Exchange of CO2 with CO as Reactant Switches Selectivity in Photoreduction on Co−ZrO2 from C1–3 Paraffin to Small Olefins Authors: Tarik Loumissi, Rento Ishii, Keisuke Hara, Tomoki Oyumi, Ikki Abe, Chongxu Li, Hongwei Zhang, Rumiko Hirayama, Kaori Niki, Takaomi Itoi, Yasuo Izumi Published: 2024 ABSTRACT AbstractPhotocatalytic reduction of CO2 into C2,3 hydrocarbons completes a C‐neutral cycle. The reaction pathways of photocatalytic generation of C2,3 paraffin and C2H4 from CO2 are mostly unclear. Herein, a Co0−ZrO2 photocatalyst converted CO2 into C1–3 paraffin, while selectively converting CO into C2H4 and C3H6 (6.0±0.6 μmol h−1 gcat −1, 70 mol %) only under UV/Visible light. The photocatalytic cycle was conducted under 13CO and H2, with subsequent evacuation and flushing with CO. This iterative process led to an increase in the population of C2H4 and C3H6 up to 61–87 mol %, attributed to the accumulation of CH2 species at the interface between Co0 nanoparticles and the ZrO2 surface. CO2 adsorbed onto the O vacancies of the ZrO2 surface, with resulting COH species undergoing hydrogenation on the Co0 surface to yield C1–3 paraffin using either H2 or H2O (g, l) as the reductant. In contrast, CO adsorbed on the Co0 surface, converted to HCOH species, and then split into CH and OH species at the Co and O vacancy sites on ZrO2, respectively. This comprehensive study elucidates intricate photocatalytic pathways governing the transformation of CO2 into paraffin and CO to olefins. Introduction In contrast to the irreversible consumption of fossil fuels and raw materials, the conversion of CO2 reduction into fuels and/or valuable chemicals using a sustainable energy represents a pivotal step toward establishing a new carbon‐neutral cycle.[ 1 , 2 ] Photocatalytic CO2 reduction offers a direct and simple approach; however, the range of products has been limited to CO, CH4, and CH3OH,[ 1 , 2 , 3 ] unlike the electrochemical production of a broad spectrum including formate, [4] C2H6, C2H4,[ 4 , 5 ] CH3CHO, [4] C2H5OH,[ 4 , 5 , 6 ] C3H8, C3H6, C3H7OH,[ 4 , 6 ] acetate,[ 5 , 6 ] and oxalate [5] from CO2 and/or CO, facilitated by concentrated electrons supplied from electricity. The economically viable nature that formed C2 and C3 hydrocarbons (HCs) derived from photocatalytic CO2 reduction (Table S1) has emerged as key chemicals (0.9–8 $ kg−1) compared to CO and CH4 (0.06–0.18 $ kg−1). [3] Various catalysts have been explored for this purpose, including Co−Cu/TiO2 for C2H6 and C3H8; [7] single Au/red P, [8] CdS/Cu−nanotube, [9] Nafion−Pd−TiO2, [10] Pt−graphene/TiO2–x , [11] graphene−TiO2, [12] and Au@Bi12O17Br2 for C2H6; [13] CuPt2/TiO2 nanotube, [14] Au−Pd/TiO2 {1 0 1}, [15] and TiON1−(O vacancy) for C2H6 and C2H4; [16] C/Cu2O nanorod, [17] Cu/TiO2, [18] Ag−C nanotube@TiO2, [19] CuO/CuGaS2, [20] Bi2S3@In2S3, [21] FeCoS2, [22] In2.77S4/porous polymer, [23] N, S/Fe−MOF, [24] and Cuδ+/CeO2−TiO2 for C2H4; [25] forming the corresponding alcohol, aldehyde, and acid.[ 3 , 26 , 27 , 28 ] The number of reports on the photocatalytic synthesis of C2 and C3 HCs from CO2 has dramatically increased since 2019 (Table S1a, b, g, i, j, and n–s), yet a comprehensive understanding of the reaction pathways remains elusive, hindering precise control. This study reports the switchover of photocatalytic pathways from CO2 to C1–3 paraffin versus from CO to selective C2H4 and C3H6, using a Co0−ZrO2 catalyst (Supporting Information, 1.2. Major Framework of This Study). Results and Discussion Photocatalytic 13CO2 Reduction