Probing the Nature of Bandgap States in Hydrogen-Treated TiO2

Article pubs.acs.org/JPCC

Probing the Nature of Bandgap States in Hydrogen-Treated TiO2 Nanowires Damon A. Wheeler,† Yichuan Ling,† Robert J. Dillon,‡ Robert C. Fitzmorris,† Christopher G. Dudzik,† Liat Zavodivker,† Tijana Rajh,§ Nada M. Dimitrijevic,§ Glenn Millhauser,† Christopher Bardeen,‡ Yat Li,*,† and Jin Z. Zhang*,† †

Department of Chemistry and Biochemistry, University of California, Santa Cruz, 1156 High Street, Santa Cruz, California 95064, United States ‡ Department of Chemistry, University of California, Riverside, 501 Big Springs Road, Riverside, California 92521 United States § NanoBio Interface Group, Center for Nanoscale Materials, Argonne National Laboratory, Argonne, Illinois 60439, United States ABSTRACT: Hydrogen treatment of TiO2 has been demonstrated to significantly alter its optical properties, including substantially enhanced visible light absorption that has important implications for various applications. The chemical nature of the bandgap states responsible for the increased visible absorption is not yet well understood. This work reports a detailed study of the structural, optical, electronic, and ultrafast properties of hydrogentreated TiO2 (H:TiO2) nanowires (NWs) using a combination of experimental techniques including high-resolution transmission electron microscopy (HRTEM), electron spin resonance spectroscopy (ESR), time-resolved fluorescence (TRF), and femtosecond transient absorption (TA) spectroscopy in order to explain the origin of the strong visible absorption. The combined TEM, ESR, TRF, and TA data suggest that the presence of a localized mid-bandgap oxygen vacancy (VO) occupied by a lone electron in an antibonding orbital situated at a surface site is likely responsible for the visible absorption of the material. The data further indicate that while untreated TiO2 NWs are fluorescent, the hydrogen treatment leads to quenching of the fluorescence and highly efficient charge carrier recombination from the VO state following excitation with visible light. With UV excitation, however, the charge carrier recombination of the H:TiO2 NWs exhibits a larger component of a slow decay compared to that of untreated TiO2, which is correlated with enhanced photoelectrochemical performance. Both the treated and untreated samples exhibit a fast decay that dominates the TA signals, which is likely caused by a high density of surface trap states. A simple model is proposed to explain all the key optical and dynamic features observed. The results have provided deeper insight into the chemical nature and photophysical properties of bandgap states in chemically modified TiO2 nanomaterials. characteristics of TiO2 since ∼45% of the solar spectrum is composed of visible-wavelength photons. To that end, an extensive amount of research has been performed on improving and gauging the efficiency of TiO2 in various architectures. Varying platforms have been explored, including TiO2-quantum dot (QD) heterostructures8−10 and elemental doping by elements such as nitrogen,11,12 chromium,13 and other transition metals.14−16 One recently developed strategy for improving the photoelectrochemical properties of TiO2 involves the treatment of the TiO2 with molecular hydrogen at elevated temperatures which renders the TiO2 a black color.17−20 It has been proposed that the dark coloration of H:TiO2 is due to the formation of either H or H2 impurities in the bandgap from the hydrogen treatment.21,22 It is also suggested that the black color is caused by the existence

1. INTRODUCTION In 1972, Fujishima and Honda discovered that titanium dioxide (TiO2) was a promising material as a photoanode for photoelectrochemical (PEC) water splitting.1 In the intervening four decades, research on the material has exploded. Much effort has been devoted to providing a fundamental understanding of its photocatalytic properties as well as enhancing its overall efficiency and performance as a PEC material. In particular, research on TiO2 has gained significant traction due to the advantageous characteristics of the material, including its strong optical absorption, favorable band alignment, resistance to chemical degradation and photobleaching, abundance, low toxicity, and low cost.2−6 One of the limitations to TiO2 has been its wide bandgap that prevents it from absorbing the majority of the solar spectrum. This is shown by its low photoconversion efficiency (2.2%) under 1 W, 1.5 AM global solar irradiation.7 Because of this, an emphasis is being placed on improving the visible absorption © 2013 American Chemical Society

Received: October 3, 2013 Revised: November 27, 2013 Published: December 2, 2013 26821

dx.doi.org/10.1021/jp409857j | J. Phys. Chem. C 2013, 117, 26821−26830

The Journal of Physical Chemistry C

Article

vapor deposition (CVD) system in which the CVD tube was filled with ultrahigh purity hydrogen gas (Praxair). 2.3. UV−vis Measurements, Electron Microscopy, and Electron Spin Resonance. Absorption measurements were performed on an HP 8452A diode array spectrophotometer with spectral resolution set to 2 nm. High-resolution transmission electron microscopy (HRTEM) was carried out on a Philips CM300-FEG at the National Center for Electron Microscopy at Lawrence Berkeley National Laboratory with the accelerating voltage set to 300 kV. X-band continuous wave electron spin resonance (ESR) spectroscopy experiments were conducted at the Center for Nanoscale Materials at Argonne National Laboratory on a Bruker Elexsys E580 spectrometer equipped with an Oxford CF935 helium flow cryostat with an ITC-5025 temperature controller. Samples were purged with nitrogen prior to measurements to remove air/oxygen. Excitation of the sample in the ESR spectrometer was carried out with a broadband xenon lamp equipped with water as IR cutoff filter. The g-factors were calibrated for homogeneity and accuracy by comparison to a coal standard, g = 2.00285 ± 0.00005. 2.4. Time-Resolved Fluorescence. Time-resolved fluorescence experiments were performed on a regeneratively amplified Spectra-Physics Spitfire laser system with a repetition rate of 40 kHz. The 400 nm pump beam was generated by frequency doubling the 800 nm fundamental with a barium borate (BBO) crystal. Residual fundamental was removed with a hot mirror and a Schott glass BG39 filter. Pump power dependence study was conducted by setting the per-pulse fluence to 0.4 and 1.3 μJ/cm2. The samples were kept under vacuum in a Janus ST100 cryostat and excited in the front-face geometry. Pump scatter was filtered from the signal using a 400 nm long-wave-pass filter and two Schott glass OG420 filters. The signal was detected with a Hamamatsu C4334 Streakscope. 2.5. Femtosecond Transient Absorption System. The transient absorption (TA) system utilized in this study has been described previously.26 2.5 mW and 110 fs seed pulses with a repetition rate of 33 MHz were obtained from a frequencydoubled Er-doped fiber oscillator and amplified in a Ti-sapphire regenerative amplifier using chirped-pulse (multipass) amplification. The seed pulse from the oscillator was stretched temporally to ∼200 ps using a grating stretcher, regeneratively amplified at a repetition rate of 752 Hz by an intracavity-doubled, Q-switched Nd:YLF laser, and recompressed by a grating compressor. The final output pulses that were obtained were typically 150 fs with a pulse energy of 1 mJ and centered at 795 ± 10 nm. The amplified output was beam split such that 90% of the light was directed into an optical parametric amplifier (OPA) while the remaining 10% was passed through a sapphire crystal to generate a white light continuum (WLC) probe pulse with a spectral range from 450 to 750 nm. The pump pulse repetition rate was halved by an optical chopper which was monitored by a CCD and used to produce a differential absorption spectrum. The delay between the pump and probe pulses was managed by a motor-controlled optical delay stage with a temporal resolution of 10 fs. The pump and probe pulses were focused on a 5 cm focal length curved mirror to overlap spatially and temporally on the surface of the sample. For UVpumped experiments, a 380 nm pump pulse was selected from two phase-matched bismuth borate (BiBO) crystals in the OPA, while for visible-pumped experiments a 450 nm excitation was chosen. The TA spectra were analyzed via a singular value decomposition (SVD) global fitting procedure written at UCSC for use in Matlab, in which the data at all wavelengths and all

