Electromigration in Power Devices: A Combined Effect of Electromigration and Thermal Migration

Abstract

In the power semiconductor industry, there is continuous development toward higher maximum current capability of devices while device dimensions shrink. This leads to an increase in current density which the devices have to handle, and raises the question if electromigration (EM) is a critical issue here. Generally, an EM failure can be described by the Black’s equation with temperature and current density as the main influencing factors. Normally, the current that the power packages need to handle lies in the range of 100 A. However, it should be noted that power devices exhibit asymmetric sizes of drain and source contacts. This may lead to higher current density at the source leads (area ratio drain/source: ~9× for PQFN 5 × 6). Nevertheless, the source lead area is still much larger than that of the flip chip bumps (i.e., 28 times larger than a 100-μm microbump). This typically enhances the safety of the power device with respect to EM. However, with regard to future development toward higher maximum current capability, we intended to investigate further on the EM of power devices. In the present work, we focused on the PQFN 5 × 6 package to study the EM behavior of a power device soldered on a printed circuit board (PCB). We employed the highest current (120 A) and temperature (150°C) that the stress test system could handle to study EM in accelerated mode. First fails occurred after ~1,200 h, which was much earlier than expected from previous flip-chip investigations. In addition, we found separation gaps in the solder joint between drain contact and PCB, which experienced the lowest current density in the whole test. Contradictorily, we observed only minor solder degradation at the source interface, regardless of the higher current density there. Nevertheless, the separating metal interfaces still correlated well with the current direction. Thermal simulations revealed that due to the self-heating of the device by the high current applied, both the drain and source leads were exposed to much higher temperatures (Tmax = 168°C) than the PCB board which was kept under temperature control at 150°C. This temperature difference resulted in a thermal gradient between the device and PCB, which, in turn, triggered thermal migration (TM) in addition to EM. As TM for the drain contact occurred in the same direction as EM, it enhanced the degradation effect and therefore led to a shorter time-to-failure at the drain. In contrast to this, such an enhanced effect did not occur at the source side. As a result, we observed higher solder degradation at the drain side, which we did confirm by switching the current direction in the test. To minimize the TM effect, a special EM test vehicle, which used a Cu plate instead of the metal-oxide-semiconductor field-effect transistor chip, was designed and fabricated. Thermal simulation verified that the device operated at similar temperatures as the PCB board. Using this setup, it was possible to study EM in an accelerated mode and, thus, investigate the pure EM behavior of the power device.