MAGNETISM AND SUPERCONDUCTIVITY IN CECU2SI2 SINGLE-CRYSTALS

Single crystals of CeCu2Si2 were grown from metal solvents and their magnetic properties were systematically varied by partial substitution of Cu by Sn or In. The susceptibility χ is anisotropic and at high temperatures it is dominated by the crystal field split 4f state. At low temperatures the magnetic moments are drastically reduced. This reduction, as well as the Curie–Weiss temperature, depend sensitively upon the chemical composition. The observed variations both in the magnetic and the resistance behavior point to a Kondo‐like coupling between local moments and conduction electrons. The strength of this interaction varies as the Fermi level is shifted by chemical means.Single crystals of CeCu2Si2 were grown from metal solvents and their magnetic properties were systematically varied by partial substitution of Cu by Sn or In. The susceptibility χ is anisotropic and at high temperatures it is dominated by the crystal field split 4f state. At low temperatures the magnetic moments are drastically reduced. This reduction, as well as the Curie–Weiss temperature, depend sensitively upon the chemical composition. The observed variations both in the magnetic and the resistance behavior point to a Kondo‐like coupling between local moments and conduction electrons. The strength of this interaction varies as the Fermi level is shifted by chemical means.


INTRODUCTION
CeCu 2 Si 2 is the first compound in which "heavy fermion" superconduetivity was established. 1 The valenee of Ce is close to 3 + and the magnetic properties at temperatures higher than 10 Kare determined by weil defined local 4 f moments. Below this temperature a nonmagnetic state develops and the eleetrons at the Fermi level acquire huge effective mass ( > 200 m e ). 2 • 3 These heavy fermions then condense into a superconducting state at Tc ' .:::::' .0. 6 K. Because all the original work was done on polycrystalline samples and because it was obvious from the very beginning that the low temperature properties depend sensitively on details of the preparation process 1 -<> we started a systematic study of single crystals. The aim was to gain insight into the formation of the heavy fermion liquid by a controlled variation of the chemical eomposition. We find that the degree of "demagnetization" in the low temperature range is sharply reduced when Sn or In atoms Substitute for Cu. W e suggest that this is consistent with the picture of the heavy fermions owing their origin to a Kondo-Iike interaetion between the local 4 f moments and the conduction eleetrons.

EXPERIMENTAL RESUL TS
The erystals were grown from a Sn, In, or Cu solvent. When stoichiometric amounts of the CeCu 2 Si 2 is added to In or Sn, the resulting crystals contain some In or Sn. The presence of In or Sn is eonfirmed by x-ray fluorescence. By increasing the Cu concentration it is then possible to reduce the In or Sn eontent. W e find this is a useful method to vary the physical properties, as we will describe below. The higher the Cu concentration, however, tbe more diffieult it is to extract the erystals. A SQUID magnetometer (S.H.E. Model 905) was used to measure the magnetie properties in fields up to 50 kOe and between 1. 5 and 400 K.

TEMPERATURE DEPENDENCE OF X
A general overview of the temperature dependence of the susceptibility Xis given in Fig. 1. The inverse of X for two crystals, as measured in a field of 1 kOe, is plotted versus tbe temperature. The two crystallographic directions are denoted as lla and l a, where a is in the basal plane of tbe tetragonal unit eell. This is also the plane in which the Ce atoms lie. The two samples represent two extreme cases and we call tbem "Cu rieb" ( = grown from Cu flux; round symbols) and "Cu poor" (grown from In; triangles). The anisotropy of Xis obvious and its sign and temperature dependence is consistent with tbe proposed CF level scheme. 3 At temperatures higher tban 250 K and Hl a data fall on a straight line with a slope that corresponds witbin 0.2% to the effective moment of a free trivalent Ce ion (µelf = 2.536 µg ). For Hll a, the free ion value of µetr is not yet reacbed below 400 K. The main diff erence between tbe two samples is the vertieal shift of the x -1 curves by 60 ± 2 mol/ emu. Tbis holds true for X lla at T> 20 K and for xi a at T> 100 K. The Curie-Weiss temperatures () deduced from the X l a data are -50 ± 1 K for tbe Cu rieb sample and 0 ± 2 K for the Cupoor sample. Additional differences between the two samples are given in Fig. 2. Whereas the Cu-rieb one exhibits no The chemical composition of the solvent detennines the amount of moment reduction, as described in the text. The full and broken lines refer to the same samples as in Fig. 2 than anticipated for the CF ground state.s: three doublets at 0-140 K and 364 K, respectively. The local 4/moments are partially compensated. From a systematic variation of the ftux composition we were able to establish a relation between the Cu content and the magentization. Apparently an increase ofthe Cu concentration results in a stronger moment compensation .