Using H2. The photocatalytic reduction tests of 13CO2 were first performed using ZrO2−823R, Co−ZrO2−823R (where 823 denotes the pretreatment temperature (K) of the photocatalysts with H2 and R represents reduced), and a Xe arc lamp guided through a quartz light conduit (142 mW cm−2; Table 1a–i). ZrO2 mostly reflected/scattered 94.5 %±0.3 % of light (300 nm5 nm) grown at 973 K transformed into a stable hcp phase. The HR‐TEM results were well consistent with the X‐ray diffraction (Figure S10A), which showed diffractions from monoclinic ZrO2. However, only very weak traces of the shoulder owing to hcp Co (0 1 0) and (0 0 2) were identified (Figure S10B) because of their small size of a few nanometers. Co K‐edge XANES spectra confirmed the speciation of Co: Co3O4, CoO, and Co0 metal for fresh Co (7.5 wt %)–ZrO2, Co (7.5 wt %)–ZrO2−723R, and −973R, respectively (Figure S11A and SI, 3.2. Characterizations), consistent with UV/Visible absorption spectroscopy (Figure S6A). For Co (7.5 wt %)–ZrO2−823R, the Co site comprised a mixture of CoO (15 %) and Co0 (85 %) (Figure S11B). The CoO amount based on rapid CO2 physisorption (extra 9.1 μmol of CO2 corresponding to CoO 36 %; Table S5a and b) was greater than 15 %, evaluated through XANES, probably because slower CO2 physisorption at the interface between Co and ZrO2 surface was also included. In the comparison between Co state and photocatalytic CO2 reduction, while the fresh Co (7.5 wt %)–ZrO2 and Co (7.5 wt %)−ZrO2−723R photocatalysts comprising Co3O4 and CoO, respectively (Figure S11A), exhibited poor activity for CO2 photocatalytic reduction (Tables 1a and b and Figure S2A), Co (7.5 wt %)–ZrO2−823R and −973R comprising metallic Co (Figure S11A) were the most and the second most active in the CO2 photocatalytic reduction (Scheme S1A and B, Table 1c and d, and Figures 1 and S2 C), respectively. Thus, metallic Co sites played an essential catalytic role combined with Vo⋅⋅ sites at ZrO2 surface[ 31 , 32 , 33 ] as dual active sites, similar to our previous reports for Ni−ZrO2 by experiments [32] and DFT calculations. [31] Three plausible reasons can be listed: (i) fcc Co0 surface was more conducive to dissociating COH species than hcp one, [30] (ii) CoO could work as a promoter in contact with Co0 (Figure 5A2) owing to visible light absorption and charge separation with an E g value of 2.9 eV (Figures S6A–c and S7B and Scheme 1C), and (iii) Co0 particle size may increase at 973 K. However, the major reason cannot be identified because these potential changes occurred simultaneously at 973 K, and the Co0 particle size mixed with CoO at 823 K was difficult to determine via EXAFS and HR‐TEM (Figure 5A). Photothermal Monitoring and the Control Thermal 13CO2/13CO Reduction. Related to the latter reaction steps of CO2 photoreduction over Co sites followed by CO2 activation at Vo⋅⋅ sites on ZrO2 surface, the local electronic and geometric structures associated with the thermal behavior of active metallic Co sites essential for CO2 photoreduction were monitored via Co K‐edge EXAFS spectroscopy. The analysis was conducted under UV/Visible light irradiation using CO2 (2.3 kPa), H2 (21.7 kPa), and Co (7.5 wt %)–ZrO2−973R. To precisely analyze Co0 sites not mixed with CoO, the second‐best catalyst was chosen (Table 1d and Scheme S1B). In the Fourier transform of EXAFS before UV/Visible light irradiation, the curve‐fit analysis revealed an intense peak observed at 0.21 nm (uncorrected for phase shift; Figure S12). This peak demonstrated a Co−Co interatomic