IFP Energies nouvelles International Conference / Les Rencontres Scientifiques d ’ IFP Energies nouvelles PHOTO 4 E – Photocatalysis for energy PHOTO 4 E – Photocatalyse pour l ’ énergie

Photocatalytic conversion of CO2 into fuels is an attractive option in terms of both reducing the increased concentration of atmospheric CO2 as well as generating renewable hydrocarbon fuels. It is necessary to investigate good catalysts for CO2 conversion and to clarify the mechanism irradiated by natural light. Layered Double Hydroxides (LDH) have been attracting attention for CO2 photoreduction with the expectation of sorption capacity for CO2 in the layered space and tunable semiconductor properties as a result of the choice of metal cations. This study first clarifies the effects of Cu doping to LDH comprising Zn and Al or Ga. Cu could be incorporated in the cationic layers of LDH as divalent metal cations and/or interlayer anions as Cu(OH)4 2 . The formation rates of methanol and CO were optimized for [Zn1.5Cu1.5Ga(OH)8] + 2Cu(OH)4 2 mH2O at a total rate of 560 nmol h 1 gcat 1 irradiated by UV–visible light. Cu phthalocyanine tetrasulfonate hydrate (CuPcTs ) and silver were effective as promoters of LDH for CO2 photoreduction. Especially, the total formation rate using CuPcTs-[Zn3Ga(OH)8] + 2CO3 2 mH2O irradiated by visible light was 73% of that irradiated by UV–visible light. The promotion was based on HOMO– LUMO excitation of CuPcTs by visible light. The LUMO was distributed on N atoms of pyrrole rings bound to central Cu ions. The photogenerated electrons diffused to the Cu site would photoreduce CO2 progressively in a similar way to inlayer and interlayer Cu sites in the LDH in this study. Résumé — Conversion photocatalytique du dioxyde de carbone par des hydroxydes doubles lamellaires de Zn–Cu–Ga promus par la phtalocyanine de Cu : nécessité du contact entre le Cu et le réactif gazeux pour la synthèse du méthanol — La conversion photocatalytique de CO2 est une option attractive pour limiter la concentration du CO2 atmosphérique quand elle a pour objectif la production de produits hydrocarbonés utilisés comme carburants renouvelables. Néanmoins, des études sont encore nécessaires pour étudier les catalyseurs de conversion de CO2 et clarifier les mécanismes réactionnels. Des Hydroxydes Doubles Lamellaires (HDL) sont des catalyseurs intéressants pour la photoréduction de CO2 et l’on s’attend à obtenir une capacité d’adsorption du Oil & Gas Science and Technology – Rev. IFP Energies nouvelles, Vol. 70 (2015), No. 5, pp. 841-852 S. Kawamura et al., published by IFP Energies nouvelles, 2015 DOI: 10.2516/ogst/2015020 This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. CO2 dans l’espace interlamellaire et des propriétés semi-conductrices adaptables par le choix des cations métalliques. La présente étude démontre tout d’abord les effets du dopage du Cu dans les HDL comprenant du Zn et de l’Al ou du Ga. Le cuivre pourrait être incorporé dans les couches cationiques des HDL sous forme de cations métalliques divalents et/ou d’anions inter-feuillets sous forme de Cu(OH)4 2 . Les taux de formation du méthanol et du CO ont été optimum pour le [Zn1.5Cu1.5Ga(OH)8] + 2Cu(OH)4 2 mH2O correspondant à une vitesse totale de 560 nmol h 1 gcat , sous irradiation UV visible. Le tétrasulfonate-hydrate de phtalocyanine de Cu (CuPcTs ) et l’argent se sont avérés être des promoteurs efficaces de HDL pour la photoréduction du CO2. En particulier, le taux de formation total en utilisant du CuPcTs-[Zn3Ga(OH)8] + 2CO3 2 mH2O irradié par la lumière visible représentait 73% de celui irradié par de la lumière UV visible. La promotion a été attribuée à une excitation HOMO–LUMO du CuPcTs par la lumière visible. Le LUMO répartie sur l’atome d’azote des cycles des pyrroles lié à l’ion central Cu. Les électrons photogénérés diffusés sur le site de Cu photoréduiraient le CO2 sur les sites cationiques (dans les couches) ou anioniques (interfeuillets) des HDL.


