Afridi
Zamader‡
ab,
Bertrand
Reuillard
*a,
Julien
Pérard
a,
Laurent
Billon
c,
Gustav
Berggren
*b and
Vincent
Artero
a
aUniv Grenoble Alpes, CNRS, CEA, IRIG, Laboratoire de Chimie et Biologie des Métaux, 17 rue des Martyrs, 38000 Grenoble, France. E-mail: bertrand.reuillard@cea.fr
bDepartment of Chemistry – Ångström Laboratory, Uppsala University, Box 523, SE-75120 Uppsala, Sweden. E-mail: gustav.berggren@kemi.uu.se
cUniversite de Pau et Pays de l’Adour, E2S UPPA, IPREM, Bio-inspired Materials Group: Functionalities & Self-Assembly, 2 avenue Angot, 64053 Pau, France
First published on 23rd August 2023
Integration of molecular catalysts inside polymeric scaffolds has gained substantial attention over the past decade, as it provides a path towards generating systems with enhanced stability as well as enzyme-like morphologies and properties. In the context of solar fuels research and chemical energy conversion, this approach has been found to improve both rates and energy efficiencies of a range of catalytic reactions. However, system performance still needs to be improved to reach technologically relevant currents and stability, parameters that are heavily influenced by the nature of the incorporated molecular catalyst. Here, we have focused on the integration of a biomimetic {Fe2(μ-adt)(CO)6} (–CH2NHCH2S–, azadithiolate or adt2−) based active site (“[2Fe2S]adt”), inspired by the catalytic cofactor of [FeFe] hydrogenases, within a synthetic polymeric scaffold using free radical polymerization. The resulting metallopolymers [2Fe2S]adtk[DMAEMA]l[PyBMA]m (DMAEMA = dimethylaminoethyl methacrylate as water soluble monomer; PyBMA = 4-(pyren-1-yl)-butyl methacrylate as hydrophobic anchor for heterogenization) were found to be active for electrochemical H2 production in neutral aqueous media. The pyrene content was varied to optimize durability and activity. Following immobilization on multiwalled carbon nanotubes (MWNT) the most active metallopolymer, containing ∼2.3 mol% of PyBMA, could reach a turnover number for hydrogen production (TONH2) of ∼0.4 ×105 over 20 hours of electrolysis at an overpotential of 0.49 V, two orders of magnitude higher than the isolated catalyst counterpart. The study provides a synthetic methodology for incorporating catalytic units featuring second coordination sphere functional groups, and highlights the benefit of the confinement within the polymer matrix for catalytic performance.
In nature, [FeFe]-hydrogenase enzymes are capable of H2 production at high catalytic rates (∼104 moles of H2 per mole of enzyme per second) with negligible overpotential requirement at neutral pH using only earth-abundant metals.7 Their remarkable activities are enabled by a hexa-nuclear iron cofactor denoted the “H-cluster”.8 The latter consists of a [4Fe–4S] cluster, coupled to a diiron active site ([2Fe]H) where the catalysis takes place. The Fe ions of [2Fe]H are each coordinated by one terminal CO ligand and one terminal CN− ligand, and a third CO ligand bridges the two ions.9 The two Fe ions in [2Fe]H are further bridged by an azadithiolate (–SCH2NHCH2S–, adt2−) ligand, in which the bridgehead amine is critical for efficient proton delivery during catalysis.10–13 In recent years, these fascinating biomolecules have been extensively studied on electrode surfaces and their potential application in “biohybrid devices” investigated.14–16 In parallel, [FeFe]-hydrogenases have served as blueprints for the development of molecular bioinspired catalysts.7,17–22 In particular, extensive efforts have been invested in [2Fe]H models of the general structure {Fe2(μ-SR)2(CO)6} (RS− = thiolate ligand).7,23 Nevertheless, few catalysts in this di-iron family have been