CO2 Reduction by an Iron(I) Porphyrinate System: Effect of Hydrogen Bonding on the Second Coordination Sphere

Transforming CO2 into valuable materials is an important reaction in catalysis, especially because CO2 concentrations in the atmosphere have been growing steadily due to extensive fossil fuel usage. From an environmental perspective, reduction of CO2 to valuable materials should be catalyzed by an environmentally benign catalyst and avoid the use of heavy transition-metal ions. In this work, we present a computational study into a novel iron(I) porphyrin catalyst for CO2 reduction, namely, with a tetraphenylporphyrin ligand and analogues. In particular, we investigated iron(I) tetraphenylporphyrin with one of the meso-phenyl groups substituted with o-urea, p-urea, or o-2-amide groups. These substituents can provide hydrogen-bonding interactions in the second coordination sphere with bound ligands and assist with proton relay. Furthermore, our studies investigated bicarbonate and phenol as stabilizers and proton donors in the reaction mechanism. Potential energy landscapes for double protonation of iron(I) porphyrinate with bound CO2 are reported. The work shows that the bicarbonate bridges the urea/amide groups to the CO2 and iron center and provides a tight bonding pattern with strong hydrogen-bonding interactions that facilitates easy proton delivery and reduction of CO2. Specifically, bicarbonate provides a low-energy proton shuttle mechanism to form CO and water efficiently. Furthermore, the o-urea group locks bicarbonate and CO2 in a tight orientation and helps with ideal proton transfer, while there is more mobility and lesser stability with an o-amide group in that position instead. Our calculations show that the o-urea group leads to reduction in proton-transfer barriers, in line with experimental observation. We then applied electric-field-effect calculations to estimate the environmental effects on the two proton-transfer steps in the reaction. These calculations describe the perturbations that enhance the driving forces for the proton-transfer steps and have been used to make predictions about how the catalysts can be further engineered for more enhanced CO2 reduction processes.


■ INTRODUCTION
The cumulative emission of CO 2 is raising environmental concerns for its role in global warming, and consequently CO 2 reduction is an urgent global issue.−18 This approach has 2-fold advantages, namely, CO 2 reduction in the atmosphere as well as reduction in the use of unrenewable fossil fuels for synthesis of these compounds.Therefore, transforming CO 2 into valuable chemicals may provide a solution to not only environmental problems caused by the emission of CO 2 but also the anticipated depletion of fossil resources.
took an iron(I) tetraphenylporphyrin (TPP) complex and inserted substituents on the ortho and para positions of one of the phenyl groups (Scheme 1). 34These porphyrin derivatives with amide groups in the second coordination sphere were shown to give an enhanced reduction in CO 2 to CO and water.They proposed a better positioning of CO 2 on the iron center that enabled more efficient proton transfer from solution.Subsequent, density functional theory (DFT) calculations of some of us on CO 2 reduction by [Fe(TPP)] − and [Fe(o-2amide-TPP)] − showed the same electronic configurations and mechanisms for the catalytic cycle with small proton-transfer barriers from phenol. 35Recent work of Chang and co-workers used a system with a urea group linked to one of the mesophenyl groups. 36In combination with bicarbonate as the proton donor, the authors measured a 1500-fold rate enhancement for CO 2 reduction compared to that of a system without bicarbonate.They also determined a crystal structure of the bicarbonate-bound o-urea-TPP with Zn 2+ as the central ion.To understand how and why bicarbonate can enhance the reaction rates for CO 2 reduction processes with iron porphyrins, we decided to do a DFT study into the mechanism of the CO 2 reduction process using an iron(I) porphyrin system with either p-urea, o-2-amide, or o-urea as substituents to one of the meso-phenyl groups of the porphyrin system with bicarbonate as the proton donor (models I−III in Scheme 1) and the models with o-2-amide and o-urea but with phenol as the proton donor (models II,phenol and III,phenol).The work shows that a hydrogen-bonding interaction in the second coordination sphere enhances proton-transfer rates.Moreover, the urea group can lock bicarbonate into a double hydrogenbonding interaction mimicking a salt bridge in enzymatic structures that stabilizes the system and provides the best network for proton relay.

