David W.
Wakerley
and
Erwin
Reisner
*
Christian Doppler Laboratory for Sustainable SynGas Chemistry, Department of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK. E-mail: reisner@ch.cam.ac.uk
First published on 6th February 2014
Electrochemical molecular catalyst screening (EMoCS) has been developed. This technique allows fast analysis and identification of homogeneous catalytic species through tandem catalyst assembly and electrochemistry. EMoCS has been used to study molecular proton reduction catalysts made from earth abundant materials to improve their viability for water splitting systems. The efficacy of EMoCS is proven through investigation of cobaloxime proton reduction activity with respect to the axial ligand in aqueous solution. Over 20 axial ligands were analysed, allowing rapid identification of the most active catalysts. Structure–activity relationships showed that more electron donating pyridine ligands result in enhanced catalytic currents due to the formation of a more basic Co–H species. The EMoCS results were validated by isolating and assaying the most electroactive cobaloximes identified during screening. The most active catalyst, [CoIIICl(dimethyl glyoximato)2(4-methoxypyridine)], showed high electro- and photoactivity in both anaerobic and aerobic conditions in pH neutral aqueous solution.
Molecular proton reduction catalysts are capable of evolving hydrogen through electro- and photocatalysis.6,7 Progressive research has focused on fabricating proton reduction catalysts based on earth abundant and therefore scalable elements, rather than precious metals.8 This effort has produced catalysts that generate hydrogen with high turnover frequencies,9 low overpotentials,10 from water11 and even under aerobic conditions.12 However, synthetic catalyst activity still falls short of the more industrially relevant catalytic rates and turnover numbers attained by Pt13,14 and H2 evolving enzymes called hydrogenases.15
A number of techniques have recently emerged that allow heterogeneous surfaces to be screened for catalytic activity, quickly yielding new, highly active materials.16–20 In contrast, the synthesis of novel molecular catalysts remains a very slow and often fruitless practice and as such a molecular screening procedure would dramatically accelerate progress in this field.
This work describes the development of electrochemical molecular catalyst screening (EMoCS). This concept extends upon combinatorial approaches21,22 and is specifically tailored to discover molecular redox catalysts for renewable fuel generation. EMoCS is achieved through tandem in situ assembly and electrochemistry of a complex and is carried out in aqueous solution, which ensures all catalysts identified are active in conditions relevant for large scale application.23
EMoCS has been used here to investigate cobaloximes (Fig. 1),24–29 an O2-tolerant12 catalyst family that display proton reduction activity at low overpotentials30 in neutral aqueous solution.31–33 Substantial improvements are required before these catalysts can be considered useful for generating hydrogen outside of the laboratory environment.
Enhancing cobaloxime activity has thus been the subject of considerable scrutiny. Proton reduction proceeds through several key oxidation states and requires protonation of a Co–H species (Scheme 1). The hydride intermediate can be made more basic by increasing electron density on the Co center34 but this also leads to more negative reduction potentials, thereby increasing the catalytic overpotential. A number of studies have explored modification of the complex to exploit this relationship, however this has focused predominantly on the equatorial macrocycles,30,34 leaving the axial ligand relatively unexplored in aqueous solution.
Scheme 1 Proposed mechanism of H2 formation.26 |
Although the native CoIII state of the cobaloxime has two axial substituents, upon reduction to lower oxidation states one dissociates, leaving a free coordination site for catalysis.35,36 The remaining axial ligand is trans to the intermediate hydride formed during proton reduction (Scheme 1) and it is well documented that the presence of an axial aromatic N-binding ligand, such as pyridine, enhances the reactivity with H2 and the reduction of protons in organic media.35,37 Attempts to augment this effect are limited in number and have drawn conflicting conclusions concerning the effect of more electron donating axial ligands on cobaloxime activity.24,35
EMoCS is capable of clarifying the role of the cobaloxime axial ligand in aqueous solution at a rate unattainable by synthetic procedures. This was used to elucidate structure–activity relationships (SARs) and subsequently the most active cobaloximes have been isolated and tested to confirm the efficacy of the EMoCS technique.
