Electronic structure in thin film organic semiconductors studied using soft X-ray emission and resonant inelastic X-ray scattering

Thin Solid Films 515 (2006) 394 – 400 www.elsevier.com/locate/tsf

Electronic structure in thin film organic semiconductors studied using soft X-ray emission and resonant inelastic X-ray scattering Yufeng Zhang a, James E. Downes b, Shancai Wang a, Timothy Learmonth a, Lukasz Plucinski a, A.Y. Matsuura a,c, Cormac McGuinness a, Per-Anders Glans a, Sarah Bernardis a, Cian O’Donnell a, Kevin E. Smith a,* a

c

Department of Physics, Boston University, 590 Commonwealth Avenue, Boston, MA 02215, USA b School of Chemical and Physical Sciences, Victoria University of Wellington, Wellington, NZ U.S. Air Force Office of Scientific Research, 4015 Wilson Blvd., Arlington, VA 22203, United States Available online 19 January 2006

Abstract The electronic structure of thin films of the organic semiconductors copper and vanadyl (VO) phthalocyanine (Pc) has been measured using resonant soft X-ray emission spectroscopy and resonant inelastic X-ray scattering. For Cu – Pc we report the observation of two discrete states near E F. This differs from published photoemission results, but is in excellent agreement with density functional calculations. For VO – Pc, the vanadyl species is shown to be highly localized. Both dipole forbidden V 3d to V 3d*, and O 2p to V 3d* charge transfer transitions are observed, and explained in a local molecular orbital model. D 2005 Elsevier B.V. All rights reserved. PACS: 78.70.En; 71.20.Rv Keywords: Organic semiconductors; X-ray emission

1. Introduction Organic semiconductors are the subject of intense study due to the challenge they pose to our understanding of the physical properties of complex solids and due to technological interest in developing carbon-based electronic devices [1,2]. Accurate determination of the electronic structure of thin film organic semiconductors is a prerequisite to developing a comprehensive understanding of these electronic materials. Phthalocyanines (Pc) form an important class of organic semiconductors due to the ease with which a wide variety of cations can bond to the phthalocyanine ligand. The ligand has a complex electronic structure in itself, and the introduction of metal cations leads to a combination of both localized and delocalized states near the Fermi level (E F). It has recently been shown that synchrotron radiation-excited resonant soft X-ray emission (SXE) spectroscopy is a very powerful tool for

* Corresponding author. Fax: +1 617 353 9393. E-mail address: [email protected] (K.E. Smith). 0040-6090/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2005.12.221

studying electronic structure in such materials [3]. SXE has a distinct advantage over photoemission spectroscopy in that resonant SXE is a non-ionizing spectroscopy, and directly measures the element specific partial density of states (PDOS) [4– 6]. Furthermore, as will be discussed below, electronic excitations from occupied valence states to unoccupied conduction band states can be measured in the variant of the emission experiment known as resonant inelastic X-ray scattering (RIXS) [7 –9]. We have used synchrotron radiation-excited SXE and RIXS to study the electronic structure of thin films of Cu – Pc and VO – Pc. Our results are in excellent agreement with theory [10], but differ significantly from previously published X-ray emission and photoemission results [11,12]. The films were discovered to be highly susceptible to synchrotron radiation beam damage. We successfully circumvented this effect by continuous translation of the films during measurement. This application of resonant SXE has important consequences for the determination of band gap energies in organic molecular crystal systems, since it allows determination of the nonionized electronic structure.

