A Robust Pyridyl-NHC-Ligated Rhenium Photocatalyst for CO₂ Reduction in the Presence of Water and Oxygen

Abstract

Re(pyNHC-PhCF₃)(CO)₃Br is a highly active photocatalyst for CO₂ reduction. The PhCF₃ derivative was previously empirically shown to be a robust catalyst. Here, the role of the PhCF₃ group is probed computationally and the robust nature of this catalyst is analyzed with regard to the presence of water and oxygen introduced in controlled amounts during the photocatalytic reduction of CO₂ to CO with visible light. This complex was found to work well from 0–1% water concentration reproducibly; however, trace amounts of water were required for benchmark Re(bpy)(CO)₃Cl to give reproducible reactivity. When ambient air is added to the reaction mixture, the NHC complex was found to retain substantial performance (~50% of optimized reactivity) at up to 40% ambient atmosphere and 60% CO₂ while the Re(bpy)(CO)₃Cl complex was found to give a dramatically reduced CO₂ reduction reactivity upon introduction of ambient atmosphere. Through the use of time-correlated single photon counting studies and prior electrochemical results, we reasoned that this enhanced catalyst resilience is due to a mechanistic difference between the NHC- and bpy-based catalysts. These results highlight an important feature of this NHC-ligated catalyst: substantially enhanced stability toward common reaction contaminates.

1. Introduction

The photocatalytic transformation of CO₂ into a useful fuel is one of humanity’s paramount challenges [1,2,3,4]. Driving this process with widely abundant sunlight is highly desirable. The use of a catalyst to drive the multi-electron reduction of CO₂ is needed as the direct one-electron reduction is energetically inhibitive [5,6]. Thus, catalysts that can absorb sunlight and deliver multiple electrons simultaneously are in high demand for addressing two key challenges. Despite more than 30 years of exploration, very few catalysts meet these criteria with a single metal center, and nearly all of them are based on one of five frameworks: Fe-p-TMA, Ir(tpy)(ppy)X, Ir(tpy)(bpy)X, Re(bpy)(CO)₃X, or Re(pyNHC-R)(CO)₃X (Figure 1) [7,8,9,10,11,12]. Among these catalysts, the Re(pyNHC-R)(CO)₃X catalyst is unique in allowing the reduction of CO₂ to occur at the first reduction potential of the neutral catalyst (Figure 2) [10,13]. Typically, CO₂ reduction photocatalysts first undergo photoexcitation followed by electron transfer from a sacrificial electron donor (SED) to give the anionic catalyst (Figure 2a). An SED is used in place of a complete photoelectrochemical system to reduce complexity and allow for the study of a single catalyst. The anionic complex then receives a second electron ultimately from a SED. The catalyst can then reduce CO₂ to a lower-oxidation-state carbon-based product such as CO and regenerate the initial neutral catalyst. Alternatively, after the first reduction sequence, the catalyst can attack CO₂ before the second reduction event (Figure 2b). Catalysis occurring after the first reduction is a property that is commonly observed for NHC catalysts and rarely observed for pyridyl-based systems [10,14,15,16,17,18,19]. Reactions operating off the first reduction potential for activation of CO₂ can be beneficial in designing catalysts with lower overpotentials (the excess energy beyond what is needed to reduce CO₂) which will be important in complete photoelectrochemical systems. In these systems, the energetics of the active catalyst which attacks CO₂ are readily measured and can be rationally tuned [18].

In addition to a key mechanistic difference during photocatalytic CO₂ reduction reactions, NHC-ligated catalysts have long been heralded as robust, highly reactive systems in general [20,21,22,23]. Concerning photocatalytic CO₂ reduction with sunlight, only one catalyst framework has been put forward utilizing an NHC ligand [10]. Interestingly, when an electron-deficient aryl-CF₃ group was used at the NHC wingtip, the catalyst was found to be faster reacting and more durable than Lehn’s bipyridyl analogue under anhydrous conditions (Figure 2). In this study, we probe the role of the critical-to-reactivity CF₃ group computationally and compare the properties of the two catalysts through excited-state lifetime measurements. Additionally, we evaluate the resilience of each catalyst to common reaction contaminates such as water and ambient atmosphere (O₂).

