Noncovalent Integration of a Bioinspired Ni Catalyst to Graphene Acid for Reversible Electrocatalytic Hydrogen Oxidation.

Efficient heterogeneous catalysis of hydrogen oxidation reaction (HOR) by platinum group metal (PGM)-free catalysts in proton-exchange membrane (PEM) fuel cells represents a significant challenge toward the development of a sustainable hydrogen economy. Here, we show that graphene acid (GA) can be used as an electrode scaffold for the noncovalent immobilization of a bioinspired nickel bis-diphosphine HOR catalyst. The highly functionalized structure of this material and optimization of the electrode-catalyst assembly sets new benchmark electrocatalytic performances for heterogeneous molecular HOR, with current densities above 30 mA cm-2 at 0.4 V versus reversible hydrogen electrode in acidic aqueous conditions and at room temperature. This study also shows the great potential of GA for catalyst loading improvement and porosity management within nanostructured electrodes toward achieving high current densities with a noble-metal free molecular catalyst.

Introduction

The use of hydrogen as a sustainable energy vector requires the development of efficient, low-cost but robust means to produce and oxidize H2 at the cathode of electrolyzers and the anode of fuel cells.[1,2] In state-of-the-art proton exchange membrane fuel cells (PEM-FC), the hydrogen oxidation reaction (HOR) is performed using scarce and expensive Pt- or Pt group metal (PGM) catalysts, which are critical raw materials, thus severely hampering their future deployment.[3−6] In recent years, several studies have reported the use of PGM-free catalyst for HOR, in particular with Ni-based materials under alkaline conditions[7−9] and metal-carbide-based materials in acidic media.[10−14] Although some of these non-PGM based materials could demonstrate good activities for HOR with good resistance to poisoning, they still suffer from limited stability under operation.

Nature’s catalysts for reversible H2 production and oxidation, hydrogenases (H2ase), are able to perform these reactions with very high TOFs (up to 10 000 s–1) using only cheap and widely available Fe and Ni metals at their catalytic site.[15−17] Despite their sensitivity to a number of inhibitors and their high molecular weight, these enzymes have been extensively used for their impressive electrocatalytic properties in H2–O2 biofuel cells over the past decade.[18−20] Strategies involving the use of protecting redox polymer,[21−23] protein–surface orientation[24−26] or electrode nanostructuration[27,28] have been developed to overcome H2ases intrinsic limitations. However, their fragility and sensitivity to various conditions and inhibitors combined with their large molecular footprint remain problematic for catalyst loading optimization and implementation in market-ready PEMFC setup.

Nevertheless, H2ases have provided synthetic chemists with valuable blueprints that allowed synthesizing biomimetic and bioinspired PGM-free compounds able to electrocatalytically generate and/or oxidize H2.[29−33] In particular, the family of the mononuclear Ni-based bis-diphosphine [Ni(P2RN2R)2]2+ complex first described by DuBois’s group represents a unique class of bidirectional molecular catalysts for electrocatalytic H2/H+ interconversion.[34−36] The 1,5-diaza-3,7-diphosphacyclooctane ligand (P2RN2R) provides an electron rich environment to the metal center, while mimicking the 2-azapropanedithiolate bridge found in [FeFe]-H2ases and acting as proton shuttle.[37−40] Over the past decade, an extensive body of work has focused on the expansion of the outer coordination sphere beyond the cyclic tertiary amine, in particular, through the incorporation of amino acid residues to the ligand.[41−44] This allowed the design of an arginine-containing derivative [NiII(P2CyN2Arg)2]7+ (NiArg), owing the strongest bias for HOR of the series, with reported TOFs of up to 106 s–1 under 100 bar of H2 at 72 °C, while also retaining catalytic reversibility for H2/H+ interconversion over a broad pH range (Figure a).[45−47]

Figure 1

Schematic representations of (a) NiArg simplified chemical structure, (b) GA sheets bearing −CO2H anchoring functions, and (c) GA|NiArg composite modified electrode.

In order to reach technological relevance, molecular catalysts for fuel cells or solar fuels production need to be integrated onto electrode surfaces before being eventually implemented within functional devices.[48] This relatively recent research field has been particularly active over the past few years.[49−51] Past developments of covalent or noncovalent grafting strategies for molecular catalysts have allowed great control over catalyst concentration at the material–electrolyte interface,[52,53] electrode-catalyst electron transfer rates and redox properties,[54−56] catalyst activity and stability[57−61] as well as in some cases, catalyst selectivity,[62−64] through the possibility to tune the direct environment of the immobilized catalyst.

