Novel synchrotron-based techniques provide an opportunity to look again at the classic problem of intermediate valence (IV) in lanthanide materials. Traditional electronic probes such as photoemission (PES) and X-ray absorption spectroscopy (XAS), although quite powerful, suffer from specific drawbacks: the energy resolution of XAS is often limited by the short lifetime of deep core holes, and the small probing depth of PES makes it very sensitive to surface conditions. Consequently, comparison of the experimental results with theory, e.g. the Anderson Impurity Model or its lattice extension, is not free from ambiguities. On the other hand, photon in-photon out spectroscopies like “high-resolution” partial fluorescence yield XAS (PFY-XAS) or resonant inelastic X-ray scattering (RIXS) [1], and hard X-ray photoemission (HAXPES) [2], have recently demonstrated the ability to probe changes in the bulk 4f configuration occurring at the characteristic low-energy “Kondo” scale kBTK. Our experiment performed at beamline ID16 on selected Yb compounds was an attempt to compare the information available from these complementary techniques [3].

Figure 5a shows PFY-XAS data for the IV compound YbInCu4, measured at the Yb L3 (2p3/2) edge by recording the intensity of the Yb L (3d5/2 2p3/2) fluorescence. The spectra reveal distinct Yb2+ and Yb3+ features, split in the final state by the Coulomb interaction with the core hole. Their intensities are proportional to the weights of the Yb2+ and Yb3+ configurations in the hybrid ground state. The spectral weight transfer observed between 300 K and 20 K in YbInCu4, but not in the Yb2+ reference material, demonstrates the Yb valence change predicted by the Kondo scenario [4]. The temperature changes are even more visible in the RIXS data of Figure 5b, which illustrate the energy distribution of the photons emitted in the 3d5/2 2p3/2 radiative recombination. The excitation energy was set at the maximum of the 2+ XAS feature, which led to a selective resonant enhancement of the Yb2+ contribution.

Fig. 5: a) PFY-XAS spectra of YbInCu4 and of an Yb2+ reference compound at 300 K and 20 K. b) L RIXS spectra measured at the maximum of the Yb2+ resonance.

A temperature-dependent intensity transfer between the Yb2+ and Yb3+ spectral features is also observed in the HAXPES data of Figure 6a, measured at ID16 with the VolPE spectrometer [5]. The experimental 4f spectrum of Figure 6b was obtained by subtracting the data for the sister compound LuInCu4, after removal of the core-like Lu 4f doublet. The Kondo resonance located at kBTK below the Fermi level EF and its spin-orbit satellite are remarkably free from the broad “surface satellites” that are typically observed in conventional PES data. The contribution from the first surface layer is estimated to be less than 5%. We have observed a similar spectral weight transfer in the Yb 3d5/2 core levels (at EB=1528 eV; not shown).

Fig. 6: Valence band HAXPES spectra of YbInCu4. The dashed line is the spectrum of LuInCu4 after removal of the atomic-like 4f doublet at ~8 eV. The difference spectrum is shown in (b) for T=20 K.


Figure 7 summarises the RIXS and HAXPES results. The temperature dependence could be measured continuously by RIXS, thanks to high detection efficiency. The RIXS data are in good agreement with thermodynamic and magnetic measurements, and capture the sudden valence drop at 42 K, in correspondence of a first-order electronic phase transition. They confirm that RIXS is a reliable bulk probe of the 4f configuration. HAXPES is directly sensitive to the 4f spectrum, and the data confirm a lower valence at low temperature. Nevertheless there is a quantitative discrepancy, especially for the valence band data. We have observed similar discrepancies for the other Yb compounds we have investigated. This suggests that the HAXPES data are still influenced by a perturbed surface layer extending over a range of a few tens of angstroms below the scraped surface. Its influence is likely to be reduced on cleaved single crystal surfaces, and especially by further increasing the photon energy, and therefore the probing depth, in HAXPES. Experiments at energies as high as ~10 keV are now possible by VolPE, and will be performed in the future.

Fig. 7: Yb valence of YbInCu4 determined from RIXS and HAXPES (valence band (VB) and Yb 3d5/2 core level).


[1] C. Dallera et al., Phys. Rev. Lett. 88, 196403 (2002).
[2] H. Sato et al., Phys. Rev. Lett. 93, 246404 (2004).
[3] L. Moreschini et al., Phys. Rev. B 75, 35113 (2007).
[4] N.E. Bickers et al., Phys. Rev. B 36, 2036 (1987).
[5] P. Torelli et al., Rev. Sci. Instrum. 76, 020939 (2005).


L. Moreschini (a), C. Dallera (b), J.J. Joyce (c), J.L. Sarrao (c), E.D. Bauer (c), V. Fritsch (c), S. Bobev (c), E. Carpene (b), S. Huotari (d), G. Vankó (d), G. Monaco (d), P. Lacovig (e), G. Panaccione (e), A. Fondacaro (f), G. Paolicelli (e), P. Torelli (g) and M. Grioni (a).
(a) IPN, Ecole Polytechnique Fédérale (EPFL), Lausanne (Switzerland)
(b) cnr-INFM, Dipartimento di Fisica, Politecnico di Milano (Italy)
(c) Los Alamos National Laboratory, Los Alamos, New Mexico (USA)
(d) ESRF
(e) Laboratorio TASC, INFM, Basovizza (Italy)
(f) INFM and Dipartimento di Fisica, Università di Roma III (Italy)
(g) LURE, Université de Paris-Sud, Orsay (France)