Plasmonic photosynthesis of C1–C3 hydrocarbons from carbon dioxide assisted by an ionic liquid

Photochemical conversion of CO2 into fuels has promise as a strategy for storage of intermittent solar energy in the form of chemical bonds. However, higher-energy-value hydrocarbons are rarely produced by this strategy, because of kinetic challenges. Here we demonstrate a strategy for green-light-driven synthesis of C1–C3 hydrocarbons from CO2 and H2O. In this approach, plasmonic excitation of Au nanoparticles produces a charge-rich environment at the nanoparticle/solution interface conducive for CO2 activation, while an ionic liquid stabilizes charged intermediates formed at this interface, facilitating multi-step reduction and C–C coupling. Methane, ethylene, acetylene, propane, and propene are photosynthesized with a C2+ selectivity of ~50% under the most optimal conditions. Hydrocarbon turnover exhibits a volcano relationship as a function of the ionic liquid concentration, the kinetic analysis of which coupled with density functional theory simulations provides mechanistic insights into the synergy between plasmonic excitation and the ionic liquid.


Supplementary Figures
Supplementary Fig. 1 | Time course of hydrocarbon (CH4, C2H4, C2H2, C3H6, and C3H8) production in CO2 reduction reactions photocatalyzed by Au NPs with (a) 0 mol%, (b) 1 mol%, (c) 3 mol%, (d) 5 mol%, (e) 7 mol%, (f) 10 mol%, (g) 25 mol%, (h) 50 mol%, and (i) 100 mol% EMIM-BF4 solutions. A CW 532 nm laser with an intensity of 1 W cm −2 was used as the light source. Each data point is the average of results from three identical trials, and the error bar represents the standard deviation of these measurements. Each dashed line is a linear fit to the plots of TON for each product vs. illumination time. The slope yields the average TOF of each product, which is noted with the corresponding product in each plot. These TOF values are plotted in Fig. 1c and Fig. 4. For allowing appropriate comparison, all the plots have the same y-axis scale as the plot in panel (d), which represents the EMIM-BF4 concentration at which the highest activity was measured.     Supplementary Fig. 2. Thus, the chromatograms only show a trace of CH4 as an impurity originating from CO2 gas. No significant generation of hydrocarbon products was observed under these photoreaction conditions. The TON vs. illumination time plot resulting from the analysis of these chromatograms is shown in Supplementary Fig. 1a.  Supplementary Fig. 1d. Extended retention time-range GC-TCD chromatogram obtained at the end of a 10 h long Auphotocatalyzed CO2 reduction reaction with 5 mol% EMIM-BF4 and CW irradiation of 532 nm light (1 W cm −2 ). Three identical trials were performed, the results from which are shown in the three panels above. The GC peaks were assigned to H2, O2, and N2 by comparison with the reference chromatogram in Supplementary Fig. 13. No peaks associated with CO or other gaseous products were observed in the extended chromatograms.   Fig. 21 | Experimentally measured mass spectra at retention times of (a) 4.6 min and (b) 8.2 min of the TIC (Fig. 2a). The TIC and mass spectra were obtained by GC-MS analysis of the headspace gas collected after 240 h of a Au NP-photocatalyzed 13 CO2 reduction reaction with 5 mol% EMIM-BF4 and CW irradiation of 532 nm light (1 W cm −2 ). For the mass spectrum shown in (a), fragments at m/z = 14 and 18 were suppressed to remove the mass peaks contributed by N2 and moisture, respectively. The fragment at m/z = 28 in (b) has relatively high abundance as compared to that of the reference fragmentation pattern (Fig.  2c) due to the contribution of N2 from the atmosphere (Fig. 2a). were assigned to 13 CH4 and 13 C2H2 by comparison with these reference TICs and by the characteristic mass fragmentation patterns of these isotopologues. From the peak position of 12 C2H4 in the TIC, we deduce that the TIC peak expected for 13 C2H4 is overshadowed by the broad, intense peak of 13 CO2 in the TIC (Fig. 2a).   (6)).
3CO2 + 20H + + 20e − → C3H8 + 6H2O E˚ = 0.14 V (5) The E˚ values were determined from the standard Gibbs free energy of formation, ΔGf˚ of the molecules involved in each reaction 2 and are provided with reference to the standard hydrogen electrode (SHE) and at a pH of 0. The Fermi level, EF, of Au is at a potential of 0.66 V vs SHE at a pH of 0. The photon energy, hν, used for electronic excitation is 2.33 eV (λ = 532 nm). Electrons at the Fermi surface can be photoexcited to a potential as high as 0.66 V -2.33 V = -1.67 V, which is sufficiently more negative than the E˚ for the above-listed reactions. In other words, these electrons are sufficiently energetic for driving the abovelisted reduction half-reactions under standard conditions. Aside from the thermodynamic feasibility, the kinetics of hot electron transfer from the Au NP to CO2 will depend on the free energy barrier. In the presence of EMIM-BF4, the free energy barrier or overpotential is expected to be considerably lower, as per our estimates (Fig. 3d) and previous work on electrochemical CO2RR [3][4][5][6][7] .
Note that the accompanying photogenerated hole needs to be energetic enough for the oxidation of H2O. The photon energy of 2.33 eV is larger than the ~1.8 eV threshold for inter- H2O2 + 2H + + 2e -→ 2H2O E˚ = 1.83 V (7)

