On the role of hydroxide species in sulphur- and nitrogen-doped cobalt-based carbon catalysts for the oxygen evolution reaction

Ali Shahraei ab, Markus Kuebler a, Ioanna Martinaiou ab, K. Alexander Creutz a, W. David Z. Wallace a, Mohammad A. Nowroozi c, Stephen Paul ab, Natascha Weidler a, Robert W. Stark d, Oliver Clemens c and Ulrike I. Kramm *ab
aCatalysts and Electrocatalysts Group, Department of Materials- and Earth Sciences, Department of Chemistry, Otto-Berndt-Str. 3, 64287 Darmstadt, Germany. E-mail: kramm@ese.tu-darmstadt.de
bGraduate School of Excellence Energy Science and Engineering, Otto-Berndt-Str. 3, 64287 Darmstadt, Germany
cMaterialdesign durch Synthese Group, Department of Materials- and Earth Sciences, Alarich-Weiss-Str. 2, 64287 Darmstadt, Germany
dPhysics of Surfaces Group, Department of Materials- and Earth Sciences, Alarich-Weiss-Str. 16, 64287 Darmstadt, Germany

Received 16th June 2018 , Accepted 3rd October 2018

First published on 19th October 2018


The influence of high S/Co ratios on the structural composition and oxygen evolution reaction (OER) activity of a group of cobalt-based carbon catalysts was investigated. Catalysts were prepared from polyaniline, cobalt acetate and dicyandiamide as precursors for active site formation and as structure forming agents. The sulphur to cobalt ratio was investigated in a range of S/Co = 10 to 32. On the basis of a comprehensive structural characterisation by XRD, Raman, XPS, TEM and N2 sorption measurements it was possible to show that the S/Co ratio has a significant impact on the carbon morphology. In fact, with increasing S/Co ratio the carbon morphology continuously changes from highly amorphous carbon to carbon-nanotubes, with increasing diameter. Besides the anticipated CoN4 sites and cobalt sulphite species, the catalysts also contained cobalt nanoparticles as well as cobalt hydroxide species. The most active catalyst required 0.37 ± 0.01 V overpotential to reach 10 mA cm−2 and even increased in activity during galvanostatic treatment and cycling-illustrating its very good performance. A faradaic efficiency of >35% was determined. A detailed analysis of the activity and stability in combination with Raman and XPS provides two explanations for observed Tafel slope changes, that might also be coupled to each other, namely a change in the carbon oxidation rate depending on preparation and potential or a variation in the coverage by hydroxide and oxidic species of the metal, whereas hydroxide species seem to enable a higher OER activity.


Introduction

Water electrolysis to produce hydrogen is among the most promising technologies considering the need for appropriate storage technologies for renewable energies.1 This is rather important as for future society we have to move from energy provided by fossil fuels (and the harmful effect of greenhouse gases) to renewable energies. Since renewable energy sources are not available on a continuous basis, appropriate storage technologies have to be developed.

In water electrolyzers, the state-of-the art materials are often precious metals. These systems achieve high conversion rates at relatively low loadings. However, precious metals such as platinum, ruthenium and iridium (all typical catalysts for water splitting reactions) were classified as critical raw materials by the European Commission2 due to their low availability and high costs. Non-precious metal catalysts therefore play a crucial role in future energy supply.3 Especially, the group of Me–N–C catalysts with catalytically active MeN4 sites have received much attention as they are active for the oxygen reduction reaction (ORR),4–9 CO2 reduction,10–13 and the hydrogen evolution reaction (HER)14–18 as well as the oxygen evolution reaction (OER).19–24 Me–N–C catalysts that are highly active for the ORR and OER are also promising candidates as bifunctional catalysts for metal air batteries.25,26 Regarding their synthesis, Me–N–C catalysts can be prepared by pyrolysing a precursor mixture consisting of carbon, nitrogen and metal sources. Most of the preparation routes that result in highly active electrocatalysts also contain inorganic species like carbides, nitrides or metallic nanoparticles that can contribute to the overall performance of the resulting catalyst in various electrocatalytic reactions.

In terms of structural composition and activity, sulphur addition was found beneficial for Me–N–C catalysts.14,27–29 However, if the sulphur content (or sulphur to metal ratio) becomes too large, metal sulphides are formed.14,30–32 Recently, Qiao and his/her team prepared a bifunctional catalyst of cobalt sulphide nanoparticles embedded in nitrogen and sulphur co-doped graphene activating both the ORR and OER.25 Also others attributed catalytic activity towards the OER to transition metal sulphides.25,33–37

Therefore, in the context of this work, sulphur addition might result in OER active metal sulphides and/or enable a variation of the CoN4 contribution.

Nevertheless, during the last few decades, oxidic surfaces, as found in transition metal oxides/hydroxides such as perovskites,38 cobalt oxides39–41 and nickel-based hydroxides,42–45 have been known as the most promising OER catalysts. Weidler et al. studied various cobalt oxides under OER conditions and reported a transition of an oxide surface to oxide-hydroxide sites as active species that were formed on the surface of the catalysts.41 During the first twelve minutes of operation, a continuous improvement of the OER activity was observed. Afterwards no further change in activity was found. Step-by-step investigation of the as-prepared and “aged” thin film electrodes by X-ray photoelectron spectroscopy (XPS) showed that induced by operation, the contribution of cobalt oxide-hydroxide CoOx(OH)2 increased.41 Indeed, the potential induced formation of oxyhydroxides could not only be shown for oxides and sulphides but also for CoN4 centres in non-pyrolyzed cobalt porphyrin that were active for the OER.22

In particular, these described effects as observed by Weidler et al. and Daniel et al. make it questionable to what extent the OER activity in Co–N–C catalysts can be attributed to CoN4 species (or Co sulphides, in the case of S-addition).

Inspired by the very promising OER activities observed for Co–N–C and cobalt sulphides embedded in carbon we aimed to investigate whether the OER can indeed be assigned to one of the two active sites (CoN4 and/or Co sulphide) or if also here oxidic species are the true catalytic sites for the OER. Thus, we prepared samples of cobalt-based carbon catalysts with ultrahigh amounts of sulphur and nitrogen in the precursors. The catalysts were thoroughly characterized in terms of morphology, chemical composition and OER activity and stability. We found that cobalt hydroxide species improved the OER activity of the catalysts. This result gives an important indication for the development of more active and stable non-precious metal catalysts by a rational design of the synthesis.

