Photoelectrochemical behaviour of anatase nanoporous films: effect of the nanoparticle organization

Abstract

The photoelectrochemical behaviour of anatase thin films with different nanoarchitectures and the same active surface area (or thickness) has been studied in acidic media in the absence and in the presence of formic acid. The electrodes were composed of either wire-like nanocrystal aggregates or commercial TiO₂ nanoparticles. Cyclic voltammetry in the dark reveals a larger trap concentration in the band gap for the nanoparticulate (NP) electrodes, which can be ascribed to a larger number of intergrain boundaries. Also under illumination, the behaviour for both types of anatase structures significantly differs: water photooxidation arises at more negative potentials for the nanocolumnar (NC) electrodes. In the presence of an efficient hole acceptor such as HCOOH, significantly larger photocurrents were noted for the NC films as compared with those for the NP electrodes, with the photocurrent onset also shifted towards more positive potentials for the latter. These results point to a diminished electron recombination, which can be related with a smaller concentration of intergrain boundaries, together with a more efficient HCOOH hole transfer for the wire-like nanocrystal aggregate architecture. In addition, the oxygen reduction reaction is also favoured in the case of NC electrodes.

---

Introduction

In the last decade, the number of publications on the materials science aspects of titanium dioxide has considerably increased. Particularly, nanoporous TiO₂ thin films have attracted great attention due to their particular physical and chemical properties, which differ from their analogous compact films.^1,2 In addition, their applicability in multiple fields such as energy generation and storage,³ environmental remediation⁴ or even, as sensors in analytical devices⁵ renders their study particularly attractive. For most of these applications, one-dimensional nanostructures such as nanorods, nanowires and nanotubes possess, in principle, several advantages, including considerable exposed surface area, fast electron transport⁶ and long electron lifetimes.⁷ In fact, three-dimensional nano- and micro-structures with well-controlled morphology and architecture have been a hot research topic in the last few years.

Many different methods and techniques have been developed to prepare one-dimensional TiO₂ related materials. Among them, the anodization of titanium metal has been reported to be a simple route to obtain well-defined and vertically aligned nanotubes directly grown on a conducting substrate.^8,9 The fabrication of TiO₂ nanowire arrays has been also developed via the assistance of anodic alumina membrane templates.^10,11 On the other hand, it is known that TiO₂ nanotubes and nanowires can be synthesized in relatively large quantities by the reaction of a TiO₂ precursor and a concentrated NaOH solution under moderate hydrothermal conditions. In such a case, the one-dimensional structures are obtained as a nanopowder. Only recently have some publications appeared, in which nanowire TiO₂ thin films have been prepared from a calcinated Ti foil in NaOH aqueous solution.^12–14 However, the practical use of these nanopowders as suspensions in photocatalytic applications is not straightforward. From an engineering point of view, in spite of their superior properties, their morphology and size would complicate their separation from the effluent after decontamination. For this reason, the use of a thin nanostructured film is advantageous.

A common strategy for the generation of thin films from the as-prepared nanoparticles consists in the wet deposition of the corresponding nanoparticle suspension on a support, followed by a thermal treatment and sintering, most of the time at 450 °C, for half an hour in air. This thermal treatment is required to favour physical contact among the nanocrystals as well as between them and the conducting substrate. In fact, the lifetime and the diffusion coefficient of the photogenerated charge carriers increase as a result of annealing treatment.¹⁵ Nevertheless, the control of the particle–particle interconnections is not straightforward despite their importance. Most of the time, the sintering of a random nanoparticle agglomeration results in a network of crystallographically misaligned nanocrystals. Recently, an electrochemical study of nanocrystalline rutile thin films has put in evidence that the particle connections, also called grain boundaries, can act as electron trap states.¹⁶ Obviously, trapping of electrons increases the residence time of the electrons in the nanoporous film and thus, also the probability of losses due to recombination processes (such as with photogenerated holes or with electron acceptor species from the electrolyte) in their way to the back-contact. Accordingly, grain boundaries are expected to act as regions of enhanced recombination because they can behave as electron trap states.

On the other hand, during photocatalytic processes, both oxidation and reduction reactions take place simultaneously. For this reason, it is advisable to independently study both processes by electrochemical methods. In principle, in this way, it would be possible to determine the limiting reaction rate. In addition, the electrochemical methods allow us to obtain information not only about the location of electron trap states but also about their concentration. However, the task of comparing different samples is not straightforward. It is known^17,18 that several factors such as particle size, surface area, crystallinity, morphology and trap state concentration affect the photocatalytic activity of TiO₂ and some of these parameters may change when preparing a thin film.^19,20 Therefore, the precise control of these parameters can be considered essential for drawing any reasonable conclusions.

In this contribution, we study the photoactivity of two different anatase nanocrystalline electrodes along the lines of a previous work on rutile quantum-size nanowire electrodes:²¹ (i) a thin film electrode composed of randomly distributed wire-like nanocrystal aggregates prepared by the hydrothermal treatment of a TiO₂ precursor in NaOH, followed by neutralization and another hydrothermal treatment of the intermediates, and (ii) an electrode composed of commercial TiO₂ nanoparticles. We have focused on the effect of the morphology on the photoelectrocatalytic behaviour. We studied not only water photooxidation but also the photooxidation of HCOOH, which has been chosen as a model hole acceptor. The role of the grain boundaries in the photoelectrochemical behaviour is discussed, employing samples with either similar active surface area or film thickness. We find that the use of wire-like nanocrystal aggregates minimizes the number of detrimental grain boundaries that an electron needs to cross when travelling to the back-contact, thereby improving the resulting photoelectrochemical behaviour.

