Journal of the Physical Society of Japan Vol. 74, No. 9, September, 2005, pp. 2538–2543 #2005 The Physical Society of Japan
Photoemission Spectroscopy of Sm4 As3 Using Soft and Hard X-rays Atsushi Y AMASAKI, Akira S EKIYAMA, Shin I MADA, Masanori T SUNEKAWA, Claudia D ALLERA1 , Lucio B RAICOVICH1 , Tien-Lin LEE2 , Akira O CHIAI3 and Shigemasa S UGA 1
Graduate School of Engineering Science, Osaka University, Toyonaka, Osaka 560-8531 INFM-Dipartimento di Fisica, Politecnico di Milano, Piazza Leonardo da Vinci 32, I-20133 Milano, Italy 2 European Synchrotron Radiation Facility (ESRF), BP 220, F-38043 Grenoble, France 3 Graduate School of Science, Tohoku University, Aoba-ku, Sendai 980-8578 (Received April 25, 2005; accepted June 21, 2005)
The bulk and surface electronic structures of Sm4 As3 have been investigated by hard and soft X-ray photoemission spectroscopies (PESs). PES with a wide range of photon energies (h’s) between 220 and 2450 eV demonstrates the absence of valence mixing in both high- and low-temperature phases, that is, the valence of Sm ions is definitely trivalent in the bulk and divalent on the surface. Atomic multiplet calculations taking into account the bulk and surface spectral weights at each h well reproduce the experimental valence band PES and Sm 3d core-level spectra, supporting the valence difference between bulk and surface in Sm4 As3 . KEYWORDS: hard X-ray, soft X-ray, photoemission spectroscopy, bulk sensitivity, Sm compounds DOI: 10.1143/JPSJ.74.2538
1.
Introduction
Recently, hard X-ray photoemission spectroscopy (HAXPES) has attracted much attention for investigating the bulk electronic structures of transition-metal and rareearth compounds.1–3) As has already been reported, soft Xray (SX) PES with a resonant photoemission technique can examine the 3d and 4 f electronic states in the bulk for 3d transition-metal and rare-earth compounds, respectively.4–9) The HAXPES has a longer probing depth than the SXPES due to the increase in the mean free path of photoelectrons and has shown to be insensitive to surface contamination. However, the HAXPES requires the high photon flux supplied by third-generation synchrotrons and high-performance electron analyzers to detect weak signals because the photoionization cross section decreases rapidly at high photon energies (h’s).10) These severe conditions hindered the application of this technique until recently. As has been mentioned above, both the photoionization cross section and the mean free path of photoelectrons, that is, the bulk sensitivity, are known to depend significantly on the kinetic energy of photoelectrons. Therefore, we can investigate the contributions of the bulk and surface spectral weights of a specific state by changing h. For this study, we focus on a particular Sm compound since some Sm compounds do not have only different electronic structures but also different valences in the bulk and on the surface. For a few decades, the valence of Sm metal has been intensively studied using, for instance, lowenergy PES with h’s up to 200 eV.11–18) Although the valence of an evaporated Sm changes sensitively from a pure trivalent to an intermediate valence in the bulk due to the presence of various substrates and their orientations, the valence on the surface favors a divalent state.14,15) Samarium pnictide Sm4 X3 exhibits various physical properties, such as a mixed valence and a charge order in X ¼ Bi, and a ferromagnetic order in X ¼ As and Sb.
E-mail:
[email protected]
Sm4 As3 , in which Sm ions are thought to be almost trivalent in the bulk, is one of the most suitable candidates for a test specimen in the present experiment because Sm divalent peaks are distinctly observed in low-energy photoemission experiments. The thus-far reported PES study of Sm4 As3 suggests the existence of divalent Sm ions on the surface.19) Furthermore, some experimental results have implied the possible existence of Sm2þ ions even in the bulk.20) In addition, the mean valence of Sm ions on the surface is still a controversial issue. Sm4 As3 has an anti-Th3 P4 crystal structure. It shows a ferromagnetic transition at TC ¼ 160 K.20) Below TC , the magnitude of the magnetic moment of Sm3þ is smaller than that of a theoretical saturation moment by a factor of 10, suggesting that a weak magnetization is possibly induced by a dense Kondo state. In the temperature dependence of resistivity, a Kondo-like anomaly is observed at around TC . In this paper, we report the difference in electronic structure between the bulk and surface of Sm4 As3 in detail. To systematically investigate the bulk and surface electronic structures, we have carried out the SX and HAXPESs in the h range between 220 and 2445 eV. The PES spectra obtained in this h range indicate a change in valence between the bulk and surface. Our theoretical analysis demonstrates the applicability of a simple model for Sm4 As3 and the usefulness of the PES performed in the wide energy range for the investigation of bulk and surface electronic structures. 2.
