Size dependency of nanocrystalline TiO2 on its optical ... - UD Physics

Applied Catalysis B: Environmental 68 (2006) 1–11 www.elsevier.com/locate/apcatb

Size dependency of nanocrystalline TiO2 on its optical property and photocatalytic reactivity exemplified by 2-chlorophenol H. Lin a, C.P. Huang a,*, W. Li b, C. Ni b, S. Ismat Shah b,c, Yao-Hsuan Tseng d a

Department of Civil and Environmental Engineering, University of Delaware, Newark, DE 19716, United States Department of Materials Sciences and Engineering, University of Delaware, Newark, DE 19716, United States c Department of Physics and Astronomy, University of Delaware, Newark, DE 19716, United States d Energy and Environment Laboratories, Industrial Technology Research Institute, Hsinchu 301, Taiwan, ROC

b

Received 19 January 2006; received in revised form 7 July 2006; accepted 18 July 2006 Available online 1 September 2006

Abstract Anatase TiO2 nanocrystallines (17–29 nm) were successfully synthesized by the metal–organic chemical vapor deposition method (MOCVD). Moderate manipulation of system parameters of MOCVD can control the particle size. The electro-optical and photocatalytic properties of the synthesized TiO2 nanoparticles were studied along with several commercially available ultra-fine TiO2 particles (e.g., 3.8–5.7 nm). The band gap of the TiO2 crystallines was determined using the transformed diffuse reflectance technique according to the Kubelka–Munk theory. Results showed that the band gap of TiO2 monotonically decreased from 3.239 to 3.173 eV when the particle size decreased from 29 to 17 nm and then increased from 3.173 to 3.289 eV as the particle size decreased from 17 to 3.8 nm. The results of band gap change as a function of particle size agreed well with what was predicted by the Brus’ equation, i.e., the effective mass model (EMM). However, results of the photocatalytic oxidation of 2-chlorophenol (2-CP), showed that the smaller the particle size, the faster the degradation rate. This is attributed in part to the combined effect of band gap change relative to the spectrum of the light source and the specific surface area (or particle size) of the photocatalysts. The change of band gap due to particle size represents only a small optical absorption window with respect to the total spectrum of the light source, i.e., from 380 to 400 nm versus >280 nm. Consequently, the gain in optical property of the larger particles was severely compromised by their decrease in specific surface area. Our results clearly indicated the importance of specific surface area in controlling the photocatalytic reactivity of photocatalysts. Results also showed that the secondary particle size grew with time due mainly to particle aggregation. The photocatalytic rate constants decreased exponentially with increase in primary particle size. Primary particle size alone is able to predict the photocatalytic rate as it is closely related to the electro-optical properties of photocatalysts. # 2006 Published by Elsevier B.V. Keywords: Size effect; TiO2; Photocatalyst; Particle size; 2-Chlorophenol; Size quantization effect; Electro-optical property

1. Introduction Since the discovery of photovoltaic property of titanium dioxide TiO2 by Fujishima and Honda [1], great efforts have been focused on elucidating the electronic structure [2–5], catalytic reactivity [6–8] and surface property [9] of TiO2. Inexpensive and thermal-dynamically stable at room temperature, this semiconductor material has been widely used in heterogeneous photocatalysis and proven to be capable of decomposing a host of organic pollutants such as phenolic

* Corresponding author. Tel.: +1 302 831 8428. E-mail address: [email protected] (C.P. Huang). 0926-3373/$ – see front matter # 2006 Published by Elsevier B.V. doi:10.1016/j.apcatb.2006.07.018

compounds [10,11], metal ethylene diamine tetra acetate (EDTA) complexes [12,13], airborne microbes [14] and odorous chemicals [15]. Most of these studies involved ultraviolet (UV) photons as the major exciting light sources. Considering that there is only 5% of solar irradiation within the UV range, intuitively it is desirable to enhance the photocatalytic performance of TiO2 by enabling it to utilize photons from the near-visible to visible region. It has been suggested that this can be achieved by manipulating the particle size of photocatalyst [3,6,16] or doping the TiO2 with foreign ions [17–19]. The initiation of a photocatalytic reaction requires a minimum photon energy that exceeds the band gap of the material in order to trigger the interband transition of electrons

