Chemical Changes of Titanium and Titanium Dioxide under Electron Bombardment

Thin metal oxide films over metal and semiconductor surfaces are probably the most widely studied structures in basic physics and technology. The reason is quite simple; their applications cover areas of interest such as catalysis microelectronic devices, medicine, and aircraft construction. Metal-metal oxide structures involving titanium (Ti) are among the most interesting and used ones. Thus, the surface oxide passivation of Ti makes possible the biocompatibility of Ti and its use in medicine, and the production of protective coatings against corrosion. Since most of the useful properties are determined by phenomena occurring at the surface, the kinetic and thermodynamic features of the oxide films growth and chemical reactions at the surfaces are key facts in most of these processes. These chemical reactions can be greatly affected by the incidence of energetic particles or photons. While sometimes these reactions are looked for wanted effects, such as in electron lithography, in other cases they represent experimental artifacts, like in surface analysis techniques based on electron spectroscopies. The chemical reactions can be either oxidation or reduction depending on the particular substrate. Thus, while in electron lithography the desired effect is the metallic reduction, the electron bombardment can enhance by several orders of magnitude the oxidation kinetic of a semiconductor. The identification of the different products appearing in the growth of oxide films, as well as in the different reaction occurring at surfaces, allows us a better understanding of the different processes and, at the same time, opens the possibility of designing reliable growth and characterization methods. In this paper, using Auger electron spectroscopy (AES) and factor analysis (FA), applied in the sequential way, we study the oxidation process of metallic titanium and the reduction process of a TiO 2 thin film under electron irradiation. In both cases we determine the chemical state of the different compounds formed over the surface. The application of FA to the Auger spectra allows us to obtain the Auger line shapes of the different species formed along the irradiation process. 2. Experimental details


Introduction
Thin metal oxide films over metal and semiconductor surfaces are probably the most widely studied structures in basic physics and technology.The reason is quite simple; their applications cover areas of interest such as catalysis 1 microelectronic devices 2,3 , medicine 4,5 , and aircraft construction 6 .Metal-metal oxide structures involving titanium (Ti) are among the most interesting and used ones.Thus, the surface oxide passivation of Ti makes possible the biocompatibility of Ti and its use in medicine, and the production of protective coatings against corrosion 7 .Since most of the useful properties are determined by phenomena occurring at the surface, the kinetic and thermodynamic features of the oxide films growth and chemical reactions at the surfaces are key facts in most of these processes.These chemical reactions can be greatly affected by the incidence of energetic particles or photons.While sometimes these reactions are looked for wanted effects, such as in electron lithography [8][9][10] , in other cases they represent experimental artifacts, like in surface analysis techniques based on electron spectroscopies.The chemical reactions can be either oxidation or reduction depending on the particular substrate.Thus, while in electron lithography the desired effect is the metallic reduction 11 , the electron bombardment can enhance by several orders of magnitude the oxidation kinetic of a semiconductor 12,13 .The identification of the different products appearing in the growth of oxide films, as well as in the different reaction occurring at surfaces, allows us a better understanding of the different processes and, at the same time, opens the possibility of designing reliable growth and characterization methods.
In this paper, using Auger electron spectroscopy (AES) and factor analysis (FA) 14 , applied in the sequential way 15,16 , we study the oxidation process of metallic titanium and the reduction process of a TiO 2 thin film under electron irradiation.In both cases we determine the chemical state of the different compounds formed over the surface.The application of FA to the Auger spectra allows us to obtain the Auger line shapes of the different species formed along the irradiation process.