INTRODUCTION
Photocatalytic conversion of CO 2 into fuels has emerged as an attractive option, in terms of both reducing the increased concentration of atmospheric CO 2 as well as generating renewable hydrocarbon fuels that can be directly supplied to our present energy infrastructure (Costentin et al., 2013;Corma and García, 2013;Genevese et al., 2013;Habisreutinger et al., 2013;Indrakanti et al., 2009;Izumi, 2013;Kubacka et al., 2012;Lewis and Nocera, 2006;Lv et al., 2012;Roy et al., 2010).The photocatalytic conversion of CO 2 involves two reaction steps: It is important to investigate good catalysts for CO 2 conversion into fuels using hydrogen as a reductant (Eq.2), which is potentially obtained from photocatalytic water splitting (Eq. 1) (Izumi, 2013;Ahmed et al., 2011).
In this study, catalysts for converting CO 2 into methanol using hydrogen and UV-visible light were investigated.Layered Double Hydroxides (LDH) [M II 1Àx M III x (OH) 2 ] x+ X 2À x/2 ÁmH 2 O (M II = Zn, Cu; M III = Al, Ga; X = CO 3 , Cu(OH) 4 ; m~1/2) were chosen with the expectation of: sorption capacity for CO 2 in the layered space, tunable semiconductor properties as a result of the choice of metal cations (Ahmed et al., 2013;Cavani et al., 1991;Fan et al., 2014;Li et al., 2014;Sideris et al., 2008;Zümreoglu-Karan and Ay, 2012).Furthermore, to utilize visible light as the major part of the solar spectrum, dyes/nanoparticles were mixed with the LDH.Especially, the effects of Cu phthalocyanine tetrasulfonate hydrate (CuPcTs 4À ) combined with LDH were studied.

EXPERIMENTAL SECTION
A LDH compound of [Zn 3 Ga(OH) 8 ] + 2 CO 3 2À ÁmH 2 O was synthesized using a reported procedure from metal nitrates, Na 2 CO 3 , and NaOH controlled at pH 8 (Ahmed et al., 2011).This compound is abbreviated as Zn 3 Ga|CO 3 .Similarly, LDH comprising various compositions of [M II 3 M III (OH) 8 ] + 2 X 2À ÁmH 2 O (X = CO 3 , Cu(OH) 4 ) were synthesized from nitrates of Zn, Cu, Al and Ga, and sodium carbonate/ammonium tetrachlorocuprate dihydrate at pH 8 (Ahmed et al., 2012).The formula is abbreviated as M II 3 M III |X.Na + 4 CuPcTs 4À was purchased from Aldrich (purity >85%) and used without further purification.For the ion exchange method, we followed the procedure presented in reference (Parida et al., 2007).Firstly, 0.50 g of Zn 3 Ga|CO 3 powder was immersed in an aqueous solution of Na + 4 CuPcTs 4À (0.60 mM, 25 mL) in a flask and magnetically stirred at 900 rotations per minute (rpm) for 24 h.The blue precipitates that were obtained were filtered using a polytetrafluoroethylene-based membrane filter (Omnipore JGWP04700, Millipore) with a pore size of 0.2 lm and washed well with deionized water (<0.06 lS cm À1 ; total 250 mL).The precipitates that were obtained were dried under vacuum at 290 K for 24 h.The blue sample was denoted as CuPcTs-Zn 3 Ga|CO 3 .The loading of Cu was 0.19 wt%.
Anion exchange of Zn 3 Ga|NO 3 with Na + 4 CuPcTs 4À was performed following the procedure in the literature (Abellán et al., 2012).All of the procedure until the drying under vacuum was conducted under an argon atmosphere.0.30 g of Zn 3 Ga|NO 3 powder was placed in a flask and an aqueous solution of Na + 4 CuPcTs 4À (5.0 mM, 200 mL) was slowly added and magnetically stirred at 900 rpm.Then, 30 mL of ethylene glycol and 30 mL of ethanol were added to the flask.The mixture was agitated by ultrasound (430 W, 38 kHz) for 10 min and then was magnetically stirred at 900 rpm for 3 days.
The precipitates that were obtained were filtered using an Omnipore JGWP04700 filter and washed well with ethanol (total 250 mL) and deionized water (total 250 mL).The precipitates that were obtained were dried at 290 K for 24 h under vacuum.The sample was denoted as Zn 3 Ga|CuPcTs.
