PERGAMON
Solid State Communications 118 (2001) 563±567
www.elsevier.com/locate/ssc
Electronic structure of CoSe2 studied by photoemission spectroscopy using synchrotron radiation H. Sato a,*, F. Nagasaki a, Y. Kani a, S. Senba b, Y. Ueda c, A. Kimura a, M. Taniguchi a,b b
a Graduate School of Science, Hiroshima University, Kagamiyama 1-3-1, Higashi-Hiroshima 739-8526, Japan Hiroshima Synchrotron Radiation Center, Hiroshima University, Kagamiyama 2-313, Higashi-Hiroshima 739-8526, Japan c Kure National College of Technology, Agaminami 2-2-11, Kure 737-8506, Japan
Received 1 November 2000; received in revised form 9 March 2001; accepted 5 April 2001 by H. Akai
Abstract The electronic structure of the valence-band of CoSe2 has been investigated by means of synchrotron radiation photoemission spectroscopy. A main peak 1.0 eV below the Fermi level (EF) is attributed to the emission of the Co 3d states with t2g symmetry, while a shoulder of the main peak near EF originates from the Co 3d states with eg symmetry. The features of the photoemission spectrum are quite similar to those of CoS2. Electron correlation effects are stronger for CoSe2 than for CoS2. q 2001 Elsevier Science Ltd. All rights reserved. PACS: 71.20.Be; 79.60.2i Keywords: D. Electronic band structure; E. Synchrotron radiation
Pyrite-type 3d transition metal dichalcogenides exhibit a wide variety of electrical and magnetic properties [1]. Among them, CoSe2 is a metallic compound and its magnetic ground state is known to be an exchange-enhanced Pauli paramagnet [2,3]. The magnetic susceptibility obeys the Curie±Weiss law at temperatures above ,150 K with the Curie constant corresponding to a magnetic moment around 1 mB per Co atom [4]. CoSe2 is an end-point material of the Co(SxSe12x)2 system with x 0. CoS2 with x 1 is a ferromagnetic metal with the Curie temperature TC of 124 K [5]. TC rapidly decreases by a substitution of Se for S and the system becomes paramagnetic by the 12% substitution [5,6]. In the region of 0.28 # x # 0.6, the metamagnetism has been found [5±7]. Thus, the chalcogen elements play an important role for the magnetism of these pyrite-type 3d transition metal dichalcogenides through the p±d hybridization. The electronic structure of CoS2 has been investigated extensively so far. From the electroscopic viewpoint, it
* Corresponding author. Tel.: 181-824-24-7471; fax: 181-82424-0719. E-mail address:
[email protected] (H. Sato).
has been pointed out that the electron correlation effect is important in CoS2 [8,9], though this compound is, in general, classi®ed into the itinerant compound. Fujimori et al. have analyzed the photoemission spectrum of CoS2 using a con®guration interaction calculation with a (Co[S±S]6) 2 10 cluster model and commented that CoS2 is located near the boundary region between the charge-transfer and Mott± Hubbard regimes [8]. On the other hand, a few experimental and theoretical works have been reported for the electronic structure of CoSe2. Valence-band electronic structure of CoSe2 has been investigated by ultraviolet and X-ray photoemission spectroscopies [10,11]. However, the valenceband electronic structure has not been revealed in detail because of the insuf®cient energy resolution of the experiments. Information on the electronic structure of CoSe2, in particular, Co 3d contribution to the valence bands, is expected to give a clue to understand the ferromagnetism of CoS2. In this paper, we report the valence-band electronic structure of CoSe2 investigated by resonant photoemission spectroscopy in the Co 3p±3d excitation region using synchrotron radiation. The Co 3p±3d resonant photoemission experiments for CoSe2 were carried out at beamline BL-3B of the Photon Factory, High Energy Accelerator Research Organization. A combination of a 24-m spherical grating monochromator
0038-1098/01/$ - see front matter q 2001 Elsevier Science Ltd. All rights reserved. PII: S 0038-109 8(01)00172-7
564
H. Sato et al. / Solid State Communications 118 (2001) 563±567
Fig. 1. An absorption spectrum of CoSe2 in the Se 3d±4p and Co 3p±3d excitation region measured with a total electron yield mode.
