Adsorbed cobalt porphyrins act like metal surfaces in electrocatalysis

Electrodes chemically modified with molecular active sites are potent catalysts for energy conversion reactions. Such electrodes are typically presumed to operate by the same redox mediation mechanisms as the analogous soluble molecules, with electron transfer and substrate activation in separate elementary steps. Here we uncover solvent-dependent concerted reaction mechanisms for cobalt porphyrins attached to glassy carbon electrodes by flexible aliphatic linkages. In acetonitrile, outer-sphere CoII/I reduction mediates H2 evolution in a stepwise sequence. However, in aqueous media, outer-sphere reduction is not observed and H2 evolution proceeds instead by concerted proton–electron transfer pathways typical of metal surfaces. Consequently, catalysis is not defined by the reduction potential of the parent molecule, but rather by the free energy of hydrogen binding. We attribute these mechanistic changes to electrostatic coupling between the molecule and the surface arising from adsorption. Our results motivate a re-examination of the reaction mechanisms of and design principles for molecularly modified electrodes﻿. Heterogenized molecular catalysts are often assumed to operate via analogous mechanisms to their homogeneous counterparts. Here, the authors demonstrate that a tethered cobalt porphyrin exhibits either molecule-like or metal-like behaviour depending on the strength of adsorption between the molecule and the electrode surface.

The design of CMEs usually centres on the optimization of the molecular catalyst because the electrode support is generally viewed as an inert source or sink of electrons. The synthesis of active CMEs thus follows two general steps: the selection of a known, highly active molecular catalyst for the reaction of interest; and the development of a heterogenization method. This synthetic logic assumes that activity trends in homogeneous molecular catalysts will correspond to the same trends in electrode activity upon heterogenization. However, there are flaws in these implicit assumptions. For solubility reasons, most molecular electrocatalysts are evaluated and optimized in aprotic polar organic solvents, such as acetonitrile and dimethylformamide (DMF) with added proton donors. In contrast, catalysis by CMEs is typically evaluated in the aqueous medium relevant to device operation, which has starkly different solvation properties and proton conductivities, both of which are expected to dramatically impact the mechanisms of the proton-coupled electron transfer reactions. Further, in molecular electrocatalysis, substrate activation takes place in solution, far from the enormous electric fields native to the electrode surface [35][36][37][38][39][40][41][42] . Since electric fields impact the reactivity of enzymes [43][44][45] , surfaces [35][36][37][38][39][40][41][42]46,47 and molecules [48][49][50][51][52][53] , there is every reason to expect that they may influence the reactivity of CMEs 28,35,41 . An improved understanding of the mechanistic implications of the impact of solvation and local electric fields has the potential to expose design principles for the rational development of more active CMEs.
Recent observations lend credence to the notion that heterogenization does alter catalytic activity 24,26,27,[29][30][31]35 . For example, immobilization of an iron porphyrin results in a 26-fold rate enhancement versus the same species dissolved in the same reaction media 30 . Other studies have shown that the reactivity of an appended catalyst is tuned by diverse factors, such as immobilization method 24,31 , electrode surface chemistry 26,27 and polymer binder identity 16,27 . Although these factors are as important as the catalyst structure in defining its reactivity, the origins of these effects remain unclear 16,24,26,27 . In an extreme example, pyrazine linkages that engender strong electronic coupling between molecules and the surface prevent the immobilized catalyst from operating by the stepwise pathways characteristic of molecular catalysts, and instead enforce the concerted mechanisms unique to heterogeneous metallic catalysts 29,54,55 . Clearly, the rational development of CMEs requires greater understanding of the molecule-surface interactions that play a key role in defining their reactivity.
In this study, we directly probe the impact of strong surface interactions on catalysis by adsorbed molecules. Counterintuitively, instead of directly adsorbing the molecular species to the surface, we anchored a cobalt tetraphenylporphyrin (CoTPP) to an oxidized glassy carbon electrode with a covalent aliphatic linkage. We chose this linkage to minimize convolution from multilayer adsorption (Supplementary Note 1), prevent dynamic aggregation on the surface 22 and permit the comparison of reactivity across diverse reaction media while maintaining a constant surface population of active sites. Importantly, this flexible cyclohexyl tether allows the appended CoTPP to adopt either an adsorbed or solvated configuration, depending on the solvating properties of the electrolyte. Thus, by examining catalytic trends across different reaction media, we isolate the critical role of surface interactions on catalysis by CMEs. We identify that surface adsorption of CoTPP confers electrostatic coupling to the electrode surface, with dramatic implications for reaction mechanism and catalyst design. We identify these results by comparing the hydrogen evolution reaction (HER) activity of CoTPP covalently linked to a glassy carbon electrode in acetonitrile, aqueous and mixed aqueous-pyridine electrolytes. We further compare the surface-bound CoTPP to a water-soluble analogue in aqueous electrolyte. We demonstrate that the stepwise reaction pathways observed for soluble molecules occur for this CME in media in which CoTPP is soluble, but that the concerted reaction pathways typical of metallic electrodes occur for this CME in media in which CoTPP is insoluble. Finally, we provide a model to explain how this change in reaction mechanism could arise from adsorption onto the surface. These results have broad implications for the design of CMEs and suggest that the design criteria for optimizing soluble molecular catalysts may not apply upon adsorption to the electrode surface.
