Electrodes with self-assembled monolayers (SAMs) have broad applications in the fields of sensing, catalysis, and energy storage and conversion. Understanding the electron transfer dynamics between the electrode substrate and the attached monolayer is crucial in designing interfaces for these applications (Kellon et al., 2019; Xie et al., 2019). Electrodes with tethered nanoparticles by the SAM feature low loadings and high atomic utilization of nanoparticles, both of which are desirable in the field of catalysis (Young et al., 2016; Shermukhamedov et al., 2020; Lachmanová et al., 2021). Chazalviel and Allongue (2011) established a theory in describing the relationship between the insulating layer (SAM) thickness and electron transfer dynamics in the substrate-SAM-nanoparticle, where the electron transfer rate is unhindered until a threshold thickness of the SAM is reached. Later, Hill et al., 2015 proposed a model to qualitatively calculate the current of the substrate-SAM-nanoparticle assembly by considering the electron tunneling, reaction kinetics, and mass transport (Hill et al., 2015). Evidences supporting these theories using gold nanoparticles as the electron mediator are extensively reported (Wang et al., 2010; Barfidokht et al., 2013; Kizling et al., 2018; Liu et al., 2019). However, gold nanoparticles are catalytically inert in most cases, and nanoparticles with catalytic activities are desirable to be assembled onto the electrode surface in catalysis.

The main bottle-neck of hydrogen production by electrochemical water splitting is the anodic oxygen evolution reaction (OER, 2 H₂O → 4 H⁺ + O₂ + 4 e⁻, E ^o = 1.23 V). The four coupled electron and proton transfer processes cause the sluggish kinetics of the OER, and active electrocatalysts and interfaces are required to reduce the large overpotential of the OER (Walter et al., 2010). The nanoparticle-attached electrodes have been constructed via SAM for OER. For example, iridium oxide nanoparticles (IrO_x) and ruthenium oxide nanoparticles (RuO_x) have been attached to the indium tin oxide-coated glass slide (ITO) surface via esterification reaction using polycarboxylic acids and pyrophosphoric acid as the linker, and the resulting electrodes are applied to catalyze the OER in acid (Gambardella et al., 2012; Tian et al., 2021). Consistent with the theory established by Chazalviel and Allongue (2011) the apparent electron transfer rates of the IrO_x-modified electrodes are unhindered using the linkers with short carbon-chains, and the electron tunneling rate does not limit the OER rate (Gambardella et al., 2012; Tian et al., 2021). Since iridium is a precious metal, interfaces with nonprecious metal nanoparticles are highly desirable to catalyze the OER. Spinel-type cobalt oxide nanoparticles (Co₃O₄) have been adopted to catalyze the OER in acid and are a promising candidate in replacing in part the precious metal catalysts (Lai et al., 2021; Natarajan et al., 2021). Co₃O₄ is usually electrodeposited on the substrate surface or is bonded to the substrate surface using Nafion (Liu et al., 2013). These methods inevitably lead to low percentage of utilization of electrochemically active Co atoms as the active sites, as some nanoparticles are not in direct contact with the electrolyte. The substrate-SAM-Co₃O₄ assembly offers a promising route in fine-tuning the surface structure of the catalytic active surface, with a maximum percentage of Co exposed. Therefore, we report the construction of the ITO-SAM-Co₃O₄ (SAM = dicarboxylic acid) assembly, and the constructed interface is active toward the OER in acid. The influences of SAM chain lengths and the average diameters of the Co₃O₄ on the electron transfer kinetics and on the electrocatalytic OER activities are investigated. Also, comparisons to the electrode with Nafion-bound Co₃O₄ are made, and the Co₃O₄ tethered by SAM show significantly enhanced number of electrochemically active Co sites and increased OER activity per active site.

