Four-pronged Attack Promises Ultra-high Efficiency Lighting
The efficacy of white LEDs is advancing at breakneck speed. In 2004 engineers were proud to quote values of almost 60 lm/W at 350 mA, but today the research record is more than 130 lm/W.
This lighting efficiency, which is almost double that of compact fluorescents, and an order of magnitude better than incandescents, has positioned the white LED as a very promising candidate for general lighting applications. However, further improvements could strengthen the case for solid-state lighting, while leading to substantial energy savings and reduced lighting bills.
Unfortunately, the headroom left for improving conventional white-LED performance is vanishing all the time. The process that turns blue or ultraviolet emission from a chip into yellow light via a down-converting phosphor wastes energy and ultimately limits efficacy. Nichia, for example, estimates a theoretical maximum of 263 lm/W for a phosphor pumped with blue light, and just 203 lm/W for an ultraviolet-sourced equivalent.
Higher efficacies are possible with alternative LED-based schemes, according to Jeffrey Tsao and colleagues at Sandia National Laboratories. This team has shown that it is theoretically possible to deliver efficacies of more than 400 lm/W at a color rendering index (CRI) of 90 by color mixing four carefully selected LEDs emitting at different wavelengths. A CRI of 90 is excellent, says Tsao, and would satisfy virtually all white-light applications.
At first glance this all seems very encouraging, but Tsao makes it clear that getting there is no cakewalk. For starters, electrical-to-optical power efficiencies of almost 100% are required to get close to the theoretical efficacy. And, although infrared lasers with efficiencies of 80% have been made, two of the wavelengths selected by Tsao, 530 and 573 nm, are in a region of relatively poor performance known as the "green gap". Even the best LEDs in this spectral range are incapable of delivering modest efficiencies (figure 1).
Substantial improvements in the performance of green-gap LEDs are obviously required and the first step towards this involves establishing a clear understanding of what is hampering device output. This question is a hot topic of debate within the scientific community and Tsao suggests that various types of defects could be playing a role.
The InGaN/GaN epilayers upon which these devices (and their blue brethren) are based are grown on foreign substrates, such as sapphire and SiC, which cause strain and high threading dislocation (TD) densities that are typically in the range 5 × 108 to 5 × 109 cm–2. Despite these high defect densities, InGaN LEDs can produce external quantum efficiencies of up to 70%. However, some evidence suggests that defects might still limit efficiency: cathodoluminescence studies have revealed that TDs are non-radiative, while calculations have indicated that screw dislocations can induce strain fields that can localize one type of carrier and ultimately restrict radiative recombination.
In addition, point defects, such as gallium and nitrogen vacancies and carbon and oxygen impurities, could act as non-radiative recombination centers. Positron annihilation studies of GaN show that certain defects are incorporated with gallium vacancies, which can limit photoluminescence efficiency. Since point defects are more prevalent at lower growth temperatures – which are required for the higher indium compositions used to fabricate green and yellow LEDs – they may account for the lower LED efficiencies at longer wavelengths.
Green and yellow LED efficiency is also impacted by intrinsic polarization fields, which get stronger with higher indium content. The polarization helps to red-shift the emission, but the benefit is offset at higher drive currents by carrier-induced screening of the internal fields. This means that the emission wavelength varies with temperature, which is a major hindrance for color-mixing approaches.
One promising route that could overcome the polarization-related problems involves a switch to growth on GaN s non-polar planes. Work in that direction is still in its infancy, but researchers at the University of California, Santa Barbara, have made significant progress in the past year and produced devices with external quantum efficiencies of 45%. However, even if this figure can improve substantially, there will still be the problem of wafer size. These devices are made on Mitsubishi Chemical s 1 cm2 substrates, and it is not yet clear whether this particular process could be scaled up to the larger diameters demanded by LED production.
InGaN templates that feature a thin InGaN layer on another material, such as sapphire, could also provide a basis for bright-green LEDs. Compared with conventional LEDs, this material system would suffer less from intrinsic polarization fields, and 2 inch diameter substrates are being sampled by Technologies and Devices International of Silver Springs, MD.
Another problem lurks, however: "As well as the green gap there is a red gap," said Tsao. The most common compound used to access this spectral range is AlInGaP, which is grown on GaAs. In the deep-red spectrum this material is estimated to have an internal quantum efficiency of almost 100% but performance drops rapidly at shorter wavelengths, such as 614 nm – the ideal emission wavelength for the orange-red component of a white-light source.
Tsao believes that this red gap may be an even harder problem to solve than the green one. The material system used for these devices has an indirect bandgap at a high aluminum content, and carriers start to occupy indirect valleys at compositions far removed from this cross-over point. Magnesium-doped layers in high-aluminum-content LEDs also suffer from electron leakage, which reduces internal efficiency. Lastly, the output falls off at the higher device temperatures associated with high-power LEDs and the emission wavelength is very temperature dependent.
According to Tsao, the great deal of effort already expended on addressing these issues has brought very little, if any, improvement. However, he believes that more radical approaches might be able to deliver a breakthrough. One possibility is the creation of hybrid systems that unite AlInGaP with a wide-bandgap material, and another is the development of InGaPN LEDs on GaP substrates.
Back to phosphors
Substantially improving the efficiency of 530, 573 and 614 nm LEDs is going to be tough, to say the least. However, it might be possible to produce an ultra-efficient white-light source by using different combinations of chips and phosphors. "If you could hit a home run by addressing the green gap, through the use of a primary semiconductor like InGaN, then maybe you could use a phosphor for the red, [alongside a blue LED]," explained Tsao.
This approach might seem absurd, since the motivation for moving on from the blue LED and yellow phosphor is the elimination of down-conversion losses. However, energy loss can be minimized if there is just a small difference in wavelength between the excitation source and the phosphor s emission – the case for a green chip and red phosphor.
Tsao and his colleagues have considered several different combinations of chips and phosphors, and made calculations assuming 95% conversion efficiency, less a Stokes down-conversion loss. Using a combination of red and blue emitters and a broad-emission, green phosphor can deliver an overall efficiency of 70% (286 lm/W) if the primary emitters are 80% efficient and the red emission is at 615 nm. But if this source is pushed out to 626 nm, the primary semiconductors must be 90% efficient to hit the same overall efficiency.
What s clear is that it s going to take substantial improvements in LED technology to make ultra-efficient white-light sources, regardless of whether they are based on a combination of different colored LEDs, or a mixture of LEDs and phosphors. Achieving this will not only require advances in the materials themselves, but also improvements in extraction efficiency that could require the development of new device designs. The potential reward for these efforts is a great motivator, but the path to get there looks long and hard.
J M Philips et al. 2007 Laser and Photonics Review 1 307.