and a double-stage cylindrical-mirror analyzer was used to measure the angle-integrated photoemission spectra. Total energy resolution was estimated to be ,0.3 eV around the photon energy (hn ) of 80 eV. Throughout this paper, we de®ne a binding energy with respect to the Fermi level (EF). Samples used in the present experiments were polycrystalline CoSe2. An appropriate amount of Co and Se enclosed in a quartz ampoule was kept at 9008C for 3 weeks. The grown samples were characterized by X-ray powder diffraction measurements. No signal ascribed to other phases was detected. The clean surfaces of the samples were obtained by scraping with a diamond ®le in an ultra-high vacuum chamber with a base pressure below 5 £ 10 210 Torr. After scraping, the photoemission spectra were in situ measured at room temperature. Fig. 1 shows the core-absorption spectrum of CoSe2 in the photon energy region from 50 to 70 eV measured with the total electron yield mode. One notices the Co 3p±3d absorption structure from hn 58 to 64 eV and a broad dip around 65 eV after the Co 3p±3d absorption. The whole feature from 58 to 70 eV is similar to that of CoS2 [8] including the broad dip. The feature of a prominent peak around 55 eV due to the Se 3d absorption is very similar to that of the trigonal Se with the 4p states in the conduction bands [12]. In comparison with that, the doublet structures are assigned to the spin-orbit splitting states: the most intense peak at 55.6 eV and a shoulder at 54.7 eV are due to the Se 3d3/2 and Se 3d5/2 states, respectively [13]. The peak structure means that the Se 4p states contribute to the conduction bands of CoSe2. Fig. 2 shows a series of valence-band photoemission spectra of CoSe2 measured at the excitation photon energies from 30 to 125 eV including the Co 3p±3d and Se 3d±4p excitation regions. The spectra have been normalized by the photon ¯ux except for the spectra measured at hn 30, 80 and 125 eV. The photoemission spectrum at hn 30 eV
Fig. 2. A series of the valence-band photoemission spectra of CoSe2 taken at excitation photon energies from 30 to 125 eV.
exhibits a main peak at 1.0 eV with a shoulder on the lower binding-energy side and two broad structures around 3 and 6.5 eV, which are indicated by vertical bars in the ®gure. With increasing the excitation photon energy from 30 to 50 eV, the feature of the photoemission spectra drastically changes. The main peak at 1.0 eV is remarkably enhanced, while two broad structures are considerably suppressed. These changes are due to energy-dependent photoionization cross sections [14]. The valence bands of CoSe2 are composed of mainly the Co 3d and Se 4p states. At hn 30 eV, the cross section of the Se 4p states is about half of that of the Co 3d states. Thus, the photoemission spectrum at hn 30 eV provides a few information on the Se 4p states other than the Co 3d states in the valence bands. At hn 50 eV, on the other hand, a ratio of the cross sections of the Se 4p and Co 3d states is [Se 4p]/[Co 3d] , 1/20 and the contribution of the Se 4p states to the spectrum is negligibly small compared with that of the Co 3d states. Thus, the main peak at 1.0 eV is mainly due to the Co 3d states, and the two broad structures around 3 and 6.5 eV are mainly attributed to the Se 4p bands. The valence bands extend over the top 8 eV below EF. Above hn 50 eV, the feature of the photoemission
H. Sato et al. / Solid State Communications 118 (2001) 563±567
Fig. 3. CIS spectra of CoSe2 for selected valence-band initial states, which are speci®ed by their initial state energies Ei.