Scheme 1 illustrates the procedure to fabricate the ITO-SAM-Co₃O₄ (SAM = dicarboxylic acid). In acetone solution containing the dicarboxylic acid, surface hydroxyl groups of ITO can react with the carboxylic acid groups to form ester bonds. Then, the ITO-SAM (SAM = dicarboxylic acid) is immersed in the acidic solution containing Co₃O₄ to allow the esterification between the adsorbed carboxylic acid groups and the hydroxyl groups of Co₃O₄. Hydroxyl groups on the Co₃O₄ surface have been observed previously (Anantharaj et al., 2019), and the interaction of carboxylic acid groups with the Co₃O₄ surface hydroxyl groups has also been reported (Kollhoff et al., 2018). Dicarboxylic acids with different numbers of carbon chains are adopted. The Co₃O₄ with various average diameters (3.5, 11, and 19 nm) are synthesized, and the TEM (Supplementary Figure S3) and XRD (Supplementary Figure S1) of the Co₃O₄ powders are shown in the Supplementary Information. All XRD patterns show diffraction peaks that correspond to the face-centered cubic phase of Co₃O₄ (JCPDS 09-0418) (Dong et al., 2007), and Co₃O₄ with larger average diameters show increased XRD peak intensities, which suggests their better crystallinities. The pH of the Co₃O₄-containing solution (pH = 2.0) is lower than the pKa₁ values of the polycarboxylic acids adopted [glutaric acid 4.34 (Canari and Eyal, 2003), 1,8-octanedioic acid 4.5 (Hullar and Anastasio, 2011), 1,16-hexadecanedioic acid 4.65 (Kanicky and Shah, 2002), and 1,20-eicosanedioic acid 5.5 (Mukerjee and Ostrow, 2010)]; therefore, carboxylic acid groups remain in the -COOH form to allow the esterification reaction to happen.

We first used cyclic voltammetry (CV) to probe the surface change during the fabrication of the ITO-Oct-Co₃O₄. Figure 1 shows the results. For the ITO-Oct-Co₃O₄, a couple of redox peaks are shown at ∼ 1.58 V, which can be assigned to the Co^3+/4+ redox couple in acid (Xiao et al., 2020; Han et al., 2021; Natarajan et al., 2021). The charge under the anodic Co^3+/4+ peak is integrated, and the number of electrochemically active Co atoms (Γ) is estimated to be 4.6 × 10⁻¹⁰ mol cm⁻² based on the integrated charge, assuming a 1 e⁻ transfer process. The total amount of Co on the ITO-Oct-Co₃O₄ is 3.31 × 10⁻⁸ mol cm⁻² acquired using ICP-AES. These values indicate that 1.4% of the Co on the electrode is electrochemically active. Starting at 1.65 V, the OER process happens. In comparison, the CV of the bare ITO and ITO-Oct is identical (Figure 1A inset) and lack the characteristic Co^3+/4+ redox peaks and the OER process. Electrodes prepared by simply immersing the ITO into the Co₃O₄ colloidal solution (ITO-Co₃O₄) give rise to a higher CV current than bare ITO and show redox features, but the current is still negligible compared to the ITO-Oct-Co₃O₄. We also observed a strong dependence of the coverage of the Co₃O₄ on the pH of the esterification reaction (Supplementary Figure S4), which implies that the ester bonds formed are the major driving force to anchor the Co₃O₄ to the surface, while the ionic interaction, hydrogen bonding, and chelation between the acid and the defected cationic centers contribute insignificantly to the surface bonding as they are less dependent on pH. Immersing the ITO-Oct in CoCl₂ solution (ITO-Oct-CoCl₂) leads to formation of Co²⁺ coordinated to the carboxylic acid groups at the surface, and the CV lacks the unique feature for Co^3+/4+ redox couple but shows a slight increase in the OER current at high potentials. These show that the Co₃O₄, rather than Co ionic species, are tethered to the ITO in case of the ITO-Oct-Co₃O₄. Figure 1B displays the CV of the ITO-Oct-Co₃O₄ at various scan rates in 0.1 M HClO₄. All the CV show distinct Co^3+/4+ redox peaks, and by plotting the log i _p versus log v (Supplementary Figure S5), the slope is close to 1, which implies that the redox active species are confined to the electrode surface (Silva et al., 2014).