distance (R) of 0.249 nm with an associated coordination number (N) of 9.8. Furthermore, the N and R values did not change considerably during the photocatalytic reaction (Figure S13A and B). The N value corresponds to a particle size of 2.9 nm, assuming a spherical fcc nanoparticle model with a surface dispersion (D) was 0.43. [40] The Debye–Waller factor (σ) was calculated using the correlated Debye model[ 41 , 42 ] for bulk and surface Co sites (vertical motion versus surface),[ 32 , 33 ] considering the Debye temperature for bulk θ D(Bulk) (445 K) [43] and surface θ D(Surface,⊥) for the vertical motion of freedom (211 K; [44] Figure S1). We approximated the mean temperature of Co nanoparticle (T nanoparticle) as the arithmetic mean value, considering θ D(Surface,⊥) weighted using 1/2 ⋅ 1/3 D for vertical translational motion at a free hemisphere surface and θ D(Bulk) weighted using (1−D)+(1/2)D+(1/2 ⋅ 2/3)D for the bulk site, nonfree hemisphere in contact with ZrO2 surface, and horizontal translational motion at a free hemisphere surface. (4) The σ value for samples (σ sample) was calculated using Eq. 5, taking the contribution of structural disorder (σ disorder) and the difference of the σ sample value from the Co metal foil (σ XDAP) into account: (5) At 296 K, σ sample and σ Co metal, correlated Debye values were 0.007 (28) and 0.007 (05) nm, respectively, for the nanoparticle model above based on the correlated Debye model (Figure S1) and Eq. 4. The σ XDAP value was given as −0.000 (32) nm. Thus, σ disorder was evaluated as 0.001 (82) nm using Eq. 5. Then, the temperature of the Co site was inversely evaluated based on the σ sample value obtained from Eq. 5 and the arithmetic mean value based on Eq. 4 (Figure 6). This specific approach directly monitored the local Co site temperature.[ 32 , 33 ] The Co sites were initially at 296 K before exposure to UV/Visible light. Upon irradiation, the temperature of the Co sites quickly increased to 362±21 K, gradually decreasing once the light was turned off after 138 min. The gradual drop in temperature post‐illumination cessation (Figure 6) is probably attributed to the presence of a minor CoO layer between the Co0 nanoparticle and the ZrO2 surface, which slows heat dissipation. Time course of the Debye–Waller factor σ derived from the Co K‐edge EXAFS analysis and the determined temperature of Co nanoparticles in Co (7.5 wt %)–ZrO2−973R using CO2 (2.3 kPa), H2 (21.7 kPa), and UV/Visible light irradiation (142 mW cm−2), based on the correlated Debye model. The error bars were drawn based on data reproducibility in three runs and the fit errors. Compared with the critical change of σ sample values triggered by UV/Visible light irradiation (Figure 6), the N(Co−Co) and R(Co−Co) values negligibly varied during UV/Visible light exposure and cessation, except for rapid initial quick changes upon the light activation (Figure S13). This initial change may result from the oxidation of minor CoII sites by the effect of CO2 (Figure S11B), which were subsequently reduced under H2 and UV/Visible light. The reduced state of CoII maintained a dynamic equilibrium under these conditions, with N values gradually decreasing upon light deactivation (Figure S13A). The CH4 formation rate reached 196 μmol h−1 gcat −1 (Table 1d) under UV/Visible light irradiation leading to a temperature rise of 11.9 K within 10 min thermodynamically. However, this increase was smaller than the initial observed increase of 66 K within 10 min (Figure 6; SI, 