■ INTRODUCTION

The photocatalytic conversion of CO 2 to reduced carbon states with visible light is a promising direction in renewable energy production. [1][2][3][4][5][6][7][8] The photocatalytic CO 2 reduction reaction (PCO 2 RR) is often driven with a photosensitizer (PS) and a transition-metal-based CO 2 reduction reaction (CO 2 RR) catalyst. 9,10 Molecular catalysts help to easily make precise structural modifications with atomistic-level control, which enables the rapid iterative design of CO 2 RR catalysts through understanding the structure-function relationship. Au-based N-heterocyclic carbene (NHC)-ligated molecular catalysts are underexplored in the PCO 2 RR despite these complexes being of interest for a large variety of transformations such as C-H activation, alkyne hydration, nitrile hydration, and alkene activation among others. [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27] In general, NHC ligands are strong sigma donors with variable stereochemical and electronic features, capable of covalently bonding to a gold center to give, in many cases, more reactive catalysts with enhanced durability which are needed characteristics in the PCO 2 RR. 13,28 In general, molecular Au catalysts in the PCO 2 RR are underexplored, and our group reported the first examples. 29 This is in sharp contrast to the large number of studies on Au heterogeneous catalysts and Au nanoparticles in the CO 2 reduction literature. [30][31][32][33][34][35][36][37] Among the five molecular catalysts originally reported, one NHC-ligated Au complex was evaluated, showing relatively moderate performance in the PCO 2 RR with promising electrochemical reactivity observed under a CO 2 atmosphere. In this work, we seek to understand the effects of NHC ligand tuning both sterically and electronically on the PCO 2 RR with Au complexes (Figure 1). Additionally, varying the ligand which dissociates to open a reactive site has not been studied with Au-NHC complexes, and we seek to understand the effects of this ligand on catalyst activity in the PCO 2 RR. This report studies the performance of eleven Au-NHC complexes that differ in the wingtip substituents on the NHC ligand, the saturated or unsaturated backbone in the NHC ligand, and the labile ligand (counterion) on the Au center (Figure 1).

Six of the complexes evaluated in the PCO 2 RR have an imidazole NHC ligand with 2,6-diisopropylphenyl (IPr) wingtips. These complexes are all numbered as 1, with superscripts denoting the labile ligands on each Au complex. Specifically, chloride (1 Cl ), hydroxide (1 OH ), bis-(trifluoromethane)sulfonamide (1 NTf2 ), tetrafluoroborate (1 BF4 ), carbazole (1 Cz ), and a hydroxide-bridged bimetallic complex (1 OH 1 BF4 ) were selected to evaluate the effects of a range of labile ligand binding strengths and counterions on the PCO 2 RR reaction. Interestingly, 1 OH and 1 OH 1 BF4 combined are known to trap CO 2 stoichiometrically. 38 Complexes 2 Cl , 3 Cl , and 4 Cl vary the wingtip groups as mesityl (Mes), 2,6dibenzhydryl-4-methyl phenyl (IPr*), and adamantyl (Ad), respectively. These wingtip changes seek to probe the effects of varied steric and electronic characteristics on the Au catalysts during the PCO 2 RR. 39 4 Ph is also evaluated in the PCO 2 RR to compare with 4 Cl to analyze the difference in the dissociation of a chloride versus benzene (once protonated). Finally, 5 Cl is comparable to 2 Cl since both complexes have chloride ligands and Mes wingtips on the NHC ligand but vary at the imidazolylidene ring with 5 Cl being saturated along the ligand backbone.

