Report

Electrostatic Modulation of Superconductivity in Ultrathin GdBa2Cu3O7-x Films

See allHide authors and affiliations

Science  14 May 1999:
Vol. 284, Issue 5417, pp. 1152-1155
DOI: 10.1126/science.284.5417.1152

Abstract

The polarization field of the ferroelectric oxide lead zirconate titanate [Pb(ZrxTi1–x)O3] was used to tune the critical temperature of the hightemperature superconducting cuprate gadolinium barium copper oxide (GdBa2Cu3O7–x) in a reversible, nonvolatile fashion. For slightly underdoped samples, a uniform shift of several Kelvin in the critical temperature was observed, whereas for more underdoped samples, an insulating state was induced. This transition from superconducting to insulating behavior does not involve chemical or crystalline modification of the material.

Superconductor-insulator transitions have been investigated theoretically and experimentally in a variety of systems, including three-dimensional (3D) ionic crystals, 2D films, and Josephson junction arrays. These transitions are examples of the modulation of superconductivity by means of control parameters such as pressure, magnetic field, disorder, and thickness (1–4). For applications, superconducting bolometers operate on changes in the temperature, and superconducting quantum interference devices rely on the application of small magnetic fields that modulate the critical current of a Josephson junction.

Here we report on nonvolatile, reversible modulation of superconductivity in the high-temperature copper oxide superconductor GdBa2Cu3O7–x (GBCO), using an electrostatic field as the control parameter. A unique feature of these cuprate materials, as compared to the elemental BCS superconductors, is that their properties depend strongly on the carrier concentration, as shown by their generic temperature-doping phase diagram (Fig. 1), which depicts an antiferromagnetic insulating state at low doping levels that is separated from a superconducting state at higher doping levels. This phase diagram suggests, in the underdoped regime, a superconductor-insulator quantum phase transition controlled solely by the doping level (or chemical potential), as outlined in the recent SO(5) theory of Zhang (5).

Figure 1

Schematic temperature-doping phase diagram of the copper oxide superconductors. As the doping level is increased, a transition occurs between an antiferromagnetic insulating state and a superconducting ground state at T = 0.T* is the signature of the pseudogap in these materials. AFI and SC refer to the antiferromagnetic insulating Néel state and the superconducting state, respectively. The small circles are experimental data from (22). Large circles indicate the experimental T c onset data of this work, and the arrows indicate the shift in carrier density induced by the field effect. The horizontal axis is plotted on a log scale.

In our experiment, we controlled the doping level using the polarization field of a ferroelectric oxide, Pb(ZrxTi1–x)O3 (PZT), grown epitaxially on ultrathin films of GBCO, to change the carrier concentration and modulate the electronic properties of the superconducting layer without introducing any chemical or microscopic structural disorder. This approach is both reversible and nonvolatile, with no external electric field being applied across the insulating ferroelectric during the measurement of the properties of the superconducting layer (6).

The field effect has been used before to shift the critical temperature of elemental superconductors by ∼1 mK (7). It has also been applied to the copper oxide superconductors to shift the critical temperature by a few kelvin, as well as to study vortex dynamics by changing the effective thickness of the superconductor (8). In these studies, the volume and number of charge carriers were too substantial to change the metallic character of the superconductor. On the applied side, proximity field-effect devices and Josephson field-effect transistors have been fabricated (9). For all of these studies, a conventional field-effect device, in which a continuous voltage is applied across the gate insulator, was used.

