Testing Printed Circuit Boards for Creep Corrosion in Flowers of Sulfur Chamber

The iNEMI technical subcommittee on creep corrosion is developing a flowers-of-sulfur (FOS) based qualification test for creep corrosion on printed-circuit boards (PCBs). The test setup consists of a 300-mm cube chamber with two means of mounting the test specimens and flowing air over them to expose them to constant, predefined humidity and temperature conditions and sulfur and other contaminants. The FOS chamber performance has been evaluated using copper and silver foils and preliminary test runs have been conducted on PCBs from a manufacturing lot known to have failed in service. The effect of air velocity on the copper and silver corrosion rates was quite linear. The effect of humidity on copper and silver corrosion rates in the low air velocity range of less than 0.1 m/s showed a strong dependence on relative humidity. In the high velocity range of 1 m/s, there was no clear dependence of humidity on copper and silver corrosion rates. A means has been developed for applying controlled concentration of ionic contamination on selected local areas of test PCBs. Preliminary test runs have shown that ionic contamination found in fine dust may be a necessary condition for copper creep corrosion. Printed circuit boards from a manufacturing lot that suffered creep corrosion in service, with and without dust contamination applied to them, were tested in a FOS chamber at 60C with 1 m/s air flowing over them. The PCBs with no dust contamination did not suffer creep corrosion in the 3-day test; whereas, the PCBs with dust contamination suffered creep corrosion with morphology similar to that occurring in the field. Introduction Qualification of electronic hardware from a corrosion resistance standpoint has traditionally relied on stressing the hardware in a variety of environments. Before the development of tests based on mixed flowing gas (MFG), hardware was typically exposed to temperature-humidity cycling. In the pre-1980s era, component feature sizes were relatively large. Corrosion, while it did occur, did not in general degrade reliability. There were rare instances of the data center environments releasing corrosive gases and corroding hardware. One that got a lot of publicity was the corrosion by sulfur-bearing gases given off by data center carpeting. More often, corrosion was due to corrosive flux residues left on as-manufactured printed circuit boards (PCBs) that led to ion migration induced electrical shorting. Ion migration induced failures also occurred inside the PCBs due to poor laminate quality and moisture trapped in the laminate layers. The two common modes of environmentally induced corrosion related failures electronic hardware is currently suffering are the corrosion of silver terminated surface mount resistors and the creep corrosion on printed circuit boards (PCBs). The corrosion of silver terminations in surface mount resistors has been brought largely under control by improving the packaging of the resistors and by developing a flowers of sulfur test consisting of exposing the resistors to dry flowers of sulfur in a chamber at 105C for 20 days. The flowers of sulfur chamber is kept dry by not intentionally introducing moisture in to the chamber; the reason for introducing no moisture is that silver corrosion has little dependence on humidity and, anyway, adding and controlling humidity to a chamber at 105C would be problematic. The creep corrosion of PCBs is commonly believed to be mostly on RoHS compliant circuit boards [Fu 2011]. RoHS is the European Union Restriction of Hazardous Substance (RoHS) directive issued in February 2003 that took effect on 1 July 2006, banning the use of lead in solder joints [ROHS 2003]. The silver and copper containing tin solder that replaced the lead tin (PbSn) solder has two shortcomings: One is its higher melting range that necessitates the change of PCB laminate epoxies to ones with higher glass transition temperatures; the second is the poor wetting of the copper metallization on the PCBs by the lead free solder, necessitating the use of various surface finishes on the copper metallization to enhance wetting by the lead free solder. The net result is that the RoHS compliant PCBs are more prone to creep corrosion. Creep corrosion is the corrosion of the copper (and sometimes silver) metallization on PCBs and the creeping of the corrosion product (mostly sulfides of copper and sometimes silver) on the board surfaces which may lead to the electrically shorting