Metal-free electrocatalytic hydrogen oxidation using frustrated Lewis pairs and carbon-based Lewis acids† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c5sc04564a

The synergistic interaction of a carbon-centred Lewis acid and borane “hydride shuttle” offers a metal-free, CO tolerant pathway to hydrogen oxidation.


Introduction
As the demand for sustainable and carbon-neutral sources of electricity increases, there is a need for new technologies that allow the efficient storage and utilization of energy. 1 H 2 is attractive as an energy vector since energy from renewable sources may be stored in its chemical bond, and then cleanly and safely released as electricity using fuel cell technology. 2 Unfortunately, in the absence of a suitable electrocatalyst, the conversion of H 2 into two protons and two electrons is slow and must be driven by a large overpotential (voltage). Precious metal electrodes (such as Pt) provide an electrocatalytic effect that is indicated by a marked increase in current and a shi in the electrode reaction to a lower potential (voltage). 3,4 However, the high cost and low abundance of such materials presents a signicant barrier to the wide-spread adoption of current H 2 fuel cell technology. There is clearly a need to develop new H 2 oxidation electrocatalysts that are free from precious metals. Progress has been made in this area using bioinspired catalysts 5-7 that contain either Ni [8][9][10] or Fe [11][12][13] centres. However, a signicant weakness of existing electrocatalysts (Pt and the majority of hydrogenase enzyme mimics) is that they are highly sensitive to CO binding and inhibition. 5,14 Trace amounts of CO are inevitably present in H 2 that is commercially produced from hydrocarbon feedstocks. Worse still, for indirect methanol fuel cells (a combined H 2 fuel cell and MeOH reformer) a CO removal process is oen necessary to prevent electrocatalyst poisoning. 15 An alternative metal-free strategy uses frustrated Lewis pairs (FLPs) to activate H 2 . Since their discovery by Stephan's group in 2006, 16 research involving FLPs has grown apace. [17][18][19][20][21][22][23] FLPs, formed from the combination of suitably sterically encumbered Lewis acids (LA) and bases, are precluded from forming classical Lewis adducts; such systems can heterolytically cleave H 2 to generate hydridic and protic components. The hydrogenation of a wide range of functional groups including imines, enamines, nitriles, 24-27 aldehydes, 28 and ketones 29-33 using FLPs has been reported.
In 2014, Wildgoose and Ashley pioneered a new metal-free route to H 2 oxidation using a combined "electrochemical-FLP" approach. 34,35 This enables the conversion of H 2 into two protons and two electrons at cheap and ubiquitous carbon electrodes. Using the archetypal t Bu 3 P/B(C 6 F 5 ) 3 system (Fig. 1a), 34,36 the voltage (driving energy) required to oxidize H 2 was decreased by 610 mV (ca. 118 kJ mol À1 ). Later, we applied this "electrochemical-FLP" approach to Stephan's NHC-stabilized borenium cation (Fig. 1a), 35,37 which decreased the voltage required for H 2 oxidation by 910 mV (ca. 176 kJ mol À1 ). However, a detailed mechanistic study of both these electrochemical-FLP systems revealed several limitations that signicantly hindered their catalytic turnover, efficiency, and application as replacement electrocatalysts for energy applications. This included the side-reaction of radical intermediates with solvent/electrolyte during electrolysis, and the deactivation of electrocatalyst via its reaction with electrogenerated protons. Whilst the borenium cation offered an improvement over the borane system, the rate of H 2 cleavage by this borenium-FLP is far too slow.
Whilst the majority of research involving FLP H 2 activation has been focused on boron-centred Lewis acids, the Ingleson group have recently reported a FLP derived from salts of the Nmethylacridinium cation (1 + ), a carbon-centred Lewis acid, and the Lewis base 2,6-lutidine (lut). 38,39 1 + is inexpensive, easy to synthesise, and is similar in structure to the NADH/NAD + coenzyme system that is found in biological redox systems. [40][41][42][43][44] Furthermore, in 1990, Savéant and co-workers elucidated all the pertinent non-aqueous mechanistic parameters of the 1 + /Nmethylacridane (1-H) redox couple both in the presence and the absence of a Brønsted base. 43,44 The oxidation of 1-H involves an ECE-DISP1 mechanism and results in the net formation of two electrons and an electrogenerated proton (Scheme S1 †). Compared to either of the boron-based electrochemical-FLP systems reported previously, the standard potential of the 1-H/ [1-H]c + couple is relatively low (+0.48 AE 0.01 V vs. Cp 2 Fe 0/+ in MeCN). Also, 1-H is insufficiently hydridic to react with any electrogenerated H + produced, so no competing H 2 evolution reaction (the reverse reaction of H 2 cleavage by the FLP) occurs. Together, these attributes (ease of synthesis, high hydride affinity of 1 + , favourable oxidation potential of 1-H and the lack of side-reactions during electrolysis) combine to make the carbon-based 1-H/1 + system a highly attractive candidate for electrochemical-FLP studies. The only limitation of the 1 + /lut FLP is that the rate of H 2 cleavage is very slowrequiring >9 days for almost complete H 2 activation at 60 C and 4 bar. 38 Fortunately, a solution to this nal problem is available to us. We have recently examined the possibility of using tris[3,5bis(triuoromethyl)phenyl]borane (BArF 18 ) as the Lewis acidic component of an electrochemical-FLP system. 45 The activation of H 2 by BArF 18 -containing FLPs is rapid and favours the formation of the bridging hydride, [(m-H) (BArF 18 ) 2 ] À . 46,47 However, the oxidation potential of [(m-H) (BArF 18 ) 2 ] À is too positive to be useful for electrochemical-FLP applications (ca. +1.55 V vs. Cp 2 Fe 0/+ ) and resembles that of molecular H 2 -BArF 18 is not electrocatalytic towards H 2 oxidation.
