Electronic state characterization of SiOx thin films prepared by evaporation

Electronic state characterization of SiO x thin films prepared by evaporation A. Barrancoa兲 Nanotech@surfaces Laboratory, Materials Science and Technology (Empa), Feuerwerkerstrasse 39, CH-3600 Thun, Switzerland and Physics Department, University of Fribourg, Pérolles, CH-1700 Fribourg, Switzerland F. Yubero and J. P. Espinós Institute of Materials Science of Seville [Consejo Superior de Investigaciones Cientificas (CSIC)–University of Seville], Avenida Américo Vespucio s/n. 41092 Sevilla, Spain P. Groening Nanotech@surfaces Laboratory, Materials Science and Technology (Empa), Feuerwerkerstrasse 39, CH-3600 Thun, Switzerland and Physics Department, University of Fribourg, Pérolles, CH-1700 Fribourg, Switzerland A. R. González-Elipe Institute of Materials Science of Seville [Consejo Superior de Investigaciones Cientificas (CSIC)–University of Seville], Avenida Américo Vespucio s/n. 41092 Sevilla, Spain 共Received 9 November 2004; accepted 6 April 2005; published online 8 June 2005兲 SiOxthin films with different stoichiometries from SiO1.3 to SiO1.8 have been prepared by evaporation of silicon monoxide in vacuum or under well-controlled partial pressures of oxygen 共P⬍10−6 Torr兲. These thin films have been characterized by x-ray photoemission and x-ray-absorption spectroscopies, this latter at the Si Kand L2,3 absorption edges. It has been found that the films prepared in vacuum consists of a mixture of Si3+ and Si+species that progressively convert into Si4+ as the partial pressure of oxygen during preparation increases. From this spectroscopic analysis, information has been gained about the energy distribution of both the full and empty states of, respectively, the valence and conduction bands of SiOxas a function of the O/Si ratio. The characterization of these films by reflection electron energy-loss spectroscopy 共REELS兲has provided further evidences about their electronic structure 共band gap and electronic states兲as a function of the oxygen content. The determination of the plasmon energies by REELS has also shown that the films prepared by evaporation in vacuum consist of a single phase which is characterized by a density 共1.7 g cm−3兲lower than that of SiO2共i.e., 2.2 g cm−3兲or Si 共i.e., 2.4 g cm−3兲. The optical properties 共nand k兲of the films as a function of the O/Si content have been deduced from the analysis of REELS spectra in the energy range from 4 to 20 eV. It has been also shown that the O/Si ratio in the films and several spectroscopic parameters such as the Auger parameter or the energy of bulk plasmons present a linear relationship and that this linear dependence can be used for a rapid characterization of SiOxmaterials. By contrast, the band-gap energy changes differently with the O/Si ratio, following a smooth linear increase from about 3.8 eV for SiO1.3 to ca. 5.0 eV for SiO1.7 and a jump up to 8.7 eV for SiO2. These results indicate that the random-bonding model does not apply to thin films prepared by evaporation under our experimental conditions. Other distributions of Sin+states can be induced if the films are excited with an external source such as heat or photon irradiation. In this case the electronic properties vary and the previous linear correlations as a function of the oxygen content do not hold any longer. ©2005 American Institute of Physics.关DOI: 10.1063/1.1927278兴 I. INTRODUCTION Nonstoichiometric silicon oxides 共SiOxwith x⬍2兲thin films have been widely studied during the last decades due to their numerous technological applications. The interest in these films began with their use as passivation layers, dielectric films, and interlayers in microelectronics.1,2 SiOxlayers have been also used as protective coatings in metallic mirrors, antireflection coatings, or low index layers in the midinfrared range.3–5 SiOxthin films are also being employed for the fabrication of light-emitting silicon-based structures.6,7 These latter consist of silicon nanocrystals embedded in a silica matrix and are obtained by disproportionation reaction of the Sin+共0⬍n⬍4兲species present in the SiOx. Annealing in vacuum or in an inert gas8–10 or exposure to a laser7,11 or an intense source of ultraviolet 共UV兲light12 have been used to induce such a reaction. Different models have been proposed to account for the distribution of Sin+species in SiOxthin films.13–15 These models were originally developed to account for the distribution of Sin+states at the interface between Si and a thermally grown silicon oxide overlayer, although they have also been used to describe the distribution of Sin+species in SiOx a兲Author to whom correspondence should be addressed; FAX: ⫹34954460665; electronic mail: [email protected] JOURNAL OF APPLIED PHYSICS 97, 113714 共2005兲 0021-8979/2005/97共11兲/113714/8/$22.50 © 2005 American Institute of Physics97, 113714-1 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp thin films. In the mixture model 共MM兲SiOxis described as a mixture of Si0and Si4+ species 共i.e., Si–Si4and SiO4tetra- hedral