Photocatalyzed Conversion of CO2 to CH4: An Excited-State Acid

Article pubs.acs.org/JPCA

Photocatalyzed Conversion of CO2 to CH4: An Excited-State Acid− Base Mechanism Edward G. Look and Harry D. Gafney* Department of Chemistry, City University of New York, Queens College, Flushing, New York 11367, United States ABSTRACT: Photoinduced reduction of CO2 by H2O occurs in nanoporous Vycor glass doped with tungsten oxides derived from 312 nm photolysis of physisorbed W(CO)6. In polished forms of Vycor, a hydrated precursor to monoclinic WO3 is formed, whereas in the unpolished glass, the WO3 precursor and mixed valence tungsten oxides that exhibit lower energy intervalence charge transfer (IVCT) transitions are formed. The relative amounts of the different tungsten oxide are attributed to structural differences between the polished and unpolished forms of PVG. Chemisorption converts CO2 to a formic acid-like species, and population of the conduction band of the WO3 precursor with 312 nm light or population of the IVCT state of the mixed valence oxide with ≥437 nm light photocatalyzes the reduction of chemisorbed by coadsorbed water yielding CH4 and O2. Regardless of the excitation wavelength, electronic spectra and the dependence of methane yield on absorbed energy and the energetics of the conversions implies that neither metal oxide acts as a source of reducing equivalents. Instead, excitation of the metal oxide induces a charge polarization that creates local acidic and basic regions about the oxide in which the reduction of chemisorbed CO2 and oxidation of chemisorbed H2O occur exergonically. The dependence of methane yield on surface pH, and the pH dependencies of the respective half reactions show that the pH gradient needed for the reactions to occur spontaneously fall within the range of known photoinduced changes in acid−base properties. Within this excited-state acid−base mechanism, the metal oxide is not a source of electrons, but a conduit of electrons and protons between two exergonic processes.

■

INTRODUCTION Photocatalzing multielectron, multiproton reactions are essential to solar energy conversion,1 and achieving photocatalysis with low-cost, abundant, thermodynamically robust materials is a scientific challenge. One avenue of investigation focuses on utilizing visible light to drive the reduction of CO2 to CH4 as a means to increase energy production while reducing greenhouse gas emission. The photocatalyzed conversion has been known to occur on various metal oxide surfaces2−17 and within doped nanoporous silicas18−30 for more than twenty-five years, yet these systems continue to require UV excitation and methane yields remain small. In both systems, photoexcitation populates the conduction band of the metal oxide, but how that one-photon, one-electron process converts into the eight electron, four proton reduction of CO2 is not clear. Detailed mechanistic information is needed to identify approaches to improving the efficiency of the conversion, and for driving the conversion with visible light. Our interest in this chemistry stems from the photochemistry of W(CO)6 or Mo(CO)6 physisorbed into Corning’s code 7930 porous Vycor glass (PVG) or base-catalyzed (NH3) tetramethoxysilane/methanol/water (TMOS/CH3OH/H2O) xerogels.22−30 A ≤350 nm photolysis of the physisorbed complexes leads to CO evolution followed by H2 and CH4 evolution. Stoichiometric measurements, 13C/12C and H/D labeling, dependence on coabsorbed water, and the detection of © 2013 American Chemical Society

32−74% of the stoichiometrically expected O2 correspond to the overall reaction CO2[ads] + 2H 2O[ads] → CH4[g] + 2O2[g]

(1)

22−27

where ads designates an adsorbed species. CO2 readily adsorbs onto PVG, but the initial evolution of 12CH4 in all experiments, including photolyses of W(13CO)6 or under 300 Torr of 13CO2 indicate a carbonaceous impurity, thought to be a C1 oxide within the silica matrix, is initially reduced to CH4.2,23 Exhaustive 312 nm photolysis depletes the 12C impurity sites, and subsequent exposure to 13CO2 leads to immediate 13CH4 evolution on continued photolysis implying 13 CO2 adsorbs into the depleted impurity sites and is reduced to methane.22,23 The detection of IR bands attributable to formaldehyde and methanol suggests reaction 1 occurs via the Fischer−Tropsch sequence22,23 2e‐

2e‐

2e‐

2e‐

CO2[ads] ⎯⎯⎯→ HCOOH → H 2CO ⎯⎯⎯→ CH3OH → CH4 + + 2H

2H

(2)

Photolysis decarbonylates the physisorbed complex, and when the photolyzed complex achieves an average stoichiomReceived: February 20, 2013 Revised: October 9, 2013 Published: October 11, 2013 12268

