Ultraviolet inverse photoemission from FeS2

Ultraviolet inverse photoemission from FeS2

Solid State Communications, Vol. 82, No. 6, pp. 489-491, 1992. Printed in Great Britain. 0038-1098/92 $5.00 + .00 Pergamon Press Ltd ULTRAVIOLET INV...

301KB Sizes 0 Downloads 67 Views

Solid State Communications, Vol. 82, No. 6, pp. 489-491, 1992. Printed in Great Britain.

0038-1098/92 $5.00 + .00 Pergamon Press Ltd

ULTRAVIOLET INVERSE PHOTOEMISSION FROM FeS2 E. Puppin, M. Finazzi and F. Ciccacci Istituto di Fisica, Politecnico di Milano, Piazza Leonardo da Vinci 32, 20133 Milano, Italy (Received 4 D e c e m b e r 1991 by E. Molinari)

We present a set of angle integrated inverse photoemission spectra from FeS2 taken in the ultraviolet region (12-25 eV) in the isochromat mode; this is the first application of photon energy dependent isochromat inverse photoemission to a transition metal sulfide. It will be shown that the spectra can be simply interpreted in terms of the available singleparticle ground state calculations. The data will also be discussed in the light of present theories on correlated systems. THE ELECTRON states of transition metal (TM) oxides and sulfides have been the matter of a longstanding controversy which gained a renewed interest in recent years also due to its connections with high Tc superconductivity [1]. Such systems are now being investigated by new and powerful spectroscopic tools, among which isochromat inverse photoemission (IP) plays a fundamental role. This spectroscopy consists in scanning the energy of the monochromatic electron beam sent onto the sample by detecting, at the same time, the outcoming photons at a fixed value of hv [2]. The most simple way to interpret the isochromat spectra is to consider them as replicas of the density of states (DOS) above Er. This naive model is adequate for a qualitative discussion of the isochromat IP spectra of elemental TM and alloys. In systems characterized by a strong contribution of correlation to the cohesive energy, as in many TM oxides and sulfides, other features may occur in the spectra which have no counterpart in the DOS. These structures are determined by the electron removal (direct photoemission) or addition (IP) effects as well as the anion-to-cation charge transfer mechanisms. For instance, in the IP spectra from a TM compound having a 3d n electronic configuration in the ground state, a satellite peak having 3dn+2L-~ character may occur at higher energy above EF with respect to the main 3d "+1 peak. In the last notation, L - ~indicates a ligand hole promoted via anion-to-cation charge transfer. Such a peak has been observed in IP from NiO [3, 4]. The photon energy dependence of the isochromat spectra can give valuable information on the orbital character of the empty electron states. In fact, it is well known that the photoionization cross-section (o) increases between 10 and 25 eV for 3d states, whereas it decreases for p states [5]. The corresponding spectral features, therefore, show a photon energy dependence

which represents a strong fingerprint of their orbital orion. A characteristic hv dependence is also expected for satellite peaks, if they are present. More precisely, the intensity of the addition peak connected with the final state configuration d"+2L - 1 is expected to be the same as the d "+1 peak. Transition metal sulfides constitute another relevant class of correlated systems. Among them iron pyrite (FeS:) has a particular interest due to its potential application as a solar-cell material [6]. FeS~ is a diamagnetic semiconductor with cubic structure having a gap in the order of 0.9 eV. The electron states of this material are rather well understood from a theoretical point of view in the framework of single particle band structure calculations. An experimental confirmation of this picture came from Folkerts et al. who measured direct and inverse photoemission spectra at hv = 1486.7 [7]. Photon energy dependent isochromat IP spectroscopy has been applied to the investigation of Fe oxides [8] and FePS3 [9]. Here we present for the first time IP data on a TM sulfide; the photon energy range covered is in the UV region, namely between 12 and 25eV. Our choice fell on FeS2 because the basic features of its electron states are well known and therefore it is better suited for this preliminary investigation whose major goal is to assess the possibilities and the limitations of !P in the case of sulfides as compared to the most largely investigated family of TM oxides. The grating ultraviolet inverse photoemission apparatus is described in more detail elsewhere [10]. The emitted photons are dispersed by a grating on a flat field where a microchannel plate detector is located. In this way it is possible to measure at the same time up to 12 isochromat spectra in the photon energy range 10-25 eV.

489

490

Vol. 82, No. 6

ULTRAVIOLET INVERSE PHOTOEMISSION FROM FeS2 I

I

I

I

I

I

Fe S 2

I

:%C'-.