2. Results and Discussion

2.1. Computational Analysis: Role of the Ph-CF₃ Substituent

Among known NHC-ligated photocatalysts, the phenyl-CF₃ NHC substituent was previously found to give the most reactive catalyst when compared with phenyl, p-hexyloxyphenyl, or methyl substituents on the NHC ring [10]. Here, we reasoned that this group allows better delocalization of the added electron on the singly reduced catalyst across the ligand framework since the lowest energy excitation is thought to be a metal-to-ligand charge transfer (MLCT) followed by addition of an electron [25,26,27]. To analyze the role of the phenyl-CF₃ substituent, the geometry of Re(pyNHC-PhCF₃)(CO)₃Br was first optimized with DFT at the M06-L/6-31++G(d,p) level. Analysis of the highest occupied molecular orbital (HOMO) distribution shows the orbital primarily situated on the Re-metal center, the CO ligands, and the Br ligand with no contribution from the pyNHC-PhCF₃ ligand (Figure 3). The lowest unoccupied molecular orbital (LUMO) distribution is primarily located on the pyridyl and NHC rings of the pyNHC-PhCF₃ ligand which confirms the presence of an MLCT event. Interestingly, the PhCF₃ group has a substantial dihedral angle (~50°) relative to the NHC ring. This angle is large enough to inhibit contribution of the PhCF₃ group to the LUMO. This explains the prior observation that the NHC group substituent has only a minor role in controlling catalyst reduction potentials, but the origin of the increased durability of the CF₃ complex is not obvious from HOMO/LUMO orbital images. However, the same computational analysis of the singly reduced catalyst species shows a substantial involvement of the phenyl-CF₃ group in the single occupied molecular orbital (SOMO) distribution. In addition to a clear contribution from the phenyl-CF₃ group to the SOMO, the phenyl–CF₃–NHC dihedral angle is dramatically reduced (from ~50° to ~35°). This observation suggests that the primary contribution of the phenyl-CF₃ group is to stabilize the highly reactive reduced catalyst. Presumably, this stability preserves more active catalyst in solution leading to a more durable catalyst (higher turnover number (TON)) compared to other NHC photocatalysts.

The computational results also aid in understanding the unique irreversibility of the NHC catalysts toward single-electron reduction when compared with the reversible first reduction of Re(bpy)(CO)₃Br [10]. The Re–Br bond was found to substantially lengthen upon reduction by 0.08 Å when the neutral catalyst is compared with the singly reduced catalyst. The remaining Re–ligand bonds change a relatively minor amount with the Re–CO and Re–N bonds slightly shortening and the Re–NHC bond slightly increasing in length by ≤0.03 Å. This suggests that the irreversible event occurring upon the first reduction of Re(pyNHC-PhCF₃)(CO)₃Br is the loss of Br. The increased lability of this halide is potentially affected by the electron-rich nature of the NHC ring both increasing electron density at the Re center and weakly (relative to a pyridyl) stabilizing the ligand centered anion of the reduced complex. This increased electron density at the Re metal center likely leads to the lengthening of the Re–Br bond to give a more facile dissociation of the anionic bromide atom. The Br dissociated complex is known to react with CO₂ after the first reduction of Re(pyNHC-PhCF₃)(CO)₃Br, and this reduction/dissociation is a key step in accessing the active catalytic intermediate earlier in the catalytic cycle for NHC-based catalysts. It is apparent that the NHC ligand plays a key role in controlling the photoexcitation energy and electron storage after reduction, and likely indirectly aids in halide dissociation. Further studies are needed to better understand the role of the NHC ligand relative to bpy ligands.

2.2. Excited-State Lifetime: Mechanistic Implications

Accessing a highly reactive catalyst intermediate after a single reduction is additionally beneficial when considering the excited-state lifetimes of Re(pyNHC-PhCF₃)(CO)₃Br versus Re(bpy)(CO)₃Br in acetonitrile. We have previously reported an enhanced reactivity of Re(pyNHC-PhCF₃)(CO)₃Br over Re(bpy)(CO)₃Br for the photocatalytic conversion of CO₂ to CO in terms both of rate and durability. The first step of the catalytic cycle for both of these catalysts is photoexcitation followed by reduction of the catalyst with an SED. Thus, the excited-state lifetimes of the complexes are an important factor in controlling the amount of reduced complex being generated in solution. Through the use of time-correlated single photon counting (TCSPC) measurements on the two catalysts, the excited-state lifetime was found to be 4.2 ns for Re(pyNHC-PhCF₃)(CO)₃Br and 150 ns for Re(bpy)(CO)₃Br in N₂ degassed acetonitrile (Figure 4).

