There are several reasons this book was written. One was to provide a survey of all of the information obtained to date on cuprates by photoelectron spectroscopy, a technique that has been one of the more productive techniques for providing information on the electronic structure of cuprates. Thus the book serves as a review of research, but such reviews soon become dated in a rapidly moving field. Another reason was to provide a textbook, albeit of a limited nature, for persons entering the field of photoelectron spectroscopy. This aspect of the book should be useful for a longer time. We have limited our discussion to only the techniques used in photoemission on cuprates, excluding other applications of photoelectron spectroscopy, but, in fact, not much has been omitted. Finally, we hope that experienced theorists and experimentalists from fields other than photoelectron spectroscopy will learn something of the difficulties with the photoemission experiments and problems with the interpretation of the data. Photoemission data are widely quoted and we hope to provide a better understanding of the phenemona and the experimental difficulties for those who use such data in the future. In the past few years there has been much activity in the study of complex oxides, e.g., manganates, nickelates, ruthenates, often by photoelectron spectroscopy. We hope those whose interests lie with these materials rather than cuprates will find much of value in this book.

Introduction

In this chapter we present photoelectron spectra of several materials. We present valence-band data for all of them, and in most cases core-level spectra are discussed. Na is a simple metal, about as simple a metal there is. If the photoelectron spectra of Na cannot be understood it is difficult to be comfortable with interpretations of photoelectron spectra of the cuprates. Cu and Ni are more complicated, having 3d bands near and spanning the Fermi level, respectively. Then we present results on NiO, Cu2O, and CuO which introduce new effects. The photoelectron spectra of NiO and especially CuO bear some resemblance to those of the cuprates. The studies reported on the three oxides do not represent all of the work on these materials. Studies on related materials, e.g., Cu- and Ni halides, have been useful, for one can see the influence on the photoelectron spectra of changes in ionicity, number, type, and disposition of nearest neighbors, and distances between nearest neighbor transition metal atoms. Studies of a series of 3d transition-metal oxides are also helpful, but we do not discuss these here. Another approach, also not discussed here, is to start with a clean single-crystal surface of a transition metal and let it oxidize in stages in the experimental chamber, taking spectra and LEED patterns at various stages. In some cases, e.g., NiO, the final stage may be several layers of the same oxide that is studied in bulk samples.

Introduction

As soon as the cuprates were discovered, measurements of all types were made on the samples then at hand. Photoelectron spectroscopy was employed early to learn something of the electronic structure of these materials. As there were many more XPS spectrometers than ultraviolet, many XPS papers appeared in the first year of the “cuprate era.” One of the early goals was to seek evidence of Cu3+ ions in those cuprates that were superconducting, thereby demonstrating that the holes introduced by doping resided, at least partially, on the Cu sites. Cu 2p spectra existed for a series of compounds in which the Cu ions had a known valence, mostly +1 and +2, but at least one, NaCuO2, in which it was trivalent. The Cu 2p apparent binding energy increased as the valence increased, so the binding energy could be used to determine the valence to about one significant figure. (The shift is not linear with valence. The binding energy difference between Cu2+ and Cu3+ is rather small (Karlsson et al., 1992).) The search was not successful. Core-level spectra from all the other elements in the cuprates were studied, not just the Cu 2p. Valence-band spectra were measured for comparison with densities of states from band calculations. Considerable agreement was found, but not enough to say definitively that LDA calculations did or did not give an accurate picture of the electronic structure of the cuprates.

Between 1911 and 1986, superconductivity was strictly a low-temperature phenomenon. The highest critical temperature Tc for any superconductor was 23.2 K for Nb3Ge. For any possible applications, the only useful refrigerants were liquid helium and liquid hydrogen. For much of this period, an understanding of the microscopic origin of superconductivity was lacking. Bardeen, Cooper, and Schrieffer published the BCS theory in 1957, and the more general Eliashberg theory soon followed. Theoretical predictions of the highest Tc one could hope for were not very useful for finding new materials with Tc above 23.2 K. In 1986, Bednorz and Müller discovered that La2CuO4 went superconducting when Ba substituted for some of the La. Later doping studies with Ca, Sr, and Ba showed that in this system Tc reached 38 K. Soon thereafter, Wu et al. found that suitably doped YBa2Cu3O7 had a Tc of 92 K, well above the boiling point of liquid nitrogen. In the intervening years several other classes of high-temperature superconductors were found. All of these had several properties in common. All contained one or more planes of Cu and O atoms per unit cell, all had structures which were related to the cubic perovskite structure, and all were related to a “parent compound” which was antiferromagnetic and insulating. Doping this parent compound (in one of several ways) produced a metal which was superconducting and, for some optimum doping, Tc could be very high, usually above the boiling point of liquid nitrogen.