Experimental
Single crystals of Sm4 As3 were fractured in situ for SX and HAX valence band (VB) PESs, core-level spectroscopy (core-XPS), and X-ray absorption spectroscopy (XAS) at the SX beamline BL25SU of SPring-8 (h ¼ 220 {1750 eV) and at the HAX beamline ID32 of ESRF (h ¼ 2445 eV). The total energy resolutions (E) of the SX and HAX VB-PESs were set to be h=E 4000 and 5000, respectively. The configuration of the apparatus was set so as to collect the normal emission of the photoelectrons from the samples.
2538
J. Phys. Soc. Jpn., Vol. 74, No. 9, September, 2005
A. YAMASAKI et al.
Table I. Slater integrals and spin–orbit coupling constants obtained by HF method.21Þ The values are given in (eV). Slater integrals are reduced to their 70% (80%) values in the calculation for PES (XAS and core-XPS) spectra. 4f5
4f6
3d 9 4 f 5
3d9 4 f 6
F ð4 f 4 f Þ F 4 ð4 f 4 f Þ
14.625 9.229
13.643 8.563
12.515 7.803
15.313 9.675
14.389 9.046
F 6 ð4 f 4 f Þ
6.655
6.161
5.600
6.980
6.513
F 2 ð3d4 f Þ
9.627
9.010
F 4 ð3d4 f Þ
4.493
4.169
G1 ð3d4 f Þ
6.872
6.352
G3 ð3d4 f Þ
4.029
3.722
G5 ð3d4 f Þ
2.783
2.570
4 f
0.174
3d
0.160
0.145
0.200
0.185
10.298
10.305
Sm4As3 hv=2445 eV
Intensity (arb. units)
4f4 2
2539
1200 eV
700 eV 220 eV Cal.
The samples were cooled and maintained at 20 and 180 K for the PESs at the beamlines BL25SU and ID32, respectively. 3.
3+
Sm
15
Models
3.1 Atomic full multiplet calculation For the quantitative analysis of localized rare-earth compounds, atomic full multiplet calculation is well-known to be applicable. The required parameters such as Slater integrals and spin–orbit coupling constants are obtained using Cowan’s Hartree–Fock (HF) program, as summarized in Table I.21) The Slater integrals are obtained by reducing the HF values down to 70 or 80% to account for intra-atomic relaxation effects.22) Multiplet calculations of spectra were performed using Tanaka’s program.23) The calculated spectrum is broadened by a Gaussian and a Lorentzian (the full width at half maximum (FWHM); ¼ 60, 100, and 300 meV, ¼ 0:7 {1:6, 0.8, and 2.0 eV for the PES, XAS, and core-XPS, respectively), considering the broadening effects of the experimental resolution and lifetime of a photohole, respectively. 3.2 Mean free path of photoelectrons and bulk sensitivity When one discusses the bulk sensitivities of specific compounds at various h’s, the ratio of the bulk spectral weight to the surface spectral weight at each h is required. The bulk and surface spectral weights are given by s exp : bulk ð1Þ cos s 1 exp : surface; ð2Þ cos where s, , and denote the thickness of the surface layer, the inelastic mean free path of photoelectrons in solid, and the emission angle, respectively. In the present work, the emission angle is set to be normal to the surface ( ¼ 0). is expressed by Ek ¼ 2 ; ð3Þ Ep 0 lnð0 Ek Þ where Ek is the kinetic energy of photoelectrons and Ep ¼ 28:8ðNv =AÞ1=2 .24) Here, , A, and Nv are the molecular
Sm
2+
10 5 0 Binding Energy (eV)
Fig. 1. VB-PES spectra of Sm4 As3 at various h’s. These spectra are normalized at the intensity of the main peak in the Sm3þ multiplet. The bottommost spectrum is the calculated spectrum, in which the bulk and surface spectral weights at h ¼ 220 eV are considered (see Fig. 5). Broken lines indicate the positions of Sm 4 f multiplet peaks. Arrows show the electronic states related to the As 4s, 4p, and Sm 5d states.
