Kinetic and Mechanistic Effects of Bipyridine (bpy) Substituent, Labile

Jan 23, 2018 - ‡Department of Chemistry and §School of Engineering and Applied Science, Princeton University, Princeton, New Jersey 08544, United S...
0 downloads 0 Views 1MB Size
Research Article Cite This: ACS Catal. 2018, 8, 2021−2029

pubs.acs.org/acscatalysis

Kinetic and Mechanistic Effects of Bipyridine (bpy) Substituent, Labile Ligand, and Brønsted Acid on Electrocatalytic CO2 Reduction by Re(bpy) Complexes Melissa L. Clark,† Po Ling Cheung,† Martina Lessio,‡ Emily A. Carter,§ and Clifford P. Kubiak*,† †

Downloaded via KAOHSIUNG MEDICAL UNIV on November 3, 2018 at 23:16:06 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

Department of Chemistry and Biochemistry, University of California, San Diego, 9500 Gilman Drive MC 0358, La Jolla, California 92093, United States ‡ Department of Chemistry and §School of Engineering and Applied Science, Princeton University, Princeton, New Jersey 08544, United States S Supporting Information *

ABSTRACT: In order to help develop robust and deployable molecular electrocatalysts for the reduction of CO2 to CO, we must understand the effects of tuning their structure and catalytic conditions. To this end, we quantify how modifications to the catalyst fac-Re(4,4′-R-bpy)(CO)3X (bpy = 2,2′-bipyridine, R = OCH3, CH3, tBu, H, CN, CF3; X = Cl, Br, py(OTf), or CH3CN(OTf)) with and without an added proton source (phenol, acetic acid, 2,2,2-trifluoroethanol) affect the catalyst stability, activity, and overpotential. Through cyclic voltammetry experiments, we found that the substituents and proton source had a large effect on both overpotential and activity. Substituents with moderate electron-donating ability (tBu and CH3) increased activity and overpotential in comparison to the unsubstituted complex Re(bpy)(CO)3Cl. In contrast, substituents resulting in too much electron density distributed over the bpy ligand, either from too-strong electron-donating ability (OCH3) or from the requirement of a third reduction to activate the complex (CN and CF3), destabilized the catalyst. An added proton source both increased the activity and decreased the overpotential by 200 mV for all catalyst derivatives, shifting the catalytic mechanism from an electron-first pathway to a proton-first pathway. We used binding energies calculated via density functional theory to help understand the substituent effect on the catalyst affinity for CO2 and other intermediates relevant to the catalytic mechanism. Catalyst activity was quantified using intrinsic rate constants determined through the utilization of catalytic plateau currents, as well as the application of a foot of the wave analysis, which yielded incongruent values. Of those complexes tested, Re(4,4′-tBubpy)(CO)3Cl with an added 1 M phenol yielded the most active catalytic system (kcat = 6206 s−1) at an overpotential of 0.67 V. KEYWORDS: electrocatalysis, CO2 reduction, bipyridine, cyclic voltammetry, catalytic Tafel plot



molecular systems for the same reaction.15,16 Savéant and coworkers have demonstrated how cyclic voltammetry (CV) can be utilized to gain both kinetic and thermodynamic parameters that are intrinsic to the catalytic activity of a complex, thus limiting variations due to experimental design.16−18 Ideally, a catalytic CV would show an “S-shaped” wave that is independent of the scan rate, where the plateau current can be used to determine an intrinsic catalytic rate constant (kcat). More common, however, is a peak-shaped response, which occurs due to events such as deactivation of the catalyst, consumption of the substrate, or inhibition by products.18 An S-shaped wave can sometimes still be obtained through traversing the kinetic zone region by increasing the scan rate, decreasing catalyst concentration, or increasing the substrate

INTRODUCTION Electrochemically converting CO2 to liquid fuels is a promising pathway to help mitigate anthropogenic climate change while contributing to the world’s energy supply. A viable route for this process is the electrocatalytic reduction of CO2 to CO, which can be used in Fischer−Tropsch technologies to create hydrocarbon fuels.1 An organometallic catalyst family demonstrated to be selective and active toward the reduction of CO2 to CO is the one based on the fac-Re(2,2′-bipyridine)(CO)3Cl (Re-bpy) motif.2−6 This catalyst and derivatives thereof have been investigated for their electrochemical7 and photochemical8−10 catalytic properties, as well as incorporated into higher-order systems11,12 and attached to surfaces.13,14 Most of these studies involve tailoring the Re-bpy system, although the variance in experimental conditions and measurements complicate the comparison of catalysis across reports. A recent trend in the literature has been toward standardizing the reporting of catalysis so that one can directly compare © 2018 American Chemical Society

Received: November 21, 2017 Revised: January 15, 2018 Published: January 23, 2018 2021

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis concentration,19 but this is not always the case. In such instances, a foot of the wave analysis (FOWA) is more informative, as it utilizes an onset catalytic current that is not yet affected by side phenomena.20 FOWA has given rate constants comparable to those of stopped-flow UV−vis spectroscopy for the catalytic reduction of oxygen by iron(0) tetraphenylporphyrin (FeTPP), demonstrating the validity of a CV approach to obtain kinetic information.21 With these intrinsic kinetic data and the thermodynamic descriptor of catalytic overpotential, CO2 to CO catalysts can be directly compared across reports via catalytic Tafel plots, providing a clear picture of current progress in the field.22 These CV techniques have not been applied to Re-bpy, despite CV being a primary characterization tool for these complexes. Herein, the descriptors of intrinsic catalytic rate constant and overpotential are used to report the effects of the 4,4′-substituent (OCH3, CH3 tBu, H, CN, CF3), labile ligand (pyridine, acetonitrile, chlorine, bromine), and Brønsted acid (phenol, 2,2,2-trifluoroethanol, acetic acid) on Re-bpy-based catalysis (Scheme 1). While the modifications to Re-bpy have

complexes were chosen with relatively unreactive electrondonating/-withdrawing substituents in the 4,4′-positions of the bpy: Re(4,4′-R-bpy)(CO)3Cl, where R = OCH3, tBu, CH3, H (bpy-H), CF3, and CN (Scheme 1). The above order is in decreasing electron-donor ability, as described by the parasubstituted Hammett parameter (σp, Table 1).23 As detailed previously,2,24−26 each complex displays a reversible oneelectron wave followed by an irreversible or quasi-reversible one-electron wave (Figure 1). These have been confirmed

