Oxidative Deposition of Manganese Oxide Nanosheets on Nitrogen-Functionalized Carbon Nanotubes Applied in the Alkaline Oxygen Evolution Reaction.

The development of nonprecious catalysts for water splitting into hydrogen and oxygen is one of the major challenges to meet future sustainable fuel demand. Herein, thin layers of manganese oxide nanosheets supported on nitrogen-functionalized carbon nanotubes (NCNTs) were formed by the treatment of NCNTs dispersed in aqueous solutions of KMnO4 or CsMnO4 under reflux or under hydrothermal (HT) conditions and used as electrocatalysts for the oxygen evolution reaction (OER) in alkaline media. The samples were characterized by X-ray photoelectron spectroscopy, X-ray diffraction, transmission electron microscopy, and Raman spectroscopy. Our results show that the NCNTs treated under reflux were covered by partly amorphous and birnessite-type manganese oxides, while predominantly crystalline birnessite manganese oxide was observed for the hydrothermally treated samples. The latter showed, depending on the temperature during synthesis, an electrocatalytically favorable reduction from birnessite-type MnO2 to γ-MnOOH. OER activity measurements revealed a decrease of the overpotential for the OER at a current density of 10 mA cm-2 from 1.70 VRHE for the bare NCNTs to 1.64 VRHE for the samples treated under reflux in the presence of KMnO4. The hydrothermally treated samples afforded the same current density at a lower potential of 1.60 VRHE and a Tafel slope of 75 mV dec-1, suggesting that the higher OER activity is due to γ-MnOOH formation. Oxidative deposition under reflux conditions using CsMnO4 along with mild HT treatment using KMnO4, and low manganese loadings in both cases, were identified as the most suitable synthetic routes to obtain highly active MnO x /NCNT catalysts for electrochemical water oxidation.

Introduction

The most efficient and sustainable method for storing renewable energy is in chemical bonds, for instance, by producing molecular hydrogen, an energy carrier with the highest gravimetric energy density.[1−4] Thus, electrochemical water splitting into hydrogen and oxygen is a promising alternative complementary to other energy storage technologies.[5] The bottleneck of electrochemical water splitting is the slow kinetics of the oxygen evolution reaction (OER). From the thermodynamic point of view, the standard potential to drive water splitting is 1.23 V.[6] However, much higher anodic potentials have to be applied to achieve dioxygen formation at a meaningful rate because of inevitable overpotentials.[7] Therefore, lowering the overvoltage for water splitting, particularly the overpotential of the OER, is one of the most important goals in electrocatalysis research.[8] Up to now, precious metal oxides, typically RuO2 and IrO2, are the best OER catalysts under acidic conditions,[9,10] making them unappealing for large scale use. To reduce costs, the development of nonprecious, highly abundant, and nontoxic electrocatalysts is essential. Therefore, transition-metal oxides such as nickel oxide, cobalt oxide, or manganese oxides have been widely investigated to serve as OER catalysts under alkaline conditions.[11] Inspired by the ability to evolve oxygen at a CaMn4O5 cluster in the natural protein complex photosystem II, manganese oxide-based catalysts for the OER have been a topic of intense investigation.[12] The high abundance, low toxicity, and unique redox properties of manganese render manganese-based catalysts promising electrode materials.[13]

Most manganese oxides are built from differently connected [MnO6] octahedra, leading to a multitude of Mn oxides/hydroxides with different crystal structures.[14] Numerous manganese oxides have been synthesized and applied as OER electrocatalysts.[15−21] It has been shown that many structural features of manganese oxides such as the oxide phase, a large extent of di-μ-oxo ligands within the oxide, and the simultaneous presence of Mn3+ and Mn4+ play a crucial role in influencing the electrocatalytic performance of manganese oxides.[22] It was shown by Takashima et al.[23,24] that high activities can only be achieved in the presence of Mn3+. Therefore, an alkaline environment becomes necessary because Mn3+ disproportionates into Mn2+ and Mn4+ in acidic media. Huynh et al.[25] proposed a proton-coupled one-electron transfer (PCET) pathway in alkaline media and a Mn3+ disproportionation mechanism in acidic media. In the neutral regime, these two mechanisms compete. It was shown by Jin et al.[26,27] that Mn3+ might be stabilized on MnO nanoparticles, accompanied by the occurrence of the PCET. However, Pokhrel et al.[28] showed that the water oxidation activity of manganese oxides strongly depends on the applied technique, that is, electrochemical, chemical, or photochemical water oxidation, where α-MnO2 having a 2 × 2 tunnel structure is the most active bulk manganese oxide for alkaline electrochemical water oxidation so far, exhibiting a potential of 1.72 V at a current density of 10 mA cm–2. The OER performance of manganese oxides is diminished by their low intrinsic electrical conductivity. This can be mitigated by the deposition of MnO nanoparticles on carbon supports. It has recently been shown by Melder et al.[29] that sp2-hybridized carbon is more favorable than sp3 carbon, as it is less prone to carbon corrosion. Therefore, multiwalled carbon nanotubes (CNTs) are considered as promising catalyst support materials.

