Galvanic displacement of Co with Rh boosts hydrogen and oxygen evolution reactions in alkaline media

The growing energy crisis put an emphasis on the development of novel efficient energy conversion and storage systems. Here we show that surface modification of cobalt by a fast galvanic displacement with rhodium significantly affects the activity towards hydrogen (HER) and oxygen evolution reactions (OER) in alkaline media. After only 20 s of galvanic displacement, the HER overpotential is reduced by 0.16 V and OER overpotential by 0.06 V. This means that the predicted water splitting voltage is reduced from 2.03 V (clean Co anode and cathode) to 1.81 V at 10 mA cm−2 (Rh-exchanged Co electrode). During the galvanic displacement process, the surface roughness of the Co electrode does not suffer significant changes, which suggests an increase in the intrinsic catalytic activity. Density Functional Theory calculations show that the reactivity of the Rh-modified Co(0001) surface is modified compared to that of the clean Co(0001). In the case of HER, experimentally observed activity improvements are directly correlated to the weakening of the hydrogen-surface bond, confirming the beneficial role of Rh incorporation into the Co surface.


Introduction
Water splitting has always occupied the attention of the electrochemical community, but now more than ever, it seems to be the focus of research. The situation is caused by a global energy crisis whose ending cannot be foreseen. Thus, the improvement of electrocatalysts for hydrogen evolution (HER) and oxygen evolution reactions (OER) in either acidic or alkaline media is both important and urgent.
The best electrocatalysts for both HER and OER (and other electrocatalytic reactions) come from the group of noble metals, platinum and its neighbors in the Periodic Table of Elements [1][2][3][4]. While noble metals are highly active and stable under harsh electrochemical conditions of HER and, particularly, OER in an acidic environment, they are both deficient and very expensive to be used at industrial scales [5], particularly having in mind the expected future boost in hydrogen production as a way of energy storage of renewable energy sources. Thus, alternatives for precious metals must be searched. When one deals with alkaline solutions, where HER is generally slow even on noble metals, transition metals and their alloys are used. On the other hand, OER is favored in alkaline media [6]. Thus, cathodes for HER in alkaline media mainly consist of Ni, Co, Fe, and their combinations. However, numerous other classes of materials have been investigated, including supported metallic nanoparticles, chalcogenides, hydroxides, pnictides, and carbides [7][8][9][10][11], as well as single-atom catalysts [12].
On the other hand, OER takes place at high anodic potentials, where metals are covered with oxide layers. The most active catalysts for OER are RuO 2 and IrO 2 [13], which are benchmark OER catalysts, but are expensive [14]. Thus, depending on the environment, different, nonprecious classes of materials have been used for OER, such as Ni-oxides [15,16], Co-oxides [17,18], bimetallic and Bojana Nedić Vasiljević and Aleksandar Z. Jovanović contributed equally.
Electrocatalysis is a heterogeneous process. Thus, exposing the more active site to the reactants speeds up the overall rate of the reaction. This strategy is employed when it comes to practical applications. Thus, high surface area catalysts are made, or the particles are reduced to nanometric dimensions, providing more active sites and a higher surface-to-area ratio. The latter strategy is applied particularly for noble metal-based catalysts as this also significantly reduces the price by minimizing noble metal loading. Nevertheless, for more abundant metals, like Ni and Co, electrochemical deposition is a frequent procedure for preparing anodes and cathodes for water splitting, where large quantities of metal are deposited to form electrochemically active layers with large surface areas [24][25][26][27].
Besides increasing the electrochemically active area, enhancing the intrinsic activity of the catalysts can have only beneficial effects on the economy of the water splitting processes. However, such activity enhancements also have to be acceptable cost-wise. Thus, making, for example, bulk alloys or intermetallic where the expensive solute is uniformly dispersed in a low-cost host is not the best option for the production of affordable materials for large-scale applications. However, if the modifications of low-cost materials can be restricted to the surface, where a large fraction of an expensive phase will be exposed to reactants, such an approach could be rather attractive for practical use. An example of such modifications is the galvanic displacement of a non-precious metal with a more noble one [28]. The process is spontaneous and limited to the surface layer, which is highly beneficial, allowing for obtaining cost-effective, highly active electrocatalysts [29][30][31][32].