■ METHODS
The model was based on the CO 2 -bound optimized geometry of our previous studies on CO 2 reduction by an [Fe I (TPP)] complex, where TPP = meso-tetraphenylporphyrin, 35 whereby we manually inserted the p-urea, o-2-amide, or o-urea groups into one of the meso-phenyl substituents to create models I−III (Scheme 1) and replaced the other three meso-phenyl groups by a hydrogen atom.In addition, a molecule of bicarbonate was added to the distal site to create a cluster model with overall charge of 3− and odd multiplicity.To test whether the large negative charge in the model influences the structure and kinetics, we also used phenol as a proton donor because it has a pK a value similar to that of bicarbonate, which gave the model an overall charge of 2−.−40 An all-electron def2-SVP 41 basis set was selected for all atoms during the geometry optimizations, analytical frequency calculations, and constraint geometry scans designated as BS1.Single-point energy calculations were performed with the def2-TZVP basis set on all atoms designated as BS2.Although the experimental work was performed in a dimethylformamide (DMF) solution, we tested several implicit solvent models with a range of dielectric constants of 37.2 (DMF) and 78.4 (water). 42In general, a change in the dielectric constant for the solvent model gives minor differences to the energies and optimized geometries and therefore does not change the results dramatically; see the Supporting Information.Calculations were performed on all low-lying triplet and quintet-spin-state surfaces.These methods and approaches were used before in our group and reproduced experimental free energies of activation within a few kilocalories per mole and predicted product selectivities in line with experimental observation. 43,44

RESULTS AND DISCUSSION
The catalytic cycle for the reduction of CO 2 to CO and water on an iron(I) porphyrinate complex as calculated here is shown in Scheme 2. The cycle starts from an iron(I) porphyrinate system with nearby bicarbonate, structure A or [Fe I (HCO 3 − )] 2− .This complex is reduced from iron(I) to iron(0) to form B and triggers the binding of a molecule of CO 2 and its reaction to give C or [Fe I (CO 2 − )(HCO 3 − )] 3− .An internal proton transfer from bicarbonate to CO 2 is expected to lead to protonated CO 2 and carbonate in complex D: [Fe II (OCOH − )(CO 3 2− )] 3− .The carbonate picks up a proton from the solvent/buffer and assists with the second protonation step of CO 2 that leads to the formation of CO and water in complex E. The cycle is closed with the release of CO and the reduction of iron(II) to iron(I).In our previous studies, 35 we showed that the first two steps in the CO 2 reduction cycle, i.e., A → B and B → C, are little affected by second-coordination-sphere effects and substitution of the meso-phenyl groups of the porphyrin complex.Therefore, our work presented here is mostly focused on the protonation steps in the cycle and on how bicarbonate is involved in the process.In particular, we created CO 2 -bound iron(II) porphyrinate complexes with p-urea, o-2-amide, and o-urea as substituents to one of the phenyl groups on the meso position: structures C I , C II , and C III .Optimized geometries of structures C of models I−III in the triplet and quintet spin states are shown in Figure 1.The spin multiplicity is given as a superscript in front of the label and the model type as a subscript after the label.
