J. Phys. Chem. 1995,99, 1484-1490
1484
Spectroscopic Characterization of Quantum-Sized Ti02 Supported on Silica: Influence of Size and TiOz-SiOz Interface Composition G. Lassaletta, A. Fernfindez," J. P. Espin6s, and A. R. Gonzhlez-Elipe Instituto de Ciencia de Materiales de Sevilla and Dpto. Quimica Inorgbnica, P.0. Box 1I1 5, 41 080-Sevilla, Spain Received: May 11, 1994; In Final Form: September 22, 1994@
In the present paper we study Ti02 thin films prepared by evaporation of Ti in the presence of oxygen onto silica substrates. Samples have been analyzed by UV-vis absorption spectroscopy, LEIS (low-energy ion scattering), X P S (X-ray photoelectron spectroscopy), and REELS (reflection electron energy loss spectroscopy). The supported titania phase shows a significant blue shift in its band gap UV-vis absorption spectra which is higher than the value expected based on the thickness of the Ti02 layer. XPS analysis of the samples shows that the TiOz-Si02 interface is formed by cross-linking Ti-0-Si bonds, thus decreasing the positive charge of the Ti atoms at the interface, the mobility of the electrons in the titania phase, and the polarizability of the oxygen atoms at this interface. The influence of these effects on the observed band gap shift is expected to be significant and we have attributed the increase in the band gap to a combination of size quantization effects and interface phenomena. Comparison of the results obtained on silica with those obtained for Ti02 evaporated on graphite demonstrates the importance of support effects on the photoemission behavior and permits the characterization of the TiOz-SiO2 interface.
Introduction The electronic band structure of semiconducting oxides is size dependent if the dimensions of the studied crystallites are comparable with the exciton (Bohr) radius in the semiconductor (1-10 nm). These effects have been extensively studied for small particles in colloidal solutions (Q-particles) and thin solid films (Q-layers or Q-well~)'-~and various applications in optoelectronicsare f ~ r e s e e n .Although ~ these effects have been well characterized by the study of the optical absorption properties of the size effects in X-ray photoelectron spectroscopy (XPS) have not been reported for metal oxide particles. We note, however, that the shifts observed in the binding energy (BE) of photoelectron peaks and in the kinetic energy (KE) of the Auger peaks of small metal particles6-I0 have been generally recognized as being due to size effects. Further, in bulk oxides the shape of the oxygen Auger spectrum has been related to the different covalency of the M - 0 bond and to the different electronic structure of the oxides." In addition, for the first elements of the transition series, differences in the polarizability of the ligands have been related to the magnitude of the Auger parameter of the cations (Le., difference between the photoelectrons and Auger electrons).I2 In a previous paperI3 we presented a comparative study by XPS of Ti02 as bulk material, colloidal particles, and small particles supported on Si02 and found that the photoemission processes in these oxide materials are more influenced by support effects than by the size of the particles. In the present work we describe the study of Ti02 evaporated on a silica substrate and on a graphite single crystal by means of X P S , LEIS, REELS, and UV-vis absorption spectroscopy. Significant shifts in the Ti2p and 0 1s photoelectron peaks, as well as changes in the Auger parameters, have been found to be related to the nature of the TiOz-Si02 interface. Although many studiesI4 deal with the fundamental aspects of the metal-ceramic interfaces, less work has been done to
* To whom correspondence @
should be sent. Abstract published in Advance ACS Abstracrs, December 15, 1994.
0022-36541992099-1484$09.00/0
understand the ceramic-ceramic contacts. On the other hand silica-titania mixed oxides are important materials from two points of view: (i) It is well-known that these oxides can be used as acidic catalyst^'^,'^ and might be of interest as supports for industrial ~ata1ysis.l~(ii) They constitute very interesting multication ceramic oxides materials.18 Previous studies by X P S of these kind of materials are included in ref 18 and 19. The results shown in the present work permit the characterization of the electronic structure of the TiO2-SiO2 interface. In particular it is worth mention that the exhaustive characterization of the TiOa-SiO2 interface carried out in the present work clearly shows the formation of cross-linking Si-0-Ti bonds at the interface. These bonds strongly modify the electronic structure of the oxygen and titanium atoms at this interface, in comparison to bulk TiO2. Possible implications of these effects on the changes in the band-gap electronic structure of Ti02 are discussed.