In previous studies, doping CNTs with oxygen and nitrogen functional groups improved the deposition and dispersion of metal oxide nanoparticles because these groups act as strong anchoring sites.[30−32] In addition, it was shown that nitrogen doping increases the electrical conductivity.[33] Different methods for MnO deposition were applied, such as grinding of MnO2 with CNTs,[34] impregnation and calcination with Mn(NO3)2[35] or Mn(CH3COOH)2,[36] and comproportionation of KMnO4 and Mn(NO3)2.[37]

Strong metal–support interactions and the presence of thin layers of manganese oxide are requirements for high electrocatalytic activity.[38,39] The most promising deposition method is based on the method developed by Hummers and Offeman[40] for the synthesis of graphite oxide using KMnO4. In accordance with this method, we aimed at thin layer deposition of MnO2 through a redox reaction between CNTs and KMnO4, as described by eq

Although this deposition technique has been investigated intensively in the literature, it has not yet been applied for the synthesis of MnO/NCNT catalysts.[29,41−44]

In this study, we investigated the oxidative deposition of manganese oxides on NCNTs and their application in the OER. The deposition was performed either by hydrothermal (HT) treatment or under reflux conditions with different manganese loadings and using different precursors (CsMnO4 and KMnO4). For the hydrothermally deposited samples, the effect of different HT treatment temperatures and duration on the resulting MnO structure and OER performance was investigated. Furthermore, we addressed the influence of residual CNT growth catalysts on their electrochemical performance by performing extensive purification.

Results

Influence of the Residual Growth Catalyst on the Electrocatalytic Performance

Prior to the oxidative deposition, the electrocatalytic behavior of the CNTs and the influence of the purification method used to remove the residual growth catalysts were investigated. The purification sequence involved partial oxidation of the as-received CNTs in HNO3 vapor first (a′), followed by purification in 1.5 M HNO3 (a″), or vice versa, that is, washing the as-received CNTs in 1.5 M HNO3 first (b′) prior to the HNO3 vapor treatment (b″). Figure shows the influence of the pretreatment procedure on the OER activity of the resulting materials.

Figure 1

Linear sweep OER voltammograms in O2-saturated KOH (1 M) at a scan rate of 5 mV s–1 and a rotation speed of 1600 rpm of the as-received CNTs, compared to the effect of different pretreatment procedures consisting of (a) gas-phase HNO3 treatment followed by washing in 1.5 M nitric acid or (b) washing as the first step followed by HNO3 treatment.

The as-received CNTs displayed the lowest activity, achieving a current density of 10 mA cm–2 at a potential of 1.85 V. After stirring the as-received CNTs in 1.5 M HNO3 (washing), the potential decreased to 1.68 V (b′) and further to 1.60 V after 48 h HNO3 vapor treatment (b″). When the as-received CNTs were first oxidized with HNO3 vapor (a′), the potential directly decreased to 1.60 V and increased to 1.70 V after subsequent washing in 1.5 M HNO3 (a″). Elemental analysis was used to determine the content of the residual growth catalysts. The as-received CNTs contained 1000 ppm Fe and 600 ppm Co, whereas Ni was not detected. The Fe and Co content increased to 1800 and 900 ppm, respectively, after oxidation in HNO3 vapor because of the carbon loss, and decreased to fully encapsulated 400 ppm Fe and 100 ppm Co after washing the OCNTs in 1.5 M HNO3. Therefore, the significantly more efficient purification sequence involving partial oxidation of the as-received CNTs in HNO3 vapor followed by washing in 1.5 M nitric acid was chosen for the subsequent investigations.

Bulk Characterization

Elemental Analysis

Elemental analysis was applied to determine the manganese loading either after reflux or after HT deposition. The samples having K(Cs)MnO4/NCNT ratios of 1:2, 1:1, and 2:1 yielded manganese loadings of 9.8 ± 0.1, 16.2 ± 0.2, and 23.3 ± 0.1 wt %, respectively, irrespective of the deposition technique. Upon increasing the temperature to 140 °C and the treatment time to 24 h at 110 °C for the 1:1 ratio, the loading increased from 16.2 to 17.1 ± 0.1 wt %. A further temperature increase to 180 °C led to an increase of the manganese loading to 19.0 wt % because of carbon loss.

X-ray Powder Diffraction (XRD) Characterization

XRD was applied to determine the manganese oxide phase after oxidative deposition under the chosen conditions. Figure a shows the XRD patterns of the samples synthesized under reflux conditions with the KMnO4/NCNT ratio varying from 1:2 to 2:1, and a sample synthesized with CsMnO4 with a CsMnO4/NCNT ratio of 1:1. Characteristic graphite reflections at 25.9° (002) and 42.9° (100) were observed in all the samples. The intensity of the (002) graphite reflection was found to decrease with the KMnO4 content. The patterns show two reflections, which can be assigned to birnessite-type manganese oxide, one at 37.0° and the broad second one in the low 2θ region between 5° and 15°. Although XRD patterns of layered manganese oxides tend to be rather featureless because of poor long-range order,[21] MnO2 with a birnessite structure (JPCDS 42-1317) was identified. It can be seen that the reflection in the low 2θ region of the sample synthesized with CsMnO4 is shifted to 8.0°, suggesting a higher interlayer distance between the edge-sharing layers formed by the MnO6 octahedra.

