Letter pubs.acs.org/JPCL
Instability of Hydrogenated TiO2 Manjula I. Nandasiri,† Vaithiyalingam Shutthanandan,† Sandeep Manandhar,† Ashleigh M. Schwarz,† Lucas Oxenford,† John V. Kennedy,‡ Suntharampillai Thevuthasan,† and Michael A. Henderson*,§ †
Environmental Molecular Sciences Laboratory, Pacific Northwest National Laboratory, PO Box 999 MS K8-93, Richland, Washington 99352, United States ‡ National Isotope Centre, GNS Science, Lower Hutt 5010, New Zealand § Fundamental and Computational Sciences Directorate, Pacific Northwest National Laboratory, PO Box 999 MS K8-87, Richland, Washington 99352, United States S Supporting Information *
ABSTRACT: Hydrogenated TiO2 (H-TiO2) is touted as a viable visible light photocatalyst. We report a systematic study on the thermal stability of H-implanted TiO2 using nuclear reaction analysis (NRA), Rutherford backscattering spectrometry, ultraviolet photoelectron spectroscopy, and X-ray photoelectron spectroscopy. Protons (40 keV) implanted at a ∼2 atom % level within a ∼120 nm wide profile of rutile TiO2(110) were situated ∼300 nm below the surface. NRA revealed that this H-profile broadened toward the surface after annealing at 373 K, dissipated out of the crystal into vacuum at 473 K, and was absent within the beam sampling depth (∼800 nm) at 523 K. Photoemission showed that the surface was reduced in concert with these changes. Similar anneals had no effect on pristine TiO2(110). The facile bulk diffusivity of H in rutile at low temperatures, as well as its interfacial activity toward reduction, significantly limits the utilization of H-TiO2 as a photocatalyst.
T
he wide optical bandgap of TiO2 (≥3 eV) represents a significant limitation to its use in solar energy conversion. Much effort has been invested in engineering the TiO2 bandgap to increase visible light absorptivity, particularly through doping.1−6 Recently, hydrogenated TiO2 (H-TiO2) thin films and nanostructures have attracted considerable attention because of their photocatalytic activity in the visible region,7−23 which is generally ascribed to an upward shift of the valence band maximum (VBM).7−9,15,18,20,22−25 However, visible light photocatalytic performance for H-TiO2 depends significantly on the preparation methods,13 with an apparent balance existing between the extent of hydrogen incorporated and the undesirable formation of certain bulk/interfacial defects.7,19,26 In this study, we show that hydrogen implanted into rutile TiO2(110) diffuses to the surface at low temperatures (∼373 K) and is completely depleted from the near-surface region (≤800 nm) by 523 K. The outward diffusion and depletion of H from TiO2 is accompanied by extensive surface reduction within the probe depths of X-ray photoelectron spectroscopy (XPS) and ultraviolet photoelectron spectroscopy (UPS). This reduction most likely results from the reaction of H with surface oxygen, followed by formation and desorption of water. The presence of surface Ti3+ persists until the thermally induced surface-to-bulk diffusion of Ti3+ interstitials is initiated above 550 K. Nuclear reaction analysis (NRA) was performed on Himplanted samples to determine the hydrogen depth profiles as a function of annealing temperatures (Figure 1). The as© 2015 American Chemical Society
Figure 1. Hydrogen depth profiles measured by the resonant 1H(19F,αγ)16O nuclear reaction analysis for the hydrogen-implanted TiO2(110) single-crystal sample after annealing in vacuum. Inset displays the integrated area in the profile as a function of annealing temperature.
