Tuning Product Selectivity for Aqueous CO2 Reduction with a Mn(bipyridine)-pyrene Catalyst Immobilized on a Carbon Nanotube Electrode

The development of high-performance electrocatalytic systems for the controlled reduction of CO2 to value-added chemicals is a key goal in emerging renewable energy technologies. The lack of selective and scalable catalysts in aqueous solution currently hampers the implementation of such a process. Here, the assembly of a [MnBr(2,2′-bipyridine)(CO)3] complex anchored to a carbon nanotube electrode via a pyrene unit is reported. Immobilization of the molecular catalyst allows electrocatalytic reduction of CO2 under fully aqueous conditions with a catalytic onset overpotential of η = 360 mV, and controlled potential electrolysis generated more than 1000 turnovers at η = 550 mV. The product selectivity can be tuned by alteration of the catalyst loading on the nanotube surface. CO was observed as the main product at high catalyst loadings, whereas formate was the dominant CO2 reduction product at low catalyst loadings. Using UV–vis and surface-sensitive IR spectroelectrochemical techniques, two different intermediates were identified as responsible for the change in selectivity of the heterogenized Mn catalyst. The formation of a dimeric Mn0 species at higher surface loading was shown to preferentially lead to CO formation, whereas at lower surface loading the electrochemical generation of a monomeric Mn-hydride is suggested to greatly enhance the production of formate. These results emphasize the advantages of integrating molecular catalysts onto electrode surfaces for enhancing catalytic activity while allowing excellent control and a deeper understanding of the catalytic mechanisms.


■ INTRODUCTION
Efficient, low-cost, and scalable electrocatalytic reduction of CO 2 is currently under consideration as a viable means to produce useful chemicals while limiting the rise of CO 2 levels in the atmosphere. 1−5 During the past few years, transition metal complexes have been extensively studied for the electro-and photoreduction of CO 2 , mainly to carbon monoxide (CO) and formate (HCOO − ). 6−9 The molecular nature of these complexes allows specific fine-tuning of their structure using synthetic chemistry to rationally control the activity, 10 selectivity, 11 and stability of the catalysts via a profound understanding of the catalytic mechanisms involved. 12,13 In order to replace expensive Re, 14−17 Ru, 18,19 and Ir 20 based catalysts, a range of different catalytic structures incorporating more abundant first-row transition metals such as Ni, 21−23 Co, 24−27 Fe, 28,29 and Mn 30−32 have been described for catalytic CO 2 reduction.
The integration of these molecular species onto electrodes in many cases gives an enhancement of the catalytic activity and provides new insights into the involved mechanisms. 33−36 Immobilization also enables catalysts that would otherwise be insoluble to operate in water and overcomes limitations from diffusion-controlled electrocatalysis with catalysts in the bulk solution. 37−41 In this respect, carbon nanotubes (CNTs) are an established platform for the immobilization of molecular electrocatalysts. 42−45 Their high surface area and excellent conductivity allow grafting of large amounts of electrocatalytically active species while retaining good electron transfer properties. 46 The possibility to chemically modify CNTs using a variety of covalent or noncovalent techniques is another advantage of these nano-objects. 47 Specifically, the immobilization of pyrene-modified catalysts via π−π interactions with the CNT sidewalls has emerged as a straightforward strategy. 48−51 Using this technique toward efficient electrochemical CO 2 reduction, recent examples have described the immobilization of Re, 52 Ir, 53 and Fe 54 based catalysts onto CNT matrixes, the latter two examples affording catalytic activity in aqueous solution.
The complex [MnBr(bpy)(CO) 3 ] (bpy = 2,2′-bipyridine) has become a noble metal-free model catalyst for CO 2 reduction thanks to the versatile and simple structure of the bpy ligand. 30 This catalyst exhibits high activity (turnover frequency up to 480 s −1 ) in MeCN, and its catalytic mechanism has been studied in depth by changing the nature of the substituents on the bpy ligand. 12,13,55−61 The Mn catalyst was previously integrated onto CNTs using Nafion, 39,62 onto p-Si via polymerization, 63 and onto TiO 2 through a phosphonate anchoring group. 33 The latter system exhibited a record turnover number (TON) for the Mn catalyst of 112 in MeCN. 33 UV−vis spectroelectrochemistry (SEC) revealed that the grafted Mn catalyst dimerized on the electrode surface, 33 which has also been described in solution upon electrochemical reduction for this class of catalyst. 12 Although these studies showed enhancement of the catalytic properties of the Mn complex upon immobilization, the reported activity is still limited to organic solvents and low TONs (maximum of 101) in aqueous conditions. 39,62 In this work, a novel fac-[MnBr(bpy)(CO) 3 ]-type complex bearing a pyrene anchoring unit was synthesized (Mn pyr , Schemes 1 and 2). The pyrene unit allowed stable immobilization onto CNTs. The electrocatalytic activity of the compound was first studied toward the reduction of CO 2 in homogeneous organic solution (MeCN + 5% H 2 O) and then in fully aqueous solutions after heterogenization on the CNT surface. The Mn catalyst modified electrodes were investigated using cyclic voltammetry (CV) and controlled-potential electrolysis (CPE). The surface loading of the Mn complex was found to have a distinct effect on the selectivity toward CO or HCOO − production (Scheme 1). Using transmission UV− vis and surface-sensitive IR SEC in the attenuated total reflection (ATR) mode, the different catalytic intermediates involved were investigated in situ. Upon decreasing the surface loading, the previously described dimer formation was lowered. Instead, formation of a putative hydride species was detected. The formation of one or the other catalytic intermediate is assigned to be decisive for selectivity toward either CO or HCOO − at different surface loadings. 55,61,64 ■ RESULTS AND DISCUSSION Synthesis and Characterization of Mn pyr . 4-Methyl-4′-(5-(pyren-1-yl)pentyl)-2,2′-bipyridine (bpy pyr ) was synthesized by reacting a lithiated 4,4′-dimethyl-bipyridine intermediate with 1-(4-bromobutyl)pyrene, as previously described (Scheme 2). 65 The ligand bpy pyr was then refluxed with [MnBr(CO) 5 ] in dry Et 2 O for 4 h to yield fac-[MnBr(bpy pyr )(CO) 3 ] (Mn pyr ) with an overall yield of 64% (Scheme 2). The composition and purity of Mn pyr were confirmed by 1 H, 13 C NMR, and FT-IR spectroscopy as well as mass spectrometry and elemental analysis (see the Experimental Section for details).