In all structures, the triplet and quintet spin states are within 1.5 kcal mol −1 , which implies that both may be accessible.Structures II and III have the same spin-state ordering and spin ground state; hence, these are not affected by the modification in the second coordination sphere.In the p-urea system (I), we find the quintet spin below the triplet spin state by 1.3 kcal mol −1 for structure C. Nevertheless, for all structures C, the triplet−quintet energy gap is within 1. mol −1 ; consequently, the spin states are degenerate, and both will be accessible at room temperature conditions.We also calculated the CO 2 binding energies for all complexes.In previous work, we reported CO 2 binding energies of ΔG = 11.3 kcal mol −1 for the difference in energy between [Fe I (CO 2 − )(o-2-amide-TPP)] 2− and an isolated CO 2 and [Fe 0 (o-2-amide-TPP)] 2− complex.With additional bicarbonate in the model in hydrogen-bonding interactions with CO 2 in the C II and C III structures, CO 2 release requires a free energy of ΔG = 9.6 and 9.2 kcal mol −1 in the triplet spin state, whereas we find ΔG = 4.6 and 3.5 kcal mol −1 in the quintet spin state.As such, the double hydrogen bond from urea to bicarbonate does not appear to dramatically influence the CO 2 binding energy compared to the o-2-amide system.Despite the fact that the bicarbonate anion forms hydrogen-bonding interactions with the CO 2 group bound with o-urea and o-2amide, there does not appear to be a major stabilization energy as a result of this.As a matter of fact, our previous study also had a hydrogen bond to the bound CO 2 group in [Fe II (CO 2 )(o-2-amide-TPP)] 2− but directly with the amide proton.However, the hydrogen-bonding network will facilitate the CO 2 binding and lock it in position.
The structures shown in Figure 1 give tight binding of CO 2 and bicarbonate between the iron(II) and hydrogen-bonding o-amide and o-urea groups.In the p-urea structure, the hydrogen-bonding groups are relatively far from the iron center by more than 10 Å, and, consequently, a bridge will require multiple water molecules, which were not included in the model.Nevertheless, the hydrogen-bonding network to the CO 2 group only has a minor effect on the Fe−C bond lengths, which is 2.085 Å in 5 C I and 2.018 Å in 3 C I , whereas the quintet-spin-state structures for 5 C II and 5 C III have values within 0.02 Å.These Fe−C distances are in line with those reported previously for the interaction of iron with a first-row element such as oxygen, nitrogen, or carbon. 45The optimized geometries compare well to those calculated before using the B97-D approach with strong hydrogen-bonding interactions between bicarbonate and the two NH groups of the urea substituent and the oxygen atom of CO 2 . 36−50 In model III, the bicarbonate ion is locked in position through two hydrogen-bonding interactions with the urea N− H groups, while its OH group forms a hydrogen bond to the CO 2 group.−53 However, different from a salt bridge in a protein structure, the urea system investigated here is charge-neutral; hence, the binding strength of the interaction with bicarbonate will be  much weaker than that in a typical salt bridge.The two urea hydrogen-bonding interactions are different in length; namely, the one closest to the porphyrin is shortest at 1.690 Å in 5 C III , while the other one has a distance of 1.737 Å.At the same time, the OH group of bicarbonate forms a hydrogen bond with one of the oxygen atoms of CO 2 at 1.718 Å.These distances are very similar for the o-2-amide complex, but because it can form only two hydrogen bonds, the bicarbonate is lesser restraint.
Next, we studied the proton-transfer step from 5,3 C (models I−III) to form an iron(II) with a protonated CO 2 group (structures IM1) via transition state TS 1 .Thereafter, we took structures 5,3 IM1 and added a proton to the carbonate group to form the 5,3 RC2 complexes and searched for a protontransfer transition state (TS2) to form water and CO products (structures IM2).The optimized transition-state structures, i.e., TS1 and TS 2 , for the model III pathway are shown in Figure 2, together with the calculated energy landscapes.Both proton-transfer free energies of activation are small, namely, ΔG ⧧ = 5.4 kcal mol −1 for the first proton transfer with respect to 5 C III and ΔG ⧧ = 5.7 kcal mol −1 with respect to 5 RC2 III (or ΔG ⧧ = 11.0 kcal mol −1 with respect to 5 C III ).On the triplet spin state, the barriers are considerably higher in energy, namely, ΔG ⧧ > 14 kcal mol −1 for the first proton transfer to reach 3 IM1 III in an endergonic step of 13.1 kcal mol −1 .On the triplet spin state, we were not able to fully characterize it transition state, but a constraint geometry scan shows it to be only slightly higher in energy than the value for 3 IM1 III ; hence, a value of <14 kcal mol −1 is reported.The second proton transfer, similar to the quintet-spin-state surface, is 5.7 kcal mol −1 .This is not surprising as no electron transfer happens in these steps and all group spin densities remain the same.Interestingly, for the IM2 III complex, the quintet-spin-state structure is higher in energy than the corresponding triplet spin complex.Thermal collisions of 5 IM2 III in solution, however, are likely to revert the spin to the triplet spin ground state for the CO-bound structures IM2.