implanted hydrogen profile was peaked at a depth of 300 nm, consistent with that predicted by SRIM simulations,27 although the profile shape was not Gaussian as expected. These simulations (see Figure S1 in the Supporting Information) Received: October 5, 2015 Accepted: November 6, 2015 Published: November 6, 2015 4627
DOI: 10.1021/acs.jpclett.5b02219 J. Phys. Chem. Lett. 2015, 6, 4627−4632
Letter
The Journal of Physical Chemistry Letters estimated a defect formation probability from 45 keV H+ of 1 in 1000 for displacement of an O atom and virtually no probability for Ti displacement.28 Given the dose of H+ (see Experimental Methods), the probability of structural defect formation from the irradiated samples used in this study is on the order 1 per every 5000 TiO2 unit cells. A broad shoulder extending deeper in the crystal is suggestive of hydrogen diffusion during the implantation process. The hydrogen profile was significantly altered by annealing at 373 K, with the peak diminished in height and skewed toward the surface in a nearly uniform concentration gradient. Little or no hydrogen was lost from the crystal (into the vacuum) during the 373 K anneal, as shown by the integrated NRA signal (inset to Figure 1). Interestingly, although the H profile broadened toward the surface during heating, there was no significant broadening toward the bulk. A significant amount of hydrogen was lost into the vacuum after annealing at 473 K (see Figure 1 inset). This was particularly evident in the region located below the original Hprofile (i.e., deeper than 400 nm) which was depleted to the background level. A remnant of the original profile remained after the 473 K anneal. This signal may be due to H trapped at structural defects generated during the implantation process (see below). Further annealing at 523 K completely removed all of the implanted H within the sampling depth of the NRA experiment. Accumulation of lattice damage from H-implantation, as well as the subsequent thermal recovery, was studied using Rutherford backscattering spectrometry (RBS/C). By “damage”, we refer to any structural abnormality or irregularity in the lattice that would impact the channeling of RBS He+ ions to an extent not seen in the pristine sample. Such signal attenuations could arise from a variety of effects that include incorporated species, displaced lattice atoms or lattice defects. For example, association of an implanted H with a lattice oxygen forming an OH could displace the oxygen from its normal position. In Figure 2a, channeling spectra from the H-implanted sample are presented after several anneals that match those used in Figure 1. For comparison, a random spectrum from the H-implanted sample is also shown (black line). (The term “random” indicates that the ion beam was not oriented along a lattice vector reflective of the bulk crystal structure.) The arrows in the figure indicate the energy positions expected for helium backscattered from Ti and O atoms at the surface, as well as from the damage region,29 that is, the region most likely to experience structural distortions as a result of the atoms in the lattice “stopping” the 40 keV H+ beam. The damage peak broadened and shifted toward the surface as the H-implanted sample was annealed at 373 K. As 373 K is too low for intrinsic point defect diffusion in rutile TiO2,30 the broadening of the damage peak detected by RBS can be ascribed to the diffusion of hydrogen toward the surface, in agreement with the NRA data (see Figure 1). Note also in Figure 2b that the Ti peak reflective of changes in the surface region increased slightly after the 373 K anneal. We ascribed this to the arrival of H at the surface from the bulk. High concentrations of surface H (>0.5 ML) can be formed on the TiO2(110) surface by dosing H atoms,31,32 by electron irradiation of ice films,33 or by photochemical reactions.34 In the latter two cases, temperature-programmed desorption shows that highly H-covered TiO2(110) is unstable above ∼400 K with the vast majority of H being accounted for in desorption of water (rather than diffusion into the bulk).
Figure 2. (a) RBS/C spectra (2 MeV He+) from H-implanted TiO2(110) after various anneals. (b) Surface region of the channeling spectra extracted from panel a.
Additional annealing of the H-implanted surface between 473 and 523 K decreased the damage significantly. From these results, it is evident that some degree of the lattice distortions resulted from the presence of H in the bulk, but that annealing alleviates this through the diffusion of H out of the crystal, in agreement with the NRA profile measurements. Interestingly, scanning electron microscopy (SEM) analysis after heating Himplanted SrTiO3(100)35,36 or ZrO2(100)37 (at comparable levels and depths) reveals extensive exfoliation as a result of H2 bubble formation in the implanted region. However, no such macroscopic damage occurred for H-implanted TiO2(110) (see Figure S2). This suggests that the dynamics of H diffusion in rutile is facile compared to that in SrTiO3 and ZrO2. In the TiO2 case, H is delivered to the surface (for removal by desorption) before bulk H2 bubble formation can occur. This, of course, does not preclude formation of “nanobubbles” that would be incapable of causing mechanical damage in TiO2(110). The diffusion of H within the bulk of rutile TiO2 has been studied both experimentally and theoretically with general agreement. The barrier for H-hopping along the ⟨001⟩ lattice direction (i.e., the so-called “c” channel) is roughly 0.6 eV, which is nearly half that for motion perpendicular to this direction (i.e., along ⟨100⟩ directions or “a” channels).38−40 Diffusion along the ⟨001⟩ direction is detectable at ∼400 K38 and is 104 times faster than that along ⟨100⟩ at temperatures above 525 K.41 Using electrostatic calculations, Bates et al.39 proposed that the slow rate for H hopping along ⟨100⟩ directions was due to pivoting of internal OH groups to allow 4628
DOI: 10.1021/acs.jpclett.5b02219 J. Phys. Chem. Lett. 2015, 6, 4627−4632
Letter