Scheme 1. Variants of Re(4,4′-R-2,2′-bipyridine)(CO)3X Studied for the Proton-Assisted Electrochemical Reduction of CO2 Figure 1. Cyclic voltammograms under Ar of Re(4,4′-R-bpy)(CO)3Cl (where R is labeled on the graph) taken at 0.1 V/s. Reduction potentials are given in Table S1 in the Supporting Information.

spectroscopically and crystallographically as a reversible bpycentered reduction succeeded by a metal-based reduction (ReI/0) that results in the loss of the labile ligand, in this case chloride.27 A third irreversible reductive wave is shown at more negative potentials (600−810 mV negative of the second reduction, Figure S3). This has been experimentally confirmed as bpy-centered reduction for CN;13 this assignment is most likely valid for the other five complexes as well. The electron-donating substituted complexes OCH3, tBu, and CH3 have reduction potentials slightly more negative than that of bpy-H (−120, −80, and −110 mV, respectively), reflecting an increased difficulty in reducing the more electron rich compounds. In contrast to the electron-donating substituents, the complexes with electron-withdrawing substituents, CF3 and CN, have reduction potentials much more positive than that of bpy-H by 440 and 590 mV, respectively. The difference in the first reduction potentials between the most electron donating OCH3 (σp = −0.27) and electronwithdrawing CN (σp = +0.66) is ca. 700 mV. The reduction

been previously reported,2,7,23 full comparisons, especially with respect to the electrocatalytic mechanism, have not been made. To this end, we used density functional theory (DFT) calculations to describe how modifications to the catalyst affect the mechanism and affinity for CO2 binding. These descriptors, coupled with careful analysis by CV techniques, allow for the comparison of Re-bpy with other CO2 reduction electrocatalysts on the basis of catalytic Tafel plots, which reveal the unique nature of the Re-bpy system for selective catalysis.



RESULTS Cyclic Voltammetry under Ar. We first compared the complexes via CV under an inert atmosphere (Ar) to investigate the thermodynamic effect of the bpy substituents’ electron-donating nature, as well as the labile ligand. Six

Table 1. Catalytic Descriptors for Re(4,4′-R-bpy)(CO)3Cl under CO2a R σp Ecat/2 (V)b ηc FECO (%) icat/ip kcat (υ, s−1)f

OCH3

tBu

CH3

H

CF3

CN

−0.27 −2.18 −1.64 59 40.7 4088 (20)

−0.20 −2.20 −1.66 100d 32.1 2601 (10)

−0.17 −2.14 −1.60 100d 32.6 3336 (20)

0 −2.09 −1.55 100d 8.4 155 (5)

+0.54 −2.57 −2.03 30 15.0 8606 (10)

+0.66 −2.31 −1.77 18e 8.3e 3487 (25)

a

Benzene para-substituted Hammett parameter (σp), catalytic potential (Ecat/2 vs Fc+/0), overpotential (η), Faradaic efficiency to produce CO (FECO), ratio between peak current under CO2 and peak current under inert conditions (icat/ip), and intrinsic catalytic rate constant (kcat). b Determined by the half wave height method.28 cDetermined using E°CO2/CO(CH3CN) = −0.541 V vs Fc+/0.29 dValues from ref 2. eValues from ref 13. fScan rate (υ) used is given in parentheses (V/s). 2022

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis potentials for all of the substituted complexes are reported in Table S1. We also considered the effect of the labile ligand in addition to changes in the bpy substituent. Four tert-butylbipyridinesubstituted complexes with differing labile ligands were compared: Re(tBu-bpy)(CO)3X (X = Cl (tBu), Br, pyridine (triflate counterion, py(OTf)), or acetonitrile (CH3CN(OTf)).7 We found that the identity of the labile ligand affects the reduction potentials in comparison to tBu, although it is subtle in comparison to the substituent effects (Figure 2, inset;

Figure 2. Cyclic voltammograms under catalytic conditions (CO2 atmosphere) of the four labile ligand complexes Re(bpy-tBu)(CO)3X, where X = Cl (Ar, black; CO2, red), Br (blue), pyridine (green), or CH3CN (orange). Scans were taken at 0.1 V/s. The inset (A) gives an enlargement of the boxed area to highlight the reductions before the catalytic wave.

Figure 3. Cyclic voltammograms of the six 4,4′-substituted complexes Re(R-bpy)(CO)3Cl under Ar (gray), CO2 (colored), and CO2 with 1 M PhOH added (black). Scans were taken at 0.1 V/s.

reduction potentials are reported in Table S2). The Cl and Br complexes have almost identical CV responses, with the Br complex having a slightly more positive second reduction potential (by 130 mV) than Cl, attributed to the increase in lability of Br over Cl. The neutral ligands, CH3CN and py, result in complexes that are easier to reduce, as the first and second reduction potentials are more positive by ca. 150 and ca. 300 mV, respectively, in comparison to the complexes with Cl or Br. Of note is that the reductions of the CH3CN complex are not at the same potentials as any of the other complexes, ruling out a coordination of CH3CN solvent after a labile ligand loss on the CV time scale. Cyclic Voltammetry under CO2. Each complex was measured under saturated CO2 conditions (0.28 M in CH3CN) to investigate the effect of the 4,4′-substituent and labile ligand on the catalytic current response without an added proton source.30 In the case of the complexes with electrondonating substituents (OCH3, CH3, and tBu), a current enhancement is observed upon the onset of the second reduction potential (Figure 3). Due to this reduction being irreversible under inert conditions, the catalytic potential (E1/2) for each complex was determined from the potential corresponding to the value at half of the catalytic current (Ecat/2, Table 1), which occurs at or near the steepest part of the catalytic wave.28 While this could involve some variation due to nonideal catalytic behavior (i.e., influence from side phenomena), the variation in Ecat/2 is smaller than if the potential from the peak current (icat) was used. The overpotential (η) was determined from the difference between E cat/2 and E°CO2/CO(CH3CN) of −0.541 V vs Fc+/0 for the reaction in Scheme 1, when no explicit proton source is present, as recently described by Savéant and co-workers.29 A rough measure of the catalytic current response can be taken by comparing the peak current under CO2 (icat) versus

