Earth's ozone layer completely blocks solar light of very low ultraviolet wavelengths. Biological organisms on Earth have therefore never developed a tolerance for this ‘UVC’ radiation, and artificially generated UVC light has become a useful tool in the treatment and destruction of bacteria, yeast, viruses and fungi. Mercury, xenon and deuterium lamps are currently the main sources of UVC light, but the high operating voltages and large size of such lamps, together with the environmental dangers of using mercury, preclude their use in techniques for disinfection, for air and water purification and in biomedicine. On page 325 of this issue, Taniyasu et al.1 report the potentially revolutionary production of UVC light using semiconductor light-emitting diodes (LEDs) based on aluminium nitride, which could be powered by low-voltage solar cells.

An LED generally comprises a junction between semiconductor materials with two different types of conductivity, electron conductivity and hole conductivity, that are equivalent to the movement of negative and positive charge carriers, respectively. At the junction, electrons and holes recombine, emitting energy in the form of light. Highly efficient blue–green LEDs, using semiconducting aluminium indium gallium nitride (AlInGaN), were initially reported2 in the early 1990s. Thus, for the first time, miniature low-voltage light sources in all three primary colours — red, green and blue — became available, opening up a multi-billion-dollar market for the lighting and display industries. LEDs that emit at even lower, ultraviolet wavelengths, divided into UVA, UVB and UVC, soon followed3,4 (Fig. 1).

Figure 1: Diodes in the solar radiation spectrum.
figure 1

The blue line indicates the approximate intensity of radiation that penetrates Earth's atmosphere at wavelengths from the ultraviolet (UV) through the visible to the infrared. Light at low UVC wavelengths is completely absorbed by the atmospheric ozone layer, so organisms on Earth have developed no tolerance to it. A small-scale, low-power light source emitting in the UVC range could therefore be useful for sterilization applications. Previous work had led to indium gallium nitride and aluminium gallium nitride light-emitting diodes (LEDs) that emit at blue2 and UVB (ref. 4) wavelengths, respectively. But, until now1, difficulties controlling the conduction properties of aluminium nitride prevented the production of LEDs at UVC wavelengths.

The main obstacle to the development of both blue–green and ultraviolet LEDs has been the difficulty in manipulating the conduction properties of various AlInGaN materials that possess the appropriate optical properties. The level and type of conductivity of such materials can be controlled by incorporating impurities into them such as silicon and magnesium, in a process known as doping. Indium nitride (with no aluminium or gallium) absorbs visible and ultraviolet light, and is typically a good conductor. Manipulating its conductivity by doping is also relatively easy. The same is true for gallium nitride, but this compound only absorbs ultraviolet radiation, and is completely transparent to visible light.

Typically, the wavelength of light emitted from a semiconductor LED is nearly the same as that of the light that it absorbs. And as the proportion of aluminium in the alloy increases, so does its transparency down to lower ultra-violet wavelengths. Thus, whereas an LED based on indium gallium nitride emits visible radiation, those based on aluminium gallium nitride and aluminium nitride generate ultraviolet light with a wavelength in the UVB and the UVC portion of the solar spectrum, respectively. But unfortunately, as the aluminium fraction increases, so too does the difficulty of doping. It is hardest of all for aluminium nitride — which is, in fact, an insulator.

There are several reasons why it is difficult to turn aluminium nitride or aluminium gallium nitride with large aluminium fractions into conductors. It requires the incorporation of many impurity atoms to generate additional positive and negative current carriers, but thermal vibrational effects caused by the high-temperature conditions required for growing high-quality crystalline aluminium nitride layers work against this incorporation process. In addition, the layers are deposited over non-nitride materials, such as sapphire, generating a large number of defects that hinder doping. Defects are also generated if the number of incorporated dopant species becomes too numerous, resulting in a self-compensation effect that actually reduces conductivity. Finally, the effect of doping material on the conduction properties can be compensated by that of other materials such as hydrogen gas, used as part of the manufacturing process.

Taniyasu and colleagues' crucial contribution1 is to overcome these obstacles by using a variation of the standard growth conditions. They control the amount of dopant precisely to avoid self-compensation, and use annealing procedures to avoid compensation effects from the reactant gases such as hydrogen. They are thus able to impart to aluminium nitride sufficient conductivity for both positive and negative current carriers. By combining layers of the two conducting types, they fabricated LEDs emitting at a UVC wavelength of 210 nanometres — at present the shortest reported wavelength of any LED when injected with electrical current.

Developments are needed in two areas to improve such UVC LEDs to the point where they can be used in devices: first, an increase in their efficiency by a factor of at least a million; and second, a reduction in their operating voltage to well below the 25 volts used by Taniyasu and colleagues1. The first development will require substantial improvements in the crystalline quality of the aluminium nitride layers; the second will necessitate more-efficient doping to bring about a near thousandfold increase in their room-temperature conductivity. Both challenges are difficult: further bold and innovative research will be required if the promise of UVC LEDs is to be fulfilled.