Figure 2A shows the deconvoluted XPS spectrum of the Co 2p region of the ITO-Oct-Co₃O₄. The Co 2p_3/2 peak can be deconvoluted into two components at 781.5 and 779.5 eV, which originate from the Co²⁺ and Co³⁺ in the Co₃O₄, respectively (Zhang et al., 2018). The deconvoluted Co 2p_1/2 peak also shows the contribution from the Co²⁺ (796.8 eV) and Co³⁺ (794.7 eV) components. The deconvoluted C 1s spectrum in Figure 2B displays three peaks at 284.6, 286.1, and 288.2 eV, which correspond to the C-C, C-O, and C=O bonds, respectively (Dwivedi et al., 2015). The deconvoluted O 1s peak in Figure 2C shows three components at 529.4, 532.0, and 533.4 eV, which match the Co-O bond in the Co₃O₄, surface -OH structure, and the C-O and C=O bonds, respectively (Yang et al., 2009). The existence of C=O and C-O structures in both C 1s and O 1s spectra indicates the formation of ester bonds. The XRD pattern of the ITO-Oct-Co₃O₄ shows only crystalline peaks assigned to the ITO substrate (Supplementary Figure S2), suggesting the loading of crystalline Co₃O₄ is low. Both the XPS and CV show that the Co₃O₄ are successfully tethered to the ITO surface using the dicarboxylic acid as SAM.

Figure 3A compares the OER activity of the ITO-Oct-Co₃O₄ with bare ITO, ITO-Oct, ITO-Co₃O₄, and ITO-Oct-CoCl₂ in 0.1 M HClO₄ at 5 mV s⁻¹. Significantly higher OER current is observed at the high-potential region (>1.65 V) for the ITO-Oct-Co₃O₄, while the ITO and ITO-Oct exhibit negligible current in the similar region. This indicates that the attached Co₃O₄ is responsible for the observed OER activity. The OER onset overpotential of the ITO-Oct-Co₃O₄ is 374 mV, and the overpotential to reach 1, 5, and 10 mA cm⁻² current densities is 497, 562, and 570 mV, respectively. The Tafel plot and the corresponding Tafel slope value of the ITO-Oct-Co₃O₄ are displayed in Figure 3B. The Tafel slope value acquired is 70 mV dec⁻¹, consistent with the reported Tafel slopes of Co₃O₄ in acid (Mondschein et al., 2017; Han et al., 2021). The turn-over frequency (TOF) at 1.72 V is calculated based on Eq. 1 TOF = j / 4 F Γ (1) to be 4.06 s⁻¹. The key electrochemical parameters of the ITO-Oct-Co₃O₄ are summarized in Table 1.