Initial experiments explored this premise by monitoring the in situ formation of cobaloximes. Starting from a solution of [Co(H2O)6]Cl2 in aqueous phosphate buffer (Pi, 0.1 M, pH 7), cyclic voltammograms (CVs) were recorded using a glassy carbon (GC) working electrode as constituent ligands were added (Fig. 2).
The grey trace in Fig. 2 is a CV of [Co(H2O)6]Cl2 (2 mM), which shows no pronounced electrochemical features between −0.8 and +0.7 V vs. NHE. Sonication of [Co(H2O)6]Cl2 with two equiv. of dimethylglyoxime (dmgH2) results in the spontaneous formation of a brown solution, which corresponds to dissolved [Co(dmgH)2(H2O)2] as confirmed by UV-vis spectroscopy and mass spectrometry (Fig. S1 and S2, ESI†). This compound displays a small proton reduction wave with an onset potential of approximately −0.65 V vs. NHE (Fig. 2, red trace), consistent with previous studies.30
Addition of pyridine to the solution of [Co(dmgH)2(H2O)2] formed [Co(dmgH)2(pyridine)2], which was confirmed by 1H-NMR spectroscopy and mass spectrometry (Fig. S3 and S4, ESI†). Introducing pyridine resulted in the appearance of a reversible CoIII/CoII couple at approximately −0.03 V vs. NHE and a catalytic wave at a lower overpotential (onset at −0.54 V vs. NHE) in the CV (Fig. 2, blue trace). The catalytic wave occurs upon reduction of CoII to CoI and its size correlates to catalytic activity.38 The wave height was found to increase until approximately four equivalents of pyridine were added (Fig. S5, ESI†), at this point it was assumed all water ligands were substituted.
Addition of four equiv. of 2,6-dimethylpyridine to [Co(dmgH)2(H2O)2] produced no significant change in the CV (Fig. 2, green trace) compared to [Co(dmgH)2(H2O)2] (red trace). 2,6-Dimethylpyridine is sterically hindered and cannot coordinate to the Co ion, confirming that the observed increase in activity upon addition of pyridine was a result of ligation to the cobaloxime to form a more active catalyst. Similar accounts of in situ formation have been documented for other catalysts,39–43 however this assembly has not been exploited to methodically screen for catalytic activity.
CVs generated from EMoCS are shown in Fig. 3 and Fig. S6 (ESI†) and the data is summarised in Table 1. The screening procedure was highly reproducible with a small standard deviation in all experiments (typically <5%). It is apparent that pyridines with electron donating groups, such as 4-methoxypyridine (2) and 4-methylpyridine (3), form highly effective catalysts that surpass the activity of the pyridine (4) substituted cobaloxime. 3,5-Dimethylpyridine (1) (in contrary to 2,6-dimethylpyridine) can coordinate to the Co center and also forms a cobaloxime that displays a high level of activity. The CV of this species decays rapidly upon consecutive scans due to poor electrostability, which was not observed for the other cobaloximes (Fig. S7, ESI†).
No. | Axial ligand | E 1/2 CoIII/CoII/V vs. NHE | Catalytic onset/V vs. NHE | Catalytic current/μA | E L/V |
---|---|---|---|---|---|
a No well-defined reversible CoIII/CoII couple was observable, preventing the accurate determination of E1/2 CoIII/CoII. b This value rapidly decayed upon consecutive scans. c See ref. 46. d Values determined in this study from Fig. S8 (ESI) and eqn (2). e E L values converted from corresponding Hammett parameters (σp) (ref. 50, R. Museo and O. Sciacovelli, J. Org. Chem., 1997, 62, 9031–9033 and C. Hansch, A. Leo and R. W. Taft, Chem. Rev., 1991, 91, 165–195.). f E L value calculated based on the deprotonated phosphonic acid (hydrogen phosphonate anion) or protonated form of the amine substituent (ammonium cation). | |||||
1 | 3,5-Dimethylpyridine | −0.06 | −0.52 | −116b ± 4.3 | 0.21c |
2 | 4-Methoxypyridine | −0.07 | −0.54 | −105 ± 2.6 | 0.22d |
3 | 4-Methylpyridine | −0.07 | −0.54 | −103 ± 1.8 | 0.23c |
4 | Pyridine | −0.03 | −0.54 | −90.7 ± 1.1 | 0.25c |
5 | 4-Pyridinephosphonic acid | —a | −0.52 | −79.7 ± 3.3 | 0.27e,f |
6 | 4-Cyanopyridine | 0.14 | −0.52 | −50.5 ± 1.1 | 0.32c |
7 | 3,5-Dichloropyridine | 0.19 | −0.53 | −45.8 ± 4.4 | 0.33c |
8 | 3-Cyanopyridine | 0.15 | −0.53 | −39.4 ± 0.8 | 0.33d |
9 | 4-Aminopyridine | —a | −0.60 | −42.0 ± 0.8 | 0.32e,f |
10 | 4-Dimethylaminopyridine | —a | −0.61 | −36.9 ± 5.1 | 0.33e,f |
Other noteworthy ligands include the more withdrawing cyanopyridines (6, 8) and 3,5-dichloropyridine (7), which produce catalysts that function at lower overpotentials, albeit with relatively low catalytic currents (Table 1). The 4-pyridinephosphonic acid (5) substituted complex also shows an appreciable catalytic current.