of neighboring features on PCBs. The problem of creep corrosion has been largely brought under control by selecting finishes, by trial and error, that have less propensity to creep corrosion. But the challenge of a reliable qualification test for creep corrosion remains, though some progress has been made developing the mixed flowing gas (MFG), the Chavant clay and the flowers of sulfur tests. Before we discuss the tests for the qualification of creep corrosion, let us briefly examine the three gases (H2S, NO2 and Cl2) considered to have the most influence on the corrosion of copper and silver [Abbott 1988]. H2S: Hydrogen sulfide is a colorless gas with a rotten-egg odor. Some can smell hydrogen sulfide at levels as low as 0.5 parts per billion (ppb). Most hydrogen sulfide in the air comes from natural sources. It is produced when bacteria break down plant and animal matter, often in stagnant waters with low oxygen content such as bogs and swamps. Volcanoes, hot springs and underwater thermal vents also release hydrogen sulfide. Industrial sources of hydrogen sulfide include petroleum and natural gas extraction and refining, pulp and paper manufacturing, rayon textile production, chemical manufacturing and disposal of construction and demolition debris contain large quantities of wallboard [NY State Dept of Health 2013]. NO2: Nitrogen dioxide belongs to the NOx family of highly reactive gases that are generated when fuel is burned at high temperatures in motor vehicles, stationary sources such as electric utilities and industrial boilers. A suffocating, brownish gas, NOx is a strong oxidizing agent that reacts in the air to form corrosive nitric acid, as well as toxic organic nitrates [EPA 2013]. Cl2: Chlorine is among the ten highest volume chemicals manufactured. It is produced commercially by electrolysis of sodium chloride brine. Chlorine is used to disinfect water and is part of the sanitation process for sewage and industrial waste. During the production of paper and cloth, chlorine is used as a bleaching agent. It is also used in cleaning products, including household bleach which is chlorine dissolved in water. Chlorine is also used in the preparation of chlorides, chlorinated solvents, pesticides, polymers, synthetic rubbers, and refrigerants. Mixed flowing gas test is the most researched test for the study of atmospheric corrosion of copper. The major development of the test and the definition of the performance characteristics of the test occurred at Battelle-Columbus in the 1980s [Abbott 1988]. According to this study, the commonly occurring gases that have the most influence on copper corrosion were found to be H2S, Cl2 and NO2. Sulfur dioxide, SO2, a common contaminant in most industrially polluted environments, was not considered to be corrosive to copper. Creep corrosion of copper on porous Au-Ni-Cu plating was found to be accelerated by small amounts of NO2, but large variations in concentration had little added effect. During the early 1980s, PCBs were soldered with PbSn solder and therefore did not suffer creep corrosion. The Battelle-Columbus study had no reason to address creep corrosion on PCBs. The need for a reliable qualification test for copper creep corrosion on PCBs was highlighted by a 2007 publication from Dell Incorporated with the surprising observation that immersion silver (ImAg) finished PCBs, that had passed a battery of tests, such as HALT, shock and vibration, torsion and thermal cycling, failed in the field due to creep corrosion, some within 4 weeks of being put into service [Schueller 2007]. The field experience was especially surprising because earlier in 2004, a Table 1: MFG test conditions developed/used by various organizations. For organizations with multiple test conditions, only the most aggressive condition is listed here. Organization Reference Temp (C) RH (%) H2S (ppb) Cl2 (ppb) NO2 (ppb) SO2 (ppb) Cu corrosion rate, nm/day Battelle Columbus Abbott 1988 502 752 20010 505 20025 --Rockwell Automation Veale 2005 28.50.5 75 100 20 200 200 350 AlcatelLucent Xu 2009 40 69 1700 20 200 200 600 Univ of Maryland Zhang 2010 321 751 105050 ----iNEMI Fu 2011, Fu 2012 40 70-75% 1200 20 200 200 500 typical MFG test at Underwriters Laboratories on PCBs with immersion silver had not shown any evidence of creep corrosion [Cullen 2004]. One of the most cited early work on PCB creep corrosion testing in MFG environment was done at Rockwell Automation [Veale 2005]. The MFG gas composition and temperature and humidity condition was able to cause creep corrosion