that under inert conditions (ip, Table 1). In this work, higher values are measured in comparison to those previously reported for all catalysts,2 likely due to using a freshly polished working electrode for each scan. Electrode fouling, especially when a proton source was used, resulted in decreased catalytic current responses and often a slight shift in catalytic potentials. We found it critical to use a freshly polished working electrode in order to obtain reproducible catalytic voltammograms. Of note is OCH3, which had been previously reported as noncatalytic2 but has the highest icat/ip value in this study (icat/ip = 40.7). Controlled-potential electrolysis (CPE) was run to determine the activity and selectivity of OCH3. Over 1 h (∼3.9 turnovers, where a turnover corresponds to 2 equiv of electrons being passed per mole of catalyst), analysis of the headspace corresponded to Faradaic efficiencies (FE) of 59% ± 5% for CO, where the current decreased throughout the course of the experiment. The complexes tBu, CH3, and bpy-H have all been previously reported as selective CO2-to-CO catalysts, with %FE values for CO close to unity and a steady current over time.2 The catalytic current responses for electron-withdrawing CN and CF3 do not occur until the third reduction of the complexes (Figure 3). This is due to the inability of the doubly reduced species to favorably bind CO2 (vide infra). The icat/ip values are also much lower than those for the complexes with electron-donating substituents (icat/ip: CF3 = 15.0; CN = 8.3). CN was found to have an 18% FE for CO, owing partly to evidence that the cyano groups react with CO2 to form carboxylamides.13 The modest current enhancement at the third reduction for CF3 was investigated via CPE, where analysis of the headspace corresponded to an FE value of 30 ± 2% for CO, with 2% H2 detected over three runs. Similar to the case for CN and OCH3, the current response for CF3 decreased 2023

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis through the course of the experiment (CPE traces are given in the Supporting Information). While the labile ligand affected the reductions of the complexes under Ar, this is not the case under catalytic conditions. The catalytic onset potentials and peak catalytic waves obtained for complexes with Br, py, and CH3CN labile ligands are almost identical to those for tBu (Cl labile ligand), as are the icat/ip values (Figure 2, values in Table S3). This is of note for both the CH3CN and py derivatives, whose respective second reductions are less negative than the onset of the catalytic wave. Determination of kcat. Although consideration of icat/ip is a useful tool in first identifying catalysis by CV, it does not convey intrinsic catalytic properties. This is because when a peak shape is seen in a catalytic wave, “side phenomena” (substrate consumption, deactivation of the catalyst, etc.) are inhibiting a full catalytic response.16,18 An ideal catalytic CV response (when the substrate is in gross excess) is S-shaped, where the forward and return sweeps are identical to each other and the catalytic current is independent of scan rate. For the Re-bpy system, which undergoes a two-electron, two-step EECC mechanism where the first chemical step (reaction with CO2) is limiting (k2 ≫ k1), a catalytic rate constant can be determined from this plateau (S-shaped) current by eq 1:19,31 icat 1 = ip 0.446

RT n′kcat nFυ

Figure 4. Catalytic scan rate dependence studies for the six Re(4,4′-Rbpy)(CO)3Cl complexes. (1)

third reduction of the complex for catalytic turnover indicates a change in mechanism that may not be accounted for in these values. FOWA allowed for the use of the CV scans at 0.1 V/s. One obstacle for FOWA was the fact that a true E1/2 value could not be determined; thus the estimated Ecat/2 was utilized. The values obtained from the two methods were vastly different, ranging orders of magnitude across triplicate experiments for every complex (e.g., bpy-H kcat from an Sshaped wave is 155 s−1, kcat from FOWA is 19711 s−1; the rest of the values can be found in Table S4). Addition of Proton Source. The increase in catalysis due to the addition of a Brønsted acid for Re-bpy species has been well documented since the first report by Hawecker, Lehn, and Ziessel, where an increase in the current was seen when 10% water was added to a DMF solution of the catalyst.4 Water, however, is a weak proton donor (pKa = 38−41 in CH3CN)33 and severely limits the solvent window of CH3CN. We thus first looked to quantify the effects of three different weak Brønsted acids of varying strength on the catalytic behavior and selectivity of tBu: 2,2,2-trifluoroethanol (TFE) with pKa(CH3CN) 35.4, phenol (PhOH) with pKa(CH3CN) 29.1, and acetic acid (AcOH) with pKa(CH3CN) 23.5.33−36 The most favorable of these acids then would be used to study the effects on the other 4,4′-substituted complexes. We optimized the acid concentration by taking CVs at incremental acid additions until the catalytic current no longer increased. It was found that the lower the acid pKa, the lower the acid concentration needed (0.75 M AcOH, 1 M PhOH, and 1.5 M TFE). A catalytic current response of tBu with each acid is seen in Figure 5, where all acids are seen to shift the current response to more positive potentials by ∼170 mV (−2.32 V to ca. − 2.15 V, Table 2). The current response with added TFE or PhOH matched the peak height of that without a proton source (icat/ip(0.1 V/s) ≈ 30), while AcOH gave a slight increase in current (icat/ip = 38.8). CPE experiments revealed that both TFE and PhOH did not change the selectivity of the

where R is the universal gas constant, T is the temperature, F is Faraday’s constant, υ is the scan rate, n is the number of unique electron transfer processes that occur at the electrode per catalyst (n = 2), n′ is the catalyst equivalents required per turnover (n′ = 1), and kcat is the intrinsic catalytic rate constant. In this case, kcat is equal to the maximum turnover frequency (TOFmax, eq 2):32

