Theorists question Auger as the primary cause of LED droop
Band–to-band Auger recombination does not account for droop, the decline in nitride LED efficiency at increasing drive currents, according to team of theorists based in the US and Italy.
The claim of this team is at odds with previous calculations of Chris Van de Walle’s group at the University of California, Santa Barbara. Last year these West-coast academics performed first-principles, density-functional electronic structure simulations on GaN and InN, and used these results to determine that Auger recombination was a likely cause of LED droop.
These calculations, like those of the US-Italian team, considered the Auger coefficient for non-radiative processes involving two electrons and one hole at a range of bandgaps.
Van De Walle’s team found that the strongest Auger process revolves around resonant electron scattering from the lowest to the second lowest conduction band. This Auger coefficient for this process peaks at 2 x 10-30 cm6 s-1 for InGaN with a bandgap of 2.5 eV.
In contrast, perturbation theory calculations by three theorists from Boston University and the Politecnico di Torino, Italy, indicate that the Auger coefficient for the resonant interband coefficient is far, far lower – less than 10-32 cm6 s-1 – and occurs at a higher energy, 2.8 eV.
This team of Francesco Bertazzi, Michele Goano and Enrico Bellotti has also calculated the Auger coefficient for processes involving two holes and one electron. This process has no resonance peak, steadily decreases as the bandgap increases, and is less than 10-32 cm6 s-1 for an InGaN bandgap greater than 2 eV.
The US-Italian team has tried to fathom why its value for the Auger resonance peak is 0.3 eV higher than that provided by UCSB simulations.
They believe that this discrepancy stems from employing different values for the energies of the lowest and second lowest conduction bands in InGaN. They began by adopting a ‘nonlocal empirical pseudopotential method’ to calculate the band structure for InN and GaN, an approach that allows parameters to be tweaked so that the electronic structure can replicate the main features obtained by either experiment or first principles calculations. Application of a modified virtual crystal approximation yielded the electronic structure for the conduction and valence bands in InGaN.
In comparison, the UCSB team first determined the energy difference between the lowest and second lowest conduction bands in InN and GaN, before linearly extrapolating values for InGaN.
The US-Italian team says that the difference between the value of the energy of the resonance peak calculated by them and the UCSB team impacts the calculation of the Auger coefficient. However, its impact is not large enough to account for the three orders of magnitude difference between the two calculations.
According to them, this massive difference could stem from the West coast team’s failure to fully account for the symmetry of the electronic states that are involved in the calculation of the Auger recombination strength in InGaN alloys.
Both teams have only considered band-to-band Auger recombination in bulk InGaN. When this non-radiative process occurs in the quantum well layers of an LED, it could increase the population of higher energy electrons that could leak out of the wells. Further work is needed to investigate Auger recombination in these types of structure, including second order processes mediated by phonons, defects and material inhomogeneities, according to the US-Italian team.
These researchers detail their work in a paper that will soon appear in Applied Physics Letters.