dx.doi.org/10.1021/jp401812k | J. Phys. Chem. A 2013, 117, 12268−12279

The Journal of Physical Chemistry A

Article

where the mixed valence tungsten oxides exhibit lower energy intervalence charge transfer (IVCT) states, population of these states with ≥437 nm light also leads to methane evolution. The dependence on absorbed energy when the samples are excited with 312 nm light, along with the ability to drive the conversion via population of the IVCT state of the mixed valence oxide implies the metal oxide not acting as a one-photon, oneelectron source of the eight electrons needed for reaction 1. Rather, the dependence of methane yield on the pH of the glass surface, the pH dependencies of the respective half reactions, and the energetics of the conversion relative to the absorbed energy lead us to propose that the photocatalyzed reduction of CO2 occurs by an excited-state acid−base (ESAB) mechanism. The ESAB mechanism is a proton-coupled, electron-transfer mechanism,42−45 except the light energy is not used to drive electron transfer from the tungsten oxide to the CO2, but to create a pH gradient where the oxidation of coadsorbed water and the reduction of chemisorbed CO2 occur exergonically. Population of the conduction band of WO3, or the lower energy IVCT state of the mixed valence oxides polarizes charge, and analogous to the polarization created in transition metal complexes,46−50 creates regions of high acidity and basicity about the tungsten oxide. The tungsten oxide is not the source of reducing equivalent per se, but instead, by creating acidic and basic regions, shuttles protons and electrons between the two exergonic half reactions occurring within these regions.

etry of M(CO)4.0±0.2, H2 and CH4 evolution accompany CO evolution.22,23 Unable to achieve stability via octahedral coordination to the surface silanols, the tungsten photoproduct undergoes oxidation. In the absence of air, the evolution of H2 points to coadsorbed water, ubiquitous in these nanoporous silicas, as the oxidant.22,23 The zero valent photoproduct or the hydrogen evolved from its oxidation by coadsorbed water do not appear to be involved in the evolution of methane. Adding H2 (100 Torr) does not increases the yield of methane, and the evolution of methane is not stoichiometric with respect to oxidation of the tungsten carbonyl.22,23 CH4 evolution occurs from PVG samples impregnated with (NH4)2WO4, although the absorption spectrum of the oxide requires 254 nm excitation, which is also absorbed by the glass (50% T at 259 nm).25−27 With 312 nm excitation, turnover numbers, i.e., the moles of CH4 evolved/mol of W(CO)6 adsorbed, range from 5 to 53, the fraction of CO2 converted to CH4 ranges from 10 to 40% with turnover frequencies of 0.3 × 106 s−1 to 3.2 × 106 s−1, and quantum efficiencies reaching a high of 0.14 ± 0.03.22−30 All are considered lower limits since only CO2 adsorbed into a depleted 12C impurity site (vide infra) is converted to methane, and the availability of these chemisorption sites appears to be a major limitation on the conversion efficiency. The appearance of strong absorbance onsets in the general area of the MoO3 and WO3 band gaps, 2.6−3.0 eV,31−40 concurrent with the decline in the metal carbonyl absorptions during 312 nm photolysis attest to the formation of MoO3 and WO3.22 However, EPR also reveals the formation of a Mo(V) hydrous oxide, MoO2OH, during 312 nm photolysis.25−27 No EPR signal is detected with tungsten, but the stoichiometry of hydrogen evolution, 0.31 to 2.43 mols of H2 evolved per mol of W(CO)6 consumed, implies less than complete oxidation of some W atoms.22,23 WO3 and MoO3 are the spectroscopically evident oxides, but these other oxides, which contain the less than fully oxidized molybdenum or tungsten, could serve as a source of electrons for the reduction of CO2 (vide infra). These experiments, which focus on the W(CO)6−PVG system, were undertaken to gain additional insights into the conversion mechanism; specifically, the tungsten oxides formed and the factors that influence their formation within PVG, the chemisorption of CO2 onto PVG, and the energetics of the conversion. Spectroscopy and O2 sensitivity reveal two types of tungsten oxides are formed in PVG with the relative amounts of each dependent on whether the glass is polished or not. In polished PVG, a small, hydrated, amorphous aggregate containing W6+ that converts to monoclinic WO3 on heating is the spectrally evident oxide. This species is also formed in unpolished PVG along with the concurrent formation of mixed valence tungsten oxides. The different amounts of the different oxides formed in polished and unpolished PVG is attributed to structural differences on the length scale of the correlation length of the silica matrix, 22 ± 1 nm.41 Consistent with past 13 C/12C labeling experiments,22,23 diffuse reflectance FTIR (DRIFT) spectra reveal that the small fraction of CO2, which chemisorbs onto PVG and spectroscopically resembles formic acid, is converted to CH4. Methane evolution occurs from both polished and unpolished PVG impregnated with W(CO)6. In the polished samples, where WO3 is the spectrally evident product, excitation wavelengths of ≤350 nm are necessary. Although 312 nm excitation is energetically sufficient to populate the conduction band of WO3, methane evolution from these polished samples exhibits a first-order dependence on the absorbed energy. In addition, in the unpolished samples,

