News & Views | Published:

Nanotechnology

Beyond the silicon roadmap

Silicon will eventually fail to satisfy the 'smaller, faster, cheaper' drive in technology. Nanoscale techniques could take over, and a recent conference reviewed the prospects for computing and electronics.

When and how will nanotechnology make computers work better? At a conference held last monthFootnote 1, Jim Hutchby (Semiconductor Research Co.) predicted that nanoscale electronics would appear on the International Technology Roadmap for Semiconductors1 in 2011–2016 — and would tear it up by 2050.

Existing silicon-based CMOS technology in computers cannot navigate the roadmap as far as the nanoscale. At that scale, quantum effects start to become significant, and even if silicon components were shrunk to these dimensions the result would not be a 'quantum computer': heat dissipation (around 100 W cm−2, according to L. Manchanda, Semiconductor Research Co.) would not only lead to the decoherence of fragile quantum states, it would melt the silicon substrate. So what does the science of today suggest about the computing of tomorrow?

In the short term, the existing von Neumann architecture of today's computers should survive and carry silicon a little beyond its natural shelf-life, thanks to spectacular advances in molecular electronics. For example, binary information can be written into an 'atomic switch' between crossed and nearby nanometre-size wires of platinum and silver if the latter are coated with the solid electrolyte Ag2S (M. Aono, RIKEN). The silver in this material is mobile2, and a few atoms can be electrically induced to migrate to the platinum wire and close the switch. It is possible to combine switches such as these to form logic gates, the building-blocks of computers, and to perform computations.

Alternatively, rotaxane molecules can be used as binary memory bits. Rotaxanes have a molecular ring surrounding a central linear molecule, like a bead on a wire. The molecular ring can be repositioned by an applied voltage3, altering the rotaxane's longitudinal electrical conductivity. To address a nanoscale array of these molecules using platinum wires requires a clever patterning trick. One possibility is to differentially etch the edge of a GaAs/AlGaAs superlattice, deposit platinum, and transfer the platinum nanowires thus formed to another substrate (J. Heath, UCLA). A 64-bit rotaxane-based memory device produced within one square micrometre has now been demonstrated (S. Williams, Hewlett– Packard). This level of bit density lies quite a few years down the silicon roadmap, and could in fact be increased by a few orders of magnitude by incorporating patterning tricks such as the one described above.

Various groups have demonstrated that single-wall carbon nanotubes can also be used to make an impressive variety of room-temperature electronic logic devices (for example, refs 4 and 5). In contrast, DNA has been found to be an unsuitable building-block for computers because it is electrically insulating (C. Dekker, Delft Univ.). Although basic logic functions have now been performed in a range of nanostructures, fault-free mass fabrication is, at least for now, thought to be impossible. So computers that incorporate nanotechnology in the near future will have to have some degree of fault tolerance, and for this, novel architectures are required. One strategy, used in the Hewlett–Packard 'Teramac'6, is to build in a slight redundancy, so that perfect performance can be achieved provided there are sufficiently few faults7.

But what about more radical departures from the conventional architecture? Perhaps phase-based logic will overtake amplitude-based logic (L. Manchanda). This leads us to consider quantum computing. A quantum computer with just a few bits has already been produced experimentally using nuclear magnetic resonance and a solution of organic molecules8. Just one of many recent ideas (J. C. Egues, Basel Univ.) is based on controlling electron spin using gate-induced spin-orbit coupling9. In fact, many possible routes to quantum computing have been suggested, but the most promising are solid-state implementations — most famously, nuclear spins of phosphorus atoms in a silicon matrix10 — because they can be scaled up to generate the massive parallelism required for useful computation.

The counter-intuitive rules of quantum mechanics imply that, unlike classical computers, quantum computers should perform best with a slow clock speed11 (in other words, the devices within them should switch slowly). This necessarily suggests that rapid decoherence of the quantum states encoding information is unacceptable. Entanglement between two solid-state quantum bits, or qubits, was reported for the first time at this conference — albeit with a rather rapid decoherence time of 300 picoseconds (J. S. Tsai, NEC). The qubits used were based on low-temperature superconducting tunnel junctions. Perhaps this breakthrough indicates where the future of quantum computing actually lies.