Results and discussion

Structural characterisation of the Co–N–C catalysts

In Fig. 1 a scheme of the synthesis is shown. Similar to our previous work,14 PANIevap was used as the main nitrogen precursor.
image file: c8ta05769a-f1.tif
Fig. 1 Schematic of the synthesis of sulphur-doped Co–N–C catalyst.

PANIevap was prepared by the oxidative polymerisation of aniline in the presence of ammonium peroxidisulphate (APS). The precipitate was dried without the application of washing and filtering procedures, and thus the final product was highly acidic and contained polyaniline as well as the residuals from APS. Therefore, it is labelled PANIevap. Following this procedure, PANIevap is a carbon, nitrogen, sulphur and oxygen source within the synthesis. Catalysts prepared with PANI obtained very good stability in fuel cell operation46 and the carbon formed during the synthesis might therefore also be of stronger corrosion resistance as required for the OER. With the given cobalt content in the precursor (3 wt%), even without addition of elemental sulphur (Selemental), the molar S/Co ratio equaled ten (S/Co = 10).

In order to evaluate the impact of Selemental addition during the synthesis on the structure and morphology of the catalysts, N2 sorption measurements and transmission electron microscopy (TEM) images were obtained as shown in Fig. 2.


image file: c8ta05769a-f2.tif
Fig. 2 (a) Results of BET and meso- and micropore surface area (SA) obtained from N2 sorption measurements. In (b) exemplary TEM images of the catalysts are shown, the scale bar is 100 nm.

It is evident from the N2 sorption measurements that the addition of Selemental leads to a decrease in BET surface area (SA) while on the other hand, the micropore SA tended to increase. In ESI Fig. S1, the cyclic voltammograms of the catalysts are shown as well as a correlation of the double layer capacity (taken from the CVs at 0.5 V) and the overall BET surface area.

The change in specific surface area was associated with a change in the carbon morphology from amorphous carbon to more defined carbon as visible from TEM images in Fig. 2b and in ESI Fig. S2.

In fact, for the catalysts with S/Co ratios of 15, 24 and 32 the formation of carbon nanotubes was identified. It seems that higher S/Co ratios led to more and larger nanostructures. This is in agreement with previous reports on a positive effect of sulphur-addition on the growth of carbon nanofibers.47,48

Also visible in the TEM images (see Fig. 2b) is the presence of nanoparticles, which seem to be surrounded by multiple layers of carbon, as also distinctly visible in Fig. S2. Most likely, these shells also acted as the protectant of the metal during the acid-leaching step after the first heat-treatment. The presence of these nanoparticles was also confirmed by X-ray diffraction (XRD), shown in Fig. 3a. The refinement of the diffraction data identified cobalt nanoparticles in all of our catalysts together with smaller amounts of cobalt sulphide species.


image file: c8ta05769a-f3.tif
Fig. 3 (a) X-ray diffraction data including refinement of the four investigated catalysts. In (b) Raman spectra are shown in the range <1000 cm−1. As the inset, the 1st order region assigned to carbon is shown.

However, in the case of cat A Co3S4 was formed, all other catalysts contained Co9S8. This is interesting, as from the S/Co ratio in the precursor, a higher sulphur content would have been expected for cat B to D.

As described in the Introduction, both Co3S4 and Co9S8 have been reported as catalysts for the oxygen evolution reaction (OER) and hydrogen evolution reaction (HER).33,36,49

The Raman spectra for the range <1000 cm−1 of the catalysts are displayed in Fig. 3b. There are four well-pronounced bands at 463 cm−1, 507 cm−1, 601 cm−1 and 664 cm−1 and a weakly pronounced band at about 200 cm−1 that were assigned to metallic cobalt and/or cobalt oxide particles.14,50 A comparison of the observed bands with model systems reported in the literature suggests the presence of CoOx and Co3O4 species.51,52 However, the small intensity of the band at 200 cm−1 and the results concluded from X-ray photoelectron spectroscopy (XPS) (see below) indicate that the presence of CoOx is more likely, while it seems that Co3O4 can be excluded.

As TEM indicated some protective carbon layers around the nanoparticles and also the absence of reflections which could be assigned to cobalt oxides/hydroxides in XRD, we assume that these were small particles with a lower degree of order and thus X-ray amorphous. Indication of partial oxidation of the cobalt on the surface was also given by X-ray photoelectron spectroscopy (XPS, see later in this section).

In ESI Fig. S3, the first order region related to carbon blacks is deconvoluted into four bands, namely, the G band (ca. 1585 cm−1), D band (1355 cm−1), D3 band (ca. 1500 cm−1) and the D4 band (ca. 1200 cm−1). Following the same order, they are assigned to vibrations within the graphene plane, vibrations at the edges or at curvations, vibrations induced by heteroatoms and vibrations of lower hydrocarbons.53–55 The ratio of ID/IG is inversely proportional to the graphene layer extension La,53 while in our previous publications we found indications that MeN4 centres as well as pyridinic nitrogen contributed to the D3 band intensity.56 A tendency for higher ID3/IG ratios is found for catalysts with mir NMeN and Npyrid, as is shown in ESI Fig. S4.

Influence of sulphur modification on the composition

In order to get additional insights into the near-surface structure and composition of the catalysts, XPS was performed (Fig. 4). In Fig. 4a the survey scan of cat D is shown. Similar to all other catalysts, cobalt, oxygen, nitrogen, carbon and sulphur were detected. The respective peaks are indicated in the survey spectrum. For further analysis, the fine scans of the different regions were analysed. In Fig. 4b–e the Co 2p 3/2, O 1s, N 1s and S 2p regions are shown for all four catalysts. The shape of the curves in the Co 2p, N 1s and O 1s region seemed very similar for all catalysts. In ESI Fig. S5 the overall Co 2p energy range is shown as well as measurements of reference samples. Furthermore, in Fig. S6 the fit of the Co 2p 3/2 region of the most active catalyst of this study (cat D) is exemplarily shown. A good fit was obtained assuming the presence of metallic cobalt/Co9S8, CoN4, CoO and Co(OH)2. Due to the similarities of the Co spectra, qualitatively the same species were attributed to the other three catalysts. Note: it was visible by XRD that the catalysts contained crystalline cobalt sulphides. However, as a reference measurement is missing to implement it in our fit, a Co 2p spectrum from the literature was considered as a reference. Alstrup et al.57i showed in their work that the Co 2p spectra of Co9S8 and cobalt single crystal are almost identical; hence in the fit in Fig. S6 it can only give a minor contribution in agreement with the calculated value for S in CoSy in Table 1.
image file: c8ta05769a-f4.tif
Fig. 4 XPS results giving an exemplary survey scan of cat D (a) and finescan regions related to Co 2p 3/2 (b), O 1s (c), N 1s (d) and S 2p (e) for cat A to D (from top to bottom).
Table 1 Summary of the elemental composition (at%) derived from XPS measurements of the investigated samples. Besides the overall composition, the contents assigned to sulphur, respectively, nitrogen in CoSy and CoNx are given as well
At% N C O Co S CoSyx CoNxx
Cat A 13.5 ± 0.3 76.0 ± 0.5 7.2 ± 0.3 2.3 ± 0.3 1.0 ± 0.1 0.24 ± 0.02 1.8 ± 0.1
Cat B 11.4 ± 0.4 72.5 ± 0.6 10.0 ± 0.3 2.7 ± 0.4 3.4 ± 0.1 0.47 ± 0.02 2.1 ± 0.2
Cat C 11.8 ± 0.4 74.9 ± 0.6 8.0 ± 0.4 2.6 ± 0.4 2.7 ± 0.2 0.42 ± 0.03 2.1 ± 0.1
Cat D 10.2 ± 0.3 71.2 ± 0.5 11.4 ± 0.3 3.1 ± 0.3 4.0 ± 0.1 0.34 ± 0.01 1.4 ± 0.1