Recently, we reported significant HER activity improvements of Ni in alkaline media upon fast galvanic displacement with Rh [33]. The displacement reaction was performed in highly acidic media (thus shifting Ni potential to low values) in a concentrated RhCl 3 solution (thus increasing the exchange rate). As a result, HER activity surpassed that of platinum after only 45 s of galvanic exchange. Here we extend this research to the case of cobalt and address both HER and OER in alkaline media, showing significant activity improvements after a short galvanic exchange with Rh. To better understand observed activity improvements, theoretical calculations are performed for different Rh-modified Co surfaces. The results suggest that Co modification with Rh could be a viable strategy for making highly efficient anodes for OER, while HER activity improvements are also quite high.

Working electrode preparation
The electrode preparation and measurement protocols are similar to the ones we used in previous work on Rh-modified Ni electrodes and are described here for completeness [33]. The measurements are performed using smooth polycrystalline Co (Co-poly, 99.9% Goodfellow) rotating disk electrodes (RDE). The Co disk had a diameter of 3 mm and was inserted in a Teflon cylinder with a 10-mm diameter. After polishing the disk to mirror-finish using alumina powder, the electrode was sonicated for 1 min, washed in deionized water, and diluted with HCl, in deionized water, and finally with 1 mol dm −3 KOH solution. It was then quickly transferred to the electrolytic cell, and HER/OER measurements were done to obtain HER/OER activities of pure Co-poly. Galvanic displacement was done by immersing polished and washed Co-poly RDE into 0.1 mol dm −3 RhCl 3 solution in 0.1 mol dm −3 HClO 4 for a given amount of time (up to 30 s). After the exchange, the electrode was rinsed in deionized water, 1 mol dm −3 KOH solution, covered with the droplet of the same solution, and quickly transferred into the electrochemical cell. The transfer to the cell typically took under 30 s. The electrodes are denoted as Rh(t)@Co, where t is the duration of the galvanic displacement experiment in seconds (10, 20, or 30).

Electrochemical measurements
Electrochemical measurements were done using IVIUM Vertex.One or Gamry Interface 1011e potentiostats in one compartment three-electrode electrochemical cell with double-junction saturated calomel electrode (SCE) as a reference electrode and a 3 × 3 cm Ni foam (Goodfellow Cambridge Limited, England) as a counter electrode. KOH solution (1 mol dm −3 , Sigma-Aldrich), prepared with ultrapure deionized water, was used as the electrolytic solution in all experiments. Measurements were done at room temperature. The electrolyte was purged with N 2 gas stream (99.999%, Messer, Serbia) during the experiments. To calculate potentials versus reversible hydrogen electrode (RHE), the potentials are converted to the RHE scale using the formula E RHE = E SCE + 0.244 V + 0.059 V × pH (pH = 13.45). Electrolyte resistance was corrected using hardware settings with a positive feedback scheme. Up to 75% of the resistance value, determined using singlepoint impedance measurement at −1 V vs. SCE (approximately 0 V vs. RHE), was corrected to avoid instabilities during the measurements. HER and OER measurements were done using cyclic voltammetry at a common sweep rate of 10 mV s −1 . After the transfer to the electrochemical cell and before the potential sweep, the electrode was held −1 V vs. SCE until the current density dropped below 1 μA cm −2 . Then, the electrode potential was swept in a cathodic or anodic direction to measure HER/OER activity. HER was always measured on the electrodes which were not previously subjected to potentials higher than −1 V vs. SCE (approximately 0 V vs. RHE). In order to prevent surface blockage by gas bubbles formed during the HER/OER measurements, the electrode was rotated at a constant rate of 1800 rpm to remove evolved H 2 /O 2 . For this purpose, RDE710 Rotator (Gamry, USA) was used.
Here we report currents normalized per geometric surface area. Namely, the roughness factor variations, determined in separate measurements described below in more detail, were small enough to be excluded from calculations in the analysis of electrocatalytic activity.