The electrochemical properties of Mn pyr were investigated by CV in MeCN containing TBAPF 6 (0.1 M, TBA = tetrabutylammonium) at a glassy carbon working electrode (Figure 1a,b). In dry MeCN and under a N 2 atmosphere, the Mn I complex exhibits two successive irreversible reduction processes at red E p = −1.68 and −1.86 V vs Fc + /Fc (see below for assignment of the reversible redox process at −2.48 V). According to literature assignments, 30 the first cathodic wave ( red,1 E p ) is attributed to the reduction to Mn 0 , followed by dimerization and formation of a Mn 0 −Mn 0 species. The second cathodic process ( red,2 E p ) is assigned to the one-electron reduction of the dimer to a mononuclear anionic species (Figure 1a). 30 The redox wave ( red,3 E 1/2 ) observed at −2.48 V vs Fc + /Fc is attributed to the reversible redox response of the bpy ligand. 30 Two anodic waves were observed following scan Immobilization and Performance of Mn pyr . Immobilization of Mn pyr onto a multiwalled carbon nanotube (MWCNT) electrode enabled the study of its electrocatalytic activity in fully aqueous solution. The Mn complex was anchored onto MWCNTs following an established two-step procedure: 51,66 First, a dispersion of MWCNTs in N-methylpyrrolidone (20 μL, 5 mg CNT mL −1 ) was drop-cast and dried on a glassy carbon disk electrode (diameter ø = 5 mm), giving a MWCNT electrode with 5 μm thickness. The catalyst was subsequently immobilized by immersing the MWCNT electrode into an anhydrous DMF solution of Mn pyr (10 mM) for 30 min in the dark and at room temperature. The resulting CNT|Mn pyr electrode was subsequently rinsed successively with DMF and water before CV experiments were undertaken (see below for more characterization details of the Mn-loaded MWCNT electrode).
CV scans recorded with CNT|Mn pyr in aqueous KHCO 3 (0.5 M, pH 8.2) electrolyte solution under N 2 show an irreversible cathodic wave at red E p = −1.0 V vs SHE ("shoulder" of solid black trace in Figure 2), which was assigned to the one-electron reduction of Mn I to Mn 0 . This Mn reduction wave overlaps with a dominant cathodic current that is mainly attributed to proton reduction ( Figure 2, see below for more details). Following scan reversal, an irreversible anodic process, assigned to the reoxidation of the Mn 0 species, 39 can be observed at ox E p = −0.24 V vs SHE. In the presence of CO 2 (pH 7.4), the current observed at red E p is enhanced by a catalytic wave from reduction of CO 2 . The observed onset potential (E cat = −0.91 V vs SHE) with the immobilized Mn pyr complex corresponds to a small overpotential (η = 360 mV for CO and/or η = 390 mV for formate production). 67 It is noteworthy that the irreversible anodic process at ox E p is still present in the presence of CO 2 but displays a decrease in peak current. This observation may highlight that some of the immobilized and electrochemically reduced Mn pyr does not readily react with CO 2 and is reoxidized at the electrode surface instead, which could possibly be due to limited diffusion of CO 2 within the CNT film. The bare (Mn pyr -free) MWCNT electrode does not show notable cathodic current enhancement in the presence of CO 2 , confirming that the current enhancement observed with the CNT|Mn pyr electrode solely arises from the presence of the catalyst on the electrode surface.
To confirm the formation of the dimeric Mn 0 species on the electrode surface upon reduction at red E p , transmission UV−vis SEC was carried out using Mn pyr deposited onto MWCNT films on a transparent FTO glass substrate (see the Experimental Section for details). CV of CNT|Mn pyr on FTO-coated glass was carried out to confirm the immobilization of the complex ( Figure S1), and UV−vis absorption spectra of the electrodes without and with applied potential (E appl = −1.1 V vs SHE) were compared in aqueous electrolyte solution ( Figure 2, inset). Two bands appear at 655 and 850 nm under negative applied bias, which are comparable to the bands described previously for the dimer in solution and on a TiO 2 electrode in MeCN. 30, 33 The UV−vis SEC analysis therefore supports generation of the Mn 0 −Mn 0 dimer at the surface of the MWCNT electrode upon reduction at red E p .