density, the molecular weight, and the total number of valence electrons per molecule, respectively.25) 0 and 0 are empirical formulae containing , Ep , and the energy-band gap of the molecule Eg . In the case of Sm4 As3 , Eg is zero since it is metallic under our experimental conditions. Then, one can estimate the bulk sensitivity of Sm4 As3 in the SX and HAXPESs. 4.
Results and Discussion
4.1 Valences of Sm ions in bulk and on surface Figure 1 shows the VB-PES spectra of Sm4 As3 at various h’s. The photoionization cross section of Sm 4 f states at h ¼ 220 eV is larger than those of As 4s, 4p, and Sm 5d states by a factor of more than 60.10) The cross sections of the latter states relatively increase with h and become comparable to that of Sm 4 f states at a few thousands of eV. Therefore, the spectrum at h ¼ 220 eV originates predominantly from Sm 4 f electronic structures. These spectral features in the binding energy EB range between 0.7 – 4.0 eV and 5.5 –10.5 eV can be assigned to the multiplet of Sm divalent (j4 f 5 i final state) and trivalent (j4 f 4 i final state) components. Three components, which are located at EB ¼ 11:4, 3.1, and 0.5 eV, become prominent with increasing h. These are related to the As 4s, 4p, and Sm 5d states, considering the increases in their photoionization cross sections. The As 4p and Sm 5d states hybridize with each other and form bonding and antibonding bands.26) The peaks at EB ¼ 0:5 and 3.1 eV can be assigned to these antibonding and bonding bands in the initial state.
2540
J. Phys. Soc. Jpn., Vol. 74, No. 9, September, 2005
A. Y AMASAKI et al.
1.0
Bulk Spectral Weight
Intensity (arb. units)
Sm4As3 Sm 3d5/2 hv 1310eV 1750eV 2445eV
(EK) (~225eV) (~665eV) (~1360eV)
9
9
3d 4f 2+ (Sm )
5
3d 4f 3+ (Sm )
5
6
0
-5
-10
Sm4As3 0.8
2445eV 1200eV
0.6 0.4 0.2
-15
700eV 220 eV
1/4 of Unit Cell Unit Cell
0
Relative Binding Energy (eV)
0
5
10
15
20
Surface thickness (Å)
Fig. 2. Sm 3d core-XPS spectra of Sm4 As3 at various h’s.
Fig. 4. Surface-thickness dependence of bulk spectral weight in Sm4 As3 at various Ek ’s.
Intensity (arb. units)
Sm 3d-4f XAS 3d5/2
3d3/2
Sm4As3 (Exp.) 3+
pure Sm (Cal.)
-20
-10
0
10
20
30
Relative Photon Energy (eV) Fig. 3. Sm 3d–4 f XAS spectrum of Sm4 As3 . Calculated Sm 3d–4 f XAS spectra of a purely trivalent Sm ion are also shown.