of a lone electron occupying a Ti 3d state within the bandgap of the TiO2.18 Because the hydrogen gas used to treat the sample has a strong reductive ability and generates a high density of electron donors, this Ti 3d state has been termed an “oxygen vacancy” (VO).18,23 Electronically, this VO serves as a potential gateway for optical transitions, which contribute to the strong optical absorption, thereby contributing to the black color of the H:TiO2. Because of the hydrogen treatment, the solar-to-hydrogen efficiency of TiO2 was significantly enhanced.18 The improvement was attributed to elevated charge transport due to increased donor density from the VO. It resulted in a 5-fold enhancement of the incident-photon-to-current-conversion efficiency (IPCE) in the UV region. Despite the promising results that have been obtained by using H:TiO2 as a photoanode, there is still some dissension with regard to the chemical nature of the black color, specifically if the photons absorbed can be utilized for PEC water splitting. Because H:TiO2 samples display consistently improved PEC performance relative to pristine TiO2, an improved understanding of the fundamental reason behind this improvement is of great interest.17,18,23−25 In this work, we carried out structural, optical, electronic, and ultrafast transient absorption characterizations of the midbandgap states by systematically investigating hydrogen-treated and pristine TiO2 NWs using transmission electron microscopy (TEM), electron spin resonance (ESR), time-resolved fluorescence (TRF), and transient absorption pump−probe spectroscopy (TA). The TEM indicated that the smooth, dense, and untreated NWs gave yield to the presence of a roughened surface structure upon hydrogen treatment. ESR analyses indicated the presence of a lone electron localized at a Ti 3d state, characterized by an axial g-tensor at g⊥ = 1.975 and g∥ = 1.943, typical of radicals in which an unpaired electron acquires orbital angular momentum. Meanwhile, TRF and TA studies were used to probe the dynamics of the charge carrier in the VO and related states which provide insight into the constitution of the midbandgap state and how it gives rise to improved PEC performance. From all of these results in conjunction, we obtained a more complete picture about the electronic states that give rise to the color and the overall improved PEC response of the H:TiO2 relative to the untreated TiO2.

2. EXPERIMENTAL SECTION 2.1. TiO2 Nanowire Synthesis. The TiO2 NW arrays were synthesized based on a previously reported hydrothermal method.5 Briefly, 15 mL of deionized water was mixed with 15 mL of concentrated (37%) HCl in a 100 mL beaker with continuous stirring. Subsequently, 0.5 mL of titanium n-butoxide was added to the dilute HCl solution. The prepared solution and a piece of fluorine-doped tin oxide (FTO) glass substrate were transferred to a 40 mL Teflon-lined stainless steel autoclave with the FTO being fully submerged in the solution. The sealed autoclave was heated for 5 h in an oven set at 150 °C and allowed to cool naturally to room temperature. A uniform white film composed of TiO2 was found coated on the FTO glass. The sample was then washed sequentially with ethanol and water. Finally, the sample was annealed in air at 550 °C for 3 h in order to increase the crystallinity of the NWs and increase their contact with the FTO substrate. 2.2. Hydrogen Treatment of TiO2 Nanowires. The asprepared TiO2 NWs were annealed in a hydrogen atmosphere at a temperature range from 350 to 550 °C for 30 min. The hydrogen treatment was performed in a home-built chemical 26822



Article

times are fit simultaneously to a multiexponential process, ensuring that the minimal number of exponentially decaying intermediates needed to fit the spectra can be determined.30,31

3. RESULTS 3.1. UV−Vis Electronic Absorption Spectra and Electron Microscopy Images. Figure 1 shows representative

Figure 1. UV−vis spectra of TiO2 and H:TiO2 NWs annealed at various temperatures.

