- Email: [email protected]
Influence of Surface OH Groups and Traces of Water Vapor During the Preparation of TiO2-SiO2 Samples
G. Poncelet,P.A.Jacobs,P.Grange and B. Delmon (Editors),Preparation of Catalysts V 0 1991 Elsevier Science PublishersB.V., Amsterdam - Printed in The Netherlands
INFLUENCE OF SURFACE OH GROUPS AND TRACES DURING THE PREPARATION OF Ti02-Si02 SAMPLES
OF
627
WATER VAPOR
A. MUBOZ-PAEZ and G. MUNUERA
Dept. of Inorganic Chemistry and Instituto de Ciencia de Materiales (UNSE-CSIC) P.0.Box 1115, 41071 Sevilla SPAIN SUMMARY
TiOz-SiOz samples have been prepared by impregnation of silica support with n-hexane solutions of titanium alcoholate controlling the hydration/ hydroxylation degree of the silica surface. Once calcined, the samples were characterized by IR, XRD, SEM/EDAX and XAS. The mechanism proposed for the decomposition of the alcoholate involves reaction with adsorbed molecular water in a first step, followed by anchoring by reaction with acid OH- groups. The amorphoustitaniumoxide obtained after calcination shows a layered open structure of clusters formed by a few octahedra sharing edges and corners. INTRODUCTION Titania has been widely used as support in metal catalyst due to its ability to modify the catalytic properties of the metal (ref. 1). As a consequence, the study of the interactions taking place at the metal-titania interface has attracted the interest of several research groups (ref. 2). Nevertheless it is very difficult to obtain high surface area titania (>lo0 m2/g), and while studying the metal-titania interactions it is difficult to get information from the support because the bulk properties of the Ti02 mask those of the surface, the unique part of the titaniaaffectedby the metal. To overcome both problems, inert oxides like silica, have been used as support to obtain high surface area dispersed titania, by grafting to the SiO, support (refs. 3,4)throughthe impregnation from n-hexane solutions of Ti alcoxides, that by hydrolysis and calcination would produce the final coated Ti02-Si02 powders. EXPERIMENTAL Preparation af catalysts The surface oxide was prepared by impregnation of silica Aerosil-200 (SBE,=200 m2/g), with a n-hexane solution of a Ti-alcoholate (tetraisopropyltitanate, Ti(OPr’)4 from Tilcom, 16.9% Ti). The TiOz percentage (by weight) used has been c.a. 12%. This value correspondsroughly to the amount required to form a monolayer of titania on this type of silica (14.7%, 5.5 Ti/nm2) and
628
is close to the amount needed to allow the grafting of each Ti atom to one hydroxyl group of the silica surface (13 %, 5 OH/nm2) (ref. 5). In the standard procedure, the desired amount o f Ti(OPri)4 was dissolved in dried n-hexane (25 ml/g of silica in Methods 1-3 and 6 ml/g of silica in Method 4) and the solution was allowed to react with the surface of the silica for several hours. After that, the solvent was removed at room temperature by flowing Nz, and the sample heated in N 2 up to 673K. Subsequently the solids at 873K. Four differents methods were used as were calcined i n air follows: Method 1. Reaction under Nz at 300K for 20h of undried SiOz. Method 2. Reaction under Nz at 300K for 20h of Si02 dried at 388K for 2h. Method 3 . Reaction under Nz at 350K for 5h of SiOz dried at 440K for 2h, cooling down to room temperature, filtering and washing with n-hexane and subsequently with water. Method 4. Incipient impregnation at 300K in the air of undried silica with a n-hexane solution of Ti Tests were made along the preparation (Methods 1 to 3) by gas chromatography to detect i-PrOH in the liquid phase as well as the presence of Ti(OPr’)4 by hydrolysis in aliquots of the liquid phase. Characterization of solids IR spectra were carried out at 300K on a wafer of the sample mounted in a cell that allows in situ thermal treatments under controlled atmospheres up to 773K, using a Perkin-Elmer 684 spectrometer fitted to a 3600 data station. X-ray diffractograms were recorded in a Phillips 1730 diffractometer and scanning electron micrographs using an IS1 microscope model SS-40, with an energy dispersive X-ray analyzer (EDAX) KEVEX, model 8000 fitted to it. XAS experiments were performed on the EXAFS station 8.1 in the Synchrotron Radiation Source at Daresbury Laboratories with ring energies of 2 GeV and ring currents of 250 mA. The EXAFS spectrum was recorded at 140K in an “in situ” cel1,wherethe sample was placed after being pressed with BN into a wafer with an absorbance ( p x ) of 2.5 at the Titanium K-edge assuring an optimum signal to noise ratio. Data analysis was carried out by fitting in kand R-space using the phase and amplitude corrected Fourier transforms to identify the different contributions (ref. 6). Phase shift functions and backscattering amplitudes were obtained from reference compounds. RESULTS Assuming that the two reactions taking place during the decomposition of the alcoholate to TiOZ are grafting through OH- groups on the surface of the silica and hydrolysis of the Ti-alcoholate by water to produce colloidal
629