Experimental Section Ti02 supported on a fused quartz substrate (Si02 99.9%, Goodfellow) or on an oxidized silicon wafer has been prepared by evaporation of Ti metal from a heated wire in the presence of 2 x Torr of oxygen. The X P S analysis of the samples shows that under these conditions Ti02 is formed as indicated by the presence of fully oxidized Ti4+ species. In addition the Ti/% atomic ratio, as determined by XPS, has been taken as a relative measurement of the coverage. The samples will be designated hereafter as TiO2-SiO2-P where P is the TUSi atomic ratio. For comparison, the same experiments have been carried out on graphite single crystals (Carbonne, Lorraine). As above, samples are denoted Ti02-C-P, where P is the Ti/C atomic ratio measured by XPS. Anatase P-25 from Degussa was used as bulk Ti02 reference for UV-vis absorption spectroscopy and XPS measurements. It is a polycrystalline material having an average grain size of 300 8, and a ratio of anatase to rutile of three. LEIS data were recorded with a LHS-10 spectrometer (from Leybold) working at a constant retardation ratio of three and 0 1995 American Chemical Society
Spectroscopic Characterization of Ti02
J. Phys. Chem., Vol. 99, No. 5, 1995 1485 TiOz-SiO2- P
I
TiO2-SiO2-P
s40kI ,
20
0 0
0.1
-,
0.2 Ti/Si( P )
,
0.3
, _
0.L
Figure 1. Ti and Si 1-keV He+ ion scattering signals normalized to provide a measure of the fractional surface coverage plotted as a function of the TUSi atomic ratio as measured by XPS.
with the electron energy analyzer polarized to detect positive charges. The ion gun, an extractor ion source (IQE 12/38 from Leybold) supplied with helium and working with an acceleration voltage of 1 kV, was swept a surface of 2 x 2 mm to prevent extensive etching of the sample. UV-vis absorption spectra were recorded for bulk titania in the reflectance mode (R,) and transformed to a magnitude proportional to the extinction coefficient (k) through the Kubelka-Munk function [F(R,)]. The TiOz-SiO2-P samples, prepared on fused quartz substrates, were measured in the transmission mode so that the absorbance (A) (which is also proportional to k) has been recorded. To evaluate the band gap of the different Ti02 samples, we have plotted (Ahv)2 or (F(R,). h ~ (where ) ~ hv is the photon energy) against hv. The linear portion of the curve has been extrapolated to absorption equal to zero?O XPS spectra were recorded by using a VG-Escalab 210 spectrometer working in the AE = cte mode with a pass energy of 50 eV and Mg K a radiation as the excitation source. The binding energy (BE) references were taken at the C 1s and Si 2p peaks of the substrates at 284.2 and 103.4 eV, respectively. An estimated error of kO.1 eV can be assumed for all measurements. Sensitivity factors supplied with the instrument were employed to calculate the atomic ratios. REELS spectra of TiOz-SiO2-P samples were measured by using the same instrument, the incidence and takeoff angles being 60" and 0", respectively, to the surface normal. All spectra have been normalized and calibrated with respect to the elastic peak.
Results and Discussion I. Low-Energy Ion Scattering. Growth of Ti02 thin films on silica substrates was first characterized by recording the enhancement of the Ti signal, and the attenuation of the substrate Si signal, in both X P S and LEIS experiments for different TiO2-SiO2-P samples. Specifically, Figure 1 shows the 1-keV Hef ion scattering results from such an experiment, the Si and Ti scattering signals being expressed as percentages of the sum of both as a function of the P value of the sample. This mode of display should give a reasonable guide to the proportion of the surface covered by each species.21 Figure 1 shows that attenuation of the substrate Si signal and growth of the Ti overlayer signal are both essentially linear for low coverages; such behavior is characteristic of a layer by layer growth of the Ti02 overlayer.21 Further, since LEIS is particularly surface sensitive and one complete monolayer of Ti would be expected to obscure the substrate Si signal (see Figure l), a fast coverage of the silica substrate by growth of the Ti02 overlayer is indicated.
h? ( e V ) Figure 2. W-vis absorption spectra in the band-gap region of Ti02SiO2-P samples and bulk Ti02 (powdered sample).