2

H. Lin et al. / Applied Catalysis B: Environmental 68 (2006) 1–11

between the lowest unoccupied molecule orbital (LUMO) and the highest occupied molecule orbital (HOMO); that is, the incident wavelength needs to be smaller than the wavelength of the band gap threshold, lbg. Thus, it is speculated that reducing the band gap of TiO2 can enhance its photocatalytic performance through more efficient utilization of lower energy photons. TiO2 has three distinct crystalline structures: rutile, anatase, and brookite. Most studies on the photocatalytic reactivity were conducted with either rutile or anatase, which reported that the indirect band gap was 3.0 eV (or lbg 413 nm) and 3.2 eV (or lbg 387 nm), respectively. Although rutile has a lower band gap than anatase, it has been demonstrated that anatase-structure TiO2 exhibits a better photocatalytic performance than that of rutile [20–22]. This is attributed in part to a wider optical absorption band and smaller electron effective mass of rutile than those of anatase, which leads to higher mobility of charge carriers in rutile than anatase [5]. The band gap of TiO2 is commonly believed to be indirect. In contrast to direct interband transition, indirect transition requires phonons (lattice vibration) to compensate for the change in wave vector during electron transition. It has been reported that the band gap of semiconductor crystalline is a function of the particle size [3,4,23]. Below a certain threshold, the density of point/surface defects of semiconductor crystalline increases with decrease in particle size. Due to mild delocalization of molecular orbitals on the surface, defects in the bulk semiconductor create deep and shallow traps near the band edge of its electronic state, which brings about reduction in band gap, that is, red-shift in absorption spectrum [3,4]. When the size of semiconductor particle decreases from its bulk to that of Bohr radius, e.g., the first excitation state, the size quantization (Q-size) effect arises due to the spatial confinement of charge carriers. Consequently, electrons and holes in the quantum sized semiconductor are confined in a potential well and do not experience the delocalization that occurs in the bulk phase. Therefore, the band gap of ultra-fine semiconductor particle increases with the decrease in particle size when it is smaller than the band gap minimum [3,4]. This phenomenon has been described by the Brus’ effective-mass model (EMM) [3]. Size quantization effect has been studied using various semiconductors including CdS [24], HgSe, PbSe, CdSe [25], ZnO [26], Cd3S2 [27], and TiO2 [6,16,28]. The reported Q-size effect of semiconductor clusters appears to be between 1 and 12 nm. Results of these studies were mostly obtained from liquid phase UV–vis absorption spectrum, and few had focused on the photodegradation of organic compound. A good photocatalyst must have high photon conversion efficiency in addition to high specific surface area. In fact, the primary particle size of photocatalysts determines both the specific surface area and the photon conversion efficiency. Although the size dependency of band gap has been studied theoretically, only a few investigations have been conducted on the change of band gap as a function of particle size using TiO2 [6,16,29]. Anpo et al. [6] studied the change of band gap of Ti02over a wide range of particle sizes (e.g., 3.8–200 nm) and found significant blue shifts of the absorption edge by 0.093 and

0.156 eV for rutile and anatase crystalline, respectively, when particle size was less than 12 nm. Kormann et al. [16] observed the quantum confinement effect upon illumination of TiO2 colloids (e.g., 99.9% purity from Aldrich). Constant air purging was provided at a flow rate of 100 sccm using a porous diffuser. The initial pH value of the mixture solution was set at 10.0 to minimize the evaporation of 2-CP during experiments (note that the pKa of 2-CP is 8.52). The residual concentrations of the parent and intermediate compounds were measured using high-performance liquid chromatography (HPLC) (Model HP 2100, Agilent, CA, USA). The instrument was equipped with a Zorbax Eclipse XDB-C18 column (4.6 mm 250 mm) from Agilent. The eluent consisted of 60% acetonitrile and 40% phosphate (pH 3) buffer. The system flow rate was maintained at 0.5 ml/min with total injection volume of 10-ml. Spectra were acquired by a diode array detector (l = 210 nm). 3. Results and discussion 3.1. Optical properties From the diffuse reflectance plot (Fig. 5a), we can see that A20, A17 and A23 have better light absorbance than A3.8, A4.9, A5.7, and A29 in the range of wavelength (l) between 400 and 350 nm. The damping of the F(R1) for A3.8 and A4.9 was observed when the wavelength was below 360 nm. We suspect that is due to the size quantization effect and defects induced deep energy traps that are more discretely created (compared to A5.7). The damping of the diffuse reflectance may be caused by attenuation of excited electron de-excited during the scanning of different wavelength. Thus, the luminescence is created during the de-excitation and amplified the signal of the reflectance. The absorption coefficient of an indirect semiconductor near the absorption threshold can be expressed as a¼

