At this February’s Photonics West Conference, Joachim Piprek did a quick survey asking the audience which of the currently proposed mechanisms they thought to be responsible for the efficiency droop in GaN-based devices. The vast majority supported Auger as the dominant cause. I guess, this should not be surprising, seeing how this theory is pushed much more forcefully than any other.
In 2007, Philips Lumileds Lightning Co. declared it had “fundamentally solved” the droop problem, later revealing that they identified Auger losses as the culprit. In 2009, when intraband Auger processes were suggested as dominant process by K.T. Delaney, et al. (Appl. Phys. Lett. 94, 191109 (2009)), the authors concluded: “The calculated values … confirm that Auger recombination is a key loss mechanism in nitride light emitters.” Two years later, calculations by E. Kioupakis, et al. shifted the blame to indirect Auger processes (Appl. Phys. Lett. 98, 161107 (2011)). One of the authors, C. Van de Walle, declared: “With Auger recombination now established as the culprit [for the droop] …”.
Recently, J. Iveland et al. performed their well publicized experiment aimed at measuring directly the Auger losses in Nitride materials (Phys. Rev. Lett. 110, 177406 (2013)). They proceeded to present their work under titles like: “Auger effect identified as main cause of efficiency droop in LEDs,” (SPIE Newsroom (2014). DOI: 10.1117/2.1201406.005109) and with statements like: “Cause of LED Efficiency Droop Finally Revealed” and abstracts like “LED droop: Overwhelming Evidence for Auger” (Compound Semiconductor (2013)).
One prominent proponent of the Auger theory, Mike Krames, summarized the level of confidence in the thesis with the quote: “The debate goes on in other people’s heads, but it doesn’t go on in mine,” (New York Times, Feb. 20 2012).
However, the proposed proofs have been met with skepticism all along the way. It was shown that direct intraband Auger processes are too small to explain the droop. A re-evaluation of intraband Auger processes found these to be orders of magnitude smaller than original proposed. Among the Auger contributions, the indirect processes have been confirmed to be the dominant ones in Nitrides, but their amplitude is still found too small in calculations that can determine the radiative losses correctly. The latter calculations have also shown that the temperature dependence of the droop is fundamentally different from that of any Auger processes.
As for the experimental evidence, calculations by F. Bertazzi, et al., indicate that internal scattering processes should not allow for Auger electrons to be measured in the Iveland set-up (Appl. Phys. Lett. 106, 061112 (2015)). We recently examined another experiment that also aimed at the direct measurement of Auger losses in Nitrides (J. Comp. Electron.14, 425 (2015)). We found that the results of the experiment by M. Binder et al. (Appl. Phys. Lett. 103, 071108) could well be explained without consideration of any Auger processes. I should stress that neither Bertazzi et al., nor our own work can disprove the importance of Auger losses, nor are they aimed at doing that.
So, in my head, the debate does go on. However, the recent experiments by Iveland and Binder provide a highly welcomed, refreshing new look at the problem. While the first installations have shortcomings, improvements on them could be made which should help to eliminate uncertainties in their results. E.g., for the Iveland experiment, calculations as performed by Bertazzi et al. should be used to guide modifications in the used structure which unequivocally would allow Auger electrons to reach the detector. We are currently looking into modifications of the structure used in the Binder experiment which could eliminate uncertainties there. In both cases, checking for known dependencies of Auger processes besides the density dependence, like temperature or external pressure, could also bring additional insights.