Synthesis and Characterization of Mn₃O₄:M Nanoparticles

To study the effect of the incorporation of redox-inactive metal substituents into oxides on their activity for ORR, we have selected the Mn₃O₄ oxide family as a model system owing to the ability of the spinel structure to readily accommodate different elemental substitutions,³² the high activity of Mn₃O₄ for ORR,^{30, 33} and the fact that the bulk Mn₃O₄ stoichiometry does not undergo noticeable changes at ORR potentials, at least over relatively short times.³⁰ The metal substituents for the Mn₃O₄:M (Figure 1a) series (M²⁺=Sr²⁺, Ca²⁺, Mg²⁺, Zn²⁺, Cu²⁺) were selected to cover elements with pK_a([M(H₂O)_n]²⁺) and enthalpies of MO oxide formation (Δ_fH°) (see Table S1 in the Electronic Supporting Information, ESI) that exceed and fall below the respective values of Mn²⁺ (i.e., pK_a([Mn(H₂O)_n]²⁺) and Δ_fH°(MnO)). The Mn₃O₄:M nanoparticles were prepared by colloidal synthesis following a literature procedure³⁴ using a nominal mass loading of 2.5 at.% of the M substituent with respect to the manganese content. For comparison, Mn₃O₄:Sr nanoparticles with a higher loading (i.e., 5 at.% of Sr) were also synthesized. Note that although increasing concentrations of substituent will likely have even stronger impact on the activity, higher substituent concentrations were not explored in Mn₃O₄:M series, as in this case the synthesis yielded particles of irregular size and shape distribution that would not allow for the reliable comparison of intrinsic activities and tracking of the reactivity trends, which was the main goal of this study (rather than finding the optimum composition). Barium substitution (Ba²⁺ is the only M²⁺ ion with basicity exceeding that of Sr^2+[35]) was not investigated due to the large mismatch of the ionic sizes of Ba²⁺ and Mn²⁺ making incorporation of Ba²⁺ into Mn₃O₄ lattice challenging. Below, we focus on a detailed structural analysis of Mn₃O₄:Sr and Mn₃O₄:Cu, as these compositions demonstrate the highest and lowest ORR activity, respectively, within the Mn₃O₄:M series.

Annular dark-field scanning transmission electron microscopy (ADF-STEM) and high-resolution transmission electron microscopy (HR-TEM) imaging of the as-synthesized nanoparticles (Figure 1b–d) revealed their high monodispersity with an average particle size of the different M substituted nanoparticles in the range 3–5 nm (Figure S1, Table S2). Figure 1d (left panel) shows an atomic resolution ADF-STEM image of a Mn₃O₄:Sr nanocrystal oriented along the <100> zone axis. The bright spots correspond to atomic columns of Mn and Sr, whereas the oxygen sub-lattice was not registered by the ADF detector due to its low scattering power and the different angular distribution of the scattered electrons. Direct identification of the Sr sites proved to be challenging, as we could not observe any noticeable contrast change in the local atomic rows corresponding to the M²⁺ sites due to the low amount of Sr incorporated and its high dispersion. The atomic arrangement is consistent with a spinel model for the cation sub-lattice as visualized by the overlay of the ADF-STEM image with the respective structural model (Figure 1d, right panel). Along the <100> zone axis, the spinel structure contains close-packed planes with [MnO₆] octahedral (Mn_oct) layers alternating with mixed [MnO₆] and [MnO₄] tetrahedral (Mn_tet) layers. The detailed ADF-STEM analysis of the particle surface (Figure 1d, S2) revealed that the majority of the terminating layers were composed exclusively of Mn_oct. Similar terminations of closed packed layers with cations in octahedral positions have been observed previously for several isostructural systems, such as γ-Ga₂O_3, γ-Fe₃O₄ and γ-Al₂O₃.^36-38 In line with experimental observations, DFT calculations (see discussion below) indicate that the surface terminated by Mn_oct is thermodynamically the most stable surface.

X-ray diffraction (XRD, Figure S3) and electron diffraction (Figure 1e) patterns also confirm the presence of a nanocrystalline spinel structure and the absence of any detectable amounts of secondary phases. The low content of M incorporated into the spinel structure (ca. 2.5 at.% from inductively coupled plasma (ICP) spectroscopy analysis, see Table S2) precludes a noticeable shift of the diffraction peaks. In addition, energy dispersive X-ray spectroscopy (EDX) of the sample Mn₃O₄:Sr evidenced a homogeneous distribution of the elements in the synthesized material (Figure S4), providing further evidence for the successful incorporation of M²⁺ ions into the Mn₃O₄ spinel structure. Additionally, Raman spectroscopy that is sensitive to variations of the manganese oxides composition³⁹ confirmed the formation of the Mn₃O₄ spinel structure (Figure S5).

