GaN-on-GaN LED technology can deliver better light quality in terms of color rendering and accurate bright whites while also providing a lumen-per-wafer advantage over conventional LEDs, write AURELIEN DAVID, FRANK SHUM, and MIKE KRAMES.
Gallium nitride (GaN) LEDs based on sapphire and silicon-carbide substrates have advanced over the last 20 years but generally still fail to meet the light quality aspects we associate with natural lighting (daylight and incandescence). The shortcoming is partly due the LED industry's focus on increasing luminosity at the expense of aspects such as color and whiteness rendering, beam pattern, and total light output. A different approach would be to develop high-efficiency LEDs and leverage the performance to provide the quality aspects of natural lighting. Let's discuss ways to characterize light quality and how it can be delivered by LEDs built on homogeneous substrates.
The latest GaN-on-GaN LEDs will afford the headroom from a performance perspective to allow a focus on better light. The third-generation devices exhibit peak wall-plug efficiency of 80% (at 85°C), and 65% at 150 A/cm2, a current density five (or more) times higher than for LEDs based on foreign substrates. The practical benefits of this recent performance leap are twofold. First, it enables a compelling LED-based solution in systems running at high temperature and high current densities, where conventional LED technology has struggled. And the headroom provided by increased efficiency can be spent on light quality.
Indeed, Quality of Light (QoL) must be considered as an essential property of light sources. QoL has many aspects, but perhaps the best known is color fidelity, measured by the color rendering index (CRI) Ra and the special index R9. The latter is important for deep reds but also, crucially, the tones of human skin. Most LED products aim moderately for Ra in the range of 80 and R9 above 0. This target is based on the often-held belief that such values are sufficient. On the other hand, higher-end options are also available with values as high as Ra = 95 and R9 = 95. Such higher color rendering requires photons at the ends of the spectrum (violet and deep red) and therefore comes at cost of efficiency. The superior efficiency of GaN-on-GaN technology will be crucial in supporting high CRI, as the efficiency headroom more than makes up for the penalties inherent at the high QoL levels.
Optical brightening agents
An equally important and often disregarded measure of light quality is whiteness rendering. While rendering of the color white may seem trivial, this is not the case due to the decades-old and near-ubiquitous use of optical brightening agents (OBAs) in manufactured white products. OBAs absorb short-wavelength light present in natural daylight, as well as in halogen and incandescent lamps, and convert the energy to blue light. The OBA essentially causes a color shift that is perceived as a bright and clean white. Many white objects including fabrics, paper, plastics, and others contain these whiteness-enhancing OBAs.
The OBA trick doesn't play well with many LEDs. Conventional LEDs use primarily blue emission and do not deliver short-wavelength radiation. They are thus unable to excite OBAs, making white objects appear dull and yellowish. A solution is the use of high-efficiency violet pump LEDs that correctly excite OBAs and render whites accurately. The amount of violet light can be precisely tuned to emulate the whiteness rendering of natural light sources.
Fig. 1 depicts the color science behind the use of OBAs. The material in question converts the short-wavelength energy as shown in the graph on the left. (For background information on color science see the multipart series that concluded in LEDs Magazine last year.) The figure compares how the human visual system would perceive the color of commercial white paper illuminated by different light sources with identical 3000K CCTs. The example compares a halogen source, a conventional LED (#1) with no OBA excitation, and a high-QoL LED (#2) with OBA excitation. The violet energy in LED #2 is tuned to match the halogen lamp and the pair result in color uniformity within a 1-step MacAdam ellipse (also called SDCM or standard deviation of color matching). LED #1, meanwhile, results in a 7-step color error — meaning that the paper appears cream colored rather than bright white.
In fact, not all high-CRI LEDs perform similarly when you consider all aspects of QoL. Table 1 compares GaN-on-GaN, -sapphire, and -silicon-carbide LEDs side-by-side for similar 800-lm flux levels. The data came directly from the manufacturer datasheets in the case of the Cree and Nichia LEDs. We see that for similar input power and efficacy levels, the QoL for the violet-primary GaN-on-GaN LED is much higher. The comparison shows that the raw performance of the GaN-on-GaN LED can overcome the inherent efficiency penalties of using a primary violet emitter resulting in Stokes' loss and the use of deep red phosphor that radiates some energy outside the human visual sensitivity range. The GaN-on-GaN LED delivers the QoL aspects we associate with natural lighting.
Does such superior color and whiteness rendering truly matter to users? To assess this, Professor Kevin Houser at Penn State University recently organized a study comparing the perceptual response to two LED MR16 lamps. The study included a conventional LED lamp (Ra = 84, R9 = 24), and a high-QoL LED lamp (Ra = 97, R9 = 88). Participants were presented with two side-by-side booths displaying a retail environment. They were asked to make visual comparisons and indicate their preference for various colors. The results, summarized in Fig. 2, show that users easily perceived a difference and strongly preferred the high-QoL option — both for colors and whites.