Figure 2

XRD patterns of the MnO/NCNTs samples, comparing (a) synthesis under reflux conditions using different K(Cs)MnO4/NCNT ratios, (b) HT synthesis (110 °C; 8 h) with different K(Cs)MnO4/NCNT ratios, (c) HT synthesis at different temperatures 110 °C (for 8 and 24 h), 140 °C (8 h), and 180 °C (8 h) for a fixed KMnO4/NCNT ratio of 1:1, and (d) HT synthesis with cesium and potassium at 110 and 140 °C (8 h).

Figure b–d compares the XRD patterns of the MnO/NCNT samples obtained by oxidative deposition under HT conditions, with (b) different loadings, (c) temperatures/durations, and (d) cations. Figure b exhibits reflections similar to those of the reflux-treated samples, assigned to the graphitic CNT structure and birnessite MnO2. The reflection in the low 2θ region at 12.4° is comparatively sharper, indicating higher crystallinity, whereas the sample having a 2:1 KMnO4/NCNT ratio shows only birnessite and CNT reflections, the samples with a 1:1 ratio exhibit additional reflections in the 39°–42° 2θ range, attributed to the formation of γ-MnOOH (manganite). For the sample synthesized with a 1:2 ratio, clear reflections in the 34°–36°, 39°–42°, and 51°–55° 2θ range can be observed, which also suggest the formation of γ-MnOOH. Figure c shows the XRD patterns of the hydrothermally treated samples obtained at different temperatures and durations. The samples treated for 24 h at 110 °C and for 8 h at 140 °C exhibit reflections that can be assigned to γ-MnOOH, while the sample treated for 8 h at 180 °C manifested reflections ascribed to Mn3O4 as well as γ-MnOOH. The reflections appear to be relatively sharp, indicating high crystallinity; meanwhile, the NCNT reflections disappeared almost completely.

The use of CsMnO4 for the HT deposition led to similar observations as for the KMnO4-derived samples and its reflux counterparts. Thus, the XRD pattern of the sample treated for 8 h at 110 °C exhibited reflections characteristic of birnessite-type MnO2. The low-angle reflection at 9° suggests a higher interlayer distance as compared to the sample prepared with KMnO4. The treatment at 140 °C also led to the formation of γ-MnOOH.

Raman Spectroscopy

Raman spectroscopy was applied to obtain information regarding the graphitic structure of the NCNTs. Figure shows a Raman spectrum of nitrogen-functionalized CNTs compared with a representative Raman spectrum after oxidative MnO deposition under reflux conditions. The spectrum of the NCNTs contains bands located at 1577 cm–1 (G-band) and 1334 cm–1 (D-band) originating from the Raman-active in-plane atomic displacement E2g mode and disorder-induced features of the CNTs, respectively.[45] Additionally, a band located at 2681 cm–1 is the overtone of the D band (2D band). The Raman spectrum of the MnO/NCNT sample shows an upward shift of the G, D, and 2D bands to 1346, 1585, and 2692 cm–1, indicating that MnO2 is covalently bound to the NCNTs. The upward shift was monitored for all samples after MnO deposition.

Figure 3

Representative Raman spectra of NCNTs (dashed line) and MnO/NCNT (bold line).

Because the XRD patterns exhibited low crystallinity of the manganese oxides, Raman spectra were recorded to provide further structural information. Figure displays the high-resolution Raman spectra in the range of 300 to 850 cm–1. Figure a shows the Raman spectra of the samples synthesized under reflux conditions with the KMnO4/NCNT ratio varying from 1:2 to 2:1 and a sample synthesized with CsMnO4 at a ratio of 1:1. The spectra show three Raman bands located at 630, 571, and 495 cm–1. According to the literature, the band at 630 cm–1 can be assigned to the symmetric stretching vibration of Mn–O and the band at 570 cm–1 to the stretching vibration in the basal plane of MnO6 groups in birnessite-type MnO2.[46]

Figure 4

Raman spectra of MnO/NCNT samples, comparing (a) synthesis under reflux conditions using different K(Cs)MnO4/NCNT ratios, (b) HT synthesis (110 °C; 8 h) with different K(Cs)MnO4/NCNT ratios, (c) HT synthesis at different temperatures 110 °C (for 8 and 24 h), 140 °C (8 h), and 180 °C (8 h) for a fixed ratio of 1:1, and (d) HT synthesis with cesium or potassium at 110 and 140 °C (8 h).

Figure b depicts the Raman spectra of the samples with different KMnO4/NCNT ratios synthesized under HT conditions at 110 °C for 8 h. Clearly, the samples with a higher Mn content show typical bands at 637, 573, and 500 cm–1 related to birnessite-type MnO2. The sample with a 1:2 ratio shows multiple peaks between 480 and 700 cm–1 with superimposing bands, suggesting the presence of at least two manganese oxide phases. Birnessite-type bands are located at 630, 573, and 485 cm–1, while γ-MnOOH bands are positioned at 530 and 558 cm–1.[46,47] The band located at 659 cm–1 is attributed to traces of amorphous manganese oxide (AMO)[48] or Mn3O4 impurities.[49]Figure c shows the Raman spectra of the samples derived from HT synthesis at different temperatures and durations. It shows that upon heating to 180 °C or increasing duration of the treatment at 110 °C to 24 h, three bands in the high wavenumber region at 624, 558, and 524 cm–1 and two bands in the lower region at 386 and 358 cm–1 emerge. The bands can be assigned to γ-MnOOH, in good agreement with the literature results.[47] For the samples treated at 180 °C, a band attributed to traces of Mn3O4 emerges at 659 cm–1.