In particular, over the last decade, efforts have focused on interfacing this series of Ni-based molecular catalysts with carbon-based electrodes for the development of molecular HER cathodes and HOR anodes.[65−70] Covalent or noncovalent modifications of carbon nanotube (CNT) electrodes with [Ni(P2RN2R)2]2+ allowed to reach current densities of 1–2 mA cm–2 measured at 0.3 V versus reversible hydrogen electrode (RHE) and room temperature in 0.5 M H2SO4 for HOR.[65,66] Recent works reported improved performances reaching current densities of up to 16 mA cm–2, at 0.3 V versus RHE, through design and structuration of the CNT-based electrodes.[68,69] Importantly, such bioinspired electrodes are tolerant to CO,[66] compatible with proton-exchange membrane (PEM) technology, and they have been successfully integrated into fully functional proof-of-concept fuel cell devices.[69,71,72]

In this context, graphene acid (GA) has recently appeared as a particularly appealing platform material for catalysis as it can be easily obtained from commercially available fluorographite. It yields material with high levels of carboxyl functionalization, about 10% (atomic content), directly grafted on the basal plane, allowing the graphene sheet to maintain excellent electron conduction properties (Figure b).[73] These properties were very recently exploited in order to covalently incorporate redox centers through peptide coupling or bind metal nanoparticles in order to perform C–H bond insertion, C–C bond coupling, or even alcohol oxidation through heterogeneous catalytic processes.[74−76]

Here, we describe the use of GA nanosheets as an original electrode material for the noncovalent grafting of NiArg through electrostatic interactions. The highly functionalized and conductive GA provided a large amount of anchoring sites for the catalyst, while ensuring excellent electronic wiring of the molecular catalyst, thus allowing the development of an efficient molecular-based anode for HOR in PEM fuel cells (Figure c).

Results and Discussion

GA synthesis was carried out as previously reported,[73] and the GA-modified electrodes were prepared through vacuum filtration of a 0.05 mg mL–1 of GA dispersion in EtOH directly at the surface of a gas diffusion layer (GDL) coated with a hydrophobic microporous layer (MPL) (area = 10 cm2). The volume of GA dispersion filtrated was varied in order to obtain several GA loadings (from 0.05 to 0.8 mg cm–2 of GA, see Experimental Section). The obtained GDL|GA films were characterized before and after modification with NiArg using X-ray photoelectron spectroscopy (XPS) and scanning electron microscopy (SEM) coupled with energy dispersive X-ray spectroscopy (EDX) mapping (Figure ).

Figure 2

(a) C 1s XPS region of the GA sample (b) SEM micrograph of a GA film deposited at the surface of a GDL.

The C 1s XPS region spectrum of the GA sample shows an important contribution of the carboxylic groups (Figure a). This high content in −CO2H functions (9.9% atomic content determined by XPS, see Table S1) is expected to lead to an efficient grafting of NiArg at the electrode surface through electrostatic interactions with the guanidinium moieties of the catalyst. SEM characterization depicted flake-type microstructures for the GA deposit, with sheet sizes reaching up to several μm (Figures b and S1). EDX mapping of the films clearly shows a high oxygen content on the GA surface, where much lower levels are observed on the MPL layer part (Figure S2).

Catalyst deposition was carried out by drop casting 2 μL of a 5 mM solution of NiArg in deionized water at the surfaces of the GDL|GA electrode with different thicknesses (area = 0.125 cm2) (see Supporting Information for experimental details). The deposit was then dried for 10 min before being rinsed with deionized water to remove unbound catalyst. The modified electrodes were characterized using cyclic voltammetry (CV) in 0.5 M H2SO4 with a constant flow of H2 (5 mL min–1) at the back of the GDL-based homemade working breathing electrode (Figure S3). For all modified electrodes, similar S-shape CV traces could be observed, characteristic of the catalytic behavior of NiArg for reversible H2 production and oxidation under aqueous conditions, with a strong bias for HOR, as previously reported (Figure a).[69] During turnover and following the addition of H2, the catalyst is thought to be reduced from NiII to NiII–H (with a protonated pendant amine) before being reoxidized in a two-electron process at the electrode, giving rise to the observed catalytic current.[45,46,69]

Figure 3

(a) CV traces of GDL|GA electrodes at different GA loadings (0; 0.05; 0.1; 0.2; 0.4, and 0.8 mg cm–2) modified with 2 μL of NiArg (5 mM) and (b) current densities for HER at −0.2 V vs RHE and HOR at 0.1 and 0.4 V vs RHE obtained from CVs in 0.5 M H2SO4 solution purged with argon with a constant flow of H2 at the back of the GDL (5 mL min–1) (ν = 20 mV s–1).