A comparison of the pair distribution functions (PDF) obtained from electron diffraction data is shown in Figure 1f. The PDF peaks located at 1.44 Å and 2.48 Å are due to the presence of an amorphous carbon film in the TEM sample grid. The first pair correlation peak at ca. 2 Å is due to Mn−O distances, the second PDF peak at ca. 3 Å is due to edge-sharing Mn³⁺−Mn³⁺octahedra, while the third and the fourth PDF peaks (3.45 Å and 3.75 Å) are mainly due to the contribution from Mn−Mn corner-sharing polyhedra. Notably, the PDF peak positions and their intensities ratio are the same for Mn₃O₄, Mn₃O₄:Sr, and Mn₃O₄:Cu indicating that the local atomic arrangement remained unperturbed by the substitution of Mn by M.

X-ray absorption spectroscopy (XAS) on the Mn K-edge was applied to probe the average oxidation state and local environment of the manganese sites in Mn₃O₄:M. The shape of the white line in the X-ray absorption near edge structure (XANES) spectrum of Mn₃O₄:M nanoparticles showed only slight differences when compared to that of the commercial bulk Mn₃O₄ reference (Figure S6). We observed a small yet distinguishable (ca. 0.2 eV) shift of the position of the absorption edge towards high energies for all M-substituted Mn₃O₄:M materials with respect to the unsubstituted Mn₃O₄ nanoparticles, as well as the rise of the white line upon substitution (Figure 1g), consistent with the replacement of Mn²⁺ sites in the spinel structure with M²⁺ ions and an associated increase of the average oxidation state of manganese (owing to an increased ratio of Mn³⁺:Mn²⁺ sites). In line with this argument, the smallest change in the edge position and increase in the intensity of the white line was observed for the Mn₃O₄:Cu sample that contained the smallest quantity of the substituent in Mn₃O₄:M series (1.0 at.%, Table S2). A slight decrease in the pre-edge peak for the M-substituted materials suggests an increase in the average symmetry of the sites, possibly due to the increased fraction of Mn_tet sites (that are less distorted than the Mn_oct sites).

Extended X-ray absorption fine structure (EXAFS) analysis was used to probe the local environment of the manganese sites in the Mn₃O₄:M materials (Figure 1h, S7). The fitting of the experimental EXAFS data (Figure 1h) for the Mn₃O₄:M series is in good agreement with the crystallographic model of the Mn₃O₄ spinel⁴⁰ (Figure 1a) and our PDF analysis (Figure 1e). The structural model consists of two Mn sites, in tetrahedral (Mn_tet) and octahedral (Mn_oct) coordination with O, which were used to generate two series of scattering paths. Fixing the coordination numbers, allowed us to determine the Mn−O (within the polyhedra) and Mn−Mn (between the polyhedra) distances. The first coordination shell is composed of three Mn−O scattering paths corresponding to Mn_tet−O and two non-equivalent Mn_oct−O subshells that are due to the highly distorted octahedra, while the second sphere is dominated by Mn_oct−Mn_oct and Mn_oct−Mn_tet paths (Table S3, Figure S7). Although, the element substitution is expected to affect the covalency of the Mn−O bond (hence also the bond length) due to the inductive effect of M,⁴¹ structural changes associated with the substitution of Mn by M are likely too subtle to be traced by EXAFS analysis: indeed, for the entire Mn₃O₄:M series, all Mn−O bond lengths are virtually undistinguishable within experimental uncertainty (0.01–0.03 Å). Similarly, small changes in the XANES caused by the substitution of Mn²⁺by M²⁺ (i.e., shift of the edge position) make it challenging to quantify the effect of differences in the electron withdrawing strength of M on the electronic states of manganese. In line with the Mn K-edge XAS analysis, Mn 2p and Mn 3p core levels X-ray photoelectron spectroscopy (XPS) (Figure S8, S9) and O K-edge XAS spectroscopy (Figure S10) do not reveal any pronounced effect of substituent M in Mn₃O₄:M on the Mn states.

