Implementing a Simple Corrosion Test Method for Early Detection of “Black Pad” Phenomenon In ENIG Plating BabHui Lee 11th Dec 2003 HKPCA-IPC Conference
Introduction <ul><li>Forms during ENIG </li></ul><ul><li>plating, but manifests </li></ul><ul><li>itself at the board </li></ul><ul><li>assembly stage </li></ul><ul><li>Brittle, interfacial </li></ul><ul><li>solder joint failure </li></ul><ul><li>Selective and occurs </li></ul><ul><li>at low ppm level </li></ul><ul><li>Can not be predicted </li></ul><ul><li>or detected before </li></ul><ul><li>assembly, causing </li></ul><ul><li>catastrophic failures </li></ul>What’s “Black Pad”?
Introduction What’s “Black Pad”? On bare board level… Electroless nickel pad surface appears “black” after stripping off immersion gold layer under visual inspection
Introduction What’s “Black Pad”? On bare board level… Electroless nickel pad surface exhibiting “mud-crack” signature after stripping off immersion gold layer under SEM, top scan (6,000X)
Introduction What’s “Black Pad”? On assembly level… Brittle solder joint showing separation Pin lifted from SMT pad Solder joint failure Solder joint failure
Introduction What’s “Black Pad”? On assembly level… Pin side (SEM 6,000X) Pad side (SEM 6,000X) Nickel-like nodular interfaces with “mud cracks” are exposed
Introduction What’s “Black Pad”? Cross-Section/SEM Large regions of severe black pad with spikes protruding into nickel layer
Introduction Upon solder reflow, interconnects are wetted and solder joints appear in normal form What’s “Black Pad”? Cross-Section/SEM
Introduction Failure occurs after reflow, resulting in open joints What’s “Black Pad”? Cross-Section/SEM
Introduction <ul><li>The root cause is not completely understood, but the following conclusions can be made: </li></ul><ul><li>The Black Pad defect is the result of an interaction </li></ul><ul><li>between process control and other factors, such </li></ul><ul><li>as circuit board design (ITRI) </li></ul><ul><li>It results in corrosion of the Ni layer during </li></ul><ul><li>accelerated Au deposition in the immersion </li></ul><ul><li>process (N. Biunno, ITRI) </li></ul><ul><li>Ni corrosion occurs under galvanic cell conditions </li></ul><ul><li>(N. Biunno, P. Snugovsky, K. Johal) </li></ul>What causes “Black Pad”?
Introduction The Ni layer is more susceptible to corrosion if • It is thin, < 120 µ” • P < 6 wt % • High level of micro- structure defects such as nodule layer grain boundaries What causes “Black Pad”?
Background <ul><li>Not all PCB shops have the luxury of SEM/EDX available </li></ul><ul><li>for process control & troubleshooting </li></ul><ul><li>Often, this is an after-the-fact, ie. when “black pad” occurs, </li></ul><ul><li>it’s already too late to take preventive actions </li></ul><ul><li>Cross-section preparation is both time-consuming & costly </li></ul><ul><li>Cyanide-stripping of Au layer can post health hazard if not </li></ul><ul><li>performed under controlled environment </li></ul>Constraints on current common “Black Pad” detecting methods:
Objective To devise & implement an optimum acid corrosion test methodology for early detection of “black pad” phenomenon prevalent in the PCB industry pertaining to the use of ENIG. The preferred method should have the following characteristics: <ul><li>Simple & yet sensitive, a quick “Acid-Test” </li></ul><ul><li>Repeatable (least variation & reliance on operator) </li></ul><ul><li>Predictable & with strong correlation to failure mode </li></ul><ul><li>Serves as an in-process control checkpoint </li></ul>
Test Methodology <ul><li>Nickel Strips </li></ul><ul><li>20mm x 60mm, 0.2mm thick </li></ul><ul><li>Cut from 18”x24” EN-plated dummy </li></ul><ul><li>panel per Ni bath MTO </li></ul><ul><li>Ni Bath MTO: 0~5 </li></ul><ul><li>Ni thickness: 3~4 µm </li></ul><ul><li>Nitric Acid </li></ul><ul><li>Conc: ~40% (v/v) </li></ul><ul><li>Volume: ~60 ml in 100-ml </li></ul><ul><li>glass beaker </li></ul><ul><li>Temperature: 24~25 o C </li></ul><ul><li>Single use per Ni strip </li></ul><ul><li>Test Procedure </li></ul><ul><li>18 strips per MTO were used </li></ul><ul><li>Each strip was immersed in the beaker of nitric acid as illustrated </li></ul><ul><li>Time (sec) for the strip to turn “black” was recorded using a stopwatch </li></ul><ul><li>Note: The hold time for freshly prepared Ni strips should not exceed 12-hour before testing </li></ul>20 ml 40 ml 60 ml 80 ml 100 ml 2.0 cm 6.0 cm 2.0 cm
