Structure-Reactivity Relationships of Anatase and

Solid State Phenomena Vol. 162 (2010) pp 203-219 © (2010) Trans Tech Publications, Switzerland doi:10.4028/www.scientific.net/SSP.162.203

Structure-Reactivity Relationships of Anatase and Rutile TiO2 Nanocrystals Measured by In Situ Vibrational Spectroscopy Lars Österlund1,2,a, 1

FOI: The Swedish Defence Research Agency, Cementvägen 20, S-901 82 Umeå, Sweden.

2

Dep. Engineering Sciences, The Ångström Laboratory, Uppsala University, P. O. Box 534, S-751 21 Uppsala, Sweden. a

[email protected], [email protected]

Keywords: TiO2, anatase, rutile, nanocrystal, Fourier transform infrared, spectroscopy, formic acid, formate.

Abstract. A comprehensive analysis of structural-reactivity relations on TiO2 nanocrystals is presented. Using an interplay between TEM, X-ray diffraction and vibrational spectroscopy of welldefined anatase and rutile TiO2 nanocrystals correlations between the adsorbate structure of formic acid and the corresponding photo-induced decomposition rate are described. It is demonstrated that the detailed bonding configuration determines the decomposition rate. Generalizations and implications of the findings are discussed. Introduction Photocatalysis is a broad research field which lies at the heart of modern sustainable technologies such as air and water cleaning, solar hydrogen production, wet solar cells, self-cleaning and antibacterial surface coatings. A detailed understanding of photocatalytic processes is not possible without explicit model studies of the detaield photocatalyst structure, adsorbate structures and surface reaction mechanisms. The most common crystal modifications of TiO2 are anatase and rutile. They have different electronic properties with slightly different energy band dispersion with different band gaps and symmetries of band gap interband transitions. The surface atomic arrangement and free energy of a crystal vary with the crystallographic orientation. Therefore, the electronic properties and reactivity as well as other physical and chemical properties of crystals depend on their shape. The {101} and {001} crystal faces of TiO2 are reported to have the lowest surface energy of anatase, while the {110} crystal faces of rutile have much lower surface energy than other faces [1-4]. Thus different families of crystal planes are expected to terminate anatase and rutile nanoparticles to minimize the particle surface energy. Anatase which has several low index crystal faces with similar surface energy exhibits different crystal morphologies exposing different fractions of low index faces . The minimum free energy morphology of an anatase crystal calculated by the Wulff construction exhibits a truncated tetragonal bipyramidal structure exposing {101} and {001} faces, where more than 90 per cent of the surface are {101} faces [1]. Recent reports suggest that there is an intimate coupling between the particle morphology, the exposed surface faces and the particle reactivity [5-7]. Hence it is of utmost importance to be able to control and modify the morphology of crystals to tune their reactivity. The (101) surface of anatase is reported to be quite unreactive and do not promote dissociative water, methanol and formic acid adsorption [8-10]. In contrast it has been reported that the (001) surface is reactive: It spontaneously reconstruct under ultra-high vacuum conditions and water spontaneously dissociated on the non-reconstructed (001) surface [11]. The photocatalytic activity of anatase TiO2 crystal with different morphologies, face distribution and surface atomic arrangements other than that dictated by thermodynamics have been the subject of only few reports.

All rights reserved. No part of contents of this paper may be reproduced or transmitted in any form or by any means without the written permission of the publisher: Trans Tech Publications Ltd, Switzerland, www.ttp.net. (ID: 150.227.15.253-12/05/10,08:29:54)