To grow the PZT/GBCO heterostructures, we carried out off-axis radio frequency magnetron sputtering onto single crystalline SrTiO3 (STO) (001) substrates, a process that is described in detail elsewhere (10). Because the carrier density of the copper oxide superconductors is relatively large, ∼1021/cm3, the electrostatic Thomas-Fermi screening length is on the order of a few angstroms, which necessitates the use of extremely thin GBCO layers, ∼1 to 2 unit cells thick (12 to 24 Å), in order to obtain substantial field effects. To grow such thin layers, a buffer layer of insulating (at low temperature) PrBa2Cu3O7 (PBCO), 72 Å thick, was first grown on the STO substrate, as has been used previously for the growth of ultrathin YBa2Cu3O7–xfilms and multilayers (11). X-ray diffraction revealed epitaxial growth of PZT (001), GBCO (001), and PBCO (001) on STO (001) substrates (12). For (001)-oriented PZT, the two ferroelectric polarization states lie perpendicular to the GBCO layer, and Sawyer-Tower measurements on these samples reveal a remnant polarization of ∼10 μC/cm2 and a coercive field of 100 kV/cm. The heterostructures were then patterned into four point resistivity paths, with Au pads being deposited for the current and voltage contacts, as well as for the poling contact on top of the PZT layer.

We first examined the ferroelectric field effect in these PZT/GBCO heterostructures by measuring their room-temperature resistance as a function of the poling field used to establish the polarization state of the ferroelectric layer. A voltage pulse (∼1 s) was applied across the ferroelectric, after which the resistance was measured. By repeating this procedure for a series of different voltages, the ferroelectric hysteresis [resistance versus field (R-E)] was revealed (Fig. 2) for a 3000 Å PZT/ 20 Å GBCO/72 Å PBCO heterostructure. The measured coercive field was 100 kV/cm, and the difference in resistance of the GBCO/PBCO bilayer between the two polarization states of PZT was ∼10%, with the sign of the resistance change agreeing with the hole-doped character of GBCO. The resistance states after application of the voltage pulses were nonvolatile for periods of hours (the length of the experiments) and were also reversible, as demonstrated by the hysteresis. For comparison, we show in the inset to Fig. 2 a ferroelectric hysteresis loop [polarization versus field (P-E)] obtained through a standard Sawyer-Tower measurement on this sample, showing a coercive field of 100 kV/cm and a remnant polarization of 10 μC/cm2. Although the coercive fields of both measurements agree, extracting a quantitative value for the remnant polarization is not straightforward from the R-E measurement, because the resistance change does not directly reveal the magnitude of the ferroelectric polarization. A free electron estimate, however, based on assumingdσ/σ ∼ dn/n, where σ is the conductivity of the GBCO/PBCO bilayer and n is the carrier concentration, yields a value consistent with a polarization of 10 μC/cm2. The hysteresis loop obtained with the resistance measurements (R-E) is considerably sharper than the P-E hysteresis loop obtained by the Sawyer-Tower measurement. The rounded nature of the P-E loop is attributable to leakage currents through the PZT layer, which are not present during the measurement of the R-E loop, because no voltage is applied across the ferroelectric during the measurement of the resistance.

Figure 2

Ferroelectric hysteresis loop on a 3000 Å PZT/20 Å GBCO/72 Å PBCO heterostructure obtained with resistance measurements (R-E hysteresis). (Inset) A P-E hysteresis loop of this sample obtained through a Sawyer-Tower measurement.

We then studied the low-temperature transport properties of these heterostructures. The resistivities for two different PZT/GBCO/PBCO heterostructures as a function of temperature for both polarization states of the PZT layer were measured (13,14). A shift of the superconducting transition temperature (T c) of 7 K (Fig. 3, A and B) and switching between superconducting and insulating behavior (Fig. 4) were observed. For Fig. 3A, we measured at room temperature a resistivity change of ∼15% upon switching the PZT polarization. As the temperature decreased to ∼200 K, the resistivities of both curves increased, as did the size of the field effect; this increase can be attributed to the insulating behavior of PBCO. Below 200 K, where the contribution of PBCO to the overall conductivity becomes negligible, metallic behavior was observed, with the size of the field effect increasing to ∼50% at the superconducting transition. At the transition, the observed shift in temperature of the two resistivity curves of 7 K was maintained throughout the region of the transition, as demonstrated in Fig. 3B. For the upper curve, the polarization vector points toward the substrate, corresponding to a removal of holes from thep-type GBCO that results in a depression ofT c, whereas the lower curve corresponds to an addition of holes to the system. The only difference between these two curves is the state of the ferroelectric polarization. No other modification to the sample was carried out. As with Fig. 2, both states are reversible and nonvolatile.