The quantitative analysis of cat D is in relatively good agreement with contributions in the N 1s and O 1s spectra. However, due to the need to integrate several species in this fitting, the error might be rather large. Instead of integrate, the difference spectra of the catalysts were analysed and will be discussed in relation to Fig. 8, below.

Considerable changes were visible in the S 2p region. Here, the relative intensity of the peak located at about 168 eV was significantly higher for cat B and cat D compared to cat A and cat C. The opposite trend was observed for a shoulder at ca. 166 eV.

The deconvoluted N 1s and S 2p spectra are shown in Fig. 5a and b. The peaks at 168 eV and 166 eV, respectively, were assigned to sulphate and C–S–O bonds (integrated in carbon). We will refer to this again, later. In addition to these two species, cobalt sulphide (CoSy, ca. 161.8 eV) and sulphur integrated in carbon (C–S, 163.7 eV) were identified in the S 2p spectra. Within the N 1s region (Fig. 5a) oxidic (404–406 eV), graphitic (ca. 402 eV), pyrrolic (ca. 400.5 eV), pyridinic (ca. 398.5 eV) and Me–N interactions (ca. 399.5 eV) were found.58 From the relative areas and the nitrogen and sulphur contents, the fraction of atoms bound to nitrogen and sulphur could be determined for each species.


image file: c8ta05769a-f5.tif
Fig. 5 Deconvoluted N 1s (a) and S 2p (b) spectra of cat D. In addition, the contents assigned to each N 1s and S 2p species were determined and are plotted in (c) and (d), respectively.

These values are given in Fig. 5c and d, respectively, for the N 1s and S 2p regions. Already here, the decrease of pyridinic nitrogen and pronounced increase of sulphur heteroatoms in carbon are visible. The trends get even better pronounced in ESI Fig. S7 where these values are plotted as a function of the S/Co ratio in the precursor.

Thus, the higher initial S/Co ratio caused the release of nitrogen and favoured the integration of sulphur as a dopant into the carbon network.

The overall elemental composition of the catalysts is summarised in Table 1; also here the estimated amounts of nitrogen in CoN4 and sulphur in cobalt sulphide are added. This indicates that all catalysts contained large fractions of nitrogen and oxygen, while the sulphur content – considering its large amounts in the precursor – was relatively small, but increased when Selemental was added during the synthesis.

Impact of the composition on oxygen evolution reaction (OER) activity and stability

Being aware of the chemical composition and local structural environments, the electrocatalytic applicability of the catalysts was investigated. Therefore, in Fig. 6 the linear scan voltammograms (LSVs) for evaluating the OER activity in 0.1 M KOH as well as the corresponding Tafel plots are shown for a loading of 1 mg cm−2 (Fig. S8 shows the OER activity trend for a smaller loading of 0.5 mg cm−2 (0.1 M KOH) and Fig. S9 the comparison of two measurements of each catalyst for high loading). Considering the low onset potentials, it is evident that the catalysts displayed a good catalytic activity for oxygen evolution in agreement with previous findings.33,36,49
image file: c8ta05769a-f6.tif
Fig. 6 (a) Linear scan voltammetry (LSV) of the catalysts in 0.1 M KOH at 1500 rpm (catalyst loading 1 mg cm−2). The potential was corrected for iR drop. In (b) the Tafel plots related to these measurements are given. For reasons of comparison a commercial IrO2 catalyst was measured (loading 250 μg cm−2 at 10 mV s−1).

In Fig. S10 measurements with a RRDE electrode were performed for the catalysts to distinguish between contributions of the desired oxygen evolution reaction and the undesired carbon oxidation reaction. In contrast to Li et al.59 the applied potential at the ring was 0.45 V (SHE) rather than −0.6 V in their work. This potential of 0.45 V was chosen to avoid contributions of the HER or CO2 reduction. Our selected potential is similar to the recommendations for using the RRDE for efficiency determination by Diaz-Molares et al.45 and McCrory et al.60 It should be pointed out that these data were measured on aged samples (stored in air) and were not iR corrected as the related bipotentiostat does not offer iR correction. In addition, efficiency determination by the RRDE on a porous catalyst might underestimate the real OER contribution as part of the formed O2 might be trapped in the pores.

Nevertheless, there are some important observations to be made:

(1) Due to the storage in air the activity of the samples became less, consequently the onset for the OER was shifted to higher potential values (see Fig. S11 for the time-dependent change of the most active cat D).

(2) The faradaic efficiency εFaraday towards the OER (Fig. S12), for the recommended value60 of j = 1 mA cm−2, was in the case of the anodic scan about 20–25% (>50% for cat C) and in the cathodic scan about 50% (cat A, cat D) or 70% (cat B, cat C).

The OER activity was found to be highest for cat D followed by cat A > cat B ≈ cat C. Also the Tafel slope (determined in a current density range 1–10 mA cm−2) varied for the catalysts. A value of 95 ± 5 mV dec−1 was observed for cat D, 125 ± 2 mV dec−1 for cat B, about 128 ± 1 mV dec−1 for cat C and 190 ± 5 mV dec−1 for cat A. The error was determined by considering variation in the current density range between 0.9 and 20 mA cm−2.

