J. Phys. Chem. 1993,97, 5085-5094
5085
An IR and NMR Study of the Chemisorption of T i c 4 on Silica Suvi Haukka,'*+** Eeva-Liisa Lakomaa,+ and Andrew Roots Microchemistry Ltd., P.O. Box 45, SF-021 51 Espoo, Finland, Department of Chemistry, Analytical Chemistry Division, University of Helsinki, Vuorikatu 20, 00100 Helsinki, Finland, and Scientific Services, Analytical Research, Neste Oy,P.O. Box 310, SF-06101 Porvoo, Finland Received: October 13, 1992; In Final Form: January 26, 1993
The atomic layer epitaxy (ALE) reactions, Le., the saturating gas-solid reactions of Tic14 a t 175 and 450 OC with silica preheated a t 200-820 OC, were studied by nuclear magnetic resonance and Fourier transform infrared spectroscopy, while etching experiments with sulfuric acid were carried out to determine the amount of amorphous titanium species. At 175 OC Tic14 reacted directly with the O H groups of silica either monofunctionally or bifunctionally depending on the preheat temperature of silica. A bifunctional reaction took place with strongly H-bonded O H groups as verified by using hexamethyldisilazane (HMDS) to prevent Tic14 from reacting with isolated O H groups. The titanium species bound at 175 OC was amorphous in nature, and anatase-like Ti-OH groups formed upon water treatment. At 450 OC, agglomeration into Ti02 having both anatase and rutile crystal structures occurred, together with the formation of amorphous titanium species. A constant Cl/Ti ratio of 2 was obtained independent of the preheat temperature of silica, which in this case, because of the agglomeration, did not indicate a bifunctional reaction. After water treatment only Ti-OH due to amorphous titanium species was observed. The agglomeration into Ti02 is proposed to be due either to the interaction of HCl with the O H groups of silica or direct chlorination of the OH groups with TiC14.
Introduction The gas-phase reaction of TiC14with hydroxyl groups of silica has been studied by several groups,ld mainly by infrared spectroscopy and chemical analysis. IR studies have not shown conclusively whether it is the isolated or the H-bonded hydroxyl groups that are more reactive toward TiC14. Kunawicz et a1.I observed that Tic14 reacts with H-bonded OH groups at room temperature,whereas Armistead et a1.2proposed that the reaction of Tic14 at room temperature occurs with all the hydroxyl groups of silica. Damyanov et aL3observed a simultaneous decrease in intensity of the IR bands of isolated and H-bonded OH groups after exposure to Tic14 at 150 OC. In recent studies, using Raman and FTIR spectroscopy, Morrow and McFarlan4concluded that at 20 OC T i c 4 reacts with both of these groups and a bifunctional reaction occurs with H-bonded OH groups. On the basis of studies by infrared photoacoustic spectroscopy Kinney and Staley5 proposed the reaction of TiC14with strained siloxane bridges, the validity of which was later questioned by Ellestad and Blindheima6 Thus, the possible reactions of TiC14 with the silica surface are the following: Si-OH Si-OH
+ Tic4
+ Tic4 -Si-OH
Si-0-Si
+ Tic4
-D
4
Si-O-TiCI3 S i 'TiCI,
/
+
HCI
monofunctional
(1)
+
2HCI
blundional
(2)
+
Si-CI
wlhsiloxane
(3)
S i
Si-O-TCIa
The reaction mechanism has been elucidated by the Cl/Ti ratio, which Damyanov et al.3 found in a review of the literature to range from 1 to 3.4. No correlation was observed between the preheat temperature of silica, which determines the number of OH groups, and the CI/Ti ratio. However, the use of the Cl/Ti ratio has been criticized due to the possible interaction of hydroxyl groups with the HCl evolved during the main r e a ~ t i o n .The ~ common feature of all the studies on this mechanism was the low reaction temperature used to bind titanium to silica. The reaction was carried out under 200 OC and most often at room temperature. Microchemistry Ld. of Helsinki. s Neste Oy. +
1 University
