Diana
Hötger
a,
Markus
Etzkorn‡
a,
Claudius
Morchutt
ab,
Benjamin
Wurster
a,
Jan
Dreiser
c,
Sebastian
Stepanow
d,
Doris
Grumelli
e,
Rico
Gutzler
*a and
Klaus
Kern
ab
aMax Planck Institute for Solid State Research, 70569 Stuttgart, Germany. E-mail: r.gutzler@fkf.mpg.de
bInstitute de Physique, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland
cSwiss Light Source, Paul Scherrer Institute, 5232 Villigen PSI, Switzerland
dDepartment of Materials, ETH Zürich, 8093 Zürich, Switzerland
eInstituto de Investigaciones Fisicoquímicas Teóricas y Aplicadas (INIFTA), Facultad de Ciencias Exactas, Universidad Nacional de La Plata, CONICET, 1900, La Plata, Argentina
First published on 11th January 2019
Transition metal atoms stabilised by organic ligands or as oxides exhibit promising catalytic activity for the electrocatalytic reduction and evolution of oxygen. Built-up from earth-abundant elements, they offer affordable alternatives to precious-metal based catalysts for application in fuel cells and electrolysers. For the understanding of a catalyst's activity, insight into its structure on the atomic scale is of highest importance, yet commonly challenging to experimentally access. Here, the structural integrity of a bimetallic iron tetrapyridylporphyrin with co-adsorbed cobalt electrocatalyst on Au(111) is investigated using scanning tunneling microscopy and X-ray absorption spectroscopy. Topographic and spectroscopic characterization reveals structural changes of the molecular coordination network after oxygen reduction, and its decomposition and transformation into catalytically active Co/Fe (oxyhydr)oxide during oxygen evolution. The data establishes a structure–property relationship for the catalyst as a function of electrochemical potential and, in addition, highlights how the reaction direction of electrochemical interconversion between molecular oxygen and hydroxyl anions can have very different effects on the catalyst's structure.
The FeTPyP + Co layer on Au(111) served as ORR catalyst (working electrode). Characterization of the stability was investigated after ten linear sweep voltammograms (LSV) from +0.1 V to −0.5 V with a scan rate of 50 mV s−1. For OER, a new sample of FeTPyP + Co on Au(111) served as working electrode for ten full range cyclic voltammograms (CVs) from −1.2 V to +0.7 V, also at a scan rate of 50 mV s−1. Prior to the electrochemical experiments the sample surface is characterized by STM at room temperature for real-space information on the catalyst structure, and with XAS for chemical characterization. After EC, samples were rinsed with milliQ water (R = 18.2 MΩ) to wash off most of the remaining electrolyte residues, and were then transferred back to the STM. Images were analysed using WsXM.23 In a second step, the samples were transported in a vacuum suitcase to the X-Treme beamline24 (Swiss Light Source, Paul Scherrer Institute, Villigen, Switzerland) for characterization by XAS. X-ray absorption spectra were recorded at 300 K with linear horizontal (σh) or linear vertical (σv) polarization and background subtraction was performed. The XAS were normalized to the integral of the corresponding spectrum. X-ray linear dichroism (XLD) spectra are calculated by taking the difference between two linearly polarized spectra (σv–σh) and normalized to the integral of the corresponding XAS . The experimental incidence geometry of the X-rays was 60° from the surface normal and the spectra were recorded in total electron yield. A negligibly small magnetic field of 50 mT was applied, assisting the electrons in leaving the surface and thereby increasing the signal-to-noise ratio of the measured drain current. The Fe/Co ratio was estimated comparing the area of the Fe and Co XAS weighted by the respective number of holes in the d-shell; other parameters such as differences in X-ray absorption cross-section were not taken into account. Chemical and structural integrity of the FeTPyP + Co catalyst was investigated after ten linear sweep voltammograms (LSVs) driving ORR (+0.1 V to −0.5 V, scan rate 50 mV s−1) and ten full range cyclic voltammograms (CVs) (−1.2 V to +0.7 V, scan rate 50 mV s−1) driving OER.
X-ray photoelectron spectroscopy (XPS) was performed using a SPECS Phoibos 150 analyzer with a non-monochromatic Mg Kα source (hν = 1253.6 eV). Signals were fitted using a Gaussian after subtracting the measured background of the clean sample.