The Journal of Physical Chemistry Letters H to move from one O to another. In contrast, direct hopping could occur along c with no distortion of the lattice. In order for bulk H to reach the surface of a (110)-oriented crystal, hops between ⟨001⟩ channels42 must occur along ⟨110⟩ channels (which has not been addressed in bulk studies) or combined hops along ⟨100⟩ channels. Although some controversy exists regarding simulations of subsurface H diffusion from bridging sites on the TiO2(110) surface,43,44 it would seem that the barrier for net motion along ⟨110⟩ lattice directions is comparable to, if not slightly less than, that for motion along ⟨100⟩ channels.31,34,45 Results from these H diffusion studies are consistent with our observations that the implanted H profile would be redistributed (by diffusion) in the temperature range near 373 K. However, these results do not explain why the profile shifted toward the surface to a greater extent than toward the bulk. The question arises as to whether lattice defects resulting from the implantation process could facilitate H diffusion from the profile toward the surface rather than deeper into the bulk (i.e., toward a region not subjected to the H-implantation process). In other words, bulk defects may promote H hopping. As it turns out, bulk diffusion studies indicate that intrinsic defects in rutile (resulting from vacuum reduction) do not enhance H diffusion and may actually inhibit hopping by acting as trap sites.38,41 A defect inhibiting effect would be consistent with the retention of a weak as-implanted profile after annealing at 473 K (see Figure 1) despite the significant removal of a majority of the H. However, it is possible that the defects caused by the H+ beam (at the 0.02% level, see above) are fundamentally different from the intrinsic defects in rutile, although annealing should (eventually) negate differences. We propose that the gradient shifted toward the surface, perhaps through stabilization of H at or near the surface, followed by irreversible removal of H into the vacuum, as discussed below. UPS measurements were carried out on the H-implanted TiO2(110) surface to gauge the extent of surface reduction during vacuum annealing. The VBM region is shown in Figure 3. As the H-implanted TiO2(110) surface was annealed from 300 to 463 K (Figure 3a), the only change in the spectra was a shift in the VB edge by ∼0.3 eV closer to the Fermi level. We associate this shift to the arrival at the surface of implanted H, which caused either band bending and/or generation of new occupied states in the TiO2 gap. The latter would be consistent with a decreased the optical bandgap for HTiO2.7−9,15,18,20,22−25 As the H-implanted TiO2(110) was further annealed (Figure 3b), a bandgap state appeared at ∼0.7 eV which is commonly attributed to generation of surface Ti3+ defect sites.42 This feature reached maximum intensity after annealing at ∼588 K, coincident with the growth of the core-level Ti3+ state in XPS (see Figure S3). The VB edge also shifted back with the annealing from 463 to 588 K (see arrow Figure 3b). Both of these effects, the formation of surface Ti3+ and the shifting of the bandgap back to its original value, reflect surface reactions associated with the depletion of H from the crystal through the interface, in agreement with the NRA data of Figure 1. Annealing above 588 K (Figure 3c) had little effect on the VB edge but slowly diminished the intensity of the gap state at 0.7 eV peak due to diffusion of Ti3+ interstitials from the surface to the bulk.30,46 Desorption of H2 can not be excluded as a depletion mechanism for bulk H, but this reaction could not be responsible for the observed surface reduction. Similarly, reactions of H with the surface C contamination would likely
Figure 3. UPS spectra for H-implanted TiO2(110) annealed from (a) 300 to 463 K (b) 463 to 588 K, and (c) 588 to 688 K. Arrows in panels a and b indicate the direction of the valence band edge shift with increasing temperature. Regions near the Fermi edge are shown in the insets.
not result in surface reduction. Heating pristine surfaces contaminated with carbon does not result in surface reduction at temperatures used in this study. In contrast, extraction of surface oxygen, forming water, would reduce the surface. The NRA data suggests that most of the implanted H appeared to transit into vacuum through the interface during the annealing process. The near-surface Ti3+ concentration increased to only ∼12% during annealing from 463 to 588 K (see inset of Figure S3c), though approximately 200 ML of H was implanted. This suggests that most of the implanted H departed as H2. The overall changes in the surface Ti3+ concentration during annealing of the H-implanted TiO2(110) are summarized in Figure 4 from the deconvoluted Ti 2p XPS data (circles) and the intensity of the UPS gap state (triangles). Similarly, surface Ti3+ concentrations during annealing of the pristine TiO2(110) are displayed in Figure 4 (squares). The latter did not 4629
DOI: 10.1021/acs.jpclett.5b02219 J. Phys. Chem. Lett. 2015, 6, 4627−4632
Letter
The Journal of Physical Chemistry Letters
loading into the XPS chamber. Despite this treatment, the surfaces of both samples were contaminated by C impurities. Under normal circumstances, this carbon contamination could easily be removed by ion sputtering and high-temperature annealing;47 however, such treatments would compromise the integrity of the implanted H. The surface C is not anticipated to affect the bulk properties of H in TiO2(110) but may exert nominal influence on the surface as the samples were heated. Samples were annealed in the XPS/UPS chamber at 5 × 10−9 Torr for 30 min at each temperature, with spectra collected at RT. Unlike measurements within the NRA/RBS chamber where the thermocouple was pressed against the sample, temperatures in the XPS/UPS system were obtained from a thermocouple attached to the sample holder. As a result, cited temperatures for the XPS/UPS system are higher than those from the NRA/RBS chamber. For all samples, an “accumulated” annealing approach was employed because of limited availability of implanted crystals. XPS was performed using a Kratos Axis Ultra DLD spectrometer, consisting of a highperformance Al Kα monochromatic X-ray source (1486.6 eV) and a high-resolution hemispherical analyzer. The X-ray source was operated at 150 W, and emitted photoelectrons were collected normal to the sample surface. Data was acquired in a 700 × 300 μm2 area with a pass energy of 20 eV, which produced a full width at half-maximum of 0.59 eV for the Ag 3d5/2 core level of a reference Ag surface. Charge neutralization was achieved with low-energy electrons (