particles (refs. 3 , 4 ) , the only competitor to decompose the alcoholate would be the water vapor from moisture. Thus, we have used several preparation methods in which moisture was carefully avoided. Therefore, in the first case (Method l ) , the grafting would involve only OH- groups and/or water molecules adsorbed on the silica surface. Taking into account that water physisorbed on the surface of the silica could produce mainly ungraphted titania, we have carried out a second preparation method (Method 2) in which adsorbed water was avoided by submitting the silica to a previous outgassing treatment at 388K that would remove at least physisorbed water. A new method (Method 3 ) was designed in which all molecular water was removed and, considering that the hydrolysis process could be very slow at room temperature in such extremely dry conditions, the reaction temperature was raised up to the boiling point of the n-hexane. In this case, after five hours of reaction the liquid phase, still containing Ti(OPr’)4, was filtered off and the sample was thoroughly washed with n-hexane, to remove the unreacted alcoholate, and then with water to get a complete hydrolysis of the grafted alcoholate. During the washing with water, formation of a thin, opaque white layer was clearly observed which, unlike the transparent silica, remained stuck on the surface o f the filter. This white coating, presumably TiOz, once dried, calcined and weighed,turn out to be c.a. 50% of the total amount of titanium oxide that should be formed by decomposition of all the alcoholate employed in this preparation. Control test during the preparation in Methods 1 and 2 showed a complete hydrolysis of Ti(OPri)4 at the end of the reaction time (20h), while in Method 3 the liquid still contained the alcoholate in spite of the more drastic thermal conditions used in this case. Finally, a fourth type of preparation was carried out consisting in the well known incipient impregnation method, using the n-hexane solution of Ti (OPr i , and the si 1 ica support without any drying pretreatment. In principle, the degree of success of the anchoring process in our preparations could be followed by checking the changes in the concentration o f one of the reactants (i.e.surface OH-/HzO at the silica support) using I R spectroscopy, since silica aerosil shows a characteristic sharp band at 3750 -1 cm due to basic free hydroxyls together with a broader band due to more acidic OH- groups at 3680 cm-’ (ref. 7). So, their reaction can be followed by changes in their intensities as previously observed in similar preparations (ref. 8). Thus, figure 1 shows IR spectra o f the SiOp support and of a sample prepared by method 1 containing only 1% Ti02 before submitting the samples to any thermal or outgassing treatment. The unique change observed after the addition of the alcoholate is the decrease of the intensity in the range
630 100.
b
a
Fig. 1. I R spectra in OH stretching the region of a sample l%TiOz-SiOz (a) and of si 1 ica support (b) registered in the atmosphere (solid line) and after outgassing at 673K for 2h.
%A
50 I
\
I
\
I
\
I
I 0
.
I
4000
'
1% Ti0 S i O
\,
2
\\,
cm
.% ' -1
2
I
3000
4000
cm
--I
3000
cm-', where the bands due to the more acidic OH- groups and/or molecular water appear, what suggests that these species are those mainly involved in the interaction of the alcoholate with the SiOz support. The IR spectra in the same figure, recorded after outgassing at 673K to remove the water readsorbed upon exposure to air, clearly show the decrease in the intensity of the band at 3680 cm-' assigned to the more acidic OH- groups of the silica, thus suggesting their participation in the decomposition of the alcoholate. cm-' of the samples Figure 2 shows I R spectra in the range 2900-3400 prepared with c.a. 12% Ti02 by the four methods (spectra have been normalized using u s i - o at 1830 cm-' from the bulk of the silica, to make them comparable). Except for the sample prepared by method 3 , in all other cases the intensity o f the band at 3750 cm -1 remains nearly unchanged with respect to that of the silica support pretreated under similar conditions, while changes in intensity and/or position are observed in the band at 3680 cm-' in all the preparation methods. The increase in intensity o f the IR bands in the 0-H stretching region in sample prepared by Method 3 is probably related with the final washing with water used in this method. oxide phases XRD was used to check the crystallinity of the titanium formed after calcination by decomposition of the hydrolyzed alcoholate. Only small shoulders appear in sample 1, 2 and 3 in the position o f the most 3700-3550
63 1
h
8 b
0
m
I Ti0 S i O -4 2
0
-c---c--t-cm
+
-1
Fig. 2 . IR spectra in the range 3900-3400 cm-' of the silica support and the samples 12%TiOz-SiO2 prepared by the four methods outgassed at 673K for 2h. intense peak of anatase, while no peaks were visible at the positions of the most intense diffraction lines of anatase, rutile or brookite in sample 4 . Nevertheless, when the alcoholate was hydrolyzed with water in the absence of silica and calcined under similar conditions, strong peaks appear in the positions of the most intense diffractions of anatase, thus indicating that the hydrolysis o f the pure alcoholate produces crystalline phases. Analysis o f the samples using SEM/EDAX was carried out to examine thehomogeneityof the titanium distribution on the TiOz-Si02 samples, and the homogeneity in grain shape and size. Thus, in sample prepared by Method 2, (using dried Si02) the grains have angular shapes and the local concentration of Ti changes drastically when going from one grain to another. The changes are less drastic, although still remarkable in sample 1, that shows round grains. In sample 3 the particle size was bigger than in the other cases, and the existence of different types of particles (opaque and transparent) could be seen without the aid of the microscope. The most homogeneous sample, considering grain shape and size, as well as titanium dispersion was sample 4 , that has a very homogeneous spongy appearance with constant concentration of T i in all the grains. From the previous results, we deduced that method 4 is the best one, so this sample was studied by XAS to get a deeper insight into the structure around Ti ions. The XANES region of this sample has been plotted in figure 3 , where the corresponding spectra o f anatase and rutile, measured as a