Experimental

Anatase TiO₂ nanocrystals, constituting wire-like aggregates, were synthesized by a hydrothermal soft chemical transformation in neutral solution of large diameter hydrogen titanate wires, as previously described.²² The electrodes were prepared from the as-synthesised nanopowder by mixing 15 mg of TiO₂ with 5 μL of acetylacetone, 5 μL of Triton X-100 and 60 μL of water. Approximately 5 μL of the resulting slurry were deposited over a ∼3 cm² FTO substrate (Pilkington TEC15, 15 Ω cm⁻² resistance). The TiO₂ films were sintered at 350 °C for 30 min in air. The electrical contact was made using silver epoxy cement (Electron Microscopy Sciences) and a copper wire, and protected with an epoxy resin (Poxipol® 10 min).

Nanoparticulate electrodes were also prepared by spreading an aqueous slurry of commercial anatase TiO₂ nanoparticles (Alfa Aesar) in a similar way. For these samples, after TiO₂ deposition, the films were annealed and sintered for 1 h at 450 °C in air. Afterwards, the coated conducting glass substrates were electrically contacted, as previously described.

For the photoelectrochemical experiments in the presence of oxygen, a TiO₂ compact layer was prepared on the FTO substrate by spin coating (1000 rpm) a 0.1 M Ti(IV)isopropoxide solution in isopropanol, and subsequent thermal treatment at 500 °C for 0.5 h in air, prior to the deposition of the nanoporous anatase film.

Morphological analysis of the electrodes was carried out by Transmission Electron Microscopy (TEM) employing a JEM-2010 (JEOL) microscope, equipped with a MegaView II camera (SIS) after detaching the oxide films from the substrates. Scanning Electron Microscopy (SEM) images were obtained with a HITACHI S-3000N equipment. Raman spectra were recorded with a LabRam spectrometer (Jobin-Yvon Horiba) at 632.8 nm. XRD patterns of the nanoporous TiO₂ thin films were collected using a Philips X'Pert diffractometer, using Cu Kα radiation. Diffuse reflectance spectra were obtained with a Shimadzu UV-2401 PC spectrometer, equipped with an integrating sphere. In such a case, BaSO₄ (Wako) nanopowder was used as the background.

Photoelectrochemical measurements were performed at room temperature in a three-electrode cell equipped with a fused silica window. All potentials were measured and referenced with respect to a Ag/AgCl/KCl(sat) reference electrode, whereas a Pt wire was used as a counter electrode. An Autolab PGSTAT30 potentiostat was used to perform the measurements. The full output of a Xenon arc lamp of 150W was used as the illumination source (Spectral irradiance data can be found in the ESI.†) The electrodes were illuminated from the electrolyte side (EE configuration). The applied light irradiance was measured with an optical power meter (Oriel model 70310) equipped with a bolometer (Ophir Optronics 71964). In all experiments, a N₂-purged 0.1 M HClO₄ solution (Merck, 70%, Suprapur) in ultrapure water (Millipore Elix 3) was used as the working electrolyte. The Incident Photon to Current Efficiency spectra were measured using a 300 W Xe arc lamp (Oriel) equipped with a monochromator (Oriel model 74000) and a photodetector (Thermo Oriel 71608) connected to the optical power meter (Oriel model 70310) to measure the light intensity at each wavelength.

Results

Nanostructured electrodes consisting of anatase wires were prepared using the as-synthesised powder obtained from a hydrothermal treatment of hydrogen titanate structures. The nanopowder is characterized, on the one hand, by its pure anatase crystal structure and on the other hand, by its well-defined morphology.²² Films with a thickness about 5 μm were prepared by wet deposition of the powder onto a FTO conducting substrate and subsequent thermal treatment at 350 °C. This treatment allows the film to sinter, thereby improving its mechanical properties, because of the formation of grain necks and/or particle–particle interconnections, without a significant change of the particle morphology (see below).

Fig. 1a shows a TEM image of TiO₂ wire-like aggregates after detaching them from a thin film supported on FTO-coated glass (once sintered). Different types of one-dimensional structures can be observed. Both nanowires and nanorods with smooth surfaces and small sizes can be distinguished from others, larger in diameter and noticeably rougher, together with some nanoparticle clusters. The morphology of the electrode sample agrees well with the morphology of the powder used to prepare the thin film (see Fig. S1 of the ESI†). Fig. 1b illustrates an individual large wire of about 1.5 μm length and 240 nm in diameter. It is composed of small particles, with some as well-defined nanorods. Fig. 1c shows that these nanorods are formed by nanocrystals interconnected and attached to each other, while keeping the same orientation, as revealed by the observed lattice fringes. As previously demonstrated for the nanopowder, anatase TiO₂ nanoparticles, constituting the wire-like aggregates, are essentially aligned along the [101] direction.²² The wet deposition method employed in the preparation of the films preserves the original morphology of the constituent 1D-like agglomerates and yields a film composed of randomly oriented columns as demonstrated by SEM (Fig. 1d).

Fig. 1 TEM micrographs of TiO₂ wire-like aggregates with different magnifications (a, b and c) and commercial TiO₂ nanoparticles (Alfa Aesar) (e). (d) shows a SEM image of the cross section of a TiO₂ wire-like aggregate electrode.

