LARGE MAGNETIC-FIELD EFFECT ON THE HEAVY FERMION SUPERCONDUCTOR U1-XTHXBE13 JUST ABOVE TC

Magnetic fields as small as 4 kOe substantially increase the resistivity of U 1 _ x Thx Be 13 for x = 0.0175 and 0.0378 between l.4 K and Tc, the superconducting transition temperature ofthis heavy fermion material. At the intermediate value of x = 0.026, Tc reaches a local maximum, and no normal-state field dependence is observed up to 4 kOe. At x = 0.0603, away from the region of nonmonotonic dependence of Tc on x, again no field dependence is seen. The observations are discussed with regard to specific-heat measurements and in tenns of the interaction of superconductivity with narrowband features.


INTRODUCTION
The intennetallic compound UBe 13 is a member ofthe small but growing cJass of so-called heavy fermion materials. Suchsystems at low temperatures are characterized by large, temperature-independent susceptibilities and large linear electronic specific-beat coefficients r. These enhanced r values correspond to an effective electronic mass of the order of 100 electron masses. Such enhanced properties suggest a description of the low-temperature state of these systems as an electronic Landau-Fermi liquid, the classic example of which is normal 3 He.
Three of these heavy fermion systems undergo a superconducting transition below 1 K, notably CeCu 2 Si 2 , 1 UBe 13 , 2 and UPt 3 • 3 What is remarkable, focusing on UBe 13 in particular, is that the magnitude of the specific-heat jump at the superconducting transition temperature is enormous, and of the same order of magnitude as the low-temperature r value of 1.1 J/mol K 2 • This indicates that it is indeed an electronicaUy enhanced r and that the electrons responsible for the unusual normaJ-state properties are the same electrons which bring about superconductivity.
There have been suggestions that the nature of the Cooper pairs in tnese heavy fermion superconductors is not of the conventional, singlet BCS type. There have been proposa.ls that UBe 13 4 and UPt/· 6 are anisotropic (L #0) superconductors by analogy with the L = 1 triplet polar state of superfiuid 3 He at very low temperature. In contrast, Ce-Cu2Si2 has been modeled as a Kondo lattice system with a conventionai BCS ground state. 7 • 8 Nonmagnetic impurities have been shown to substantially modify both the normal and superconducting properties of UBe 13 . 9 • 10 Substitution of thorium, which has no 5/ electrons, for uranium in compounds of the form U 1 -x Th ... Be 13 has been the most heavily studied case. The Tc is strongly depressed by Th in the x = 0--0.06 concentration limit, and between about x = 0.017 and 0.026 shows an extremely unusual increase of Tc, as determined by resistivity or (less dramatically) by ac susceptibility . 9 In addition, as x is increased, the sboulder in the resistivity at about 25 K becomes a broad peak and moves to higher temperature. This feature looks as if it is caused by crystal field scattering, but no crystal field levels have been observed to date. 9 Tue narrow low-temperature peak, at 2.5 K for pure UBe 13 , is similar to a feature in the resistivity of CeCu 2 Si 2 , 11 which has been ascribed to the development of the highly correlated Fermi liquid state. This peak moves to lower temperatures rapidly on Th addition, disappearing below Tc for x>0.026. The net effect of the motion of these two features is that at a particular low temperature, the resistivity over a range of x values decreases with increasing x.
More unusual behavior yet has been observed recently in the Specific-heat cp data Qf single-phase u t _ X ThxBe13 below Tc. 12 Between about x = 0.0216 and 0.0378 a second transition is observed below Tc. lt is argued that this can only be a transition between superconducting configurations with different L =FO pairing states, although there is some c.oncern about how to interpret entropy balance for these sampies below Tc. Very recent NMR results 13 also reject the possibility that the second transition is structural or magnetic in nature. The spin-lattice relaxation rate l/T 1 is roughly proportional to T 3 for x = 0 and x = 0.033. This is the expected temperature dependence for an energy gap that vanishes on one or more lines on the Fermi surface, which is the case for an L = 1 triplet polar state superconductor.
Because small Th additions produce large effects in UBe 13 , we thoug.ht that modest magnetic fields might also strongly affect the properties of U 1 _ x Tb" Be 13 , and indeed they do.

H. EXPERIMENT Al PROCEDURE
High-purity U, Th, and Be were melted together in a conventional argon atmosphere arc furnace. The buttons were turned and remelted at least seven times. Compositions were corrected for weight loss using the measured ratio of constituents in the arc fumace residue. 12 Samples were subjected to x-ray diffraction and metallographic analysis. All specimens were measured in the as-cast condition. Semicircular slices were spark cut from the buttons. A standard four-probe ac resistivity technique in a dilution refrigerator and superconducting magnet were employed in the measurements.