204

Solid State Chemistry and Photocatalysis of Titanium Dioxide

Yang et al [12] reported on a method to synthesize 1.6 µm sized anatase crystals with a large fraction of {001} faces by using fluorine-terminated surfaces to change the relative stability of {101} and {001} surfaces. They synthesized uniform anatase TiO2 crystals with a high percentage (47 per cent) of {001} faces using hydrofluoric acid as a morphology controlling agent. Moreover, they showed that fluorated surface of anatase single crystals can easily be cleaned using heat treatment to render a fluorine-free surface without altering the crystal structure and morphology. Byun et al [13] reported on a CVD method to prepare anatase TiO2 thin films with preferred orientation. The orientated films resulted in the larger surface area for photocatalytic reaction by forming columnar structure with deeper voids on the film surface. Among the films they prepared those that exhibited orientation showed the highest photocatalytic reaction rate for benzene decomposition and was attributed to columnar structure and larger surface area of these films. In a later study Kim et al [14] compared SEM micrographs of the surface morphology and cross sectional images of and -oriented anatase TiO2 films. They concluded that surface of the oriented film was denser than that of the oriented film. The oriented film was reported to consist of {100} and {004} faces and exhibited an open columnar structure perpendicular to the substrate surface, while the oriented film exhibited densely aggregated columns. Jhin et al [15] reported a CVD method for preparation of anatase TiO2 films with enhanced orientation on soda-lime glass by plasma pretreatment of the substrate. Tokita et al [16] prepared anatase films on non-epitaxial substrates by a CVD method. The films exhibiting , , and crystal orientation showed high photocatalytic activity for methylene blue reduction. Among them the oriented film had the highest reactivity. Ohno et al [17] studied the crystal face dependent photocatalytic reactivity of 1 µm large rutile and anatase crystals obtained from Toho Titanium Company. The truncated tetragonal bipyramidal shaped anatase particles in this study exposed predominantly {101} and {001} faces. Photocatalytic reduction of hexachloroplatinate resulted in Pt deposits on all faces of the anatase crystals only when isopropanol was added to the solution. By monitoring the amount of deposits on various crystal surfaces they showed that the anatase {001} faces pre-coated with Pt were more reactive for Pb2+ oxidation to PbO2 than the {101} faces. They thus concluded that the {001} and {101} faces provide oxidative and reductive sites, respectively. In another study Taguchi et al [18] etched the micro-sized anatase particles. By comparing the structure determined by SEM before and after etching they showed that by this process the edge between two {101} faces is selectively etched thus forming new {112} faces. Large PbO2 deposits are observed in SEM on the {112} faces when Pb2+ was photocatalytically oxidized on the Pt-deposited TiO2 particles. The oxidative activity of the {112} face was considered to be stronger than that of the {001} face, which acts as the oxidative site on the TiO2 particles before etching because no PbO2 deposits were seen on the {001} face of the etched particles. In this paper we combine in situ vibrational spectroscopy with solution based nanofabrication methods to probe the adsorbate structure of formic acid on TiO2 nanoparticles with different structure and particle size. We show that the TiO2 surface structure strongly affects the bonding and photo-reactivity. There are hitherto no reports available that describes the photocatalytic reactivity of the (112), (100) and (111) anatase surfaces for organic molecules. In particular, there are no reports that treats the photocatalytic activity of anatase TiO2 nanoparticles in the 10-60 nm size range that exposes {112}, {111}, and {100} faces. Even though reports of large anatase crystals and minerals exhibiting a truncated ditetragonal bipyramidal terminated by {100}, {112} or {001} faces have published elsewhere [12, 17, 18], these studies are not related to catalysis, in particular photocatalysis, nor to the particle size regime below 100 nm. Our results suggest that even though TiO2 photocatalysis has been studied extensively during the past decades, the understanding of the bonding and reactivity of even the simplest organic molecules with TiO2 is still very much rudimentary and there are plenty of room for improvements in many applications. Advanced

