SEMICONDUCTORS are commonplace. Without them the modern world would look entirely different. No computer chips, no Internet, no cash machines, not even CD players or transistor radios. Semiconductors have earned their place in everyday life because of the way they influence how electrons behave. From this comes precise, rapid control over electric currents 鈥� and the entire microelectronics industry, from Silicon Valley onwards.
But electrons may not be the only particles that can be controlled in this way. Physicists believe they can pull off the same sort of trick with photons, the particles that make up light. Their hopes lie with exotic materials called photonic crystals which are to photons what semiconductors are to electrons. Theory says the idea should work. Now the experiments are coming along to show that it does.
The practical problems, however, are fiendish. Silicon is a natural material, readily available, but a photonic crystal is an intricate structure that has to be manufactured. It is deliberately constructed in such a way that light with a particular range of frequencies cannot pass through it. But things get really interesting if there is a flaw in the crystal structure. Then light of the forbidden frequency will be trapped at the defect, unable to move anywhere else in the crystal. Trapped light such as this could be the basis for a miniature laser or photonic switch. A line of defects could act as a precision waveguide to transport light from one device to the next.
Advertisement
So how do semiconductors and their optical equivalents work? In a pure semiconductor crystal, there is a particular range of energies that no electrons are allowed to have. This forbidden zone is known as a band gap. All the permitted energy states with lower energy than the band gap are filled. So because electrons can move around the material only if they can swap into different energy states, this severely restricts their movement and hence the material鈥檚 conductivity. Adding individual atomic impurities to the semiconductor creates new, localised energy states in the band gap. These states make it possible to control the silicon鈥檚 electronic properties by defining exactly where and when its electrons can move.
The band gap arises because electrons do not just behave as particles, they also behave as waves that scatter off the silicon atoms. The atoms are arranged in a regular pattern throughout the lattice, and at certain electron energies the waves associated with the scattered electrons cancel each other out. This reduces to zero the probability of finding an electron with that energy 鈥� in other words there is a gap in the energies that electrons can have.
In principle, there is nothing to stop light behaving in the same way. But for this to work, the dimensions of the photonic crystal鈥檚 repeating lattice have to be comparable to the wavelength of the light. Modern optical communications systems tend to operate at near-infrared wavelengths of 1.3 and 1.5 micrometres. So at these wavelengths the lattice would need spaces of around 0.5 micrometres, more than a thousand times the lattice spacing of ordinary crystals.
But although this is large compared with an atom or molecule, it is still only about one-hundredth the diameter of an average human hair 鈥� minuscule compared with the sizes that engineers normally deal with. Even with the lithographic techniques perfected by chip makers, a three-dimensional structure on a scale as small as this is extremely difficult to make. To make the task as easy as possible, the first photonic crystals attempted were for microwaves-wavelengths around 1 centimetre. A photonic crystal that works with microwaves would have to have a lattice that would be measured in millimetres.
In 1991, Eli Yablonovitch of Bellcore, the telecommunications research company in New Jersey, became the first to make a photonic crystal. He did it in the simplest way imaginable. Starting with a solid slab of a commercial material called Stycast-12, he used an ordinary workshop drill to bore three sets of long, slanted holes through the top surface of the block. Yablonovitch chose Stycast-12, which is manufactured by the Massachusetts company Emerson and Cumming, because it is transparent to microwaves. The holes he drilled intersected below the surface to produce an intricate, periodic, three-dimensional pattern, which is what forms the photonic crystal. Just drilling the holes turns the material into a perfect mirror to reflect the microwaves.
Yablonovitch has since moved to the University of California at Los Angeles where, together with Axel Scherer of Caltech, he has been trying to reduce the size of the structure to a matter of micrometres. There are plenty of lithographic and etching techniques that can be used to make tiny holes in solid materials. But the smaller the diameter being aimed for, the harder it is to control, especially for holes more than a few micrometres deep. At a meeting this summer in Crete, Yablonovitch reported that his team has managed to drill micrometre-sized holes through the top few layers of a gallium arsenide slab. They are now trying to refine their methods to make a complete structure.