Bi ðhn Eg Þ2 hn

(3)

where Eg is the band gap of indirect allowed transition (eV), h the Planck’ s constant (J s), Bi the absorption constant, and n is the frequency of the light (s1). Therefore, a transformed Kubelka–Munk function can be constructed by plotting [F(R1)]0.5 against the energy of excitation source to obtain the band gap of TiO2 particles (Fig. 5b) [41,44]. The band gaps for A3.8, A4.9, A5.7, A17, A20, A23, and A29 were determined to be 3.289 0.02, 3.251 0.02, 3.275 0.02, 3.173 0.02, 3.179 0.02, 3.224 0.02, and 3.239 0.02 eV, respectively. Results showed that the band gap

Fig. 5. (a) Diffuse reflectance spectra of various TiO2 anatase particles according to the Kubelka–Munk equation; (b) the transformed Kubelka–Munk function vs. energy of the excitation source.

decreased (e.g., red shifted) to a certain minimum value (i.e., critical size) with decrease in particle size from an estimated bulk value of 29 nm. Further decrease in particle size from the critical size (corresponding to band gap minimum) caused the band gap to increase (i.e., blue-shift). A plausible explanation for this change in band gap with respect to the size of the particle is that the bulk defects induce delocalization of molecular orbitals in the conduction band edge (e.g., LUMO) and create shallow/deep traps in electronic energy, in turn causing the red-shift of the absorption spectra. When crystalline size decreased below its size at the band gap minimum, the traps shifted to higher energy, which resulted in blue shifting of the absorption spectra (e.g., size quantization effect) [3,4]. Based on the Bras’ EMM model, the first excitonic energy of semiconductor cluster, E*, can be expressed as a function of particle size, as given by the following expression: h2 p2 1 1 1:8e2 E ﬃ Eg þ þ eR 2R2 me mh

(4)

where Eg is the band gap of the bulk semiconductor, the second term is the energy induced by the quantum confinement effect,


7

Fig. 6. Band gap shift vs. the particle size of TiO2. Solid circles are experimental results. Lines are predicted by the Brus model at different reduced effective mass of exitions, m (1/m = 1/me + 1/mh) and dielectric constant k = 184.

and the third term is the shift of energy due to columbic attraction between electron and hole pairs. £ is the Plank’ s constant (J s), R the radius of the cluster (m), me the effective mass of the electron (kg), mh the effective mass of the hole (kg), e the charge of electron (C), and e is the electric permittivity (C2 N1 m2) of the material (a product of dielectric constant k and permittivity in free space e0). As shown in Eq. (4), the band gap increased as particle size decreased to below a certain threshold. The shifts in band gap were significantly affected by the reduced mass of exciton, i.e., m1 ¼ m1 þ m1 and the e h dielectric constant of the material k. The reported effective mass of the electron, me was in the range of 5–13me [16,29,45]. The effective mass of the hole, mh was reported to be 2me [16]; the dielectric constant of TiO2 anatase crystalline was reported to be between 23 and 184 [29,46–49]. Eq. (4) was plotted by using the reported values for the dielectric constant f of 184 and effective mass of exitons (me ) from 5 to 13me and mh of 2me (Fig. 6). This was equivalent to m values of between 1.43 and 1.73me. We can see the decrease (left shift) in size of band gap minimum and increase in band gap shift (DEg) when m increases from 1.43 to 1.73. In contrast, decrease in dielectric constant from 184 to 23 will result in greater reduction in band gap minimum and significantly enlarge the band gap shift in the negative direction (result not shown). By comparing the observed band gap values to those calculated by the EMM model, it is seen that the observed values agreed very well with those predicted by the model. 3.2. Photoreactivity If the energy of the incident photon is greater than the band gap, photoexcitation will occur and yield the electron and hole pairs. The trapped charge carriers are formed within a 20-ps pulse but have a life time in the nanosecond (ns)-range at Ti3+ sites within the semiconductor bulk [50]. This mechanism has been confirmed by direct and indirect in situ electron paramagnetic resonance (EPR) measurements (e.g., examination of paramagnetic species on hydrated TiO2 surfaces)