Note, as the faradaic efficiency was obtained only for the aged samples and without iR correction, the absolute values of the Tafel slopes would change, whereas we assume that the relative trend between the four samples would remain. It is interesting to note that while the capacity correlated with the overall BET SA, the OER activity correlates with the micropore SA, as indicated in Fig. S1c. The error bars from double measurements of the OER activity are included in this graph.

In Fig. 7 both the durability (in terms of activity changes induced by potential cycling) and the stability (in terms of galvanostatic treatment) of the best performing catalyst (cat D) were evaluated. The former was performed in order to check for performance changes induced by carbon oxidation, the latter to confirm stable performance during constant current operation. It is evident that the current density was stable or even slightly increased during these treatments. In addition to this, a decrease of the Tafel slope from 95 ± 5 mV dec−1 to 65 ± 2 mV dec−1 was observed.


image file: c8ta05769a-f7.tif
Fig. 7 (a) LSVs of cat D in 0.1 M KOH at 1500 rpm (catalyst loading 1 mg cm−2) at the beginning of life (B.o.L.) and after 500 and 2000 potential cycles. In (b) the Tafel plots related to these measurements are shown. In (c) the LSVs at the B.o.L. and End of Life (E.o.L.) of the galvanostatic treatment at 10 mA cm−2 (d) are given.

In Fig. S13 the change in the CVs during this durability cycling is shown with only minor effects on the capacity induced by cycling.

Discussion of possible origins of Tafel slope changes

Here we propose and discuss two different explanations for the observed trends that might even be coupled. More work is required to enable final conclusions.
Explanation 1. The obtained value of the Tafel slope could depend on the degree of overlap between the OER and carbon oxidation. Under this assumption, the contribution of carbon oxidation is different for the different catalysts, causing the variation in current density and Tafel slopes. In this case, it could be expected that the Tafel slopes were related to the properties of the carbon, e.g. the degree of graphitization (as expressed by the ID/IG ratio) or edge-exposed heteroatoms. (In the work of Charreteur et al.61 it was shown for other Me–N–C that the full width at half maximum (fwhm) increased as more heteroatoms were integrated at the edges of graphene layers.)

The degree of graphitization is inversely proportional to the ID/IG ratio. Thus if ID/IG increases, a larger contribution of carbon oxidation to the current density is expected. Consequently, Tafel slopes might increase with increasing ID/IG ratio. To check this statement, the related graph is shown in Fig. S14a. Indeed, related changes in the Raman spectra were observed by Zana et al.62 for carbon black, that was continuously cycled between 1.0 and 1.5 V. However, for the catalysts in this work Raman spectra were not measured after conditioning.

Similarly, it can be expected that heteroatoms at the edges of graphene layers get oxidized more easily; to evaluate this the Tafel slope is correlated with the fwhm of the D band, in Fig. S14b. Also here, catalysts that could be assumed to exhibit a stronger carbon oxidation current show larger Tafel slope values. Following this explanation, the trends observed within the stability tests in Fig. 7 could be explained, for example by a decreasing ratio of ID/IG (as expected in relation to Zana's work62).

Explanation 2. There is a strong relationship between the rate determining step (RDS) as well as surface coverage and the observed Tafel slopes for multi-electron transfer reactions.63–65

For the OER in alkaline electrolyte on single metal sites, the following mechanism is proposed:63,64,66

 
M + OH ⇌ M–OHads + e(1)
 
M–OHads + OH ⇌ M–Oads + H2O + e(2)
 
M–Oads + OH ⇌ M–OOHads + e(3)
 
M–OOHads + OH ⇌ M + H2O + O2↑ + e(4)

In these equations M indicates the metal of the catalytically active species. In terms of the catalysis relevant surface coverage, the equations indicate that empty sites, Co–OH, Co–O and Co–OOH, will be formed during the oxidation cycle. There are two extreme cases in terms of the dimension of theoretically expected Tafel slopes. In general, a Tafel slope of 120 mV dec−1 is expected when the species that is formed before the RDS is dominating the surface coverage, e.g. if the fourth step is the RDS a surface coverage dominated by OOHads would be expected.

However, smaller values are observed, if the surface coverage is shifted to species formed at an earlier stage in the oxidation cycle. For example, if the third step would be rate determining, it is expected that the surface of the cobalt-based catalytic sites is mainly determined by Oads. A decrease of the Tafel slope could then also indicate that the surface coverage with Oads becomes less, but more OHads is found. Depending on which species are mainly found on the surface of catalytic sites, the value might even go down to about 20 mV dec−1 when for instance empty sites remain the dominating surface sites (M without adsorbate). This has been nicely illustrated by Shinagawa64 (see Fig. 4b–f shown in their work).

How does the second explanation relate to the results obtained in this work?

From Fig. 6 and 7 it was possible to draw the following conclusion with respect to the activity trend: the Tafel slope of the best performing catalyst was 95 mV dec−1 and became larger for the catalysts with the lowest performance. In accordance to this, during the stability and durability measurements the Tafel slope decreased from 95 ± 4 mV dec−1 to 75 ± 4 mV dec−1 and further down to 65 ± 2 mV dec−1 while the activity got enhanced. Thus, in this work more active catalysts displayed smaller Tafel slopes.

As described above, in terms of surface occupation, a decrease of the Tafel slope could be indicative of a dominance of species formed prior to the RDS within the overall oxidation cycle.

To get further conclusions on the mechanism, the O 1s spectra were deconvoluted into their different components. The spectra were fitted with Co–O, Co–OH (assigned to hydroxide and defective oxide), H2Oads and organic species (C–O, C[double bond, length as m-dash]O). Fig. 8 shows the deconvoluted O 1s spectrum of the most active catalyst and correlates the concentration of Co–OHads (from O 1s) with the amount of cobalt in the catalysts. Furthermore, it shows that the amount of Co–OHx increases and the amount of Co–O decreases, when the Tafel slope decreases.


image file: c8ta05769a-f8.tif
Fig. 8 Deconvoluted O 1s spectrum of the most active catalyst (cat D) of this work (a). In (b) the oxygen content assigned to Co–OH is given as a function of the as-measured cobalt content in the catalysts. Illustration of the changes in Tafel slope as a function of O in Co–O (c) and in Co–OH (d).