The NHC ligand dramatically shortens the excited-state lifetime by 36 times when compared with the bipyridyl analogue. This decreased time scale, however, did not correlate to diminished reactivity which suggests that this catalyst is remarkably active toward CO₂ addition. Energetically, a slightly larger ∆G exists for the transfer of an electron from 1,3-dimethyl-2-phenyl-2,3-dihydro-1H-benzo[d]imidazole (BIH; the SED used in our study) to the Re(bpy)(CO)₃Br catalyst (1.03 V versus 1.14 V) [10]. The outer sphere electron transfer from the SED to the photoexcited catalyst (cat*) is presumably governed by Markus theory kinetics, with the ∆G values not being large enough to typically lie conclusively within the inverted region [28]. Thus, a larger ∆G should correlate to a more facile electron transfer event. This suggests that a dramatically larger amount of reduced Re(bpy)(CO)₃Br is present in solution relative to reduced Re(pyNHC-PhCF₃)(CO)₃Br due to both a shorter excited-state lifetime for the NHC catalyst and a smaller ∆G for the electron transfer from BIH to the NHC catalyst. However, it is important to recall that the Re(pyNHC-PhCF₃)(CO)₃Br catalyst reacts CO₂ after a single electron transfer, which means that a CO₂ reactive species is being generated after a single photoexcitation and SED electron transfer. The reduced Re(bpy)(CO)₃Br catalyst likely undergoes a second electron transfer event before CO release. The relative concentration of the species being reduced is difficult to predict, although the singly reduced Re(bpy)(CO)₃Br will be lower in concentration than neutral Re(pyNHC-PhCF₃)(CO)₃Br, both of which require one electron transfer to become reactive toward CO₂ and produce CO. Thus, the higher reactivity of the Re(pyNHC-PhCF₃)(CO)₃Br catalyst (relative to Re(bpy)(CO)₃Br) toward photocatalytic CO₂ reduction could be due to a higher concentration of CO₂ reactive species in solution or due to a more reactive catalyst or both.

2.3. Catalyst Sensitivity: Water Concentration

We reasoned that the differences in mechanisms could lead to an NHC-ligated catalyst that is more robust to water and O₂. Initial anhydrous, O₂ free experiments reveal a closer match in TON values for the Re(bpy)(CO)₃Cl and Re(pyNHC-PhCF₃)(CO)₃Cl complexes (2–3 TON difference) than the Br derivatives (~15 TON difference). Thus, the Cl complexes were examined toward water and oxygen sensitivity from a common TON starting point. Both Re(pyNHC-PhCF₃)(CO)₃Cl and Re(bpy)(CO)₃Cl are stable to ambient air and moisture for prolonged periods with no noticeable loss of reactivity provided that light exposure is avoided. However, the anionic catalysts are presumably much more reactive toward water and oxygen. As a key difference in these catalysts, though, Re(bpy)(CO)₃Cl must access a doubly reduced species to be competent toward CO₂ reduction. Thus, the stability of both the first and second reduced catalyst species related to Re(bpy)(CO)₃Cl toward water and oxygen must be considered, while only the Re(pyNHC-PhCF₃)(CO)₃Cl anion needs to be considered due to the NHC ligand. This led us to hypothesize that the NHC catalyst would be more resilient to these common contaminates.

To probe this hypothesis, we compared the head-to-head photocatalytic reduction of CO₂ reactions with Re(pyNHC-PhCF₃)(CO)₃Cl and Re(bpy)(CO)₃Cl with controlled addition of water or ambient atmosphere to the reaction. We monitored the changes in rates of reactivity (turnover frequency, TOF) and the overall durability of the catalysts (turnover number, TON) for these comparisons. Concerning water, seven low-reaction solvent volume ratios were analyzed starting from dried and distilled anhydrous MeCN up to 3.2% water concentration. For the benchmark Re(bpy)(CO)₃Cl complex, strictly anhydrous conditions show a relatively low TON value (26) compared to the highest TON value of 72 with the addition of 0.05% water volume (Figure 5). This represents at least a 64% loss in reactivity relative to peak performance.

Interestingly, a near trace amount of water is needed to observe high reactivity with the Re(bpy)(CO)₃Cl catalyst, but addition of water beyond 0.05% leads to a slight decrease in durability up to 0.80% volume water before a large drop in reactivity is observed. It should be noted that the anhydrous TON value is reported as an average of several experiments with a very large variability in TON values between 62 and 4 TON (Figure 5c). We suspect that this catalyst requires a very small amount of water to perform the photocatalytic reduction of CO₂ and that the variability is due to trace amounts of adventitious water entering our reactions despite our attempts to run these reactions under strictly anhydrous conditions.