on lead free PCBs with morphology similar to that occurring in the field. The test conditions are listed in Table 1. The usefulness of the test was demonstrated by the finding that the mean time to failure of immersion silver (ImAg) finished boards was an order of magnitude less than that of organic surface preservative (OSP) finished boards, in agreement with the relative creep corrosion propensities of the two surface finishes in the field. The next major work on MFG testing to be commonly cited in the literature was conducted at Alcatel-Lucent [Xu 2007]. As Table 1 shows, this was probably the first study to dramatically raise the concentration of H2S. The awareness that creep corrosion on Pb-free PCBs is highly surface sensitive was reported. [Xu 2007 and 2009]. The following general observations were made: PCBs with clean FR4 and clean solder mask surfaces were found not to support creep corrosion; organic acid wave soldering flux residues supported creep corrosion; rosin-based wave soldering fluxes and rosin-based solder paste are resistant to creep corrosion; and MFG testing provides a realistic accelerated test for creep corrosion. . Two recent papers by iNEMI [Fu 2012 and 2012] took the lead from the Alcatel-Lucent work and conducted a major study on four finishes: immersion silver (ImAg), electroless nickel-immersion gold (ENIG), lead-free hot air surface leveled (Pb-free HASL) and organic solderability preservative (OSP). Of these four finished, the Pb-free HASL was found to be the least prone to corrosion. University of Maryland did extensive work developing MFG test method using various gas compositions and temperature and humidity test conditions [Zhang 2009, Zhang 2010]. One interesting observation coming out of the work was that a MFG test using single H2S gas can cause creep corrosion of PCBs. This is contrary to the general understanding that creep corrosion requires a synergy amongst various gases and may be surface contamination on PCBs. Other tests for creep corrosion that have been researched are the Chavant clay test and the flowers of sulfur test. Mazurkiewicz was probably the first to observe and/or publish creep corrosion on PCBs exposed to sulfur-bearing gases in an automobile clay modeling design shop [Mazurkiewicz 2006]. Schueller published a copper creep corrosion test based on Chavant type J-525 modeling clay [Schueller 2007]. The PCBs under test were sealed in a container along with 2-4 pounds of wet clay. Sulfur gas was generated by heating the clay to approximately 50C twice a day. Creep corrosion was typically visible on ImAg PCBs after 2 days and was quite pronounced after 5 days. Others, including Zhou at the University of Maryland, have used the clay test for the reliability assessment of PCBs [Zhou 2009]. The Chavant clay test has not become popular probably because the sulfur concentration in the air is difficult to control. Modified versions of the flowers of sulfur (FOS) test based on ASTM B 809-95 (2008) have been used for testing the corrosion protection provided by conformal coatings [Hindin 2003] and for testing the corrosion resistance of miniature surface-mount resistors [Cole 2011]. The FOS test has also been successfully used to cause creep corrosion on PCBs [Kondos 2013]. The advantage of FOS testing is that air flow over the test PCBs and the temperature of the chamber are the only variables required to control the test. For every temperature, under equilibrium condition, there is a known, constant concentration of sulfur vapor (S8) in the air. The humidity, achieved by having a large surface area saturated salt solution in the chamber, is also controlled by temperature. The simplicity of control is a very attractive reason to study the use of FOS for the qualification of PCBs for resistance against creep corrosion. The focus of this paper is the design of a FOS chamber and its performance evaluation using copper and silver foils and the preliminary testing of the procedure using PCBs from a manufacturing lot known to have failed in service. Role of fine dust on creep corrosion A survey of the quality of air associated with the occurrence of creep corrosion and the corrosion of surface mount resistors was conducted and published by ASHRAE in the 2011 white paper on particulate and gaseous contamination guidelines for data centers [ASHRAE 2011]. The paper concluded that for the aforementioned corrosions to occur, the copper and silver corrosion rates must be greater than 300 and 200 angstroms/month, respectively. Though the survey did not