A view of the nanoworld

SANDEEP SAXENA

Interview with Eli Yablonovitch, Professor, Department of Electrical Engineering, UCLA.

In August 2002, Intel Corp. announced its new version of the Pentium-4 processor chip, called Prescott. With individual components measuring just 90 nanometres (nm) or 90 billionths of a metre, it heralded a higher level of circuit integration from the present minimum feature size of 130 nm. One nanometre is about 100,000 times smaller than the diameter of a human hair. This new chip is expected to hit the market in mid-2003. It is a watershed development because it marks the transition from microelectronics to nanoelectronics, an important facet of nanotechnology. Nanotechnology refers to the fabrication, study and use of materials, structures, devices and systems in the scale of 1 to 100 nm. With Prescott, nanotechnology would have entered our daily lives.

One of the front-ranking scientists shaping the course of nanoelectronics is Eli Yablonovitch of the University of California, Los Angeles (UCLA).

He visited the Tata Institute of Fundamental Research (TIFR), Mumbai, last December under the Sarojini Damodaran Fellowship Programme instituted by the TIFR Endowment Fund. Yablonovitch's enriching presence was a highlight of the international conference, Photonics-2002, held in Mumbai, at which he gave a plenary talk and a public lecture on the emergent trends in nanoelectronics.

Yablonovitch's current research interests are: optoelectronics, high-speed optical communications, high-efficiency light-emitting diodes and nano-cavity lasers, photonic crystals, quantum computing and quantum communication. Indeed, he is credited with the discovery of `photonic crystals' - the photonic analogue of the electronic semiconductors - which are likely to form an important component of integrated circuits (ICs) of the nanoworld. Frontline's Science Correspondent, R. Ramachandran, spoke with Yablonovitch in Delhi. In a wide-ranging discussion on the emerging nanoworld, Yablonovitch talked about nanoelectronics, nanophotonics and quantum computing and, as he said, "put a real face to it". Excerpts from the interview:

As a scientist intimately connected with the developments in nanotechnology, what is your perspective of the field? There is a lot of talk but where do we stand today?

Most of nanotechnology is hype. Some of it is unrealistic. But the industry is already in the nanoworld. And in a month or two, Intel will start shipping the next version of the Pentium-4 chip, which is going to have features of 90 nm size. And it will just get smaller and smaller after that.

And those things are mass-produced, in huge quantities. So, it is not just a laboratory curiosity. This chip is at present the prime example of mass production in the nanoworld.

Why do you say most of it is hype?

Well, I think everyone knows that we are heading for the nanoworld but sometimes people make promises that are unrealistic. For example, that we are going to have nanorobots that wander through our bodies. I do not think that seems to be on the horizon. On the other hand, we are making many other nanomachines. For example, the nano-IC, I mentioned, which is perhaps the most important.

In the nano-ICs, are the materials and structures identical to what one had at the microlevel? What are the special problems that one encounters at these scales?

With every generation everything must be redesigned again from scratch.

It is getting harder and harder to progress into the nanoworld as we have been doing for the past 44 years since the IC was invented in 1958. Now what is happening is that we are going to continue, but it is not going to be as easy because instead of just technology we are also going to need new science, new scientific insights to make that possible. Up until now it has been much easier; we just went ahead and made things smaller and smaller.

What kind of new science?

In the ICs of today we are printing images that are much smaller than the wavelength of the light used. Now IC printing is essentially a photographic process. A photographic print of the IC is made and the minimum feature size in the prints now is less than the wavelength of light used for photography, which is about 200 nm. People used to think that this was impossible because of what is called the Rayleigh Criterion. But it is possible and it is being done. It is not that Lord Rayleigh was wrong but our interpretation of his criterion was.

I am one of those who believe that photolithography will never end. It will just go on forever in spite of the Rayleigh Criterion. None of the other fancied things like X-ray lithography or extreme Ultra Violet (UV)-lithography, in whose R&D (Research and Development) billions of dollars has been spent, will ever supersede. But it is going to require new mathematics to print things that are progressively smaller.

The reason is that when you print below a wavelength, it is very difficult to know what the negative should look like that would print correctly. In formal mathematical terms it is what is called the Inverse Problem. You know what the final image should be and the problem is to design the negative (or the IC mask) to give that. It is an unsolved problem in image processing.

Just to design the negative will require a lot of mathematics and a lot of software. Mathematics will be a key asset for the nanoworld.

Apart from problems with printing, are there significant changes in the choice of materials, for instance?

I think it will still be based on silicon. But some of the other materials used might change a little bit. Indeed, they have started changing.