Ag nitrate was impregnated from aqueous solution with Zn 3 Ga|CO 3 .The loading of Ag was 0.36 wt% (Kawamura et al., 2015).The sample was denoted as Ag-Zn 3 Ga|CO 3 .Separately, the assembly of Au nanoparticles with Zn 3 Ga| CO 3 was obtained using the structural reconstruction of LDH in the aqueous solutions containing Au 3+ (Carja et al., 2013).Hence, 1.20 g Zn 3 Ga|CO 3 powder was calcined in an oven at 773 K for 8 h.Freshly calcined Zn 3 Ga|CO 3 powder was directly added to 200 mL aqueous solution of Au(III) acetate (>99.9%,Alfa Aesar; 0.10 g) stirring at a rate of 900 rpm.The pH of the solution was adjusted to 8.0 by the addition of NaOH aqueous solution (0.10 M).The reaction mixture was stirred at the rate of 900 rpm and 290 K for 20 min and at the rate of 150 rpm and 313 K for 5 h.Then, the precipitate was centrifuged at the rate of 10 000 rpm and dried in an oven at 353 K for 1 h.The color of the catalyst obtained was light purple.The sample was denoted as Au-Zn 3 Ga|CO 3 .
Optical spectroscopic measurements were performed using a UV-visible spectrophotometer (model V-650, JASCO) using D 2 and halogen lamps for wavelengths between 200 and 340 nm and 340 and 800 nm, respectively.An integrating sphere (model ISV-469, JASCO) was used for the Diffuse Reflectance (DR) measurements.The samples were set in contact with the quartz window glass in gas-tight DR cells.Measurements were performed at 290 K within the wavelength range 200-800 nm using 70 mg of sample.DR spectra were converted to absorption spectra on the basis of the Kubelka-Munk equation (Ahmed et al., 2011(Ahmed et al., , 2012)).The band-gap (E g ) value was evaluated on the basis of either simple extrapolation of the absorption edge or the fit to the Davis-Mott equation: in which a, h, and m are the absorption coefficient, Planck's constant and the frequency of light, respectively, and n is 1/2, 3/2, 2 and 3 for allowed direct, forbidden direct, allowed indirect and forbidden indirect electronic transitions, respectively (Wooten, 1972).The Brunauer-Emmett-Teller surface area (S BET ) was calculated on the basis of eight-point measurements between 10 and 46 kPa in the N 2 adsorption isotherm at 77 K.The X-Ray Diffraction (XRD) pattern was observed using a D8 ADVANCE diffractometer (Bruker) at the Center for Analytical Instrumentation, Chiba University, at a Bragg angle of 2h B = 3-60°with a scan step of 0.01°and a scan rate of 5 s per step.The measurements were performed at 40 kV and 40 mA using Cu Ka emission and a nickel filter.
Cu K-edge X-ray Absorption Fine Structure (XAFS) spectra were measured at 290 K in transmission mode in the Photon Factory Advanced Ring at the High Energy Accelerator Research Organization (Tsukuba) on beamline NW10A.The storage ring energy was 6.5 GeV and the ring current was 46.6-36.1 mA.A Si (311) double-crystal monochromator and platinum-coated focusing cylindrical mirror were inserted into the X-ray beam path.The X-ray intensity was maintained at 65% of the maximum flux using a piezo translator set to the crystal.The slit opening size was 1 mm (vertical) 9 2 mm (horizontal) in front of the ionization chamber.Part of the XAFS measurements were performed for a sample in a reactor equipped with polyethylene naphthalate windows (Q51-16, Teijin) irradiated by a xenon arc lamp (Morikawa et al., 2014a;Izumi et al., 2007).The Cu K-edge absorption energy was calibrated to 8 980.3 eV for the spectrum of Cu metal foil (Bearden, 1967).
The XAFS data were analyzed using an XDAP (X-ray absorption fine structure Data Analysis Program) package (Vaarkamp et al., 2006).The pre-edge background was approximated by a modified Victoreen function C 2 /E 2 + C 1 /E + C 0 .The background of the post-edge oscillation was approximated by a smoothing spline function and calculated by an equation for the number of data points, where k is the wavenumber of photoelectrons: Multiple-shell curve-fit analyses were performed for the Fourier-filtered k 3 -weighted Extended X-ray Absorption Fine Structure (EXAFS) data in k-and R-space using empirical amplitude extracted from the EXAFS data for Na + 4 CuPcTs 4À (Izumi et al., 2005(Izumi et al., , 2009)).The interatomic distance (R) and its associated coordination number (N) for the Cu-N pair were set to 0.1950 nm with a N value of 4 (Carrera et al., 2004).The many-body reduction factor S 0 2 was assumed to be equal for both the sample and the reference.