In this paper we combine the rapid H 2 cleavage kinetics of BArF 18 -derived FLPs with the stability and efficiency of the carbon-centred Lewis acid, 1 + . Using this approach, the bridging hydride, [(m-H) (BArF 18 ) 2 ] À , effectively functions as a redox inactive "hydride shuttle" to generate 1-H from 1 + (Fig. 1b). As we demonstrate herein, the "hydride shuttle" combines the rapid cleavage of H 2 by the BArF 18 /lut FLP with the favourable electrochemical properties of 1-H. This provides an improved electrocatalytic system, with numerous advantages over previous electrochemical-FLP systems: a ca. 1 V decrease in the voltage for H 2 oxidation at a carbon electrode; a metal-free system that is catalytic in 1 + , turns over efficiently and can be recharged multiple times; no undesirable H 2 evolution sidereactions and a marked improvement in FLP H 2 cleavage kinetics compared to carbon-based Lewis acids alone. We also demonstrate that, in stark contrast to conventional H 2 oxidation electrocatalysts, this electrochemical-FLP system is tolerant of CO.
For proof of concept, a sample of 1[BArCl] (1.0 equivalent), BArF 18 (2.3 equivalents) and 2,6-lutidine (1.7 equivalents) in CD 2 Cl 2 were combined. 2,6-Lutidine was chosen as the Lewis base because it is known to be compatible with 1 + and allows direct comparison to previous work. 38 Importantly, in a control experiment 2,6-lutidine was found to be compatible with BArF 18 as a FLP, with no evidence for adduct formation observed by NMR spectroscopy when an equimolar mixture of BArF 18 and 2,6-lutidine was le for 2 days in CD 2 Cl 2 (Fig. S8-S10 †). On exposure of the three component mixture to H 2 (4 bar) at room temperature, the progress of 1-H formation was monitored by the disappearance of the CH signal of 1 + (at d 9.4 ppm) and the appearance of the CH 2 signal of 1-H (at d 3.9 ppm) in the 1 H NMR spectrum ( Fig. 3 and S4 †). Aer only 25 minutes at 20 C, 30% of 1[BArCl] had been converted to 1-H; quantitative conversion was achieved aer 17 hours. This represents a signicant improvement in H 2 cleavage rate compared to 1 + / lut in the absence of BArF 18 , which requires over 9 days of heating at 60 C before it approaches completion. 38 Additionally, no evidence for CO binding was observed via NMR spectroscopy when the three component mixture (1 + /BArF 18 /lut) was sparged with pure CO gas for 30 seconds (Fig. S5 and S7b †). On admission of excess H 2 to the sample headspace, the usual formation of 1-H occurred with no discernible signals corresponding to a formyl-borate species (Fig. S7 †). 48 This suggests that, in contrast to Pt or bioinspired organometallic electrocatalysts for H 2 oxidation, 5,14 our electrochemical-FLPs are CO tolerant and are not poisoned or otherwise inhibited, even in the presence of signicant CO.