UV−vis spectra of the TiO2 and H:TiO2 samples. The overall extinction coefficient increases from the red to the UV region. The “apparent” peak at 330 nm is not real but a result of inaccurate measurement at wavelength shorter than 300 nm. The signal at wavelengths redder than the expected bandgap of TiO2 is mostly from scattering due to the fact that the nanowires have length scales comparable to the wavelength of light. There may also be some minor contributions from absorption of bandgap states. Figure 2a presents HRTEM images which show the crystal lattices of the untreated TiO2 sample. A lattice spacing of 3.5 Å, indicative of the d-spacing of the (001) plane of rutile-phase TiO2, has been highlighted. Also shown in Figure 2a (inset) is the electron diffraction pattern of the samples. After hydrogen treatment, the H:TiO2 samples show an increased roughness when compared to the untreated TiO2 (Figure 2b). The image in Figure 2b was a sample annealed at 450 °C. As the annealing temperature of the H:TiO2 samples increased, the observed surface roughness likewise increased. The rough surfaces are notably absent in the untreated TiO2 samples. Finally, Figure 2c shows a 3D representation of what we believe the morphological changes manifest as on the unit cell scale. Figure 2c (left) is a schematic of untreated TiO2, while Figure 2c (right) is a schematic of H:TiO2 NWs, in which the VO are present. Because VOs are generated when an oxygen atom at a bridging oxygen site renders two subsurface Ti3+ ions exposed, the schematic representation of the right side of Figure 2c takes on a twisted shape. 3.2. Electron Spin Resonance (ESR) Spectra. ESR was performed in order to validate the notion that the VO was composed of a lone electron in the localized21,27 Ti 3d state of the H:TiO2. In Figure 3, we present the ESR spectra of hydrogentreated TiO2 in different crystalline modifications: rutile, P25, and nanowires (a), rutile nanoparticles (b), hydrogen-treated P25 (c), and hydrogen-treated rutile TiO2 NWs (d). All hydrogen-treated samples display two sets of signals: signals for electrons localized on Ti centers (g < 2.00) and those for oxygen centered radicals (g ≥ 2.00).28,29 The electrons on the hydrogen-treated rutile sample were seen to localize mainly on Ti rutile lattice sites (g = 1.972, g∥ = 1.948) and a small fraction on

Figure 2. Representative HRTEM images of (a) TiO2 NWs, showing a dense and smooth architecture, (b) TiO2 NWs hydrogen treated at 450 °C, showing a rough surface, and (c) a proposed image of (left) pristine TiO2 and (right) hydrogen-treated TiO2 in which bridging oxygen atoms have been removed, thereby exposing subsurface Ti3+ ions. White spheres correspond to Ti and red spheres to O.

rutile surface Ti sites (g = 1.953, g∥ = 1.975) and was the most intense of the spectra obtained. The hydrogen-treated P25 displayed a small rutile-like signal which was somewhat shifted from the lattice signal (g⊥ = 1.978, g∥ = 1.943 compared to rutile g⊥ = 1.975, g∥ = 1.940), indicative that the signal is localized at the interface between anatase and rutile crystalline structures, and a large anatase surface signal (g = 1.926), which indicates some surface-trapped electrons on anatase sites, consistent with the hydrogen reduction process. Additionally, a small oxygencentered radical with the same characteristics as holes on anatase TiO2 is observed. Finally, hydrogen-treated TiO2 NWs displayed a rutile lattice signal (g⊥ = 1.975, g∥ = 1.943). Moreover, an oxygencentered radical is observed with the g-tensor indicative of an “extra” electron in the antibonding orbital (g⊥ = 1.999, g∥ = 2.044). 3.3. Time-Resolved Fluorescence and Ultrafast Transient Absorption. Time-resolved fluorescence (TRF) was 26823



Article

Figure 3. (a) ESR plot of hydrogen-treated TiO2 in the forms of rutile (black curve), P25 (blue curve), and nanowires (red curve). (b) ESR plot of hydrogen-treated rutile (black) and hydrogen-treated rutile under differing illumination times: intermediate (red) and long (orange). (c) ESR comparison of untreated P25 (blue curve), dark H:P25 (black curve), and illuminated P25 (orange). (d) ESR comparison of illuminated TiO2 nanowires (orange), H:TiO2 nanowires (light blue curve), and 15 MA, 2 MW H:TiO2 (black curve).

collected on the H:TiO2 NW samples following a 400 nm excitation from a frequency-doubled 800 nm fundamental. Figure 4 represents the time-resolved fluorescence (TRF) of H:TiO2 NWs annealed at various temperatures: 350, 400, 450, and 500 °C. A global fitting algorithm was employed to determine a single set of time constants that would fit the kinetic traces across all probe wavelengths for a given sample with a minimal number of exponentials.30,31 For both treated and untreated samples, all decays were fit well with a triple exponential. The TiO2 sample had a decay that was fit to time constants of 35 ps, 120 ps, and >1 ns regardless of pump fluence. For the 0.4 μJ/cm2 fluence runs, the H:TiO2 samples displayed a faster decay for the fast component with increasing annealing temperature with lifetimes of 30 ps for the samples. The middle component fit well to a 200 ps lifetime, and the long component had a lifetime of greater than 1 ns for all samples. For the 1.3 μJ/cm2 fluence runs, the fast component lifetime fit well to a triple exponential of 25 ps for the samples. Again, the middle component fit to a 200 ps lifetime and a slow component of greater than 1 ns for all samples. Additionally, as seen in Figure 4b, the global fitting revealed that the relative intensity of the fast component of the decays increased from 68 to 77% as the annealing temperature increased from 350 to 550 °C. Despite that the integrated intensity

of the fast component remained at 30% for all treatment temperatures. Ultrafast TA data were collected of the TiO2 and H:TiO2 NWs at UV pump (380 nm, 475 nm probe) and visible pump (450 nm, 505 nm probe) wavelengths and are shown in Figure 5. For the untreated TiO2 sample, visible wavelength excitation produced no noticeable signal as the bandgap of pristine TiO2 is greater than the wavelength of excitation light used; thus data could not be collected on that sample at that pump wavelength. However, for the UV pump, the TiO2 exhibited a fast initial decay followed by a medium decay and a long-lived slow decay manifesting as a persistent y-offset. The data (summarized in Table 1) were fit to a triple exponential with time constants of 30 ps, 100 ps, and >1 ns. The UV-pumped H:TiO2 displayed a slower medium decay than the UV-pumped TiO2; yet, as with the untreated TiO2 NW sample, the H:TiO2 also retained the persistent offset. This is indicative of long-lived electron−hole pairs. All of the data of the UV-pumped H:TiO2 samples were fit well to a triple exponential with time constants of 60 ps, 200 ps, and >1 ns. For the 450 nmpumped H:TiO2, a fast initial decay was noticed, followed by a nearly complete relaxation to baseline, indicative of very efficient electron−hole recombination. Those data, essentially independent of annealing temperature, were fit to a double exponential 26824



