Investigating dendritic growth: getting to the root of capacitor shorts.

A recent failure analysis presented with a clear failure mechanism of a dendrite shorting along the surface of the component. Our job was to determine what happened.

[ILLUSTRATION OMITTED]

[TEXT NOT REPRODUCIBLE IN ASCII]

A simple dendrite grew across the surface of a 3V capacitor (Figure 1), but curiously it grew below the coating and did not break the surface. No visible condensing water was present on this field failure, which was used as a controller assembly in a controlled environment for 30 days. No other components showed visible residue issues on the component surface.

[FIGURE 1 OMITTED]

Process conditions. The acrylic coating thickness was uniformly 0.003" to 0.004" thick over the board surface. The mixed technology assembly included bottom-side SMT and selective soldering. A low-residue, no-clean approach was used to build the assembly. The bottom-side SMT parts were reflowed with a no-clean solder paste, while through-hole components were soldered with a VOC-free, no-clean flux. A selective pallet protected SMT components from the solder wave.

Analysis techniques. Analysis was conducted using x-ray fluorescence and ion chromatography using a localized extraction tool: the C3 tester. XRF showed tin and lead metalization through the coating only. Scraping samples viewed with XRF also showed tin and lead metalization in the dendrite.

Ion chromatography analysis of the entire assembly would not reveal what caused the localized corrosion cell on one capacitor. Thus, a focused extraction was conducted by first using a scalpel to roll back the conformal coating and C3 to isolate an extraction area of 0.1 i[n.sup.2].

The C3 tester uses DI water steam to solublize anything that would come into solution under normal operating conditions as an extraction solution, and then conducts an electrical impact assessment of the extracted solution. By using a Y electrode with a 0.050" pitch and a 10V bias, the electrical impact of the contaminants in the DI water can be assessed and a solution determined. DI water can consistently cause copper dendrites to short in 450 to 500 sec., while corrosive ionic residues cause shorts much faster. (For example, 9.2 [micro]g/i[n.sup.2] of chloride has constantly caused shorts in 19 sec.).

Table 1 (online) shows ion chromatography analysis of the C3 extraction sites. The area of the corrosion site shows a high level of weak organic acids (WOA) only on the component next to the selective solder area. Using the current production assembly for process comparison, it also shows a high level of WOA only on this location. Each area tested shows a low chloride, bromide, nitrate and WOA, except the area next to the selective wave. Bare-board cleanliness appears to be low from the HASL cleaning process, and because the capacitor from the same vendor's reel nearby (but not next to) the selective wave also showed low levels of ionic residue, it was not a result of these contributors.

The root cause of this failure is the presence of water-based no-clean flux trapped below the conformal coating covering the surface of the capacitor. This flux when not completely heat-activated has a pH of 2.5 and is very corrosive, enough for dendrite growth below a coating. For the biased circuit with sufficient corrosive flux (not completely heat-activated), this failure mechanism caused a dendrite to grow across the surface of the capacitor.

Terry Munson is with Foresite Inc. (residues.com); [e-mail]. His column appears monthly.