3.2. Characterizations). Thus, the temperature elevation in Figure 6 stemmed from the transformation of light energy into heat at the Co0 surface, quickly reaching the heat equilibrium and dissipating into the reactor/EXAFS cell. Consistent with this evaluation of warming of Co0 sites by light under CO2, the time course of Fourier transform (Figure S14), σ value, and Co0 site temperature under CO, H2, Co (7.5 wt %)–ZrO2−973R (Scheme S1B), and UV/Visible light irradiation (Figure S15) reached 350±8 K and behaved very similarly to Figure 6, irrelevant to reactants. To identify the origin of photocatalytic CO reduction in this study, control thermal reaction tests were performed using Co (7.5 wt %)–ZrO2−973R, 13CO, and H2 at a reaction temperature between 363 and 423 K (Table S6B). The total formation rate at 350 K was lower by a factor of 71 000 compared to that under UV/Visible irradiation (Scheme 1D and Table 1i), wherein the Co nanoparticles reached 350 K during the CO photoreduction test (Figure S15). This indicates that simple thermal catalysis did not occur solely owing to heat transformed from UV/Visible light energy. Similarly, the total formation rate at 362 K using Co (7.5 wt %)–ZrO2−823R as evaluated for Co (7.5 wt %)–ZrO2−973R (Figure 6), 13CO2, and H2 was lower by 200 times compared to that under UV/Visible irradiation (Table S6A and Scheme 1D). The C2,3 HC formation was especially suppressed in the thermal reaction, suggesting different reaction pathways under UV/Visible light irradiation (SI, 3.1. Kinetic Results). The higher apparent activation energy (E act) from CO (151 kJ mol−1) compared to that from CO2 (83.4 kJ mol−1) contradicts the possibility of higher E act from CO2 to CO that then follow common pathway. Instead, the fact suggests that CO followed a different thermal pathway from that from CO2, related to the difference of photocatalytic pathways between Scheme 1A and B. Reaction Mechanism. FTIR spectroscopy investigated the specific photocatalytic reaction mechanism, starting from the conversion of CO to C2H4 and C3H6 using the optimal Co (7.5 wt %)–ZrO2−973R photocatalyst (Scheme S1B). Analysis of the 13C/12C isotopic distribution of the photocatalytic products, generated from 13CO2 and 13CO as starting materials (Table S4b–i) and 13CO2 exchange reaction (Figure 4), revealed that a part of the 12CO2 initially adsorbed from the air onto Vo⋅⋅ sites at the ZrO2 surface (one per the area of 61 nm2) remained after pretreatment under H2. Subsequently, this 12CO2 was then incorporated into the reaction pathway toward HCs (Scheme 1B–a). Migration of 12CO2 from Vo⋅⋅ sites to the surface of Co0 nanoparticles and its reaction with the surface (Eq. 6) demonstrated by the presence of 12CO stretching vibration (ν 12CO) peaks observed at 2142 and 2126 cm−1 (Figure 7A‐a) under 13CO and H2: (6) FTIR monitoring of reaction intermediates using the Co (7.5 wt %)–ZrO2–973R photocatalyst (71 mg) in the (A) carbonyl and (B) C−H stretching vibration regions and the peak assignments. The reaction conditions were as follows: (a) under 13CO (2.3 kPa) and H2 (2.3 kPa) in the dark for 2 h, (b–d) under 13CO (2.3 kPa), H2 (2.3 kPa), and UV/Visible light irradiation at 0 h (b), 12 h (c), and 34 h (d), and (e, f) under vacuum and UV/Visible light irradiation at 0 h (e) and 2 h (f). (C) The time course of CH2 and CH3 stretching vibration peak intensity (including convolution of 13C and 12C peaks) by FTIR and (D) the corresponding time course of photocatalytic products monitored using GC−MS. A minor part