Although the multiplet peaks of the Sm3þ component are still observed at h’s higher than 220 eV, those of the Sm2þ component are strongly suppressed with increasing h, that is, with the enhancement of the bulk sensitivity. This suggests that the observed Sm2þ peaks originate from the surface electronic structures. Sm 3d core-XPS spectra support this suggestion as shown in Fig. 2. The intensity of the j3d9 4 f 6 i final state (Sm2þ component) decreases relative to that of the j3d 9 4 f 5 i final state (Sm3þ component) with an increase in h. In addition, we have not observed the intensity of the bulk Sm2þ component, which should increase with h as opposed to the decrease in the intensity of the surface Sm2þ component. The XAS is known to be a more bulk sensitive technique than the PES. Figure 3 shows the Sm 3d– 4 f XAS spectrum of Sm4 As3 . The spectrum can be well reproduced by the calculated spectrum of a pure Sm3þ ion. This also supports the experimental evidence that Sm is trivalent in the bulk. 4.2 Bulk sensitivity for Sm4 As3 One of the experimental features in the present work is the wide tuning of the incident photon energy. In Fig. 4, we show the surface-thickness dependence of the calculated
bulk spectral weight for Sm4 As3 at various Ek ’s (’ h’s in VB-PES spectra) obtained using eqs. (1) and (3). At h ¼ 2445 eV, a large spectral weight from the bulk is observed in the VB-PES spectrum. The bulk spectral weight decreases with h and markedly decreases with increasing thickness of the surface layer at a low h of 220 eV. To estimate the bulk sensitivity for Sm4 As3 at these h’s, one must assume the thickness of the surface layer. In the surface layer of rare-earth compounds, the electron correlation is generally increased due to the reduced atomic coordination on the surface. Therefore, the valence of Sm4 As3 should be changed to divalent. We have assumed the thickness of the surface layer s to be a quarter of the lattice In the surface layer, only the constant (a ¼ 8:833 A). topmost Sm atom is contained, that is, the second and deeper Sm atoms from the surface have bulk electronic structures. The calculated mean free paths and bulk sensitivities at various h’s are summarized in Table II. Then, one can directly compare the calculated Sm 4 f spectrum with the experimental VB-PES spectrum at h ¼ 220 eV, in which the photoionization cross section of the Sm 4 f states is more prominent than those of the other states. According to the present experimental results, we conclude that the valences of the Sm ions in the bulk and surface layers are purely trivalent and divalent, respectively.
Table II. Calculated inelastic mean free paths (IMFPs) and bulk spectral weights (BSWs) for Sm4 As3 at various h’s (or kinetic energies Ek ’s). Surface thickness is defined as a quarter of the lattice constant. State Valence
Sm 3d core
h (eV)
Ek (eV)
220 700
IMFP (A)
BSW (%)
220
8.5
77.0
700
18.4
88.7
1200
1200
27.4
92.3
2445
2445
47.8
95.5
1310 1750
225 665
8.6 17.7
77.3 88.3
2445
1360
30.2
92.9
J. Phys. Soc. Jpn., Vol. 74, No. 9, September, 2005
A. YAMASAKI et al.
Exp.
Sm 5p
(hv=220 eV)
Cal. 3+
Bulk (Sm ) : 77% 2+ Surface (Sm ) : 23%
3+
Sm
30
25
hv 220eV 700eV 1200eV 2445eV
Sm4As3
20
15
10
5
2+
Sm
0
-5
Intensity (arb. units)
Intensity (arb. units)
Sm4As3
2541
3+
Sm
Sm
(Bulk) (77%) (89%) (92%) (96%)
2+
Binding Energy (eV)
We have calculated the VB- (Sm 4 f ) PES spectrum, considering the bulk and surface spectral weights at h ¼ 220 eV. The energy between the centers of gravity of the Sm3þ and Sm2þ components (G ) of the Sm 4 f states, and the of spectra ( ¼ 0:7 eV for Sm2þ and 1.6 eV for Sm3þ ) are parameters for fitting. G ð4 f Þ is fitted to be 5.6 eV. The features of the experimental spectrum are quantitatively in good agreement with those of the calculated spectrum, as shown in Fig. 5. This suggests that our hypothesis concerning the surface thickness is valid. It might be due to the unique high-density crystal structure of Sm4 As3 that the effective thickness of the surface layer in Sm4 As3 is smaller 12) than that in the Sm metal (s ¼ 3 A). Figure 6 shows the calculated Sm 4 f PES spectra, considering the bulk Sm3þ and surface Sm2þ spectral weights depending on h. The surface Sm2þ spectral weight is strongly suppressed relative to the bulk Sm3þ one with increasing h. This Sm2þ spectral weight behavior is consistent with the experimental one (see Fig. 1). When performing a quantitative analysis for the spectrum, one should note that even at h of a few thousands of eV, the spectral weight originating from the surface electronic structures contributes still to the experimental spectrum. We have further calculated the Sm 3d core-XPS spectra to quantitatively compare them with the experimental spectra since the calculated VB-PES spectrum cannot be directly compared with the experimental one at high h’s due to the increases in the spectral weights of other states relative to the Sm 4 f states. The calculated Sm 3d core-XPS spectra are shown in Fig. 7 (G ð3dÞ ¼ 8:2 eV and ¼ 2:0 eV for both Sm3þ and Sm2þ ). Although the calculated Sm2þ spectral weight at h ¼ 1310 eV is slightly larger than the experimental one, the overall spectral feature is reproduced by only the bulk Sm3þ and surface Sm2þ components. As empirically known, a core-XPS spectrum is more sensitive to the surface condition than a VB-PES spectrum. Thus, the surface spectral weight might decrease due to surface
15
10
5
0
Binding Energy (eV) Fig. 6. h dependence of Sm trivalent (bulk) and divalent (surface) spectral weights in calculated VB-PES spectrum. The variation in thermal broadening and experimental resolution at each h is not considered.
