Thousand-fold increase in O2 electroreduction rates with conductive MOFs.

Molecular materials must deliver high current densities to be competitive with traditional heterogeneous catalysts. Despite their high density of active sites, it has been unclear why the reported O2 reduction reaction (ORR) activity of molecularly defined conductive metal-organic frameworks (MOFs) have been very low: ca. -1 mA cm-2. Here, we use a combination of gas diffusion electrolyses and nanoelectrochemical measurements to lift multiscale O2 transport limitations and show that the intrinsic electrocatalytic ORR activity of a model 2D conductive MOF, Ni3(HITP)2, has been underestimated by at least 3 orders of magnitude. When it is supported on a gas diffusion electrode (GDE), Ni3(HITP)2 can deliver ORR activities >-150 mA cm-2 and gravimetric H2O2 electrosynthesis rates exceeding or on par with those of prior heterogeneous electrocatalysts. Enforcing the fastest accessible mass transport rates using scanning electrochemical cell microscopy revealed that Ni3(HITP)2 is capable of ORR current densities exceeding -1200 mA cm-2 and at least another 130-fold higher ORR mass activity than has been observed in GDEs. Our results directly implicate precise control over multiscale mass transport to achieve high-current-density electrocatalysis in molecular materials.

Introduction

Achieving synthetic molecular control over electrocatalytic materials is a longstanding challenge in electrocatalysis. Molecular materials need to deliver high current densities to be competitive with heterogeneous electrocatalysts, but this is rare.[1] Electrically conductive metal–organic frameworks (MOFs) offer a way to bridge this gap, as they are molecularly defined and are both intrinsically porous and conductive.[2] They are fundamentally distinct from electrocatalysts made from sacrificial MOF precursors (such as single-atom catalysts accessed via thermolysis or electrolytic degradation of MOFs), because they retain their molecular definition.[3,4,13,14,5−12] As such, the structure space available to conductive MOFs renders them an ideal platform to tune the atomic structure for performance. We and others have previously shown that a family of 2D MOFs with the general formula M3(HITP)2 (HITP = 2,3,6,7,10,11-hexaiminotriphenylene, M = Co, Cu, Ni) (Figure A) are active for the O2 electroreduction reaction (ORR), a transformation central to H2O2 electrosynthesis, metal/air batteries, and fuel cells.[4,15,16] These and other conductive MOFs typically exhibit intrinsic surface areas (∼300–900 m2 g–1) at least 10 times larger than that of dense metallic nanoparticles and conductivities comparable to that of graphite, yet their geometric current densities for ORR rarely exceed −1 mA cm–2, implying a surprisingly low intrinsic electrocatalytic activity.[2,4,5,15,17−19]

Figure 1

Controlling mass transport during ORR electrocatalysis with conductive MOFs: (A) atomic structure and connectivity of M3(HITP)2; (B) schematic of transport gradients during ORR catalysis in a conventional electrochemical H-cell using RRDEs; (C) schematic of transport dynamics in a GDE combined with flow fields; (D) mass transport of O2 across a nanodroplet in SECCM.

The performance of ORR electrocatalysts is most commonly measured using rotating ring disk electrodes (RRDEs) immersed in an electrolyte within two-compartment “H-cells”.[4,20,21] During ORR catalysis in an H-cell, a region of depleted O2 concentration (the concentration boundary, or diffusion layer) is formed adjacent to the catalyst layer (Figure B), because O2 is reduced to H2O2 or H2O at the electrode/electrolyte interface. In combination with the low saturation concentration of O2 in water (∼1 mM at 1 bar of O2 and 298 K), O2 must diffuse over distances of ca. 100 μm from the bulk electrolyte in order to reach the electrode/electrolyte interface.[22] Concentration gradients are further exacerbated in porous electrodes, where diffusion within the porous layer can be severely restricted;[23,24] the resulting transport resistance depresses the mass activity of the electrocatalyst and leads to underutilization of the active sites.

Under these circumstances, it is unclear whether the −1 mA cm–2 limit arises from mass transport limitations or in fact reflects intrinsically slow ORR kinetics with molecular materials. To probe the fundamental limitations of ORR catalysis with MOFs and potentially unlock a much higher intrinsic activity, we pursued a campaign to lift mass transport limitations by integrating Ni3(HITP)2 with gas diffusion electrodes (GDEs) and by leveraging the rapid transport environment afforded by scanning electrochemical cell microscopy (SECCM).

