Valence Transition of YbInCu 4 Observed by Photoemission Spectroscopy

First-order valence transition of YbInCu 4 at T V =42 K has directly been observed by photoemission spectroscopy (PES) using wide excitation photon energies from extremely low-energy region ( h(cid:23) =7 eV) to soft x-ray region ( h(cid:23) =800 eV), and further to hard x-ray region ( h(cid:23) =6 keV). The valence transition is more sharply observed in the hard x-ray PES spectra than the soft x-ray PES spectra. We demonstrate that the Yb 3 d hard x-ray PES is very suitable for the estimation of the Yb valences with high accuracy, and they have been derived to be ∼ 2.90 in the high-temperature phase and ∼ 2.74 in the low-temperature phase. In the low-energy PES spectra, on the other hand, the hybridization band between the conduction-band (CB) and Yb 4 f states is observed as a remarkable structure at 47 meV below the Fermi level. This structure suddenly appears below T V , indicating that a degree of the c - f hybridization increases in the low-temperature phase. [DOI: 10.1380/ejssnt.2011.90]


I. INTRODUCTION
YbInCu 4 exhibits isostructural first-order valence transition at T V =42 K [1][2][3]. The crystal structure is C15btype as shown in Fig. 1 and the volume expands by 0.45 % below T V with no change of the crystal structure. In accordance with the valence transition, abrupt changes are observed in the electrical resistivity, magnetic susceptibility, and the other physical properties. The Kondo temperature, which is a characteristic parameter for Yb compounds, also increases from T K+ ∼25 K above T V to T K− ∼400 K below T V [4]. The Yb valence of z ∼ 3 in the high-temperature phase decreases to z ∼ 2.9 in the lowtemperature phase, which are estimated indirectly from the magnetic susceptibility of the Curie-Weiss behavior in the high-temperature phase and the lattice expansion at T V . On the other hand, the Yb L III -edge absorption spectroscopy measurements provide z ∼ 2.9 above T V and † Corresponding author: jinjin@hiroshima-u.ac.jp z ∼ 2.8 below T V [2]. In order to observe the valence transition of YbInCu 4 directly, a great number of photoemission spectroscopy (PES) experiments has been carried out since its discovery of this compound [5][6][7][8][9][10][11][12][13][14][15][16]. Reinert et al., for the first time, carried out the temperature-dependent vacuum ultraviolet PES (VUV-PES) experiments precisely with an excitation energy of hν=43 eV with the energy resolution of 80 meV [7,17]. A Yb 2+ 4f 7/2 -derived peak showing up in the vicinity of the Fermi level (E F ) is indeed enhanced with decreasing temperature. The Yb valence estimated from the intensity ratio of the Yb 2+ and Yb 3+ 4f structures in the valence-band region, however, continuously decreases from z ∼ 2.85 at 220 K to z ∼ 2.56 at 20 K with no indication of the valence transition in contrast to the abrupt change expected from the thermodynamic data, that is, the magnetic susceptibility and lattice expansion. Furthermore, the estimated Yb valences are smaller than those from the thermodynamic data. In consideration of the experimental results, they proposed an existence of a subsurface region with different physical properties from the bulk, within small probing depth (λ) of VUV-PES at hν=43 eV [7], where λ is estimated to be only λ ∼ 5Å.
As is well known, λ depends on kinetic energy (E K ) of photoelectrons detected in the PES measurements and reaches minimum around E K ∼ 50 eV, which roughly corresponds to the case of the valence-band VUV-PES at hν=43 eV. In order to settle the discrepancy between the VUV-PES and thermodynamic results for YbInCu 4 , it is important to probe the electronic states in much deeper region from the surface by increasing (decreasing) E K using the higher (lower) excitation energy in the PES measurements.
In this review article, we present our PES results of YbInCu 4 using wide excitation energies from the extremely low-energy region (hν=7 eV) [15] to soft x-ray region (hν=800 eV), [12] and further to hard x-ray region (hν=6 keV) [13]. Hereafter, we abbreviate these exper- The Surface Science Society of Japan (http://www.sssj.org/ejssnt) iments to LE-PES, SX-PES, and HX-PES, respectively, and refer the VUV-PES results at hν=43 eV by Reinert et al. [7] for comparison. The λ-value in the case of YbInCu 4 is estimated to be λ ∼ 15Å for SX-PES and λ ∼ 75Å for HX-PES according to Ref. [18], and is expected to be more than 100Å for LE-PES [19,20]. The valence transition is the most abruptly observed in the HX-PES and also LE-PES spectra. We demonstrate that Yb 3d HX-PES is the powerful method to estimate the Yb valence with high accuracy, while LE-PES near E F is useful to investigate the conduction-band (CB) states hybridized with the Yb 4f states.

