Tailoring Pore Size and Catalytic Activity in Cobalt Iron Layered Double Hydroxides and Spinels by Microemulsion-Assisted pH-Controlled Co-Precipitation
Graphical Abstract
Bubble plays: Through application of microemulsions during computer-controlled co-precipitation at constant pH, nano-scaled cobalt iron layered double hydroxides and spinels with controllable pore size were synthesized. Pores smaller than a certain threshold size turned out to be beneficial for electrochemical oxygen evolution reaction.
Abstract
Cobalt iron containing layered double hydroxides (LDHs) and spinels are promising catalysts for the electrochemical oxygen evolution reaction (OER). Towards development of better performing catalysts, the precise tuning of mesostructural features such as pore size is desirable, but often hard to achieve. Herein, a computer-controlled microemulsion-assisted co-precipitation (MACP) method at constant pH is established and compared to conventional co-precipitation. With MACP, the particle growth is limited and through variation of the constant pH during synthesis the pore size of the as-prepared catalysts is controlled, generating materials for the systematic investigation of confinement effects during OER. At a threshold pore size, overpotential increased significantly. Electrochemical impedance spectroscopy (EIS) indicated a change in OER mechanism, involving the oxygen release step. It is assumed that in smaller pores the critical radius for gas bubble formation is not met and therefore a smaller charge-transfer resistance is observed for medium frequencies.
Introduction
Hydrogen is generated by electrocatalytic water splitting as a clean energy carrier.1 The anodic half reaction, the oxygen evolution reaction (OER), exhibits sluggish kinetics and therefore represents the bottleneck of the overall reaction, which results in a high overpotential.2 A vast number of publications on decreasing the OER overpotential have been published including both screening studies and mechanistical investigations on a broad variety of electrode materials. In alkaline medium, 3d transition metal-based catalysts are up-and-coming materials.3 Undoubtedly, it is desirable to find and further develop catalysts based on earth abundant and therefore cheaper materials with low overpotential and high stability. Layered double hydroxides (LDHs) as well as cobalt-based spinels have proven to be promising candidates for this endeavor.4 We recently studied the systematic cobalt substitution in spinel catalysts, showing the impact of composition and surface facets on the OER.5 The therein observed effect of exposed crystallographic facets, as well as many other literature reports, underline that the mesostructure of electrocatalysts, i. e., the particle size, shape and porosity, influences the performance significantly,6 offering great potential for improved OER catalysts and calling for a systematic investigation and development of synthetic tools to control the mesostructure.
Mesostructure, in particular porosity, is known to play a major role in thermal catalysis, for example in reactions catalyzed by zeolites, metal-organic frameworks, covalent organic frameworks, and metals as well as oxides on host structures.7 Observed positive impact can be assigned to the change of adsorption energy, electronic effects or altered mass transport. Such effects could also play a role in electrocatalysis. For example, the control over catalyst pore size allows the investigation of electrocatalysis under spatial confinement in the pores. In this rapidly progressing field, so-called confinement effects shall alter and improve kinetics, mass transport, and ion specific selectivity of electrochemical reactions.8 Based on detailed understanding of such effects and their interaction, confinement could be strategically employed to design superior electrocatalysts. In this respect, OER catalysts represent a key area of research due to both, the high technological relevance of electrochemical water splitting and the complexity of the electrochemical process comprising a multi-step reaction mechanism and transport phenomena related to gas evolution. Despite the recent progress made in the fabrication of nanostructured OER catalysts,9 the synthesis of well-defined nano- or mesoporous systems remains challenging.
For this purpose, we further developed the catalyst synthesis by the crystalline precursor decomposition approach based on Co-containing LDH precursors forming pseudo-morphous spinels through thermal treatment.5 LDHs are layered materials that usually exhibit a typical platelet morphology and a homogeneous cation distribution,10 and are well-known OER catalysts,11 especially in form of nickel iron containing LDHs.3d, 12 Cobalt iron containing LDHs and derived (oxy)hydroxides exhibit a comparable activity in OER and are also highly suitable candidates for OER catalysts.13 Through thermal decomposition, LDH precursors can be transformed into mixed metal oxides.14 Given the LDH is at least partly built from oxidizable cations allowing a ratio of bivalent to trivalent cations of 1 to 2, spinels can topotactically form as a single oxide phase,5, 15 which also can be attractive OER catalysts.5 To allow controlling the mesostructure of LDH and LDH-derived spinel OER catalysts, we present a synthesis method using water in oil microemulsions as nanoreactors in combination with computer-controlled constant-pH co-precipitation in this work. Surfactant-stabilized microemulsions16 can form reverse micelles as a confined reaction space for co-precipitation to limit the particle growth. Such microemulsions have been used as a reaction medium for, for example, metal nanoparticles,17 metal salt aggregates,18 core shell nanoparticles,19 polymers,20 hydroxides,21 and oxides,22 including LDHs.23 While confinement of particle growth has been accomplished in these works, exact control to tune particle and pore size by specific synthesis parameters is often lacking, but highly desirable for the investigation of mesostructure-reactivity relationships and confinement effects.
The herein presented work establishes the microemulsion-assisted co-precipitation (MACP) under highly controlled conditions, allowing a systematic variation of important synthesis parameters, especially the pH value of co-precipitation. The constant-pH nature of the co-precipitation is based on simultaneous dosing of the metal salt microemulsion and the precipitating agent microemulsion, allowing in principle a continuous process that can tackle the notorious up-scaling problem of syntheses involving microemulsions. A cobalt to iron ratio of 2 to 1 was chosen to achieve the formation of phase pure LDHs, which can be thermally transformed to spinels. The impact of the synthesis pH values during such MACP on the mesostructure of the as-prepared LDHs and the corresponding Co2FeO4 spinels has been investigated in detail and compared to conventional aqueous co-precipitation. A special focus falls upon the investigation of conceivable confinement effects during OER.