To gain more insight into the overall activity, selectivity, and stability of the CNT|Mn pyr electrode, CPE was performed at E appl = −1.1 V vs SHE under CO 2 and the electroreduction products were analyzed using gas and ion chromatography ( Figure 3). The CNT|Mn pyr electrodes showed high current densities between −5 and −1.5 mA cm −2 during the first hour of CPE before stabilizing at approximately −0.5 mA cm −2 after 8 h ( Figure S2), which coincides with a decrease in CO production activity (Figure 3a). Indeed, the electrodes showed good activity for CO production during the first hour, before stabilizing at 46 ± 11 μmol cm −2 CO. This decrease in CO production can be attributed to deactivation of the immobilized catalyst as it is commonly observed for molecular catalysts. 68 On the other hand, H 2 evolution remained stable over time and

Journal of the American Chemical Society
Article reached a value of 108 ± 17 μmol cm −2 after 8 h. Some HCOO − was also detected at the end of CPE measurements (16 ± 5 μmol cm −2 ).
Control CPE measurements on a bare (Mn-free) MWCNT electrode under CO 2 showed only small current densities of −0.3 mA cm −2 , with linear production of H 2 (26 μmol cm −2 after 8 h) and only traces of CO ( Figure S3a). In addition, no CO was observed for a CNT|Mn pyr electrode under N 2 , whereas H 2 production was higher than that with the bare MWCNT electrode (47.2 μmol cm −2 after 8 h; Figure S3b). This observation confirms that the CO production solely arises from Mn pyr on the electrode and in the presence of dissolved CO 2 in the electrolyte solution, whereas some H 2 is produced by the MWCNT scaffold under reductive conditions as well as by Mn pyr . The high current densities and CO evolution rates during the first hours of CPE translated into a reasonable Faradaic yield for CO (Figure 3c), with a maximum yield of (34 ± 4)% being obtained after 1 h of CPE, before decaying to (25 ± 2)% cumulated after 8 h. In contrast, the efficiency for H 2 slightly increases over time and Faradaic yields of (45 ± 8) and (59 ± 8)% were observed after 1 and 8 h, respectively. For formate, a Faradaic yield of (8 ± 2)% was determined after 8 h, which results in a close-to-quantitative overall yield of (93 ± 4)% for all detected products.
The carbon source for the generated CO and HCOO − was confirmed by gas-phase IR spectroscopy and 1 H NMR spectroscopy, respectively. CPE under 13 CO 2 resulted exclusively in 13 CO and H 13 COO − ( Figure S4). A TON Mn of 1400 ± 230 was obtained for CO ( Figure 3d), which is more than 10 times higher than previous results obtained electrochemically 30 or photochemically 59 using Mn(bpy)-type catalysts in organic 33 or aqueous 39,62 solution. Thus, the CNT| Mn pyr electrode displays high activity for catalytic CO 2 to CO reduction in aqueous solution and compares well with noble metal-free solid-state electrocatalysts in terms of the TON at a low overpotential. 37,38,54,69,70 A respectable TON Mn of 460 ± 120 was obtained for HCOO − production.
In order to verify the importance of immobilization and the molecular nature of the catalytically active species, a series of control experiments were carried out using anchor-free fac-  Figure S5d). These results support that the high activity observed for CNT| Mn pyr is not caused by possible degradation products of the molecular Mn catalyst. In addition, CV scans performed on CNT|Mn pyr after CPE showed the appearance (although at reduced intensity) of identical redox behavior compared to preelectrolysis measurements ( Figure S6). The presence of the irreversible peak at ox E p at −0.24 V vs SHE (due to dimer reoxidation in aqueous conditions 39 ) and the catalytic wave at E cat = −0.91 V vs SHE confirm that at least some of the molecular species remains on the surface throughout the CPE experiment without clearly observable degradation.
Tuning of Selectivity by Varying the Surface Loading. To study the dimerization of the heterogenized Mn pyr catalyst in more detail, the influence of catalyst surface loading was investigated by incubating the MWCNT electrodes in solutions containing different concentrations of Mn pyr . The electrochemical behavior of the electrodes with different surface loadings of Mn pyr was subsequently examined. From integration of the oxidation wave in CV, the amount of electroactive Mn pyr immobilized on the electrode can be estimated. The surface loading and the position of the oxidation peak of Mn pyr immobilized onto MWCNTs was shown to be dependent on the Mn pyr incubation concentration (Figure 4).
