Recently, increasing awareness of energy crisis has prompted people to treat reduction of CO2 emission and energy consumption as a crucial global concern. According to statistics released by the US International Energy Agency, illumination accounted for 20% of the global electricity consumption in 2007. Therefore, improving the electrooptical conversion efficiency of lighting equipment is vital for reducing energy consumption as well as CO2 emission. Compared with conventional lighting sources, GaN-based light-emitting diodes (LEDs) have following advantages: no mercury pollution, small foot prints, high electrooptical conversion efficiency, and long lifetime. In 2010, GaN-based white LEDs manufactured by combining blue LEDs and fluorescent powders yielded more than 150 lm/W, and the electrooptical conversion efficiency could reach to approximately 60%. To further increase the efficiency of the high brightness LEDs faces great challenges. Among them, efficiency droop, an emission efficiency reduction while cranking up the injection current density, requires well explanation for its origin and a way to deal with it properly. Possible roots accounting efficiency droop have been raised, such as polarization field due to the quantum confined Stark effect (QCSE), carriers over flow, non-uniform carrier injection and distribution within the active region, and Auger recombination.
Despite so many candidates for roots of efficiency droop have been proposed and debates between them are still going on, it’s still worth turning our attention to the defects inherited in this peculiar material system. The GaN-based LEDs are commonly grown hetero-epitaxially on the sapphire substrates, which tends to generate lots of threading dislocation (TD) with density of approximately 10^8 – 10^10 cm-2 due to the mismatch of lattice and thermal expansion coefficients. Several methods have been demonstrated to reduce the defect density of GaN layer on the sapphire substrates, such as using epitaxially lateral overgrown (ELOG) or hydride phase vapor epitaxially (HVPE) grown GaN template but with a relatively higher fabrication cost. Blue/violet LEDs grown on free-standing GaN substrates have shown superior anti-drooping performance at very high current density but the price of GaN substrates is still expensive. Nowadays, the commercial GaN LEDs are largely grown on pattern sapphire substrates (PSS) to reduce the threading dislocation and enhance the photon extraction efficiency, simultaneously. Even so, the as-grown GaN layer still exhibits lots of threading dislocation (TD) with density of approximately 10^7 – 10^8 cm-2, which is equivalent to one TD in one micron square. Since the diffusion length of carriers is on the order of micrometer, it’s inevitable to neglect the influence of TD on the luminescence efficiency. TDs play an important role in the carrier leakage paths and trap centers. Dark areas would form within the radius of carrier diffusion length around the TD because the carriers could diffuse into TD deep-level states as non-radiative centers to reduce the efficiency of the radiative recombination emission.
(1) Formation of V-shape defects along the threading dislocations
Interestingly, a kind of V-shape pits (V-pits) defects extended as the inverted hexagonal pyramid will be created in GaN-based LEDs. The V-shaped hexagonal pits form an open core at the apex of TD with those facets inclining to six sidewalls on (10-1) planes at an oblique angle of approximately 60˚. In 1997, people started to find that the TDs could easily induce the formation of V-pits during the growth of InGaN multiple quantum wells (MQW) for the active layer of LEDs. Especially, the high Indium content and lower temperature growth of InGaN/GaN MQW facilitate the formation of V-pits much easily. The pit formation might be related to the Indium segregation on the (10-1) surfaces around the dislocation core to provide a strong driving force to induce pits. Most important fact about the V-pits is the growth rate of sidewall MQWs along (10-1) plane on V-pits is generally slower than those grown on the c-plane epitaxial film, resulting in narrower MQWs on the inclined V-pit semi-polar planes with a larger energy gap.
(2) Relationship between InGaN LED performance and V-shape defects
In 2005, Hangleiter et al. reported that the emission energy peaks of localized position on sidewall MQW is higher than the c-plane QW. Because the narrower MQW on the inclined V-pit planes along (10-1) face create higher energy of quantum confinement effect, compared with planar MQWs. The narrower MQWs provide energy barriers for the carriers to be blocked to enter TDs for non-radiative recombination, which satisfactorily explained a relatively high emission efficiency of InGaN LEDs with such a high defect density. In 2014, Kim et al. observed the carrier blocking capability of V-pits with different barrier heights. One-dimensional model with k.p approximation was used to calculate the relation between the droop effect and energy barrier height created by V-pits. It was shown that the higher energy barrier of V-pit can indeed alleviate the efficiency droop by decreasing the tunneling possibility at high injection current. Furthermore, the formation of inverted hexagonal pyramid could help the light scattering and be beneficial to the electrical characteristics of InGaN LEDs because the Mg incorporation of p-type GaN capping on top of V-pits could be hindered, which could increase the local resistance to effectively resist the carrier transportation to TDs.