11. UV-vis Absorption Spectroscopy. Figure 2 shows the transformed W-vis absorption spectra of three different Ti02SiO2-P samples together with the band-gap values, evaluated by linear extrapolation. The spectrum of bulk Ti02 is also included for comparison. The above samples have been prepared by evaporation of Ti in the presence of oxygen on a transparent fused quartz substrate to allow recording of the Wvis absorption spectra in transmission mode. For samples with lower coverage values we approached the sensitivity limit of the spectrometer, so no data are available for P < 0.4. The observed blue shift of the band-gap absorption edge may be explained in terms of the quantum size effects previously observed for small titania particle^.^^^^^^ Specifically, if we assume a layer by layer growth during the evaporation of Ti02 on silica, consistent with our previously described LEIS results, the band-gap shift (AE,) is given by23-25
Here pq and pz are the reduced effective masses of electronhole pairs in the plane of the layer and perpendicular to it, and Lq and L, are the corresponding dimensions. For an ideal thin layer, the first term in eq 1 can be neglected since Lq is large. Consequently, the band-gap shift depends only on L,. Despite the fact that the reduced effective masses suffer from a significant uncertainty in the literature and are also related to the band-gap energy,26 the dimensions of the Ti02 overlayer may be estimated by using eq 1. Using a reduced effective mass of electron-hole pairs &) ranging from 2.72 to 1.5 a,27 we calculate a deposit dimension in the range 5-8 and 4-6 8, for the band gap shifts of 500 and 800 meV measured for the samples TiO2-SiOz- 1.2 and Ti02-Si02-0.4, respectively. In A E S and X P S studies the decay in intensity of photoelectrons is given by eq 2.28
I = I , exp( -d/A)
(2)
d is the displacement of the point of origin of the photoelectrons from the sample surface (d = 0) and 1 is the inelastic mean free path for such photoelectrons. According to this expression, which assumes layer by layer growth, if the silica substrate is covered by a Ti02 layer of thickness d, the intensity of the Si
Lassaletta et al.
1486 J. Phys. Chem., Vol. 99, No. 5, 1995
TiOz-Si02 -P
, Ti L3 M2&
.///
I
0.030
0.0
J 390
&bo
110
K.E.(eV 1 Figure 4. TiL3M23V Auger peak of Ti02 evaporated on silica for different TYSi atomic ratios as measured by XPS.
!
,
475
,
!
465
I
,
455
B.E.(eV) Figure 3. Ti 2p photoelectron peak of Ti02 evaporated on silica for different Ti/Si atomic ratios as measured by XPS.
2p peak should be attenuated exponentially and d can be evaluated by using eq 3.
d = I In Id1
(3)
1 for the Si 2p photoelectrons of ca. 1150 eV in the Ti02 matrix can be estimated from the empirical function (1 = BE”2, E > 150 eV) described by Seah and Dench.28 A value of ca. 32 8, has been obtained. The Id1 ratio has been substituted by the ratio of the Si 2p XPS signal (referenced to the oxygen peak intensity) of the clean silica substrate and the corresponding TiOz-Si02-P sample. We obtain from eq 3 a d value of 27 and 11 8, for samples Ti02-Si02-1.2 and Ti02-Si02-0.4, respectively. Based on these values the band-gap shift observed in the optical absorption spectra appears higher than expected. A possible explanation is that the coverage layer is formed by crystallites of very small size (4-5 8, in radius), leading to the observed blue shift of the optical absorption. The LEIS results show, however, a layer by layer instead of islandlike growth mode. On the other hand the value of 4-5 A appears extremely small, corresponding to one Ti-0 unit. In fact in the next section of the paper the XPS study shows that the electronic characteristics of the samples are highly influenced by the TiOa-SiOz interface, which may also influence the optical absorption characteristics of the samples. 111. X-ray Photoelectron Spectroscopy. The TiL3M23V Auger and Ti 2p photoelectron peaks have been recorded for different Ti02 samples obtained by evaporation of titanium in the presence of oxygen, either on Si02 or on a single crystal of graphite. These spectra, shown in Figures 3 and 4, exhibit significant shifts in the position of the maxima. In Figure 5 we have represented the changes in Ti 2p BE and the modified Auger parameter (a’= a hv, a‘ = BE of the Ti 2p peak
+
+
0
0.2
0.4
0.6
0.8
1.0
2
Ti/Si
Figure 5. Evolution of the Ti 2p BE and the modified Auger parameter
(a’)of Ti02 evaporated on silica substrates as a function of the Ti/Si atomic ratio as measured by X P S . (Substrate: 0 fused quartz, 0 oxidized silicon.)
