Iron and Manganese Containing Multi‐Walled Carbon Nanotubes as Electrocatalysts for the Oxygen Evolution Reaction ‐ Unravelling Influences on Activity and Stability

Hydrogen economy is a central aspect of future energy supply, as hydrogen can be used as energy storage and fuel. In order to make water electrolysis efficient, the limiting oxygen evolution reaction (OER) needs to be optimized. Therefore, C‐based composite materials containing earth‐abundant Fe and Mn were synthesized, characterized and tested in the OER. For pyrolysis temperatures above 700 °C N‐rich multi‐walled carbon nanotubes (MWCNT) are obtained. Inside the tubes Fe3C particles are formed, Fe and Mn oxides are incorporated in the carbon matrix and metal spinel nanoparticles cover the outer surface. The best catalyst prepared at 800 °C achieves a low overpotential of 389 mV (at 10 mA/cm2) and high stability (22.6 h). From electrochemical measurements and characterization it can be concluded that the high activity is mainly provided by MWCNT, Fe3C and the metal oxides in the conductive carbon matrix. The metal spinel nanoparticles in contrast protect the MWCNT from oxidation and thereby contribute to the high stability.


Introduction
Environmental pollution and a rising energy consumption of the growing world population represent major global challenges. Renewable energies such as wind, solar or water power facilitate a sustainable energy supply; however, due to their temporal and spatial intermittency, concepts for energy storage are needed. [1][2][3] Electrochemical water splitting to generate hydrogen as portable energy source represents one possible solution. [4][5][6] The reaction is limited by the high overpotential (OP) of the oxygen evolution reaction (OER) (Scheme 1). [7][8][9] RuO 2 and IrO 2 were the first examples for highly active electrocatalysts in the OER at high pH value. [10][11][12] Drawbacks of these catalysts are the high raw material price and their unsatisfying stability. An alternative are non-noble mixed metal oxides. [13][14] However, such materials often suffer from low specific surface area, porosity and electrical conductivity. [15] To enhance dispersion and electron transfer, the metal oxides were supported on conductive carriers such as graphene and carbon nanotubes (CNT). [16,17] CNT provide a good electrocatalytic performance due to their conductivity, mechanical stability, chemical resistance and high specific surface area. [18][19][20][21][22] The stability of a metal-free CNT catalyst is yet reported to be low as deactivation occurs in under 2 h of reaction. [21] The combination with metal oxides allows an improved activity in electrochemical reactions. [23][24][25][26][27][28][29] Antoni et al. [30] prepared Mn x O y nanoparticles of different oxidation state on O-and Nfunctionalized CNT finding the highest activity for Mn in a high oxidation state. However, the reported stability was low. Masa et al. [31] synthesized Mn x O y and Co x O y nanoparticles embedded in an N-doped carbon matrix. They demonstrated that the interaction of N-doped carbon and metal oxide enhances the electrocatalytic performance. In addition, they observed a rapid degradation of the catalyst after few cycles of double-pulse chronopotentiometry. Wen et al. [32] and Ma et al. [33] reported that metal nanoparticles encapsulated in a carbon matrix possess an excellent OER activity due to electronic interaction between metal and carbon tuning the redox properties of the surrounding carbon layer. Also Fe 3 C has been reported to have the ability to catalyze electrochemical reactions. [34][35][36] Barman et al. [37] report an Fe/Fe 3 C encapsulated N-doped CNT catalyst with long term stability over 1000 cycles. Many studies address the synthesis and characterization of electrocatalysts composed of metal oxides, metal carbides and conductive carbon as well as their application in OER. [38][39][40] Some authors are able to reach high activity and outstanding stability, however, these studies often lack a detailed investigation of the exact origin of the high performance of the composite materials. Herein, we report a simple strategy to prepare a catalyst consisting of Fe, Mn and N-rich carbon. Thorough characterization elucidates the catalyst structure and composition, which is used to explain influences on the OER activity and stability in an alkaline electrolyte solution.

Characterization of the electrocatalysts
To synthesize the electrocatalysts, MnCl 2 , FeCl 2 , and FeCl 3 were used as metal precursors and ground together with dicyandiamide (DCD). The choice of metals and metal ratios is based on a previous study, whose results are briefly summarized in Figures S1 and S2 in the ESI. The mixture was pyrolyzed under nitrogen atmosphere at temperatures between 600°C and 900°C. After pyrolysis, the catalysts were ground and washed with NaOH and H 2 O to remove excess metal salts (detailed experimental procedures can be found in the experimental section). The materials were named Fe,Mn@DCD p -T and characterized by N 2 -physisorption, CHN, ICP-OES, Raman spectroscopy, XPS, STEM(-EDX), and XRD. All synthesized materials are porous according to N 2 -physisorption analysis ( Figure S3 in the ESI). They show a type IV isotherm after IUPAC and a specific surface area of up to 136 m 2 /g ( Table 1). The N-content of the materials obtained from elemental analysis decreases with increasing pyrolysis temperature from 20 % (600°C) to 1.5 % (900°C). Simultaneously, the C-content of the materials increases with increasing synthesis temperature from 17 % to 37 %. ICP-OES analysis of the materials shows a decreasing Mn-content (19 to 11 wt%) and an increasing Fe-content (32 to 42 wt.%) with increasing pyrolysis temperature.
