Comparison of In Situ and Postsynthetic Formation of MOF-Carbon Composites as Electrocatalysts for the Alkaline Oxygen Evolution Reaction (OER)

1. Introduction

Nickel-based materials have emerged as promising electrocatalysts for the oxygen evolution reaction (OER), one of the two half reactions in electrochemical water splitting (OER, alkaline conditions: 4 OH^- → O₂ + 4 e^- + 2 H₂O, E° = 1.23 V vs. the reversible hydrogen electrode, RHE) [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23]. The OER is the bottleneck of electrochemical water splitting and requires a higher activation energy than the hydrogen evolution reaction (HER, alkaline conditions: 4 H₂O + 4 e^- → 2 H₂ + 4 OH^-, E° = 0.00 V vs. RHE) [2]. To produce one oxygen molecule in the OER four electron-proton coupled reaction steps, which require a larger applied potential than the theoretical potential of 1.23 V for electrochemical water splitting, have to be gone through [3,4]. The difference of the applied potential (E) to the theoretical potential of E° = 1.23 V, which still needs to be corrected for the internal resistance (IR) because of the ohmic drop, is called overpotential η (η = E – E° (1.23 V) – IR (1)) [5]. The overpotential is a parameter to check the efficiency of an electrocatalyst and is determined at a defined current density, with often 10 mA/cm² being used as a reference point [2,6].

Currently noble-metal-based catalysts like ruthenium oxide (RuO₂) and iridium oxide (IrO₂) are seen as benchmark electrocatalysts for the OER [7,8]. RuO₂ is known to efficiently drive the OER in alkaline and acidic media [9,10,11]. However, the high costs due to the scarcity of the noble-metals, impede a large-scale industrial utilization of those electrocatalysts [12]. Nickel presents an attractive option for the OER due to its abundance, its lower cost and electrocatalytic activity [3,12,13]. Ni³⁺ is assumed to be the main active site in nickel-based materials, leading to an improved conductivity and accelerated reaction kinetics [8,12,14,15]. The incorporation of iron into nickel-based materials can be advantageous for electrocatalytic water splitting, because iron promotes the nickel oxidation from Ni²⁺ to Ni³⁺ [15,16,17]. In recent years, various nickel(-iron)-based materials, including oxides, (oxy)hydroxides, phosphides, sulfides, nitrides and metal-organic frameworks (MOFs), have been extensively studied for their electrocatalytic ability in the OER [1,3,18]. MOFs usually act as pre-catalyst that only form the active phase under the electrochemical reaction conditions [19]. The resulting electrocatalysts, often metal (oxy)hydroxides, can potentially retain some of the porosity of the MOFs, which help in lessening diffusion limitations and increasing the electrocatalytic activity of the material [12,1][–[1]. For mixed-metal MOFs the intimate and uniform mixture of the involved metals is retained at the nanoscale in the derived electrocatalyst.

MOFs and also the derived metal (oxy)hydroxide electrocatalysts have low electrical conductivity, which needs to be increased through the combination with a carbon material [22,23,24]. The admixture of a carbon material, that is, the formation of a MOF/carbon composite improves the efficiency of charge transfer processes in electrochemical applications [25]. Ketjenblack (KB) or carbon nanotubes (CNTs) are often utilized to produce composite materials. KB is a carbon black material and displays good electrical conductivity, a high surface area and thermal stability [22,26,28]. CNTs are cylindrical nanostructures composed of carbon atoms arranged in a hexagonal lattice, forming seamless tubes with diameters typically on the order of nanometers. CNTs have good mechanical stability, high thermal stability, high electrical and thermal conductivities, which make them interesting to use as an alternative to carbon black materials in composites [25,29,30,31]. Preceding research presented different highly active MOF-composite-derived materials for the alkaline OER [32,33,34,35].

The MOF and the carbon material can be combined in situ, that is, by carrying out the MOF synthesis in the presence of the carbon material or postsynthetically, that is, by combining the separately prepared MOF with the carbon material through physical mixing [36]. Both routes are used in the literature but mostly without direct comparison. Studies on in situ prepared Ni-MOF/carbon composites [37] are, of course, difficult to compare with studies on postsynthetically prepared physical mixtures [38]. Few Ni-MOF/carbon works on OER include a comparison of both routes [39].

Here we present the direct comparison of Ni(Fe)-MOF/carbon composites which were prepared in situ and postsynthetically with the same weight percent of carbon with respect to their electrocatalytic activity as electrode material in the oxygen evolution reaction. The parent MOF was Ni-BTC of formula [Ni₃(BTC)₂(Me₂NH)₃]·(DMF)₄(H₂O)₄ (BTC = benzene-1,3,5-tricarboxylate, DMF = N,N-dimethylformamide) [40].

