PIONNERING EXPERIMENT
MYSTERY of NITRIDE LEDs
the device. After this step, measurements could be performed at room temperature. Electron energy distribution curves( EDCs) are measured using a cylindrical electron energy analyzer( Figure 2 and top photo). Varying the cathode potential( V cath) allows sweeping the energy. Figure 1 summarizes the energy scale. Note that a hole in the analyser allows taking the electroluminescence emitted by the LED during the electron emission measurement( Figure 2). The electron energy resolution is approximately 50 meV. The first measurements were carried out on a blue LED composed of eight InGaN quantum wells containing 18 % indium [ Iveland ]. Taking into account both quantum confinement and internal electric fields in the quantum wells, the transition energy in these wells is 2.75 eV( blue light at 450 nm). The device also includes an Al₀. ₁₅Ga₀. ₈₅N EBL, and a 200-nanometer-thick p-doped top layer. When the LED is forwardbiased and the current is gradually increased, one, then two, and finally three distinct peaks appear( figure 3). The first, at low energy, corresponds to electrons mainly emitted from the metal excited by the blue light generated by the LED. This first peak extends up to about 0.5 eV, which, when added to the vacuum level, gives an energy of 2.8 eV above the Fermi level, consistent with the photon energy of the emitted light. The second peak, separated from the first through spectral analysis, extends up to around 1.45 eV( estimated at 64 mA) and corresponds to the emission of electrons in the Γ valley. It originates from electrons that have crossed the active region and the EBL as well as from electrons initially in the L valley that thermalized into the Γ valley on their way to the surface. Finally, the third peak, extending up to 2.4 eV, corresponds to a higher-energy valley attributed to the L valley. These“ hot” electrons are believed to be a signature of Auger processes that excite them into the L valley before they diffuse through the p-type GaN to the surface. Their energy distribution broadens in the final, strongly curved region of the band structure. The energy separation between the Γ and L valleys is about 0.95 eV in this case. As the diode’ s voltage and current increase, parasitic resistances in the p-contact region cause both the Γ and L valley peaks to shift by roughly the same amount. Tracking these two peaks makes it possible to determine an average separation between the valleys, ultimately estimated at 1 eV. Note that similar electron emission studies were performed on LED structures by optical pumping, i. e. photoemission [ 4 ]. When the exciting photon energy exceeds 4.2 eV( 0.8 eV above the gap), the L-valley signal emerges confirming a value of approximately 1 eV for the Γ – L separation in GaN( and low-indium InGaN). Studying how the intensity of the peak associated with the L valley varies with diode current also proves highly informative. At the same time, the emitted light intensity is measured( see Figure 4). Initially, the light output increases linearly with current, but then grows more slowly, with a sublinear behavior that reflects the“ efficiency droop.” By extrapolating the initial linear region, one can estimate the emission that would be obtained if this droop did not occur. Comparing the two curves, with and without the droop, makes it possible to determine the current
REFERENCES
[ 1 ] A. David et al., ECS J. Solid State Sci. Technol. 9, 016021( 2020) [ 2 ] E. Kioupakis et al., Appl. Phys. Lett. 98, 161107, 2011; Phys. Rev. B 92, 035207( 2015) [ 3 ] J. Iveland et al., Phys. Rev. Lett. 110, 177406( 2013) [[ 4 ] M. Picardo et al., Phys. Rev. B 89, 235124( 2014) that did not give emission. This“ missing current” corresponds to the portion of the injected current feeding the droop process, i. e. Auger recombination. When the intensity of the L-valley peak is plotted as a function of this missing current, a perfect linear relationship appears. In other words, the Auger effect, revealed through the presence of electrons in the L valley, is directly responsible for the loss of quantum efficiency in high-current nitride LEDs. As further proof of the link between the Auger effect in LEDs and hot electrons, the same measurement was repeated on a simple pn junction without quantum wells where both radiative and Auger recombination are reduced due to the absence of carrier confinement. In that case, only the Γ-valley peak was observed [ 5 ]. Thus, roughly a decade of electron emission measurements solved the mystery of the efficiency droop. Have they provided a solution? Partially, yes. Although the problem is intrinsic, it can be mitigated by reducing charge densities. Strategies such as ensuring a more uniform current distribution, increasing the number of quantum wells, or widening them are all feasible approaches and several have already been widely implemented.
Acknowledgements
The author aknowledges Prof. Claude Weisbuch( École Polytechnique, France and University of California at Santa Barbara, USA.
[ 5 ] W. Ho et al. Appl. Phys. Lett. 122, 212103( 2023); Appl. Phys. Lett. 119, 051105( 2021)
28 www. photoniques. com I Photoniques 137