For the Co-poly electrode, roughness factor (RF) measurements were done on freshly polished electrodes using double-layer capacitance measurements by cyclic voltammetry in a narrow potential window. Measurements were done in 1 mol dm −3 KOH solution by cycling the electrodes between −0.1 and 0 V vs. RHE. Before the measurements, the electrodes were subjected to intensive HER at deep negative potentials to reduce oxide content. Electrode capacitance, determined from the slope of the current vs. potential scan rate line, was divided by 20 μF cm −2 [34,35] to obtain the real electrochemically active surface area (ESA). RF was determined as the ratio of ESA and the geometrical crosssection of a disk. RF measurements were done only for clean, polished Co-poly electrode, while the surface roughness changes upon galvanic displacement were checked using SEM 3D surface reconstruction.

Surface morphology and chemical composition
Morphology analysis and chemical composition were probed using SEM-EDX with Phenom ProX Scanning Electron Microscope (Phenom, Netherlands). SEM characterization was done using an acceleration voltage of 10 kV, while the chemical composition was probed with an acceleration voltage of 15 kV. The high conductivity of the samples allowed for the SEM-EDX characterization without any deposition of a conductive layer on the electrodes.

DFT calculations
The first-principle Density Functional Theory (DFT) calculations were performed using the Quantum ESPRESSO package [36]. The Generalized Gradient Approximation (GGA) in the parametrization by Perdewet al. was used [37]. Cutoff energy of 40 Ry was employed, while the charge density cutoff was 16 times larger. We modeled Co(0001) surfaces using the corresponding p(2 × 2) or p(4 × 4) cells of given surfaces with four-layer slabs. Rh was inserted in the surface layer (Rh surf ) or the subsurface layer (Rh sub ) to the concentration of 25 at.% and the top 2 surface layers' surfaces were allowed to relax fully. Also, we have investigated the case of a full Rh monolayer on top of the Co(0001). A Monkhorst-Pack Γ-centered 4 × 4 × 1 k-point mesh was used to integrate the first irreducible Brillouin zone [38]. For the studied surface, we have investigated the adsorption of H, OH, and O to link the reactivity changes to the observed electrocatalytic activity changes. The adsorption is quantified using binding energies of given adsorbates (E b (A), A = H ads , OH ads , O ads ), calculated as: where E SURF+A , E SURF , and E A stand for the total energy of the surface with an adsorbate, the total energy of the clean surface, and the total energy of an isolated adsorbate.

Electrocatalytic activity and surface composition of Rh-exchanged Co-poly electrodes
In order to estimate the HER and OER activity improvement, we measured HER and OER and compared the overpotentials needed to reach 10 mA cm −2 (η 10 ), as suggested by McCrory et al. [39]. For a smooth Co-poly disk, HER η 10 is found to be −0.43 V (uncertainties of determined overpotentials are below ± 30 mV). The corresponding Tafel slope at higher current densities is −216 mV dec −1 , while the change of the Tafel slope is observed when going from low to high current densities (Fig. 1). This indicates that Heyrovsky reaction could be the rate-determining step although the values of determined Tafel slopes significantly deviate from theoretical values (−40 mV dec −1 and −120 mV dec −1 , for low and high surface coverage by H ads ). The situation is rather similar for the Rh-exchanged Co-poly, except that a tremendous reduction of the HER overpotential is observed. Upon 10 s of exchange, the overpotential is reduced by 0.1 V, down to −0.33 V. After 20 s of exchange, the HER overpotential is −0.27 V, and the corresponding Tafel slope at higher current densities is −150 mV dec −1 . After a further increase of the galvanic exchange time to 30 s, we have not observed further improvements in HER activity (Fig. 2, left).