2. Results and Discussion

2.1. Synthesis and Characterization of the MOF Composites

For the in situ synthesis the nickel and nickel-iron MOF composites were synthesized by a simple one-pot solvothermal reaction at 170 °C for 48 h from Ni(NO₃)₂·6 H₂O or a mixture of Ni(NO₃)₂·6 H₂O, Fe(NO₃)₃·9 H₂O (molar Ni:Fe ratio 10:1 or 5:1), 1,3,5-benzenetricarboxylic acid (H₃BTC), 2-methylimidazole (2-MeImH), and ketjenblack (KB) or carbon nanotubes (CNT) in N,N-dimethylformamide (DMF) (Figure 1a). The amount of carbon materials was set to give about 10% of KB or CNT in the composite.

For the postsynthetic admixture the neat Ni-, Ni₁₀Fe and Ni₅Fe-MOFs were generated through the same procedure as noted above [40] but without the presence of a carbon material. The neat MOFs were physically mixed with 10 wt% of KB or CNT with respect to the MOF amount with a vortex mixer at 40 Hz for 20 min (Figure 1b).

A hyphen is used for the in situ MOF-composites, i.e., Ni_XFe-CNT or Ni_XFe-KB depending on the used carbon material. A slash indicates the postsynthetically physically-mixed MOFs/carbon materials, i.e., Ni_XFe/CNT or Ni_XFe/KB.

To assess also the influence of iron on the electrocatalytic performance the Ni-only MOF and its carbon composites were prepared and investigated in comparison to the Ni_XFe materials.

The verification of the MOF synthesis had to be carried out for the neat MOFs and for the in situ prepared MOF-carbon composites. For the latter it had to be seen if and to what extent the presence of carbon would influence the MOF crystallinity and the porosity. The crystallinity was assessed by powder X-ray diffraction (PXRD), the surface area determination and porosity were calculated by BET and NLDFT on the volumetric gas sorption isotherms. For the postsynthetically admixed MOF/carbon composites PXRD and gas sorption measurements confirmed the superposition of the MOF and carbon contribution.

It was verified by energy-dispersive X-ray spectroscopy in the scanning electron microscope (SEM-EDX) of the reaction products that the used molar 10:1 and 5:1 nickel to iron ratios in the in situ synthetic procedure were also present in the nickel-iron containing MOFs and their composites (Table S1, SI). SEM-EDX element mapping displays a homogeneous distribution of Ni and Fe inside the samples (Figure 2 and Figure S2 to S4, SI). The morphology of the MOF particles ranges from an octahedral to a cuboctahedral shape. The change in shape was also seen by Qiao et al. and correlated with the iron content in their samples [41]. Noteworthy, the average particle size for the in situ MOF-CNT composites (Ø(Ni-CNT) = 16 µm, Ø(Ni₁₀Fe-CNT) = 6 µm, Ø(Ni₅Fe-CNT) = 6 µm) is smaller than for the neat MOF (Ø(Ni) = 27 µm, Ø(Ni₁₀Fe) = 15 µm, Ø(Ni₅Fe) = 7 µm) and the in situ MOF-KB composites (Ø(Ni-KB) = 27 µm, Ø(Ni₁₀Fe-KB) = 14 µm. Ø(Ni₅Fe-KB) = 15 µm) (Figure 2, S2 to S9).

Powder X-ray diffraction (PXRD) patterns of the neat MOFs, the in situ and postsynthetic formed composites present the same reflexes as the simulated pattern of the structurally authenticated Ni-MOF (Ni-BTC) in Figure 4, S1 and S10 (see Figure S1, SI for the building unit and packing diagram of the parent MOF Ni-BTC). Thereby, the in situ incorporation of KB had less influence on the MOF crystallinity than did CNT, that is the width of the reflections and signal-to-noise ratio of the KB samples was similar to those in the neat MOF diffractograms For the in situ CNT samples the reflexes were broader and the noise increased. The lower crystallinity of the in situ CNT samples correlates with the seemingly smaller crystallites which formed in the presence of CNT (Figure 2, Figure S2 to S9, SI). In addition, there could be more defects in the MOF-CNT composite samples [42,43,44,45]. Mazlan et al. reviewed recent works on the growth of MOFs in the presence of graphene oxide and also showed that depending on the MOF, on the functional groups of the graphene oxide (GO) and on the synthesis method of the composite, the crystal growth of the MOFs could be influenced to result in ordered or disordered (nanosized) MOF-GO structures [45]. No additional other reflexes than for Ni-BTC were detected in the Ni_xFe-MOF samples, which indicates that iron was well incorporated into the structure of the Ni-MOF. KB presented three broad diffraction peaks at 8, 26 and 44° matching to the (100), (002), (101) planes of graphitized carbon, respectively [22,46]. CNT displayed its two broad diffraction peaks at 26° (002) and 44° (101) [47].

