Nanocomputing

views updated

Nanocomputing

Nanocomputing describes computing that uses extremely small, or nanoscale, devices (one nanometer [nm] is one billionth of a meter). In 2001, state-of-the-art electronic devices could be as small as about 100 nm, which is about the same size as a virus. The integrated circuits (IC) industry, however, looks to the future to determine the smallest electronic devices possible within the limits of computing technology.

Until the mid-1990s, the term "nanoscale" generally denoted circuit features smaller than 100 nm. As the IC industry started to build commercial devices at such size scales since the beginning of the 2000s, the term "nanocomputing" has been reserved for device features well below 50 nm to even the size of individual molecules, which are only a few nm. Scientists and engineers are only beginning to conceive new ways to approach computing using extremely small devices and individual molecules.

All computers must operate by basic physical processes. Contemporary digital computers use currents and voltages in tens of millions of complementary metal oxide semiconductor (CMOS) transistors covering a few square centimeters of silicon. If device dimensions could be scaled down by a factor of 10 or even 100, then circuit functionality would increase 100 to 10,000 times.

Furthermore, if such a new device or computer architecture were to be developed, this might lead to millionfold increases in computing power. Such circuits would consume far less power per function, increasing battery life and shrinking boxes and fans necessary to cool circuits. Also, they would be remarkably fast and able to perform calculations that are not yet possible on any computer. Benefits of significantly faster computers include more accuracy in predicting weather patterns, recognizing complex figures in images, and developing artificial intelligence (AI) . Potentially, single-chip memories containing thousands of gigabytes of data will be developed, capable of holding entire libraries of books, music, or movies.

Modern transistors are engineering marvels, requiring hundreds of careful processing steps performed in ultraclean environments. Today's transistors operate with microampere currents and only a few thousand electrons generating the signals, but as they are scaled down, fewer electrons are available to create the large voltage swings required of them. This compels scientists and engineers to seek new physical phenomena that will allow information processing to occur using other mechanisms than those currently employed for transistor action.

Future nanocomputers could be evolutionary, scaled-down versions of today's computers, working in essentially the same ways and with similar but nanoscale devices. Or they may be revolutionary, being based on some new device or molecular structure not yet developed. Research on nano-devices is aimed at learning the physical properties of very small structures and then determining how these can be used to perform some kind of computing functions.

Current nanocomputing research involves the study of very small electronic devices and molecules, their fabrication, and architectures that can benefit from their inherent electrical properties. Nanostructures that have been studied include semiconductor quantum dots, single electron structures, and various molecules. Very small particles of material confine electrons in ways that large ones do not, so that the quantum mechanical nature of the electrons becomes important.

Quantum dots behave like artificial atoms and molecules in that the electrons inside of them can have only certain values of energy, which can be used to represent logic information robustly. Another area is that of "single electron devices," which, as the name implies, represent information by the behavior of only one, single electron. The ultimate scaled-down electronic devices are individual molecules on the size scale of a nm.

Chemists can synthesize molecules easily and in large quantities; these can be made to act as switches or charge containers of almost any desirable shape and size. One molecule that has attracted considerable interest is that of the common deoxyribonucleic acid (DNA), best known from biology. Ideas for attaching smaller molecules, called "functional groups," to the molecules and creating larger arrays of DNA for computing are under investigation. These are but a few of the many approaches being considered.

In addition to discovering new devices on the nanoscale, it is critically important to devise new ways to interconnect these devices for useful applications. One potential architecture is called cellular neural networks (CNN) in which devices are connected to neighbors, and as inputs are provided at the edge, the interconnects cause a change in the devices to sweep like a wave across the array, providing an output at the other edge.

An extension of the CNN concept is that of quantum-dot cellular automata (QCA) . This architecture uses arrangements of single electrons that communicate with each other by Coulomb repulsion over large arrays. The arrangement of electrons at the edges provides the computational output. The electron arrangements of QCA are controlled by an external clock and operate according to the rules of Boolean logic .

Another potential architecture is that of "crossbar switching" in which molecules are placed at the intersections of nanometer-scale wires. These molecules provide coupling between the wires and provide computing functionality.

The fabrication of these nanoscale systems is also a critical area of investigation. Current ICs are manufactured in a parallel process in which short wavelength light exposes an entire IC in one flash, taking only a fraction of a second. Serial processes, in which each device is exposed separately, are too slow as of early 2002 to expose billions of devices in a reasonable amount of time. Serial processes that are capable of attaining nanometer, but not molecular, resolution include using beams of electrons or ions to write patterns on an IC. Atomic resolution can be achieved by using currents from very sharp tips, a process called scanning probe lithography, to write on surfaces one atom at a time, but this technique is too slow for manufacturing unless thousands of tips can be used in parallel.

It is reasonable to search for nanoscale particles, such as molecules, that do not require difficult fabrication steps. An alternative to the direct patterning of nanoscale system components is that of self assembly, a process in which small particles or molecules arrange themselves. Regardless of the method used to create arrays of nanostructures, organizing the nanodevices into useful architectures, getting data in and out, and performing computing are problems that have not yet been solved.

In summary, nanocomputing technology has the potential for revolutionizing the way that computers are used. However, in order to achieve this goal, major progress in device technology, computer architectures, and IC processing must first be accomplished. It may take decades before revolutionary nanocomputing technology becomes commercially feasible.

see also Artificial Life; Central Processing Unit; Generations, Computers; Molecular Computing; Neural Networks.

Gary H. Bernstein and Wolfgang Porod