Strong electronic correlations in a new class of Yb-based compounds: YbXCu4 (X=Ag,Au,Pd).

A new series of heavy-electron Yb compounds YbXCu& with X = Ag, Au, or Pd are reported which have electron effective masses as high as -60m„where m, is the mass of the free electron. Whereas the compounds with X =Au and Pd order magnetically below 1 K, the compound with X =Ag is nonmagnetic. The magnetic susceptibility and specific heat of YbAgCu4 can be account-ed for by the Bethe-ansatz solution of the Coqblin-Schrieffer impurity model without any adjustable parameters.

The remarkable physical properties of compounds of rare earth and actinide elements which are associated with the hybridization between conduction-electron and localized f-electron states continue to attract the attention of experimentalists and theoreticians alike. A significant amount of effort has been expended in characterizing exemplary f-electron materials and developing theoretical models with which their extraordinary properties can be compared. ' Compounds of the rare-earth elements Ce and Yb are especially well suited for this purpose, since the 4f shell of Ce and Yb can accommodate at most one electron or hole, respectively, which considerably simplifies relevant theoretical models. In this paper we report transport, thermal, and magnetic measurements on a new series of Yb compounds with the formula YbXCu4, where X=Ag, Au, or Pd. These compounds appear to be nearly ideal materials for testing theoretical models since their crystal structure has cubic symmetry, the Yb 4f shell contains approximately one hole, and interactions between the Yb magnetic moments are relatively weak. Our results reveal that the compounds with X=Au and Pd order magnetically below 1 K, whereas YbAgCu4 has a nonmagnetic ground state.
Moreover, the magnetic ground state in YbAuCu4 and the nonmagnetic groundstate in YbAgCu4 both develop out of a heavy Fermi liquid characterized by an effective mass of -50m"where m, is the mass of the free electron. We have found that the magnetic susceptibility 7 as a function of temperature T and the value of the coefficient y of the specific heat C =yT of the nonmagnetic compound YbAgCu4 can be described by the Bethe-ansatz solution of the Coqblin-Schrieffer Hamiltonian for angular momentum J = -, , which is appropriate for the Hund's rule groundstate multiplet of trivalent Yb, without any adjustable parameters. The polycrystalline YbXCu4 samples were prepared in three ways: by melting the pure elements in (1) an argon arc furnace on a Cu hearth, (2) a BeO crucible, or (3) a sealed Ta tube, in order to minimize losses of the volatile Yb. The physical properties of the samples were not influenced by the different cooling rates involved in these three techniques. X-ray powder diffraction and metallographic analyses yielded no evidence of impurity phases.
In contrast to YbCuz, which has the hexagonal CaCu5type crystal structure and in which Yb is divalent, the three YbXCu4 compounds with X=Ag, Au, and Pd have the cubic AuBe5 structure (space group C15b or F 43m)and Yb appears to be trivalent, or nearly so. The values of the lattice parameters are 7.0696, 7.0519, and 7.0396 A for X=Ag, Au, and Pd, respectively. The Yb-Yb distance in these materials is of the order of 5 A, a relatively large value which seems to favor a magnetic and/or heavy-electron ground state as proposed for the actinides by Hill. Electrical resistivity p measurements were made with a self-balancing four-wire impedance bridge operating at a frequency of 16 Hz between 80 mK and 295 K. The data below 1.2 K were taken in a conventional He-He dilution refrigerator. The magnetic susceptibility 7 was measured with an SHE superconducting quantuminterference device (SQUID) magnetometer in a magnetic field of 1 T between -2 and 300 K. Heat-capacity C measurements were carried out in a semiadiabatic He calorimeter between 0.45 and 16 K.
Normalized electrical resistivity p(T)/p(295 K) versus T data between 80 mK and 295 K for the three YbXCu4 compounds, displayed in Fig. 1, exhibit different types of behavior for each of the three substituent X elements.
Little temperature dependence is observed for all three samples down to -80 K. However, below -60 K, the resistivity of YbAgCu4 falls rapidly and follows a T power-law dependence between -1 and 28 K which is characteristic of Fermi-liquid behavior or spin fluctuations. In contrast, the resistivity of YbAuCu& increases with decreasing temperature below -100 K and displays two maxima, one at -20 K and a second one at -1.1 K (see inset of Fig Normalized electrical resistivity p(T)/p(295 K) vs temperature T for YbXCu4 compounds with X=Ag, Au, and Pd. Shown in the inset are the curves below 6 K. The presence of a maximum at 1.1 K for X=Au and at 0.7 K for X=Pd is associated with freezing out of magnetic scattering due to the onset of magnetic order.
peak is due to the onset of magnetic ordering (probably antiferromagnetic) as verified by specific-heat measurements which are discussed below. The drop in p(T) below -0.7 K for the compound YbPdCu4 is also associated with magnetic order. In these latter two compounds p( T) seems to exhibit Kondo-like behavior with a typical minimum located around 8S K for the Au-based sample and -60 K for the Pd-based material. At room temperature, the electrical resistivity of these compounds ranged between 30 and 100 pA cm.  Fig. 3 are X(T) data for YbAgCu4 which follow a Curie-Weiss law between room temperature and -100 K (see Fig. 2) and then pass through a maximum at -3S K before reaching a constant value at low temperature. This behavior of X(T) is rather interesting and similar to that reported for YbCuA1. Rajan has calculated X( T) and the specific heat C ( T) in zero magnetic field for angular momentum values from J = -, ' to -, by solving the exact thermodynamic equations of the Coqblin-Schrieffer impurity model with the Bethe-ansatz. Calculations based on the Coqblin-Schrieffer model ' have been applied successfully by Hewson et al. ' to YbCuA1 which suggests that this compound is a good example of a Kondo lattice. Kondo-lattice behavior is expected for compounds containing rare-earth ions with integral valence such as Ce3+ (e.g., CeA13), Sm + (e.g. , SmB6), or Yb + which have a ground-state multiplet and tendency towards mixed valence, and where the antiferromagnetic exchange coupling between the rare-earth ion and the conduction electron spins dominates over the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction. Such compounds are usually paramagnetic at T =0 K or have very low ordering temperatures which appears to be the case for the new YbXCu4 compounds discussed herein.
