Gaining light from silicon

Article metrics

Silicon lasers would help computers operate faster by replacing electrical connections with optical ones. The trouble is, normal silicon doesn't glow. Densely packed silicon nanocrystals could be the answer.

Silicon technology is so powerful that it pervades our everyday lives. Silicon chips are in our homes, our cars and even in some people's bodies. These microcircuits help us wake up on time, surf the Internet, and keep elderly hearts beating in a rhythmic fashion. Indeed, the scientific literature contains more than 250,000 papers relating to silicon. But it has been known since the 1960s (shortly after the invention of the integrated circuit) that bulk silicon is extremely inefficient at emitting light, and so could play only a minor role in optoelectronics — the high-speed future of electronic circuits. So, to make lasers and high-speed telecommunications devices, scientists turned to more complex semiconductors, such as gallium arsenide and indium phosphide. These are good at emitting light but are more expensive than silicon and are hard to integrate into silicon microchips. If an all-silicon laser could be created it would revolutionize the design of supercomputers and lead to new types of optoelectronic devices.

The findings reported by Pavesi et al.1 on page 440 of this issue suggest that silicon lasers may yet emerge. If this paper had appeared more than ten years ago, it would have met with disbelief in the semiconductor community, but today it should cause more excitement than scepticism. This is because our perception of silicon has changed over the past few years. It is now known that the properties of silicon are sensitive to its structure at the nanometre scale.

In particular, silicon nanocrystals and highly porous silicon can emit red, green and even weak blue light when stimulated by light of shorter wavelength (Fig. 1). The efficiency of silicon nanocrystals (measured in photons emitted per incident photon) can exceed 1%, which is 10,000 times better than bulk silicon. The origin of visible light from porous silicon has been hotly debated, but for light-emitting diodes (LEDs) made from porous silicon, the efficiency (the ratio of photons emitted per injected electron) has also risen steadily over the past decade2, and now stands at about 1%. This is still too low for practical devices, but is closer to the efficiency of a few per cent needed for integrated light-emitting displays.

Figure 1: First light for silicon.

Ten years ago, researchers managed to get nanocrystalline silicon to emit red light by stimulating it with an argon-ion laser beam (green light). Scientists hope that by the year 2010 they will be able to use a nanocrystalline silicon laser instead of the argon one. Such tiny silicon lasers would revolutionize the growing field of optoelectronics.

There is still some way to go before LEDs made from porous or nanocrystalline silicon are commercially attractive. One of the problems is the nature of the optical bandgap — the gap in the electronic energy levels that controls light emission. Although silicon nanocrystals have an energy bandgap wide enough to produce visible light (2–3 electron volts rather than 1 electron volt in bulk silicon), they retain some of the 'indirect bandgap' nature of normal silicon. This means that light emission can occur only when accompanied by an additional process that transfers energy. As a result, light production in these nanocrystals is slower (microseconds) than that of a 'direct bandgap' material such as gallium arsenide (nanoseconds). This means that processes that generate heat rather than light can compete with the emission of photons and severely limit light output. It also means that silicon LEDs can be switched on and off only at rates of 1 megahertz, rather than the 100 megahertz or 1 gigahertz rates needed for high-speed optical connections3.

Theoretical calculations show that the size of silicon nanocrystals will need to decrease to 1 nanometre to emit blue light as efficiently as compound semiconductors such as gallium nitride. So many researchers are investigating ways of making even smaller and more uniform silicon nanocrystals4. Others have tried sandwiching layers of luminescent nanocrystals between tiny mirrors to make an optical cavity that can amplify the light emission. For porous silicon, this has led to significant narrowing of the emission spectrum, which is one of the features of 'coherent' laser-like light (when all the photons are in step with one another). But so far there has been no improvement in the speed of emission, which keeps the intensity of emitted light low. The best solution would be to convert the normal spontaneous luminescence process within silicon nanocrystals into the sort of 'stimulated emission' associated with lasers. In a laser, each emitted photon stimulates the emission of another photon with the same frequency, resulting in light output being both amplified and coherent.

Which brings us to the remarkable findings of Pavesi and co-workers1. They have created a densely packed assembly of silicon nanocrystals, 3 nanometres wide, embedded in an oxide matrix. This type of silicon nanostructure, like porous silicon, was first produced some time ago5, but its luminescent properties lay dormant until recently. In the experiment of Pavesi and colleagues, the nanocrystals form a thin layer, just below the surface of an oxidized silicon wafer. The authors use two standard methods for measuring the 'optical gain' (when light output exceeds light input), and obtain a remarkably high gain in both cases — comparable to that from assemblies of compound semiconductor nanocrystals, such as gallium arsenide.

