On the reliability of power semiconductor modules in corrosive environments

Abstract

Power electronics devices are used in increasingly harsher environmental conditions as they are commonplace in driving all kinds of electric motors. The harshest conditions feature corrosive gases and high humidity, and these together pose a danger to the device operation. Dendritic corrosion growth has been detected in the isolation trenches of power semiconductor modules, caused by such corrosive environments. The dendrites can, over time, form a resistive short-circuit in the isolation trench, leading to degraded component performance and eventually failure. The power semiconductor modules are separated from the environment by gas-permeable packaging, meaning that the transfer of corrosive gases and humidity to the component surface are only slowed down and not prevented by the package. This is the root cause of the dendrite growth problem. An additional disadvantage of power electronics is their high voltage levels, which will enable and accelerate electrochemical corrosion reactions.

The work included in this doctoral dissertation is experimental. The dendrites grow over long time periods spanning over years, and the data gathered are very limited as, e.g., measurements of corrosive gases in the ambient are not available. Therefore, the dendrite growth phenomenon has to be accelerated with accelerated corrosive gas tests. This dissertation uses mixed flowing gas tests to accelerate dendrite growth in the copper–alumina isolation trenches of power semiconductor modules. The existing standard mixed flowing gas tests were found not to be applicable to testing of potted power semiconductor modules. The slow time constant associated with gas permeation causes a time window at the beginning of an accelerated corrosion test, during which no dendrite-forming corrosion reactions occur on the component surface under the packaging and potting materials. As part of the dissertation work, further developments to the mixed flowing gas test were made. As a result, a cyclic test profile is introduced that aims to increase the gas permeation speed into the component surface to minimize the delay at the beginning of the corrosive gas test.

The chemical or electrochemical reactions forming the dendrites have not been studied extensively in the existing literature, but the effect of increasing humidity and bias voltage has been documented to accelerate the dendrite growth. In this dissertation, the effect of local heating as a means to inhibit dendrite growth was studied based on its strong humidity dependence. Heating was found to significantly increase the component lifetime.

The research on this topic will continue after this dissertation. In future work, the phenomenon of dendrite growth will be investigated further, as the effect of parameters such as bias voltage and its waveform has not yet been thoroughly addressed. Furthermore, as the component packaging is a major contributor to the reliability in corrosive environments, particular emphasis should be placed on the analytical characterization of the component potting materials to construct gas permeation models.