Sm4As3
Intensity (arb. units)
Fig. 5. Calculated VB-PES spectrum of Sm4 As3 . Dotted and dashed spectra show the contributions of the Sm trivalent and divalent components at h ¼ 220 eV, respectively. The experimental spectrum at h ¼ 220 eV is also shown. The Shirley background27) obtained from the experimental spectrum is added and the Fermi Dirac function (T ¼ 20 K) is multiplied in the calculated spectrum.
Sm 3d5/2 hv 1310eV 1750eV 2445eV
(Bulk) (77%) (88%) (93%)
2+
3+
Sm
Sm
5
0
-5
-10
-15
Relative Binding Energy (eV) Fig. 7. h dependence of Sm trivalent (bulk) and divalent (surface) spectral weights in Sm 3d core-XPS spectrum. Dots indicate the experimental spectra shown in Fig. 2.
contamination or oxidization. This effect should be reduced at high h’s because of the large spectral weight of the bulk. In fact, the discrepancy of the spectrum between the calculation and experiment becomes smaller at h ¼ 2445 eV than at 1310 eV. One should note that the deviation between the calculations and experiments in the Sm3þ peak does not arise from the Sm3þ component of the surface, because the spectral shape is never changed by the variation in h. 4.3 Possibility of realizing ordinary Kondo state in Sm4 As3 Valence mixing is often observed in Sm compounds such as Sm3 Se4 , SmB6 , Sm4 Bi3 , and others.19,28,29) In these compounds, the 4 f states of Sm2þ are located just below the Fermi level (EF ). Meanwhile, the bulk 4 f states of Sm2þ in Sm4 As3 are located above EF , because the Sm 5d states are
2542
J. Phys. Soc. Jpn., Vol. 74, No. 9, September, 2005
partially occupied instead of the 4 f states of Sm2þ , as shown in Fig. 1. The temperature dependence of the resistivity, magnetization, and susceptibility indicates that Sm4 As3 might be a dense Kondo system, in which the Kondo temperature (TK ) is about 150 K.20) It has been pointed out that the unoccupied 4 f states, which are located very close to EF , play an important role in realizing the dense Kondo character of Sm 4 f electrons. Namely, As 4p electrons transfer from occupied conduction states to unoccupied 4 f states due to the effective c– f hybridization. In other words, the minor state, which is mixed with the major j4 f n i state in the configuration of the initial state in Sm4 As3 (n ¼ 5), is not the j4 f n1 i state in Ce (n ¼ 1) Kondo systems but the j4 f nþ1 i state in Pr (n ¼ 2) Kondo systems.2,6) Let us compare the spectral feature in Sm4 As3 with those in the Ce and Pr Kondo systems. Although a high-TK system such as that proposed for Sm4 As3 has not thus far been discovered in Pr compounds, the Kondo resonance ‘‘peak’’ is observed just below EF in PrFe4 P12 , which has a relatively low TK (of the order of tens of K).6) In a Ce Kondo system, 1 CeNi, in which TK is around 150 K, has a prominent f5=2 feature, the so-called ‘‘tail of the Kondo resonance peak’’ in the vicinity of EF in the Ce 3d–4 f resonant PES spectrum.30) Any signs of the Kondo resonance are, however, not observed in the VB-PES spectra of Sm4 As3 . In addition, there is no prominent feature near EF