As-synthesized and preheated samples of the LDH CuPcTs-Zn 3 Ga|CO 3 , Ag-Zn 3 Ga|CO 3 and Au-Zn 3 Ga|CO 3 S. Kawamura  were tested for the photocatalytic conversion of CO 2 (Kawamura et al., 2015).The tests were conducted in a closed circulating system (171 mL) equipped with a photoreaction quartz cell that had a flat bottom (23.8 cm 2 ). 100 mg of the LDH catalyst was uniformly spread in the photoreaction cell and was evacuated by rotary and diffusion pumps (10 À6 Pa) at 290 K for 2 h until the desorbed gas was detected by an online Gas Chromatograph (GC).2.3 kPa of CO 2 (0.177 mmol) and 21.7 kPa of H 2 (1.67 mmol) were introduced to both intact and pretreated LDH photocatalysts and were allowed to circulate for 30 min in contact with the catalyst to attain sorption equilibrium before irradiation.
The photocatalyst was then irradiated with UV-visible light from the 500-W xenon arc lamp (Ushio, model UI-502Q) upward through the flat bottom of the quartz reactor for 5 h.The distance between the bottom of the reactor and the lamp house exit window was set to 20 mm.The light intensity was 42 mW cm À2 at the center of the sample cell and 28 mW cm À2 at the periphery of the bottom plate of the sample cell.The temperature was within the range 305-313 K at the catalyst position during the illumination for 5 h.As a comparison, the photocatalyst was irradiated with visible light using one of the following UV-cut filters between the light exit of UI-502Q and the photocatalyst.Photocatalytic CO 2 reduction tests were performed systematically using L42, Y52, O58, R62, R66 and W-R715 sharp cut filters (Hoya) between the irradiation light exit and photoreactor.The thickness was 2.5 mm except for W-R715 (2.0 mm).These filters pass the light of wavelengths greater than 420, 520, 580, 620, 660 and 715 nm, respectively.The transmittance of nonfiltered light was more than 88% for L42, Y52, O58, R62 and R66, and almost 100% for W-R715 (manufacturer's web information (1) ).Products and reactants were analyzed using packed columns of a 13X-S molecular sieve and PolyEthylene Glycol (PEG-6000) supported on Flusin P (GL Sciences) set in the online GC equipped with a thermal conductivity detector (Shimadzu, model GC-8A).
The molecular orbitals for CuPcTs 4À were calculated using Gaussian 09 (Gaussian, Inc., Wallingford, Connecticut, USA) employing a polarized basis set of 6-31G(d) and in density functional theory mode to calculate the electron correlation using the functional B3LYP (Rauf et al., 2012).

Characterization
The interlayer interval and S BET of the LDH synthesized were between 0.751-0.792nm and 35-70 m 2 g À1 , respectively, except for Zn 3 Ga|CuPcTs.The metal cations existed in nearly identical octahedral MO 6 coordination environments in cationic sheets based on the similarity of the Zn K, Cu K and Ga K-edge X-ray Absorption Near-Edge Structure (XANES) (Ahmed et al., 2011(Ahmed et al., , 2012)).
Besides the absorption edge for the LDH, an intense absorption peak appeared at 678 nm for CuPcTs-Zn 3 Ga| CO 3 (Fig. 2b-line, solid line).The UV-visible spectrum for CuPcTs-Zn 3 Ga|CO 3 basically resembled that for aqueous solution of Na + 4 CuPcTs 4À (Fig. 2a-line); however, the peak intensity ratio in the range 600-700 nm was different: 678 nm (s), 630 nm (sh) and 612 nm (sh) for CuPcTs-Zn 3 Ga|CO 3 versus 630 nm (s) and 660 nm (sh) for aqueous solution of Na + 4 CuPcTs 4À .As the peaks at 678-660 nm or 630 nm would be assigned to HOMO (a 1u )-LUMO (e g ) electronic transition (Q-band, Fig. 3), the HOMO and/or LUMO of CuPcTs 4À were significantly perturbed by the interaction with the Zn 3 Ga|CO 3 surface.
As compared with the Q-band, the peaks at 337-340 nm were assigned to transition from a 2u to LUMO (e g ) (Soret band).The peak at 220 nm was only observed for aqueous solution of Na + 4 CuPcTs 4À (Fig. 2a-line).This peak would be a transition from a slightly deeper level than a 2u to LUMO (e g ) or from a 2u to a slightly shallower level than LUMO (e g ) (Marom et al., 2008).
The UV-visible spectrum for CuPcTs-Zn 3 Ga|CO 3 used for photocatalytic tests under CO 2 + H 2 for 5 h was also measured (Fig. 2b-line, dotted line).The spectrum was essentially identical to that for the fresh sample (solid line), indicating the stability of CuPcTs 4À dispersed over Zn 3 Ga| CO 3 under the photocatalytic reaction conditions.