with metal-bound pyridinic nitrogen and the structural data are most consistent with square-planar cobalt (see below). Thus, we attribute this feature to the reaction of pyridine with reactive surface chlorides that result from the thionyl chloride treatment. Further, this feature is not observed, or is greatly attenuated, after electrochemistry in aqueous media, where such species would be expected to hydrolyse ( Supplementary Fig. 6). The high-resolution spectrum of the Co 2p peak manifold reveals a single 2p doublet at 779.9 and 795.1 eV, indicative of only a single cobalt environment on the surface (Fig. 2b). The observed Co 2p binding energies are consistent with literature reports on cobalt porphyrins 31,61 . The relative integrations of the amide and porphyrin N 1s features together against the Co 2p peak manifold yields a N:Co ratio of 8 (expected ratio is 6), suggesting that some demetallation occurs during the surface functionalization procedure. Together, the XPS data support the formation of the amide linkages and the presence of intact porphyrin units on the surface. X-ray absorption spectroscopy (XAS) further supports surface modification with intact CoTPP units. Cobalt K-edge extended X-ray absorption fine structure analysis (EXAFS) provides a fingerprint for the molecular structure about the cobalt centre. EXAFS of CH-CoTPP is nearly identical to that of the molecular analogue, CoTPP, but does not fit well to CoClTPP, which bears an axial ligand on the Co centre (Fig. 2c). These data reveal a square-planar coordination environment for CH-CoTPP. The X-ray absorption near-edge structure (XANES) provides complementary electronic structure information. Consistent with previously examined covalently grafted porphyrins on graphitic carbon 31 , the XANES spectrum of CH-CoTPP exhibits pre-edge features consistent with both the Co II and Co III standards, which may indicate mixed valency character of the Co centre on the surface (Fig. 2d). The origin of the changes to the pre-edge features remains undetermined. Together with the XPS data, the XAS data support ligation of intact CoTPP units to the glassy carbon surface.
Outer-sphere electron transfer is observed in acetonitrile. In acetonitrile electrolyte, the cyclic voltammogram (CV) of CH-CoTPP exhibits a clear outer-sphere Co II/I feature, consistent with analogous systems 58 . In 0.1 M tetrabutylammonium hexafluorophosphate (TBAPF 6 ) in acetonitrile, this feature occurs at −0.76 V versus decamethylferrocene (Fc* + /Fc*) (Fig. 3). The peak current density of this feature scales linearly with scan rate, consistent with a surface-bound species (Supplementary Fig. 7). Importantly, the charge passed in this feature correlates well to the amount of cobalt on the surface, as determined by inductively coupled plasma mass spectrometry (ICP-MS). We find a Co-to-electron ratio of 1.05 ± 0.07, which further supports our assignment of this feature as a Co II/I redox process. The observation of this feature in electrolyte without strongly coordinating anions suggests that it is not ion coupled. Furthermore, this feature exhibits no dependence on chloride concentration, indicating that the feature is not tied to ion coordination or dissociation ( Supplementary Fig. 8). Together, these data support that, in acetonitrile media, CH-CoTPP exhibits outer-sphere electron transfer (ET) behaviour. Catalysis in acetonitrile operates by redox mediation. CH-CoTPP is a catalyst for HER in acetonitrile. CVs of CH-CoTPP were collected in acetonitrile containing 0.1 M TBAPF 6 and donors with defined proton activities. The proton activity was defined by addition of 25 mM each of proton donor and its conjugate base. In the presence of highly acidic donors, CH-CoTPP catalyses the HER but decomposes too rapidly for in-depth analysis. In the presence of a chloroacetic acid buffer (pK a = 15.3; CH 2 ClCO 2 H-[TBA + ] [CH 2 ClCO 2 − ]) 62 , CH-CoTPP catalyses the HER, with a sloping S-shaped wave observed at a scan rate of 5 mV s −1 (Fig. 4a). The onset of this wave is tied to the Co II/I redox couple. Such behaviour indicates mediated catalysis that operates by stepwise outer-sphere pathways 63,64 . At faster scan rates, which exceed the rate of the catalytic turnover, we observe the Co II/I redox wave. Integration of this wave in the chloroacetic acid buffer accounts for ~75% of the total loading of Co on the surface, indicating that CH-CoTPP does not substantially decompose under these reaction conditions ( Supplementary Fig. 9).
To further support the assignment of a mediated stepwise mechanism, catalysis was evaluated in the presence of a less acidic donor. Redox-mediated reactions are pinned to an outer-sphere reduction potential, but E RHE shifts with the pK a of the proton donor, according to the Nernst equation 65,66 . For a stepwise reaction, selection of a sufficiently basic donor can move E 1/2 to an underpotential of E RHE and catalysis will shut off. Such behaviour is diagnostic of a redox-mediated sequence proceeding by outer-sphere stepwise electron transfers 55 . CVs of CH-CoTPP were collected in the presence of an acetic acid buffer (pK a = 22.3; AcOH-[TBA + ][AcO − ]) 65 , where E 1/2 (Co II/I ) lies at an estimated underpotential of ~100 mV (see Supplementary Note 2). In the acetic acid buffer, no catalysis was observed for the Co II/I couple and the redox feature was recovered (Fig. 4b). Importantly, the observation of clear Co II/I waves in the acetic acid buffer demonstrates that CH-CoTPP does not decompose under these more basic conditions. Together, the observation  of a catalytic wave tied to the Co II/I redox couple with chloroacetic acid buffer and the lack of catalysis with the less acidic AcOH support a stepwise, redox-mediated sequence initiated by outer-sphere electron transfer.

Electron transfer waves are not observed in aqueous media.
In contrast to its behaviour in acetonitrile, CH-CoTPP does not exhibit clear Co-based redox features in aqueous media. To aid interpretation of the aqueous electrochemistry of CH-CoTPP, we first examined the water-soluble cobalt meso-tetra In acidic media, CoClTSP mediates HER catalysis preventing observation of the Co II/I wave ( Supplementary Fig. 12). Consistent with our measured value for the E 1/2 (Co II/I ) for CoClTSP, heterogenized cobalt protoporphyrin IX exhibits a Co II/I couple at −0.6 V versus the standard hydrogen electrode when incorporated into a film on pyrolytic graphite 67 .