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

Thin films of Cu –Pc and VO –Pc were grown in an ultra high vacuum organic molecular beam deposition (OMBD) system (base pressure 2 T 10 9 Torr), and transferred under vacuum into the spectrometer system (with base pressure < 5 T 10 10 Torr). This is important if the intrinsic properties of CuPc are to be measured, since thin film samples can be strongly p-doped due to oxygen contamination [13]. Films were deposited on P:Si (100) substrates which were ultrasonically cleaned in acetone, then heated to 900 -C for 10 min in the OMBD chamber. 200 nm of Pc was deposited from a well outgassed source with the substrate at 150 -C. Film thickness was determined using a quartz microbalance. SXE spectroscopy involves the measurement of the photon emitted when a valence electron makes a radiative transition into a hole on a localized core level created by excitation of a core electron with monochromatic synchrotron radiation. Since strong dipole selection rules associated with the core level govern the transition, SXE directly measures the orbital angular momentum resolved partial density of states (PDOS) [6,14]. Furthermore, since the core level is associated with a specific element in a compound, the element specific PDOS is measured [6,14]. Additional information can be extracted from emission spectra if the incident monochromatic synchrotron radiation is tuned to the energy close to a core absorption threshold. In this resonant case, the excited electron resides in a conduction band or lowest unoccupied molecular orbital (LUMO) state, and the system does not become ionized. Two additional phenomena can be observed with resonant excitation. First, if the system exhibits core level shifts due to different chemical bonding or site symmetry, holes can be resonantly created on each core level and the PDOS associated with a particular bond or chemical environment can be measured. For example, in a system with C –C, C –N, and C – O bonds, the C 2p PDOS of each can be measured by sequentially creating holes on the respective C 1s level. This variant is referred to as resonant SXE (RSXE). The second phenomenon that can appear with resonant excitation are features in the emission spectrum associated with excitations near E F; i.e the incident photon, resonant with a core level, excites an electron hole pair around E F, and scatters from the system with an energy shifted to lower values by the characteristic energy of the valence excitation. This variant of SXE is known as resonant inelastic X-ray scattering (RIXS) [7– 9]. RIXS features are losses from the main excitation energy, and thus change their photon energy in step with changes in the incident photon energy. This is in contrast to PDOS features seen in SXE or RSXE, which are fixed at the photon energy associated with the transition into the core hole. This makes the identification of PDOS and RIXS features in SXE spectra straightforward. Note also that at resonance, the excited core electron effectively screens the empty core-hole state in large delocalized molecular systems [15]. This reduces the effect that the core-hole has on the energies of the valence states and reduces the strength of final state effects in SXE spectra in comparison to photoemission measurements.

The experiments were performed at undulator beamline X1B at the National Synchrotron Light Source, which is equipped with a spherical grating monochromator. SXE/RIXS measurements were performed with a grazing incidence grating spectrometer [14]. The energy resolution for SXE spectra presented here was approximately 0.3, 0.6 and 0.74 eV for C KO, N KO, and O KO emission, respectively. The energy resolution of the incident beam was set to 0.5 eV for RSXE spectra. Photoemission measurements were made using a 100 mm Scienta hemispherical analyzer. For soft X-ray absorption spectra (XAS), the resolution of the monochromator was set to 0.4 eV, and the spectra were obtained by recording the sample drain current (total electron yield). The incoming radiation was normal to the surface. Both SXE and XAS experiments were performed with the samples at room temperature. Density functional theory (DFT) calculations were performed for the electronic structure of VO –Pc. The calculations used general gradient approximation GGA:PW91 exchange-correlation functionals and a spin-unrestricted llelectron model. The calculated DOS presented here have been broadened to account for instrument resolution and core-hole lifetime effects to facilitate comparison to experiment. 3. Results and discussion 3.1. Cu – Pc High resolution SXE measurements require a small photon spot (circa 40 Am) with a high photon flux (1013 ph/s) on the sample, and long collection times (30 – 60 min). This combination can lead to significant beam induced damage in organic systems [16]. This problem has been solved by continuously translating the films in front of the beam (at 40 Am/s) as the spectra are being recorded. Fig. 1 shows the C Ka SXE spectrum reflecting the C 2p PDOS from both a stationary and continuously translated CuPc film [3]. The excitation energy was 294 eV, and is well above the absorption threshold for the C Kα SXE hν = 294.0 eV

Translated Stationary

Intensity

2. Experimental method

395

260

270

280

290

Emission Energy (eV) Fig. 1. Comparison of non-resonant C Ka SXE spectra reflecting the C 2p PDOS for both stationary and translated films of CuPc. The inset shows a schematic of the CuPc molecule.