As for the less effective catalysts, pyridines bearing donating 4-aminopyridine (9) and 4-dimethylaminopyridine (10) substituents show surprisingly low activity. This observation was attributed to the amines being protonated at pH 7 (pKa of 4-aminopyridine = 9.12 and dimethylaminopyridine = 9.6044) to form electron withdrawing ammonium-substituted pyridines. Compounds containing non-aromatic nitrogen atoms did not appear to bind to the cobaloxime and displayed nearly identical CVs to the [Co(dmgH)2(H2O)2] precursor upon addition (Fig. S6, ESI†).27 A similarly small catalytic influence was observed when adding 4-hydroxypyridine, which may be due to the formation of a 4-pyridone tautomer.45
An imidazole substituted catalyst showed high catalytic currents (Fig. S6, ESI†), but the electron-rich cobaloxime makes reduction of CoII more difficult,35 consequently producing a less attractive catalytic overpotential for this species. The cobaloxime substituted with a withdrawing pyrazine ligand similarly showed a notable catalytic current (Fig. S6, ESI†).
Lever's parameterisation approach was employed to rationalise the EMoCS data.46 This electrochemical parameterisation was originally developed to predict an Mn/Mn−1 redox potential by assuming that all electronic contributions from ligand to metal are additive (eqn (1)).
E1/2(Mn/Mn−1)/V vs. NHE = SM·(ΣEL) + IM | (1) |
The slope, SM, and intercept, IM, are dependent on the metal, oxidation state, geometry of the complex, the spin state, stereochemistry and complex net charge in aqueous solution.46–49 The electrochemical ligand parameter, EL, reflects the net electron-donating character of the ligands.46 The unique advantage of using EL values for each ligand over other parameters, such as pKa values, is that they reflect both the σ and π interactions of a given metal–ligand bond. An EL value defines how these interactions will change the electronic properties of a complex. More electron donating ligands have more negative EL values, as the complex to which they are ligated has a more negative redox potential.
E L values for each axial ligand used in EMoCS were taken from the literature46 or converted from respective Hammett parameters50 (Table 1). A plot of ΣEL against the CoIII/CoII reduction potential then allowed analysis of the underlying electronic change introduced by each axial ligand to the catalyst (Fig. S8, ESI†). The observed correlation produced eqn (2), with SM and IM values similar to those found for RuIII/RuII (SM = 1.14 and IM = −0.35) in aqueous solution.46Eqn (2) can be employed to predict the CoIII/CoII couple of any related aqueous cobalt complex and was subsequently used to generate unreported EL values for 4-methoxypyridine and 3-cyanopyridine (Table 1).