632 C
D 0.1
h
' " 0 Lr
-0.1
J I
-20
I
I
0
1
E(eV) 20
I
40
Fig. 3 . Ti k-edge XANES spectra of TiOz rutile (a), TiOz anatase (b) 12%TiOz-SiOz prepared by method 4. Fig. 4. Ti k-edge, Fourier transforp of the EXAFS syectrum of sample lZ%TiOz-SiOz prepared by method 4. (k , Ak=3.12-11.00 A- ) . Arrows indicate the ranges for Fourier filtering used during the data analysis. reference, have been included as well. In addition to the round shape of features C and D, typical of amorphous compounds (ref. 9), it has to be pointed out the appearance of the triplet A,,A2,A3 characterisitic of octahedral symmetry (ref. 10) that indicates that the absorbing atom is six fold coordinated. Nevertheless, there is a remarkable change in the intensity ratio between peaks AZ and AB, that is close to 1 in anatase or rutile and close to 0.5 in sample 4 where it shows a shape similar to the spectra of uncalcined TiOz colloids prepared from hydrolyzed Ti(OPr')4 (ref. 11). A similar shape has been observed in the spectra of titania-silica glasses prepared by gelation in air of Ti and Si alcoxides by Emili et a1 (ref. 9 ) , who have assigned it to the existence of Ti ions in tetrahedral environment. The Fourier Transform o f the EXAFS signal yields the radial distribution function shown in figure 4, where we can see an intense peak at around 1.7 A due to backscattering from the first shell of oxygen atoms. For higher distances there is a drop in intensity that, in principle, could be assigned to the lack of higher coordination shells. Nevertheless no good fit could be obtained with only one or two shells. So, we have performed the data analysis shell by shell, doing inverse Fourier Transforms of increasing ranges, shown
633
TABLE 1
1
TiOz-Si02-4 Shel 1
N R(A) -
I
Anatase ~
Ao2(AZ)
5.8
1.93
0.011
1.0 6.7
3.09 3.78
-0.005 0.03
7.2
4.37
0.06
3.3
5.34
0.00
Shel 1
NxR(A)
Ti-Ol
4x1.93 2x1.98
Ti-Ti Ti-02 Ti-Ti2 Ti-03
4x3.04 8x3.86 4~3.78 8x4.25 8x4.27 4x4.75
Ti-Ti3
ax4. 85
number octahedr lS'
qrd
by the arrows in figure 4. We started the analysis considering the basic octahedra of anatase Ti06 RTi-O= 1.95 A (fit range 0.16- 2.3 A). Afterwards, we expanded the range up to 3 A, and included a shell Ti-Ti. When the fitting range was expanded to 4.1 A , two new Ti-0 bonds were required to reach a good fit. Finally, to reach the final values the range for the Fourier filtering was 0.16-5.4 A requiring the inclusion of a new Ti-0 bond at 5.3 A . The parameters of this fit are summarized in Table 1, that includes the number of the neighboring atoms, N, the absorbing atom-neighbor distance, R, Debye-Waller factor, Ao', related to static and thermal disorder, as well as the structural parameters of crystalline anatase appearing in a cluster of 4 octahedra (ref. 12). A plot of the raw data and the best fit in k and R space for the wider range has been included in Fig 5. The first peak in the Fourier transform may be attributed to the six Ti-Ol bonds of the basic octahedron, as already predicted from the XANES data. The distance is the same that the short bond of the distorted octahedra in anatase. The peaks between 3 and 6 A are a complex result of the overlap of four different features. The first one, Ti-Ti at 3.09 A, is very similar to the distance observed in anatase between two octahedra sharing edges (3.04 A), while the next one, Ti-02 at 3.78 A, is very close to the distance of the oxygen atoms in the second octahedron (3.86 A). The shell Ti-03 would correspond to oxygen atoms in a third octahedron in an anatase-like structure. The shell Ti-04 has no correspondance in a cluster of anatase structure including four octahedra. In relation with the similarities with the anatase strucutre in the other four shell, it has to be pointed out the low coordination number of the Ti-Ti bond at 3.09 A , as well as the lack
634
Fig. 5 . Ti k-edge EXAFS spectrum and Fourier transform (kl, Ak=3.5-10.5 A-') of the raw data (solid line) and best fit (dotted line) of sample 12%TiOz-Sioz prepared by method 4. of Ti-Ti bonds for higher shells. Both facts indicate that only small clusters of TiOs octahedra are present on the SiOz support. DISCUSSION Formation in our conditions of colloidal particles of Ti02 grafted to the high surface area SiOz can be assumed t o occur according to one of the two following schemes: Scheme 1
Ti (OR)
+
(-Si-O)n-Ti(OR)4-n
-
Si-OHb __ > (-Si-O)n-Ti(OR)4-n+
+
(4-n)HzO
n ROH
>(-Si-O)n-Ti(OH)4-n+(4-n)ROH
(1) (2)
Scheme 2
Ti (OR) -
Si-OHa
+ +
4 HZOads
Ti(OH)4
------> Ti (OH)4 + > Si-O-Ti(OH)3
4
ROH
(3)
+ H20
(4)
where -Si-OHa and -Si-OHb stand for basic and acid OH- groups at the SiOz surface, Ti (OR)4 for Ti(OPri)4 monomers and HZOads for physisorbed/chemisorbed water. In the first case, grafting should involve in a first step the more basic OH- groups of the silica through a hydrophylic attack, and in a second step hydrolysis by reaction with adsorbed water or moisture. According to scheme 2, hydrolysis of the alcoholate by adsorbed water at the SiOz support is postulated, leading t o T i hydroxide colloidal particles in a first step, which must be followed by anchoring to the SiOz surface through reaction with more acidic OH- groups, a process that should be enhanced by the final thermal treatment during the calcination used in the preparation of the samples.