KE of the TiL3M23V Auger peak) for a series of TiOz-SiO2-P samples as a function of P. Most of these data were recorded by using fused quartz substrates thereby allowing simultaneous recording of W - v i s optical absorption spectra. Parallel experiments were performed on oxidized silicon wafers, to have better conducting samples, and the results were fully equivalent (see Figure 5). For low Ti02 coverages the BE of Ti 2p is higher by ca. 0.7 eV and a’ is smaller by ca. 2.6 eV than those of the steady state (high P values or bulk TiO2). A similar experiment carried out on a graphite single crystal is represented in Figure 6, showing practically no dependence of these parameters on the amount of evaporated TiOz, thus indicating an important role of the Si02 substrate on the effects described in Figure 5. In addition, if we assume that the increase in the P value of the samples is equivalent to an increase in the Ti02 deposit dimensions, the experiment on graphite suggests that size effects do not significantly affect photoemission. In fact
Spectroscopic Characterization of Ti02
J. Phys. Chem., Vol. 99, No. 5, 1995 1487 01s
Si02
az90
I
873
,
,
,
.
,
.
,
.
-
-
.
28
7 2 L
>
.
ti02
- 8 . 871
-
0
0.2
0.L
a6
0.8
1.0
I
n.
536 5.3 532 Yo 528
TilC
Figure 6. Evolution of the Ti 2p BE and the modified Auger parameter (a') of Ti02 evaporated on a graphite single-crystal substrate as a function of the Ti/C atomic ratio as measured by XPS.
it appears that the TiOz-SiOz interface is the most important parameter influencing photoemission. Similar results have been reported by us in a previous paper13 in which we compare the photoelectron spectra of Ti02 in the form of bulk material, colloidal particles, and small clusters supported on SiO2. These results contrast, however, with those obtained for small metallic particles,6-Io in which the shifts in BE and in the Auger parameters can be the result both of interaction of the particle with the support and of differences in the electronic characteristics of the particles according to their size and/or shape.6-'0 In the simple approximation introduced by Thomasz9 and Wagner,30 the chemical and Auger parameters shifts can be expressed as in eqs 4 and 5.
ABE = A€ - AR
(4)
Aa' = 2AR
(5)
t is a term related to the eigenvalue of the level undergoing photoemission and the initial state charge distribution, R is the extraatomic relaxation energy of the photohole, and a' is the modified Auger parameter. By use of eqs 4 and 5, the AR and A t values with respect to the bulk Ti02 can be evaluated. For the TiO2-SiO2-P samples with low Ti02 coverages (low P values), where the TiOz-SiOz interaction should be higher, we obtain from Figure 5 a ABE of $0.7 eV and a ha' of -2.6 eV, which in turn gives values of -1.3 and -0.6 eV for AR and A€, respectively. The greatest differences occur for the relaxation energy (i.e., the Auger parameter), which in principle can be due to a particle size effect. However, as already mentioned, the absence of changes in a' for Ti02-C-P samples demonstrates that the nature of the support plays a more important role than layer dimension in decreasing the extraatomic relaxation energy of the photohole. Therefore, if we take into account the layer by layer growth mode of the Ti02 deposit, the photoelectrons created in the Ti atoms bonded to Oz- ions shared by the Si02 at the TiOz-Si02 interface must be less efficiently screened than those generated in Ti ions of a bulk material. This is reasonable if we consider that electrons in silica, as an insulator, are less mobile than electrons in a semiconductor oxide like Ti02 and is also related to the fact that the oxygen atoms present at the TiOz-SiOz interface should be less polarizable in comparison to oxygen atoms of bulk Ti02 (silica has a dielectric constant of ca. 2 in comparison to TiO2,