Structurally, the proton-transfer transition state 5 TS1 III has the transferring proton midway between the donor and acceptor groups at distances of 1.184 and 1.246 Å.The O− H−O angle is close to linearity (177°), and the imaginary frequency of the transition state is i871 cm −1 for the protontransfer mode.This is a relatively low imaginary frequency for proton transfer because usually values well over i1000 cm −1 are found. 54−56 For 5 TS2 III , the imaginary frequency is similar (i730 cm −1 ) and represents simultaneous proton transfer and C−O cleavage, i.e., water release.The distance of the proton from the carbonate in 5 TS2 III is 1.242 Å, while the distance to the OH group is 1.178 Å.Note that the Fe−C distances in 5 TS1 III and 5 TS2 III are similar (2.138 vs 2.129 Å, respectively), and consequently no electron transfer has happened in this process.Both energy landscapes were calculated with either water or DMF as an implicit solvent model.Both sets of data are shown in Figure 2, and as can be seen, the energy and structural differences are small when the solvent model is changed from water to DMF.These results match previous conclusions drawn from changing the value of the dielectric constant in implicit solvent calculations. 57o find out how bicarbonate as a proton donor reacts compared to alternative proton donors, we removed bicarbonate from structures 3,5 C III and replaced it with a phenol molecule: 3,5 C III,phenol .Subsequently, the protontransfer pathways of CO 2 were investigated, leading to CO and water as products.Figure 3 shows the optimized transition-state geometry for the first and second proton transfer on the quintet-spin-state surface with phenol as a proton donor.The obtained landscape for the phenol-bound structures compared to the bicarbonate reaction is shown in Figure 3, and the data for all local minima and transition states for all calculated models are reported in Table 1.The free energy of activation for the first proton transfer from 5 C III,phenol is ΔG ⧧ = 8.0 kcal mol −1 , while the second proton transfer was calculated at ΔG ⧧ = 6.6 kcal mol −1 with respect to 5 C III,phenol .Previously, for proton transfer from phenol to CO 2 bound to Fe II (o-2-amide-TPP), we found a proton-transfer free energy of activation of 7.0 kcal mol −1 for the second proton-transfer step in the catalytic cycle, in agreement with what we find here with different methods and models. 35Taken together, the two consecutive proton-transfer steps for 5 C III have a maximum barrier of 11.0 kcal mol −1 using bicarbonate as a proton donor, while with phenol as the proton donor, the barrier is reduced to 8.0 kcal mol −1 .Although the energy differences between the bicarbonate and phenol models are small, they do indicate a slowing down of the rate by a factor of 600 for the replacement of phenol by bicarbonate.This is not a surprising result because the pK a value of the free bicarbonate/carbonate couple is 10.3, while the deprotonation of phenol has a pK a value of 9.95. 58As such, free phenol should react with lower protontransfer barriers than free bicarbonate, as indeed is observed here.These pK a values, of course, will be affected upon binding to a hydrogen-bonding scaffold, such as o-urea, which can hold bicarbonate stronger than phenol.Despite the strong binding of bicarbonate to the urea scaffold, we see little effect on the rates based on the pK a differences with phenol.The pK a values of the o-urea and o-2-amide groups were calculated and found to be much higher than those for phenol.Therefore, the o-urea and o-2-amide groups cannot serve as proton donors in the reaction mechanisms because their N−H bonds are too strong to break.In none of the geometry optimizations reported here was a proton transfer from the o-urea or o-2-amide groups to either bicarbonate, CO 2 , or phenol observed.Therefore, the calculations show that a proton donor is needed and that phenol and bicarbonate are strong enough bases to provide protons for the CO 2 reduction reaction.