We also prepared the Co₃O₄ (3.5 nm in average diameter bound by Nafion on ITO, and the amount of Co deposited on the ITO is controlled to 3.31 × 10⁻⁸ mol cm⁻², same as the ITO-Oct-Co₃O₄. Figure 4A shows the CV of the electrode at 50 mV s⁻¹ in 0.1 M HClO₄. Weak Co^3+/4+ redox peaks are shown centered at 1.63 V. By integrating the charge under the Co^3+/4+ anodic peak of the Co₃O₄/Nafion electrode, assuming a 1 e⁻ transfer process, the Γ is estimated to be 6.66 × 10⁻¹¹ mol cm⁻². This indicates that the ITO-Oct-Co₃O₄ exposes seven times higher amount of the electrochemically active Co atoms than the Co₃O₄/Nafion electrode with the same Co₃O₄ loading. This is consistent with previous observations using IrO_x and Au nanoparticles and can be caused by the inhomogeneous distribution of Nafion bound Co₃O₄ at the ITO surface, as aggregation is constantly observed on the Nafion bound nanoparticles (Moghaddam et al., 2015). In addition, different substrate–nanoparticle interactions and parallel interactions can account for the observed differences in the Γ. Figure 4B shows the LSV of the Co₃O₄/Nafion electrode. The OER process is also observed starting ∼480 mV overpotential, but the OER current at 1.72 V is significantly lower than that of the ITO-Oct-Co₃O₄, in part attributed to the lower Γ. The TOF at 1.72 V for the Co₃O₄/Nafion electrode is calculated to be 0.058 s⁻¹, 69 times lower than that of the ITO-Oct-Co₃O₄. Figure 4C shows the Tafel plot of the Co₃O₄/Nafion electrode. The Tafel slope value of the Co₃O₄/Nafion electrode is 165 mV dec⁻¹. This suggests sluggish OER kinetics of the Co₃O₄/Nafion electrode. Both the lower TOF and the higher Tafel slope value for the Co₃O₄/Nafion electrode indicate inferior electrocatalytic activity per active Co site for the electrode with Nafion-bound Co₃O₄. As we adopted the same Co₃O₄, the surface structure of the active sites is similar. The observed inferior activity per active Co site is probably caused by the hindered electron or proton transport for the OER process. This proposal agrees with the results acquired by Young et al. (2016), where molecular tether facilitates the Au nanoparticle-mediated electron transfer process (Young et al., 2016). Therefore, the dicarboxylic acid-tethered Co₃O₄ exposed significantly higher amount of electrochemically active Co atoms than the Nafion bound Co₃O₄, and the electrocatalytic OER activities of nanoparticles acquired using the Nafion as the binder can be a severe underestimation.

The surface modification process is further conducted using carboxylic acids with different chain lengths. Figure 5 A–C shows the CV at various scan rates of the ITO-Glu-Co₃O₄, ITO-Hex-Co₃O₄, and ITO-Eic-Co₃O₄, respectively, in 0.1 M HClO₄. All electrodes show distinct Co^3+/4+ redox peaks, which suggests the successful attachment of the Co₃O₄ to ITO. The CV at 50 mV s⁻¹ was used to estimate the Γ, and the results are summarized in Table 1. Figure 5D displays the LSV for the ITO-Glu-Co₃O₄, ITO-Hex-Co₃O₄, and ITO-Eic-Co₃O₄ from 1.3–1.9 V at 10 mV s⁻¹. All electrodes display electrocatalytic OER activities that originate from the Co₃O₄. The corresponding Tafel plots are shown in Figure 5E, and similar Tafel slope values for all three electrodes are observed, which suggests similar OER mechanistic pathways. The Tafel slope values, overpotentials at 1 and 5 mA cm⁻², and the TOF at 1.72 V of these electrodes are summarized in Table 1. The electron transfer rate constants (k) are analyzed using the peak separations from CV at various scan rates based on Laviron’s approach (Lavagnini et al., 2004; Lachmanová et al., 2021). The E _peak−E ^0’ is related to ln v by Eq. 2, E peak − E 0 ′ = − R T α n F ln ( α n F v R T k ) , (2) where E _peak is the redox peak potential, E ^0’ is the formal redox potential, α is the transfer coefficient, n is the number of electrons transferred, and other symbols have their standard meanings. By plotting the E _peak−E ^0’ against the ln v, the fitted line intercepts with the x-axis, and the k can be calculated based on Eq. 3. k = α n F v R T . (3)

Figure 5F shows the E _peak−E ^0’ versus ln v plot for the ITO-Glu-Co₃O₄, ITO-Hex-Co₃O₄, and ITO-Eic-Co₃O₄. The calculated k values are close and are summarized in Table 1.

