IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY

Transcription

1 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY A Comparison of Hourly Versus Daily Testing Methods for Evaluating the Reliability of Water Soluble Fluxes W. Jud Ready and Laura J. Turbini, Senior Member, IEEE Abstract This paper presents background on surface insulation resistance (SIR) testing and reports on its application to six water-soluble fluxes, and two hot-air solder leveling (HASL) fluids. This work evaluates two electrical reliability-testing methods. One method recorded SIR readings on a daily basis, the other on an hourly basis. During the twenty-eight day test, increased frequency of testing was more successful in detecting discrete electrical events, such as surface dendrites but not subsurface conductive anodic filaments (CAF). These failure mechanisms are sporadic in nature as they initiate, grow, short out and disintegrate, reform, then short out again. Unless a measurement is taken during the time when shorting is about to occur, they will not be detected electrically. Surface dendrites due to water condensation were easily detected with the hourly SIR method but not with the daily measurements. Unfortunately, neither the daily nor hourly SIR methods detected the presence of conductive anodic filaments (CAF). These results demonstrate the importance of visual examination of circuit assemblies after environmental testing, as well as highlighting the need for a more sensitive electrical reliability test method. Index Terms Conductive anodic filament (CAF), dendrites, electrochemical migration, hot-air solder leveling (HASL) fluids, surface insulation resistance (SIR), water-soluble flux. I. BACKGROUND A. Surface Insulation Resistance Testing ACOMMON test for evaluating printed wiring board (PWB) reliability is the surface insulation resistance (SIR) test. When it is done using a bias voltage, it is also an electrochemical migration (ECM) test [1] [3]. Some ECM testing is done at low voltage (i.e., 5 V) whereas SIR testing uses V bias. If the SIR and ECM tests are carried through to failure, they are accelerated life tests (ALT). The conditions for ALT typically consist of elevated temperature and humidity (i.e., 85 C/85%RH [4], 35 C/85%RH minimum [5], etc.) and a bias voltage. SIR tests provide important data used to characterize PWB laminates, soldering fluxes, solder masks and conformal coatings on PWB s. They are used to study the effect of aging a Manuscript received February 22, 1999; revised December 9, This work was supported by the United States Army Missile Command, Contracts DAAH01-92-D-R and DAAH01-92-D-R , Concoat, Ltd. Surrey, U.K., and Alpha Metals. W. J. Ready is with the School of Materials Science and Engineering, Georgia Institute of Technology, Atlanta, GA USA ( jud.ready@mse.gatech.edu). L. J. Turbini is an International Reliability Consultant, 31 28th Street NW #1, Atlanta, GA USA ( ljturbini@mindspring.com). Publisher Item Identifier S (00) Fig. 1. IPC-B-24 test coupon. sample at accelerating conditions to determine if there are any detrimental effects to the electrical properties. Although the insulation resistance readings are a combination of both bulk and surface resistance, for the case of FR-4, 99.9% of the current leakage will occur across the surface of the laminate [6]. Most SIR testing follows a standard test method and is used as a short-term test of the integrity of a material or process. Thus, it is usually performed as an accelerated aging test, but not an accelerated life test because it does not test to failure. SIR testing is performed in order to accelerate normal failure modes through the use of elevated stress conditions. Joint Industry Standard (J-STD-004) Requirements for Soldering Fluxes [4] specifies that a SIR test be performed in accordance with IPC-TM-650, TM In this test method, processed interdigitated comb patterns such as the IPC-B-24 (Fig. 1) are placed under a bias voltage of 145 V to V. Periodically (at 24, 96, and 168 h) the bias is removed and the resistance between the comb fingers is measured under a 100 V test voltage. This test is performed while coupons are in an environmental chamber at C C, and %RH %RH. Following the test, the coupons are inspected under a microscope for signs of degradation. SIR electrical data along with microscopic examination of comb patterns after testing provide important information on the flux/substrate interactions [4]. SIR is an extrinsic property of the material system under investigation. SIR results will be affected by the test pattern chosen, the temperature, humidity and bias (THB) conditions and duration of test, as well as the contamination associated with previous processing steps. A potential failure mode during SIR tests is electrochemical migration, which causes dendritic growth, due to the presence of localized condensation on the PWB surface associated with flux /00$ IEEE