The optimized transition-state structures 5 TS1 III,phenol and 5 TS2 III,phenol are also shown in Figure 3.The first protontransfer transition state ( 5 TS1 III,phenol ) has a small imaginary frequency of i183 cm −1 for proton transfer from phenol to CO 2 .The structure is relatively central, with the transferring proton midway between the donor (by 1.190 Å) and acceptor (by 1.140 Å) oxygen atoms, respectively.The angle O−H−O is close to linearity (173°), as is expected of a proton-transfer transition state.In the second proton-transfer transition state, the structure is more upright and the C−O bond has elongated significantly to 2.097 Å.The imaginary frequency of i660 cm −1 indicates dominant proton transfer from phenol to oxygen.The transferring proton is at a distance of 1.258 Å from phenolate and 1.162 Å from the accepting oxygen atom.Nevertheless, the calculations presented in Figures 2 and 3 show that both bicarbonate and phenol are strong enough acids to donate a proton to an Fe II -CO 2 complex efficiently and initiate the CO 2 reduction process.We observe a lowering of the first proton-transfer barrier for o-urea versus o-2-amide by 0.4 kcal mol −1 , which would correspond to a rate enhancement of a factor of 2.4.Our work, therefore, is in good quantitative agreement with the experimental work of Chang et al., which showed enhanced reactivity by 2-fold between the two complexes with either o-urea or o-2-amide in the second coordination sphere. 36verall, the calculations presented in this work indicate that an o-urea substituent to a TPP scaffold enables strong hydrogen-bonding interactions that hold and position distal ligands to the iron center better than o-2-amide.To fully understand the positional differences, we created an overlay of the 3 C II and 3 C III optimized geometries and present the results in Figure 4.As can be seen, the two structures are seemingly similar and have most groups in a similar position and orientation.However, there are differences, as highlighted in the figure.Both of these differ by more than 12°and show that bicarbonate is better positioned for proton transfer in 3 C III than in 3 C II .
To find out how these iron porphyrin systems can be improved for more efficient proton transfer, we decided to apply electric-field-effect perturbations to the optimized geometries of 5 C III , 5 IM1 III , 5 RC2 III , and 5 IM2 III .In particular, single-point calculations at the UB3LYP-GD3/BS2 level of theory were run in Gaussian with an electric field located along the molecular x, y, or z axis, with magnitudes ranging from −200 to +200 au, and the results are shown in Figure 5.These electric-field perturbations were used previously in our group and shown to influence charge distributions in complexes and bifurcation patterns in chemical catalysis. 59,60In particular, recent work showed that charged groups in proteins can influence the strength of the C−H bonds in substrates and direct a reaction selectivity to a specific bond in a substrate. 61,62Thus, an electric-field perturbation has a major effect on the thermodynamics for proton transfer and the reaction energy.To be specific, an electric-field effect along the negative z axis is along the proton-transfer axis and makes the first proton-transfer step more exergonic, while a field in the opposite direction makes the reaction more endergonic, i.e., less likely.Interestingly, electric-field effects in the x and y directions do not appear to have a major effect on the first proton-transfer step unless very large electric fields are used.For the second proton-transfer step, a field along the positive y axis is favorable, while in the negative y direction, the reaction becomes more endothermic.The vector in the imaginary frequency of TS2 for this reaction step is along the y axis and shows motions for the C−O stretch vibration, i.e., cleavage of the C−O bond as well as proton transfer.Also, for the second proton-transfer step, fields orthogonal to the proton transfer show few effects on the reaction thermochemistry.Overall, these calculations show that the first proton-transfer step is improved with an electric-field effect along the negative z axis, while the second proton-transfer step is probably faster with a field along the positive y axis.Therefore, engineering of the iron porphyrin complex by, for instance, latching the porphyrin to a metal surface through its axial ligand, may help with the first proton-transfer reaction in the CO 2 activation reaction, while perturbations with charged groups along the y axis will enhance the second proton-transfer step.