units兲.13,15 The more widely accepted model is the so-called random-bonding model 共RBM兲. It considers that the films are built up by several tetrahedral units Si–OxSi共4−x兲 共x=0,...,4兲that follow a simple statistical distribution.13–15 Although the structure and the electronic and optical properties of SiOxmaterials have been extensively discussed within the frame of this RBM, some controversy has arisen about its general applicability. This controversy stems from experimental evidences showing that in SiOxthin films with a given O/Si ratio, different distributions of oxidation states can be found according to the method of preparation.12,16,17 Thus, in previous works, we have found that SiOxthin films deposited at room temperature by evaporation of silicon monoxide do not follow the distribution of Sin+states predicted by the RBM.10,12,18 We also showed that these SiOx films were in a metastable state and that either by light excitation12 or ion-beam irradiation19 the distribution of Sin+ species may change while keeping constant the overall O/Si ratio. Most experimental procedures employed for the preparation of SiOxthin films are based on chemical or physical phenomena that, directly or indirectly, may induce some kind of excitation of the growing layers. This is the case of sputtering13 or plasma processes16,20,21 where ion bombardment, UV-light irradiation, or the supply of thermal energy might lead to a redistribution of oxidation states in the SiOx thin films and, therefore, to changes in their properties. Since the influence of these processes is not well controlled or even recognized, a comparison of the published results on SiOx prepared by different procedures is not always straightforward. As a continuation of our previous investigations on SiOx thin films prepared by evaporation of silicon monoxide we present here a multitechnique study of their electronic properties and on how these properties change with the O/Si ratio. The analysis is done by means of surface sensitive techniques, such as x-ray photoemission spectroscopy 共XPS兲, x-ray-absorption spectroscopy 共XAS兲, and reflection electron energy-loss spectroscopy 共REELS兲. A first objective of this investigation is to provide experimental correlations between the O/Si stoichiometry and electronic properties of well-characterized SiOxthin films prepared without external excitation. A second objective is to provide reference data to evidence possible changes in the electronic properties of these materials due to alternative distributions of Sin+states induced by an external excitation. II. EXPERIMENT Silicon oxide 共SiOxwith x⬍2兲thin films were prepared by thermal evaporation of silicon monoxide brown powder 共Goodfellow兲with a Knudsen cell. Full details about the deposition method and experimental conditions can be found elsewhere.12 The SiOxthin films were deposited at room temperature on flat one-side-polished Si共100兲wafers. The stoichiometry of the SiOxfilms was varied by evaporation of the silicon monoxide compound in increasingly higher partial pressures of oxygen from 10−8⬍P⬍10−6 Torr. In some cases the evaporated SiOxlayers were heated under ultrahigh-vacuum 共UHV兲conditions 共P⬍10−8 Torr兲at 750 K for 3 h; this treatment produces dissociation of the structure and the formation of Si0species. In the text, this sample rich in Si0will be labeled as SiOx关Si0兴. Chemical and electronic information of the topmost 共⬃5nm兲surface region of the obtained thin films was studied in situ after deposition. The films were characterized by electron spectroscopies such as XPS, photoemission spectroscopy 共PES兲with synchrotron radiation, and REELS. XPS spectra were acquired with a VG-ESCALAB 210 spectrometer. It was operated at a constant pass energy with a value of 20 eV. Unmonochromatized Mg K ␣ radiation was used as excitation source. The Si 2pand O 1sphotoemission peaks and the Si KLL and O KLL Auger peaks were recorded. The resolution was such that the width of the Si 2p of a clean silicon sample was 1.2 eV. For the characterization by REELS, primary beam energies between 200 and 2000 eV with an energy resolution of ⬃0.7 eV were used. The measurements were done in the XPS chamber by using a VG electron source 共LEG 6000兲. The angle of incidence was 60° with respect to the normal surface. The reflected electrons were analyzed in an angle normal to the surface. The nand kfunctions in an energy range from 4 to 20 eV and the absorption coefficients 共 ␣ 兲 were extracted from the REELS spectra following a procedure described in a previous paper.22 The optical band gaps were also estimated from the REELS spectra by considering the photon energy at which ␣ 艋104cm−1.22–24 Photoemission with synchrotron radiation 共PES兲was done at the KMC beam line of the BESSY synchrotron in Berlin. Si L2,3 XAS spectra were recorded in the same line by total electron yield measurements. XAS spectra at the Si Kedge were recorded at the SA32 beam line of the Super- Aco synchrotron 共LURE, Orsay兲. An InSb共111兲double- crystal monochromator was used for selection of the