Superstructures and Superconductivity

wavy fringes

exhibits marked sequence similarity with a family of genes that include several transcriptional regulators. Thus, other putative repair helicases may also function in transcription.
The new findings presented in this issue suggest that the transcription machinery may serve a second role. Rather than using distinct proteins for scanning transcribed regions and identifying damaged DNA. the cell appears to use an existing enzyme to carry out these tasks: the RNA polymerase elongation complex.

Zachary Fisk and Gabriel Aeppli
A n area of research now largely hidden in ehe shadow of high transition temperature (T<) cuprates is that of heavy fermion superconductivity. Heavy fermion materials-so named because their conduction electrons behave as though they had extra mass--are like the cuprates in that they exhibit unusual superconducting properties. By the time the cuprates had been discovered, a good understanding of these materials was in hand. Unlike theories of high-T< superconductivity, however, ideas about heavy fermions have not been ehe subject of great controversy. Thus, most of the effort in this backwater of condensed matter physics has focused on certain derails of ehe behavior of one particularly wellstudied compound, UPt 3 • The cause for sustained interest was that the process of developing ever more elaborate explanations for ever more elaborate experiments did not seem to converge. A recent paper by Midgley et al. ( 1) reporting modulations in the crystal lattice of UPt 3 suggests that theory and experiment might finally converge in a way that, while it does not threaten the broad understanding of heavy fermion syscems, involves a degree of freedom ignored until now even in ehe face of past experience with elemental metallic uranium.
The heavy fermion materials are intermetallic compounds with effective conduction electron masses of order l 00 times that of the free electron. This is seen, for example, in the enhancement of the electronic-specific heat coefficient measured at low temperatures. The origin of this large mass lies in the compensation of a magnetic moment on one of the atomic constituents, typically uranium o r cerium, by conduction electrons in the compound: the antiferromagnetic interact ion between conduction electron spin and atomic magnetic moment results in a "many-body" covalent state.
Much of the interest in heavy fermion compounds deri ves from the d iscovery of several superconductors in their midst. Steglich and collaborators found the first such , CeCu 2 Si 2 , in l 979 and other groups discovered rwo more examples, UBeu and UPt 3 , in the first half of the l 980s (2). Their Tc's are all below l K, and hence of little foreseeable technological interest, but unusual in that they were superconducting at all. The Bardeen-Cooper-Schrieffer (BCS) theory of superconductivity, involving the condensation of electron pairs below Tc, provided a good explanation for a large body of experimental results on conventional superconductors, including the catastrophic effe<:ts of magnetic momentbearing ions such as cerium and uranium on superconductivicy. Specifically, magnetic moments generally break pairs, owing to the tendency of the magnetic moment to make the rwo spin members of each pair parallel rather than antiparallel. In fact, the heavy fermion materials at high temperatures contain uncorrelated magnetic impurities at un- precedented density for superconductors.
Thus, the discovery of superconductivity in this unexpected place summoned forth hordes of rheorists. Motivated by pioneering experimental and theoretical work on superfluid 3 He (3), attempts were made to account for the heavy fermion superconductivity via different kinds of Cooper pairing, generally produced by a mechanism other than electron-phonon (4).
lt has been a long cherished hope that some mechanism other than phonons might give rise to pairing, leading to higher Tc's. The characteristic energy scales for phonons in merals is the Debye temperature, usually around room temperature. Transition temperarures might be expected to reach values an orderof magnitude smaller chan this, say 30 K. Other mechanisms with higher energy scales could be expected to have correspondingly h igher T c's, and many believe that they are relevant for cuprates. The heavy fermion superconductors held out the first solid hope for a new type of pairing and a new mechanism. UPt 3 has been in many ways the darling of the heavy fermion superconductivity community. lt has a simple cryscal structure and large single crystals are easily prepared. Lornarich's group at Cambridge has mapped out muchof the Fermisurface (5), which is in general possible only in a nearly perfect material. A n extensive body of experimental data on the properties of UPt 3 accumulated rapidly. ln particular, measurements of ultrasonic absorption ( 6) and, later, magnetic penetration depths (7) and vortex lattices (8) (see figu re), indicated that thesuperconducting state is anisorropic eo an unprecedented degree for a relatively isotropic material such as UPt 3 • Also, substan tial antiferromagnetic flucruations were found to appear in the coherent, metallic state (9). All of these results fit neatly into a picture of unconventional pairing mediated by antiferromagnetic fluctuations.