in the X-ray bremsstrahlung isochromat spectrum of Sm4 As3 ,31) in contrast to CeNi.32) Furthermore, the degree of mixing of configurations between the j4 f 5 i and j4 f 6 i states in the initial state must be estimated in order to discuss the potentiality of the realization of the 4 f Kondo character in this system. If the degree of mixing of the two configurations, j4 f 5 i and j4 f 6 i states, is sufficiently large to realize the Kondo state with TK ’ 150 K, the spectral weight of the Sm2þ should reach up to 15%, according to the analogy with CeNi (estimated by Ce 3d– 4 f resonant PES and noncrossing approximation calculation based on the single-impurity Anderson model).33) However, we have never observed any spectral weights originating from the bulk Sm2þ (4 f 6 ) component in the Sm 3d core-XPS and VB-PES spectra both below and above TC (T ¼ 20 and 180 K). We would conclude even above TC that the degree of mixing of the j4 f 5 i and j4 f 6 i configurations is negligible (less than 1% of the j4 f 6 i configuration that has been estimated by Sm 3d core-XPS). Therefore, the Kondo-like anomaly that has been observed in other measurements20) does not originate from the ordinary Kondo mechanism observed in other rare-earth compounds. Alternative mechanisms such as the magnetic scattering and/or the relatively large spin to the orbital magnetic moment34) might be responsible for the Kondo-like anomaly and small magnetic moments in high- and low-temperature phases. 5.
Conclusion
We have demonstrated that photoemission spectroscopy (PES) with a wide range of photon energies is promising for studying the bulk and surface electronic structures of rareearth compounds. The PES spectra of Sm4 As3 at h ¼ 220 { 2445 eV indicate that Sm ions are trivalent and divalent in the bulk and surface layers, respectively. Even in the nonmagnetic high-temperature phase, the degree of mixing
A. YAMASAKI et al.
of the valences between di- and trivalent states is negligible. Therefore, the Kondo-like anomaly observed in other measurements does not arise from the ordinary Kondo mechanism. Acknowledgment We would like to thank A. Tanaka for providing the multiplet calculation program. The research was performed at SPring-8 (Proposal No. 2000A0114-NS-np) and ESRF (Proposal No. HE1422) under the support of a Grant-in-Aid for COE Research (10CE2004) and Creative Scientific Research (15GS0213) from the Ministry of Education, Culture, Sports, Science and Technology, Japan.
1) L. Braicovich, N. B. Brookes, C. Dallera, M. Salvietti and G. L. Olcese: Phys. Rev. B 56 (1997) 15047. 2) A. Yamasaki, S. Imada, A. Sekiyama, M. Tsunekawa, C. Dallera, ¯ nuki L. Braicovich, T.-L. Lee, H. Sugawara, H. Sato, R. Settai, Y. O and S. Suga: J. Phys. Soc. Jpn. 74 (2005) 2045. 3) K. Kobayashi, M. Yabashi, Y. Takata, T. Tokushima, S. Shin, K. Tamasaku, D. Miwa, T. Ishikawa, H. Nohira, T. Hattori, Y. Sugita, O. Nakatsuka, A. Sakai and S. Zaima: Appl. Phys. Lett. 83 (2003) 1005. ¯ nuki and 4) A. Sekiyama, T. Iwasaki, K. Matsuda, Y. Saitoh, Y. O S. Suga: Nature (London) 403 (2000) 396. 