A nanopyramid structure for monolithic white-light LEDs

The majority of white LEDs are currently made using a phosphor coating that converts emissions from blue to yellow. Luminous efficacy is limited in devices like these because, despite the high efficiency of blue LEDs, energy loss is unavoidable in an optical down conversion (from higher- to lower-energy photons). A monolithic white LED that emits multiple spectra—such as blue-yellow or three primary colors—has been considered an ultimate solution to this problem. However, the efficiency of an indium gallium nitride (InGaN) blue LED decreases with increasing wavelength, whereas the efficiency of an indium gallium phosphide (InGaP) red LED decreases as the wavelength decreases. This leads to an absence of efficient green and yellow emitters (the so-called green gap), making it difficult to achieve monolithic white light.

LED researchers have long believed that replacing the inefficient Edison lightbulb with LEDs would result in lower electrical usage, thereby decreasing the need for new power plants. However, the cost is currently too high for this to be a viable solution. When it comes to reducing cost, increasing luminous efficacy is much more effective than reducing the manufacturing cost. The luminous efficacy of phosphor-converted white LEDs that are made today well exceeds 200 lumens per watt (lm/W) and keeps increasing, but the theoretical maximum efficacy is limited to 263lm/W by two factors. Firstly, when a wavelength is converted from blue to yellow, there is an energy loss (Stokes shift) of ∼20% primarily owing to heat loss. Secondly, as luminous efficacy is a measure of how well a light source produces visible light and stimulates human vision, the absence of green emission—for which the eye has the strongest sensitivity—makes the efficacy low. If these two problems were solved, the maximum efficacy would exceed 400lm/W. Thus, the challenge is to achieve efficient emission in the green and yellow spectral ranges, enabling the production of monolithic white LEDs.

The green gap of InGaN LEDs arises from piezoelectric fields. These fields induce a spatial separation of the electron and hole wave functions in the well, decreasing the radiative recombination rate. Using InGaN with higher indium content results in longer wavelengths and increased strain because of a reduction in the band gap and a larger lattice mismatch, respectively. Since the piezoelectric field is proportional to strain, this causes the field to increase. Using GaN on semi-polar (or non-polar) sapphire or a free-standing non-polar GaN substrate to relieve the piezoelectric field has been extensively studied. However, the former suffers from stacking fault defects whereas the latter is too expensive for LED production. By contrast, GaN nanostructures provide facets for semi- or non-polar growth directions and their small features are expected to be helpful in reducing strain and crystal defects. Accordingly, they have attracted a great deal of attention as another possible route toward closing the green gap and providing the added benefit of enhanced light extraction.

To produce efficient emissions in the photoluminescence spectrum between green and red, we grew InGaN/GaN multi-quantum-wells (MQWs) on nanosized GaN hexagonal pyramids formed via selective growth on patterned c-plane (polar) GaN templates. The emission wavelength is controlled by the growth parameters, including gas flow ratio, temperature, and nanopyramid spacing. Figure 1 shows the photoluminescence spectrum from blue to red. The internal quantum efficiencies of the green, yellow, and red emissions are highly enhanced and measured to be 61%, 45%, and 29%, respectively. The high efficiency is attributed to effective suppression of the piezoelectric field by semi-polar growth direction and elastic strain relaxation.¹ Reduced crystal defects were evidenced by our transmission electron microscopy study. We also identified that the emission can be characterized by the quantum dot, wire, and well properties.²

Figure 1. Photoluminescence spectra from indium gallium nitride multi-quantum-wells on a nanopyramid structure. Insets are photoluminescence images of blue, green, yellow, and red emissions. Eye sensitivity function is overlapped for reference. a.u.: Arbitrary units.

An interesting feature of the nanopyramid structure is that the composition of selectively grown InGaN can be varied not only by the growth parameter, but also by pyramid spacing. This has very important implications: the wavelength of InGaN on nanopyramids is different from that of InGaN on a planar substrate, which makes it possible to realize multiple-wavelength emission. We achieved a white emission composed of blue and yellow peaks—see Figure 2(a)—using InGaN MQWs on a nanopyramid structure with partial openings for planar growth: see Figure 2(b). This micro-photoluminescence study revealed that the blue and yellow peaks are from InGaN MQWs on the planar area and nanopyramids, respectively. The enhanced presence of indium species—owing to the lateral surface migration of indium on the selective growth mask—causes a longer wavelength emission from the nanopyramids. Finally, we fabricated LED devices with the hybrid structure: see Figure 2(c). Using this LED structure, the color temperature and color rendering index can be controlled by altering the ratio between the nanopyramid and planar areas, and the spacing of the nanopyramids.

Figure 2. (a) Photoluminescence spectra from the hybrid structure. For a comparison, a spectrum of a phosphor-converted white LED (PL) is shown as a dotted line. (b) Scanning electron micrograph of nanopyramids and the microplanar hybrid structure for a white LED. (c) Optical microscope image of a white LED made from the hybrid structure.

In summary, our efforts highlight the tremendous potential of LEDs built from nanopyramid arrays. High luminescence efficiency can be delivered over a very wide spectral range owing to the combination of reduced defects, relaxed strain, and a suppressed piezoelectric field. However, challenges still remain, the biggest among which is in realizing uniform current flow in the InGaN layer; it currently crowds through the shortest paths due to the 3D geometry of the structure. We are also working to lower the operating voltage using high p-type doping.