Even with tremendous improvements in HER activity, the Rh-exchanged electrodes do not present exceptional catalysts. For example, we have previously reported HER η 10 of clean Ni-poly more negative at −0.4 V and a corresponding Tafel slope of −140 mV dec −1 , with the electrode having RF of 2.8 [33]. In the results of McCrory et al. [39], η 10 for the Ni surface with a roughness factor of 20 ± 10 was approximately −0.25 V. However, after 45 s of Ni exchange with Rh, η 10 amounted to only −110 mV, without a significant change of RF. As a reference, Pt-poly can be used, and η 10 reported in the literature amount − (100 ± 20) mV for smooth Pt-poly surface (RF 6 ± 2), and ~ −40 mV for Ptpoly with RF = 90 ± 20 [39], while we report η 10 of −110 mV for RF around 8 [39]. Nevertheless, the data for clean Co indicate η 10 of −0.22 ± 0.2 V, but for an electrode with RF of 1100 ± 400 [39]. We have used smooth Co-poly with RF found to be 9.9 ± 1.5, thus with 100 times lower electrochemically active surface, which explains the differences in observed η 10 .
Nevertheless, it is important to understand whether the HER activity improvements are connected with the increase of the intrinsic activity of the electrodes upon exchange with Rh or whether the effect is associated with the change of the electrochemically active surface, i.e., the surface roughness. First, we have considered the blank cyclic voltammograms of Co-poly and Rh-modified electrodes (Fig. 2, right) and observed that there are some changes in the potential window corresponding to extensive surface oxidation before the OER start. However, the current response in the potential window between hydroxide formation and oxide buildup (roughly between −0.5 and −0.25 V vs. RHE) is similar in size (20% relative Fig. 1 The HER polarization curves for Co-poly and Rh-exchanged electrode (HER overpotentials are indicated, left) with corresponding Tafel plots (right). 1 mol dm −3 KOH, potential sweep rate 10 mV s −1 , electrode rotation rate 1800 rpm Using SEM (combined with the 3D surface reconstruction; Fig. 3), it was confirmed that the morphology and active surface of the Co-poly electrodes have not changed upon galvanic displacement with Rh. After integrating the 3D surface profiles (Figs. S1 and S2, Supplementary  Information). We have actually found that the surface area is decreasing. The integration suggested RFs of 1.33, 1.25, and 1.14 for clean Co-poly, Rh(10)@Co, and Rh(20)@ Co, respectively. These results indicate that the HER activity improvements are due to increased intrinsic catalytic activity. We note that the RF values obtained using SEM 3D reconstruction are much lower compared to the value obtained using electrochemical measurements, but this is due to the resolution of SEM, which is much lower compared to the resolution of electrochemical techniques (which have, colloquially said, atomic or solvated ion resolution).
The activity increase is linked with the Rh incorporation into the Co lattice, which is confirmed using EDX elemental mapping (Fig. 4). It can be seen that the oxygen surface content decreases while Rh content increases with the exchange time. However, for the first 20 s of exchange, Rh content increases rapidly to 1.96 at.%, while after additional 10 s, it reaches only 2.06 at.% (Fig. S3, Supplementary Information). This result indicates a certain saturation of the surface with Rh, which is also in line with the observation that HER activity was not improved upon extending the exchange time from 20 to 30 s (Fig. 2, left). Moreover, a similar percentage of Rh was found by EDX as in Jovanović et al. [33], which translated to approx. 60 at.% of Rh found by XPS, thus restricted to the surface layer. For this reason, it is safe to expect that similar surface concentrations of Rh are present after galvanic displacement in the case of Co-poly. We note that in the case of Rh(30)@Co electrode, RF determined using SEM 3D reconstruction increased slightly compared to other electrodes, to 1.41, which is not a very large change. Thus, we believe it is safe to conclude that the RF is not largely affected by the galvanic displacement procedure, which was also observed in the case of Rh-modified Ni-poly electrodes [33].
Next, we analyze the effects of Co galvanic displacement by Rh on the OER activity (Fig. 5). In general, Co is considered a good OER catalyst [22,40], in contrast to its HER catalytic activity. In the case of OER, the absolute changes in η 10 are not as prominent as in the case of HER. However, the activity is significantly improved, particularly because the Tafel slope is also significantly reduced for Rhmodified Co electrodes.
Here we see that the value of η 10 shifts from 0.37 V (clean Co-poly) to 0.31 V (Rh(20)@Co). Moreover, the Tafel slope is reduced from 102 mV dec −1 (Co-poly) to 56 mV dec −1 (Rh(20)@Co). As reference values, one can mention the ones reported in McCrory et la. [39], from which we tabulated selected cases for easier comparison (Table 1). Obviously, the activities reported here for Rh-modified Co electrodes are exceptional, suggesting such electrode materials could be considered anodes for alkaline water electrolysis.