of the CO formed on the CoII site in Eq. 6 further moved laterally (Figure 5A2) to the Co0 site (ν 12CO 2028 cm−1, Figure 7A–a). Because 13CO was not in direct equilibrium with 12CO2 at the Vo⋅⋅ sites, species derived from 13CO were not observed at 0 h under 13CO and H2 (Scheme 1B and Figure 7A–a and B–a). When the UV/Visible light irradiation started, leading to a gradual decrease in the ν 12CO peak intensity on CoII sites, the corresponding ν 12CO peak on the Co0 site observed at 2028 cm−1 progressively grew (Figure 7A–b and c). It was followed by a progressive increase in the isotopically‐labeled terminal and bridging ν 13CO peaks observed at 1978, 1921, 1874, and 1839 cm−1 (Figure 7A–c and d), [45] demonstrating that the 12CO species derived from 12CO2 adsorbed at Vo⋅⋅ site when the light was turned on (0 h) were gradually replaced by gas phase 13CO (0–12 h), with 13CO on Co0 sites becoming the major species at 34 h under light (Figure 7A–d and Scheme 1B–b). However, the isotopic ratio of the ν 12CO/ν 13CO peak intensity in FTIR (Figure 7A–b and c) and the ratios of 12C2H4/(12C2H4+13C2H4) and 12CH4/(12CH4+13CH4) (7.9 mol %, Tables 1i and S4i; Figure 7A–c and d) obtained from gas chromatography–mass spectrometry (GC−MS) were significantly greater than 12CO ratio in reagent (1 %). This discrepancy can be explained if the 13CO adsorbed on Co0 sites was in equilibrium with 12CO2 on Vo⋅⋅ site at ZrO2 surface (Scheme 1B–a; Eq. 6 and Figure 7A–a–d), and the CO adsorbed on Co0 sites should be the intermediate species toward HCs (Scheme 1B–b). Synchronized with the growth of the ν 12CO peak (Figure 7A‐b–d), the antisymmetric and symmetric vibration of 12CH2 (ν as 12CH2 and ν s 12CH2) peaks increased at 2936 and 2858 cm−1, respectively, accompanied by weak vibration peaks for 12CH3 (ν as 12CH3 2958 cm−1; ν s 12CH3 2876 cm−1) under 13CO, H2, and UV/Visible light (Figure 7B–c). Then, following the isotopic replacement by multiple ν 13CO peaks (Figure 7A–b–d), the CH2/CH3 vibrational pairs of ν as 12CH n and ν s 12CH n (n=2, 3) progressively shifted to the corresponding 13C‐isotopic pairs of ν as 13CH n and ν s 13CH n (2927 and 2850 cm−1 when n=2 and 2948 and 2869 cm−1 when n=3, respectively, Figure 7B–d). Although the stretching vibration peak intensity ratio of CH2/CH3 in FTIR was almost constant at 3 : 1 during the 34‐hour photocatalytic test (Figure 7C), the major product obtained using GC−MS transformed consecutively transformed from 13C2H4 to 13CH4 (Figure 7D). This indicates that the ratio of CH x to H adsorbed at a specific local site, e.g., the interface site between Co0 and the ZrO2 surface, determined the product selectivity (Scheme 1B–b–e). In contrast, the 12CO3 at the ZrO2 surface was replaced by 13CO3 species (1513 and 1308 cm−1), while bidentate 13C‐formate at the ZrO2 surface gradually increased (1523 and 1344 cm−1) [32] until 34 h of reaction under 13CO, H2, and UV/Visible light (Figure S16), both irrelevant to HC generation over the Co0 surface. Then, the Co (7.5 wt %)–ZrO2−973R photocatalyst subjected to vacuum and the UV/Visible light irradiation. In contrast to the quick disappearance of all the weakly adsorbed CO peaks (Figure 7A–e and f), approximately half of the reaction intermediate CH2 and CH3 species remained at 2 h under vacuum (Figure 7B–e, f and C), consistent with the HC preference using this photocatalyst. Spin‐polarized periodic DFT+U calculations were conducted using the Vienna Ab initio Simulation Package code version 6.2.1 [46] (SI, 2. Experimental