635
IR data in figures1 and 2 suggest that Scheme 2 (hydrolysis by adsorbed water followed by grafting) is the most likely in the conditions used in our preparative work, since the band at 3750 cm-’, due to more basic OH- groups, is not modified during the whole process. Moreover, changes in the band at 3680 cm-’ due to more acidic hydroxyls, can be explained by assuming that the grafting involves this type of hydroxyls of the silica surface. It is worth noting that preparation by Method 3 , where adsorbed water and probably part of the acidic OH- groups have been removed from the S i 0 2 support before reaction, only allows ca. 50% reaction of the Ti-(OPr1)4 in spite of the presence of all the basic OH- groups. This fact again excludes these groups from the process (reaction (1)). In fact, hydrolysis of the alcoholate remaining at the SiO2 surface in this case only occurs by washing with water what probably also produces breaking of siloxane bridges at the Si02 surface, (partially dehydroxylated) as detected by the much larger intensities of the IR bands for this sample in figure 2. If we assume Scheme 2 , grafted colloidal titania particles, similar to those obtained from simple hydrolysis of Ti-(OPr’)4 with water, should be obtained and therefore their structure should not be very different from that recently proposed by Leaustic et a1 (ref.11). In fact, the XANES spectrum of our sample is very similar to the spectrum recorded by these authors for such colloidal particles. However, there are big differences in the EXAFS region that can be explained by the smaller size of the titania particles obtained in our system. Moreover, after heating at 373K, these authors obtain crystalline anatase, as previously did Kozlowski et al.(ref. 12) and Reichmann et a1 (ref. 3 ) during the preparation of similar systems, while the crystalline structure of anatase could not be detected by XRD in our samples, even after calcination at 873K thus implying that the layered open structure, remains stabilized on the surface of the silica. It is not surprising that the best preparation method for this type of ultradispersed Ti02-SiOz systems was the incipient impregnation, since in this conditions the lack o f an excess of solvent will probably prevent the growth of the original nuclei t o bigger colloidal particles. Additionally, this method has the advantage that it is the easiest and provides homogeneous and well dispersed amorphous samples. The analysis of the XAS spectrum of this sample is far from easy. Thus, although the XANES region of titanium oxides (anatase and rutile) has been the object o f several experimental and theoretical studies (refs. 9-14) the definitive explanation of all the features appearing in this region has not been given yet. Nevertheless, by comparing it with the spectra of previously studied compounds, we can use this region of the spectrum as a finger print.
636
Thus, from the comparison with the Ti k-edge XANES spectra of several alcoholates previously measured (ref. 1 4 ) , we can discard the presence of tetrahedral or square planar geometry around the Ti centers, as well as the long range order typical of crystalline structures, like anatase, rutile or brookite (refs. 12,13), confirming in this way the conclusions reached by XRD. The EXAFS results point to the existence of a phase similar to anatase but, since the distances Ti-Ti2 and Ti-Ti3 are missing, the coordination numbers for Ti-Til, Ti-02 and Ti-03 are very small, and there is a new distance Ti-04 above 5A, it seems that the new structure is more open and has grown in two dimensions. The parameters obtained are compatible with a structure similar to the Ti02-B, proposed by Brohan et al. (ref.15) and more recently by Reichmann and Bell (ref.16) as a precursor o f anatase in the decomposition of TiC14. In conclusion, incipient impregnation of SiOz with a n-hexane solution of Ti(OPr')4 leads to TiOz coated material with an extremely high dispersion where very small clusters of Ti06 octahedra (probably 3-4 octahedra sharing edges and corners) are formed. The process involves hydrolysis by physisorbed/chemisorbed water followed by anchoring during calcination. ACKNOWLEDGEMENTS. The authors wish to thank Prof. D.C.Koningsberger for the use o f his EXAFS analysis programs, CICYT and Junta de Andalucia for financial support, and the staff in the SRS (Daresbury lab., SERC) for help during the XAS measurements. REFERENCES i
G.C Bond and R.Burch, Catalysis (Specialist Periodical Report).Chem.Soc.,
6 (1983) 27-60.
2 K.Foger. Catalysis,Science and Technol. 6 (1984) 227-305. 3 M.G.Reichmann and A.T.Bel1, Appl.Catal., 32 (1987) 315-326. 4 C.Morrison and J.Kiwi, J.Chem.Soc.,Faraday Trans.1, 85(5) (1989) 1043-1048. 5 J.B.Peri and A.Hensley, J.Phys.Chem., 72 (1968) 2926 6 J.B.A.D van Zon, D.C.Koningsberger, H.F.J. van't Blik, and D.E.Sayers, J.Chem.Phys., 82 (1985) 5742-5754. 7 J.B.Peri, Catalysis,Science and Technol., 5 (1984) 171-220. 8 E.T.C. Vogt, M.de Boer, A.J. van Dillen, and J.W.Geus, Appl. Catal., 40 (1988) 255-275.
M.Emili, L.Incoccia, S.Mobilio, G. Fagherazzi, and M.Guglielmi, J.Non Crys.Solids, 74 (1985) 129-146. 10 L.A.Grunes, Phys. Rev. 8, 27(4) (1983) 2111-2131. 11 A.Leaustic, F.Babonneau and J.Livage, Chem. Mat., 1 (1989) 248-252. 12 R.Kozlowski, R.F.Pettifer, J.M.Thomas, J.Phys.Chem., 87 (1983) 5172-5176. 13 G.A.Waychunas, J.de Physique Colloque C8, 47(12) (1986) 841-844. 14.F.Babonneau, S.Doeuff, A.Leaustic, C.Sanchez C.Cartier, and M.Verdaguer, Inorg.Chem., 27 (1987) 3166-3172. 15 L.Brohan, A.Verbaere, M.Tourneaux and G.Demazeau, Mat.Res.Bul1 .,
9
16
17 (1982) 355.
