Magnetic excitations and quasielastic linewidths in YbBe13 from neutron scattering

Abstract Inelastic magnetic neutron spectra were taken on polycrystalline YbBe 13 between 1.2 and 300 K. At low temperatures the formation of a magnon corroborates the existence of a magnetic phase transition observed also by static susceptibility. At higher temperatures two distinct CF-transition lines are observed, which lead to the cubic CF-parameters x = 0.911 and W = 0.1543 meV . The magnetic quasielastic linewidth increases linearly with temperature and is anomalous large compared to TbBe 13 but small everywhere compared to k B T .


Magnetic susceptibility measurements
of YbBe,, [l] have established an effective magnetic moment of pert = 4.54~~ of the 4f-shell in Yb, which is exactly the theoretical value of the stable trivalent 4f l3 configuration in the Hund's rule groundstate.
Until recently, this compound was the only intermetallic compound with Yb known to order magnetically.
This singular case is in contrast to the large number of magnetically ordering Ce-compounds on the other end of the rare earth series.The reason for the rare occurrence of the magnetic ordering for Yb is expected to lie in the instability of the 4f-shell.Actually, one expects more symmetry in the behaviour between Ce and Yb.Indeed, three more magnetically ordering Yb-compounds have been found recently, namely YbPd [2], Yb,Pd, [3] and YbIr, [4].When Yb does actually order magnetically in a metal, its properties should carry some memory of the valence instability, which would lead to the kind of interplay between valence instability and magnetic order, which has attracted so much attention in the magnetic ordering of Ce-compounds, e.g.CeAl, and CeAg, in the past.Indeed, recent measurements of the magnetic quasielastic (QE)-linewidth of YbPd and Yb,Pd, [5] show qualitatively the same behaviour as CeAl, and CeAg.In the work we present here, we have investigated the magnetic neutron spectrum of YbBe,, to find out whether we here also have a case of magnetic order at the verge of valence instability.

Experimental and discussion
YbBe,, crystallizes in the cubic NaZn,, structure (Fm3c).While 8 Yb-ions are placed at equivalent cubic sites 104 Be-ions occupy two different types of sites (monoclinic, cubic) thus forming a kind of cage around the Yb-ions.
We have prepared polycrystalline YbBe,, with lattice parameter a, --10.195A and have measured its inelastic neutron spectrum at the time-of-flight spectrometer IN6 (ILL, Grenoble, E, = 3.1 meV) in the temperature region between 1.2 and 300 K.
Fig. 1 shows the scattering function as measured from YbBe,, at low temperatures.At T = 1.2K the full line is a fit to the measured spectrum (open circles) with only one inelastic line centred at about 0.13 meV.Since this inelastic line becomes quasielastic above 1.5 K, we interpret it as a magnon.Support is given to this assumption by the fact that only a Gaussian line shape can be fitted to those data.The existence of a magnon at 1.2 K is in agreement with the existence of an antiferromagnetic phase transition at TN = 1.28 K observed by magnetic susceptibility [l].However, it should be pointed out that the spin fluctuations (SF) associated with the magnon a 1.2 K do not die out quickly above TN as expected.Instead, they survive as critical spin fluctuations up to temperatures at least ten times TN as a large additional QE-SF-line (see fig. 4) with a typical Gaussian lineshape (for T = 10 K (fig. 1) only a fit with Gaussian lineshape gives satisfactory results).For example at T = 10 K the QE-SF-intensity turns out to be 2.8 barn, whereas the standard QE-CF-intensity is 1.6 barn.(The latter could be determined by the CF-parameters extracted from higher temperatures (see below) with the temperatures (fig.2, T= 40 K) two distinct inelastic restriction that the inelastic (IN) CF-lines at T= 10 K transitions can be observed at about 1.2 and 3.2 meV.If (not shown here) imply that the magnetic moment is we take these lines as CF-transitions and interpret the reduced by a factor of 3 at that temperature).
At higher QE-scattering as CF-like, too, we only get consistent fits The CF-lineshapes could only be fitted with a Lorentzian shape function as usual with a single generating QE-linewidth for all QE-and IN-lines.The dependence of this QE-linewidth is shown in fig. 4 by the full circles.It is important to mention that for temperatures above 40 K the fits require an additional QE-intensity analogous to T< 40 K described above.However, incontrast to the former case this additional QEintensity is of Lorentzian type and thus will be interpreted as additional QE-CF-scattering, i.e. in addition to the scattering which occurs in the limit of very small linewidths compared to the CF-splitting (standard QE-CF-scattering).
This additional QE-CF-intensity increases rapidly with increasing temperatures and at 300 K becomes the same order of magnitude as the standard QE-CF-scattering.
The growth of the additional QE-CF-intensity goes hand in hand with a decrease of the inelastic intensity, such that the total magnetic cross section remains constant at the value corresponding to the trivalent Yb-ion.This feature can be attributed to the renormalization of the CF-eigenr, --z'7' states, if their linewidths become comparable or greater than the CF-splitting.Indeed the application of the BFK-theory [7] by a program written by Keller, which takes this renormalization effect into account (in this theory this effect is induced by the exchange or Coulomb scattering of the conduction electrons at the 4f-shell) leads to the requested additional QE-CF-intensity.However, a fit with that theory fails for the calculation of the inelastic transitions.
Here it generates wrong relative linewidths both in the case of pure exchange and Coulomb scattering.In addition, the best BFK-fit leads to an exchange constant N(O)J,, = 0.17, which is very large compared to the expected value of about 0.01, which is a typical value in metallic RE-compounds [8], e.g.REAL,.This anomalous large constant which implies an anomalous large QE-linewidth can also be compared with the linewidth of the reference sample TbBe,,, which has also been measured [9] and which is indicated by full triangles in fig. 4. exchange scattering between the conduction electrons and the 4f electrons but should originate from another scattering effect such as the hybridization of 4f with 5d electrons as in the case of intermediate valence or of the Kondo effect.However, we believe that the occurrence of the additional QE-intensities should be understood at low temperatures from the classical theory of critical phenomena near magnetic phase transitions for stable ions with long range (metallic) interactions.

Fig. 1 .Fig. 2 .
Fig. 1.Scattering function (open circles) of YbBe,, at 1.2 and 10 K at 0 = 60' mean scattering angle and E, = 3.1 meV.The solid line is a fit to the spectra with one magnon line at T = 1.2K and one QE-CF-line plus one additional QE-SF-line for T=lO K.