photon energy 共resolution of 0.75 eV at 2000 eV兲. The spectra were recorded by measuring the drain current through the sample 共total electron yield detection兲. III. RESULTS AND DISCUSSION A. Distribution of Si n +species in SiO x thin films: Stoichiometry and local structure around Si A set of SiOxthin films with thickness of the order of 100 nm were prepared by evaporation of silicon monoxide in the UHV chamber at increasing oxygen partial pressures between 10−8 and 10−6 Torr. The stoichiometry of the films was controlled by in situ XPS at several stages of the film preparation. In previous works we have reported about the distribution of Sin+states in this type of films deduced by XPS, PES, extended x-ray-absorption fine structure 共EXAFS兲,12 x-ray-Absorption Near-Edge Spectroscopy 共XANES兲, and resonant Auger spectroscopy.10 Thus, it was found that the structural data obtained by EXAFS assuming the presence of Si3+ and Si+species were in agreement with the fitting analysis of Si 2pphotoemission spectra. These results and the calculation procedures will not be discussed 113714-2 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp any longer here but will be directly used to ascertain the distribution of Sin+states in our films as a function of the oxygen content. Figure 1共a兲shows the percentage of the different oxidation states of silicon determined by XPS in previous publications10,12 for films with different O/Si ratios prepared under oxygen partial pressures in the range of 10−9 to 10−6 Torr. Figure 1共b兲shows the theoretical distribution of Sin+species according to the MM and RBM as a function of this ratio.13 A comparison between the two figures shows clearly that the distribution of the Sin+species in our films does not follow the distribution predicted by any of the two models. For example, in the MM a film of global stoichiometry SiO1.3 contains 35% of Si0and 65% of Si4+ species 关cf. Fig. 1共b兲兴. Meanwhile, according to the RBM, a film of this stoichiometry would consist of a mixture of 2% Si0, 11% Si+, 31% Si2+, 38% Si3+, and 18% Si4+ 关cf. Fig. 1共b兲兴. By contrast, the experimental composition determined for a film of that stoichiometry is 23% Si+and 77% Si3+ 关cf. Fig. 1共a兲兴. The most outstanding result from these experiments is the practical absence of Si0and Si2+ species in all the layers.10,12 The particular distribution of species in these films was discussed in terms of kinetic constraints that hinder the disproportionation of Si3+ and Si+into Si0and Si4+ species.12 Meanwhile, it was shown that the disproportionation of Si2+ species into Si+and Si3+ species is a reaction thermodynamically favored and with a relatively low-energy activation barrier, two features that explain the absence of the Si2+ species in our films.12 The existence of activation barriers and kinetic constraints in SiOxthin films grown at room temperature contradicts the basic assumption of the RBM 共i.e., equal probability for the different types of fourfold-coordinated Si atoms兲and explains why we do not obtain the distribution of Sin+states predicted by that model. B. Auger parameter and thin-film stoichiometry For a given element, its Auger parameter 共 ␣ ⬘兲in a compound is defined as the sum of the binding energy 共BE兲of a photoelectron peak and the kinetic energy 共KE兲of an Auger peak.25,26 This parameter can be taken as a measurement of the extra-atomic relaxation energy of photoholes and therefore depends on the electronic properties of the material hosting that element. The exact definition of ␣ ⬘refers to a given oxidation state of an element and therefore its application to SiOxwould imply a precise deconvolution of both the Si 2pphotoemission and Si KLL Auger spectra. Particularly for the latter, such a deconvolution is not straightforward and, following the approach of Alfonsetti et al.,26 we prefer to apply the Auger parameter concept to the centroids of the Si 2pand Si KL2,3L2,3 spectra. This definition is not subjected to ambiguities and can be directly correlated with the stoichiometry and electronic properties of SiOxcompounds. Figure 2 shows the experimental values obtained for this averaged Auger parameter as a function of the content of oxygen in the films. The plotted straight line is a linear regression to the experimental points and can be used as a phenomenological relationship to identify the stoichiometry of SiOxcompounds according to the expression ␣ ⬘共Si兲= 1713.5 + 2.3共2−x兲.共1兲 This linear correlation agrees quite well with that found by Alfonsetti et al.26 for SiOxthin films prepared by a similar preparation procedure. In the present investigation we have also carried out the evaluation of the Auger parameter for the oxygen peaks. This analysis is interesting because, contrary to the Si 2pand Si KL2,3L2,3 peaks that are broadened because of the superposition of the components of the different oxidation states of silicon 共i.e., Sin+species兲, the oxygen peak virtually maintains the same width irrespective of the actual O/Si ratio. This behavior can be explained by considering that in the SiOxthe oxygen atoms are always bonded as a bridge between two silicon atoms. The evolution of the Auger parameter of oxygen 关 ␣ ⬘共O兲=O 1sBE+O KVV KE兴 as a function of stoichiometry is also represented in Fig. 2. The slope defined by these points and therefore the difference between the values for SiO1.3 and SiO2关i.e., ⌬ ␣ ⬘共Si兲 FIG. 1. 