Synthesis and characterization of Au nanoparticles (NPs)
The synthesis of 12 nm Au NPs was carried out as described previously 9  band centered around 520 nm. The molar concentration, C, of Au NPs in the stock colloid was estimated using Beer-Lambert's law: where A is the absorbance at the peak wavelength of the LSPR band, ε is the extinction coefficient at this peak wavelength for ~12 nm diameter Au NPs (1.891 × 10 8 M −1 cm −1 ) 10 , and b is the path length of the solution in the UV-vis spectrophotometry cuvette (1 cm). The molar concentration, C, was used to determine the number of moles of Au NPs in a given volume of the stock colloid.
Prior to use of the IL, moisture and oxygen dissolved in the IL were removed. Dehydration and degassing were achieved by applying a vacuum (< 100 mTorr) to a flask containing the IL heated at 100 °C for 4 h and then refilling the flask with Ar (> 99.999%) gas. Three mL of the aqueous IL solution were transferred to the Pyrex reaction cell.
A few mL of Au colloid containing a net ~3.175 × 10 −11 mol of NPs (i.e., 1.912 × 10 13 NPs), was drop-coated on a 1.5 cm × 1.5 cm cotton cloth substrate (Texwipe TX 306) that was free of chemical additives. With the aim of cleaning off surface ligands and potential contaminants, the NP-coated substrate was washed with 60 °C DI water and then rinsed with room temperature DI water three times, followed by drying overnight in an oven at 70 °C. Control studies were also performed: one without CO2, one without Au NPs, and another without light. These studies were performed in a 5 mol% EMIM-BF4 solution, the IL concentration at which the activity was found to be the highest in this work. For the control study without CO2, the reaction mixture was instead bubbled with Ar gas for 30 min to 1 h, but all other conditions were kept the same as the photoreaction tests. For the control study without Au NPs, a bare substrate (with no Au NPs) was employed, whereas all other conditions, including the presence of a CO2 saturated solution, were kept the same as the photoreaction tests. For the control study without light excitation, all conditions were kept the same as the photoreaction tests, however the reaction mixture was maintained in the dark at reaction mixture in the photoreaction tests under 1 W cm −2 , 532 nm laser excitation.