spectra is almost unchanged. Broad structures in the photoemission spectra at hn 60±67 eV denoted by vertical arrows in Fig. 2 move to the higher binding-energy side with hn , and are attributed to the emission of the Co MVV Auger electrons. In order to investigate the resonance behavior of the Co 3d emission in the Co 3p±3d excitation region including the
Fig. 4. Co-pDOS of CoSe2 derived by the band-structure calculation [20] (solid line in the lower part). A dashed line presents the occupied pDOS convoluted with the Lorentzian and Gaussian functions for the life time and instrumental resolution, respectively. The experimental spectrum measured at hn 50 eV is also shown by dots. In the upper part, the corrected spectrum of the theoretical Co-pDOS with the v -dependent self-energy with g 10 and r 3 is drawn by a solid line (see text).
565
Se 3d absorption region in detail, we have measured the constant initial states (CIS) spectra for selected initial states of CoSe2. The CIS spectra for Ei 1.0, 2.8 and 6.4 eV are shown in Fig. 3, where Ei represents the binding energy of the initial states. The CIS spectra show the normalized photoemission intensities of the speci®ed initial states as a function of hn . The CIS spectra for Ei 1.0 and 2.8 eV slightly exhibit a resonance enhancement in the Co 3p±3d excitation region (hn 58~64 eV), indicating an existence of the localized component of the unoccupied 3d states. It is noticed that the line shape is close to the Fano type [15] since the resonance shape has the small dip and is asymmetric in the excitation region. The resonance enhancement is weak [16] in comparison with, for instance, the manganese chalcogenides [17,18]. In the CIS spectrum for Ei 6.4 eV, one notices a broad peak around 61 eV. This peak is mainly dominated by the Co MVV Auger emission as also seen from Fig. 2. The gradual decrease in intensity from hn 50 to 70 eV, which is well recognized in the CIS spectrum for Ei 1.0 eV, is due to the decrease of the photoionization cross section of the Co 3d states [14]. In the CIS spectrum for Ei 1.0 eV, one also notices a weak structure around the Se 3d core-absorption region around 55 eV. This implies that the Se 4p states are strongly hybridized with the Co 3d states and therefore they also contribute to the main peak at 1.0 eV in the photoemission spectra. The broad structure due to the Se MVV Auger electrons is observed around 55 eV in the CIS spectrum for Ei 6.4 eV. The whole feature of the photoemission spectrum at hn 50 eV in Fig. 2 is quite similar to that of CoS2 measured using a He discharge lamp (He II:hn 40.8 eV), reported by Mamiya et al. [19]. In the spectrum of CoS2, a main peak appears at 1.0 eV together with the shoulder on the lower binding-energy side. In CoS2, a Co atom is octahedrally surrounded by six covalently bonded S±S dimers. The Co 3d states split into the triplet t2g and the doublet eg states under the crystal ®eld. The Co 3d electrons are in the low-spin state and the nominal con®guration is 6 1 t2g eg . In comparison with their result, the main peak at 1.0 eV in the photoemission spectrum of CoSe2 is attributed to the fully occupied t2g bands, while the shoulder to partially ®lled eg bands. The electronic states in CoSe2 are similar to those in CoS2, where the electron correlation is important [8,9] though CoS2 is known to be an itinerant compound in general. To our knowledge, only one band-structure calculation for paramagnetic CoSe2 is reported by Yamada et al. [20]. The calculations have been carried out by means of the linear muf®n-tin orbital (LMTO) method within the atomic sphere approximation (ASA). They compare the theoretical densities of states (DOS's) of CoSe2 with that of CoS2 in the paramagnetic phase in the region from 23.4 to 4.1 eV relative to EF in order to discuss the stable magnetic structure of these materials. Fig. 4 shows their results together with the photoemission spectrum measured at hn 50 eV. A solid line in the lower part represents the theoretical Co
566
H. Sato et al. / Solid State Communications 118 (2001) 563±567
