Synthesis and Structure of an Ion-Exchanged SrTiO 3 Photocatalyst with Improved Reactivity for Hydrogen Evolution

. After illumination, the amount of hydrogen in the vial headspace was analyzed with a gas chromatography. The light intensity of the LEDs was measured with a photodiode. The resulting photon ﬂuxes were integrated over the bottom area of the vial ( 4 . 9 cm 2 ), with the irradiation of the photocatalyst estimated to be 5 . 6 × 10 16 photons s − 1 . The AQY was calculated according to the following equation:


BaTiO heated in an excess of SrCl at
°C converts to SrTiO through an ion exchange reaction.The SrTiO synthesized by ion exchange produces hydrogen from pH water at a rate more than twice that of conventional SrTiO treated identically.The apparent quantum yield for hydrogen production in pure water of the ion exchanged SrTiO is .% under nm illumination.The catalyst resulting from ion-exchange differs from conventional SrTiO by having ≈ % residual Ba, inhomogeneous Cl-doping at a concentration less than %, Kirkendall voids in the centers of particles that result from the unequal rates of Sr and Ba diffusion together with the transport of Ti and O, and nanoscale regions near the surface that have lattice spacings consistent with the Sr-excess phase Sr TiO .The increased photochemical efficiency of this nonequilibrium structure is most likely related to the Sr-excess, which is known to compensate donor defects that can act as charge traps and recombination centers.

. Introduction
Photocatalytic water splitting, a reaction that can directly convert water into hydrogen and oxygen under illumination, has attracted significant attention in past decades. [ , ]At present, two of the major issues that limit its application are the rapid recombination of photogenerated charge carriers [ ] and surface back (TEM) because Sr-excess is known to compensate recombination centers.The observations demonstrate that ion exchange reactions make it possible to synthesize photocatalysts with structures not accessible by conventional approaches.S , Supporting Information), BaTiO has converted to SrTiO .The ion-exchanged SrTiO derived from BaTiO will be referred to as ix-SrTiO throughout the remainder of this paper.The equilibrium constant for the reaction:

. Results and Discussion
is greater than at °C (with a free energy change of -. kJ mol − in the standard condition), so it should not be surprising that SrTiO forms under these conditions.This is analogous to ion exchange reactions recently reported by O'Donnell et al. [ ] Pristine SrTiO without any treatment and SrTiO that has been heated in molten SrCl using the same conditions as the BaTiO ion exchange reaction will be referred to as SrTiO and s-SrTiO , respectively.The s-SrTiO and ix-SrTiO (both treated in molten SrCl in the same conditions) were loaded with a .wt% Rh-Cr-O cocatalyst and the rate of hydrogen production was measured using a parallelized and automated photochemical reactor (PAPCR). [ ]The Rh-Cr-O cocatalyst serves as electron accumulation sites and is kinetically active for charge transfer to H + /H redox couples, and the O /H O anodic half reaction is supposed to take place on the SrTiO surfaces. [ ]The catalysts were in DI water whose pH was adjusted to pH , , or , or in a % methanol solution with unadjusted pH.Photoreactions were stimulated by nm light from two LED chips.The results, shown in Figure c, clearly show that the ix-SrTiO produced more hydrogen than the conventional s-SrTiO under all conditions tested.The ix-SrTiO sample in pH water produced hydrogen at the greatest rate ( µmol h − ), which is more than twice the rate of hydrogen production from the conventional s-SrTiO ( µmol h -).The apparent quantum yield (AQY) of ix-SrTiO was determined using a larger reactor with a more well-defined illumination geometry described in the Experimental Section and Figure S (Supporting Information) and found to be .% in nm light.Note that the highest measured AQY at nm SrTiO is .%, but these particles had controlled shapes with two different cocatalysts selectively deposited on different facets of each crystal. [ ]Both s-SrTiO and ix-SrTiO produce hydrogen at a higher rate at pH than that at pH and , which is consistent with our previous work reported for undoped SrTiO single crystals [ ] and Al-doped SrTiO particles. [ ]This is because the amount of surface charge influences band bending, and pH corresponds to a surface charge where neither the photocathodic nor the photoanodic reaction limit the overall reaction rate.For ix-SrTiO , the rate is higher in pH water than that in % methanol solutions, indicating that the anodic half reaction is no longer the rate-limiting reaction.To check the photostability, the same samples have been tested for another run under the same conditions, and the hydrogen evolution data are shown in Figure S (Supporting Information), indicating that the catalysts are stable after at least h of illumination.To better understand the reaction that forms ix-SrTiO and its effect on the reactivity, the BaTiO precursor was exposed to molten SrCl at different temperatures for different times, with the mildest conditions being min at °C.The XRD patterns and SEM images from these materials are shown in Note that in all conditions, the patterns indexed are cubic, meaning that the critical concentration of Sr needed to convert the solid solution to a cubic phase (x = .in Sr x Ba -x TiO ) [ ] occurred within min at °C.However, the refined lattice parameters (see Figure b) indicated that the samples heated at and °C did not complete the conversion to SrTiO in min.Evidence for the incomplete and probably inhomogeneous reaction is found in the tail of the ( ) diffraction peak.The tail to smaller θ angles arises from regions of the crystals richer in Ba that have a larger lattice spacing.These annealing conditions, which did not result in a complete conversion to SrTiO , impacted the rate of hydrogen production in pH water (see Figure c).The samples heated at and °C did not produce hydrogen at a rate detectible by the PAPCR.However, all ix-SrTiO samples heated at °C produced measurable rates of hydrogen, with the samples heated for and h having the highest and similar rates of hydrogen production, suggesting that no significant composition changes occur with additional annealing at h.Note that the molten SrCl treatment also modifies the SrTiO surface and improves the reactivity; the poor reactivities at and min suggest that the surface modification is incomplete at such a short time, although the ion exchange is nearly complete.
The bulk composition was studied with inductively coupled plasma -optical emission spectrometry (ICP-OES) analysis and the results are shown in Table S (Supporting Information).Both s-SrTiO and ix-SrTiO have small but detectible amounts of Al.Approximately % Ba was detected in the ix-SrTiO , indicating the ion exchange reaction is nearly complete.To determine if Cl exchanged into the samples while exposed to SrCl , we used X-ray fluorescence (see Table S   the center of the particle.The presence of a Cl gradient is consistent with the XRF and XPS results; XPS, which is surface sensitive, detects no Cl (see Figure S , Supporting Information), while XRF, which is sensitive to deeper regions within the crystal, does detect Cl.