II. EXPERIMENTAL
SX-PES at hν=800 eV, HX-PES at hν=6 keV, and LE-PES at hν=7 eV for YbInCu 4 were carried out at the undulator beam lines BL-25SU of SPring-8, BL-29XU of SPring-8, and BL-9 of Hiroshima Synchrotron Radiation Center (HSRC), respectively. For all PES experiments, hemispherical photoelectron analyzers (Scienta SES200 (SX-PES), SES2002 (HX-PES, LE-PES)) were used to measure the angle-integrated PES spectra. The respective total energy resolutions estimated from the Fermi edge of Au-spectra were around 100, 270, and 8 meV. Binding energy is referred to E F , also determined from the Au-spectra. Clean surfaces were in situ obtained by fracturing under ultrahigh vacuum. After the fracturing, the PES experiments were carried out only on the first cooling across the valence transition in order to prevent a formation of defects in the sample by repetition passing through the transition [21].
YbInCu 4 single crystals were grown by the flux growth method similar to that described by Sarrao et al. [21]. The crystal structure with the C15b-type was confirmed by means of x-ray powder diffraction measurements. The temperature width of the valence transition at T V =42 K was within 2 K from the magnetic susceptibility measurements. Fig. 2 (a) shows the valence-band SX-PES spectra of YbInCu 4 measured at 30 K with hν=800 eV [12]. We find an intense peak in the vicinity of E F originating from the Yb 2+ 4f 7/2 states with its spin-orbit partner of the 4f 5/2 peak at 1.45 eV. Surface contributions next to the two Yb 2+ 4f peaks are still observed in the spectrum as indicated by vertical bars, though those are fairly reduced in comparison with the VUV-PES spectra at hν=43 eV [7]. The Yb 3+ 4f -derived multiplet structures due to the Coulomb interaction between two 4f holes in the PES final states, are widely spread over the 5.5-12 eV region. A doublet-like structure at 2.5-5 eV is ascribed to the Cu 3d states. The spectrum almost reflects only the Yb 4f states except for the Cu 3d states taking into account the photoionization cross sections [22]. Figures 2 (b) and (c) show temperature dependences of the Yb 2+ and Yb 3+ 4f photoemission regions of the SX-PES spectra in Fig. 2  temperature from 100 to 50 K, the Yb 2+ 4f 7/2 and 4f 5/2 peaks in Fig. 2(b) gradually increase in intensity. Between 50 and 40 K, we notice a remarkable enhancement in those peaks, which should reflect the valence transition of YbInCu 4 . The Yb 3+ 4f multiplet structures in Fig. 2(c) are, on the other hand, suddenly reduced with the energy shift toward deeper side by ∼ 50 meV. The abrupt change of the SX-PES spectra across the valence transition is contrast to the gradual change of the VUV-PES spectra, obviously indicating that the increase of λ is essentially important to probe the valence transition of YbInCu 4 .

III. RESULTS AND DISCUSSION
We point out here that the valence-band SX-PES has an additional advantage other than its larger λ compared with the valence-band VUV-PES. Since the Yb 4f states except for the Cu 3d ones dominantly contribute to the SX-PES spectra as seen from Fig. 2(a), the Yb 2+ and Yb 3+ 4f components required for the estimation of the Yb valence are extracted with higher accuracy compared with the VUV-PES spectra, where, in particular, the Yb 3+ 4f states are observed only as substantially weak multiplet structures with the large background contributions.
In order to estimate the Yb valence, we have carried out the fitting analysis of the SX-PES spectra after removing the Cu 3d contribution appropriately and the background contribution due to secondary electrons according to the conventional Shirley's method [23]. Two surface Yb 2+ 4f 7/2 components are extracted by assuming the Gaussian functions and the same assumption is applied for the surface Yb 2+ 4f 5/2 components. The fitting analysis using the theoretical Yb 3+ 4f multiplets [24] leads to the Yb valences plotted in Fig. 5 together with those obtained from the other PES measurements. We find that the obtained Yb valence becomes closer to trivalent in comparison with the VUV-PES results and a sudden drop across T V is succesfully detected. We also notice, however, the gradual decrease of the Yb valence from 100 to 50 K already in advance of the valence transition.