Article

Figure 4. (a) Intensity vs time (ps) of H:TiO2 NWs along with fitting curves and (b) global fitting results of the PL lifetime of H:TiO2, annealed at 350, 400, 450, and 500 °C, showing the contribution from each lifetime component to the overall integrated spectrum at each wavelength. The H:TiO2 samples required triple-exponential functions to fit their decays.

with time constants of 30 and ∼125 ps. For all of the samples studied, no power dependence was observed. Figures 5b and 5c are three-dimensional false-colored plots of the H:TiO2 and TiO2 TA data with UV excitation which indicates the spectral location of the maximum transient bleach feature. Noticeably, the transient bleach is seen to extend for a longer time period for H:TiO2 (Figure 5b) than for pristine TiO2 (Figure 5c). Additionally, it is apparent from Figures 5b and 5c that there is an overt spectral evolution of TA signals over time, and the

maximum bleach feature for both is consistently situated at probe wavelengths of 475 nm. Figure 5d illustrates a three-dimensional false-colored plot of H:TiO2 following visible pumping. From Figure 5d, one can see that the bleach feature is centered near 460 nm and has components red-shifted to 600 nm. The bleach feature is also seen to recover quickly, regardless of wavelength, and no overt spectral evolution is noticeable. As with the time-resolved fluorescence data, SVD fitting indicated that the relative intensity of the fast component 26825



Article

Figure 5. (a) Transient bleach relaxation traces of TiO2 and H:TiO2 NWs following 450 nm excitation (red curve) and 380 nm excitation (blue and green curves). No noticeable differences in the bleach relaxation traces were evident based upon annealing temperature. (b, c) 3D plot of H:TiO2 and TiO2 NWs, respectively, with optical density plotted as a function of wavelength (nm) and time (ps) following UV excitation. (d) 3D plot of H:TiO2 NWs following visible pumping. (e−g) Example of SVD fit results for H:TiO2 under visible illumination, TiO2 under UV illumination, and H:TiO2 under UV illumination, respectively. Shown for (e), (f), and (g) are the B spectra showing the wavelength dependence of the initial amplitude of the various time constants.

the fitting revealed a component centered at 455 nm with a 30 ps lifetime. This feature is also noticeable in Figure 5a in which the fast component is mixed with the substrate response and does not appear to have an overt time-dependent red-shift. The fast

increased with increasing annealing temperatures. Finally, Figures 5e−g show the global fitting results representative of H:TiO2 under visible illumination, TiO2 under UV illumination, and H:TiO2 under UV illumination, respectively. In Figure 5e, 26826



Article

position of the signal is in accordance with previously reported ESR signatures from Ti3+.37,38 Illuminated rutile samples displayed a broad weak signal for photogenerated holes and a sharp signal for electrons in rutile environment. As the rutile was illuminated at liquid helium temperatures, only lattice electrons are formed as they do not have enough energy, at that temperature, to be transferred via a hopping mechanism to surface sites. The hydrogen-treated rutile displayed three interesting characteristics: (i) a rutile lattice signal that grew with illumination (g = 1.972), (ii) a fraction of tetrahedral sites that decrease upon illumination, suggestive that they are located near the surface (g = 1.977) and disappear in the reaction with surface generated holes, and (iii) surface rutile sites that disappear upon illumination also due to recombination with surface holes. When the H:P25 was investigated, interfacial rutile and surface anatase Ti3+ signals were observed. More importantly, surface trapped electrons on anatase sites were observed, consistent with chemical reduction of H2 that should affect surface sites primarily. Interestingly, oxygen radicals were observed and associated with holes on anatase TiO2. When this sample was illuminated, new electrons on lattice rutile centers were formed (g = 1.975) in addition to anatase photogenerated holes while surface-level anatase electrons disappeared in the reaction with photogenerated holes. The H:TiO2 NW signal was very small and was composed primarily of rutile signal. The sample contained no hole-like oxygen-centered radicals that were observed in illuminated samples but rather oxygen-centered radicals, indicative of an “extra” unpaired electron in an antibonding orbital. When this sample was illuminated, more rutile centers were formed and a signal typical of photogenerated holes was observed. This signal resembles the signal obtained by illumination of untreated NWs. Effectively, the H:TiO2 NWs are composed mainly of rutile with a small fraction of anatase (smaller than that of P25, at least) which is still photoactive. All ESR experiments indicate that hydrogen treatment of TiO2 nanoparticles, independent of their crystalline structure, results in VO associated with the presence of extra electrons primarily localized on Ti sites both within the nanoparticle lattice and on the surface. The charge migration between electron trapping sites under illumination shows behavior typical of the stabilized Ti3+ centers. 4.3. Charge Carrier Dynamics. Traditionally, the PL from rutile is considered to be weak and is likely mediated by a high density of defect states.39,40 Interestingly, the weak PL was reduced further upon hydrogen treatment: 35 ps for the TiO2 NWs and ∼30 ps for the H:TiO2. This suggests that hydrogen treatment fundamentally alters the quantity and energetic position or depth of trap states within the bandgap since a higher density of states within the bandgap is expected to quench PL and reduce the lifetime observed. This is in line with the HRTEM results, indicating an increase in surface defect density in the crystalline lattice of the samples upon hydrogen treatment. However, the fact that the medium component becomes slower from TiO2 (120 ps) to H:TiO2 (200 ps) indicates that the PL decay process is slowed on this time scale due to some longerlived trap state with hydrogen treatment, likely deeper trap states. The overall slower decay of the TA of the H:TiO2 sample with UV excitation as compared to the untreated sample indicates that hydrogen treatment lengthened the lifetime of the charge carriers, most likely in trap states or the VO state. Based on the PL peak position (∼460 nm, or 2.7 eV), there is one fluorescent trap state estimated to be ∼0.3 eV below the CB since the bandgap of rutile is known to be 3.0 eV. While there are many trap states,

Table 1. Summary of Fitting Time Constants to the TA Data sample

τ1 (ps)

τ2 (ps)

τ3 (ns)

TiO2 (UV illumination) H:TiO2 (UV illumination) H:TiO2 (visible illumination)

30 60 30

100 200 125

>1.0 >1.0

30 ps decay gives way to the 125 ps slow component which is centered at 462 nm. Importantly, the fast component is seen to account for ∼85% of the overall signal, indicating that the H:TiO2 NW dynamics are driven primarily by surface disorderoriented states. For Figure 5f, the fitting revealed a fast component centered at 470 nm, a middle component of 100 ps centered also at 470 nm, and a long-lived component in excess of 1 ns. Here, the fast component comprised ∼65% of the overall signal. Finally, in Figure 5g for UV-pumped H:TiO2, a fast component of 60 ps centered at 450 nm, a middle component of 200 ps centered at 455 nm, and a slow component of >1 ns were present. In this sample, the fast component accounts for ∼75% of the overall signal.