DISCUSSION
Because the main purpose of our study is to shed some light on the formation ofthe heavy fermion which ultimately can condense into a superconducting state, we concentrate the discussion on the main trends in our data. These trends, we suggest, are due to a variation of the strength of the interaction between the 4 f moments and the conduction electrons. In particular we would like to argue that the chemical variations affect the position of the Fermi energy with respect tot he 4 f Jevel of Ce. When the crystals are grown from Sn or In the Cu concentration in the fiux determines how many Sn or In atoms can occupy the Cu sites in the crystal. Since both Sn or In supply more electrons than Cu does, the Fermi level is higher in the "Cu poor" samples. The dominant parameter in the given situation is presumably the same as in the single Kondo impurity case, namely, the position of the 4 f level with respect to EF. Thus even small shifts of E,..
will have pronounced consequences of the kind observed here: decrease of the moment compensation as the Fermi level rises. Equivalently, one can describe the low temperature state of CeCu 2 Si 2 in terms of a Fermi liquid. In this language the replacement of Cu by In or Sn causes a decrease of the characteristic temperature T,.. . Even though this paper deals mainly with the magnetic properties we should mention some other physical properties which we have studied. The electrical resistivity is anisotropic and islargest in the basal plane(/ JJa). More important, however, is the difference in the temperature dependence, as shown in Fig. 4. Both p 11 and p 1 go through a broad maxim um around 100 K, and p 11 continues to rise upon further cooling to a maximum of at Tii = 11.S K, whereas p 1 reaches a much lower maximum at Tt = 15-16 K . At even lower temperatures, the resistivity drops precipitously, presumably due to the formation of the coherent Kondo-lattice ground state. We would like to note that the difference between Tjl' and T"{ makes the interpretation of T• in polycrystalline samples somewhat ambiguous, particularly if T • is used as a quantity related to the Kondo temperature. For our single crystals Tjl' varied from -8 K for the "Cu-poor" sample to -14 K for the "Cu-rich" one. This trend is in line with the ideas discussed earlier. At present we are studying the superconducting properties and a more detailed account will be given elsewhere. We only note that ac susceptibility and resistance measurements indicate superconductivity in the "Cu-rich" sample at T c = 0.5 K . In addition we also study the specific heat properties as we vary the degree of moment compensation. First measurements of the longitudinal magnetoresistance up to 80 kOe revealed a quite complex behavior. Particularly striking is the magnitude (.:::SR / R --40% for I llc,.'1R / R -10% for I lla) and thevery pro- nounced nonlinearity at T -1. 5 K . The magnetoresistance appears tobe determined by an interplay between CF effects and the suppression of Kondo-type interactions.

CONCLUSION
We were able to grow single crystals of CeCu 2 Si 2 and to inftuence their magnetic properties in a controlled way through chemical means. The magnetic susceptibility is anisotropic and its temperature dependence is qualitatively consistent with the proposed CF-level scheme. The most prominent difference among the various single crystals is found in the value for the Curie-Weiss temperature and the 2003 J. Appl. Phys., Vol. 55, No. 6, 15 March 1984 degree of moment compensation oflowest temperatures. We argue that these Observations are indicative of a variation of the local moment-conduction electron coupling strength. Resistivity data are supporting this view. lt is quite interesting to observethe maximum in x (T) at T = 3.5 Kinthe"Cupoor" sample (Fig. 2). The transition f rom a nonmagnetic to an antiferromagnetic-type ground state is predicted to occur in a Kondo lattice as the strength of the antiferromagnetic fd coupling T 1 d increases. 7· 8 A significant variation of T 1 d has tobe inferred from the magnetization data in Fig. 3, and thus the present series of CeCu 2 Si 2 based compounds may represent an example for the nonmagnetic-to-magnetic transition in a Kondo lattice. The data presented here help to understand the origin of the heavy fermions in CeCu 2 Si 2 , and the additional work, which is in progress, allows to draw quantitative conclusions.

ACKNOWLEOGMENTS
We like to thank C. M. Varma, F. Steglich, and D. Wohlleben for interesting discussions. We also thank G. P. Espinosa for experimental crystal growth, H. Barz for arc melting some samples and S. Vincent for x-ray ftuorescence analysis.