From Raman spectroscopy two characteristic signals for carbon based materials can be identified (Figure 1): The G-band at around 1580 cm À 1 originates from the Raman-active E 2g mode of graphitic carbon, while the D-band at 1350 cm À 1 can be referred to amorphous regimes in the carbon matrix. [41] The ratio of the intensity of D-and G-band provides information about the materials degree of graphitization and therefore its ability to conduct electrons. The ratio of I D /I G decreases for the materials with increasing pyrolysis temperature. While for a pyrolysis temperature of 600°C an I D /I G ratio of 1.1 is obtained, it decreases to 0.7 for the material prepared at 900°C proving a higher crystallinity and therefore more graphitic nature. Additionally, Raman modes at low and high wave numbers are observed.
The high frequency signals at 2650 cm À 1 and 2900 cm À 1 are known as the 2D and D + D' signals corresponding to overtones. [42] The 2D signal originates from a second-order overtone of different in-plane vibrations. For samples prepared at 800°C and 900°C some additional signals appear at low wave numbers of 215 cm À 1 , 277 cm À 1 , 389 cm À 1 , 474 cm À 1 and 583 cm À 1 . They can be referred to two T 2g , one E g and one A 1g Raman mode according to literature. [43] These modes indicate the presence of Fe 3 O 4 spinel on the surface of the materials.
To further elucidate the structure of the materials XPS analysis was performed. According to the XPS survey spectra of the electrocatalysts synthesized at 700°C to 900°C Cl, C, N, O, Mn and Fe are present on the surface ( Figure S4 in the ESI). The carbon concentration decreases with increasing pyrolysis tem- Table 1. Physical properties of Fe,Mn@DCD p -materials: surface area was obtained from N 2 -physisorption analysis using the BET-method, atomic ratio of Fe/Mn was determined via ICP-OES, elemental analysis yielded the fractions of C, H and N, the wall thickness of the multi-walled carbon nanotubes (MWCNT) was extracted from STEM images.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57 perature. Simultaneously the surface concentration of O and Fe increases. Therefore, the surface composition of the materials can be altered by tuning the pyrolysis temperature. From the element spectra ( Figure S5 in the ESI), four different carbon species can be identified from the C 1s signals corresponding to sp 3 -and sp 2 -hybridized carbon, ether/ester-and carbonylgroups. Two different nitrogen species corresponding to pyridinic and pyrrolic nitrogen can be identified from the N 1s spectrum. The Mn 2p 3/2 and Mn 2p 1/2 signals appear at a binding energy of 641.5 and 652.7 eV, respectively. Due to the low surface concentration of Mn, the signals were not fitted. Nevertheless, the binding energy indicates the presence of bior trivalent Mn species on the surface. [44] The Fe 2p 3/2 and Fe 2p 1/2 peaks show binding energies of 710.2 and 723.5 eV, respectively. The peak position, the shape of the obtained signals and the fitting indicates the presence of metal oxides, i. e. Fe 2 O 3 and/or Fe 3 O 4 , on the surface. [45,46] STEM analysis of the synthesized materials reveals the formation of an unstructured carbon at low pyrolysis temperature of 600°C (Figure 2a and b). The highest I D /I G value of 1.1 obtained from Raman spectroscopy for Fe,Mn@DCD p -600°C suggests the formation of amorphous carbon ( Figure S6 in the ESI). Additionally, XRD of the sample reveals no distinct peaks due to the high signal-to-noise ratio typically observed for amorphous carbon ( Figure S7 in the ESI). For an increased pyrolysis temperature the formation of CNT is observed (Figure 2 c-h). These tubes are multi-walled (MWCNT) and the wallthickness increases with increasing pyrolysis temperature to maximally 102 nm (Table 1) To clearly identify the formed metal species inside and outside of the MWCNT, the materials synthesized at 700°C to 900°C were analyzed by XRD (Figure 3). The broadened diffraction peaks at 2θ values of 25.8°and 43.1°can be attributed to graphitic MWCNT. [47] Also Fe 3 O 4 (30.0°(220), 35.0°( 311), 46.0°(400), 54.0°(422), 56.0°(511), 64.0°(440)) (JCPDS no. 74-0748) [45] as well as Mn 3 O 4 spinels can be identified (18.0°( 101), 32.5°(103), 37.4°(211), 52.5°(105)) (JCPDS no. 24-0734). [42,48] The reflex at 44.7°additionally indicates the presence of Fe 0 (JCPDS no. 87-0722). [32] When taking a closer look into the region of 39°to 50°for materials synthesized at 700°C and 800°C, many small diffraction signals are visible. These signals indicate the presence of  [49] It is literature known that Fe 3 C undergoes decom-position to Fe 0 and graphitic carbon at elevated temperatures, explaining why Fe 3 C was not found for materials synthesized at 900°C. [50] Figure S12 in the ESI). The analysis clearly proves that metal oxides are not only formed on the outer surface of the tubes but are also incorporated into the carbon matrix.