4. DISCUSSION 4.1. Structural Properties. Pristine rutile TiO2 is composed of O anions and Ti cations in the native 4+ state with a 6-fold coordination to oxygen atoms.23 It is known that heating the sample at elevated temperatures in excess of 350 °C results in desorption of surface-adsorbed oxygen atoms, which results in oxygen vacancies, or VO.18,32,33 In terms of structure, an absent oxygen atom at a bridging oxygen site renders two subsurface Ti3+ ions exposed. As has been established previously,17,18 the inclusion of Ti3+-based disorder into the lattice of pristine, rutilephase TiO2 during the high-temperature treatment with molecular hydrogen has the effect of reducing Ti4+ to Ti3+ through the occupation of vacant Ti 3d orbitals of TiO2 that are otherwise empty. The generation of Ti3+ states is therefore believed to follow a stoichiometric equation of x H+ + TiO2 + x e− → HxTiO2 − x

These isolated and localized states are known to exist ∼0.75 eV below the CB of TiO2 that is composed of 3d Ti3+.21,27,34 Structurally, it is the TiO2 (110) plane that has been thought to have defect stoichiometry leading to a photocatalytically active Ti3+ state.23,35,36 There also exists a number of defects in the lattice. One other interesting feature of the H:TiO2 NWs, relative to the TiO2 NWs, is that the high-temperature annealing yields substantial roughness in the nanowires. Prior to annealing, the nanowires were largely smooth and dense. The high-temperature annealing afforded an abrupt change to a much more polycrystalline form. 4.2. Electronic States Probed by ESR. All fully oxidized samples were ESR silent in the dark while upon illumination they displayed signals for photogenerated electrons and holes in different crystalline environments. In previous studies, it was found that photogenerated electrons in P25 are localized on both rutile and anatase sites, and the position of the signals in illuminated P25 can be used as a standard for decoding the crystalline environment in which photogenerated electrons and holes reside. Illuminated P25, which should contain no unpaired electrons indicative of the VO, displayed an ESR signal indicative of surface holes localized on anatase sites and electrons in both rutile and anatase environment. The signal in water is however was found to be shifted to higher g-values, which is consistent with the existence of tetrahedral interfacial sites. The spectral 26827



Article

component as with UV excitation. The lack of a long-lived component is consistent with lack of photocurrent in the visible due to the very fast charge carrier recombination. 4.4. Nature of the Bandgap State in H:TiO2. Figure 6 presents a model to illustrate the relevant energy levels including the bandgap states of TiO2 and H:TiO2 and associated charge carrier relaxation process following UV or visible excitation. For non-hydrogen-treated TiO2, above-bandgap excitation leads to an electron−hole pair. The initial relaxation that the electron will experience is into a state approximately 0.3 eV below the CB which is determined to be a fluorescent state. Subsequent to that relaxation occurs through the manifold of trap states endemic to metal oxides before a final recombination occurs with the hole. Upon hydrogen treatment of TiO2, however, an increasing number of defects are introduced into the system, thereby serving to quench the PLboth in intensity and in PL lifetime. Additionally, the treatment of the TiO2 with molecular hydrogen has the ability to donate electrons, thereby reducing some Ti4+ ions into Ti3+ ions and creating a VO, within the bandgap, approximately 0.75 eV below the bottom of the CB.45 Upon probing the state with UV-pumped transient absorption, an electron−hole pair is photogenerated, the electron of which relaxes down 0.75 eV into the VO. Compared to untreated TiO2, the relaxation is relatively slow since the TiO2 has a fluorescent state located ∼0.3 eV below the CB, making the initial electronic relaxation into that state relatively fast. Visible light excitation directly populates the VO state. Because of the high-temperature treatment and thereby the generation of a dense manifold of PLquenching states below the VO, the recombination of the visibly pumped electron and hole is fast and efficient. This is shown schematically in Figure 6. However, further work may be done to provide a more insightful look at the model. It has been shown before that when Ti3+ centers are introduced into a TiO2 system, the Fermi level is raised and electron accumulation occurs on the surface.21 This accumulation of electrons on the surface of the TiO2 will not only reduce the rate of surface recombination but also act to decrease the quantum yield, which is what is seen with the decrease in PL upon hydrogen treatment. Meanwhile, hydrogen treatment of TiO2 was found to be capable of prolonging hole lifetimes by reducing the number of bulk recombination centers.46 For visible pumping of the H:TiO2, the electronic transition involved in the excitation of the H:TiO2 should therefore result in the transfer of an electron from the valence band to the VO state, which is a singly occupied orbital of Ti3+. Surface Ti3+ states, which act as electron traps, aid in charge separation by slowing charge recombination, as discussed earlier, resulting in not only slower electron−hole recombination relative to pristine samples but also by resulting in a persistent, long-lived offset of the data, both of which are seen in Figure 5a. For UV pumping, in which the conduction band states are the electron acceptor, the rate of relaxation depends on the density of conduction band states. It has been argued before that because TiO2 has a large density of such states, the overall relaxation event for pristine TiO2 should be quick. For hydrogen-treated samples, however, since the initial relaxation dynamics of UV-pumped samples is slower, it can be argued that the density of these states has diminished or at least that the energetic distance separating them has increased. If one considers that TiO2 has, overall, a large number of electron traps and a relatively low number of free electrons, then for any recombination to occur, an electron must be excited to the conduction band before migrating to the surface of the nanostructure. From this, the lifetime will increase as the ratio of