In all experiments, the photosensitizer (PS) is Ir(ppy) 3 (where ppy is 2-phenylpyridine) since this complex is a potent reductant upon addition of an electron to give [Ir(ppy) 3 ] - (Figure 2). 40 Ir(ppy) 3 has a higher driving force for electron transfer than many PSs used in the PCO 2 RR, which allows for the probing of a wide range of catalysts under similar conditions. 1,3-Dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH) is used as a sacrificial electron donor (SED) in all of the PCO 2 RRs due to the strongly reducing nature of BIH relative to other commonly used SEDs such as alkylamines. 40 Notably, triethylamine is added to these reactions to deprotonate BIH •+ after electron transfer to photoexcited Ir(ppy) 3 to promote an irreversible electron transfer event (Figure 2). 5,41 All reactions evaluated here are photosensitized as a prior study demonstrated a PS is necessary for the product to be observed with an Au-NHC complex. 29 A proposed catalytic cycle is shown in Figure 2 with the roles of Ir(ppy) 3 , BIH, TEA, and the Au catalyst depicted. Addition of a Au catalyst to CO 2 upon opening a reactive site is proposed based on prior cyclic voltammetry studies on 1 Cl and common observations in the literature that NHC-ligated metals favor the addition of CO 2 after a single electron reduction of the metal complex leading to the opening of a reactive site. [42][43][44][45][46] The series of catalysts being probed in this study also allow analysis of how varied labile ligands promote entry into the catalytic cycle. Additionally, NHC ligand variations allow for probing of how the addition of Au complexes to CO 2 and the subsequent protonation, C-O bond cleavage, and CO dissociation steps are affected in the catalytic cycle.

■ RESULTS AND DISCUSSION Catalyst Synthesis. All complexes were synthesized via literature methods and their purity confirmed through elemental analysis. [47][48][49][50][51][52][53] See the Experimental Section below for specific complex references.

Photocatalytic CO 2 Reduction. Unless otherwise noted, all PCO 2 RRs were conducted with 1 μM Au complex, 100 μM Ir(ppy) 3 , 0.02 M BIH, and 5% v/v TEA in dry acetonitrile. The reaction mixtures were irradiated with a white LED (380-750 nm range) set to a power density of 63.6 mW/cm 2 . All of the BI and + BI has been previously discussed in the literature. 5,41 reaction mixtures were analyzed for CO, CH 4 , H 2 , and CO 2 H -, and only CO and trace H 2 (<1 TON) were observed, which indicates high selectivity toward a single carbon-based reduction product. As controls, removal of CO 2 , BIH, Au catalyst, or Ir photosensitizer from the reaction led to no CO production. The catalysts were first evaluated without any additives in the reaction mixture for turnover number (TON) values at 72 h (moles of CO/moles of Au complex) and for the turnover frequency per hour (TOF h -1 ) at 1 h (TOF 1 ), from the final 48 to 72 h time segment (TOF F ) and at the maximum value observed (TOF M ) (Table 1). TON versus time plots are provided to compare each of the catalysts (Figure 3).

Without an activator, the chloride complexes gave TOF 1 and TON values in the following order: 2 Cl > 5 Cl > 1 Cl > 4 Cl > 3 Cl . The two mesitylene wingtip-bearing complexes had the highest TOF and TON values among the series of chloride complexes with the unsaturated backbone catalyst having appreciably higher TOF 1 and TON (8.4 h -1 and 45 vs 6.0 h -1 and 37, respectively). The exceptionally bulky wingtip ligand of 3 Cl is correlated to the lowest TOF 1 and TON of the series. The two complexes (3 Cl and 4 Cl ) which were slow to dissolve in MeCN were also analyzed in N,N-dimethylformamide (DMF) as solvent (Figures S1, S2, andS8). In both cases, DMF gave higher TON values by > 5 turnovers, but the TOF values were significantly higher in the first 2 h with MeCN (Figure S2). Due to the relative ease of using distilled MeCN and the higher initial TOF values, MeCN was used as the comparison solvent for all of the catalysts examined in this study. The catalysts examined here are relatively robust to prolonged photocatalysis compared to systems based on benchmark rhenium NHC and bipyridine complexes that last only 1-2 h under analogous conditions. 43,54 To probe the reason why catalysis halts around 72 h for some catalyst such as 2 Cl , an experiment with 2 Cl was conducted where an additional portion of Ir(ppy) 3 , BIH, or CO 2 was added at 24 h to probe if any of these reagents were depleted in the reaction (Figure S13). Ir(ppy) 3 addition at 24 h was found to significantly increase the rate of catalysis during the 24 to 48 h time frame with a higher overall TON obtained. This indicates the photosensitizer decomposes under these conditions, leading to reduced turnovers.