Operationally, what kind of new things does one expect in nanoelectronic devices?

A couple of changes are expected in the near future. First of all, everything is going to go wireless. Every laptop, for instance, will have a wireless built in. People will be able to receive the Internet that way. You must have heard about 802.11b. These are standards that make it possible just to go from place to place and always have a wireless link.

The other thing that is expected to be a dominant application in about five years is that voice recognition will be perfected. They expect to have 10,000 times as many transistors as we have today because it is so difficult. But, it is expected that it will be built into microprocessors starting about five years. And that will be a very major development, because people will be able to talk to their computers and the computers will actually be able to understand without making any errors. Another thing is nanophotonic circuits, circuits with not just electrons carrying information as it is in the ICs of today but also photons in the same circuit. And that is going to be a very big step. It will increase the speed of chips by a factor of about a thousand. That is coming soon.

When do you expect to see the entire IC to be photon driven?

See, that is the thing. People believe that the entire circuit will be photonic. I do not believe that. In fact, many people are asking me when are we going to have the optical transistor? I do not believe that we are going to have the optical transistor. What we are going to have is optics coexisting in the same manufacturing process with electronic transistors. So it will be optics and electronics - all in one chip indistinguishable from one another. The logic and memory functions will come from electrons and the telecommunication functions will come from photons. They will coexist in one technology.

How do you make these distinctions on a single chip?

Well, it is really a matter of design. You have to design the chip so that certain parts will perform the electronic functions and other parts will perform the photonic functions. The most important thing in electronics, these days, is not the manufacturing of the chip but rather the designing of it.

It is even more important than manufacturing and most of the value added to the chip - almost 90 per cent - is in design. India is very strong in software and so there is a great opportunity for India to contribute in a major way. Already there are many electronic software design companies that have operations here. Recently, there was a newspaper report that Metrographics is doubling its operations in India. Metrographics is not a graphics company. It is a company that makes software for chips, including the software that makes it possible to go into the nanoworld.

You have been credited with the discovery, or shall we say invention, of the interesting concept of photonic crystals. Will the nanophotonic circuits that you mention be based on these?

Yes. Photonic crystals are going to be used for these nanophotonic circuits.

Can you briefly describe the idea of photonic crystals?

It is the analogue of the semiconductor. But instead of having a semiconductor that controls electrons, it is an artificial semiconductor that controls photons. It is purely the product of human design and human imagination. The first one was created in 1991. The first unsuccessful attempt was in 1987. So it took four years to go from the idea to create the first successful one. Now there are many of these. The concept is being simplified and made more practical every year.

It has reached a point now that we can make these part of the normal integrated circuit process.

So what exactly is it?

It is an artificial crystal where you trap light at various parts of the crystal by controlling photons to form an artificial periodic structure.

What it means is that a trap like that can act as a floater for light, which is very useful for telecommunications because then you can separate different telecom channels. It is a trap and so light intensity can get very large which makes it much easier to modulate light using these. So what has arrived is that all the components that we need for processing optical signals can now be made using normal IC processing. Wavelength modulators, detectors, filters - all the things you need for an optical channel are possible now. This means that you do not have to invent something new. You can just come up with a good design and make it as if it were a normal IC.

When you say localised light what kind of dimensions are we talking of?

The dimensions are the same as the Pentium chips that will hit the market now. The fact that they are going to be 90 nm is exactly the right size for photonic crystals. Just at this very point in history we are finally able to make nanophotonic circuits and these are photonic crystal circuits.

Have any kind of prototypes been already developed as functioning photonic crystal circuits?

No. Some of the individual components have been demonstrated. That is all I am prepared to say as this is a topic of very high competition in the industry.

What specific new advantages do these photonic circuits offer?

Nanophotonics is going to be revolutionary. It will help overcome the present recession in the telecom industry. Part of the reason for the recession is that things got too expensive in relation to the functionality that they provided. And here we are going to make chips for $5 that is going to provide the functionality that costs a million dollars today. The processor frequency or speed will be controlled by the electronic circuits. The photonic circuits will duplicate into many different channels and enable you to transmit over large distances. For example, today you get a Pentium chip that runs at 4 Giga Hertz (GHz). It is expected that three years down the road it will run at 40 GHz. That is going to be one electrical channel. In photonic circuits you can have a thousand of those channels. That means you have a thousand times the signal processing capability.

When you say that nanoelectronic devices of tomorrow would be essentially operated through wireless...

That would be one of the big applications...

Are you saying this because part of it would be photonics-driven?