At higher surface loading, above 30 nmol cm −2 from incubation with 10 and 20 mM Mn pyr , the dimer reoxidation peak was detected at ox E p = −0.24 V vs SHE, close to the previously reported value. 39 At lower loading, below 20 nmol cm −2 from incubation with 0.5 and 1 mM Mn pyr , a small irreversible reduction wave was observed at red E p = −1.0 V vs SHE and the corresponding reoxidation wave was shifted considerably to ox E p = −0.34 V vs SHE (100 mV more cathodic than the response at higher surface loadings). At intermediate surface loadings, between 20 and 30 nmol cm −2 from incubation with 2.5 and 5 mM Mn pyr , a broadened oxidation wave was attributed to a convoluted response from both anodic processes. For all Mn pyr surface loadings, a strong reduction wave followed the reduction of the Mn I to Mn 0 ( red E p = −1.0 V vs SHE), which is assigned to the reduction of protons to H 2 under a N 2 atmosphere (see above). This catalytic reduction Γ Mn increased with the concentration of the incubation solution until reaching a plateau between 5 and 10 mM, following a Langmuir isotherm (see the Experimental Section and Figure 4b). 66,71 The best fit (using Mn concentrations, [Mn], up to 10 mM) was obtained with a saturation surface concentration, Γ Mn,max , of 36.5 nmol cm −2 and with a K Mn of 950 L mol −1 , in line with previously reported values for pyrenebearing coordination complexes. 66 The surface loading after incubation with the highest Mn pyr concentration (20 mM) does not align with the fit, which can be rationalized by oversaturation of the surface at higher incubation concentrations. At Γ Mn,max , the probability for Mn pyr to dimerize on the surface during electroreduction is high. On the other hand, for lower Γ Mn , the probability of two Mn pyr units being close enough to each other to form a dimer is substantially decreased. The altered surface loading therefore gives a possible explanation for the different electrochemical response of Mn pyr . At lower surface concentration, impeded dimerization would lead to the formation of another reduced species and the oxidation peak observed at ox E p = −0.33 V vs SHE could possibly be attributed to the one-electron oxidation of a single Mn 0 complex back to the Mn I complex.
As the CV response of Mn pyr changed as a function of catalyst loading, the possible impact of Γ Mn on the catalytic activity and selectivity was studied by CPE coupled to H 2 , CO, and formate analysis. Figure 5 summarizes the results and shows clear trends (see Figures S7 and S8 for the complete data set). Lowering the surface loading of Mn pyr on MWCNTs resulted in a decrease in overall CO production from 62 ± 15 μmol cm −2 at high Γ Mn (50 ± 3 nmol cm −2 ) to 26 ± 5 μmol cm −2 at low Γ Mn (14 ± 1 nmol cm −2 ), which is consistent with less Mn dimer available for CO 2 reduction (Figure 5c). The TON Mn for CO remained between 1500 and 2000 independent of the surface loading. However, the production of HCOO − increased substantially, from 19 ± 1 μmol cm −2 at high Γ Mn (50 ± 3 nmol cm −2 ) to 54 ± 7 μmol cm −2 at low Γ Mn (14 ± 1 nmol cm −2 ), thus leading to a high TON Mn for HCOO − generation of 3920 ± 230 ( Figure S9). This increase of HCOO − production appears to be inversely correlated to the surface loading of the catalyst and starts to become predominant when Γ Mn reaches values lower than 25 nmol cm −2 (Figures 4b and  S8), which is significantly lower than Γ Mn,max estimated from the Langmuir isotherm. In addition to the increased production of HCOO − , the production of H 2 is also greatly enhanced at lower catalyst loading, from 90 ± 4 μmol cm −2 at high Γ Mn (50 ± 3 nmol cm −2 ) to 225 ± 27 μmol cm −2 at low Γ Mn (14 ± 1 nmol cm −2 ).
Several fac-[MnBr(bpy)(CO) 3 ] systems have already been described for the photocatalytic 57,59,64 or electrocatalytic 55,61 reduction of CO 2 to HCOO − , and for this transformation, a Mn-hydride is thought to be the active catalytic intermediate. This Mn−H intermediate was recently characterized using IR-SEC in solution, and it was shown that the HCOO − production is greatly enhanced by the addition of strong Brønsted acids through more favorable formation of the transient Mn−H species. 61 In agreement, at low surface loadings of Mn pyr , dimerization is less likely, giving rise to a monomeric reduced Mn 0 species that is suggested to form a transient Mn−H upon further reduction in aqueous solution (below the red E p of −1.0 V vs SHE). The Mn−H can then preferentially produce H 2 and HCOO − over CO (Scheme 1).
Spectroelectrochemistry of Mn pyr Dimerization. IR SEC was employed to obtain structural information on the catalytically relevant states of the immobilized Mn pyr at different surface loadings. Specifically, the CO stretching vibrations, ν(CO), were monitored as their frequencies strongly depend on the electronic properties (i.e., oxidation state and coordination environment) of the metal atom. 72,73 Although ATR IR SEC has previously been used for the in situ study of active sites of adsorbed enzymes (on graphite nanoparticles 74 and Au-modified surfaces 75,76 ) as well as for probing bulk catalytic reactions, 77,78 only a few examples applied this method

Journal of the American Chemical Society
Article to investigate heterogenized synthetic molecular catalysts. 79,80 An IR signal enhancing Au layer is usually required to obtain the necessary signal intensity to analyze a surface-attached cocatalyst. To tackle the limitations of established approaches, we employed a modified ATR IR SEC setup to allow an indepth study of the immobilized molecular Mn catalyst on the MWCNT-coated Si prism as the working electrode under in situ conditions ( Figure S10).