The Kirkendall voids and the inhomogeneous Cl distribution result from kinetic limitations in the ion exchange reaction.Based on the conclusion that Sr diffuses inward more slowly than Ba diffuses out, it is possible that the surface region is enriched in Sr.While there was no obvious Sr enrichment in the EDS maps, particles were also examined by much higherresolution TEM (see Figure ).In the high-resolution image (see Figure a), the spacing between the lattice fringes in the very outer region is measured to be .nm, consistent with the spacing of the { } planes of SrTiO .The selected area diffraction pattern (see Figure b) confirms that the image is near the [ ] zone axis.Below this, there are additional lattice fringes with twice the spacing.This is a lattice spacing not found in SrTiO ; it is also noteworthy that the lattice fringes with the wider spacing are misoriented by ≈ ° from the SrTiO { } planes.As illustrated in Figure c, this doubled period and small inclination angle are consistent with the { } planes of Sr TiO , if it were epitaxially oriented on SrTiO such that both the [ ] and [ ] axes of both structures were aligned.Sr TiO is referred to as a Ruddlesden-Popper phase and adopts the K NiF structure. [ ]Using the ideal lattice parameters, the misorientation between { } p and { } rp should be .°, where the subscripts p and rp denote the perovskite and Ruddlesden-Popper structure, respectively.
The observation of one set of lattice fringes is not sufficient for phase identification with certainty.However, Sr TiO is known to form from SrTiO in SrO-rich conditions and grow epitaxially on SrTiO with this orientation relationship, [ ] so this is a plausible explanation for the contrast in the TEM images.Note that the diffraction pattern in Figure b is formed from the entire field of view of Figure a and does not have spots corresponding to the larger lattice spacing.If the aperture is reduced to include only the area with the larger interplanar spacing, there is not enough intensity to form a pattern, consistent with the absence of spots in the pattern from the larger field of view.In an ideal case, we would collect images and diffraction patterns from a combination of low index orientations, but here we are constrained to studying only small intergrowths in the electron transparent edges of irregularly oriented crystals.Given these constraints, a likely interpretation of the larger interplanar spacing is that there are small intergrowths of the Sr-rich phase, Sr TiO near the SrTiO surface.
Schematics illustrating the differences between s-SrTiO and ix-SrTiO are presented in Figure .Among these differences, the residual Ba, Kirkendall voids, and Cl impurities seem unlikely to have a significant effect on the reactivity.The residual Ba, for example, is an isovalent impurity on the A-site and is not electrically active.Furthermore, the similarity of the bandgaps of SrTiO and BaTiO suggests that a small amount of Ba will not influence the absorption edges. [ ]This is verified by the fact that the diffuse reflectance spectra of ix-SrTiO and conventional s-SrTiO are indistinguishable.One might hypothesize that the Kirkendall voids increase light scattering in a way that increases absorption, but this was also not evident in the diffuse reflectance spectra.Finally, Cl substituted on an O site (Cl O • ) has a positive effective charge and this has the potential to influence the reactivity, but it is also present in the s-SrTiO .If the ix-SrTiO has more Cl O • in the center of the particle, as indicated by the EDS mapping, then the core might be relatively positively charged with respect to the surface and this would encourage the transport of photogenerated holes to the surface, promoting the oxidation reaction that is usually thought to be rate limiting. [ , ]However, the electrical char- Therefore, the effect of substitutional Cl on the reactivity is less clear.
Based on past work, the Sr-excess phase in the near surface region seems more likely to be responsible for the increased photocatalytic reactivity. [ -]For example, Kato et al. [ ]  reported that SrTiO with excess Sr obtains the highest watersplitting activity.More recently, Vijay et al. [ ] have compared the photocatalytic reactivity of SrTiO , Sr TiO , and Sr Ti O and found that both Sr-rich compounds evolved more hydrogen than SrTiO .While the mechanism by which excess Sr leads to improved reactivity is not entirely clear, Yamada et al. [ ] have shown that as the Sr/Ti ratio increases, charge carrier trapping decreases, and they associated this with an increase in the rate of water splitting.Moreover, charge separation could occur at the phase boundary between SrTiO and Sr TiO .Based on the DFT calculations by Bilc et al., [ ] the valance band edge of Sr TiO is about .eV higher than SrTiO along the [ ] direction; this can promote the migration of photogenerated holes to the surface.The construction of such heterojunction is known to contribute to the overall reactivity. [ , ]A schematic .