Comparing varied labile ligands on gold complexes with the common ligand 1 shows the following trend of TOF ranging from 5.6 to 0.0 h -1 : 1 Cl > 1 MeCN BF4 > 1 NTf2 > 1 OH = 1 OH 1 BF4 > 1 Cz (Figure 3 and Table 1). Chloride-ligated complexes being faster CO 2 reduction catalysts than the presumably MeCN-ligated complexes has been previously attributed to a more easily dissociated anionic ligand upon reduction of the complex. 42 After 72 h, all catalysts retain some activity with TOF F values ranging from 0.1 to 0.5 h -1 . A chloride-labile ligand is compared to a phenyl group via complexes 4 Cl and 4 Ph . This comparison shows a higher TOF 1 for the 4 Cl catalyst (3.6 h -1 vs 0.5 h -1 ) but a higher TON value for the 4 Ph catalyst (35 vs 14 TON). Thus, while the phenyl group does not impart increased rates of reaction, this group does impart increased catalyst durability. Among all catalysts, only 1 Cz and 4 Ph do not have TOF 1 values approximately equal to TOF M values, which indicates a significant induction period is present until the active catalyst can be formed (Figures S3-S6). Interestingly, 1 OH and 1 OH 1 BF4 show general improvements in changes to TOF over time for the time segments after 24 h of irradiation (e.g., the 24-72 h time period relative to the preceding segment from 4 24 h).

The presence of an induction period with 1 Cz and 4 Ph prompted us to explore the role of opening a reactive site on the Au catalysts. Specifically, AgBF 4 was used with the chloride complexes to remove the chloride ligand via a metathesis reaction prior to initiating catalysis (Figure 4 and Table 1). 55 HBF 4 *Et 2 O was used as an acid activator to facilitate the protonation of the labile groups on catalysts 1 OH , 1 Cz , and 4 Ph to open a coordination site (Figure 4). 53,56 Despite the chloride complex of 1 being a faster CO 2 reduction complex than other chloride-free complex analogues of 1, chloride complexes 4 Cl and 5 Cl gave a dramatically positive change in TOF 1 rates (∼1.8×) upon AgBF 4 activation, which illustrates that the chloride-ligated complexes are not strictly faster catalysts across all ligand frameworks. This indicates that the complexes with alkyl-substituted wingtips or saturated NHC rings are relatively slow to be reduced or dissociate a chloride during the PCO 2 RR. However, no dramatic differences are noted in the cyclic voltammetry studies with respect to reduction potential, which suggests the rate of reactivity is not likely due to a sluggish reduction reaction (Figure S7). Thus, the rate of reactivity change may be due to the rate at which a coordination site opens. However, in several cases, the chloride-ligated complexes are faster catalysts such as when a series based on 1 (no activators), 2 with and without an activator, 4 Cl , and 4 Ph are compared. As an example, 2 Cl has appreciably higher TOF I and TON values without the addition of AgBF 4 (8.4 h -1 and 45 vs 3.3 h -1 and 20). Cyclic voltammetry conducted on 2 Cl before and after exposure to AgBF 4 shows only a moderate shift in the reduction potential onset (∼100 mV), which indicates that the initial electron transfer reaction to reduce the complex is not the likely reason for the nonobvious change in the reactivity observed (Figure S7). A possible explanation for the higher reactivity of 2 Cl in the absence of AgBF 4 could be due to the transition state which dissociates a Cl -being lower in energy than a transition state dissociating a MeCN group (present after metathesis) from a single electron reduced complex, as has been observed during electrocatalytic CO 2 reduction studies via density functional theory. 42 Acid activation was explored with respect to the PCO 2 RR with 1 OH , 1 Cz , and 4 Ph . In all cases, addition of HBF 4 *Et 2 O led to a faster TOF 1 with the most dramatic effects being observed with 1 Cz and 4 Ph . Additionally, the TOF 1 values and TOF M values were in close agreement after addition of an acid additive, indicating that the induction period found with 1 Cz and 4 Ph without HBF 4 *Et 2 O is no longer present. Higher TON values are observed for all three catalysts through 24 h; however, after 72 h, higher TON values are observed for 4 Ph and 1 OH without a HBF 4 *Et 2 O treatment. A plausible explanation for this observation could be due to catalystcatalyst deactivating interactions which are suspected in the PCO 2 RR due to lower concentrations of catalysts commonly being observed to give much higher TON values in many systems. [57][58][59][60] Without the presence of an acid activator, the active catalyst could be slow to form and could be slowly released over time, limiting the number of reactive catalyst species present in solution.