By using CsMnO4 for the HT deposition at 110 and 140 °C, similar inferences as those for the XRD observations in Figure d can be made for the KMnO4-derived samples. The Raman spectra for the samples treated at 110 °C show the three typical bands of birnessite-type manganese oxide, while higher temperatures led to the formation of γ-MnOOH. A small band at 660 cm–1 is also noticeable in the 140 °C sample, which is assigned to traces of Mn3O4.

Surface Characterization

X-ray Photoelectron Spectroscopy (XPS)

XPS was used to investigate the surface chemistry of the samples after oxidative deposition of MnO on NCNTs. Figure a–d shows the Mn 3s region with the characteristic multiplet splitting with one peak centered at 89.1 eV for all the samples.

Figure 5

Mn 3s XP spectra of MnO/NCNT samples, comparing (a) different K(Cs)MnO4/NCNT ratios synthesized under reflux conditions, (b) HT conditions (110 °C; 8 h) with different K(Cs)MnO4/NCNT ratios, (c) HT conditions applying 110 °C (for 8 and 24 h), 140 °C (8 h), and 180 °C (8 h) at a fixed ratio of 1:1, and (d) HT conditions with cesium or potassium at 110 and 140 °C (8 h).

The energy separation of the multiplet splitting was used to estimate the average surface oxidation state. The samples synthesized under reflux and HT conditions at 110 °C for 8 h with different ratios show a splitting value of 5.1 eV (see Figure a,b), indicating an +4/+3 mixed oxidation state, as it has been reported for birnessite-type MnO2.[50] By increasing the temperature to 140 °C/180 °C and the treatment time at 110 °C to 24 h, the splitting value increased to 5.2 and 5.7 eV (Figure b,c), respectively, for the KMnO4-derived samples. The increase indicates a surface reduction from the +4 to the +3 oxidation state. The CsMnO4-derived sample treated at 140 °C shows a splitting of 5.6 eV, suggesting the presence of Mn3O4.

Figure displays the C 1s high-resolution spectra of the samples treated hydrothermally at different temperatures and durations.

Figure 6

XPS C 1s region of the samples treated under HT conditions at 110, 140, and 180 °C for 8 and 24 h.

The C 1s profile consists of a main peak at 284.5 eV originating from sp2-hybridized graphitic carbon. Additionally, a shoulder arises on the lower binding energy side upon increasing the treatment time to 24 h at 110 °C. By increasing the temperature to 140 °C, the shoulder gets more pronounced, leading to a distinct peak at 282.5 eV. The shoulder and the peak at 282.5 eV suggest the formation of manganese carbide.

Transmission Electron Microscopy (TEM)

TEM and scanning TEM (STEM) were applied to investigate the morphology of the NCNTs and the MnO species after oxidative deposition under reflux and HT conditions.

Figure a,b shows the STEM image and the corresponding EDS elemental mapping of manganese (green) and carbon (red), clearly revealing that the multiwalled CNTs remain intact with MnO nanosheets randomly and to some extent nonuniformly distributed on them. Figure c shows the TEM image of the samples synthesized under reflux conditions with a 1:1 ratio of KMnO4/NCNT. Small and very thin crystalline MnO nanosheets at the outer surface of the tubes can clearly be seen. The sample synthesized with a higher KMnO4/NCNT ratio of 2:1 (Figure d) shows NCNTs with a higher surface coverage of MnO nanoparticles having both crystalline and amorphous properties.

Figure 7

STEM (a) and elemental mapping (b) of the MnO/NCNT samples treated under HT conditions at 110 °C for 8 h with KMnO4 at a 1:1 (KMnO4/NCNT) ratio. TEM (c,d) images of the samples synthesized under reflux conditions with a KMnO4/NCNT ratio of 1:1 (c) and 2:1 (d).

Electrochemical Characterization

Oxygen Evolution Reaction

Figure shows iR-corrected linear sweep voltammograms (LSVs) of the MnO/NCNT samples during oxygen evolution in 1.0 M KOH. To compare the activity, the potentials required to reach a current density of i = 10 mA cm–2 are summarized in Table S1. When comparing the OER performance of the samples synthesized under reflux conditions with the bare NCNT support, a decrease of the required potential from 1.70 to 1.64 V is observed. Surprisingly, no evident activity trend was found as a function of the KMnO4/NCNT ratio in spite of manganese loadings ranging from 9.8 to 23.3 wt %, suggesting that an optimal manganese oxide distribution is necessary to achieve an enhanced electrocatalytic performance. When using CsMnO4, a further decrease of overpotential down to 1.61 V is observed. The OER overpotential at 10 mA cm–2 for hydrothermally treated samples depended on the applied temperature and treatment time. A decrease of the potential required to reach 10 mA cm–2 was also observed for the hydrothermally treated sample compared with blank NCNTs. Among the hydrothermally treated samples, the sample treated for 8 h at 110 °C with the highest manganese loading exhibits a slightly inferior performance, with 1.63 V at 10 mA cm–2. By decreasing the KMnO4/NCNT ratio to 1:2, the potential decreases to 1.62 V. An increase of the treatment time to 24 h at 110 °C and the temperature to 140 °C for 8 h led to a further decrease to 1.61 and 1.60 V, respectively. The samples treated with CsMnO4 at 140 °C and KMnO4 at 180 °C reached overpotentials of 1.63 and 1.68 V, respectively.