include temperature and humidity observations, in general most mission critical data centers do maintain the temperature and humidity well within the ASHRAE recommended limits of 18-27C and relative humidity less than 60%. For corrosion products such as sulfides and/or for metallic ions to creep on PCB surfaces there must be an adsorbed moisture film on the surface forming an electrolytic path through which material transport can occur. But since a typical mission critical data center has well controlled and low relative humidity, for creep corrosion to occur, there must be dust with low deliquescent relative humidity settled on the PCB surface which can absorb moisture from the air and get wet even under low relative humidity conditions. It turns out that high levels of particulate contamination do occur along with high levels of gaseous pollution. Airborne dust in data centers can be characterized in to two size ranges: Coarse dust with particle size greater than 2.5 m and fine dust with particle size less than or equal to 2.5 m. Coarse dust is mainly from mineral sources. It is low in ionic salts, which makes it less corrosive, and it can be very effectively kept from entering data centers by common filtration practice. Fine dust is of particular concern because the water soluble ions represent a significantly greater fraction of its mass compared to coarse dust which has much lesser ionic content [Sinclair 1985]. The source of fine dust is both anthropogenic and natural. Nitrogen dioxide from automobile exhaust, ammonia from agriculture and sulfur dioxide from coal fired power plants can combine in the atmosphere possibly on submicron carbon particles to produce fine particles of ammonium sulfate, ammonium hydrogen sulfate and ammonium nitrate [Zhang 2004]. Of the three, ammonium hydrogen sulfate has the lowest deliquescent relative humidity (DRH) of 40% [Frankenthal 1993]. Depending on the composition, fine dust can have an effective DRH of 50-65% [Litvak]. In contrast, the effective DRH of most clean metal surface is in the range 70-80% [Frankenthal 1993]. Fine dust on surfaces can have three detrimental effects: (1) If the relative humidity in the room is above the DRH of the dust, the dust will get wet and support corrosion; (2) The increased surface area due to the fine dust will provide additional area on which gases can be adsorbed, thus increasing the rate of corrosion of the underlying metal; (3) Dust particles can increase the rate of corrosion through differential aeration. It has been suggested that in order to mimic the field conditions, especially those in the more polluted geographies, the qualification test for electronic devices should include both the influence of atmospheric particulate and gaseous contamination [Reid 2010]. A research addressing the contributions of both particulate and gaseous contamination was done at the AT&T Bells Labs [Frankenthal 1993]. In this research a novel but complicated means of generating submicron particles was described and it was concluded that the average acceleration factor from particle deposition was 100. The effect of fine dust on electronic hardware, without any contribution from gaseous contamination was studied by a team at the Lawrence Berkeley National Lab [Litvak 2000] using a novel means of generating fine ammonium sulfate particles and measuring the degradation of electrical surface insulation resistance as a function of relative humidity and voltage. The insulation resistance was found to decrease by several orders of magnitude with increasing relative humidity, even when the room humidity was below the DRH of the fine particles on the test circuit boards. Electrostatic forces were found to enhance the segregation of particles on the test circuit boards. In summary, the direct effect of particulate contamination on hardware corrosion is the absorption of moisture by the fine dust particles especially when the room relative humidity is higher than the deliquescent relative humidity of the dust which (a) (b) Figure 1: (a) Schematic and (b) picture of paddle-wheel setup in FOS chamber (a) (b) Figure 2: Forced air setup. (a) Schematic of the setup; (b) Airflow pattern. for fine particles is in the range 50-65%. The fine dust chemical constituent of main concern is probably the ammonium hydrogen sulfate because of its low deliquescent relative humidity of 40%. The high ionic content of the settled fine particles provides the electrolytic path for the metallic ions and corrosion products to creep on the PCB surfaces. Particles