396

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

C 1s states. The feature marked ‘‘a’’ at 294 eV visible in both spectra is elastically scattered light. Significant differences in the measured electronic structure of identical films when stationary or translated are visible in Fig. 1. We note that the spectrum obtained here for the damaged, stationary film is similar to that published by Kurmaev et al. for CuPc [12], and by Tegeler et al. for the related compounds PtPc and MgPc [11]. In particular, the feature marked ‘‘y’’ in our spectrum from the translated sample is missing from both the spectrum we obtained from a stationary film, and from the published spectrum [12]. The significance of this observation will be addressed below. The remainder of SXE spectra presented in this paper are from samples translated during data acquisition. The left panel of Fig. 2 shows a series of C Ka SXE spectra from CuPc recorded as a function of excitation energy. The right panel of Fig. 2 shows the results of a DFT calculation for the electronic structure of the valence band within approximately 3 eV of E F [10]. Also shown are the binding energies for the ring and pyrrole C 1s core levels as measured using photoemission from our translated films. To first approximation, there are two distinct chemical environments for C atoms in CuPc: C bonded to other C in the exterior rings (for example, atoms 1,2, 3 in the inset to Fig. 1) and C bonded to N in the pyrrole structure (for example, atom 4 in the inset of Fig. 1). The spectra in Fig. 2 show only the states near the top of the valence band and are presented with an expanded vertical scale; the spectrum in Fig. 1 corresponds to spectrum (viii) in the series of Fig. 2. Consider spectrum (i), excited with a photon energy of 284.6 eV. The feature marked ‘‘a’’ in the series is the elastic emission feature, and is manually truncated in the figure for clarity. At an excitation energy of 284.6 eV, only electrons from ring C atoms are excited into the conduction band states. Thus the emission spectrum reflects

valence electron transitions into this hole. DFT calculations indicate that the uppermost occupied states in CuPc (D4h point group) consist of a half filled HOMO state at E F of b1g symmetry, a filled state of a1u symmetry 0.5 eV below E F, and a manifold of states roughly 2 eV below E F marking the onset of the main valence band [10]. The feature marked ‘‘h’’ corresponds to transitions from the manifold of states at the top of the valence band into a hole on a ring C atom. As the excitation energy is increased to 285.9 eV (spectrum (iii)), two weak emission features now are visible between features ‘‘a’’ and ‘‘h’’. These features, labeled ‘‘g’’, are not RIXS loss structures since they do not move in energy as the excitation energy is increased to 286.3 eV, spectrum (iv). Although weak, these features are well above the noise, as can be seen in the inset to Fig. 2, which expands spectrum (iv). Based on comparison with the DFT calculations [10], we assign these features to transitions from the b1g and a1u states to a hole on a 1s level in a ring C. The higher energy feature is assigned to the half-filled b1g symmetry HOMO state at E F [3]. Note that emission from these states is not visible from damaged films. The doublet is not observed in the threshold spectrum (i), since the excitation and emission involve the same half filled b1g state, and thus the doublet lies in the elastic emission manifold. We note that the predicted doublet separation is 0.5 eV [10], and the observed separation is 0.8 eV. As the excitation energy is increased above the threshold for creation of C 1s holes on the pyrrole C atoms, we expect to observe the same features (valence band emission ‘‘h’’, and HOMO emission doublet ‘‘g’’) due to transitions to the pyrrole C 1s hole. However, since holes are still being created on the ring C 1s states, the emission features from transitions into both holes will overlap. Thus emission feature ‘‘y’’ corresponds to emission from the same pair of b1g and a1u states (‘‘g’’), but

-3

hνexcite 284

286

288

-2

284.6 eV

Intensity

285.9 eV 286.3 eV 286.9 eV 288.1 eV 290.3 eV 294.0 eV 280

285

290

Binding Energy (eV)

-1 285.3 eV

b1u

0 1 2 282 283 284 285 286

2eg b1g (E ) f a1u 1eg b2g a 2u b2u

ring C 1s pyrole C 1s

295

Photon Energy (eV) Fig. 2. Left panel: Series of SXE spectra from CuPc showing emission from states near E F. Right panel: Electronic structure of CuPc near E F as predicted by DFT calculations [10], with measured C 1s core level binding energies indicated. Note that the DFT calculation energies have been rigidly shifted such that the binding energy of the b1g state at E F is set to zero, allowing comparison to the measured XPS spectra and the excitation energies in the RSXE spectra. See text for details.