reported to display substantial activities when heterogenized on electrodes or incorporated within synthetic macromolecular chains mimicking proteins. A wide range of strategies have been reported in the literature by Darensbourg et al.,24,25 Talarmin et al.,26 Artero et al.,27 Pickett et al.,28 Liu et al.29–31 targeting immobilization at carbon-based or metal electrodes via covalent grafting or simple physisorption. However, most of these systems showed rapid loss of activity either following the degradation of the linkage or the active sites. More recently Dey and coworkers reported exceptional improvements in the activity of immobilized {Fe2(μ-adt)(CO)6} based complexes,32–34 following physisorption32,33,35,36 and covalent grafting34 on edge plane graphite (EPG) electrodes in aqueous media. However, both strategies yielded relatively low catalyst loadings (lower than 1 nmol cm−2). Recently, our groups demonstrated the simple and efficient immobilization of a pyrene containing naphthalene bridged {Fe2(μ-SR2)(CO)6} complex on multiwalled carbon nanotubes (MWNT).37 The presence of pyrene enabled catalyst immobilization through π–π interactions, resulting in a largely improved surface loadings (up to 9.5 nmol cm−2) and excellent stability of the anchoring group on the electrode during electrolysis at neutral pH. However, the latter system displayed moderate activity due to the degradation of the active site during catalysis.37
In parallel, the integration of such diiron based complexes within organic polymer scaffolds has emerged as a promising approach to improve their stability and catalytic properties.38–43 Among such metallopolymers, Brezinski et al. reported a striking example with the integration of a naphthalene bridged {Fe2(μ-SR)2(CO)6} complex inside a poly-(2-(dimethylamino)ethyl methacrylate) (PDMAEMA) polymeric scaffold. The resulting metallopolymer [2Fe2S]-PDMAEMA showed excellent activities at neutral pH under homogeneous conditions.42,43 These outstanding activities were rationalized by the presence of the PDMAEMA which was proposed to provide not only water solubility but also ensuring efficient proton transfer during catalysis as a polybase. We could further demonstrate that redesigning of the aforementioned metallopolymers to include a pyrene anchoring group, yielding [2Fe2S]-P(DMAEMA-r-PyBMA) (PyBMA = 4-(pyren-1-yl)-butyl methacrylate) afforded a system showing high electrocatalytic H2 production activities at neutral pH once immobilized at MWNT based electrodes.44 However, the catalytic performance of the latter system was limited by the steady degradation of the diiron catalyst over the course of the electrolysis.
In this study, we explored the possibility of improving the performance of such pyrene containing metallopolymers for electrochemical H2 production through the re-design of the di-iron catalyst “active site”. More specifically, we strived to incorporate a potentially more robust catalytic core within the polymer based on previously reported molecular designs,32,33 and we report the successful integration of a {Fe2(μ-adt)(CO)6} based catalyst within a polymer scaffold combining both DMAEMA and PyBMA co-monomer units via free radical polymerization using azobisisobutyronitrile (AIBN). To the best of our knowledge, this is the first report of a {Fe2(μ-adt)(CO)6} derivative active site successfully incorporated within a multifunctional synthetic macromolecular chain. The resulting original metallopolymers were active for H2 production at neutral pH (Fig. 1A). The influence of the polymer scaffold was elucidated by comparing the activity of the metallopolymers to that of a molecular pyrene modified isolated analogous diiron catalyst.