共a兲Distribution of Sin+species as a function of the oxygen content determined experimentally for SiOxthin films prepared by evaporation. 共b兲 Distribution of Sin+species as a function of the oxygen content predicted by the MM and RBM. FIG. 2. Evolution of the Auger parameters of Si and O as a function of the oxygen content in SiOxthin films prepared by evaporation. 113714-3 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp ⬃1.4 and ⌬ ␣ ⬘共O兲⬃1.2 eV兴are quite similar, likely indicating that for the two atoms the relaxation processes depend on similar physical variables. In contrast with the similar evolution with xof ␣ ⬘共Si兲and ␣ ⬘共O兲, BEs of Si 2pandO1speaks present a different behavior. Thus, while the centroid of the Si 2ppeak shifts by ⬃1.3 eV from SiO1.3 to SiO2共note that the BE difference between Si0and Si4+ species in, respectively, metallic silicon and SiO2is ⬃4eV兲,27 theO1speak remains practically constant at a fixed BE of 532.6 eV. Differences in the Auger parameter 共⌬ ␣ 兲are related to differences in the extra-atomic relaxation energies 共REs兲in the final state by the expression25 ⌬ ␣ =2⌬RE. 共2兲 Meanwhile, the difference in BEs accounts for differences in both the initial 共␧兲and final-state energies according to25 ⌬BE = ⌬␧ +⌬RE. 共3兲 For the analyzed samples the maximum differences in the value of the different energetic terms estimated through 共2兲 and 共3兲are as follows: ⌬RE共Si兲=0.7 eV, ⌬RE共O兲=0.6 eV, ⌬␧共Si兲=−0.1 eV, ⌬␧共O兲=−0.6 eV. The similar values of ⌬RE determined for both silicon and oxygen indicate that the medium surrounding the two atoms relaxes similarly upon the creation of a photohole. This similarity supports that our samples are homogeneous and consist of a single phase. The higher value of RE obtained for the SiO1.3 samples must be related to the higher electron density determined by plasmon analysis 共see below in Sec. III E兲. By contrast, ␧changes by −0.6 eV when determined for the oxygen and only by −0.1 eV when estimated for the silicon. According to 共3兲, this means that for oxygen ⌬␧ must compensate ⌬RE almost exactly. Changes in ␧occur because of differences in the density of charge at the oxygen atom or in the value of the Madelung potential around it. Oxygen atoms in SiOxare bridging between two Sin+ions whose actual densities of charge will depend on their formal charge n.Asnincreases 共i.e., towards Si4+兲the positive Madelung potential due to the neighboring silicon ions increases. We attribute the decrease of −0.6 eV in ␧共O兲measured for SiO1.3 with respect to SiO2 to a larger Madelung potential and an increasingly higher density of negative charge at the oxygen sites for samples with the SiO2stoichiometry. Similarly, intermediate deviations are expected for samples with 1.3⬍x⬍2. C. Valence-band states: XPS versus PES The occupied states of the valence band of SiOxthin films were analyzed by XPS and PES. Valence-band spectra of SiOxthin films excited with photon energies of 140 eV 共PES兲and 1253.6 eV 共XPS兲are shown in Figs. 3 and 4. The comparison of these two types of spectra permits to identify the different element and orbital contributions to the valence band. In fact, owing to the different photoexcitation cross sections for the O 2p,Si3p, and Si 3sstates that upon hybridization defined the occupied levels of the valence band of SiOxcompounds, the intensity of the different bonds of the spectra change with the photon energies used for photoexcitation.13,28 From the comparison of these two figures and taking into account the cross sections, it appears that the VB states at ⬃8 eV whose intensities are enhanced in PES have a predominant O 2pcharacter. Meanwhile, the structures between 10 and 17 eV whose intensities are enhanced by a factor ⬃10 by using higher photon energies in XPS must be attributed to Si 3pand Si 3sstates.28 Along with this preliminary attribution four main structures can be identified in the spectra of Figs. 3 and 4. The feature labeled O1and located at ⬃7.5 eV corresponds to the O2plone pairs of the SiOxstructure.13,14 Besides, two other features, labeled OII and OIII and located at ⬃12 and ⬃15 eV, correspond to O 2pstates strongly hybridized with Si 3pand Si 3slevels, respectively. The feature at ⬃4eV corresponds to Si–Si bonding states in SiOx.13,14,23 The Si 3p FIG. 3. PES spectra 共h ␯ =140 eV兲of two SiOxthin films with different O/Si ratios. The shadowed curve corresponds to the difference spectra between the two plots. FIG. 4. XPS spectra 共h ␯ =1253.6 eV兲of two SiOxthin films with different O/Si ratios. The dashed curve is the spectrum of a fully stoichiometric SiO2, while the shadowed curves are the difference spectra between the spectra of SiOxand SiO2. 