Improving the efficiency of gallium nitride-based LEDs for high-power applications

Over the past decade, solid-state lighting has been a rising star for next-generation illumination sources. It has the advantages of saving energy and providing high-quality lighting, as well as offering a range of flexible design features.¹ In particular, there has been much interest in the potential of solid-state lighting using gallium nitride (GaN)-based LEDs. However, for high-power applications, LEDs need to be operated at a very high current density, which can diminish device efficiency.²

In LEDs, the device structure gives rise to a separation of charge carriers, that is, electrons and holes. When a bias voltage is applied, the electrons and holes recombine, emitting light. Carrier overflow---electron overflow out of the radiative-recombination region---has been identified as the main cause of the drop in efficiency at high current densities.^{3, 4} An aluminum gallium nitride (Al_xGa_1−xN) ‘electron-blocking layer’ (EBL) has been widely adopted in LED structures to to suppress carrier overflow. However, it has been reported that the large polarization field in the Al_xGa_1−xN EBL reduces the effective energy barrier height for electrons,⁵ weakening the stopping power of this barrier layer. To make the situation worse, the polar nature of the group III nitride material and lattice mismatch give rise to a polarization field at the interfaces of the GaN and EBL. The polarization field bends the electronic energy band structure. In addition, there is a difference in the energy of the outermost electrons: a valance band offset. The band bending and offset are thought to retard the injection of holes, preventing recombination and further reducing efficiency.^{2, 5}

Figure 1. Schematic diagram of the concept of band engineering. EBL is the electron blocking layer and MQWs are multiple quantum wells. The dotted blue circle indicates the triangular valence band structure. Electrons are injected from the n-side and holes from the p-side. n, p: Semiconductor types.

By adapting the concept of band engineering, we designed a graded-composition EBL (GEBL) for GaN-based LEDs. The GEBL not only suppressed the electron overflow from the active region but also enhanced hole injection. Our concept of band engineering stems from observations of the band diagram of GaN-based LEDs. For conventional LEDs operated under a forward bias, the valence band of the EBL has a triangular shape due to the internal polarization field and externally applied forward bias⁶ (see Figure 1, bold black line). Electrons are injected from the n-type semiconductor side, where the band structure has a series of dips known as multiple quantum wells (MQWs) that confine the electrons and holes and allow them to radiatively recombine. The MQWs are sandwiched by the p- and n-type layers. The valance band of the EBL slopes upward from the n-GaN side toward the p-GaN side, where holes are injected. The upward slope retards the holes from transport across the triangular barrier. But if the aluminum composition in the EBL increases gradually from the n-GaN side toward the p-GaN side, the bandgap gradually broadens. As a result, the barrier in the valance band could level out or even slope in the opposite direction, while the slope of the conduction band could be enhanced. We expect an improvement in hole transport across the EBL as well as improved electron confinement as a result.

Figure 2. Calculated (a) hole concentration distribution and (b) electron concentration distribution of conventional EBL and graded EBL (GEBL) LEDs at a current density of 100A/cm².

We simulated the profiles of the hole and electron concentration distribution at a current density of 100A/cm² (see Figure 2). The simulations clearly show that injected holes can spread more uniformly in the EBL region with a GEBL compared with a conventional one. In addition, the hole concentration in the MQWs is significantly increased, as expected. Moreover, the electron concentration in the MQWs is also increased, while the electron distribution within the GEBL region and p-GaN is decreased by over two orders of magnitude. These results indicate that the GEBL can suppress the electron overflow out of the active region more effectively than a conventional EBL, even though the conduction band offset between the last GaN barrier and the GEBL is diminished.

We also grew LEDs with EBL and GEBL structures on c-plane sapphire substrates by metal-organic chemical vapor deposition. The growth temperature of the conventional EBL and the GEBL structures was the same (870^°C), and the aluminum composition profile of the GEBL was graded from 0 to 25%. Finally, the LED chips were fabricated by a regular chip process. The chips incorporated an indium tin oxide current spreading layer and nickel/gold contact metal. The effective area of the LED is 300×300μm². The emission wavelength was around 450nm at 22A/cm² for both LEDs.

Figure 3. Forward voltage and output power as a function of current density for conventional EBL and GEBL LEDs.

We measured the forward voltage and power as a function of current density for both the conventional EBL and GEBL LEDs (see Figure 3). The values for the forward voltage and the series resistances of the GEBL LED at 22A/cm² are 3.28V and 7Ω, respectively, lower than the values of 3.4V and 8Ω measured for the conventional LED. We attribute the reduced forward voltage and series resistance to the improvement in the hole injection and the higher-efficiency p-type doping in the GEBL.⁷ The output power was enhanced by 40 and 69% at 100A/cm² and 200A/cm², respectively. We attribute this enhancement to the negligible tunneling of holes at high current density. Consequently, hole transport into the MQWs is dominated by diffusion.⁵ The diffusion process takes place more readily in the GEBL LED than in the conventional device due to the flatter valence band and smaller band offset at the interface of the last GaN barrier and the EBL. The normalized efficiency of the conventional and GEBL LEDs as a function of current density was also investigated (see Figure 4). The drop in efficiency, defined as (η_peak−η⁻²_@200A/cm²)/η_peak (where η_peak is the highest efficiency achieved and η_@200A/cm² is the efficiency at a current density of 200A/cm²) was reduced from 34% in the conventional LED to only 4% in the GEBL LED. We attribute these significant improvements in efficiency mainly to the enhancement in the hole injection and electron confinement, especially for high current densities.

Figure 4. Normalized efficiency as a function of current density for conventional EBL and GEBL LEDs.

This work shows that the transport of holes can be modified by bandgap engineering in the epitaxial layer design. The performance of our LEDs is improved as a result. Further investigation of this method could be an important step toward highly efficient LEDs. More work on hole transport in the MQWs is required. We will then focus on additional improvements with the aim of achieving higher efficiency at high current density.