Even though these data were obtained ex situ, the relationship between Tafel slopes and both Co–O and Co–OH seems to indicate that the surface occupations of both species was of importance for the OER on our catalysts. Indeed also the Co 2p difference spectra (in relation to the most active cat D) in Fig. S15 point in the same direction: while the contribution of CoO decreased in the order cat A > cat B ≈ cat C > cat D, the contribution for the binding energy associated with Co(OH)2 was largest for cat D while it was approx. similar for cat B, cat A and cat C.

Usually a Tafel slope of 120 mV dec−1 is assigned to the first electron transfer step as the RDS. However, it seems that on this group of catalysts, the surface occupation might be more important for changes of the Tafel slope instead of a change of the RDS. Comparing our trends in Fig. 8c and d to the relationship of Tafel slope and surface occupation proposed by Shingagawa et al.,64 the fourth electron transfer step could be assigned as the RDS.

These results are further supported by the fact that during OER stability and durability testing (Fig. 7) a decrease of the Tafel slope could be observed.

Therefore, also for this group of sulphur- and nitrogen-doped cobalt-based carbon catalysts the concentration dependence of surface hydroxide on the OER41 was strongly indicated.

To what extent the CoN4 and CoSy species remain intact during the OER or whether they transform to hydroxides is an important question that needs to be addressed in more detail in future work. It is also important to mention that carbon oxidation and the formation/presence of hydroxide species at the surface of the catalyst might be related to each other. For example, the formation of hydroxide species might be initiated by carbon oxidation and exposure of cobalt species to the surface of the catalyst.

Conclusions

A group of multi-heteroatom doped carbon-based catalysts were synthesised from PANIevap., DCDA, Selemental and cobalt acetate by a pyrolysis process. The catalysts revealed a heterogeneous composition of cobalt nanoparticles, cobalt sulphide, CoN4 sites, and oxidised cobalt. Regarding the carbon morphology, it was found that with a molar ratio of S/Co > 15 nanotube formation took place, whereas at the same time the overall BET surface area decreased. More importantly, the diameter and relative contribution of the carbon nanotubes to the overall carbon became larger. The catalysts obtained high catalytic activity towards the oxygen evolution reaction and even improved performance with time. The faradaic efficiency is minimum 35% at 1 mA cm−2 for the investigated catalysts. The remaining current contribution is most probably given by carbon oxidation.

The analysis of Tafel plots in relation to the 1st order region of carbon in Raman spectroscopy and the near-surface composition obtained by XPS let us conclude that either carbon oxidation or a beneficial formation of cobalt hydroxide species at the surface dominated the observed Tafel slope trends. Also a combined mechanism is possible. In future work, emphasis should be given to a more detailed study of structural changes induced by the OER for this group of carbon-based catalysts by post mortem analysis.

Experimental

Synthesis procedures

Synthesis of PANIevap. The synthesis of PANIevap was already described in our previous article.14 Briefly, PANIevap was synthesised by oxidative polymerisation of aniline with ammonium peroxidisulfate (APS) in hydrochloric acid. The molar ratio was 1[thin space (1/6-em)]:[thin space (1/6-em)]3. After synthesis, the solution was evaporated so that the product contains polyaniline, the residuals of APS and HCl. The quantities of PANIevap and the other precursors are given in Table 2.
Table 2 Mass quantities for the preparation of N- and S-doped cobalt-based carbon catalysts
[mg] Selemental CoAc·4H2O PANIevap DCDA S/Co (molar)
Cat A 0 98.7 680 2280 10
Cat B 98 98.7 680 2280 15
Cat C 180 98.7 680 2280 25
Cat D 362 98.7 680 2280 32


Procedure for synthesis of Co–N–C catalysts. In order to prepare the sulphur- and nitrogen-doped cobalt-based carbon catalysts (CoAc), 99 mg cobalt acetate tetrahydrate and 680 mg PANIevap were mixed and ground with a mortar and pestle. In the next step, 2268 mg dicyandiamide (DCDA) is added as a structure forming agent (SFA). The elemental sulphur (Selemental) was added to the mixture after pre-grinding and the overall mixture was ground further until a homogeneous mixture was obtained. Without considering the mass of the SFA and Selemental, the precursor contains 3 wt% cobalt.

The mixed powder was filled into quartz boats and subjected to the pyrolysis in a furnace under constant flow of nitrogen. The synthesis procedure includes two heat treatment steps with an acid leaching step in between. The first heat treatment involves heating from room temperature (RT) to 800 °C at 5 °C min−1, with intermediate temperature equilibrations at 300 °C for 30 min, at 500 °C for 30 min and finally at 800 °C for 60 min. Subsequently, the mixture was cooled down to RT and then transferred to 2 M HCl for the acid leaching step (60 min in an ultrasonic bath, followed by stirring over-night). The suspension was washed with water, dried and transferred again to the furnace for the second heat treatment. Also here, the catalysts were continuously heated in a nitrogen atmosphere but with a significantly larger ramp of 38 °C min−1 up to 600 °C. The heating ramp was then slowed down to 5 °C min−1 up to 800 °C with a dwell time of three hours. After cooling down, the catalyst powder was ground and stored for further characterisation.

Electrochemical evaluation

Preparation of the working electrode. In order to prepare the working electrodes, a catalyst suspension was prepared and then drop-cast on a glassy carbon disc. For the ink preparation, 5 mg of catalyst powder was mixed with 25 μl Nafion (5 wt%), 142 μl of ethanol, and 83.2 μl of water. The suspension was placed for 30 min in an ultrasonic bath. Before use, the ink was homogenised with an ultrasonic homogenisator. Then, 10 μl of the ink was drop-cast on the glassy carbon disc to obtain a loading of 1 mg cm−2. It should be noted that the OER activity was also evaluated with a smaller loading (0.5 mg cm−2), therefore 5 μl of the same ink were used in the working electrode preparation.
Evaluating OER activity. The measurements were accomplished by using a standard three electrode configuration. The setup contains the RDE electrode with a glassy carbon disc equipped with the catalyst, a glassy carbon rod (counter electrode), and a Hg|HgO|1 M NaOH reference electrode. First, for activation of the electrodes, cyclic voltammetry (CV) was carried out in 0.1 M KOH from 0.0 V to 1.2 V with a sweep rate of 100 mV s−1 with 10 cycles and afterwards in the range 1.2–1.8 V for 30 cycles with a sweep rate of 300 mV s−1. To evaluate the OER activity, linear sweep voltammograms were measured from 1.2 V to 1.9 V with a sweep rate of 5 mV s−1 at a rotation speed of 1500 rpm. The reported potentials refer to the standard hydrogen electrode and are corrected for iR drop.
Evaluation of the faradaic efficiency εFaraday. The faradaic efficiency was determined by rotating ring disc electrode (RRDE) experiments with a glassy carbon disc (A = 0.238 cm2) with the deposited catalysts (5 μl ink) and a platinum ring. Potential control of the disc and ring was made with a Versastat 3/3F bipotentiostat with the same counter and reference electrodes as given above. After activation of the catalysts, RRDE experiments for selectivity evaluation were performed. The faradaic efficiency for oxygen formation during the OER was determined using the following equation:
 
εFaraday = IRingIdisc−1N−1(5)
where IRing is the measured ring current, Idisc is the as-measured disc current and N the collection efficiency. For the given electrode design N equals 0.38.