Figure 8

Linear sweep OER voltammograms in O2-saturated 1.0 M KOH at a scan rate of 5 mV s–1 and a rotation speed of 1600 rpm using (a) different K(Cs)MnO4/NCNT ratios synthesized under reflux conditions, (b) HT conditions (110 °C; 8 h) with different K(Cs)MnO4/NCNT ratios, (c) HT conditions applying 110 °C (for 8 and 24 h), 140 °C (8 h), and 180 °C (8 h) at a fixed ratio of 1:1, and (d) HT conditions with CsMnO4 at 110 and 140 °C (8 h).

To explore the kinetics of the OER, Tafel analysis was performed. Figure shows the Tafel plots with the respective slopes of four selected samples derived from the linear sweep voltammograms. The bare NCNTs exhibited the highest slope (164 mV dec–1), which decreased to 91 and 99 mV dec–1 after deposition under reflux conditions using KMnO4 and CsMnO4, respectively. Lower Tafel slopes of 74 and 75 mV dec–1 were obtained after applying HT conditions, indicating an improvement of the OER kinetics. The barely visible semicircle in the Nyquist plot of the electrochemical impedance spectroscopy (EIS) measurements (Figure S1) indicates that the conductivity of the samples is very high. Rotating ring disk electrode (RRDE) measurements were performed for the (K)MnO/NCNT-reflux sample with a 1:1 KMnO4/NCNT ratio to determine the faradaic efficiency toward the OER. First, an LSV (Figure S2a) was recorded while simultaneously monitoring the ring and disk currents. As can be seen in Figure S2a, in the potential region between 1.30 and 1.55 V, a current is measured at the disk, while no O2 is detected at the ring. In the higher potential regime (above 1.55 V), an oxygen reduction current is detected at the ring, which increases with the disk current, indicating that the OER occurs at the disk. A potential of 1.634 V versus RHE, corresponding to a ring current of −0.5 mA cm–2, was selected for determining the faradaic efficiency because, as seen in Figure S2a, it is high enough to drive the OER, but not too high to cause turbulences due to bubble formation, thus allowing collection of reliable faradaic efficiency data. Figure S2b shows a chronoamperometric measurement, where the ring current, the disk current, and the Faraday efficiency are juxtaposed as a function of time, while potentials of 1.184 and 1.634 V versus RHE were applied at the disk for 5 and 10 min, respectively, and a constant potential of 0.234 V was held at the ring. Upon increasing the disk potential from 1.184 to 1.634 V versus RHE, a constant current of −0.5 mA cm–2 was monitored over 10 min at the ring. The disk current levels off within the first 5 min, reaching a plateau, which corresponds to a current density of approximately 1.84 mA cm–2. The faradaic efficiency increased from 75 to 90% after 10 min. The initial high disk current and low faradaic efficiency are ascribed to the oxidation of manganese to higher oxidation states, which, after being accomplished almost completely within 10 min, leads to a higher faradaic efficiency.

Figure 9

Tafel plots of bare NCNTs (black) and four selected samples with a MnO4–/CNT ratio of 1:1 synthesized under reflux conditions with KMnO4 (red) and CsMnO4 (green), and under HT conditions with KMnO4 at 110 °C (blue) and 140 °C (cyan) for 8 h.

Discussion

Depending on the sequence of washing and oxidation, differences in the electrocatalytic behavior of the manganese oxide-supported NCNTs were observed. The as-received CNTs display a low activity because of hydrophobic amorphous carbon present on the surface. Washing the as-received CNTs prior to oxidation led to an improved performance, as the amorphous carbon was removed. The further decrease of the potential to 1.60 V after oxidation indicates that the removal of the residual growth catalysts was not effective. Surface oxidation of the CNTs by HNO3 vapor removes amorphous carbon layers and exposes the residual growth catalyst, which then affects the OER. Washing in 1.5 M HNO3 after oxidation results in the dissolution of the residual growth catalyst and therefore leads to an increase of the OER potential at 10 mA cm–2 to 1.70 V. The decrease from 1800 to 400 ppm iron and from 900 to 100 ppm cobalt detected by elemental analysis shows that it is more effective to remove the residual growth catalysts by first applying HNO3 vapor oxidation followed by washing in HNO3. The residual 100 ppm cobalt is assumed to be fully embedded in the NCNTs and hence does not participate in the OER. The contribution of iron to the measured OER can be excluded because it is known to be an inefficient OER catalyst.[51] NCNTs were used as a substrate because previous studies had shown that N-related defect sites are chemically more active and, therefore, interacting more strongly with precursor molecules.[52−54] A detailed deconvolution of the high-resolution XP O 1s and N 1s spectra before and after nitrogen functionalization is shown in Figure S3.