The Re(pyNHC-PhCF₃)(CO)₃Cl complex shows a similar trend with the anhydrous conditions giving about 20 TON and the peak catalyst performances observed with 0.20% water volume (35 TON, Figure 5b). While water does enhance the reactivity of Re(pyNHC-PhCF₃)(CO)₃Cl, the loss of reactivity (43%) under anhydrous conditions is much less than that of Re(bpy)(CO)₃Cl. Again, addition of larger amounts of water (up to 3.2%) shows a dramatic loss in catalyst reactivity. Interestingly, the variation in catalyst performance of about +/− 5 TON under anhydrous conditions with Re(pyNHC-PhCF₃)(CO)₃Cl is no larger than under peak performance conditions and compares favorably to the +/− 30 TON observed for Re(bpy)(CO)₃Cl under anhydrous conditions (Figure 5d). This observation helps to explain the wide range of TON values observed with Lehn’s catalyst (Re(bpy)(CO)₃Cl) from various research groups under “anhydrous” conditions, since even rigorously anhydrous experimental technique results are substantially variable.

2.4. Catalyst Sensitivity: Air Concentration

Under practical conditions, CO₂ volumes may only make up 3% of the total gas present in exhaust from fire power plants [29]. Thus, a catalyst which operates under reduced CO₂ concentrations is highly desirable [30]. To prove the reactivity of each catalyst with reduced CO₂ concentrations, ambient atmosphere (air) was added as the makeup gas. Through this study, we were able to evaluate the efficiency of these catalysts both under lowered substrate concentrations and toward resilience to O₂. Air was added in quantities ranging from 0 to 40% prior to initiating photocatalysis with 0.20% water present to ensure reproducible results for Re(bpy)(CO)₃Cl. Both catalysts work best with 0 added air (Figure 6).

However, the NHC catalyst only shows a 66% reduction in TON for all air concentrations from 1% to 40%. On the other hand, the bpy catalyst shows a continual loss of TON values as air in the reaction headspace increases to 40% for a 91% reduction in TON. Despite starting with a lower TON value at 0% air, the Re(pyNHC-PhCF₃)(CO)₃Cl catalyst outperforms Re(bpy)(CO)₃Cl at 20% air volume and higher due to no relative loss of reactivity under these conditions. This highlights the high relative stability of Re(pyNHC-PhCF₃)(CO)₃Cl to O₂, which is potentially due to the mechanistic differences in reactive anions present for the two catalysts, as described above. The TOF of these reactions was analyzed at the earliest time point to see the maximal TOF value under each air percentage. Re(pyNHC-PhCF₃)(CO)₃Cl is fastest with no air present (35 TON/h) while the Re(bpy)(CO)₃Cl catalyst is significantly slower at 16 TON/h (Figure 7). The Re(bpy)(CO)₃Cl catalyst gradually loses its initial TOF rate as more air is added. Surprisingly, after an initial loss of TOF as air increases, Re(pyNHC-PhCF₃)(CO)₃Cl shows an increase in TOF at air concentrations of 20% or more. This effect is even more dramatic if the TOF is normalized to the percent of CO₂ present in the reaction vessel. In this case, at 40% air, Re(pyNHC-PhCF₃)(CO)₃Cl is as fast as with 0% air with a gradual loss of TOF at intermediate air volumes. The origin of this increase in TOF and the rationale for such a high stability of Re(pyNHC-PhCF₃)(CO)₃Cl toward O₂ is not apparent. However, we can conclude that Re(pyNHC-PhCF₃)(CO)₃Cl is significantly more robust to O₂ than Re(bpy)(CO)₃Cl.

In conclusion, the NHC-based ligand on Re(pyNHC-PhCF₃)(CO)₃Br was shown to have SOMO delocalization onto the PhCF₃ ring after the catalyst was reduced with a planarization of the ligand π-system. This result illustrates the importance of substituent selection for the NHC wingtip. Through time-correlated single photon counting studies, the Re(pyNHC-PhCF₃)(CO)₃Br complex was found to have a dramatically shorter excited-state lifetime than the benchmark Re(bpy)(CO)₃Br complex. Presumably due to a difference in catalyst mechanisms, Re(pyNHC-PhCF₃)(CO)₃Cl maintains comparable or higher reactivity than Re(bpy)(CO)₃Cl for the photocatalytic reduction of CO₂ despite the tremendously shortened excited state. Concerning common contaminates, added water was found to be critical for reproducible reactivity with Re(bpy)(CO)₃Cl while Re(pyNHC-PhCF₃)(CO)₃Cl was significantly less sensitive to anhydrous conditions. Both catalysts show substantially reduced reactivity when water reached ~1% of the solvent volume. With regard to reduced CO₂ concentration with O₂ incorporation from ambient atmosphere as the makeup gas, Re(pyNHC-PhCF₃)(CO)₃Cl was found to be significantly less sensitive than Re(bpy)(CO)₃Cl as ambient atmosphere concentrations increased. Interestingly, Re(pyNHC-PhCF₃)(CO)₃Cl was found to have a similar turnover frequency at 0% ambient atmosphere (100% CO₂) as at 40% ambient atmosphere once the amount of CO₂ was normalized. Future studies will focus on the examination of NHC ligand designs to find a more durable catalyst given the unique properties of NHC-ligated complexes in photocatalytic CO₂ reduction reactions. Additionally, we plan to evaluate new metal centers with future ligand designs as the combined role that the ligand and metal center play in CO₂ reduction is likely important but not fully understood.