Binding Energy (eV)
Figure 7. 0 1s peak for different TiOz-Si02-P samples recorded for a takeoff angle of 15' with respect to the surface normal. The Si02 and Ti02 references are included together with the fitting analysis
of the spectra. which has values ranging from 48 to 173 depending on the crystal structure and direction). In fact, this interpretation follows previous papers by Moretit2 that demonstrate that in the case of a nonlocal screening mechanism (as it is the case for Ti), an increase in the ligand electronic polarizability is expected to increase the screening energy for the central atom undergoing photoemission. Although the total shift in the Ti 2p BE is dominated by the final state effect of the extraatomic relaxation energy of the photohole (AR = -1,3), this effect is partially compensated by changes in the t values (At = -0.6). These changes may arise from a decrease in the positive charge on Ti atoms in the initial state due to the formation of mixed Ti-0-Si bonds at the interface. In fact such cross-linking bonds have been postulated to be formed in TiO2-Si02 mixed oxide g l a s ~ e s . ~To ~ - assess '~ this hypothesis we have carried out a careful analysis of the OKLL Auger and the 0 1s photoelectron peaks of the different TiOz-SiO2-P samples, recording the spectra with a takeoff angle of 15" with respect to the surface normal to increase the surface sensitivity. In Figure 7 we have shown the 0 1s spectra obtained together with their corresponding fitting analysis. From this figure we can conclude that for intermediate P values the experimental spectra cannot be fitted with two single peaks at 532.9 and 530.7 eV, corresponding respectively to oxygen in Si-0-Si and Ti-0-Ti bonds (values obtained for bulk Si02 and high Ti02 coverages on silica). A new peak at an intermediate BE value is clearly present at the interface, which is assigned to oxygen in Si-0-Ti cross-linking bonds. This agrees with a decrease in the negative charge of the oxygen (or a decrease in the polarizability) and a decrease in the ionic nature of the bonds at the interphase in comparison to the Ti-0-Ti bonds in bulk titania. The presence of this intermediate oxygen is also demonstrated in Figure 8 where we have depicted the 0 1s spectra recorded for two different P values at different takeoff angles (from 90" to 15"). For low coverages the presence of an intermediate oxygen peak different from those typical for Ti and Si is clearly seen. In addition Figure 9 shows the OKLL Auger line series at increasing P values. Apart from the characteristic peaks for silica and titania at kinetic energies of 506.9 and 511.4 eV, respectively, corresponding to a' Auger parameters of 1039.8 and 1042.1 eV, the appearance of an intermediate peak is clearly observed, particularly at low
Lassaletta et al.
1488 J. Phys. Chem., Vol. 99, No. 5, 1995 CTi02-Si02- 0,OL
I
,-an,
0.20}
0.20
a05 ,lo
t
t
B.E. (eV 1 Figure 8. 0 1s peak for two different TiOz-SiOz-P samples recorded at the indicated takeoff angles.
I Ti02-Si02 - P
O(KLL)
1
0.3
A
j 0.2
-a' UI E
I
3
0.1
.
L
m
,
520
I
Kinetic Energy(eV1 Figure 9. X-ray excited OKLL Auger line for the different Ti02SiO2-P samples recorded for a takeoff angle of 15" with respect to the surface normal. Si02 and Ti02 references are included.