Here, we show that conductive MOFs enable geometric ORR current densities greater than −150 mA cm–2 if the mass transport of O2 is carefully controlled. When it is supported on a GDE (Figure C), Ni3(HITP)2 exhibits ORR activity and H2O2 electrosynthesis rates >100-fold higher and >740-fold higher, respectively, than in an H-cell. At low mass loadings in a GDE, the gravimetric rates of H2O2 electrosynthesis using Ni3(HITP)2 rival those of the highest rates reported for state of the art heterogeneous electrocatalysts. Together with efficient O2 mass transport, metal ion substitution revealed that the intrinsic porosity and conductivity of M3(HITP)2 are the major drivers of activity during ORR catalysis. By enforcing the fastest accessible mass transport rates using SECCM (Figure D), we find that Ni3(HITP)2 is capable of at least another 130-fold higher mass activity than has been observed in GDEs. Our results directly implicate precise control over mass transport to achieve high-current-density electrocatalysis in molecularly defined, conductive MOFs.

Results and Discussion

Due to its high conductivity, intrinsic porosity, and established activity for the ORR, we prepared Ni3(HITP)2, an archetypal 2D conductive MOF, to understand mass transport effects on ORR catalysis.[19] Powder X-ray diffraction (PXRD) patterns (Figure S1) and X-ray photoelectron spectra (XPS; Figure S2) were consistent with literature precedent and indicated the formation of a monophasic and highly crystalline 2D framework.[19,25] N2 adsorption measurements of Ni3(HITP)2 (Figure S3) at 77 K after activation at 373 K under dynamic vacuum revealed a high Brunauer–Emmett–Teller (BET) surface area of 802 ± 0.8 m2 g–1. SEM imaging (Figure S4) indicated that the Ni3(HITP)2 powders were composed of crystallites measuring 50–200 nm agglomerated into 1 μm wide clusters.

We drop-cast a suspension of as-synthesized Ni3(HITP)2 particles sonicated with Nafion (to act as a binder; see the Supporting Information and Figure S5) onto glassy-carbon electrodes (GCEs) for a total mass loading of 0.4 mg cm–2 of Ni3(HITP)2. We then evaluated the performance of the Ni3(HITP)2-loaded GCEs for the ORR in a two-compartment H-cell combined with a RRDE setup (Figure B). We measured cyclic voltammograms (CVs) in O2 and N2, using 1.0 M sodium chloride electrolyte buffered with 0.3 M sodium phosphate (NaPi) at pH 7. Polarization from +0.21 to −0.29 V versus the standard hydrogen electrode (SHE, to which all potentials are referenced), indicated that Ni3(HITP)2 exhibited less than −1 mA cm–2 O2 electroreduction activity. The onset of catalytic activity was observed at −0.05 V, and a plateau in the CV (around −0.6 mA cm–2) was observed beginning at −0.2 V (Figure B, inset). Polarization to more negative potentials or an increase in the rotation rate did not significantly increase the current density for O2 reduction (jORR; Figure S6A,B). We quantified the partial current density for H2O2 synthesis (jH) under potentiostatic conditions while applying an oxidizing potential of 0.91 V at the Pt ring to simultaneously detect H2O2.[15] (Figure C,D). The Faradaic efficiency (FE) for H2O2 peaked at 60% (−0.121 mA cm–2) at 0.09 V and decreased upon application of more cathodic potentials, dropping to 21% (−0.085 mA cm–2) at −0.54 V (Figure S6C). The small limiting current values (jORR = −0.4 mA cm–2) are less than expected from the Koutecky–Levich equation (jd ≈ – 2.5 mA cm–2 at 1500 rpm for the 2e– process), indicative of slow O2 mass transport not only from the bulk solution to the Ni3(HITP)2 catalyst layer but also within the immersed Ni3(HITP)2/Nafion catalyst layer.[20]