As expected, direct deposition of the catalyst on an unmodified GDL electrode leads to the lowest electrocatalytic responses (9 ± 2 mA cm–2 at 0.4 V versus RHE). For the GDL|GA|NiArg electrodes, catalytic current responses scaled up almost linearly with the GA loading GDL before leveling off at 0.4 mg cm–2 of GA, giving maximum current densities for HOR of 33 ± 6 mA cm–2 at 0.4 V versus RHE, setting up a new benchmark for molecular HOR (Figure b). Maximum HOR current densities of 37 ± 5 mA cm–2 were obtained at 0.4 V versus RHE with 0.8 mg cm–2 of GA films, but they showed limited mechanical stability during catalyst deposition and electrochemical testing because of material leaching off the electrode surface. Bare GA electrodes and GA electrodes modified with Ni nanoparticles did not show any HOR activity (Figure S4a). Although still not competing with low-loaded Pt electrodes (Figure S4b), GDL|GA|NiArg electrodes outperform all previously reported bioinspired nanomaterials for catalytic HOR under acidic conditions.

In order to study the impact of the amount of catalyst deposited at the GDL|GA (0.4 mg cm–2) electrode surface on HOR catalysis, concentration of NiArg in the deposition solution was varied from 1.25 to 10 mM. CVs performed in neutral buffer conditions (0.2 M potassium phosphate, pH 7) and under argon allowed observation of the reversible redox signature of NiArg at E1/2 = 0.03 V versus RHE, corresponding to the 2e–/2H+ Ni-centered redox process, (Figure a) as previously reported.[69] Integration of the oxidation wave allowed to estimate the catalyst loading (Γ) ranging from 9 nmol cm–2 to about 19 nmol cm–2, depending on the concentration of NiArg deposited for the GLD|GA with 0.4 mg cm–2 of GA deposited (Figure b). Higher amounts of Ni (from 14 ± 3 to 45 ± 3 nmol cm–2) were quantified from digested GDL|GA|NiArg films in nitric acid using inductively coupled plasma atomic emission spectroscopy (ICP-AES). These results indicate that only a part (∼40%) of the overall grafted catalyst (detected by ICP) is electrochemically active (detected by CV). XPS measurements (Table S2) on the modified film confirmed the presence of Ni at the surface of the electrode, and EDX mapping could show that NiArg was mainly grafted on the GA modified surface, as expected from the high surface concentration of −CO2H groups (Figure S2b). As a result of the drop cast and the drying method used to deposit the catalyst, some nonspecific interactions can potentially be expected. Thus, this would explain the presence of small amounts of Ni on the hydrophobic MPL and formation of aggregates, as seen on Figure S2c (although uneven distribution shown by EDX could potentially be caused by degradation under the electron beam). Interestingly, the amount of Ni measured by ICP-AES after 10 cycles of CVs under electrocatalytic conditions (Figure b) is closer to the one measured by CV and thus indicates a slow leaching of the catalyst out of the GA film in the electrolyte over the course of time and operation.

Figure 4

Electrochemical characterization of GDL|GA (0.4 mg cm–2) electrodes at different NiArg loadings (2 μL of 1.25; 2.5; 5; and 10 mM) (a) CV traces in 0.2 M phosphate buffer pH 7 under argon (ν = 20 mV s–1) (b) NiArg surface loadings determined from CV experiments (red dots) and from ICP-OES measurements before (black squares) and after (blue square) CV measurements (c) CV traces and (d) corresponding current densities for HER at −0.2 V vs RHE and for HOR at 0.1 and 0.4 V vs RHE obtained from CVs in 0.5 M H2SO4 solution purged with argon and with a constant flow of H2 at the back of the GDL (5 mL min–1) (ν = 20 mV s–1) (see Figure S4 for other GA loadings).

The catalytic performances of the GDL|GA electrodes (0.4 mg cm–2) with different NiArg loadings (Figure c) increase only slightly with the concentration of the NiArg deposition solution up to 5 mM, reaching 33 ± 6 mA cm–2 at 0.4 V versus RHE.

Γ values on GDL|GA|NiArg with lower GA loadings were also obtained from CVs experiments in neutral pH (Figure S5). As expected, similar trends were extracted for thinner GA films of 0.05, 0.1, and 0.2 mg cm–2 but with lower maximum Γ values of 4.8, 9.8, and 12.4 nmol cm–2, respectively (Figure ). The surface loading Γ increases with solution concentration in NiArg, following a simple Langmuir binding isothermwhere A is the density of binding sites available at the GA electrode (nmol mg–1), qGA the amount of GA deposited (mg cm–2), K the association constant between NiArg and the GA electrode surface (L mol–1) and [NiArg], the concentration of NiArg in the deposited solution (mol L–1). The data could be fitted with a single K affinity constant for all series, underlining the effective grafting of the molecular catalyst to the GA-modified electrode (Figure a). The calculated K value of 640 L mol–1 is a low affinity binding constant, in coherence with nonspecific electrostatic interactions between the guanidinium groups of NiArg and the carboxylate of the GA surface. It is also possible that π-cation interactions take place between the guanidinium moieties and the π-conjugated parts of the GA surface, as already suggested with CNTs.