■

EXPERIMENTAL SECTION Materials. W(CO)6 (Alfa Aesar) was used as received since electronic spectra agreed with published spectra.22,51 WO3 (Pfaltz & Bauer, 99.75%), methane (Aldrich, Electronic grade, 99.998%), and CO2 (Matheson Tri·Gas, Lot# 1029100452B1) were used as received. 1 × 1 inch pieces of 2 mm thick polished and 4 mm thick unpolished Corning’s code 7930 porous Vycor glass (PVG) were cleaned and dried as previously described.22,23 The samples were impregnated with W(CO)6 by previously described solution adsorption techniques and dried under vacuum at room temperature.22,23 The amount of W(CO)6 adsorbed, which ranged from 10−6 to 10−5 mol/g of PVG, was calculated from the decrease in absorbance of the impregnating solution.22,23 The mixed valence form of tungsten oxide, which is also called photochromic tungsten oxide52 (vide infra) was created in situ by 312 nm photolysis of W(CO)6ads (ads designates an adsorbed species) in vacuo. The impregnated PVG samples were mounted upright in previously described rectangular quartz cells,22 degassed to p ≤ 10−4 Torr, and photolyzed in a Rayonet Reactor equipped with four 312 nm lamps to form the photochromic tungsten oxide and/or bronze in situ. The photoinduced conversion was monitored by UV−visible spectroscopy and continued until electronic spectra showed no further increase in absorbance at 632 nm. At this point, the different samples ranged from light yellow-green to blue with absorbances at 632 nm ranging from 0.32 to 0.85. The gases evolved during generation of the photochromic were removed under vacuum, p ≤ 10−4 Torr; the evacuated samples were charged with CO2 (50 to 100 Torr) and then photolyzed with visible light, ≥437 nm (vide infra). Provided the samples are not exposed to air or O2, the photochromic formed in PVG is quite stable. Samples stored in vacuo or handled in a glovebag filled with Ar, for example, show little or no decline in absorbance at 632 nm over a period of at least six months. 12269



Article

Photochemical Procedures. All samples were mounted upright in previously described rectangular quartz or pyrex cells and either irradiated in vacuo (p ≤ 10−4 Torr) or under specific pressures of CO2 introduced via a vacuum line.22 UV photolyses were carried out in a Rayonet reactor equipped with 312 nm bulbs. The intensity of 312 nm excitation, quanta/ sec·cm2, incident onto the glass samples was measured by ferrioxalate actinometry as previously described. 22 The uncertainty in the intensity incident onto the sample is 5− 8%. Visible light photolyses utilized a 1000 W Oriel Optics Xe lamp. The lamp output was collimated and passed through an eight inch plexiglass trough filled with distilled water that also serves as a mount for a Corning 3389 glass absorption filter. The combination of plexiglass, filter, and water limited the excitation incident onto the samples to 437−1140 nm light. The relative intensities of the visible light were also measured by ferrioxalate actinometry using an average value of 0.93 for the quantum efficiency of Fe2+ formation, in the 437−550 nm region,53 which corresponds to the light transmitted by the filters and that absorbed by the ferrioxalate. These measurements yield an average intensity of 1.8 ± 0.4 × 1017 quanta/s· cm2 incident onto the mixed valence oxides. Samples containing the photogenerated photochromic, mixed valence tungsten oxide were evacuated (p ≤ 10−4 Torr), then charged with pressures of CO2 (50 to 100 Torr) and irradiated with the 437−1140 nm light. Physical Measurements. UV−visible spectra were recorded on a Cary 5000 spectrometer adapted to hold the rectangular quartz cells. IR spectra were recorded on a Nicolet 20/5DX FTIR equipped with an intensified IR source, an MCTB detector maintained at 77 K, and a Harricks diffuse reflectance (DRIFT) accessory. To obtain the spectrum of chemisorbed CO2, a mixture of dried, crushed PVG (60−80 mesh), and IR grade KBr were placed in the DRIFT sample compartment, evacuated to p ≤ 10−4 Torr at room temperature, and the DRIFT spectra recorded before and after exposure to CO2. The sample was then evacuated to p ≤ 10−4 Torr to remove the physisorbed CO2, and the DRIFT spectrum recorded again. The spectrum of chemisorbed CO2 was obtained by subtracting the spectrum of PVG recorded prior to exposure to CO2 from the spectrum of PVG after exposing it to CO2 and then removing the physisorbed CO2 under vacuum at room temperature. GC analyses were performed on Shimadzu GC-8A gas chromatograph equipped with a 1/8″ diameter, 80/100 mesh, Restek 5A° molecular sieve column maintained at room temperature, and a thermal conductivity detector. The GC was calibrated by introducing known amounts of methane (Aldrich, Electronic grade, 99.998%) into the vacuum line, collecting and compressing the gas with the Toepler pump, and transferring the collected gas to the GC by means of a sample loop. Two cycles of expanding the gas into the Toepler pump and compressing the gas into the sample loop are required for quantitative transfer. Methane evolved during the photochemical experiments was quantitated in the same manner. The photolysis cell was attached to the vacuum line, and the evolved gases were expanded into the Toepler pump, compressed into the sample loop, and transferred to the GC for analysis.22,23 On the basis of the reproducibility of methane calibration, the uncertainties in the amounts of evolved CH4 is 5% for smaller peaks to 15% for larger peaks as per output from a Shimadzu GC-8A gas chromatograph.