Figure 3

(A) Resistivity versus temperature of a PZT/20 Å GBCO/72 Å PBCO heterostructure for the two polarization states of the PZT layer. The upper curve corresponds to a removal of holes from thep-type GBCO, resulting in an increase in the normal state resistivity and a depression of T c by 7 K. (B) Expanded view of the data presented in (A).The resistivities ρ of the two curves have been normalized at 100 K.

Figure 4

Resistivity versus temperature of a PZT/20 Å GBCO/72 Å PBCO heterostructure. The upper curves, in which holes have been depleted from the system, are insulating throughout the temperature range investigated, except for a dip below 20 K due to nonuniformities in the field effect. The thick curves are zero field measurements, whereas the other curves are measurements in magnetic fields of 1, 4, and 7 T. (Inset) Fits of the low-temperature insulating behavior to variable range hopping and to ln 1/T behavior.

For the PZT/GBCO/PBCO heterostructure of Fig. 4, the thick lines are measurements taken in zero applied magnetic field, whereas the other curves (below 60 K) are measurements in magnetic fields of 1, 4, and 7 T applied perpendicular to the plane of the film (15). For the lower zero-field curve, the resistivity is essentially temperature-independent from 200 K down to ∼50 K, where the onset of the superconducting transition occurs. The resistivity of the sample at this temperature is 2.3 × 10–3ohm·cm. The application of magnetic fields of up to 7 T has essentially no effect above 50 K, other than the small magnetoresistance that is expected in the normal state of GBCO. Below 50 K, however, the resistive transition broadens substantially with increasing magnetic field because of the motion of field-induced vortices in the superconducting state. At the lowest temperatures and for all the magnetic fields, the transition sharpens.

In contrast, the upper curves, which again correspond to a depletion of holes by the ferroelectric polarization, are insulating throughout the temperature region investigated, except for a dip observed below 20 K. This dip on the insulating side probably occurs because the GBCO thickness is not uniform over the entire film. Atomic force microscopy imaging of the PBCO and GBCO layers reveals the presence of unit cell steps; because of the small Thomas-Fermi screening length in these materials, these thickness variations prevent the polarization field from acting throughout the entire film thickness, leaving regions of the sample that are still superconducting, as confirmed by the sensitivity of the dip to the magnetic field. In addition, within this Thomas-Fermi screening length, there is an exponentially decaying vertical gradient of the carrier concentration. At low temperatures, the insulating behavior is not simply thermally activated. Rather (as shown in the inset to Fig. 4), it can be fitted to variable range hopping (VRH), as well as to the ln (1/T) behavior that has been observed by Ando and co-workers in their high magnetic field experiments on La2–xSrxCuO4 (LSCO) (2,16). Although the limited temperature range of investigation precludes a definitive determination based on the fits alone, the density of localized states deduced from the fit to VRH is unphysically large (∼1024/cm3, assuming a tunneling length of 10 Å), which implies that it is not the appropriate conduction mechanism to consider here. As shown in recent magnetic field experiments on Pr2–xCexCuO4, measurements to lower temperatures (<100 mK) are important to more completely address the nature of the insulating state (4).

To quantify the ferroelectric field effect, we have carried out Hall effect measurements to determine the sign and carrier concentration of these films (17). For samples such as the ones shown in Fig. 4, we measure a low-temperature carrier density of ∼7 × 1020 holes per cubic centimeter, which is below the optimally doped value of ∼4 × 1021 holes per cubic centimeter that we find for thick GBCO films, revealing that these samples are in the underdoped region of the high-T c phase diagram. Using the measured remnant polarization of 10 μC per cubic centimeter, we see that the polarization field of the ferroelectric is sufficient to remove the carriers from the material and induce an insulating state. For samples such as the ones shown in Fig. 3, we measure a low-temperature carrier density of ∼2 × 1021 holes per cubic centimeter, and thus one would not expect to be able to induce an insulating state. These results are summarized in Fig. 1, where the data of this work are plotted on the experimental phase diagram obtained by chemical doping. The arrows indicate the expected shift in carrier density due to the polarization field of the ferroelectric.