The bulk structure of the support and the deposited MnO species were investigated by XRD and Raman spectroscopy. The shift of the D, G, and 2D Raman bands to slightly higher wave numbers indicates that the manganese oxide species are covalently bound to the NCNTs as a result of the oxidative deposition introducing surface defects in the carbon structure (eq ). The STEM and TEM images show that the MnO nanosheets are evenly distributed on the NCNTs at 16.2 wt % Mn loading and suggest an increase of the layer thickness for a Mn loading of 23.3 wt %. Table S2 lists the surface concentration of the elements derived from the XPS regions, where birnessite-type MnO2 nanosheet formation was observed. The hydrothermally derived samples generally exhibited higher manganese surface concentrations, indicating a higher dispersion. However, the manganese surface concentration does not increase linearly with loading, indicating that a higher loading led to an increase of the MnO nanosheet thickness rather than a higher coverage on the NCNTs. The XRD and XPS results identify birnessite-type MnO2 containing Mn4+/3+ for the deposition under reflux and HT conditions at 110 °C for 8 h. Temperature increase during HT deposition led to partial reduction from the 4+ to the 3+ oxidation state. The XRD reflections of the reflux-derived samples are rather broad, indicating low crystallinity. The reflection in the low 2θ region can be utilized to estimate the interlayer distance between the birnessite sheets, which ranges from 7.8 to 9.5 Å in the KMnO4-derived reflux samples. Compared with the unsupported K-birnessite, which has an interlayer distance of 7.03 Å according to JPCDS database 42-1317, the KMnO4-derived samples exhibited comparatively larger interlayer distances, probably because of a higher amount of hydrate water between the layers. The CsMnO4-derived sample shows a larger distance of about 10.5 Å owing to the intercalation of Cs ions between the birnessite sheets. The hydrothermally deposited samples with a birnessite structure exhibit a sharper reflection around 12.2°, which corresponds to an interlayer distance of 7.2 Å. The CsMnO4 sample synthesized under HT conditions at 110 °C again shows a higher distance of 9.5 Å.

Raman spectroscopy demonstrates an improved signal-to-noise ratio for the hydrothermally derived samples compared with the reflux-derived samples. Specifically, the intensity of the band at 575 cm–1 is more pronounced for the hydrothermally treated samples. Because this band relates to the layered structure of birnessite, structural degradation during the reflux treatment could have taken place.[55] This observation further confirms that the reflux-derived samples have a lower degree of crystallinity. The Raman spectra of the HT treated samples at 180 °C, at 110 °C with a KMnO4 to NCNT ratio of 1:2, and the 140 °C CsMnO4-derived sample show a band located at 660 cm–1. Both amorphous MnO and Mn3O4 exhibit a band at this wavenumber. For the KMnO4 1:2 sample, it is likely that AMO is formed when using lower Mn loadings on the NCNT support because lower loadings favor the formation of AMO.[41] However, the possibility that MnO2 is partially reduced to Mn3O4 (which is more Raman-active) during spectra acquisition cannot be ruled out. In contrast, for the CsMnO4-derived samples, Mn3O4 is the likely species because the XRD results (Figure d) contain reflections at 36.0° and 32.5° 2θ corresponding to Mn3O4.

The different deposition techniques did not affect the structure of the CNT support. The features and intensities of the D and G bands remain essentially unchanged after MnO deposition under reflux and HT conditions (Figure S4). The samples treated under HT conditions at a temperature of 140 and 180 °C or 24 h at 110 °C cannot be considered as homogenous depositions of MnO on the NCNTs because large discrepancies were observed in the intensities of the D and G bands depending on the illuminated spot (see Figure S5). Further, Figure S6 shows the presence of needle-like γ-MnOOH crystals and NCNTs after the HT treatment at 140 °C. Additional evidence for minor degradation of the NCNT support associated with the reduction of Mn4+/3+O2 to γ-Mn3+OOH is indicated by the increase in the Mn content to 17.1 and 19.0 wt % after HT treatment for 24 h at 110 °C and for 8 h at 180 °C, respectively.

Comparing the Raman spectroscopy, XRD, and the XPS results leads to the conclusion that the deposition under reflux and HT conditions at 110 °C for 8 h resulted mainly in the formation of birnessite-type MnO2 with a mixed-valent surface oxidation state of +4/+3. Increase of the temperature and duration of the treatment leads to partial reduction of the bulk and the surface species from δ-Mn4+/3+O2 to γ-Mn3+OOH and Mn33+/2+O4 (see eq ), as indicated by the XRD reflections of γ-MnOOH and Mn3O4 and the Mn 3s multiplet splitting of 5.2 and 5.7 eV, respectively.