Incubation of the MWCNT electrode with Mn pyr (10 mM in dry DMF) afforded intense IR signals of the catalyst. In particular, well-resolved ν(CO) bands in the region from 1800 to 2050 cm −1 were observed (Figure 6a). As the evanescent IR wave exhibits a penetration depth of about 0.5 μm in this experimental configuration, 81 this spectral feature should exclusively result from immobilized Mn pyr molecules with ν(CO) frequencies at 2022, 1930, and 1912 cm −1 for the facially coordinated tricarbonyl metal complex. 73 These frequencies match well those of unmodified [MnBr(bpy)-(CO) 3 ] in MeCN solution, 73 indicating once again that introduction of the pyrene linker does not significantly influence the electronic properties of the Mn center ( Table  1). The observation of intense and distinctly resolved ν(CO) bands demonstrates that ATR IR is sufficiently sensitive to probe Mn pyr on the MWCNT electrode in this experimental configuration.
Negative potentials were applied in aqueous KHCO 3 solution (0.5 M, pH 8.2) to monitor the reduction of the immobilized Mn pyr catalyst through the surface-sensitive ATR IR SEC approach. Figure 6b displays IR difference spectra at increasingly negative electrode potentials (a spectrum at 0 V vs SHE was used prior to each potential step as a reference spectrum). Major spectral differences were only observed at potentials more negative than −0.7 V vs SHE, upon which the appearance of bands located at 1968, 1921, 1870, and 1844 cm −1 was accompanied by a decrease of bands at 2025, 1938, and 1903 cm −1 ( Table 1). Note that the absolute frequencies of the latter two bands are difficult to determine as the band at 1921 cm −1 increases upon reduction. Accordingly, the stated frequencies are estimated. Plotting the intensity of the positive bands at 1968 and 1870 cm −1 as the function of poised electrode potential yielded sigmoidal curves that could be fitted with the Nernst equation affording matching midpoint potentials at −0.84 ± 0.05 V vs SHE for both redox transitions (see the Experimental Section and Figure S11). The concomitant rise and the matching midpoint potentials indicate that these bands belong to the same redox species.
The band at 1921 cm −1 was not considered because two neighboring negative bands at 1938 and 1902 cm −1 masked its intensity (see above). The appearance of the four ν(CO) bands upon applying reductive conditions is indicative of the formation of a Mn 0 dimer species, as previously reported. 56,73 A fifth band (predicted at around 1963 cm −1 from previous measurements in THF 73 ) is known to exhibit only weak IR absorption; therefore, it is most likely masked by the more intense adjacent 1968 cm −1 absorption. 56,73 A frequency downshift (up to −15 cm −1 compared to literature values in organic solvents listed in Table 1) is observed in this work, which may be attributed to decreased electron density at the CO bond resulting from hydrogen bonding to the carbonyl oxygen atoms in aqueous electrolyte solution. 82 The assignment of these bands to a Mn 0 dimer is also supported by the UV−vis SEC experiments above (Figure 2, inset). No indication for the presence of another reduced species was found by IR SEC under the applied conditions. Reoxidation of the dimer was monitored with IR SEC by a stepwise increase of the potential from −0.8 to 0 V (fully oxidized state) vs SHE. Figure 7a shows the resulting difference spectra (a reference spectrum was recorded under fully reduced conditions, E appl = −1.1 V vs SHE, prior to each potential step). The previously observed set of bands located at 2025, 1938, and 1903 cm −1 re-emerges upon applying E appl > −0.5 V vs SHE at the expense of bands assigned to the dimer species (see above). Upon comparison with literature reports, the emerging ν(CO) bands are assigned to formation of a (monomolecular) Mn I complex upon reoxidation of the dimer. As the initial generation of the dimer complex involves removal of the bromide ligand, the state formed at more positive potentials most likely represents a Mn I complex with the bromido ligand being replaced by the solvent water (Table 1). 13,73 Thus, the

Journal of the American Chemical Society
Article observed ν(CO) frequencies are slightly shifted compared to those found for the bromide-bound Mn pyr complex (Figure 6a, Table 1). 64 Figure 7b shows the normalized intensity of the bands at 2025 and 1968 cm −1 , representing the two different redox species (i.e., Mn 0 −Mn 0 dimer vs monomolecular Mn I species) as a function of electrode potential. Both intensities follow the same trend with opposite signs, and the data sets can be fitted satisfactorily using the Nernst equation with matching apparent midpoint potentials (Figure 7b). This observation points to a coupled redox events, in which dimer oxidation directly affords formation of the Mn I species with marker bands identified at 2025, 1938, and 1903 cm −1 .