Fig. 7. Photocatalytic degradation of 2-CP under UV radiation: (a) changes of normalized concentration as a function of time; (b) least square best fitting for the first-order reaction rate constant k (s1). Symbols: P25 (solid squares), A29 (solid circles), A20: (open circles), A17 (open squares), A5.7 (solid diamonds), A4.9 (open triangles), and A3.8 (open diamonds).

[6,51,52]. Generated electrons are readily trapped on Ti4+ sites and form Ti3+. The trapped electron can be readily scavenged by oxygen [51]. Localized holes can be scavenged by either reacting with hydroxide ions or through electron transfer with water to form hydroxyl radicals, which are strong oxidants that can oxidize organic substances nonselectively [10,31,51,52]. It is generally accepted that photogenerated hydroxyl radicals are primarily responsible for the heterogeneous degradation of organic compounds, such as 2-CP [10,31,53,54]. Fig. 7 shows the photocatalytic reactivity of 2-CP over TiO2 particles at various particle sizes. Results indicated that the degradation reaction followed pseudo firstorder kinetics regardless of the particle size (Table 2). It also showed that the reaction rate constants increased with decrease in primary particle size. This is in contrast to the band gap measured (Fig. 5b), where a minimum band gap was observed with a specific primary particle size of 17 nm. That is, the particles with size in the range of 17–29 nm, exhibited larger lbg values than the smaller ones. Photoreactivity under UV irradiation showed that the smaller the particles, the higher the 2-CP decomposition rate

8


Table 2 List of the least square best fitting parameters for different primary particle sizes

P25 A29 A20 A17 A5.7 A4.9 A3.8 a b c d

Diametera

Specific surface areab

k 1000c

R2

QE 1000d

33 29 20 17 5.7 4.9 3.8

47.7 5 55.7 5 77.9 5 92.2 5 282.5 5 312.7 5 396.1 5

16.36 0.17 17.47 0.02 20.31 0.03 21.17 0.04 24.55 0.10 30.03 0.11 29.23 0.15

0.9565 0.9997 0.9995 0.9993 0.9974 0.9982 0.9901

12.75 13.49 13.70 13.71 13.87 14.02 13.95

Diameter based on TEM, primary size (nm). BET specific surface area (m2/g). First-order reaction rate constant obtained by least squares best fitting (s1). Quantum efficiency at 180 min of experiment.

under otherwise identical experimental conditions. In our study, commercially available P25 appeared to have the worst degradation rate of 2-CP (k = 0.01636 s1), whereas A3.8 yielded the highest rate (k = 0.02923 s1). The specific surface area of particles increases with decrease in particle size (Table 1). Obviously, the advantage in photocatalytic activity gained through the enhancement in optical response from the smaller band gap exhibited by larger particles (e.g., 17–29 nm) was severely compromised by the decrease in specific surface area. Furthermore, the variation in photon absorption due to size effect was lbg between 381 and 395 nm. Considering the emission spectrum of the excitation source (Fig. 8), utilizable photons are in the range between 280 nm (note that the Pyrex glass cuts off the photon emission at 280 nm, as indicated in Fig. 8) and lbg of TiO2. Clearly, the change of light absorption band due to size effect (e.g., 381–395 nm) is relatively small compared to the fully utilizable light spectrum (e.g., 280– 395 nm) available in this study. In addition, the quantum efficiency of the photocatalytic reactivity appeared to have an