N SXE C SXE UPS

-5

510

515

520

525

530

535

Photon Energy (eV) Fig. 4. V L-edge and O K-edge XAS data from VO – Pc. Inset: schematic structure for the VO – Pc molecule. The arrows indicate the excitation energies used to excite the RIXS spectra presented in Fig. 6.

a1u. This is ascribed to the large difference in the degree of localization of the two states. If an electron is removed from a delocalized orbital it has less of an effect on the energy of that level than if the orbital is localized. Thus, strong final state effects modify the binding energy of the b1g state, as observed by UPS, and photoemission from this state overlaps in energy that from the a1u state producing a broad asymmetric photoemission feature that is identified in UPS

Intensity

shifted by 1.4 eV to higher photon energies because of the chemical shift of the pyrrole C sites, and emission feature ‘‘(’’ corresponds to ‘‘h’’, but again shifted by 1.4 eV. Further details can be found in the primary report [3]. SXE spectra were also recorded around the N Ka edge, reflecting the N 2p PDOS. Fig. 3 presents, on a common binding energy scale, the N Ka spectrum recorded just above threshold, the C Ka spectrum (iv) from Fig. 2, and a published ultraviolet photoemission spectroscopy (UPS) spectrum for CuPc [17]. Details of the alignment can be found elsewhere [3]. The elastic emission in the N Ka spectra is very weak. The low energy asymmetric emission feature ‘‘j’’ visible in the N Ka spectrum, is a RIXS energy loss feature. This feature lies approximately 1 eV below the excitation energy. Since the SXE spectra represent the C 2p and N 2p PDOS, the combination plotted in Fig. 3 represents the dominant fraction of the total density of states for CuPc, and can be compared to the UPS spectrum. Since the UPS spectrum represents the total DOS, it can be placed on a common binding energy scale with the SXE spectra by aligning the three prominent valence band features in the UPS spectrum with the equivalent features in the N and C 2p PDOS. The inset in Fig. 3 shows the detailed electronic structure close to E F, where the C Ka and UPS spectra are plotted. (The N Ka spectrum in this region is dominated by the RIXS loss feature,‘‘a’’, and is not plotted in the inset.) As is clear, the lowest energy feature in the UPS spectrum aligns exactly with the second lowest, a1u, feature in the SXE spectrum. i.e. the SXE spectrum reveals that the HOMO state at E F in CuPc lies 0.8 eV higher in energy than measured by photoemission spectroscopy. The difference lies in the fact that while UPS is an ionizing spectroscopy, RSXE is not. Liao and Scheiner note that ionization of the CuPc molecule first occurs from the ligand a1u state rather than the lower binding energy ligand-metal shared b1g state [10]. The ionization energy of the b1g state is calculated to be 0.7 eV higher than that of the

397

Absorption

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

0

Energy Relative to HOMO (eV)

515 -25

-20

-15

-10

-5

0

5

Energy Relative to HOMO (eV) Fig. 3. Comparison of resonant C and N KO spectra and UPS spectra from CuPc. The SXE excitation energy corresponds primarily to the C/N 1s 6 2eg resonance. The UPS spectrum is from Chasse et al. [17]. See text for details.

520

525

530

535

Photon Energy (eV) Fig. 5. O K-edge SXE spectra reflecting the occupied O 2p PDOS in VO – Pc, 1.5% Cr-doped V2O3 [5], and molecular O2 [23]. The nominal excitation energies are indicated. Elastic emission from the VO – Pc film is suppressed for the chosen experimental geometry. Also included are the results of a DFT calculation. See text for details.

398

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

a Excitation Energy 517.6 eV

Intensity

517.2 eV

516.8 eV

516.3 eV

515.6 eV

515 eV

514.5 eV -15

-10

-5

0

5

Energy Loss (eV)