Spectroelectrochemistry at Low Mn pyr Loading. The change in selectivity of product formation from high to low loading of adsorbed Mn pyr and the suggested change in mechanism for CO 2 reduction were also probed by IR SEC. An obvious challenge in this SEC analysis was associated with the monitoring of very low concentrations of Mn pyr and the resulting weak ν(CO) frequencies, which made a reliable evaluation difficult. We have therefore incubated from a relatively high concentration of Mn pyr (10 mM in DMF) and followed the decreasing Mn pyr surface loading over time by performing repetitive CV scans (scanning between −1.1 and 0 V vs SHE at ν = 50 mV s −1 ) of the CNT|Mn pyr electrode in aqueous KHCO 3 solution (0.5 M, pH 8.2). The progress of slow Mn pyr loss from the MWCNT electrode can be analyzed by following the decreasing anodic wave ( ox E p ; see above), and the catalytic intermediates are monitored through IR SEC by tracking the intensity of the strongest dimer band at 1968 cm −1 formed upon reduction (E appl = −1.0 V vs SHE, reference spectrum recorded at 0 V vs SHE). The electroactive Mn pyr surface concentration was reduced by approximately 50% after 400 CV scans ( Figure S12). The Mn pyr film loss may be due to desorption or decomposition of the immobilized Mn pyr complex. Only two reduced Mn species carrying CO ligands were observed, which can be consistently assigned to reduced species of the intact Mn pyr catalyst on the surface (see below). Possible decomposition of the molecular Mn pyr would not interfere with the IR SEC experiment as ν(CO) frequencies were selectively monitored. Figure 8 shows the surface-sensitive IR difference spectra after incubation as a reference and after 200 and 400 CV scans. After incubation, reduction of Mn pyr at E appl = −1.0 V vs SHE afforded the spectral pattern as discussed above (Figure 6b), indicating the formation of the Mn 0 −Mn 0 dimer species. Lowering of the concentration of Mn pyr is consistent with a decrease in the overall signal intensity of the dimer bands (at 1968, 1870, and 1844 cm −1 ) and the appearance of spectral features at 1991 and 1880 cm −1 . Stepwise reduction by applying potentials from −0.7 to −1.0 V vs SHE revealed that the new spectral features in the spectrum in Figure 8 were almost reversibly formed only upon reduction of the catalyst ( Figure  S13). Further decrease of Γ Mn afforded the increase of the latter bands, which dominate the spectrum upon reduction.
To verify that these two bands belong to the same redox species, band analysis for the spectra recorded after 200 and

Journal of the American Chemical Society
Article 400 CV scans was performed (Figure 8). Lorentzian band profiles were chosen for the positive bands, and their intensity and half-width iteratively varied until full reconstruction of the (positive parts of the) experimental spectra ( Figure S14a). The bands at 1991 and 1880 cm −1 do not alter in relative intensity in the red and blue spectra in Figure 8, indicating that they belong to the same spectral component, that is, redox species. For comparison, such behavior is also observed for the bands at 1968 and 1844 cm −1 ( Figure S14), which is expected as these bands are assigned to the dimer species, as demonstrated above (see Figure S11). The bands at 1991 and 1880 cm −1 likely represent a Mn-hydride complex that is formed upon reduction under these conditions, which is supported by the striking match between observed and literature-reported band frequencies for measured and calculated Mn-hydride complexes ( Table 1). 61 Furthermore, the species is reversibly formed only upon reduction, and the presence of an anionic species (i.e., the [Mn(bpy pyr )(CO) 3 ] − species) can be ruled out as this species would be indicated by two strong ν(CO) bands located at approximately 1910 and 1811 cm −1 . 73 While the first band may be masked by the negative bands of the oxidized Mn pyr , no indication for the latter is observed in any of the recorded spectra upon reduction at potentials down to −1.1 V vs SHE. Finally, the assignment of a Mn-hydride species is also in accordance with preferential formation of formate and H 2 upon CPE experiments using lowered Mn pyr surface concentrations, providing thereby an explanation for the switch in reactivity. Accordingly, the SEC IR experiments demonstrate that low concentrations of Mn pyr favor formation of the hydride over the dimer species, and the long-term CPE measurements reflect the catalytic consequences of this change in mechanism. The application of our surface-sensitive ATR IR SEC configuration allowed the effective in situ detection of catalytically relevant states of the immobilized Mn pyr complex. Formation of the Mn 0 −Mn 0 dimer species as well as the monomeric reduced Mn−H complex could be characterized on a MWCNT electrode surface. Notably, the presence of the two species could be directly linked to the electrocatalytic product selectivity (CO via the dimer and HCOO − and H 2 via the Mnhydride).

■ CONCLUSIONS
This work describes the synthesis, electrochemistry and SEC characterization of a Mn CO 2 reduction catalyst immobilized through a pyrene anchoring group on a MWCNT electrode. Despite the low Faradaic yield for CO and formate, the hybrid electrode displays high activity for electrocatalytic reduction of aqueous CO 2 with TONs of up to 1790 ± 290 for CO and up to 3920 ± 230 for formate at E appl = −1.1 V vs SHE (corresponding to overpotentials of η = 550 mV for CO and η = 590 mV for formate production). Product selectivity can be fine-tuned by controlling the surface loading of the catalyst on the CNT sidewalls. Low Mn catalyst loadings afford selectivity for the generation of dissolved formate, whereas high loading result in higher selectivity for gaseous CO. Control over product selectivity is a first step to rationally synthesize desired high-value products from CO 2 on demand.
