer. Sample 11 with the laminated substrate inside the cavity did not show any significant increase in hydrogen. 125°C/168 hr pre-bake of the laminated substrate appeared to be quite effective in removing any trapped hydrogen in the nickel plating of the carrier and the substrate. Sample 12 with the silicone rubber sheet did not show any increase in hydrogen level either. This was not unexpected since the silicone rubber did not contain any hydrogen. However, the RGA results for the package showed high level of CO2 (~9500 ppm) and methane (~2500 ppm) which may create other reliability hazards. 4.0 Life test 1000 hour Operating life test was performed as a part of Telcordia hybrid qualification in accordance with TR-NWT-000930, Generic Requirements for Hybrid Microcircuits Used in Telecommunications Equipment. All qualification tests (see Table 3) were successfully completed, but only the operating life test and post-operating life RGA test data are presented here. Table 3. Telcordia Qualification Test Requirement TEST Electrical Sampling Plan * MIL-STD-883 NOTE Method Condition LTPD SS C 30+ - Physical Mech Shock 2002 Vibration Thermal Shock (15 cycles) Temp Cycling (100 cycles) 2007 1011 Operating Life 1005 Low Temp. Storage Int Moisture RGA 1010 - 5 - Cond B 10 38 1 Cond B 10 10 38 38 1 1 A, B 10 38 1 B, C, D 1000 hr 10 38 1 E -40 C/100 hr 20 11 0 Cond B, -40 C/125 C Cond B, -40 C/125 C 1018 10 F * LTPD= Lot tolerance percent defective, SS = Sample Size, C= Max number of failures allowed. NOTES A. The same set of samples shall be used for all mechanical tests. The total number of failures is counted against accept/reject criteria. The mechanical test may be performed in any order. B. Hermeticity tests (fine/gross leak) consistent with MIL-STD-883, Method 1014 shall be performed in addition to electrical tests. C. The lower temperature limit shall be -40 C as specified in Bellcore TR-NWT-000930. D. Samples completing Mechanical tests shall be used for this test. E. The ambient temperature shall be chosen to maintain the maximum IC junction temperature during Operating Life Test in the 155 5 C range. F. Parts completing Operating Life tests shall be used in these tests. 4.1 Test Sample A 12.5 Gbit/Sec Modulator Driver Amplifier, PSPL P/N 5865 (Figure 2), that is fairly representative of the PSPL amplifier product family, was used as the test vehicle. It contained a 3.3mmX 2.3mm GaAs PHEMT amplifier IC containing 9 cascaded pairs of dual gate FETs (0.5 m Technology). Packages were not subject to any thermal bake and a sample size of 38 (Telcordia Qualification requirement) was used for the life test. Figure 2: Test Device showing the supply voltages used for Life test. (No connection to pins QC, CP, and VB) 4.2 Test Details Figure 3a. Test Chamber (left) with Power Supply and Test Monitor (right) An air circulating Environmental Chamber was used for the life test (Figures 3). Life test was performed under steady state condition using a supply voltages of +8V for drain and –5V for the gate (Figure 2). No heat sink was used to cool the devices during life test. Under the test condition, the PHEMT IC dissipated 1.92 watt of power. Based on the thermal resistance data provided by the IC vendor, this corresponds to a Junction-to-Case temperature difference of 70 C. The case temperature during life test was maintained at 85 5 C to prevent IC junction temperature from going much over the recommended maximum IC junction temperature of 150 C. An automated test program monitored the current drawn by each device and saved the information every 60 minutes. The test circuit had a built-in safety feature that shut down the power going to any specific device if its current exceeded 325mA. Electrical measurements were made at room ambient after 0, 168, 504, and 1000 hours of life test. Measured parameters are shown in column 1 in Table 4. A device was considered to have “failed” if it did not meet the electrical specification for the hybrid or if any of the following conditions occurred: Risetime10-90% changed by more than 6 ps. Jitter changed by more than 6 ps. Eye amplitude (Voutp-p ) changed by more than 10% ((Input= 0.5 Vp-p). 