■ CONCLUSIONS
In this work, a computational study is presented on CO 2 reduction to CO and water on several iron porphyrinate complexes.The calculations show that an iron porphyrinate system is an efficient catalyst for CO 2 reduction reactions and particularly systems with a hydrogen-bonding donor in the second coordination sphere because that helps to position and tighten the substrate.The tight binding of the CO 2 substrate enables two low-energy proton-transfer steps to form CO and water efficiently.Furthermore, bicarbonate can be locked in a position with an o-urea group attached to the ortho meso position of the ligand to provide a second-coordination-sphere environment for locking the proton donor (bicarbonate) and substrate (CO 2 ) in a tight orientation for efficient proton shuttle and ultimately CO 2 reduction purposes.Finally, electric-field calculations were performed on the complexes for the successive proton-transfer steps in the catalytic cycle.These calculations predict that the proton-transfer steps will be sensitive to local perturbations, and an electric-field-effect perturbation can influence driving forces for these steps and make the CO 2 reduction reaction more exothermic and efficient.
■ ASSOCIATED CONTENT * sı Supporting Information

Figure 1 .
Figure 1.UB3LYP-GD3/BS1-optimized geometries of structure C for models I−III.Relative energies include energies at the BS2 level of theory and zero-point corrections in kilocalories per mole.Bond lengths are in angstroms.

Figure 2 .
Figure 2. UB3LYP-GD3/BS2//UB3LYP-GD3/BS1 calculated the free energy profile for the two proton-transfer steps from the CO 2 -bound iron(II) porphyrin complex [Fe II (CO 2 )(o-urea-TPP)(HCO 3 − )] 3− .Free energies (ΔG) were calculated at 298 K and include energies at the BS2 level of theory and zero-point, solvent, thermal, and entropic corrections in kilocalories per mole relative to 5 C III , whereby calculations in the water solvent are the data outside parentheses and those in the DMF solvent are the data inside parentheses.Optimized transition-state geometries for the first and second proton transfer for model III are shown with bond lengths in angstroms, angles in degrees, and the imaginary frequency in reciprocal centimeters.In RC2, a proton is added to the model of IM1 and the energy is set to the value of IM1 for comparison.

Figure 3 .
Figure3.UB3LYP-GD3/BS2//UB3LYP-GD3/BS1 calculated the free energy profile for the two proton-transfer steps from the CO 2 -bound iron(II) porphyrin complex [Fe II (CO 2 )(o-urea-TPP)(phenol)] 2− .The data for bicarbonate are in blue, whereas the data for phenol are in green.Free energies (ΔG) were calculated at 298 K and include energies at the BS2 level of theory and zero-point, solvent, thermal, and entropic corrections in kilocalories per mole relative to 5 C III,phenol .Optimized transition-state geometries for the first and second proton transfer for model III are shown with bond lengths in angstroms, angles in degrees, and the imaginary frequency in reciprocal centimeters.In RC2, a proton is added to the model of IM1, and the energy is set to the value of IM1 for comparison.
Thus, the O−H−O angle is close to linearity in 3 C III at 172°, while the angle is 168°in 3 C II .The closer to linearity the O−H−O angle is, the easier it is for proton transfer to take place.Indeed the smallest proton-transfer barrier is found for 3 C II .Further differences between 3 C II and 3 C III relate to the positioning of the bicarbonate ion in the complex, as seen from the dihedral angles Fe−C−O−H and O−C−O−H in Figure 4.

a
ΔG(BS2) values in kilocalories per mole.b Data from ref 35.
3 kcal Scheme 1. Models Investigated in This Work Scheme 2. Catalytic Cycle of CO 2 Reduction on a Bicarbonate-Bound Iron Porphyrinate System Studied in This Work a a Formal oxidation states of iron are given in square brackets.

Table 1 .
Free Energies (BS2//BS1) for Calculated Reaction Pathways for Double Protonation of CO 2 Bound to Various Porphyrin Models a