The highest TON and TOF 1 catalyst (2 Cl ) without an activator was studied via a series of experiments to probe for homogeneous versus heterogeneous catalyst behavior since Au nanoparticles/nanomolecules are well documented in CO 2 RRs as mentioned in the Introduction. The Maitlis test was conducted by initiating a typical PCO 2 RR for 4 h, followed by filtration of the reaction mixture over a bed of Celite under an inert atmosphere via an in-line filtration to remove any suspended nanoparticles. 61,62 The solution was then degassed with CO 2 and photolyzed for 68 more hours (72 h in total; Figure 5). A similar reaction profile is observed with both the unperturbed reaction and the filtered reaction with very similar TON values (56 vs 60 TON at 72 h, respectively). We stress that proving an active catalyst is molecular in nature is not possible; 62 however, this data is consistent with a homogeneous molecular catalyst in the PCO 2 RR.

Next, a Hg 0 test was conducted as a second homogeneity test. 62 Importantly, Au-Hg amalgams are well documented, and miscibility of the catalyst metal and Hg is required for the use of this test. 63 A PCO 2 RR was initiated with 2 Cl for 4 h to generate an active catalyst and then Hg 0 was added to the reaction mixture with the mixture photolyzed for 68 more hours (Figure 5). If the active catalyst was an Au nanoparticle, it is expected that the Au nanoparticle would be dissolved in the Hg 0 layer upon formation and would result in substantially less CO formation. Again, a similar TON value (45 vs 42 TON at 72 h) and a similar reaction curve profile are observed with and without Hg 0 present. While this experiment does not confirm a molecular-active catalyst, the data is consistent with the molecular species.

Finally, dynamic light scattering (DLS) studies were done on the PCO 2 RRs after 72 h of irradiation. Only two of the lowest TON value catalysts (4 Cl and 1 OH 1 BF4 ) show a moderate DLS signal between 2 nm and 800 nm sizes (Figure 5). A small signal is observed near 1 nm for 1 NTf2 and 1 OH 1 BF4 as well, although we note this is near the detection limit of the instrument, and assuming a spherical nanoparticle would only correspond to approximately 8 Au atoms if the entire particle was Au atoms face-centered-cubic-packed with a unit cell volume of 0.0679 nm 3 . 64 Notably, the Au atom count of a particle would likely be smaller than this since DLS actually measures the hydrodynamic radius. 64 Combined, these three tests for heterogeneity suggest the catalysts may be functioning as molecular systems since the highest performing catalysts do not show evidence of forming heterogeneous species across three common heterogeneity tests. Lastly, the rate of catalysis was found to increase as the concentration of the 2 Cl decreased (Figure 6). A TOF of up to 28.2 h -1 was reached at 0.1 μM of 2 Cl with a TON of 270 obtained at this concentration. The increase in TOF suggests that the rate of catalysis is limited by the amount of reductant [Ir(ppy) 3 -] in solution and that the catalysis being observed is homogeneous since a lower concentration would be expected to form nanoparticles more slowly than a higher concentration.

■ CONCLUSIONS

Eleven gold complexes bearing N-heterocyclic carbene ligands were tested for CO 2 reduction reactivity in a photosensitized system. The complexes were selected with an array of wingtip groups on the carbene ligand and with variations in labile ligands on the gold atoms. In general, the complexes were found to be durable in the photocatalytic CO 2 reduction reaction with catalysis continuing in many cases for ∼50 h with CO generation as the only product at >1 TON. The Au complexes with chloride atoms as the labile ligand were found to undergo rapid catalysis without an induction period, which indicates the chloride atom is rapidly dissociated upon reduction of the complex to give an active catalyst species. More tightly bound ligands such as phenyl and carbazole groups were slow to access the active catalyst without the use of acid additives. Additionally, the most active complex in the photocatalytic CO 2 reduction reaction was analyzed via the Maitlis test, the Hg 0 test, and DLS, which showed no evidence of heterogeneous behavior. The tunability of these catalysts with respect to TOFs and TON numbers via ligand substitutions is attractive toward future studies. Future efforts will focus on increasing the rate of photocatalysis while retaining the high durability.