Ag-Zn 3 Ga|CO 3 and Au-Zn 3 Ga|CO 3 exhibited a major absorption peak at 411 and 555 nm, respectively (not shown).These peaks are ascribed to Surface Plasmonic Resonance (SPR) of Ag and Au metallic nanoparticles.The XANES spectra for Na + 4 CuPcTs 4À and CuPcTs-Zn 3 Ga|CO 3 are shown in Figure 4.The whole spectrum pattern changed negligibly by the dispersion of CuPcTs 4À over the Zn 3 Ga|CO 3 surface as compared with Na + 4 CuPcTs 4À crystallines, suggesting that the framework structure of CuPcTs 4À was retained upon the dispersion over the Zn 3 Ga|CO 3 surface.In contrast, a sharp shoulder peak at 8 988 eV became relatively weaker upon the dispersion over the LDH surface.This may be related to the perturbation of LUMO by the interaction with the LDH surface based on UV-visible spectrum change (Fig. 2).
The 1s-3d electronic transition peak appeared at 8 982 eV for both Na + 4 CuPcTs 4À and CuPcTs-Zn 3 Ga|CO 3 (Fig. 5-I).The transition is allowed for the Cu II state of the 3d 9 configuration, whereas the peak disappeared for the Cu I state of the 3d 10 configuration (Morikawa et al., 2014a).Utilizing this difference, the reduction of Cu II sites in Na + 4 CuPcTs 4À and CuPcTs-Zn 3 Ga|CO 3 was monitored irradiated by UVvisible light.
The Cu amount in Na + 4 CuPcTs 4À charged was 118 lmol.The decreasing rate of the 1s-3d peak irradiated by UVvisible light for 64 min was 18.0% h À1 or 21.2 lmol h À1 .In contrast, the Cu amount in CuPcTs-Zn 3 Ga|CO 3 charged was 13.5 lmol.The decreasing rate of the 1s-3d peak irradiated by UV-visible light for 180 min was 20.2% h À1 or 2.72 lmol h À1 (Fig. 5-II).The photoreduction to the Cu(I) state was slightly faster for CuPcTs-Zn 3 Ga|CO 3 than for Na + 4 CuPcTs 4À ; however, exact comparison of the diffusion rates of photogenerated electrons to the Cu(II) site (Morikawa et al., 2014a) was difficult in this case due to the difference in the shape and light absorbance of the samples: dense crystallines of Na + 4 CuPcTs 4À versus fine powder of CuPcTs-Zn 3 Ga|CO 3 .At least, as the UV-visible light absorbance was apparently higher for Na + 4 CuPcTs 4À , the possibility of faster electron accumulation for CuPcTs-Zn 3 Ga|CO 3 than Na + 4 CuPcTs 4À is high due to the injection of excited electrons in Zn 3 Ga|CO 3 irradiated by UV light into CuPcTs 4À at the surface.
The EXAFS v oscillation changed negligibly when Na + 4 CuPcTs 4À was dispersed over the Zn 3 Ga|CO 3 surface (Fig. 6A).This fact supported the retention of the CuPcTs 4À framework over that of the Zn 3 Ga|CO 3 suggested by XANES.The curve-fit results for the Cu-N interatomic pair provided a distance of 0.198 nm (fit error ±0.011 nm) with a N value of 3.3 (fit error ±1.3) (Fig. 6D, E).

Photocatalytic Tests under CO 2 + H 2
In the photocatalytic reaction tests under CO 2 + H 2 irradiated by UV-visible light, Zn 3 Ga|CO 3 produced CO and methanol (Tab.2A, entry a).Using Zn 1.5 Cu 1.5 Ga|CO 3 , the methanol formation rate increased by a factor of 3.3 (Tab.2A entry g).Zn 3 Al|CO 3 was more active than Zn 3 Ga|CO 3 , but the major product was CO (entry h).When Zn 1.5 Cu 1.5 Al|CO 3 was compared with Zn 3 Al|CO 3 , the methanol formation rate was promoted by a factor of 3.3 (Tab.2b entry i) similarly to the Cu substitution into Zn 3 Ga|CO 3 .The methanol formation rate using Zn 3 Ga|Cu(OH) 4 was enhanced by a factor of 5.9 compared with Zn 3 Ga|CO 3 (Tab.2Aentries a, j).The methanol selectivity was nearly the same as that obtained with Zn 1.5 Cu 1.5 Ga|CO 3 (71-68 mol%).Using Zn 1.5 Cu 1.5 Ga| Cu(OH) 4 , the methanol formation rate and selectivity were further improved to 0.49 lmol h À1 g cat À1 and 88 mol% (Tab.2A entry k).
