CRYSTAL-FIELD LEVELS IN YBBIPT

Abstract The new massive-electron system YbBiPt has been studied by means of neutron inelastic scattering at temperatures between 3.3 and 77 K, using the crystal-analyser spectrometer LAM-40. Crystal field excitations have been observed at approximately 6 meV and at very low energy. There is also significant quasielastic scattering and the higher-energy level softens slightly with increasing temperature. Although these observations are in qualitative agreement with the bulk susceptibility measurements, our data suggest that the first excited crystal-field level is much lower in energy than had been deduced from bulk measurements.


Introduction
The cubic heavy-fermion compound YbBiPt was discovered in 1991 by Canfield et al. [1,2] and has a very large linear specific-heat coefficient 7 = 8 J mol-~ K 2. It has previously been characterised by means of electrical-resistivity, magnetoresistance, ac and dc magnetic susceptibility and specific-heat measurements, and in order to explain these data, it was suggested that the cubic crystal field splits the 4fv/2Yb 3÷ Hund's rule state into two doublets I" 7 and ['6 and a quartet 1 '8, in that order, with the two doublets split by approximately 1 meV. The main purpose of the work reported here has been to clarify this situation by observing the crystal-field levels directly using neutron scattering. YbBiPt forms in the CI b or MgAgAs structure type (space group F43m), which is a common Heusler alloy structure [3], and in which some uranium-based ternaries like UNiSn [4] form. The Yb atom lies on the same site as Mn or U in these compounds, with point group symmetry 43m.

Method
as reported previously [1,5]; 9.6g of polycrystalline material was sealed under helium gas in a thin-walled aluminum tube and mounted within a helium cryostat. Neutron powder diffraction analysis showed the presence of a small amount of elemental bismuth in the sample, presumably from the flux. The cryostat was mounted in the time-of-flight crystal-analyser spectrometer LAM-40 at KENS, the spallation neutron source of the Japanese National Laboratory for High-Energy Physics (KEK). This spectrometer [6] views a solid-methane moderator and employs seven largesolid-angle pyrolytic-graphite focusing analysers, each with its own beryllium filter and detector. The final energy is 4.6 meV. The analyser arms were positioned at scattering angles of 8 °, 24 °, 40 °, 56 °, 72 °, 88 ° and 104 ° , although in our analysis we have only used the four higher-angle detectors, because of contamination from the cryostat walls. The instrumental resolution has a width of 0.35 meV FWHM, with some asymmetry due to the pulse shape of the incident neutrons, and one can measure in neutron energy loss down to 0.3 meV with insignificant influence from incoherent inelastic scattering. spectrum and the kf/k~ term in the cross section, and the resultant resolution-broadened scattering functions S(Q, to) measured at 3.3, 10, 30 and 77K are shown in fig. 1. At low temperature, there is a clear excitation at 5.7 meV, and it softens to 5.1 meV at 77 K. Its intensity decreases with temperature, indicating that it is magnetic in origin and not a phonon. No other strong magnetic peaks were observed above 6 meV. In addition, there is very significant quasielastic scattering, which we were unable to fit using a single Gaussian (or Lorentzian) response function. However, we can obtain reasonable fits to the observed spectra with a two-component quasielastic response, one with a width roughly twice that of the resolution function and the other ten times broader. The second component is quasielastic to within 0.05 meV. Gaussian lineshapes were assumed in all cases and the fits are shown in fig.  1. We tried Lorentzian lineshapes in some cases, but the agreement with observation was always worse. In the subsequent discussion, we assume that the imaginary part of the generalised susceptibility has the following form:

Im x(Q,ho~O)) __ CI e_(h,~/,~,)2 + ~ ci e_(~(~_~o,)/~i ?
where the first term is the true quasielastic response, and the second two terms are inelastic, with hto 2 ~0  and ht0 3 ~ 6 meV. While the quasielastic linewidth hardly changes with temperature, the two inelastic excitations broaden monotonically as temperature is increased. There is also an indication, in Fig. 1, of some magnetic scattering at about 8 meV, but it is very weak, too weak for the published matrix elements [7,8], and at this point, we have little idea as to its origin.

Discussion
Our data clearly indicate that the characters of the quasielastic spectra are more Gaussian-like than Lorentzian-like. This means that the spin fluctuations are mainly due to inter-site spin-spin correlations, rather than single-site spin relaxation. In other words, there are strong (para)magnetic correlations at least up to 30 K. From the width of the broad quasielastic component, we estimate that the correlation energy is approximately 20K. Second, within our model of a three-component Gaussian lineshape response function, we can integrate with respect to energy to obtain the susceptibility x(Q). We do not yet have an absolute intensity normalisation, so we have scaled our result to the bulk measurement [2] at 77 K, and the results are shown in fig. 2. If this scaling is correct, the agreement is quite good.
Although our analysis is preliminary, it does seem that there are three components in the spectrum, and it is natural to associate these with the three crystal field states of the Yb 3+ ion in a cubic potential, ['6, F v and F s. The interesting thing is that the lower two   [9], and the ratio of our observed intensities for the 6 meV excitation (at 3.3 and 77 K) is in good agreement with a sixfold degenerate ground state, i.e. F 7 with F 8 directly above. Considering the fact that there is no magnetic long-range order above 0.4 K, in spite of strong magnetic correlations with characteristic energy 20 K or so, we conclude that there is a strong Kondo effect. Qualitatively, it seems that this strong Kondo suppression, together with the high multiplicity of the ground state, gives rise to the giant enhancement of 7. We are currently in the process of performing a complete quantitative analysis similar to that made on other cubic Yb compounds [7,8], and this will be the subject of a subsequent publication.