Figure 2

Polarization of M3(HITP)2 in an H-cell and using a GDE flow electrolyzer. All current densities reported are normalized to the geometric surface area. (A) Schematic of the gas diffusion flow electrolysis cell used in this study. (B) Cyclic voltammograms of 0.4 mg cm–2 Ni3(HITP)2 GDEs in 1 M NaCl, 0.3 M sodium phosphate (NaPi) electrolyte at pH 7. Inset CVs show RRDE data at the same 0.4 mg cm–2 mass loading, but where the current scale is much smaller. Scan rate: 50 mV s–1. The RRDE was rotated at 1000 rpm. Potentials vs SHE reported in (B) are not corrected for the system iR drop. (C) Geometric current densities for O2 reduction during potentiostatic polarization for both the RRDE and GDE electrolyses. The RRDE was rotated at 1500 rpm. (D) Partial current densities for H2O2 synthesis during potentiostatic polarization for both the RRDE and GDE electrolyses. The RRDE was rotated at 1500 rpm. (E) SEM images of M3(HITP)2. Scale bar: 200 nm. (F) BET surface area derived from N2 adsorption measurements and electrochemical roughness factors (RF) represented as multiples of geometric surface area, calculated from CVs obtained in N2. (G) CVs of the three isoreticular M3(HITP)2 GDEs in O2. Scan rate: 50 mV s–1.

To investigate whether the ORR activity of Ni3(HITP)2 would improve under a rapid bulk mass transport regime, we drop-casted Ni3(HITP)2 particles onto GDEs, at the same mass loading of 0.4 mg cm–2 (SEM imaging in Figure S7 and S8). Composed of a carbon fiber support and a hydrophobic microporous conductive coating, GDEs enhance gas mass transport by providing a gas flow pathway unimpeded by electrolyte through the back of the electrode (Figure C). Ni3(HITP)2 GDEs were interfaced into a custom-built gas diffusion flow electrolyzer (Figure A, detailed in the Supporting Information)[26−29] with a microfluidic pocket that limits the contact area (∼0.8 cm2) between the MOF-loaded GDE and electrolyte (Figure A). The Ni3(HITP)2 GDE was compressed against a conductive, interdigitated flow field that rapidly transports O2 to the Ni3(HITP)2/electrolyte interface. Electrolyte flowed through the cell and into a collection vial, enabling quantification of electrogenerated H2O2. Enhancing O2 transport to the Ni3(HITP)2/electrolyte interface led to orders of magnitude higher ORR current densities using Ni3(HITP)2 GDEs. Across the same potential range as was used with RRDEs, CVs of a Ni3(HITP)2 GDE indicated that the geometric jORR was ca. 1–2 orders of magnitude larger with the GDE relative to the RRDE (Figure B). At –0.29 V, whereas Ni3(HITP)2 exhibited a jORR value of only −0.6 mA cm–2 on the RRDE, its activity on the GDE was −62 mA cm–2. Control experiments of both the GDE support under an O2 atmosphere (Figure S9) and the Ni3(HITP)2 GDE under an N2 atmosphere (Figure B) indicated that essentially all of the observed current could be attributed to ORR catalysis (i.e., no H2 evolution) occurring at the Ni3(HITP)2 sites. These data indicate that Ni3(HITP)2 was starved of O2 during polarization in the H-cell, which led to a vast underestimation of its intrinsic electrocatalytic performance.

We evaluated the ORR performance of the Ni3(HITP)2 GDEs by measuring the current during step-potential polarization (Figure C). Unlike the case in the H-cell, the geometric jORR value using the GDE increased monotonically with the applied potential in a broader range, reaching a maximum current density of −103 mA cm–2 at −0.36 V, an approximately 310-fold improvement in jORR relative to those measured in the RRDE/H-cell. Similarly, the jH value using the Ni3(HITP)2 GDE increased as a function of applied potential from −1.2 mA cm–2 at 0.01 V to a maximum current density of −88.5 mA cm–2 (or 85% FE for H2O2) at −0.36 V, a 740-fold improvement over the maximum jH value measured using the RRDE/H-cell at the same mass loading (Figure D). With 0.2 mL min–1 of electrolyte flowing through the cell, we measured a 108 mM (∼3270 ppm) H2O2 product stream at −0.36 V. These high ORR current densities corresponded to a mass activity of 259 A g–1 at −0.36 V, which is competitive with state of the art, H2O2-producing heterogeneous electrocatalysts in a neutral electrolyte.[30] PXRD patterns and XPS spectra obtained immediately after polarization indicated that Ni3(HITP)2 retained its crystallinity, and we found no evidence for the formation of metallic Ni from reduction of framework Ni2+ (Figures S10 and S11). As H2O2 concentrations were likely even higher within the Ni3(HITP)2 pores, these data indicate that Ni3(HITP)2 is stable to high local concentrations of electrogenerated H2O2.