■ EXPERIMENTAL SECTION General Information. MeCN was freshly distilled over CaH 2 and kept under N 2 before use. DMF was distilled by vacuum distillation with heating to not more than 100 °C with no drying agent present. The presence of drying agents (CaH 2 or 4 Å MS) accelerates decomposition. DMF was distilled in 500 mL portions with the first 100 mL and the last 100 mL being discarded. The remaining 300 mL was kept under N 2 in a freezer and was not used for a period of more than 2 weeks before being discarded. Ir(ppy) 3 was purchased from Strem and used as received. Complexes 1 Cl , 47 2 Cl , 47 3 Cl , 48 4 Cl , 47 5 Cl , 47 1 NTf2 , 49 1 OH , 49,50 1 OH 1 BF4 , 51 1 Cz , 52 and 4 Ph53 used herein were all prepared according to previously reported methods. BIH was prepared as previously described and recrystallized from hot EtOH/ H 2 O (2:1) prior to use. 43 TEA was freshly distilled over CaH 2 and kept under N 2 prior to use. Celite was obtained from Celite Corp. (545, not acid-washed) and used as received.

Activation with AgBF 4 . The literature procedure was generally followed with some modifications as listed below. 55 AgBF 4 (1.25 equiv.) was added to a stirred solution of [Au(NHC)Cl] (1.0 equiv.) in dichloromethane (∼0.03 M). The reaction mixture was stirred, avoiding the presence of light, at room temperature for 5 min and then filtered over Celite. The solvent was evaporated by passing N 2 over the solution, and the complex was then used directly in catalysis experiments.

Activation with HBF 4 *Et 2 O. The literature procedure was generally followed with some modifications as listed below. 53,56 HBF 4 *Et 2 O (1.1 equiv.) was added to a MeCN (∼0.03 M) solution of 1 Cz or 4 Ph (1.0 equiv.), and the solution was stirred for 5 min at room temperature. The solvent was then removed under vacuum, diethyl ether was added, and the solution was filtered. The solid obtained was washed with Et 2 O and dried. The above procedure was followed for 1 OH except that the complex was first dissolved in toluene (0.09 M in catalyst), then acid was added, then MeCN (0.14 M in catalyst) was added, and the solution was stirred for 2 h. Pentane was added to precipitate 1 MeCN BF4 . The liquid was removed by cannula and 1 MeCN BF4 was then dried under N 2 .

PCO 2 RR Setup. Photocatalytic experiments were performed in a 17 mL Pyrex culture tube (no. 99447, threaded GPI 15-415, 16 × 150 mm) with a stir bar by adding BIH (0.005 g, 0.02 mmol, 1 × 10 -2 M), MeCN (6 mL), Ir(ppy) 3 (0.2 mL of a 1 × 10 -3 M MeCN stock solution in a MeCN solution; final reaction concentration after degassing: 1 × 10 -4 M), and catalyst (0.2 mL of a 1 × 10 -5 M MeCN stock solution; final reaction concentration after degassing: 1 × 10 -6 M). Each tube was sealed with a Teflon-coated rubber septum inside an open top screw cap with the parafilm sealed around the cap-to-tube joint. A white rubber septum (14/20 septum inverted) was placed over the cap, and the rubber septum was parafilm sealed to the Pyrex tube (see Figure S9). The mixture was then bubbled vigorously with CO 2 until the solution volume reached 1.9 mL (typically 15 min) and then 0.1 mL of N 2 degassed distilled triethylamine was injected into the mixture. A neutral white LED (Thorlabs, MNWHL4, color temperature: 4900 K) mounted with a collimation adaptor (Thorlabs, SM2F32-A, lens: Thorlabs, ACL50832U-A, with an antireflective coating over the range of 350-700 nm) was used as the light source for the photocatalytic experiments. The photocatalytic reaction vessel was set at a distance to adjust the power equal to 1 sun intensity across the white light emission range (380-750 nm), which was determined by using a power meter (Coherent PowerMax PM10 detector with a FieldMate 1098297 instrument). The sun intensity across the 380-750 nm range was calibrated using cut-off and cut-on filters with an AAA rated solar simulator with a reading of 180 mW on the power meter equaling 63.6 mW/cm 2 when accounting for the detector area.