Results and Discussion
Two series of cobalt iron layered double hydroxides with the same cobalt to iron ratio of 2 to 1 were synthesized and thoroughly characterized. Computer-controlled synthesis conditions were applied for all reactions, ensuring reproducibility (Figure S1). The first series was prepared by conventional co-precipitation using only aqueous solutions without any further additives. Five samples were prepared by varying the constant pH during precipitation from pH 7.5 to pH 9.5 in steps of 0.5 and are labeled according to the corresponding precipitation pH c75 to c95 throughout this manuscript.
The second series was prepared by MACP, confining the reaction space within the reverse micelles of a microemulsion. A suited formulation was adapted from literature,24 ensuring stable microemulsions containing water, metal salt solution or the alkaline precipitation agent as aqueous phase. Using these microemulsions five syntheses at constant pH values from pH 7.5 to pH 9.5 were performed as described above. The resulting samples are labeled m75 to m95. Further details on the synthesis procedures can be found in the Experimental Section and synthesis protocols for both synthesis approaches are shown in Figures S2 and S3.
Through thermal decomposition and oxidation of a fraction of Co2+ to Co3+ the layered double hydroxides prepared by conventional co-precipitation and MACP were transformed into Co2FeO4 spinels and are marked with the suffix “s” accordingly.
Micelles
The mesostructure of the as-prepared MACP LDHs and the corresponding spinels reported in this work is based upon the confinement of the reaction space within reverse micelles, limiting the particle growth. To prove this confinement and to gain further insight about the size of the micelles, dynamic light scattering (DLS) and TEM after plasma treatment were performed. More details can be found in the Experimental Section. Figure 1 shows scanning transmission electron microscopy (STEM) images of the template microemulsion, in other words the microemulsion containing only water, after drying and plasma treatment. Well defined spherical carbon walls with a cavity in the center are observed. A diameter of 13.1±4.6 nm was found by TEM, which agrees within the margin of error with the result of 16 nm from dynamic light scattering.
STEM images of the dried-out template micelles after plasma treatment.
Even though high-quality images of well-defined micelles were obtained for the template microemulsion, imaging of the microemulsions containing the metal salt solution and alkaline precipitation agent was more challenging. Due to higher ionic strength of the aqueous phase within these microemulsions, aggregation of the micelles upon deposition on the substrate took place and no reliable size determination was possible. The resulting TEM images can be found in Figure S4. Next to size evaluation with TEM, dynamic light scattering was conducted. Micelle diameters from both methods are listed in Table 1 and clearly show small micelles sizes of around 20 nm, which are suitable to confine the reaction space during co-precipitation. To investigate a potential change in plain micelle size, pH-dependent measurements were performed. The results can be found in Figure S5 but showed no significant difference in the investigated pH range and margin of error. Particle growth during co-precipitation is assumed to mainly take place within the reverse micelles, accomplished by stirring and a coalescence-exchange mechanism.25
Aqueous phase within micelles |
DiameterTEM [nm] |
DiameterDLS [nm] |
|---|---|---|
template |
13.1±4.6 |
16 |
metal salt solution |
– |
25 |
alkaline precipitation agent |
– |
17 |
Layered double hydroxides
Figure 2a shows the powder X-ray diffraction (PXRD) patterns of the as-prepared conventional and microemulsion-assisted samples at pH 8.5. Diffraction patterns of the complete series are shown in Figure S6 and no differences with precipitation pH within the groups are observed.
a) PXRD patterns of the conventionally prepared sample c85, the corresponding microemulsion-assisted sample m85 and the hydrotalcite reference (ICSD. No. 629626); b) IR spectra of c85 and m85; c) SEM micrograph of the conventionally prepared LDH c85; d) SEM micrograph of m85.
For the conventionally prepared samples a typical crystalline layered double hydroxide is obtained, and no crystalline by-phases are identified. The series from MACP on the other hand shows extremely broadened reflections. This can be caused by highly nano scaled particles, which consist of only a few brucite-like layers. Even though the main reflections of the hydrotalcite reference can barely be identified, theoretic calculations anticipated this behavior with decreasing slab size.27 With just a few layers remaining, most of the reflections are not visible anymore and the formation of a nano scaled, almost X-ray amorphous layered double hydroxide phase is therefore assumed. Exemplary IR spectra of the samples prepared at pH 8.5 are shown in Figure 2b. IR spectra of the complete series and a reference spectrum of the surfactant Triton X-100 can be found in Figures S9 and S10. Independent of the synthesis approach the spectra mainly exhibit the same features. The broad bands with a maximum at 3390 cm−1 are assigned to water and OH−, and the interlayer water exhibits a band at around 1640 cm−1.28 At 1350 cm−1 the vibration of the interlayer carbonate is visible and the bands below 1000 cm−1 are assigned to M−O, M−O−M, O−M−O, and M−OH vibrations.29 All bands of the conventional samples are significantly more pronounced than the bands of the MACP analogs, agreeing with the more crystalline nature of the former. For the MACP samples, the typical C−H stretching bands of the surfactant are absent, indicating that the organic part has been effectively removed from the sample by washing and drying. An additional broad band at around 1015 cm−1 is observed. With the help of UV-Vis spectroscopy of m85, a minor phosphate contamination was identified (Table S1) and the band at 1015 cm−1 was therefore mainly assigned to the P−O vibration.30 Additional minor carbonaceous residues cannot be completely excluded, given the slightly increased content of C and H. As the conventionally prepared LDHs do not exhibit this additional band, the phosphate is assumed to be brought in during synthesis using the microemulsion. The IR spectrum of the used pure Triton X-100 in Figure S10 shows the most prominent band at 1100 cm−1 with a broad shoulder in the wavenumber range of the additionally observed band in the MACP LDH spectra, making it a probable source for the phosphate. Partial intercalation of phosphates between the brucitic layers of the LDH was observed in literature,31 which might explain the continuing presence of the phosphate after washing. Scanning electron microscopy (SEM) micrographs of the corresponding samples in Figure 2c and d underline the findings from X-ray diffraction and IR spectroscopy. The conventional sample shows the expected anisotropic platelet morphology opposed by the aggregates of very small primary particles of m85 that appear rather isotropic at the resolution of the SEM. For all other precipitation pH, the SEM images are shown in Figures S7 and S8. Conventional co-precipitation yielded anisotropic platelets in all cases. For the MACP LDHs, significantly smaller and more isotropic nanoparticles are formed and an elongation of the particles with increasing precipitation pH is observed. Elemental analysis proved a cobalt to iron ratio close to the desired ratio of 2 to 1 for all samples (Table 2).
