Focus: Size Matters

Published October 8, 1998  |  Phys. Rev. Focus 2, 16 (1998)  |  DOI: 10.1103/PhysRevFocus.2.16

Size Dependent Thermopower in Mesoscopic AuFe Wires

C. Strunk, M. Henny, C. Schönenberger, G. Neuttiens, and C. Van Haesendonck

Published October 5, 1998
Figure 1
Phys. Rev. Lett. 81, 2982

Heat flow in the deep freeze. A small current through the long, snaking wire heats the center of the chip and creates a small temperature difference across each of the four perpendicular wires of different widths (and lengths). The temperature difference leads to a different voltage for each wire, demonstrating that the Kondo effect depends on wire size. All wires are gold doped with iron. The arrow at the top represents 50 µm.

As the circuits on computer chips continue to shrink in size, engineers have grown more interested in the quantum mechanical phenomena that affect wires less than a micron in width. But physicists have also been studying these effects, to learn about fundamental physics. In the 5 October PRL, a team demonstrates a new combination of precise measurement techniques and claims to resolve a controversy regarding a small-wire effect. They have won admiration for combining skillful measurements of temperature and electron heat transport on wires narrower than a micron and colder than 1 K.

The controversy involves a classic problem in condensed matter physics, called the Kondo effect. Although cooling a metal usually reduces its electrical resistance, adding only 0.005% magnetic impurity material, such as iron, causes the resistance to increase when cooling below about 10 K. In the early 1960s Kondo explained the effect by calculating the interaction of the magnetic spin with the metal’s electrons. But many aspects of the theory continue to be studied today, and the theoretical approach has permeated many areas of condensed matter physics. The dispute in recent years has centered on this question: Does the effect change in very small wires, or is it simply a property of the material, regardless of wire size? Based mainly on their resistance measurements, experimentalists have disagreed on the answer, and the theory is ambiguous as well. Researchers hope the resolution of the question will bring a deeper understanding of the Kondo phenomenon.

To address the question from a new direction, Christoph Strunk and Christian Schönenberger of the University of Basel in Switzerland and their colleagues decided to measure thermopower–the voltage difference between the ends of a wire induced by a temperature difference. Thermopower, like resistance, is a property of the conduction electrons and should show the same size-dependent effects, according to Schönenberger, but is much more sensitive than resistance. The team fabricated a microchip containing four AuFe wires with widths ranging from 105 to 305 nm, applied the same temperature difference across each one, and compared the voltages they generated.

The main difficulty in such experiments is creating and measuring suitable temperature differences on a micron scale. The slight temperature differences required were too small for traditional methods, so Strunk and his colleagues heated the electrons directly, using microamps of current. They measured the local electron temperature by employing the simple Nyquist formula, which relates the temperature and resistance of a wire to the random voltage fluctuations across it. Measurement of these fluctuations determined the temperature to within 50 mK . “It’s the most direct measurement of the absolute electron temperature you can think of,” says Schönenberger. Using the precise temperature measurements, the differences in thermopower between the different widths of wire were clear, and Schönenberger believes the measurements should settle the controversy. If the Kondo effect were size-independent, as some have claimed, the thermopower would have been identical for all of the wires. Schönenberger adds that the new combination of techniques expands the limited number of tools for characterizing small samples–now thermopower can be measured along with resistance and magnetic properties.

Nicholas Giordano of Purdue University says the experiments are “very difficult” and “very elegantly done.” He agrees that the measurements seem to settle the controversy, especially because “it gets at the same physics from a completely different angle” than the previous resistance measurements.


New in Physics