Section). CO2 was adsorbed at the Vo⋅⋅ site (Scheme 1A–a and Figure S8A and B). [31] When H was provided, the adsorbed CO2 transformed to OCOH with an E act of 1.3 eV (Scheme 2A–b). If the Vo⋅⋅ site neighbored a Co0 site, the terminal O atom of the OCOH group occupied the Vo⋅⋅ site (species b), and the COH moiety hopped onto the metallic Co0 atom with an E act of 0.28 eV (species c). A very similar reaction mechanism was also proposed for Pd−ZrO2. [47] The population of Vo⋅⋅ sites adjacent to the Co0 nanoparticle was anticipated to be relatively higher compared to the mean population of one Vo⋅⋅ per ZrO2 surface area of 61 nm2 (Table S5b). This is because the Fermi level of Co (work function 5.0 eV) [48] is lower than the energy level of Vo⋅⋅ in ZrO2, [31] enabling electron acceptance, [49] thereby facilitating the generation of Vo⋅⋅ sites neighboring to Co0 nanoparticles. Energy Diagram over Monoclinic ZrO2 (1 1 1) Surface Combined with Co19 Cluster Exposing hcp (0 1 0) Surface Calculated (A) under CO2 and H2 and (B) under CO and H2. Ethene Route Was Also Drawn in (A′) and (B) Compared to Methyl Route. Three‐Dimensional Illustration of Each Species Is Presented in Scheme S2. Throughout the pathway to CH4, the hydrogenation to OCOH required the highest E act of 1.3 eV, followed by the E act of 0.93 eV for CH3 hydrogenation, and 1.2 eV for OH hydrogenation by the elimination at the interface (Scheme 2A–a, b, d, and e, Methyl route), thereby supporting the preferential formation of CH4 from CO2 (Table 1b–e). The regeneration energy of the Vo⋅⋅ site can be minimized to 1.7 eV by replacing H2O with CO2 at the Vo⋅⋅ site. [31] C2H6 was formed via the Ethene route (Scheme 2A′) as a byproduct, preferably via C2H4 (E act of 0.70 eV) rather than the coupling of two CH3 (E act of 1.0 eV; Scheme 2A–d, Figure S4, and Eq. S1), different from the situation using Ni−ZrO2 photocatalyst, where CH4 was exclusively formed. [31] The dissociation of HOCH2 into hydroxy and CH2 at the interface was also suggested (Scheme 2A), in contrast to considerably stable HOCH and dissociation on the Ni0 surface. [31] The reaction route of CO photoreduction, especially to C2H4, was also investigated by DFT calculations. In contrast to CO2 adsorption at the Vo⋅⋅ site of the ZrO2 surface,[ 31 , 32 , 33 ] CO is adsorbed on the Co0 site (the adsorption energy (E ads) of 2.23 eV; Schemes 1B–b and 2B–b), followed by the formation of CHO and HCOH species at the interface of the ZrO2 surface and Co0 (Schemes 1B–c and 2B–c). It is energetically advantageous when the hydroxy group fills the neighboring Vo⋅⋅ site to form CH (Schemes 1B–d and 2B–d). In the following Methyl route, the barrier for CH3 hydrogenation to form CH4 is highest (0.75 eV). In comparison, the Ethene route via the coupling of CH2 was more favorable (E act of 0.70 eV, Scheme 2B–e and f), which aligns with the specific formation of 13C2H4 and 13C3H6 from CO within the first 0–4 h of the reaction (Table 1g–i). The isotopic shuffling of 13C/12C was also plausible between CO adsorbed on the Co0 surface and 12CO2 adsorbed at the Vo⋅⋅ sites from the air (Scheme 1B–a–c). The advantagenous CO adsorption in the first stage transitions to competitive adsorption of CO (E ads of 1.36 eV) and H (E ads of 1.32 eV; Scheme 2B–b and b′) in the second step of consecutive photocatalysis. The increased population of H on Co0 sites kinetically explains the transition to form C1–3 paraffin and C3H6 after 4 h of the reaction (Figures 2 and S3). In contrast, under CO2 and H2, H adsorption