~-

-

i

hv (eV) 21.8

i " ~ / ~

19.5

.

15.o

"

,,,,~

J." :jf J..'4k.Z Ec 4

8

12.1

i

i

i

i

i

~

i

I

I

I

hv = 14.0 eV ~ B

I

I

Ec 4

I

I

8

I

12 16

Energy above E c (eV) 12

Energy above Ec (eV)

Fig. 1. Inverse photoemission spectra from FeS2 taken in the isochromat mode at different photon energies hr. The energy scale is referred to the bottom of the conduction band Ec. The spectra have been normalized to the same height of the main peak. The natural iron pyrite single crystal sample has been scraped in Ultra-High-Vacuum (base pressure in the low 10-~° Torr range) with a diamond file. Due to the characteristics of the electron source, a thoriated tungsten filament, the spectra are averaged in the reciprocal lattice, i.e. they are k-integrated. In order to avoid any damage of the sample due to the electron bombardment, the electron current has been kept below 1 #Amm -2. A stack of selected spectra is shown in Fig. 1. The spectra have been normalized to the same amplitude since the following discussion will be exclusively based on the photon energy dependence of the spectral shape. The zero of the energy scale is represented by Ec, the bottom of the semiconductor conduction band. In order to discuss the evolution of the spectral shape it is convenient to refer first to a single isochromat. This is done in Fig. 2 where the isochromat at 14.0 eV is compared with the partial density of the Fe 3d-, S 3p- and S 3d-derived empty states, as taken from [7]. Three spectral features (labelled A, B and C, respectively) are dearly observed in the spectrum of Fig. 2, and their energy position is in good correspondence with the structures of the partial DOS. This comparison confirms the spectral assignment for features A and B done in [7] in the isochromat at 1487.6eV, which is basically the same as the one in Fig. 2. A consideration of the photon energy dependence of the spectral shape gives further support to the above assignment. As noted above, we refer to the relative amplitudes of the spectral features. Peak B has

Fig. 2. Comparison between the inverse photoemission spectrum from FeS2 at hv = 14.0eV (dots, top curve) and the calculated projected density of empty states (after [7]). a decreasing intensity, at increasing hv, if compared to feature A. This fact represents a strong confirmation of the p-character of feature B and of the d-character of feature A. In fact, the photoionization cross section of the sulphur 3p-derived states has a decreasing trend between 10 and 25 eV whereas the Fe 3d cross section increases in the same energy range. The energy dependence of the photoionization cross section has been calculated from the isolated atoms by Yeh and Lindau [5]. The atomic values can be used in a qualitative discussion of the photon energy dependence of a also in the solid state since the increasing or decreasing trend is not modified by the so called "solid state effects", consisting in the modifications introduced in the atomic wavefuctions by the formation of the solid [11]. The same arguments have already been applied in discussing the electron states of a related system, FePS3, and also in that case the isochromat spectra show two distinct features whose relative intensity versus hv follows the same trend as in Fig. 1 [9]. Regarding feature C the most simple interpretation, based on the energy correspondence observed in Fig. 2, would assign it S 3d character. Another argument in favour of this assignment is the hv dependence of the intensity of this structure. Feature C, in fact, roughly maintains the same relative intensity with respect to feature A, which, as stated above, has Fe 3d character. The atomic value of a is not available for S 3d states. However, it is reasonable to assume that it increases between 10 and 25 eV since this trend is followed by all the filled 3d orbitals, from Sc to Zn [5]. The hv dependence of the spectral shape therefore excludes a relevant p character of feature C. From the above analysis, it comes out a picture of isochromat IP in TM sulfides quite different from that found in TM oxides. In the last case, in a one-electron