Experimental setup
All measurements were performed at room temperature in a commercial ultra high vacuum (UHV) surface analysis system with a base pressure in the lower 10 -10 Torr range, equipped with Auger electron spectroscopy (AES) facilities.Differentiated Auger spectra of the transitions Ti LMM , Ti LMV and O KLL were acquired using a single-pass cylindrical mirror analyzer with a resolution of 0.6% and modulation amplitude of 4 and 2 V p-p for the oxidation and reduction processes, respectively.The incidence angle of the electron beam was 30º with respect to the surface normal.The primary electron beam energy was 3 keV, the irradiated area ~7 x 10 -5 cm 2 , the electron gun was permanently switched on during the oxidation and reduction processes, and the acquisition of the Auger spectra was performed within the same irradiation conditions.The current density was 48 and 35 mA.cm -2 , for the oxidation and reduction processes, respectively.
The sample used for the oxidation process was high-purity polycrystalline titanium (with 125 ppm of impurities including Fe, Si, Mn, Mg and Sn).The sample surface was cleaned by ion bombardment using Ar + at 4 keV until carbon and oxygen contamination was below the AES detection limits.The UHV chamber was filled with high purity oxygen (better than 99.997%) at a pressure of 2 x 10 -8 Torr.The pressure was measured by means of a non-corrected Bayard-Alpert gauge and the oxygen exposure was maintained at least until the saturation of the oxygen Auger signal was obtained.
The TiO 2 film for the reduction process was grown over a Cu(100) single crystal, cleaned by cycles of Ar + ion bombardment (1 keV) followed by annealing at 850 K, until carbon and oxygen contamination was below the AES detection limit.The TiO 2 film was prepared by Ti thermal evaporation from a high purity polycrystalline Ti bar heated through electron bombardment in an O 2 environment (2 x 10 -7 Torr).The Cu substrate was kept at room temperature during the co-adsorption process, which leads to the formation of stoichiometric amorphous TiO 2 films with an abrupt interface with the Cu substrate 17 .The evaporation rate was determined through the attenuation of the substrate signal with deposition time, being a typical value 0.15 mL/min.The analyzed films were ~10 mL thick.This thickness was large enough to attenuate any Auger signal coming from the substrate, but small enough to prevent sample charging.
The irradiation dose φ was calculated as φ = J a ⋅ t i , where J a is the current density and t i the total irradiation time in seconds.Since, the same irradiation conditions were used during the AES spectra acquisition, the total dose was obtained by the addition of irradiation and acquisition times.A typical spectrum acquisition time was 30 seconds and the whole irradiation time until saturation ~6 hours.

Data treatment
The factor analysis (FA) method 14 has been extensively discussed in previous works 12,15,16,18,19 , and it is currently included as a standard data treatment system in most data treatment packages.Therefore we will limit ourselves here to a brief description.The first step in FA is the determination of the number of linearly independent factors, i.e. the minimum number of pure components required to describe the complete series of spectra corresponding to the evolution under study.In doing that, we compare the error obtained, by fitting the experimental data with a minimum set of factors, with the experimental error (EE).This procedure is performed as Auger spectra are added in a sequential way 15 , and each time the fitting error surpasses the experimental one, a new factor appears in the process.
Once the number of independent factors is known, the Auger line shape of each pure component (or base) is determined through a least square fit procedure called target transformation (TT) 14,15 .
The final step in FA is the correlation among the obtained bases and actual compounds, which is performed through a simple fingerprint procedure.Since the experimental Auger spectra of all Ti compounds involved in this study are available in the literature [19][20][21][22][23][24] , this final step is easily performed.

