Superconductivity in d-and f=Band Metals

possible.

d- AND f-BAND SUPERCONDUCTIVITY - SOME EXPERIMENTAL ASPECTS¹

T.H. Geballe,     Department of Applied Physics and Center for Materials Research, Stanford University, Stanford, California; Bell Laboratories, Murray Hill, New Jersey

Publisher Summary

A branch of superconductivity, called d- and f-band superconductivity, was recognized as having an identity of its own. Nature provides a rich variety of superconductors in which the d- and f-bands play an important role. This chapter provides an overview of the major themes of d- and f-band superconductivity. New superconducting systems are continuously being found in d- and f-band compounds, although it is not as easy to do as it once was. As a result there has been a slow rise in the highest known transition temperature as a function of time. The vast number of solid solutions and intermediate phases, formed by the transition metal elements, shows that systematic superconducting behavior could be predicted simply by finding a ratio of valence electrons to atoms. The wide range of homogeneity in d- and f-band systems offers the opportunity for relating superconductivity to the normal state microscopic parameters. The mechanical properties of superconductors are limited.

INTRODUCTION

Ours is a subject with roots that go back to the beginnings of superconductivity. In the first edition of Shoenberg’s Superconductivity (1938), following Kurti and Simon (1935), the known superconductors were grouped according to the Periodic Table into transition metals and nontransition metals. Along the way to our present state, some art, some fiction, and quite a bit of science has been produced. The Periodic Table has proved to be a most useful vehicle for predicting new superconductors as evidenced in the work of Matthias, Hulm and others (Roberts, 1976). Douglass in initiating these conferences (1972, 1976) recognized that a branch of superconductivity exists called d- and f-band superconductivity which has an identity of its own. Nature provides a rich variety of superconductors in which the d- and f-bands play an important role.

An attempt to identify the major themes of d- and f-band superconductivity is given in Fig. 1. The discussion given below follows from the outline.

FIGURE 1 Attributes which define d- and f-band superconducting research.

OCCURRENCE AND VARIATIONS OF Tc

New superconducting systems are continuously being found in d- and f-band compounds, although it is not so easy to do so as it once was. As a result there has been a slow rise In the highest known transition temperature (Tc) as a function of time. It is reasonable to assume that modest increases in Tc will continue. The most likely route is via further stabilization of one of the presently known high-Tc metastable A15 phases, or some derivative. However, there is always the remote possibility that some completely unexpected new superconductor will provide much higher Tc ’s. CuS must have been a complete surprise when Meissner found it to be superconducting 50 years ago and I suspect if CuCl were to be shown to be superconducting today it would evoke even more surprise. The large diamagnetic anomalies which have been observed above liquid nitrogen temperatures as transient signals in CuCl under pressure (Brandt et al., 1978; Chu et al., 1978) have not yet been proven to be associated with superconductivity. A summary of the status of CuCl research will be given by Chu later in this conference.

The vast number of solid solutions and intermediate phases formed by the transition metal elements show systematic superconducting behavior that Matthias first pointed out could be predicted simply by averaging over the Periodic Table and finding ratio of valence electrons to atoms. The behavior is very well summarized by Hulm and Blaugher (1972) in their paper at the first d- anf f-band conference. Little can be added except for the work going on in the 5f series which is reviewed by Smith in the next session.

The simple dependence of Tc upon the electron to atom ratio is applicable to many structures when other variables such as degree of order either do not affect Tc, or can be maintained constant. However, there can be a strong dependence of Tc upon structure, and frequently is, when the structure gives rise to flat bands, soft phonons and nesting Fermi surfaces that affect the microscopic superconducting parameters. One only has to note that amorphous and heavily disordered Nb3X (X = Ge, A1, Sn) have transitions near 3K whereas the corresponding compositions with the A15 structure have transitions above 18K. A further illustration of the effect of structure, or rather lack of structure, is gained by comparing the Colliver-Hammond (1973, 1977) curve for Tc ’s found for the amorphous 4d and 5d series with their crystalline counterparts. The former give a Tc vs. composition curve that maximizes at or near the half-filled d-band, in contrast to the minimum found in the corresponding crystalline series. The Tc minimum and the adjacent maxima at e/a −4.8 and 6.7 correlate well with the density of states at the Fermi-level in the crystalline series as determined experimentally and can be understood from the theoretical model of Varma and Dynes (1976). Empirical considerations similar to those used in Varma and Dynes (1976) applied to the amorphous 4d and 5d series (i.e., the constancy of /M<ω²> in a given class) would again lead one to expect Tc to correlate with the density of states. The maximum would now be expected to be where the spherical Fermi surface has its maximum area − near the half-filled band.