113714-4 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp character of this latter feature is demonstrated by the enhancement of its intensity in the XPS spectra 共cf. Figs. 3 and 4兲.13 Figures 3 and 4 also show a comparative analysis of the spectra of SiOxsamples with different O/Si ratios. The shadowed curves are the differences between the two spectra in each figure and represent the changes due to the formation of direct Si–Si bonds in SiOx. These difference curves, particularly for the XPS spectra, show that the intensity of the peak at 4 eV increases as the oxygen content in the film decreases 共i.e., difference curve with negative sign兲. A decrease in intensity is also observed in the XPS spectra at the valley around 10-eV BE. Meanwhile, for low stoichiometries, the OIII feature depicts a net increase in intensity and a shift in position. Similar tendencies are also found for OII, although the changes are less pronounced. No noticeable changes occur for feature OI. The separation ⌬between the maximum of the O 2p–lone pair 共feature OI兲and the O2p–Si3pstates took the values of 6.0, 6.8, and 7.2 eV for x=1.3, 1.7, and 2.0 respectively. This shift, as well as the enhancement of the gap between the O 2p–lone pair states and the O 2p–Si 3p features, is consistent with previous results reported in literature for other SiOxthin films.13 The spectra in Figs. 3 and 4 also show that the intensity of the Si 3pfeatures at ⬃4 eV decreases in intensity as x increases, although it is still important for O/Si ratios as high as 1.7. Previous results in literature for SiOxthin films with a distribution of oxidation states following the RBM show that the intensity of Si 3pstates abruptly vanishes for xvalues in the range of 1.3–1.5.13,14,23 On the other hand, theoretical calculations by Martinez and Yndurain14 based on this model predicts that a minimum length of ⬃10 atoms in the Si–Si chains is necessary to observe Si–Si bonding states 共i.e., Si 3pstates兲in the valence band. Taken these two assertations into account, our VB spectra provide an additional confirmation that the distribution of Sin+species in our films differs from that predicted by the RBM. This is in fact supported by the presence of a noticeable peak at 4 eV in the VB spectra of films with a relatively high O/Si ratio 共cf. SiO1.7 films in Figs. 3 and 4兲. This high intensity supports the existence in these samples of long chains where the Si is bonded to another Si ion and contradics the predictions of that model. D. Conduction-band states: XAS characterization of SiO x The unoccupied states of the conduction band of the SiOxthin films have been investigated by XAS. Si L2,3 edge and Si Kspectra were recorded for several SiOxthin-film samples. These spectra are shown in Figs. 5 and 6. According to the dipolar selection rule, Si L2,3 and Si Kspectra can be considered, respectively, as a representation of the distribution of Si 3s,Si3d, and Si 3pstates in the conduction band. A discussion about the electronic origin of the different structures appearing in the Si L2,3 and Si Kabsorption spectra of SiO2thin films can be found in the bibliography.29 Si L2,3 spectra of SiOxthin films with different stoichiometries have been previously reported by Baba et al.30 Si K-edge absorption spectra of SiOxsamples exposed to the air have been also reported by several authors.31 In a previous publication about the resonance behavior of the Auger peaks of SiO1.3 thin films, we also discussed about the shape of the Si K-edge spectra,10 although in that report, as in the present study, the samples were prepared and examined in situ. From the comparison of the different XAS results in literature for samples measured ex situ, it is evident that the correlation of the spectra with the O/Si ratio assumed for the films has to be done very cautiously because the samples, particularly at the most external layers, may undergo changes in stoichiometry by reaction with oxygen and water vapor of the atmosphere. In the Si L2,3 spectra in Fig. 5 for SiO1.3 and SiO1.7 samples, particularly in the spectrum of this latter, there are a series of well-defined peaks and broadened structures located at 106.2 共LI兲, 108.3 共LII兲, 114.1 共LIII兲, and ⬃126.0 eV 共LIV兲. In a theoretical study by Chaboy et al.,29 these authors attributed similar features found in SiO2to the t2representation of d-like orbitals 共i.e., LIV feature兲and the erepresentation of the Si 3dlevels hybridized with orbitals of the ligands 共LIII兲. The difference between the spectra in Fig. 5 for samples SiO1.7 and SiO1.3 shows that the intensity of the two white line peaks decreases for the sample with the lowest oxygen content, while new available states develop at energies E ⬍106.2 eV. Some increase in intensity is also observed in this sample for an energy E⬎109 eV, although the difference spectrum might not be completely accurate in this spectral region because of the large slope of the backgrounds of the two spectra in this figure. Baba et al.30 have studied the Si L2,3 spectra of SiOxthin films produced by low-energy oxygen-ion implantation in Si共100兲whose