Electrochemical-FLP experiments
Cyclic voltammetry was performed at a glassy carbon electrode (GCE) on solutions of 1-H in CH 2 Cl 2 containing 0.1 M [ n Bu 4 N] [B(C 6 F 5 ) 4 ] as a weakly-coordinating supporting electrolyte. In the absence of Brønsted base, cyclic voltammograms (CVs) of 1-H exhibit a single-electron oxidation wave that is devoid of a back-peak (appears to be irreversible) until scan rates exceed 300 mV s À1 (Fig. S11 †). In the presence of excess 2,6-lutidine, electrochemical reversibility is lost at all scan rates (Fig. S12 †) and the peak current obtained for 1-H approximately doubles (Fig. 4) a 2-fold increase in peak current is observed at 50 mV s À1 and a 1.7-fold increase is observed at 2000 mV s À1 . This effect is highly indicative of an underlying ECE-DISP1 mechanism, as reported by Savéant and co-workers previously. 43 A peak potential of +0.47 V vs. Cp 2 Fe 0/+ was obtained for 1-H at the 100 mV s À1 scan rate. This is represents a 1 V decrease in the potential that is required for H 2 oxidation at a GCE, a very signicant energy saving that is equivalent to ca. 197 kJ mol À1 , and provides a further 110 mV improvement over the previous most suitable borenium-based electrochemical-FLP system. 35 Note that 2,6-lutidine has the added benet of being electrochemically inactive within the potential window of our electrolyte system. This is unlike the phosphine and aliphatic amine bases used in our previous electrochemical-FLP studies which oxidize at similar potentials to the borohydrides, leading to electrode passivation and failure of the system.

Applied H 2 oxidation and electrocatalyst recyclability
The 1 + /BArF 18 /lut system was next applied towards the in situ oxidation of H 2 with the intention of investigating whether the electrocatalyst (1 + ) could participate in successive charging and discharging cycles. The advantage of using BArF 18 as a hydride shuttle is that the oxidation potential of [(m-H) (BArF 18 ) 2 ] À is on the limit of the oxidative potential window, and does not interfere with the measurement of 1-H concentration at the electrode surface.
A sample of 1[B(C 6 F 5 ) 4 ] was electrosynthesised via the controlled-potential bulk electrolysis of 1-H (0.1 equivalent) in the presence of excess 2,6-lutidine (11 equivalents) at a Toray carbon paper electrode. An initial CV scan of 1-H (recorded at a GCE) produced a peak current of 153 mA, and 8.03C of charge was passed during the initial bulk electrolysis stepthis data is represented by the dotted line in Fig. 5. The formation of 1 + was further indicated by the solution turning bright yellow.
An equivalent of BArF 18 was added (relative to the catalyst, 1 + , which is present at 10 mol%) and the sample was sparged with H 2 gas for 20 minutes before a CV was recorded at the GCE. The CV clearly demonstrated the regeneration of considerable amounts of 1-H, even at this short sparging time, with the peak current for this rst H 2 activation cycle at 46% (in agreement with the NMR studies above) of that passed for the original 1-H sample prior to bulk electrolysis. The sample was electrolyzed back to 1[B(C 6 F 5 ) 4 ], passing 3.28C of charge (41% of that passed for the original sample). H 2 activation was then repeated for the second time (again, with only a 20 minute sparge) at which point the observed peak current was comparable to that obtained for the rst H 2 activation. On repeating the H 2 activation a third time, the peak current and charge passed for 1-H during bulk electrolysis was somewhat diminished compared to the initial two attempts (to ca. 20% of the original sample values). A  fourth H 2 activation attempt was unsuccessful, with no regeneration of 1-H.
It was suspected that the system was no longer turning over due to the depletion of 2,6-lutidine via its sequestration by protons generated during the bulk electrolysis of 1-H and also in the H 2 activation cycles by the FLP. In a fuel cell, H 2 oxidation constitutes only one half-reaction of the redox couple; the other half-reaction, O 2 reduction, would consume any protons that are generated during H 2 oxidation and regenerate the Brønsted base. Thus, at this point in the experiment, the number of regeneration cycles was limited by the quantity of available 2,6lutidine. To overcome this issue, an additional 10 equivalents of 2,6-lutidine were added to the sample, which was then subjected to a further 20 minute sparge with H 2 . Reassuringly, this h H 2 activation run successfully regenerated 1-H in similar concentrations (ca. 45% of the original sample concentration aer a 20 minute sparge) to those obtained during the rst two H 2 activation attempts. The sample was then subjected to bulk electrolysis.