E1/2(CoIII/CoII)/V vs. NHE = 1.08·(ΣEL) − 0.60 | (2) |
E L values could not be used to predict the onset potential of the catalytic wave as it does not shift significantly between the most active cobaloximes (Fig. 4a and Table 1). This observation suggests that there is only a weak interaction between the low valent Co center and the pyridine at this stage of the mechanism (Fig. 5).35,51
Fig. 4 (a) Overlaid catalytic waves from CVs of pyridine-substituted cobaloximes shown in Fig. 3. (b) Correlation between the ligand parameter, EL, of each axial ligand with the baseline-corrected catalytic currents from Table 1. (c) Summary of the analysed pyridine ligands. |
Fig. 5 A mechanistic overview of hydrogen generation from cobaloximes illustrating the proposed effect of the axial ligand on the rate of proton reduction. |
Conversely, EL values do correlate well with the size of the catalytic wave, as more electron donating pyridines produce higher catalytic currents (Fig. 4). This trend is in line with previous mechanistic studies (Fig. 5).52 The reduction of CoII yields a highly nucleophilic CoI species that is quickly protonated to form CoIII–H. CoIII–H is subsequently reduced to a highly basic CoII–H, which is protonated to form H2. Protonation of a Co–H and release of H2 has been proposed to be rate limiting,53 therefore catalytic activity can be tuned by adjusting the basicity of this species. Adding more electron donating substituents to the cobaloxime yields a more basic CoII–H, explaining the linear trend in Fig. 4b.
Eqn (3) was extracted from the plot in Fig. 4b and provides an effective means to predict the activity of a substituted cobaloxime from the respective EL value, which will be a valuable tool for future cobaloxime design. The correlation also demonstrates that activities found in non-aqueous solvents are not directly applicable to an aqueous system. The 4-dimethylaminopyridine substituted Co catalyst was previously described as the most active cobaloxime in organic solvents with Et3NH+ as a proton source.35 However, protonation of the amine residue in water produces the opposite effect and forms an inactive cobaloxime.
Catalytic current/μA = 627·EL − 247 | (3) |
Fig. 6 Single-crystal X-ray structure of compound A with atom ellipsoids of 50% probability showing the atom-labeling scheme for all non-C atoms. Solvent molecules and hydrogen atoms (except O–H) have been omitted for clarity. Monoclinic P21/n, a = 8.000(1), b = 13.5322(2), c = 19.2803(3) Å, β = 92.183(1)°. R1 = 3.87%, wR2 = 11.64%. More information can be found in Table S1 (ESI†). |
Electrochemical analysis was undertaken of each isolated complex. An additional advantage of the EMoCS procedure is that it produced water-soluble cobaloximes, whereas complexes A to C were isolated as complexes with a chlorido-ligand and required dissolution in organic solvents before adding water. All electrochemical experiments of isolated complexes consequently contain 10% acetone for solubility.
Linear sweep voltammograms (LSVs) of each cobaloxime (Fig. S9, ESI†) demonstrated the exact trend in catalytic current seen in Fig. 3, which fully authenticated the EMoCS procedure. Subsequent controlled potential electrolysis (CPE) experiments generated measurable quantities of H2, confirming that the electrochemical currents were a result of prolonged proton reduction. The electrocatalytic activity was quantified through CPE on a GC electrode held at −0.7 V vs. NHE in a degassed 1 mM solution of each complex. The H2 produced per hour was deduced via gas chromatography after electrolysis (Table 2). The amounts of H2 generated again agree with the EMoCS; compounds A and B produced more hydrogen per hour than the less active catalysts, C and D.
Catalyst | H2 μmol h−1 in 0% O2b | H2 μmol h−1 in air |
---|---|---|
a CPE performed at −0.7 V vs. NHE in 9 mL TEOA/Na2SO4 buffer (0.1 M each, pH 7) and 1 mL acetone. b Measured in 2% CH4 in N2. c Percentage in brackets shows Faradaic efficiency. | ||
A | 0.92 ± 0.05 (77 ± 2%)c | 0.89 ± 0.07 (65 ± 6%)c |
B | 0.84 ± 0.06 (75 ± 7%)c | 0.69 ± 0.05 (64 ± 10%)c |
C | 0.78 ± 0.09 (68 ± 4%)c | 0.76 ± 0.03 (59 ± 6%)c |
D | 0.65 ± 0.04 (62 ± 4%)c | 0.62 ± 0.02 (43 ± 2%)c |
CPE was also performed on a Hg-pool electrode for compounds A to D to confirm that a molecular catalyst is responsible for H2 formation. Hg forms amalgams with metallic deposits, thereby preventing the formation of a heterogeneous electroactive deposit on the electrode.57 Hg-pool electrode CPE at −0.7 V vs. NHE yielded at least 10 turnovers from each cobaloxime in the bulk solution, confirming that hydrogen evolution was a result of homogenous cobaloxime catalysis.