Precipitation |
Co/Fe ratio |
C+H content [at %] |
||
|---|---|---|---|---|
pH |
conventional |
MACP |
conventional |
MACP |
7.5 |
2.10±0.03 |
2.07±0.06 |
7.4±0.6 |
12.1±0.6 |
8.0 |
2.12±0.04 |
2.21±0.08 |
7.2±0.8 |
10.9±0.6 |
8.5 |
2.16±0.03 |
2.29±0.17 |
7.1±0.0 |
10.5±0.9 |
9.0 |
2.27±0.01 |
2.15±0.05 |
7.4±0.5 |
9.6±0.3 |
9.5 |
2.29±0.00 |
2.23±0.05 |
7.2±0.8 |
8.4±1.2 |
Some samples exhibit a slightly larger ratio, which can be explained by the hygroscopic nature of the iron nitrate precursor, leading to weighing errors and too small amounts of iron in the metal salt solution. CHN elemental analysis showed no nitrogen indicating the complete removal of the nitrate counter ions by the washing procedure. The sample series prepared by microemulsion-assisted co-precipitation shows a somewhat higher content of carbon and hydrogen implying incomplete removal of the surfactant. On the other hand, IR spectroscopy does not show any bands assigned to the surfactant, pointing towards minor amounts or a changed adsorption behavior due to the nano structuring of the MACP samples.
The textural properties were analyzed by N2 physisorption experiments and BET theory. Figure 3 depicts the adsorption/desorption isotherms as well as the pore size distribution for both series. The micropore volume and total pore volume can be found in the Supporting Information (Tables S2 and S3). As already seen in PXRD, IR, and SEM, enormous differences in terms of adsorption capacity and hysteresis shape are encountered. For the conventional samples no trend with precipitation pH in the type IV isotherms with an H3 hysteresis loop is observed. This behavior is typical for mesoporous solids with slit shaped pores and nonuniform size or shape, agreeing with the platelet morphology observed in SEM. On the contrary, the type IV isotherms of the series derived from MACP exhibit a clear trend with synthesis pH. The H2-like hysteresis as well as the pore size distribution increases with increasing pH.32 Surface areas for the microemulsion-based LDHs are found to be up to 4 times higher than those of their conventionally prepared counterparts, congruent with the nano sized particles found by PXRD and microscopy. Nevertheless, surface areas do not show any trend with precipitation pH (Table 3). Instead, a trend of the pore size distribution maximum of the MACP samples as determined by the Barrett-Joyner-Halenda (BJH) method is observed. The pores are substantially smaller than in the conventionally prepared LDHs and fall into a regime between 5.7 and 10.4 nm. In this size range, a steady increase with increasing co-precipitation pH is observed.
Top: N2 adsorption/desorption isotherms of the conventionally prepared LDHs (left) and microemulsion-assisted LDHs (right). Bottom: pore size distributions of the c (left) and m samples (right) synthesized at different pHs. Note the different y axis scaling.
Precipitation |
Surface area [m2 g−1] |
Pore size distribution max. [nm] |
||
|---|---|---|---|---|
pH |
conventional |
MACP |
conventional |
MACP |
7.5 |
97 |
281 |
21.3 |
5.7 |
8.0 |
75 |
295 |
21.5 |
6.9 |
8.5 |
69 |
255 |
32.8 |
7.9 |
9.0 |
57 |
255 |
21.4 |
9.0 |
9.5 |
69 |
233 |
32.5 |
10.4 |
Figure 4b shows a clear and linear relationship between particle properties of the MACP LDHs, in this case the maximum of the pore size distribution (open red symbols), and the synthesis parameter pH, by which the MACP LDH mesostructure can be controlled. The pH during synthesis influences the coalescence of the micelles in the microemulsion. Higher pH means a higher ionic strength and with that presumably more exchange due to increasing instability of the reverse micelles,33 leading to stronger growth of the particles and apparently widening of the inter-particle voids. For the conventionally prepared LDHs (Figure 4a, closed red symbols) no such control was provided. Consequently no trend of mesostructure with precipitation pH is observed.
Maximum of the pore size distribution PSDmax of all LDHs and spinels for all pHsynthesis. a) LDHs and spinels from conventional co-precipitation. b) MACP LDHs and corresponding spinels. Note the different scaling of the y axis.