Thermogravimetric analysis (TGA) of the neat MOFs and their in situ composites was performed on the as-synthesized samples without activation and under N₂ atmosphere (Figure S11a-c, SI). Two major mass loss steps were recorded, where the first one 350 °C can be attributed to the evaporation of coordinated and crystal solvent molecules from the porous Ni-BTC structure (~20–30% mass loss) [40,48,49,50,51]. The second mass loss step after 350 °C can be ascribed to the decomposition of the BTC-linker and the disintegration of the MOF structure (~45–55% mass loss) [40,48,49,50,51]. The carbon materials KB and CNT show a good stability until 600 °C under the measurement conditions (Figure S11d, SI). In general, the residual masses of the MOFs without any carbon material are lower than their corresponding composites, which can be used to calculate the amount of carbon material inside of the in situ composites via subtraction of the residual mass of the neat MOFs from the residual mass of their composites Table S2, SI):

Fourier transform infrared (FT-IR) spectra (Figure S12, SI) of the in situ composite materials display the characteristic bands of the neat Ni-BTC MOF (Table S3, SI). All spectra show a broad band in the region of 3600–3000 cm^–1, which can be ascribed to the stretching vibration of the hydroxy group from adsorbed or coordinated water (Figure S1, SI) [40,52,53]. The samples were not activated before the FT-IR measurement. The bands around 1629–1559 cm^–1 and 1440–1345 cm^–1 can be assigned to the asymmetric and symmetric stretching vibrations of carboxylate groups [24,49,52,54]. Characteristic metal-oxygen bond vibrations can be seen for a mixed-metal-oxygen (FeNi-O, Fe₂Ni-O) bond around 722–713 cm^–1 (Lit. ~720 cm^–1 [55,56]) and a Fe/Ni-O bond around 589–453 cm^–1 (Lit. 587–461 cm^–1 [24,35,55,57]) in all spectra.

To examine the porosity of the materials N₂-sorption measurements were conducted. The N₂-adsorption isotherms at 77 K were used to calculate the specific Brunauer-Emmett-Teller (BET) surface areas (S(BET)) and pore volumes of the materials (Figure 5 and S13 to S15, ) and are summarized in Table S4 and S5, SI.

The isotherm shape of the neat Ni-MOF and Ni₁₀Fe-MOFs are close to type I with a slight type II increase with pressure and a slight H4 hysteresis. The distinct uptake at low relative pressure originates from the filling of the micropores. The type II branch and the hysteresis loops derive here from the texture effect of the physisorption in the meso- and macroporous voids of the aggregated crystallites [58,59]. Ni₅Fe demonstrates a steeper uptake approaching a relative pressure of 1 with a more pronounced H3 hysteresis loop than the other neat MOF samples, which could be due to a stronger capillary condensation in this sample and is an indication that the sample exhibits more aggregates and larger pores from defects, which are not completely filled with pore condensate [59].

The isotherms of the neat carbon materials are displayed in Figure 5d. KB is a mesoporous carbon material (S(BET) = 1415 m²/g, pore volume: 1.59 cm³/g, pore size: ~ 4 ± 2 nm, Figure S13d, SI). KB shows a mixed adsorption isotherm consisting of an IUPAC type I and II branch at low and high relative pressure, respectively, with an H3 hysteresis at p/p₀ >0.9 and H4 hysteresis between 0.4 < p/p₀ < 0.9, [59]. CNT is a carbon material, where the inner surface area of the tubes is not accessible and only their outer surface area was measured (S(BET) = 117 m²/g, Figure S6d, SI) [60]. CNT gives a type III isotherm with a slight H3 hysteresis, characteristic of nonporous or macroporous materials [59].

The MOF-KB isotherms are evidently a superposition of the neat MOF and KB isotherm and have the same KB-characteristic type II branch at high relative pressure with the H3 and H4 hysteresis upon desorption. There are also bimodal pore size distributions with maxima at ~ 2 nm and ~ 4 nm (Figure S13 and S15, SI). The BET surface areas for the MOF samples and their in situ composites fit into the range of published values for Ni-BTC and their analogs (Lit. 86-596 m²/g [35,57,61], Ni-MOF: 648 m²/g, Ni-KB: 414 m²/g, Ni-CNT: 301 m²/g, Ni₁₀Fe: 859 m²/g, Ni₁₀Fe-KB: 526 m²/g, Ni₁₀Fe-CNT: 387 m²/g, Ni₅Fe: 698 m²/g, Ni₅Fe-KB: 599 m²/g, Ni₅Fe-CNT: 67 m²/g) [28]. The sorption isotherms of the postsynthetic composites show a similar curvature as the in situ synthesized composites (Figure S14, Table S5, SI).

2.2. Electrocatalytical Results

Multiple electrochemical measurements were performed in order to assess the electrocatalytic ability of the neat MOF, in situ and postsynthetically-mixed carbon composite samples in the alkaline (1 mol/L KOH electrolyte, pH = 14) towards the oxygen evolution reaction (OER). A three-electrode setup with an ink-coated carbon paper as the working electrode was utilized (see experimental section for details). MOFs are usually of low stability in alkaline medium where they form metal hydroxides. Metal hydroxides are recognized as electrocatalytically active species. Mixed-metal MOFs allow to achieve a defined mixed-metal hydroxide composition with a uniform distribution of the metals on the nano-level [62,63,64,65]. The electrochemical parameters overpotential (η) (Figure 7 and S23, SI) and Tafel slope (b) (Figure 9 and S24a, SI) were derived from activated linear sweep voltammetry (LSV) measurements (Figure 6). A comparison of the overpotentials before and after 1000 cyclic voltammetry cycles (CVs) and a 12 h chronopotentiometry (Figure S25, SI) was carried out to test the stability of the electrocatalysts. The electrochemical parameter charge transfer resistance (R_ct), was achieved through electrochemical impedance spectroscopy (EIS) (Figure 10 and Figure S24b, SI). Measurement details are given in the experimental section.