In the case of ternary compounds which have more than two sites and elements per unit cell there is no good way that I am aware of for predicting superconductivity. Empirical generalizations will be forthcoming as new ternary systems are found and, as Fischer shows in this conference, they can already be made within given systems. Within the large unit cells that exist there is room for ionic and covalent, as well as metallic bonding and a wide range of chemical substitution can frequently be accomplished. The contact interaction of the conduction electron wave function at a given site varies when, by the substitution, localized ionic or covalent orbitals are caused to mix with the conduction band. The valence electron to atom ratio then becomes an arbitrarily defined variable. The mixing of localized magnetic states gives rise to interplay between superconductivity and magnetism in the Chevrel and RE-boron phases about which much fascinating information will be presented in the following sessions.

There is also the possibility of keeping the essential bonds fixed, while changing volume. TaS2 can be intercalated with the series of aliphatic amines containing from 1 to 18 carbon atoms (Gamble, et al., 1971). At first, for the shorter chains which lie with their axes parallel to the layers, the number of bonds per unit area between the nitrogen of the amine and the metallic layers surrounding it decreases as the length of the aliphatic chain increases; Tc gradually decreases with chain length. On the other hand the larger chains stand perpendicular to the layers, the number density of nitrogen-layer bonds is constant; Tc remains constant while volume per bond (layer spacing) increases with increasing chain length. More recently metallocences (Gamble and Thompson, 1978) with and without uncompensated spins have been intercalated between the layers. The former follow an expected Curie law with little effect upon Tc. The insensitivity of Tc to the presence of dense layers of spins suggests that the chalcogenide layers may not be Josephson-coupled, leaving open the question of how the superconducting, and the eventual magnetic ordering required by the third law of thermodynamics, take place.

MECHANISMS AND MICROSCOPIC PARAMETERS

The wide range of homogeneity in d- and f-band systems offers the opportunity for relating superconductivity to the normal state microscopic parameters. Heat capacity measurements are particularly convenient when new and/or difficult-to-prepare superconductors are studied. They can be used to establish whether or not observed electrical and magnetic transitions are representative of the bulk of the sample as well as for extracting averaged normal state and superconducting microscopic parameters. We shall hear of one such investigation in the next session (Stewart et al., 1979).

Superconducting tunnel junctions as discovered by Giaever (1960) can have I-V characteristics that directly measure the energy gap, and that can be analyzed, as shown by Rowell and McMillan (1969) to give the spectral function of α² f(ω), and the Coulomb pseudopotential μ*. Unfortunately the great reactivity of the d- and f-band materials makes it very difficult to fabricate junctions that can be analyzed satisfactorily. Shen (1972) presented the first satisfactory tunnel junction data for Nb3Sn at the first d- and f-band conference. Progress has been made since then using the coevaporation techniques developed by Hammond (1975, 1978). Some recent data are shown in Fig. 2. The 50-odd Nb3Sn junctions studied as a function of composition by Moore, et al. (to be published), are seen to become strongly coupled only at and near the ordered stoichiometric structure. This is a microscopic manifestation that special features of the A15 structure which enhance the superconductivity are a property of the ordered stoichiometric structure. Application of the analysis of Wolf and Arnold (to be published), which allows for degradation of the A15 structure at the interface via the proximity effect, gives reasonable values for μ*. Also in Fig. 2 are some data for Nb3Al junctions being studied by Kwo et al. (1979). Well-behaved junctions for off-stoichiometry compositions have been obtained which suggest that Nb3Al is even more strongly coupled than Nb3Sn.

FIGURE 2 Compositional dependence of energy gaps for A15 phases of Nb-Sn ( Moore et al., 1979) and Nb-Al (Kwo et al., 1979).

The difficulty of fabricating junctions of compounds such as Nb3Sn has prevented their being considered in the present stage of development of superconducting circuitry where reproducibility of the individual elements is of paramount importance. Nevertheless, the larger gaps and potentially faster switching characteristics of A15 junctions should be of increasing interest as more is learned about fabrication. When the rather heroic methods used by Shen to prepare the first junction are compared with the now routine methods used for the preparation of those shown in Fig. 2, it is evident that substantial progress is being made.

INTRINSIC PROPERTIES

The intrinsic superconducting properties of type II transition metal superconductors become of intense interest following the discovery of the existence of high-field, high-current superconductivity by Kunzler and coworkers (1961). The possibility of finding improved properties that can directly be put to practical use offers support for this field of research. Again the large homogeneous compositional ranges available make systematic studies and quantitative comparison with the theory of type II superconductivity possible. Recent work of Orlando et al. (to be published) for Nb3Sn shows that it is necessary to bring strong-coupling and the electron phonon enhancement in on an equal footing with the spin-orbit scattering before the effects of Pauli paramagnetism at high fields can be understood on a quantitative basis.