III. RESUL TS
Tue main results are summarized in Fig. 1, which shows the resistivities of U 1 _ x Thx Be 13 as a function of T for several values of applied magnetic field H. For x = 0.0175, the low-temperature peak in the resistivity moves to lower temperatures in an applied field until it appears to move below Tc for H-4 kOe. Putting it another way, tbe lowtemperature decrease in the resistivity is inhibited by a magnetic field. For x = 0.026, the sample with x nearest to the relative maxirnum in Tc, the magnetic fieid has essentially no effect on the resistivity. Near the end of the anomalous region in x at x = 0.0378, the H = 0 resistivity shows no lowtemperature peak; the slowly falling resistivity as T decreases is the low-temperature side of the broad, high-temperature ( -30 K) feature. Tbere does, however, appear to be a rather more rapid decrease in the resistivity below l K than for x = 0.0175. The application of a fieid inhibits thisdecrease, asit does in thex = 0.0175 case. Finally, for x = 0.0603 (not shown) there is a decrease in the resistivity below the 30-K peak all the way to Tc. This is in a region where only one transition is observed. in CP data, and here only a very slight increase ( -1 % ) of the normal-state resistance is observed in a 4-kOe field just above Tc.
All of the compositions studied have a very large ( -dHc 2 /dT)r. Magnetic field corrections to the thermo-, metry account for more than one half of the measured Tc shift. Thus the determination of the Tc depression is somewhat inaccurate. Furthermore for x = 0.0175, as seen in Fig.  1, it is rather difficult to assign a Tc because the nonnal-state resistance is dropping so rapidly. Therefore, we can only state that ( -dHc 2 / dT Jr for 0.0178 < x < 0.0603 is the c same order of magnitude as the value for pure UBe 13 , which is 420 kOe/K. 14

IV. SAMPLE QUALITY CONSIDERATIONS
We have rejected the possibility that the nonnal-state magnetic field effects observed here may actually be caused by free Th filaments in the sample. Pure thorium has a Tc of 1.374 K and Hc 2 (0) of 162 Oe. This is approximately the same temperature for which field-dependent etfects occur for x = 0.0175 and 0.0378. Ifit is Th superconductivity, then the observed "Hci" is much too large and the transition width is extraordinarily wide, greater than 0.8 K, which would be quite broad for a precipitated element. Furthermore, one would expect the amount of free Th to increase with x, but no excess conductivity is observed for x = 0.0089, 9 0.026, 0.0603, and 0.0675. 9 In addition, the shape of the resistivity curves for x = 0.0378 and x = 0.034 of Ref. 9 are very similar near Tc, suggesti.ng a systematic trend with x rather than random variations with free Th.
MetaUography and x-ray diffraction analysis also reveal these samples tobe essentially single phase. Very minor ( -l % ) amounts of U0 2 and Be were observoo by x-ray diffraction. Tue presence of free Be implies that the stoichiometry is on tbe (U, Th) poor side ofthe U 1 _x11t.., Be 13 compound, and thus free U or Th would be very unlikely. In contrast, flux-grown "single crystals" were of much poorer quality. Fractionation of Th in the melt was clear from lattice parameter detennination, and an additional unidentified phase was found. SmaU amounts of free superconducting Al (the ftux) were observed in the resistivity measurements. These transitions were very sharp and were quenched by very small magnetic fi.elds on the order of 100 Oe. Again, all of the measurements reported here are on arcmelted material. The above remarks lead us to believe that tbe observed resistivity curves are intrinsic to U 1 _ x Th.., Be 13 and are not caused by second phases, in agreement with the conclusions in Ref. 12.

V. D!SCUSSION
The nature ofthe 2.5-K feature in the resistivity is not known; however, it should be noted that the position of a similar peak in CeCu 2 Si 2 , which is attributed to Kondo lattice scattering, has a strong infiuence on superconductivity in that heavy fermion system. 11 • 1 5 Superconductivity is de- stroyed in CeCu 2 Si 2 if this peak moves frorn near 20 K to below 6.5 K, which occurs for some stoichiometries; in UBe 13 , Tc is also depressed as the 2.5-K feature moves to lower temperature upon Th substitution. Although neither the extreme magnetic field dependence of the low-temperature resistivity nor the two specific heat transitions observed for this small range of x are clearly understood, we suggest that these two phenomena may be related. The apparent Jack of sensitivity to magnetic field for x = 0.026, the midpoint of the anomalous x region with the local Tc max.im um by resistivity and with approximately equal entropies in the two specific-heat anomalies, is also not fully understood. W e speculate that these phenomena may be caused by a subtle interplay between superconductivity and the source of the scattering mechanism related to the low-temperature resistivity peak as the relative temperatures or energies of these two interactions change with x. Volovik and Khmel'nitskii very recently suggested 16 that some ofthese unusual effects may be caused by the fonnation of a "superconducting glass" state with short range order just below Tc followed by a transition to normal superconductivity with Iong range order at lower temperatures. A crucial point to bear in mind is that the energy scales and bandwidths relating to these phenomena are extremely small; as a result, modest changes to the system, either by impurity additions or magnetic fields, can cause substantial modification of the physical properties of UBeu.
Specific-heat measurements on U, _" Th"Be, 3 in an applied magnetic field are currently under way to further elucidate this subtle interplay and to probe the nature of the specific-heat anomalies observed in zero magnetic field.