The electrocatalytic performance of the electrocatalysts derived from the neat MOFs, from the in situ synthesized MOF-carbon composites and from the postsynthetic mixtures of Ni-MOF, Ni₁₀Fe and Ni₅Fe were contrasted to the performance of the RuO₂ benchmark.

The LSV plots in Figure 6 demonstrate that Ni₅Fe-CNT achieves the highest current density (j) of all samples before and after the stability test. All samples except for Ni-MOF reached higher current densities than the commercial RuO₂ and the neat carbon materials KB and CNT. With regards to the current density a higher iron content in the composites allows to attain higher current densities at a given potential. Regarding the type of carbon material in case of the postsynthetically mixed Ni₁₀Fe composites KB managed to gain a slightly higher current density than CNT, but in case of the Ni₅Fe-BTC composites the CNT composite reaches a higher current density (Figure S22). For most samples the current densities before and after the stability test were similar or only slightly lower after 1000 CV cycles except for the Ni-MOF composites, where higher current densities were attained after the 1000 CV cycles. The oxidation of Ni²⁺ to Ni³⁺ is visible by the peak around and above 1.4 V vs. RHE (Figure 6), and shifts to higher potentials when iron included into the samples which indicates a stabilization of the Ni²⁺ state [55,66].

From the potential at the current density of 10 mA/cm² (E₁₀), which is often chosen for comparison, the overpotential (η₁₀) is derived by subtracting the standard potential of 1.23 V, i.e., η₁₀ = E₁₀ – 1.23 V. The overpotentials before and after the 1000 CV cycle stability test are summarized in the bar diagrams in Figure 7 (values are listed in Table S6, SI). The initial overpotential of Ni-MOF is 372 mV and after the stability test it has slightly increased to 379 mV. With the addition of a carbon material, the overpotential for Ni-MOF decreases, and unexpectedly is lowered further after the CV cycles, both in situ and the more so when postsynthetically mixed [35,67].

In general, the addition of Fe into the Ni-MOF at a 5:1 molar Ni:Fe ratio led to a significant and anticipated [66,68] decrease of the overpotential, down to below 300 mV before the activation cycles (Ni₅Fe: η₁₀ = 296 mV, Ni₅Fe/KB: 297 mV, Ni₅Fe-CNT η₁₀ = 289 mV, Ni₅Fe/CNT: 294 mV) (Figure 7 and Table S6, SI). Activation with 1000 CV cycles then gives the normally seen increase in overpotential to values between 300 and 340 mV. Still, all Ni_xFe-MOF-carbon samples outperformed the benchmark RuO₂ (η₁₀ = 360 mV, before to 354 mV, after activation). The overall lowest overpotential, i.e., highest activity for OER is given by Ni₅Fe-CNT with η₁₀ = 289 mV, before to 301 mV, after activation. It can be clearly seen that an increase in Fe content from the Ni₁₀Fe to the Ni₅Fe series is advantageous with the latter giving lower overpotentials.

Judged from the lower overpotential, the in situ synthesized CNT composites perform better than the in situ KB composites (Figure 7a, Table S6, SI). Comparing the in situ and the respective postsynthetically-mixed composites, for KB the postsynthetic ones and for CNT the in situ ones have the lower overpotential (Figure 8).

Figure 7. Overpotentials η₁₀ calculated from LSV curves in Figure 6 (a-d) of (a) MOFs and in situ MOF-carbon samples, (b) MOFs and postsynthetic MOF/carbon mixtures; each with KB, CNT and RuO₂ for comparison. No η₁₀ value of KB could be determined after 1000 CVs, because KB did not reach 10 mA/cm² anymore. Error bars are derived from the highest and lowest achieved η₁₀ value in multiple measurements.

Figure 7. Overpotentials η₁₀ calculated from LSV curves in Figure 6 (a-d) of (a) MOFs and in situ MOF-carbon samples, (b) MOFs and postsynthetic MOF/carbon mixtures; each with KB, CNT and RuO₂ for comparison. No η₁₀ value of KB could be determined after 1000 CVs, because KB did not reach 10 mA/cm² anymore. Error bars are derived from the highest and lowest achieved η₁₀ value in multiple measurements.

Figure 8. Overpotentials η₁₀ calculated from LSV curves in Figure 6 of in situ MOF-carbon samples and postsynthetic MOF/carbon mixtures for comparison. For clarity, only the data after 1000 CVs is presented (see Figure S23, SI for the additional data before the 1000 CVs). Error bars are derived from the highest and lowest achieved η₁₀ value in the multiple measurements.