b

Partial Density of States

Calculated V 3d and O 2p PDOS for VO-Pc

V 3d

˚ above the N-atom plane, with the oxygen atom sits 0.58 A isolated above V ion [19]. (A schematic molecular structure is presented in Fig. 4.) The V ion is in a 4+ state, with nominally ˚ from the O one electron in a V 3d state. The V ion is 1.58 A ˚ from the N atoms to which it is atom, but an average of 2.03 A also bonded [19]. In thin films of phase II VO – Pc, the closest ˚ and the average distance intermolecular distance is 3.308 A ˚ [19]. Thus to a between adjacent vanadium ions is 5.378 A good approximation VO – Pc presents a system with a well isolated vanadyl species. SXE spectra were recorded for the C 2p, O 2p, N 2p and V 3d valence band PDOS. A full analysis of the C 2p and N 2p PDOS and comparison to that obtained from CuPc [3], will be presented elsewhere. The discussion here focuses on measurement of the V and O states. Fig. 4 presents the XAS results for VO – Pc in the region of the V L-edge and O K-edge. As the incident photon energy is swept from 510 to 540 eV, absorption from spin-orbit split V 2p states and O 1s states is observed. The XAS spectrum in Fig. 4 is generally similar to that observed in other vanadium oxides, such as V6O13, V2O3, and NaV2O5 [20 –22]. However a clear difference in the V L-edge and O K-edge SXA for VO – Pc compared to the bulk oxides is the existence of two clear pre-peaks (at 514.5 and 515.6 eV), and significantly narrower absorption features, consistent with a more molecular environment for the vanadyl species. The O K-edge SXE spectrum reflecting the O 2p PDOS in VO – Pc is presented in Fig. 5. The spectrum is recorded with an excitation energy of 531.5 eV, above the absorption threshold. Fig. 5 also includes the results of our DFT calculation of the occupied O 2p PDOS for VO – Pc. The agreement between the measured and calculated O 2p PDOS is excellent. The calculation successfully reproduces both the measured O 2p band width, and the shape of the spectral

CT & 3d/O2p hybridization d-d & 3d PDOS 8 ----- DOS fluorescence 9

O 2p 7 -5

0

5

Binding Energy (eV) Fig. 6. a) Series of V L3-edge RIXS spectra from VO – Pc. The spectra are plotted on an energy loss scale relative to the elastic peak. The excitation energies used are indicated on the XAS spectrum in Fig. 4. b) Calculated PDOS for the V 3d and O 2p states.

spectra as the HOMO. The presence of the excited electron in the system in resonant SXE strongly screens the core-hole state and leaves the arrangement of states close to that of the neutral, ground state molecule. 3.2. VO – Pc

Loss Energy (eV)

-10

6 5 4 3 2 1 0 514.5 515.0 515.5 516.0

We have also performed a RSXE/RIXS study of the electronic structure of thin films of vanadyl phthalocyanine (VO – Pc) grown in situ [18]. The vanadium cation in VO – Pc

516.5 517.0 517.5 518.0

Photon Energy (eV) Fig. 7. Dispersion of the V L3 loss features as a function of excitation energy. The dashed lines indicate the expected behavior of simple fluorescent features.