M.G.Reichmann and A.T.Bel1, Langmuir, 3 (1987) 111-116.
INFLUENCE OF SURFACE OH GROUPS AND TRACES DURING THE PREPARATION OF Ti02-Si02 SAMPLES
OF
627
WATER VAPOR
A. MUBOZ-PAEZ and G. MUNUERA
Dept. of Inorganic Chemistry and Instituto de Ciencia de Materiales (UNSE-CSIC) P.0.Box 1115, 41071 Sevilla SPAIN SUMMARY
TiOz-SiOz samples have been prepared by impregnation of silica support with n-hexane solutions of titanium alcoholate controlling the hydration/ hydroxylation degree of the silica surface. Once calcined, the samples were characterized by IR, XRD, SEM/EDAX and XAS. The mechanism proposed for the decomposition of the alcoholate involves reaction with adsorbed molecular water in a first step, followed by anchoring by reaction with acid OH- groups. The amorphoustitaniumoxide obtained after calcination shows a layered open structure of clusters formed by a few octahedra sharing edges and corners. INTRODUCTION Titania has been widely used as support in metal catalyst due to its ability to modify the catalytic properties of the metal (ref. 1). As a consequence, the study of the interactions taking place at the metal-titania interface has attracted the interest of several research groups (ref. 2). Nevertheless it is very difficult to obtain high surface area titania (>lo0 m2/g), and while studying the metal-titania interactions it is difficult to get information from the support because the bulk properties of the Ti02 mask those of the surface, the unique part of the titaniaaffectedby the metal. To overcome both problems, inert oxides like silica, have been used as support to obtain high surface area dispersed titania, by grafting to the SiO, support (refs. 3,4)throughthe impregnation from n-hexane solutions of Ti alcoxides, that by hydrolysis and calcination would produce the final coated Ti02-Si02 powders. EXPERIMENTAL Preparation af catalysts The surface oxide was prepared by impregnation of silica Aerosil-200 (SBE,=200 m2/g), with a n-hexane solution of a Ti-alcoholate (tetraisopropyltitanate, Ti(OPr’)4 from Tilcom, 16.9% Ti). The TiOz percentage (by weight) used has been c.a. 12%. This value correspondsroughly to the amount required to form a monolayer of titania on this type of silica (14.7%, 5.5 Ti/nm2) and
628
is close to the amount needed to allow the grafting of each Ti atom to one hydroxyl group of the silica surface (13 %, 5 OH/nm2) (ref. 5). In the standard procedure, the desired amount o f Ti(OPri)4 was dissolved in dried n-hexane (25 ml/g of silica in Methods 1-3 and 6 ml/g of silica in Method 4) and the solution was allowed to react with the surface of the silica for several hours. After that, the solvent was removed at room temperature by flowing Nz, and the sample heated in N 2 up to 673K. Subsequently the solids at 873K. Four differents methods were used as were calcined i n air follows: Method 1. Reaction under Nz at 300K for 20h of undried SiOz. Method 2. Reaction under Nz at 300K for 20h of Si02 dried at 388K for 2h. Method 3 . Reaction under Nz at 350K for 5h of SiOz dried at 440K for 2h, cooling down to room temperature, filtering and washing with n-hexane and subsequently with water. Method 4. Incipient impregnation at 300K in the air of undried silica with a n-hexane solution of Ti Tests were made along the preparation (Methods 1 to 3) by gas chromatography to detect i-PrOH in the liquid phase as well as the presence of Ti(OPr’)4 by hydrolysis in aliquots of the liquid phase. Characterization of solids IR spectra were carried out at 300K on a wafer of the sample mounted in a cell that allows in situ thermal treatments under controlled atmospheres up to 773K, using a Perkin-Elmer 684 spectrometer fitted to a 3600 data station. X-ray diffractograms were recorded in a Phillips 1730 diffractometer and scanning electron micrographs using an IS1 microscope model SS-40, with an energy dispersive X-ray analyzer (EDAX) KEVEX, model 8000 fitted to it. XAS experiments were performed on the EXAFS station 8.1 in the Synchrotron Radiation Source at Daresbury Laboratories with ring energies of 2 GeV and ring currents of 250 mA. The EXAFS spectrum was recorded at 140K in an “in situ” cel1,wherethe sample was placed after being pressed with BN into a wafer with an absorbance ( p x ) of 2.5 at the Titanium K-edge assuring an optimum signal to noise ratio. Data analysis was carried out by fitting in kand R-space using the phase and amplitude corrected Fourier transforms to identify the different contributions (ref. 6). Phase shift functions and backscattering amplitudes were obtained from reference compounds. RESULTS Assuming that the two reactions taking place during the decomposition of the alcoholate to TiOZ are grafting through OH- groups on the surface of the silica and hydrolysis of the Ti-alcoholate by water to produce colloidal
629