![]() | ||
Fig. 1 (A) Schematic representation of synthesized compounds, i.e., complex 1 (synthetic precursor of the metallopolymers), metallopolymer 1a–c and complex 2, (B) UV-vis spectra of 1a–c (197; 50 and 20 μg mL−1) and 2 (11 μg mL−1) in DMF (C) Infrared spectroscopy of 1, 1a–c and 2 in CHCl3 using transmission mode. The intensity of FTIR traces of 1a were multiplied by 3 and 1b and 1c multiplied by 10 each to have between comparison with 1 and 2. Here k![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() |
The synthesized compounds (i.e., metallopolymers 1a–c and complex 2) were characterized using a range of spectroscopic techniques. In all three metallopolymers, the 1H NMR exhibited peaks between 7.0–7.3 ppm due to the aromatic group present at amine bridgehead of the diiron active site. Compared to spectra collected for complex 1 (Fig. S1†), 1a–c showed negligible shifts in peak positions which suggest that the polymerization did not cause significant changes in the electronic distribution at the metal center. In addition, 1a–c showed broad peaks between 7.8–8.4 ppm, characteristic of pyrene, and multiple peaks in the aliphatic region (0–5 ppm), originating from DMAEMA (Fig. S2–S5†). Integration of the previously mentioned signature peaks allowed to calculate the ratio between each co-monomer (k:
l
:
m), as shown in Table 1, and estimate the composition of the metallopolymers' backbones. This analysis indicated the presence of ∼0.9, ∼2.3 and ∼6.6 mol% of pyrene in 1a, 1b and 1c respectively (Table 1). As expected, the 1H NMR spectrum of the complex 2 revealed distinct peaks ranging from 3.5 to 5.5 ppm, 7.2 to 7.9 ppm, 8.0 to 8.5 ppm, attributable to the protons present on the aliphatic, bridgehead aromatic and pyrene groups, respectively (Fig. S7†).
Metallopolymer or complex | DMAEMA (l)/PyBMA (m)/diiron site (k) final molar ratioa | mol% of PyBMA | Fe contentb (mgFe gpol−1) | M n (g mol−1) | Đ |
---|---|---|---|---|---|
a Calculated from 1H NMR spectrum. b Measured by ICP-OES. c M n = molecular weight of the metallopolymers and Đ = dispersity, both measured by SEC. | |||||
1a | 37/0.35/1 | 0.9 | 9.6 ± 0.2 | 14![]() |
2.7 |
1b | 62/1.5/1 | 2.3 | 4.10 ± 0.06 | 16![]() |
2.19 |
1c | 53/3.8/1 | 6.6 | 4.0 ± 0.1 | 17![]() |
2.25 |
2 | — | — | 0.16 ± 0.01 | — | — |
UV-vis spectroscopy further supported the presence of the pyrene in both the metallopolymers 1a–c and complex 2 by showing multiple sharp peaks between 255–355 nm attributed to the pyrene group (Fig. 1B). As the relative pyrene content in 2 is much larger compared to 1a, 1b and 1c, a noticeable enhancement of the intensity of the corresponding pyrene peaks was observed. Infrared spectroscopy allowed to clearly observe the CO stretching bands at 2074, 2036 and 2001 cm−1, confirming the presence of the organometallic active site inside the polymers (Fig. 1C). The peak positions were identical to those observed for complex 1, again indicating that the polymer had negligible effects on the electronic structure of the di-iron complex, in agreement with the 1H NMR data. Calibrating the absorbance of these infrared bands with that of complex 1 revealed the presence of ∼1.6 active sites per mol of 1a and ∼1 active site per mol of both 1b and 1c (Fig. S10, S11 and Table S2†). In addition, a strong peak at 1726 cm−1 was observed for all polymers, attributable to carbonyls from the ester groups present in the polymeric backbone, as expected this band was not present in the precursor 1 (Fig. S12†).44 The DMAEMA groups in this type of metallopolymers have been previously shown to have a pKa of ≈8.4.43 To further elucidate the protonation chemistry of 1a–c, the basicity of the adt-amine in the metallo-polymers was probed using 1b as a representative model, utilizing attenuated total reflection ATR-FTIR spectroscopy. It has been previously reported that below pH 4.4 in aqueous media, and under strong acidic condition in organic media, the bridgehead amine of Fe(μ-adt)(CO)6 and closely related complexes begin to undergo protonation, resulting in a distinct hypsochromic shift of 20 cm−1 in the stretching band of CO ligands coordinated to Fe.46–48 Interestingly, within the pH range of 2 to 8, no obvious shift of the CO bands is observed, indicating that the bridgehead amine remains in deprotonated state. However, at pH 0, a marginal hypsochromic shift of 3 to 8 cm−1 was noticed (Fig. S13†). To identify the origin of the latter shift, a comparative study was carried out using a similar metallopolymer, [2Fe2S]-P(DMAEMA-r-PyBMA),44 where the active site lacks the bridgehead amine. The results revealed that the later metallopolymer exhibited only a minor 1–2 cm−1 hypsochromic shift in the CO bands when transitioning from pH 7 to pH 0 (Fig. S14†). This finding suggests that observed shift (3–8 cm−1) for 1b could be due to the protonation of the bridgehead nitrogen at pH 0, albeit we note that the shift is unusually small.