on Co0 is rarely hampered by CO2, and CH2 is easily transformed to CH3 (E act of 0.30 eV, Scheme 2A–d) throughout the photocatalytic tests (SI, 3.3. DFT Calculations). Selective C2,3 olefin formation proceeded during intial 4 h of reaction (Figures 2 and 3) while on later stage, CH4 was a major product with minor C2H6, C3H6, and C3H8 (Figure 2). This contrast was understandable if we assume the adsorbed CO/H balance on Co0 surface decreased as the time elapsed (Scheme 2B–b and b′). Compared to the two different stages of reaction starting from CO, COH species hopped from Vo⋅⋅ site to Co0 surface followed by down‐hill steps mostly to CH4 (Scheme 2A–c and d). The consecutive reaction steps on separated sites on ZrO2 and Co0 starting from CO2 proceeded efficiently (Scheme 2A) compared to competitive steps mostly on Co0 starting from CO (Scheme 2B and Figure 7). Photocatalytic CO2 reduction has been proposed to proceed via various ways, including the coupling of two CO molecules to form C2H4 at Vo⋅⋅ sites of TiON, [16] In–(S vacancy)–Bi, [21] and Cu δ+ sites, [25] the reaction of CO with CHOH to form OC−CHOH and then to C2H4 at CuI−CuII sites, [20] and the reaction of CO with COH to form C2H4 at In2.77S4 surfaces [23] (Table S1j, n, o, q, and s). However, herein, CO did not directly participate in forming the C−C bond. Instead, the coupling of two species to form C2H4 at the FeCoS2 surface, [22] the coupling of two CHOs to form CH−CHOH and then to C2H4 at the Fe−N surface, [24] and the coupling of CH3 with CH3 or C2H5 to form C2H6 and C3H8 at Cu2O [7] and Cu surfaces, [9] have been reported. Additionally, the coupling of CH3 at TiO2 surfaces[ 11 , 12 ] and Au surfaces [13] to form C2H6 has been reported via C−C formation, involving oxygenate and/or HC intermediates (Table S1a, c, e–g, p, and r). In these reports, the combination of reduced active sites with partially oxidized sites resembles the Co0 site and the neighboring Vo⋅⋅ site at the ZrO2 surface to form C2,3‐olefins from CO (Scheme 1B) and C1–3‐paraffin from CO2 (Scheme 1A). Photocatalytic CO2 Reduction Using H2O. Compared with the photosynthesis of 13C2H6 and 13C3H8 from 13CO2 and H2 (3.7–4.5 μmol h−1 gcat −1; Table 1c and d and Figure 8A) and the photosynthesis of 13C2H4 and 13C3H6 from 13CO and H2 (5.5–5.9 μmol h−1 gcat −1; Table 1g and i and Figure 8B), CO2 photoreduction was attempted using H2O as a one‐step sustainable reaction under reaction conditions similar to those reported in Table S1.[ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] Summary of major findings of C2,3 photosynthesis rates using (A, C) 13CO2, (B) 13CO, (D) CO2, (A, B) H2, (C, D) H2O, and Co (7.5 wt %)–ZrO2−823R (open symbol) or −973R (filled symbol) photocatalyst irradiated by UV/Visible light at each intensity. The key factors for paraffin synthesis were the metallic Co0 surface and the adsorbed H species on the Co0. Using the Co−ZrO2−823R photocatalyst (Scheme S1A) and swiftly evacuating gas‐phase H2, H2O (2.2 kPa) served as a reductant for CO2 (2.3 kPa) under UV/Visible light irradiation to form 13C2H6 and 13C3H8 at a rate of 0.21 μmol h−1 gcat −1 (Table 1f and Figures 8C and S17). Conversely, O2 was not formed above the detection limit because the formed O2 reacted with adsorbed H species on the Co surface to regenerate H2O at the solid/gas interface. This challenge was addressed using liquid H2O for the photocatalytic CO2 conversion to release O2 from the surface. C2H6 and C3H8 were photogenerated at rates of 2.0–2.7 μmol h−1 gcat −1 at