The photoelectroactivity of the wire-like aggregates was studied and compared with that of commercial TiO₂ anatase nanoparticles (Alfa Aesar) of similar particle size. Fig. 1e shows a representative TEM image of the Alfa Aesar sample, composed of crystallites of spherical shape of about 20 nm in diameter.²³

The crystalline structure of both samples (e.g. particles and wire-like aggregates) was verified by Raman spectroscopy. The spectra are essentially identical and show the presence of the characteristic bands expected for anatase, namely 395 cm⁻¹ (B_1g), 514 cm⁻¹ (B_1g), and 638 cm⁻¹ (E_g)²⁴ (see Fig. S2 of the ESI†). In agreement with the Raman spectra, XRD patterns of the nanoporous thin films also show the characteristic peaks of the hexagonal anatase phase (space group I4₁/amd; JCPDS File No. 21-1272) (see Fig. S3 of the ESI†). No other peaks corresponding to other titania phases such as rutile or brookite are present in either the Raman spectra or the X-ray diffractograms, indicating that the samples were composed of pure anatase phase.

In Fig. 2, we show some voltammetric profiles of nanoparticulate (NP) and nanocolumnar (NC) anatase electrodes obtained in 0.1 M HClO₄ in the dark. Fig. 2a corresponds to electrodes having the same capacitance in the low potential region. This was achieved by adjusting the relative thickness. Fig. 2b corresponds to electrodes of approximately the same thickness. A rather symmetric profile is recorded in all cases. The reversibility is maintained when working with different scan rates (see Fig. S4 of the ESI†), which indicates the absence of any significant resistance or faradaic process.

Fig. 2 Cyclic voltammetry of a nanostructured NC electrode (5 μm thickness) and a NP electrode (Alfa Aesar) with the same active surface area (0.7 μm in thickness) (a) and with the same film thickness (5 μm) (b) in the dark. Electrolyte: nitrogen purged 0.1 M HClO₄, scan rate: 20 mV s⁻¹. Cyclic voltammetry data with a magnification factor of 20 were measured at a scan rate of 10 mV s⁻¹.

Two possible interpretations can be found in the literature for this capacitive behaviour. On the one hand, these charges have been related to the accumulation of electrons in the oxide conduction band (CB) and/or to the filling of a surface state distribution just below the CB.²⁵ Interestingly, the density of states deduced from the experimental cyclic voltammetric response of the nanocolumns follows an exponential distribution (Fig. S4a†). This feature agrees well with the behaviour previously observed for anatase nanoporous electrodes.²⁶ Nevertheless it differs from the characteristics of quantum-sized rutile nanowire electrodes whose density of states, as deduced from the cyclic voltammetry data, is far from following an exponential distribution.¹⁶ Such a difference in behaviour can be correlated with the different crystal structure²⁷ and/or more probably, to the quantum size effect.^16,26 In any case, due to the charge accumulation, protons from the electrolyte are either adsorbed or inserted into the crystalline structure in order to compensate for the electronic charge.²⁸

Recently, we have demonstrated that the integrated charge in the accumulation region is proportional to the active interfacial area of the TiO₂ nanostructure in contact with the electrolyte.¹⁶ In fact, this relationship is compatible with both explanations (filling of either CB or surface states), as adsorption is a surface-related phenomenon and proton insertion is expected to take place in a region near the oxide surface. Fig. 2a shows the cyclic voltammetry curves for a NP electrode of about 0.7 μm in thickness and for a NC electrode of about 5 μm in thickness. Despite the different thickness, the accumulated charge is very similar for both samples. Consequently, we can expect that both electrodes have the same active surface area. For comparison, Fig. 2b shows the cyclic voltammetry of an anatase NP electrode with a similar film thickness as that of the NC electrode (∼5 μm). In such a case, the currents associated with the accumulation region are much larger for the NP electrode, which agrees well with the increase in surface area.

Interestingly, in addition to the reversible charge accumulation region (low potential region), a voltammetric signal associated with band gap states can be clearly observed at ∼0 V_Ag/AgCl, particularly in the 5 μm NP film. These states are also present in the thin NP film, but in a smaller amount, as shown in the magnification (Fig. 2a). Nevertheless, they are virtually absent in the case of the NC electrode.

Previous studies indicate that such electron traps can be associated with interparticle grain boundaries (GBs).^16,21,23 Although the wire-like aggregates are mainly composed of small nanoparticles as shown in the TEM images, the concentration of traps is negligible. This can be understood on the basis of the well-defined orientation of the nanocrystals constituting the nanocolumns. As indicated above, the nanocrystals are interconnected without the apparent presence of misorientations. Consequently, there are no real GBs between the nanocrystals.

It is important to mention that the NP electrodes were submitted to a thermal treatment at a temperature higher than that employed with the NC electrodes, which in principle would favour the formation of grain necks and good interparticle connections. The treatment at 450 °C, instead of 350 °C, was necessary in order to assure a good interparticle connectivity and an ohmic contact with the FTO substrate. A treatment at 350 °C yielded electrodes with a measurable resistance associated with the TiO₂/FTO interface (see Fig. S5 of the ESI†). In spite of the fact that annealing at a high temperature could improve the contact among the particles, the density of electron traps is larger for the NP film, even when comparing films with the same active surface area (Fig. 2a).

