Photocatalytic decomposition of fatty stains by TiO2 thin films

Anatase TiO2 nanoparticles have been prepared as aqueous colloidal solutions by a conventional sol-gel process. Thin films have been deposited onto glass substrates by dip-coating and eventually further crystallized by heat-treatment up to 300 ◦C. The effect of the TiO2 film thickness, porosity and particle size on the reaction kinetics has been studied as well as the starting thickness of the fatty acid layer. The photocatalytic activity of the films has been measured through the decomposition of artificial stains formed by fatty acids deposited onto the TiO2 layer (adipic or stearic acid). The kinetics of stain photodecomposition has been studied in details by in-situ infrared spectroscopy absorption studies. This allows to discuss the role of UV light absorption and the interdiffusion of radicals and oxygen across the interfaces. The mechanism for stain decomposition is shown to be associated to a shrinkage of the solid film at the surface of the porous oxide film after a diffusion of the excited radicals from the oxide.


INTRODUCTION
The semiconducting properties of titanium oxide as well as its low price and non toxicity are making it a suitable photocatalyst for the degradation of organic or biological pollutants [1,2] or for the photocleavage of water [3]. TiO 2 has been used as colloidal dispersions [4,5] or as transparent thin films supported on glass [6][7][8][9].
Self cleaning windows have been investigated more recently for the decomposition of stains deposited either mechanically or by condensation of vapors or fogs in industrials areas.
The decomposition of fatty acid deposited onto the TiO 2 film is a complicated solid-solid reaction mediated by the diffusion of radicals from the interface. We discuss the effect of the TiO 2 film morphology as well as the thickness of the organic layer.

Synthesis. Colloidal dispersions of TiO 2 were
prepared by peptization of hydrous titania similar to that described by Bartlett, et al. [10] but using titanium tetrabutoxide. 17 ml of this precursor were dissolved in 100 ml of butanol (solution 0.5 M). They were hydrolyzed with 8 ml of a solution of nitric acid (0.2 M) slowly added. After completion of the hydrolysis (30 min) the resulting transparent gel is dispersed in 100 ml of nitric acid 0.2 M and homogeneously stirred for 24 hours at various temperatures (20 • C, 40 • C, 60 • C and 80 • C) for 24 hours. Butanol and water are not miscible. The two liquids slowly separate with the colloidal dispersion in the aqueous layer. Thermogravimetric analysis confirms that the concentration of the colloidal particles in water is 0.5 M TiO 2 .
Dynamic light scattering experiments performed on these solutions indicate that the highest reflux temperatures yield particle growth and ripening ( Figure 1). The average particle size ranges from 20 nm for the sample without heat-treatment up to 150 nm for the 80 • C heattreated samples. However, X-ray diffraction does not show evidence of such a large difference between all samples. The crystalline phase is predominantly anatase with small amounts of brookite. One can estimate the size of the diffraction domain from the width of the diffraction peak at 25 • which is the only one in the X-ray pattern that originates from a single d-spacing (common to the anatase and brookite forms). All other peaks are composite peaks. For all colloids we obtain a constant Vol. 05 cristallized domain size of 4.0 nm. Heating the powders at 300 • C slightly increases the crystallite size up to a diffraction coherence length of 7.0 nm (Figure 2(c)).

Thin films. Thin films have been obtained by
dip-coating soda lime microscope slides in the colloidal solution (withdrawal rate 4.6 cm/min). The film quality is improved by addition of 1% of dispersing agent (Tween 80). Scratch resistant films are obtained after calcination of the films at 300 • C for 30 min. However, the films remain very porous as shown by scanning electron microscopy ( Figure 3). On substrates without chemical barrier, the annealing temperature of 300 • C is the highest temperature that can be reached in order to avoid reaction between some glass component and the coating. Indeed, diffusion of Na + ions into the films deteriorates the photoreactivity [11].
The equivalent thickness of the film can be estimated after comparing its absorption at 312 nm with that of the parent colloidal solution diluted to 1.10 −3 mol/l (and correction from the glass contribution). After one coating sequence the estimated mass of TiO 2 in the film is 50 µg/cm 2 in good agreement with the mass uptake measured on the slide after coating. This is equivalent to 125 nm of dense TiO 2 film.
The photocatalytic activity has been measured with a method described by Y. Paz, et al. [8]. The films were covered by dip-coating with a solution of stearic acid in ethanol (solution concentration between 5 g/l and 60 g/l, withdrawal rate 4.6 cm/min). Films were placed in an infrared transmission spectrometer (Perkin Elmer 783) at an angle of 45 • with the IR beam. They were simultaneously irradiated with a 312 nm light (measured power: 2.85 W/m 2 ). The change of transmittance at 2910 cm −1 which is the characteristic frequency of the C − H stretching vibrations of the stearic acid deposited onto TiO 2 films is recorded as a function of time. The equivalent thickness l of the film is related to the change of transmission ∆T = T − T ∞ between initial and infinite time (when all organics have been removed) by the relation ∆T = e −εl where ε is the absorption coefficient of stearic vibrations. The choice of the UV radiation wavelength was based on the availability of a commercial lamp. A longer wavelength (such as 365 nm) would be more representative of the application with solar light but it is less absorbed by titanium oxide (See Figure 4) and the radicals will be generated in a larger depth of the TiO 2 layer. This effect, under study, changes quantitatively the decomposition kinetics.