Recently Kooyman et al.' reported the results from treating the silica with Tic14 at 450 "C. Upon examination of the Ti/silica samples using IR spectroscopy and XRD, no changes could be detectedafter theTiC14treatment, but SEM micrographs showed particles to be present. Besides Tic&,other metal chlorides (SiCl4,SnCL, GeCl4, CC14, AlC13,BC13)have been used as a means of determiningthe surface structure and the number of isolated and H-bonded OH groups of silica.8-12 Reaction temperatures up to 800 OC have been used. In no case, however, was chlorination of the silica surface by HCl observed; instead, there appeared to be direct chlorination by the metal chlorides in which chloride, but not the metal, was attached to silica. On hydrolysis with water vapor some of the OH groups reappeared, but no IR peaks attributable to M-OH wereobsend8 TheCl/metal ratioor theCl/A(OH) replacement ratio has been used to evaluate the reaction mechanism of these other metal chlorides. For tetrachlorides, values above 3 for both ratios have been proposed to indicate a strongsiloxanecontribution while values lower than 3 indicate hydroxyls having very close neighbors. There was a difference in the reactivity of metal chloridestoward the siloxane bridges. For example, no evidence for the reaction of Sic14 with siloxanes was found, but depending on the reaction temperature, C C 4 reacted with the OH groups or siloxanes or both. By studyingthe reaction of S i c 4 with silica, Peri and HensleyI2proposed that some OH groups exist as pairs even after heat treatment of silica at 800 "C. Contradictory results have been published by McDaniel? however. Information concerning the nature of the OH groups on silica can be also be obtained using 29SiCPMAS (cross-polarization magic angle spinning) NMR.13 Here the different silicon sites on the surface can be identified. In general, three peaks are observed at -91, -101, and -111 ppm corresponding to 02Si(OH)2,03SiOH,and Si04 groups, respectively. This technique has been used to study both the morphology14of silica and the dehydration/rehydration of silica.Is From the latter study, it has been shown that so-called geminal or paired groups like 0 2 Si(OH)2 can be present on silica heated to 800 OC. Complementary information about the OH groups on silica can also be obtained by using solid-state IH NMR. IH CRAMPS (combined rotation and multiple pulse spectroscopy)l6Sl7studies have
0022-3654/93/2097-5085%04.00/0 0 1993 American Chemical Society
5086 The Journal of Physical Chemistry, Vol. 97, No. 19, 19'93
revealed'8,'9 that the f H NMR spectrum consists of two basic areas: one broad peak due to H-bonded OH groups, and one narrower peak due to isolated OH groups. A subsequent comparison of the CRAMPS method, with MAS only, showed20 that the spectra obtained are essentially the same and in most cases the less demanding MAS experiment is sufficient to obtain acceptable resolution. Catalyst systems can also be investigated*] with IH MAS NMR when the surface species is not paramagnetic, and in favorable circumstances the metal-OH groups can be identified. Atomic layer epitaxy (ALE)22is a method for growing thin films or single crystals one atomic layer per reaction cycle. Each reaction cycle consists of a saturated surface reaction in which chemical bonds are formed between a gaseous reactant and the surface. Physisorption is avoided by elevated growth temperatures. In our previous paper2j the atomic layer growth of Ti02 on silica was studied at 175 and 450 OC. The formation of Ti02 clusters with both anatase and rutile crystal structure after a single reaction cycle of Tic14 and H 2 0at 450 "C was confirmed by X-ray diffraction. The reaction temperature of 175 OC led to the formation of amorphous TiOz. In the present paper we report more detailed information on the reactions of T i c 4 at 175 and 450 OC with silica preheated at 200-820 "C. In addition to FTIR spectroscopy, the reactions were also studied using solidstate NMR, a technique not previously used in this area. The blocking of isolated OH groups with hexamethyldisilazane (HMDS)24and etching of titanium with sulfuric acid are also reported, and the agglomeration mechanism is discussed.
