A Superconducting Field-Effect Switch

See allHide authors and affiliations

Science  28 Apr 2000:
Vol. 288, Issue 5466, pp. 656-658
DOI: 10.1126/science.288.5466.656

This article has been retracted. Please see:


We report here on a novel realization of a field-effect device that allows switching between insulating and superconducting states, which is the widest possible variation of electrical properties of a material. We chose C60 as the active material because of its low surface state density and observed superconductivity in alkali metal–doped C60. We induced three electrons per C60 molecule in the topmost molecular layer of a crystal with the field-effect device, creating a superconducting switch operating up to 11 kelvin. An insulator was thereby transformed into a superconductor. This technique offers new opportunities for the study of superconductivity as a function of carrier concentration.

The basic idea of an ideal electric “valve” goes back to the late 1920s and involves switching between high- and low-resistance regimes by an applied electric field (1). However, reliable devices could not be prepared until 30 years later (2), surface states of inorganic semiconductors, such as silicon or germanium, being the major hurdle. Since then, the silicon field-effect transistor (FET) has become the cornerstone of modern semiconductor industry and technology. In addition, there has been an ongoing effort to modulate superconductivity in thin films by an applied static electric field (3). Shifts of the transition temperatureT c of up to 30 K have been observed in high-T c cuprate films, caused by changing the concentration of charge carriers in the electronically active CuO2 layers (4, 5). However, complete field-induced switching between superconducting and insulating states remains a desirable goal. Here we report on a novel C60-based field-effect device, an ultimate switch between the insulating and superconducting regimes of chemically pure C60.

Working with C60 as the starting material is advantageous because so much is known about transport, and ultimately superconductivity, in C60 when electrons are induced by chemical doping (6, 7). Undoped C60 has a band gap of approximately 2 eV and is therefore insulating, whereas alkali metal–doped C60 (A3C60) exhibits metallic conductivity and, at low temperatures, superconductivity. Furthermore, working with van der Waals–bonded materials, such as molecular crystals, offers the inherent advantage of low surface state densities because of the absence of dangling bond–type surface states. Previous studies on organic semiconductors such as pentacene have shown that in an organic FET the chemical potential can be shifted easily across the band gap of the semiconductor, leading to n- as well as p-channel activity (8).

We have grown C60 single crystals that are several mm3 in size in a stream of hydrogen in an apparatus similar to that used for the growth of other organic semiconductor crystals (9). Multiply sublimed material was used as a starting material. In order to prepare field-effect devices on C60single crystals, we evaporated gold source and drain contacts on smooth growth surfaces through a shadow mask. Channels were typically 25 to 50 μm long and 500 to 1000 μm wide. Sputtered Al2O3 with a capacitanceC i of 185 nF/cm2 was used as the gate dielectric. Finally, a gold gate electrode was deposited on top of the oxide (Fig. 1). FET measurements were carried out in vacuum at temperatures between 4 and 300 K. Additional space charge–limited current measurements were used to determine the number of electrically active defects in these high-quality single crystals (10). Trap concentrations (deep levels) as low as 3 × 1012 cm−3 (one per 5 × 108 C60 molecules) are estimated. This level is significantly lower than those reported earlier for other vacuum-grown crystals (11).

Figure 1

Source-drain current versus gate voltage at room temperature for a C60 single-crystal FET. Top,p-channel operation; bottom, n-channel operation. The inset shows the schematic structure of a single-crystal FET.