we tentatively assign the 30 ps decay observed in the TA data to electron relaxation from the bottom of the CB to the fluorescent trap state. With hydrogen treatment, the fast decay is slowed down to 60 ps, and this could be due to contribution of the VO state created or change in the location or density of the fluorescent trap states. It is challenging to pinpoint this at this point. As can be seen in Figures 5b and 5c also, the TA intensity of H:TiO2 at red wavelengths decreases more slowly than that of pristine TiO2, indicative of longer-lived charge carriers, consistent with the single wavelength analysis. The VO state is estimated to lie ∼0.75 eV below the CB.18 Electrons in this deep trap state could be longer lived than in the 0.3 eV shallow fluorescent trap state. This would be consistent with the longer lifetime observed for the H-TiO2 sample. This would be consistent with the localized nature of the VO state.21,27 Of course, other states in the bandgap could contribute to the charge carrier dynamics. For example, interstitial hydrogen, introduced during the hydrogen treatment, is shown to locate in positions close to surface oxide ions.41−43 These hydrogen− oxygen species, taking the form of −OH functionalities, are known to have to primary energetic locations at 0.6 eV below the CB and 0.82 eV below the CB.44 As with the VO, these bands are also greatly localized42 which will aid in the slower recombination time subsequent to UV pumping. For visible-pumped H:TiO2, we suggest that the visible excitation takes the electron from the VB to the VO. We cannot completely rule out the possibility of the transition being from the VO to the CB. The overall fast decay of the charge carriers with visible excitation is likely due to a dense manifold of states under the VO, as illustrated in Figure 6. There is no long-lived

Figure 6. Proposed model for energy levels related to the optical properties and dynamics studies. CB and VB represent the conduction band and valence band, respectively. For H:TiO2 NWs, visible pumping excites an electron to the VO state which is followed by nonradiative exciton decay mediated by a manifold of trap states. UV pumping for the same sample excites an electron across the bandgap which is followed by relaxation to the VO state and subsequent relaxation through the manifold of trap states. For TiO2 NWs, UV pumping (visible pumping not possible) likewise excites an electron across the bandgap, which is followed by relaxation into a fluorescent trap state situated ∼0.3 eV below the CB. Subsequent recombination was fast due to trap states in the bandgap. 26828



■

trapped to free electrons increases, which changes exponentially as the Fermi level moves through the bandgap.47,48 This is most true when recombination occurs primarily from the conduction band and becomes different when considering surface-statemediated recombination.49,50 Assuming that the surface defect states have a similar distribution as that of bulk trap states, the distributed density of trap states near the TiO2 surface can be described by the volume fraction of bulk traps near enough to the surface of the nanostructure to participate in electron relaxation.

REFERENCES

(1) Fujishima, A.; Honda, K. Electrochemical Photolysis of Water at a Semiconductor Electrode. Nature 1972, 238, 37−38. (2) Linsebigler, A. L.; Lu, G.; Yates, J. T., Jr. Photocatalysis on TiO2 Surfaces: Principles, Mechanisms, and Selected Results. Chem. Rev. 1995, 95, 735−758. (3) Bak, T.; Nowotny, J.; Rekas, M.; Sorrell, C. C. Photoelectrochemical Properties of the TiO2-Pt System in Aqueous Solutions. Int. J. Hydrogen Energy 2002, 27, 19−26. (4) Biswas, S.; Hossain, M. F.; Takahashi, T. Fabrication of Gratzel Solar Cell with TiO2/CdS Bilayer Photoelectrode. Thin Solid Films 2008, 517, 1284−1288. (5) Liu, B.; Aydil, E. S. Growth of Oriented Single-Crystalline Rutile TiO2 Nanorods on Transparent Conducting Substrates for DyeSensitized Solar Cells. J. Am. Chem. Soc. 2009, 131, 3985−3990. (6) Lin, Y. J.; Zhou, S.; Liu, X. H.; Sheehan, S.; Wang, D. W. TiO2/TiSi2 Heterostructures for High-Efficiency Photoelectrochemical H2O Splitting. J. Am. Chem. Soc. 2009, 131, 2772−2773. (7) Murphy, A. B.; Barnes, P. R. F.; Randeniya, L. K.; Plumb, I. C.; Grey, I. E.; Horne, M. D.; Glasscock, J. A. Efficiency of Solar Water Splitting Using Semiconductor Electrodes. Int. J. Hydrogen Energy 2006, 31, 1999−2017. (8) Fitzmorris, R. C.; Larsen, G.; Wheeler, D. A.; Zhao, Y.; Zhang, J. Z. Ultrafast Charge Transfer Dynamics in Polycrystalline CdSe/TiO2 Nanorods Prepared by Oblique Angle Co-Deposition. J. Phys. Chem. C 2012, 116, 5033−5041. (9) Baker, D. R.; Kamat, P. V. Disassembly, Reassembly, and Photoelectrochemistry of Etched TiO2 Nanotubes. J. Phys. Chem. C 2009, 113, 17967−17972. (10) Hensel, J.; Wang, G. M.; Li, Y.; Zhang, J. Z. Synergistic Effect of CdSe Quantum Dot Sensitization and Nitrogen Doping of TiO2 Nanostructures for Photoelectrochemical Solar Hydrogen Generation. Nano Lett. 2010, 10, 478−483. (11) Asahi, R.; Morikawa, T.; Ohwaki, T.; Aoki, K.; Taga, Y. VisibleLight Photocatalysis in Nitrogen-Doped Titanium Oxides. Science 2001, 293, 269−271. (12) Liu, G.; Wang, X.; Wang, L.; Chen, Z.; Li, F.; Lu, G. Q.; Cheng, H. M. Drastically Enhanced Photocatalytic Activity in Nitrogen Doped Mesoporous TiO2 with Abundant Surface States. J. Colloid Interface Sci. 2009, 334, 171−175. (13) Li, Y.; Zhang, J. Z. Hydrogen Generation from Photoelectrochemical Water Splitting Based on Nanomaterials. Laser Photonics Rev. 2011, 4, 517−528. (14) Doan, N. M.; Estrellan, C. R.; Purnomo, A.; Gallardo, S.; Salim, C.; Hinode, H. Characterization and Photocatalytic Activity of NanoTiO2 Doped with Iron and Niobium for Turquoise Blue Dye Removal. ASEAN J. Chem. Eng. 2012, 1, 34−41. (15) Ye, G. X.; Wu, B.; Chen, T.; Zhang, L. K.; Wang, M.; Chen, L.; Liu, H. L.; Huang, C. R.; Li, J. L. Crystal, Electronic and Magnetic Structure of Co and Ag Doped Rutile TiO2 from First-Principles Calculations. Adv. Mater. Res. 2012, 399, 1789−1792. (16) Zhong, M.; Shi, J.; Xiong, F.; Zhang, W.; Li, C. Enhancement of Photoelectrochemical Activity of Nanocrystalline CdS Photoanode by Surface Modification with TiO2 for Hydrogen Production and Electricity Generation. Sol. Energy 2012, 86, 756−763. (17) Chen, X.; Liu, L.; Yu, P. Y.; Mao, S. S. Increasing Solar Absorption for Photocatalysis with Black Hydrogenated Titanium Dioxide Nanocrystals. Science 2011, 331, 746−750. (18) Wang, G.; Wang, H.; Ling, Y.; Tang, Y.; Yang, X.; Fitzmorris, R. C.; Wang, C.; Zhang, J. Z.; Li, Y. Hydrogen-Treated TiO2 Nanowire Arrays for Photoelectrochemical Water Splitting. Nano Lett. 2011, 11, 3026−3033. (19) Wheeler, D. A.; Wang, G.; Fitzmorris, R. C.; Adams, S. A.; Li, Y.; Zhang, J. Z. Ultrafast Charge Carrier Dynamics and Photoelectrochemical Properties of Hydrogen-Treated TiO2 Nanowire Arrays. MRS Proc. 2012, 1387, e04−07. (20) Chen, X.; Liu, L.; Liu, Z.; Marcus, M. A.; Wang, W.-C.; Oyler, N. A.; Grass, M. E.; Mao, B.; Glans, P-.A.; Yu, P. Y.; et al. Properties of