PCO 2 RR Product Detection. Product analysis via gas chromatography (GC) was done after 1, 2, 4, 24, 48, and 72 h of photolyzing the reaction. Then, headspace analysis was conducted with a VICI gastight syringe with a stopcock and a custom Agilent 7890B GC instrument equipped with an Agilent Porapak Q 6ft, 1/8 in. O.D. column, a methanizer, an FID detector, and a TCD detector. The amount of CO and CH 4 formed was quantified by using the FID detector signal, while H 2 was quantified using the TCD detector signal. All GC response calibrations were done using standards purchased from BuyCalGas.com. GC samples were prepared by taking headspace samples (300 μL) using a VICI gas-tight syringe, compressing the sealed sample to 250 μL, then submerging the tip of the sealed gas tight syringe in diethyl ether, opening and closing the valve to adjust the pressure in the syringe to atmospheric pressure, and injecting the sample into the GC instrument (see Figure S10 for a representative GC trace). Formate analysis was conducted as previously described by 1 H NMR spectroscopy (see Figure S11 for a representative formate analysis 1 H NMR trace). 40 Formate analysis was only performed at the end of the photocatalytic reaction as it requires significant reaction volume. No formate or methane was observed in these experiments. Only trace H 2 was observed, equaling less than 1 TON. All data points were collected in at least duplicate via two separate reactions. Data points agreed within a maximum of ±10% with a typical agreement of < ±5% being observed. Control experiments were conducted. Removal of the Au catalyst, CO 2 , or BIH led to no CO formation. Removal of Ir(ppy) 3 led to small amounts (∼1 TON as the highest observed) of CO formation after prolong irradiation with no appreciable CO forming in the first 24 h (Figure S12).

Heterogeneity Test Procedures. Maitlis Test. 61,62 A PCO 2 RR reaction was setup as described above with GC monitoring. After 4 h, with GC evidence of typical reaction behavior indicating the presence of the active catalyst, the reaction was transferred via a cannula through a rubber septum (white rubber, 14/20 size) onto an in-line filter with a bed of Celite (∼50 mL) resting on a fine frit with a filter paper on top of the Celite to reduce disruption of the Celite bed. The in-line filter with Celite and filter paper in place was previously thoroughly flame-dried under vacuum and then kept under N 2 for the duration of the experiment. The in-line filter was connected to an N 2filled, flame-dried, two-neck flask with vacuum applied. Celite was rinsed with dry, distilled MeCN (5 mL). The reaction solution was then transferred via a cannula to the Pyrex reaction tube and degassed with CO 2 until the solution reached a volume of 2 mL. The solution was then photolyzed and monitored as described above. Notably, the concentration of triethylamine was likely diminished to some extent during the degassing phase, given the odor of the vapors coming off the exit needle.

Hg 0 Test. 62 A PCO 2 RR reaction was done as described above with GC monitoring. After 4 h, with GC evidence of typical reaction behavior indicating the presence of the active catalyst, Hg 0 (0.5 mL) was injected into the reaction vessel via a syringe. The solution was then photolyzed and monitored as described above.

DLS Analysis. DLS analysis was conducted with a Malvern Instruments device (model: ZEN3600, serial no: MAL 1037566) by collecting a sample (2 mL) after the reaction had been photolyzed for 72 h. The sample was analyzed from 1 to 800 nm in a semimicrodisposable cuvettes (dimensions: 12.5 × 12.5 × 45 mm).

■ ASSOCIATED CONTENT

* sı Supporting Information

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.inorgchem.2c03487.