For Fe,Mn@DCD p -materials prepared at temperatures of 700°C or above N-rich MWCNT are formed. They are composed of C, N and metal oxides. Inside the MWCNT, Fe 3 C is formed,  Figure S13 in the ESI).

Electrochemical performance of Fe,Mn@DCD p -materials
The electrochemical OER activity of the prepared Fe,Mn@DCD pmaterials and a commercial NiCoO x -catalyst was evaluated by linear sweep voltammetry (LSV) in 1 M KOH. The working electrode was a rotating disc electrode (RDE) coated with the catalyst and operated at 1600 rpm. The LSV results are depicted in Figure 4a. With an increasing synthesis temperature, the OP decreases to a minimum of 389 mV at a current density of 10 mA/cm 2 for the material prepared at 800°C. This is comparable to the performance of NiCoO x , reaching an OP of 390 mV. In addition, the H 2 SO 4 treated material without spinel particles on the outer surface of the MWCNT reaches a low OP of 397 mV as well. For a further increasing pyrolysis temperature, the OP increases again. The poor performance of the material prepared at 600°C is attributed to the absence of MWCNT and therefore to a low conductivity. With increasing pyrolysis temperature additionally, the formation of Fe 3 C is facilitated. At further increased pyrolysis temperatures, anyhow, Fe 3 C is decomposed to Fe 0 and graphitic carbon resulting in a decreased OER activity. The presence of the Mn and Fe spinel particles on the outer surface of the MWCNT in contrast do not have a major influence on the OER activity, as Fe,Mn@DCD p -800°C and Fe,Mn@DCD p -800°C-H 2 SO 4 show a comparable performance. Fe and Mn oxides are anyhow known to be active compounds for the OER, so their role cannot be neglected. [51,52] The STEM-EDX and XRD analysis of the samples confirmed that the Fe and Mn oxides are not only present on the catalyst surface but also incorporated in the carbon matrix. Therefore, metal oxides as well as MWCNT are identified to be active species and responsible for a good activity. Additionally, an increased activity is found for materials possessing Fe 3 C instead of Fe 0 particles inside the tubes, so that Fe 3 C is considered as active species as well. It can on the one hand contribute to the conductivity of the electrode material and on the other hand function as adsorption site for the substrate OH À . Even though the particles are located inside the tubes, they can be accessed by the substrate as the tubes are open at one end as seen in STEM ( Figure S14 in the ESI). As temperature controls the formation of Fe 3 C and MWCNT, it strongly influences the electrochemical activity of the catalysts.
For a deeper understanding of the behavior of the composite materials, electrochemical impedance spectroscopy (EIS) was carried out as shown in Figure 5a. The impedance spectra consist of a complete semicircle for all materials, which can be simulated using a Randles-circuit. In accordance with the catalytic performance, Fe,Mn@DCD p -800°C and Fe,Mn@DCD p -800°C-H 2 SO 4 exhibit the lowest charge transfer resistance r CT of 45 Ω and 31 Ω, respectively. The lower charge transfer resistance after the etching of the MWCNT surface using H 2 SO 4 can be referred to the removal of weakly conductive metal oxide species. We further calculated the roughness factor r F , from which the electrochemical surface area ECSA is derived according to equations S1 and S2 (see ESI). For the calculation of r F the double layer capacitance c DL of the composite materials is needed, which can be extracted from CV curves (1.0 to 1.1 V vs. RHE) at different scan rates (Figure 5b).  The obtained c DL values vary between 0.08 and 3.8 mF/cm 2 , where the highest c DL value is obtained for Fe,Mn@DCD p -800°C. r F was then used to normalize the activity and derive an intrinsic catalyst activity J int ( Table 2). The calculation of J int is shown in equation S3 in the ESI. Fe,Mn@DCD p -800°C exhibits the highest ECSA (13.2 cm 2 ), the highest J int of 0.87 mA/cm 2 is obtained for Fe,Mn@DCD p -700°C. From the data, it becomes clear that the specific surface area has no influence on J int .