The selectivity measurements were made in a potential range of 1.9 V to 1.2 V for the disc electrode with a sweep rate of 10 mV s−1 at a rotation speed of 1500 rpm. The ring potential was fixed to U = 0.45 V (SHE).

The internal resistivity (ohmic) was determined using a Nordic electrochemistry potentiostat in the corresponding OER potential window at a constant frequency (5 kHz), obtained values are depicted in Fig. S16. Fig. S17 shows the comparison of the non- and iR-corrected OER polarization curve for cat D. Compensation for ohmic losses was made according to eqn (6):

 
UiR-corr. = UexpiR(6)

Fig. S17 also gives the non-corrected OER data of the four catalysts.

All potentials in this work were converted to the standard hydrogen electrode (SHE) by measuring the potential shift between the reference electrode in 0.1 M KOH towards a commercial SHE (giving a potential difference of 0.9 V).

This value fits nicely to the theoretically assumed value of 0.907 V.

 
ESHE = Emeasured + E0Hg|HgO|1 M NaOH + 0.059 V pH(7)
 
ESHE = Emeasured + 0.140 V + 0.767 V = Emeasured + 0.907 V(8)

Evaluating OER stability and durability. The durability test was performed with cycling between 1.1 and 1.8 V vs. SHE and a sweep rate of 0.3 V s−1, and the stability test was carried out galvanostatically at 10 mA cm−2 at a rotation speed of 1500 rpm.

Structural characterisation

X-ray diffraction. The measurements were performed at RT using a Bruker D8 Advance diffractometer in Bragg–Brentano geometry employing a monochromatised Cu Kα radiation source and a VANTEC detector. Data were recorded in an angular range between 5° and 50° (2θ) for a total measurement time of one hour using a step size of ∼0.007°, a step time of 0.5 step per s and a fixed divergence slit of 0.3°. Due to the low signal to noise ratio, assignment of the reflections was made using a logarithmic plot of the data and the presence of the phases could be confirmed by Rietveld analysis.
TEM imaging. TEM characterisation was performed with a FEI CM20 STEM (Eindhoven, the Netherlands) microscope equipped with a LaB6 cathode and a Gatan double tilt holder at a nominal acceleration voltage of 200 kV. The catalyst powder was dissolved in ethanol and ultrasonicated for 30 seconds. Large agglomerates and particles were allowed to settle down, before a small quantity was dropped on a cupper grid with a carbon film (Baltec MED010).
Raman spectroscopy. The measurements were performed using an alpha 300R confocal Raman microscope from WiTec (Ulm, Germany) with a grid of 600 lines mm−1 and an excitation laser of λ = 532.2 nm with a power of 1 mW. The catalyst was suspended and then dropped on a silica disc and left to dry. The data were obtained for three different spots in a wavenumber range of 0 to 2000 cm−1. An integration time of 10 s was used for each data point. The as-given spectra are the sum-spectra of the three measurements obtained for each catalyst. The main paragraph text follows directly on here.
N2 sorption measurements. N2 sorption measurements were performed with an Autosorb test station from Quantachrome. Prior to the measurements, the samples were degassed at 200 °C for 16 h. From N2 sorption data the BET specific surface area as well as the micropore surface area from Vt-plots were determined. Micropores are pores with a diameter < 20 nm. The given data of the mesopore surface area are calculated by subtracting the micropore surface area from the overall BET SA.
X-ray photoelectron spectroscopy. XPS measurements were performed using a Specs Phoibos 150 hemispherical analyzer and a Specs XR50M Al Kα X-ray source (E = 1486.7 eV). For the survey scans, an energy step of 1 eV has been applied and two scans were overlaid. In addition, fine scans were obtained for the Co 2p, O 1s, N 1s, C 1s and S 2p regions. For the fine scans an energy step of 0.05 eV has been used. For the fitting of the spectra, the software Casa XPS was used. The standard deviation corresponding to each species was determined by using Monte Carlo simulation and error propagation.

Statement of contributions

Concept of the work A. S. and U. I. K., preparation of catalysts A. S. and K. A. C., electrochemical characterization A. S., W. D. Z. W., and N. W., physicochemical characterization and data analysis M. K., I. M., M. A. N., S. P., and O. C., and all authors contributed to the writing process.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

Financial support by the German Research Foundation within the Graduate School of Excellence Energy Science and Engineering (GSC1070) is gratefully acknowledged by A. S., I. M., S. P. and U. I. K., U. I. K. and W. D. Z. W also acknowledges financial support via the grant KR3980/4-1. We would also like to thank W. Jaegermann and his group for the possibility to use the XPS system DAISY-FUN for characterisation of our catalysts.

