Surface Density-of-States Engineering of Anatase TiO2 by Small Polyols for Enhanced Visible-Light Photocurrent Generation.

Enhancement of visible-light photocurrent generation by sol-gel anatase TiO2 films was achieved by binding small polyol molecules to the TiO2 surface. Binding ethylene glycol onto the surface, enhancement factors up to 2.8 were found in visible-light photocurrent generation experiments. Density functional theory calculations identified midgap energy states that emerge as a result of the binding of a range of polyols to the TiO2 surface. The presence and energy of the midgap state is predicted to depend sensitively on the structure of the polyol, correlating well with the photocurrent generation results. Together, these results suggest a new, facile, and cost-effective route to precise surface band gap engineering of TiO2 toward visible-light-induced photocatalysis and energy storage.

Introduction

An increasing number of important processes are now being transitioned to using sunlight as their energy source,[1,2] including pollutant degradation,[3−7] photocurrent generation,[8−11] and high energy content chemical synthesis.[12−16] In designing devices for these applications, a fine balance must be struck between the energy-conversion efficiency of the device, the lifetime of the end product, and other factors, such as material cost/abundance, toxicity, and physicochemical stability.[17] Because of its good performance in the latter categories, titanium dioxide (TiO2) is commonly used as a photoactive material in devices.[18−20] Furthermore, the energy levels of the valence and conduction bands (CBs) of TiO2 envelop the oxidation and reduction energies of H2O, respectively, making it especially suitable for water-splitting applications.[12,14,15]

One of the biggest drawbacks in the TiO2 applications is its wide band gap (∼3.2 eV for the most active crystal form, anatase TiO2). This limits the spectral range of absorbable solar photons to the UV region (λ < 400 nm); only about 5% of the light reaches the earth’s surface.[17,21] Much effort has been expended to increase the visible-light activity of TiO2, including elementary doping,[22−25] dye sensitization,[26−28] plasmonic enhancement,[15,29] and surface modification by organic molecules.[30−32] Although visible-light activity was achieved in these approaches, sacrifices were also made in terms of material performance. For example, although elementary doping engineers the band gap in a way that enables visible photoactivity, it usually also reduces the physicochemical stability of TiO2. In the case of surface modification with organic molecules and dye sensitization, most of the progress was made using aromatic compounds as charge-transfer agents.[32,33] These compounds are generally not very soluble in water, limiting their application potential, and their degradation during photocatalytic processes is a constant issue. A more serious drawback when using aromatic charge-transfer agents is that they are often toxic and can negatively impact the environment.

In this work, we present a solution to the latter problem, demonstrating that visible-light activity of TiO2 can be enhanced by adding simple, relatively nontoxic, water soluble organic surface-binding agents, small polyols. Despite the lack of charge-transfer moieties and/or absorption bands in the visible region in these polyols, an increase in TiO2 visible-light activity by up to 2.8 times is observed in their presence in photocurrent generation experiments. The mechanism of enhancement is revealed by density functional theory (DFT) calculations, which show that polyol binding to TiO2 results in generation of mid-band-gap energy states that can interact with visible photons. The energy of these mid-band-gap states is found to depend on the polyol structure, allowing surface DOS engineering and optimization of TiO2 visible photoactivity in a facile and cost-effective manner.

Results and Discussion

The effect of the molecular adsorption on the photoactivity of sol–gel derived anatase TiO2 films on fluorine tin oxide (FTO) in visible light was found in photocurrent generation experiments, an analysis tool successfully employed in similar studies on semiconductor photoactivity.[34,35] The photocurrent was measured using an electrochemical analyzer and a three-electrode cell consisting of the TiO2 film as the working electrode (W.E.), a Pt wire counter electrode (C.E.), and an AgCl/Ag reference electrode (R.E.), all immersed in a 0.1 M KOH aqueous electrolyte bath (Figure ). The cell was designed such that a 150 W solar simulator lamp could irradiate the TiO2 film with a 430 nm long-pass filter available to isolate the contribution of visible light (λ > 430 nm, Figure S1).

Figure 1

Schematic representation of the used setup in photocurrent generation experiments.