can also contribute to degradation of hardware reliability two other ways: (1) Particles on metal surfaces may increase the rate of metal corrosion by differential aeration induced localized corrosion and (2) by increasing the surface area for gas adsorption. Experimental Setup and Procedure The flowers of sulfur chamber is a 300-mm cube with a front-loading door sealed with a silicone coated rubber gasket. Humidity is maintained using a saturated salt solution contained in two beakers with 100-mm diameter. Sulfur concentration is maintained by sulfur contained in two beakers with 80-mm diameter. Temperature in the chamber is maintained at 60C with as little a temperature fluctuations as possible. There are two approaches for mounting the test specimens in the chamber and for stirring the air: One uses a paddle wheel as shown in Figure 1, with 8 paddles, to mount the 8 test PCBs and move the air; the other is a forced air setup that uses a blower to draw the air over 8 stationary test PCBs, as shown in Figure 2. The paddle wheel has a rotations per minute (RPM) limit of 25. The paddles do not move the air as readily as one’s intuition might expect. At the maximum possible RPM, fluid flow simulation estimates the air velocity over the specimens is about 0.1 m/s. The air velocity over PCBs in computer racks is in the range 0.5-2 m/s. To achieve these higher airflow velocities, a new insert was designed and built, as shown in Figure 2. In this setup, a blower draws air from the front of the setup that has 8 test PCBs. The action of pulling the air up through the setup and having the air take a 180 degrees turn results in quite a uniform air flow across the 8 test PCBs as shown in Figure 3a. Air velocities of up to 2 m/s can be achieved. The air velocity is controlled by the RPM of the blower which in turn is controlled by the dc voltage input to its electric motor. Figure 3b shows a relationship of the air velocity over the PCB test specimens and the dc voltage applied to the blower motor. Figure 4 shows that in the paddle wheel setup, the temperature and humidity can reach steady state within ~2 hours. The time to achieve steady state in the forced air setup would be even shorter because of the higher air velocity in the chamber. As mentioned in the previous section, it is probable that the necessary conditions for creep corrosion in data centers, with well controlled temperature and humidity conditions, may be the high levels of gaseous and particulate contamination. The Voltage 10 15 20 25 30 A ir v e lo c it y , m /s 0.4 0.6 0.8 1.0 1.2 1.4 (a) (b) Figure 3: (a) Forced air setup airflow pattern and (b) airflow velocity versus voltage applied to the blower. The air velocity versus voltage relationship was obtained for each forced air setup. Figure 4: The time for temperature and humidity to reach steady state in a FOS chamber with paddlewheel setup. most corrosive salt in fine dust is probably ammonium hydrogen sulfate because of its low deliquescent relative humidity of 40%. Besides the obvious need for reduced sulfur, the early and expansive work by Abbott pointed to the need for NOx and Cl2/chlorides in the environment to enhance copper corrosion rates. In our exploratory work, these two components were represented by ammonium nitrate and by NaCl in the form of dust. We covered the test PCBs with dust consisting of equal parts of (NH4)2SO4, NH4HSO4, NH4NO3 and NaCl before subjecting them to the FOS vapors in the test chamber. The challenge was to device a convenient way to apply a controlled amount of the dust uniformly distributed on the test PCB surfaces prior to stressing the PCBs in the FOS chamber. The objective was to achieve 10-500 g/cm total salt concentration on the test PCB surfaces. The procedure explored to spread the dust on the PCBs was to mix a solution of 25% by mass of each of the 4 salts, (NH4)2SO4, NH4HSO4, NH4NO3 and NaCl, pour a controlled amount of the aqueous solution on the test PCB and let the solution dry, leaving behind the salts on the PCB surface. Uniform distribution of salt could not be achieved because the water-PCB surface tension kept puddling up the water into large droplets, thus, concentrating the salt in the areas last to dry. We failed to achieve a uniform salt distribution across the complete PCB surface. So, instead, we decided to apply small droplets of the dilute salt solution locally on areas of interest on the PCBs. By controlling the volume of the drop, Relative humidity, % 40 50 60 70 80 90 100 C o rr o s io n r a te , a n g s tr o m s /d a y