5. CONCLUSION We have carried out a systematic study in order to gain fundamental insight into the chemical nature of the bandgap states in hydrogen-treated TiO2 NWs that results in visible light absorption. Pristine, rutile-phase TiO2 strongly absorbs UV light due to its bandgap of 3.0 eV. Hydrogen treatment, however, introduces mid-bandgap states, attributed to oxygen vacancies or VO, as well as nonfluorescent trap states situated below the localized VO state. This VO state is partly responsible for the visible absorption of the hydrogen-treated TiO2. Moreover, the manifold of states below the VO leads to fast recombination of charge carriers following visible excitation into the VO state. ESR data, collected on hydrogen-treated rutile, P25, and NWs, indicate that the VO state is a singly occupied state of largely Ti3+ character. In addition, an oxygen-centered radical, possibly obtained by thermalization of some electrons on surface oxygen sites, is observed which is attributed to an electron in the antibonding orbital (i.e., an extra electron). Ultrafast TA measurements suggest the charge carrier relaxation to the VO state with above-bandgap excitation using UV light is relatively slower as compared to that of untreated TiO2 NWs. Upon excitation with visible light, however, the charge carrier recombination of the electron−hole pair is very fast, attributed to trap states below the VO. The new bandgap states introduced from hydrogen treatment led to PL quenching due to increased nonradiative decay. The overall short lifetime of charge carriers generated with visible light excitation is unfavorable for device applications such as PEC since they are too short-lived for efficient collection. It would be highly desirable to design and create bandgap states with long lifetime for PEC or photovoltaic applications. It is unclear at this point if and how the bandgap states probed in TA are responsible for PEC. Correlation between the lifetime of bandgap states and the PEC performance is likely more complicated than anticipated due to the multiple steps or time scales in charge transport and the multiple time scales of charge carrier relaxation.

■

Article

AUTHOR INFORMATION

Corresponding Authors

*E-mail [email protected], Ph (831)-459-1952 (Y.L.). *E-mail [email protected], Ph (831) 459-3776 (Z.Z.). Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS J.Z.Z. acknowledges the BES Division of the U.S. DOE (DEFG02-ER46232) for financial support. Y.L. acknowledges the support of this work partially by U.S. NSF (DMR-0847786) and UCSC faculty startup funds. Use of the Center for Nanoscale Materials was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract DE-AC02-06CH11357. 26829



Article

(42) Bjorheim, T. S.; Kuwabara, A.; Norby, T. Defect Chemistry of Rutile TiO2 from First Principles Calculations. J. Phys. Chem. C 2013, 117, 5919−5930. (43) Koudriachova, M. V.; de Leeuw, S. W.; Harrison, N. M. FirstPrinciples Study of H Intercalation in Rutile TiO2. Phys. Rev. B 2004, 70, 165421−165425. (44) Deak, P.; Aradi, B. l.; Frauenheim, T. Polaronic Effects in TiO2 Calculated by the HSE06 Hybrid Functional: Dopant Passivation by Carrier Self-Trapping. Phys. Rev. B 2011, 83, 155207−155213. (45) Wang, G. M.; Wang, H. Y.; Ling, Y. C.; Tang, Y. C.; Yang, X. Y.; Fitzmorris, R. C.; Wang, C. C.; Zhang, J. Z.; Li, Y. Hydrogen-Treated TiO2 Nanowire Arrays for Photoelectrochemical Water Splitting. Nano Lett. 2011, 11, 3026−3033. (46) Harris, L. A. Titanium Dioxide Hydrogen Detector. J. Electrochem. Soc. 1980, 127, 2657−2662. (47) Bisquert, J.; Cahen, D.; Hodes, G.; Ruhle, S.; Zaban, A. Physical Chemical Principles of Photovoltaic Conversion with Nanoparticulate, Mesoporous Dye-Sensitized Solar Cells. J. Phys. Chem. B 2004, 108, 8106−8118. (48) Zaban, A.; Greenshtein, M.; Bisquert, J. Determination of the Electron Lifetime in Nanocrystalline Dye Solar Cells by Open Circuit Voltage Decay Measurements. ChemPhysChem 2003, 4, 859−864. (49) Hamann, T. W.; Farha, O. K.; Hupp, J. T. Outer-Sphere Redox Couples as Shuttles in Dye-Sensitized Solar Cells: Performance Enhancement Based on Photoelectrode Modification via Atomic Layer Deposition. J. Phys. Chem. C 2008, 112, 19756−19764. (50) DeVries, M. J.; Pellin, M. J.; Hupp, J. T. Dye-Sensitized Solar Cells: Driving-Force Effects on Electron Recombination Dynamics with Cobalt-Based Shuttles. Langmuir 2010, 26, 9082−9087.