The authors attribute this high gain to the extremely high density of silicon nanocrystals in the structures, because their estimates of the gain per nanocrystal are much lower than those of direct bandgap semiconductors. Demonstrating optical gain is a crucial step towards making a silicon laser, but it is not the end of the story. A true laser has to produce coherent as well as amplified light, to achieve the narrowness and intensity required of a laser beam.

Why has optical gain not been seen in the much more intensely studied porous silicon? One difference is clear from the optical absorption spectrum of the nanocrystals, (see Fig. 1 on page 440 ), where there is a feature that has not been reported for porous silicon. Pavesi et al. suggest that their gain is a consequence of the high quality of their nanocrystal–oxide interface, which has many 'surface states' that emit light per nanocrystal. By contrast, in unoxidized porous silicon the luminescence process does not involve such surface states, and is much more dependent on the size and shape of the nanocrystal. As with lasers made from compound semiconductor nanocrystals, high optical gain would require a much more uniform nanocrystal size distribution than etching or Pavesi et al.'s technique has achieved.

What next? To build a silicon laser these silicon nanostructures must be made to work inside an optical cavity, in which mirrors can bounce the light back and forth until it becomes coherent (Fig. 2). The other key feature missing from Pavesi and colleagues' system is the electrical stimulation of light emission. Replacing the existing optical stimulation with an electrical process is essential for silicon lasers to be integrated easily into electronic circuits (Fig. 2). Here, attention may focus on the interconnected nanowire structure typical of porous silicon, rather than silicon nanocrystals, because the electrical properties of the silicon nanowires have been crucial to achieving high LED efficiency2. Future studies of optical gain should also be carried out below room temperature, because the temperature dependence of these processes can yield considerable insight into the gain mechanism, and lasing is normally first achieved at low temperatures, where competing non-luminescent processes are less effective.

Figure 2: How an electrically driven silicon laser might work in the future.

Electrically driven emission of light from semiconductors is the basis of tiny light-emitting diodes (LEDs) and lasers. Traditional semiconductor LEDs are formed from p-type and n-type semiconductors, which donate positively charged 'holes' and negatively charged electrons, respectively, when a voltage is applied across the structure. Recombination of an electron and hole, within the semiconductor, produces a photon and leads to the emission of light. If the efficiency of light emission is high enough and the whole structure is placed between two highly reflective mirrors, the LED can be turned into a miniature laser. Silicon — the semiconductor of choice for the electronics industry — used to be considered a poor emitter of light, but Pavesi et al.1 have demonstrated good optical emission from a layer of silicon nanocrystals stimulated by pulses of ultraviolet light (not electrons). Now the challenge is to electrically stimulate these nanocrystals into producing a beam of laser light.

The work of Pavesi et al.1 and others over the past decade has only recently changed our view of silicon as a poor emitter of light. But if some astrophysicists are correct then nanocrystalline silicon has been emitting light throughout the galaxy for billions of years6,7. They argue that the properties and cosmic abundance of silicon could account for the emission of red light seen by astronomers in clouds of interstellar dust. So silicon nanocrystals may be rather more abundant than previously thought, though it will be some time before such natural structures can be closely examined. In the meantime, once the results of Pavesi et al. have been reproduced, they will, I'm sure, be seen as a milestone in attempts to develop silicon-based optoelectronics. It appears that we can still teach the 'old dog' of semiconductors a few tricks. It just needs to be restructured on the nanoscale.


  1. 1

    Pavesi, L., Dal Negro, L., Mazzoleni, C., Franzo, G. & Priolo, F. Nature 408, 440–444 (2000).

  2. 2

    Gelloz, B. & Koshida, N. J. Appl. Phys. 88, 4319–4324 (2000).

  3. 3

    Goodman, J. W. et al. Proc. IEEE 72, 850– 866 (1984).

  4. 4

    Holmes, J. D. et al. Science 287, 1471– 1473 (2000).

  5. 5

    Nesbit, L. A. Appl. Phys. Lett. 46, 38–40 (1985).

  6. 6

    Ledoux, G. et al. Astron. Astrophys. 333, L39– L42 (1998).

  7. 7

    Zubko, V. G. et al. Astrophys. J. 511, L57– L60 (1999).

Download references

Author information

Rights and permissions

Reprints and Permissions

About this article

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.