To evaluate the practical applicability of an OER catalyst, also the stability plays an important role and was investigated by applying a constant current density of 10 mA/cm 2 (Figure 4b). The potential obtained over Fe,Mn@DCD p -600°C increases directly and the catalyst deactivates within 1.3 h of reaction. Also the commercial NiCoO x shows a low stability as it deactivates within the first 2 h of reaction. With increasing pyrolysis temperature, an increasing stability of up to 22.6 h for Fe,Mn@DCD p -800°C is obtained. Further increase of the pyrolysis temperature to 900°C or the treatment of the material with H 2 SO 4 leads to a decreased stability of 4.7 h and 8.7 h, respectively. Compared to literature reported results for carbonbased materials, the prepared Fe,Mn@DCD p -materials show lower or comparable OP and a higher stability (Table 2). [30,53 The stability of the prepared materials is even higher than obtained for IrO x , [9] the OP is not as low.
The especially high stability of Fe,Mn@DCD p -800°C can be explained by taking a look at the STEM images, revealing that the complete surface of the MWCNT is covered with metal spinel particles protecting the surface of the MWCNT. As instability due to oxidation in corrosive electrolyte is a frequently reported issue of carbon-based materials in the OER, the spinel nanoparticles seem to be beneficial to protect the MWCNT from decomposition. [54] Thus, catalytic activity and stability of the materials are both highly influenced by the variation of the pyrolysis temperature. At temperatures above 800°C, the formed Fe 3 C inside the MWCNT decomposes to Fe 0 and graphitic carbon. The improved stability of the materials results from Mn 3 O 4 and Fe 3 O 4 spinel nanoparticles, covering the outer surface of the MWCNT and thereby protecting the catalyst from oxidation. The activity instead rather results from the Nrich MWCNT, the metal oxides incorporated in their structure, and the Fe 3 C particles inside.

Conclusion
In the presented study a series of carbon-based composite materials containing different Fe and Mn species were synthesized and extensively characterized by N 2 -physisorption, CHN, ICP-OES, Raman spectroscopy, XPS, STEM(-EDX), and XRD. The complementary methods reveal that a minimum temperature of 700°C is needed to ensure the formation of highly graphitic and conductive N-rich MWCNT. The materials are decorated on their outer surface with Mn 3 O 4 and Fe 3 O 4 spinel nanoparticles. Importantly, Fe and Mn oxides are also firmly incorporated into the carbon matrix. Furthermore, the tubes contain Fe 3 C particles inside that decompose to Fe 0 particles and graphitic carbon at elevated pyrolysis temperatures. Electrochemical measurements of the composite materials revealed that the MWCNT with incorporated N-moieties and Fe and Mn oxides as well as the Fe 3 C particles contribute to a high OER activity. The most active catalyst synthesized at a pyrolysis temperature of 800°C achieves a low OP of 389 mV. Additionally, the spinel nanoparticles covering the outer surface of the electrocatalyst ensure a stability of up to 22.6 h, which is high for a carbon-based OER electrocatalyst and a step forward to practical application. The study shows that the adjustment of different material properties can lead to an optimized composite material, which enables the OER with simultaneously high activity and stability.
The ink was ultra-sonicated for 15 min. 2.5 μL of catalyst ink was pipetted onto the glassy carbon rotating disk electrode (RDE) and dried at RT for 30 min. The tip possesses a geometric surface area of 0.1396 cm 2 . Therefore, a catalyst film with a loading of 89.5 μg catalyst /cm 2 resulted.
Electrochemical measurements were performed in a glass cell using a three electrode setup at RT. The modified glassy carbon RDE was used as working electrode. A glassy carbon and an Ag/AgCl (3 M KCl) electrode were used as the counter and the reference electrode, respectively. 1 M KOH was used as electrolyte. The electrocatalytic activity was investigated by LSV using an Autolab potentiostat by Metrohm. Potentials were referenced to the reversible hydrogen electrode (RHE). Before data aquisition, the catalysts underwent continuous potential cycling (CV) until steady voltammograms were obtained (100 CVs @ 0.1 V/s). Afterwards LSV was carried out in a potential range of 1.00 to 1.86 V at a rotation speed of the working electrode of 1600 rpm and a scan rate of 5 mV/s. c DL was measured by CV scans at various scan rates in the range 1.0 to 1.1 V vs. RHE. EIS was performed at an OP of 179 mV with an amplitude of the sinusoidal voltage perturbation of 10 mV over a frequency range of 100 kHz to 50 mHz. The long-term performance was investigated by applying a steady current density of 10 mA/cm 2 .  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57