TOFmax = kcat

(2)

FOWA can also be used to determine kcat if an S-shaped wave cannot be achieved. This method utilizes the current at the onset of the catalytic wave, where “side phenomena” have not yet taken place. For Re-bpy, FOWA is governed by eq 3:16,18,19

( nFRTυ )n′kcat ( RTnF )(E − Ecat/2)⎤⎦

2.24 icat = ip 1 + exp⎡⎣

(3)

kcat is obtained by plotting icat/ip vs 1/(1 + exp[(nF/RT)(E − Ecat/2]), where the slope is defined by 2.24[(RT/nFυ)n′kcat]1/2 (see the Supporting Information for details). We applied both methods to each complex and then compared the resulting kcat values to garner which method would be most applicable to the Re-bpy system. All complexes had a catalytic current that was scan rate dependent at 0.1 V/s, which was used to determine the icat/ip values. To achieve an Sshaped wave, we increased the scan rate until the current response plateaued (Figure 4). The kcat values did not trend with σp. For the complexes with electron-donating substituents, OCH3, tBu, and CH3, kcat was determined to be 4088, 2601, and 3336 s−1, respectively. bpy-H had the lowest kcat value at 155 s−1 while the complexes with electron-withdrawing substituents had larger rate constants at 8606 and 3487 s−1 for CF3 and CN, respectively. The rate constants for CF3 and CN are considered separately (vide infra), as the need for a 2024

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis

preference is again noted in this work, where H+ binding is predicted to be much more favorable than CO2 (the more negative the binding energy value, the more favorable the binding, based on the definition of binding energy reported in the Supporting Information). The binding of hydroxycarbonyl (CO2H) also was considered to garner a more complete comparison. Previous computational and experimental work showed that the CO2H-bound species is the most stable intermediate in the catalytic cycle, making it a fair marker for comparison.37,39 The favorability of binding for all three species (H+, CO2, and CO2H) does not correlate precisely with the σp of the substituent. Just like the complexes’ reduction potentials, the binding energies for CO2 in the electron-donating bpysubstituted cases trend OCH3 < CH3 < tBu < bpy-H. In contrast, tBu is less exergonic than bpy-H when the binding of CO2H is considered. The endergonic binding energy toward CO2 for CN (+6.8 kcal/mol) is congruent with the CV experiments, where there was no catalytic activity at the second reduction potential.

Figure 5. CVs of Re(tBu-bpy)(CO)3Cl (tBu) under Ar and CO2 with different acids added. The return sweep is removed for clarity.

catalyst for CO production; however, the addition of 0.75 M AcOH resulted in 35% CO and 35% H2 production. The determination of kcat was not as straightforward with the added Brønsted acids. For one, we could not use FOWA because the foot of the catalytic wave overlapped with a slight increase in current at the first reduction potential of the catalyst. When the scan rate was increased to achieve an S-shaped wave, the catalytic current still increased through 25 V/s. Faster rates were not measured due to instrument limitations. As such, we used 25 V/s to estimate kcat via eq 1, and we note that this is most likely an underreporting of the intrinsic rate constants under these conditions. Values can be found in Table 2, where the magnitude of kcat correlates with increasing acid strength. Since PhOH gave the largest enhancement in kcat for tBu without affecting selectivity, we used it as the proton source for the remainder of the complexes (Table 3). For the complexes with electron-donating substituents, the addition of 1 M PhOH increases kcat by ∼2-fold and again shifted the catalytic potential to more positive potentials by ca. 200 mV. The kcat of bpy-H increased by 26-fold (155 to 3965 s−1); however, it is still less active than tBu (6206 s−1). CPE experiments of bpy-H and CH3 showed that selectivity for CO production was not affected by the added PhOH. CF3 and CN show a moderate increase in kcat but have a much larger shift in catalytic reduction potential between that of no added acid to 1 M PhOH (+400 mV). CPE runs of CN and CF3 with 1 M PhOH give FE values of 3% CO and 1% H2 for CN and 3% CO and 22% H2 for CF3. Density Functional Theory (DFT) Calculations. Binding energies of relevant ligands to the key catalytic intermediate were determined via DFT to further understand how the 4,4′bpy substituents affect the reactivity of the Re-bpy species toward CO2 (Table 4). Previous studies indicated that the Rebpy catalysts have a thermodynamic preference for the binding of protons over CO2;37,38 however, the catalyst reacts almost 10 times faster with CO2 than with protons.7 This thermodynamic



DISCUSSION

Labile Ligand. The effect of the labile ligand on the spectroscopic properties of Re-bpy systems was first characterized by Kurz and co-workers, where they studied Re(bpy)(CO)3X (X = Cl, Br, H2O, SCN, and CN) with respect to the photochemical reduction of CO2.10 Congruent with their results, the cationic complexes with CH3CN or py as the labile ligand are easier to reduce than the neutral complexes (Cl and Br ligands; reduction potentials are listed in Table S2 in the Supporting Information). The calculated densities corresponding to the singly occupied molecular orbital (SOMO) of the singly reduced bpy-H complex and the highest occupied molecular orbital (HOMO) of the doubly reduced bpy-H complex show that the electron density is distributed respectively first on the bpy ligand and then throughout the complex in the sequential reductions (Figure S7 in the Supporting Information);40 hence the larger changes in second reduction potentials from the change in labile ligand. In all cases, a catalytically active Re0bpy− state is expected after the second reduction, where subsequent CH3CN ligation is not expected to occur under these conditions.41 The identity of the labile ligand does not affect the activity or overpotential of the catalyst. While there were small differences in the icat/ip values between the four complexes, the kcat values were within experimental error of each other (2686 ± 102 s−1). The same was true for η, i.e., the catalytic wave does not begin until ca. −2.10 V vs Fc+/0 even for the py and CH3CN derivatives, where the reduction potential corresponding to ligand loss occurs at more positive potentials (ca. −1.70 V vs

