HIGH-T SUPERCONDUCTING ELECTRONICS : IMPLICATIONS OF REDUCED DIMENSIONALITY

First, we present experimental evidence that Bi-Sr-Ca-Cu-O superconductors are effectively quasi-two dimensional, in both the normal and the superconducting state. Next, we analyze how this unusual property of the material influences the quest for active electronic components based on high-To superconductors, with focus on trilayer Josephson junctions and ferroelectric/superconductor field-effect devices.


I. INTRODUCTION
The first speculations about the two-dimensional (2D) nature of superconductivity in cuprates may be as old as the high-T superconductors themselves.More refined theoretical arguments have been offered since2, but nevertheless the question about the effective dimensionality of the normal and the superconducting states in cu- prates is still vividly debated.
The first clear indication that the normal state may be 2D is obtained from polarized reflectance spectroscopy 3a on single crystals of Bi2Sr2CaCu208.As seen from Figure 1, the reflectance is metallic-like for the in-plane polarization, and distinctly insulator-like for the out-of-plane field orientation.
Optical spectroscopy also provides some indication that the superconducting state may be 2D as well.Namely, it enables one to accurately monitor the carrier density in oxide metals, and this technique has been applied 3b to La.8Sr0.1CuO4,YBa2Cu307, Bi2Sr2CaCu208, and TI2Ba2Ca2Cu3OI0.Tc is found not to scale with the volume carrier density N, but rather with the molecular-layer density n (i.e., the number of mobile charge carriers, per unit area, in one molecular layer), as shown in Figure 2.This suggests a picture in which thin high-T superconducting slabs (such as CuO-Ca-CuO2 layers within BiSr2CaCuOs) are separated by some 'spacer' layers (e.g., BiO-SrO), which can be insulating, semiconducting, or me- tallic.The point is that, in this picture, T is a property of individual superconducting slabs, while the way in which they are coupled should not matter very much.
Other indications for 2D superconductivity have also been found, for example, in the critical current, angular-dependent magnetoresistance, and torque magne- tometry data.However, neither of these experiments is quite conclusive; for ex- ample, in highly anisotropic systems the out-of-plane resistance is very sensitive to planar defects such as stacking faults and microcracks.n (10 '4 cm"2)

HIGH-To
FIGURE 2 Tc as a function of n, the molecular-layer carrier density, in Lal.8sSr0 15CUO4 (L), YBazCu307 (Y,, the contribution from the CuO2 planes only, solid rectangle; Y,, the sum of both contributions from planes and chains, open rectangle), Bi2SrzCaCuzO8 (B), and Tl2Ba2Ca2Cu30o (T).

II. EXPERIMENTAL
To overcome such criticism, we have synthesized 4 'diluted' superlattices in which ultrathin layers of Bi2Sr2CaCu2Os, which has Tc 85 K, are separated by up to 120/ thick layers of Bi2Sr2CuO6, with Tc 15 K. Indeed, these two compounds are very well matched chemically, crystallographically and electronically.Their a and b lattice constants are nearly identical, which allows for excellent heteroepitaxy.Furthermore, their charge carrier densities do not differ very much, as seen, for example, from similarity of their optical spectra, and in particular from the fact that their plasma edge frequencies are not very different.In consequence, one does not expect much carrier depletion or accumulation to occur in multilayer structures assembled by stacking layers of these two phases.
The technique we have employed was sequential (atomic-layer-by-layer) mo- lecular beam epitaxy (ALL-MBE) deposition, which we will not describe in detail here since it has been reviewed recently5.It is sufficient to point out here that this slow, well-controlled technique provides single-crystal films with atomically flat surfaces and interfaces, and with transport properties, without any post-treatment, that are equal or better than those of bulk single crystals.Furthermore, ALL-MBE enables one to stack molecular or even atomic layers in any desired sequence, and hence it is ideally suited for fabrication of highest quality multilayers and superlattice structures.