2 286 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY 2000 TABLE I TEST FLUX PROPERTIES TAKEN FROM MATERIALS SAFETY DATA SHEETS residues or water spots caused by poor humidity control in the chamber. A less frequently encountered failure mode is conductive anodic filament (CAF) formation due to mechanisms that have been published [7] [9]. B. Electochemical Migration In most SIR testing the comb pattern is placed under a bias voltage. Periodically this voltage is removed, a test voltage is applied and the resistance between traces is measured using an electrometer. Under both voltage conditions, electrochemical migration may be induced. However, it is only detectable electrically during the comparatively brief measurement phase of the SIR test. Electrochemical corrosion of metallic conductors, the migration of metal ions between anode and cathode and reduction of these ions to metallic dendrites at the cathode can lead to circuit failure. It is important to understand the cause of these failures in order to select materials and processes for soldering and cleaning which will minimize the occurrence of these failures. For electrochemical migration to occur, a pathway must exist for ions to move from the anode to the cathode. Surface moisture can provide that pathway. This may come from the high humidity used in testing, or from water condensation. Certain soldering flux residues can cause localized condensation when the critical relative humidity is exceeded due to the nature of the contaminant [10]. Polyglycols and other nonionic surfactants which absorb into the PWB during the soldering process can also cause increased moisture adsorption by the PWB causing reduced SIR levels [11] [13]. In the presence of moisture, the following electrochemical reactions can occur at the anode: H O O H (1) Cu Cu (2) Cu Cu (3) Sn Sn (4) Sn Sn (5) Pb Pb (6) The preferred species of copper will depend upon the anion that is present. In water, Cu is the preferred species except when Cl is present. In this case the formation of the CuCl will favor [14] Cu rather than Cu. In aqueous solutions, Sn is the preferred species. At the cathode, the following reactions are possible: O H O OH (7) H O OH H (8) Cu Cu (9) Cu Cu (10) Sn Sn (11) Sn Sn (12) Pb Pb (13) For electrochemical dendritic growth, electrolytic dissolution of the metal occurs at the anode and reduction of the metal ions occurs at the cathode. As surface dendrites grow between

3 READY AND TURBINI: RELIABILITY OF WATER SOLUBLE FLUXES 287 Fig. 2. Thermal profile experienced by comb patterns in laboratory reflow oven. cathode and anode, their effect on the total SIR reading is minimal until they are very close to the anode. At the point of bridging, the dendrite will burn out quickly due to the high current that abruptly flows through the fine and fragile dendrite. The presence of dendrites is not easily determined by the electrical SIR readings. The readings required by J-STD-004 [4] are not taken frequently enough (three times in seven days) to insure that a measurement will be made exactly when the dendrite bridges. Thus, it is necessary to examine SIR samples under the microscope after the test is terminated to observe visually if dendrites have formed. C. CAF In the mid 1970's a new failure mode was observed as a result of increased wiring density and hostile environments in which circuits were required to perform. This failure mode is characterized by an abrupt, unpredictable loss of SIR between conductors that are held at a potential difference [7]. The loss of resistance is caused by the growth of a subsurface filament from anode to cathode between the two conductors. The filament, termed conductive anodic filament (CAF), is a result of an electrochemical corrosion process that initiates at the anode and proceeds along separated fiber/epoxy interfaces [8], [9], [15], [16] CAF is easily differentiated from dendrite growth. In dendritic growth, metal ions go into solution at the anode, but plate out at the cathode, growing in tree-like dendrites across the surface of the PWB. In contrast, CAF growth emanates from the anode. Furthermore, the filament is a copper salt containing chloride or bromide ions in addition to copper ions rather than pure metal as in the case of dendrites. Additionally, dendrite growth is a surface phenomenon, while CAF is a subsurface phenomenon. TABLE II FINAL SIR VALUES (GEOMETRIC MEAN &STANDARD DEVIATION) SIR testing is used to measure the electrical properties of a processed comb pattern and to determine if CAF or electrochemical migration has occurred during the test period. The objective of this study is to evaluate two SIR test methods. One method recorded SIR data on a daily basis, the other on an hourly basis. Eight different water-soluble flux formulations were included in the study along with control coupons. II. EXPERIMENTAL PROCEDURE A. Test Coupons Industry standard, IPC-B-24 test coupons (Fig. 1) were used throughout this work. The dimensions of the IPC-B-24 are 4.5 in 4.0 in in. Each coupon contains four comb patterns with 35 interdigitated 0.4 mm wide copper lines with 0.5 mm spacing between the lines. The coupons were made of FR-4. Some samples had bare copper, while others had tin-lead soldercoated copper metallization. A model 290 Dremel engraver was