Figure 8. Overpotentials η₁₀ calculated from LSV curves in Figure 6 of in situ MOF-carbon samples and postsynthetic MOF/carbon mixtures for comparison. For clarity, only the data after 1000 CVs is presented (see Figure S23, SI for the additional data before the 1000 CVs). Error bars are derived from the highest and lowest achieved η₁₀ value in the multiple measurements.

Figure 9. Tafel plots with Tafel slope b (given in mV/dec) of (a) MOFs, in situ MOF-carbon samples and RuO₂ and (b) postsynthetically mixed MOF/carbon composites.

Figure 9. Tafel plots with Tafel slope b (given in mV/dec) of (a) MOFs, in situ MOF-carbon samples and RuO₂ and (b) postsynthetically mixed MOF/carbon composites.

The correlation of the overpotential from the LSV curves in Figure 6 with the decadic logarithm of the current density is the Tafel plot (Figure 9 and S24a) in which the data points can be fitted by the linear Tafel equation (η = a + b × log(j) (2)) with slope b (tabulated in Table S6, SI) [69]. A smaller Tafel slope b represents a steeper increase in current density together with a smaller increase in overpotential. Both suggest a faster reaction rate and a better electrocatalytic performance [70]. The Tafel plots of RuO₂ and the MOF samples are presented in Figure 9. The Tafel plot of KB and CNT is displayed in Figure S23a, SI. The smallest Tafel slope is seen for Ni₅Fe-CNT with 58 mV/dec but for all MOFs and their carbon composites the Tafel slopes lie in a narrow range of 64 to 76 mV/dec, except for neat Ni-MOF (100 mV/dec) and Ni-CNT (87 mV/dec). Thus, the Tafel slopes of the Ni_xFe-MOF carbon composites, both in-situ and postsynthetically mixed, are much smaller than for RuO₂ (b = 91 mV/dec [58]) or the neat carbon materials (KB b = 134 mV/dec, CNT b = 137 mV/dec). The Tafel slopes from our study agree with published values of comparable materials such as Ni(Fe)(OH)₂/KB b = 66 mV/dec [28], FeNiNH₂BDC MOF@2-6 wt% CNT b = 69-84 mV/dec [55] and Ni-BTC b = 114 mV/dec [71].

The Tafel slope provides also information on the rate determining step (rds) according to eqn. (3)–(6) (M = active site) when assuming Krasil’shchikov’s OER mechanism [69,72]. The Tafel slope of 58 mV/dec for Ni₅Fe-CNT and the range of Tafel slopes of 64 to 76 mV/dec for the other Ni_xFe-MOF composites suggests that the metal hydroxide deprotonation reaction $\mathrm{MOH}+{\mathrm{OH}}^{\text{–}}\text{⇄}{\mathrm{MO}}^{–}+{\mathrm{H}}_2\mathrm{O}$ (eqn. 4) with b = 60 mV/dec is the rate limiting step.

$$\mathrm{M}+{\mathrm{OH}}^{\text{–}}\text{⇄}\mathrm{MOH}+{\mathrm{e}}^{\text{–}}\mathrm{b}=120\mathrm{mV}/\mathrm{dec}$$

(3)

$$\mathrm{MOH}+{\mathrm{OH}}^{\text{–}}\text{⇄}{\mathrm{MO}}^{–}+{\mathrm{H}}_2\mathrm{O}\mathrm{b}=60\mathrm{mV}/\mathrm{dec}$$

(4)

$${\mathrm{MO}}^{–}\to \mathrm{MO}+{\mathrm{e}}^{\text{–}}\mathrm{b}=45\mathrm{mV}/\mathrm{dec}$$

(5)

$$2\mathrm{MO}\to 2\mathrm{M}+{\mathrm{O}}_2\mathrm{b}=19\mathrm{mV}/\mathrm{dec}$$

(6)

The charge transfer resistance (R_CT) gives insight into the OER kinetics. A low R_CT indicates a fast charge transfer rate during a redox reaction [73]. The R_CT values were obtained by electrochemical impedance spectroscopy (EIS) which was conducted at 1.5 V vs. RHE in a frequency range of 0.01 Hz to 10 kHz at an alternating current (AC) with a potential amplitude of 10 mV. The resulting Nyquist plots, fitted to a simple Randles cell model, are presented in Figure 10 and Figure S23b, SI [74]. The diameter of the displayed semi-circles give insight into the R_CT value, a smaller semi-circle indicates a lower R_CT and following faster OER kinetics [20,75].

Figure 10. Nyquist plots of RuO₂, pristine MOFs and (a) in situ and (b) postsynthetically mixed composites. The inset is a magnification between Re(Z) = 0 – 40 Ω.

Figure 10. Nyquist plots of RuO₂, pristine MOFs and (a) in situ and (b) postsynthetically mixed composites. The inset is a magnification between Re(Z) = 0 – 40 Ω.