The effect of the molecular adsorption on the photoactivity of sol–gel derived anatase TiO2 films on fluorine tin oxide (FTO) in visible light was found in photocurrent generation experiments. To investigate the effect of polyol molecules on photocatalytic activity in visible light, the activity was compared before and after addition of polyol molecules to the electrolyte. During a measurement, the lamp light was alternatively blocked or allowed to impinge on the TiO2 film, for alternating 30 s periods, creating a block-wave pattern in the measured photocurrent (Figure a). The initial photoactivity was determined using three on–off cycles of 60 s and averaging the step heights between the light and dark periods. Subsequently, the polyol was added as indicated with the arrow in Figure a so that the change in the photocatalytic activity of the same sol–gel film can be compared. The final concentration of the polyol in the electrolyte is set to be ∼1.8 M. The newly created system was allowed to mix for 120 s, and then the activity was resolved using another three cycles. Note that although about 200 times smaller compared with activity in full solar spectral (full) light, including UV light (Figure S1), our TiO2 films are slightly visible-active even before addition of molecules. This could be due to oxygen vacancies on the surface that have emerged during preparation or calcination.[23,36] This activity, however, is crucial for the quantitative analysis of the effect of the added polyols because the enhancement factor (EF) in this study is given by the ratio between the photocurrent before and after polyol addition to the same sol–gel TiO2 film.

Figure 2

(a) Typical photocurrent generation experiments before and after adding polyols in the electrolyte. The result with glycerol is shown as an example. (b) Photocurrents found before (red) and after (blue) addition of the specific molecule in the electrolyte. (c) Enhancement factors of each molecule derived from the values in (b).

To understand this effect, we have studied several polyol molecules, as shown in Scheme . Among them, we first investigated the effect of the number of −OH groups and thus the hole-scavenging effect of ethanol (EtOH), ethylene glycol (EG), and glycerol (GC). Higher enhancement can be predicted when having more OH groups due to increase of the hole-scavenging power, expected to be in the order of EtOH < EG < GC. The photocurrent activities for these molecules and H2O as control are displayed in Figure b. As can be seen in the figure, every sol–gel film shows different photocurrent activity even before polyol addition due to the heterogeneous feature of the sol–gel film. As mentioned before, for the quantitative analysis, the ratios between activities before (red bars) and after (blue bars) polyol addition are given as the enhancement factor (EF) for each molecule in visible light (Figure c). It is clear that visible activity is enhanced upon addition of all used molecules. Addition of EG shows the highest EFs (>2.8), whereas EtOH and GC show similar EFs, less than 2. This result suggests that the number of −OH groups is not responsible only for the enhancement.

Scheme 1

Chemical Structures of the Investigated Molecules

To investigate the influence of the molecules on the visible photocatalytic activity more precisely, the experiment was conducted using TiO2 films that were pretreated with polyol molecules; namely, the sol–gel film was immersed in each polyol for 68 h and subsequently rinsed with water and isopropanol several times and dried with Ar gas. These pretreated sol–gel films were subjected to the photocurrent experiment using pure 1 M KOH electrolyte without adding polyols to the electrolyte. In such a way, we can discuss only the effect of adsorbed polyol molecules on the TiO2 surface and can avoid the unexpected effect of the polyols in the electrolyte, such as the continuous hole-scavenging effect by the molecular adsorption and desorption dynamics.

In this pretreatment study, the series of polyol molecules was expanded with propylene glycol (PG) and 1,3-propanediol (PD) (Scheme ). PD has one extra carbon compared with EG, which could affect adsorption of polyols on the TiO2 surface. PG is very similar to EG and thus could similarly adsorb on the surface as EG but it is a stronger electron donor than EG. Figure shows that the photocurrent EF obtained on TiO2 sol–gel films is pretreated with H2O, EtOH, EG, PG, PD, and GC. Note that the sol–gel film treated with H2O was used as the control. In this pretreatment experiment, after the photocurrent experiments with visible light, the sample was irradiated with UV light until all of the adsorbed molecules decomposed. Then EF was calculated using the value of the photocurrent on the freshly treated film and that from the UV-irradiated film. In such a way, quantitative estimation of EF was realized on the same sol–gel film. As shown in Figure , the films treated with EtOH and polyols exhibit higher EF compared to the control, indicating these molecules strongly adsorb and remain on the surface of TiO2 even after the several rinsing processes and affect on the photocurrent activity. The EFs for these molecules are found in the order of EG > PG > GC > PD > EtOH > H2O.