No. Photonics would be very useful in providing backbone telecom capacity for the wireless nodes. So, we will have very many wireless nodes and wherever we go in most of the built-up world, we shalls have a local wireless service, short-range wireless service. But this is going to require very many optical links to wire up all these wireless nodes and each of them will have to get its information from somewhere. It is called the optical backbone. So the wireless will be backed up by an optical backbone.

Is the light frequency at which these photonic circuits will operate a crucial factor?

The industry has standardised a certain wavelength, which is 1550 nm. So we have to accept that standard.

Why that constraint?

It became a standard because optical fibres were most transparent at this wavelength. But there are new types of optical fibres that are being developed that are transparent at many other wavelengths and eventually we may be able to run it at all wavelengths and I think that is going to happen.

You mean that in principle you could make a photonic crystal at any wavelength?

Absolutely. A very wide range of wavelengths. All the way up to the UV wavelengths. So far we do not have any idea how to make an X-ray photonic crystal. But we can go up to UV wavelengths.

Conceptually how do you figure out that a given structure would be a photonic crystal?

That's the hardest thing. You can have an idea but you could have no idea whether it will be good or not.

But even to have an idea, it would involve, I would imagine, actually being able to calculate mathematically...

Right. If you know what structure it is, you can always calculate. But the difficulty is you know what your objective is. But we have no idea about the structure that achieves that objective. In the early days we just used our imagination and tested it by making it by trial and error. It is still trial and error largely but now they run it on the computer. There is still no magic recipe. It comes from the human imagination and it takes time. Every new category of photonic crystal has taken about four or five years from the conception to the first successful example.

Even then, what is the basis at the starting point?

The starting point is you just take a very good guess. If it does not work you try to change it. Honestly that has been the way it has worked so far. People may be shocked that in the 21st century this is the best we can do. But, there is still a very big role for human imagination. It is always very exciting when you start out, you have no idea whether it has any hope of success.

What kind of experience in other fields gives you this intuition?

There is not much by way of examples except from the semiconductors that nature has created for electrons. Nature did not create any useful semiconductors for light. So, some of the first inspirations were by saying this worked for electrons, may be it will work for light. In some cases they were lucky and in some others not.

Are there not supposed to be some natural systems in which photonic crystals occur?

Yes. That is the case. The gemstone opal, which is white but flashes different colours is nature's attempt to make a photonic crystal. But it is rather incomplete. That's the best that nature could do. Another example in nature is in the colour of butterfly wings. Some of those are also two-dimensional photonic crystals. Again a very incomplete photonic crystal; enough to make colours but not enough to be the equivalent of a semiconductor.

Have these structures been analysed?

Oh yes. They have been analysed. People have looked all over the natural world. In many types of animals including undersea animals, shells, in the bodies, they have looked everywhere. But these are the best two examples that one has - butterfly wings and opals. Even these were discovered only later.

Nature is unable to achieve a perfect photonic crystal because it requires a refractive index greater than 2. And as human beings we have created that.

Ordinary semiconductors have a refractive index greater than 3. But in nature there is nothing that has a refractive index greater than 2.

But all these are basically silicon based, is that right?

Well. The preferred material is silicon. But there are other materials that you can use as well. But it is curious that exactly the same material that we use for an electronic semiconductor is also used for photonic crystal. Nature seems to have decreed that the diamond structure is ideally suited for a photonic crystal and it is amazing that the diamond structure is also the structure of silicon on which electronic semiconductors are based.

You take any structure and can you tell me where does the light localisation actually occur.

Well. These have the property of expelling the light. What you do is you create any kind of a defect inside the structure. May be make one of these holes somewhat bigger (see picture). Then the light gets trapped inside. And each one of those traps can be a filter, a modulator or any other. Most of the functions are based on these traps.

Then you have an optical fibre link connected to the electronic part?

No. No. That's the point. There is no optical fibre. They are made side by side in the same semiconductor. So you don't need an optical fibre to link these. The semiconductor itself is a waveguide for optics.

So in the same structure or pattern I can put in the electronic functions as well as the photonic functions.

Exactly. It is manufactured by the same process in such a way that the manufacturer is not even aware.

Do you also have concepts of doping as in electronic semiconductors?

Yes. It's quite amazing. When you plug one of these holes you add a dielectric material; it is donor dopant. It is completely equivalent to what is donor doping in the electronic case. When you make one of these holes bigger, it is like acceptor doping. So you can do both donor and acceptor doping. Just completely analogous to an electronic semiconductor.

How do the trapping functions differ in the two cases?