Experimental Section Reagents. TiC14, used without further purification, was vaporized at 25 OC. Hexamethyldisilazane (HMDS) was vaporized at 50 OC. Silica (EP 10, Crosfield Ltd.) with an area of 300 m2/g, pore volume of 1.75 cm3/g, and mean particle size of 100 pm was used as support. Nitrogen (99.999%) was used as carrier gas in the ALE reactor. ALE Equipment. The reaction chamber consists of a quartz cup which can hold up to 10 g of silica. The reactant vapor is led in a flow of nitrogen through heated quartz tubes onto the silica bed. The reaction chamber is kept at 6-10 kPa, and the pumping takes place from the bottom of the silica bed supported on a sinter. The reaction chamber and the tubes leading the gases into the chamber are resistively heated. The heating and the flow of reactant and carrier gas are computer controlled. ALE Procedure. Silica (5-8 g) preheated at 200, 450, 560, 750, and 820 OC was used in T i c 4 reactions at 175 and 450 "C. Preheating of silica at 200 "C was carried out directly in the reaction chamber in a nitrogen flow at 6-10 kPa for 16 h. Preheatingof silica at 450,560,750, and 820 OC was first carried out in air in a muffle furnace for 16 h. Care was then taken to remove any water physisorbed during transfer from the muffle furnace by further heating in the reaction chamber at 450 OC for 4 h in a nitrogen flow at 6-10 kPa. To avoid changes in the silica during the exposure to TiC14,the reaction temperature was never allowed to exceed the pretreatment temperature of the silica. An excess of TiC14was led through the silica bed during 2 h to ensure surface saturation. To confirm that surface saturation had been achieved, the Ti concentration at the surface was compared with that in the bottom layers of the silica bed. Each reaction was followed by a nitrogen purge for 2 h at the same temperature as the Ti-binding reaction to remove unreacted Tic14 and the released HCl from the silica bed. The HMDS reaction with silica preheated at 200 and 450 OC was carried out at 175 OC. HCl reaction was carried out at 175,300,350,450, and 550 OC. The sample was cooled in a nitrogen flow and transferred inertly to NMR and IR measurement sample holders. Element Determinations and Etching Tests. Titanium was determined by UV/vis spectrophotometry after quantitative
Haukka et al. release of the titanium species from the silica, and by instrumental neutron activation analysis (INAA) as described by Haukka and Saastamoinen.25 Samples for chloride determinations were weighed into polyethylene vials and 10 mL of 3 M H2S04 was added immediately to avoid any volatilization of chloride as HCl. Potentiometric titration with AgNOj was used to quantify the chloride. Carbon was determined with a Leco carbon analyzer. To study the strength of the bonding between the silica surface and titanium, two different etching tests were carried out on each water-treated or non-water-treated sample. The number of titanium atoms released with 3.5 mol dm-3 sulfuric acid without decomposition of silica was measured and compared with the total number of titanium atoms. FMR Measurements. An FTIR spectrometer (Galaxy Series 6020) equipped with a diffuse reflectance accessory was installed in a glovebox, into which the samples were inertly transferred. Burneau et a1.Z6have shown that silica spectra obtained by diffuse reflectance of silica powders and by transmission through selfsupported silica disks are in general agreement. No sample preparation was carried out and the spectra were recorded directly from the surface using a spectral resolution of 2 cm-I. The accumulation time was about 4.5 min, corresponding to 1000 scans. Spectra were presented in a diffuse transmittance format. NMR Measurements. All measurementswere carried out using a JEOL GSX 270-MHz NMR spectrometer. The probe used was home built with a low IH NMR background signal. In the IH NMR experiments, 100 transients were acquired using a 2-ps (45O) pulse, 10-s recycle delay, and a 50-kHz spectral width. To ensure that the 'H spin system had reached equilibrium after a recycle delay of 10 s, the experiment was repeated for the least and most densely populated proton systems with a recycle delay of 20 s. There was no detectable signal increase for the populated system, the 150 OC dried EP10, and an increase of 9% for the least populated 820 OC treated EP10. It was concluded that the accuracy at low pretreatment temperatures is f 5 % while at the higher temperatures the proton population may be slightly underestimated by 10%. All experiments were carried out using MAS at a spinning speed in excess of 6 kHz with dry nitrogen as the drive and bearing gas. In all cases a background signal, run under identicalconditions with an empty rotor, was subtracted from the spectra. Samples were dried at 180 OC under vacuum (CO.1 Pa) for at least 8 h before measurement