Fig. 8. Photonflux spectrum of the medium pressure mercury lamp used. The two vertical lines, between 381 and 395 nm, indicate the band gap changes observed of the TiO2 samples used in this study. Generally, the band gap of the bulk TiO2 was around 387.2 nm. The band gap red-shifted to ca. 395 nm as the particle size decreased to about 17 nm then the band gap blue-shifted to ca. 381 nm as the particle size continued to decrease to about ca. 3.8 nm. Inset is the plot of particle size vs. its corresponding wavelength of band gap threshold lbg (nm).

exponential relationship to the incident wavelength near the optical adsorption edge. Zang et al. [55] reported that the incident photon conversion efficiency (IPCE) could vary by an order of magnitude comparing excitation wavelength at 400 and 550 nm during the photocatalytic degradation of 4chlorophenol (4-CP). Puma and Yue [54] observed that the reaction rate of 2-CP increased by 3.9 times (1.73 mM min1 versus 6.78 mM min1) when the system was under UV-A versus UV-ABC (i.e., the full UV spectrum) irradiation. These observations imply that even when the incident photon energy is higher than the band gap of the photocatalyst, quantum efficiency varies greatly below the wavelength of the band gap threshold, lbg. This variation is especially large when the incident wavelength is near the adsorption edge. Thus, sizeeffect-enhanced optical absorption (381–395 nm in our study) is likely to contribute an unnoticeable improvement on the photocatalytic degradation of the 2-CP. As a result, it is not surprising to see photocatalysts with lower band gap and smaller specific surface area yield lower photodegradation efficiency than those with higher band gap and larger specific surface area. The quantum yield (F) was computed based on Eq. (5) [48]: Fl ¼

mole of 2-CP degraded Einstein of incident photons

(5)