Figure 3

Enhancement factors of photocurrent generation obtained with TiO2 sol–gel films pretreated without and with EtOH, EG, PG, PD, and GC.

Further, the very similar EF values found in the pretreatment experiment (Figure ) and those in the addition experiment (Figure c) suggest that the polyol–TiO2 interaction, once established, is of similar nature in both cases.

To investigate this interaction in more depth, we have tried to obtain additional evidence for molecular functionalization and the type of adsorption of the molecules to the surface using Fourier transform infrared (FTIR) spectroscopy and Raman spectroscopy. The restricted sensitivity of these spectroscopic tools combined with the limited layer thickness of the functionalized layer, however, made it too hard to obtain the extra evidence in this way.

Polyols adsorbed on semiconductor surfaces have been studied in the past. In 2004, a number of articles were published regarding hole-scavenging by polyol molecules on TiO2 surfaces.[37,38] It is known that hole-scavenging decreases the recombination rate of the generated excitons and thus improves the photocatalytic activity. More recently, increased photocatalytic activity in UV light was shown for Au/TiO2 systems in the presence of several polyols, where increased photocatalytic activities were recorded with hole-scavenging being the likely explanation.[39] Interestingly, in both experiments on the hole-scavenging efficiency by polyols (found to increase with the number of hydroxyl groups per molecule)[38] and catalytic performance studies of TiO2 in the presence of polyols,[41] the order of enhancement factors was found to be GC > EG > EtOH > H2O under UV irradiation. Further, on the basis of their similar hole-scavenging efficiencies, PD and EG should show comparable EFs.[38] Our results obtained only with visible light (>430 nm), however, are not consistent with these trends.

To understand the trend found in this study, we calculated the surface local density of state (DOS) of anatase TiO2 with and without molecule adsorption. Recently, we have predicted that EG adsorption onto the TiO2 surface can give rise to midgap energy states in its surface density of state (DOS).[40] Such midgap energy states can facilitate electron excitations into the conduction band (CB) of TiO2 by light with lower energy and thus improve the photoactivity in visible light. Here, we present an expansion of the DFT calculations with the investigated molecules in this report. The DFT calculation has been conducted within the generalized gradient approximation (GGA) using the Perdew–Burke–Ernzerhof (PBE) functional[39] in the SIESTA computational code.[40] A double-ζ basis set of localized atomic orbitals and Troullier–Martins pseudopotentials were used for the valence and core electrons, respectively.[41] Two layers of the TiO2 slab is used to model the adsorption of molecules, as depicted in Scheme on the {001} and {101} facets. All of the structures were relaxed until less than 0.05 eV/Å. A Monkhorst–Pack grid was used for the density-of-state (DOS) calculations.

Figure shows the surface DOS of TiO2 with adsorption of H2O, EtOH, EG, PG, PD, and GC, where only the results for a {001} facet are shown because the more thermally stable yet less photoactive {101} facet yields no midgap states likely due to the inability of the molecules to attach dissociatively to the surface.[19,44] As shown in Figure , an extra energy state is found for EG, GC, and PG but not for EtOH, H2O, and PD (Figure ). Moreover, the position of the localized midgap state induced by EG is ∼2 eV below the CB, whereas the positions for PG and GC are ∼2.5 eV below it. The presence of these midgap states could associate with efficient electron excitation under visible light besides the hole-scavenging effect. Indeed, the maximum EF was found with EG, whereas PG and GC show the second highest EF among the molecules. The enhancement observed for EtOH and PD is smaller than that of EG, PG, and GC and yet higher than that of H2O, which is most likely due to the hole-scavenging effect. Therefore, taking into account both the extra midgap state and hole-scavenging effect, EFs found in the order of EG > PG > GC > PD > EtOH > H2O nicely correlate with our calculation.

Figure 4

DFT-calculated density-of-states (DOSs) plot for the investigated molecules (H2O, EtOH, EG, PG, PD, and GC) dissociatively adsorbed on a {001} TiO2 facet. The right end of the figure shows the position of the newly created midgap states in the electronic structure of the TiO2 surface.