NP electrodes as well as NC electrodes with the same electrochemical active surface areas are photoactive towards water photooxidation in nitrogen purged 0.1 M HClO₄ solution, as shown in Fig. 3. At positive potentials, the photocurrents measured for the nanoporous electrodes are similar for both morphologies, although the film thicknesses are considerably different. Importantly, the photocurrent onset potential shifts towards more positive potentials for the NP electrode, and the photocurrent starts to level off at the potential where the grain boundary peak appears.

Fig. 3 Cyclic voltammetry of a NC electrode and a NP electrode (Alfa Aesar) with the same active surface area under polychromatic illumination (660 mW cm⁻²) and in the dark. Electrolyte: nitrogen purged 0.1 M HClO₄; scan rate: 20 mV s⁻¹; film thicknesses: 5 μm (NC) and 0.7 μm (NP).

Fig. 4 shows a series of voltammograms for NC and NP electrodes having same active surface areas (Fig. 4a and b) or approximately the same film thickness (Fig. 4a and c), recorded under UV-Visible illumination in the presence of different formic acid concentrations. As observed, the introduction of the organic substance gives rise to the development of larger photocurrents. Under these conditions, the charge recombination is reduced, because the holes can be effectively captured by the organics. In addition, the photooxidation of formic acid generates intermediate radicals that are able to inject an electron into the CB of the TiO₂ semiconductor (current-doubling effect).²⁹ Moreover, the introduction of formic acid leads to a shift of the photocurrent wave towards more negative potentials.

Fig. 4 Linear voltammograms of a NC electrode (6 μm in thickness) (a) and NP (Alfa Aesar) (1 μm in thickness) electrodes with the same surface area (b) and the same electrode thickness (6 μm in thickness) (c) in the presence of different formic acid concentrations (indicated in mM), under polychromatic illumination (660 mW cm⁻²). Electrolyte: nitrogen purged 0.1 M HClO₄; scan rate: 20 mV s⁻¹.

It is worth mentioning that in the case of the NC and the NP thin film electrodes (with similar active surface areas), the photocurrent shows a well-defined plateau in the positive potential region, which indicates an effective transport of the organics within the nanoporous structure. Nevertheless, the transport of organics in the NP structure seems more difficult. This fact is revealed by comparing the linear voltammograms of NP and NC films with the same thickness. As the formic acid transport is slower than the hole photogeneration at the NP film, in this case, the photocurrent shows a peak prior to the plateau (Fig. 4c).

Photocurrents are larger for the NC electrodes than for the NP ones regardless of working with samples possessing either similar active surface area or film thicknesses. In addition, for a defined HCOOH concentration, the onsets of the photocurrents are shifted towards more negative values in the case of the NC electrode (Fig. 5) which is in agreement with the results obtained for water photooxidation. It is remarkable that the photocurrent onset is also dependent on the NP film thickness (in agreement with the potential where the GB states are filled). It should also be highlighted that the potential needed to attain the saturation photocurrent is larger for the NP electrodes than for the NC ones.

Fig. 5 Photocurrent onset values deduced from Fig. 4 as a function of the formic acid concentration for NP and NC electrodes, having the same film thickness or active surface area.

As the photocatalytic properties of metal oxide nanoparticles depend not only on the photooxidation activity properties but also on the catalytic properties for the oxygen reduction reaction, we have studied this process in the dark. In Fig. 6a, cyclic voltammograms for both types of anatase electrodes with the same active surface area in 0.1 M HClO₄, and in the presence of oxygen are shown. As observed, oxygen reduction occurs more easily on NC electrodes as judged from the more positive current onset (lower oxygen reduction overpotential). In agreement with this result, a 28% diminution of the photocurrent for water photooxidation in the presence of oxygen is measured for the wire-like aggregates, which is considerably larger than the 12% diminution found for the nanoparticles (see Fig. S6 of the ESI†). A larger diminution in the photocurrent corresponds to an enhanced recombination of the photogenerated electrons with oxygen molecules on their way to the back-contact.

Fig. 6 Cyclic voltammetry of a NC electrode (∼5 μm thickness) and a NP electrode (Alfa Aesar, ∼0.7 μm thickness) in the dark in the presence and in the absence of oxygen (a). The inset shows the detail of the current onset potential for the NP electrode. Open circuit potential measurements of a NC and a NP electrode in O₂-purged 0.1 M HClO₄ in the presence of formic acid (b). The experiments were performed using a compact TiO₂ blocking layer deposited on the FTO substrate.

To complete the photoelectrochemical study, we have measured the photopotential decay in the presence of molecular oxygen and formic acid. Fig. 6b shows the results obtained for both types of samples. Upon illumination, the open circuit potential shifts towards more negative values, which reflects the accumulation of electrons in the TiO₂ nanostructure. When the light is cut off, the potential relaxes back to the original value, with the relaxation process occurring faster in the case of the NC electrode. In the presence of HCOOH, the electrons accumulated in the TiO₂ columns also exhibit a higher reactivity towards oxygen, in agreement with results obtained from voltammetry (Fig. 6a).

Discussion

In this work, two types of nanoporous anatase electrodes with different morphologies have been studied. Both are composed of aggregated nanoparticles, but their assemblage is very different. In the case of commercial quasi-spherical nanoparticles, they are randomly distributed, while in the case of synthesized nanoparticles, they are self-assembled, forming wire-like aggregates. In addition, for the latter, the nanocrystals are essentially aligned along the [101] direction and no crystallographic mismatches between adjacent nanoparticles have been observed by TEM.