and were thereafter handled in dry nitrogen. The samples were loaded into 7-mm 0.d. zirconia rotors with push-on caps. Leaving a sample in the probe overnight and running the spectrum again did not give any detectable increase in signal intensity within the limits of experimental accuracy (is%). The number of OH groups/g of the dried, uncalcined EPlO was determined from a weight loss experiment in which the amount of water lost on heating under vacuum T) at 180 OC for 12 h was determined. The intensity loss in the NMR spectra of a carefully weighed sample after heating could then be calibrated to a known number of protons in the sample before heating. This was then used to calculate the number of protons, and therefore OH groups, in the dried sample. The results of this calibration gave a value for dried EPlO of 19.5 X lOZOOH/g or 6.5 OH/nm2 (S = 300 m2/g). In the subsequent experiments, the intensity of the NMR signal was measured relative to this reference dried EPlO sample and the number of OH groups calculated from this. It has been shownI8that the IH MAS NMR spectrum of silica generally consists of two peaks: one broad peak, due to the H-bonded O H groups, and one sharper peak, due to the so-called isolated OH groups. Although the residual MAS line shape can be ~alculated2~ for an inhomogeneous dipolar interaction, this is more difficult for a homogeneous dipolar interaction such as in this case. Therefore,the lineshapes used for simulating the spectra were derived empirically from the best fit to the data. From the variable-temperature IH MAS NMR spectra shown below it
Chemisorption of T i c 4 on Silica
The Journal of Physical Chemistry, Vol. 97, No. 19, 1993 5087
ISOLATED I 9 OH/nm'
uy ,1,bL
1000
I I
I ,I I I I 1 I 1
am
I
I I 1 1 I L I "
6.00
I U"" I
400
''
2.00
"""
' !'
' I "
OM
1
' a
!'
I"'
-200
t '
+
I LY
-4.00
PPM
Figure 1. 'HMAS NMR spectrum of untreated, dried silica EPlO.
appears that there are at least two contributions to the broader peak since one almost disappears when heated to 450 OC. Therefore, this broad asymmetric resonance was simulated with two peaks. The best fit, in terms of the residual of the sum of the squaresand the minimum number of fitted peaks, was obtained by simulating the spectra using two broad Gaussian peaks of width 1100 and 530 Hz and two narrower Gaussian peaks of width 200 and 90 Hz as shown in Figure 1. Fitting with Lorentzian line shapes for some or all of the peaks always gave a poorer fit. The assignment of these four components to physically significant species on the surface is somewhat arbitrary since there are probably varying degrees of hydrogen bonding on the surface of the silica. There may even be strongly hydrogen-bonded water molecules within micropores on the surface, and it is tempting to assign the very broad peak to these. However, in this paper the two broader peaks are assigned to the hydrogen-bonded OH groups, while the two narrower are assigned to the isolated OH groups. From these simulations the total number of OH groups can then be divided into those due to hydrogen-bonded groups and those due to isolated groups as shown in Figure 1. 29Si CPMAS (cross-polarization magic angle spinning) experiments were carried out on the samples without drying using a 5-ms contact time (50-kHz 29Si and IH radio frequency fields), 5-s recycle delay, and between 20 000 and 50 000 transients.
4000
3750 3500
3250
3000
Wavenumberslcm '
Figure 2. FTIR spectra of silica EPlO after heat treatment at the indicated temperatures. TOTAL OH/n"
SIMULATION
EXPERIMENTAL
R€?3dtS
Heat Treatment of Silica. Because of the surface controlling nature of ALE the starting surface of the heat-treated silica must be well characterized. To be able to detail the changes on the silica surface during heat treatment, both FTIR and IH NMR spectra were recorded after five heat treatments of 200,450,560, 750, and 820 OC (Figures 2 and 3). From the IR spectra the decrease in the number of OH groups between 200 and 450 OC seems to be due to the removal of strongly H-bonded OH groups, seen in the disappearance of the peak at 3500-3550 cm-1. The remaining broad peak at 3650-3680 cm-I belongs to the weakly H-bonded OH groups and the narrower peak at about 3743 cm-' to the isolated OH groups.I.2,28 In some studies2328 the peak at 3650-3680 cm-' has been attributed to inaccessible internal or bulk hydroxyl groups. The same general behavior is seen in the IH MAS NMR spectra where the total number of OH groups drops from 6.5 to 4.l/nm2. In the simulated spectra the broader H-bondedcomponent is almostcompletely lost, and it is tempting to assign this to the same strongly H-bonded OH groups as in the IR spectra. Heat treatment above 450 OC initiates the removal
10
8
6
4
2
PPM
0 -2 -4
8
6
4
2 0 - 2 4
PPM
Figure 3. 'HMAS NMR spectra of silica EPlO treated at the indicated temperatures in air.