III. RESULTS AND DISCUSSION
In Figure 3, we show the four-point-contact resistance data for five Bi-Sr-Ca-Cu- O thin films.The film A contains the Bi2Sr2CuO6 phase only, B is a pure Bi,,Sr2CaCu208 film, and C, D, and E are 1:1, 2:1, and 5:1 superlattices, respec- tively, of Bi2Sr2CuO6 and Bi2Sr2CaCu208.Notice that in the film E, one molecular layer (15.4A thick)of Bi2Sr2CaCu208 alternates with five molecular layers (the total of 5 x 12.3 A 61.5 A) layers of Bi2Sr2CuO6.Nevertheless, there is no 0.5 significant decrease of Tc as we increase the number of the spacer layers, which indeed suggests that the interslab coupling is not essential here.Finally, the most direct approach is to synthesize an ultrathin high-T film, ideally containing just one superconducting slab.Generally, this task is quite tedious because of interdiffusion of the substrate material from the bottom side, and oxygen loss, interaction with the atmosphere, etc., from the top side.Yet, ALL-MBE has proven to be up to such a task; using Bi2Sr2CuO6 for both the buffer layer and the protecting overlayer, we have succeeded in achieving a remarkably high Tc in a one-unit-cell thick film of Bi2Sr2CaCu208, as shown in Figure 4.The transition midpoint, at about 95 K, is actually higher than what we observe in thicker films grown under similar conditions.This result leaves very little room for doubts about the essentially 2D nature 6 of high-T superconductivity, at least in Bi-Sr-Ca-Cu-O.
Having established this important fact about the superconducting state in cu- prates, it is natural to inquire about its implications for various devices that could conceivably be made of these materials.To begin with, let us consider Josephson junctions, focusing on what is commonly referred to as the sandwich-junction geometry.Such junctions, in which a thin insulating barrier layer separates two superconducting layers, have been fabricated quite successfully with conventional, low-T superconductors.The principal advantage of this geometry is that, with good epitaxy and growth control, it should provide very reproducible and uniform junctions.
The key technological problem here is to ensure that the superconductor remains undeteriorated in the vicinity of the barrier.What constitutes 'vicinity' turns out to depend on the material, i.e., on its superconducting coherence length, which in low-T superconductors is of the order of 10-100 nm.In the case of a high-T superconductor such as Bi2Sr2CaCu208, the coherence length is much shorter, in particular in the c-axis direction (i.e., perpendicular to CuO2 planes).Actually, we have already seen that in this material the superconducting slabs are more or less independent, so it will largely be the last molecular monolayer, the one closest to the barrier, that has to ensure that the order parameter is conveyed across the junction.From the materials science point of view, it means that crystalline per- fection has to be maintained on an atomic scale, in particular near the barrier.Furthermore, the barrier must be made very thin--comparable, perhaps, to one molecular layer of the superconductor--and structurally as perfect itself.Indeed, this is not an easy task; one has to deposit an insulating layer, merely 2-3 nm thick, without any pinholes over large, macroscopic areas.
In Fig. 5, we show some of our data on trilayer Josephson junctions manufactured by ALL-MBE5.The top and the bottom electrodes consist of Bi2Sr2CaCu2Os, with T > 80 K after device fabrication.The barrier consists of a single molecular layer of a phase that is not thermodynamically stable, but which nevertheless can be grown by virtue of atomic layering.Its structure is that of Bi2Sr2CaTCU8020, but trivalent cations (Dy and Bi) are doped to selected Ca sites, to suppress free carriers.[This may be a good illustration of material engineering, by ALL-MBE, at and beyond an atomic monolayer level.]The device structure is illustrated in Fig. 5a; the area of the active mesa is 30/zm x 30/zm.The junction shows near ideal I- V characteristic, as displayed in Fig. 5b.This way, we believe, the first demon- stration of a hysteretic junction made of cuprate superconductors.Very clear mi- O (U) d3NIVLSIS I HIGH-To SUPERCONDUCTING ELECTRONICS 183 crowave (Shapiro) steps, as shown in Fig. 5c, are observed in this and other similar devices.
With ALL-MBE, it is possible to control the level of Dy or Bi doping within the barrier, and thus to vary the barrier resistance.[This can be considered as barrier engineering, on an atomic scale.]In this way, the junction critical current and normal state resistance are scaled by over three orders of magnitude, while their product remains nearly constant (IRn 0.5 mV).For the most heavily doped barriers, resistance exceeds 105 f at low temperatures.This shows that there is no leakage due to pinholes or second phase inclusions, over a relatively large area of (30/m)2, although the barrier thickness is only 25/.
So, we have shown that good trilayer or 'sandwich' Josephson junctions can be manufactured using high-To cuprate superconductors.However, admittedly this is not an easy task, and it may call for sophisticated deposition techniques such as ALL-MBE, because, in a sense, one has to work against the 2D nature of the material.[Nevertheless, this avenue undoubtedly deserves to be pursued further, in view of potential payoff of superconducting electronics at liquid nitrogen tem- peratures.]On the other hand, there exists another class of devices, which we discuss next, in which the reduced dimensionality of high-T superconductors is an advantage, and even may be indispensable.
The simplest such device is schematically represented in Fig. 6.This is a three- terminal device, quite similar to a field-effect transistor, in which one is using the electric field to modulate the carrier concentration within the high-Tc layer.Here we actually benefit from the fact that layered cuprates such as Bi2Sr.CaCu_O8 are bad conductors in the c-axis direction, so that the screening length is long and the electric field parallel to that axis can penetrate deeply.Besides, for superconductivity at a high temperature (say, at 77 K), we may need no more one or two molecular layers of Bi:Sr2CaCuO8, provided one can grow such layers with enough perfection to ensure good connectivity over macroscopic area, and without dete- rioration of the material's properties.We have seen above that this has been accomplished already.Finally, the carrier density is much lower in cuprates than in the conventional, low-Tc superconductors.It is the fortunate coincidence of these unusual propertiesin addition to the potential for operation at liquid nitrogen temperaturethat has greatly revived interest in field-effect superconducting devices7, which were actually proposed long ago.
In the device shown in Fig. 6, we have incorporated two improvements.First, we have added a thick BiSr:CuO6 buffer layer.It separates the active layer from the substrate, improves the epitaxy, and facilitates device fabrication, while it does not perturb much the ultrathin high-Tc BiSr:CaCu:O8 layer above it.Furthermore, in this way one has to displace electrons by just few angstroms in order to accomplish carrier depletion or accumulation in the active Bi_Sr_CaCuO8 layer; the speed of the device is thus increased significantly.
The second modification consists of replacing a simple high-dielectric-strength insulator, as utilized commonly, by a ferroelectric material such as PbTiO3.In a good ferroelectric, spontaneous polarization is large, and the surface charge density can exceed 0.5 C/m:, while the switching field remains relatively low.In this way, one can considerably reduce the voltage needed to switch the device.With that state of the ferroelectric can be switched by a current pulse, and it can be inferred from the voltage across the device.Ideally, if one can sufficiently modulate the carrier concentration within the high-T layer, one could switch it between the superconducting state for one polarization of the ferroelectric, and the normal- metal state for the opposite one.In this way, one could read without altering the polarization state, i.e., without erasing.Notice further that a very thin layer of the ferroelectric may be sufficient; the key advantage is that such a layer can be switched very fast, with a voltage as low as 0.1-1 V.This is indeed very convenient for portable systems, for example, so that these and similar devices, still in their infancy, may develop into a mature and useful technology.The future of superconducting electronic components seems bright.