yol. 82,,No. 6

ULTRAVIOLET INVERSE PHOTOEMISSION FROM FeS2

framework no strong d-character is expected in the spectra up to several eV above Ee since the electron states have a dominant sp nature. For this reason they can be easily identified in the spectra by changing the isochromat photon energy. Actually, at increasing hv their weight decreases and eventually, above hv = 20 eV, they cannot be observed any more. This situation is clearly observed also in Fig. 1 where, at the highest photon energy, feature B (originated from S 3p states) has disappeared. In TM oxides at high photon energies, then, we have an energy window where only a small contribution is expected from oneelectron states. Therefore, the presence of additional peaks in this energy region having d-character would be easily identified. In fact, a satellite-like peak of this type would have a photon energy dependence which closely follows the one of the lowest energy d-related peak (3d n+l peak). In other words, a 3dn+2L -j satellite, if present, would always be visible in a stack like the one of Fig. 1. Such satellites are not expected to occur in isochromat IP spectra from FeS2. The argument in favour of this assumption [7] is that the 3d states of the Fe 2÷ ions are split by a strong octahedral crystal field so that the d electrons (n = 6) are accommodated in the t2s states while the completely empty eg states lie at higher energies. Therefore, FeS2 has no partially filled sub-bands so that the electron added in IP has no open-shell electrons to correlate with and the spectra can be directly compared with the DOS calculated using the ground-state charge density. In other sulfides the situation might be different: for instance NiS2 is a good candidate for the observation of satellitepeaks [7]. The energy position of satellites in isochromat spectra can give valuable information on the physics correlated systems. For instance, isochromat IP allowed to estimate the values of U, the Hubbard correlation energy, and of A, the charge transfer energy, in NiO [3, 4]. Actually, isochromat IP is the only experimental technique presently available which allows the direct measurement of these quantities. Therefore, it would be extremely important to extend IP to other systems, such as, for instance, TM sulfides. However, in light of the above discussion in the presence of a strong and broad structure having S 3d character, looks like a very general limitation exists in applying isochromat IP to the investigation of TM sulfides. In fact, the presence of a structure like feature C would probably hide possible satellite peaks and the photon energy dependence would not be useful due to the same trend of the cross section for S 3d and satellites. A mor e systematic investigation of TM sulfides must be done in order to verify if this is a general rule

491

or not. A word of caution, however, must be spent at this point. In this connection we mention that in IP from FePS3 the isochromat spectra measured with the same apparatus used for the present investigation [9] do not show any evidence of spectral features in the region of feature C. This is rather surprising since in FePS3 and in other TM phosphorus trisulfides the phosphorus electron states are very lightly influenced by the nature of the metal [12] and a behavior quite similar to FeS2 is therefore expected. In summary, we have presented isochromat IP data from FeS2 taken in the isochromat mode at different photon energies in the ultraviolet. The spectra have been discussed in terms of the modem theories of TM compounds. It is found that the spectra can be directly interpreted by comparison with the DOS. The photon energy dependence of the isochromat spectra allows to unambiguously identify the orbital character of the empty electron states. We have also shown that for sulfides IP cannot be as powerful as it is for oxides in distinguishing satellite-like addition peaks due to the presence of S 3d-derived states. Acknowledgements - The authors thank L. Braicovich for fruitful discussions. This work has been supported by the Ministero dell'Universit~ e della Ricerca Scientifica e Tecnologica through the Consorzio INFM. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8.

9. 10. 11. 12.

J. Zaanen, G.A. Sawatzky & J.W. Allen, Phys. Rev. Lett. 55, 606 (1985). V. Dose, Appl. Phys. 14, 117 (1977). S. Hufner, J. Osterwalder, T. Riesterer & F. HuUinger, Solid State Commun. 52, 793 (1984). G.A. Sawatzky & J.W. Allen, Phys. Rev. Lett. 53, 2339 (1984). J.J. Yeh & I. Lindau, At. Data Nucl. Data Tables 32, 1 (1985). A. Ennaui, S. Fiecther, W. Jaegermann & H. Tributsch, J. Elettrochem. Soc. 133, 97 (1986). W. Folkerts, G.A. Sawatzky, C. Haast, R.A. de Groot & F.U. Hillebrecht, J. Phys. C20, 4135 (1987). M. Sancrotti, F. Ciccacci, M. Finazzi, E. Vescovo & S.F. Alvarado, Z. Phys. !184, 243 (1991); F. Ciccacci, E. Puppin, L. Braicovich & E. Vescovo, Phys. Rev. !!44, 10444 (1991). E. Puppin, M. Scagiioti & C. Chemelli, Solid State Commun. 78, 905 (1991). M. Sancrotti, L. Braicovich, C. Chemelli, F. Ciccacci, E. Puppin, G. Trezzi & E. Vescovo, Rev. Sci. Instrum. 62, 639 (1991). G. Rossi, I. Lindau, L. Braicovich & I. Abbati, Phys. Rev. B28, 3031 (1983). Y. Ohno & S. Nakai, J. Phys. Soc. Jpn. 54, 258 (1986).