Figure 5

(a) Γ determined from CV experiments for GDL|GA electrodes with 0.05 (black squares), 0.1 (red dots), 0.2 (blue triangles), and 0.4 (green triangles) mg cm–2 of GA as a function of the [NiArg] in the soaking solution; the dashed traces correspond to the fitted binding isotherm for each electrode thickness (b) evolution of the site density as a function of the amount of GA at the surface of the GDL (c) evolution of the values of HOR and HER catalytic currents (see Figure S4) from CV at −0.2 (black squares); 0.1 (red dots) and 0.4 V vs RHE (blue triangles) with the Γ extracted from CV at pH 7 from GDL|GA electrodes with 0.05, 0.1, 0.2, and 0.4 mg cm–2 of GA incubated with 10 mM NiArg.

Interestingly, the A value, corresponding to the number of available anchoring sites per surface area, is decreasing with increasing GA loadings (Figure b). This indicates that anchoring site availability decreases with thicker GA deposits, which stems from a loss of porosity and clogging up of thicker GA deposit. More important, the maximum catalytic current densities for both HER and HOR increase linearly with maximum Γ taken at individual GA loadings (Figures c and S6). However, in the present series, further increase of GA loading leads to a decrease of the relative grafting site availability (Figure b), preventing a linear improvement of the catalyst loading and thus of the obtained catalytic currents (Figure a).

The stability of the best performing and mechanically stable GDL|GA|NiArg electrodes (0.4 mg cm–2 of GA and modified with a 5 mM solution of NiArg) was then studied in chronoamperometry (CA) for HOR at 0.1 and 0.3 V versus RHE (Figure a).

Figure 6

(a) Averaged current values of the CA triplicates of the GDL|GA|NiArg-modified electrodes at 0.3 V (black trace) and 0.1 V (red trace) and (b) CV traces of the GDL|GA|NiArg before (black trace) and after (red trace) CA at 0.3 V vs RHE in 0.5 M H2SO4 solution purged with argon and with a constant flow of H2 at the back of the GDL (5 mL min–1) (ν = 20 mV s–1).

At both applied potentials, a steady decrease could be observed for the GDL|GA|NiArg electrodes over the course of a 1 h experiment, going from 31 ± 6 to 22 ± 5 mA cm–2 at 0.3 V versus RHE and from 11 ± 1 to 7 ± 1 mA cm–2 at 0.1 V versus RHE, retaining, respectively, 71 and 64% of catalytic activity (Figures a and S7). As mentioned above, loss of the catalyst through solution leaching can partly explain this decay and is consistent with the low calculated K value. XPS spectra recorded after electrocatalytic experiments also show that some of the catalyst could have potentially been oxidized at the phosphine ligand (Figure S9), which is another possible reason for nickel releasing out of the electrode. On the other hand, minimal changes are observed in the Ni 2p core level spectra, consisting of a small broadening of the main Ni 2p3/2 peak on the high binding energy side. In addition, CVs performed before and after CA measurements showed slight modification in the electrocatalytic response, with a decrease of the HOR contribution combined with an increase of HER (Figure b). This hints that the decrease in HOR performance could also be due to electrode flooding overtime, hampering H2 diffusion within the active layer rather than decomposition of NiArg. After 1 h of electrolysis at 0.3 V versus RHE, TOFHOR between 3.3 and 11.9 s–1 can be estimated, taking the maximum and minimum values of Γ obtained through ICP measurements, respectively (see above).

Conclusions

We described the use of GA to prepare porous electrodes suitable for supported molecular electrocatalysis. The high degree of functionalization of GA allowed efficient incorporation of a bioinspired nickel-based molecular catalyst through noncovalent electrostatic interactions. The high catalytic current densities, setting up a new benchmark for molecular HOR, were achieved by optimization of the electrode design and catalyst loading. Further optimization of the catalytic layer is underway in order to (i) develop multivalent grafting to prevent leaching, (ii) maximize the number of available active sites through increased film thickness with retention of high porosity as well as (iii) rationalize formulation to prevent electrode flooding and increase the stability of the anode overtime. In the broader context of electrode nanostructuration, this work legitimates the use of GA as a versatile platform for supported molecular electrocatalysis with upscale potential similar to other graphene derivatives for preparation of large surface electrodes.