The results in Table 2 were independently reported in this study and references (Ahmed et al., 2011(Ahmed et al., , 2012;;Kawamura et al., 2015).However, the rates and selectivity are comparable because common reaction apparatus connected to common online GC were used in these studies.The reproducibility of rates was checked for Zn 3 Ga|CO 3 in this study and references (Ahmed et al., 2011(Ahmed et al., , 2012;;Kawamura et al., 2015), and the variation of formation rates of methanol and CO was always within 5%.
Furthermore, we reported control reaction tests in darkness, in the absence of a photocatalyst, and in the absence of CO 2 .No products were found in these control tests except for water from the interlayer space of the LDH.When the reactant was switched from CO 2 to CO 2 + H 2 , methanol and CO began to evolve using Zn 1.5 Cu 1.5 Ga|CO 3 (Ahmed et al., 2011).When the Zn 1.5 Cu 1.5 Ga|CO 3 and Zn 1.5 Cu 1.5 Ga|Cu(OH) 4 photocatalysts were recycled four times (in total 20 h of reaction), the methanol and CO formation continued and the selectivity was kept at 68-57 mol% and 76-84 mol%, respectively (Ahmed et al., 2011(Ahmed et al., , 2013)).
The addition of CuPcTs 4À to Zn 3 Ga|CO 3 improved the total (methanol and CO) formation rates by a factor of 1.6 and the methanol selectivity increased to 48 mol% (Tab.2A entry b; Fig. 7a-line).The formation of CO and methanol continued for more than 5 h, demonstrating the stability of dispersed CuPcTs 4À over the LDH irradiated by UV -visible light under CO 2 + H 2 .In contrast, the performance of Zn 3 Ga|CuPcTs was lower than that of Zn 3 Ga|CO 3 (Tab.2A entry c and Fig. 7b-line).It should be noted that unsupported Na + 4 CuPcTs 4À generated CO at a rate of 48% of that using Zn 3 Ga|CO 3 (Tab.2A entry d).
In a control kinetic test using the physical mixture of 3.0 wt% of Na + 4 CuPcTs 4À and 97 wt% of Zn 3 Ga|CO 3 LDH prepared by mixing using a mortar and pestle for 30 min, the formation rates of CO and methanol were 41 and 28 lmol h À1 g cat À1 , respectively (Tab.2A entry b), suggesting the importance of close contact of CuPcTs 4À with the LDH surface for the CO 2 photoconversion using CuPcTs-Zn 3 Ga|CO 3 (Fig. 7a-line).
The addition of Ag to Zn 3 Ga|CO 3 similarly affected the addition of CuPcTs 4À .Total formation rates increased by a factor of 1.7 compared with Zn 3 Ga|CO 3 and methanol selectivity of 54 mol% (Tab.2A entry e).The addition of Au to Zn 3 Ga|CO 3 promoted the total formation rates by a factor of 1.8, but the methanol selectivity became only 13 mol% (Tab.2A entry f).
Next, photocatalytic reduction tests of CO 2 by H 2 were conducted irradiated by visible light (k > 420 nm).Although the Zn 3 Ga|CO 3 LDH showed poor photoactivity irradiated by visible light (Tab.2B entry a), the total formation rate using CuPcTs-Zn 3 Ga|CO 3 irradiated by visible light was 73% of that irradiated by UV-visible light (Tab.2B entry b).The methanol selectivity was maintained at 48-51 mol% (Fig. 7 c-line).Again, under the condition of visible light irradiation, the formation of CO and methanol continued for more than 5 h, demonstrating the stability of dispersed CuPcTs 4À over the LDH under CO 2 + H 2 .
The total formation rate using Ag-Zn 3 Ga|CO 3 irradiated by visible light was 56% of that irradiated by UV-visible light (Tab.2A entry c, Tab.2B entry d).The selectivity to methanol decreased from 54 mol% (irradiated by UVvisible light) to 29 mol% (irradiated by visible light).Thus, Ag was less effective than CuPcTs 4À as a promoter under the condition of visible light irradiation.In comparison with the addition of CuPcTs 4À and Ag, Au-Zn 3 Ga|CO 3 did not exhibit photocatalytic activity above the detection limit of GC irradiated under visible light (Tab.2B, entry e).