particles (refs. 3 , 4 ) , the only competitor to decompose the alcoholate would be the water vapor from moisture. Thus, we have used several preparation methods in which moisture was carefully avoided. Therefore, in the first case (Method l ) , the grafting would involve only OH- groups and/or water molecules adsorbed on the silica surface. Taking into account that water physisorbed on the surface of the silica could produce mainly ungraphted titania, we have carried out a second preparation method (Method 2) in which adsorbed water was avoided by submitting the silica to a previous outgassing treatment at 388K that would remove at least physisorbed water. A new method (Method 3 ) was designed in which all molecular water was removed and, considering that the hydrolysis process could be very slow at room temperature in such extremely dry conditions, the reaction temperature was raised up to the boiling point of the n-hexane. In this case, after five hours of reaction the liquid phase, still containing Ti(OPr’)4, was filtered off and the sample was thoroughly washed with n-hexane, to remove the unreacted alcoholate, and then with water to get a complete hydrolysis of the grafted alcoholate. During the washing with water, formation of a thin, opaque white layer was clearly observed which, unlike the transparent silica, remained stuck on the surface o f the filter. This white coating, presumably TiOz, once dried, calcined and weighed,turn out to be c.a. 50% of the total amount of titanium oxide that should be formed by decomposition of all the alcoholate employed in this preparation. Control test during the preparation in Methods 1 and 2 showed a complete hydrolysis of Ti(OPri)4 at the end of the reaction time (20h), while in Method 3 the liquid still contained the alcoholate in spite of the more drastic thermal conditions used in this case. Finally, a fourth type of preparation was carried out consisting in the well known incipient impregnation method, using the n-hexane solution of Ti (OPr i , and the si 1 ica support without any drying pretreatment. In principle, the degree of success of the anchoring process in our preparations could be followed by checking the changes in the concentration o f one of the reactants (i.e.surface OH-/HzO at the silica support) using I R spectroscopy, since silica aerosil shows a characteristic sharp band at 3750 -1 cm due to basic free hydroxyls together with a broader band due to more acidic OH- groups at 3680 cm-’ (ref. 7). So, their reaction can be followed by changes in their intensities as previously observed in similar preparations (ref. 8). Thus, figure 1 shows IR spectra o f the SiOp support and of a sample prepared by method 1 containing only 1% Ti02 before submitting the samples to any thermal or outgassing treatment. The unique change observed after the addition of the alcoholate is the decrease of the intensity in the range
630 100.
b
a
Fig. 1. I R spectra in OH stretching the region of a sample l%TiOz-SiOz (a) and of si 1 ica support (b) registered in the atmosphere (solid line) and after outgassing at 673K for 2h.
%A
50 I
\
I
\
I
\
I
I 0
.
I
4000
'
1% Ti0 S i O
\,
2
\\,
cm
.% ' -1
2
I
3000
4000
cm
--I
3000
cm-', where the bands due to the more acidic OH- groups and/or molecular water appear, what suggests that these species are those mainly involved in the interaction of the alcoholate with the SiOz support. The IR spectra in the same figure, recorded after outgassing at 673K to remove the water readsorbed upon exposure to air, clearly show the decrease in the intensity of the band at 3680 cm-' assigned to the more acidic OH- groups of the silica, thus suggesting their participation in the decomposition of the alcoholate. cm-' of the samples Figure 2 shows I R spectra in the range 2900-3400 prepared with c.a. 12% Ti02 by the four methods (spectra have been normalized using u s i - o at 1830 cm-' from the bulk of the silica, to make them comparable). Except for the sample prepared by method 3 , in all other cases the intensity o f the band at 3750 cm -1 remains nearly unchanged with respect to that of the silica support pretreated under similar conditions, while changes in intensity and/or position are observed in the band at 3680 cm-' in all the preparation methods. The increase in intensity o f the IR bands in the 0-H stretching region in sample prepared by Method 3 is probably related with the final washing with water used in this method. oxide phases XRD was used to check the crystallinity of the titanium formed after calcination by decomposition of the hydrolyzed alcoholate. Only small shoulders appear in sample 1, 2 and 3 in the position o f the most 3700-3550
63 1
h
8 b
0
m
I Ti0 S i O -4 2
0
-c---c--t-cm
+
-1
Fig. 2 . IR spectra in the range 3900-3400 cm-' of the silica support and the samples 12%TiOz-SiO2 prepared by the four methods outgassed at 673K for 2h. intense peak of anatase, while no peaks were visible at the positions of the most intense diffraction lines of anatase, rutile or brookite in sample 4 . Nevertheless, when the alcoholate was hydrolyzed with water in the absence of silica and calcined under similar conditions, strong peaks appear in the positions of the most intense diffractions of anatase, thus indicating that the hydrolysis o f the pure alcoholate produces crystalline phases. Analysis o f the samples using SEM/EDAX was carried out to examine thehomogeneityof the titanium distribution on the TiOz-Si02 samples, and the homogeneity in grain shape and size. Thus, in sample prepared by Method 2, (using dried Si02) the grains have angular shapes and the local concentration of Ti changes drastically when going from one grain to another. The changes are less drastic, although still remarkable in sample 1, that shows round grains. In sample 3 the particle size was bigger than in the other cases, and the existence of different types of particles (opaque and transparent) could be seen without the aid of the microscope. The most homogeneous sample, considering grain shape and size, as well as titanium dispersion was sample 4 , that has a very homogeneous spongy appearance with constant concentration of T i in all the grains. From the previous results, we deduced that method 4 is the best one, so this sample was studied by XAS to get a deeper insight into the structure around Ti ions. The XANES region of this sample has been plotted in figure 3 , where the corresponding spectra o f anatase and rutile, measured as a