Solid State Phenomena Vol. 162

205

Infrared (IR) surface spectroscopy is an important characterization technique in this context that provide information of adsorbate structures, and mechanism of molecular surface processes occurring on metal oxides. IR spectroscopy allows for studies under a wide range of experimental conditions: from ultra-high vacuum to high pressures, from cryogenic temperatures to ∼800 K, in gas or liquids, on single crystals, powders, thin films, or colloidal solutions. This allows for true in situ or operando studies, which can provide valuable information that can help bridge pressure and structure gap problems known to hamper our understanding of e.g. catalytic reactions on solid surfaces, including photocatalytic reactions [6, 19-22]. Experimental Nanoparticles were prepared by solution based methods employing alkoxide precursors [23, 24] or hydrothermal treatments of microemulsions [25, 26] as described in detail elsewhere. The particles thus obtained were all well-crystallized, mono-disperse and spectroscopically contaminant free. The anatase samples are denoted “A25”, “A40” and “A60”, respectively. The latter is the commercially available anatase obtained from BDH Ltd. Sample “P25” is the commercially available TiO2 powder sample P25 provided by Degussa AG which is typically used as a reference photocatalyst. P25 is not phase pure; it contains 70-80 wt.% anatase primary particles mixed with rutile primary particles. The sample denoted “Rutile 6x80” is pure rutile TiO2 nanoparticles [25, 26]. Structural, optical and chemical properties of the synthesized materials were characterized by a range of different techniques as described in detailed elsewhere [24, 25, 27]. Briefly, phase analysis was made by the powder X-ray diffraction patterns obtained with a Guiner-Hägg focusing camera employing Cu Kα1 radiation and Si as internal standard using a Siemens D5000 instrument. The Xray diffraction patterns all showed that the materials were of the tetragonal anatase or rutile modification, respectively, with characteristic 2Θ diffraction peaks. The particle diameter determined from a Scherrer analysis of the (101) diffraction peak are shown in Table 1. Scanning electron microscope (SEM) images were obtained with a FEG-SEM Leo 1550 Gemini instrument. Transmission electron microscopy (TEM) was done with a Jeol 2000 FXII instrument equipped with an EDS module (Link AN 1000). High-resolution transmission electron microscopy (HRTEM) and selected area diffraction (SAED) were performed with a transmission electron microscope JEOL JEM 3010 operated at 300 kV (LaB6 cathode) giving a point resolution of 0.17 nm [28]. A copper grid coated with a holey carbon support film was used to prepare samples for the TEM observation. The grid was coated by dispersing powder samples in ethanol and subsequent treatment of the suspension in ultrasound for 10 minutes. The TEM analyses showed that the entire sample was crystalline with well developed crystal faces [7]. Morphology analysis of the particles was done by orienting a computer simulated crystal morphology (tetragonal symmetry, point group 4/mm) according to the orientation deduced from the HRTEM and SAED analysis of TEM micrographs and subsequently fitting the crystal surface energy (vectorial distance from crystal origin to surface plane) to yield a best two-dimensional fit with the experimental images. This procedure was made for typically 2-4 different particles from each batch with sufficient data quality. For consistency validation, morphology analysis was also made on TEM images of particles where no a priori knowledge of particle orientation was available. Complementary validation of the particle dimensions was made by comparing the size, dhkl, of the particles from XRD data calculated from the Scherrer equation, viz. dhkl=0.9λ/Bhklcosθ, by analysis of the full-width at half maximum of the reflection from the (101), (004) and (200) planes for anatase and the (110) from rutile, respectively. The major exposed surface faces determined from the morphology analyses are presented in Table 1 (see Results section below). Visualization of the crystals was made with the WinMorph software [29]. The accuracy of the dimensions determined form this analysis is estimated to be better than 15%. Sufficient HRTEM data of the A25 particles

206

Solid State Chemistry and Photocatalysis of Titanium Dioxide

with appropriate orientation with respect to the electron beam prevents however an unambiguous determination of (112) planes, whose existence can only be iniferred indirectly from low magnification TEM data and XRD dimensions. Particle size Major exposed surfaces* [nm]§ A25 anatase 25 {101}, {001}, {100}, ({112}) A40 anatase 40 {101}, {112}, {111}, {100} A60 anatase 66 {101}, {001} R6x80 rutile 20 {110}, ({101}) P25 anatase/ rutile 21 n.a. Table 1 Physical properties of the TiO2 nanoparticles used in the present study [21, 24, 25, 27]. §) Particle size determined from Scherrer analysis of (101) peak. *) Exposed surfaces are listed in order of relative abundance. Surfaces in parenthesis are not conclusively evidenced from morphology analysis of HR-TEM images. Sample