Results and Discussion
In Figure 1, we present two sets of spectra showing the evolution of the Auger line shape corresponding to the Ti LMM , Ti LMV and O KLL transitions as a function of the electron irradiation dose during the Ti oxidation (a) and TiO 2 reduction (b) processes.The total irradiation dose required to reach saturation during the oxidation process is one order of magnitude larger than that for the reduction.As the irradiation dose increases, the spectra show the growth/decrease of the amount of oxygen at the surface for the oxidation/reduction process and changes in the Auger line shapes of Ti in both cases, especially in the region corresponding to the LMV transition.As expected, this last result suggests changes in the Ti electron density of states due to the oxidation/reduction process.
To analyze the results in a more quantitative way, we depict in Figure 2 the peak intensities corresponding to the Ti LMV (~418 eV) and O KLL (~510 eV) Auger transitions as a function of the electron bombardment time for both experiments already mentioned.The intensities were normalized to the O KLL initial yield of the TiO 2 spectrum.
In Figure 2a, the Ti LMV and O KLL Auger yields show a typical oxidation behavior, with two clearly different evolution stages: an initial fast decrease/increase of the titanium/oxygen amount at the For comparison purposes, we added a curve obtained from measurements made keeping the electron bombardment to a minimum () 19 .
surface followed by a slowing down of the kinetics until a plateau is reached.In Figure 2b, we depict the results for the reduction process.Although a substantial change is observed for the Ti LMV signal during the first 5 minutes of bombardment, i.e. a large monotonic increase, little modifications are visible for the O KLL signal in this scale.Thus, a zoom for the latter is shown in the inset of Figure 2b.As we can see, the oxygen depletion begins as soon as the irradiation starts, showing a large monotonic decrease along the whole process.
A more detailed analysis can be performed based in the fact that only one of the main Ti Auger transitions involves valence electrons (Ti LMV ) 25 .Thus, it has been proposed that the ratio between both Ti Auger transitions yields, Ti LMV /Ti LMM hereinafter named as R VM , can be used as a good indicator of the Ti oxidation state 26,27 .Moreover, comparing the evolution of this parameter with factor analysis results 19 , we showed that the evolution of the R VM ratio can be used as a first indicator for a qualitative analysis.
In Figure 3, we show the evolution of the R VM ratio as a function of the oxygen Auger yield, for the experiments already mentioned.The arrows indicate the direction in which each experiment has been done.We can observe that the Ti oxidation state for the oxidation experiment strongly depends on the electron irradiation; in fact changes in the R VM ratio take place for lower O KLL yields in comparison with measurements made keeping the electron bombardment to a minimum 19 .The same effect has been observed by decreasing/increasing the oxygen pressure/ electron beam density for a given electron density/pressure.Thus, under electron irradiation we can get larger changes in the R VM evolution for lower oxygen Auger yields 28 .These results can be explained based only in the oxygen enhanced diffusion induced by electron irradiation.In the case of the reduction experiment, the R VM shows a monotonic increase originated in the electron transference to the Ti valence band (chemical reduction process), which is correlated with the monotonic decrease of the O KLL yield along the process.The dotted line, added in the figure as a guide to the eye, seems to suggest that the oxide reduction under electron irradiation might follow the same path as the Ti oxidation in a reversible way.As we will see, this is not true, but in order to prove this fact we need to identify the different Ti compounds that appear along both experiments, by applying the factor analysis method 14 to the titanium Auger line shape evolutions.
In order to determine the number of independent components present in the oxidation and reduction processes, we apply the PCA in the sequential way 15,16,19 .In Figure 4, we depict the evolution of the relative error attained in reproducing the experimental data, acquired along the irradiation processes, using one and two factors.We found that both, the oxidation and reduction processes, can be described in terms of two independent components.Since metallic titanium and titanium dioxide are present as the starting surface component for the oxidation and reduction processes, respectively, under these particular analyzed conditions the results imply that only one Ti oxide is created by electron irradiation in both cases.
In Figure 5, we compare the shape of the Auger spectra obtained through FA for the oxidation (a) and reduction (b) processes with known experimental Auger spectra of Ti 20 and different Ti oxides (c) 21,22,29,30 .The spectrum ascribed to the first component (Base 1) obviously corresponds to pure Ti for the oxidation process and TiO 2 for the reduction process.The second component spectrum seems to be the same for both experiments.A fingerprint procedure allows us to identify this second independent component as Ti 2 O 3 .The slight discrepancies observed among the bases coming from the different experiments, may point out the presence of small amounts of a third component.Within the FA capability of compound identification, the oxidation and reduction processes are characterized by the appearance of only one Ti oxide, i.e.Ti 2 O 3 .
In Figure 6, we show the evolution of the weight of each independent component in the total Auger line shape of the experimental spectra acquired along the irradiation processes under study.One of the processes is characterized by a rapid metallic Ti depletion and the other one by TiO 2 chemical reduction, both giving place to the formation of Ti 2 O 3 .Although, the oxide reaches a similar maximum weight in both experiments, it is clear that the reduction is a faster process.Another interesting issue to be noticed is that for both experiments the Ti 2 O 3 weight reaches a steady stage at a saturation value of about 80%, which prevents the complete disappearance of metallic Ti or TiO 2 in the oxidation and reduction processes, respectively.This last result can be explained as a consequence of the deposition of little amounts of carbon at the surface, favored by the electron bombardment, which tends to stabilize the oxide preventing us from reaching the complete disappearance of metallic Ti or TiO 2 .From the fact that at saturation in the reduction process we still observe the presence of TiO 2 , we conclude that the reduction length towards the bulk is below 20 Å (typical escape depth of Ti Auger electrons in an oxide).
Although the electron stimulated desorption is present in both experiments, it is of crucial importance in the chemical reduction of TiO 2 to Ti 2 O 3 due to electron irradiation.This process can be explained on the basis of the Knotek and Feibelman model (KF) [31][32][33] , which is the most suitable for electron stimulated desorption on ionically bonded surfaces.The KF model introduces the Auger stimulated desorption, which in a maximal valence ionic compound is produced by the repulsive potential that appears when the sign of the anion charge changes due to an Auger transition.In an ionic compound, like TiO 2 , electrons are spatially accumulated around the O anions, while the Ti is nominally ionized down to the noble gas configuration Ti 4+ (maximal valence).If an energetic electron ionizes a titanium (oxygen) inner shell, the dominant decay will be an inter-atomic (intra-atomic) Auger process involving two or three O electrons.This will leave both O and Ti positively charged giving place to the O expulsion.In the sub-oxide Ti 3+ , however, valence electrons do reside on the Ti cation.Consequently, the dominant decay will be an intra-atomic Auger process, which will make the Ti more positive and thus will generally increase surface stability.