Considering large improvements in OER activity and relatively good HER activity improvements, one might consider the overall reduction in water splitting voltage when Rh-modified Co electrodes are used. For example, if the current density of 10 mA cm −2 is taken into account, the use of pure Co electrodes would require a theoretical voltage of 2.03 V (iR-free). However, if Rh(20)@Co electrodes are used, the voltage would be 1.81 V, which is an 11% reduction in the overall voltage. However, combining the Rh-modified Co anode with a more efficient cathode is also possible. For example, if Rh-modified Ni cathode [33] is used (with η 10 of −0.11 V), the overall voltage would be 1.70 V, which is a 16% reduction in voltage compared to the combination of pure Co anode and Co cathode. Going to a higher current would impact HER overpotential more than OER overpotential as the Rh(20)@Co electrode displays a low-value Tafel slope (56 mV dec −1 ; Fig. 5) compared to typical values of Tafel slopes for HER alkaline media (−120 mV dec −1 ). Furthermore, if the Ni-modified cathode and the benchmark Ir anode are used, the overall water splitting voltage would be higher by 0.08 V compared to the previously discussed hypothetical cell with Rh-modified Co anode (at 10 mA cm −2 ; Table 1). On the other hand, replacing the benchmark Ru anode with an Rh-modified Co anode would keep the voltage roughly unchanged if the cathode is unchanged. Namely, based on the overpotentials presented in Table 1, the activity of Rh (20)@Co is close to that of Ru anode, taking into consideration uncertainties in measured activities.
As discussed in our previous work on Rh-modified Ni [33], the catalyst modification occurs only in the surface layer, and a very small amount of Rh is deposited. Therefore, even though the Rh price is extremely high (439 $ per gram [41]), it can be estimated that a replacement of a full monolayer of Co with Rh would increase the price by only 1.4 $ per square meter of a real surface area.

Theoretical calculations
In order to elucidate the effects of modification of Co surface with Rh, we have performed a series of DFT calculations. First, we have considered the formation of Rh islands and surface Co-Rh surface and subsurface alloys. Using a large p(4 × 4) unit cell of Co(0001), we compared the energies  Table 1 The comparison of OER overpotentials (η 10 ) for Rh-modified Co electrodes, and the reference data for Co, Ir, and Ru, as reported in McCrory et la. [39] *This work Electrode RF η 10 /V Co-poly* 9.9 ± 1.5 0.37 ± 0.03 Rh(10)@Co* 9.9 ± 1.5 0.33 ± 0.03 Rh(20)@Co* 9.9 ± 1.5 0.31 ± 0.03 Co 11 ± 5 0.41 ± 0.03 Ir 130 ± 10 0.39 ± 0.01 Ru (1) 70 ± 20 0.29 ± 0.03 Ru (2) 200 ± 20 0.34 ± 0.03 of the formation of Rh ad-island and Rh ad-atoms (coverage of 0.25 ML). It was found that a compact Rh island is energetically more favored over separated Rh ad-atoms by 0.53 eV per Rh atom. It is likely due to strong Rh-Rh interactions in the adsorbed layer, which additionally stabilize the system [42]. However, the surface alloy is slightly more stable (0.03 eV per Rh atom) when Rh atoms are separated and surrounded by Co atoms, i.e., dispersed in the surface layer, rather than making a compact Rh island in the surface layer. Moreover, the calculations show that a surface alloy (Rh atom in the surface layer) is energetically more stable than a subsurface alloy (Rh atom in the subsurface layer) by 0.45 eV. This result indicates that Rh would segregate (or remain in the surface layer once inserted). This trend is due to the differences in surface energies of Rh and Co, and our calculations agree with the predictions of Ruban et al. [43]. Based on the considerations given above and the fact that we were not able to observe Rh aggregates on the Co surface, we have focused on the reactivity of Rh-Co surface (Rh surf -Co) and subsurface alloy (Rh sub -Co), with 0.25 ML of Rh in surface or subsurface layer, and also considered the case of an Rh monolayer epitaxially formed on the Co(0001) surface (Rh ML -Co) (Fig. 6a). The results concerning H ads , O ads , and OH ads formation are given in Table 2. Furthermore, we considered the adsorption for a full monolayer of mentioned adsorbates to better match the calculations to the experimental conditions. Namely, at high overpotentials, HER proceeds with high surface coverage by H ads , while OER proceeds on an oxidized surface.