Unlike for the soluble molecular analogue, voltammetry of CH-CoTPP does not reveal redox waves attributable to an outer-sphere Co II/I redox process. The CV of CH-CoTPP in 0.1 M HClO 4 reveals one feature near 0.6 V versus NHE (Fig. 5a), which is also observed in the CV of bare glassy carbon in this electrolyte ( Supplementary Fig. 13). We attribute this feature to surface quinone proton-coupled electron transfer (PCET) features or ion-intercalation processes native to oxidized glassy carbon 68 . In 0.1 M phosphate buffer at pH 6.4, this feature is not observed and the CV of CH-CoTPP is featureless within the solvent window (Fig. 5b). Finally, in 0.1 M NaOH, the CV of CH-CoTPP displays a broad redox wave centred at −0.75 V versus NHE (Fig. 5c). The magnitude of this wave is highly variable across electrode preparations and the integrated charge in this wave does not correlate with the Co surface loading. Additionally, this feature is also observed in the CVs of CH-H 2 TPP ( Supplementary Fig. 14), thus we conclude that it is not associated with a Co II/I redox process. Importantly, CVs of CH-CoTPP in acetonitrile still show a pronounced Co II/I wave after collection of these aqueous electrochemistry data ( Supplementary  Fig. 15), indicating that the attached CH-CoTPP units remain intact even though they do not exhibit a clear Co II/I redox wave in water.
Interestingly, an outer-sphere redox process for CH-CoTPP in aqueous media can be recovered in a mixed solvent system. Initially we hypothesized that axial coordination of the Co species could recover the Co II/I wave. In non-aqueous media, the introduction of pyridine is known to induce a Nernstian shift in the Co II/I wave of heterogenized Co porphyrins to more negative potentials 58 , indicative of pyridine ligation to Co II . Additionally, Co porphyrins have also been heterogenized via axial ligation to surface-bound pyridine units [69][70][71] , suggesting that pyridine binds strongly to CoTPP in aqueous media. Thus, we collected CVs of CH-CoTPP in aqueous 0.1 M NaOH containing 0.1 M pyridine ( Supplementary  Fig. 16). In this media, no new waves were observed. However, further addition of pyridine to a final concentration of 3 M (~13% v/v) results in the appearance of a new redox feature (Fig. 5d). The new redox feature is centred at −0.91 V versus NHE and the charge integration accounts for approximately 90% of the surface Co. This potential is 0.27 V more negative than the E 1/2 (Co II/I ) potential of water-soluble CoTSP, consistent with the more donating phenyl substituents and, perhaps, due to the interfacial solvation environment and/or pyridine coordination. Nonetheless, on the basis of the similarity in potentials and the correspondence in charge integration value, we assign this feature as a Co II/I redox process. Importantly, the concentration of pyridine necessary to recover this wave in aqueous media is also enough to sparingly solvate the parent CoTPP molecule ( Supplementary Fig. 17). Given the literature precedence and the high concentration necessary to reveal this Co II/I wave, it is likely that both the solvation of the CoTPP unit by pyridine and axial ligation are responsible for the change in the electrochemical response of CH-CoTPP.

Catalysis operates by inner-sphere pathways in aqueous media.
In aqueous media, CH-CoTPP shows Nernstian scaling for the onset of HER catalysis. In aqueous media over a pH range of 0.3 to 12.8, CH-CoTPP displays catalysis substantially above the background voltammograms of the free base CH-H 2 TPP (Fig. 6a-c). To investigate the mechanism of catalysis, we collected steady-state current-voltage (Tafel) data in buffers across a wide pH range. Control experiments establish that these data are not convoluted by mass transport artefacts or deactivation on the time scale of the measurement (Supplementary Figs. [18][19][20][21][22][23]. These data therefore correspond to activation-controlled rates for CH-CoTPP-catalysed HER. Although high-resolution XPS scans of CH-CoTPP after catalysis at each pH reveal minor amounts of demetallation during catalysis, particularly in acidic media, they show no peaks for Co 0 across all evaluated conditions ( Supplementary Fig. 24). Since the post-catalysis XPS data indicate no detectable formation of Co nanoparticles during catalysis, we attribute the HER activity to intact CH-CoTPP.
Extraction of per-site turnover frequencies (TOFs) of 1 s −1 from the Tafel data reveal that CH-CoTPP catalyses HER with nearly constant overpotential across the pH range (Fig. 6d, red circles). Indeed, the potential required to sustain a TOF of 1 s −1 shifts 58 mV per pH unit (Fig. 6d, red dashed line), in line with the Nernstian scaling  To gain further insight into the mechanism of HER catalysis by CH-CoTPP, we analysed the Tafel slope and measured the H/D isotope effect for HER. In aqueous 0.5 M HClO 4 and 0.1 M NaOH, CH-CoTPP displays Tafel slopes of 110 and 103 mV dec −1 , respectively. These values correspond to transfer coefficients of 0.5 and 0.6, which are consistent with a rate-limiting charge transfer step ( Fig. 7a and Supplementary Fig. 22) 72,73 . Control experiments demonstrate no noticeable deactivation of the catalyst during these measurements, and that these values are not convoluted by mass transport limitations ( Fig. 7a and Supplementary Fig. 23). Importantly, these values are inconsistent with mechanisms invoking pre-equilibrium reduction of the Co centre before rate-limiting proton transfer, which would yield a transfer coefficient of 1.0 and a Tafel slope of 59 mV dec −1 (refs. 72,73 ). Additionally, comparison of HER data collected in 0.5 M HClO 4 in H 2 O and D 2 O reveals an H/D isotope effect of 2.9 ± 0.1, indicative of a proton transfer involved in the rate-limiting step ( Fig. 7b and Supplementary Fig. 25). Together, the Tafel and H/D isotope data are consistent with an HER mechanism by CH-CoTPP involving rate-limiting concerted proton-electron transfer (CPET). Although CPET reactions of water-soluble small molecules and coordination compounds can be driven by outer-sphere ET from an electrode, these mechanisms are generally restricted to a relatively narrow pH or potential range, over which both possible stepwise pathways are energetically disfavoured 63,[74][75][76][77] . Thus, the Tafel and H/D isotope effect data, taken together with the observation of Nernstian scaling and similar mechanistic profiles across the entire pH range, lead us to invoke an inner-sphere, non-mediated concerted mechanism for HER by CH-CoTPP in water.