Criss-crossed
At the same meeting, Gregor Feiertag of the Institute for Microtechnology, a company in Mainz, Germany, reported on a technique he and various colleagues have been working on together with researchers from the Greek Foundation for Research and Technology. Instead of gallium arsenide, Feiertag and his colleagues use a polymethylacrylate polymer commonly used in X-ray lithography. This material has a rather low refractive index, so the contrast between the material and air is not enough to make photons scatter effectively and create a full photonic band gap. So having drilled their holes, in just the same way that Yablonovitch did, Feiertag and his team filled them with a material that has a much higher refractive index. They then burnt away the outer structure, leaving a crisscrossing network of columns that acts as a photonic crystal for light with a wavelength of 200 micrometres. So where Yablonovitch had air, Feiertag鈥檚 group has solid material, and vice versa. So far, Feiertag鈥檚 columns are only a few micrometres long, but the researchers are working on ways to improve this.
Meanwhile, Kai-Ming Ho and colleagues at Iowa State University have been using a different approach. They take thin silicon wafers and cut away thin strips of material on each one to leave a grating. Then they stack these gratings one on top of another with the second layer at right angles to the first. The third layer is oriented the same way as the first but with an offset, and so on (see Diagram). The result is a three-dimensional photonic crystal that excludes light with a wavelength of around 600 micrometres 鈥� the shortest wavelength yet for a working three-dimensional rystal.
In our own work at MIT together with Shan Hui Fan we are using a different strategy. Instead of searching for the ideal structure and then working out whether it can be fabricated at a suitable scale, we are aiming for a photonic crystal that we are confident can be made using existing microlithography techniques. First we plan to deposit two different materials (say silicon and silicon dioxide) layer by layer, using lithography to form two-dimensional patterns that will scatter light from the intersections between the materials. To form the three-dimensional crystal we plan simply to drill holes straight through the layered structure (see Diagram). Because we would be drilling only one set of holes, and these holes could be vertical rather than slanted, this should be easier than other drilling techniques. We have calculated that 10 layers should be enough to produce a photonic band gap and are working on a prototype crystal using this design that would operate at a wavelength of 1.5 micrometres.
Fine tuning
After we succeed in making a photonic crystal with the appropriate band gap, the next step will be to add or remove some material to create a defect in the structure 鈥� a microcavity in which to trap light. Adding material should trap light whose energy lies near the top of the photonic band gap, while removing material would trap light closer to the bottom. We expect to be able to fine-tune the frequency by changing the size of the defect.
Microcavities open up the possibility of building extremely efficient microlasers and light-emitting diodes. To see why, it is necessary to understand what happens when a few atoms are placed inside the cavity. The electrons in the atom can occupy certain clearly defined energy levels, and if the atom absorbs energy one of its electrons can be excited into a higher energy level. As it then drops back to its original level, it will give out light, whose frequency corresponds to the energy difference between the two levels. In a laser, some initial photons, emitted spontaneously in this way, stimulate other excited atoms to emit photons in a coherent cascade of light. (see Diagram).
The efficiency of a laser depends on the probability that excited atoms will emit light of the right energy. This probability is related to the number of different shapes that the photon鈥檚 associated wave could adopt if it was emitted 鈥� the more different shapes, or 鈥渕odes鈥�, it can choose from, the more likely it is to be emitted. In free space, the possible modes for a photon with a given energy are spread out over a very large volume. Confining the photon dramatically increases the density of the modes 鈥� it鈥檚 as if all of the modes spread out over the whole of free space were now squeezed into this one small box. This makes it much more likely that the atoms in the cavity will emit a photon rather than staying in their excited states. In a microcavity, virtually all the absorbed energy should be emitted at the required frequency, so the efficiency should be close to 100 per cent. These devices should therefore run on very little power. Because the cavities are so small, large numbers of them could be coupled together onto a single photonic integrated circuit to produce a compact, efficient laser. The frequency of the light should be adjustable by tailoring the cavities.