Fig. 9 shows the apparent quantum yield versus the primary particle size at 180 min. It is seen that the quantum yield increases almost monotonically with decreases in primary particle size. Again, this is due to the advantage of the large specific surface area of the ultra-fine particles, the spectrum of the excitation source, and the wavelength-dependent nature of photocatalytic reactivity. In photocatalytic slurry systems, two major factors will greatly impact the reaction: (a) specific surface area and (b) band gap. The specific surface area determines the available sites for reactions to take place, whereas, the band gap of the semiconductor will define the amount of photons that are available for quantum conversion. Under our experimental conditions, results showed that ultra-fine particles (e.g., 10 nm). However, the ultrafine particles only exhibited a reaction rate roughly by a factor of two and slightly higher apparent quantum efficiency (F) in 180 min than that of the larger particles. It seems that several factors might offset the surface area advantage of ultra-fine particles: (1) size quantization effect (e.g., less than 10 nm in our case), (2) increase in surface electron–hole recombination, (3) particle aggregation, and (4) spectrum of the excitation source. Quantization effect yields a larger band gap as the particle size decreases. As a result a light source with higher energy, e.g., blueshifted, is required to separate the excitons [3,4]. When the particle size decreases, the density of recombination center increases which encourages holes and electrons recombination [35,38]. As discussed in Section 3.3, ultra-fine particles (e.g., 10 nm and binomially distributed when the primary particle size was l > 320 nm) provides sufficient energy for exciton separation of higher band gap energy in particles (e.g., A3.8, A4.9, and A5.7). We therefore suggest that unless a well-confined light source is provided (e.g., l = 380–400 nm), TiO2 particles in the size range of 17– 29 nm cannot perform better in terms of photocatalytic reactivity than those in the 3.8–5.7 nm size range, under otherwise identical experimental conditions. Furthermore, even in the presence of light at wavelength between 380 and 400 nm, the photocatalytic performance of large particles (e.g., 17– 29 nm) will be compromised due to their relatively low specific surface area. Our results clearly indicated that particle aggregation did take place readily in aqueous solutions. Unless a constant energy input is provided throughout the entire experiment, there is no way to maintain the primary particle size. In light of the rapid aggregation among the ultra-fine particles, especially those that are smaller than 10 nm, primary particle size alone was able to predict the photocatalytic reaction. Intuitively, one would reason that the primary particle size holds key to the electro-optical property and photocatalytic reactivity of photocatalysts due mainly to the change in band gap energy (e.g., result in shifts in absorption spectrum) and its specific surface area. Acknowledgement The authors wish to acknowledge our two anonymous reviewers for their excellent comments. This work was supported by a NSF grant NIRT #0210284. References [1] A. Fujishima, K. Honda, Nature 238 (1972) 37–38. [2] R. Asahi, Y. Taga, W. Mannstadt, A.J. Freeman, Phys. Rev. B 61 (2000) 7459–7465. [3] L. Bras, J. Chem. Phys. 80 (1984) 4403–4409. [4] L. Bras, J. Phys. Chem. 90 (1986) 2555–2560. [5] S.D. Mo, W.Y. Ching, Phys. Rev. B 51 (1995) 13023–13032. [6] M. Anpo, T. Shima, S. Kodama, Y. Kubokawa, J. Phys. Chem. 91 (1987) 4305–4310. [7] A. Hagfeldt, H. Lindstrom, S. Sodergren, S.E. Lindquist, J. Electroanal. Chem. 381 (1995) 39–46. [8] B. Oregan, M. Gratzel, Nature 353 (1991) 737–740. [9] K. Bourikas, T. Hiemstra, W.H. Van Riemsdijk, Langmuir 17 (2001) 749– 756. [10] I. Ilisz, A. Dombi, K. Mogyorosi, A. Farkas, I. Dekany, Appl. Catal. B: Environ. 39 (2002) 247–256. [11] S. Nevim, H. Arzu, K. Gulin, C. Zekiye, J. Photochem. Photobiol. A: Chem. 146 (2002) 189–197. [12] S.V. Muhammad, A.P. Davis, Water Res. 34 (2000) 952–964. [13] J.K. Yang, A.P. Davis, Environ. Sci. Technol. 35 (2001) 3566–3570.

H. Lin et al. / Applied Catalysis B: Environmental 68 (2006) 1–11 [14] S. Takashi, K. Yoshiyuki, T. Masahito, Biochem. Eng. J. 14 (2003) 149–152. [15] M.C. Canela, R.M. Alberici, W.F. Jardim, J. Photochem. Photobiol. A: Chem. 112 (1998) 73–80. [16] C. Kormann, D.W. Bahnemann, M.R. Hoffmann, J. Phys. Chem. 92 (1988) 5196–5201. [17] R. Asahi, T. Morikawa, T. Ohwaki, K. Aoki, Y. Taga, Science 293 (2001) 269–271. [18] W. Choi, A. Terrain, M.R. Hoffmann, J. Phys. Chem. 98 (1994) 13669– 13679. [19] T. Umebayashi, T. Yamaki, H. Itoh, J. Phys. Chem. Solids 63 (2002) 1909– 1920. [20] J. Augustynski, Electrochim. Acta 38 (1993) 43–46. [21] X.H. Tang, Y. Zhang, G.F. Yin, D.L. Zhou, H. Liu, C.Q. Zheng, Rare Met. Mater. Eng. 33 (2004) 864–868. [22] K. Tanaka, M.F.V. Lapule, T. Hisanaga, Chem. Phys. Lett. 187 (1991) 73–76. [23] J.P. Rino, N. Studart, Phys. Rev. B 59 (1999) 6643–6649. [24] M. Haase, H. Weller, A. Henglein, J. Phys. Chem. 92 (1988) 4706–4712. [25] J.M. Nedeljkovic, M.T. Nenadovic, O.I. Micic, A.J. Nozik, J. Phys. Chem. 90 (1986) 12–13. [26] U. Koch, A. Fojtik, H. Weller, A. Henglein, Chem. Phys. Lett. 122 (1985) 507–510. [27] A. Fojtik, H. Weller, A. Henglein, Chem. Phys. Lett. 120 (1985) 552–554. [28] L. Kavan, T. Stoto, M. Gratzel, D. Fitzmaurice, V. Shklover, J. Phys. Chem. 97 (37) (1993) 9493–9498. [29] N. Serpone, D. Lawless, R. Khairatdinov, J. Phys. Chem. 99 (1995) 16646–16654. [30] A.J. Maira, K.L. Yeung, C.Y. Lee, P.L. Yue, C.K. Chan, J. Catal. 192 (2000) 185–196. [31] C.B. Almquist, P. Biswas, J. Catal. 212 (2002) 145–156. [32] J.A. Byrne, B.R. Eggins, P.S.M. Dunlop, S. Linquette-Mailley, Analyst 123 (1998) 2007–2012. [33] H.D. Jang, S.J. Kim, S.K. Kim, J. Nanoparticle Res. 3 (2001) 141–147. [34] H.J. Nam, T. Amemniya, M. Murabayashi, K. Itoh, J. Phys. Chem. B 108 (2004) 8254–8259.