Discussion
The foregoing electrochemical data indicate distinct ET behaviours and HER mechanisms for CH-CoTPP in aqueous and acetonitrile  electrolyte. In acetonitrile, we observe a clear outer-sphere Co II/I redox process that is in quantitative agreement with the total cobalt population on the surface. That E 1/2 (Co II/I ) couple mediates HER catalysis and the pK a dependence supports a stepwise mechanism initiated by outer-sphere reduction of the cobalt centre. In water, clear Co-based surface redox waves are not observed but are recovered on addition of pyridine. Investigation of the mechanism of HER catalysis by CH-CoTPP in aqueous media supports a non-mediated mechanism in which proton transfer and electron flow are concerted. Additionally, we find that a water-soluble homogeneous CoTPP analogue operates via stepwise HER pathways in aqueous media, suggesting that the change in mechanism for CH-CoTPP requires surface interactions rather than simply resulting from the change in the reaction medium.
The non-mediated, concerted mechanism we observe for CH-CoTPP in aqueous media historically has been attributed solely to metal surfaces, and recently has been implicated for graphite-conjugated catalysts (GCCs), which are molecules conjugated to carbon through aromatic linkages. The ET behaviour of CH-CoTPP thus parallels the differences previously observed for GCCs and their soluble analogues 29,31,54,55,[78][79][80] but with the key distinction that the same surface-bound site displays mediated and non-mediated reaction mechanisms depending on the electrolyte environment. Given the parallels to GCCs, we rationalize the solvent-dependent behaviour of CH-CoTPP by adapting the model put forth for GCCs 29,54,55 . Specifically, we invoke differential interactions with the electrode as a function of the electrolyte to explain the behaviour of CH-CoTPP. Before discussing the specific surface interactions that might be at play, we discuss how the electrochemical behaviour of CH-CoTPP in acetonitrile and aqueous electrolytes can be rationalized by distinct positioning of the Co centre within the electrochemical double layer. A cartoon schematic energy diagram of the double layer is presented in Fig. 8. In each panel, the rectangle represents the band states of the electrode, with filled states in grey and the Fermi level, E F , marked in black. The position of E F is sensed and directly modulated by the potentiostat during polarization. The red lines correspond to the electrostatic potential drop between E F and the solution. The potential of zero free charge, E PZFC , is the potential at which there is no electrostatic potential drop between the electrode and solution 81 . In Fig. 8a,b, we denote a situation in which the Co centre does not interact with the electrode and is instead freely solvated, which places it outside the double layer electrostatic potential drop (red line) (see ref. 38 , pages 771-894 and 1068-1070). Alternatively, in Fig. 8c,d, we depict a situation in which the Co centre is not freely solvated, but is adsorbed to the surface sufficiently strongly to reside within the electrochemical double layer (EDL) (see ref. 38 , pages 771-894 and 1068-1070).
The data for the CH-CoTPP in acetonitrile point to the model shown in Fig. 8a,b. This model details the classical behaviour expected for dissolved molecules and molecules appended to but not interacting with an electrode (see ref. 63 , pages 1-28). In the left of Fig. 8b, the electrode is polarized at E PZFC , where E F lies at an underpotential of E 1/2 (Co II/I ) and no ET occurs. Polarization of the electrode generates an interfacial electric field and raises the electronic levels of the electrode relative to those of the solution outside the EDL, which are fixed. When E F approaches or shifts to more negative values than the fixed E 1/2 (Co II/I ), electrons from the electrode tunnel across the EDL to the CoTPP unit and more electrons are provided by the external circuit to fill the resulting holes in the electrode band states (Fig. 8b) 81 . This is exactly the behaviour observed for dissolved molecules that lie outside the EDL 29 (see also ref. 63 , pages 1-28). In acetonitrile, CH-CoTPP exhibits a clear outer-sphere Co II/I redox process (Fig. 2) that is not proton-coupled and is consistent with the ET behaviour shown in Fig. 8a,b. Further, the catalytic wave tied to the E 1/2 (Co II/I ) redox couple and the pK a -dependent changes to HER activity (Fig. 3) indicate that E 1/2 (Co II/I ) redox couple mediates catalysis and imposes a rate-overpotential scaling relationship 29,[82][83][84][85][86][87][88] for HER activity by CH-CoTPP, which causes catalysis to turn off under conditions in which E RHE is close to or more negative than E 1/2 (Co II/I ). Importantly, applying a more negative potential cannot recover catalytic activity by CH-CoTPP in the presence of strongly basic donors because the free energy of the reaction of Co I with proton is pinned by E 1/2 (Co II/I ), regardless of the applied potential. Together, these data indicate that the energy levels of the Co centre are well isolated from those of the electrode in acetonitrile, and that ET proceeds by the tunnelling of electrons across the EDL in direct analogy to the ET mechanisms observed for dissolved molecules. Thus, the model in which the Co centre in CH-CoTPP resides outside the EDL potential drop is consistent with the experimental data in acetonitrile and implies that the Co centre behaves like a dissolved molecule in this medium. In aqueous media, the data for CH-CoTPP points to the model shown in Fig. 8c,d. Surprisingly, this is the same model that is invoked for metallic surfaces, not molecular species. Here, any shifts in E F also result in tandem movements of the energy levels of the CoTPP unit. As a result, there is insufficient driving force for outer-sphere ET from the electrode to the Co centre. Since this redox process is inaccessible, any mediated reaction pathways that proceed via outer-sphere ET steps are excluded. However, catalysis still occurs via inner-sphere mechanisms because there is a potential drop between the Co centre and the solution. Instead of the tunnelling of an electron, polarization drives the movement of an ion across the EDL to subsequently bind to the Co centre. Thus, unlike the redox-mediated mechanism described above, varying the strength of the proton donor across the pH range can be compensated for by polarization of the Co centre to a more negative potential, thereby maintaining HER catalysis across the entire pH range (Fig. 6). The charge of the proton brought to the Co centre is compensated by concerted electron flow from the external circuit to maintain the electrode potential. This CPET could be described as a one-electron, one-proton transfer to a Co II surface site to make a formally Co III -H species. The catalytic cycle is completed when a soluble donor protonates this Co-H intermediate and an electron flows from the circuit to balance the charge of the proton that crossed the EDL, to result in the overall two-electron stoichiometry of the HER and regenerate a formally Co II state. This model (Fig. 8c,d) in which the Co centre in CH-CoTPP resides inside the EDL potential drop is consistent with the Tafel, H/D isotope effect and pH dependence data in water and requires that the CoTPP unit is interacting strongly with the electrode in this medium.