Complex 2 also exhibited the expected carbonyl ligand bands, with wavenumbers of 2075, 2037, and 2002 cm−1 along with a weak carbonyl stretching band at 1659 cm−1 attributed to the presence of the amide bond (Fig. 1C and S15† orange trace). Inductively coupled plasma optical emission spectrometry (ICP-OES) was further employed to quantify the concentration of Fe in the polymers and 9.6 ± 0.2 mg of Fe per gram polymer (mgFe gpol−1) was estimated for 1a while 4.10 ± 0.06 and 4.0 ± 0.1 mgFe gpol−1 were estimated for 1b and 1c, respectively. The discrepancy between the Fe content estimated by NMR relative to the ICP quantification (about 36–55% of expected Fe could be detected in ICP) can be rationalized by the relatively small NMR signals used for the quantification. However, degradation of a small fraction of the Fe–CO core during the polymerization reaction cannot be fully excluded, as previously proposed.45 Similarly, complex 2 exhibited 0.16 ± 0.01 g of Fe per gram of complex, within the error margin of the expected value (0.15 g of Fe per gram of complex). Size exclusion chromatography (SEC) of metallopolymer 1a–c showed that the molecular weight varied between 14200–17
300 g mol−1 with dispersity (Đ) ranging from 2.2 to 2.7. The dispersity values indicate a degree of length inhomogeneity in the metallopolymers (i.e. Đ > 1), which can be ascribed to the nature of the employed polymerization method, i.e., free radical polymerization (Fig. S16,† and Table 1). The combination of ICP-OES and SEC data also provides a second measure of active site incorporation, suggesting ∼1.2 active sites per mole for 1a and ∼0.6 active sites per mole of both 1b and 1c, in line with the estimates from infrared spectroscopy discussed above (Tables 1 and S2†).
Metallo polymer | Homogeneous condition | Heterogeneous condition | |||
---|---|---|---|---|---|
Onset potentiala (V vs. RHE) | Current density at −0.65 V vs. RHEb (mA cm−2) | Onset potentiala (V vs. RHE) | Current density at −0.6 V vs. RHE (mA cm−2) | Loading as per electrochemistry (nmol cm−2) | |
a The onset potential is calculated as a potential where catalytic current deviates ∼10% from the capacitive current of respective traces. b The standard deviations for homogeneous condition were marginal. c nd = not determined. | |||||
1a | −0.34 | −2.42 | −0.32 | −36 ± 1 | 3.7 ± 0.3 |
1b | −0.35 | −1.55 | −0.33 | −28 ± 1 | 2.7 ± 0.4 |
1c | −0.36 | −1.17 | −0.33 | −24 ± 2 | 2.3 ± 0.4 |
2 | ndc | nd | −0.37 | −8 ± 2 | 13.9 ± 0.6 |
The electrocatalytic properties of the modified electrodes were then characterized in 0.2 M phosphate buffer at pH 7. From CV measurements, the 1a–c/MWNT modified electrodes displayed an onset potential ranging from −0.32 V to −0.33 V vs. RHE, and reached high current densities ranging from −24 ± 2 to 36 ± 1 mA cm−2 at −0.6 V vs. RHE. As observed in homogeneous conditions, the highest currents densities were observed with 1a followed by 1b and 1c (Fig. 2B and Table 2). This can be rationalized by a higher surface loading of electrochemically active diiron catalyst. Importantly, switching from homogeneous to heterogeneous electrocatalysis with the metallopolymer, resulted in a significant drop in overpotential requirement and increase in catalytic current (Table 2). This is expected as diffusion limitations of the catalyst are circumvented following the immobilization of the metallopolymer at the electrode,44 and could also indicate improved electronic communication between the electrode surface and a large fraction of the catalytic sites. In contrast, the 2/MWNT composite exhibited a notable cathodic shift in the onset potential of approximately 40–50 mV (i.e., −0.37 V vs. RHE), along with a significantly lower current density of −8 ± 2 mA cm−2 at −0.6 V vs. RHE compared to the metallopolymers under equivalent conditions. The substantial enhancement in the activity of the metallopolymers, despite a lower loading of active sites by 5–8 times compared to that of complex 2 (as per loading from ICP-OES), emphasizes the crucial role of the outer coordination sphere provided by the polymeric scaffold (Fig. 2B, orange trace, Table 2).44