of isolated OH groups and further loss in the hydrogen-bonded OH groups until at 750 O C the IH MAS N M R show that essentially all the OH are isolated. The IR spectra show the same behavior, and the wavenumber of the peak maximum of the isolated OH groups shifts from 3741 to 3746 with increasing preheat temperature from 200 to 820 'C, in agreement with previous studies,29JO and showed the increasingly isolated character of the OH groups of silica preheated at elevated temperatures. The same phenomenon is seen in the IH MAS NMR spectra where the narrower component of the isolated OH group, at the lowest frequency, becomes the dominant component at the
5088 The Journal of Physical Chemistry, Vol. 97, No. 19, 19'93 0,SiOH
Haukka et al. 2.5
A
1 3.5 _ _- _ +- -_ -+ .-
...+'
- 2.5
SiOI
-A\ O,Si(OH),
3.0 0 '=
.-2
. 1.5
0.0
110
'
200
350
500
650
800
Preheat temperature of stlica/T
Figure 5. TiC14reaction at 175 OC with 200-820 OC silica: Ti/nmz and (+) Cl/Ti ratio.
n
I \
-J l " " l " " l " ~ ' " " ' " " ' 1
-80
-90
-100
-110
-120
-130
PPM
Figure4. 29SiCPMASspectra of silica EP10: (A) untreated, (B)treated at 450 OC, and (C) treated at 750 OC.
TABLE I: Numbers of Different OH Croups/nm2 of Silica EPlO Determined by 'H MAS NMR. heat treatment temperature ("C)
total no. of OH groups/nm2
isolated OH groups/nm2
H-bonded OH groups/nm2
200 450 560 750 820
6.5 4.1 (3.4) 2.1 (2.2) 1.1 (1.1) 1.1
1.9 2.0 (2.1) 1.6 (1.6) 1.1 (1.1) 1.1
4.6 2.1 (1.3) 0.5 (0.5)
0 The results shown in parentheses were obtained after further heat treatment at 450 OC in a nitrogen flow at 6-10 kPa prior to the measurement.
highest temperatures. However, even at the high pretreatment temperatures, some of the OH groups exist as pairs, or geminal OH groups 02Si(OH)2. This is shown in Figure 4 where the 29Si CPMAS NMR spectra reveal the presence of the geminal groups on silica preheated to 750 OC. This is in agreement with previous observations on Fisher S-157 silica gel.Is The numbers of different OH groups/nm2 in the heat-treated silica samples as measured by IH NMR are listed in Table I. The table shows two sets of values for the number of different OH groups of silica preheated at 450,560, and 750 OC. The values shown in parentheses were recorded from silica samples heat treated as explained in the Experimental Section. Thus, the samples were first heat t r e a t 4 in air in a muffle furnace and then further heat treated at 450 O O in a nitrogen flow at 6-10 W a for 3 h. The other values were obtained from samples heat treated in air in a muffle furnace and then under vacuum at 180 OC prior to the measurement. The biggest difference between the values is seen in silica samples heat treated at 450 OC. The extra heat treatment at 6-10 Wa at 450 OC in a nitrogen flow allowed further removal of H-bonded OH groups but the number of isolated OH groups stayed constant. Our values for total number of OH groups/nm2 measured by 'H MAS NMR are somewhat higher than values given by McDanie19using the CHjMgI method for silica Davison 952 (surfacearea 300g/m2). Thesedifferences probably arise because NMR is a bulk method, measuring all
(0) reacted
OH groups present, while the CH3MgI reaction only measures those OH groups able to react with the probe reagent. Thus internal or inaccessible OH groups may not be measured and, in areas of high OH group density, steric hindrance may inhibit the reaction of all exposed OH groups. The same was observed in a comparison with the values obtained by Zhuravlev31 by the deuterio-exchangemethod. Zhuravlev has stated that themethod of deuterio exchange is advantageous in that only surface OH groups enter into the reaction of isotopic exchange, and the structural water (or OH groups) inside the silica particles does not. Chlorination of Sica with HCI. The criticism of the use of the Cl/Ti ratio even after reaction of TiC14 with silica at temperatures below 200 OC encouraged us to study the possible interaction of HCl with the OH groups of silica. After the HCl reaction at 175 "C no chloridewas found at the surface, and after the HCl reaction at 300 and 350 OC with silica preheated at 300 and 560 OC, respectively, the chloride coverage was less than 0.1 Cl/nmz. The HCl reaction at 450 "C with silica preheated at 450,750, and 820 OC left0.45,0.6, and 0.6 Cl/nm2, respectively, on the surface. The number of OH groups after HC1 reaction at 450 OC with silica preheated at 820 OC was measured by IH MAS NMR to be 0.4 OH/nm2. This number together with the number of C1 atoms found corresponded closely with the total number of OH groups before reaction (1.1 OH/nm2). We thus conclude that the reaction occurs with the OH groups, not the siloxane bridges, since more OH