The polarization curve of FeTPyP + Co/Au(111) catalysing ORR is depicted in Fig. 2a. The on-set potential of ORR is −0.08 V resulting in a peak potential Ep of about −0.19 V. Similar overpotentials for the same current densities are reported for Fe/Co conjugated microporous porphyrin polymer electrocatalysts for ORR in acidic media.29 After 10 LSVs in 0.1 M NaOH O2 saturated electrolyte, STM images depict major structural changes of the FeTPyP + Co network after ORR (Fig. 2b). The order of the monolayer is largely disrupted leading to a disordered appearance in the STM image. The molecules remain adsorbed on the Au(111) surface under reaction conditions, but reassemble in non-crystalline monolayers. Only a few small domains retain the squared structure typical for preEC FeTPyP + Co (yellow circle in Fig. 2b). Moreover, initial cluster formation with a size of about (6 ± 2) nm is visible (orange arrow). Additional 100 LSVs leave the catalytic activity unaltered but lead to more apparent disorder in STM images (Fig. S1, ESI†) and increased cluster formation.
A completely different picture arises during the anodic branch of OER. The on-set potential of OER is ∼0.6 V and unlike the oxygen reduction the process is not diffusion limited (Fig. 3a). The cathodic cycle demonstrates ORR onset at −0.10 V with a peak position of −0.22 V. Both the on-set potential and the peak position are slightly more negative compared to the experiment focusing exclusively on ORR. Whereas the catalytic activity of FeTPyP + Co persists for >10 CVs,13 the FeTPyP + Co/Au(111) surface structure changes considerably during OER (Fig. 3b, see Fig. S2 in ESI,† for a non-3D version of the STM topograph). After 10 CVs, no molecules can be resolved and part of the surface is covered with islands with a lateral extension of the order of tens of nanometres and an average height of about 4 Å (Fig. 3c). These islands exhibit a rather rough basal plane and appear structurally similar to the (oxyhydr)oxide nanoislands formed during OER with CoO nanoislands on Au(111),30 which agrees well with the almost identical on-set potential of ∼0.6 V for OER for both structures. The remaining surface shows the typical reconstruction of the clean Au(111) surface (Fig. 3b).
In order to disentangle the contributions to catalytic performance we studied OXR for the separate systems FeTPyP, CoTPyP and Co, always with a clean Au(111) surface as support. The ORR voltammetry is shown in Fig. 4a, where the grey dashed curves are clean Au(111) and free-base porphyrin (H2TPyP) are depicted as reference. Au(111) shows a higher ORR activity compared to the H2TPyP monolayer, possibly because the ORR inactive molecules block catalytically active sites of the clean gold surfaces such as step edges and other defects. ORR activity is increased by the presence FeTPyP, CoTPyP and by Co, which clusters at the elbow sites of the Au(111) herringbone reconstruction. The turquoise polarization curve of CoTPyP shows the highest ORR activity in particular in the reduction peak around ∼−0.2 eV. FeTPyP compares slightly worse and show a much broader reduction signal. The same is observed for Co. This observation is in accord with expected higher ORR activity of Co-porphyrins compared to Fe-porphyrins.31,32 However, O2 is expected to be reduced using two electrons to peroxide on Co-porphyrin and using four electrons to water on Fe-porphyrin.32,33 By comparison of these curves to the ORR curve of FeTPyP + Co in Fig. 2a it becomes apparent that ORR activity of FeTPyP + Co resembles that of CoTPyP, created during Co deposition. The characteristic curve shape of CoTPyP with the pronounced reduction peak masks the broad contributions of FeTPyP and Co, which nevertheless contribute to overall catalytic activity.
Fig. 4 OXR activity of the single constituents of the FeTPyP + Co catalyst. (a) ORR, 50 mV s−1 in 0.1 M NaOH, O2 saturated solution, (b) OER, 50 mV s−1 in 0.1 M NaOH, Ar saturated solution. |
The OER voltammetry in Fig. 4b provides a somewhat different picture and was studied in detail in ref. 13. Again, clean Au(111) and the H2TPyP covered surface exhibit the smallest catalytic activity. FeTPyP and CoTPyP are much more OER active above 0.6 V, with very similar activity.13 Co is the more active catalyst.