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

distribution. Fig. 2 also presents the measured O 2p SXE spectra from a 1.5% Cr-doped V2O3 [4,5], and from gas phase O2 [23]. (The measured O 2p PDOS for a 1.5% Cr-doped V2O3 is identical to that of pure V2O3.) The O 2p PDOS in VO –Pc bears little resemblance to that from the crystalline 3d1 vanadate, but is remarkably similar to that obtained from gas phase molecular O2. The main part of the VO – Pc spectrum is quite symmetric, with a full width at half maximum (FWHM) of approximately 1.7 eV. This can be compared to the equally symmetric emission in molecular O2 with a FWHM of approximately 1.5 eV [23]. In contrast the main emission feature from Cr-doped V2O3 is highly asymmetric, and has a FWHM of approximately 3 eV. The similarity observed in the PDOS of O 2p states between molecular O2 and VO –Pc, and the difference observed between a typical 3d1 solid vanadate and VO – Pc, is consistent with the premise that the vanadyl species is spatially isolated in VO – Pc thin films, and is strong evidence for the lack of significant intermolecular interaction. We note that hybridization of O and V states of this vanadyl species is also visible in the O 2p spectra of Fig. 2. There is a discrete shoulder at lower binding energies (higher emission energies, approximately 527 eV) in the O K-edge emission from VO –Pc that is not observed in molecular O2. This emission is due to O 2p states hybridized with V 3d states near E F, and has not previously been directly observed in VO – Pc. Similar hybridization has been observed in other vanadium oxides, as is illustrated by the spectrum for Cr-doped V2O3 in Fig. 2 [4,5,20], and also appears in the calculated PDOS. SXE has been shown to be a very sensitive method of measuring hybridized states in both the valence band and in shallow core levels in many transition metal oxides [24 – 26]. Fig. 6a presents a series of RIXS/SXE V L3-edge spectra from VO – Pc as a function of incident photon energy, while Fig. 6b presents the calculated V 3d and O 2p PDOS [18]. The photon energies used to excite the spectra in Fig. 6a are marked on the XAS spectrum in Fig. 4. The spectra are plotted on a loss energy scale, relative to the elastic emission feature. Three discrete features are visible in the spectra. The elastic emission feature lies at 0 eV, and is observed to resonate as the excitation energy is tuned through the L3 V 2p3 / 2 edge. The remaining two features are centered at 2.2 eV (labeled ‘‘A’’) and 5.6 eV (labeled ‘‘B’’) in the spectrum recorded with the lowest exciting photon energy, 514.5 eV. As the excitation energy is increased, these two features disperse to higher loss energies. The energy of the centroid of the features is plotted as a function of excitation energy in Fig. 7. In principle, the origins of these spectral features are easily understood. At the absorption threshold (resonance), features ‘‘A’’ and ‘‘B’’ are identified as RIXS features, with feature ‘‘A’’ resulting from dipole forbidden V 3d –V 3d* excitations, and feature ‘‘B’’ resulting from O 2p – V 3d* charge transfer excitations. Comparison to the calculated V 3d PDOS presented in Fig. 6b reveals good agreement between the energy of the measured V 3d to 3d* loss feature and the energy of the occupied and unoccupied V 3d states in our DFT calculation. Similarly, there is good agreement between the energy of the charge

399

transfer loss and the calculated binding energies of the occupied O 2p and unoccupied V 3d* states (Fig. 6b). The RIXS features observed in VO – Pc have an unusual dependence on the excitation energy. Simple RIXS features should appear at a constant energy offset (loss) to the excitation energy. This clearly is not the case here (Figs. 6 and 7). However, neither do the features behave as simple fluorescence near the threshold (i.e appear at a constant emission energy). Obviously, well above the L3 absorption threshold (for example at an excitation energy of 517.6 eV), feature ‘‘A’’ should simply be fluorescence from the occupied V 3d states into the V 2p3 / 2 hole, and feature ‘‘B’’ should be the fluorescence from hybridized V 3d/O 2p states. It is possible that the apparent anomalous behavior of the features with excitation energy is a resolution effect, compounded by interference in the spectra between the RIXS process and simple fluorescence. Thus both features ‘‘A’’ and ‘‘B’’ can be considered as made up of two separate components: a RIXS component at constant loss energy that loses intensity as the excitation energy moves away from the absorption threshold, and a PDOS fluorescence component at constant emission energy that gains in intensity as the excitation energy is increased. Clearly a higher resolution study would be necessary to resolve these components. Acknowledgements The authors gratefully acknowledge the assistance of Dr. Steven Hulbert. This work was supported in part by the donors of the Petroleum Research Fund, administered by the ACS, by the DOE under DE-FG02-98ER45680 and by the NSF under DMR-0304960. The SXE spectrometer system was funded by the U.S. ARO under DAAD19-01-1-0364 and DAAH04-95-0014. The experiments at the NSLS are supported by the U.S. DOE, Division of Materials and Chemical Sciences. The authors thank Laurent Duda for numerous helpful discussions. References [1] S.R. Forrest, J. Quantum Electron. 6 (2000) 1072. [2] S.R. Forrest, Chem. Rev. 97 (1997) 1793. [3] J.E. Downes, C. McGuinness, P.-A. Glans, T. Learmonth, D. Fu, P. Sheridan, K.E. Smith, Chem. Phys. Lett. 390 (2004) 203. [4] K.E. Smith, C. McGuinness, J.E. Downes, P.J. Ryan, S.L. Hulbert, J.M. Honig, R.G. Egdell, Mater. Res. Soc. Symp. Proc. 755 (2003) DD1.1.1. [5] K.E. Smith, Solid State Sci. 4 (2002) 359. [6] J. Nordgren, N. Wassdahl, J. Electron Spectrosc. Relat. Phenom. 72 (1995) 273. [7] J. Nordgren, J. Phys. IV 7 (1997) 9. [8] J. Nordgren, P. Glans, K. Gunnelin, J. Guo, P. Skytt, C. Sathe, N. Wassdahl, Appl. Phys., A 65 (1997) 97. ˚ gren, Phys. Rev., B. 57 (1998) 2780. [9] F. Gel’mukhanov, H. A [10] M.-S. Liao, S. Scheiner, J. Chem. Phys. 114 (2001) 9780. [11] E. Tegeler, M. Iwan, E.-E. Koch, J. Electron Spectrosc. Relat. Phenom. 22 (1981) 297. [12] E.Z. Kurmaev, S.N. Shamin, V.R. Galakhov, A. Moewes, T. Otsuka, S. Koizume, K. Endo, H.E. Katz, M. Bach, M. Neumann, D.L. Ederer, M. Iwami, Phys. Rev., B 64 (2001) 045211. [13] Q. Zhou, R.D. Gould, Thin Sol. Films 317 (1998) 432.