The absence of this correlation in other studies on polyol-enhanced TiO2 photocatalysis can be attributed to the light source employed. Other studies used UV light or sunlight including UV light, whereas the present study used only visible light (430 < λ < 750 nm). Because of the high photoresponse of TiO2 in UV light, the response of the system to those wavelengths could conceal the effect of the midgap states. Because we use only visible light (λ > 430 nm) in our irradiation experiments (Figure S1), the effect of the midgap states is more pronounced in the photocurrent generation experiments. The differences in enhancement when using either wavelength range is demonstrated in Figure S2 for the sol–gel TiO2 surfaces treated with EG.

Supported by infrared (IR) spectroscopy[45] and X-ray absorption near edge structure (XANES) spectroscopy measurements,[38] a possible interaction of polyol molecules with defect sites on the TiO2 surface was suggested previously, involving an octahedrally coordinated Ti-atom chelated by two hydroxyl groups of the polyol. In other studies, the interaction between hydroxyl groups and TiO2 was used to anchor larger molecules to the semiconductor surface.[32,33] From our results, we conclude that the interaction of polyols with the surface is of great importance to surface DOS engineering. According to our calculation, all molecules showing midgap energy states also feature a noncoordinating hydroxyl group (Figure S4B–D). When the number of backbone carbon atoms is increased above two, two different Ti atoms can be bound by the hydroxyl groups on either end of the molecule, converting the free hydroxyl group into a binding oxygen atom (Figure S4E). The presence of a free hydroxyl group for EG, PG, and GC but not for PD and the presence and absence, respectively, of a midgap energy state, as predicted by DFT, further supports this scenario.

Conclusions

In conclusion, the photocatalytic performance of TiO2 films in visible light was improved by applying several small polyols to the surface. For a select number of polyols with similarities in the carbon backbone, the performance increase was found to be higher due to the appearance of an extra mid-band-gap energy state. With these results, the efficiency of photocatalysis, energy storage, or direct energy production using solar light can be further increased, supporting a worldwide transition from conventional to sustainable energy sources.

Experimental Section

Anatase TiO2 films on FTO were prepared using a sol–gel method, with Ti(OPr)4 as the precursor salt and the annealing temperature of 475 °C.[46] Full saturation of the TiO2 surface with polyol molecules was achieved by either immersing the samples in the pure compound for 68 h or by adding the pure compound to the electrolyte medium during photocurrent generation experiments, as will be discussed further below.

In a typical experiment, the photocurrent was measured using an electrochemical analyzer and a three-electrode cell consisting of the TiO2 film as the working electrode (W.E.), a Pt wire counter electrode (C.E.), and an AgCl/Ag reference electrode (R.E.), all immersed in a 0.1 M KOH aqueous electrolyte bath (Figure ). The cell was designed such that a 150 W solar simulator lamp could irradiate the TiO2 film with a 430 nm long-pass filter available to isolate the contribution of visible light (OD > 5 below 430 nm, Figure S5). The Ohmic contact between the TiO2 film and the wiring was improved using Cu tape and an In/Ga paste.

All of the polyols employed here as well as the Ti(OPr)4 were purchased from Sigma-Aldrich, with a purity of at least 98%, and used without further purification. Both the electrochemical analyzer and the AgCl/Ag reference electrode were purchased at CH instruments. The solar simulator, model number 67005, was provided by Newport, Oriental Intruments. The 430 nm long-pass filter, HQ430LP, was purchased at Chroma.

We used density functional theory (DFT) calculations within the generalized gradient approximation (GGA) within the Perdew–Burke–Ernzerhof (PBE) form,[41] as implemented in the SIESTA code.[42] The core electrons were replaced by Troullier–Martins pseudopotentials.[43] A double-ζ basis set of localized atomic orbitals was used for the valence electrons. A mesh cutoff energy of 300 Ry has been imposed for real-space integration. All structures have been relaxed until forces were less than 0.05 eV/Å. In the calculations, a vacuum interval of more than 15 Å was used to avoid the interaction between the periodic slabs. The surface area of anatase TiO2 {001} used in the calculations is 7.57 Å × 7.57 Å. When relaxing the atomic structures, the sampling of the Brillouin zone was restricted to the Γ point and a (10 × 10 × 1) Monkhorst–Pack grid was used later on for DOS calculations.[47,48]