4 288 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY 2000 (a) (b) Fig. 3. (a) Typical SIR plot for samples processed with Flux 1 in the manufacturing environment, values measured once daily for 28 days. (b) CAF that formed to right of solder plated copper track was not detected electrically. used to etch a serial number into the identification box in the upper left of each test coupon. All test coupons were precleaned in a Zero Ion System containing a mixture of 75% isopropyl alcohol and 25% deionized water. Following this, they were handled using latex gloves to minimize contamination. B. Flux A variety of water-soluble fluxes were used in this study. The majority of the fluxes were applied to the comb pattern side of the test coupon surface using a spray bottle. Approximately ten milliliters of flux was required to completely cover the comb pattern side of the test coupon. For the more viscous HASL fluxes, the flux was spread across the board using a plastic applicator. For both of these methods, the excess flux was allowed to run off the board into a suitable waste container. The physical properties of the various test fluxes are listed in Table I for comparison. Control coupons for this study consisted of IPC-B-24 test coupons that underwent the thermal processing steps without flux application or post-process cleaning. Some of the test coupons had eutectic tin/lead applied by the board manufacturer using a hot air solder leveling (HASL 1) fluid. The physical properties of this HASL fluid are presented in Table I for comparison with the other test fluxes. Flux 4 is also a HASL fluid. C. Manufacturing Wave Soldering and Cleaning Procedure After fluxing, the boards were placed on the conveyor which moved them across the preheater which was set at (nominally) 100 C on both top and bottom board surfaces. The solder wave consisted of a dual-wave system with an initial turbulent wave, followed by a secondary lambda wave. The molten solder was held at (nominally) 250 C. The boards moved at three feet per minute along the 14 foot Electrovert wave soldering machine. After soldering, the coupons were then cleaned using an in-line Electrovert Clean-Aquapak 526 cleaner. There were three zones of cleaning: 1) a vigorous 70 C water wash followed by 2) a wash with Loncoterge 560 (3.5 vol%), followed by 3) another vigorous 70 C water rinse. The PWB s were forced-air dried at 80 C.

5 READY AND TURBINI: RELIABILITY OF WATER SOLUBLE FLUXES 289 (a) (b) Fig. 4. (a) Typical SIR plot for samples processed with Flux 5 in laboratory environment, values measured hourly for 28 days. (b) CAF that formed at tip of copper track was not detected electrically. D. Laboratory Reflow and Cleaning Procedure A bench-top OK Industries model JEM-310N convection reflow oven was used to process additional samples in the laboratory. The thermal profile experienced by the test coupons is given in Fig. 2. The cleaning procedure consisted of a vigorous tap water rinse (at room temperature) for one minute. This was followed by a five minute ultrasonic cleaning in a Branson 5210 ultrasonic bath with deionized water at 65 C. A final deionized water rinse completed the cleaning process. E. Accelerated Testing To simulate aging of the sample, the test coupons were placed under a bias voltage in an accelerating atmosphere of 85 C/85%RH (Model SM-8C Thermotron oven) for twenty-eight days. SIR measurements were taken using two types of electrical test equipment. Initially, tests were conducted using an Alpha Metals Model 200 SIRometer with bias and test voltages of 100 VDC. The 100 V was chosen over the IPC specified 50 V in order to evaluate the potential for CAF formation. Every 24 h, SIR electrical measurements were recorded automatically during a 28 day test. To examine the effect of an increased sampling rate a second SIR measuring system was used. Unlike other instruments, the Concoat AutoSIR has the capability to make SIR readings on a nearly continuous basis. (The only factor that prevents continuous readings is the time required to record measurements on the instrument's 128 channels approximately 45 s.) The unit was set to take hourly readings for 28 days. The bias and test voltages were both 100 VDC. III. RESULTS A. Manufacturing Scale Wave Soldering Table II presents the final SIR values for the test coupons processed with the various test fluxes. The value is a geometric average of the SIR on day twenty-eight for each of the comb patterns. The geometric average is much less affected by extreme values of individual data points, which are characteristic of SIR measurements. Fig. 3(a) represents a typical SIR plot for the manufacturing scale coupons. These SIR readings were taken on a daily basis. Although there is no indication from the electrical readings, all comb patterns for Flux 1 [Fig. 3(a)] exhibited