The OER results show an improvement because of the oxidative deposition under reflux conditions manifested by a decrease of the OER potential at a current density of 10 mA cm–2 from 1.70 to 1.64 V with respect to the bare NCNTs. The potential necessary to reach a current density of 10 mA cm–2 further decreased to 1.61 V when using CsMnO4. The higher activity is presumably caused by the higher interlayer distance of the birnessite layers, as reported in the literature.[56] The same trend is observed for the hydrothermally derived samples using CsMnO4. For the hydrothermally derived samples, a weak correlation between the OER activity and low manganese loadings was observed. It is expected that the lower Mn loading favors the formation of γ-Mn3+OOH, as revealed by the XRD patterns and the Raman spectroscopy data, which is beneficial for electrocatalysis of the OER. Dismukes et al.[47] reported that γ-MnOOH is a better performing OER catalyst than layered MnO. This observation is further supported by the high OER activity of the samples formed by HT treatment for 24 h at 110 and 140 °C for 8 h, where γ-MnOOH formation was observed. These samples afforded 10 mA cm–2 at low potentials of 1.60 and 1.61 V, among the lowest attained, and exhibited lower Tafel slopes compared to the KMnO4 reflux-derived samples. HT treatment at 180 °C led to further reduction to Mn3O4 as well the formation of manganese carbide, accompanied by the increase of the OER potential (at 10 mA cm–2) to 1.68 V. Thus, oxidative deposition under reflux conditions (100 °C) using CsMnO4 and mild HT treatment (110–140 °C) using KMnO4, and low manganese loadings in both cases, resulted in the most active MnO/NCNT catalysts for electrochemical water oxidation among the studied samples. The Faraday efficiency increased with time during polarization of the disk at 1.634 V reaching 90% after 10 min, indicating that within the first minutes, the disk current has a significant contribution from the oxidation of manganese to higher oxidation states. The long-term stability of MnO/NCNT obtained by HT treatment with a 1:1 (KMnO4/NCNT) ratio at 140 °C was investigated chronoamperometrically at a potential of 1.65 V versus RHE for 2 h (see Figure S7). The current decreased continuously with a higher initial slope, dropping from 9 mA cm–2 to below 4 mA cm–2 within the first 30 min, reaching 1 mA cm–2 after 2 h, which may be caused by catalyst detachment.

Further studies employing operando X-ray absorption spectroscopy are in progress to elucidate structural changes of the deposited MnO nanosheets under applied potential.

Conclusions

The oxidative deposition of manganese oxide on nitrogen-doped CNTs under reflux and HT conditions using KMnO4 and CsMnO4 was applied to obtain MnO/NCNT composites for the electrocatalytic OER. The deposition under reflux conditions led to the formation of birnessite-type MnO2 with a partly amorphous character covalently bound on the NCNTs. TEM and STEM showed that the manganese oxide nanosheets were evenly distributed on the NCNTs. The optimized MnO/NCNT catalyst formed by reflux treatment afforded a current density of 10 mA cm–2 at a 60 mV lower overpotential (1.64 V) compared to the purified NCNTs (1.70 V), and exhibited a significantly lower Tafel slope of 91 mV dec–1 in relation to 164 mV s–1 for the NCNTs. The optimal hydrothermally treated sample required 100 mV lower potential to achieve the same current density and had a lower Tafel slope of 75 mV dec–1. XRD, XPS, and Raman results showed that bulk and surface reduction from birnessite MnO2 to γ-MnOOH took place when the temperature or duration time was increased beyond 110 °C or 8 h, respectively, during the HT synthesis. The oxidative deposition under reflux conditions using CsMnO4 along with mild HT treatment using KMnO4 and low manganese loadings in both cases were identified as the most suitable synthetic routes to obtain highly active MnO/NCNT catalysts for electrochemical water oxidation.

Experimental Section

Synthesis of CsMnO4

KMnO4 (1.0 g, Sigma-Aldrich) and 1.02 g of CsCl (Sigma-Aldrich) were dissolved in 150 mL of deionized water and heated to 65 °C. After cooling, the solution was kept at room temperature overnight. Green/bronze-colored crystals (0.95 g) were separated from the solution by filtering. The crystals were recrystallized once by dissolving 0.95 g of CsMnO4 in 80 mL of deionized water, heating to 65 °C and subsequent cooling to room temperature overnight. After filtering, 0.68 g of CsMnO4 crystals was obtained. XRD reflections (Figure S8) agree well with those reported for CsMnO4 (ICSD: 01-070-3167).

Sample Preparation

CNTs with an average diameter of 9 nm and a length of 1.5 μm were obtained from NANOCYL SA (Sambreville, Belgium). The as-received CNTs (2 g) were treated with HNO3 (65%, Sigma-Aldrich) vapor for 48 h at 200 °C and then dried overnight at 80 °C to obtain oxygen-functionalized CNTs (OCNTs).[57] The OCNTs were purified by washing in 1.5 M HNO3 under continuous stirring for 72 h at room temperature. The suspensions were filtered, washed with deionized water to a pH of 7, and dried overnight at 80 °C to obtain purified OCNTs. In order to produce nitrogen-functionalized CNTs (NCNTs), 500 mg of the OCNTs was thermally treated for 6 h at 400 °C with 10% NH3 in He (99.999% purity, Air Liquide) at a flow rate of 50 sccm.