Ablation of the photolyzed sample and secondary ionization mass spectral (SIMS) analysis of the ablated material were performed as previously described.54−56 All measurements were made under slight pressures (10−2 Torr) of O2 to convert the ablated tungsten to the corresponding monoxide cation, WO+, to improve collection efficiency.57 The depth of the sputtering craters was measured with a Veeco Instruments Stylus Profilometer Detak II with a precision of ca. 1 nm.57,58

■

RESULTS The UV−visible spectrum of W(CO)6ads in both unpolished and polished PVG closely resembles the electronic spectrum of the complex in hexane and acetonitrile.22,23,51 An intense absorption at 290 nm followed by less intense, overlapping absorptions with maxima at 315 and 350 nm are assigned to charge transfer and ligand-field transitions, respectively.51 The intensity of the 290 nm absorption relative to the 350 nm absorption for the adsorbed complex, 16 ± 2:1, is within experimental error of the relative intensities in fluid solution, 17:1,51 while the CO stretching vibration of W(CO)6ads, 1986 cm−1, is equivalent to that in hexane, 1986 cm−1.22,23 The similarity of the electronic and DRIFT spectra establish W(CO)6 physisorbs into both polished and unpolished PVG without disruption or significant distortion of its primary coordination sphere.22−27 Impregnation of PVG produces a uniform distribution of W(CO)6 in the outer volumes of PVG. With ≤10−4 mol of W(CO)6ads/g of PVG, the adsorbed complex uniformly impregnates the outermost 4 ± 0.2 × 105 nm of PVG.59,60 A 312 nm photolysis, however, changes the distribution of tungsten within the silica matrix. Ablation of the sample surface facing the excitation light with an Ar+ beam and SIMS analysis of the ablated material as a function of depth shows that the four prominent tungsten isotopes are uniformly distributed to a depth of 800−850 nm followed by a sharp decline in W content. There is a natural variation in the distribution of adsorbates in PVG;41,61 however, the distribution of tungsten after photolysis is well beyond that.61 Photolysis is also expected to produce the largest amount of W in the outermost volume since the excitation intensity is highest in these volumes, but not at the expense of the amount of W in interior volumes. Rather, the redistribution of tungsten resembles that found with Fe(CO)5 in PVG and is attributed to CO pressure gradients that develop within PVG during photolysis.61 Photoinduced release of one CO within the 7 ± 3 nm diameter pores in PVG, for example, corresponds to a CO pressure of 0.2 atm. W(CO)6 is not as volatile as Fe(CO)5, but like Fe(CO)5,62 it physisorbs into PVG and exhibits a high quantum efficiency of CO loss, Φ = 0.84 ± 0.09.22,23 A 312 nm photolysis of W(CO)6 physisorbed into polished or unpolished PVG produces initially equivalent spectral changes. Declines of the 315 nm charge-transfer and 350 nm ligand-field bands of W(CO)6ads are accompanied by the appearance of an intense absorption onset at ≤320 nm in both samples. GC analyses of the surrounding gas phase reveal the presence of CO and CH4. The quantum efficiencies of CO loss are within experimental error in both forms of PVG, Φ = 0.84 ± 0.09, and the total amount of CO evolved in both forms of PVG correspond to 97 ± 1% of the initially coordinated CO.22,23 Exposing both the polished and unpolished photolyzed samples to air or O2 has little effect of the UV absorption onset implying the species giving rise to this absorption onset 12270



Article

unpolished samples, exhibits an intense absorption onset in the UV within experimental error of the band gap of WO3, 2.6−3.0 eV,31−40 shows little or no absorption beyond 400 nm, and shows no measurable sensitivity to air or O2. This species is assigned to an oxide aggregate containing only W6+, i.e., WO3 (vide infra). The other oxide, which is spectrally evident in the unpolished samples, exhibits, in addition to the UV absorption, lower energy visible and near IR absorptions (Figure 1) similar to those of photochromic tungsten oxide52 and WO2.83, WO2.92, WO2.72, and WO2.63 These visible absorptions, which are bleached by air and O2, are assigned to IVCT absorptions of a mixed valence oxide aggregate that contains a mixture of W6+ and tungsten in lower oxidation states. Both types of oxides are formed in polished and unpolished PVG, but the relative amounts of each differ. In the polished samples, the fully oxidized oxide is the spectroscopically evident product, whereas in the unpolished glass, the fully oxidized oxide and the mixed valence oxide are spectrally evident. Since the structure of WO3 in aqueous solution is pH dependent,64,65 spectra of WO3 were recorded as a function of pH to further probe the oxides formed in PVG, which has a surface pH of 4−5.66 In aqueous solution, the composition of the oxide is described by the equilibrium65

in both the polished and unpolished PVG contains fully oxidized tungsten, i.e., W6+. While the initial UV spectral changes in polished and unpolished PVG are equivalent, the spectral changes in the visible and near-IR are quite different. The polished samples remain essentially colorless with ΔA beyond 400 nm ≤0.01, whereas the unpolished samples range in color from greenishyellow to dark blue (Figure 1, inset). The bluer samples show

WO4 2 − + 12H+ → W12O4212 − + 6H 2O

Figure 1. UV−visible spectrum of W(CO)6 photoproduct in rolled PVG. The dotted and dashed absorptions on the left are the absorption spectra of the cutoff filter and Plexiglas filter holder.