4.3 Test Results 4.3.1 Electrical Results A summary of electrical test results is shown in Table 4. No evidence of any degradation of electrical properties has been observed over the operating life; observed changes were well within the range of measurement error. Figure 3b. Units under Test in the Test Chamber Table – 4. Electrical Test Results (sample size=38) PARAMETER Eye Amplitudep-p (Volt) Risetime10-90% (ps) Eye Jitterp-p (ps) Noise Figure (dB) S21, S11, S22 4.3.2 0 hr 168 hr 504 hr 1000 hr Av. Max Min Av. Max Min Av. Max Min Av. Max Min 8.1 8.6 7.8 29.9 32.4 28.0 13.0 14.2 10.7 5.3 5.7 5.2 8.0 8.3 7.8 29.9 32.4 28.0 12.7 14.2 10.2 5.4 5.7 5.2 8.2 8.6 7.9 29.5 33.3 27.6 12.8 14.2 11.1 5.3 5.6 5.1 8.4 8.7 8.2 29.8 32.4 28.0 12.8 14.2 10.7 5.3 5.6 5.2 No statistically significant change. (plots not presented here) RGA Results RGA testing was performed on 10 units that completed the life test. Results are shown in Table 5. An Ar/He (75/25) mixture was used for the welding environment and it is reflected in the RGA data. Hydrogen level inside the packages ranged from 1279 to 5986 ppm, while moisture content ranged from 1375 to 2807 ppm. A fair amount of nitrogen was also detected in the packages. This occurred due to inadequate flushing of the nitrogen from the welding chamber prior to welding of the packages, and has since been corrected. Packages also showed Argon (%) CO2 (%) Moisture (ppm) Hydrogen (ppm) He (%) Fluoro Carbons Others (ppm) K. Decker performed an extensive GaAs hydrogen sensitivity study with PHEMT devices (200µm X 0.25 µm test FETs) and concluded that electrical bias did not effect the hydrogen degradation mechanism and it was the ambient temperature (not channel temperature) that drives the hydrogen degradation mechanism [3]. His data fitted the model developed by Adams, et. al. [6] that is given by the Equation O2 (ppm) Table – 5. RGA results (post Operating Life) N2 (%) ppm level seen in the aluminum test housings. High temperature exposure of the package during solder lead attach and device screening (temperature cycling, and 168 hour open burn-in) eliminated about two-thirds of the hydrogen from the plated packages. Sample # CO2 in the 0.67% to 1.90% range. According to the RGA Test Lab, this is not uncommon for packages containing FR4 and Duroid boards which the test hybrids had. High level of fluorocarbon in one of the units was attributed to contamination from gross leak test performed prior to RGA. 1 2 3 4 5 6 7 8 9 10 0.68 ND 75.4 0.93 1,770 5109 22.3 ND <100 0.79 124 75.1 0.72 2,384 5986 22.5 ND 455 0.67 ND 74.6 1.13 2807 5816 22.7 ND 158 4.70 301 74.3 1.29 2622 1279 19.3 220 ppm <100 3.44 107 65.8 1.90 2212 3706 27.2 1.78 % <100 0.63 ND 75.2 1.23 2561 3090 22.4 ND <100 2.49 2133 74.9 0.88 2584 3101 20.9 491 ppm <100 0.58 ND 76.0 0.67 1875 1809 22.4 ND <100 0.72 390 75.1 0.90 2046 2709 22.7 ND <100 0.58 ND 75,.8 0.77 1375 3603 22.4 ND 110 5. Discussions Computer analysis has shown that any change in Ids of the GaAs PHEMT from possible hydrogen poisoning will affect S21 and eye amplitude parameters for the hybrid as shown in Figs 4 and 5 respectively. There was no evidence of any such change during life test. S21 (dB) Nominal Delta Ids = + 10% Delta Ids = - 10% Delta Ids = - 20% 30 25 20 t = A Pnexp(Ea/kT) (1) where t = median life (hr) for 10% Idss degradation A = a proportionality constant (5.46E-6) P = hydrogen partial pressure (%H2 X 7.6) n = -0.7935 Ea = activation energy (0.73 eV) K = Boltzman Constant (8.615E-5) T = ambient temperature (ºK) If the model is assumed to be valid for the PHEMT devices used in this study, then for ambient temperature of 85ºC and hydrogen level of 5,000 ppm, (conditions representative of the present life test study), testing has to be continued for 36,000 hours (4.1 yr) before the device fails (10% degradation in Ids). It is, therefore, not surprising that we did not see any hydrogen poisoning effect in the 1000 hr life test. 15 10 0 1 2 4 6 8 10 12 14 16 18 20 Frequency (GHz) Chnage in Eye Amplitude Figure 4. Effect of change in Ids on S21 10% Hybrids used in this study contained various epoxy adhesives, soldered parts and duroid/FR-4 composite boards. The fact that moisture content in the hybrids remained below 3000 ppm (allowed maximum 5000 ppm) suggests that there is no evidence of hydrogen reacting with outgassed products or other oxides inside the hybrid creating excessive moisture. 