Table 2. Catalytic Descriptors for Re(tBu-bpy)(CO)3Cl (tBu) with Added Brønsted Acid acid amount (M) pKa Ecat/2 (V vs Fc+/0) FECO (%) FEH2 (%) icat/ip (0.1 V/s) kcat (υ, s−1)

none

TFE

PhOH

AcOH

−2.32 89 0 32.1 2601 (10)

1.5 35.4 −2.15 102 0 29.3 4861 (25)

1.0 29.1 −2.11 96 0 31.1 6206 (25)

0.75 23.5 −2.16 35 35 38.8 19861 (25)

2025

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis Table 3. Catalytic Descriptors for Re(4,4′-R-bpy)(CO)3Cl under CO2 with 1 M PhOHa R Ecat/2 (V) ΔEcat/2 (V)b FECO (%) icat/ip kcat (s−1)c Δkcatb

OCH3

tBu

CH3

H

CF3

CN

−1.97 +0.21 70 21.4 5,357 ×1.3

−2.03 +0.17 96 31.1 6,206 ×2.3

−1.98 +0.16 100 33.9 5,869 ×1.8

−1.89 +0.20 100 30.0 3,965 ×25.6

−2.17 +0.40 3 7.0 10,957d ×1.3

−1.91 +0.40 3 4.6 6,244d ×1.8

a

See Table 1 for a definition of the different descriptors. bChange from no added phenol to added phenol. cScan rate is 25 V/s. dCalculated using the current increase at the third reduction.

frequencies of the Re0bpy− anions of bpy-H, CH3, and tBu are all within 1 cm−1 of each other, suggesting a very similar electron density on the metal center.27 Nonetheless, DFT calculations show that the electron density in this highest occupied molecular orbital (HOMO) is spread throughout the molecule, where both the metal and bpy ligand are involved in activating CO2.40 Thus, the availability of the electron density on the bpy ligand is still crucial for the selectivity of these complexes. This is demonstrated in the electron-withdrawing cases of CN and CF3, where the movement of the electron density away from the metal center renders these complexes inactive in their doubly reduced state, quantified by the endergonic binding energy for CO2 (CN +6.8 kcal/mol). Even though CN and CF3 have the most positive first and second reduction potentials, the need for a third electron for catalytic activity leads to the largest overpotentials in this study. Coupled with the change in overpotential is the change in activity and selectivity of the complexes. Of the six bpysubstituted complexes studied in this work, tBu was the most active while maintaining selectivity and durability, closely followed by CH3. Previous reports cite tBu as the most active variant due to its steric bulk, which prevents deactivation by catalyst dimerization.2 Considering the catalytic similarity between tBu and CH3, the electron-donating ability of these substituents is a more compelling reason for the large increase in activity over the unsubstituted bpy-H. However, there can be too much electron density on the bpy ligand. CN, CF3, and OCH3 all show a rapidly decreasing amount of current over time (monitored over 1 h), suggesting that these catalysts quickly deactivate. We ruled out the reductive disproportionation of CO2 to CO and carbonate since it involves a bimetallic mechanism,43 which would not be present for these complexes. It is more likely that the extra electron density on the bpy ligand (due to the third reduction of CN and CF3 to achieve a catalytically active state and the highly electron donating nature of OCH3) is highly destabilizing, leading to a catalyst that quickly degrades over time. Acid Cosubstrate and Mechanism. Re-bpy complexes have a second-order dependence on protons for the overall electrocatalytic reduction of CO2 to CO and H2O.7 Without an explicitly added proton source, the catalyst must scavenge protons from “adventitious water” in the solution, electrolyte (TBAPF6), or CH3CN solvent.19,33 When a proton source is explicitly added, the strength of the proton donor affects the selectivity and activity of the catalyst. Our DFT studies reiterate past computational and experimental work showing that H+ binding to the catalytically active state is more thermodynamically favorable; however, CO2 reacts much faster,7 lending to the product selectivity of these complexes. Stronger proton donors would result in H+ favorability,37 although the presence

Table 4. Thermodynamic Binding Energies (ΔG, kcal/mol) to the Re0bpy− Catalytically Active State for Five of the 4,4′bpy Substituted Complexes R

ΔG(H+)

ΔG(CO2)

ΔG(CO2H)

OCH3 tBu CH3 H CN

−46.2 −41.8 −43.8 −39.4 −24.5

−12.8 −7.2 −8.3 −4.6 +6.8

−40.1 −36.8 −37.9 −37.1 −18.6

Fc+/0). This illustrates that ligand loss does not determine the catalytic potential. This is in contrast to Re-bpy-based photocatalytic reactions, where the loss of the labile ligand at the first reduction is key to reactivity.42 4,4′-bpy Substituents. The substituents on the bpy ligand greatly affect the electronic distribution in the Re-bpy complex, which in turn impacts the overpotential, activity, and selectivity of the catalyst. The change in overpotential is apparent not just from catalysis but also from a look at the CVs under noncatalytic conditions. Under Ar, the first reduction potentials of the 4,4′-substituted Re-bpy complexes trend almost linearly with the values of the para-substituted Hammett parameter, OCH3 > CH3 > tBu > bpy-H > CF3 > CN (Figure 6), and

Figure 6. Trend between the electron-donating character of the 4,4′bpy substituent and the first reduction potential of the corresponding Re complex.

span 700 mV from OCH3 to CN. This trend shows the huge effect that the bpy ligand’s electronics has on the reductions of the resulting Re complex. The linear relationship between σp and the reduction potential can be used to predict the reduction potentials of other complexes with nonreactive 4,4′substituents. However, we find that reactive substituents, such as those that can hydrogen bond, do break this trend.43 The second reduction potential and corresponding Ecat/2 values are less affected by the 4,4′-substituent. Previous structural studies have shown that the carbonyl stretching 2026