(3)

WO42−,64,65

In basic solution, tungsten exists as the tetrahedral and the electronic spectrum of a pH = 8 aqueous solutions of WO3 exhibits an intense absorption with a maximum at ca. 210 nm with a nondescript, weak broader absorption in the 240− 340 nm region. At pH = 6, the spectrum retains the 210 nm maximum, but a distinct shoulder at ca. 260 nm appears, while at pH = 2, the intense absorption shifts to 225 nm, and two distinct shoulders at 265 and 330 nm appear. The electronic spectrum of PVG impregnated with 3.8 × 10−6 mol of WO3/g by adsorption from a pH = 8 aqueous solution of WO3 and dried under vacuum at room temperature shows a strong absorption with a maximum at 230 nm. Increasing the WO3 loading to 1.7 × 10−5 mol/g shifts the absorption maximum to 256 nm. Instead of the distinct shoulders at 265 and 330 nm evident in pH = 2 and pH = 6 aqueous solution, however, the adsorbate exhibits a low intensity, broad, featureless absorption extending out to ca. 400 nm. Further increases in WO3 loading yield an absorption onset in the vicinity of 320 nm similar to the spectral changes accompanying 312 nm photolysis of W(CO)6ads. XRD and TEM analyses of the polished or unpolished samples after 312 nm photolysis of polished PVG containing 10−5 mol of W(CO)6g failed to produce measurable reflections attributable to a tungsten species or any indication of distinct tungsten oxide particles ≥1 nm in diameter. Heating the photolyzed, polished samples at 625 °C, however, produces a series of weak, broad reflections with the most prominent at 2θ = 23−24° and 33−35° that most closely resemble the most intense reflections of monoclinic WO3 (JCPDS pattern number 00−005−0386).58 Heating the photolyzed, unpolished samples, however, led to a loss of the low energy visible absorptions (Figure 1) without distinguishable reflections at 2θ = 23−24° and 33−35°. DRIFT spectra show that adsorption of WO3 into polished and unpolished PVG produces similar results; the intensity of the 3744 cm−1 free silanol band declines relative to that in calcined PVG.27 These declines, which range from 12 to 42%, are accompanied by slight increases in a broad 3450 cm−1 band assigned to adsorbed water.27 The increase parallels the

distinct maxima at 632 and 875 nm (Figure 1) with absorbances at 632 nm ranging from 0.32 to 0.85 for the different unpolished samples. An immediate question is whether the differences in color correspond to the formation of different oxides that absorb in the visible (Figure 1) or different amounts of the same oxides. In spite of the color differences, all of the unpolished samples exhibit similar photoand thermochemical behavior and common spectroscopic properties. Exposing the samples to air or O2, for example, leads to a loss of the lower energy, visible absorptions with little change in the UV implying the species that exhibit the visible absorptions (Figure 1) contain less than completely oxidized tungsten. The visible and near IR absorptions indeed resemble those of the photochromic form of tungsten oxide,52 which is thought to be oxide aggregate containing W6+ and lower oxidation states of tungsten, principally W5+.52 A preliminary fit of the visible and near-IR absorptions in Figure 1 is also consistent with the presence of WO2.83, WO2.92, WO2.72, and WO2,63 Removing the air under vacuum and re-exposing the O2 bleached, unpolished samples to 312 nm light regenerates the visible absorptions. Reversibility is initially close to quantitative but slowly declines with increasing number of photolysis−O2 exposure−photolysis cycles. Lastly, in spite of the color differences, the ratio of the absorbance at 632 nm relative to that at 875 nm, R632/875, for all unpolished samples examined fall in the range 1.17 ± 0.08. This narrow range of R632/875 values and the similarities in their photo- and thermochemical behavior suggest the differing colors arise from the formation of different amounts of the same oxides, as opposed to the formation of different oxides. Collectively, the spectral changes and the differing sensitivities to O2 distinguish two general types of tungsten oxides formed in PVG during 312 nm photolysis of W(CO)6ads. One type, which forms in both the polished and 12271