partial DOS (Co-pDOS) of CoSe2. To compare the theoretical Co-pDOS with the photoemission result, we have convoluted the theoretical occupied DOS with the Lorentzian function with the binding-energy (EB)-dependent FWHM of 2G 0.15EB 1 0.05 eV and the Gaussian function with the FWHM of 0.3 eV, to represent the life time broadening and the instrumental resolution, respectively. The convoluted result is also shown by a dashed line in the lower part of the ®gure. Here we neglect the Se-pDOS because the photoemission spectrum at hn 50 eV re¯ects almost only the Co 3d states as mentioned above. The theoretical DOS at EF is mainly composed of the eg components of the Co 3d states strongly hybridized with the antibonding Se 4p states and the peak at ,2 eV is due to the t2g bands. On the basis of the band-structure calculation, the main peak and the shoulder in the experimental spectrum are again assigned to be the t2g and eg bands, respectively. Although the basic feature of the spectrum is in qualitative agreement with the convoluted curve with respect to two structures near EF and 1.0 eV, the energy position of the t2g bands of the theoretical Co-pDOS is markedly higher than that obtained from the experimental result. The band-structure calculation by means of the LMTO method with ASA is generally considered to be insuf®cient in accuracy compared with those by means of, for example, the Full-Potential Augment Plane Wave (FLAPW) method. As one of the reasons of the disagreement of the energy positions of the t2g bands between the theoretical and experimental DOS's, the accuracy of the calculations in Fig. 4 is suggested. The calculations with the FLAPW method have not been performed for CoS2 and CoSe2. However, the energy position of the t2g bands in the theoretical DOS of paramagnetic CoS2 derived from LMTO-ASA is around 1.7 eV [20], while its energy position is around 1.75 eV in the theoretical DOS with the Korringa±Kohn±Rostoker method [21]. In the ferromagnetic phase of CoS2, the calculated energy positions of the majority and minority t2g bands with the LMTO-ASA are around 2.0 and 1.2 eV, respectively, and they are consistent with those derived from the self-consistent linear combination of atomic orbital (2.0 and 1.2 eV) [22]. The disagreement of the energy position of the t2g bands between the experimental and theoretical Co-pDOS's suggests that the electron correlation is important for CoSe2. Since the t2g bands are located at 2 eV in the theoretical DOS, they have a localized character more or less. The similarity of the photoemission spectra of CoSe2 and CoS2 also suggests an importance of the correlation in CoSe2. A self-energy correction to the theoretical DOS [23,24] to take into account the correlation, leads to the energy shift of the 3d bands toward EF side and gives rise to a broad satellite structure on the higher binding-energy side [23]. Here we have carried out the self-energy correction to the theoretical Co-pDOS of CoSe2 [20]. In the calculation, we have assumed the energy (v )-dependent
self-energy to be represented by a simple analytical function P of
v gv=
v 1 ir 2 [23,24], where g and r are adjustable parameters. The corrected spectrum with the parameters of g 10 and r 3 is also shown by a solid line in the upper part of Fig. 4. One notices that the main peak shifts toward EF side and its energy position is 1.0 eV in the corrected spectrum. In addition, a tail to the higher binding-energy side of the main peak well reproduces the experimental spectrum. The dip between the peak near EF and the main peak at 1.0 eV is, however, not seen in the experimental spectrum. One of the reasons would be due to the contribution of secondary electrons from the peak near EF to this binding-energy region. Muro et al. [25] reported the Co 3p±3d high-resolution resonant photoemission spectrum of CoS2, where the peak structure near EF shows up and the dip is also observed. High-resolution spectroscopy on CoSe2 is expected to separate the main peak and the shoulder on the lower binding-energy side, and the shoulder would be observed as a peak. From the parameters of g 10 and r 3 a mass enhancement factor (m p/mb)s is estimated to be ,2.1, where m p and mb are the enhanced effective mass and the bare band mass of a conduction electron at EF, respectively, and a