Test Results – MTO vs. HNO 3 Time From the box plots, there is a clear influence of Ni MTO on the time to fail (ie. for the nickel strip sample to turn “black”), esp. at 5 MTO – this is affirmed with one-way ANOVA analysis with Tukey’s pairwise comparisons. 25 34 33 25 32 33 24 35 34 22 35 36 24 32 26 22 41 44 25 33 33 23 33 28 24 36 37 27 33 32 25 32 37 23 25 43 24 28 36 28 34 35 27 30 44 23 33 31 24 27 38 27 25 39 5.0 MTO 2.5 MTO 0 MTO
Statistically, 5 MTO samples have lower mean time to failure than 0 & 2.5 MTOs (around 10 sec less). The general trend is that the acid corrosion resistance of the as-plated EN surface reduces as Ni MTO rises. Repeatability is also better at 5 MTO from the obvious lower standard deviation attained. This may be attributed to a more steady state reached towards the end of the useful electroless nickel bath life (just an attempted theorized explanation). One-way ANOVA: Time (sec) versus Ni MTO Analysis of Variance for Time (sec) Source DF SS MS F P Ni MTO 2 1130.1 565.1 38.59 0.000 Error 51 746.7 14.6 Total 53 1876.8 Individual 95% CIs For Mean Based on Pooled StDev Level N Mean StDev ----+---------+---------+-------- 0 MTO 18 35.500 4.997 (----*---) 2.5 MTO 18 32.111 3.984 (---*----) 5 MTO 18 24.556 1.756 (---*----) ----+---------+---------+---------+---------+ Pooled StDev = 3.826 24.0 28.0 32.0 36.0 Test Results – MTO vs. HNO 3 Time
Test Results – Reliability Plots As expected, 5 MTO has the most hazardous function, and all 5 MTO samples are expected to fail within 30 sec.
More than 90% of the samples (across the full Ni MTO range) should be able to survive the 40% nitric acid test for at least 20 sec. Test Results – Reliability Plots
SEM Scan/EDX Analysis (%P) For each Ni MTO (0, 2.5 & 5) and at an interval of 5 sec (up to 30 sec) of 40% nitric acid dip test, the following were obtained: These shall be used (in parallel with the predicted reliability data) as a basis for the acceptance level of the degree of nickel corrosion subjected to nitric acid test, and as a quick indication of the corrosion resistance of the as-plated electroless nickel surface under normal production conditions. <ul><li>SEM Scan at 2,500X – Surface topography depicting Ni grain boundary structure and any sign of nickel attack / corrosion. </li></ul><ul><li>EDX for surfcae %P content to show any correlation with Ni bath MTO and resulting attack from nitric acid dip test. </li></ul>
SEM/EDX – As Is (Before Test) 0 MTO, 9.41%P 2.5 MTO, 7.81%P 5 MTO, 9.46%P Clean nickel surfaces
SEM/EDX – 10 sec HNO 3 Dip 0 MTO, 7.82%P 2.5 MTO, 9.66%P 5 MTO, 8.05%P Signs of minor attack at 5 MTO
SEM/EDX – 15 sec HNO 3 Dip 0 MTO, 8.46%P 2.5 MTO, 9.81%P 5 MTO, 10.42%P Signs of slight attack at 5 MTO
SEM/EDX – 20 sec HNO 3 Dip 0 MTO, 7.87%P 2.5 MTO, 9.36%P 5 MTO, 10.36%P Signs of some attack at 5 MTO
SEM/EDX – 25 sec HNO 3 Dip 0 MTO, 8.28%P 2.5 MTO, 10.64%P 5 MTO, 8.59%P Signs of heavier attack at 5 MTO
SEM/EDX – 30 sec HNO 3 Dip 0 MTO, 10.38%P 2.5 MTO, 11.94%P 5 MTO, 10.20%P Signs of heavier attack at 5 MTO
Test Results - %P vs. MTO, HNO 3 Statistically, the Ni MTOs & nitric acid dip time do not seem to affect %P content significantly in the Ni deposits within the ranges under test. However, caution should be exercised as residual analysis only indicates some reasonable fit. General Linear Model: %P versus Ni MTO, HNO 3 Time Factor Type Levels Values Ni MTO fixed 3 0.0 2.5 5.0 HNO3 Tim fixed 7 0 5 10 15 20 25 30 Analysis of Variance for %P, using Adjusted SS for Tests Source DF Seq SS Adj SS Adj MS F P Ni MTO 2 3.9341 3.9341 1.9670 2.26 0.146 HNO3 Time 6 10.5961 10.5961 1.7660 2.03 0.139 Error 12 10.4230 10.4230 0.8686 Total 20 24.9531
Nonetheless, from the main effects & interaction plots, it’s discernible that prolonged nitric acid at 30 sec tends to induce phosphorus enrichment at the attacked nickel surface (which is not difficult to predict by intuition) – this is especially pronounced at 2.5 Ni MTO. Test Results - %P vs. MTO, HNO 3
Conclusion & Summary <ul><li>At different nickel bath MTOs (Metal Turnover), the extent of nickel corrosion varies as the exposure time to nitric acid increases – this is especially pronounced towards the end of bath life at 5 MTO, where the attack is more severe and the time to failure (complete attack, manifested by “blackened” surface almost spontaneously) is significantly lower than fresher EN baths. </li></ul><ul><li>As contrary to the general belief, the %P variation of the nickel deposits as the nickel bath ages is not significant, and %P alone cannot explain why the nickel surface is more susceptible to attack as the MTO rises. There could be some interaction effects from the bath contaminants build-up and the balance of other organic/inorganic additives, stabilizers &/or complexors, etc. which may all contribute to the observed phenomenon. This could be subject for future study/research. </li></ul><ul><li>A quick & easy nitric acid test methodology for assessing the corrosion resistance nature of as-plated electroless nickel surface can be practically implemented as an additional monitoring & control item for early detection & prevention of “black pad” that has serious deteriorating impact and consequences on solderability & reliability performances. </li></ul>