Despite the pronounced stability of the bulk Mn₃O₄ structure, evidenced by Mn K-edge XAS measurements,³⁰ some studies report on the instability of Mn₃O₄ surface during electrochemical cycling at ORR potentials in harsh alkaline environment,⁴³ as well as effect of Nafion binder used for the preparation of the catalyst inks on the Mn electronic states.⁴⁴ However, for Mn₃O₄:M materials studied here, XPS analysis of the samples deposited from the Nafion-containing inks did not evidence any noticeable differences in the envelope of the Mn 3p core level region compared to the pristine samples (Figure S9) indicating low reactivity of Mn₃O₄:M towards Nafion. Interestingly, Mn²⁺/Mn³⁺/Mn⁴⁺ ratio (oxidized Mn⁴⁺ species are present on the surface of all as-synthesized Mn₃O₄:M materials, in line with other reports on XPS analysis of Mn₃O₄ oxides^{43, 45}) did not vary over 20 CV cycles (0.4 V_RHE–1.0 V_RHE, 100 mV s⁻¹, 0.1 M KOH) (Figure S9, Table S4), providing evidence for the high stability of the spinel structure over the period of the catalytic measurements. Selected area electron diffraction (SAED) patterns of the Mn₃O₄:M materials after potential cycling (Figure S11) confirm the presence of a nanocrystalline spinel structure and the absence of any detectable amounts of secondary crystalline phases, whereas TEM analysis (Figure S12) does not reveal any noticeable changes in particles morphology, size, or crystallinity indicating that surface changes (if any) can only be confined to surface atomic layer(s). (The long-term stability of Mn₃O₄:M is however an issue as the ORR activity reduces significantly after 1000 CV cycles, Figure S13).

XPS measurements of the O 1s core levels allowed us to confirm the inductive effect of M as evidenced by a shift of the binding energy of the lattice oxygen component (by ca. 0.2 eV) to lower binding energies when comparing Mn₃O₄:Sr to Mn₃O₄:Cu (Figure S14). This shift is indicative of a larger negative charge on the oxygen sites in Mn₃O₄:Sr compared to Mn₃O₄:Cu, which is in line with an increased electron donating strength (i.e., a reduced Lewis acidity) of strontium with respect to copper (the trend holds well for the entire substitution series, Figure S15).

Overall, our detailed structural characterizations of the series of Mn₃O₄:M nanoparticles demonstrate the formation of phase-pure, M-substituted Mn₃O₄ oxides. The incorporation of M had a negligible effect on the crystal structure, the nanoparticle morphology and the manganese valence states, making this series of materials a reliable model system to decouple and investigate the individual effect of the different M substituents on the electrocatalytic activity of the material.

Oxygen Reduction Activity of the Mn₃O₄:M Series

The ORR activity of the series of Mn₃O₄:M nanoparticles was assessed in a conventional three-electrode electrochemical setup using 0.1 M KOH as the electrolyte, whereas the H₂O₂/H₂O selectivity was estimated using rotating ring-disk electrode measurements (RRDE, see Supporting Information for the experimental details). Intrinsic ORR activities were assessed using steady-state galvanostatic measurements (to eliminate the contribution of redox processes within the bulk of the electrode material, which can interfere with the measured catalytic currents in conventional CV studies,⁴⁶ Figure S16). The currents were normalized to the geometric surface area of the nanoparticles estimated by TEM (see Figure S1, Table S2). As the CV curves for Mn₃O₄:M samples do not feature a well-defined limiting current plateau, our analysis focused on the current range that was much smaller than the mass transport limit, where the mass transport correction can be neglected (for Tafel plots for extended current range, see Figure S17).

The Mn₃O₄:M nanoparticle series studied here feature mass and electrode surface area normalized activities (Figure S18) that are comparable to previously reported activities of MnO_x catalysts.⁴⁷ The specific activities (per oxide surface area) are similar to other high surface area MnO_x oxides (ca. 1 μA cm_oxide⁻² at 0.8 V_RHE),^{48, 49} but below the activities typically reported for non-porous highly crystalline materials (>10 μA cm_oxide⁻² at 0.8 V_RHE).⁴⁷ Remarkably, we found that the measured ORR activity depends strongly on the type of the M substituent. Indeed, incorporation of 2 at.% of Sr increased the ORR activity by almost an order of magnitude compared to Mn₃O₄:Cu (Figure 2a).