(3) Manipulating V-shape defects by InGaN/GaN superlattices
Although previous studies have shown various benefits of V-pits formation to passivate considerable amount of TDs, the quantitative impact of V-pit structures is still unclear. In our study, we systematically investigated the relationship between the emission efficiency of InGaN/GaN MQWs and the nanoscale structure of V-pits along TDs. The LED structure embedded InGaN/GaN superlattices (SLS) with low Indium composition under the active layer of InGaN/GaN MQWs as a template to assist the V-pits formation along the TD. By varying numbers of SLS pairs, the V-pit size can be manipulated.
Figure 1 shows the top view SEM image of V-pit distribution on MQWs. Although the distribution of V-pits on MQWs seems random, the spatial frequency mapping of V-pits after the Fourier transformation shows exact hexagonal lattices as the underlying PSS pattern, indicating that the V-pits are highly correlated to the substrate. The V-pit density was approximately 1.7x10^8 cm2 and a similar order of V-pit density was observed in all other samples.
As shown in the cross sectional TEM images of Figure 2, it can be clearly observed that the TD would extend along the c direction originating from the flat c-plane region of PSS and triggered the V-pit formation from the termination of threading edge dislocation. The presence of V-pits actually started to appear on the surface of SLS layer and the V-pit size increased with increasing the pairs of SLS. In addition, we found that the growth rate of sidewall MQW would gradually increase as the V-pit size expanded.
The power dependent PL measurement for MQWs with from10 to 60 pairs SLS were measured at 10K. As shown in Fig. 3(a), at a lower excitation power, the light emission mainly came from planar MQWs. At a higher excitation power, an extra shoulder peak appeared with a higher emission energy than that of planer MQWs, which was the emission from narrow sidewall MQWs of V-pits. Since these sidewall MQWs on V-pits serve as energy barriers to prevent carriers in planar MQWs into non-radiative recombination centers of TDs, it can be seen that the internal quantum efficiency (IQE) decreases as the barrier energy decreases with increasing the V-pit size on InGaN MQWs as shown in Fig. 3(c) and (d). In addition, the normalized PL emission efficiency of InGaN MQWs with 10 to 60 pairs is shown in Fig. 3(b). It can be found that the efficiency droop of InGaN MQWs started at the onset of the emission from the sidewall MQWs on V-pits. At a high excitation power, those hot carriers with higher energy could have more chances to jump over those energy barriers around the V-pits, leading to the occurrence of serious droop phenomenon of PL emission. As a result, a proper V-pit structure could be engineered to improve the droop behaviors.
It’s quite puzzling to note that the InGaN MQWs with 10 pairs SLS had a smaller pit size and a larger energy barrier but exhibited the lowest IQE value, comparing with 15, 30 and 60 pairs SLS. It implied that the barrier energy of V-pit insufficiently explained the relation of IQE and V-pit size. We then used a more comprehensive two dimensional simulation model for the spatial carrier distribution to understand the influence of TD and the formation of V-pit based on the carrier rate equation and the diffusion effect by carefully considering the diffusion length, carrier generation rate and carrier life time. The spatial distribution of emission intensity for InGaN MQWs with different v-pit size can be obtained as shown in Figure 4. The emission intensity gradually decreases near the non-radiative recombination centers of TD. Despite the InGaN MQWs with V-pit size less than 170 nm had a higher energy barrier, the shorter diffusion distance between the TD center and V-pit boundary tended to reduce the capability for passivating the TDs in the case of 10 pairs SLS. On the other extreme, the thicker sidewall quantum wells along with a larger V-put size would lower local energy barriers and have a smaller emission area, which would reduce the IQE values. The theoretical analysis indicated that the optimized nanoscale structure of V-pit had a higher energy barrier and weaker carrier diffusion effects to passivate the non-radiative centers of TD. In our experiment, an optimized V-pit size of about 220 nm could be obtained in InGaN MQWs with 15 pairs SLS, showing the highest IQE value of 70%.
The commercial GaN LEDs grown on PSS still exhibit lots of TDs. Fortunately, the formation of V-pit structure in InGaN MQWs shall be effective to block carriers to enter TDs to sustain the radiative emission efficiency. We systematically investigated the structural characteristics of V-pits and the relation between emission efficiency and nanoscale V-pits size. Two dimensional simulation model for the spatial carrier distribution could well interpret the influence of TDs on IQE and an optimized V-pit structure with a higher barrier energy could provide better carrier blocking capability to mitigate the efficiency droop. Of course, growth of low-defect density materials is the ultimate cure of efficiency droop, however, similar concept of creating energy barriers around TDs can be applied to UV and green GaN-based light emitters to boost their emission efficiency with a relatively lower fabrication cost.