As shown in Fig. 3 the numerical results of Ref. 8 (solid line in the figure) provide an excellent description of the X(T) data for YbAgCu4. The theoretical curve is based on only one parameter, the characteristic temperature To which is given by To=v(v 1)g~p~/24mk~X-(0), where v is the degeneracy of the ground state, gJ is the Lande g factor of the Hund's-rule ground-state multiplet with angular momentum J, and X(0) is the experimental value of the susceptibility at zero temperature. The value To --167.S K used to generate the theoretical curve in Fig. 3 was calculated from this relation using X(0) = 1.96&& 10 cm /mol for YbAgCu4 and v =2J +1=8 for Yb + Specific heat C versus T data are displayed in Fig. 4 for the three compounds with X=Ag, Au, and Pd. The large peak in C(T) at the respective temperatures of -0.8 K and -0.6 K for the specimens with X=Au and Pd is presumably associated with the anticipated long-range antiferromagnetic order. The absence of an anomaly in C ( T) for YbAgCuq confirms that this compound does not exhibit any magnetic order, at least down to 0.45 K. The plots of C/T versus T presented in Fig. 5 reveal large y(0) values of -245 mJ/mol K for YbAgCu4 and -200 mJ/mol K for YbPdCu4. For this latter compound, a linear extrapolation of the C/T versus T data to T =0 to estimate y(0) seems reasonable in spite of the magnetic contribution at low T. In the case of YbAuCu4, there does not seem to be a reliable way of determining y(0) even if a fairly large value were to be expected. The values of y(0), the Debye temperature OD, the Neel temperature T~, the effective mass m' and the temperature at which the entropy S reaches R ln2 are summarized in Table I. The effective mass was estimated from the relation m*=6 kFy(0)/mk~(Z/fI ) where kF --(3n Z/II )' with Z =4 the number of 4f holes per unit cell (assuming Yb to have a valence 3+ ) and fl the volume of the cubic unit cell.
According to Ref. 8, the zero-temperature, zero-field limit of the specific heat is given by C ( T~0)/T = y = ( v -1 )rrkg l6 Tp. Using the value To --167.5 K determined above from the zerotemperature limit of the magnetic susceptibility X(0), we obtain a value of 182 mJ/mol K for y which is in reason-  '' YbPd, ' YbX (X=N, P,As) (Ref. 13) and YbXCu4 (X=Au, Pd)] with the exceptions of YbBe&3 (Tz --1.28 K) (Ref. 14) and Yb3Pd4 (T& --3 K). ' In the Ce and U compounds, T& is usually closer to 10 K [e g. , CeAlq ( T~= 3.5 K), UCd» ( T~= 5 K) or U~Zn» ( T~= 9.7 K)]. Another reason which makes the present family of Yb compounds interesting is the rather low entropy associated with the ground state which is expected to be a magnetic I 6 or I"7 doublet. It is apparent from Table I that for both YbAuCu4 and YbPdCu4 an entropy of S = R ln2 is not recovered until the temperature is raised from 0 K to about 10 times Tz, or in other words, only about 30 -40% of R ln2 is released by the magnetic transition. Similar behavior has been reported for YbX compounds with X=N, P, and As (Ref. 13). It seems to be a general observation in magnetically ordered heavy-electron systems that the highly correlated electronic ground state removes a sizable amount of the entropy over a scale of the order of the Kondo temperature Tz. Since Tz is usually larger than T&, only a small fraction of the expected magnetic entropy R ln2 is recovered at T&. The competition between the magnetic ground state and the nonmagnetic heavy-electron ground state has been addressed by several authors, ' but the criteria for the occurrence of magnetic order in a heavy-electron system remains an unresolved problem for all Ce, U, and Yb materials.
In conclusion, we have shown that the hexagonal YbCu5 compound, which is a nonmagnetic normal metal, can be transformed into a face-centered-cubic structure by replacing one of the Cu atoms with an X atom of Ag, Au, or Pd, yielding dramatic changes in its electronic properties. In particular, the three YbXCu4 compounds exhibit strong electronic correlations at low temperature with high y values and large effective masses m'. Whereas YbAuCu4 and YbPdCu4 order magnetically below 1 K, YbAgCu4 does not display any cooperative phase transition. The excellent description of the temperature dependence of the magnetic susceptibility X(T) and the electronic specific-heat coefficient y of YbAgCu4 by the Bethe-ansatz solution of the Coqblin-Schrieffer model seems to prove that the conduction electron-impurity spin-exchange interaction is much larger in this compound than the intersite interaction and any possible crystalline electric field terms in the effective Hamiltonian.
Finally, it is interesting to note that the original motivation for this study was based on some previous work performed in the Ce-Cu and U-Cu phase diagrams. For example, the hexagonal compound CeCu5 orders magnetically at low temperature and has a rather low electronic specific-heat coefficient, ' whereas, in contrast, the two neighboring compounds CeCu4 (Ref. 17) and CeCu6 exhibit heavy-electron behavior. ' Another interesting case consists of the magnetically ordered UCu& and UAgCu4 systems in which the formation of a heavy-electron state has also been reported. ' The present study clearly demonstrates that new efforts should be made to include Yb-based compounds in the fascinating and rapidly growing family of heavy fermion systems.