We attribute the starkly divergent ET and catalytic behaviour in the above model to distinct positions of the CoTPP unit with respect to the EDL potential drop in each medium. This position defines whether there is a substantial electrostatic potential drop between the Co actives sites and the electrode (Fig. 8a,b) or not (Fig. 8c,d). In acetonitrile, the data indicate that there are sufficient layers of solvent and ions between the surface and the CoTPP units to screen the Co centres from the charge on the electrode (Fig. 8a). These intervening layers comprise the EDL and the majority of the electrostatic potential drop occurs across these solvent and ion layers (see ref. 38 , pages 771-894 and 1068-1070). This model implies that, in acetonitrile the relative free energy of solvation of the molecular fragment and affinity of the solvent for the charged surface outcompete the free energy of adsorption of the molecule to the surface (see ref. 38 , pages 895-983). In line with this reasoning, the parent CoTPP complex is sparingly soluble in acetonitrile ( Supplementary Fig. 17) and Co II/I waves can be recovered in aqueous media on addition of 3 M pyridine (Fig. 5d), which sparingly solvates CoTPP (Supplementary Fig. 17).
In contrast, the aqueous data indicate that the catalytically active CoTPP units are not screened from the electrode charge in aqueous electrolyte and that there is no substantial electrostatic potential drop between the electrode and the Co centres. This lack of screening implies that there are insufficient layers of solvent and ions located between the electrode and the CoTPP units over the entire aqueous pH/potential window examined for HER (Fig. 8c). The majority of the electrostatic potential drop that occurs in the EDL thus lies between the Co active sites and the bulk solution. In short, the active CoTPP units are co-solvated with the electrode. In line with this reasoning, we note that the parent CoTPP complex is insoluble in water (Supplementary Fig. 17). We posit that interactions with the electrode surface, such as π-π interactions and/or axial coordination of the Co centres to oxidic edge terminations, stabilize the CoTPP fragment within the EDL (Fig. 8e). In addition to ensuring the

Fig. 7 | CH-CotPP catalyses Her by concerted mechanisms in aqueous media. Kinetic data for Her by CH-CotPP in aqueous acidic media.
a, Potential versus activation-controlled rate (Tafel) plot. The Tafel slope is 110 mV dec −1 . The error bars represent the standard deviation of n = 2 separate electrodes. Data were collected from higher to lower overpotential, followed by recollection at the most negative applied potential to generate the red point. The similarity between the first point collected and the red point indicate that there is minimal change in activity over the course of data collection. b, Cyclic voltammograms of the same CH-CoTPP electrode in 0.5 M HClO 4 in H 2 O (red) and D 2 O (black) recorded with a scan rate of 25 mV s −1 . For both panels, the data were measured while the electrode was rotated at 2,000 r.p.m. strong electrostatic coupling between the CoTPP fragment and the electrode, these surface interactions also appear to afford sufficient quantum mechanical coupling to allow for concerted electron flow to the Co site on the time scale of proton transfer. Given the generality of these surface interactions for adsorbed porphyrins and phthalocyanines on graphitic carbons, the observation of non-mediated pathways for porphyrins heterogenized with alkyl tethers motivates a re-evaluation of the reaction mechanisms of a wide variety of electrochemical transformations by heterogenized macrocycles.

Conclusion
We report the ligation of CoTPP to a graphitic carbon electrode through an aliphatic tether and examined its electron transfer and catalytic behaviour in acetonitrile and aqueous electrolytes. We show that the ET behaviour follows a typical outer-sphere, redox-mediated, stepwise mechanism for the H 2 evolution reaction in acetonitrile but transitions to a non-mediated concerted reaction pathway in aqueous electrolytes. This concerted mechanism by CH-CoTPP in aqueous media is akin to those of metallic electrode  surfaces and bypasses the redox intermediates that pin the reactivity of molecular electrocatalysts. The starkly disparate reaction mechanisms in the two media are attributed to the different solvating properties of acetonitrile and water for the CoTPP unit. The preference for solvation over surface adsorption of the porphyrin in acetonitrile causes it to reside outside the potential drop of the EDL, whereas poor solvation in water favours strong adsorption to the surface, causing the molecule to reside inside the EDL. Thus, in aqueous media, preferential surface interactions lead to strong electrostatic coupling of the CoTPP units to the electrode, which, in turn, drives inner-sphere, concerted reaction mechanisms. Importantly, our model invokes nothing special about the covalent aliphatic amide linkage. While this covalent, flexible anchor was critical for us to expose distinct reaction mechanisms in aqueous and non-aqueous media, there is expected to be negligible electronic coupling between the surface and the appended molecule through this aliphatic tether. Thus, the linkage is not expected to contribute to the strong surface interactions that give rise to non-mediated concerted reaction mechanisms in water. Instead, the flexibility of the linkage permits other surface interactions to occur, if favourable. It is these other surface interactions that give rise to inner-sphere mechanisms in water and these native surface interactions are likely to exist for a wide variety of metallomacrocycle/carbon composite electrodes. Thus, these results motivate a re-examination of the electrochemical reaction mechanisms of adsorbed planar macrocycles.