Importantly, as described above, this was achieved without changing the crystal structure, particle morphology, nor the valence state of manganese. This suggests that there exist direct correlations between certain intrinsic properties of the substituting elements and the catalytic performance of Mn₃O₄:M. In fact, the influence of the incorporation of redox-inactive substituents on the redox properties of biological systems⁵⁰ and molecular complexes has been reported^51-55 and the observed trends have been rationalized typically by an inductive effect triggered by the substituent/dopant altering the covalency of the bonds between the parent metal centers and adjacent oxygen.⁴¹ Although many studies have explored the effect of metal substitution in oxides (e.g., perovskites⁵⁶) on their catalytic performance, changes in the catalytic activity have been attributed typically to changes in the valence state of the active site (for aliovalent substitution),^57-59 changes in the crystal structure of the oxide,⁶⁰ or the introduction of crystallographic strain,⁶¹ i.e., effects originating from the mismatch of the size or the formal charge of the substituent and the substituted site (note that in this context, we do not consider systems for which the substitution results in a change of the reaction site, e.g., from Fe to Mn in Mn_xFe_3−xO₄⁶²). (Specifically for Mn₃O₄, metal substitution was used as a tool to tailor electrocatalytic activity.^{43, 63, 64}) On the other hand, there are very few experimental works that aim to leverage electronic, or inductive, effects arising from the incorporation of substituents to rationally tailor the activity of electrocatalysts (examples can be found in references^65-68). Hence, establishing a unified description of the effect conferred by the introduction of redox-inactive metals into a host oxide on their catalytic ORR/OER activity (irrespective of their structure) would provide a new perspective (and experimental tool) for the design of materials for targeted electrochemical transformations.

Reactivity Trends: Theoretical Considerations

Next, we aim to assess the correlation(s) between the ORR activity of the Mn₃O₄:M nanoparticles (Figure 2a) and the properties of the substituting element M. First, we observed a strong correlation between the ORR activity of Mn₃O₄:M and the Lewis acidity of the substituting M²⁺ ions (expressed as pK_a of the corresponding hydrated ion, [M(H₂O)_n]²⁺, Figure 2b), implying that the Mn−O bond covalency (tuned by the inductive effect of M) controls the catalytic activity of the material. As discussed above, XPS O 1s core level measurements confirmed the inductive effect of M reflected in a larger negative charge on the oxygen sites in Mn₃O₄:Sr compared to Mn₃O₄:Cu owing to the greater electron donating strength (i.e., a reduced Lewis acidity) of strontium with respect to copper. This robust correlation between the ORR activity and the pK_a([M(H₂O)_n]²⁺) provides clear evidence for the relevance of the Lewis acidity of the redox-inactive substituent on the electrocatalytic activity of Mn₃O₄ (note that in a related case, the oxygen exchange kinetics of Pr_0.1Ce_0.9O_2-δ was also shown to depend on the acidity [on the Smith scale] of the surface-infiltrated oxides species,⁶⁹ however the catalytic activities in ORR/OER reported for a wide row of oxides did not demonstrate a correlation to the Smith acidities of the respective oxides⁷⁰). Thus, to construct a clear and direct link between the properties of M and the catalytic activity of an oxide in terms related to the mechanism of oxygen electrocatalysis, we further analyzed the activity and selectivity (Figure 2c) trends.

As the ORR mechanism includes the formation/cleavage of metal-oxygen bonds (whereby their binding strength controls the catalytic activity),¹³ it is expected that the energy of oxygen vacancy formation [E(O_vac)] also scales with catalytic activity.²⁹ Indeed, the calculated values for E(O_vac) (for O atoms bound to octahedrally coordinated Mn atoms serving as the active sites in spinel oxides⁷¹) in the vicinity of M in a Mn₃O₄:M model slab (Figure 3a) demonstrate a good correlation with the experimentally measured activities (Figure 3b), hence confirming a relationship between the catalytic activity and the oxygen binding strength on the catalyst surface. As a consequence, we observe an excellent correlation between the ORR activity of Mn₃O₄:M nanoparticles and the experimental enthalpy of formation of the binary oxide, MO (Δ_fH°(MO), Figure 2b), which is rooted in the dependence of the binding strength of the oxygen species on the oxide surface on Δ_fH°(MO) (Figure 3c): incorporation of an element with a more negative Δ_fH°(MO) (i.e., a thermodynamically more stable oxide) increases the binding strength of oxygen species (*O/*OH/*OOH) on the surface of the multimetallic compound (Mn₃O₄:M). These findings are fundamentally distinct from the correlations involving oxide formation energies of metals representing the actual redox-active site discussed for binary oxides²⁵ or single-atom catalysts.^{72, 73}