Supporting conductive MOFs on GDEs is a general strategy that allowed us to probe the effect of metal ion substitution in M3(HITP)2 under a high O2 flux. We integrated two additional, isostructural HITP-based frameworks, namely Cu3(HITP)2 and Co3(HITP)2 with GDEs (Figure E–G; characterization in Text 1 in the Supporting Information and Figures S12–S17), to evaluate their ORR activity under high-mass-transport conditions.[25] Polarization in O2 revealed that the total jORR and jH values both depend on the identity of the MOF and vary in the order Ni > Co > Cu (Figure G). This reflected the trend in electrochemical surface area (ECSA, a composite value of intrinsic surface area and conductivity) among the three MOFs (Ni > Co > Cu; Figure F and Figure S18). Ni3(HITP)2 exhibits more than 6-fold higher ECSAs in comparison to the Cu or Co analogues, characteristic of its high conductivity and porosity. On a mass activity basis, Ni3(HITP)2 exhibits the highest activity at the lowest driving forces (Figure S17A). These data provided a simple model to rationalize the observed jORR value: Ni3(HITP)2 is the most active of the three M3(HITP)2 because it possesses an intrinsically higher ECSA and therefore a higher density of active sites. Given that crystallinity generally engenders high conductivity and surface area in conductive MOFs, and noting that as-synthesized Ni3(HITP)2 is intrinsically more crystalline than either Co3(HITP)2 or Cu3(HITP)2 (cf. Figures S1 and S12), these results suggest that high conductivity, porosity, and crystallinity are the keys to unlocking high rates of ORR catalysis in MOFs.[2]

Motivated by the apparent dependence of jORR on M3(HITP)2 ECSA in the GDE and because maximizing jORR is technologically desirable, we sought to understand how much of the ECSA in Ni3(HITP)2 could be productively recruited for catalysis. To this end, we varied the Ni3(HITP)2 mass loading from 0.1 to 0.8 mg cm–2 (Figure ; chronoamperograms are given in Figure S20). The geometric jORR value generally increased with higher catalyst loading (Figure A). For instance, at −0.36 V, the geometric jORR value at 0.4 mg cm–2 was ∼2-fold higher than that at 0.1 mg cm–2. CVs in N2 showed that the ECSA also increased with increased mass loading (Figure S21 and Table S1). These mass-dependent increases in ECSA correlate with the increase in jORR and indicate a larger number of active sites available for ORR catalysis. Additionally, higher values of jH are correlated with higher mass loadings of up to 0.4 mg cm–2 (Figure B). At –0.36 V, jH increased ∼3.3-fold as the mass loading increased from 0.1 to 0.4 mg cm–2. We did not observe a systematic correlation between mass loading and FE (Figure S22A), suggesting that 2e– reduction of H2O2 to H2O, or framework-catalyzed decomposition of H2O2 to O2, does not accelerate with higher mass loadings under the conditions employed here.

Figure 3

ORR activity limits in Ni3(HITP)2 GDEs. (A) Geometric current densities for the ORR during potentiostatic polarization with different catalyst mass loadings: (◆) 0.1 mg cm–2; (▲) 0.2 mg cm–2; (●) 0.4 mg cm–2; (■) 0.8 mg cm–2. (B) Partial current densities for H2O2 synthesis during potentiostatic polarization. (C) Comparison of total ORR mass activities of the four different mass loadings.