Specifically, these findings suggest that chemically modified electrodes should not be presumed, a priori, to operate via stepwise mechanisms initiated by outer-sphere ET; the surface redox waves observed for CMEs may correspond to only a fraction of the surface loading and catalytic activity may not arise solely from those apparently electroactive surface sites; and molecular fragments with redox potentials misplaced relative to the thermodynamic potential of the target reaction may, nonetheless, be active as CMEs under electrolyte conditions that foster strong interactions with the electrode surface.
This last point suggests that the reduction potential, E 1/2 , is not as useful a descriptor for identifying the molecular constituents of high-performance CMEs, particularly for molecules that are expected to adsorb strongly to electrode surfaces. Despite the diminished role of the molecular redox potential, the metal identity and local structure 11,12,20,22,32,[89][90][91][92][93][94] , as well as the local environment 26-31 of the adsorbed molecule, remain effective handles for tuning catalysis. We posit that changing the structure influences the free energies of key intermediates 80 , which are common descriptors used to optimize heterogeneous catalysts [95][96][97][98] . Thus, tuning catalyst structure to modulate the binding of key intermediates-instead of the redox potential-may be particularly valuable in identifying candidate molecules for high-performance CMEs. Given that the concerted reactivity observed here circumvents the redox intermediates endemic to mediated catalysis, our findings open the door to an expansion of the use of strong molecule/surface interactions to enhance electrocatalysis on chemically modified electrodes.

Methods
Chemicals and materials. All syntheses were performed in solvents of ACS grade or better. Chloroform and methanol were obtained from BDH; dichloromethane, hexanes, DMF and toluene were obtained from Macron Fine Chemicals; 200 proof ethanol was obtained from Koptec and pyridine was obtained from Acros Organics; and all were used as received unless otherwise noted. Dry dichloromethane and toluene were degassed and dried using a Glass Contour Solvent Purification System built by SG Water and were stored under an atmosphere of N 2 over 4 Å molecular sieves. Dry pyridine and dry dichloroethane were both obtained from DriSolv EMD Millipore. All aqueous synthetic manipulations used deionized water, while all aqueous electrochemical preparations and measurements used reagent-grade water (Millipore Type 1, 18.2 MΩ cm resistivity). Sulfuric acid (OmniTrace, 95.5-96.5%) and hydrochloric acid (OmniTrace, 34-37%) were obtained from EMD Millipore and were used as received. Nitric acid (68-70%) was obtained from BDH and was used as received. Electrolytes were prepared from the following: perchloric acid (Suprapur, Sigma-Aldrich, 70%), sodium hydroxide (Sigma-Aldrich, 99.99%), sodium formate (Sigma BioUltra, >99%), sodium phosphate monobasic (Sigma, 99.999%), sodium chloride (Alfa Aesar, 99.99%), sodium tetraborate (Alfa Aesar, 99.95%), sodium perchlorate monohydrate (Sigma-Aldrich, 99.99%), tetrabutylammonium hexafluorophosphate (Sigma-Aldrich, >99.0%), tetrabutylammonium chloride (Sigma-Aldrich, >99.0%), dry acetonitrile (Sigma-Aldrich, 99.8%) and tetrabutylammonium acetate (>99.0%). Cobalt meso-tetra(p-sulfonatophenyl) porphyrin chloride was obtained from Frontier Scientific and used as received. Glassy carbon disk electrodes were obtained from Pine Research Instrumentation. Hg/HgO and Hg/HgSO 4 reference electrodes were obtained from CH Instruments. Non-aqueous Ag/AgCl reference electrodes with Vycor frits were assembled from kits purchased from BASi. Platinum wire (99.9%) and platinum mesh (99.9%) were obtained from Alfa Aesar. Sources and purities of other chemical reagents used in syntheses are included in the protocols below or in the Supplementary Methods.

General electrochemical methods.
All electrochemical experiments were performed under ambient conditions (21 ± 1 °C) using a Biologic VSP 16-channel potentiostat and EC-Lab software (v.11.43). Rotation experiments were performed using a Metrohm Autolab RDE-2. A Hg/HgO reference electrode (stored in 1 M NaOH, 99.999% semiconductor grade, Sigma-Aldrich) was used for all aqueous experiments in pH >10. All other aqueous experiments used a Hg/HgSO 4 reference electrode (stored in saturated K 2 SO 4 , 99.997% metals basis, Alfa Aesar). Both reference electrodes were periodically checked against pristine electrodes to ensure against potential drift. Electrode potentials for experiments conducted in aqueous media were plotted versus the reversible hydrogen electrode (RHE, E RHE = E Hg/HgO + 0.140 + pH × 0.059 V or E RHE = E Hg/HgSO4 + 0.640 + pH × 0.059 V). All non-aqueous measurements used a Ag/AgCl reference electrode that was filled with and stored in acetonitrile containing 0.1 M TBAPF 6 . At the end of each non-aqueous measurement a small portion of decamethylferrocene (obtained from Sigma-Aldrich) was added to the solution. All non-aqueous potentials were referenced to the Fc* + /Fc* redox couple. A Pt mesh counter electrode was used for all experiments. All electrochemical measurements were recorded in a custom five-neck cell equipped with a sparge tube and counter compartment separated by a glass frit. All glassware used for electrochemical measurements was soaked in aqua regia for at least 30 min and thoroughly washed with reagent-grade water before use. For non-aqueous measurements, the glassware was subsequently dried in an oven and then either brought immediately into an N 2 -filled glovebox or cooled while a stream of MeCN-saturated N 2 was passed through the cell. No iR correction was applied to any measurement.

Preparation of CH-MTPP.