At the beginning of the CA experiment, 1a/MWNT showed an initial current density of −2.7 ± 0.2 mA cm−2 (at t = 1 min to minimize capacitive current contribution). The current density was found to decline to −0.3 ± 0.1 mA cm−2 during the initial six hours and then remained fairly stable, however quite close to the current obtained for bare MWNT (−0.17 mA cm−2). This corresponds to retention of only 8 ± 1% of the initial current density (Fig. 3A, red trace, Table S5†). Increasing the pyrene content of the metallopolymer to ∼2.3 mol%, resulted in a significant improvement in stability of the catalytic current. 1b/MWNT showed an initial current density of −2.8 ± 0.1 mA cm−2, comparable to 1a/MWNT, which dropped to −0.66 ± 0.12 mA cm−2 during the initial three hours. Following this initial current loss, the current density remained stable and 24 ± 2% of the initial current density was retained after 20 hours of electrolysis (−0.65 ± 0.10 mA cm−2). Further increase in pyrene content, up to 6.6 mol% (1c/MWNT) caused a substantial drop in initial activity to −0.9 ± 0.1 mA cm−2 (Fig. 3A, blue trace, Table S5†). The electrocatalytic activity of 1c/MWNT then followed a similar trend to 1b/MWNT with a steady decrease of current during the three first hours to then stabilize around −0.35 ± 0.08 mA cm−2. Interestingly, this current could be retained for at least 20 hours, corresponding to 38 ± 13% of the initial current density (Fig. 3A, green trace, Table S5†). Thus, following an initial current loss during the first three hours, the current density was remarkably stable for 1b and 1c. Indeed, in the case of 1b and 1c about 90–95% of the current densities observed after three hours, were retained after 20 hours. This represents a substantial improvement from the recently studied metallopolymers by our group where stable catalytic activity could not be reached within 20 hours of CA.44 CA of the 2/MWNT exhibited an initial current density of −0.9 ± 0.1 mA cm−2 (at t = 1 min), gradually decreasing to −0.20 ± 0.01 mA cm−2 (very close to blank i.e. −0.17 mA cm−2). Overall, retention of 22 ± 4% of the initial current density was noted after 20 hours of CA (Fig. 3A, orange trace, Table S5†). The Faradaic efficiencies were estimated between 98–100% for all 1a–c, 2/MWNT based electrodes. As a result, 30 ± 1, 54 ± 1, 35 ± 2 and 16 ± 1 μmol of H2 corresponding to TONH2 of 1.3 ± 0.1 × 104, 3.8 ± 0.3 × 104, 3.0 ± 0.4 × 104 and 0.6 ± 0.1 × 103 were estimated for 1a, 1b, 1c and 2 respectively, after 20 hours of electrolysis (Fig. 3B, S25, Tables S5 and S6†). The corresponding turnover frequency (TOFH2) values, determined after the first hour of CA (reached the maximum value), were 0.8 ± 0.4, 0.95 ± 0.20 and 0.8 ± 0.2 for 1a, 1b and 1c, respectively. In contrast, complex 2 reached a maximum TOFH2 of 0.008 ± 0.001 s−1, almost two orders lower than metallopolymers during CA. The TOFH2 of the metallopolymers on longer time-scales can be found in the ESI† (Fig. S30 and Table S6†). It should be noted that the values of TONH2 and TOFH2 were estimated considering the total diiron active sites detected as per ICP-OES measurements and are thus most likely conservative.
