R. Stanley Williams
3500 Deer Creek Road, MS 26-U
Palo Alto, CA 94304-1392
The direction of modern technological innovation is clear - smaller, cheaper, faster, smarter. These issues are strongly interelated, and they cut across the broad spectrum of U.S. industries: chemical, measurement, communications, and computation. At the present time, the size scale of electronic technology is significantly below 1 micron, with the current generation of integrated circuits incorporating 0.35 micron features, and the generation of devices scheduled for release at the end of 1998 based on 0.25 micron features. During the past 24 years, the number of transistors integrated onto a single chip has increased by a factor of 16,000, or by a factor of four every 3.4 years. This phenomenal exponential increase in circuit capability is now known as Moore's Law, and it has been at the foundation of the great success enjoyed by many of this country's largest and most profitable companies. Compared to the ENIAC, which was first powered up fifty years ago, the present generation of microprocessors can compute 60,000 times as fast while using just .0001 of the power with very much greater reliability.
Many times in the past two decades the end of Moore's law of scaling has been forecast, with the predicted limiting factors being the increasing complexity of the integrated circuits or the inability of optical lithography to scale to dimensions smaller than the wavelength of light. So far, ingenious engineering and the commitment of significant resources have been able to overcome these limitations by utilizing increasingly sophisticated computer algorithms to aid in the design of the circuits and by applying various optical and chemical tricks to lithography. However, as feature sizes in electronic circuits move into the nanometer size regime, fundamental physics limitations begin to emerge. These limitations include rather obvious issues, such as the impossibility of requiring a fraction of an electron to continue the scaling of a device into the nanometer size range, as well as more enigmatic quantum mechanical issues such as size quantization that opens up significant gaps between allowed energy levels and the ease of electron tunneling through ultrathin insulating barriers.
Thus, if the capabilities of electronic devices are to continue scaling exponentially after the year 2010, which is as far as the Semiconductor Industry Association Roadmap forecasts, there will have to be dramatic changes in device design and operating principles. At this point, one must inquire if absolute limits of electronic circuitry exist. After all, if semiconductor technology is already near those limits, then it makes most sense for industrial organizations to concentrate on software considerations in order to increase the performance and value of their products. However, a simple analysis supplied by Richard Feynman in Lectures on Computation (1996) shows how to estimate the ultimate performance of a Boolean logic computing machine that is limited only by fundamental physical constants and the laws of thermodynamics. This analysis shows that it is physically possible to perform calculations one billion times faster than a present generation microprocessor while expending only one watt of power. Thus, the age of electronic computers has not yet even begun. There are clearly huge rewards in store for pushing into the nanometer size range and learning how to harness quantum phenomena as the operating basis for electronic, optical, magnetic, and chemical technologies.
The critical parameters for the successful introduction of any significantly new product to the market are performance and cost. A new technology must either offer a substantial performance improvement (by perhaps an order of magnitude) at the same cost, or offer the same performance at a greatly reduced cost (by a factor of three) in order to replace an entrenched technology and the infrastructure that exists to support it. Functional nanostructures offer the potential of both performance and cost enhancements, and therefore their market potential is enormous. However, a huge amount of research in nanoscience must precede the successful introduction of nanotechnology, and there will certainly be a significant transition period where both classical and quantum circuitry will be present in the same systems and will have to be compatible.
In terms of performance issues, nanostructures offer the potential for enormous increases in information storage capacity and computational throughput, just because of their packing density and the short information transit times from site to site. They offer greatly decreased voltage and power requirements for all types of optical sources, from high performance communications lasers to general illumination. In order to deliver these benefits, functional nanostructures will have to be fabricated in huge quantities with extremely uniform and controlled size, shape, and composition distributions to optimize their properties for a particular application. They will have to be assembled into larger structures with ever increasing complexity, hierarchy, and functionality. Utilizing the types of subtractive processes used today to create microstructures will be far too costly to manufacture nanostructures. Moore's second law states that the cost of building fabrication facilities (or fabs) for electronic circuits based on lithography and subtractive processes increases by a factor of two every generation. The present generation of logic and memory chips are being built in billion dollar fabs, which means that a fab line for true nanostructures made by current strategies would cost many tens or even hundreds of billions of dollars to construct. On the other hand, additive processes, such as chemical synthesis and self-assembly or guided-assembly techniques, hold the promise of making nanostructures very inexpensively and appear to be necessary for the successful development of nanotechnology.
Thus, from the viewpoint of both performance and cost, nanostructures should revolutionize several industries in the next century.
Functional nanostructures will be first introduced to the market gradually over the next decade. The first applications will be passive nanostructures, such as phosphors, pigments, and photographic emulsions, which are not addressed individually but do respond to an external stimulus based on their properties as individual entities. In fact, in the next couple of years such passive applications of nanostructures will account for several billion dollars in revenues for U.S. companies as they are introduced as improvements in existing products rather than as completely new technologies. Nanostructures will be in general use by ordinary people in the industrialized nations on a daily basis by 2001. In most cases, only a very few corporate technologists will know about their use, since they will be held by their companies as trade secrets and proprietary intellectual property. The general public will see dramatic improvements in familiar products that they purchase, but they will not know about and companies will not advertise these first applications of passive functional nanostructures. This is an unfortunate result of increasingly stiff competition and short product life-cycles, because the importance of nanostructures to the economy will not be widely known or understood by the public or the government. In many cases, academic scientists and the research agencies that fund them will have to take it on faith that the results of their work are having a major impact on the economy, even though it cannot be measured directly, and that nanotechnology is indeed a vibrant field in industry, even though there are few publications and patents.
For active nanostructures, those that will be specifically addressed individually in creating the output of a device, the appearance in the market will be much later and will require the combined research efforts of a large and interactive community. Examples of such applications include
This is where the most visible research will be and where little return on investment will occur for at least the next decade. In academic and governmental circles, the primary attraction will be the intellectual challenge of understanding quantum effects and attempting to express them at ordinary temperatures. There will be repeated cycles of enthusiasm as new discoveries are made and dejection as the time scales for these discoveries to reach the market appear to lengthen (even though they may be in the market already because of the stealth approach being taken by manufacturers). The primary issue for funding agencies will be to provide a substantial and predictable level of funding for the long term, and not to respond to short-term pressures to switch all funding into the latest hot area. In industrial laboratories, the balance will be between the near-term bottom line of the company and the promise of substantial profits to be made by applications of active nanostructures. The major difficulties and slow progress will dissuade some major companies from investing in this area, but there will be a new wave of innovation and invention from small companies, with the creation of a few new giant corporations and the disappearance of a few of the old ones. This is because the use of additive processes for making nanostructures should be cheap enough to allow smaller companies to play seriously in this new arena.
Feynman, R.P. 1996. Lectures on computation. Ed. A.J.G. Hey and R.W. Allen. Reading, MA: Addison Wesley Longman.