The structural changes during OXR observed in STM are chemically further analysed using XAS. Just like the Fe L2,3-edge preEC FeTPyP + Co (Fig. 5a, grey), the postORR FeTPyP + Co signal lacks a clear multiplet structure (red), indicative of the metallic state of Fe.34 The predominantly metallic Fe L2,3-edge is traced back to a partial exchange of the central Fe atom by co-deposited Co: the coordinated Fe in the macrocycle shows a somewhat more defined XAS with a shoulder at lower energy. This is masked by the Fe that is expulsed from the porphyrin macrocycle, which sits between molecule and Au(111) substrate and exhibits a featureless XAS signal due to the missing ligand field provided by the pyrrolic nitrogen.25 The resulting weak XLD signal is further reduced after ORR (Fig. 5b), presumably due to the reorganisation into disordered structures that do not allow Fe or Co coordination at the peripheral pyridyl groups. In contrast, the XAS Co L2,3-edge shows a multiplet structure both preEC (Fig. 5c, grey) and postEC (blue), which clearly demonstrates the exposure of Co to a coordinative environment,25,35 provided either within the porphyrin macrocycle or by the pyridyl nitrogen. The main peak is located at 776.6 eV, the second peak lies at 778.5 eV, whereas a spectrum lacking a multiple structure with one maximum at ∼777.9 eV is expected for metallic Co.36 The signal does change slightly postORR, however, the XLD remains unchanged (Fig. 5d). The minor differences can be explained by the use of different samples with ratios of Co/Fe ≈ 4.1 preORR and Co/Fe ≈ 1.2 postORR. The larger abundance of Co gives rise to a larger fraction of Co in a metallic environment. Therefore, the metallic contribution to the XAS spectrum that lacks a fine structure is enhanced, which explains the modified ratio of the double peak. XLD at the N K-edge (Fig. S3, ESI†) confirms that the organic backbone of the molecules is intact. The shape of the XLD of monometallic TPyP and the sample exposed to ORR are similar, confirming that the molecules are lying planar on the surface. The obvious structural changes between the pristine and postORR network observed in STM are accompanied by only minor changes in XAS. The absence of major changes in XAS together with the appearance of porphyrin molecules in STM confirms that the catalyst for ORR is indeed the intact FeTPyP/CoTPyP molecule with possible contributions from Co. The absence of order observed in STM of postEC FeTPyP + Co results from perturbed intermolecular interactions between porphyrin molecules during ORR and also from the rupture of metal–pyridyl coordination bonds, which otherwise stabilise the square structure. Increasing the number of LSVs (Fig. S1, ESI†) shows further reorganisation of the catalyst. While STM images after 1 and 11 LSVs are comparable, the catalyst surface changes significantly after 111 LSVs. Although the amount of molecules appears significantly reduced and clusters are forming, the electrocatalytic current density varies only slightly over the number of repetitions, leaving on-set and peak potentials (−0.08 V and −0.19 V) unchanged. The appearance of larger clusters in STM could be a consequence of agglomeration of the molecules and Co centres, but might as well be residues from the electrolyte which might not have been removed completely during the rinsing procedure. These clusters could hence incorporate sodium atoms and hydroxyl groups (see discussion of XPS data further down).