a steady state for 48 h (Figures S18A and B), with up to 60 mol % of C2,3 HC formation using H2 (Table 1c, k, and l and Figure 8A and D), accompanied by O2 formation at the rates higher than 66 μmol h−1 gcat −1 using Co−ZrO2−823R and −973 R photocatalysts. The relatively high formation rates to C2,3 HCs compared to that to CH4 (31–60 mol %; Table 1k and l) were attributed to competitive adsorption of C‐species and H2O on Co0, rather than the 1.4–1.8 mol % using H2 preferably adsorbed on Co0 (Table 1c and d). However, sufficiently high light intensity (222 mW cm−1) was needed for CO2 photoconversion in H2O (Tables 1j, l and S8a–e; SI, 3.4. Photocatalytic Conversion of CO2 Using H2O). 13CO2 (2.3 kPa) photoreduction using D2O (2.2 kPa), H2 (21.7 kPa), and Co−ZrO2−823R formed 13C‐methane with a D ratio of 9.2 mol %, which agrees with a D ratio in the reactants (9.1 mol %; Figure S19 and Table S7b; SI, 3.5. Photocatalytic Conversion of 13 CO2 Using D2O). This suggests that D2O and H2 reached equilibrium more rapidly and were shuffled over Co0 more efficiently than the progressive hydrogenation steps to C2,3 paraffin common under either H2 or H2O (Scheme 1A–a–d). Such a photocatalytic pathway from CO2 to C2,3 paraffin using either H2 or H2O[ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 ] and the photoformation from CO to C3H6 have been rarely reported. This study paves the way to precisely explore further active photocatalysts to selectively produce C2,3‐HCs utilizing unsaturated/lower‐dimensional semiconductors to regenerate the Vo⋅⋅ sites with an E act of <2.6 eV (Scheme 2B). This can be achieved in combination with metal nanoparticles/single atoms using either H2 or H2O, surpassing the capabilities of recent photocatalysts (Table S1).[ 7 , 8 , 9 , 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] Conclusions The Co (7.5 wt %)–ZrO2−823R photocatalyst formed C1–3 paraffin at a total formation rate of 330±20 μmol h−1 gcat −1 using CO2 and H2. In contrast, Co (7.5 wt %)–ZrO2−973R formed C2H4 and C3H6 at a total formation rate of 6.0±0.6 μmol h−1 gcat −1 with an olefin selectivity of 70 mol % using CO and H2. CO2 was adsorbed on the Vo⋅⋅ sites at the ZrO2 surface, and the intermediate COH species hopped onto the Co0 surface, where it was progressively hydrogenated to form CH3 and/or CH2 species, which subsequently coupled to form C1–3 paraffin. Conversely, CO adsorbed on Co0 was hydrogenated to form HOCH and was most effectively dissociated to CH at the interface with the ZrO2 surface comprising the Vo⋅⋅ site. Preferential CO adsorption and favorable CH2 coupling until 4 h were followed by the competitive adsorption of CO and H on Co0 sites, resulting in consecutive CH4 and C3H6 formation. Predominant photocatalytic formation of C2H4 and C3H6 (61–87 mol %) was achieved from CO after repeated tests for 4–10 h, followed by evacuation. The dual mechanism involved electron donation from Vo⋅⋅ to OCOH/HOCH species using CO2/CO, followed by consecutive hydrogenation steps on the Co0 surface at 362/350 K, facilitated by UV/Visible light energy, at a rate of 200/71 000 times higher than that observed via a thermal reaction at 362/350 K, respectively, without light. CO2 photoreduction to C2H6 and C3H8 was possible in H2O (l) to release O2 from the surface via a very similar pathway involving CH3/2 activation on Co0. Conflict of Interests The authors declare no conflict of interest. Supporting information As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors. Supporting Information