In situ ATR IR and UV−vis SEC characterization of the molecular catalyst on the electrode surface allowed distinguishing between two different catalytic pathways by detecting the involved intermediates. The identified Mn dimer and Mnhydride species formed on the electrode surface could be directly linked to the occurrence of distinct catalytic products. The implementation of such surface-sensitive in situ spectroscopy therefore proves to be of importance for understanding catalytic reactions of surface-bound molecular catalysts, stimulating future interdisciplinary research also beyond CO 2 reduction and MWCNT electrodes.
Overall, the present study highlights the benefits of rational incorporation of molecular catalysts onto electrode surfaces. Our strategy allowed enhancement of the catalytic activity and stability, as well as tuning of the selectivity of the immobilized catalyst, while enabling a detailed understanding of the catalytic mechanisms involved through the use of a powerful SEC approach.

Journal of the American Chemical Society
Article Synthesis of bpy pyr . Diisopropyl amine (0.33 mL, 2.35 mmol) was added to THF (10 mL). The mixture was cooled to −10°C with an ice/acetone bath. A solution of n-BuLi (1.6 M, 1.5 mL, 2.4 mmol) was slowly added to the reaction mixture and stirred for another 5 min at −10°C under a N 2 atmosphere. Then, bpyMe 2 (230 mg, 1.25 mmol) in THF (40 mL) was added dropwise to the mixture (over 15 min) and further stirred for 10 min at −10°C under a N 2 atmosphere. Then, 1-(4-bromobutyl)pyrene (400 mg, 1.2 mmol) in THF (40 mL) was added slowly to the mixture (over approximately 30 min), and stirred for another 30 min at −10°C under N 2 atmosphere. Cold water (60 mL) was slowly added, followed by CH 2 Cl 2 (80 mL). The product was extracted twice with CH 2 Cl 2 (80 mL), dried over MgSO 4 , and the solvent evaporated to dryness. The crude product was purified by column chromatography on silica gel using CH 2 Cl 2 /MeCN (9:1), yielding the product as an off-white solid. Yield: 410 mg, 77%. 1  Electrodes for CV and CPE. MWCNT/glassy carbon electrodes were drop-cast by adding 20 μL of a 5 mg mL −1 dispersion of MWCNTs in N-methylpyrrolidone onto a glassy carbon electrode, and the deposit was dried under high vacuum before use. A 5 μm thick homogeneous MWCNT film was obtained as reported previously (0.2 cm 2 geometrical surface, ø = 5 mm). 66 The MWCNT/glassy carbon electrodes were soaked in a solution of Mn pyr in dry DMF for 30 min in the dark. The modified electrode was subsequently rinsed with DMF and deionized water. Control experiments were performed with [MnBr(bpyMe 2 )(CO) 3 ], MnCl 2 /bpy pyr , and [MnBr(CO) 5 ] instead of Mn pyr .
Electrodes for UV−Vis SEC. A MWCNT film was prepared as previously described. 84 Briefly, 2 mg of MWCNTs were sonicated in 500 mL of deionized water for 30 min and left to settle overnight. Then 200 mL of the supernatant were filtered over a cellulose nitrate filter (Millipore, 0.45 μm pore size, overall diameter of ø = 3.5 cm), leading to the homogeneous MWCNT film. The MWCNT film was subsequently deposited on a 2 × 1 cm glass slide with the top 1 × 1 cm coated with FTO (so that the MWCNT film covered both FTOcoated and bare glass areas) by carefully dissolving the cellulose filter with several washings of acetone, leading to the MWCNT film on FTO glass. The Mn pyr modified electrode was obtained by dropcasting 10 μL of a 10 mM solution of Mn pyr onto the MWCNT/FTO electrode (geometrical surface area of 0.25 cm 2 ) in the dark. After 10 min, the electrode was rinsed with deionized water and the FTOcoated portions of the electrode were not exposed to the electrolyte solution to prevent reductive degradation of FTO from influencing the spectra.
Electrodes for ATR IR SEC. The MWCNT films were deposited onto the Si substrate by carefully dissolving the cellulose filter with several washings of acetone, leading to the MWCNT film. The Mn pyr modified electrode was obtained by drop-casting 10 μL of a 10 mM solution of Mn pyr onto the CNT surface in the dark. After 10 min, the electrode was rinsed with DMF and with deionized water.
Electrochemistry. All electrochemical measurements were performed on a PalmSens EmStat or MultiEmStat 3+ potentiostat in the dark. Electrochemistry in organic media was performed in MeCN containing 0.1 M TBAPF 6 in a three-electrode configuration using a Ag/AgCl wire as the reference electrode and a platinum mesh as the counter electrode. Ferrocene was used as the internal reference at the end of the measurements.
Aqueous electrochemistry was performed in KHCO 3 (99.95%, 0.5 M) electrolyte solution using a three-electrode configuration using a Ag/AgCl (sat. KCl) reference electrode and platinum mesh as the counter electrode. The potentials were converted from Ag/AgCl to standard hydrogen electrode (SHE) by adding +0.199 V. The CV and CPE measurements were carried out in a three-necked twocompartment cell separated by a Nafion membrane under N 2 or CO 2 . The working electrode compartment was stirred during the CPE measurement. All experiments were carried out at room temperature (∼22°C).