The mixed-metal Ni_xFe-MOF samples and their carbon composites exhibit a much lower R_CT values (<33 Ω) than neat Ni-MOF and its composites (R_CT 80 Ω, Table S7, SI). The mixed-metal Ni_xFe-MOF samples and their carbon composites differ primarily by the Ni:Fe ratio, with the Ni₁₀Fe materials between 17-33 Ω and the Ni₅Fe materials between 7-15 Ω, such that the carbon composites vary around the value of the neat MOF (Ni₁₀Fe, R_CT = 20 Ω and Ni₅Fe, R_CT = 10 Ω). Overall, the in situ synthesized Ni₅Fe-CNT has the lowest R_CT = 7 Ω corroborating our above low overpotential and low Tafel-slope results that assigns Ni₅Fe- CNT the highest electrocatalytic OER activity. RuO₂ has an R_CT of 39 Ω, which matches the literature (RuO₂ R_CT = 40 Ω [76]).

Since the mixed-metal samples with the highest amount of iron (Ni₅Fe) have the best OER performance they were further characterized with their faradaic efficiencies (FE) (using the method of Galán-Mascarós et al. [77]; see materials and methods for details). Ni₅Fe-BTC achieved the highest FE with 96%, closely followed by Ni₅Fe-CNT with FE = 95% and Ni₅Fe-KB with FE = 94% (Table S7, SI). The postsynthetically mixed Ni₅Fe/KB reached a FE of 92% and Ni₅Fe/CNT a FE of 86%. All samples, with the exception of Ni₅Fe/CNT (FE = 86%), attained a higher FE than RuO₂ (FE = 91%) (Lit. FE ~93% [78]). The neat carbon materials KB achieved a FE of 72% and CNT a FE of 77%. A graphical illustration of the FEs is displayed in Figure S26, SI.

As mentioned in the introduction and at the beginning of this electrochemical section, MOFs typically act as pre-catalysts, that form the active phase (often metal (oxy)hydroxides) in the alkaline electrolyte and under the electrochemical reaction conditions due to their instability in the electrolyte medium [12,19,20,21]. The behavior of the MOFs and their carbon composites in the alkaline electrolyte was investigated through soaking the materials in 1 mol/L KOH for 24 h. In the PXRD patterns of all samples (Figure 11) a change from the MOF structure to their metal (oxy)hydroxides can be observed through reflexes, which fit to β-Ni(OH)₂ (ICSD: 169978) and/or γ-NiOOH (COD: 9012319). Similar structural changes and loss of crystallinity of Ni_xFe-MOFs and their composites were already observed in the reported literature [14,20,28,35]. Noteworthy, the crystallinity of the hydrolysis products varies strongly from largely amorphous for the Ni-MOF-carbon composites and Ni₅Fe-KB to well crystalline for the Ni₁₀Fe-MOF and its carbon composites. In a 12 h chronopotientiometric measurement a good stability of all derived electrocatalysts and RuO₂ could be shown.

3. Materials and Methods

3.1. Materials

The used chemicals were obtained from commercial sources and no further purification was carried out. The carbon nanotubes were multi-walled carbon nanotubes (OD: 10-30°nm; L: 5-15°µm; 95+%) from IOLITEC Ionic Liquids Technologies GmbH (Heilbronn, Germany). Ketjenblack EC 600 JD was purchased from AkzoNobel, The Netherlands. The MOFs Ni-BTC, Ni₁₀Fe, Ni₅Fe and their carbon composites were synthesized according to our previous work Ni_xCo-MOF work with some modifications [35].

3.2. Synthesis of the In Situ MOF-Composites

For the Ni-BTC sample 378 mg (1.3 mmol) Ni(NO₃)₂·6 H₂O, 205 mg (0.98 mmol) H₃BTC and 55 mg (0.67 mmol) 2-MeImH were dissolved in 15 mL of DMF at RT and stirred for 30 min [40]. For the in situ Ni-KB or CNT samples the same amounts were used and 10°wt% KB or CNT with respect to the educts for the MOF were added.

For the Ni₁₀Fe sample 349 mg (1.2 mmol) Ni(NO₃)₂·6 H₂O, 48 mg (0.11 mmol) Fe(NO₃)₃·9 H₂O, 205 mg (0.98 mmol) H₃BTC and 55 mg (0.67 mmol) 2-MeImH were dissolved in 15 mL of DMF at RT and stirred for 30 min. For the in-situ Ni₁₀Fe-KB or -CNT samples the same amounts were used and 10°wt% CNT or KB in regards to the weighted educts for the MOF were added.

For the Ni₅Fe sample 314 mg (1.08 mmol) Ni(NO₃)₂·6 H₂O, 89 mg (0.22 mmol) Fe(NO₃)₃·9 H₂O, 205 mg (0.98 mmol) H₃BTC and 55 mg (0.67 mmol) 2-MeImH were dissolved in 15 mL of DMF at RT and stirred for 30 min. For the in-situ Ni₅Fe-KB or -CNT samples the same amounts were used and 10°wt% CNT or KB in regards to the weighted educts for the MOF were added.