Article

out that 254 nm excitation is insufficient to populate the conduction band of the oxide.16 Small amounts of methane are also detected during 254 nm photolysis of unimpregnated PVG, which, corresponding to 4.9 eV, is significantly less than the band gap of bulk and thin film SiO2, 8.8 eV.67,68 Although the efficiencies of methane evolution differ, collectively, the results indicate that it is possible to drive the conversion, particularly in PVG, with less than band gap excitation, and by implication, without creation of a mobile electron/hole pair. Visible, ≥437 nm, light is absorbed by the lower energy IVCT absorptions, which arise from the presence of W6+ and lower oxidation states of tungsten in the oxide aggregate. In the mixed valence oxide, the electronic spectra suggest the presence of W5+52 and W4+.63 Any of these lower oxidation states, or W°, since the PVG is impregnated with W(CO)6, could serve as a source of reducing equivalents for the reduction of CO2. Previous experiments, however, preclude W° since CH4 evolution is not stoichiometric with oxidation of W(CO)6.22−27 Similarly, CH4 evolution from the unpolished samples occurs without a corresponding change in the visible absorptions of the mixed valence oxide. Photolysis of the unpolished samples containing the mixed valence oxide under 100 Torr of CO2 with either 312 or ≥437 nm light leads to CH4 evolution with no significant change in the intensities of the visible absorptions. The evolution of as much as 10−5 mol of CH4 from the different unpolished samples examined, for example, occurs with a ≤0.002 change in absorbance at 632 nm. Some samples actually show slight increases in absorbance, while the values of R632/875 for all samples remain within the range 1.17 ± 0.08. The evolution of CH4 without a corresponding change in the 632 and 875 nm IVCT absorptions of the mixed valence oxides52,63 precludes a stoichiometric reaction in which less than fully oxidized tungsten within the mixed valence oxides provides the reducing equivalents. Whether initiated by 312 nm induced population of the conduction band of the hydrous W6+ oxide or 437 nm induced population of the IVCT state of the mixed valence oxide, the absence of a stoichiometric reaction and the similar dependencies on coadsorbed water and CO2 pressure suggest the same mechanism; chemisorption of CO2 into a depleted C1 impurity site within the silica matrix and these metal oxides photocatalyzing its reduction by coadsorbed water. Adsorbing CO2 onto PVG shifts the CO2 asymmetric stretch from 2345 cm−1 in the gas phase to 2360 cm−1 when adsorbed onto PVG.24−27 Accompanying slight hypsochromic shifts of the free and associated silanol bands suggests the majority of CO2 physisorbs onto PVG via hydrogen bonding to the silanol groups. Consistent with a weak adsorbate−adsorbent interaction, physisorbed CO2 readily desorbs under vacuum at room temperature. Subtracting the spectrum of PVG from that after desorption reveals overlapping absorptions in the 1600−1650 cm−1 range and overlapping bands with maxima at 1697 and 1709 cm−1 (Figure 2a). These absorptions are assigned to chemisorbed CO2, and consistent with past 12C/13C labeling experiments, which indicate that only chemisorbed CO2 is reduced to methane,22,23 312 nm photolysis reduces the intensities of the 1697 and 1709 cm−1 bands (Figure 2b). Chemisorption of CO2 onto hydroxylated silicas is thought to produce carbonate-, bicarbonate-, and/or formic acid-like species.69−73 Assuming aqueous acid−base equilibrium constants, the pH of the glass surface, 4−5,66 suggests the majority of the carbonate-like species exist predominantly as the bicarbonate. Bicarbonates exhibit a strong absorption in vicinity

WO3 loading suggesting that adsorbed WO3 interacts with the surface silanol groups and adsorbed water present in the silica matrix. The inability to detect distinct particles by TEM, the broadness of the XRD reflections after heating, the changes in the DRIFT spectrum, and the lack of a sensitivity to O2 indicate that the photoproduct exhibiting the UV absorption onset in the vicinity of the band gap of WO3 is a small, ≥1 nm in diameter, hydrated, amorphous oxide aggregate that on drying forms monoclinic WO3.58 Similar changes in the UV and insensitivity to O2 during 312 nm photolysis of W(CO)6ads in unpolished PVG suggests that the same or very similar oxide aggregate precursor to monoclinic WO3 is also formed in unpolished PVG. In addition to this oxide, which contains only W6+, the appearance of visible and near-IR absorptions (Figure 1) and their sensitivity to O2 indicate the concurrent formation of mixed valence tungsten oxides in unpolished PVG. The differences in the relative amounts of the two oxides formed in polished and unpolished PVG are not consequences of changes in W(CO)6 on adsorption, differences in its photoreactivity, or differences in the composition or the hydroxylated surfaces of polished and unpolished PVG. Elemental analyses of the polished and unpolished glass are equivalent within experimental error, and DRIFT spectra of dried (550 °C) samples exhibit equivalent free silanol bands at 3750 ± 5 cm−1 with shoulders at 3650 ± 5 cm−1 assigned to associated or hydrogen-bonded silanol groups.22,23 Normalized to the 3750 cm−1 band, the relative intensities of the 3650 cm−1 shoulders indicate the relative ratios of free and associated silanols are basically the same in the polished and unpolished glasses. In addition to equivalent compositions and hydroxylated surfaces, both forms of PVG are dried and impregnated in the same manner. The difference is not due to changes in W(CO)6 or its photoreactivity. The electronic and IR spectra of W(CO)6 adsorbed into both forms of PVG are identical, and the quantum efficiencies of CO loss in both forms of PVG are within experimental error.22−28 A 312 nm photolysis is needed to generate the mixed valence oxides in the unpolished glass, but once formed, excitation of these oxides with visible light, ≥437 nm (Figure 1), under CO2 leads to CH4 evolution. Unpolished PVG samples containing W(CO)6ads were photolyzed with 312 nm light until spectra recorded periodically during photolysis showed no further increase in absorbance at 632 nm. At that point, the photolyzed sample was evacuated under vacuum, p ≤ 10−4 Torr, and charged with 60 to 100 Torr of CO2. Photolysis of the charged samples with ≥437 nm light leads to methane evolution. The amount of methane evolved increases with increasing CO2 pressure, and on the basis of the light transmitted by the filters (Figure 1) and absorbed by the ferrioxalate actinometer, ca. 437−550 nm, the quantum efficiency of methane evolution from the different unpolished samples, under 100 Torr of CO2, falls within the range, ΦCH4 = 2.1 ± 2.0 × 10−4. The values of ΦCH4 obtained with visible excitation are smaller than those obtained during 312 nm photolysis of the same samples under 100 Torr of CO2, ΦCH4 = 4.0 ± 1.0 × 10−3. Considering the width of the visible excitation wavelengths and the breadth of the visible absorptions (Figure 1), one could readily argue that the >437 nm light populates the conduction band of a form of tungsten oxide with a band gap of ≤2.8 eV since the reported band gaps of WO3 range from 2.6 to 3.0 eV.31−40 However, driving the conversion with less than band gap excitation is not without precedent. Yoshida and Kohno, for example, report the conversion occurs on ZrO2 with 254 nm excitation, and point 12272