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis

“reduction first” mechanism (B → C). The computationally determined catalytic potentials for this step for bpy-H match within expected error to our experimental results (experimental −2.09 V vs computational −2.25 V vs Fc+/0).36 When an explicit proton donor is added, a “proton first” mechanism can be achieved, where protonation of the bound CO2 can occur before further reduction of the complex (B → D). This results in a lower overpotential for catalysis by about 200 mV, and again our experimental results match the computational prediction for bpy-H (experimental with 1 M PhOH −1.89 V vs computational −1.94 V vs Fc+/0).36 Comparing Catalysis. The intrinsic rate constants (kcat) for the 4,4′-substituted complexes were calculated by both achieving an S-shaped wave by increasing the scan rate and using FOWA. Reproducible values were only obtained from the former method, whereas FOWA gave widely varying values across triplicate experiments. While FOWA is ideal for cutting out “side phenomena” affecting the catalytic current response, in the case of Re-bpy complexes, there are other factors that render FOWA an inapplicable method to determine kcat. These reasons include having to estimate E°cat as well as the close proximity of the two reductions to each other, where there is also a slight increase in the current at the first reduction potential. These reasons were even more applicable when a proton source was added to the system, as there was a distinct overlap between the two reductive features in the CV. When the scan rate method to achieve a plateau current is considered, a perfect S-shaped wave could not be achieved for any complex, as the return wave never matched the forward wave in the CV. Instead, curve-crossing is observed for OCH3, tBu, and CH3, indicative of other chemical reactions (side phenomena) occurring around the electrochemical time scale. When PhOH was added, a plateau current could not be achieved within the scan rate limits of our potentiostat. The kcat values obtained from changes in scan rate are most likely underreporting catalysis, while those from FOWA overreport catalysis. We note the difficulty in using CV for determining kcat when most complexes, including Re-bpy, do not display ideal electrochemical responses. The instability of OCH3, CN, and CF3 also calls into question the use of kcat as a means of describing catalysis when the complexes themselves are not stable. However, there is a need across the literature for more standard and rigorous approaches to reporting catalysis. The values determined in this work were difficult to compare to previously reported rate constants for Re-bpy complexes, due to the variety of methods and values reported. Nevertheless, the kcat equivalent value determined by stopped-flow UV−vis spectroscopy for the chemically reduced [Re(tBu-bpy)(CO)3]− with no added acid was 2800 s−1,7 which is within range of what was calculated in this work using the plateau current and eq 1 (2601 s−1). Our values are larger than those reported previously by CV (131 s−1),2 where scan-rate-independent currents had not been used to determine kcat. More rigorous tandem comparisons between spectroscopic studies and CV approaches are needed to verify the use of plateau currents and eq 1 for determining Re-bpy catalytic rate constants. Catalytic Tafel plots can be used to benchmark these catalysts against other molecular catalysts, which depict the relationship between a catalyst’s defining thermodynamic and kinetic parameters: η and TOFmax. Plots are constructed using eq 4, where E°CO2/CO(CH3CN) is the standard reduction potential for the catalytic reduction of CO2 to CO with added Brønsted

of a proton donor helps to achieve the stable intermediate of the catalytic cycle, Re-CO2H. Thus, a balance must be found between a proton donor strong enough to facilitate protonation of a bound CO2 but weak enough to not shift to the favorability of catalyst protonation.36 The effect of an added proton source was first investigated with [Re(bpy)(CO)3(py)]OTf, where TFE, MeOH, PhOH, and water were compared.30 In agreement with this latter work, we found that the efficiency of the acid follows the acidity: i.e., the lower the pKa, the higher the catalytic current. As verified with tBu, PhOH and TFE did not affect the selectivity of the catalyst, but AcOH changed the selectivity to higher H2 production. The loss of selectivity for the CO2 reduction toward proton reduction in the presence of AcOH has also been observed with FeTPP.44 The addition of a proton donor also affects the catalytic mechanism, as observed from the consistent shift of the catalytic potential by +200 mV that is independent of the proton source, labile ligand, and bpy substituent. As described by previous studies of the Re-bpy mechanism, once a CO2 molecule is activated on the metal center (Scheme 2, B), Scheme 2. CO2 Reduction Catalytic Cycle for [Re(Rbpy)(CO)3]− Active Catalyst, Highlighting the ProtonDependent Pathwaysa

a

Potentials correspond to Re(bpy)(CO)3Cl, where those shown in black are the computational values from ref 36 and those in green are the experimental values.

another proton and electron are needed in order for the ratelimiting step to occur (breakage of a C−OH bond coupled with a proton addition to create H2O and bound CO, E → F).3,37,38 The order of the proton and electron additions at this step in the mechanism has not been previously probed experimentally; however, this question was addressed computationally in 2015 by Riplinger and Carter.36 They found that, without an explicitly added proton source, additional driving force was needed to abstract protons from the solvent, CH3CN. This is manifested in both a higher overpotential for catalysis and a 2027

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis acid (−1.36 V vs Fc+/0).29 An ideal catalyst would have a low η and high TOFmax, which would fall on the upper-left-hand side of a catalytic Tafel plot.

by changing the order of protonation and reduction following CO2 activation in the catalytic cycle, as predicted in earlier work.36 Utilizing more rigorous approaches to determine the catalytic rate constants from CV allowed for a more accurate and direct comparison of Re-bpy-based complexes to other molecular CO2 reduction electrocatalysts. Both FOWA and the use of Sshaped catalytic waves had not been utilized previously for Rebpy, and the results show increased rates over previous CV studies. While the kcat values obtained from FOWA were inconsistent, the use of increased scan rates to obtain a pseudoplateau current gave more reliable kcat values that were in agreement with those obtained for tBu by stopped-flow UV−vis spectroscopy. The effects of added Brønsted acid and Re-bpy catalysts with electron-donating substituents had not been investigated previously, and it was found that CH3 is very similar to tBu in activity. This study highlights the importance of considering the effects on the electronic communication throughout a molecule when making structural changes to a molecular electrocatalyst. The ease of transferring electron density between the ligand and metal center is of paramount importance, which should be the first factor considered when modifying or designing multielectron catalysts. The interplay between the overpotential and activity is also a focal point, where changes that increase activity often also increase overpotential. These fundamental principles should be considered in the design of more effective electrocatalysts for CO2 reduction, which are critical in helping to mitigate the negative effects of climate change due to anthropogenic CO2 emissions.

