Nature 354, 515 - 518 (26 December 1991); doi:10.1038/354515a0

A silicon neuron

Misha Mahowald^* & Rodney Douglas^†‡

^*Computation and Neural Systems Laboratory, California Institute of Technology, Pasadena, California 91125, USA
^† MRC Anatomical Neuropharmacology Unit, University of Oxford, Oxford 0X1 3TH, UK

BY combining neurophysiological principles with silicon engineering, we have produced an analog integrated circuit with the functional characteristics of real nerve cells. Because the physics underlying the conductivity of silicon devices and biological membranes is similar, the 'silicon neuron' is able to emulate efficiently the ion currents that cause nerve impulses and control the dynamics of their discharge. It operates in real-time and consumes little power, and many 'neurons' can be fabricated on a single silicon chip. The silicon neuron represents a step towards constructing artificial nervous systems that use more realistic principles of neural computation than do existing electronic neural networks.

References

1.	Murray, A. F. & Smith, A. V. W. Electronics Lett. 23, 642−643 (1987).
2.	Linares-Barranco, B., Sánchez-Sinencio, E., Rodríguez-Vázquez, A. & Huertas, J. L. IEEE J. Solid-state Circuits 26, 956−965 (1991). | Article |
3.	McCormick, D. A. in The Synaptlc Organization of the Brain, 3rd edn (ed. Shepherd, G.) 220−243 (Oxford University Press, New York, 1990).
4.	Hodgkin, A. L. & Huxley, A. F. J. Physiol. (London) 117, 500−544 (1952). | ChemPort |
5.	Mead, C. Analog VLSI and Neural Systems (Addison-Wesley, Reading, Massachusetts, 1989).
6.	Mahowald, M. A. & Mead, C. A. Scient. Am. 264, 76−82 (1991). | ChemPort |
7.	Hopfield, J. J. Network 1, 27−40 (1990).