By utilizing Co₃O₄ with different average diameters, ITO-Oct-Co₃O₄ (11) and (19) are constructed. Figures 6A,B display the CV at various scan rates in 0.1 M HClO₄ for the ITO-Oct-Co₃O₄ (11) and (19), respectively. Both electrodes show Co^3+/4+ redox peaks, and the Γ are calculated and summarized in Table 1. Figure 6C shows the LSV of these electrodes. Both electrodes are active toward OER, and the OER onset overpotential and overpotentials to reach 1 and 5 mA cm⁻² and the TOF at 1.72 V are all summarized in Table 1. Figure 6D shows the Tafel slope of these two electrodes. The Tafel slope values for the ITO-Oct-Co₃O₄ (11) and (19) are 112 and 84 mV dec⁻¹, respectively. Figure 6E shows the E _peak−E ^0’ versus ln v plots for the ITO-Oct-Co₃O₄ (11) and (19), together with the ITO-Oct-Co₃O₄ (the one with 3.5 nm average diameter. The calculated k values are summarized in Figure 6F and in Table 1.

From Table 1, we observe a dependence of k on the nanoparticle size, as k increases with the increased average diameter of the Co₃O₄. Based on the Marcus theory of electron transfer, the k is related to the reorganization energy (λ) and the extent of electron coupling. For an NP with a radius r, the λ is related to nanoparticle radius (r) by Eq. 4 (Chazalviel and Allongue, 2011), λ = q 2 4 π ε 0 r , (4) where q is the unit charge and ε ₀ is the vacuum permittivity. A larger r of the nanoparticle would entail a lower λ, which increases the rate of electron transfer. The dependence of k on the carboxylic acid chain lengths is not obvious as all k of the ITO-Glu-Co₃O₄, ITO-Oct-Co₃O₄, ITO-Hex-Co₃O₄, and ITO-Eic-Co₃O₄ are in the range of 1.10–1.14 s⁻¹. This phenomenon is consistent with Chazalviel’s theory describing the relationship between the insulating layer thickness and the k in the substrate–SAM–nanoparticle assembly (Chazalviel and Allongue, 2011). Based on their theory, nanoparticles with 3.5 nm in diameter in the substrate–SAM–nanoparticle assembly would require approximately 20 −CH₂- units in the carbon chain to reach the critical point, above which electron transfer would become hindered. In our case, no hindered electron transfer is observed using carboxylic acids with 18 or fewer −CH₂- units. Also, this observation is in accordance with Bard’s model where the tunneling current, though it decreases with increased distance between the substrate surface and the nanoparticle, is still higher than the kinetic current of the Co^3+/4+ oxidation process. The k values reported are close to the literature value where Wang et al. fabricated the electrodeposited Co₃O₄ nanosheet on Ti foil in 1 M KOH and observed the k at 0.29 s⁻¹ (Xiao et al., 2020).

The electrocatalytic OER mechanism of Co₃O₄ is proposed through four consecutive proton-coupled electron transfer (PCET) processes in acid as shown in Eqs 5–8 (Shinagawa et al., 2015; Srinivasa et al., 2020; Xiao et al., 2020). Co + H 2 O → Co - OH + H + + e − , (5) Co - OH → Co - O + H + + e − , (6) Co - O + H 2 O → Co - OOH + H + + e − , (7) Co - OOH → Co + O 2 + H + + e − . (8) where Co represents an electrochemically active Co site and the −O, −OH, and −OOH represent the surfaced-adsorbed oxo, hydroxyl, and peroxyl intermediates, respectively. The Tafel slope at 120 mV dec⁻¹ represents that the adsorption of -OH is rate-limiting (Eq. 5), while the Tafel slope of 40 mV dec⁻¹ suggests that the deprotonation of −OH to form −O is rate-limiting (Eq. 6). Tafel slopes between these two values arise owing to different relative rates of the first two elementary steps in the mechanism. On the Co₃O₄ surface, the Tafel slope of our electrodes and literature reported values all lie in the range of 79–120 mV dec⁻¹, which suggests that the adsorption of −OH is rate-limiting (Mondschein et al., 2017; Han et al., 2021; Natarajan et al., 2021). This is also consistent with the volcano plots calculated by DFT (Song et al., 2020). Other interpretations of the Tafel slope include the influence of the dissolution of Co₃O₄ during the OER that leads to a Tafel slope value that deviates from the typical one (Mondschein et al., 2017).