subscript of `s' means that this value is derived from the g- and rvalues of the v -dependent self-energy function. On the other hand, (m p/mb)t, is estimated from g exp/g b, where a subscript of `t' denotes `transport' and g exp and g b are the electron speci®c heat coef®cients derived from the experiments and the band-structure calculations, respectively. From the theoretical DOS of CoSe2 [20], the g b-value is estimated to be ,4.3 mJ mol 21 K 22, while Ogawa and Nishihara reported the g exp-value of CoSe2 to be 11.0 mJ mol 21 K 22 from the speci®c heat measurements [26]. These values provide (m p/mb)t , 2.6, which is consistent with the (m p/mb)s-value. This consistency suggests that the present self-energy correction to the band theory is appropriate for CoSe2. The band theory predicts that the center of gravity of the t2g bands shifts to the higher binding-energy side by about 0.3 eV from CoS2 to CoSe2 [20]. In comparison with the photoemission spectra of CoS2 [19,25], we have not observed the energy shift of the t2g bands. We have also carried out the self-energy correction for the theoretical Co-pDOS of CoS2 in the paramagnetic phase [20]. The corrected spectrum with the parameters of g 7 and r 3 reproduces that of CoSe2 in Fig. 4. These values provide the (m p/mb)s , 1.8. On the other hand, (m p/mb)t is estimated to be ,1.7 using g exp 9.5 mJ mol 21 K 22 [26] and g b , 5.5 mJ mol 21 K 22 [20]. Also for CoS2, the (m p/mb)s-value is consistent with the (m p/mb)t-value. The increase of (m p/ mb)s or (m p/mb)t-values from CoS2 to CoSe2 suggests that the electron correlation effect is more important for CoSe2 than for CoS2. In order to investigate the electron correlation effect in CoS2 and CoSe2, high-resolution photoemission experiments on these compounds with the just same experimental
H. Sato et al. / Solid State Communications 118 (2001) 563±567
condition are strongly required. They will reveal whether the energy position of the t2g bands shifts between CoS2 and CoSe2. In addition, the parameters of g and r are also expected be determined precisely by comparing the emission intensities of the peak near EF. They would give a clue how the chalcogen elements in the cobalt dichalcogenides affect the p±d hybridization and electron correlation. In summary, the photoemission spectra of CoSe2 are quite similar to those of CoS2 with respect to the basic features and the energy positions of the structure. The valence-band DOS spreads over the top 8 eV region below EF. The main peak at 1.0 eV is assigned to the fully occupied Co t2g bands, while the shoulder on the lower binding-energy side of the main peak to the partially ®lled Co eg bands. The similarity of the photoemission spectra between CoSe2 and CoS2 as well as the disagreement between the photoemission spectrum and the theoretical DOS [20], suggest that electron correlation effects are important for CoSe2, which is also suggested from the large g exp-value. The spectrum with the self-energy correction to the theoretical DOS of CoSe2 [20] reproduces the experimental spectrum with the parameters g 10 and r 3 of the v -dependent self-energy function. The (m p/mb)s-value derived from g- and r-values is qualitatively consistent with the (m p/mb)t-value, implying that the self-energy correction to the band theory is essential for CoSe2. The larger (m p/mb)s- or (m p/mb)t-values of CoSe2 suggest that the electron correlation effects are stronger for CoSe2 than for CoS2. Acknowledgements The authors are grateful to Prof. A. Kakizaki and Prof. Y. Azuma of Photon Factory for their technical supports, to Dr T. Maruyama for his help for the photoemission spectrometer, to Dr M. Usui, Dr K. Shimada and Dr M. Arita for stimulating discussion, to K. Murakami and T. Noda for the sample preparation and to Prof. M. Koyama for some advices for the sample growth. They would like to thank the referee for valuable advices for the self-energy correction to the band theory. This work is supported by the Grant-in Aid for Scienti®c Research from the Ministry of Education, Science and Culture, Japan. References [1] J.A. Wilson, Adv. Phys. 21 (1972) 143. [2] S. Waki, N. Kasai, S. Ogawa, Solid State Commun. 41 (1982) 835.