The shift in the transition temperature (Fig. 3) and the switching between superconducting and insulating behavior (Fig. 4) represent the largest reversible modulation of superconductivity that has been achieved using a field effect. We attribute these results to the ferroelectric field-effect approach taken here, in which the large polarization field of the ferroelectric applied to ultrathin films induces a substantial modulation of the carrier density. Furthermore, because the effect is nonvolatile, with no voltage being maintained across the insulating ferroelectric, it is not susceptible to leakage currents or nonlinearities in the dielectric properties of the insulator (8). From the point of view of applications, these experiments demonstrate a route to fabricating a nonvolatile, reversible superconducting switch.

On the theoretical side, we note that the copper oxide superconductors are distinct from the elemental superconductors in that they possess a small superfluid density, which implies that they are sensitive to phase fluctuations as well as to changes in the chemical potential, as evidenced by their temperature-doping phase diagram (Fig. 1). This phase diagram is central to the SO(5) theory of Zhang, in which the antiferromagnetic and superconducting states have a common origin, with their T = 0 electronic properties being controlled solely by the chemical potential (5).

Chemical substitution can also be used to change the doping level, and numerous experiments have been performed. However, this approach is often irreversible and alters in an unavoidable and uncontrolled way the microscopic disorder and crystalline structure, which can itself induce a superconductor-insulator transition. The extreme sensitivity of the electronic properties of these materials to even small amounts of Zn substitution that do not modify the doping level is one extreme example of the effect of impurities (3). For the field-effect approach discussed here, no chemical or structural disorder is introduced, allowing us to identify the key role of the electronic doping level in determining the physical properties of these materials.

In the underdoped superconducting regime, the SO(5) theory of Zhang, as well as the theory of superconductivity in bad metals of Emery and Kivelson, predict that classical and quantum phase fluctuations are responsible for depressing T cwell below the pairing temperature, which, in these theories, is represented by T* in Fig. 1 (5,18). As the chemical potential μ is increased, these phase fluctuations decrease because of improved phase stiffness and screening, leading to an increase in T c. In particular, Emery and Kivelson calculate the T cdependence for the minimum metallicity required to suppress the phase fluctuations that prevent superconductivity from occurring. This limiting value is nonuniversal and depends on the material, the dimensionality of the system, and the details of the short-range screening. Work on YBCO single crystals, in which radiation damage was used to induce an insulating state, yields a value of ∼2.5 × 10−3 ohm·cm (19, 20). In this work, we changed the metallicity directly by changing the number of carriers, finding a normal state resistivity of 2.3 × 10−3ohm·cm. Because YBCO is highly anisotropic and quasi-2D in the underdoped regime, an equivalent 2D sheet resistance, given by the resistivity per copper oxide plane (6Å), is sometimes considered. The equivalent sheet resistance in this case is ∼40 kohm per square, which is significantly larger than the quantum of resistance for Cooper pairs (6.45 kohm per square).

Finally, regarding the insulating state we observed, the results suggest that because the insulating state is not consistent with variable range hopping, which is what is observed at low doping levels in the antiferromagnetic Néel insulating state up to high temperatures (∼50 K), the samples may instead be in a quantum disordered state that exists between the Néel and the superconducting ground states (21). The observed ln (1/T) behavior over the temperature range investigated suggests that this quantum disordered state may also be the state that Ando and co-workers achieved by applying large magnetic fields to underdoped LSCO (2).

REFERENCES AND NOTES

View Abstract

Navigate This Article