The presence of GB is clearly evidenced by cyclic voltammetric experiments in the dark,¹⁶ where the mismatches in the crystal structure appear as electron traps at the band gap. As shown in the Results section, these states are present in the NP TiO₂ films, but they are virtually absent in the case of NC electrodes. We expect that these GB play an important role in governing the photocurrent. On the one hand, the presence of electron trap states may retard the electron transport, thereby increasing the residence time in the TiO₂ structure. On the other hand, they may act as recombination centres.

Discussion section is organised as follows. First, the photocurrent properties for NC and NP electrodes will be analysed and discussed on the basis of the same active surface area or film thickness (i.e., similar light absorption). The comparison of the magnitude and the location of the photocurrent onset allows one to identify the deleterious effect of the GB and consequently, the better performance of the wire-like nanocrystal aggregates as compared with the photoelectrode material. Secondly, the oxygen reduction experiments will be analysed, in order to elucidate if the NC sample is also a good candidate for conventional photocatalysis.

Both electrodes, namely NC and NP thin films, develop a photocurrent under UV-visible illumination in 0.1 M HClO₄, which can be ascribed to water photooxidation. The main feature that distinguishes their behaviour is the potential corresponding to the photocurrent onset: the photocurrent appears at more positive potential in the NP electrodes (Fig. 3). Interestingly, this potential correlates well with the energetic location of the GB electron traps. The interpretation of such a result should consider that the collected electrons (i.e., the photocurrent) are driven to the back-contact by diffusion according to the multiple trapping/detrapping mechanism.³⁰ Within this model, the electron transport is thermally activated: the electrons are trapped and detrapped at band gap states, and only free electrons contribute to the photocurrent (Fig. 7). As the chemical capacitance, i.e. the charge integrated in the accumulation region, is proportional to the density of states,³¹ we can expect that these states exhibit a similar distribution for the NC and the NP TiO₂ samples of Fig. 2a. However, a relatively high concentration of quasi-monoenergetic states associated with GB is also present in the NP electrode (according to the sketch in Fig. 7). When a sufficiently negative potential is applied, electrons populate these states, leaving GB negatively charged. This charge can be only partially screened by species in solution, due to its particular location. A potential barrier will exist at the vicinity of the GBs, thereby blocking the diffusion of majority carriers and thus, the photocurrent. This potential barrier will also attract minority carriers. As photogenerated holes possess a non-negligible mobility,³² even surface trapped holes could recombine at GB. This recombination pathway will be particularly important in the case of the water photooxidation process, which requires the accumulation of surface trapped holes.^33,34 Therefore, a significant photocurrent will develop for NP electrodes once the applied potential is positive enough to provoke the emptying of the GB band gap states. By contrast, the GB states are virtually absent for NC electrodes which causes the photocurrent to develop at more negative potentials.

Fig. 7 Schematic illustration of the effect of band gap grain boundary traps on both anatase NP (a) and NC (b) electrodes.

Obviously, the recombination of electron–hole pairs at the GBs can be drastically diminished if the holes are effectively consumed in a competitive process. This is the case in the presence of formic acid, which acts as an efficient hole acceptor. In contrast with water photooxidation, the accumulation of surface trapped holes is not needed for the reaction to proceed. Formic acid can be specifically adsorbed at the Ti(IV) sites, competing with water molecules and facilitating the direct interfacial hole transfer.³⁵ Consequently, the photocurrent onset potential shifts towards increasingly negative values for the NP electrode as the formic acid concentration grows (Fig. 5). The photocurrent can now be detected even at potentials at which the GB states are already filled. In the case of NC electrodes, a shift in the photocurrent onset potential can also be observed, although to a smaller extent.

On the basis of the discussion above, one should expect that the GB traps would act as recombination centers under illumination (even at potentials positive to the GB states) as photogenerated electrons can accumulate at the GB states. In such a case, the photocurrent should be smaller for NP than for the NC electrodes. However, the comparison of the photocurrents is not straightforward. In the case of electrodes with the same surface area, the electrode film thicknesses are considerably different, their light absorption capacity is dissimilar, and as a result, the electron–hole pair generation will be altered. On the other hand, if the film thickness is kept constant, the active surface area will also be very different, according to Fig. 2. In addition, another important related feature that renders even more complex the interpretation of the photocurrent is the possible shift of the light absorption onset. Although both NC and NP samples are composed of anatase nanoparticles, a different band gap value was calculated from their reflectance (R) spectra. Assuming that the films are thick enough, the Kubelka-Munk transformation can be applied (f(R)), and, as titanium dioxide is an indirect band gap semiconductor, we can expect that:

(f(R)hν)^1/2 = A(hν − E_g) (1)

with hν as the photon energy. Therefore, a plot of (f(R)hν)^1/2 against hν should be linear, allowing us to derive the band gap (E_g) (Fig. S7†). A similar behavior is expected when measuring the Incident Photon to Current Efficiency (IPCE) instead of f(R) as shown in Fig. S7.† A band gap about 3.1 eV was experimentally found for the NP, while a value as high as 3.3 eV is found for the NC thin film. In any case, it is remarkable that in the presence of high enough HCOOH concentrations, the photocurrent is larger for the NC than for the NP sample, although the NC thin film harvests a smaller portion of the incident light spectrum. This observation illustrates the better behaviour of the wire-like nanocrystal aggregates as a photoanode material.