A comparison between the Yb valences derived from the SX-PES and VUV-PES strongly suggests that the valence transition, taking place in the bulk, is more sharply observed by PES measurements with further higher excitation energy, which is also expected to provide the more genuine Yb valence. We have, thus, performed HX-PES at hν=6 keV on YbInCu 4 [13]. The valence-band HX-PES spectra measured at 220, 50 and 20 K are depicted in Fig. 3. At a glance, we notice a tremendous enhancement of the Yb 2+ 4f 7/2 and 4f 5/2 peaks between 50 and 20 K with the reduction of the Yb 3+ 4f structures. In addition, almost no Yb 2+ 4f peak appears at 220 K, which indicates that the Yb valence in the high-temperature phase is indeed close to trivalent. Although we can also estimate the Yb valence from the valence-band HX-PES spectra, non-Yb 4f states with the complicated feature largely contribute to the spectra, due to the relatively large photoionization cross sections of the sp states at several keV [22], and it is difficult to unambiguously remove its contribution.
We demonstrate here that the Yb 3d HX-PES, not the valence-band HX-PES, can be the powerful method to estimate the Yb valence with high accuracy. It is noted that we cannot excite the Yb 3d core electron using the conventional Mg and Al K α lines (1253.6 and 1486.7 eV) because of its deep binding energy. Figure 4 shows the temperature dependence of the Yb 3d HX-PES spectra of YbInCu 4 measured between 220 and 10 K. The Yb 3d spectrum is split into the 3d 5/2 region at 1515-1540 eV and 3d 3/2 region at 1560-1585 eV due to the spin-orbit interaction. Each component is further split into the Yb 2+ and Yb 3+ -originated structures. The Yb 2+ 3d 5/2 (3d 3/2 ) state is observed as a single peak at 1519.5 (1567) eV, while the Yb 3+ 3d 5/2 (3d 3/2 ) state as multiplet structures at 1524-1536 (1572-1584) eV due to the Coulomb interaction between the Yb 4f and 3d holes. Broad structures at 1543 and 1591 eV come from plasmon excitations. We notice a dramatic enhancement of the Yb 2+ 3d peaks between 50 and 30 K with the sudden reduction of the Yb 3+ multiplet structures. In addition, the spectral feature is almost unchanged in the high-temperature phase and also in the low-temperature phase. We would like to emphasize here that the Yb 3d HX-PES is very suitable to estimate the Yb valence directly with respect to the following points; 1) The Yb 2+ and Yb 3+ -derived structures are well separated, 2) Except for the plasmon structures, no other core states from the In and Cu ions, and no other contribution such as the Auger electrons disturb the spectra, 3) The hybridization effect is negligible in contrast to the valence-band PES spectra, where the contribution of the CB states appears just in the same region of the Yb 2+ 4f peak through the hybridization. The analysis of the Yb 3d spectra is obvious and the Yb valence can solely be derived from the intensity ratio of the Yb 2+ and Yb 3+ components. The Yb 3d HX-PES can thus be regarded as the straightforward method for the quantitative estimation of the Yb valence with high accuracy.
In order to estimate the Yb valence, we have carried out the fitting analysis of the Yb 3d HX-PES spectra after removing the background contribution [23] and the plasmon contributions assuming the Gaussian functions. The Yb valences derived from the fitting analysis of the Yb 3d HX-PES using the theoretical multiplets, are compared with those from the valence-band SX-PES and VUV-PES spec- HX-PES at hν=6 keV [13], valence-band SX-PES at hν=800 eV [12], and valence-band VUV-PES at hν=43 eV [7], together with the expected values from thermodynamic data. After Ref. [13]. tra in Fig. 5. The Yb valence at 220 K is z ∼ 2.90, almost constant down to 50 K, and then sharply drops to z ∼ 2.74 between 50 and 30 K across the valence transition. Here we recall that λ ∼ 5, ∼ 15, and ∼ 75Å for HX-PES, SX-PES and VUV-PES, respectively. With increasing λ, the estimated Yb valence becomes close to trivalent as expected from the thermodynamic data, and its change is the sharpest for the HX-PES results among the three PES measurements. These λ-dependent or hν-dependent results clearly indicate the existence of the subsurface region as proposed by Reinert et al. [7] and that the Yb ions in the subsurface region exist with the valence closer to divalent compared with the bulk Yb ions. The gradual decrease of the Yb valence with decreasing temperature above T V in the SX-PES and VUV-PES results, suggests the higher transition temperature in the subsurface region than in the bulk. Recently, Suga et al. reported the Yb 3d HX-PES spectra of YbInCu 4 taken at hν=8 keV and the slightly larger Yb valences than our HX-PES results at hν=6 keV; z ∼ 2.93 above T V and z ∼ 2.76 below T V [16]. They further deduced the Yb valences in the bulk from the comparison of the Yb 3d HX-PES results at hν=8 and 6 keV, and the valence-band SX-PES results at hν=800 and 700 eV using that the PES spectral weight from a depth region of d from the surface is given by exp[−d/λ]. With the surface Yb ions fixed to divalent, they determine the thicknesses of the surface and subsurface, and the Yb valences in the subsurface and bulk from the four PES results with the different excitation energies. The derived Yb valences in the bulk are z ∼ 2.98 above T V and z ∼ 2.84 below T V .
