ELECTRONIC-STRUCTURE OF THE GOLD BI2SR2CACU2O8 AND GOLD EUBA2CU3O7-DELTA INTERFACES AS STUDIED BY PHOTOEMISSION SPECTROSCOPY

High-resolution photoemission has been used to probe the electronic structure of the gold/ Bi 2 Sr 2 CaCu 2 0 8 and gold/EuBa 2 Cu 3 0 7 s interface formed by a low-temperature ( 20 K) gold evaporation on cleaved high quality single crystals. We find that the metallicity of the EuBa 2 Cu 3 0 7 -15 substrate in the near surface region ( - 5 Ä) is essentially destroyed by the gold deposition, while the near surface region of Bi 2 Sr 2 CaCu 2 0 8 remains metallic. This has potentially wide ranging consequences for the appHcability of the different types of superconductors in real devices.

A key requirement in both utilizing and understanding the high-temperature superconductors is the fabrication of very high quality materials. With the advent of high quality single crystals and films, this requirement is at least now partly fulfilled. One very irnportant ncxt step is a knowledge of the properties of interfaces of these materials with others. Photoemission spectroscopy is a natural tool to probe the interfacial region because of its surface sensitivity (in the range of 3-100 A) and ability to detect changes in the local chemical environment of an atom. lt has aiready been very successfuily applied to the study cf the interfacial chemistry of a wide variety of adatoms on the different types of high Tc substrates. l-4 Typically, the effects of the adatoms on the Cu 2p and 0 ls line shapes were emphasized most strongly, since the Cu-0 planes are thought to be crucial to the superconductivity. In particular, a single 0 ls peak centered at a binding energy of around -529 eV and a strong Cu 2p satellite/main line intensity ratio are thought to be especially important. Higher binding energy 0 Is components are usually associated with the formation of some oxide while the disappearance of the Cu 2p satellite signals a decrease of the Cu vatency frnm 2 + to 1 + .
These earlier experiments showed that gold formed the cleanest interfaces with each of the cupratc superconductors; it goes down smoothly with no evidence for islanding and does not adversely affect the Cu 2p or 0 ls line shapes. 1 • 2 Other noble metals such as Ag, Pd, and Cu did not tend to fare as weil. 3 • 4 This letter addresses primarily the states at the Fermi level, and how they are affected by the application of gold adatoms. This is a more sensitive and direct test of the interfäcial region than the core level line shape studies for a number of reasons. First, we are considering those states that are directly responsible for the superconductivity. Second, we have used higher resolution (200 meV) than is possible with x-ray photoemission spectroscopy (XPS) core level studies, and, since the photoelectron kinetic energies are lower, we are more surface sensitive ( escape depths on the order of 3-10 A as opposed to 5-50 Ä for XPS) . 5 This surface sensitivity is actually very crucial because of the very short supcrconduct.ing coherence Jengths of these materials, particularly along the c axis ( -3 Ä). 6 The coherence length is greater aiong the a or b axes, but surfaces for a-or b-axis junctions are much more difficult to prepare. Also, chemical intuition tells us that we are not as likely to have as clean an interface along the a or b axes as along the c axis. Our junctions were formed on crystals cleaved normal to the c axis (the easy cleavage direction).
The experiments were performed at the Mark II grasshopper monochromator at the Synchrotron Radiation Center in Stoughton, WI in an angle-resolved vacuum sc.ience workshop ( VSW) chamber with base pressure in the 10-11 Torr regime. During the gold evaporations, the chamber pressure never rose above 2X 10·· lO Torr. All experiments were performed at normal emission, although we will interpret the spectra in this letter as if they were essentiaily angle integrated. Tlüs is because the photon energies are relatively high, which makes the uncertainty in the momentum k high, and because the cleaves of our single crystals were not perfect over a distance the size of the photon beam-about 3 square mm. The incident angle of the photon beam was 75• from normal, and the combined energy resolution of the photon source and spectrometer was held constant at 200 meV.