The promotion effect of CuPcTs 4À to Zn 3 Ga|CO 3 was further investigated by plotting the in-profile (action) spectrum of total formation rates (methanol and CO) based on the photocatalytic tests irradiated by the light of wavelengths progressively greater than 420, 520, 580, 620, 660 and 715 nm using sharp cut filters (Fig. 8).The in-profile spectrum and UV-visible absorption spectrum coincided well in the wavelength ranges 310 and 688 nm, demonstrating that the visible light absorption by CuPcTs 4À dispersed over Zn 3 Ga|CO 3 (Fig. 2b-line) led to electron excitation and then the reduction of CO 2 .

DISCUSSION
Zn 3 Ga|CO 3 exhibited photocatalytic reduction of CO 2 by H 2 and UV-visible light at a total formation rate (methanol and CO) of 130 nmol h À1 g cat À1 (Tab.2A, entry a).When the amount of photocatalyst varied between 25 and 100 mg, the formation rate (the unit: mol h À1 ) was essentially proportional to the amount (Ahmed et al., 2011;Yoshida et al., 2012).
Copper ions were doped as a part of M II cations in the cationic layer [M II 1Àx M III x (OH) 2 ] x+ of LDH and/or Cu(OH) 4

2À
anions between the cationic layers.Both inlayer and interlayer Cu sites were effective for photocatalytic reduction of CO 2 (Ahmed et al., 2011(Ahmed et al., , 2012)).As a result, the total formation rate using Zn 1.5 Cu 1.5 Ga|Cu(OH) 4 was 560 nmol h À1 g cat À1 (Tab.2A, entry k).Copper sites were primarily active for photocatalytic reduction of CO 2 by H 2 .The drawback of LDH comprising Cu was the limitation of the wavelength for excitation light.The estimated E g values for the LDH used in this study were 5.6-3.0 eV (Tab.1), indicating that only UV light was effective for CO 2 photoreduction using LDH comprising Cu.To overcome this drawback, CuPcTs 4À , Ag or Au was doped to Zn 3 Ga|CO 3 .CuPcTs 4À , Ag or Au in CuPcTs-Zn 3 Ga|CO 3 , Ag-Zn 3 Ga| CO 3 and Au-Zn 3 Ga|CO 3 photocatalysts were effective at boosting methanol and CO formation rates by a factor of 1.6-1.8compared with Zn 3 Ga|CO 3 , both irradiated by UV-visible light (Tab.2A).The interlayer distance of the LDH changed negligibly for these photocatalysts (Fig. 1), demonstrating CuPcTs 4À , Ag and Au were over the exterior surface of the LDH rather than the interlayer space (Kawamura et al., 2015).Under UV-visible light, band-gap excitation of the LDH occurred by UV light (Tab.1), while HOMO-LUMO excitation (2.2 eV, Fig. 9; Giraudeau et al., 1980;Shang et al., 2011) of CuPcTs 4À and SPR of Ag and Au nanoparticles occurred by visible light.
The photocatalytic effects of UV and visible light were clarified by the photocatalytic tests irradiated by visible light only (Tab.2B), and also the in-profile spectrum of CO 2 photoreduction (Fig. 8).The formation rates of methanol and CO irradiated by visible light were 73-56% of those irradiated by UV-visible light using CuPcTs-Zn 3 Ga|CO 3 and Ag-Zn 3 Ga|CO 3 , whereas Au-Zn 3 Ga|CO 3 did not exhibit photocatalytic activity above the detection limit of GC under the irradiation of visible light.Thus, CuPcTs 4À was the best promoter for the LDH irradiated by visible light.The stability of CuPcTs 4À during photocatalytic tests was demonstrated (Fig. 2, 7a-c-line).Because the E g value for Zn 3 Ga| CO 3 was 5.6 eV (Tab.1), the LDH did not participate in the catalysis under visible light but just dispersed CuPcTs 4À molecules on the external surface.