Disorder-Engineered Black Titanium Dioxide Nanoparticles Through Hydrogenation. Sci. Rep. 2013, 3, 1−7. (21) Cronemeyer, D. C. Infrared Absorption of Reduced Rutile TiO2 Single Crystals. Phys. Rev. 1959, 113, 1222−1226. (22) Fabregat-Santiago, F.; Barea, E. M.; Bisquert, J.; Mor, G. K.; Shankar, K.; Grimes, C. A. High Carrier Density and Capacitance in TiO2 Nanotube Arrays Induced by Electrochemical Doping. J. Am. Chem. Soc. 2008, 130, 11312−11316. (23) Lu, G.; Linsebigler, A.; Yates, J. T., Jr. Ti3+ Defect Sites on TiO2 (110): Production and Chemical Detection of Active Sites. J. Phys. Chem. 1994, 98, 11733−11738. (24) Wang, H.; Wang, G.; Ling, Y.; Lepert, M.; Wang, C. C.; Zhang, J. Z.; Li, Y. Photoelectrochemical Study of Oxygen Deficient TiO2 Nanowire Arrays with CdS Quantum Dot Sensitization. Nanoscale 2012, 4, 1463−1466. (25) Liu, H.; Ma, H.; Li, X.; Li, W.; Wu, M.; Bao, X. The Enhancement of TiO2 Photocatalytic Activity by Hydrogen Thermal Treatment. Chemosphere 2003, 50, 39−46. (26) Newhouse, R. J.; Wang, H.; Hensel, J. K.; Wheeler, D. A.; Zou, S.; Zhang, J. Z. Coherent Vibrational Oscillations of Hollow Gold Nanospheres. J. Phys. Chem. Lett. 2011, 2, 228−235. (27) Zuo, F.; Wang, L.; Wu, T.; Zhang, Z.; Borchardt, D.; Feng, P. SelfDoped Ti3+ Enhanced Photocatalyst for Hydrogen Production Under Visible Light. J. Am. Chem. Soc. 2010, 132, 11856−11857. (28) Hurum, D. C.; Agrios, A. G.; Gray, K. A.; Rajh, T.; Thurnauer, M. C. Explaining the Enhanced Photocatalytic Activity of Degussa P25 Mixed-Phase TiO2 Using EPR. J. Phys. Chem. B 2003, 107, 4545−4549. (29) Hurum, D. C.; Gray, K. A.; Rajh, T.; Thurnauer, M. C. Recombination Pathways in the Degussa P25 Formulation of TiO2: Surface Versus Lattice Mechanisms. J. Phys. Chem. B 2005, 109, 977− 980. (30) Gul, S.; Cooper, J. K.; Corrado, C.; Vollbrecht, B.; Bridges, F.; Guo, J.; Zhang, J. Z. Synthesis, Optical and Structural Properties, and Charge Carrier Dynamics of Cu-Doped ZnSe Nanocrystals. J. Phys. Chem. C 2011, 115, 20864−20875. (31) Hug, S. J.; Lewis, J. W.; Einterz, C. M.; Thorgeirsson, T. E.; Kliger, D. S. Nanosecond Photolysis of Rhodopsin: Evidence for a New BlueShifted Intermediate. Biochemistry 1990, 29, 1475−1485. (32) Bowker, M.; Bennett, R. A. The Role of Ti3+ Interstitials in TiO2 (110) Reduction and Oxidation. J. Phys.: Condens. Matter 2009, 21, 474224−474224. (33) Diebold, U.; Lehman, J.; Mahmoud, T.; Kuhn, M.; Leonardelli, G.; Hebenstreit, W.; Schmid, M.; Varga, P. Intrinsic Defects on a TiO2(110)(1× 1) Surface and Their Reaction with Oxygen: A Scanning Tunneling Microscopy Study. Surf. Sci. 1998, 411, 137−153. (34) Lisachenko, A.; Kuznetsov, V.; Zakharov, M.; Mikhailov, R. The Interaction of O2, NO, and N2O with Surface Defects of Dispersed Titanium Dioxide. Kinet. Catal. 2004, 45, 189−197. (35) Shklover, V.; Nazeeruddin, M.-K.; Zakeeruddin, S.; Barbe, C.; Kay, A.; Haibach, T.; Steurer, W.; Hermann, R.; Nissen, H.-U.; Gratzel, M. Structure of Nanocrystalline TiO2 Powders and Precursor to Their Highly Efficient Photosensitizer. Chem. Mater. 1997, 9, 430−439. (36) Kumar, P. M.; Badrinarayanan, S.; Sastry, M. Nanocrystalline TiO2 Studied by Optical, FTIR and X-ray Photoelectron Spectroscopy: Correlation to Presence of Surface States. Thin Solid Films 2000, 358, 122−130. (37) Howe, R. F.; Gratzel, M. EPR Observation of Trapped Electrons in Colloidal TiO2. J. Phys. Chem. 1985, 89, 4495−4499. (38) Ookubo, A.; Kanezaki, E.; Ooi, K. ESR, XRD, and DRS Studies of Paramagnetic Ti3+ Ions in a Colloidal Solid of Titanium Oxide Prepared by the Hydrolysis of TiCl3. Langmuir 1990, 6, 206−209. (39) Ghosh, A. K.; Wakim, F.; Addiss, R., Jr. Photoelectronic Processes in Rutile. Phys. Rev. 1969, 184, 979−988. (40) Hachiya, K.; Kondoh, J. Photoluminescence from Localized States in Rutile by Ar+-Ion Laser Excitation. Physica B: Condens. Matter 2003, 334, 130−134. (41) Bjorheim, T. S.; Stolen, S.; Norby, T. Ab Initio Studies of Hydrogen and Acceptor Defects in Rutile TiO2. Phys. Chem. Chem. Phys. 2010, 12, 6817−6825. 26830


Probing the Nature of Bandgap States in Hydrogen-Treated TiO2

Recommend Documents