A larger photocurrent is a result of several factors: (i) a better electron transport inside the film (the GBs may retard the electronic transport since they correspond to deep traps and the electron residence time in traps will increase with their depth); (ii) an enhanced hole transfer or (iii) a diminished electron–hole recombination rate (due to the absence of GB). In order to rationalize the results, we have used a previously published kinetic model that assumes a direct hole transfer for HCOOH photooxidation and takes into account the possibility of trapping holes at terminal oxygen lattice sites.^21,35–37 In such a case, the continuity equation can be analytically solved for obtaining the maximum photocurrent value (in the plateau) according to:²¹

with J the electron flux (cm⁻² s⁻¹) (directly proportional to the photocurrent), Φ₀ the incident photon flux (cm⁻² s⁻¹), a the surface density of formic acid adsorption sites (cm⁻²), k^d_ox the rate constant for direct hole transfer oxidation (cm² s⁻¹), N the concentration of hole trapping sites at the surface (cm⁻²), k₁ the rate constant for surface hole trapping (cm² s⁻¹), K the Langmuir isotherm adsorption constant (cm³), [HCOOH] the formic acid concentration in solution (cm⁻³), α the absorption coefficient (cm⁻¹), and d the electrode film thickness (cm).

Fig. 8 gathers the experimental values for the saturation photocurrent density corresponding to the oxidation of HCOOH on a NP electrode and a NC electrode with the same film thickness, respectively, as taken from Fig. 4. The lines correspond to the best fit of the experimental results obtained according to eqn (2). It should be emphasized that only one parameter is fitted, (ak^d_ox/Nk₁), which can be interpreted as the ratio of the cross sections for hole capture at an adsorbed formic acid molecule and at an intrinsic surface site (e.g. terminal oxygen lattice sites). A larger value is obtained for the NC (ak^d_ox/Nk₁ = 0.078) as compared with the NP electrode (ak^d_ox/Nk₁ = 0.046). As this result is not dependent on the electron diffusion coefficient, it indicates an enhanced hole transfer towards formic acid photooxidation at the NC electrodes. (It is worth noting that despite the different experimental energy band gap, we have considered the same absorption coefficient for both samples, as both are composed of anatase nanoparticles. If we had considered the difference in the light absorption edge, a slightly larger value for the NC ak^d_ox/Nk₁ parameter would have been found).

Fig. 8 Saturation photocurrent densities at 0.4 V as a function of formic acid concentration for NC and NP electrodes with the same thickness (∼6 μm) in N₂ purged 0.1 M HClO₄ (Fig. 4). The lines indicate the best fit for the photocurrent, with α = 5 × 10³ cm⁻¹, Φ₀ = 6.6 × 10¹⁶ cm⁻² s⁻¹, K = 1.0 × 10⁻²⁰ cm³.

One may attribute the difference in activity to the different adsorption properties. However, we have fitted the results with the same adsorption constant for both samples. Without taking into account surface states, major differences in the geometry of the surface sites could exist, depending on the exposed crystal faces. However, for more spherical nanoparticles, one would expect a larger fraction of surface sites with tetrahedral coordination for titanium atoms. Species adsorbed at titanium surface sites of low coordination are probably more reactive than those adsorbed at sites of high coordination. However, this argument would favour the photocatalytic activity for the nanoparticles, which is contrary to both the experimental and simulated results. In conclusion, the larger photoactivity for the NC sample can be explained by considering simultaneously their particular three-dimensional network configuration virtually free of GB as well as a more effective hole transfer in the case of formic acid photooxidation.

There are other parameters related with the electron transport that could also help to explain the larger photocatalytic efficiency of the NC sample. Each time that an electron is either photogenerated or transferred to a wire-like aggregate, it can travel up to one micrometre without crossing any GB. Even when the NCs are randomly distributed, the electrons will follow a less tortuous path towards the back-contact, crossing smaller number of GB than for a NP network. There are also other parameters that may be relevant and that are specifically dependent on the geometry. When the wire-like aggregates are immersed in solution, the outer surface of the aggregates is depleted of carriers, causing a surface band bending in the radial direction, while the central region of the aggregates becomes a conducting tubular region.³⁸ This structure is ideal for channelling the electrons far away from the photogenerated holes, preventing recombination at the surface. In the case in which the wires are directly connected to the substrate, the electron collection is improved. This fact can also explain the better behaviour of the NC samples as photoanodes. In any case, transport probably is not the overall determining factor, as in the presence of formic acid, the electron diffusion length should be large. In addition, the model used in the photooxidation is not dependent on transport either.

According to the previous discussion, one will always expect a larger photoactivity for NC photoanodes. Nevertheless, it should be mentioned that in the case of water photooxidation, the particularities of the mechanism at the molecular level could induce a deleterious effect in the photocurrent of the NC samples. As proposed by Nakato, the oxygen photoevolution reaction is initiated by a nucleophilic attack of a H₂O molecule onto a surface-trapped hole, accompanied by bond breaking.^33,34 This particular mechanism makes necessary the presence of kinks and steps and the accumulation of photogenerated holes. Probably, in the case of NC electrodes, due to the particular morphology of the sample, the number of sites that can photooxidize water is highly diminished, by comparison with commercial anatase nanoparticles, which are more spherical. In fact, in the case of smooth well-defined surfaces such as single crystals, water photooxidation induces the generation of defects (photoetching).³⁹ This could be the reason why the photocurrent is similar for water photooxidation when comparing electrodes with the same active surface area (Fig. 3) and only slightly larger for the NC electrode when they have the same film thickness (ESI,† Fig. S8). Moreover, the photocurrent developed for water photooxidation is also affected by the reactivity of CB electrons with either photoevolved oxygen or oxygenated species, which is also dependent on the anatase morphology (see below).