632 C
D 0.1
h
' " 0 Lr
-0.1
J I
-20
I
I
0
1
E(eV) 20
I
40
Fig. 3 . Ti k-edge XANES spectra of TiOz rutile (a), TiOz anatase (b) 12%TiOz-SiOz prepared by method 4. Fig. 4. Ti k-edge, Fourier transforp of the EXAFS syectrum of sample lZ%TiOz-SiOz prepared by method 4. (k , Ak=3.12-11.00 A- ) . Arrows indicate the ranges for Fourier filtering used during the data analysis. reference, have been included as well. In addition to the round shape of features C and D, typical of amorphous compounds (ref. 9), it has to be pointed out the appearance of the triplet A,,A2,A3 characterisitic of octahedral symmetry (ref. 10) that indicates that the absorbing atom is six fold coordinated. Nevertheless, there is a remarkable change in the intensity ratio between peaks AZ and AB, that is close to 1 in anatase or rutile and close to 0.5 in sample 4 where it shows a shape similar to the spectra of uncalcined TiOz colloids prepared from hydrolyzed Ti(OPr')4 (ref. 11). A similar shape has been observed in the spectra of titania-silica glasses prepared by gelation in air of Ti and Si alcoxides by Emili et a1 (ref. 9 ) , who have assigned it to the existence of Ti ions in tetrahedral environment. The Fourier Transform o f the EXAFS signal yields the radial distribution function shown in figure 4, where we can see an intense peak at around 1.7 A due to backscattering from the first shell of oxygen atoms. For higher distances there is a drop in intensity that, in principle, could be assigned to the lack of higher coordination shells. Nevertheless no good fit could be obtained with only one or two shells. So, we have performed the data analysis shell by shell, doing inverse Fourier Transforms of increasing ranges, shown
633
TABLE 1
1
TiOz-Si02-4 Shel 1
N R(A) -
I
Anatase ~
Ao2(AZ)
5.8
1.93
0.011
1.0 6.7
3.09 3.78
-0.005 0.03
7.2
4.37
0.06
3.3
5.34
0.00
Shel 1
NxR(A)
Ti-Ol
4x1.93 2x1.98
Ti-Ti Ti-02 Ti-Ti2 Ti-03
4x3.04 8x3.86 4~3.78 8x4.25 8x4.27 4x4.75
Ti-Ti3
ax4. 85
number octahedr lS'
qrd
by the arrows in figure 4. We started the analysis considering the basic octahedra of anatase Ti06 RTi-O= 1.95 A (fit range 0.16- 2.3 A). Afterwards, we expanded the range up to 3 A, and included a shell Ti-Ti. When the fitting range was expanded to 4.1 A , two new Ti-0 bonds were required to reach a good fit. Finally, to reach the final values the range for the Fourier filtering was 0.16-5.4 A requiring the inclusion of a new Ti-0 bond at 5.3 A . The parameters of this fit are summarized in Table 1, that includes the number of the neighboring atoms, N, the absorbing atom-neighbor distance, R, Debye-Waller factor, Ao', related to static and thermal disorder, as well as the structural parameters of crystalline anatase appearing in a cluster of 4 octahedra (ref. 12). A plot of the raw data and the best fit in k and R space for the wider range has been included in Fig 5. The first peak in the Fourier transform may be attributed to the six Ti-Ol bonds of the basic octahedron, as already predicted from the XANES data. The distance is the same that the short bond of the distorted octahedra in anatase. The peaks between 3 and 6 A are a complex result of the overlap of four different features. The first one, Ti-Ti at 3.09 A, is very similar to the distance observed in anatase between two octahedra sharing edges (3.04 A), while the next one, Ti-02 at 3.78 A, is very close to the distance of the oxygen atoms in the second octahedron (3.86 A). The shell Ti-03 would correspond to oxygen atoms in a third octahedron in an anatase-like structure. The shell Ti-04 has no correspondance in a cluster of anatase structure including four octahedra. In relation with the similarities with the anatase strucutre in the other four shell, it has to be pointed out the low coordination number of the Ti-Ti bond at 3.09 A , as well as the lack
634
Fig. 5 . Ti k-edge EXAFS spectrum and Fourier transform (kl, Ak=3.5-10.5 A-') of the raw data (solid line) and best fit (dotted line) of sample 12%TiOz-Sioz prepared by method 4. of Ti-Ti bonds for higher shells. Both facts indicate that only small clusters of TiOs octahedra are present on the SiOz support. DISCUSSION Formation in our conditions of colloidal particles of Ti02 grafted to the high surface area SiOz can be assumed t o occur according to one of the two following schemes: Scheme 1
Ti (OR)
+
(-Si-O)n-Ti(OR)4-n
-
Si-OHb __ > (-Si-O)n-Ti(OR)4-n+
+
(4-n)HzO
n ROH
>(-Si-O)n-Ti(OH)4-n+(4-n)ROH
(1) (2)
Scheme 2
Ti (OR) -
Si-OHa
+ +
4 HZOads
Ti(OH)4
------> Ti (OH)4 + > Si-O-Ti(OH)3
4
ROH
(3)
+ H20
(4)
where -Si-OHa and -Si-OHb stand for basic and acid OH- groups at the SiOz surface, Ti (OR)4 for Ti(OPri)4 monomers and HZOads for physisorbed/chemisorbed water. In the first case, grafting should involve in a first step the more basic OH- groups of the silica through a hydrophylic attack, and in a second step hydrolysis by reaction with adsorbed water or moisture. According to scheme 2, hydrolysis of the alcoholate by adsorbed water at the SiOz support is postulated, leading t o T i hydroxide colloidal particles in a first step, which must be followed by anchoring to the SiOz surface through reaction with more acidic OH- groups, a process that should be enhanced by the final thermal treatment during the calcination used in the preparation of the samples.