The energy diagram and proposed electron flow during CO 2 photoreduction using CuPcTs-Zn 3 Ga|CO 3 are depicted in Figure 9.As the HOMO-LUMO gap positions between Valence Bands (VB) and Conduction Bands (CB) of Zn 3 Ga|CO 3 , band-gap excited electrons at CB and resultant holes at VB transfer to LUMO and HOMO of CuPcTs 4À , respectively, irradiated by UV light.In fact, photogenerated electrons diffused to the central Cu(II) site of CuPcTs 4À at the rates of 2.7-21 lmol h À1 for Na + 4 CuPcTs 4À and CuPcTs-Zn 3 Ga|CO (Fig. 5).Na + 4 CuPcTs 4À was able to photocatalyze CO 2 to CO by H 2 (Tab.2A, entry d).When CuPcTs-Zn 3 Ga|CO 3 was irradiated by visible light only, the band-gap excitation (5.6 eV) of Zn 3 Ga|CO 3 did not take place and only HOMO-LUMO excitation (2.2 eV) of CuPcTs 4À was possible.The decrease in the formation rates of methanol and CO under visible light to 73% of those under UV-visible light (Tab.2) can be explained by the decrease in the electron and hole supply from Zn 3 Ga|CO 3 to CuPcTs 4À .
The LUMO of CuPcTs 4À does not exactly populate Cu sites but N atoms neighboring Cu sites (Fig. 3).Thus, the electrons excited to the CB of the LDH would transfer to LUMO (and/or an unoccupied level near LUMO) of CuPcTs 4À .The energy level diagram of the LDH and CuPcTs 4À supported this hypothesis (Fig. 9).
The HOMO dominantly distribute on C atoms of pyrrole rings, while the LUMO dominantly distribute on both C and N atoms of pyrrole rings for CuPcTs 4À (Rauf et al., 2012).The N atoms of pyrrole are bonded to central Cu 2+ ions and thus able to transfer the photo-excited electrons to Cu 2+ (Fig. 3).CO 2 would be progressively reduced by the reduced Cu + in a similar way to that by inlayer and interlayer Cu sites (Morikawa et al., 2014a;Ahmed et al., 2011Ahmed et al., , 2012)).On the other hand, H 2 would donate an electron to the position around HOMO (Fig. 3) to form H + .The proton combines the electron and CO 2 at the Cu site and finally transforms into CO and methanol.
The SPR effect of Ag nanoparticles was already discussed in reference (Kawamura et al., 2015).In contrast, SPR of Au was not effective for the photoreduction of CO 2 due to the greater work function of Au (5.31-5.47eV) as compared with that of Ag (4.52-4.74eV).Irradiated by UV-visible light, the Au surface played the role of an electron trap from Zn 3 Ga|CO 3 .

CONCLUSIONS
LDH comprising Zn and Al or Ga photoreduced CO 2 to CO and methanol by H 2 and the irradiation of UV-visible light.The energy diagram and proposed electron flows in CuPcTs-Zn 3 Ga|CO 3 during photocatalytic reduction of CO 2 .
When Cu ions were doped as cations in the cationic layer and/or as interlayer anions, the CO 2 photoreduction rates increased to as high as 560 nmol h À1 g cat À1 (Zn 1.5 Cu 1.5 Ga| Cu(OH) 4 photocatalyst).Due to the wide band-gap nature of these LDH, UV light was effective for CO 2 photoreduction.The doping of CuPcTs 4À and Ag to the Zn 3 Ga|CO 3 LDH boosted CO 2 photoreduction by a factor of 1.6-1.8 by H 2 and the irradiation of UV-visible light.CuPcTs 4À was especially effective doped to the LDH for CO 2 photoreduction irradiated by visible light only.The LUMO of CuPcTs 4À was distributed on N atoms of pyrrole rings bound to central Cu 2+ ions.The photo-excited electrons diffused to central Cu 2+ would progressively reduce CO 2 finally to CO and methanol in a similar way to the inlayer and interlayer Cu sites in the LDH in this study.
Figure 6 A) Cu K-edge EXAFS v oscillation and B) k 3 -weighted EXAFS v oscillation for Na + 4 CuPcTs 4À diluted by boron nitride (a) and CuPcTs-Zn 3 Ga|CO 3 (b), C) its associated Fourier transform, and D, E) best-fit results in k-space D) and R-space E) for CuPcTs-Zn 3 Ga|CO 3 .Solid line: magnitude and dotted line: imaginary part C, E).Red (thick) line: experimental and blue (thin) line: fit D, E).
Figure 9 et al. / Photocatalytic Conversion of Carbon Dioxide Using Zn-Cu-Ga Layered Double Hydroxides Assembled with Cu Phthalocyanine: Cu in Contact with Gaseous Reactant is Needed for Methanol GenerationPhotocatalytic Conversion of Carbon Dioxide Using Zn-Cu-Ga Layered Double Hydroxides Assembled with Cu Phthalocyanine: Cu in Contact with Gaseous Reactant is Needed for Methanol Generation

TABLE 2
Photocatalytic rates of CO 2 reduction with H 2 using LDH(a)