To investigate the effect of exposing the sample to H 2 for longer periods of time, the sample was le sealed under H 2 for 2.5 days. To great surprise, the resulting CV (cycle 6) exhibited a 1.5-fold increase in peak current compared to the original 1-H sample. It is likely that excess [lutH][(m-H) (BArF 18 ) 2 ] builds up in solution once all 1 + (present at 10 mol% cf. the borane) has been converted back to 1-H. As 1-H undergoes oxidation at the electrode surface, the electrogenerated 1 + is rapidly converted back to 1-H via reaction with the excess [(m-H) (BArF 18 ) 2 ] À . This leads to an enhancement in the peak current of the 1-H oxidation wave i.e. a perceived electrocatalytic effect. This effect was conrmed experimentally by treating a sample of 1-H with increasing quantities (0, 0.5, 1, and 2 equivalents) of the hydride donor [ n Bu 4 N][HB(C 6 F 5 ) 3 ] (Fig. S13a †). The addition of [ n Bu 4 N] [HB(C 6 F 5 ) 3 ] resulted in a proportional increase in the peak current of the 1-H wave (Fig. S13b †). Note that whilst [HB(C 6 F 5 ) 3 ] À is redox active, its peak potential is observed at +0.88 V vs. Cp 2 Fe 0/+ and therefore does not interfere with the 1-H oxidation wave. The fact that the peak current of 1-H increases, with no observable wave corresponding to [HB(C 6 F 5 ) 3 ] À , suggests that hydride shuttling occurs within the timescale of the electrode processi.e. the system is not only rechargeable, but it is catalytic and turning over many times per H 2 -charge cycle. Digital simulation of this electrochemical data determined the turnover frequency of the hydride shuttling process to be 2.7 AE 0.2 Â 10 4 s À1 .
Henceforth, excess 2,6-lutidine (10 equivalents) was added aer each bulk electrolysis step to ensure that that system recyclability was not limited by the concentration of Brønsted base. Following bulk electrolysis, the sample containing 1 + was subjected to further H 2 activation (recharging, cycle 7) and bulk electrolysis cycles (discharging) until 1-H could no longer be regenerated. Only one successful regeneration cycle was performed before no further 1-H formation was observed. Since the 2,6-lutidine concentration was not the limiting factor, it is likely that the deactivation of the electrocatalytic system resulted from the decomposition of the boron-based Lewis acid, BArF 18 (whose concentration was not altered from the initial experiment in the series), over the course of several charging and discharging cycles. Indeed, BArF 18 is relatively sensitive to trace amounts of adventitious air and moisture. Despite this, the 1 + / 1-H carbon-based Lewis acid system was conrmed to still be fully active when the addition of [ n Bu 4 N][HB(C 6 F 5 ) 3 ] resulted in successful recovery of the oxidation wave corresponding to 1-H.

Conclusions
The 1 + /BArF 18 /lut system provides a new and improved electrochemical-FLP approach to H 2 oxidation by combining the best attributes of two different Lewis acids: one carbon-based with excellent electrochemical attributes, and one boron-based with excellent H 2 activating attributes as part of a FLP. Unlike conventional, precious metal or biomimetic electrocatalysts, this system is highly tolerant to CO. The pre-activation of H 2 in the form of 1-H results in an astonishing 1 V decrease in the potential that is required for H 2 oxidation at ubiquitous carbon electrodes. This represents a signicant decrease in the required energetic driving force for H 2 oxidation (equivalent to ca. 197 kJ mol À1 ) and a further 110 mV improvement over previous electrochemical-FLP systems. In addition to this (and in contrast to our previous electrochemical-FLP systems) there are no H 2 evolution side-reactions due to the reaction of incoming hydride with electrogenerated H + ; this leads to a marked improvement in efficiency and recyclability.
The completely metal-free system is electrocatalytic with respect to the carbon-based Lewis acid 1 + and can be turned over multiple times without any loss of activity. The "hydride shuttle" effect provided by the synergistic interaction of BArF 18 and 1 + gives rise to a signicant improvement in the overall Fig. 5 The peak current obtained at a GCE (left y-axis, red) and the charge passed at a Toray carbon paper electrode (right y-axis, blue) after sparging a freshly generated 1[B(C 6 F 5 ) 4 ] (3.7 mM, 10 mol%) solution in CH 2 Cl 2 with H 2 for 20 minutes in the presence of 2,6lutidine (41 mM, 11 equivalents) and BArF 18 (37 mM, 1 equivalent). The dotted line represents the peak current/charge passed for the original 1-H sample, which was converted to 1[B(C 6 F 5 ) 4 ] via bulk electrolysis, and provides a reference point for the following H 2 activation cycles. Additional 2,6-lutidine was added from cycle 4 onwards.
rates of H 2 cleavage and the generation of 1-H by the carbonbased FLP.
We see two routes to further improve this electrochemical-FLP system. One, to develop a boron-based FLP that exhibits a greater stability to air and moisture whilst retaining the ability to rapidly cleave H 2 and to function as an efficient "hydride shuttle". Indeed, we have already demonstrated that solutions of B(C 6 F 5 ) 3 in 1,4-dioxane can be rendered water tolerant simply by operating at increased pressures of H 2 . 33 Alternatively, an analogous carbon-based Lewis acid is required that is capable of rapid H 2 activation when combined with a suitable Lewis base, without requiring the presence of any additional boron-based Lewis acid as a hydride shuttle. Both approaches form part of our ongoing research efforts.