The effect of efficiency droop in light emitting diodes (LEDs) is a huge roadblock for consumer lighting adoption. To prevent loss of efficiency from droop, LEDs must be operated at lower current density, requiring a larger epitaxial area and increasing the overall cost. Proposed mechanisms responsible for droop include, among others, carrier delocalization from indium rich regions, overshoot and leakage current, and Auger recombination. Current methods used to understand droop mechanisms are indirect, and often rely on models that have no unique solution. It is clear that a direct measurement method and a deeper understanding of the fate of injected carriers not contributing to radiative recombination is needed to focus improvement efforts on materials and structures to help identify and mitigate the relevant droop mechanism(s). By analyzing the energy of vacuum emitted electrons from a forward biased LED, we can gain direct information of their origin internally.

The study of vacuum emitted electrons has existed for almost 130 years with the discovery of the photoelectric effect. Advancements in electron energy analysis techniques have led to the direct measurement of conduction band structures and transport properties in many commonly used semiconductors such as: InP, GaAs, Si, and recently GaN.

The kinetic energy of the vacuum emitted electrons from an InGaN/GaN LED was analyzed and three peaks were identified: First, a low energy peak, resulting from photoexcited electrons generated by diode light. Second, a mid-energy peak, generated by the accumulation of thermalized electrons at the bottom of the Gammaconduction band valley. Third, a high-energy peak is generated by an accumulation of electron at the bottom of a low lying side-valley "L". Auger recombination is uniquely identifiable as it is the only proposed droop mechanism capable of generating hot carriers and solely responsible for the population of electrons found in the L-valley.

Two control experiments were carried out to strengthen our interpretation. First a simple GaN pn junction was measured and generated only a single Gamma-valley peak. Second, selective detection of photoemission under modulated light from an LED in forward bias confirms that only the low energy peak is photogenerated and that LED light incapable of generating the higher energy Gamma or L-valley peaks. Lastly, we discuss the new UCSB electron energy analyzer as well as some proposed future experiments to advance the electron emission technique.