High quality single crystals of Bi 2 Sr 2 CaCu 2 0 8 and EuBa 2 Cu 3 0 7 " ~· of transition temperatures of -90 K each, were cooled to 20 K before being cleaved in the ultrahigh vacuum environment, and werc maintained at that temperature for the fuH duration (including the gold deposition) of the experiments. The low temperature is important for a number of rcasons. lt had been shown to be necessary for getting accurate spectra for the 123's, partic- ularly in the near Fermi surface regime. 7 Although the Bi 2 Sr 2 CaCu 2 0 8 samples appear much more stable at room temperaturc 8 and there are recent reports showing high quality spectra from 123's cleaved at room temperature, 9 we elected to perform all of our measurements at low temperature fcr safety and consistency. The low-temperature evaporation is also superior for a number of reasons. rnterfacial processes such as chemical reactions and the creation of defects will be greatly inhibited. In addition, the decreased surface mobility of the adatoms at lowtemperature favors Frank-van der Merwe or layer-by-layer growth, instead of the Volmer-Weber growth mode, wbere the deposited atoms aggregate to form clusters or islands on the substrate surface. Examination of the spectra shows very uniform exponential decay of the Bi 5d core Ievel intensity with gold coverage, implying true layer-by-layer growth. The very short escape depth of 3 A in the gold layer confirms that no intermixing or alloying of the gold with the substrate surface has occurred. Figure l shows the valence bands and Bi 5d core Jevels of Bi 2 Sr 2 CaCu 2 0 8 for different gold coveragt---s. As depicted in the figure, the spectra were all normalized to give equal Bi Sd core level intensity. Thus the contribution to the valence band from the Bi 2 Sr 2 CaCu 2 0 8 states shou]d to first order be essentially the same in each of the spectra presented, since the signal from t he underlying states should fall off as a simple exponentiaI with distance, governed by the photoelectron escape deptb through the gold. Thus we see that with even as little as 2.0 Ä of gold on the surface (approximately one monolayer), the valence band signal is almest completely dominated by the gold states. This is because of the extreme surfäce sensitivity of the spectroscopy. The detailed effects that the gold adatoms have on the superconduciing substrate are best seen for Iow gold coverages (below one monolayer). For these coverages, the gold overlayer will not yet havc reached its fuU "metallicity," which in cases where the substrate is nonmetallic, is easily discernable by the lack of a Fermi edge. Other signals of a film that have not yet reached metallicity are ( adatom) core ievels that are broader and at higher binding energy than those of the bulk material. The behavior of our core levels confirms that the 0.5-Ä-thick gold films are not metallic. Blowups of the valence band and near Fermi level region of just the 0.0 and 0.5 A goid coverages on Bi 2 Sr 2 CaCu 2 0 8 are shown in Figs. ! (b) and 1 ( c }, with the same normalization as in Fig. 1 ( a). A clear Fermi edge is observed for both the clean substrate and the substrate with the 0.5 A gold overlayer. 10 Since the Fermi edges could not have been derived from the gold overlayers, it is clear that the surface of Bi 2 Sr 2 CaCu 2 0 8 remains metai!ic, which is of course a prerequisite for the superconductivity. Thus we can conclude that there is very iittle chemical reaction between the gold and the Bi 2 Sr 2 CaCu 2 0 8 Substrate, and that the Bi 2 Sr 2 CaCu 2 0 8 derived density of states at the Fermi level is essentially unaltered by the go1d deposition, at least within the scale of our energy resolution (-200 meV). Ultrahigh resolution (-35 meV) angle-resolved measurements have recently been completed in our lab, and have in fact shown that there is some alteration of the BiO states (in the cleavage plane) at the Fermi level from a deposition of gold. The states which originate from the CuO planes, which are a few angstroms below t. he surface, are essentia!ly unaltered. 11 The goid/EuBa 2 Cu 3 0 7 _ 6 interface is unfortu.riately not quite so nice, as is shown in Fig. 2 , which presents the valence bands of a single cleaved crystal of EuBa 2 Cu 3 0 7 _ 6 for different gold coverages. The normalization was per· formed in a similar way as done earlier on Bi 2 Sr 2 CaCu 2 0 8 , except that the Ba 4d core 1evels were used instead of the Bi 5d's. The cmcial piece of information is obtained from the blowups of the near Fermi level region in Fig. 2 ( b) (There is a sloping density of states that does just cut the Fermi level in the 0.5 A spectrum, although it is much broader than the system resolution of 200 me V and the midpoint is pushed at least 100 meV below the Fermi energy) . As the gold overlayer becomes metallic at larger coverages, a gold-derived Fermi edge appears. Thus we can conclude that the gold deposition alters the states near the Fermi level in the near surface region of the EuBa 2 Cu 3 0 7 ... 8 substrate and renders it nonmetallic. Since we formed the interface under the ideal conditions of very low temperature and ultrahigh vacuum, with the most inert of overlayers, this is potentially haz..ardous for any device on these materials (particularly c-axis films) that intends to make use of the proximity or Josephson elfects. Better choices may be to make the junctions along the a or b axes, where the superconducting coherence length is much greater, 6 or to use the Bi 2 Sr 2 CaCu 2 0 8 material, which appears to be much less affected by the gold deposition.
In conclusion, low-temperature evaporations of gold, which should form the cleanest and most abrupt junctions with the high Tc cuprates, appear to alter the near Fermi level states of EuBa 2 Cu 3 0 7 -li at the interface enough to render the surface essentially nonmetallic, making the feasibility of 123/gold junctions along the c axis questionable at best. The situation with Bi 2 Sr 2 CaCu 2 0 8 /gold junctions is much more favorable, as the Bi 2 Sr 2 CaCu 2 0 8 substrate rernains metallic in the near surface region, making the feasibility of devices relying on the Josephson or proximity etfects quite real.
We thank Dr. C. G. Olson for many discussions and lowship, and A. K. acknowledges the NSF through the PYI program. This work was supported by NSF grant-DMRR-89103478, the JSEP contracts DAAG29-85-K-0048 and N00014-84-K-0327, and by the Department of Energy.