Article

CO2.14 A 312 nm excitation, 4.0 eV, is energetically capable of populating the WO3 conduction band 2.6−3.0 eV above the valence band.31−40 Assuming these mobile electrons go on to reduce CO2, the Stark−Einstein Law (2nd Law of Photochemistry) states that one photon induces one event or, in this case, creates one electron/hole pair.76,77 Yet, the moles of CH4 evolved during 312 nm photolysis exhibits a dependence on absorbed energy (Figure 3), that is basically first order. This is

Figure 2. (a) Residual absorptions after subtracting spectrum of PVG from the spectrum of PVG exposed to CO2, and the physisorbed CO2 was removed under vacuum at room temperature. (b) Spectrum of sample after 312 nm photolysis.

Figure 3. Amount of CH4 evolved from polished PVG samples containing 10−5 mol of W(CO)6g as a function of the amount 312 nm excitation energy absorbed.

of 2850 cm−1.74 The absence of a band in this region, however, and the appearance of weak absorptions in the C−H region at 2967 and 2944 cm−1 point to an adsorbate more like formic acid than bicarbonate. There are two steps in the production of PVG where the chemisorption site, thought to be C1 impurity within the SiO2 matrix, could be formed. PVG derives from a 96% SiO2, 3% B2O3, 1% Na2O, and Al2O3 melt in which B2O3 phase separates from the silicates during cooling.56 Acid leaching removes the borate phase leaving a transparent material consisting of silica nodules with the intervening spaces defining the random array of interconnected 7 ± 3 nm diameter pores throughout the material.56 One possibility is during the rolling/cooling of the melt, while the other is during acid-leaching of the borates from the phase separated glass. In both cases, the carbon source is thought to be atmospheric CO2. To test the acid-leaching step, a sample of the glass was leached with carbonic acid, H2CO3. The leached glass was then dried and impregnated with W(CO)6 in the same manner as the polished and unpolished samples of PVG. Photolysis of the impregnated sample with 312 nm light, however, did not lead to a noticeable increase in methane evolution.75 This suggests that C1 oxide impurity sites are created by reaction between atmospheric CO2 and the hot glass during the rolling. Since atmospheric CO2 is expected to react with the outer surfaces of the hot glass, the majority of the C1 impurity sites are expected to be on or near the outer surfaces of the glass. Indeed, with similar W(CO)6 loadings and 312 nm excitation, the quantum yields of methane evolution from the unpolished samples are consistently higher than that from the polished samples where the outer surfaces of the glass have been removed by polishing. Photoactivated reduction of CO2 in the presence of a metal oxide and nonmetal oxides is usually interpreted within a band gap model, where optical excitation promotes an electron from the metal oxide’s valence band to its conduction band and that mobile electron is involved in the reduction of chemisorbed

significantly less than the eighth order dependence predicted by the Stark−Einstein Law if the metal oxide provides the eight electrons for the reduction of chemisorbed CO2. Furthermore, a hydrated precursor to monoclinic WO3 is formed in both polished and unpolished PVG. Assuming the reducing potential of an electron in the conduction band of this precursor is similar to that of bulk WO3, +0.31 ± 0.11 V (vs NHE),78 it does not have sufficient potential to carry out a one electron reduction of CO2 since its reduction potential is −1.90 V (vs NHE).79 In addition, 312 nm light corresponds to 389 kJ/ Einstein absorbed, while 437 nm light corresponds to 278 kJ/ Einstein absorbed, yet both drive reaction 1 where the ΔG° is 801 kJ/mol of CH4 formed.80 Collectively, the data imply that neither the hydrated precursor to monoclinic WO3 nor the mixed valence oxide acts as the source of electrons being used to reduce CO2. Rather, one-photon excitation of the hydrated precursor to monoclinic WO3 with 312 nm light or population of the IVCT state of the mixed valence oxide must create a situation in which reaction 1 occurs under much lower energy conditions.

■

DISCUSSION Elemental analyses and IR spectra confirm the compositional equivalence and the similarity of the hydroxylated surfaces of polished and unpolished PVG.22−29,54 The similarity of the electronic and DRIFT spectra with fluid solution spectra further establish that W(CO)6 physisorbs onto both forms of PVG, while the equivalence of the initial spectral changes and the quantum efficiency of CO loss imply the physisorbed complex exhibits the same initial photochemical reactions in both forms of PVG. Consequently, the different oxides formed in polished and unpolished PVG arise during the secondary thermal oxidation chemistry of the subcarbonyl photoproduct.23,27,41,54,56 12273