Phase

The HCOOH adsorbate structures were determined by in situ Fourier transform infrared (FTIR) spectroscopy employing a vacuum pumped spectrometer (Bruker IFS66v/S) equipped with a LN2 cooled narrow band HgCdTe detector. All FTIR measurements were performed in modified transmission (Specac) or diffuse reflectance, DRIFT, (Praying-Mantis, Harrick) gas reaction cells allowing for in situ reaction studies with simultaneous gas deposition, light illumination and FTIR spectroscopy of the thin solid films or powders [24, 28]. For DRIFT spectroscopy of adsorbates on TiO2 nanoparticles dilution of the solid by e.g. KBr is not necessary since IR absorption by TiO2 is negligible above 1000 cm-1 where most necessary spectral information is located. The ionic character of the Ti-O bonds in TiO2 is manifested in the strong IR absorption below ca. 950 cm-1. IR spectroscopy of adsorbed species on metal oxide surfaces is therefore in general prohibited below ca. 1000 cm-1. It should be noted that even though the particle size is in general much smaller than the irradiation wavelength for nanoparticles the IR spectrum of TiO2 (and metal oxides in general) is still not independent of particle size. Field-induced polarization effects and state of aggregation may have strong influence on the M-O vibrational stretching modes and must be considered on a case by case basis [30]. Formic acid (GC grade, Merck) was adsorbed on the various TiO2 nanoparticles by means of a homebuilt gas generator [24, 28]. The independently calibrated formic acid concentrations in the gas feed were ca. 7900 ppm. No other gases than formic acid were detected by mass spectrometry analysis of the gas effluent form the reaction cells. FTIR spectra were recorded with 4 cm-1 resolution and 135 scans (corresponding to 30 sec measurement time per spectra) and 30 sec dwell time between consecutive spectra. Prior to each measurement the samples were annealed at 673 K in synthetic air and subsequently cooled to 299 K in the same feed. After dosing the sample was kept in the gas feed for ca 15 minutes prior to illumination (except when otherwise stated). FTIR backgrounds were collected on clean samples (265 scans) in synthetic air or N2 feed at 299 K. FTIR spectra were smoothed with a Savitzky-Golay algorithm using a 9 point window. Absorbance peak areas were obtained after appropriate base line corrections and Lorenzian curve fitting of the spectral region of interest. Photo-induced degradation of HCOOH and surface reactions were measured by in situ FTIR spectroscopy by monitoring the mode resolved surface products and adducts as a function of time during illumination. Repeated FTIR spectra were measured with 4 cm-1 resolution as a function of time consisting typically of 135 scans and 30 sec dwell time between consecutive spectra. The samples were held at 299 K in a 100 ml min-1 gas flow of synthetic air (20% O2 and 80% N2) during measurements.

Solid State Phenomena Vol. 162

207

Photon irradiation was made with simulated solar light generated by a Xe arc lamp source operated at 200W employing air mass filters AM1.5 [24, 28]. Briefly light was admitted to the reaction cells through a quartz fibre bundle and focused through a lens onto the sample surface such that the whole film or powder surface was illuminated (typically 7-10 mm illuminated area). The photon power on the sample was determined to be 166 mW cm-2 for wavelengths between 200 and 800 nm, corresponding to ca 12 mW cm-2 for λ0.90 Table 2 Compilation of structural parameters obtained from morphology analysis of the anatase crystals used in the present study. *) The centre-to-face distance is the radial distance between the centre of the crystal to the surface plane labelled {hkl} employed in the Wulff construction of the particles. This number is proportional to the surface energy of face {hkl}. §) The face area ratio is defined as ([face surface area×no. faces]/[total surface area of the nanocrystals]). The numbers in parenthesis show the result if the {111} faces are included in the analysis. †) The face index of exposed surface planes and face area ratio is taken from Refs. [25, 26 ]. The face distribution of the different particles depicted is summarized in Table 2. It is seen that the fraction of {101} faces, which typically is the most abundant crystal face (AI), significantly decreases for AII (75%) and AIII (ca. 50%), while simultaneously the area of the {112}, {100} and to a lesser extent {001} faces increases. The rutile R6x80 nanoparticles expose mainly {110} surfaces in agreement with the much lower surface energy of this face compared to the other low index faces. We [25] and others [31] have shown that more than >90% of the exposed surfaces can be attributed to {110} planes. The latter particles are therefore excellent model systems to compare data obtained on single crystal TiO2(110) surfaces and thus provide a rationale to associate typical adsorption structures with observed vibrational spectra and reaction kinetics. We have previously adopted this approach to interpret reaction kinetic of formic acid, formate, acetone and propane on TiO2 nanoparticles [6, 7, 32]. In the following we extend this approach and use it to associate vibrational spectra and adsorption structure to specific crystal face distributions. Formic acid adsorption structure. In Fig. 3 is shown FTIR spectra obtained after HCOOH adsorption and subsequent solar light irradiation in synthetic air, respectively. The FTIR spectra are normalized to the initial surface coverage and the relative surface concentrations are thus intercomparable between the different TiO2 particles. Based on the measured uptake kinetics from in situ FTIR spectroscopy and considering that minor desorption may take place during evacuation of the reaction cell prior to irradiation (see below), we estimate that in all cases the HCOOH/HCOO coverage is close to saturation. In the following we simply assume that the surface coverage prior to irradiation is 1 ML without quantifying how many molecules per unit area this corresponds to. The reported desorption rates are therefore reported in units of ML min-1. Table 3 summarizes the observed vibrational frequencies and corresponding mode assignments.