6 290 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY 2000 (a) (b) Fig. 5. (a) SIR plot for sample processed with Flux 4 in laboratory environment, values measured hourly for 28 days showing vertical spikes indicative of condensation and splashes. (b) The resulting water-spot dendrite growth. CAF formation [Fig. 3(b)]. This did not show up in the electrical readings but was identified later when the coupons were examined under the microscope with back lighting. The dark shadow to the right of the metal trace is CAF. B. Laboratory Scale Reflow Soldering Table II also presents the final SIR values for the various fluxes tested with the laboratory reflow set up. Again, the final value is a geometric average of the SIR for the several comb patterns on day twenty-eight. Fig. 4(a) shows a typical SIR plot for samples processed in the laboratory. All comb patterns for Flux 4 (Fig. 4) exhibited substantial quantities of CAF formation [Fig. 4(b)] even though the electrical readings did not reveal their presence. C. Hourly SIR Readings An SIR test using hourly data collection was able to detect discrete events that daily data collection did not. This is illustrated in Fig. 5(a) where condensation of moisture and splashes of water drops on comb patterns created distinctive vertical spikes. The sensitivity of the instrument to these anomalies demonstrated the value of the hourly readings. The condensation occurred in an early experimental set-up where water droplets collected in the connector between the test coupons and the wiring to the test equipment. An experimental problem caused water condensing on the chamber window, which splashed onto some coupons, inducing water-spot dendrites. They were detected electronically when hourly readings were taken, and the water-spot dendrites were also observed optically

7 READY AND TURBINI: RELIABILITY OF WATER SOLUBLE FLUXES 291 Fig. 6. Typical SIR plot for a copper control-coupon processed in the laboratory environment, values measured hourly for 28 days. [Fig. 5(b)]. For comparison, Fig. 6 shows the hourly SIR data for a copper control coupon without condensation. IV. DISCUSSION In this experiment, twenty-eight day SIR values were measured on two time scales: hourly and daily. The daily measurements were unable to detect CAF growth [Fig. 3(a) and (b)]. More than a dozen CAF formed on the comb patterns of the boards whose SIR values are depicted in Fig. 3(a). However, there are no electrical indications that catastrophic growth of CAF [Fig. 4(b)] was occurring. By taking hourly readings, discrete events such as water splashes and condensation can be captured electrically [Fig. 5(a) and (b)]. However, CAF formation and growth was not observed electrically with hourly readings [Fig. 4(a) and (b)]. Several important comparisons between the SIR data should be noted. The first is the comparison between the SIR values of bare copper boards prepared in the manufacturing environment versus the boards prepared in the laboratory environment. For all of the samples tested except Flux 5, the final SIR of the laboratory-processed samples was approximately one order of magnitude greater than that of the manufacturing environment samples. The most probable causes for this difference are 1) the dual solder waves cause greater thermal shock and penetration of flux residues into the PWB than the reflow process and 2) the manufacturing cleaning process may not have been as effective as that of the laboratory ultrasonic cleaning process. Another comparison that should be made is between the laboratory processed bare copper samples and the laboratory processed solder-coated samples. The solder-coated samples had a final SIR value that was on average more than one half an order of magnitude lower than the bare copper. Since the thermal profiles and cleaning processes were identical, it is proposed that the difference is likely due to the HASL fluid used by the board manufacturer. It should be noted that, regardless of the differences in SIR measurements, the final SIR values for the majority of boards were well within the specifications set forth in ANSI-J-STD-004 [4]. Despite this fact, numerous CAF s were observed optically on PWB s that would have otherwise passed this reliability standard. This points to the conclusion that SIR values alone are not an acceptable criterion of electrical reliability. The SIR value represents a distributed resistance and the presence of dendrites will not be observed unless an electrical reading is taken shortly before bridging occurs. In addition, SIR is greatly influenced by the current leakage that occurs across the surface of the PWB [6]. Surface leakage is typically in the form of dendrite formation, which was easily detected with hourly SIR measurements [Fig. 4(a) and (b)]. Subsurface failures such as CAF are less likely to be detected. V. CONCLUSION The water-soluble fluxes tested in this study exceeded the electrical requirement of per J-STD-004. However, dendrites and subsurface CAF occurred which were not always identified by the electrical readings. Thus, microscopic examination of the comb patterns was necessary. While daily SIR readings were not sufficient to detect surface dendrites, the use of hourly measurements improved the sensitivity of the test and proved successful in detecting these discrete electrical events. Neither daily nor hourly SIR measurements were successful in detecting CAF growth in situ. An electrical measurement technique that is sensitive to subsurface filament growth should be designed and implemented. Such a circuit should allow for continuous in situ monitoring of CAF growth. The measurement technique should also provide