Up to now, we have demonstrated that NC anatase has promising photoanode properties: e.g. larger photocurrents with more negative onset potentials. The question that remains is whether it is also a good photocatalyst candidate. During photocatalysis, both oxidation and reduction take place simultaneously. NC anatase will be a promising photocatalyst candidate only if the oxygen reduction reaction takes place sufficiently fast. A good indication of its potential use is the fact that the current associated with oxygen reduction arises at a more positive potential for the NC electrode as compared with the NP electrode. In fact, for the latter, the oxygen reduction current begins at the potential corresponding to the filling of grain boundary trap states.

During the electrochemical oxygen reduction experiments, the Fermi level of the conducting substrate becomes sufficiently negative to inject electrons into the TiO₂ matrix. These electrons will be transferred to oxygen molecules present at the interface. It follows that an architecture that facilitates the transport of electrons in the semiconductor film, for example a one-dimensional structure without electron traps, will allow for the flow of electrons from the FTO more easily, i.e. at a more positive potential, with respect to a random assembly of nanoparticles. In the latter case, the electrons will be also injected from the back-contact, but their transport through the whole film is more difficult due to the presence of deep traps, associated to GB, which should be initially filled. This concept is in good agreement with the experimentally measured oxygen reduction current onset, which is located at the potential corresponding to the filling of the electron traps.

Photopotential measurements allow us to complement the discussion about the electron reactivity towards oxygen present in solution. The faster the photopotential relaxation is, the faster the electron concentration decay and, therefore, the larger the kinetic constant for the transfer of electrons to molecular oxygen.⁴⁰ As observed in Fig. 6, the photopotential relaxation is faster for the wire-like aggregates, which reveal a larger reactivity for oxygen reduction than that of the randomly distributed nanoparticles. In conclusion, NC thin films not only show larger photoactivity, but also, in addition, are better electrocatalysts towards oxygen reduction as compared with their NP counterparts. Both of these characteristics render NC thin films plausible photocatalyst candidates.

In this sense, it has already been reported that wire-like aggregates show better characteristics as photocatalysts towards Procion Red MX-5B oxidation than commercial anatase nanoparticles.²² This result is in good agreement with electrochemical experiments in the dark and under illumination, although the behaviour of photocatalytic thin films and suspensions is not necessarily the same.⁴¹

Conclusions

In this work, two different types of anatase nanostructures with similar active surface areas have been investigated as photoanodes. Wire-like nanocrystal aggregates and randomly distributed nanoparticles have been deposited as nanoporous thin films on conductive glass. In the dark, their voltammograms are characterized by a similar accumulation region. However, in the case of randomly distributed nanoparticles, an additional voltammetric feature associated with GB traps can be observed, which is virtually absent in the NC sample. The different sample morphology allows us to study the effect of GB on electron transport and recombination in nanoporous metal oxide semiconductor electrodes.

Under illumination, a significant shift in the photocurrent onset exists, which is particularly relevant in the case of water photooxidation. The photooxidation begins at more negative potential values for NC electrodes than for NP ones. In the presence of an efficient hole acceptor such as HCOOH, significantly larger photocurrents are developed for the NC films than for the NP films. These results point to a reduced electron recombination in the case of NC electrodes, which can be related to a lower concentration of GBs, together with a fast hole transfer in the NC architecture. In addition, the presence of these electron traps also influences the oxygen reduction reaction, with the NC sample as more active towards oxygen reduction. The apparent superior performance of the NC films towards the photooxidation of formic acid and oxygen reduction indicates that NC anatase is likely a good photocatalyst candidate.

More generally, this work shows on the one hand, the high potential of electrochemistry for uncovering the photoactivity of semiconductor oxides while, on the other, the distinctive advantages that truly (one-dimensional) nanostructured electrodes could have as functional materials in different devices.

Acknowledgements

This work was financially supported by the Spanish Ministry of Science and Innovation through projects HOPE CSD2007-00007 (Consolider-Ingenio 2010) and MAT2009-14004 and by the Generalitat Valenciana (GVPRE/2008/198). Synthesis work at Brookhaven National Laboratory was supported by the US Department of Energy Office of Basic Energy Sciences under Contract No. DE-AC02-98CH10886. Initial collaborative work on this project led to the receipt of a Royal Society of Chemistry journals grant award to SSW.