210

Solid State Chemistry and Photocatalysis of Titanium Dioxide

Fig. 3 FTIR spectra obtained after dosing formic acid on anatase and rutile nanoparticles and subsequent photo-induced decomposition.

Solid State Phenomena Vol. 162

TiO2 substrate

νas(OCO)

Mode assignment νs(OCO) ν(C=O)

211

Ref. ν(CO)

∆νas-s

Single crystal (UHV) Rutile TiO2(110) [33] * 1566 (Specie B) 1393 (Specie B) 173 µ−formate * 1535 (specie A) 1363 (Specie A) 172 µ-formate 1365 [34] µ-formate Nanoparticles Anatase A25 This 1558 1359 199 work µ-formate 1665, 1650 1270 HCOOH(a) 1717 HCOOH (liq) Anatase A40 This 1564 1366 198 work µ-formate 1662 1260 HCOOH(a) 1714 HCOOH (liq) – 1586 1346 240 HCOO Anatase A60 This 1566 1366 200 work µ-formate 1662 1289 HCOOH(a) 1720 HCOOH (liq) – 1586 HCOO P25 This 1563, 1552 1363 200, 189 work µ-formate 1679, 1663 1271 HCOOH(a) P25 [35] 1553 1379 174 µ-formate 1682, 1277 HCOOH(a) Rutile R6x80 This § 1562 (Specie B) 1371 191 work, µ-formate 1536 (Specie A) 1358 (Specie A) 178 [25] Table 3 Vibrational frequencies and mode assignment of formic acid and formate adsorbed on TiO2. *) Note the mixing of mode assignments in Ref. [33]: 1566 ↔ 1535 cm-1. §) Unambiguous assignment of peak not possible due to mixing with δ(CH) and possible other µ-formate species [25]. The FTIR spectra clearly demonstrate that the adsorbate structure is very different on the various anatase and rutile nanoparticles. In particular it is evident that no significant dissociation of HCOOH occurs on A60 upon adsorption. The vibrational signature is instead associated with liquid-like (H-bonded) HCOOH which gradually desorbs as a function of time while the sample is held in synthetic air. The evaporation rate is determined to be 0.02 ML min-1 by monitoring the decay of the ν(C=O) mode associate to the liquid-like specie. The other extreme situation occurs on rutile which exhibit spectra typical for formate adsorbed on single crystal TiO2(110) [33]. In this case HCOOH dissociates according to [33, 36, 37]: HCOOH(a) + OB→ HCOO(a)+HOB

(1)

Here OB denotes the two-fold coordinated, bridging O atoms that form rides along the [001] direction on the (110) surface. On the perfect (110) rutile surface the formate molecule adsorbs with

212

Solid State Chemistry and Photocatalysis of Titanium Dioxide

the molecular axis along the [001] direction with the O atoms bonded to four-fold coordinated Ti atoms thus forming a bridging bidentate specie (µ-formate A). Missing Ob atoms (O vacancies) have been shown to lead to another type of formate specie associated with a different vibrational structure [33]. In this case one O atom in the formate molecule interacts with the O vacancy and fill that site thus “completing” the defective (110) surface; the formate molecular axis is in this case perpendicular to the [001] direction. This specie is denoted µ-formate B. The two species co-exist under atmospheric conditions and give rise to a double ν(OCO) peak structure in FTIR spectra as shown in Fig. 4d [7, 25]. Smaller rutile particles exposing larger areas of minority surface planes can have additional formate structures; again with slightly different vibrational signatures [25]. Turning now to the smaller anatase particles A25 and A40 which exhibit more elaborate surface structures containing mixtures of crystal faces (AII and AIII in Fig. 2), we see that the vibrational adsorbate structure in Fig. 4 also becomes more complex. The smaller size also generally leads to a larger fraction of defect sites (apart from O vacancies) due to edges, corners and steps. Following the HCOOH uptake we have previously shown [7] that on small anatase particles formate first adsorbs on c.u.s. atoms associated with defect sites (νas(COO)≈1570 cm-1), which is associated to bidentate coordinated species based on comparisons with related published data and Deacon’s rules (whether it is chelating or bridging cannot be deduced based on FTIR data alone [31]). With increasing coverage these adsorption sites becomes populated. We attribute the absorption band at ∼1550 cm-1 that gradually appears in the spectra to bridging bidentate coordinated to minority anatase faces ({112}, {100}, {001}, and possibly also 111}), since this band is absent on the A60 particles which mainly expose (101) faces. Furthermore, there is a clear correlation with water formation and the absorption band at 1586 cm-1 on the A40 and A60 samples (Fig. 3). This band occurs at similar frequency as aqueous formate [31] and is therefore attributed to the νas(OCO) in HCOO ions.