635
IR data in figures1 and 2 suggest that Scheme 2 (hydrolysis by adsorbed water followed by grafting) is the most likely in the conditions used in our preparative work, since the band at 3750 cm-’, due to more basic OH- groups, is not modified during the whole process. Moreover, changes in the band at 3680 cm-’ due to more acidic hydroxyls, can be explained by assuming that the grafting involves this type of hydroxyls of the silica surface. It is worth noting that preparation by Method 3 , where adsorbed water and probably part of the acidic OH- groups have been removed from the S i 0 2 support before reaction, only allows ca. 50% reaction of the Ti-(OPr1)4 in spite of the presence of all the basic OH- groups. This fact again excludes these groups from the process (reaction (1)). In fact, hydrolysis of the alcoholate remaining at the SiO2 surface in this case only occurs by washing with water what probably also produces breaking of siloxane bridges at the Si02 surface, (partially dehydroxylated) as detected by the much larger intensities of the IR bands for this sample in figure 2. If we assume Scheme 2 , grafted colloidal titania particles, similar to those obtained from simple hydrolysis of Ti-(OPr’)4 with water, should be obtained and therefore their structure should not be very different from that recently proposed by Leaustic et a1 (ref.11). In fact, the XANES spectrum of our sample is very similar to the spectrum recorded by these authors for such colloidal particles. However, there are big differences in the EXAFS region that can be explained by the smaller size of the titania particles obtained in our system. Moreover, after heating at 373K, these authors obtain crystalline anatase, as previously did Kozlowski et al.(ref. 12) and Reichmann et a1 (ref. 3 ) during the preparation of similar systems, while the crystalline structure of anatase could not be detected by XRD in our samples, even after calcination at 873K thus implying that the layered open structure, remains stabilized on the surface of the silica. It is not surprising that the best preparation method for this type of ultradispersed Ti02-SiOz systems was the incipient impregnation, since in this conditions the lack o f an excess of solvent will probably prevent the growth of the original nuclei t o bigger colloidal particles. Additionally, this method has the advantage that it is the easiest and provides homogeneous and well dispersed amorphous samples. The analysis of the XAS spectrum of this sample is far from easy. Thus, although the XANES region of titanium oxides (anatase and rutile) has been the object o f several experimental and theoretical studies (refs. 9-14) the definitive explanation of all the features appearing in this region has not been given yet. Nevertheless, by comparing it with the spectra of previously studied compounds, we can use this region of the spectrum as a finger print.
636
Thus, from the comparison with the Ti k-edge XANES spectra of several alcoholates previously measured (ref. 1 4 ) , we can discard the presence of tetrahedral or square planar geometry around the Ti centers, as well as the long range order typical of crystalline structures, like anatase, rutile or brookite (refs. 12,13), confirming in this way the conclusions reached by XRD. The EXAFS results point to the existence of a phase similar to anatase but, since the distances Ti-Ti2 and Ti-Ti3 are missing, the coordination numbers for Ti-Til, Ti-02 and Ti-03 are very small, and there is a new distance Ti-04 above 5A, it seems that the new structure is more open and has grown in two dimensions. The parameters obtained are compatible with a structure similar to the Ti02-B, proposed by Brohan et al. (ref.15) and more recently by Reichmann and Bell (ref.16) as a precursor o f anatase in the decomposition of TiC14. In conclusion, incipient impregnation of SiOz with a n-hexane solution of Ti(OPr')4 leads to TiOz coated material with an extremely high dispersion where very small clusters of Ti06 octahedra (probably 3-4 octahedra sharing edges and corners) are formed. The process involves hydrolysis by physisorbed/chemisorbed water followed by anchoring during calcination. ACKNOWLEDGEMENTS. The authors wish to thank Prof. D.C.Koningsberger for the use o f his EXAFS analysis programs, CICYT and Junta de Andalucia for financial support, and the staff in the SRS (Daresbury lab., SERC) for help during the XAS measurements. REFERENCES i
G.C Bond and R.Burch, Catalysis (Specialist Periodical Report).Chem.Soc.,
6 (1983) 27-60.
2 K.Foger. Catalysis,Science and Technol. 6 (1984) 227-305. 3 M.G.Reichmann and A.T.Bel1, Appl.Catal., 32 (1987) 315-326. 4 C.Morrison and J.Kiwi, J.Chem.Soc.,Faraday Trans.1, 85(5) (1989) 1043-1048. 5 J.B.Peri and A.Hensley, J.Phys.Chem., 72 (1968) 2926 6 J.B.A.D van Zon, D.C.Koningsberger, H.F.J. van't Blik, and D.E.Sayers, J.Chem.Phys., 82 (1985) 5742-5754. 7 J.B.Peri, Catalysis,Science and Technol., 5 (1984) 171-220. 8 E.T.C. Vogt, M.de Boer, A.J. van Dillen, and J.W.Geus, Appl. Catal., 40 (1988) 255-275.
M.Emili, L.Incoccia, S.Mobilio, G. Fagherazzi, and M.Guglielmi, J.Non Crys.Solids, 74 (1985) 129-146. 10 L.A.Grunes, Phys. Rev. 8, 27(4) (1983) 2111-2131. 11 A.Leaustic, F.Babonneau and J.Livage, Chem. Mat., 1 (1989) 248-252. 12 R.Kozlowski, R.F.Pettifer, J.M.Thomas, J.Phys.Chem., 87 (1983) 5172-5176. 13 G.A.Waychunas, J.de Physique Colloque C8, 47(12) (1986) 841-844. 14.F.Babonneau, S.Doeuff, A.Leaustic, C.Sanchez C.Cartier, and M.Verdaguer, Inorg.Chem., 27 (1987) 3166-3172. 15 L.Brohan, A.Verbaere, M.Tourneaux and G.Demazeau, Mat.Res.Bul1 .,
9
16
17 (1982) 355.
M.G.Reichmann and A.T.Bel1, Langmuir, 3 (1987) 111-116.