The prepared solutions were transferred into a Teflon-lined stainless-steel autoclave and then heated to 170 °C for 48 h. The resulting dark green (Ni-MOF), dark olive-green (Ni₁₀Fe, Ni₅Fe) and black (Ni_xFe-KB or -CNT) precipitates were separated by centrifugation (15 min, 6000 rpm). The precipitates were washed once with DMF (15 mL, 5 h) and twice with EtOH (15 mL each, 10 h) and centrifuged again (15 min, 6000 rpm). The products were dried overnight in a vacuum drying cabinet at 120 °C and a reduced pressure of less than 50 mbar.

Yields:

Ni-MOF: 277 mg; Ni-KB: 353 mg; Ni-CNT: 318 mg

Ni₁₀Fe: 320 mg; Ni₁₀Fe-KB: 434 mg; Ni₁₀Fe-CNT: 358 mg

Ni₅Fe: 366 mg; Ni₅Fe-KB: 430 mg; Ni₅Fe-CNT: 472 mg

For the postsynthetically-mixed MOF/carbon composites, the synthesized neat MOFs were physically mixed with 10 wt% KB or CNT in regards to the weighted neat MOF amount. Specifically, 50 mg of the neat MOFs and 5.5 mg of the carbon material (KB or CNT) were mixed at 40 Hz for 20 min with a vortex mixer.

3.3. Materials Characterization

A Bruker D2 Phaser powder diffractometer (Bruker AXS, Karlsruhe, Germany) with a power of 300°W and an acceleration voltage of 30 kV at 10 mA using Cu-Kα radiation (λ = 1.5418 Å) was used to make the PXRD measurements at ambient temperature. The diffractograms were gathered on a low background flat silicon sample holder and evaluated with the Match 3.11 software. The measuring range was from 5 to 50° 2θ with a scan speed of 2 s/step and 0.057° (2θ) step size.

The FT-IR spectra were measured in KBr mode on a Bruker TENSOR 37 IR spectrometer (Bruker, Rosenheim, Germany) in the range of 4000-400 cm^–1.

Nitrogen sorption isotherms were recorded with a Nova 4000e from Quantachrome (Anton Paar QuantaTec, Boynton Beach, Florida, USA) at 77 K. The sorption isotherms were analyzed with the NovaWin 11.03 software. Before the measurements the samples were activated for 5 h under vacuum (<10^–2 mbar) at 120 °C. Brunauer–Emmett–Teller (BET) surface areas were determined from the adsorption branches of the isotherms and the pore size distributions were derived by non-local density functional theory (NLDFT) calculations based on N₂ at 77 K on carbon with slit/cylindrical pores.

TGA was performed with a Netzsch TG 209 F3 Tarsus device (Erich NETZSCH B.V. & Co. Holding KG, Selb, Germany) equipped with an Al crucible. A heating rate of 10 K/min under a N₂ atmosphere was applied for the measurements.

SEM images were acquired with a JEOL JSM-6510 LV QSEM advanced electron microscope (Jeol, Akishima, Japan) with a LaB₆ cathode at 20 kV. The microscope was equipped with a Bruker Xflash 410 silicon drift detector (Bruker, Berlin, Germany) and the Bruker ESPRIT software for the EDX analysis. The Au, Cu and Zn detected in the EDX spectra can be attributed to the sputtering of the sample with gold prior to the investigation and the brass sample holder, which was utilized.

The vortex mixer, which was used to generate the postsynthetic composites, is a REAX 2000 from Heidolph (Heidolph Scientific Products GmbH, Schwabach, Germany).

3.4. Electrocatalytic Measurements

A three-electrode setup was used for the electrocatalytic measurements, which were recorded with a SP-50e potentiostat from BioLogic (BioLogic Science Instruments, Göttingen, Germany). The counter electrode was a Pt plate, the reference electrode was a reversible hydrogen electrode (RHE) from Gaskatel (Kassel, Germany), and the working electrode was an ink-coated carbon paper. A carbon paper sheet was sliced into 1x1 cm pieces. The electrocatalyst inks contained 2.5 mg of electrocatalyst, 0.5 mL ethanol and 20 μL Nafion (5 wt.%) and were sonicated for 30 min. 100 μL ink was drop-casted onto a sliced carbon paper piece (geometric area of 1 cm²), which resulted in a catalyst loading of 0.5 mg/cm². The loading of the electrocatalyst onto the carbon paper was additionally checked with the weight gain of the carbon paper piece, which showed also a catalyst loading of 0.5 mg_sample/g_{carbon paper}.