Finally, in Fig. 6, we present the temperaturedependent LE-PES spectra in the vicinity of E F of YbInCu 4 measured between 80 and 20 K. At the excitation energy of hν=7 eV, the photoionization cross sections of the Yb 4f states is substantially small and the LE-PES spectra reflect almost the CB states. At a glance, we recognize the abrupt enhancement of the structures at 47 meV below T V . The abrupt change indicates that the LE-PES actually probes the bulk-originated electronic states as HX-PES and that the CB states change in accordance with the valence transition. From the energy position of 47 meV, roughly corresponding to k B T K in the low-temperature phase (T K− ∼ 400 K) [4], the structure showing up below T V would plausibly be related to the Kondo resonance state from the viewpoint of the CB states. An appearance of the 47-meV structure also indicates that a degree of the c-f hybridization increases in the low-temperature phase, in qualitatively agreement with the increase of T K (T K+ ∼ 25 K) [4]. The LE-PES results of Y-doped samples Y x Yb 1−x InCu 4 support that the 47-meV structure is related to the Kondo resonance state [25]. The 47-meV structure for x=0 shifts toward E F side; ∼39 meV for x=0.1 and ∼31 meV for x=0.2, in agreement with the decrease of T K− with x suggested from the magnetic susceptibility measurements [26]. The LE-PES is, thus, adequate to observe the c-f hybridization for the Yb compounds. On the other hand, the structure corresponding to the Kondo temperature of T K+ ∼ 25 K in the high-temperature phase is not detected, probably because the measurement temperature is inevitably higher than T K+ and the LE-PES spectra around E F are broadened due to a thermal effect.
Recently, we found that the Cu-derived CB electronic states change with the valence transition by means of the Cu 2p soft x-ray absorption spectroscopy and the Cu 2p HX-PES [27]. We have also performed the x-ray single crystal structure analysis of YbInCu 4 using synchrotron radiation diffraction, and found that the Cu ions behave as a Cu 4 cluster in the crystal and that its size is almost constant across T V averagely in spite of the lattice expansion [28]. In order to investigate the valence transition of YbInCu 4 from the structural viewpoint focusing on the Cu ions in more detail, the Yb L α and Cu K α x-ray fluorescence holography experiments on YbInCu 4 is now in progress.

IV. CONCLUSION
We have carried out SX-PES, HX-PES, and LE-PES in order to directly observe the first-order valence transi-tion of YbInCu 4 . With increasing λ, the valence transition is successfully detected as the abrupt change at T V in the spectra, in particular, in the HX-PES and LE-PES spectra. The comparison among the Yb valences obtained from VUV-PES, SX-PES, and HX-PES settles discrepancy between the VUV-PES and thermodynamic results and clarifies the existence of the subsurface region. We have shown that the Yb 3d HX-PES is the powerful method to estimate the Yb valence of the Yb compounds directly. The Yb 3d HX-PES has recently been a standard tool to estimate the Yb valence of the Yb compounds [29,30], since we reported the results of YbInCu 4 . We point out that in order to estimate the Yb valence in the bulk more precisely, it is important to measure the several Yb 3d HX-PES spectra with changing hν, as Suga et al.. have demonstrated [16]. On the other hand, the LE-PES is useful to observe the bulk-originated CB-states, which determine the physical properties of the Yb compounds together with the Yb 4f states.