0% -10% -20% -20% -10% 0 10% Change in Ids Figure 5. Effect of change in Ids on Eye Amplitude The amount of hydrogen detected inside the hybrids after life test was much lower than the 15000 Table 5 shows the calculated level of hydrogen from equation (1) for a median product life of 20 years for different operating ambient temperatures. Table 5. H2 level for 20 year Median Life Median Life T=40 C T=50 C T=60 C T=70 C 20 Yr 4.96% 1.73% 0.64% 0.25% Hydrogen Level Since the typical operating ambient temperature for hybrids in telecommunication application is not expected to be more than 60°C, a hydrogen level of 5000 ppm should be acceptable for such application. However, to be on the safe side, 3000 ppm would be a more reasonable value, and it can be easily achieved for aluminum package by prebaking the package at 125°C bake for 168 hours or a shorter bake time at a higher temperature in the 150°C to 175°C range. 6. CONCLUSIONS: Hydrogen desorption from Ni/Au plated aluminum packages can result in hydrogen level of more than 15,000 ppm of hydrogen in hermetic packages. Hydrogen level can be reduced to 1000 ppm level by 125°C/168hr thermal bake of the package. 1000 hour operating life test performed during hybrid qualification testing is not capable of detecting potential hydrogen problem in hermetic hybrids. Test duration is too short for the level of hydrogen seen in typical hermetic packages. For telecommunication equipment, a hydrogen level of 3,000 ppm seems to be a reasonable limit. It would meet the requirement of 20 year life at 60°C operating ambient and can be easily achieved for Ni/Au plated aluminum package by thermal bake of the package for 168 hours at 125ºC or a shorter duration at a higher temperature. Acknowledgement The author would like to thank Dr. James Andrews and Clayton Smith for their support, encouragement and helpful discussions and Dr. Jason Yoho for the computer simulation results relating the changes in Ids of the PHEMT die to the changes in S21 and Eye Amplitude of the hybrid amplifier. References [1] W.O. Camp, Jr., R Lasater, V Genova, R Hume, "Hydrogen effects on reliability of GaAs MMICs", Proceedings of 11th GaAs Integrated Circuit Symposium, pp. 203-206, 1989. [2] P. Schuessler, and D. Felciano-Welpe, “The Effects of Hydrogen On Device Reliability”; Hybrid Circuit Technology, vol. 8, No.1, pp. 19-26, January 1991. [3] K. Decker, "GaAs MMIC Hydrogen Degradation Study", GaAs Reliability Workshop Digest, 1994 [4] W. Roesch “Accelerated Effects of Hydrogen on GaAs MESFETs”; GaAs Reliability Workshop Digest, 1994. [5] M.J. Delaney, T.J. Wiltsey, Min-Wen Chiang, K.K. Yu, "Reliability of 0.25µm GaAs MESFET MMIC Process: Results of Accelerated Lifetests and Hydrogen Exposure" GaAs Reliability Workshop Digest, 1994 [6] S. Adams, J. MacDonald, W. Hu, A. Immorlica, A. Reisinger, and F. Smith, “Reliability of GaAs PHEMT MMICs in Hydrogen Ambients”; GaAs Reliability Workshop Digest, 1994. [7] W.W. Hu, T.H Parks, P.C. Chao, A.W. Swanson, "Reliability of GaAs PHEMT Under Hydrogen Containing Atmosphere", GaAs IC Symposium, 16th Annual Technical Digest, 1994, p.247-250 [8] P.C. Chao, M.Y. Kao, K. Nordheden, and A. Swanson, “HEMT Degradation in Hydrogen Gas”; IEEE Electron Device Letters, Vol 15, No.5. pp. 151153, May 1994. [9] G. Kelly, M. Cobb, D. Weir, M. Welch, and M. Weig, “The Accelerated Effects of hydrogen Gas on 0.15 um Pseudomorphic HEMTs”; 1995 GaAs Reliability Workshop. [10] D. C. Eng, R.J. Culbrtson, K.P MacWilliams "The effects of hydrogen and deuterium inccrporation on the electrical performance of a GaAs MESFET", GaAs IC Symposium, 17th Annual Technical Digest, 1995 pp. 140-143. [11] K. Decker, "GaAs PHEMT Hydrogen Sensitivity Study", GaAs Reliability Workshop Digest, 1996. [12] Y. Saito R Griese, J Kessler, R. Kono, J Fang "Hydrogen Degradation of GaAs MMICs and Hydrogen Evolution in the Hermetic Package" Microwave and Millimeter-wave Monolithic Circuits Symposium Digest, 1995, p. 119-122. [13] Shason Microwave Corporation, “Hydrogen Effects on GaAs Microwave Semiconductors”, Report # SMC97-0701, 1997. [14] Sammy Kayali, "Hydrogen Effects on GaAs Device Reliabilty", Gallium Arsenide Manufacturing Technology Conference, 1996.