They are both perfectly fine traps. You can use either. It is quite amazing. The analogy goes quite far. With doping one has been able to achieve the smallest ever electromagnetic cavities.

What other conceptually new things are emerging?

Well. Very major - and that is also related to the area of photonic crystals - is the photonic crystal fibres. The principle of trapping the light makes them far superior optical fibres. And we heard about this at the Mumbai conference from Philip Russell of the University of Bath. And these fibres are quite amazing because they would probably enable you to go from California to Japan with no amplifiers in between. Light will propogate with so little loss that a signal will go 10,000 km without any amplification.

This is totally amazing.Is the bandwidth also higher in comparison?

The bandwidths can be extraordinarily large. I think what's going to happen as a result of these fibres is that we are going to have bandwidths which people could not dream of a few years ago. Bandwidths may be 40 terahertz in one fibre.

Will these fibres also be silicon based?

Yes. Fibres have been traditionally based on silicon dioxide. And some people believe that this will simply continue. The difference will only be in the design.

Will the principles of variation of refractive index from centre to the periphery, etc., still remain the same here?

The big difference in these new types of fibres is that, since they are photonic crystals, they are able to trap light in an air tube at the centre - a low index in the centre. That's the new thing. Up until now, fibres have always had a high index at the centre. And now with the photonic crystal fibres you can have light trapped in air, in an air tube. That's the reason it goes through such long distances without attenuation. It's going through completely clear substance. That work is becoming very dominant.

Photonic crystal fibres might become the biggest application of photonic crystals. And that's already commercial.

You have also been involved with quantum computing. What is happening on that front?

Quantum information is one of the most exciting things happening in science today. In the quantum world we can store and manipulate far more information - it's so gigantic that you cannot imagine. The storage available in 94 quantum transistors is equivalent to all the hard disks manufactured in the world each year multiplied by the number of years in the age of the universe which is about 10 billion years! Only in 94 transistors.

So this is beyond astronomical. This new type of transistor will store quantum information in an electron spin. At present such a quantum transistor is just an imagination. No one has ever made one like that. And as soon as somebody makes it, it's going to be revolutionary. This is expected. The first example is expected to emerge in the twelve months. This would be the new type of transistor that would manipulate quantum information instead of simple 0s and 1s.

What kind of system is a quantum transistor likely to be?

It's so important that there are many approaches that are being pursued.

The approach I am most interested is to simply use the electron that is already present. I say `the electron' in singular because it would be a transistor that is controlled by one electron. So this is going to be an extraordinary watershed for the 21st century when we make the first transistor for quantum information or quantum computer as it is called.

Since you are referring to the electron in singular, how is the manipulation of a single electron in the chip achieved?

Surpisingly, I shouldn't say surprisingly because it follows from simple reasoning. Since the transistors are getting smaller and smaller, they are so small that they can be influenced by a single electron. So it is part of the nanoworld. It is actually very easy to make a transistor that is controlled by one electron. It is remarkable.

I understand that in terms of concepts many such systems have been described. But to make one system that functions as a quantum computer has been extremely tough. Is what you have described a realisable one?

Yes. It's realisable, in part because we are making these small transistors everyday in mass production. It's not in research, it's actually in commercial mass production. So the capabilities are there to make very many of them. And I think the situation that you have described was the situation that existed up until last year. People were very pessimistic. But now it has become clear, as a result of a lot of progress that has been made, that there are many systems now that look like that they will be quite capable of making quantum information processors. I can tell you what those are: for example, superconductors look very promising, and single photons look promising. But this is a change that has occurred recently. People who were pessimistic a year ago now have their choice of very different physical systems all of which look very good. But I am very partial to electron spins, which to me are the most promising. But there is room for many of these approaches to coexist. I am partial to electron spins mostly because it's so similar to what we are already doing. The great thing about transistors in silicon is that we know how to make great many of them at once. That's the reason for my optimism for that particular approach. There is equal optimism with the other approaches as well.

In terms of q-bits as they are called, how much information can actually be stored in these single electron spin systems?

The amount of information follows from principles of quantum mechanics.

It has not yet been demonstrated that we can store 94 of these spins, the example that I gave you. The biggest that has been demonstrated is until now at most 4. There is a big difference between 4 and 94 because it grows superexponentially. There is a very simple formula which gives us the amount of information that can be stored. It is 2 to the power of 2 to the power of n, where n is the number of spins. Grows very very rapidly. So you put n = 94, it is equal to all the hard disks that will be produced in billions of years. So people are wondering how we are going to get there. That's what makes it so exciting. We are going to get there with new science, new ideas and may be new types of transistors.