SEM images of the carbon paper with and without ink-coating before and after the electrochemical measurements for the mixed-metal samples in Figure S16 to S21, SI show a good adherence of the materials to the carbon paper substrate. 1 mol/L KOH was utilized as the electrolyte for all electrocatalytic measurements, which has been purged with N₂ for 10 min before the electrocatalytic experiments. The samples were activated by cycling the working electrode between 1.0 V and 1.7 V vs. RHE at a scan rate of 100 mV/s for 10 cycles. The LSV polarization curves were measured in a potential range of 1.0 to 1.7 V vs. RHE at a scan rate of 5 mV/s. The cycling stability was measured by comparing activated LSV curves before and after 1000 CVs between 1–1.7 V with a scan rate of 100 mV/s. EIS data was collected in a frequency range of 0.01 Hz to 10 kHz at a potential of 1.5 V vs. RHE. Chronopotentiometry was conducted following the above mentioned electrocatalytic measurements. The current density was held at 10 mA/cm² for 12 h. The overpotential was calculated as shown in equation 1: η = E – E° (1.23 V) – IR. The OER performance of the MOF samples and their composites were compared with commercial RuO₂ (Sigma-Aldrich), multi-walled CNTs (OD: 10-30°nm; L: 5-15°µm; 95+%; IOLITEC Ionic Liquids Technologies GmbH, Heilbronn, Germany) and KB (AkzoNobel Netherlands). At least two measurements were done for each sample to determine the overpotentials, Tafel slopes and R_CT in order to reduce the experimental contingency error and the averaged results were displayed in the figures and tables. All electrochemical measurements were automatically corrected with the IR.

To determine the FE the method of Galán-Mascarós et al.[77] was followed. A chronopotentiometric measurement was conducted at a fixed current density of 10 mA/cm², while an Ocean Optics NeoFOX sensor system coupled with a FOSPOR probe was used to record the O₂ level in the headspace of the electrochemical cell. The calibration of the FOSPOR probe was done through a two-point calibration with N₂ atmosphere (0% O₂) as one calibration point and ambient air (21% O₂) as the other calibration point. Before the chronopotentiometric test the electrolyte underwent thorough deaeration by continuous purging with N₂ for at least 1 h. To calculate the molar amount of O₂ evolved during the electrochemical measurement, equation 7 was used, which considers ideal gas behavior:

n_O2,exp = (% O_2,det × P_total × V_gas / R × T) / 100

(7)

where % O_2,det is the detected percentage of oxygen measured by the FOSPOR probe (corrected by the % O_2,det from a measurement without any applied current to consider the oxygen leakage from ambient air); P_total is 1 atm; V_gas (L) is the developed gas volume at atmospheric pressure; R is the universal gas constant (0.082 atm × L / K × mol); and T is the absolute temperature (293 K). The theoretically generated faradaic oxygen is obtained through equation 8:

n_O2,_far = Q / n_e × F

(8)

where, Q (measured in Coulomb, C) stands for the total electric charge passed through the system; n_e represents the needed molar amount of electrons to produce one mol of O₂ (equals to 4), and F is the Faraday constant (96485 C/mol). The FE (expressed in percentage) is determined by equation 9:

FE = 100 × n_O2,exp / n_O2,_far

(9)

4. Conclusions

In this work a series of Ni-, mixed-metal Ni_xFe-MOFs and their composites with the carbon materials ketjenblack, KB and carbon nanotubes, CNT were synthesized both in situ through a simple one-pot solvothermal reaction which contains the MOF precursors and the carbon material or postsynthetically mixed from the prepared MOF and carbon material. The in situ synthesized MOF-carbon composites were fully characterized concerning the successful MOF formation in the presence of carbon. As seen before, in the alkaline electrochemical environment, the MOFs act as precursors to the actual metal (oxide)hydroxide catalysts and provide an intimate mixture of the mixed metals and with the carbon material on the nanoscale in the derived catalyst material. The carbon materials KB and CNT increase the low electrical conductivity of the derived metal (oxide)hydroxide electrocatalysts. All of the tested Ni₁₀Fe- and Ni₅Fe-MOF-carbon electrocatalysts outperformed the RuO₂ benchmark, whereas the nickel-only Ni-BTC and its carbon composite was similar or worse as RuO₂. Thus, through an increasing amount of Fe the electrocatalytic activity regarding the OER could be improved. Concerning the assessed comparison of the in situ and postsynthetic preparation of the MOF mixtures with KB and CNT for KB the postsynthetic materials and for CNT the in situ ones have the lower overpotential while the Tafel slopes, charge transfer resistance R_CT and faradaic efficiency are rather similar. Evidently, no clear advantage can be seen for one type of MOF-carbon composite formation over the other, although a large number of experiments were carried out. At the same time, the addition of Fe to neat Ni-BTC had a significant effect on reducing the overpotential, Tafel slope and charge transfer resistance. The best performing MOF-derived electrocatalyst came from in situ synthesized Ni₅Fe-CNT with an overpotential (η) of 301 mV (RuO₂ η = 354 mV), a Tafel slope (b) of 58 mV/dec (RuO₂ b = 91 mV/dec), a charge transfer resistance (R_ct) of 7 Ω (RuO₂ R_ct= 39 Ω) and a faradaic efficiency (FE) of 95% (RuO₂ FE = 91%) but the postsynthetic materials Ni₅Fe/KB and /CNT were close behind. Yet, for the composite of Ni₁₀Fe with KB the postsynthetic mixture performed significantly better in terms of overpotential and RCT than the in situ material (314 vs 339 mV and 17 vs 33 Ω, respectively) which shows that the type of mixing can still present an additional improvement. Future research could take the effect on the OER activity of in situ and postsynthetically mixed composites with a higher amount of KB or CNT into account and other carbon materials could be further added to the comparison.