electronic structures are consistent with the MM. In those films the XAS spectra for xvalues higher than 0.4 were very similar to that of SiO2perhaps indicating that these SiOxin situ samples have undergone oxidation by air exposure. The XAS spectra of Fig. 5 are completely different from those reported by Baba et al.30 for the same O/Si ratios. This means that the conduction bands of the evaporated SiOxfilms differ from that of SiO2共Ref. 30兲even with stoichiometries as high as 1.7. In addition, the absorption feature at E⬍106.2 eV can be attributed to the intermediate oxidation states present in our layers 共see the XPS analysis of Fig. 1兲. FIG. 5. Si L2,3 absorption spectra of SiOxthin films with different O/Si ratios. The shadowed curve corresponds to the difference spectra between the two plots. 113714-5 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp Figure 6 shows a series of K-edge absorption spectra for samples SiO1.3 and SiO1.7. For the sake of comparison, spectra of the Si共100兲, SiO2, and Si关SiO2兴samples are also included in the figure. The latter sample corresponds to a SiOx thin film thermally activated in vacuum to produce the dismutation of the Sin+共n=1,3兲states into Si0and SiO2 共Yubero SS-1999兲. The main feature for SiO2is a strong resonance at ⬃1846 eV which decreases in intensity for samples with a lower content of oxygen. In the spectra of SiOxthin films the well-defined peaks at ⬃1844 and ⬃1841 eV grow in intensity with respect to the feature at ⬃1846 eV as the oxygen content in the samples decreases. The spectrum of the SiO1.3 thin film is similar to spectra previously reported by Flank et al.32 and attributed by them to stoichiometric SiO. Meanwhile, the spectrum of Si0is characterized by a broad and low-intensity band at 1840 eV. The features at photon energies higher than 1842 eV in the Si共100兲thin film are not present in polycrystalline silicon and will not be commented upon here.31,33,34 These bands must not be confounded with the sharp peak situated at a slightly lower energy in the SiOxsamples 共i.e., at ⬃1841 eV, cf. Fig. 6兲. This is clearly evidenced by comparison with the spectrum of the sample labeled as SiO关Si0兴, where there is a mixture of Si0and SiO2. Thus, the two sharp peaks at ⬃1844 and ⬃1841 eV present in the two SiOxsamples have disappeared from the spectrum of SiO关Si0兴, while the position of the absorption edge of this latter sample coincides with the edge position obtained for the Si0spectrum. E. REELS spectra of SiO x thin films Information about the distribution of the unoccupied states of a material can be also obtained by REELS. Figure 7 shows ␭K共normalized REELS兲spectra for SiOxsamples of different stoichiometries.22,35 The shape of these spectra is the result of electronic transitions from the valence band to the conduction band plus the effect of collective excitations 共plasmons兲induced by the transport of a fast electron in the SiOxmatrix. This figure also reports spectra of the SiO1.7 sample recorded with various E0energies between 100 and 2000 eV. Spectra recorded with low primary energies are more surface sensitive because of the smaller information depth of these electrons. Previous XPS characterization of the SiO1.3 films reveals that they are mainly formed by Si3+ species.10,12,18 This is corroborated by the features at 5.1 and 7.3 eV in the loss spectrum of SiOxthat, in previous theoretical calculations, have been assigned to losses induced by neutral oxygen vacancies.36 In our case these vacancies can be associated with silicon ions close to vacancies that very likely consists of Si3+ or Si+states. According to the spectra in Fig. 7 recorded with different electron energies for the SiO1.7 thin film, these states should be majority at the topmost layers of the samples that, accordingly, would be heterogeneous in depth. The energy of bulk plasmons is usually identified as the maximum of the main feature appearing in a loss spectrum. For the series of spectra in Fig. 7, the plasmon energy shifts to higher energies with the value of the O/Si ratio. The experimental relation between the bulk plasmon energy Epand the value of xplotted in Fig. 8 can be described phenomenologically with the following expression obtained by linear regression of experimental points: Ep= 22.36 − 5.62共2−x兲.共4兲 FIG. 6. Si K-edge absorption spectra of SiOxthin films of the indicated O/Si ratios. Sample SiO关Si0兴corresponds to a SiO1.3 thin film heated at 750 K for 3 h under vacuum. FIG. 7. 共a兲REELS spectra of sample SiO1.7 recorded with increasing primary energies. 共b兲REELS spectra recorded with E0=2000 eV for SiOx samples of different stoichiometries. 