TOF = 1+

TOFmax F exp⎡⎣ RT (E°CO2/CO(CH3CN) −

F Ecat/2)⎤⎦ exp − RT η

(

) (4)

In Figure 7, we compare selective molecular CO2 to CO electrocatalysts with added Brønsted acid.24,45−47 The Re-bpy

Figure 7. Catalytic Tafel plot comparing catalysts in this work with 1 M PhOH with other molecular CO2 electrocatalysts under similar conditions. Values and conditions are given in Supporting Information. The legend is in the order of decreasing log TOF (top to bottom).

derivatives are average in terms of both η and TOFmax in comparison to FeTPP29,32 and Ni(cyclam).48 There is an obvious direct relationship between η and TOFmax for all CO2 to CO catalysts, i.e., increasing TOF also increases η. Among the stable and selective Re-bpy derivatives, tBu (red) and CH3 (green) have the highest TOFmax (kcat) but also the highest η. The situation is inverse for bpy-H (purple), which has the lowest η of the complexes but also the lowest TOFmax. In comparison with Mn-bpy derivatives, Mn(tBu-bpy)(CO)3Br with 1.4 M TFE (pink) is less active than tBu under the same conditions (TOFmax = 340 s−1 for Mn and 4,861 s−1 for Re) but at a lower overpotential by 190 mV.24 Conversely, [Mn(6,6′dimesityl-bpy)(CO)3(CH3CN)](OTf) (Mn(mes-bpy), black), in which the mesityl groups prevent dimerization of the catalyst, has in a significantly higher TOFmax (5000 s−1) but a larger η that nearly matches that of tBu.45 Of the molecular catalysts that have reported TOFmax values, Ni(cyclam) (blue) and FeTPP (teal) have the lowest η and highest TOFmax of reported molecular catalysts, respectively.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acscatal.7b03971. Experimental section, additional CPE and CV figures, description of foot of the wave workup, DFT calculation details, and coordinates and energies of DFT-predicted structures (PDF)



AUTHOR INFORMATION

Corresponding Author

*E-mail for C.P.K.: [email protected]. ORCID

Melissa L. Clark: 0000-0002-6369-6047 Emily A. Carter: 0000-0001-7330-7554 Clifford P. Kubiak: 0000-0003-2186-488X



CONCLUSIONS In summary, a systematic study of the effect of the labile ligand, 4,4′-bpy substituent, and Brønsted acid on Re-bpy catalysis was conducted. The labile ligands moved the reduction potentials of Re(tBu-bpy) more positively (py > CH3CN > Br > Cl) but did not influence the catalytic activity or η. The 4,4′-bpy substituents affected both the η and activity for catalysis, with bpy-H having the lowest overpotential and tBu having the highest TOFmax. CN and CF3 were not active at their second reduction potential, as described by the DFT-predicted endergonic binding energies for CO2, and therefore had the highest overpotentials. CN, CF3, and OCH3 were all unstable catalysts, suggesting that too much electron density stored on the bpy ligand is not ideal. Finally, the addition of Brønsted acids increased activity and lowered η by 200 mV, presumably

Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors thank Dr. Matthew Sampson, Dr. Mark Reineke, and Dr. Christoph Riplinger for insight and helpful discussions and Ms. Nari L. Baughman for helpful feedback during manuscript preparation. This work was supported in part by the Air Force Office of Scientific Research through the MURI program under AFOSR Award No. FA9550-10-1-0572.



REFERENCES

(1) Dry, M. E. Catal. Today 2002, 71, 227−241. (2) Smieja, J. M.; Kubiak, C. P. Inorg. Chem. 2010, 49, 9283−9289.

2028

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029

Research Article

ACS Catalysis

(37) Keith, J. A.; Grice, K. A.; Kubiak, C. P.; Carter, E. A. J. Am. Chem. Soc. 2013, 135, 15823−15829. (38) Riplinger, C.; Sampson, M. D.; Ritzmann, A. M.; Kubiak, C. P.; Carter, E. A. J. Am. Chem. Soc. 2014, 136, 16285−16298. (39) Sampson, M. D.; Froehlich, J. D.; Smieja, J. M.; Benson, E. E.; Sharp, I. D.; Kubiak, C. P. Energy Environ. Sci. 2013, 6, 3748−3755. (40) Benson, E. E.; Sampson, M. D.; Grice, K. A.; Smieja, J. M.; Froehlich, J. D.; Friebel, D.; Keith, J. A.; Carter, E. A.; Nilsson, A.; Kubiak, C. P. Angew. Chem., Int. Ed. 2013, 52, 4841−4844. (41) Christensen, P.; Hamnett, A.; Muir, A. V. G.; Timney, J. A. J. Chem. Soc., Dalton Trans. 1992, 1455−1463. (42) Takeda, H.; Ishitani, O. Coord. Chem. Rev. 2010, 254, 346−354. (43) Machan, C. W.; Chabolla, S. A.; Yin, J.; Gilson, M. K.; Tezcan, F. A.; Kubiak, C. P. J. Am. Chem. Soc. 2014, 136, 14598−14607. (44) Costentin, C.; Robert, M.; Saveant, J. M. Acc. Chem. Res. 2015, 48, 2996−3006. (45) Sampson, M. D.; Nguyen, A. D.; Grice, K. A.; Moore, C. E.; Rheingold, A. L.; Kubiak, C. P. J. Am. Chem. Soc. 2014, 136, 5460− 5471. (46) Costentin, C.; Passard, G.; Robert, M.; Saveant, J. M. Proc. Natl. Acad. Sci. U. S. A. 2014, 111, 14990−14994. (47) Costentin, C.; Robert, M.; Savéant, J.-M. Chem. Soc. Rev. 2013, 42, 2423−2436. (48) Froehlich, J. D.; Kubiak, C. P. Inorg. Chem. 2012, 51, 3932− 3934.