At a high overpotential, the mass activity decreases as the catalyst loading increases (Figure C and Figure S22B), with the lowest mass loading of 0.1 mg cm–2 being responsible for the highest mass activity of 553 A g–1 at the most cathodic applied potential. At the highest applied potentials, the rate of H2O2 production using the 0.1 mg cm–2 electrode is equivalent to a gravimetric rate of 6570 mol H2O2 kgMOF–1 h–1, better than or competitive with the highest activities reported among state of the art H2O2-evolving electrocatalysts.[30−32]

Although low Ni3(HITP)2 mass loadings yield the highest mass activities (Figure C), they also exhibit the lowest geometric jORR and jH values (Figure A,B). This an important dilemma to address, because it implies that much of the ECSA in Ni3(HITP)2 GDEs remained underutilized at high mass loading. Indeed, an apparent plateau in the geometric jORR value (ca. −110 mA cm–2) is observed for GDEs with 0.4 and 0.8 mg cm–2 of Ni3(HITP)2 at a high driving force. The contrast between the 0.8 and 0.4 mg cm–2 electrodes is small: doubling the mass loading provides only marginal improvements at low overpotentials and essentially identical activity at high overpotentials. Moreover, the jH value for the 0.8 mg cm–2 electrode is only ∼80% of that for the 0.4 mg cm–2 electrode (Figure B). If the mass activity of the 0.8 mg cm–2 electrode were identical with that of the 0.1 mg cm–2 electrode (481 A g–1 at −0.36 V), the measured jORR value for the 0.8 mg cm–2 electrode should be >380 mA cm–2, almost 4× larger than what we observe. In fact, mass activities across the four different mass loadings were relatively uniform at low overpotentials and diverged prominently only at high overpotentials (Figure 22B). This activity plateau limited the single-pass O2 conversion rate to just 30% (Figure S22C).

High-resolution scanning electrochemical cell microscopy (SECCM; Figure A)[33−35] lent critical insight into the origin of the plateau in activity observed in our GDE studies (Figure S23 and the Supporting Information for experimental details). By confinement of the entirety of the electrode contact area to the footprint of a droplet at the end of a nanopipet, SECCM offers the fastest gas mass transport rates experimentally accessible for electrocatalysis: N2 or O2 can rapidly traverse the nanoscale droplet electrolyte and the porous Ni3(HITP)2 particles, with the maximum diffusion length of gaseous species to the catalyst surface being set by the droplet radius.[28,36] For example, a hemispherical droplet with a radius rd = 25 nm has a sub-microsecond diffusion time, about 6 orders of magnitude higher than that in the RRDE studies. By confining electrocatalytic studies to a nanoscale droplet, SECCM provides a unique platform to measure the intrinsic electrochemical mass activity in the absence of extrinsic transport limitations.

Figure 4

SECCM mapping of ORR activity on Ni3(HITP)2. (A) Schematic of the experimental geometry in SECCM using a single-barrel nanopipet. (B) Optical image of the Ni3(HITP)2 particles scanned using SECCM under air. (C) Single-pixel LSVs of Ni3(HITP)2 obtained under N2 and air. Inset: average current densities at −0.36 V for Ni3(HITP)2 under N2 and air. SECCM scans corresponding to those in N2 are shown in Figure S24. (D) SECCM map of geometric current densities (defined as the current divided by the tip droplet area) measured on Ni3(HITP)2 at −0.36 V. (E) Corresponding topographic map of Ni3(HITP)2 derived from SECCM mapping.

Using a 50 nm diameter nanopipet filled with a solution of 30 mM NaPi (pH 7) and 100 mM NaCl, we directly mapped the electrochemical activity of Ni3(HITP)2 particles under cathodic polarization under both N2 and air (PO = 0.2 bar) (Figure S24 and Figure , respectively).[37] The average current under N2 at −0.36 V was negligible at ca. –1 pA, close to the noise limit of the conditions we employ here (Figure S24, Movie S1, histograms in Figure S29). Comparison of individual linear sweep voltammograms (LSVs) in N2 versus those in air indicated that essentially all of the measured current in air arose from the ORR (Figure C). Importantly, control experiments probing the nanodroplet while it was in contact with Ni3(HITP)2 revealed that the droplet was stable while it was in contact with the MOF. Furthermore, the droplet contact areas were similar on both Ni3(HITP)2 and ITO, allowing us to estimate the maximum ECSA of Ni3(HITP)2 contacted during SECCM measurements. (Text 2 in the Supporting Information and Figures S26–S29).