Effect of colloid synthesis temperature on the photocatalytic activity.
Despite of the large difference in the particle size, the photocatalytic activity of all films is similar and does not drastically depend on the synthesis temperature ( Figure 5), although the particle sizes were found to range from 20 nm to 150 nm. Probably, the particle size obtained by dynamic light scattering relates to the porous assembly of smaller nanocrystallites (around 4 nm as estimated from the X-ray diffraction peak width), which form the basic element of the photocatalyst. The slightly better activity of the sample refluxed at higher temperature may then be associated to its better crystallinity as shown by X-ray diffraction. Some films were made with additional coating sequences (two and three layers). They do not show significantly different photocatalytic activity. At 312 nm the optical density of the films is quite high. Therefore, all the UV light is absorbed by one single coating of TiO 2 as shown in Figure 4.

Effect of stain thickness.
We have deposited various amounts of stearic acid onto a TiO 2 film. The difficulty is to measure the thickness of the stearic acid layer since part of it is found to be deposited inside the porosity of the TiO 2 film and part of it covers the film. However, thicker films can be obtained with concentrated stearic acid. They require much longer times for degradation to be completed ( Figure 6) under UV. The well-known mechanism for photodegradation of organics is the following: Under constant illumination, we may assume that the TiO 2 film generates a constant flow I 0 of radicals that have a probability µ to react with a layer dx of the stain. If this reaction probability is small and if the thickness l of the stearic acid is limited, the flow of radicals across the stearic acid will be nearly constant and equal to I 0 . The total number of reactions of type (3) in the film will be: Then, the decay of the stearic film thickness can be expected to follow a simple first order kinetics: dl/dt = −µI 0 l. This should lead to an exponential decay for the normalized thickness l/L = e −µI0t , independent on the starting thickness L. Exponential fits would probably well account for the decay of the thinner films ( Figure 6). Indeed, below 20 g/l all films appear to show a simpler exponential decay.
However, the thicker film show much slower kinetics. This may be associated to a decay of the flow of radicals across the stearic acid layer (or a decay of their activity) due to encounters and side-reactions of the 1.  radicals. Then, the radical flow at the distance x from the TiO 2 film will be I(x) = I 0 e −λx . The characteristic decay of radicals λ may be different from µ since the reaction of radicals with the stearic acid may not be their only cause of decay. The total number of reactions with the fatty acid in the layer of thickness l will then become The decay of the film becomes dl/dt = (µ/λ)I 0 (l − e −λl ) that yields upon integration on time the relationship: where l(t) is the film thickness at time t and L the starting thickness. Thus the normalized thickness of the film can be written as The lines in Figure 6 are the fits to the data points using this expression. The results are reported in Table 1.
The term λ · L is characteristic of the diffusion of radicals into the stain. Its low value below 20 g/l leads back to the single exponential decay. We observe that λ · L does not increase smoothly as a function of the concentration but suddenly rises between 15.6 g/l and 20 g/l. This is probably the concentration at which the stearic acid fills the pores of the TiO 2 film and starts to form a dense film above it.
The term µ · I 0 can be understood as the efficiency of radical production (I 0 ) and reaction with the organics (µ). There is no clear dependence of this term on the stearic acid film thickness. At first, µ · I 0 decreases. A simple explanation is that this term takes into account the fact that the fatty acid layer acts as a barrier for the diffusion of water and oxygen toward the surface of the nanoparticles. The transfer of the electron-hole pair to the radicals is therefore slower. For the thick stain layers, it drastically increases but this is probably an artefact of the model, associated to the large increase of λ · L in formula (6).

CONCLUSION
We have shown that at 312 nm, only one thin layer of TiO 2 is sufficient to photodecompose stearic acid. Since this wavelength is high compared to the energy gap of TiO 2 , it should be interesting to compare the photodegradation of stearic acid at longer a wavelengths (around 365 nm).
We have also shown that the thickness of the organic layer of pollutant is a parameter that influences the photo-efficiency of TiO 2 layers towards selfcleaning properties. The thickness of the pollutant layer, has an effect on the kinetics of photodegradation which is probably associated to the diffusion of photo-generated species, such as radicals and possibly oxygen.