(3) Grice, K. A.; Kubiak, C. P.; Michele, A.; Rudi van, E. Adv. Inorg. Chem. 2014, 66, 163−188. (4) Hawecker, J.; Lehn, J.-M.; Ziessel, R. J. Chem. Soc., Chem. Commun. 1984, 328−330. (5) Hawecker, J.; Lehn, J.-M.; Ziessel, R. Helv. Chim. Acta 1986, 69, 1990−2012. (6) Elgrishi, N.; Chambers, M. B.; Wang, X.; Fontecave, M. Chem. Soc. Rev. 2017, 46, 761−796. (7) Smieja, J. M.; Benson, E. E.; Kumar, B.; Grice, K. A.; Seu, C. S.; Miller, A. J.; Mayer, J. M.; Kubiak, C. P. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 15646−15650. (8) Hawecker, J.; Lehn, J.-M.; Ziessel, R. J. Chem. Soc., Chem. Commun. 1983, 536−538. (9) Agarwal, J.; Fujita, E.; Schaefer, H. F., 3rd; Muckerman, J. T. J. Am. Chem. Soc. 2012, 134, 5180−5186. (10) Kurz, P.; Probst, B.; Spingler, B.; Alberto, R. Eur. J. Inorg. Chem. 2006, 2006, 2966−2974. (11) Cabrera, C. R.; Abruña, H. D. J. Electroanal. Chem. Interfacial Electrochem. 1986, 209, 101−107. (12) O’Toole, T. R.; Sullivan, B. P.; Bruce, M. R. M.; Margerum, L. D.; Murray, R. W.; Meyer, T. J. J. Electroanal. Chem. Interfacial Electrochem. 1989, 259, 217−239. (13) Clark, M. L.; Rudshteyn, B.; Ge, A.; Chabolla, S. A.; Machan, C. W.; Psciuk, B. T.; Song, J.; Canzi, G.; Lian, T.; Batista, V. S.; Kubiak, C. P. J. Phys. Chem. C 2016, 120, 1657−1665. (14) Anfuso, C. L.; Xiao, D. Q.; Ricks, A. M.; Negre, C. F. A.; Batista, V. S.; Lian, T. Q. J. Phys. Chem. C 2012, 116, 24107−24114. (15) Bligaard, T.; Bullock, R. M.; Campbell, C. T.; Chen, J. G.; Gates, B. C.; Gorte, R. J.; Jones, C. W.; Jones, W. D.; Kitchin, J. R.; Scott, S. L. ACS Catal. 2016, 6, 2590−2602. (16) Costentin, C.; Drouet, S.; Robert, M.; Savéant, J.-M. J. Am. Chem. Soc. 2012, 134, 11235−11242. (17) Savéant, J.-M. ChemElectroChem 2016, 3, 1967−1977. (18) Costentin, C.; Drouet, S.; Robert, M.; Savéant, J.-M. J. Am. Chem. Soc. 2012, 134, 19949−19950. (19) Rountree, E. S.; McCarthy, B. D.; Eisenhart, T. T.; Dempsey, J. L. Inorg. Chem. 2014, 53, 9983−10002. (20) Costentin, C.; Drouet, S.; Robert, M.; Saveant, J. M. Science 2012, 338, 90−94. (21) Wasylenko, D. J.; Rodriguez, C.; Pegis, M. L.; Mayer, J. M. J. Am. Chem. Soc. 2014, 136, 12544−12547. (22) Costentin, C.; Robert, M.; Savéant, J.-M.; Tatin, A. Proc. Natl. Acad. Sci. U. S. A. 2015, 112, 6882−6886. (23) Hansch, C.; Leo, A.; Taft, R. W. Chem. Rev. 1991, 91, 165−195. (24) Smieja, J. M.; Sampson, M. D.; Grice, K. A.; Benson, E. E.; Froehlich, J. D.; Kubiak, C. P. Inorg. Chem. 2013, 52, 2484−2491. (25) Johnson, F. P. A.; George, M. W.; Hartl, F.; Turner, J. J. Organometallics 1996, 15, 3374−3387. (26) Machan, C. W.; Sampson, M. D.; Chabolla, S. A.; Dang, T.; Kubiak, C. P. Organometallics 2014, 33, 4550−4559. (27) Benson, E. E.; Grice, K. A.; Smieja, J. M.; Kubiak, C. P. Polyhedron 2013, 58, 229−234. (28) Appel, A. M.; Helm, M. L. ACS Catal. 2014, 4, 630−633. (29) Azcarate, I.; Costentin, C.; Robert, M.; Savéant, J.-M. J. Am. Chem. Soc. 2016, 138, 16639−16644. (30) Wong, K.-Y.; Chung, W.-H.; Lau, C.-P. J. Electroanal. Chem. 1998, 453, 161−170. (31) Costentin, C.; Savéant, J.-M. ChemElectroChem 2014, 1, 1226− 1236. (32) Azcarate, I.; Costentin, C.; Robert, M.; Savéant, J.-M. J. Phys. Chem. C 2016, 120, 28951−28960. (33) McCarthy, B. D.; Martin, D. J.; Rountree, E. S.; Ullman, A. C.; Dempsey, J. L. Inorg. Chem. 2014, 53, 8350−8361. (34) Ngo, K. T.; McKinnon, M.; Mahanti, B.; Narayanan, R.; Grills, D. C.; Ertem, M. Z.; Rochford, J. J. Am. Chem. Soc. 2017, 139, 2604− 2618. (35) Lam, Y. C.; Nielsen, R. J.; Gray, H. B.; Goddard, W. A. ACS Catal. 2015, 5, 2521−2528. (36) Riplinger, C.; Carter, E. A. ACS Catal. 2015, 5, 900−908. 2029

DOI: 10.1021/acscatal.7b03971 ACS Catal. 2018, 8, 2021−2029