Fig. 4 Schematic drawing of possible formatic acid and formate coordination to Ti metal atoms. The vibrational bands occurring around 1640-1660 cm-1 can be attributed to either an adsorbed HCOOH specie (HCOOH(a), type Ia in Fig. 4) or hydrogen bonded monodentate HCOO specie (type Ib in Fig. 4). The distinction between these two species is ambiguous when adsorbateadsorbate interactions and hydrogen to water and hydroxyls boding is included in the interaction scheme. It is reasonable to assume that some of the faces with higher surface energy contained

Solid State Phenomena Vol. 162

213

within the smaller anatase particles present in A40, A25 and P25 samples exhibit an open surface structure and reactive (basic) surface O atoms that may promote dissociation. Theoretical considerations suggest that it is the basicity of adjacent O atoms that determines whether HCOOH dissociates or not and the proton is transferred to the surface forming a HOb species [5]. It is likely that there is an intermediate state with the H atom bonded to both the HCOO moiety and the OB surface atom (Fig. 5). For the same reasons adsorbate-adsorbate interactions and interactions with water molecules and surface hydroxyls can lead to similar mixtures of species that may account for the broad vibrational signature at 1640-1660 cm-1. In Table 2 we label this specie HCOOH(a) having these considerations in mind. The monodentate character of the O-C-O bonding of these species agrees with rules proposed by Deacon et al [38] on splitting and positions of νas(COO) and νs(COO) for a large number of acetato complexes, as well as previous reports on the HCOOH/TiO2 adsorption system [31, 39]. It is clear from the spectra in Fig. 2 that the A40 sample presents a bonding environment that is intermediate between the two extreme adsorbate structures seen on the A60 and R6x80 samples, respectively, i.e. it favours monodentate coordination and HCOOH dissociation according to the scheme outlined in Fig. 5 but not bidentate HCOO coordination. The A25 sample and even more so P25 contain significantly more bidentate formate species. We attribute this to bidentate or higher metal coordination of HCOO bonded to c.u.s. sites, e.g. edges, steps and lattice defects which to a larger extent (than the larger particles) are present on the smaller particles (species II, III or IVb in Fig. 4).

Fig. 5 Reaction scheme depicting the conversion of specie Ia and Ib in Fig. 4 and the intermediate hydrogen bonded specie. To summarize the rutile R6x80 particles have a distinct spectral profile characteristic of a bridging bidentate HCOO adsorption structure (or more precisely mixtures of µ-formate A and B species). The A60 particles is mainly associated with liquid-like HCOOH, while the smaller A25 and A40 particles show a heterogeneous distribution of co-existing adsorbed HCOO and HCOOH species, each with its own spectral signature. The detailed surface structure on A25 and A40 manifested in a distribution of surface faces (Fig. 2 and Table 2) dictates whether dissociation occurs or not as depicted by the reaction scheme in Fig. 5. The role of c.u.s. sites becomes more pronounced for the smaller A25 (and P25) particles and can be attributed to distinct vibrational signatures associated with µ-formate species. Photo-degradation of formic acid. In Fig. 6 the concentration of adsorbed HCOOH and HCOO molecules as a function of irradiation time is shown. The surface concentration was determined by the integrated absorbance originating from the ν(OCO) and ν(C=O) bands for HCOO and HCOOH, respectively, given by the assignments in Table 3. It is clearly seen that the photo-reactivity is highest on the A40 anatase particles, while it is lowest on the R6x80 rutile and P25 particles. In the former case the photo-induced decomposition (PID) rate is determined to be 0.08 ML min-1 taking into account a minor contribution from molecular HCOOH desorption. In the latter cases only very slow degradation of HCOO occurs. On the P25 particles the monodentate species associated with the 1660-1680 cm-1 absorption bands on the anatase phase rapidly decomposes, while the bidentate species coordinated to both anatase and rutile phase particles are only slowly decomposed. Based on