References

1. A. J. Bard, M. Stratmann and S. Licht in Encyclopedia of Electrochemistry, Semiconductor electrodes and photoelectrochemistry, WILEY-VCH Verlag GmbH, vol. 6, 2002.
2. A. Hagfeldt and M. Grätzel, Chem. Rev., 1995, 95, 49.
3. M. Grätzel, C. R. Chim., 2006, 9, 578.
4. U. I. Gaya and A. H. Abdullah, J. Photochem. Photobiol., C, 2008, 9, 1.
5. O. K. Varghese and C. A. Grimes, J. Nanosci. Nanotechnol., 2003, 3, 277.
6. A. B. F. Martinson, M. S. Goes, F. Fabregat-Santiago, J. Bisquert, M. J. Pellin and J. T. Hupp, J. Phys. Chem. A, 2009, 113, 4015.
7. K. Zhu, N. R. Neale, A. Miedaner and A. J. Frank, Nano Lett., 2007, 7, 69.
8. J. M. Macak, H. Tsuchiya and P. Schmuki, Angew. Chem., Int. Ed., 2005, 44, 2100.
9. G. K. Mor, O. K. Varghese, M. Paulose and C. A. Grimes, Adv. Funct. Mater., 2005, 15, 1291.
10. S. J. Limmer and G. Z. Cao, Adv. Mater., 2003, 15, 427.
11. Y. Lin, Mater. Lett., 2008, 62, 1246.
12. S. Yang, X. Quan, X. Li and C. Sun, Photochem. Photobiol. Sci., 2006, 5, 808.
13. M. Matsuoka, M. Kitano, S. Fukumoto, K. Iyatani, M. Takeuchi and M. Anpo, Catal. Today, 2008, 132, 159.
14. M. Kitano, R. Mitsui, D. R. Eddy, Z. M. A. El-Bahy, M. Matsuoka, M. Ueshima and M. Anpo, Catal. Lett., 2007, 119, 217.
15. S. Nakade, M. Matsuda, S. Kambe, Y. Saito, T. Kitamura, T. Sakata, Y. Wada, H. Mori and S. Yanagida, J. Phys. Chem. B, 2002, 106, 10004.
16. T. Berger, T. Lana-Villarreal, D. Monllor-Satoca and R. Gómez, J. Phys. Chem. C, 2007, 111, 9936.
17. X. Chen and S. S. Mao, Chem. Rev., 2007, 107, 2891.
18. A. Fujishima, X. Zhang and D. A. Tryk, Surf. Sci. Rep., 2008, 63, 515.
19. S. Kment, P. Kluson, V. Stranak, P. Virostko, J. Krysa, M. Cada, J. Pracharova, M. Kohout, M. Morozova, P. Adamek and Z. Hubicka, Electrochim. Acta, 2009, 54, 3352.
20. M. Morozova, P. Kluson, J. Krysa, M. Zlamal, O. Solcova, S. Kment and T. Steck, J. Sol-Gel Sci. Technol., 2009, 52, 398.
21. T. Berger, T. Lana-Villarreal, D. Monllor-Satoca and R. Gómez, J. Phys. Chem. C, 2008, 112, 15920.
22. Y. Mao and S. S. Wong, J. Am. Chem. Soc., 2006, 128, 8217.
23. T. Berger, T. Lana-Villarreal, D. Monllor-Satoca and R. Gómez, Electrochem. Commun., 2006, 8, 1713.
24. V. R. Rives Arnau, Optica Pura y Aplicada, 1982, 15, 25.
25. F. Fabregat-Santiago, I. Mora-Seró, G. Garcia-Belmonte and J. Bisquert, J. Phys. Chem. B, 2003, 107, 758.
26. I. Abayev, A. Zaban, V. G. Kytin, A. A. Danilin, G. Garcia-Belmonte and J. Bisquert, J. Solid State Electrochem., 2007, 11, 647.
27. U. Diebold, Surf. Sci. Rep., 2003, 48, 53.
28. L. A. Lyon and J. T. Hupp, J. Phys. Chem., 1995, 99, 15718.
29. N. Hykaway, W. M. Sears, H. Morisaki and S. R. Morrison, J. Phys. Chem., 1986, 90, 6663.
30. J. Bisquert, Phys. Chem. Chem. Phys., 2008, 10, 3175.
31. J. Bisquert, F. Fabregat-Santiago, I. Mora-Seró, G. García-Belmonte, E. M. Barea and E. Palomares, Inorg. Chim. Acta, 2008, 361, 684.
32. T. Lana-Villarreal and R. Gómez, Chem. Phys. Lett., 2005, 414, 489.
33. A. Imanishi, T. Okamura, N. Ohashi, R. Nakamura and Y. Nakato, J. Am. Chem. Soc., 2007, 129, 11569.
34. R. Nakamura, T. Okamura, N. Ohashi, A. Imanishi and Y. Nakato, J. Am. Chem. Soc., 2005, 127, 12975.
35. T. Lana-Villarreal, R. Gómez, M. Neumann-Spallart, N. Alonso-Vante and P. Salvador, J. Phys. Chem. B, 2004, 108, 15172.
36. T. Lana-Villarreal, R. Gómez, M. González and P. Salvador, J. Phys. Chem. B, 2004, 108, 20278.
37. I. Mora-Seró, T. Lana-Villarreal, J. Bisquert, A. Pitarch, R. Gómez and P. Salvador, J. Phys. Chem. B, 2005, 109, 3371.
38. J. Bisquert, Phys. Chem. Chem. Phys., 2008, 10, 49.
39. T. Kisumu, A. Tsujiko, K. Murakoshi and Y. Nakato, J. Electroanal. Chem., 2003, 545, 99.
40. D. Monllor-Satoca and R. Gómez, J. Phys. Chem. C, 2008, 112, 139.
41. T. Lana-Villarreal, D. Monllor-Satoca, A. Rodes and R. Gómez, Catal. Today, 2007, 129, 86.

---

Footnote

† Electronic supplementary information (ESI) available: Typical spectral irradiance of a 150 W Xe arc lamp, TEM images, Raman spectra, XRD patterns, cyclic voltammograms, modified Kubelka-Munk function and Incident Photon to Current Efficiency versus wavelength. See DOI: 10.1039/c0nr00140f

---