8 292 IEEE TRANSACTIONS ON ADVANCED PACKAGING, VOL. 23, NO. 2, MAY 2000 for a prompt termination of bias when bridging occurs, so that the filament or dendrite does not blow out. ACKNOWLEDGMENT The authors wish to express their sincere gratitude to R. Vaughn and B. Kervin, Rockwell International, for the use of the manufacturing equipment. Additional thanks go to G. K. Naisbitt and D. Hague, Concoat, and G. Miller, Alpha Metals, for their technical consultations and to N. Watts, Merix Corporation, B. Palmer and B. Patti, Terradyne Corporation. They also wish to thank A. Schneider, Alpha Metals, for the contribution of soldering fluxes used in this present work. REFERENCES [1] B. N. Ellis, The correlation between short- and long-term SIR testing, Circuit World, vol. 22, no. 2, pp , Jan [2] M. E. Jozefowicz and N.-C. Lee, Flux reliability assessment: Electromigration vs SIR, in Proc. NEPCON W. 1993, [3], Electromigration vs SIR, Proc. SPIE, pp , [4] Requirements for Soldering Fluxes, American National Standard, Joint Industry Standard, ANSI/J-STD-004, Jan [5] Generic Physical Design Requirements for Physical Design and Manufacture of Telecommunications Products and Equipment, Bellcore Tech. Ref. GR-78-CORE, Sept [6] E. J. Gorondy, Surface Insulation Resistance Part I: The Development of an Automated SIR Measurement Technique,, IPC-TP-518. available from the IPC, 2215 Sanders Rd, Northbrook, IL [7] T. L. Welsher, T. L. Mitchell, and D. J. Lando, Conductive anodic filaments (CAF): An electrochemical failure mechanism of reinforced polymeric dielectrics, in Proc. Annu. Rep. Conf. Elect. Insulation Dielectric Phenom., 1980, pp [8] D. J. Lando, J. P. Mitchell, and T. L. Welsher, Conductive anodic filaments in reinforced polymeric dielectrics: Formation and prevention, in Proc. 17th Annu. Rel. Phys., 1979, pp [9] J. N. Lahti, R. H. Delaney, and J. N. Hines, The characteristic wearout process in epoxy-glass printed circuits for high density electronic packaging, in Proc. 17th Annu. Rel. Phys., 1979, pp [10] L. J. Turbini, Reliability Evaluation of Alternatives to Ozone Depleting Substances in DoD Manufacturing, Final Rep. U.S. Army Missile Command, DAAH01 92 D R005, D.O. 0013, Nov. 30, [11] F. M. Zado, Effects of nonionic water soluble flux residues, Western Electric Eng., pp , [12] J. Brous, Water soluble flux and its effect on PC board insulation resistance, Electron. Packag. Prod., pp , July [13], Electrochemical migration and fluxresidues Causes and detection, in Proc. NEPCON W., vol. 2, 1999, pp [14] H. H. Ulig, Corrosion and Corrosion Control, 2nd ed. New York: Wiley, 1971, p [15] W. J. Ready, L. J. Turbini, S. R. Stock, and B. A. Smith, Conductive anodic filament enhancement in the presence of a polyglycol-containing flux, in Proc IEEE Int. Rel. Phys., 1996, pp [16] J. DeLockery, Continuous lamination an emerging technology, in Proc. Printed Circuit World Conf. 5, 1990, pp. B6/1 B6/3. W. Jud Ready received the B.S. degree in materials engineering and the M.S. degree in metallurgical engineering from the Georgia Institute of Technology, Atlanta, in 1994 and 1997, respectivley, where he is currently pursuing the Ph.D. degree. His research is focused on developing and improving electrical test methods used for assessing printed wiring board reliability. Laura J. Turbini (A 82 SM 98) received the M.S. and Ph.D. degrees in inorganic chemistry from Cornell University, Ithaca, NY. She is presently an International Reliability Consultant specializing in processing issues in electronic assembly. She joined Western Electric Engineering Research Center (now Lucent Technologies Bell Labs) in 1977 and became the Research Leader in Soldering in She served as Editor of the Western Electric Engineer from 1982 to 1985, at which time she became an Engineering Manager at the AT&T Manufacturing Plant, Denver, CO. She joined the Georgia Institute of Technology, Atlanta, in 1989 as the Associate Director of the Manufacturing Research Center and later became a Professor of materials science and engineering.