For thermal transformation of the LDHs into spinels, a suited calcination temperature had to be defined. Therefore, thermogravimetric analysis (TGA) was performed and is shown exemplary in Figure 5. In all cases decomposition takes place in two steps, namely dehydroxylation and decarbonization and is completed below 400 °C, which was therefore chosen as calcination temperature.34 Above 900 °C, the thermal reduction of CoIII oxidic species to CoIIO can be observed. TGA for all samples is shown in Figure S11.
Thermogravimetric analysis of c85 and m85. The grey dotted line highlights the chosen calcination temperature of 400 °C.
Spinels
Figure 6 shows the PXRD patterns of the spinels derived from conventional co-precipitation and MACP exemplarily for a synthesis at pH 8.5 for the LDH precursor as well as the corresponding SEM micrographs. The diffraction patterns and SEM micrographs for the whole two series are depicted in Figures S12–S14.
a) PXRD patterns of the calcined samples c85_s and m85_s; b) IR spectra of c85_s and m85_s; SEM micrograph of c) c85_s and d) m85_s.
The pronounced difference found for the as-prepared LDHs in the PXRD patterns is not visible anymore. The crystallinity of both samples is not very high, but the typical reflections of the spinel structure can be clearly observed, and no crystalline by-phases are visible. Previous work showed a nano-scale segregation of the oxides into two structurally very similar spinels, a cobalt richer and an iron richer spinel, as implied by the cobalt iron oxygen phase diagram,35 which cannot be discriminated here due the broad XRD peaks. The small coherently scattering domains lead to the similar diffraction patterns.5 The IR spectra of the conventionally prepared spinels exhibit only the expected bands at 639 cm−1 and 538 cm−1, which are assigned to M−O vibrations.36 The spinels derived from LDHs prepared by MACP show a distinct band at around 1010 cm−1, which again is assigned to phosphate introduced during synthesis of the LDH and is stable against thermal decomposition at the applied temperature.37 As described before, the organic solvent and surfactant used for the synthesis are expected to be removed during washing and possible minor residues to be decomposed after thermal treatment at 400 °C for 3 h.38 All IR spectra are shown in Figure S15. Electron microscopy showed the retainment of the platelet morphology for the samples derived from conventional co-precipitation. Additionally, a visible pore formation due to decomposition of the anions and emission of water in the precursor has been observed in our previous work by TEM for a spinel sample derived from a conventional LDH with the same composition and precipitation at pH of 8.5, which can be seen by SEM as formation of holey platelets.5 Moreover, the nano structure of the spinels derived from MACP LDHs is retained.
For the LDHs prepared from MACP, a distinct trend between precipitation pH and pore size was found. Therefore, N2 physisorption experiments were performed for the calcined samples to investigate the pore size distribution and surface area after transformation to the spinel phase (Figure 7). For the conventional spinels, pore size distribution evaluation using the BJH method yielded broad maxima for both interparticle and intraparticle pores, which are hard to differentiate. Due to the pore formation, the surface areas of the conventional spinels are higher than those of the corresponding LDHs in most cases. The maximum of the pore size distribution and the surface areas are listed in Table 4. For the samples from MACP, the nanosized particles seem to be preserved upon calcination based on the resolution of the SEM images. Pore formation is not visible due to the small particle size. The surface area after calcination for these samples is cut in half except for m75_s, which exhibits a smaller decrease in surface area. The decrease may seem contradicting, as the surface area increased for the conventional samples, but the small particle size of the as-prepared MACP LDH makes them more prone towards sintering. Interestingly, the general trend in pore size distribution maxima is retained upon calcination (see Figure 4), meaning that the MACP can be used not only for the mesostructure control of the LDHs, but the LDHs can also be used as templates for the controlled synthesis and creation of unusual morphologies and mesostructure of the resulting spinel materials, including control over pore size.
Top: N2 adsorption/desorption isotherms of the conventionally prepared spinels (left) and microemulsion-assisted spinels (right). Bottom: pore size distributions of the c (left) and m samples (right) synthesized at different pHs. Note the different y axis scaling.
Precipitation |
Surface area [m2 g−1] |
Pore size distribution max. [nm] |
||
|---|---|---|---|---|
pH |
conventional |
MACP |
conventional |
MACP |
7.5 |
96 |
216 |
18.3 |
5.6 |
8.0 |
88 |
130 |
16.0 |
6.2 |
8.5 |
89 |
153 |
5.6 |
6.9 |
9.0 |
86 |
136 |
26.0 |
6.9 |
9.5 |
83 |
163 |
44.7 |
7.8 |
Summarizing the materials synthesis and characterization part, two sample series of LDHs and spinels were synthesized by conventional co-precipitation and MACP and characterized thoroughly. In the case of the MACP series, highly nano-sized samples were formed, whose pore size scaled with the pH value of the synthesis. For the conventionally prepared samples, no such trend was found. In contrast to the mesostructure, the crystallinity of both spinel sample series was very similar. The linear dependance of the pore size maximum on the pH value of the synthesis was maintained upon spinel formation from the MACP LDH precursors, demonstrating the tunability of the mesostructure by this synthesis parameter.
Electrocatalytic activity towards oxygen evolution
The above-described control of pore size makes the two sample series a suitable materials platform to study the effect of mesostructural confinement in electrocatalysis. Therefore, we, here, take advantage of the achieved set of LDH and spinel materials by assessing their catalytic activity towards OER under alkaline conditions.