Conclusion
In summary, ix-SrTiO , produced by ion exchange from BaTiO in a SrCl melt, produces hydrogen at a rate more than twice that of conventionally synthesized s-SrTiO photocatalysts.Among the difference between ix-SrTiO and the conventional s-SrTiO , Sr enrichment is most likely responsible for its superior photocatalytic properties.These observations show that ion exchange reactions make it possible to synthesize structures and compositions not accessible by other means; in this case, the unique SrTiO produced from BaTiO is an improved hydrogen evolution catalyst.

. Experimental Section
Photocatalyst Synthesis: All catalysts were prepared using a molten salt method.In a typical reaction, .g SrTiO (Sigma-Aldrich, .%) or BaTiO (CERAC, .%) was mixed with SrCl (Alfa Aesar, .%) at a molar ratio of : with an agate mortar.The mixture was then transferred to a covered alumina crucible (Sigma-Aldrich, Coors highalumina, mL) and annealed at °C for h in air with a ramp rate of °C min − .Next, the mixture was centrifuged four times in deionized water and four times in ethanol to remove any residue of SrCl .The powders collected through the centrifugation was dried in an oven at °C over night.RhCrO x cocatalysts were deposited ( .wt% Rh and .wt% Cr) on all samples with an impregnation method described elsewhere. [ ]A total of mg of as-prepared powdered samples was dispersed into .mL of deionized water containing appropriate amounts of Na RhCl (Sigma-Aldrich) and Cr(NO ) • H O (Sigma-Aldrich, %) to yield . wt% Rh and Cr.The suspension was evaporated in a boiling water bath under constant manually stirring.The resulting powders were collected and heated at °C for h.Characterization: The crystal phase was analyzed with an Empyrean X'pert PRO X-ray Diffractometer (PANalytical, Philips, Netherlands) in the range of -° using a step size of .°, equipped with a highintensity ( kV, mA) Cu Kα radiation source (Kα = .Å, Kα = . Å). Scanning electron microscopy (SEM) images were obtained to determine the morphologies of all samples using a FEI Quanta with a kV accelerating beam and a spot size of .An Oxford fullanalytical XMAX mm SDD EDX detector was equipped on the SEM for chemical composition analysis (EDS).The particle cross-section samples were prepared using a focused ion beam (FIB) milling process with a gallium source on an FEI NOVA .TEM images were recorded with a FEI Tecnai F or a Thermal Fisher Themis at kV. X-ray photoelectron spectroscopy (XPS) spectra were conducted using a SPECS System with a PHOIBOS Analyzer.N adsorption-desorption measurements (Nova e, Quanta-chrome, FL), used to determine the specific surface areas of powders through a Brunauer-Emmett-Teller (BET) approach, were performed at K using a multipoint method.The sample was degassed at °C for h prior to measurement.UV-vis diffuse reflectance spectra were recorded on an OL Multi-Channel Spectroradiometer (Optronic Laboratories).ICP-OES was carried out at Element Technology to determine the Al + concentration.X-ray fluorescence spectroscopy (XRF) was used to measure the concentration of Cl -in samples using an Oxford X-supreme .High-Throughput Screening of Photocatalysts: Photocatalytic hydrogen generation was measured with a high-throughput method using a Parallelized and Automated Photochemical Reactor (PAPCR), which enabled the authors to study up to samples simultaneously. [ , , ] The reactor array consisted of glass vials with mg catalyst powders added, and the illumination was provided by two highpower LED chips (Chanzon).The illumination wavelength of the LEDs was nm.To explore the solution effect on the photocatalysts, .mL of DI water whose pH was adjusted to pH , , and or % methanol solution with an unadjusted pH was injected into each vial.The headspace of the vials was covered by a layer of H -sensitive film (DetecTape, Tape-Midsun Specialty Products) whose color would change from light to dark when exposed to H .The calibration between the film darkness and local H concentration had been extensively reported in the previous publication. [ ]During a h illumination, the H -sensitive film was captured by a camera every min and the pictures were used to calculate the rate of hydrogen production.A schematic of the reactor assembly was presented in Figure S a (Supporting information).
Apparent Quantum Yield Measurement: To measure the apparent quantum yield (AQY), a gas-chromatography-based reactor setup was employed and a schematic was given (Figure S b, Supporting Information).In a measurement, mg of degassed catalyst powders, together with mL DI water, was added into a -mL EPA glass vial with argon atmosphere inside.The vial was illuminated for h by two LEDs with a peak wavelength at nm.After illumination, the amount of hydrogen in the vial headspace was analyzed with a gas chromatography.The light intensity of the LEDs was measured with a photodiode.The resulting photon fluxes were integrated over the bottom area of the vial ( .cm ), with the irradiation of the photocatalyst estimated to be .× photons s − .The AQY was calculated according to the following equation:

