Light Creation Materials
Different families of inorganic semiconductor materials can contribute to solid-state white lighting. The primary chemical systems used for LEDs are Group-III nitrides and Group-III phosphides. It is, however, possible that breakthroughs in a different material system, for example ZnO, will be important.
LEDs based on gallium nitride (GaN) and ternary alloys with indium (InGaN) and aluminum (AlGaN) as well as quaternary alloys (AlGaInN) can span the entire visible spectrum. The current applications for SSL utilize InGaN structures to produce high brightness blue and green light; longer wavelength light can be efficiently generated by AlGaInP LEDs. UV light from AlGaN LEDs could also be used to pump RGB phosphors, as mentioned above.
Nitride materials are usually grown by Metal Organic Vapor Phase Epitaxy (MOVPE)—also referred to as Metal Organic Chemical Vapor Deposition (MOCVD)—from organometallic sources (e.g., trimethyl-gallium, -indium, or -aluminum) and an excess of ammonia. A major difficulty is the lack of low-cost, single-crystal GaN to use as a growth substrate. Group-III nitrides are normally grown on poorly matched sapphire (lattice mismatch +16%, thermal expansion mismatch +39%) or more expensive silicon carbide (lattice mismatch -3.5%, thermal expansion mismatch -3.2%) substrates. As a result, the films have a great number (>108/cm2) of dislocations and other structural defects, resulting in defect-mediated nonradiative recombination of electron-hole pairs and reduced mobility because of carrier scattering from charged defect centers. An intermediate buffer (or nucleation) layer is normally grown at reduced temperature between the substrate and the n-GaN layer (a GaN nucleation layer on sapphire substrates, or AlN on SiC). This low temperature buffer reduces defect densities from up to 1012 to 109/cm2. Further defect reduction (by roughly two orders of magnitude) can be achieved by substrate patterning techniques such as Epitaxial Lateral Overgrowth, Pendeo Epitaxy, or Cantilever Epitaxy; these approaches rely on spatial “filtering,” terminating, and/or turning of threading dislocations, so they do not reach the device active region.
GaN is readily n-doped with Si (usually using a silane source). However, p-type doping with Mg (usually using the metal organic precursor bis-cyclopentadienyl magnesium, Cp2Mg) is much more difficult because of passivation by hydrogen during growth, and the magnitude of the hole ionization potential associated with Mg. Depassivation of the Mg acceptors is achieved by thermal annealing or low-energy electron beam irradiation.
Indium incorporation pushes emissions to longer wavelengths; indium fractions greater than 20% are required for green LEDs. This represents a significant challenge in material growth. Low temperatures are required for In incorporation because of lower thermal stability, leading to poorer material. As In composition increases, lattice-mismatch strain also increases, leading to a variety of strain-induced defects (e.g., point-defects, v-defects, and carbon and oxygen impurities) and lower optical efficiencies.
The (AlxGa1-x)0.5In0.5P alloys are nearly lattice matched to GaAs, and production of LEDs emitting from 555 nm (yellow-green) to 650 nm (deep red) is a relatively mature technology. The availability of single crystal GaAs substrates enables growth of high-quality phosphide material by MOVPE. But the bandgap of GaAs is 1.42 eV (870 nm) at room temperature, so this substrate absorbs emitted light below this wavelength, greatly lowering the LED efficiency.
One way to prevent absorption of emitted light by the substrate is to insert a reflective structure between the LED active region and substrate. The mirror structure is a Distributed Bragg Reflector (DBR), consisting of many (e.g., 5 to 50) alternating high-refractive index and lower-refractive index layers. Because of the differences in refractive indices, a portion of the downwardly directed light is reflected upward (and out of the device) at each layer interface. The mirror stack thicknesses are adjusted so that all of the reflected waves are in constructive interference. The DBR is highly effective for light incident normal to the DBR plane; glancing incident light, however, is only partially reflected by the DBR.
Another approach is to remove the GaAs substrate after the epitaxial layers have been grown, and then to bond the remaining structure to a transparent GaP substrate. The resulting structure is a wafer-bonded LED. Total light extraction from the wafer-bonded AlGaInP LED can be more than a factor of two greater than the LED + DBR design.
For low Al fraction, the internal quantum efficiency approaches 100%. (AlxGa1-x)0.5In0.5P is a direct bandgap semiconductor for x < 0.5; above that composition, it is an indirect-gap material. The crossover occurs at bandgap energy of about 2.23 eV (555 nm). Thus, the AlGaInP LED quantum efficiency drops precipitously at shorter wavelengths because of the approach of the direct-indirect bandgap crossover. That is, as the indirect-gap X-band becomes more populated, the radiative lifetime increases, allowing other nonradiative processes to become more dominant. Efficiency also drops at higher drive currents and operating temperatures because of poor carrier confinement in heterostructures as the direct-indirect-gap crossover is approached.
ZnO-based alloys are another possibility for generation of light from the visible to the near-UV. ZnO has a number of physical properties that make it a good potential candidate for SSL. However, progress toward making it a practical material is still at an early stage.
The material is a wide-bandgap semiconductor (3.4 eV, comparable to the 3.5 eV of GaN) with a wurtzite crystal structure. Single-crystal ZnO can now be produced, and commercial 2-inch wafers are available, offering the possibility of homoepitaxy. The material can be etched by wet chemical means, making it relatively easy to process. Because ZnO has a high exciton binding energy (60 meV, compared to less than 30 meV for GaN), higher operating temperatures are possible.
ZnO has a high intrinsic n-type conductivity, the source of which is not known. It has been difficult to obtain p-doping; although there has been good recent progress, consistency is still hard to achieve. Growth of high-quality ZnO films and heterostructures is still being developed. It may be possible to tune the bandgap of ZnO by alloying with MgO (7.9 eV bandgap) or CdO (2.3 eV). However, these two oxides have cubic crystal structures, so it may be difficult to add large fractions to ZnO without introducing dislocations. Further, the use of the heavy metal Cd in commercial LED structures may not prove acceptable because of long-term safety and environmental issues.