113714-6 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp The analysis of the REELS spectra provides a way to establish a relationship between the plasmon energy Epand the density ␳ of the material. In the free-electron approximation it holds that19,37 ␳ =1 829 MEp 2 Nv ,共5兲 where Mis the molecular weight of the compound 共M=ASi +xAO兲and Nvis the density of the valence-band electrons 共Nv=NvSi+xNvO兲. Thus, for SiO2, with Ep=22.3 eV, we obtain the expected bulk density value for silicon dioxide ␳ SiO2=2.2 g/cm3. Regarding the deposited SiO1.3, where Ep =18.6 eV, we obtain ␳ SiO1.3=1.7 g/cm3. Note that the density of Si is 2.4 g/cm3. A lower density for SiOx共x⬍2兲 compounds than that of Si and SiO2have been previously reported by several authors.1,13,19,38 From the series of samples analyzed in this study, the low density in the asdeposited SiOxfilms proves the formation of a new phase, and discards the possibility of considering the SiOxthin films prepared here as a mixture of segregated Si and SiO2 domains. For the series of thin films with variable O/Si ratios, the density is a linear function of the stoichiometry according to the following phenomenological expression derived by linear regression of the experimental points: ␳ 共SiOx兲= 2.2 − 0.8共2−x兲.共6兲 This expression must be considered as derived from expression 共4兲relating plasmon energies with the stoichiometry of the films 共i.e., refraction index and absorption coefficients兲. It is apparent in Fig. 7 that the spectral intensity in the region of low-energy losses 共E⬍9eV兲decreases until it nearly vanishes for SiO2where a clear band gap of ⬃8eVis observed. The evolution of band-gap energies estimated from the REELS spectra is plotted in Fig. 9 as a function of x. The reported values have been determined from the curves in Fig. 7 by linear fitting and extrapolation to zero intensity of the loss edges. It is interesting that while for the samples with x⬍1.7 eV, the gap follows a linear relationship with xwith a slope of ⬃3.2-eV atomic ratio, it sharply increases for x =2.0 or values close to it. This result is in agreement with previous findings of Singh et al.23 for SiOxthin films of different stoichiometries. Optical properties in the UV region for energies higher than ⬃4 eV can be obtained by the analysis of the REELS spectra. Details about this type of analysis have been given in Ref. 22 and the results are shown in Fig. 10. The determination of optical parameters for this energy region is not always straightforward and REELS may circumvent the use of synchrotron techniques, usually required for this purpose. It is apparent in this figure that in the range 4⬍E⬍20 eV, the nfunction presents a decreasing tendency with the energy. In addition, the ncurves shift to higher energies as x increases. It is interesting that the ncurve for the SiO2thin film behaves differently and presents a well-defined maximum at ⬃9.5 eV. For E⬎20 eV the nvalues are very similar for all the samples. Meanwhile, the kcurves present a FIG. 8. Position of the bulk plasmon and density of the films determined by means of expression 共5兲for samples with different O/Si ratios. FIG. 9. Band-gap energies determined from the spectra in Fig. 7 共E0 =2000 eV兲as a function of the content of oxygen in the films. The line has been plotted to guide the eyes. FIG. 10. nand kfunctions derived from the REELS spectra for the indicated SiOxsamples. 113714-7 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp maximum that shifts with composition from ⬃7 eV for SiO1.3 to ⬃12 eV for SiO2. This means that the SiOxfilms become absorbent in the region with E⬍9 eV, corresponding to the band gap of SiO2. The formation of Si–Si bonds in the samples and the subsequent generation of associated electronic states must be responsible for this behavior. In fact, in samples with Si3+ and Si+species the absorption coefficient curve shows the appearance of electronic states in the region of the SiO2band gap. However, the fact that these states do not present any well-defined and intense loss feature produces that the refraction index does not present clear maxima in this region of the spectrum. IV. CONCLUSIONS The electronic properties of SiOxthin films deposited at room temperature by evaporation from a silicon monoxide source have been characterized by a series of techniques including XPS, PES, Si Kedge and L2,3 XAS, and REELS. The films present a distribution of Sin+states that differs from the predictions of the MM or RBM. Several phenomenological expressions have been obtained that correlate the O/Si ratio of the films with their electronic, optical, or structural parameters determined by the different techniques. Empirical linear expressions have been derived for the Auger parameter values, the distance between some spectral features of the VB spectra, bulk plasmon energies, and film densities. It has been also shown that the energy distribution of occupied and unoccupied states is strongly dependent on the stoichiometry of the films. For the lowest stoichiometric thin films, the band gap amounted to ⬃3.8 eV while it increases smoothly up to a film composition of about SiO1.7. Between SiO1.7 and SiO2the band gap of the film increases sharply to about 8 eV. For a O/Si ratio of 1.3, samples where Si0clusters have been prepared by thermal annealing depict a much lower energy gap which, in this case, is controlled by the low bandgap energy of elemental silicon. Samples with different O/Si ratios are