11

[35] Z. Zhang, C.C. Wang, R. Zakaria, J.Y. Ying, J. Phys. Chem. B 102 (1998) 10871–10878. [36] N. Xu, Z. Shi, Y. Fan, J. Dong, J. Shi, M.Z.C. Hu, Ind. Eng. Chem. Res. 38 (1999) 373–379. [37] H. Gerischer, Electrochim. Acta 40 (1995) 1277–1281. [38] M.A. Grela, A.J. Colussi, J. Phys. Chem. 100 (1996) 18214–18221. [39] W.C. Hao, S.K. Zheng, C. Wang, T.M. Wang, J. Mater. Sci. Lett. 21 (2002) 1627–1629. [40] G.F.A. Kortum, Reflectance Spectroscopy: Principles, Methods, Applications, New York, 1969. [41] G. Burgeth, H. Kisch, Coord. Chem. Rev. 230 (2002) 41–47. [42] E.L. Simmons, Appl. Opt. 14 (1975) 1380–1386. [43] W. Li, M. Sung, C.P. Huang, I. Shah, J. Vac. Sci. Technol. B 20 (2002) 2303–2308. [44] S. Sakthivel, H. Kisch, Angew. Chem. Int. Ed. 42 (2003) 4908–4911. [45] J. Pascual, J. Camassel, H. Mathieu, Phys. Rev. B 18 (1978) 5606–5614. [46] W. Dong, S.P. Ruan, X.D. Zhang, W.B. Guo, C.X. Liu, S. Zhang, C.P. Jia, J.X. Pan, W.Y. Chen, Chem. Res. Chin. Univ. 19 (2003) 347–349. [47] M. Es-Souni, I. Oja, M. Krunks, J. Mater. Sci. Mater. Electron. 15 (2004) 341–344. [48] I. Oja, A. Mere, M. Krunks, C.H. Solterbeck, M. Es-Souni, Solid State Phenom. 99/100 (2004) 259–262. [49] R. VandeKrol, A. Goossens, J. Schoonman, J. Electrochem. Soc. 144 (1997) 1723–1727. [50] M.A. Fox, M.T. Dulay, Chem. Rev. 93 (1993) 341–357. [51] R.F. Howe, M. Gratzel, J. Phys. Chem. 91 (1987) 3909–3960. [52] C.D. Jaeger, A.J. Brard, J. Phys. Chem. 83 (1979) 3146–3152. [53] C.D. Dong, C.P. Huang, Advances in Chemistry Series, vol. 244, Am. Chem. Soc., Washington, DC, 1995, pp. 291–313 (Chapter 15). [54] G.L.L. Puma, P.L. Yue, Ind. Eng. Chem. Res. 41 (2002) 5594–5600. [55] L. Zang, W. Macyk, C. Lange, W.F. Maier, C. Antonius, D. Meissner, H. Kisch, Chem. Eur. J. 6 (2000) 379–384. [56] P.C. Hiemenz, R. Rajagopalan, Principles of Colloid and Surface Chemistry, Marcel Dekker, Inc., 1997.