The surface loading of the catalyst was calculated through the integration of the reoxidation wave in the CV scans Γ = q nFA Mn where Γ Mn is the surface loading (mol cm −2 ), q is the charge (C) obtained from integration of the oxidation wave, n the number of electrons in the redox process per Mn center (n = 1), F is the Faraday constant (96485 C mol −1 ), and A is the geometrical electrode area (0.2 cm 2 ).
To model the impact of the Mn pyr solution concentration over the surface loading, a single Langmuir isotherm was used where Γ Mn,eq is the equilibrium surface coverage of Mn pyr (mol cm −2 ), Γ Mn,max is the saturation surface coverage (mol cm −2 ), and K Mn (L mol −1 ) the association constant of Mn pyr with the MWCNT surface in DMF at room temperature. UV−Vis SEC. Measurements were carried out using a Varian Cary 50 Bio UV−visible spectrometer. The measurements were performed using a three-electrode configuration using a Ag/AgCl (sat. KCl) reference electrode and platinum mesh as a counter electrode in a flat custom-designed three-necked electrochemical cell suited for UV−vis measurements. The working MWCNT-FTO electrode was placed between the beam source (FTO was not in contact with the solution to prevent degradation) and the collector to measure UV−vis spectra of the immobilized catalyst at different potentials.
ATR IR SEC. Surface-sensitive IR measurements were carried out in 60°single-reflection ATR configuration using an ATR-active Si prism and an IFS66v/s FT-IR spectrometer (Bruker Instruments) equipped with a photoconductive nitrogen-cooled MCT detector. A total of 400 scans were co-added for a spectrum. The spectral region was set to a range from 400 to 4000 cm −1 . The resolution of the spectra was 4 cm −1 . The MWCNT membrane film-coated Si prism was mounted into a homemade ATR IR spectroelectrochemical cell (fill volume of 5 mL). 85 The solvent-accessible electrode surface had a geometrical area of 0.9 cm 2 . A Ag/AgCl 3 M KCl (DriRef, WPI) and a hydrogen flamecleaned Pt wire (99%, Goodfellow) were used as reference and counter electrodes, respectively. All measurements were carried about at room temperature. Potentials were applied using an Autolab PGSTAT 12 (Metrohm) potentiostat. All measurements were carried out with either Ar or CO 2 overpressure. Absorbance (A) spectra were generated according to Lambert−Beer's equation I band denotes the (normalized) intensity of the respective band. F, R, and T denote the Faraday constant, the ideal gas constant, and room temperature, respectively. n and E 0 ′ represent the number of transferred electrons and the apparent midpoint potential of the redox transition, respectively. Both values are varied in the fits. C is a dimensionless parameter, accounting for possible offsets to obtain better fits of the equation to the data set. Band fitting of the difference IR spectra was performed using a homemade software (Qpipsi). Isotopic Labeling. CPE experiments were performed in a gastight single-compartment (for 13 CO detection) or two-compartment (for 13 C formate detection) electrochemical cell equipped with a CNT| Mn pyr working electrode, Pt counter electrode, and Ag/AgCl (sat. KCl) reference electrode in phosphate buffer (0.1 M, pH 7). The solution was purged with N 2 for 10 min before the headspace gas was replaced with 12 CO 2 or 13 CO 2 (>99 atom % 13 C, Sigma-Aldrich) by evacuating three times and refilling with the respective CO 2 isotopologue. CPE was performed at E appl = −1.1 V vs SHE for 3 and 12 h for CO and HCOO − detection, respectively. Subsequently, the headspace gas was transferred to an evacuated gas IR cell (SpecAc, 10 cm path length, equipped with KBr windows), and a highresolution transmission FT-IR spectrum was collected on a Thermo Scientific Nicolet iS50 FT-IR spectrometer. For H 13 COO − or H 12 COO − detection, 35 μL of D 2 O were added to 700 μL of the electrolyte solution in an NMR tube. The spectra were recorded on a Bruker DPX-400 MHz spectrometer at room temperature, and water suppression was applied.
Product Analysis. The amount of H 2 and CO produced during CPE measurements under N 2 or CO 2 (with 2% CH 4 as an internal standard) was quantified by gas chromatography by analyzing a 50 μL volume of the working electrode headspace compartment with a Shimadzu Tracera GC-2010 Plus using a barrier discharge ionization detector (BID). The gas chromatograph was equipped with a ShinCarbon micro ST column (0.53 mm diameter) kept at 40°C, using helium carrier gas. HCOO − was analyzed by ion chromatography (Metrohm 882 compact IC plus ion chromatography system) with carbonate buffer (4 mM, pH 7.4) eluent containing acetone (50 mL L −1 ). The Faradaic yield for H 2 and CO was calculated using = × × F n Q Faradaic yield product (%) 100 (2 ) prod.
where F is the Faraday constant (C mol −1 ), n prod (mol) is the amount of H 2 /CO measured in the headspace, HCOO − in the electrolyte solution, and Q (C) is the charged passed during electrolysis. The surface loading (from CV) and product quantification (from CPE) were determined from separate sets of experiments in triplicate. The error bars shown in the graphs are the standard deviation calculated as previously described. 86 ■ ASSOCIATED CONTENT