5) A. Sekiyama, K. Kadono, K. Matsuda, T. Iwasaki, S. Ueda, S. Imada, ¯ nuki and Y. Saitoh: J. Phys. Soc. S. Suga, R. Settai, H. Azuma, Y. O Jpn. 69 (2000) 2771. 6) A. Yamasaki, S. Imada, T. Nanba, A. Sekiyama, H. Sugawara, H. Sato, C. Sekine, I. Shirotani, H. Harima and S. Suga: Phys. Rev. B 70 (2004) 113103. 7) R.-J. Jung, B.-H. Choi, S.-J. Oh, H.-D. Kim, E.-J. Cho, T. Iwasaki, A. Sekiyama, S. Imada, S. Suga and J.-G. Park: Phys. Rev. Lett. 91 (2003) 157601. 8) A. Sekiyama et al.: Phys. Rev. Lett. 93 (2004) 156402. 9) S. Suga, A. Shigemoto, A. Sekiyama, S. Imada, A. Yamasaki, A. Irizawa, S. Kasai, Y. Saitoh, T. Muro, N. Tomita, K. Nasu, H. Eisaki and Y. Ueda: Phys. Rev. B 70 (2004) 155106. 10) J. J. Yeh and I. Lindau: At. Data Nucl. Data Tables 32 (1985) 1. 11) G. K. Wertheim and M. Campagna: Chem. Phys. Lett. 47 (1977) 182. 12) G. K. Wertheim and G. Crecelius: Phys. Rev. Lett. 40 (1978) 813. 13) J. W. Allen, L. I. Johansson, R. S. Bauer, I. Lindau and S. B. M. Hagstro¨m: Phys. Rev. Lett. 41 (1978) 1499. 14) B. Johansson: Phys. Rev. B 19 (1979) 6615. 15) A. Rosengren and B. Johansson: Phys. Rev. B 26 (1982) 3068. Fa¨ldt and H. P. Myers: Phys. Rev. Lett. 52 (1984) 1315. 16) A. 17) J. N. Andersen, I. Chorkendorff, J. Onsgaard, J. Ghijsen, R. L. Johnson and F. Grey: Phys. Rev. B 37 (1988) 4809(R). 18) A. Stenborg, J. N. Andersen, O. Bjo¨rneholm, A. Nilsson and N. Martensson: Phys. Rev. Lett. 63 (1989) 187. 19) S. Suga, A. Sekiyama, S. Imada, T. Suzuki, H. Aoki and A. Ochiai: Physica B 281– 282 (2000) 158. 20) A. Ochiai, T. Suzuki and T. Kasuya: J. Magn. Magn. Mater. 52 (1985) 13. 21) R. D. Cowan: The Theory of Atomic Structure and Spectra (University of California Press, Berkeley, 1981). 22) G. van der Laan and I. W. Kirkman: J. Phys.: Condens. Matter 4 (1992) 4189. 23) A. Tanaka, T. Jo and G. A. Sawatzky: J. Phys. Soc. Jpn. 61 (1992) 2636. 24) S. Tanuma, C. J. Powell and D. R. Penn: Surf. Sci. 192 (1987) L849. We have assumed Nv to be 25) The lattice constant of Sm4 As3 is 8.833 A. 21, which consists of Sm 5d1 6s2 and As 4p3 electrons.26) 26) K. Takegahara, H. Harima and T. Kasuya: J. Magn. Magn. Mater. 52 (1985) 307. 27) D. A. Shirley: Phys. Rev. B 5 (1972) 4709. 28) S. Suga, S. Imada, T. Jo, M. Taniguchi, A. Fujimori, S.-J. Oh, A. Kakizaki, T. Ishii, T. Miyahara, T. Kasuya, A. Ochiai and T. Suzuki: Phys. Rev. B 51 (1995) 2061.
J. Phys. Soc. Jpn., Vol. 74, No. 9, September, 2005 29) J.-N. Chazalviel, M. Campagna, G. K. Wertheim and P. H. Schmidt: Solid State Commun. 19 (1976) 725. 30) A. Sekiyama, K. Kadono, T. Iwasaki, S. Imada, S. Kasai, S. Suga, ¯ nuki: Acta Phys. Pol. B 34 (2003) 1105. S. Araki and Y. O 31) S. Suga, S. Imada, A. Ochiai and T. Suzuki: Physica B 186 –188 (1993) 59.
A. YAMASAKI et al.
2543
32) F. U. Hillebrecht, J. C. Fuggle, G. A. Sawatzky, M. Campagna, O. Gunnarsson and K. Scho¨nhammer: Phys. Rev. B 30 (1984) 1777. 33) A. Sekiyama, K. Kadono, T. Iwasaki, S. Imada, S. Kasai, S. Suga, ¯ nuki: unpublished. S. Araki and Y. O 34) S. Suga and S. Imada: J. Electron Spectrosc. Relat. Phenom. 78 (1996) 231.
