16.1 GaN HEMT Lifetesting ΠCharacterizing Diverse Mechanisms

We describe a conceptually simple technique for estimating the contributions of individual degradation mechanisms to the wearout of RF GaN HEMT. This is important because most “physics of failure” studies on such devices report several distinct mechanisms that operate simultaneously. As a result, the overall degradation rate for a given GaN HEMT technology does not have a well-defined thermal activation energy (i.e. the effective Ea varies with time), so it is not possible to extrapolate MTTF’s from elevated-temperature lifetests to application temperatures. Rather, it is necessary to quantify each mechanism separately. INTRODUCTION Because GaN HEMT is a relatively new technology, there is still a variety of distinct degradation mechanisms present in devices from most laboratories. Therefore, to qualify these devices, it is not sufficient to simply lifetest them till they fail – one must measure all the competing mechanisms, and determine their net overall effect at the end of the mission life. This contrasts markedly with the presentday situation for GaAs-based technologies, where the devices from a given laboratory usually wear out by only one well-characterized mechanism. In this paper, we illustrate why simple lifetesting is not sufficient, and briefly outline a technique (based on standard lifetest methods) that allows one to develop a picture of the various mechanisms, and their contributions at any time and temperature. THE PROBLEMS One problem is illustrated in Fig. 1: if one does an RF lifetest, one mechanism might dominate the observed changFig. 1. Hypothetical RF lifetest, illustrating possible effect when two mechanisms with different Ea’s are operating. This makes extrapolation to failure at a different temperature very difficult. In this example, one mechanism has been delayed, as sometimes occurs, to emphasize the effect. es initially, but another (with a different Ea) might take over later. The net δ Pout has contributions from both, so it is very difficult to extrapolate to different temperatures and times. If one does a DC lifetest instead, this problem may still be present, and there is the additional problem of relating the DC stress to the stress that is present during RF operation. Finally, lifetests can completely hide some mechanisms; this is illustrated for a hypothetical DC lifetest in Fig. 2. In the T e m p e ra tu re ( °C )