coverages where the sensitivity to the Ti02-Si02 interface is higher. This confirms the existence of Ti-0-Si cross-linking bonds and the presence at the interface of a new type of oxygen that is changing the electronic characteristics of the Ti atoms in comparison to bulk TiOz. A similar calculation to the one carried out with the Ti 2p BE and the a' Auger parameter for Ti (according to expressions (4) and (5)) has been carried out with the 0 1s and OKLL peaks. If we compare the two extreme situations for pure silica and high titania coverages, we find for the 0 1s a ABE value of +2.2 eV and a Aa' value of -2.3 eV when we change from oxygen in Si02 to oxygen in Ti02 (charge effects may lead to uncertainties in the BE values but not in the a' parameters). These values lead to a Ac of 1.05 eV and a AR of - 1.15 eV, c o n f i i n g that in silica the screening of the photoholes is lower than in Ti02. However, the initial-state charge distribution represents a decrease in the negative charge of the oxygen in
+
silica in comparison to that in titania. Both effects may be extensive to the intermediate oxygen at the Ti-0-Si crosslinking bonds. In the case of graphite the absence of such an interface of shared oxygen leads to the absence of support effects in the XPS behavior, which again indicates the significance of the Ti-0-Si cross-linking bonds. From the above-described results we can conclude that the presence of the TiOz-SiO2 interface decreases the ionicity of the Ti-0 bonds in comparison to bulk titania as stated from the changes in the e parameter of 0 and Ti atoms (initial-state effects). At the same time, the mobility of the electron in the titania phase decreases in comparison to that in bulk Ti02 as stated from the decrease in the relaxation energy R of the created photoholes (final state effects). These effects, promoted by the presence of oxygen atoms shared by both Ti and Si, may be responsible for the observed band gap shifts together with size quantization phenomena. Another important consideration is related to the geometry of the oxygen coordination around Ti. Effectively with titanium oxides supported on Vycor glass3' and titanium oxides dispersed into Si0232and A120333matrices, it has already been reported that titanium is present in the form of tetrahedral Ti-0 units. As shown is those papers the presence of such tetrahedral coordination around Ti should induce the large blue shift observed in the band gap and also the changes in the X P S data. In the experiments presented in this paper the formation of these Ti-0 tetrahedral units should be especially significant at low coverages, where the LEIS results have shown the presence of only one Ti-0 unit at the Si02 surface. Finally, we want to emphasize that for P values higher than 0.4, the effect of the TiOz-SiOz interface on the photoelectron peaks of the supported Ti02 phase starts to be negligible. This is due to the fact that surface sensitivity of the XPS analysis decreases exponentially with d so that for high coverage we are analyzing layers that closely resemble bulk TiO2. However, using W - v i s absorption spectroscopy, an average analysis of the full titania layer is achieved. Further experiments are now being carried out on metallic substrates and will be the subject of another work. However, some experiments carried out on an alumina substrate by evaporation of Ti02 have shown a similar behavior to that obtained for silica supports. In fact at increasing titania coverages we have observed a decrease in the binding energy measured for the Ti 2p photoelectron peaks and an increase in the kinetic energy of the TiL3M23V Auger peak. All these results are in full agreement with the previously described interpretation of the nature of the TiOz-SiOz (or Ti02-Al203) interface. IV. Electron Loss Spectroscopy. To complete characterization of these systems we have carried out an electron loss spectroscopy study. Figure 10 shows the typical REELS spectra recorded for different TiOz-SiO2-P samples at EO = 200 eV (this low energy has been selected to increase surface sensitivity). For comparison, the data for clean Si02 and for a sample prepared on graphite (Ti02-C-0.12) have been included. Peaks in the loss spectra are observed at 46.9, 24.5, 13.2, 10.7, and 5.8 eV, being nearly identical to previous EELS data for Ti02 and oxidized Ti.34-37 The peaks at 24.5 and 46.9 eV have been a t t r i b ~ t e d to ~ ~volume . ~ ~ plasmon excitation and Ti 3p resonance, respectively. These are shallow core level features; so it is not surprising that the change in the nature of the substrate, as well as the increase in coverage, does not have an effect on these losses. In fact, these excitations have been previously observed to be unaffected by changes in chemical en~ironment.~'
Spectroscopic Characterization of TiOz
J. Phys. Chem., Val. 99, No. 5, 1995 1489
0.05-
0.00
.
1
'
'
"
"
'
-30 -20 -10 Energy loss (eV)
-60 -50 -LO
Energy Loss (eV1
-1
Figure 10. REELS spectra at Eo = 200 eV recorded for the different TiOe-SiOz-P samples. The Si02 and the Ti02-C-0.14 samples have been included for comparison.