Another departure from conventional computing philosophy would be to work with light rather than electrons. Whereas the periodicity of the structure of crystalline materials is comparable to the wavelength of mobile electrons, advanced materials techniques enable periodic structures to be produced in which the repeating unit matches the wavelength of light — these are photonic band-gap materials. Complex structures formed with these advanced techniques (for example, ref. 12) could be used to manipulate light and form computer logic gates. Indeed, computer simulations suggest that light can bend around cleverly constructed corners with no discernible energy loss (S. John, Univ. Toronto).

The conference was also presented with some aspects of nanotechnology that could have an indirect impact on tomorrow's computers. For example, the issue of heat dissipation mentioned earlier could be locally monitored using a 75-nm-diameter carbon nanotube containing the liquid-metal gallium, which would act like a nanoscale mercury thermometer13 in the 50–500 °C range (Y. Bando, National Institute for Materials Science, Tsukuba). Or perhaps computerized devices such as gas-specific sensors will rely on specially adapted nanotube elements — such devices can detect as little as 0.02% H2 (G. Gruner, Nanōmix, Inc. and UCLA). Or perhaps computer display devices will use electron emission from nanotubes (P. Legagneux, Thales) such as those shown in Fig. 1 — in fact, Samsung have this year reported a proof-of-principle display unit14.

Figure 1: The conference logo writ small.
figure1

Carbon nanotubes such as these could make both the computers and displays of tomorrow. (Image courtesy Kenneth Teo, Manish Chhowalla, Gehan Amaratunga and Bill Milne, Univ. Cambridge.)

So where will it all end? The graphics of S. John were reminiscent of the best that Hollywood can achieve, and the many talks relating to biology could at times make this physicist feel that he had come to the wrong conference. The nanoscale therefore seems to be the length scale at which science has become truly interdisciplinary. The question is, will it reconcile science fact with science fiction?

Notes

  1. 1.

    *Trends in Nanotechnology, Santiago de Compostela, Spain, 9–13 September 2002.

References

  1. 1

    International Technology Roadmap for Semiconductors (2001); http://public.itrs.net

  2. 2

    Terabe, K., Nakayama, T., Hasegawa, T. & Aono, M. J. Appl. Phys. 91, 10110–10114 (2002).

  3. 3

    Collier, C. P. et al. Science 285, 391–394 (1999).

  4. 4

    Bachtold, A., Hadley, P., Nakanishi, T. & Dekker, C. Science 294, 1317–1320 (2001).

  5. 5

    Huang, Y. et al. Science 294, 1313–1317 (2001).

  6. 6

    Heath, J. R., Kuekes, P. J., Snider, G. S. & Williams, R. S. Science 280, 1716–1721 (1998).

  7. 7

    Kuekes, P. J. & Williams, R. S. Demultiplexer for a molecular wire crossbar network. US patent 6,256,767 (2001).

  8. 8

    Jones, J. A., Mosca, M. & Hansen, R. H. Nature 393, 344–346 (1998).

  9. 9

    Datta, S. & Das, B. Appl. Phys. Lett. 56, 665–667 (1990).

  10. 10

    Kane, B. E. Nature 393, 133–137 (1998).

  11. 11

    Feynman, R. P. Feynman Lectures on Computation (eds Hey, A. J. G. & Allen, R. W.) (Penguin, London, 1999).

  12. 12

    Toader, O. & John, S. Science 292, 1133–1135 (2001).

  13. 13

    Gao, Y. H. & Bando, Y. Nature 415, 599 (2002).

  14. 14

    Chung, D. S. Appl. Phys. Lett. 80, 4045–4047 (2002).

Download references

Author information

Correspondence to Neil Mathur.

Rights and permissions

Reprints and Permissions

About this article

Further reading

Comments

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.