Conclusions
We have characterized through Auger electron spectroscopy and factor analysis the titanium (Ti 0 ) oxidation and titanium dioxide (Ti 4+ ) chemical reduction processes under electron irradiation.We found that the electron bombardment affects the chemical reactions at the surface depending on the experimental environment.Thus, for the oxidation process we have a competition between the enhanced reaction kinetics through the dissociation of oxygen molecules induced by the electron bombardment, and the stimulated desorption produced, for instance, through the Knotek-Feibelmann process.On the other hand, the stimulated desorption is the only process present in the oxide reduction, previous to the C enhanced deposition.The measurable output in both cases is the appearance of an intermediate Ti oxidation state, i.e., Ti 2 O 3 .

Figure 4 .
Figure 4. Evolution of the error attained in reproducing the data matrix as a function of the electron bombardment time considering one () and two () factors, for the metallic Ti oxidation (a) and TiO 2 reduction (b) processes under electron irradiation.The horizontal line at unity corresponds to a calculated error equal to the EE.(a)

Figure 5 .
Figure 5. Auger spectra of the independent bases obtained after applying the TT method to the Ti spectra acquired along the electron stimulated oxidation of metallic Ti (a) and reduction of a TiO 2 film (b) under electron irradiation.Experimental Auger spectra of Ti20 and its different compounds (c)21,22 .

Figure 6 .
Figure 6.Evolution of the weights of the different titanium compounds obtained with the TT method as a function of the electron bombardment time, for the metallic Ti oxidation (a) and TiO 2 reduction (b) processes under electron irradiation.