The increase in loss intensity at energies between 3 and 6 eV can be associated with the onset of excitation across the band gap of the titanium oxide. A peak is observed for Ti02 at 5.5 eV,34,38which has been associated with an 0 2p e; transition. The start point for this increase in loss intensity across the band gap has been estimated for the spectra in Figure 6 from the first derivative profiles, an increase in this value being observed from 3.2 eV to 3.6 eV when the P values vary from 0.58 to 0.04. This tendency agrees with the band-gap shift to higher values detected by UV-vis absorption upon decrease of the TUSi ratio in the Ti02-SiOz-P samples. The peak at 10.7 eV has been assigned to an 0 2p to Ti 3d excitation and has been associated with the presence of Ti in the 4+ (3d0) oxidation state.39 The loss at 13.2 eV has been observed for various titanium oxides39 possessing different stoichiometries and has been assigned to an 0 2p to 0 3s excitation. Regarding these two peaks we have subtracted a flat base line indicated by the dashed line in Figure 10: the results are depicted in Figure 11. The assignment of losses at 10.7 and 13.2 eV to excitations localized in Ti and 0 atoms, respectively, is confirmed by the observed increase in intensity of the peak at 10.7 eV with the Ti02 coverage. On the other hand, the peak is very narrow for low P values, suggesting that the Ti 3d states have partially lost their band character as a consequence of the size quantization and the influence of the Ti-0-Si interface. With respect to the peak at 13.2 eV, its intensity is similar for the samples prepared on graphite, for the TiOz-SiOz samples of high P values, and for bulk Ti02. However, for the samples supported in silica with low P values this peak is higher in intensity, indicating changes in the electronic distribution of the oxygen atoms at the TiO2-SiO2 interface. Again the growth of titania on the silica substrate seems to occur by anchoring of the Ti02 phase through Si0-Ti cross-linking bonds, while for inert graphite substrate growth seems to occur by the formation of intact Ti02 octahedra from the very beginning.
-
Conclusions We have concluded that Ti02 thin films on silica substrates exhibit significant blue shifts in the band-gap W - v i s absorption spectra. If we ascribe such shifts to a pure quantum size
Figure 11. REELS spectra obtained after background subtraction of the dashed line in Figure 10 for the indicated samples.
effect, the calculated crystallite size is smaller than the thickness of the coverage layer evaluated from the X P S analysis. This may be due to the fact that the coverage layer is constituted from crystallites smaller than the thickness of the deposited layer. However, the X P S analysis of the samples demonstrated that the TiOz-Si02 interface is formed by cross-linking Ti0-Si bonds, thus decreasing the positive charge of the Ti atoms at the interface and decreasing the mobility of the electrons in the titania phase. On the other hand, the LEIS analysis indicates that the Ti-0 units, formed at the Si02 surface by evaporation of Ti in the presence of oxygen, should contain Ti in tetrahedral coordination. The influence of such effects on the band-gap energy are significant and we have attributed the increase in the band-gap energy to a combination of size quantization effects and interface phenomena. In addition, the study by X P S and REELS of the samples supported on graphite shows that in this inert substrate the growth of the Ti02 phase occurs by the formation of intact Ti02 octahedra, showing the importance of shared oxygen species at the TiOz-SiOz interface. It should be mentioned here that such interface phenomena are in general not considered in the studies of quantum-sized semiconductor particles, although they should be significant in semiconductor quantum dots40 inmersed in different matrices, and even in colloidal semiconductors the adsorption of different species at the surface of the particles may have an influence in the measured band-gap values. The implication of the data reported in this paper on the knowledge of ceramic-ceramic interfaces should be also emphasized.
Acknowledgements. Authors thank the DGICYT (Project No. PB91-0835 and PB93-0183) for financial support. References and Notes (1) Brus, L. E. J . Chem. Phys. 1983, 79, 5566. (2) Kormann, C.; Bahnemann, D.W.; Hoffmann, M. R. J . Phys. Chem. 1988, 92, 5196. ( 3 ) Haase, M.; Weller, H.; Henglein, A. J . Phys. Chem. 1988, 92,482. (4) Kavan, L.; Stoto, T.; Gratzel, M.; Fitzmaurice, D.; Shklover, V. J. Phys. Chem. 1993, 97, 9493. ( 5 ) Chemla, D. S. Phys. Today 1985, 38, 57. (6) Mason, M. G. Phys. Rev. 1983, 27, 748. (7) Jirka, I. Surf. Sci. 1990, 232, 307. (8) Kohiki, S . Appl. Surf. Sci. 1986, 25, 81. (9) Cheung, T. T. P. Sufi. Sci. 1984, 140, 151. (10) Asakawa, T.; Tanaka, K.; Toyoshima, I. Langmuir 1988, 4 , 521. (1 1) Ascarelli, P.; Moretti, G. S u Interface ~ Anal. 1985, 7 , 8.