In order to understand the loss of current, a post-operando analysis using CV, UV-vis and ICP–OES was performed on the catalysts-functionalized electrodes (Fig. S31–S33†). UV-vis spectra, collected for all catalysts following desorption from the MWNT film, showed negligible loss of pyrene after the electrolysis (Fig. S31†).37,44 This observation again underscores the strength of pyrene as an anchoring group on MWNT, as observed before for related diiron complex37 and closely related metallopolymers.44 Conversely, CV traces recorded in aqueous media after 20 hours of electrolysis revealed 70–90% loss and almost complete loss of the catalytic current (at −0.49 V vs. RHE) for 1a–c/MWNT and 2/MWNT, respectively (Fig. S32†). ICP-OES measurements of the iron content evolution of 1b/MWNT at different time points of the CA showed a continuous loss of the iron concentration on the electrode i.e., ∼12%, ∼40%, ∼70% loss compared with the initial concentration after 1, 3 and 20 hours of CA, respectively (Table S6†). Thus, the long-term stability of the 1a–c/MWNT metallopolymers can be rationalized by the relatively high retention of Fe during extended electrolysis. Indeed, the apparent stability of the active sites is significantly improved in 1a–c/MWNT when compared to the previously reported quinone-based [2Fe2S]-P(DMAEMA-r-PyBMA) (PyBMA = 4-(pyren-1-yl)-butyl methacrylate) metallopolymers.44 Still, we note that the loss of current density of 1b/MWNT only correlates weakly with the loss of Fe (Fig. S33 and Table S7†). The non-linear correlation between retained active sites (Fe ions) and activity again suggests the presence of different catalyst populations showcasing slightly different activities, where a relatively small fraction of highly active catalytic sites is lost relatively rapidly while slower but more robust catalysts remain active for several hours of electrolysis. The latter likely corresponds to a population of catalyst more deeply confined within the porous MWNT network or the polymer matrix bearing the active centre(s). The stable catalytic current are observed after three hours of CA could be attributed to the presence of (a) relatively stable population(s) of active sites that are not detected using CV measurement. Transformation of the CV-detectable [2Fe2S]adt active sites to yield this second population of the active sites or to degraded products like metallic Fe, on an hour time-scale is unlikely to fully rationalize the observed currents, since 1a/MWNT displays the highest initial Fe content but the lowest long term currents.
Electrochemical experiments of the metallopolymers hinted towards the presence of at least two populations of active sites that likely differ via the nature of their electrostatic coupling with the electrode surfaces, arguably due to their various degree of confinement within the polymer matrix or the MWNT film. All metallopolymers display a rapid decrease of current density during the first hours, likely due to degradation of a more bulk exposed CV-detectable [2Fe2S]adt active site population. Still, for 1b and 1c the current density remains stable from three hours until the end of the CA experiment. This observation along with the ICP-OES measurements suggest the existence of a variety of population of active sites with long-term stability and lesser activity within the porous MWNT electrodes.
In closing, additional efforts are clearly required to increase the stability of the catalytic units as well as their relative polymer content. In parallel, efforts need to be invested in increasing the surface loading and stability of the polymers on the electrode surface, as the observed currents are still in the low mA range. Finally, we note that the outlined synthesis strategy paves the way for functionalization with a broad range of (organometallic) active sites to support other relevant energy conversion reactions.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3se00409k |
‡ Current address: Université Paris Cité, Laboratoire d’Electrochimie Moléculaire, CNRS, F-75006, Paris, France. |
This journal is © The Royal Society of Chemistry 2023 |