Photonic Crystals Enable Ultrabright LEDs

Faiz Rahman and Richard De La Rue, University of Glasgow

Semiconductor light-emitting devices have almost completely replaced incandescent bulbs in indicators and other low-level-lighting applications. It is widely believed that, over the next few years, ongoing developments will bring about a similar replacement in the space lighting arena. Such developments are being driven by the promise of higher energy efficiency from LED-based light sources.

In recent years, several companies have started commercial production of LED-based light fixtures for decorative and residential applications. The higher initial cost of these commercial offerings is offset by their long-term energy savings. Add to this the much longer life expectancy of LEDs compared with that of incandescent luminaries, and the decision to replace conventional lightbulbs looks even wiser. Many of these LED lighting units use high-power LEDs, such as the Luxeon device from Philips Lumileds Lighting Co. of San Jose, Calif. (Figure 1).

Figure 1. The Luxeon is a high-brightness phosphor-converted white LED from Philips Lumileds Lighting Co. Courtesy of Philips Lumileds.

Manufacturers in the automobile sector have been touting the advantages of solid-state lighting over conventional lighting. All lights found in typical automobiles can be replaced by LED fixtures, except for the front headlights, which are powered by halogen lights. The next generation of LEDs will feature high brightness combined with high efficiency and long lifetime. These devices will make it possible to build proper high-intensity illumination systems for applications such as room lighting and automobile headlights.

It will take a few years to reach that goal through the R&D that is being carried out today. Most of the research effort at this time is directed toward making LEDs brighter.

Untrapping light

Much of the light generated in LEDs never sees the light of day, or the darkness of night, for that matter, because light-generating semiconductors have a substantial refractive index mismatch with the air that traps light generated deep within the device material itself. Total internal reflection at the semiconductor chip-air interface means that most of the light remains confined to the interior of the device. Whereas more than 70 percent of the electrical energy supplied to a typical lighting diode gets converted into light, only about 20 percent of this light manages to come out of the confines of the device as the output radiation. Obviously, this represents a severe waste of energy because it reduces the wall-plug efficiency of LEDs.

Because of their high refractive index, the light confinement problem in LEDs made from gallium nitride and related alloy semiconductors is severe. The situation is similar to that of diamond — a high-refractive-index material that can efficiently trap light. Diamond crystals can be cut and polished to allow light to exit only through well-defined and -oriented facets, lending diamond its fire.

Gallium nitride’s refractive index is even higher than that of diamond (2.67 versus 2.42), so it is even more effective at trapping light. In this case, of course, the light is generated inside the nitride material itself. Interestingly, crystal structuring, albeit of a different type, is possible with light-emitting semiconductors as well and, indeed, this technique allows efficient light extraction in preferred directions.

This way, one can greatly increase the luminous efficiency of LEDs without changing their electrical structure. This means that more of the trapped light is extracted, making the device look brighter while the electrical energy consumption remains the same. Obviously, this is an attractive approach to developing the next generation of super-bright LEDs and, for this reason, many academic and industrial laboratories are busy perfecting techniques for large-scale production of high-extraction-efficiency structured LEDs.

Figure 2. Some of the light produced in an LED is emitted, while some is trapped.

Light generated from electron-hole pair recombination in LEDs can take a variety of routes (Figure 2). Light is principally generated in the thin, optically active quantum well multi-layer sandwiched between the n- and p-type contact and carrier transport layers. Radiation produced in this layer can follow a number of paths within the device and substrate layers, as shown in the illustration. Apart from the light that comes out of the top and bottom faces of the chip, a substantial amount exists as trapped modes in the semiconductor (GaN) and substrate (sapphire) layers. This is shown as dark lines in the figure. The trapped light is equivalent to having horizontal optical modes (shown as yellow sinusoids).

Previous methods have focused on reflecting light from the front or back faces of the chip using metal reflectors. Reflecting light in this way relies on the fact that sapphire is transparent and thus allows light to propagate through it. The result of this approach is a doubling of LED intensity with a relatively simple extension of the fabrication process. But the fact remains that more than half of the generated light still remains trapped in horizontally propagating modes.

Periodic structures

Structuring the top surface of the chip to produce a regular two-dimensional array of holes is now recognized as one of the most suitable techniques for extracting the horizontally propagating trapped modes. Such periodic structures can interact with light in a manner that mimics the way that semiconductor crystals interact with electrons and are, therefore, appropriately called photonic crystals. In much the same way that the periodicity of metallic and semiconductor crystals interacts with the wavelike periodicity of electrons and holes to cause diffraction, photonic crystals also can strongly diffract light.

In much the same fashion that electronic crystals feature a band structure with allowed energy bands separated by bandgaps, photonic crystals display their unique band structures with zones of allowed wavelengths alternating with forbidden wavelength bands. Properly integrated with LEDs, photonic crystals can be used to diffract out trapped light. In addition, they can also tailor the angular emission pattern of the device — making emitter characteristics match their application requirements. Both of these are substantial advantages but require careful design of the photonic crystal structure and its integration with the host chip.