Experimentalists, undeterred by ehe tidy phenomenology just described and emboldened by steady improvements in sample size and qualicy, persisted in collecting data on UPt 3 • They discovered two interesting facts. The first was that magnetic order sets in at 5 K, considerably above the 0.5 K supercon-ducting transition ( 10). Both the ordered moment and the antiferromagnetic correlation lengrh are parhologically small. In addition, the ordered moment decreases below Tc ( 11 ), indicatingthatthesuperconductingand magnetic order parameters are coupled.
The second discovery was that, depending on extemal field and temperature, there appeared tobe more rhan one superconducting state: acoustic measurements revealed multiple transitions as a function of field strength, while specific heat showed two transitions separated by less than 0.1 K (12). T wo transitions cannot occur for simple BCS pairing: it is a sign of more complicated pairing. The magnetice field-temperature (H-T) phase diagram for the two transitions was refined experimentally ( J 3) and theoretically ( J 4). Detailed comparisons between the two showed wonderful agreement.
Not everything was wonderful however. Virtually all theories involved a combination of small moment magnetism and structural changes in ehe von ex latrice (UPt 3 is a strongly type II superconductor) ro account for the H-T phase diagram. Neutron diffraction experimems (7) showed that the flux lanice evolves smoothly rhrough the phase boundary crossed by varying H at low T, and so make it unlikely that this boundary is due to a change in the vorrex correlations. At the same time, the known anisotropies of the magnetic and superconducting order parameters did not appear consistent with how the H-T diagram varied with field direction. Funhermore, it was difficult co see how the shorr coherence lengrh magnetic state could lift the degeneracy of the superconducting state whose coherence lengrh was also of the same size. And in one experiment the crystals showed two inductively measured transitions, something most easily explained as extrinsic, not inrrinsic: superconductors display complete diamagnetic shielding, making it difficult to see how a transition from one superconducting state co another could manifest itself in an inductive measurement. Finally, and most worrisome, was the discovery that in URu 2 Si 2 , another heavy fermion superconductor with a small antiferromagnetic moment, different superconducting transitions were associated with macroscopically different parts of the sample ( 15). the complicated superconducting phase diagram of UPt 1 derives from the intemal strain field caused by the modulation, and that this strain field lifts the degeneracy associated with unconventional pairing.
The daim of Midgley et al. is rhat their annealed samples are homogeneous over domains !arger than 10,000 A, and much !arger than ehe superconducting coherence length. Thus, the modulation can produce a resolvable double superconducting transition. The observation that in an unannealed crystal with a single broad superconducting transition the correlation lengrh for the modulation was much shoner than I0,000 A reinforces this connection.
Where does this leave us? The basic ideas relating to the superconductivity of heavy fermion materials remain intact. Bur ehe pristine way in which UPt 3 allowed incredibly detailed comparison berween theory and experimem may have vanished, at least until theorists add several more terms to their Hamiltonians. Funhermore, apart from ignorance as to ehe characteristics or even existence of the modulation in samples other than the very thin (1000 fi...) and intensely handled TEM sli vers, we have yet to determine whether the small-moment antiferromagnetism is derived from ordinary defects such as stacking faults or from walls between domains with differently oriented modulations. Also, we do not know if there are unmodulated domains: might not, for example, there be coexistence of modulated and unmodulated pans of the single crystals? More co ehe point, a URu 2 Sir like origin for ehe double superconducting transition has not been ruled out.
UBeirbased superconductors now remain ehe only materials where there is still confidence thar rwo transitions can occur reproducibly, without marked sensitivity co annealing protocol and questions about macroscopic sample homogeneity. An amusing aspect of the developments on UPt 1 is that they are reminiscem of the nearly forgotten SCIENCE • VOL. 260 • 2 APRIL 1993 PERSPECTI VES situation in elememal uranium (17). In that case, a controversy existed for years over whether or not the element was superconducting at all at ambient pressure. lt tumed out that uranium develops a charge density wave state below 40 K, and that this competes with superconductivity. Application of pressure at several kilobars to uranium suppresses the charge density wave somewhat, and superconductiviry appears at Z K.
In contrast to the confused picture of the high-Tc materials, there is substantial support for unconventional pairing and a magnetic mechanism ( 18) in the heavy fermion materials. Indeed, the most obvious microscopic explanation of the split rransition in terms of the Midgley er al. data requires unconventional pairing of the type associated with a magnetic mechanism. At the same time there is an important lesson here, one perhaps even more important in the search for answers to the same questions in high-Tc superconductivity. The electronic properties of materials such as heavy fermion intermetallics and high-Tc cuprates can be pathologically sensitive to local structural details. Suchdetails are often subde and difficult to extract experimentally. As more and more complicaced materials are invescigaced, ic is weil to keep Murphy in mind and, in ehe case ofheavy fermions, to beware of superconductors bearing uranium.