Electrode cleaning and pre-treatment. Glassy carbon button (5 mm diameter, Pine Research Instruments) electrodes were soaked in freshly made aqua regia for 10 s to remove any trace metal impurities from previous functionalization treatments. Electrodes were polished using a Buehler MetaServ 250 Grinder/Polisher equipped with a Buehler Vector LC 250 rotating head. The electrodes were placed in custom-fabricated PTFE holders and polished against a rotating (300 r.p.m.) ChemoMet (JH Technologies) surface with an alumina slurry for two min with 2 lb applied force, followed by rinsing with reagent-grade water. This process was repeated in sequence using 1.0 μm, 0.3 μm and 0.05 μm alumina slurries, each on a different ChemoMet plate. Finally, the electrodes were sonicated twice in reagent-grade water. To increase the surface area and expose more quinone moieties, glassy carbon button electrodes were anodized via potentiostatic electrolysis at 3.5 V versus RHE for 10 s in 0.1 M NaOH. Electrodes were subsequently washed with copious amounts of reagent-grade water and EtOH and dried in vacuo before electrochemical evaluation or further functionalization. Electrodes that were anodized in this way are referred to as GCox.
Preparation of CH-MTPP electrodes. This procedure modifies a literature protocol 31,[56][57][58] . Inside a N 2 -filled glovebox, a septum-capped vial was charged with GCox electrodes and 12 ml of dry toluene. The vial was removed from the glovebox and, via syringe, 3 ml of SOCl 2 (TCI, >98.0%) was added and the vial heated to 120 °C for 1 h. On cooling to room temperature, the solution was removed via syringe (Caution: if not cooled fully, the vial will be pressurized) and the vial was brought back into the glovebox. The electrodes were washed by sequential submersion in two vials of dry toluene, and were then transferred to a clean vial. To the electrodes was added 8 ml of dry toluene and 4 ml of a 1:1 mixture of dry CH 2 Cl 2 and dry pyridine, which contained ~2.5 mg of either cobalt or free-base trans-4-amino-N- (4-(10,15,20-triphenylporphyrin-5-yl)phenyl) cyclohexane-1-carboxamide. The vial was sealed, removed from the glovebox and heated to 120 °C for 3 h. On cooling, the electrodes were washed with copious amounts of toluene, then DMF and then MeOH. The electrodes were dried in vacuo and stored in a vial under ambient conditions before electrochemical studies or surface characterization.
Preparation of CH-MTPP on high-surface-area carbon. Before use, Monarch 1300 powder (Cabot) was washed according to the literature 78 . Under an inert atmosphere, Monarch 1300 powder was washed in a Soxhlet extractor with EtOH and o-dichlorobenzene for 24-72 h each, then dried in vacuo. To a 40 ml scintillation vial charged with a stir bar, 200 mg of the washed Monarch 1300 was added. The vial was placed under an Ar atmosphere and 15 ml of dry toluene was added. A 3 ml portion of SOCl 2 was added via syringe and the vial was heated to 120 °C. The mixture was stirred for 1 h, then brought into a N 2 -filled glovebox. The carbon powder was collected on a filter paper using a Hirsch funnel and washed with 4 × 5 ml of dry toluene. The carbon powder was dried by pulling a gentle vacuum through the filtration apparatus and then was transferred to a clean vial charged with a stir bar. To the vial was added 12 ml of dry toluene and 3 ml of dry pyridine containing 10 mg of CoPorNH 2 . The vial was sealed, removed from the glovebox and heated to 120 °C for 3 h. On cooling, the carbon was collected on filter paper in a Hirsch funnel. The powder was washed with toluene, DMF and MeOH. The powder was transferred to a thimble and washed in a Soxhlet extractor with EtOH for 14 h then CH 2 Cl 2 for 14 h. Elemental analysis: C, 89.95; H, 0.82; N, 1.02; Co, 0.12. Elemental analysis of unfunctionalized Monarch 1300 yields a native nitrogen percentage of 0.25%. Controlling for the native nitrogen content yields a N:Co ratio of 27:1, consistent with excess incorporation of nitrogen content by reaction of pyridine with reactive carbon chloride species generated by the SOCl 2 treatment.
X-ray photoelectron spectroscopy. XPS spectra of CH-MTPP were recorded using a Physical Electronics PHI Versaprobe II with a monochromatic AlKα X-ray source (1,486.6 eV) and a hemispherical energy analyser. Spectra were collected by fixing the glassy carbon buttons to the support plate with conductive carbon tape. Data were collected at a base pressure of 5 × 10 −9 torr using a 200 μm, 50 W focused beam at a take-off angle of 45°. Survey spectra were collected using a pass energy of 187.85 eV and a step size of 0.8 eV. High energy resolution scans, which were used for peak fitting, were collected with a pass energy of 23.50 eV and a step size of 0.1 eV. All quantification was performed using MultiPak software (v.9.6.3.B). Each spectrum was smoothed with a seven-point Savitzky-Golay method before quantification. Peak fitting was performed in CasaXPS (v.2.3.17PR1.1). All scans were smoothed with a five-point Savitzky-Golay method and referenced to the graphitic C 1s peak (284.3 eV) of glassy carbon. The N 1s peak manifolds of all compounds were fit with a Shirley-type background and fit with Gaussian/ Lorentzian line shapes of 30% Gaussian shape. The Co 2p peak manifolds were fit with a linear background and an asymmetric peak shape comprising a Gelius profile convoluted with a Gaussian/Lorentzian obtained in CasaXPS by inputting A(0.35,0.8,0)GL (30) in the entry for Line Shape in the peak-fitting window. This procedure was used to produce the data in Fig. 2a,b and Supplementary Figs. 5, 6 and 25. Summaries of surface atomic concentrations and peak binding energies are given in Supplementary Tables 1 and 2. X-ray absorption spectroscopy. X-ray absorption measurements of high-surface-area CH-CoTPP were conducted at the Co K-edge (7.708 keV) at Beamline 8-ID Inner Shell Spectroscopy at the National Synchrotron Light Source II at Brookhaven National Laboratory. Measurements were collected in fluorescence mode using a Si(111) cryogenically cooled double crystal monochromator, a Fe filter and a passivated implanted planar silicon diode detector. An ionization chamber detector was used for the incident beam. Samples were prepared by grinding into a powder, which was subsequently pressed into a pellet. The pellet was mounted on a piece of Kapton tape. The molecular standards were diluted by grinding with boron nitride. CH-CoTPP was not diluted. All spectra of molecular standards were obtained at room temperature under air. EXAFS and XANES data were processed using the Horae suite for analysis. EXAFS were processed using data within the range k = 2 to k = 12 using a Hanning window. The data were k 3 -weighted. The R-space data were not phase corrected. The EXAFS data are presented in Fig. 2c and XANES data are presented in Fig. 2d.