The correlations established here demonstrate that the substitution/doping by redox-inactive elements can be used as a highly versatile tool to rationally tune the activity of complex oxide/hydroxide catalysts for ORR (and likely for the reverse OER), whereby the most appropriate choice of the substituent/dopant depends on the rate-limiting step of the catalytic reaction. Specifically for the commonly quoted associative mechanism of the oxygen reduction reaction,¹³ the reaction rate is limited by the adsorption of either *OOH (in case of weak oxygen binding) or *OH (strong oxygen binding).⁷⁴ Therefore, in the former scenario, the incorporation of elements M with a highly negative value of Δ_fH°(MO_x) would result in an increase of the oxygen binding strength (Mn−O), hence an activity enhancement would be expected. As shown previously, ORR on a Mn₃O₄ surface is indeed limited by *OOH adsorption (the least exothermic step at U=0),⁷⁵ implying that an increase of the oxygen adsorption strength would improve its activity. Our experimental results for the Mn₃O₄:M system (with M=Sr, Ca, Mg) confirm this rationale. Note that as little as 2.5 at.% of the substituent was sufficient to yield an increase in ORR activity by an order of magnitude (Mn₃O₄:Sr compared to Mn₃O₄:Cu); increasing the concentration of Sr substitution from 2.5 at.% to 5 at.% improved the activity of Mn₃O₄:Sr further (Figure S19). (As mentioned above, materials with higher concentrations of M were not studied, as in this case the synthesis yielded particles of irregular shape and size distribution.) As a likely indication of the general nature of trends observed for the Mn₃O₄:M family, analysis⁴¹ of the literature data on the ORR activity of M_xFe_3−xO₄,⁶² MCo₂O₄,⁷¹ and LnMnO₃⁷⁶ materials revealed similar linear correlations between ORR activity of the complex oxides and properties of M substituent (specifically pK_a([M(H₂O)_z]ⁿ⁺), as shown on Figure S20 (note those trends are not discussed in the original^{62, 71, 76} experimental reports). Furthermore, our calculations on a simple MnO₂:M model structure (two-dimensional birnessite layer) provide a further indication that oxygen binding energy in complex oxides correlates with thermodynamic characteristics (e.g., (Δ_fH°(MO_x)) of the substituent/dopant, independent of the particular structure of the parent oxide (Figure S21). Yet, further studies on model systems would be desirable to convincingly demonstrate the universality of the effect of the redox-inactive substituent on the electrocatalytic performance of the complex oxide materials.

Interestingly, the substitution of Mn by M does not only affect the ORR activity, but also the H₂O₂/H₂O (2e⁻/4e⁻) selectivity. Indeed, the Faradaic efficiency towards the 2e⁻ reduction (H₂O₂ synthesis) increased from ca. 5 % for M=Sr, Ca, Mg, Mn (catalytically more active oxides in the Mn₃O₄:M series) to ca. 8 % for M=Zn, and ca. 25 % for M=Cu at high overpotentials (Figure 2c). Such a selectivity trend was predicted for ORR catalysts characterized by a weak oxygen binding strength (i.e., the reaction is limited by O₂ adsorption to form *OOH, as in the case of Mn₃O₄⁷⁵), whereas a further decrease of the adsorption energy was proposed to enhance the 2e⁻ reduction selectivity.⁷⁷ This experimental observation provides additional evidence for the interplay between the properties of the redox-inactive substituent (e.g., Δ_fH°(MO_x)), the adsorption strength of catalytic intermediates (*O/*OH/*OOH), and the ORR activity and selectivity.

General Implications for Catalysts Design

Despite its simplicity, the concept described here (incorporation of redox-inactive elements into the structure of oxide to leverage their inductive effect) has not been studied systematically or considered as a general strategy for the rational design of oxide electrocatalysts. While previous studies probing correlations between the thermochemical properties of bulk oxides and their activity in ORR/OER electrocatalysis^{25, 26, 29} attempted to explain the reactivity trends for certain oxide families, our findings can be used to instruct the design of catalyst across different material families by the manipulation of their composition through the incorporation of substituent/dopant. It should be noted here that the incorporation of substituents/dopants into the lattice of the host oxide may introduce multiple effects, such as the generation of crystal lattice strain and distortion, modification of the crystal structure, change of spin or the valence state of the active site, etc., which can have counteracting effects on the catalytic activity, and therefore have to be considered as well. Yet, for a large variety of oxide families that are capable of accommodating a broad range of cationic species without leading to noticeable structural changes, and particularly for structurally non-rigid metal (oxyhydr)oxides, the concept introduced here offers a potent tool for the design of highly active materials for oxygen electrocatalysis, and can be applied to engineer compositions with benchmark performances starting from current state-of-the-art materials (e.g., RuO₂/IrO₂/Ni−Fe oxyhydroxides, etc.), however, this is yet to be demonstrated. Further work will attempt to verify whether this principle is general and can be transferred to a broader selection of materials and processes concerning oxygen adsorption and redox reactions involving oxygen, reaching beyond electrocatalysis, such as metal-ion batteries,⁷⁸ chemical looping,⁷⁹ or heterogeneous catalysis.⁸⁰