characterized by a clear evolution of their electronic densities as determined by plasmon analysis. In parallel, it is shown that the value of the refraction index of these samples for E⬎4 eV also increases. All these evidences indicate that in the absence of Si0clusters, the stoichiometry of these SiOxmaterials can be considered as the main running parameter for the prediction of electronic and optical properties of the deposited materials. Also, it has been pointed out that the electronic properties of our films differ from those of other films with the same O/Si ratio reported in the literature 共i.e., those whose distribution of Sin+species is in agreement with the RBM or MM兲. In this regard, it must be concluded that not only the stoichiometry 共i.e., O/Si ratio兲 but also the distribution of Sin+species are key parameters for a straightforward comparison of the electronic structure of SiOxthin films deposited by different techniques. ACKNOWLEDGMENTS We thank the Ministerio de Ciencia y Tecnología 共Project No. MAT2001-2820兲for financial support and the EU for providing the access to synchrotron facilities. One of the authors 共A.B.兲gratefully acknowledges a postdoctoral fellowship from the Ministerio de Educación Ciencia y Deporte of Spain. 1M. J. O’Leary and J. H. Thomas, III, J. Vac. Sci. Technol. A 5,106 共1987兲. 2The Physics and Chemistry of SiO2and the Si–SiO2Interface,editedby C. R. Helms and B. E. Deal 共Plenum, New York, 1988兲. 3K. W. Wetcht, Appl. Opt. 30, 4133 共1991兲. 4J. A. Savage, Infrared Optical Materials and their Antireflection Coatings 共Adam Hilger, Ltd., Bristol, 1985兲. 5G. Hassand and C. D. Salzberg, J. Opt. Soc. Am. 44, 181 共1954兲. 6L. Canham, Nature 共London兲408,411共2000兲. 7B. Gallas, C. C. Kao, S. Fisson, G. Vuye, J. Rivory, Y. Bernard, and C. Belouet, Appl. Surf. Sci. 185, 317 共2002兲. 8F. Yun, B. J. Hinds, S. Hatatani, S. Oda, Q. X. Zhao, and M. Willander, Thin Solid Films 375,137共2000兲. 9T. Shimizu-Iwayama, K. Fujita, S. Nakao, K. Sayito, T. Fujita, and N. Itoh, J. Appl. Phys. 75, 7779 共1994兲. 10F. Yubero, A. Barranco, J. P. Espinós, and A. R. González-Elipe, Surf. Sci. 436,202共1999兲. 11F. Rochet, G. Dufour, H. Roulet, B. Pelloie, J. Perrière, E. Fogarassy, A. Slaoui, and M. Fromment, Phys. Rev. B 37, 6468 共1988兲. 12A. Barranco, J. A. Mejías, J. P. Espinós, A. Caballero, A. R. González- Elipe, and F. Yubero, J. Vac. Sci. Technol. A 19, 136 共2001兲. 13F. G. Bell and L. Ley, Phys. Rev. B 37, 8383 共1988兲. 14E. Martínez and F. Yndurain, Phys. Rev. B 24,5718共1981兲. 15K. Hübner, A. Stern, and E. D. Klinkenberg, Phys. Status Solidi A 136, 211 共1986兲. 16K. Haga and H. Watanabe, J. Non-Cryst. Solids 195,72共1996兲. 17N. Tomozieu, E. E. van Faassen, W. M. Arnoldbik, A. M. Vredenberg, and F. H. P. M. Habraken, Thin Solid Films 420–421, 382 共2002兲. 18A. Barranco, F. Yubero, J. P. Espinós, J. P. Holgado, A. Caballero, A. R. González-Elipe, and J. A. Mejías, Vacuum 67,491共2002兲. 19J. P. Holgado, A. Barranco, F. Yubero, J. P. Espinós, and A. R. González- Elipe, Nucl. Instrum. Methods Phys. Res. B 187, 465 共2002兲. 20A. Sassella et al., J. Vac. Sci. Technol. A 15, 377 共1997兲. 21P. G. Pai, S. S. Chao, Y. Takagi, and G. Lucovsky, J. Vac. Sci. Technol. A 4, 689 共1986兲. 22F. Yubero, V. M. Jiménez, and A. R. González-Elipe, Surf. Sci. 400,116 共1998兲, and references therein. 23A. Singh, S. C. Bayliss, S. J. Gurman, and E. A. Davis, J. Non-Cryst. Solids 142,113共1992兲. 24G. Zuther, Phys. Status Solidi A 59, K109 共1980兲. 25A. R. Gonzalez-Elipe and F. Yubero, Handbook of Surfaces and Interfaces of Materials, edited by H. S. Nalwa 共Academic, New York, 2001兲,p.147, and references therein. 26R. Alfonsetti, L. Lozzi, M. Passacantando, P. Picozzi, and S. Santucci, Appl. Surf. Sci. 70/71, 222 共1993兲. 27Y. Enta, Y. Miyanishi, H. Irimachi, N. Niwano, M. Suemitsu, N. Miyamoto, E. Shigemasa, and H. Kato, Phys. Rev. B 57,6294共1998兲. 28J. Yeh and I. Lindau, At. Data Nucl. Data Tables 32,1共1985兲. 29J. Chaboy, M. Benfatto, and I. Davoli, Phys. Rev. B 52, 10014 共1995兲. 30Y. Baba, H. Yamamoto, and T. A. Sasaki, Phys. Rev. B 48, 10972 共1993兲. 31C. Senemaud, M. T. Costa Lima, J. A. Roger, and A. Cachard, Chem. Phys. Lett. 26, 431 共1974兲. 32A. M. Flank, R. C. Karnatak, C. Blancard, J. M. Esteva, P. Lagarde, and J. P. Connerade, Z. Phys. D: At., Mol. Clusters 21, 357 共1991兲. 33N. Nagashima, A. Nakano, K. Ogata, M. Tamura, K. Sugawara, and K. Hayakawa, Phys. Rev. B 48, 18257 共1993兲. 34T. Kashiwakura, H. Arai, N. Kozuka, K. Odagawa, T. Yokohama, A. Kamata, S. Nakai, J. Electron Spectrosc. Relat. Phenom. 79,207共1996兲. 35S. Tougaard and I. Chokendorff, Phys. Rev. B 35, 6570 共1987兲. 36F. Bart, M. Gautier, F. Jollet, and J. P. Duraud, Surf. Sci. 306,342共1994兲. 37S. Tanuma, C. J. Powell, and D. R. Penn, Surf. Interface Anal. 17,927 共1991兲. 38S. M. A. Durrani, M. F. Al-Kuhaili, and E. E. Khawaja, J. Phys.: Condens. Matter 15, 8123 共2003兲. 113714-8 Barranco et al. J. Appl. Phys. 97, 113714 共2005兲 Downloaded 01 Mar 2010 to 161.111.180.191. Redistribution subject to AIP license or copyright; see http://jap.aip.org/jap/copyright.jsp