Determination of catalyst surface concentration. General methods for ICP-MS.
Spectra were collected using an Agilent 7900 ICP-MS. A calibration curve was generated from aqueous 2% nitric acid solutions containing known concentrations of Co. Erbium (Ricca Chemical Company, 1,000 ppm in 3% HNO 3 ) was used as an internal standard in the calibration curve and samples. The calibration solution series were prepared by serial dilution of a Co standard solution (Fluka, TraceCERT 1,000 ppm in 2% HNO 3 ) with 2% nitric acid (EMD Millipore, OmniTraceUltra). All volumetric flasks were soaked in aqua regia and rinsed with copious amounts of reagent grade water before use. Solutions, if not used immediately, were stored in air-tight plastic containers and shielded from light.
Determination of the ratio of integrated charge to the surface Co concentration for CH-CoTPP. Before digestion for ICP-MS, cyclic voltammograms of CH-CoTPP modified glassy carbon electrodes were collected in 0.1 M TBAPF 6 in MeCN under N 2 . Voltammograms were initiated at the open circuit potential and swept reductively, cycling three times. The third sweep was used for integration. The redox feature centred at −0.76 V versus Fc* + /Fc* was assigned to the surface Co II/I redox couple and was integrated to determine the charge passed. Following cyclic voltammetry, each electrode was left in the PTFE holder used for electrochemistry and carefully placed in a 15 ml plastic centrifuge tube containing approximately 1 ml HNO 3 (OmniTraceUltra, EMD Millipore) such that the acid only made contact with the holder and the face of the button electrode that had been exposed to electrolyte. This method ensured that only the surface of the button analysed electrochemically was digested. The electrodes were soaked overnight (~14 h) followed by dilution to a final volume of 25.00 ml by addition of reagent-grade water. The concentration of Co in the resulting solutions was measured by ICP-MS. The integrated charge in the Co II/I wave was converted to moles of electrons by dividing by the Faraday constant. The measured cobalt surface concentration was divided by the integrated charge quantities for each electrode to obtain the Co/e − ratio of 1.05 ± 0.07 reported in the main text.
Determination of Co surface concentration in CH-CoTPP. The above Co/e − ratio, to the nearest integer value, was used along with the integrated charge in the non-aqueous Co II/I waves to determine the Co surface concentration of all subsequently prepared CH-CoTPP electrodes using the following equation: where Γ Co is the surface coverage of cobalt in mol, Q Co(II/I) is the integrated charge of the Co II/I wave, R Co/e− is the experimentally determined ratio of cobalt to electrons rounded to the nearest integer and F is the Faraday constant.
Assessment of H 2 reduction activity. Evaluation of catalysis and lack thereof in non-aqueous media. Catalysis in non-aqueous media was evaluated by measurement of CVs. Electrolyte was prepared immediately before use in an N 2 -filled glovebox. If experiments were not performed inside the glovebox, the electrochemical cell was purged with MeCN-saturated N 2 immediately following addition of electrolyte. During all measurements, MeCN-saturated gas was passed through the headspace of the cell to avoid evaporation of the electrolyte. Electrodes were cycled in MeCN electrolyte containing 0.1 M TBAPF 6 to establish the presence of the Co II/I surface feature. After these initial CVs, the relevant proton donor and conjugate base were dissolved using the electrolyte, and added via syringe to the electrochemical cell to yield a final concentration of 25 mM for both acid and conjugate base. Catalysis was then evaluated by recording CVs. In some cases, electrodes were initially cycled in 0.1 M NaOH before any evaluation in non-aqueous media, with no change to the observed catalytic activity but with cleaner baselines to the CV background charging current. After all voltammograms were recorded for a given solution, the CH-CoTPP electrode was replaced with a freshly polished Au electrode, a small portion of Fc* was added, and a CV was recorded.
Evaluation of catalysis in aqueous media. Before recording electrochemical data in aqueous media, CVs on each electrode were recorded in MeCN electrolyte containing 0.1 M TBAPF 6 under N 2 to measure the Co II/I redox couple. Integration of the charge passed in this wave was used to determine the surface cobalt concentration and TOFs for each electrode. After these CVs, electrodes were rinsed with MeCN and dried in vacuo.
Measurement of electrochemistry in aqueous media. Steady-state data were collected via chronoamperometry across a series of potentials spanning the activation-controlled region. All measurements were recorded while the electrode was rotated at 2,000 r.p.m. unless otherwise noted. Chronoamperograms were collected in 25 mV increments from low to high potential (higher to lower overpotential) followed by the recollection of the lowest potential (highest overpotential). Each chronoamperogram was allowed to reach steady state (10 s) and the average current density over the last 5 s of data collection was used to determine the TOF at each potential. TOFs were calculated according to the following formula: where i is the current in mA, Γ Co is the surface coverage of Co in mol and F is the Faraday constant. The factor of 2,000 accounts for H 2 evolution being a two-electron process and for the conversion factor for dimensional analysis to change units from mA to A, such that the TOFs have units of s −1 . The procedure above was used to generate Supplementary Figs. 14-19 and 21. Using the best fit-lines from those figures, the potentials for which TOF was 1 s −1 were determined and these values were used to generate Fig. 8.