Bulk and surface valence of YbBe13

Abstract We report angle-integrated photoemission results on clean and oxidized YbBe13 samples, at and below 300 K, which show YbBe13 to have a trivalent bulk ( v = 2.9 −0.02 +0.1 ) and a divalent surface layer. The bulk valence, independently derived from temperature-dependent LIII-edge X-ray absorption measurements, is found to be three.


Introduction
The isostructural XBe13 alloys (X f rare earth or actinide element) form an interesting class of materials. For superconducting UBel, [l] and magnetic NpBe13 [2], an anomalously high electronic specific-heat value has been reported (y -1 J mol-' K-') which classifies these materials as heavyfermion syst,ems. On the other hand, CeBe,3 provides evidence of a configurational instability of the Ce 4f shell [3, 41. The subject of this paper, YbBe,+ has been found to order antiferromagnetically below TN = 1.3 K [5], but above TN it shows an anomalously large linewidth in inelastic magnetic neutron scattering, which may indicate that the Yb ions are also near a valence instability [6]. The latter could also be inferred from Mijssbauer measurements for the paramagnetic state [7]. The lack of volume dependence of resistivity features suggests, however, that YbBe13 is not mixed-valent in the ground state, but above TN resistive anomalies exist, possibly owing to "Kondo-sideband" scattering [8]. Here, we apply bulksensitive L,,,-edge X-ray absorption and photoemission spectroscopy at The overall resolution (electrons and photons) was 0.2 eV. The sample could be cleaved at room temperature as well as at 100 K. For the X-ray absorption measurement, a sample previously used for photoemission was finely powdered under a dry argon atmosphere and pressed into paraffin wax. The intensities of the incident and transmitted Xrays were monitored as a function of energy using tunable synchrotron radiation from the EXAFS-II spectrometer at the HASYLAB, DESY, Hamburg [ 151. Near the Yb-LIII edge, around 8.9 keV, the energy resolution was 2 eV. Data were taken at 300 K and 5 K.

Results and Discussion
In Fig. l Fig. l(a), the Yb3+ multiplet peaks rest on a broad emission feature with a maximum around 7 eV. By comparison with a He(I) (hv = 21.2 eV) EDC (not shown), we attribute this broad emission to s,p states of Be similar to those in UBei3 [16]. There have been discussions as to whether oxygen 2p emission could be the origin of this feature [ 17,181. To check this, we have exposed clean YbBe,3 to 0.75 L O2 (1 L = 1 Langmuir = lop6 Torr X 1 s), Fig. l(b). The changes are dramatic. The Yb3+ emission is completely suppressed by the huge increase in the 0 2p emission (note the factor, 2, in Fig. l(a) relative to Fig. l(b)), indicating the high sensitivity of the surface to 02. This leads us to believe that the freshly-fractured surface is oxygenfree. In fact, further exposure of O2 increased the 0 2p emission as can be As an additional check, we in Figs. l(c) and l(d). The saturation level fractured YbBeis at 100 K to avoid oxygen diffusion from the bulk [ 181. Again, we found an EDC similar to the one in Fig. l(a), in particular the trivalent Yb multiplet resting on top of the broad Be s,p emission. The difference between cold and warm-fractured YbBe,s is an -20% reduced intensity of the Yb2+ doublet near Er. This may indicate some Yb atoms segregating to the surface at room temperature.
Upon oxidation, Yb forms trivalent Yb203, but the multiplet structure is completely smeared out by the 0 2p emission [ 191. The fact that we can clearly discern the f12 peaks in fig. l(a) is another indication that the fractured YbBei3 surfaces are clean and that the underlying broad s,p emission is Be-derived.
We now turn to the Yb2+ (4f13) doublet near the Fermi level. It is interesting to find the 2F7,2 component at 0.9 eV and a shoulder cut by EF. For a homogeneously mixed-valent system, the bulk divalent component has to be pinned at EF, if screening effects are neglected [ 131. On the other hand, it has now been established that all mixed-valent systems (with the exception of Ce systems) have a divalent surface layer, e.g., EuPd& [lo], TmSe [ll], Sm,Y,_,S [12], and Yb,Y,_,Alz [13]. Even some trivalent systems such as Sm [9], EuPds [lo] and TmS [ll] exhibit a divalent surface layer. From the non-zero binding energy of the intense f13 doublet in Fig. l(a), we can immediately conclude that YbBe13 has a divalent surface layer. The 0.9 eV position of the surface 2F,,z component is close to the value found for Yb,Y, _xA12 [13], which is around 1.1 eV. By contrast, elemental Yb metal exhibits the 2F,,2 surface component at 1.7 eV binding energy [ 201.
The question now arises, whether a bulk 4f13 doublet with low spectral intensity could cause the shoulder near EF in Fig. l(a), as would be expected for a nearly trivalent (V < 3) system. We have therefore tried to fit the experimental spectrum in the range 0 -4.5 eV with two sets of Lorentzian broadened f13 doublets (ratio of the 5/2 -7/2 component 0.75; spin-orbit splitting 1.27 eV [13]) for the bulk and the surface. Although, in principle, the EDC could be fitted by a single surface doublet, owing to the ambiguity in subtracting the underlying Be emission near EF, 10% of the divalent spectral intensity could stem from a Yb2+ bulk doublet. Hence, the lower limit for the bulk mean valence amounts to V = 2.9+_od2. An independent experimental check could involve quenching of the surface emission by oxygen exposure, which would leave a possibly existent divalent bulk f13 emission unaffected.
While it is known that this works quite well with rareearth NaCl compounds [ll, 121, intermetallic alloys such as YbAl, [13] do not clearly show this effect, probably because of the simultaneous oxygen reaction of the non-rare-earth constituents in the alloy. For YbBe13, we have studied the behavior of the surface 4f13 emission with O2 exposure in Figs. l(b) -(d). The first 3/4 L O2 clearly reduces the divalent surface emission, as is clear from the difference spectrum of Fig.  l(c). The remaining emission, however, is still not pinned at EF in Fig. l(b), pointing to the non-existence of a bulk Yb2+ signal. Further O2 exposure results in an increase of the spectral intensity in the 0 -4 eV range (see the difference curve of Fig. l(d)), which saturates after 10 L OZ. This complex behavior cannot be used to clarify the question of a small Y b2+ signal hidden in the intense Yb2+ surface emission of Fig. l(a).
To obtain a definitive answer, we measured the X-ray absorption at the Yb LIII edge at 8.934 keV. This technique is bulk sensitive only. Our results are displayed in Fig. 2. The experimental curve measured at 5 K is identical to a spectrum of YbBe13 at 300 K. This is in contrast to YbPd2Siz with a temperature-dependent mean-valence which varies from V = 2.9 at 300 K to 2.8 at 16 K [ 151. A fit to the Lm-edge profile performed in the same way as for YbPdzSi2 finally revealed that no divalent Yb is present in bulk YbBe13, as the photoemission results could only suggest (see the discussion above).  Fig. 2. Yb Lm-edge X-ray absorption of YbBers. The dotted line represents the experiment at a sample temperature of 5 K and is identical to a spectrum taken at 300 K. The solid line is a fit using a single Yb3+ Lm edge. Deviations at higher energies are due to the EXAFS oscillations.

Conclusion
From temperature-dependent Lm-edge measurements, we derive that YbBei3 is trivalent in the bulk. With photoemission spectroscopy, we find a trivalent bulk and a divalent surface layer. The Be s,p emission produces a broad emission feature centered around 7 eV.