We have yet to make a three-dimensional photonic crystal working at the right wavelengths, but we can still go some way towards making a microcavity that would work this way. The plan is to use photonic crystals in one or two dimensions, which are easier to make than the three-dimensional kind, and to use the thoroughly traditional phenomenon of total internal reflection to confine the optical modes in the remaining dimensions. Total internal reflection occurs when light travelling through a material with a high refractive index, such as glass, strikes an interface with a material whose refractive index is lower, such as air. If it strikes at an angle that is shallower than some critical value (which depends on the exact materials), all the light is reflected back into the glass. This is how light is confined in an optical fibre: the core of the fibre has a higher refractive index than its cladding, so light stays inside, as long as the fibre is not bent beyond the critical angle.
Bridge of holes
At MIT, our team has designed a microcavity that would combine a one-dimensional photonic crystal with a more traditional two-dimensional waveguide. The waveguide will be mounted on two blocks and will form a microscopic bridge between them, surrounded by air. Because air has a much lower refractive index than silicon, total internal reflection should occur within the waveguide. Drilling holes straight through the waveguide creates a one-dimensional periodic structure that has a one-dimensional photonic band gap. Introducing a defect by, say, missing out one of the holes would trap light of a single defined frequency. Computer simulations show that the light would be very tightly confined.
Prototypes of such a structure, but without the defects, have already been made at MIT by Leslie Kolodziejski and colleagues. They deposited a 1 micrometre layer of silicon dioxide onto a silicon substrate, followed by a 0.5-micrometre layer of silicon. They then etched the silicon and silicon dioxide layers away from two parts of the slab leaving a thin strip of material in between the etched regions. The next step was to remove the silicon dioxide selectively from the bottom of this strip, leaving a silicon bridge, half a micrometre thick and 10 micrometres long, hanging in thin air. Finally they etched a series of four holes vertically into the silicon bridge, to form their one-dimensional photonic crystal. They are now in the process of making and testing similar structures, which incorporate defects that can act as microcavities.
Another line of research that we are investigating at MIT is to build a device that can pass a small amount of light from cavity to cavity in a controlled way. If the silicon bridge is short enough 鈥� in other words if we don鈥檛 use too many holes 鈥� it turns out that some of the light can leak out at the edges. This could be useful for 鈥渨iring up鈥� optoelectronic integrated circuits, in which hundreds of components may be packed together on a single chip. Light must be able to travel around these devices, from one component to another, without too much of it being lost en route. Conventional waveguides are very good at guiding light along a straight line, but they are not as efficient when it comes to a bend. Some waveguides cannot cope with curves tighter than 1-millimetre radius. This doesn鈥檛 sound too restrictive until you realise that computer chips are only a few millimetres wide. On that scale a 90掳 curve of 1 millimetre radius becomes a very broad sweep indeed. Computer simulations predict that photonic crystals could come to the rescue. A narrow channel made up of defects in a photonic crystal should make an excellent waveguide, even around tight corners, because there is no way the light in the channel could leave. A channel would be created by spoiling the periodicity of the crystal, either by adding extra holes or by filling up some of the holes with a different material.
In the future, photonic crystals could be used to build super-powerful computers based entirely on light. In today鈥檚 optoelectronic circuits, electronic components still do most of the work. But processing would be much quicker in an optical system. Because optical devices are efficient, they will require less power to operate than their electronic counterparts. And whereas decreasing the size of the circuits in electronic devices tends to increase the resistance, and hence the waste heat they generate, there is no such problem with optical systems. Furthermore, optical circuits can carry many different signals on a single line, so they can process a large amount of information very quickly.
There is as yet no multipurpose optical component like the electronic transistor. But all-optical chips, containing devices made from photonic crystals, could eventually provide one.