Data analysis
Photoreaction activity was determined by analysis of chromatograms obtained by GC-FID measurement at specific time intervals in the photoreaction, as described in the past 9 . GC peaks detected in chromatograms were first assigned by comparison with reference chromatograms (see Supplementary Fig. 2 for hydrocarbon products and Supplementary Fig.   13 for non-hydrocarbon products). The chromatogram measured at each time interval was subtracted from the initial chromatogram to remove the GC peak contribution of the trace CH4 pre-present in the CO2 gas as an impurity. In control studies, where there is little to no product formation, the latter subtraction procedure can yield a negative GC peak for CH4 due to a systematic measurement error. In such cases, the amount of CH4 produced was regarded as zero.
From each chromatogram, areas of GC peaks of specific products were determined. The integrated GC peak area was converted into a molar amount of the product by taking into account the calibration constant determined from reference chromatograms for the product.
This molar amount of the product in the injection volume was multiplied by the ratio of the headspace volume to the injection volume to obtain the moles of product in the headspace. A correction was also performed for the cumulative loss of product from the headspace due to the prior injections. The cumulative molar amount of product generated up to a specific illumination time was divided by the moles of Au NPs in the reaction mixture to obtain a turnover number (TON), i.e., the number of product molecules generated per Au NP. For this TON estimate, we used the moles of Au NPs present in the colloid, which was drop-casted to form the substrate-supported photocatalyst film. However, it is plausible that a fraction of the Au NPs were lost during the deposition process. Secondly, not all the Au NPs deposited on the substrate were located within the irradiation zone of the laser beam. Therefore, the TON we estimated represents a lower limit and the actual TON is likely higher. The slope of a linear fit to the plot of the TON vs. illumination time yielded the turnover frequency (TOF).
Product selectivity was determined by dividing the TOF of a specific product generated in the photoreaction by the sum of TOFs of all products generated in the photoreaction.
H2 and O2 produced in the reaction were quantified from chromatograms measured using GC-TCD. The measured GC peak areas of H2, O2, and N2 were converted to molar amounts of these gases by multiplying by their respective calibration constants. In the case of O2, it was necessary to account for O2 that results from air contamination and does not originate from photoproduction. From the molar amount of N2 measured and the known molar ratio of O2/N2 in the atmosphere, we determined the molar amount of O2 that simply originated from air leftover in the reactor from the reaction mixture preparation or from any air leakage that may have occurred over the course of the photoreaction or during injection of the headspace gas into the GC. The molar amount of O2 determined to be from air (and not photoproduction) was subtracted out from the total molar amount of O2 determined from the GC-TCD Therefore, it was difficult to determine if O2 was generated as an oxidation product of the reaction.

Hydrogen peroxide (H 2 O 2 ) detection
H2O2 is another possible oxidation product as per the following sequence of reactions   Fig. 17) in which a 100 μL sample taken from the reaction mixture at t = 0 (prior to photoreaction) was subject to the fluorogenic test exactly as described for the post-reaction mixture. Three independent trials of this control test were performed, the results of which are shown in Supplementary Fig. 17. In these control tests, we measured a relatively weak fluorescence emission, which saturates in intensity after 10-

Isotope labeling studies
Photoreaction studies with the 13 CO2 gas (99 atom% 13 C, Aldrich) were conducted in 5 mol% EMIM-BF4 solution. All other conditions and procedures were the same as those for the photoreactions involving regular CO2 gas. Following 240 h of photoreaction, the gaseous products in the reactor headspace were subject to gas chromatography-mass spectrometry (GC-MS) analysis. Total ion chromatogram and mass spectra were acquired on a HP 5890 series II GC system equipped with a HP 5971A mass selective detector. The mass spectrometer was operated in electron-impact ionization mode. Mass peaks at m/z = 14 and 18 were suppressed to exclude the contribution from ambient N2 and moisture, but the mass peak at m/z = 28 was included for analysis of 13 C2H2 produced in the photoreaction. The fragmentation patterns of the products measured by mass spectrometry were compared with those deposited in the National Institute of Standards and Technology (NIST) Chemistry WebBook.

Computational study
Density functional theory (DFT) calculations were performed with the Gaussian 09 program.
All species studied were geometry-optimized using the B3LYP functional and the 6- where E is the total energy of the geometry-optimized species indicated in parentheses. Free energies, G, given by the sum of electronic and thermal free energies: were determined for the geometry-optimized species CO2 and [EMIM*-CO2] and their respective 1eadducts: CO2 •and [EMIM*-CO2] •-. The adducts were generated by assigning a global charge of -1 to the geometry-optimized species, which was followed by geometryoptimization in the charged state. Free energies were also determined for [EMIM-CO2] • , which is the geometry-optimized complex formed by the 1eadduct of CO2, CO2 •-, and EMIM + . The free energy of adsorption, ∆Gads, was determined for CO2, H2O, EMIM + , and BF4on Au from DFT-calculated free energies as: where G is the free energy of the geometry-optimized species indicated in parentheses. For this calculation, a two-layer Au slab consisting of 15 atoms was modeled using the facecentered cubic (fcc) structure (Fm-3m (2 2 5), a = 4.0786 Å) with the (111) surface exposed.
The (111) facets are experimentally determined to be the most prevalent surface facet on Au NPs. Geometries of the constructed Au slab, adsorbates by themselves, and adsorbate/Au slab complexes were optimized using the B3LYP functional, the LANL2DZ basis set for outer shell electrons of Au atom, the LANL2 effective core potential (ECP) for core electrons of Au atom, and the 6-311++G(d,p) basis set for C, H, O, and N atoms. In every geometry optimization procedure, the upper layer of the Au slab was frozen to maintain the surface facet, and the other lower layer was allowed to relax.