Article

WO263 indicate that mixed valence tungsten oxides containing W6+ and lower oxidation states of tungsten are also formed along with WO3 in unpolished PVG. Apparently, the rate of aggregation exceeds the rate of oxidation in unpolished PVG and a larger fraction of the tungsten is converted to the mixed valence oxide in the unpolished glass. Both polished and unpolished PVG contain WO3 and evolve CH4 during 312 nm photolysis. Although 312 nm excitation is sufficient to populate the conduction band of WO3, the less than eighth order dependence predicted by the Stark−Einstein law76,77 for methane yield on the absorbed energy (Figure 3) is inconsistent with the WO3 precursor acting as a one-photon, one electron source of the eight electrons needed to reduce CO2. Furthermore, the reduction potential of an electron in the CB of WO3, +0.31 ± 0.11 V (vs. NHE),78 does not have sufficient potential to reduce CO2, E° = −1.90 V (vs NHE).79 As Fujita points out, the latter value makes the one electron reduction highly unfavorable and difficult to achieve with currently available photoreductants.84 Rather, the first order dependence on absorbed energy (Figure 3) and the ability to drive the conversion via population of the lower energy IVCT states (Figure 1) of the mixed valence oxide point to a different mechanism by which chemisorbed CO2 is converted to CH4 within this silica matrix. The reduction of CO2 becomes more favorable with increasing acidity and as the hydrogen content of the product increases.85−87 The standard potential for the reduction of CO2 to CH4, for example, is exergonic under standard conditions

Previous studies of Fe(CO)5 physisorbed into PVG show that the thermal chemistry of the decarbonylated photoproducts is determined by a competition between oxidation and aggregation.41,61 If the rate of oxidation exceeds the rate of aggregation, the oxide aggregate contains the fully oxidized metal, which, with tungsten, is W6+ or WO3.41 However, if the rate of aggregation exceeds the rate of oxidation, the oxide aggregate contains metal atoms in different oxidation states or a mixed valence tungsten oxide.41 W(CO)6 is not as volatile as Fe(CO)5, yet both complexes physisorb onto PVG, exhibit quantum efficiencies of CO loss close to unity, and show similar changes in metal distribution during photolysis.41,55,56,61 If the rate of CO diffusion from the interior of this amorphous, nanoporous silica matrix is less than the rate of photoinduced CO release, CO pressure gradients develop within the matrix61 and these pressure gradients influence the subsequent thermal chemistry of the zero valent tungsten photoproducts in two ways.41 First, CO pressure gradients redistribute the metal within PVG. Since W(CO)5 forms a stable surface bound adduct, (CO)5W−OHSi81 and tungsten oxides exhibit little volatility at room temperature, the redistribution of tungsten during photolysis is attributed to CO pressure gradients sweeping physisorbed W(CO)6 from the interior into the outer 800−850 nm of PVG where it undergoes photolysis. As a result, photolysis of W(CO)6ads actually occurs at surface coverages higher than that corresponding to the initial loading and distribution of W(CO)6ads within the silica matrix. This increases the probability of aggregation, and in the case of Fe(CO)5ads, is thought to be the reason for the photoinduced formation of polymetallic iron carbonyls and mixed valence iron oxides in PVG and silica gel.41,61,82,83 Second, these photoinduced CO pressure gradients also affect the rate of oxidation of the subcarbonyl photoproducts. In the absence of air, H2 evolution points to coadsorbed water as the principal oxidant of the zero valent photoproducts in PVG.22,23 If the CO pressure gradients sweep water from the matrix, the rate of oxidation of the zero valent photoproducts declines.41,61 Both the redistribution of tungsten during photolysis and sweeping adsorbed water from the matrix favor the formation of mixed valence oxides. Furthermore, comparisons of the oxides formed by photolysis of Fe(CO)5 in PVG and the structurally similar tetramethylorthosilicate/methanol/water (TMOS/MeOH/ H2O) xerogels indicate the secondary thermal chemistry of the metal carbonyl photoproducts in these nanoporous silicas reflect structural differences on a length scale of the correlation length of the silica matrix.41 Consequently, the difference in the tungsten oxides formed in polished and unpolished PVG is attributed to similar structural differences except in this case the structural differences arise from the polishing of the PVG samples. In polished samples, the strong absorption onsets in the range of the band gap of WO3 and the lack of sensitivity to air and O2 implies the rate of oxidation by coadsorbed water exceeds the rate of aggregation and the dominant oxide formed contains W6+. TEM analyses and the appearance of weak, broad XRD reflections after heating suggest this oxide is a small, ≤1 nm diameter, amorphous, hydrated precursor to monoclinic WO3. A similar strong absorption in the UV and insensitivity to O2 indicate that this hydrated precursor to WO3 is also formed in unpolished PVG. In addition, however, the appearance of the visible absorptions (Figure 1), their sensitivity to O2, and their similarity to the visible absorptions of the photochromic forms of tungsten oxide52 and those of WO2.83, WO2.92, WO2.72, and

CO2 + 8H+ + 8e− → CH4 + 2H 2O

E° = 0.169 V (4)

and becomes more favorable with increasing acidity.85 Assuming unit concentration on the PVG surface, the Nernst equation indicates reaction 4 is slightly endergonic, E = −0.07 V, at the pH of the PVG surface, 4−5,66 but becomes exergonic if the pH of the surface is

Photocatalyzed Conversion of CO2 to CH4: An Excited-State Acid

Recommend Documents