CPE was consequently repeated with each catalyst in air. The H2 evolution activity of A–D decreased by typically less than 5% in each case compared to anaerobic conditions (Table 2). The tolerance of cobaloximes towards O2 is outstanding considering the potential competition between proton and oxygen reduction at a low valent Co site,59 which suggests that competing oxygen reduction at the cobaloxime is low under the employed conditions.
On the other hand, the Faradaic efficiency (FE) undergoes a noticeable decrease between inert and aerobic atmospheres (Table 2). The observed decrease is a result of oxygen reduction at the GC electrode, which competes for electrons with the cobaloxime. It appears that the limiting factor in the efficiency of cobaloximes under aerobic conditions is therefore not the catalyst, but the electrode material.
The general decrease in FE from cobaloxime A to D (Table 2) in the presence of air was assigned to their respective activities. The rate of underlying oxygen reduction by the GC remains constant in all experiments, consequently using a complex that operates at a faster rate can lessen the relative loss of the FE. The only other FE reported in air was 52% by a Co corrole catalyst60 and as such, at 65% compound A sets a new benchmark for a discrete synthetic catalyst under this demanding environment.
We also tested the most active catalyst, compound A, (0.2 μmol) for photocatalytic H2 generation in a homogenous, noble-metal free system with an organic dye, Eosin Y (0.1 μmol), and triethanolamine (4.49 mL, pH 7, 0.1 M) as a hole scavenger.12,61,62 Visible light irradiation (λ > 420 nm, 100 mW cm−2, AM 1.5 G) of the solution in a photoreactor (3.4 mL headspace) resulted in the formation of 12.3 ± 0.6 (TONCo = 62 ± 3) and 3.9 ± 0.4 (TONCo = 19 ± 2) μmol H2 after two hours of irradiation at 25 °C under anaerobic and aerobic conditions, respectively (Fig. 7). EMoCS therefore produced a complex that acts as a very active proton reduction catalyst in both electrocatalytic and photocatalytic systems under both anaerobic and aerobic atmospheres.
Fig. 7 Visible light driven (λ > 420 nm, 100 mW cm−2, AM 1.5 G) H2 production by cobaloxime A (0.2 μmol) with Eosin Y (0.1 μmol) in TEOA (4.49 mL, pH 7, 0.1 M) and acetone (10 μL). |
Utilisation of electrochemical parameterisation to differentiate the ligands based on net electron donating character provided a rational analysis of the EMoCS data. A more electron donating ligand increases the rate of proton reduction catalysis due to the formation of a more basic Co–H species. Importantly, this parameterisation allows activity to be predicted in the future.
EMoCS led to the fast identification of a novel catalyst A, which performs better than previously known cobaloximes. Cobaloxime A was isolated and fully characterised and was found to operate electrocatalytically under aerobic conditions with negligible decrease in proton reduction activity compared to an inert atmosphere. The same catalyst also functioned with high efficiency in a homogeneous, noble metal free, photocatalytic scheme.
EMoCS is particularly attractive for labile 3d transition metal ions, such as Co, Ni and Fe, and automating this technique will allow high-throughput combinatorial screening to be undertaken across a library of ligands. The fundamentals of EMoCS are applicable to any molecular electrocatalytic process and it is therefore foreseeable that other attractive catalytic reactions, such as oxygen reduction, CO2 reduction or water oxidation, will benefit from this technique in the near future.
To assemble each cobaloxime, [Co(dmgH)2(H2O)2] (5 mL of 2 mM) in Pi (0.1 M, pH 7) was pipetted into a sample vial and 40 μL of a 1 M stock solution of an axial ligand was injected into the solution.
E 1/2 of CoIII/CoII was taken as the potential at the midpoint of the reduction wave. Standard deviation was calculated using data from at least three fresh solutions. Catalytic current was taken as the difference between the peak and baseline of each catalytic wave.
(4) |
Footnote |
† Electronic supplementary information (ESI) available. CCDC 974386. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c4cp00453a |
This journal is © the Owner Societies 2014 |