Photonic crystals come in a variety of forms. Opals are naturally occurring three-dimensionally periodic photonic crystals. Their iridescence hints at the possibilities of manipulating light with such structures. Butterflies, too, have natural photonic crystal structures in their wings that lead to interesting visual effects. Photonic crystals for LED applications are different. They are two-dimensional structures that are directly etched into the emitter chip.

The periodic design may have square or hexagonal symmetry, depending on the exact application — or it could even have a quasiperiodic pattern. No matter what the exact design, this photonic structure is formed by the refractive index contrast between the filled and empty regions of the pattern. The higher this contrast, the stronger the action of the photonic crystal in manipulating light. Another important parameter is the depth to which the holes are etched into the chip. Deeper holes make for stronger photonic crystal action but have an adverse effect on the electrical and optical properties of the diode itself.

Figure 3. A backreflecting photonic crystal LED can extract horizontally propagating modes.

Figure 3 shows a typical photonic crystal scheme. Note that the holes that form the empty regions of the photonic crystal must terminate above the light-generating quantum well region for the device to work properly. For operation in the blue-green region of the spectrum, typical holes are 200 nm in diameter and are separated by 300 nm. The interaction of the planar photonic crystal with the confined optical field inside the LED chip causes light to diffract and spill out of the top surface — enhancing the brightness by a significant amount.

Figure 4. This scanning electron micrograph of a photonic crystal lattice shows details of a hole cross section. The pattern was fabricated by electron beam lithography and then dry-etching with a methane-hydrogen gas mixture. Courtesy of Harold M.H. Chong.

A photonic lattice patterned by electron beam lithography and dry-etched into GaN is shown in Figure 4. This close-up shows the cross-sectional structure of the holes. The structure shown here was deliberately made oversized so as to study the profile of the etched holes. As mentioned above, the holes stop short of penetrating the light-generating quantum well region itself.

Crystal fabrication

Fabricating photonic crystals on LED chips is not a simple process because of the small feature sizes and tolerances involved. The problem is only compounded by the fact that the optoelectronics industry typically does not have access to the high-resolution photolithography equipment that is routinely available in silicon fabs. Laboratory demonstration chips have been fabricated with the help of electron beam lithography, which, although capable of extremely high resolution and pattern placement accuracy, is far from a mainstream manufacturing technique. Holographic lithography, which involves making two laser beams interfere to produce a regular pattern of dark and light areas, also has been used to fabricate photonic crystals but, again, this is not suitable for mass production of photonic crystal LEDs.

A contact-based lithography process called nanoimprint lithography has shown more promise as a viable technique for large-scale production of next-generation ultrabright LEDs. This process relies on nanostructured stamps to transfer patterns into special thermoplastic resins. First, a stamp carrying the inverse of the desired pattern is produced on a hard substrate such as metal or quartz, using, for example, electron beam lithography and dry-etching techniques.

The stamp is coated with a mold-release agent and pressed onto the semiconductor substrate that has been coated with a layer of contact transfer resist. Using heat or ultraviolet radiation, the resist polymer is cured so that it cross-links into its hardened form. The stamp is then removed, and the semiconductor wafer is etched in a reactive ion plasma to transfer the resist pattern into the underlying semiconductor substrate.

Nanoimprint lithography produces photonic crystal patterns with high yield and acceptable speed. Several research groups have fabricated photonic crystal LEDs in this way and have demonstrated impressive enhancements in overall luminous efficiency. More production-worthy techniques are being investigated all the time, and one variant involves the use of a cylindrical stamp that can be rolled onto a surface to produce photonic crystals over large areas inexpensively.

Thanks to these developments, the advent of inexpensive commercial LEDs integrated with photonic crystals seems closer than ever. Philips Lumileds already has produced prototype versions of ultrabright blue LEDs using a triangular symmetry photonic crystal. The devices were 1.5 times brighter than similar LEDs without photonic crystals. Sharp Corp. also is developing such devices for use in future large-screen LCD display panels. A number of other companies are involved in similar development activities that are aimed at bringing this new generation of LEDs to market within the next couple of years.

Photonic crystal devices will be the next major offering from manufacturers of high-brightness LEDs. In terms of performance parameters, photonic crystal LEDs are predicted to offer in excess of 25 lm/W, emitting blue light at around 490 nm. Phosphor-converted white LEDs will similarly benefit from ultrabright emitters and provide us with a viable alternative to conventional devices for space illumination.

An LED-lit future is well within sight and, half a decade from now, the combined market for conventional and photonic crystal LEDs could break through the $10 billion mark. Much of this market will be driven by demand from the space lighting, automotive and display illumination markets that will emerge and grow rapidly by the end of this decade.

Meet the authors

Faiz Rahman is a lecturer in the department of electronics and electrical engineering at the University of Glasgow in the UK; e-mail: frahman@elec.gla.ac.uk.

Richard De La Rue is a professor of optoelectronics in the same department.