groups would then be produced. Furthermore, the results for the reaction temperature of 450 "C indiqteda higher reactivity for the OH groupsof silica preheated at 750 and 820 "C than for the OH groups of silica preheated at 450 OC. Raising the reaction temperature up to 550 OC generated a surface concentration of 1.0 Cl/nm2 on silica preheated at 560 OC. The conditions prevailing during the HCl reaction were not, however, comparable toconditionsin which simultane+s reaction of Tic14 takes place and both the reactants are competing for the same reactive sites. Thus, the contribution of direct chlorination to the chloride concentration found after TIC4 reactions carried out at higher temperatures must also be coqidered. Reactionof TIC4 with Silica at 175 OC. The reaction of Tic14 with silica at 175 9C was studied as a function of the preheat temperatureof silica by determining the number of bound titanium atoms/nm2 and the Cl/Ti ratio. In this case the Cl/Ti ratio can be used since no reaction of HCl with silica is observed at 175 OC. AS shown in Figure 5, the number of titanium atoms bound tosilicadecreasedfrom2.0 to0.7/nmzwith theincreasein preheat temperature (20G820 OC), while the Cl/Ti ratio increased from 2.4 to 3.0, indicating that the bonding mode of titanium had changed. Furthermore, the numbers of titanium and chloride atoms found in the upper layers of the silica bed corresponded to the numbers found in the bottom layers, confirming that titanium was well distributed and that saturation of the surface had been achieved. From all these samples titanium could be
Chemisorption of TiC14 on Silica
The Journal of Physical Chemistry, Vol. 97, No. 19, I993 5089
4000
I
Y
I
3750
3500
3250
3000
Wavenumbers/cm" Figure 7. FTIR spectra of silica preheated at 450 "C: (A) original spectrum, (B) after treatment with Tic14 at 175 OC, and (C) after treatment with Tic14 at 450 OC.
4000 3750 3500 3250 3000
Wavenumberslcm" Figure 6. FTIR spectra of silica EPlO preheated at 200 OC: (A) original spectrum, (B) after treatment with Tic14 at 175 'C, and (C) after further water treatment at 200 OC.
quantitatively etched with sulfuric acid without decomposition of the silica.25 On 200 OC silica the results from three separate experiments gave a mean value of 2.0/nm2 for the number of bound titanium atoms and 2.4 for the Cl/Ti ratio. The 200 OC silica used in the Tic14 reactions at 175 OC could have been the most sensitive silica to the secondary hydrolysis, Le., the continuous release of water during the TiC14 reaction and the following nitrogen purge. However, the effect of this secondary hydrolysis would have been observed by IR and NMR measurements as the appearance of the peak due to Ti-OH. Since no Ti-OH peak appeared, the effect could be considered negligible. On 450 OC silica the number of bound titanium atoms varied between 1.4 and 1.2/nm2 and the corresponding Cl/Ti ratio between 2.4 and 3.0. On 560 "C silica the variation was between 1.2 and 1.l/nm2 for bound titanium atoms and between 2.8 and 3.0 for the corresponding Cl/Ti ratio. On 750 and 820 OC silica the values for the number of bound titanium atoms varied between 0.7 and 0.8/nm2 and the Cl/Ti ratio was constantly close to 3. In spite of the varying Cl/Ti ratio, these results indicate the decreasing ability of TiC14to react bifunctionally with silica as the preheat temperature increases. The peak of isolated OH groups was always seen after T i c 4 reaction at 175 OC, independent of the preheat temperature,indicating that not all the isolated OH groups were removed by TiC14. FTIR spectra of Ti/silica samples prepared from 200,450, and 560 OC silica are shown in Figures 6, 7, and 8, respectively. The peak of isolated OH groups was symmetric and sharp after the exposure of 750 and 820 OC silica to TiC14 at 175 OC, and it was also shifted to the higher wavenumber of 3747-3748 cm-I, relative to the corresponding wavenumber of 3746 cm-I for the original 820 O C silica. The shift was probably due to the decrease in the number of isolated OH groups. Thus, the OH groups present after Tic14 reaction at 175 OC can be considered as truly isolated. An FTIR spectrum recorded from a Ti/silica sample prepared from 750 'C silica has been presented in our previous paper.23 Again, the 1H MAS NMR spectra show the same feature, with mainly isolated OH groups left after reaction with TiC14. Panels A and B of Figure 9 show the IH MAS NMR spectra of 560 and 820 OC silica after titanium binding of 175 OC. The
4000
3750 3500
3250 3000
Wavenumbers/cm ' Figure 8. FTIR spectra of silica preheated at 560 OC: (A) original spectrum, (B)after treatment with TiC14at 175 OC,and (C)afterfurther water treatment at 200 OC.