400

Y. Zhang et al. / Thin Solid Films 515 (2006) 394 – 400

[14] J. Nordgren, G. Bray, S. Cramm, R. Nyholm, J.E. Rubensson, N. Wassdahl, Rev. Sci. Instrum. 60 (1989) 1690. ˚ gren, Y. Luo, F. Gelmukhanov, H. Jørgen, A. Jensen, J. Electron [15] H. A Spectrosc. Relat. Phenom. 82 (1996) 125. [16] N. Koch, D. Pop, R.L. Weber, N. Bowering, B. Winter, M. Wick, G. Leising, I.V. Hertel, W. Braun, Thin Solid Films 391 (2001) 81. [17] T. Chasse, C.-I. Wu, I.G. Hill, A. Kahn, J. Appl. Phys. 85 (1999) 6589. [18] Y. Zhang, S. Wang, T. Learmonth, L. Plucinski, A.Y. Matsuura, S. Bernardis, C. O’Donnell, J.E. Downes, K.E. Smith, Chem. Phys. Lett. 413 (2005) 95. [19] R.F. Ziolo, C.H. Griffiths, J.M. Troup, J. Chem. Soc. Dalton Trans. (1980) 2300. [20] T. Schmitt, L.C. Duda, A. Augustsson, J.H. Guo, J. Nordgren, J.E. Downes, C. McGuinness, K.E. Smith, G. Dhalenne, A. Revcolevschi, M. Klemm, S. Horn, Surf. Rev. Lett. 9 (2002) 1369.

[21] T. Schmitt, L.C. Duda, M. Matsubara, M. Mattesini, M. Klemm, A. Augustsson, J.H. Guo, T. Uozumi, S. Horn, R. Ahuja, A. Kotani, J. Nordgren, Phys. Rev., B 69 (2004). [22] T. Schmitt, L.C. Duda, M. Matsubara, A. Augustsson, F. Trif, J.H. Guo, L. Gridneva, T. Uozumi, A. Kotani, J. Nordgren, J. Alloys Compd. 362 (2004) 143. [23] P. Glans, K. Gunnelin, P. Skytt, J.H. Guo, N. Wassdahl, J. Nordgren, H. Aagren, F.K. Gel’mukhanov, T. Warwick, E. Rotenberg, Phys. Rev. Lett. 76 (1996) 2448. [24] C. McGuinness, C.B. Stagarescu, P.J. Ryan, J.E. Downes, D. Fu, K.E. Smith, Phys. Rev., B 68 (2003) 165104. [25] P.-A. Glans, T. Learmonth, C. McGuinness, K.E. Smith, J. Guo, A. Walsh, G.W. Watson, R.G. Egdell, Chem. Phys. Lett. 399 (2004) 98. [26] P.-A. Glans, T. Learmonth, K.E. Smith, J. Guo, A. Walsh, G. Watson, F. Terzi, R.G. Egdell, Phys. Rev., B 71 (2005) 235109.