Figure
Figure a compares the secondary electron microscope (SEM) images of SrTiO and BaTiO before and after being heated at °C for h in a ten times excess of molten SrCl in an alumina crucible.The X-ray diffraction (XRD) patterns of the resulted powders are presented in Figure b.Based on the diffraction patterns and refined lattice parameters (TableS, Supporting Information), BaTiO has converted to SrTiO .The ion-exchanged SrTiO derived from BaTiO will be referred to as ix-SrTiO throughout the remainder of this paper.The equilibrium constant for the reaction:

Figure .
Figure .SEM images of: a) pristine SrTiO , pristine BaTiO , s-SrTiO (SrTiO that has undergone a molten SrCl treatment), and ix-SrTiO (BaTiO that has undergone a molten SrCl treatment).Scale bar: µm.b) XRD patterns of the two samples with a standard SrTiO pattern plotted as black droplines (ICDD: --).c) Mean photocatalytic maximum H production rates (each determined from six samples) in % methanol solution, pH water, pH water, and pH water under nm illumination.The bars represent standard deviations.
Figure a and Figure S (Supporting Information), respectively.
, Supporting Information).Both samples contained a small amount of Cl, which likely substitutes on an O site.The ix-SrTiO sample has a greater Cl concentration than the conventional s-SrTiO sample.It is possible that the lattice expansion from the larger Ba ions makes it possible for more Cl to diffuse into the crystal.To examine the surface composition, X-ray photoelectron spectroscopy (XPS) experiments were conducted, and the results are shown in Figures S -S and Table S (Supporting Information).While both samples treated in molten SrCl have similar surface compositions, they are quite different from the untreated SrTiO sample, having more surface hydroxyls and a greater Sr/Ti ratio.The specific surface areas of both s-SrTiO and ix-SrTiO have been measured and presented in Table S and Figure S (Supporting Information), suggesting that the two materials have similar surface areas and reaction sites.To determine if there were composition gradients within the crystal, we used a focused ion beam (FIB) to cut into the middle of the sample and energy dispersive spectroscopy (EDS) to map the composition (see Figure ).To make it easier to interpret the SEM, a schematic illustration of the FIB experiment is shown in Figure a.As shown in Figure b, upon cutting the particles, it became apparent that the ix-SrTiO particles had internal cavities (labeled with arrows), while the conventional s-SrTiO did not (see Figure S , Supporting Information).These cavities likely form as a result of the well-known Kirkendall effect. [ ]If Ba diffuses out of the crystal faster than Sr diffuses in, then internal cavities can form, as long as it is also possible to diffuse Ti and O away from the center of the crystal.Based on UV-vis diffuse reflectance (Figure S , Supporting Information), these pores have no detectible effect on the optical absorption of the particles.To investigate the possibility that there was a gradient in the distribution of remaining minority components (Ba, Al, Cl), we used EDS mapping on the internal surface exposed by the FIB, as shown in Figure c.The elemental maps of the majority components (Sr, Ti, O) have a nearly uniform gradient from the upper left to the lower right, because the particle becomes thinner until it ends in the lower right-hand corner.The distributions of Ba and Al do not differ enough from the majority elements to clearly indicate a gradient.The Cl, on the other hand, appears to have a concentration maximum near

Figure
Figure .a) XRD patterns of BaTiO treated in the molten SrCl for a variety of temperature and time combinations.b) Lattice parameters refined from the XRD patterns.c) Photocatalytic H production rates of the samples in pH DI water.The bars represent standard deviations from six distinct measurements.

Figure
Figure .a) Schematic illustration of the preparation of ix-SrTiO particle cross-section samples via ion milling.b) SEM images of ix-SrTiO particle cross-section samples prepared via FIB milling.c) EDS element maps of the box region in (b).

Figure
Figure .a) High-resolution TEM image of a particle edge of ix-SrTiO .b) SAED pattern recorded from the lattice in (a).A schematic illustration of the structure in (a) is shown in: c).Note that this schematic is constructed assuming that Sr TiO has the same lattice constant for the a-lattice as SrTiO ( ) Sr TiO S rTiO 2 4 3 a a = .