Exemplary linear sweep voltammograms (LSVs) in aqueous 1 m KOH solution are displayed in Figure 8a for catalyst materials precipitated at pH 8.5 (complete series shown in Figure S16). Both sets of LDH samples exhibit low overpotentials as expected from literature.39 The two spinel series differ significantly with an increase in activity as a result of calcination for the MACP sample, but a very low activity of the conventionally prepared spinel samples. This performance difference of the spinels obtained at pH 8.5 was recently investigated in great detail and related to the favorable in situ formation of Co3+-rich domains in case of the MACP-derived catalyst.40 Here, we focus on the performance trends with variation of the synthesis pH value of the three series of electrocatalytically active samples, i. e., the conventional and MACP-LDH materials and the MACP-LDH-derived spinels. The overpotentials at 10 mA cm−2 are shown as a function of the synthesis pH values in Figure 8b and the catalyst stabilities in chronopotentiometric tests for 2 h are shown in Figure 8c (whole series and details in Figures S20 and S21). The overpotentials exhibit no clear trends with the synthesis pH (see Supporting Information for detailed discussion). Most of the materials exhibit similar overpotentials between 340 and 370 mV with lowest overpotentials for MACP spinels at high synthesis pH values. Yet, the plot exhibits a noteworthy peculiarity: three significantly higher overpotentials for MACP LDH samples synthesized at pH values ≥8.5. Figure 8d indicates the diminished electrocatalytic activity of these three MACP LDHs to correlate with maxima of the pore size distribution between 7.9 and 10.4 nm, while the other samples with respective pore sizes below 7.9 nm cluster in the range of lower overpotentials.
a) Linear sweep voltammograms in 1 m KOH, 1 mV s−1 at 1600 rpm for the LDHs synthesized at pH 8.5 and their corresponding spinels. b) Overpotential at 10 mA cm−2 for all LDHs and MACP spinels. c) Chronopotentiometry in 1 m KOH for c85 (solid red line), m85 (dashed red line) and m85_s (dashed black line). d) Correlation of overpotential with PSDmax for the MACP LDHs (open red symbols) and MACP spinels (open black symbols).
To first explain the comparable electrocatalytic activity of the investigated LDH and spinel samples, the cyclic voltammograms (CVs) of the electrochemical pre-conditioning (Figure 9) are analyzed. Prior to the LSV measurements, these 50 CVs were always recorded in a potential range more negative than the OER onset and including the Co redox range, to obtain reproducible and reliable results, which was achieved as demonstrated by the sufficiently small error bars (e. g., Figure 8b).
Pre-conditioning cyclovoltammograms of the samples synthesized at pH 8.5. a) First cycle, b) fifth cycle and c) 50th cycle. Note the difference in scaling of the y axis in a).
These pre-conditioning CVs (Figure 9) show major differences in current, peak positions and extents for different samples for the initial cycles but aligning CV shapes towards the 50th cycle. The first CVs exhibit pronounced anodic conversion signals relating to the oxidative transformation to higher-oxidized cobalt oxide species.41 This electrochemical conversion reaction is mostly irreversible as 5 to 8 times more charge is transferred during the anodic half-cycle compared to the cathodic part. In consequence, the surface or rather surface-near region changes in (mean) oxidation state during the initial cycles and may alter in phase and composition, as was already found for the MACP spinel prepared at pH 8.5 during OER by Haase et al.40 For the 50th CV, the ratio of anodic and cathodic charge approaches 1, indicating that the respective electrochemically active material fraction can now be transformed reversibly. The similarity in electrochemical response at this stage implies that the surface properties are quite similar for all samples after pre-conditioning due to the cyclovoltammetrically induced changes, explaining the small range of observed OER-overpotentials.
Figure 10a displays the ratio of anodic charge of the 50th CV in regard of the first anodic half-cycle for the microemulsion-templated materials. It is clearly seen that the charge relating to reversible transformation of the MACP-LDH-surface is small (12–15 %) compared to the initial conversion charge corresponding also to irreversible conversion. This indicates large changes compared to the surface-near region of the MACP spinel samples where the reversible transformation charge makes up 36 % to 89 % of the first anodic CV half-cycle. Thereby, the two samples exhibiting the lowest overpotential (MACP spinels with pHsynthesis=9.0 and 9.5) show a lower ratio of reversible charge to initial anodic charge than the other MACP spinel samples, indicating a possibly stronger transformation of the as-synthesized state. Besides, the MACP spinels show also higher reversible anodic charge per BET surface area with increasing synthesis pH value (and correspondingly increasing pore size) as seen in Figure 10c. Accordingly, the depth of the surface layer which can be anodically transformed and cathodically re-transformed is larger (for a specific scan rate).42
a) Ratio of Qanodic at the end to Qanodic at the start of the pre-conditioning, b) Qanodic of the 50th cycle in μC, and c) Qanodic of the 50th cycle in μC m−2 at different pHsynthesis. The surface area was derived from N2 physisorption measurements.
Conway and Liu43 reported that OER kinetics are determined by the density of electrochemically oxidized Co sites at or near the surface which form intermediate states with adsorbed or rather discharged O and OH species. They respectively associated the potential-dependent pseudo-capacitance of cobalt spinel films as a measure of such (re−)chargeable surface-near regions. Recent findings describe the active surface of (initially) crystalline Co3O4 as a sub-nm thick shell which reversibly transforms into an X-ray amorphous CoOx(OH)y,44 and show that lattice oxygen contributes to evolved O2.45 The consequently indicated 3-dimensionality of the surface-near oxide region involved in the OER was further confirmed by operando-XRD.46 Haase et al. also found a higher amount of transformed surface for the MACP spinel (pH 8.5) compared to the inactive conventional spinel.40 In this context, the increased depth of the anodically and cathodically active surface-near region observed for the more OER-active MACP spinels (Figure 10) may explain the lower overpotentials.