number of OH groups left behind, as measured from the intensity of the peaks, was 0.7 and 0.3, respectively. Since the CI/Ti ratio is close to 3, we assume that the titanium reacted monofunctionally and the number of OH groups removed during the reaction of TiC14with 560 OC silica (1.2 OH/nm2), added to the number of OH groups left after titanium binding (0.7 OH/nm2), corresponded within the limits of experimental error to the number of OH groups of the untreated 560 OC silica (2.1 OH/nm2). The same was true for silica preheated at 820 OC: the number of OH groups occupied by titanium, 0.7/nm2, added to the 0.3 OH/nm* left after titanium binding, gave 1.OOH/nm2, which corresponded to the number of OH groups of the untreatd 820 OC silica (1.1 OH/nm2). From the agreement between these values, and the observation that no agglomerationinto Ti02 occurred at 175 OC, we can conclude that Tic14 had reacted according to the reaction given in eq 1. Thus, siloxane bridges could not have been involved in the reactions.
Haukka et al.
5090 The Journal of Physical Chemistry, Vol. 97, No. 19, 1993
TABLE E. Reproducibility of the Tim Reactions at 450 "C with S i l i c a Preheated at 450-820 "C'
iAi
OHIm'
1.0 1
5
4
"
"
1
"
"
1
'
"
3
'
1
2
'
'
~
'
I
I
+ . o--.---
0.8 .
'
0
.-
-
I-
PPM
\
____ o----.o _ . _-_ 2.0_ _ _ _ - - -
- 1.5
0.6
-
+'+
0
Figure 9. 'HMAS NMR spectra of (A) 560 "C silica and (B) 820 "C
0.4 . A .
silica after treatment with TIC4 at 175 "C, and (C) 560 OC silica after treatment with Tic14 at 450 "C.
0.2 .
+
.~-. A-----..---
-
1.0
. 0.5 - - _ _ A . . .-.A ~
0.0
Blocking with HMm. BlitzZ4 showed by using HMDS to prevent TiC14from reacting with isolated OH groups that Tic14 reacts bifunctionally with H-bonded OH groups. He carried out the reactionsin organic solution using silica preheated at 200 and 600 OC. The singly and doubly bonded surface titanium species were identified by variable-temperaturediffuse reflectanceFTIR spectroscopy. The use of HMDS is based on studies by various groups reviewed by Blitz,24which showed that HMDS reacts with isolated OH groups and slightly with H-bonded OH groups. The HMDS reacts only with those H-bonded OH groups that give rise to the high frequency end of the OH band (above 3650 cm-1).j2 Here this HMDS blocking was applied in the gas-phase reaction and the bifunctionality was studied by elemental determinations. Instead of using silica preheated at 600 OC, the silica preheated at 450 "C was used. As shown in Figure 5, the number of Ti atoms/nm2 bound to 200 "C silica in the absence of HMDS blocking was 2, and the Cl/Ti ratio was 2.4. Thus, approximately half of the Ti atoms must have reacted monofunctionally and the other half bifunctionally. In the corresponding FTIR spectrum (Figure 6B), a decrease in intensity of the band of stronglyH-bonded OH groups was observed together with a decrease in intensity of the band of isolated OH groups. In a first step silica preheated at 200 and 450 OC was treated only with HMDS. Carbon determination showed the number of OH groups occupied by HMDS to be the same for both preheat temperatures: 1.6 OH/nm2, which is less than the number of the isolated OH groups measured by lH MAS NMR. In a second step the HMDS and then TiC14 was led through the silica bed. After removal of unreacted HMDS, Tic14 was reacted with the modified silica. In both silica samples, the carbon determination showed the number of OH groups occupied by HMDS to be 1.4 OH groups/nm2 after the reaction with TiC14. The number of Ti atoms/nm2that reacted after the HMDS binding was 1.O and 0.1 for 200 and 450 OC silica, respectively. For 200 OC the number of bound titanium atoms after HMDS blocking was half of the number of the bound titanium atoms without the blocking. Furthermore, the Cl/Ti ratio was 2 after blocking, indicating that TiC14had reacted bifunctionally and therefore had occupied 2.0 and 0.2 OH groups of the 200 and 450 OC silica, respectively.