The TOF of the ITO-SAM-Co₃O₄ depends on the carbon chain length of the dicarboxylic acid, with the ITO-Glu-Co₃O₄ and the ITO-Oct-Co₃O₄ being higher than the ITO-Hex-Co₃O₄ and the ITO-Eic-Co₃O₄. This might be caused by the decrease of the tunneling current with increased chain lengths of the SAM, which entails the shift from the OER kinetics-controlled current to the mixed tunneling and OER kinetics-controlled current according to Bard’s model. Also, a decrease in the TOF is observed with larger Co₃O₄ size. This suggests that the OER activity per active Co site is lower in the case of larger nanoparticles, which is related to the structure of Co₃O₄, like crystallinity and surface defects. The OER activity of the constructed ITO-Oct-Co₃O₄ is compared to that of other reported Co₃O₄ electrocatalysts in acid, and the results are summarized in Supplementary Table S3. However, the stability of the constructed ITO-Oct-Co₃O₄ is limited toward OER in acid, which is mainly attributed to the dissolution of active Co sites from Co₃O₄ as evidenced by the loss of Co features in CV during repetitive potential cycling (Supplementary Figure S6). ICP-AES on the electrolyte after 1,000 cycles of the potential cycling shows that 13% of Co was leached into the electrolyte. Also, the ITO-Oct-Co₃O₄ can only sustain the 0.05 mA cm⁻² galvanostatic measurement for 2500 s in 0.1 M HClO₄ (Supplementary Figure S8). In acid, the gradual formation of a porous hydrous oxide layer with a loosely bonded Co center is observed and is proposed to be related to OER stability (Natarajan et al., 2021). The formed hydrous oxide layers could also affect the stability of the anchoring ester groups, which could lead to loss of the Co₃O₄. Further optimization of the Co₃O₄ structure, like the incorporation of acid stable components (Huynh et al., 2017), is required to enhance the stability of the assembly.

We constructed the Co₃O₄-modified ITO electrodes using dicarboxylic acid as the bridging molecule. The ITO-SAM-Co₃O₄ were characterized using electrochemistry and XPS and are active toward the OER in acid. The ITO-Oct-Co₃O₄, with Co loading at 3.31 × 10⁻⁸ mol cm⁻² exhibits 374 mV onset overpotential and 497 mV overpotential to reach 1 mA cm⁻² OER current density in 0.1 M HClO₄. The k is not affected by the carbon chain lengths of the SAM, and an increase in the Co₃O₄ size enhances the k, which is consistent with the previous theory of the electron transfer kinetics. Enhanced TOF of the OER is observed on electrodes with shorter carbon chains and smaller Co₃O₄. Meanwhile, the stability of the ITO-SAM-Co₃O₄ is limited by the Co₃O_4, which is prone to dissolute under the OER in acid. Strategies for enhancing the stability of the Co₃O₄ in acid are essential in developing non-noble metal-based interfaces for OER. In addition, the Co₃O₄ tethered to ITO by SAM exhibits significantly higher Γ and higher TOF of the OER than the Co₃O₄ bound to ITO using Nafion, and we propose that evaluation of the nanoparticle electrocatalytic activities using Nafion as the binder in the electrode preparation would cause severe underestimation. Nevertheless, binding nanoparticle electrocatalysts with Nafion or other ionomers is currently the most practical way of fabricating the membrane electrode assembly in fuel cells and water electrolyzers. The nanoparticle-tethered electrodes are promising as platforms for evaluation of the electrochemical catalytic activities of the nanoparticles.