References

  1. P. Larscheid, L. Lück and A. Moser, Renewable Energy, 2018, 125, 599–608 CrossRef.
  2. European Commission, Report on critical Raw Materials for the EU, 2014 Search PubMed.
  3. P. Du and R. Eisenberg, Energy Environ. Sci., 2012, 5, 6012–6021 RSC.
  4. M. Shao, Q. Chang, J.-P. Dodelet and R. Chenitz, Chem. Rev., 2016, 116, 3594–3657 CrossRef CAS PubMed.
  5. U. I. Kramm, M. Lefèvre, N. Larouche, D. Schmeisser and J.-P. Dodelet, J. Am. Chem. Soc., 2014, 136, 978–985 CrossRef CAS PubMed.
  6. E. Proietti, F. Jaouen, M. Lefévre, N. Larouche, J. Tian, J. Herranz and J.-P. Dodelet, Nat. Commun., 2011, 2, 416 CrossRef PubMed.
  7. J. Y. Cheon, T. Kim, Y. Choi, H. Y. Jeong, M. G. Kim, Y. J. Sa, J. Kim, Z. Lee, T.-H. Yang, K. Kwon, O. Terasaki, G.-G. Park, R. R. Adzic and S. H. Joo, Sci. Rep., 2013, 3 Search PubMed.
  8. J. Shui, C. Chen, L. Grabstanowicz, D. Zhao and D.-J. Liu, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 10629–10634 CrossRef CAS PubMed.
  9. G. Wu and P. Zelenay, Acc. Chem. Res., 2013, 46, 1878–1889 CrossRef CAS PubMed.
  10. A. S. Varela, N. Ranjbar Sahraie, J. Steinberg, W. Ju, H.-S. Oh and P. Strasser, Angew. Chem., Int. Ed., 2015, 54, 10758–10762 CrossRef CAS PubMed.
  11. T. N. Huan, N. Ranjbar, G. Rousse, M. Sougrati, A. Zitolo, V. Mougel, F. Jaouen and M. Fontecave, ACS Catal., 2017, 7, 1520–1525 CrossRef CAS.
  12. O. Hitoshi, M. Tomomi, O. Yuji and Y. Ichiro, ChemistrySelect, 2016, 1, 5533–5537 CrossRef.
  13. F. Pan, W. Deng, C. Justiniano and Y. Li, Appl. Catal., B, 2018, 226, 463–472 CrossRef CAS.
  14. A. Shahraei, I. Martinaiou, K. A. Creutz, M. Kübler, N. Weidler, S. T. Ranecky, W. D. Z. Wallace, M. A. Nowroozi, O. Clemens, R. W. Stark and U. I. Kramm, Chem.–Eur. J., 2018, 24, 122480–12484 CrossRef PubMed.
  15. A. Shahraei, A. Moradabadi, I. Martinaiou, S. Lauterbach, S. Klemenz, S. J. Dolique, H.-J. Kleebe, P. Kaghazchi and U. I. Kramm, ACS Appl. Mater. Interfaces, 2017, 9, 25184–25193 CrossRef CAS PubMed.
  16. W. Deng, H. Jiang, C. Chen, L. Yang, Y. Zhang, S. Peng, S. Wang, Y. Tan, M. Ma and Q. Xie, ACS Appl. Mater. Interfaces, 2016, 8, 13341–13347 CrossRef CAS PubMed.
  17. A. Morozan, V. Goellner, Y. Nedellec, J. Hannauer and F. Jaouen, J. Electrochem. Soc., 2015, 162, H719–H726 CrossRef CAS.
  18. Q. Lu, G. S. Hutchings, W. Yu, Y. Zhou, R. V. Forest, R. Tao, J. Rosen, B. T. Yonemoto, Z. Cao, H. Zheng, J. Q. Xiao, F. Jiao and J. G. Chen, Nat. Commun., 2015, 6, 6567 CrossRef CAS PubMed.
  19. L. Zhang, J. Xiao, H. Wang and M. Shao, ACS Catal., 2017, 7, 7855–7865 CrossRef CAS.
  20. P. Chen, T. Zhou, L. Xing, K. Xu, Y. Tong, H. Xie, L. Zhang, W. Yan, W. Chu, C. Wu and Y. Xie, Angew. Chem., Int. Ed., 2017, 56, 610–614 CrossRef CAS PubMed.
  21. A. Abbaspour and E. Mirahmadi, Electrochim. Acta, 2013, 105, 92–98 CrossRef CAS.
  22. Q. Daniel, R. B. Ambre, B. Zhang, B. Philippe, H. Chen, F. Li, K. Fan, S. Ahmadi, H. Rensmo and L. Sun, ACS Catal., 2017, 7, 1143–1149 CrossRef CAS.
  23. F. Dai, W. Fan, J. Bi, P. Jiang, D. Liu, X. Zhang, H. Lin, C. Gong, R. Wang, L. Zhang and D. Sun, Dalton Trans., 2016, 45, 61–65 RSC.
  24. Y. Gorlin and T. F. Jaramillo, J. Am. Chem. Soc., 2010, 132, 13612–13614 CrossRef CAS PubMed.
  25. X. Qiao, J. Jin, H. Fan, Y. Li and S. Liao, J. Mater. Chem. A, 2017, 5, 12354–12360 RSC.
  26. S. Dresp, F. Luo, R. Schmack, S. Kühl, M. Gliech and P. Strasser, Energy Environ. Sci., 2016, 9, 2020–2024 RSC.
  27. U. I. Kramm, I. Herrmann-Geppert, S. Fiechter, G. Zehl, I. Zizak, I. Dorbandt, D. Schmeißer and P. Bogdanoff, J. Mater. Chem. A, 2014, 2, 2663–2670 RSC.
  28. I. Herrmann, U. I. Kramm, J. Radnik, P. Bogdanoff and S. Fiechter, J. Electrochem. Soc., 2009, 156, B1283–B1292 CrossRef CAS.
  29. W. Kiciński, B. Dembinska, M. Norek, B. Budner, M. Polański, P. J. Kulesza and S. Dyjak, Carbon, 2017, 116, 655–669 CrossRef.
  30. M. Ferrandon, A. J. Kropf, D. J. Myers, K. Artyushkova, U. Kramm, P. Bogdanoff, G. Wu, C. M. Johnston and P. Zelenay, J. Phys. Chem. C, 2012, 116, 16001–16013 CrossRef CAS.
  31. A. Janßen, I. Martinaiou, S. Wagner, N. Weidler, A. Shahraei and U. I. Kramm, Hyperfine Interact., 2018, 239, 7 CrossRef.
  32. N. R. Sahraie, U. I. Kramm, J. Steinberg, Y. Zhang, A. Thomas, T. Reier, J. P. Paraknowitsch and P. Strasser, Nat. Commun., 2015, 6, 8618 CrossRef CAS PubMed.
  33. H. Qian, J. Tang, Z. Wang, J. Kim, J. H. Kim, S. M. Alshehri, E. Yanmaz, X. Wang and Y. Yamauchi, Chem.–Eur. J., 2016, 22, 18259–18264 CrossRef CAS PubMed.