The maximum known Hc2 as a function of time has increased drastically since 1960, It is perhaps no coincidence that the highest known critical fields and critical field slopes are found for ternary compounds, for it is in these compounds with large unit cells which contain electrically inactive sites that unexpected behavior in high magnetic fields is found. This is true for the Chevrel phases (Fischer, 1978), the RE borides (Fertig et al., 1977) and the intercalated layered metal dichalcogenides (Prober et al., 1975). The field is able to penetrate perpendicular to the c-axis of the dichalcogenides with so little cost in energy that enormous critical field slopes

are found. This enormous slope has motivated theoretical studies of critical fields of the layered compounds which show how novel the superconducting properties of such open structures can be (Klemm et al., 1974). Possibly related, but still not understood are the negative curvatures found in the low field portion of the Hc2-T curves (Woollam, 1974). In the Chevrel and rare earth boride phases it is not so straightforward from the structure alone to see why the critical fields can be so high; however, the intercluster linkage must play an important role.

It is clear that we have a long way to go before magnets and other kinds of devices will be limited by an intrinsic inability to remain superconducting in high magnetic fields. The limitation is more with the mechanical properties of the superconductors, the intrinsic properties of the various kinds of pinning centers and the interaction of the flux-line-lattice with distributed pinning centers. While the fabrication of composites seeking to optimize simultaneously all these properties is more nearly at the center of gravity of the Applied Superconducting Conference, there is plenty of work still to be done of interest to this conference, in developing a quantitative understanding of each of the parameters involved. Quantitative understanding can eventually lead to much higher performance superconductors.

EXPERIMENTAL IDIOSYNCRASIES

Finally of great interest to this conference is the relationship of high temperature superconductivity with instabilities and metastable phases. It need only be noted that the three highest Tc superconductors, Nb3Ge, Nb3Ga, and Nb3Al, are all metastable. Only over a narrow temperature range at the peritectic does their equilibrium composition even approach (but not include) stoichiometry (see for example Jorda et al., 1978). It follows that quench techniques must be employed in order to obtain stoichiometry.

No convincing experimental data exist, to my knowledge, that show Nb3Al, Nb3Ge, Nb3Ga, or Nb3Si has been made homogeneous and stoichiometric in the ordered A15 structure, in either film or bulk form. It remains as a challenge to do so in order to be able to compare them with Nb3Sn, for example. Epitaxial vapor phase growth by Dayem et al. (1978) is a potentially powerful method. The intent in this method is to prepare the substrate with a surface favoring the free energy of formation of the desired phase with respect to undesired more stable phases. The work on Nb3Ge (Dayem et al., 1978) showed that growth of the desired A15 phase and composition is favored when it takes place on prepared Nb3Ir films of the proper lattice constant.

The favorable role of O2 during growth can also be such as to encourage epitaxial growth as shown by Gavaler et al. (1979). Feldman et al. (1979) have found that the presence of O2 during growth of Nb3Si films favors the formation of the A15 structure much as it does for Nb3Ge. The results obtained so far, shown in Fig. 3, are the first study of Nb3Si in the A15 phase where systematic variations can be observed and reasonably interpreted.

FIGURE 3 Lattice constant vs. composition for Nb3Ge films (Dayem et al., 1978) compared with corresponding data for Nb3Si (Hammond et al., 1972; Waterstrat et al., 1979; and Feldman et al., 1979).

Resistance measurements can be valuable for correlating Tc with the microscopic parameters of transport theory. However, in cases like Nb3A1, Nb3Ge, and Nb3Si, which heat capacity measurements show are not homogeneous, both the superconductivity and the transport can be dominated by (not necessarily the same) minority phases or compositions. A more reliable set of parameters is the combination of lattice constant, and composition. Evidence is shown by Cox et al. (1979) at this conference for Nb3Ge, and earlier for Nb3Sn by Burbank et al. (1977) that for a given composition the lattice constant can be a single valued, experimentally determined function of disorder. There still can be ambiguity in comparing different films in that different concentrations of vacancies, defects (antisite and displacement) and impurities can give the same lattice constant. Lattice constant vs. composition curves nevertheless form the most simple reliable basis for characterizing films. This is illustrated in the comparison of Nb3Si and Nb3Ge, Fig. 3, where the composition determined by electron microprobe analysis gives the ratio of Nb to X. The Nb3Si line extrapolates to ao = 5.08 Å, which is the predicted value of the revised Geller radii. The limiting factor in obtaining stoichiometry may be the packing of Nb atoms along the chains of the A15 structure. The Nb-Nb bonds may, due to core-core overlap, no longer be easily compressible for ao < 5.12. If no Nb3X compounds can be made in the A15 structure with ao < 5.12 (X is a non-transition element) then the route to higher Tc may be by partial substitution on the Si site. This has been attempted many times. The lack of success can be attributed to the fact that there are many more disordered, undesired stable or metastable phases that can be reached than the desired ordered one. It follows that a high degree of control over the growth parameters is essential. Experimentalists seeking to synthesize metastable phases have a constantly increasing arsenal of materials, techniques and characterizations available to help them. The most important ingredients do not have to be listed − they are the happy combination of inspiration and hard