3. Experimental Section

3.1. Computational Details

Density functional theory (DFT) was used to compute the optimized structures and harmonic vibrational frequencies for both the neutral (18 e⁻) and anionic (19 e⁻) systems. The M06-L [31] functional was employed for these computations. The density fitting approximation was employed with M06-L using the default auxiliary basis sets in the Gaussian 09 (Rev: E.01) software package [32]. The Hay–Wadt relativistic effective core potential and LANL08(f) uncontracted triple-ζ valence basis set was initially used for the Re atom [33] in conjunction with the 6-31++G(d, p) double-ζ split-valence basis set [34,35] for all other atoms, which has been used elsewhere to successfully characterize similar rhenium complexes [36,37]. All computations employed a pruned numerical integration grid with 99 radial shells and 590 angular points per shell and the default threshold of 1 × 10⁻⁶ for removing linearly dependent basis functions. Orbital images were prepared with Avogadro 1.0.3 with an iso value of 0.25.

3.2. Electron Lifetime Measurement Information

All sample concentrations were on the order of 10⁻⁵ M to reduce reabsorption. Luminescence lifetimes were obtained by exciting with the 485 nm line of a pulsed diode laser (fwhm < 100 ps) and detecting with a PicoQuant PDM series single photon avalanche diode (Micro Photon Devices, Bolzano, Italy).

3.3. Photocatalysis General Information

Prior to experimentation, glassware was flame-dried under vacuum, then kept under nitrogen pressure. MeCN was dried for 24 h over calcium hydride, distilled with the first and last 20% of the solvent discarded, and stored under nitrogen with dry 3 Å molecular sieves prior to being used as the solvent source. Solvent, triethylamine, and catalyst solutions were added via dry, nitrogen-flushed syringes. Photosensitizer solutions were prepared in flame-dried 10 mL round-bottom flasks. BIH was added under positive nitrogen pressure.

Irradiation for photocatalytic experiments was performed with a 150W Sciencetech SF-150C Small Collimated Beam Solar Simulator equipped with an AM1.5 filter (Sciencetech, London, ON, Canada). Headspace analysis was performed using a gastight syringe with stopcock and Agilent 7890B Gas Chromatograph (column, Agilent PorapakQ 6 ft, 1/8 OD) (Agilent, Santa Clara, CA, USA). Quantitations of CO and CH₄ were made using an FID, while H₂ was quantified using a TCD (all calibrated using standards purchased from BuyCalGas.com) (Cross Company, Greensboro, NC, USA).

3.4. Water-Control Photocatalysis Procedure

To a flame-dried 17 mL flask were added BIH (0.005 g, 0.02 mmol), dry MeCN (1.8 mL), and catalyst (0.2 mL of a 1 × 10⁻³ M in dry MeCN solution). H₂O was added via a microsyringe. The solution was then bubbled vigorously for at least 15 min, until the total volume reached 1.9 mL. At that time, degassed, dry triethylamine (0.1 mL) was added to the sealed flask by syringe. The samples were then irradiated with a solar simulator (150 W Xe lamp, AM1.5 filter).

3.5. Oxygen Control Photocatalysis Procedure

To a flame-dried 17 mL flask were added BIH (0.005 g, 0.02 mmol), dry MeCN (1.8 mL), and catalyst (0.2 mL of a 1 × 10⁻³ M in dry MeCN solution). A quantity of 4 μL of H₂O (0.2% v/v) was added via a 10 μL microsyringe. The solution was then bubbled vigorously for at least 15 min, until the total volume reached 1.9 mL. At that time, degassed, dry triethylamine (0.1 mL) was added to the sealed flask by syringe. Using a gastight syringe with stopcock, CO₂ was removed and replaced with a % v/v ambient atmosphere. The samples were then irradiated with a solar simulator (150 W Xe lamp, AM1.5 filter).