  34. P. Cai, J. Huang, J. Chen and Z. Wen, Angew. Chem., Int. Ed., 2017, 56, 4858–4861 CrossRef CAS PubMed.
  35. K. Jayaramulu, J. Masa, O. Tomanec, D. Peeters, V. Ranc, A. Schneemann, R. Zboril, W. Schuhmann and R. A. Fischer, Adv. Funct. Mater., 2017, 27, 1700451 CrossRef.
  36. M. Zhu, Z. Zhang, H. Zhang, H. Zhang, X. Zhang, L. Zhang and S. Wang, J. Colloid Interface Sci., 2018, 509, 522–528 CrossRef CAS PubMed.
  37. M. Chauhan, K. P. Reddy, C. S. Gopinath and S. Deka, ACS Catal., 2017, 5871–5879 CrossRef CAS.
  38. J. Suntivich, K. J. May, H. A. Gasteiger, J. B. Goodenough and Y. Shao-Horn, Science, 2011, 334, 1383–1385 CrossRef CAS PubMed.
  39. H. Wang, Z. Li, G. Li, F. Peng and H. Yu, Catal. Today, 2015, 245, 74–78 CrossRef CAS.
  40. W. T. Hong, M. Risch, K. A. Stoerzinger, A. Grimaud, J. Suntivich and Y. Shao-Horn, Energy Environ. Sci., 2015, 8, 1404–1427 RSC.
  41. N. Weidler, S. Paulus, J. Schuch, J. Klett, S. Hoch, P. Stenner, A. Maljusch, J. Brotz, C. Wittich, B. Kaiser and W. Jaegermann, Phys. Chem. Chem. Phys., 2016, 18, 10708–10718 RSC.
  42. D. Friebel, M. W. Louie, M. Bajdich, K. E. Sanwald, Y. Cai, A. M. Wise, M.-J. Cheng, D. Sokaras, T.-C. Weng, R. Alonso-Mori, R. C. Davis, J. R. Bargar, J. K. Nørskov, A. Nilsson and A. T. Bell, J. Am. Chem. Soc., 2015, 137, 1305–1313 CrossRef CAS PubMed.
  43. M. W. Louie and A. T. Bell, J. Am. Chem. Soc., 2013, 135, 12329–12337 CrossRef CAS PubMed.
  44. K. Fominykh, P. Chernev, I. Zaharieva, J. Sicklinger, G. Stefanic, M. Döblinger, A. Müller, A. Pokharel, S. Böcklein, C. Scheu, T. Bein and D. Fattakhova-Rohlfing, ACS Nano, 2015, 9, 5180–5188 CrossRef CAS PubMed.
  45. O. Diaz-Morales, I. Ledezma-Yanez, M. T. M. Koper and F. Calle-Vallejo, ACS Catal., 2015, 5, 5380–5387 CrossRef CAS.
  46. G. Wu, K. L. More, C. M. Johnston and P. Zelenay, Science, 2011, 332, 443–447 CrossRef CAS PubMed.
  47. P. Ghosh, T. Soga, K. Ghosh, T. Jimbo, R. Katoh, K. Sumiyama and Y. Ando, Nanoscale Res. Lett., 2008, 3, 242 CrossRef CAS PubMed.
  48. L. Ci, Y. Li, B. Wei, J. Liang, C. Xu and D. Wu, Carbon, 2000, 38, 1933–1937 CrossRef CAS.
  49. H. Wang, Z. Li, G. Li, F. Peng and H. Yu, Catal. Today, 2015, 245, 74–78 CrossRef CAS.
  50. P.-G. Yin, L. Jiang, T.-T. You, W. Zhou, L. Li, L. Guo and S. Yang, Phys. Chem. Chem. Phys., 2010, 12, 10781–10785 RSC.
  51. B. Rivas-Murias and V. Salgueiriño, J. Raman Spectrosc., 2017, 48, 837–841 CrossRef CAS.
  52. J. Zhu, L. Huang, Y. Xiao, L. Shen, Q. Chen and W. Shi, Nanoscale, 2014, 6, 6772–6781 RSC.
  53. F. Tuinstra and J. L. König, J. Chem. Phys., 1970, 53, 1126–1130 CrossRef CAS.
  54. N. Larouche and B. L. Stansfield, Carbon, 2010, 48, 620–629 CrossRef CAS.
  55. A. Sadezky, H. Muckenhuber, H. Grothe, R. Niessner and U. Pöschl, Carbon, 2005, 43, 1731–1742 CrossRef CAS.
  56. I. Martinaiou, A. Shahraei, F. Grimm, H. Zhang, C. Wittich, S. Klemenz, S. J. Dolique, H.-J. Kleebe, R. W. Stark and U. I. Kramm, Electrochim. Acta, 2017, 243, 183–196 CrossRef CAS.
  57. I. Alstrup, I. Chorkendorff, R. Candia, B. S. Clausen and H. Topsøe, J. Catal., 1982, 77, 397–409 CrossRef CAS.
  58. F. Jaouen, J. Herranz, M. Lefèvre, J.-P. Dodelet, U. I. Kramm, I. Herrmann, P. Bogdanoff, J. Maruyama, T. Nagaoka, A. Garsuch, J. R. Dahn, T. S. Olson, S. Pylypenko, P. Atanassov and E. A. Ustinov, ACS Appl. Mater. Interfaces., 2009, 1, 1623–1639 CrossRef CAS PubMed.
  59. L. Li, H. Yang, J. Miao, L. Zhang, H.-Y. Wang, Z. Zeng, W. Huang, X. Dong and B. Liu, ACS Energy Lett., 2017, 2, 294–300 CrossRef CAS.
  60. C. C. L. McCrory, S. Jung, J. C. Peters and T. F. Jaramillo, J. Am. Chem. Soc., 2013, 135, 16977–16987 CrossRef CAS PubMed.
  61. F. Charreteur, F. Jaouen, S. Ruggeri and J.-P. Dodelet, Electrochim. Acta, 2008, 53, 2925–2938 CrossRef CAS.
  62. A. Zana, J. Speder, N. E. A. Reeler, T. Vosch and M. Arenz, Electrochim. Acta, 2013, 114, 455–461 CrossRef CAS.
  63. M. T. M. Koper, J. Electroanal. Chem., 2011, 660, 254–260 CrossRef CAS.
  64. T. Shinagawa, A. T. Garcia-Esparza and K. Takanabe, Sci. Rep., 2015, 5, 13801 CrossRef PubMed.
  65. R. L. Doyle and M. E. G. Lyons, in Photoelectrochemical Solar Fuel Production, ed. S. Giménez and J. Bisquert, Springer, 2016,  DOI:10.1007/978-3-319-29641-8_2.
  66. F. Calle-Vallejo, J. I. Martinez and J. Rossmeisl, Phys. Chem. Chem. Phys., 2011, 13, 15639–15643 RSC.

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c8ta05769a

This journal is © The Royal Society of Chemistry 2018