Several techniques not closely related to photoelectron spectroscopy give results similar to some of those obtained from photoelectron spectroscopy. They may also give other types of information as well. In the following we describe a few of them, but only with respect to data that can be compared with photoelectron spectroscopy. Other useful results from these techniques are not mentioned. We finish this chapter with some results from the spectroscopies closely related to photoelectron spectroscopy: electron energy-loss spectroscopy, soft x-ray absorption, and soft x-ray emission.

Infrared spectroscopy

Infrared spectroscopy probes the sample with photons of energies below about 1 eV, down to perhaps 1 meV. The interaction Hamiltonian is the same as for photoexcitation, proportional to A · p + p · A, and similar selection rules apply. The electronic excitations usually are divided into intraband excitations, often called the Drude contribution, and interband excitations. Because infrared photons have such small wave vectors, an additional scattering mechanism, phonons or impurities, is required in the Drude contribution, leading to a second-order process, and often to a strong temperature dependence. However, this can still lead to intense absorption for systems with metallic electron densities. The Drude term is the high-frequency analog of the electrical conductivity. In a superconductor, one then expects a delta function in the optical conductivity at zero energy and, perhaps, zero absorption until an energy equal to the gap energy 2Δ, is reached.

Photoemission studies generally give information on the electronic structure of a material, not the geometrical structure, although some surface structural information can be obtained by photoelectron spectroscopy. Knowledge of the crystal structures of the high-Tc materials is important in photoemission studies in several ways. First, crystal structures are required to determine the crystal potential for the calculation of the electronic structures. Even if a simple model of the crystal is being used, as in cluster calculations, the model should resemble part of the full crystal structure. Second, angle-resolved photoemission provides information in reciprocal space: energies as a function of angle or wave vector. The Bravais lattice of the crystal determines the Brillouin zone, hence the “space” in which theory and experiment often are compared. Third, the amount and nature of the anisotropy of the crystal structure, although difficult to quantify, are helpful in determining the anisotropy one might expect in physical properties, including photoelectron spectra. Orienting single crystals for angle-resolved photoemission studies also requires some structural knowledge. Fourth, there is an important experimental consideration. Most photoelectron spectroscopy is carried out on clean surfaces produced by cleaving in ultrahigh vacuum. It often is crucial to know what the cleavage surface is. Core-level photoelectron spectroscopy can help identify the atomic character of a cleaved surface, but knowledge of the crystal structure is needed and some idea of interplanar bonding is helpful. Fortunately, a structure determination is usually one of the first studies made on any new material.

The key instrumentation necessary for photoemission consists of a radiation source, an electron energy analyzer, and a means of preparing and maintaining a clean sample surface in ultrahigh vacuum. Ancillary sample characterization capability, both in situ and ex situ, is also important, especially for non-stoichiometric samples like high-temperature superconductors. We discuss each of these in turn, emphasizing some of the important attributes of each. We do not go into enough depth to allow a reader to construct the instrumentation. We emphasize the crucial characteristics of resolution and signal-to-noise ratio, and how they are limited by instrumentation.

Radiation sources

Work functions may be as high as 5 eV, so light sources for photoemission must emit photons of energies higher than this. Since O2 and N2 absorb photons starting just above 6 eV, the entire optical path must be evacuated to a pressure below about 10–5 Torr in order to ensure significant transmission over light paths of a meter or more. For this reason, the spectral range above about 6 eV is called the vacuum ultraviolet region, Its name changes at around 50–100 eV to the soft x-ray region, for which the optical path must also be in vacuum.

Radiation sources used in the earlier valence-band studies (Berglund and Spicer, 1964b) often were hydrogen discharges. These gave a weak continuum from the H2 molecule and a strong emission line, the Lyman-α line at 10.2 eV, from the H atom (Samson 1967).

We have already mentioned some incremental improvements in photoelectron spectroscopy, increased energy- and angle resolution through improvements in instrumentation. Another is the angle-scanned photoelectron spectroscopy pioneered by Aebi et al. (1994a,b), described in Chapter 5. Lindroos and Bansil (1996) recently pointed out additional advantages of this technique when coupled with suitable calculations. They found not only the usual features produced by direct transitions from initial states on the Fermi surface, but additional features from indirect transitions whose intensities are proportional to the one-dimensional densities of initial states for transitions at a fixed k∥. They illustrated these aspects with calculations for several surfaces of Cu. A re-examination of such spectra for Bi2212 is likely to occur soon. There are several other variations of photoelectron spectroscopy already tested at some level, some of which may play a role in future work on cuprates.

Photoelectron microscopy

The photoemission studies normally carried out collect electrons from an illuminated area usually no smaller than about 100 μm × several hundred μm. There may be lateral inhomogeneities on a scale smaller than this, so photoelectrons from special sites may be averaged with those from the rest of the area sampled. This may be particularly misleading in studies of reactions with overlayers which may take place preferentially at steps. Photoelectron microscopy has been developed over the past fifteen years or so, along with x-ray microscopy, which uses absorption differences for contrast.