Recommendation From the standpoint of reliability performance, a minimum nitric acid (40% v/v) withstanding timing of 20 sec can be stipulated based on the highest Ni MTO at 5 (the normal useful life of the electroless nickel bath), and the acceptable nickel surface topography (degree of corrosion) & %P content after 20 sec exposure to the nitric acid attack. This will depict an average failure rate probability of at most 3% (from the reliability probability plot for the worst case scenario, ie. at 5 Ni MTO). Sample size determination, frequency of test and the acceptance judgement are proposed on the following slides.
Power and Sample Size One-way ANOVA Sigma = 3.826 Alpha = 0.05 Number of Levels = 3 Sample Target Actual Maximum SS Means Size Power Power Difference 12.5 13 0.8000 0.8235 5 12.5 14 0.8500 0.8545 5 12.5 16 0.9000 0.9027 5 18.0 9 0.8000 0.8046 6 18.0 10 0.8500 0.8519 6 18.0 12 0.9000 0.9175 6 24.5 7 0.8000 0.8101 7 24.5 8 0.8500 0.8704 7 24.5 9 0.9000 0.9133 7 32.0 6 0.8000 0.8407 8 32.0 7 0.8500 0.9051 8 32.0 7 0.9000 0.9051 8 40.5 5 0.8000 0.8403 9 40.5 6 0.8500 0.9178 9 40.5 6 0.9000 0.9178 9 50.0 5 0.8000 0.9103 10 50.0 5 0.8500 0.9103 10 50.0 5 0.9000 0.9103 10 Sample Size Determination A test of power was conducted based on the original test results, and it’s gathered that a sample size of 10 should suffice for a power target of 85% with detectability of maximum difference of 6 sec (which makes practical sense too, considering the variation within the samples). Hence it’s recommended that the sample size to be consisted of 10 freshly prepared Ni strips per test (in accordance to the test methodology as outlined in the beginning).
Frequency of Test A test frequency is proposed at once/week for a start – this also serves to supplement the current SEM/EDX analyses that have been put in place on a monthly basis. Based on the past historical data collected over a prolonged period of time, our current tight process control on ENIG bath does not warrant a more stringent & frequent check of this additional control item to be implemented soon. As the sample is easy to prepare, and the test itself is quick and simple, once/week test frequency will not incur too much burden on the current workload. It’s also suggested that Chem Lab shall schedule & conduct this test weekly. Of course, for troubleshooting purpose, whenever there is doubt cast on poor Ni bath performance or suspected “black pad” issue, this test should be carried out immediately.
Acceptance Criteria As mentioned earlier, a minimum of 20 sec nitric acid resistance time is proposed based on the test results and the associated reliability probability plots. The nitric acid resistance time is defined as the time taken for the nickel strip sample exposed to the 40% nitric acid to turn “black” completely under the stipulated test conditions. For a sample size of 10, the acceptance judgement is based on the fact that the min. average acid resistance of the 10 samples shall exceed 20 sec, with at most one strip below 20 sec , correlating to a failure rate of 10% over the full range of the nickel bath conditions over its useful bath life – we can also infer from the probability plots that more than 90% of the samples (across the full Ni MTO range) should be able to survive the 40% nitric acid test for at least 20 sec.
Acknowledgement <ul><li>Special thanks & acknowledgement are credited to the following persons: </li></ul><ul><li>Choice Lee (B5 PE) for preparing the samples & conducting the tests, after numerous excruciating trials & errors. </li></ul><ul><li>John Ke & Bill Slough (AMD Lab) for churning out all the SEM images & EDX analyses under extreme time constraints. </li></ul><ul><li>Derris Chew (B3 QA) for some advice in statistical analysis. </li></ul><ul><li>Jim Poon (Corporate VP,QA) for propounding & supporting this initiative. </li></ul><ul><li>All others who help in one way or another. </li></ul>