High-resolution Ce 3d-4f resonance photoemission study of heavy fermion and valence-fluctuating Ce compounds: Probing bulk 4f electronic states for the first time

(Collaborating with Dr. Y. Saitoh in SPring-8 and Prof. Y. Onuki in Osaka University)
Nature vol.403, p.396 (2000).

Here we show the power of the technique "High-energy and high-resolution photoemission spectroscopy", which has become a reality very recently at BL25SU on SPring-8, by applying it to the Ce compounds CeRu2Si2 and CeRu2. Previous photoemission studies of these compounds by using low-energy hν (~120 eV) revealed mutually similar spectra for the Ce 4f electronic states, yet it is expected that such states should be different owing to their differing degrees of hybridization with other valence bands. Our determination of the bulk Ce 4f electronic states of these compounds resolves these differences.


Introduction

Valence-band photoemission spectroscopy (PES) is a useful technique to investigate the hybridization and correlation of electronic states. Since the Ce 4f contributions are remarkably enhanced by using resonance PES (RPES), many Ce 4d-4f RPES (hν~120 eV) spectra have so far been measured with high-resolution. However, a serious drawback of the PES at low-hν (~120 eV) for the studies of the bulk states has been gradually recognized as that the 4d-4f RPES is surface-sensitive due to the short mean free path of the photoelectron as shown in Fig.1. Considering its larger bulk sensitivity, the Ce 3d-4f RPES is a useful and promising technique to examine the bulk Ce 4f states. However, the energy resolution of all previous 3d-4f RPES is as poor as ~0.7 eV, which is unsatisfactory to observe detailed 4f states near the Fermi level (EF).
In order to realize a breakthrough for studying the bulk electronic states with using PES, we have constructed a high-resolution soft x-ray (hν = 500-1500 eV) PES system combined with a varied-line-spacing plane grating (VLSPGM) monochromator on a twin-helical undulator beam line BL25SU of SPring-8. Here we show the obtained bulk-sensitive high-resolution Ce 3d-4f RPES spectra of CeRu2Si2 and CeRu2 (energy resolution of about 100 meV at hn~880 eV!). CeRu2Si2 is a typical heavy fermion system with the Kondo temperature (TK) of about 20 K. This material is thought to be located at a boundary between the localized and itinerant Ce 4f states. On the other hand, CeRu2 is a typical material of the strongly valence-fluctuating 4f systems (namely, very strongly hybridized systems), where TK is of the order of 1000 K. The high-resolution Ce 3d-4f RPES was performed at BL25SU on SPring-8. The results were compared with the high-resolution Ce 4d-4f RPES spectra measured at BL-3B of the Photon Factory with the overall resolution of 50 meV.


Results and discussion

Firstly we show the Ce 4f spectra of CeRu2Si2 and CeRu2 over a wide energy range. In the 3d-4f spectrum of CeRu2Si2, there is a sharp peak near EF and a broad tail ranging from -1 to -5 eV. According to the single impurity Anderson model (SIAM), the former corresponds to the contribution of both the tail of the Kondo peak (f5/21) and its spin-orbit partner (f7/21). The latter is ascribed to the f0 final states; in the 4d-4f spectrum on the other hand, the f0 final states located at -2.5 eV is prominent, reflecting the localized character of the surface 4f states. In the 3d-4f spectrum of CeRu2, the f0 broad tail is greatly suppressed compared to CeRu2Si2, indicating that the bulk Ce 4f states are considerably hybridized. But the surface f0 state of CeRu2 is seen as a hump near -2 eV in the 4d-4f spectrum, as in CeRu2Si2. These differences reveal that the 4d-4f spectra mainly reflect the surface 4f electronic states.


The detailed high-resolution spectra near EF are shown in Fig. 3 for the 3d-4f and 4d-4f resonances. We note that the 4d-4f spectral line-shapes of CeRu2Si2 and CeRu2 are qualitatively similar, despite the TK values of these compounds differing by orders of magnitude. There are two comparable structures, located just below EF and near -0.3 eV. However, the 3d-4f spectra of these compounds differ greatly. For CeRu2Si2, there is a prominent peak in the vicinity of EF, and a weak shoulder near -0.3 eV. These two structure originate from the bulk f5/21 and f7/21 final states. in other words, from the tail of the Kondo peak and its spin-orbit partner, as predicted from SIAM. The observation of such a strong tail of the bulk Kondo peak has been made possible by the bulk-sensitive 3d-4f RPES technique with an unprecedentedly high resolution of 100 meV. The two peaks in the 4d-4f spectrum of CeRu2Si2 can also be assigned in the same manner, but to the surface f5/21 and f7/21 final states resulting from the weaker hybridization effect in the surface.
We can also assign the two peaks in the 4d-4f spectrum of CeRu2 in Fig. 3 as the surface f5/21 and f7/21 final states. The 3d-4f line-shape of CeRu2 is, however, rather surprising as no structure is seen except for a broad feature centered at -0.5 eV. One would have expected that the tail of the Kondo peak should be much stronger in the bulk-sensitive 3d-4f spectrum for a very strongly hybridized system like CeRu2, as predicted from SIAM. The experimental results that the 3d-4f spectrum shows a rather conventional Fermi cut-off and a broad peak at -0.5 eV cannot be deduced from SIAM. Therefore we conclude that the bulk 4f spectral line-shape in CeRu2 represents itinerant "4f-band" character due to a very strong hybridization effect, beyond the framework of SIAM. Such spectral behavior, which has not been revealed by previous 4d-4f RPES studies, reflects the real TK and bulk properties.


Finally we are convinced that this new photoemission technique realizes a breakthrough for studies of genuine bulk electronic states of many rare-earth and transition-metal compounds.


Akira Sekiyama