However, there is no clear trend observed for the MACP LDHs (Figure 10). Due to the different structure and with that less structural stability in regard to applied potential,3d the LDHs show significantly higher amount of irreversible transformation during the initial cycling of the pre-conditioning. The changes of the activity compared to the as-synthesized state are more significant and, thus, the materials properties and phases are less well-known. Furthermore, the pore morphology and size distribution may affect the material transformation and complicate interpretation due to ohmic drop within pores, altered mass transport conditions, as well as wetting properties.47 However, considering the three MACP LDH samples synthesized at pH≥8.5 which exhibit significantly higher overpotentials, it seems that these effects cannot fully account for the diminished OER-activity due to the similar pre-conditioning behaviour of all LDH samples. As seen in Figure 10, the ratio of anodic charge of the 50th and first pre-conditioning CVs as well as the reversible anodic charge of the 50th CV differs only minorly within the series of MACP LDHs.
Therefore, electrochemical impedance spectroscopy (EIS) was employed to better understand the considerably higher overpotentials of the three MACP LDH samples synthesized at pH ≥8.5. Figure 11 compares the Tafel slopes and EIS results for the conventional and the MACP LDHs as well as the MACP spinel sample synthesized at pH 8.5. The analysis demonstrates that the higher overpotential of the MACP LDH is accompanied by a considerably increased charge transfer resistance (R3 in Figure 11c) and a higher Tafel slope. Both suggest a change in the OER mechanism concerning the rate-determining step. The important role of the R3 resistance is further emphasized by increased values for those three samples in the MACP LDH series exhibiting higher overpotentials (pHSynthesis≥8.5, Figure S23b).
a) Tafel slopes from a 50 mV region where the charge transfer coefficient alpha is constant. b) Electrochemical impedance data in complex plane representation (Nyquist plot) and corresponding distribution functions of relaxation times γ(
) (DRT, inset in b) for cyclovoltammetrically pre-conditioned LDHs and the MACP spinel synthesized at pH 8.5, obtained at 1.60 V vs. RHE and c) resistance and capacitance values determined from DRT analysis. The results for the complete conventional and MACP LDHs series as well as the whole MACP spinel series can be found in Figure S23.
As reported previously,48 EIS data of cobalt oxide/hydroxide surfaces where OER proceeds allows to distinguish several sub-processes of the electrocatalytic reaction. This is due to changing oxidation states of accessible cobalt sites when forming intermediates, which can be described as pseudo-capacitive (dis−)charging of the oxide/hydroxide film through faradaic reactions with hydroxide and oxygen species. So, notably different time constants are present and, hence, allow to separate three charge transfer resistances parallel to double layer or pseudo-capacitors (inset, Figure 11b), because of the order-of-magnitude-different capacitance values. In the investigated frequency range (40 kHz–2 Hz), the LDH samples exhibit a fourth signal for low frequencies which may be related to interlayer mass transport not described by semi-infinite diffusion.49
Accordingly, the equivalent circuit resistor R1 (see Supporting Information detailed dicussion) in parallel to the double layer capacitor C1 is assigned to the OH− association step, which charges the pseudo-capacitor C2 with R2 representing the subsequent charge-transfer process of the reaction mechanism. To this end, R3 corresponds to the electrochemical step where oxygen is released. This assignment was further supported by recording EIS spectra in oxygen-purged and argon-purged KOH solution where the oxygen concentration in the electrolyte mostly affected the R3 resistance. Relating the EIS investigation to the structural and morphological properties of the MACP LDH series, it is apparent that the increased resistances for the oxygen release step at pHSynthesis≥8.5 are observed for bigger pore sizes with a threshold of 7.9 nm (Figure 8d) before the pre-conditioning. Thus, we hypothesize the formation of oxygen gas bubbles in these pores of the MACP LDHs to cause the diminished electrocatalytic activity, with the above-mentioned pore size being the lower limit for this phenomenon.
The evolution of gas bubbles inside the pores may not occur, if the pores are too small and therefor the critical radius for bubble formation cannot be met. For H2 gas bubbles, the radius of curvature for nucleation was found to be around 5 nm on a Pt nanoelectrode, being in the expected order of magnitude.50 The concentration profiles of dissolved oxygen in the pores depend on the diffusion characteristics transporting oxygen away from the electrode. For large enough pores, gas bubbles may form in the pores and block the connection to the bulk electrolyte. Thus, the concentration of dissolved oxygen will build-up in the pores leading to an increased charge transfer resistance for the oxygen release step. The mass transport of produced oxygen from the catalyst to the bulk electrolyte was also identified to limit the electrocatalytic activity for single cobalt spinel nanoparticles impacting an inert target electrode.51
In sum, the presented synthesis route for cobalt iron LDHs and spinels with tailored mesostructure allowed a detailed investigation of the effects on the electrocatalytic activity for oxygen evolution. The cyclovoltammetric pre-conditioning treatment, allowing for reproducibility and stable electrocatalytic performance, was found to cause material transformations in the near-surface regions. This leads to alignment of the electrocatalytic OER-activity whereas the amount of reversibly electrochemical transformable material fraction varies. Thus, relatively comparable overpotentials and highest activities can be explained but another phenomenon must account for the significantly higher overpotentials observed for three samples of the MACP LDH series. Based on electrochemical impedance spectroscopy in conjunction with detailed mesostructure characterization, we suggest the formation of gas bubbles within larger pores as explanation.