XAS provides a different picture for the fate of the molecular catalyst after OER (Co to Fe ratio ≈ 2.3). In contrast to the Fe L2,3-edges of pristine FeTPyP + Co and postORR, the Fe L2,3-edge postOER consists of two peaks at L3 almost equal in intensity at 707.1 eV and 708.9 eV (Fig. 5e, light red). The multiplet structure is clearly replicated in the L2 peak at 720.2 eV and 722.0 eV. This multiplet structure is quite similar to spectra typically reported for Fe oxide,37–39 with a rather large first peak that can be explained by the presence of metallic Fe, for example as edge atoms in the island which are not fully oxidized. The exact nature of the oxide is difficult to extract from XAS: the spectra of hematite (α-Fe2O3), goethite (α-FeO(OH)) and akaganeite (β-FeO(OH)) are very similar and difficult to differentiate from maghemite (γ-Fe2O3) or ferrihydrite.40 Since they are in qualitative agreement with the here measured XAS, any of these phases or a combination there of can explain the XAS. We can exclude, however, the formation of magnetite (Fe3O4) and wüstite (FeO), which are characterized by an absence of the defined multiplet splitting.40 The very small XLD is a result of an isotropic chemical environment or a result of randomly oriented Fe oxide clusters (Fig. 5f, light red). Moreover, three peaks can be identified in the XAS L3 Co absorption edge at 776.0 eV, 777.2 eV and 778.6 eV (Fig. 5g, light blue). This multiplet structure is typically assigned to oxidized Co in CoO.36 A minor XLD signal is present but shifted with respect to the pristine network (Fig. 5h, light blue). The multiplet structure of both Co and Fe edges deviate from the XAS of preEC (grey) and postORR and show oxidized Fe and Co, which confirms the interpretation of the 2D islands observed in STM postOER (Fig. 3b) as mixed cobalt/iron (oxyhydr)oxide. The absence of an N K-edge XLD postOER (Fig. S3, ESI†) evidences the chemical decomposition of the molecular structure, but alternatively could also be the result of a random orientation of the porphyrin plane with respect to the surface. The stable electrocatalytic current over several CVs implies that the morphological and chemical change from bimetallic-organic network into oxide islands occurs before actually driving OER. Morphological changes in STM are observable already after applying moderate positive potentials at which OER is just setting in (Fig. S4, ESI†). Untangling the connection between catalytic activity and morphology of the catalyst is thus straight forward: it is Co/FeOx, possibly in the form of (oxyhydr)oxides incorporating minor parts of carbon and nitrogen from the porphyrin core, which serves as catalyst for OER in the anodic branch. This is confirmed by the recently observed conversion of different Co-porphyrins into CoOx on an FTO glass electrode during OER.41 ORR in the cathodic branch is likewise catalysed by the oxide islands and not like in the pure ORR experiment by the molecular network. The small differences in on-set and peak potential of ORR show that the molecular network is the slightly better catalyst. The mechanism by which Fe and Co atoms are extracted from the porphyrin remains elusive. The TPyP molecules possibly decompose due to oxidative cleavage during OER, possibly through the interaction with highly reactive singlet oxygen,42 and organic fragments partially leach into the electrolyte. Interestingly, nature found ways to protect photosystem II from singlet oxygen generated during photochemical water oxidation with the help of sacrificial reductants,43 and rational catalyst design tackles the task of improving singlet oxygen resistance.42 Alternatively, gold oxidation, which sets in at around at 0.3 V, might induce molecular decomposition. The metal centres remain on the surface and cluster into catalytically active possibly (oxyhydr)oxide nanoislands.
As described above, STM evidences the appearance of clusters on the surface after ORR (Fig. 2b and Fig. S1d, ESI,† orange arrows) and also after OER (Fig. S4c, ESI,† orange arrows, and ref. 13). These are speculated to originate from residues of the electrolyte, which despite thorough rinsing after EC with pure water are not completely removed. XPS of the O 1s and Na 1s core levels shows pronounced peaks with binding energy maxima at 531.9 eV and 1071.8 eV for O 1s and Na 1s, respectively (Fig. 6). These signals have been observed for samples after EC in NaOH irrespective of electrocatalytic process (ORR, OER, also hydrogen evolution) and for different catalysts on Au(111). The O 1s peak at ∼532 eV can be attributed to adsorbed hydroxyls.44,45 The peak position of 1071.8 eV of Na 1s sits just between the expected values of metallic Na at 1071.76 eV and NaOH at 1072.59 eV.46 Submonolayer coverage of Na on Ru places the 1s signal at ∼1072 eV,47 which agrees well with the binding energy found here. The presence of electrolyte residues on the surface is confirmed by XPS and can explain the appearance of the cluster in STM. Whether they sit on top of the FeTPyP + Co catalyst, or displaced the molecules and sit directly on the bare gold surface, escapes the resolution of our STM images due to the challenging imaging conditions after electrochemistry.
Footnotes |
† Electronic supplementary information (ESI) available: Additional STM, EC, and XAS data. See DOI: 10.1039/c8cp07463a |
‡ Present address: Institut für Angewandte Physik, TU Braunschweig, 38106 Braunschweig, Germany. |
This journal is © the Owner Societies 2019 |