The crucial importance of QoL for the broad adoption of energy-saving lighting technology is clearly exemplified by the inability of compact fluorescent lamps (with their mediocre QoL) to win over end users/customers. Hence, it is perhaps surprising that the QoL topic has not been a subject of focus from more LED manufacturers. The good news is that the disruptive performance leap offered by native GaN substrates finally enables high-performance products with uncompromising QoL.
LED surface area
Another aspect of high lumen density enabled by GaN-on-GaN is its ability to deliver small LED source sizes while maintaining high light output or brightness along with high QoL. This aspect of a light source is critical when narrow beams of light are required. It's generally easier to control the beam from smaller sources.
Let's consider the source size issue relative to a replacement lamp that is difficult to achieve with LEDs. The AR111 is a professional lamp that cannot compromise on light quality and requires very narrow beam angles. Until now, there have been few LED-based AR111 lamps on the market and those that do exist suffer from the typical LED problems of poor beam quality, low Ra, poor color uniformity, multiple shadows, and form factors too big to fit into existing fixtures.
Soraa recently introduced a GaN-on-GaN-based AR111 LED lamp with exceptional center beam candle power (CBCP), CRI Ra and R9 of 95, whiteness rendering capability, a single source beam, and mechanical design that fits within the ANSI envelope. Table 2 compares the Soraa product with some commercially available LED-based AR111 lamps and with a Cree reference design based on the company's XLamp MT-G2 LED that was published in January 2012.
The advantage of the smaller sources is evident in the CBCP of 30,000 achieved by the Soraa lamp. The design delivers on all aspects of QoL. And the small source helped the lamp designers maintain compatibility with the ANSI-specified form factor.
Despite the quality advantages evident in GaN-on-GaN products applied in demanding restaurant and high-end retail applications (see the December 2013 cover story of LEDs Magazine at http://bit.ly/1ghtX3o), some in the industry have questioned the cost of the LED technology. Conventional wisdom dictated that for GaN-based LEDs, the benefits of the native substrate would be small in consideration of the much higher prices of free-standing GaN substrates compared to sapphire, silicon-carbide, or silicon substrates.
For an accurate cost picture, however, examine the case for GaN on GaN relative to higher power densities and a reduction in $/lm achieved by an approach that yields more lm/wafer. It's an established fact that GaN substrates offer much higher reliable operating power densities. Indeed, these substrates are used in the Blu-ray industry for laser diodes operating at greater than 1000 A/cm2.
The power density directly enables more lm/wafer. Fig. 3 shows the estimated lm/wafer for GaN-on-GaN technology, compared to GaN-on-sapphire, -silicon-carbide, and -silicon approaches. The LED images in the left of Fig. 3 depict the difference in chip size. Even in LEDs with a similar light-emitting surface (LES), the actual area of the emitters in the array is far smaller in the GaN-on-GaN case. Both the Cree MT-G2 pictured on the right and the Soraa array used in the company's MR16 lamps have a 6-mm LES. But the area of the actual emitters, representative of the wafer area required to produce the requisite lumen level, is almost five times greater in the Cree LED.
The projections in Fig. 3 are based on total wafer area of 12 mm2 to produce 800 lm for GaN-on-silicon-carbide LEDs, and take into account the higher lumen density for GaN-on-GaN LEDs. Lumen density for a GaN-on-sapphire LED is assumed to be the same as GaN on silicon carbide, while a 10% efficiency penalty is included for GaN on silicon, due to the inherently higher defect density for that approach. GaN-on-sapphire and -silicon-carbide technologies offer about 1.2 million lm/wafer for 6-in. substrates. GaN-on-silicon technology offers the promise of 1.9 million lm/wafer for an 8-in. substrate, although the technology has not been commercialized to this level yet. In contrast, 4-in. GaN substrates that are available today deliver 2.7 million lm/wafer, already more than the potential for 8-in. GaN-on-silicon wafers. Future 8-in. GaN-substrate technology, which will be driven by both the LED and power electronics markets, will deliver more than 10 million lm/wafer — five times more than 8-in. silicon.
GaN-on-GaN technology was introduced in 2012 to deliver the world's first single-beam 50W-equivalent LED MR16. Within two years, it has advanced to offer the same efficiency levels as traditional GaN-based LEDs, while enabling much higher light quality. GaN on GaN also offers several times higher power density and a considerable lm/wafer advantage.
AURELIEN DAVID is senior principal scientist, FRANK SHUM is vice president of LED lamp products, and MIKE KRAMES is CTO at Soraa, Inc.