Conclusion
Two series of cobalt iron layered double hydroxides (LDHs) and their corresponding Co2FeO4 spinels were synthesized and used to study the effect of mesostructure on oxygen evolution reaction (OER) performance. The first series was synthesized via conventional co-precipitation in aqueous medium, varying the pH during synthesis in 0.5 pH steps. The second series was derived from microemulsion-assisted co-precipitation (MACP), confining the reaction space within reverse micelles, with otherwise equal synthesis conditions. The systematic variation of the synthesis pH did not lead to significant differences in properties of the LDHs from conventional co-precipitation such as composition, crystallinity, surface area, pore size or their platelet morphology. The spinels formed upon thermal treatment also exhibited comparable characteristics. For the more isotropic and nano sized MACP LDHs on the other hand, a systematic increase of the comparably small pore size with increasing synthesis pH was observed, which was preserved after thermally forming the corresponding spinels. Based on this synthesis parameter-mesostructure relationship, it was possible to tailor the pore size of the as-synthesized LDHs and spinels likely due to the modified exchange between the reverse micelles in the reaction mixture with varying the precipitation pH.
The possibility of tuning the pore size of such materials qualified them for the investigation of confinement effects in the electrochemical oxygen evolution reaction. For both LDH series low overpotentials and satisfying stability were found as well as for the MACP spinel series. The conventionally prepared cobalt iron spinels on the other hand showed extremely diminished activity. The cause for this extreme difference in activity was thoroughly investigated for one example from the series elsewhere.40 For the conventionally prepared samples, no trend of overpotential with mesostructural features was evident. However, the MACP LDHs and corresponding spinels showed a clear effect of the pore size on the overpotential. The MACP LDHs synthesized at pH≥8.5 have pore sizes larger than 7.8 nm and with that the largest pore sizes of the MACP derived samples. At the same time, they exhibit the highest overpotentials. All other samples from these two series exhibit smaller pore sizes and cluster in the range of lower overpotential.
Both observations, the surprisingly similar performance of all samples with smaller pores, and the diminished activity of those with lager pores can be explained by the electrochemical results. The performed pre-conditioning in the pre-OER region showed gradually changing CV shapes, which aligned towards the 50th cycle, indicating similar surface properties despite the different starting states of the as-synthesized catalysts. Additionally, the highly active MACP spinels showed the highest portion of reversibly transforming catalyst material during cycling and can be associated with the largest extent of electrochemical changes during pre-conditioning. In light of literature reports on the in situ formation of highly active near-surface states, this can explain the lowest overpotential observed for these samples by the highest number of active species.44, 46 The striking high overpotential for the MACP LDHs synthesized at pH≥8.5 was investigated by electrochemical impedance spectroscopy (EIS), revealing an increased charge transfer resistance and a higher Tafel slope for these samples, hinting towards a change in OER mechanism concerning the rate determining step. For all samples with diminished activity, the R3 resistance, which we assume to correspond to the oxygen release step, was found to be increased. Interestingly, only samples with pore sizes larger than 7.8 nm (before pre-conditioning) show the increased resistance R3, suggesting that gas bubble formation within these larger pores is possible, which leads to the observed increased overpotentials. The gas bubbles hinder the transport of oxygen into the bulk electrolyte and with that reduce the activity of the catalyst. If pores are smaller than the observed threshold, however, gas bubbles may not form because the critical radius for bubble formation is not met.
Experimental Section
All syntheses of layered double hydroxides were carried out in an automatic lab reactor system (OptiMax 1001, Mettler Toledo). The aqueous conventional, i. e., without microemulsion, co-precipitation was adapted from a recently published paper by our group.5 125 mL metal salt solution containing 0.266 mol L−1 Fe(NO3)3 ⋅ 9 H2O (≥98 %, Sigma-Aldrich), 0.533 mol L−1 Co(NO3)2 ⋅ 6 H2O (≥98 %, Carl Roth), and the precipitation agent consisting of 0.6 mol L−1 NaOH (98.5 %, VWR) and 0.09 mol L−1 Na2CO3 (≥99.5 %, Carl Roth) were simultaneously dosed into the reactor prefilled with 200 mL of distilled water over 1 h. The metal salt solution was dosed continuously through gravimetric control. The constant pH (pH 7.5 to pH 9.5, 0.5 pH unit steps) was controlled by an InLab Semi-Micro-L pH electrode and kept at a fixed value by automatically dosing the precipitation agent. pH 7.5 was chosen as the lowest co-precipitation pH to ensure complete precipitation of Fe3+, which requires at least slightly alkaline conditions. The temperature was held at 50 °C. Afterwards, the precipitate was aged for 1 h at 50 °C without further pH control. Thereupon the reactor was cooled to room temperature, the dispersion was washed repeatedly with distilled water until the conductivity of the supernatant was below 100 S μm−1 and subsequently dried for at least 12 h at 80 °C in static air.
The microemulsion-assisted co-precipitation was done like the procedure described above. The aqueous phases, more precisely the prefilled water template, the metal salt solution and the precipitation agent were each integrated into a water-in-oil microemulsion. 50 mL of a 0.133 mol L−1 Fe(NO3)3 ⋅ 9 H2O and 0.266 mol L−1 Co(NO3)2 ⋅ 6 H2O solution were used as the metal salt solution, and the alkaline precipitation agent contained 0.15 mol L−1 NaOH and 0.0225 mol L−1 Na2CO3. For preparation of the microemulsions Triton X-100 (≥99 %, Carl Roth), 1-hexanol (≥98 %, Carl Roth), cyclohexane (p.a., Fisher Scientific) and the corresponding aqueous phase were mixed and stirred until a clear solution formed, indicating stable microemulsions. The composition of the microemulsion resulting in 8.7 % aqueous phase by volume was adapted from literature.24 Afterwards co-precipitation and aging were carried out as described above. To remove the surfactant the dispersion was washed with acetone for five times and with ethanol for ten times. Subsequently, the precipitate was dried for at least 12 h at 80 °C in static air.