In contrast, SECCM scans of Ni3(HITP)2 under air indicated high ORR activity across the MOF particle surface, with an average current of −25 pA at −0.36 V vs SHE (Figure B,C, Movie S2, and histograms in Figure S29B). Given a droplet radius of 25 nm, this current translates to a geometric jORR value of −1273 mA cm–2, 38-fold greater than the highest current densities observed under air using the GDE (Figure S25). The larger current densities observed in SECCM versus the GDE or RRDE configurations are consistent with the short diffusion pathways enforced by the nanoscopic dimensions of SECCM.[28,35,38−51]

Using the SECCM and topography maps obtained under air, we calculated a lower-bound estimate of the intrinsic mass activity of Ni3(HITP)2, making a conservative assumption that all of the cylindrical mass of Ni3(HITP)2 under the droplet’s 50 nm footprint is recruited for catalysis (Figure S30). Strikingly, these estimates yielded a lower-bound average mass activity of 11250 A g–1 at −0.36 V vs SHE, 136-fold higher than the highest mass activities measured with GDEs in air.

The rapid transport environment of SECCM revealed that Ni3(HITP)2 is even more active intrinsically than has been observed in the GDE, suggesting that even at low mass loadings some O2 mass transport resistance persists in the agglomerated Ni3(HITP)2 GDE catalyst layer. At high Ni3(HITP)2 GDE mass loadings, much of the active material is immersed in a thick aqueous electrolyte layer through which O2 mass transport is sluggish (Figure S31), with maximum diffusion lengths likely exceeding the thickness of the flooded GDE pores and Ni3(HITP)2 catalyst layer (>10 μm; see Figure S8).

Crucially, the high mass activities observed in SECCM mean that the activity plateau observed in the GDE was extrinsic to Ni3(HITP)2. Identifying slow micrometer-scale O2 mass transport, rather than low intrinsic catalyst activity, as the origin of current density limitations observed in the GDE provides an impetus to improve the mass transport properties of the Ni3(HITP)2 GDE. As a simple proof of principle, we reformulated the catalyst ink to include 10 wt % of hydrophobic polytetrafluoroethylene (PTFE) powder that was intimately mixed with the Ni3(HITP)2 nanoparticles, reducing Ni3(HITP)2 particle agglomeration and providing continuous, hydrophobic channels through which O2 could diffuse rapidly (Figures S31–S34 and Text 3 in the Supporting Information).[26,29]

The facility of O2 transport across the hydrophobic PTFE domains led to substantial increases in ORR activity: a 0.8 mg cm–2 Ni3(HITP)2 GDE with 10 wt % added PTFE passed a total current density of −170 mA cm–2 at just −0.27 V (Figure S34A). By comparison, the previous best-performing PTFE-free 0.4 mg cm–2 Ni3(HITP)2 electrode reached a peak jORR value of −103 mA cm–2 while also requiring a 90 mV higher driving force of −0.36 V. This activity translated to a 60% higher total single-pass O2 conversion rate of 48% at an O2 flow rate of 1.5 mL min–1 (Figure S34F). With the addition of 10 wt % PTFE, the 0.8 mg cm–2 electrode exhibits 70% greater mass activity than its PTFE-free analogue: 212 vs ∼124 A g–1 (Figure S34C). Incorporating PTFE into the catalyst ink increased the apparent mass-transport-limited current from −110 to −180 mA cm–2, an increase of ∼60%. These large differences in activity demonstrate that enhancing O2 mass transport in the Ni3(HITP)2 catalyst layer allows more of the Ni3(HITP)2’s ECSA to be recruited for productive catalysis. However, the mass activity of the 10% PTFE electrode with 0.8 mg cm–2 Ni3(HITP)2 remains more than a factor of 2 below the mass activity of the PTFE-free electrode with only 0.1 mg cm–2 Ni3(HITP)2 and >50-fold below those recorded in SECCM. Further improvements in the geometric jORR value in practical devices are gated not by the intrinsic activity of Ni3(HITP)2 but by the facility of O2 mass transport. Identifying the multiscale transport bottlenecks that prevent MOFs from delivering high current densities motivates the broader exploration and deployment of these designer materials for a variety of electrocatalytic processes.