Large-scale simulations of silicon nanowires just several atoms in diameter have given device designers new clues about how these nanometer-scale devices will one day perform. The work provides a basis for anticipating how the quantum mechanical effects that dominate behavior of materials at this size scale will alter the operation of future generations of electronic devices.
Writing in the August 28 issue of Physical Review Letters, Uzi Landman, Robert Barnett and Andrew Scherbakov from the Georgia Institute of Technology, and Phaedon Avouris from the IBM T.J. Watson Research Center, report on a number of issues pertaining to the atomic structure, electronic properties and electrical transport in silicon nanowires that will have to be considered by designers using devices this small.
"It's a much-discussed expectation that devices of this size will be different, but in what ways and by how much, remained unknown," said Uzi Landman, Regents' Professor of Physics and director of the Georgia Tech Center for Computational Materials Science. "In this study, we have explored certain unique properties of systems this small through first-principles quantum mechanical simulations. Such simulations, which are to the best of our knowledge the largest ones to date, are essential for gaining reliable and predictive information about these systems. They were enabled by a combination of improved methodologies and the availability of high-powered computers."
To boost speed and reduce energy use, engineers are being pushed to make electronic devices smaller and to pack more of them onto a chip. This pressure will eventually drive them to use features as small as one nanometer (one-billionth of a meter, or a hundred-thousandth the width of a human hair). When that happens, he noted, device operation will be dominated by quantum mechanical effects -- and the expectations that have long governed device design will no longer apply.
The researchers simulated silicon nanowires etched from bulk silicon, or self-assembled from clusters containing 24 atoms of silicon. In each case, the silicon was passivated by attaching hydrogen atoms to unused bonds, and the wires were connected to aluminum leads.
The theoretical simulations produced data on the nanowires' electrical conductance, the influence of the silicon-metal interface and the role that doping with aluminum atoms may play in changing materials properties. The work also suggests new ways of doping ultra-small transistor channels that could circumvent some current technological issues.
"This work attempts to fill in some of the gaps in our knowledge in this area," said Dr. Phaedon Avouris, manager of Nanometer Scale Science & Technology for IBM's Research Division at the T.J. Watson Research Center in Yorktown Heights, N.Y. "While the wires on which we report here are significantly smaller than those likely to be used in the near future, they are particularly useful because they tell us what to expect in the fully quantum mechanical limit -- the ultimate miniaturization limit. The calculations have revealed a number of significant changes in important properties."
Carried out on an IBM SP-2 computer at Georgia Tech, the simulations revealed that:
o Electronic states formed from a combination of orbitals from the aluminum leads and the silicon wire atoms penetrate all the way through nanowires of less than about one nanometer in length, giving such silicon bridges a finite conductance. But in longer structures, these electronic states penetrate only partially into the nanowire, with the silicon retaining its semiconducting properties.
o The transfer of electrons from the aluminum to the silicon at the junction between the two materials creates a localized dipole which forms a barrier to the flow of electrons. The simulations show that the height of such Schottky barriers at nanoscale metal-to-semiconductor contacts may not be too different from those found at more familiar size scales.
"The height of the barrier depends on the nature of bonding and atomic arrangement at the contact itself and varies for the various configurations of nanowires that we studied between being 40 to 90 percent larger than the value found at the corresponding macroscale contact," said Landman. "This is good news because it means that device engineers won't have to apply dangerously large voltages across the barrier formed at nanoscale metal-to-silicon contacts, as some researchers had suspected."
o The simulations suggest a way that could overcome some of the anticipated problems involved in doping the silicon used in devices this small. Doping of semiconductors is used routinely in order to tune and optimize device characteristics. However, in nanoscale devices one may expect detrimentally large device-to-device statistical variations of the dopant concentration. This variability could cause severe problems at this size scale because electronics designers may not consistently predict the performance of a collection of such devices.
However, the simulations suggest that building nanowires from silicon clusters could offer a solution. Because the clusters form hollow cages, much like carbon fullerenes, they could be fabricated around a dopant atom. With each cluster then containing a dopant atom, device consistency may be achieved.
o The wave-like nature of the electrons could cause interference effects in the electric conductance through the silicon nanowires used as current channels. When voltage is applied to open the channel, electrons penetrating the silicon nanowire from one of the aluminum leads may bounce off the contact to the other lead and flow back toward the source contact. Upon reaching that contact, they may bounce off again, and the process may repeat itself.
This behavior results, at certain electron wavelengths and wire configurations, in interference resonances that cause the channel to appear transparent, leading to the occurrence of spikes in the current flowing through the nanoscale channel.
"When building a device, engineers would have to take this into account and either find ways to use it or avoid it," Landman added. "In macroscopic devices, this phenomenon is of no particular consequence, showing again that small devices are different in ways that go beyond simple scaling with size."
The next step is to actually fabricate and test devices this small. The research was sponsored by the U.S. Department of Energy.
Research News & Publications Office
Georgia Institute of Technology
430 Tenth Street, N.W., Suite N-116
Atlanta, Georgia 30318
Media Relations Contacts: John Toon (404-894-6986); E-mail: (john.toon@edi.gatech.edu); Fax: (404-894-4545) or Jane Sanders (404- 894-2214); E-mail: (jane.sanders@edi.gatech.edu); Fax: (404-894-6983)
Technical Contacts: Georgia Tech: Uzi Landman (404-894-3368); E-mail: (uzi.landman@physics.gatech.edu); IBM: Phaedon Avouris (914-945-2722); E-mail: (avouris@us.ibm.com)
Visuals: Molecular dynamics simulations showing the phenomena mentioned in the study.
Journal
Physical Review Letters