The cobalt iron spinels were obtained by calcination at 400 °C (2 K min−1) for 3 h in a muffle furnace (Nabertherm LE 6/11/B150).
PXRD was performed with a Bruker D8 Advance diffractometer with a Cu X-ray source in Bragg-Brentano geometry, using a LynxEye XE−T detector. The samples were dispersed in ethanol on a PMMA sample holder and diffraction patterns were recorded in the angular range from 5° to 90° 2θ with a step size of 0.01° and a counting time of 1.5 s. During the measurement, the sample holder was slowly rotated.
SEM was conducted with an Apreo S LoVac (Thermo Fisher Scientific). Prior to the measurements, the samples were sputtered with Pt/Au.
STEM measurements of the dried-out micelles were conducted on a JEOL JEM-2800 electron microscope. Used electron source is a Schottky electron gun, working at 200 kV acceleration voltage. The system exhibits a point-to-point resolution of 0.14 nm. Microemulsions were directly drop cast on a lacey carbon Cu grid from PLANO and treated with a mild plasma for STEM measurements. Plasma treatment of the prepared TEM grids was conducted on a Model 1070 nanoclean system at a power of 25 W in a 25 : 75 oxygen-argon mixture at a pressure of 30 mbar for 15 s.
UV-Vis spectroscopy was performed with a Varian Cary 300 UV-Vis.
The ratio of the incorporated metal cations was determined by atomic absorption spectroscopy (Thermo Electron Corporation, M-Series) of the as-prepared precursors. The carbon and hydrogen contents were determined by CHNS analysis. For each sample twofold measurements were performed.
TGA of the as-prepared precursors was carried out with a NETZSCH STA 449 f F3 Jupiter (NETSCH GmbH, Germany). The mass loss was recorded as a function of temperature with a linear heating rate of β=5 K min−1 in a temperature range from 30 °C to 1000 °C and a gas stream of O2 (21 mL min−1) and Ar (79 mL min−1).
BET surface areas were measured using N2 physisorption at 77 K with a NOVA3000 (Quantachrome GmbH, Germany). Prior to the measurements, the samples were degassed at 80 °C in vacuum for 2 h. Total pore volume and the pore size distribution were determined by applying the BJH method, which was applied to the desorption branch of the isotherm. Because the BJH analysis may underestimate the contribution of narrow mesopores, alternative evaluation using the NLDFT (non-local density functional theory) method was also used and found to qualitatively confirm the observed trends with variation in the absolute pore diameters around the value determined by BJH depending on the model used. We therefore regard the observed trend as reliable and base our discussion on the BJH method. The micropore volume was determined with the method by Horvath and Kawazoe.52
IR spectra were recorded with a Bruker Platinum ATR Diamond spectrometer with a spectral range of 400–4000 cm−1 with a step size of 0.1 cm−1.
Dynamic light scattering was performed with a Zetasizer Nano ZS (Malvern Instruments Ltd.). For each microemulsion threefold consecutive measurements were performed at a scattering angle of 173° with a He−Ne laser with a wavelength of 633 nm. The stated results are the mean output values for three consecutive measurements of the same sample, which have been rounded to a full nanometer value.
via Equation 1:
(1)Electrochemical impedance spectra were obtained at 1.60 V vs. RHE in the frequency range from 200 kHz to 1 Hz using a sine wave amplitude of 10 mV. The data were recorded after the LSV for OER investigation at 1 mV s−1 after an equilibration time of 60 s. Data checking and processing was performed using RelaxIS 3 (rhd instruments), applying a linear Kramers-Kronig test and a Z-HIT algorithm (40 kHz to 2 Hz range). Determination of the distribution of relaxation times (DRT)53 was based on radial basis functions (RBF) with a discretization factor of 10−4 and terms proportional to the second derivative of RBF included as extra penalty in the sum of squares minimization. OriginPro 2019 (OriginLab) was used to fit DRT functions γ(lnτ), determining resistance and capacitance values. Further, equivalent circuit fitting was performed with circuit models comprising two to four parallel RC elements (see Figure S22b) using suitable weighting modes.48
Chronopotentiometry was performed with a three-electrode setup in 1 mol L−1 KOH. A platinum counter electrode, an Ag/AgCl/3 m KCl reference electrode and a glassy carbon electrode loaded with the catalyst as the working electrode were used. For preparation of the ink 2.5 mg catalyst, 249 μL H2O, 249 μL ethanol and 2 μL Nafion were mixed and sonicated for 15 min. The glassy carbon electrode with a geometric area of 0.0707 cm2 was polished, loaded with 1.4 μL of ink, and dried for 1 h. A current density of 10 mA cm−2 was applied for 2 h and the corresponding potential was recorded. The electrode was rotated at 4000 rpm during the measurements.
Acknowledgments
This research was funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) in the framework of the Collaborative Research Center/Transregio “Heterogeneous Oxidation Catalysis in the Liquid Phase”–388390466-TRR 247 (Project C01, A02, and A09), the Mercator Research Center Ruhr (MERCUR, Pe-2018-0034), and the internal support of the University Duisburg-Essen for early stage researchers (“Förderung des exzellenten wissenschaftlichen Nachwuchses”). Furthermore, the authors would like to thank Dr. Kateryna Loza (SEM), Benjamin Mockenhaupt (BET), and Robin Meya (AAS and UV-Vis spectroscopy) for their assistance in the experimental work. Open Access funding enabled and organized by Projekt DEAL.
Conflict of interest
The authors declare no conflict of interest.
Open Research
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request.