567
[3] N. Inoue, H. Yasuoka, Solid State Commun. 30 (1979) 341. [4] K. Adachi, K. Sato, M. Takeda, J. Phys. Soc. Jpn. 26 (1969) 631. [5] K. Adachi, M. Matsui, M. Kawai, J. Phys. Soc. Jpn. 46 (1979) 1474. [6] K. Adachi, M. Matsui, Y. Omata, H. Mollymoto, M. Motokawa, M. Date, J. Phys. Soc. Jpn. 47 (1979) 675. [7] G. Krill, P. Panissod, M. Lahrichi, M.F. Lapierre-Ravet, J. Phys. C 12 (1979) 4269. [8] A. Fujimori, K. Mamiya, T. Mizokawa, T. Miyadai, T. Sekiguchi, H. Takahashi, N. Mori, S. Suga, Phys. Rev. B 54 (1996) 16329. [9] A.E. Bocquet, K. Mamiya, T. Mizokawa, A. Fujimori, T. Miyadai, H. Takahashi, M. Mori, S. Suga, J. Phys.: Condens. Matter 8 (1996) 2189. [10] G. Krill, A. Amamou, J. Phys. Chem. Solids 41 (1980) 531. [11] H. van der Heide, R. Hemmel, C.F. van Bruggen, C. Haas, J. Solid State Chem. 33 (1980) 17. [12] M. Cardona, W. Gudat, B. Sonntag, P. Yu, in: Proceedings of the 10th International Conference on the Physics of Semiconductors, 1970, p. 209. [13] I. Ono, P.C. Grekos, T. Kouchi, M. Nakatake, M. Tamura, S. Hosokawa, H. Namatame, M. Taniguchi, J. Phys.: Condens. Matter 8 (1996) 7249. [14] J.J. Yeh, I. Lindau, At. Data Nucl. Data Tables 32 (1985) 1. [15] U. Fano, Phys. Rev. 124 (1961) 1866. [16] The peak energies of the CIS spectra for Ei 1.0 and 2.8 eV are different and they are 62 and 64 eV, respectively. It might be due to the different transition probability to 3p 63d 6 ®nal states depending on the 3p 53d 7 intermediate states. [17] Y. Ueda, H. Sato, M. Taniguchi, N. Happo, T. Mihara, H. Namatame, T. Mizokawa, A. Fujimori, J. Phys.: Condens. Matter 6 (1994) 8607. [18] H. Sato, T. Mihara, A. Furuta, M. Tamura, K. Mimura, N. Happo, M. Taniguchi, Y. Ueda, Phys. Rev. B 56 (1997) 7222. [19] K. Mamiya, T. Mizokawa, A. Fujimori, H. Takahashi, N. Mori, T. Miyadai, S. Suga, N. Chandrasekharan, S.R. Krishnakumar, D.D. Sarma, Physica B 237±238 (1997) 390. [20] H. Yamada, K. Terao, M. Aoki, J. Mag. Mag. Mat. 177±181 (1998) 607. [21] S. Asano, unpublished results quoted by S. Suga et al. [J. Phys. Soc. Jpn. 52 (1983) 1848]. [22] G.L. Zhao, J. Callaway, M. Hayashibara, Phys. Rev. B 48 (1993) 15781. [23] K. Shimada, Ph. D. Thesis, Tokyo University, 1996. [24] T. Saitoh, A. Sekiyama, T. Mizokawa, A. Fujimori, K. Ito, H. Nakamura, M. Shiga, Solid State Commun. 95 (1995) 307. [25] T. Muro, A. Kimura, T. Iwasaki, S. Ueda, S. Imada, T. Matsushita, A. Sekiyama, T. Susaki, K. Mamiya, T. Harada, T. Kanomata, S. Suga, J. Elec. Spec. Relat. Phenom. 88±91 (1998) 361. [26] S. Ogawa, Y. Nishihara, J. Phys. Soc. Jpn. 42 (1977) 343.