214

Solid State Chemistry and Photocatalysis of Titanium Dioxide

comparisons with data on anatase particles exposing a high fraction of {101} faces (A60) and rutile particles exposing mainly (110) planes (R6x80) as well as rutile sample containing about equal ratios of (110) and (101) planes [6, 25], we find a correlation between the high photocatalytic reactivity of the A40 sample to the presence of minority {112}, {111}, {001} and {100} planes exposed on these particles. We have previously shown that monodentate HCOO/HCOOH species are more easily decomposed during UV irradiation than bidentate species on rutile nanoparticles [25]. It is therefore likely that bonding to minority surfaces on the A40 favors monodentate HCOOH/HCOO coordination (Fig. 5), which co-exist with adsorbed HCOOH species present on {101} surfaces. Thus the larger photo-reactivity is correlated to the concentration of monodentate HCOO/HCOOH species (species Ia and Ib in Fig. 2). On the A25 sample, a significantly larger portion of HCOOH dissociates to form bidentate coordinated HCOO species due to coordination to c.u.s. sites; these species exhibit a much smaller PID rate which is evident by comparison with the R6x80 rutile sample that almost exclusively exhibit bidentate HCOO bonding. The PID rates are summarized in Table 4.

Fig. 6 The concentration of adsorbed HCOOH and HCOO molecules on anatase and rutile TiO2 as a function of irradiation time. The photo-induced decomposition (PID) rate determined from the first 20 min of illumination assuming first-order kinetics is indicated in the figures. The desorption rate (DES) was determined from control experiments with the samples held in dark. Discussion Coordinatively unsaturated surface atoms (c.u.s.) are generally more reactive than the corresponding saturated atoms, since their valence number make them prone for additional coordination. The c.u.s. Tin+ cations at the TiO2 surface are acidic (Lewis acids sites) and want to form bonds with basic molecules, while the c.u.s. On- anions are basic (Lewis bases) and adsorb acidic molecules. While the acidic sites have proven to be fairly straightforward to characterize, the basic surfaces sites of

Solid State Phenomena Vol. 162

215

oxides have proven to be more elusive. Davydov [40] suggested that reactive adsorption of CO2 forming surface carbonates is a suitable molecule to probe the basic properties of oxide surfaces. In particular, the splitting of the asymmetric and symmetric carbonate stretching vibrations, ∆νass≡νas(COO)–νs(COO), of the carbonate molecule has been correlated to the strength of the basic sites. For example, the M-O covalency of a lattice O atom involved in monodentate coordinated carbonate is lower than that for a bidentate carbonate. According to Nakamoto ∆νas-s is about 300 cm-1 for bidentate carbonate [41]. This value increases with increasing covalent character of the MO bond, i.e. decreasing basicity of the O atom. Since the surface atoms on TiO2 nanoparticles are expected to deviate from stoichiometry to a larger degree than bulk terminated samples, e.g. with excess oxygen possessing different valency, the above considerations predict that ∆νas-s for CO2 adsorbed such that it forms a bidentate carbonate molecule should be higher if it is coordinated to a three-fold coordinate O atom than a two-fold O atom. Similarly, the surface properties should be influenced by O adatoms with comparably strong basic character. In analogy with metal-carbonate coordination, Deacon et al [38] have analyzed ∆νas-s and the positions of νas(COO) and νs(COO) for a large number of acetato complexes. By comparing ∆νas-s of the free aqueous state (“ionic”) and metal coordinated complexes (“metal coordinated”), respectively, they observed the following correlations (Deacon’s rules): ∆νas-s (metal coordinated) > ∆νas-s (ionic): monodentate coordination ∆νas-s (metal coordinated) < ∆νas-s (ionic): bidentate chelating or bridging coordination ∆νas-s (metal coordinated) ∆νas-s (ionic). In our case we have ∆νas-s (ionic)=240 cm-1 (Table 2). However, the inverse of other two correlations are not always obeyed, and it is not straightforward to distinguish the bidentate chelating from the bridging coordination solely based on the magnitude of ∆νas-s. In general, these empirical rules should be complemented by additional structural data. This was also concluded in a recent spectroscopic and theoretical study of formate and acetate adsorption on (large) rutile TiO2 nanoparticles [31]. Moreover, when studying adsorption of carboxylic acid on oxide surfaces interpretation of vibrational spectra may be complicated from a practical viewpoint by formation of surface esters and condensation products of carboxylate ions, since these have spectral features close to the carboxylate surface complexes. Additional complications arise due to the heterogeneity of possible carboxylate species discussed above (cf. Fig. 4). The above empirically based findings have recently been rationalized by A. Selloni and coworkers using density functional calculations [5, 42]. Scrutinizing their findings and translating them into the context above the following general features governing the bonding and reactivity of HCOOH on anatase TiO2 surfaces can be discerned: (i) Surfaces with a large density of coordinatively unsaturated surfaces sites promote dissociation and bidentate bonding due to interactions with basic surface O atoms present on these surfaces. (ii) Surface planes exhibiting strained configuration of surface atoms, i.e. large Ti-O-Ti bond angles within the surface plane [42] make the surface O atom become more or less basic, and hence more or less reactive. The in-plane Ti-O-Ti bond angles differs however considerably among the (101), (111), (100), (112) and (001) surfaces, with the (101) exhibiting the lowest Ti-O-Ti bond angle within the surface plane (along the [ 10 1 ] direction), while the others have a strained angle (102 vs. 155 degrees for the bulk terminated structures, respectively). (iii) Small nearest-neighbour (nn) Ti-Ti distances [5] promote bridging bidentate coordination. The nn Ti-Ti distance in bulk terminated (un-reconstructed) anatase (101) is 3.783 Å, and it is the same in the anatase (001), (100) and (112) surfaces, while it is much shorter (2.953 Å) in rutile (110), and also between rows of mirror reflected domains in brookite