Photoelectron spectroscopy, especially XPS, can be used to study some aspects of the surface chemistry of cuprates. Examples of this are the changes seen in the photoelectron spectra of Y123 as a freshly prepared surface deteriorates with time at 300 K in ultrahigh vacuum. In this, the evidence that the changes are due to loss of oxygen is not direct. A better example, to be discussed below, is the study of a metal overlayer on a cuprate. The overlayer may just sit there, making it a candidate for an ohmic contact, or it may disrupt the surface, replacing some of the atoms in the cuprate, which then appear in the surface layer. The appearance or disappearance of core-level peaks can be used to track the depths of constituents, and binding-energy shifts may give clues about oxidation states and location of atoms, i.e., surface or bulk.

Most of the studies of interface reactions have been carried out on the same type of surfaces used in the study of the cuprates themselves, cleaved single crystals or scraped or fractured polycrystalline samples. These cannot be considered technologically important surfaces unless processing in the future is carried out under ultrahigh vacuum conditions. Some of the studies reported below may need to be repeated on air-exposed surfaces, or surfaces handled in inert atmospheres or poorer vacua.

Introduction

Photoemission studies may be said to have begun in 1887 with the observation by Heinrich Hertz that a spark between two electrodes was obtained more easily if the electrodes were illuminated (HeTrtz, 1887), although this occurred before the discovery of the electron. Improved experiments during the next several decades were important to the development of quantum mechanics. The modern era of photoemission spectroscopy arguably may be said to have started in 1964 with the papers of Berglund and Spicer (1964a,b). Although there had been prior experimental work and calculations in which Bloch electrons were assumed and dipole matrix elements calculated, it was these two papers, and papers by Gobeli et al. (1964) and Kane (1964), which stimulated a large amount of work, coming as they did nearly simultaneously with the widespread calculation of accurate electronic structures of many materials and the commercial availability of ultrahigh vacuum components. The first paper of Burglund and Spicer worked out what is called the three-step model for photoemission, and the second applied it to new data on Cu and Ag.

In the following we outline several approaches to the description of photoelectron spectra, starting with the most simple conceptually, then progressing to a more sophisticated picture. As we do this, we point out some assumptions and approximations often made, sometimes tacitly, and whether or not they may be important for photoelectron spectroscopic studies of high-temperature superconductors. Experimental considerations are presented in Chapter 4.

Introduction

Although Y123 and Bi2212 both contain CuO2 planes and have similar values of Tc, there are notable differences which have affected the course of the study of each class of material. We have already mentioned two such differences, the reproducible cleavage of Bi2212 and the stability of the cleaved surface in ultrahigh vacuum. The cleavage plane(s) of Y123 is not really known and may vary from cleave to cleave. Moreover, a single plane is unlikely, and the atomic nature of the exposed surface may change at cleavage steps. Surface reconstruction, or at least relaxation, may be possible, even at low temperature. The surfaces of Y123 generally are not stable in ultrahigh vacuum except at temperatures below about 50 K. However, exceptions to this have been found by several groups, but the reasons for this stability are not yet known. Tc may be varied in Bi2212 by the addition or removal of oxygen and there is an optimum oxygen content at which Tc is a maximum. YBa2Cu3Ox also has variable oxygen stoichiometry, but between x = 6.8 and x = 7, Tc remains very close to its maximum value. In Bi2212 the oxygen content changes occur on or adjacent to the Bi–O planes. In Y123 the changes occur in the Cu–O chains, although the holes provided by the oxygen atoms are believed to reside on the CuO2 planes (Cava et al., 1990). Starting from x = 7, removing oxygen atoms from the chains should lower the hole concentration.

NCCO

Nd2–xCexCuO4–y (NCCO) was the first cuprate superconductor to exhibit n-type conduction. The substitution of Ce4+ for Nd3+ introduces electrons which go to the CuO2 planes. (There are no Cu–O chains.) Tc peaks at 24 K for x = 0.15. The importance of NCCO is that is can be used to address the question of electron–hole symmetry. The electron doping swells slightly the large Fermi surface, while hole doping in other cuprates shrinks it slightly. In the localized picture with its small Fermi surface, how do the electron and hole Fermi surfaces compare? NCCO has a simpler LDA band structure near EF than many other cuprates because there is only one CuO2 plane per unit cell. Fewer bands cross EF. What are the new states in NCCO the doping introduces along with the new electrons, or do those electrons occupy existing states as in the independent-electron picture? Other trivalent rare earths may be substituted for Nd. Why do the magnetic moments of the rare-earth ions not destroy superconductivity as they do in many other superconductors? A brief review of photoemission studies of NCCO was written by Sakisaka (1994).

Most of the studies of NCCO have been carried out on polycrystalline, i.e., sintered, samples. The earliest study on a single crystal (an epitaxial film), that of Sakisaka et al. (1990a,b), was mentioned in Chapter 6. Since then, and since the review of Sakisaka was written, there have been several more studies on single crystals.