Previous thin film C60 transistors exhibitedn-type behavior and field-effect mobilities of 0.09 cm2/V·s (12). Figure 1 shows the typical transistor characteristics at room temperature for single-crystal devices. These devices show n- as well asp-channel activity, reflecting the ambipolar transport in these high-quality single crystals and further emphasizing the low interface state density of the FET. Electron and hole mobilities of 2.1 and 1.8 cm2/V·s, respectively, are deduced from standard semiconductor equations (13). At room temperature, the channel resistance can be varied over approximately nine orders of magnitude by the applied gate bias (Fig. 2A). Figure 2A shows the channel resistance as a function of gate charge (n =C i V g/e, where n is the charge carrier density,C i is the capacitance,V g is the gate voltage, and eis the elementary charge) at room temperature and 5 K. The initial drop of the resistance at a few volts reflects the turn-on of the FET. At larger bias, charge accumulates in the channel, leading to a gradual decrease of the resistance. Finally, at very high positive gate voltages, the channel resistance drops abruptly to zero below a critical temperature T c of 11 K. The channel evidently becomes superconducting. The drop of the resistance depends both on the applied gate voltage and on the temperature. This is shown in Fig. 2B, where the channel resistance is plotted versus temperature and the gate charge. A priori we do not know the electronic nature of the superconducting channel; that is, how many molecular layers become superconducting. However, excellent agreement with known bulk superconductivity in A3C60(14) is found if we assume that only a single layer of C60 molecules accepts the electrons (15). The area density is approximately 9 × 1013C60 molecules per cm2 and the gate charge corresponds to 2.7 × 1013 electrons per cm2, which is equivalent to three electrons per molecule (C60 3−). In bulk AxC60, this carrier concentration is known to produce the optimum T c, as the Fermi level lies near the maximum in the density of states of the conduction band. A more detailed map of the superconductingT c as function of gate bias, expressed in terms of electrons per C60, is shown in Fig. 3. The critical temperature is maximal for three electrons per C60 molecule (14). Hence, we conclude that field-induced doping results in the same superconducting phenomena as does chemical doping in bulk A3C60.

Figure 2

Channel resistance of a C60 single-crystal FET as a function of applied gate voltage and temperature. (A) Resistance at 5 K and room temperature as a function of gate voltage. The transition to the superconducting state is seen as a drop in the blue curve. (B) Resistance as a function of gate charge (C i V g) and temperature, demonstrating superconductivity for concentration above 2.5 × 1014 cm−2 and belowT c = 11 K.

Figure 3

Channel resistance as a function of temperature and electrons per C60 molecule. The electron concentration is calculated assuming that only the first molecular layer accepts charge. The maximum of the transition temperature at three electrons per molecule is in accordance with measurements on chemically doped A3C60 (red, normal state; blue, superconducting state).

Modifying the lattice constant of A3C60 by doping or pressure results in a change of the lattice constant and of the electronic bandwidth, and therefore of the density of states at the Fermi level. Consequently, T c varies in accordance with Bardeen-Cooper-Schrieffer theory (6, 16,17). A systematic variation ofT c with the lattice constant is indeed observed for AxC60 (slight deviations from the cubic crystal structure are ignored here). The value ofT c = 11 K observed in our field-effect switch is slightly lower (≈3 to 4 K) than expected from the “universal curve” (Fig. 4). ThisT c reduction might be ascribed to the two-dimensional nature of the channel region as compared to the three-dimensional properties of bulk samples (15).

Figure 4

Transition temperature for bulk alkali-doped A3C60 and our C60 field-effect switch. The values for A3C60 are taken from (6).

A first characterization of the superconducting channel involves the upper critical magnetic field H c2 (Fig. 5). The slope of the critical field (dH c2/dT) is approximately −5 T/K in the range reported for bulk A3C60(6). Using the standard extrapolation toT = 0 K, we can estimate the coherence length to be on the order of 30 Å. This is further evidence that superconductivity in a C60 field-effect device is of the same type as in A3C60. The filling of the band is controlled by the applied gate voltage, and because of the strong electron-phonon interaction in this material, the channel region becomes superconducting below 11 K.

Figure 5

Transition to the superconducting state for different magnetic fields applied perpendicular to the channel. The variation of the upper critical magnetic fieldH c2 with temperature is shown in the inset (slope ≈ −5 T/K).

The possibility of investigating superconductivity as function of electron (or hole) density in a simple FET device opens up various opportunities to find superconductivity in new classes of materials, especially organic semiconductors. In addition to being able to implement the longstanding idea of an ultimate field-induced switch (insulator-superconductor transformation), this technique also opens up new ways to substantially modify the electronic state in molecular crystals.


View Abstract

Stay Connected to Science

Navigate This Article