For the oxidative deposition under reflux conditions, 100 mg of NCNTs was mixed with KMnO4 (50, 100, or 200 mg) in 30 mL of water in a 50 mL round bottom flask. CsMnO4 (160 mg) was used in order to have the same manganese loading as compared to the 100 mg KMnO4 sample. The mixture was stirred for 5 min and ultrasonicated for 5 min. Afterward, the mixture was heated to 100 °C for 1 h and stirred for additional 20 h at room temperature. Finally, the product was filtered, washed several times with water, and dried at 80 °C overnight. The HT deposition was performed in the same way, except using a 50 mL Teflon-lined stainless steel autoclave. The autoclave was then put in an oven and heated to 110, 140, or 180 °C for 8 h and later for 24 h at 110 °C.

Characterization

Raman spectroscopy was performed using a Jobin Yvon iHR550 (HORIBA) spectrometer equipped with a 532 nm laser source (Ventus 532, Laser Quantum). A laser power of 1 mW was used for all the measurements. Gratings of 600 and 1800 grooves/mm were used for low and high resolution, respectively. XRD patterns were recorded using a PANalytical theta–theta powder diffractometer with Cu Kα irradiation in the 2θ range of 10°–80°. The obtained XRD patterns were analyzed by the PANalytical X’Pert HighScore Plus v.3.0 software. XPS measurements were carried out in an ultrahigh vacuum setup equipped with a high-resolution Gammadata-Scienta SES 2002 analyzer. A monochromatic Al Kα X-ray source (1486.3 eV; anode operating at 14.5 eV and 30.5 mA) was used as incident radiation and a pass energy of 200 eV was chosen, resulting in an energy resolution better than 0.5 eV. Charging effects were compensated using a flood gun. Binding energies were calibrated by positioning the main C 1s peak at 284.5 eV. TEM was performed using a JEM-2800 (JEOL) transmission electron microscope operated at 200 kV. Elemental analysis was performed by atomic absorption spectroscopy using a Varian AA 300 spectrometer equipped with a flame ionization system.

The electrocatalytic tests were performed in a conventional three-electrode cell controlled by an Autolab potentiostat/galvanostat (PGSTAT128 N) in combination with a Metrohm RDE-2 rotator and its control unit. The measurements were performed using the control software Nova 1.10. A Hg/HgO electrode and a platinum wire served as the reference electrode (RE) and the counter electrode (CE), respectively. Polished glassy carbon (GC) disc electrodes (diameter: 3.8 mm) embedded in a Teflon cylinder were employed as working electrodes (WEs). The catalyst suspension was prepared by ultrasonically dispersing the catalyst (5.0 mg) in a mixture of water (490 mL), ethanol (490 mL), and Nafion (5%, 20 mL). The resulting catalyst suspension (5.0 μL) was drop-cast onto a polished GC electrode to obtain a catalyst loading of 210 μg cm–2. The modified electrode was dried in air at room temperature for at least 30 min. Prior to measurements, the electrolyte (1 M KOH) was purged with pure O2 for 20 min to reach O2 saturation. Each measurement started with cyclic voltammetry performed between 1.0 and 1.5 V versus RHE with a scan rate of 0.1 V s–1. Twenty cycles were applied as conditioning to achieve reproducible voltammograms. To determine the solution resistance, EIS measurements were performed at open circuit potential. The spectra were recorded between 50 kHz and 10 Hz with a 10 mV amplitude. Linear sweep voltammetry was performed at a scan rate of 5 mV s–1. Long-term stability measurements were performed chronoamperometrically by maintaining the potential at 1.65 V versus RHE for 2 h. To mitigate the inhibition by O2 bubbles during the OER, the electrode was rotated at 1600 rpm.

Faradaic efficiency measurements were performed by means of RRDE voltammetry in a four-electrode cell configuration, where the disk, GC (A = 0.1963 cm2), modified with 8.31 μL catalyst ink, served as WE 1, and a concentric platinum ring (A = 0.1532 cm2) was WE 2. A Pt mesh and a Hg/HgO electrode were used as the CE and the RE, respectively. The measurements were conducted in Ar-saturated 1 M KOH solution. An LSV was recorded at the modified disk electrode from 0.934 to 1.684 V versus RHE at a scan rate of 5 mV s–1 and 1600 rpm rotation speed; meanwhile, a constant potential of 0.234 V versus RHE was applied at the ring to reduce the produced O2 at a diffusion-limited rate. From the resulting LSV, the disk potential at which a current of −0.5 mA cm–2 was obtained at the ring was determined. Subsequently, a chronoamperometric protocol was used consisting of, firstly, applying a constant potential of 1.184 V for 5 min on the disk, and secondly, stepping the previously determined potential for 10 min, while maintaining a constant potential of 0.234 V at the ring. The faradaic efficiency was determined according to % FE = (iring/(idisk·N))·100; where iring and idisk are the currents measured at the ring and at the disk, respectively, and N is the collection efficiency factor. Potassium hexacyanoferrate(III) (5 mM, K3[Fe(CN)6]) dissolved in 1 M KOH was used to determine N by reducing Fe(III) to Fe(II) at the disk and oxidizing reduced Fe(II) back to Fe(III) at the ring. N was determined from the relation N = iring/idisk and had a value of 24% for the investigated catalyst film.