1490 J. Phys. Chem., Vol. 99, No. 5, 1995 (12) Moretti, G. Surf. I n t e ~ a c eAnal. 1990, 16, 159; 1991, 17, 352. (13) Femlndez, A,; Caballero, A.; Gonzllez-Elipe, A. R. Surf. Interface Anal. 1992, 18, 392. (14) Li, J.-G. J . Am. Chem. SOC.1992, 75, 3118. (15) Shibata, K.; Kiyoura, T.; Kitagava, T.; Sumiyoshi, T.; Tanabe, K. Bull. Chem. Soc. Jpn. 1973, 46, 2985. (16) Niwa, M.; Sago, M.; Ando, H.; Murakami, Y. J. Catal. 1981, 69, 69. (17) Vishwanathan, B.; Tanaka, B.; Toyoshima, L. Langmuir 1986, 2, 113. (18) Ingo, G. M.; Dirk, S.; Babonneau, F. Appl. Surf. Sci. 1993, 70/ 71A, 230. (19) Stakheev, A. Y.; Shpiro, E. S.; Apijok, J. J. Phys. Chem. 1993, 97, 5668. (20) Zhang, D. H.; Brodie, D. E. Thin Solid Films 1992, 213, 109. (21) Bardi, U. Appl. Surf. Sci. 1991, 52, 89. (22) Anpo, M.; Shima, T.; Kodama, S.; Kubokawa, Y. J. Phys. Chem. 1987, 91, 4305. (23) Sandorf, C. J.; Kelly, S. P.; Hwang, D. M. J . Chem. Phys. 1986, 85, 5337. (24) Sandorf, C. J.; Hwang, D. M.; Chung, W. M. Phys. Rev. 1986, 33B, 5953. (25) Smotkin, E. S.; Lee, Ch.; Bard, A. J.; Campion, A,; Fox, M. A,; Mallouk, T. F.; Webber, S. E.; White, J. M. Chem. Phys. Lett. 1988, 152, 265. (26) Liu, C.; Bard, A. J. J. Phys. Chem. 1989, 93, 3232.
Lassaletta et al. (27) Kuskinski, J. J.; Gomez-Jahn, L. A.; Faran, K. J.; Gracewski, S. M.; Miller, R. J.; Dwayne, J. J . Chem. Phys. 1989, 90, 1253. (28) Seah, M. P.; Dench, W. A. Surf.Int. Anal. 1979, I , 2. (29) Thomas, T. D. J . Electron. Spectrosc. Relat. Phenom. 1980, 20, 117. (30) Wagner, C. D. Faraday Discuss. Chem. Soc. 1975, 60, 291. (31) Anpo, M.; Aikawa, N.; Kubokawa, Y.; Che, M.; Louis, C.; Giamello, E. J . Phys. Chem. 1985, 89, 5017. (32) Anpo, M.; Nakaya, H.; Kodama, S.; Kubokawa, Y. J. Phys. Chem. 1986, 90, 1633. (33) Anpo, M.; Kawamura, T.; Kodama, S.; Maruye, K.; Onishi, T. J. Phys. Chem. 1988, 92, 438. (34) Chung, Y. W.; Lo, W. J.; Somorjai, G. A. Surf. Sci. 1977,64,588. (35) Henrich, V. E.; Dresselhaus, G.; Zeiger, H. J. Phys. Rev. Lett. 1976, 36, 1335. (36) Bertel, E.; Stockbauer, R.; Madey, T. E. Surf. Sci. 1984,141, 355. (37) Smith, P. B.; Bemasek, S. L. Surf. Sci. 1987, 188, 241. (38) Henrich, V. E.; Zeiger, H. J.; Dresselhaus, G. Electron Energy Loss Spectroscopy of Surface States on Titanium and Vanadium Oxides. In Electrocatalysis on Non-Metallic Surfaces; Natl. Bur. Std. Special Publ. 455, US GPO: Washington, DC, 1976. (39) Kurtz, R. L.; Henrich, V. E. Phys. Rev. B 1982, 26, 6682. (40) Weller, H. Adv. Mater. 1993, 5, 88. JP941147D