0.0
Figure 10. Tic14 reaction at 450 "C with 450-820 "C silica: (+) reacted Ti/nm2, (0) Cl/Ti ratio, and (A) Ti/nmz released with sulfuric acid.
Assuming that HMDS occupied all the possible accessible isolated OH groups, one can conclude that titanium reacted bifunctionally with strongly H-bonded OH groups,just as Blitz24and Morrow et al.4 suggested. This is supported by the observation in both lH NMR and IR that only the H-bonded OH groups are lost upon heating the silica from 200 to 450 OC, and therefore the smalleramount of titanium binding on 450 OC silica is presumably due to the fewer H-bonded OH groups. Reaction Temperature of 450 OC. Earlier we observed a formation of Ti02 clusters after one reaction cycle of Tic&and H20 at 450 "C on silica preheated at 450-820 0C.23 Anatase and rutile forms of Ti02 were identified by X-ray powder diffraction. Water-treated and non-water-treated samples have now been further investigated. To determine if the Ti02 agglomerates on the silica surface could have been due to the hydrolysis being carried out at a temperature of 450 OC, the Ti/silica samples reacted with Tic14 at 450 OC were now simply exposed to air moisture. The samples gave identical XRD spectra to the samples hydrolyzed at 450 OC. Thus, it seems reasonable to conclude that the tendency of titanium toward nucleation is strong at higher reaction temperature even without water treatment. The total number of titanium atoms bound in both the watertreated and non-water-treated samples decreased with the increasing preheat temperature (450-820 "C) from 1.1 to 0.5 (Table I1 and Figure 10). In the non-water-treated samples the complete removal of isolated hydroxyl groups at all preheat temperatures was verified by FTIR measurements. Only the weakly H-bonded OH groups were left unoccupied, as shown in Figure7C. The'H MASNMRspectrumof 560OCsilicareacted at 450 OC is shown in Figure 9C. The number of residual OH groups is small, 0.3/nm2, and gives rise to an asymmetric peak which can be simulated by one peak of 160 Hz and another of 350 Hz presumally due to isolated and weakly hydrogen-bonded
Chemisorption of Tic14 on Silica OH groups,respectively. In comparison with the IH MAS NMR spectra recorded after Tic14 reaction at 175 OC with 560 and 820 OC silica (Figure 9A and B), the residual OH groups after Tic14 reaction at 450 OC show more of an H-bonded character. This is consistent with the IR spectrum in Figure 7C, where a broad band mostly due to weakly H-bonded OH groups after Tic14 reaction carried out 450 O C was seen. Table I1 presents the number of chlorideatoms/nm2on the surface after TiC14reaction at 450 OC, together with the number of original isolated OH groups measured by IH NMR. The values are in suprisingly good accord and may be due to chlorination of OH groups as explained later in this paper. Because of the agglomeration and the interaction of HCl with OH groups taking place at 450 OC, the Cl/Ti ratio cannot be used to evaluatethe reaction mechanism of TiC14. Calculations of the Cl/Ti ratio nevertheless reveal an interesting constant value of 2.0 for all silica samples preheated at 450-820 "C (Figure 10). The numbers of OH groups left on the silica surface after titanium binding at 450 OC for 560 and 820 OC silica were 0.3 and