216

Solid State Chemistry and Photocatalysis of Titanium Dioxide

(210). From the data in Table 2, Fig. 2 and 6 we find that there is a correlation between crystal face distributions, nn Ti-Ti distance, in-plane Ti-O-Ti bond angles on exposes crystal faces. The data are presented in Table 4. Possible surface reconstructions will of course modify these predictions, but they do give an indication of the importance of the basicity of surface O atoms and nn Ti-Ti distance on the various crystal faces. In particular we see that a short nn Ti-Ti atoms present on the {110} surfaces correlates with a smaller ∆νas-s value, formation of bridging bidentate formate and a small PID rate. We note that it previously has been shown that a shorter nn Ti-Ti distance present also on brookite TiO2(210) in the narrow gaps along the [001] direction in the herringbone structure separating mirror reflected rows [5], stabilizes HCOO bonded in bidentate configurations. The spectra shown in Fig. 2 clearly demonstrate that the vibrational structure and ∆νas-s vary significantly among the various anatase particles (cf. Table 2 and 4). The morphology analysis suggests that the different anatase nanoparticles have different face distribution. In fact, there are strong indications that the small particles expose A40 and A25 expose surface planes not present at all on A60. Moreover, the smaller particles inevitably expose a larger fraction of c.u.s. atoms due to edges, corners, surface defects than the larger particles. Moreover from the FTIR data and the corresponding reactivity measurements (Fig. 3 and 6) it is evident that the reactivity increases in the order of increasing ∆νas-s towards monodentate bonding of HCOO/HCOOH(a) (Table 4). Thus we may propose that a decreasing basicity of the oxygen surface atoms favours a monodentate bonding scheme of adsorbed HCOO/HCOOH species. We conjecture that it is beneficial for a high photoreactivity to maintain moderate bonding of adsorbed HCOO/HCOOH, which is intermediate between anatase (101) which is non-reactive, and rutile (110) which is too reactive and form stable bidenate HCOO species. On A40, and to some extent on the A25 particles this is accomplished by exposing suitable combinations of reactive {112}, {111}, {100} and {001} faces of anatase. In contrast on the {101} planes on anatase molecular adsorption of HCOOH occurs forming hydrogenbonded, liquid-like clusters whose decomposition is limited by diffusion of photo-generated oxygen radicals. Face communication by means of surface diffusion between the different crystal faces (and the non-reactive {101} faces) may also contribute to the observed reactivity thus providing an explanation for an optimum particle size which correlates with an effective diffusion length on the nanocrystals [43]. At present we cannot ambiguously correlate the inferior activity on A25 and P25 to unfavourable crystal faces or a relatively larger concentration of c.u.s. atoms. Whatever the cause, our data do however show the intrinsic difficulties of preparing small size (d