There is a direct scaling between the binding energies of H ads on one side and O ads and OH ads on the other side (Fig. 6b), and also an excellent scaling between O ads and OH ads binding energies (Fig. 6b, inset), in line with previous  reports [44]. We also notice that the effects of the insertion of Rh in the subsurface layer are minor, and the reactivity of the Rh sub -Co system is practically the same as that of the Co(0001) surface (Table 1; Fig. 6b). Figure 6c shows subtle modifications of the electronic structure of Co atoms in the surface layer when Rh is introduced in it. However, surface reactivity is what comes to be the most important when it comes to catalytic activity. Namely, it was shown that the electronic structure descriptors could serve to estimate the activity trends, but they also can be misleading in some cases. For example, in the case of Pd overlayers on the Au(111) surface, it was shown that hydrogen binding energies correlate well with HER exchange current densities, while d-band centers suggest a wrong trend [45]. Here we also focus on HER activities, as the explanation of the OER activities requires the consideration of surface oxide phases [13,46,47], which is still elusive at this point for the case of Rh-modified surfaces. However, we note that doped Co-oxides were previously investigated as OER catalysts, and activity improvements have been reported [17]. Thus, it might be considered that Rh atoms deposited in the surface layer act as dopants for Co-oxide or form a mixed (Co, Rh)-oxide layer, which acts as an active OER catalyst due to optimized binding energies towards OER intermediates. Based on the results presented in Table 2, such optimization of oxygen species' binding energies is a rather plausible expectation. Going back to HER activities, the weakening of the H ads -metal bond can be correlated directly to the improved HER activity. While such correlations are known for acidic media [48,49], they also hold in alkaline solution [50,51], giving a characteristic volcano curve for HER if the activities are correlated to the hydrogen binding energies on a metallic surface. On this computational-based volcano curve, Co is positioned at the strong bonding branch [50,51]. Thus, weakening the H ads -metal bond would shift the activity of the surface towards its apex, where platinum is situated. Based on the results presented in Table 2, it is clear that the incorporation of Rh into the cobalt surface should positively affect the HER activity, as observed experimentally (Fig. 1). Moreover, with the increasing concentration of Rh in the surface, one can expect that the activity will increase as mixed Rh-Co sites provide adequate H ads energetics for HER. However, Table 2 also shows that the catalytic activity increase should be relatively small when going from 0.25 ML of Rh in the surface to a full Rh monolayer. Namely, from clear Co(0001) to 0.25 ML Rh surf -CO E b (H ads ) is weakened by 0.11 eV. However, upon completion of the Rh monolayer, additional destabilization of the H ads monolayer is only 0.03 eV. This result aligns well with the saturation we observed and the same HER activities of Rh (20)@Co and Rh(30)@Co electrodes (Fig. 2, left).

Conclusions
In this work, we have shown that fast galvanic displacement of Co by Rh in concentrated acidic solutions of Rh significantly improves both HER and OER activities of Co. After only 20 s of galvanic displacement, HER overpotential necessary to reach 10 mA cm −2 is reduced from −0.43 to −0.27 V. At the same time, the OER overpotential is reduced from 0.37 to 0.31 V with reduction of the Tafel slope down to 56 mV dec −1 . Such large improvements in HER and OER activities can be utilized to significantly reduce the overall water splitting voltage in the alkaline water electrolysis process. With the cost penalty below 1.5 $ per square meter of a real surface area required for the formation of a complete Rh monolayer on the Co surface, the reduction of water splitting voltage is at least 10%. DFT calculations suggest that Rh embedded in the Co surface tunes the surface reactivity towards H ads , O ads , and OH ads , which can be connected to the experimentally observed activity improvements. For the case of HER, the connection is rather straightforward and weakening the H ads -surface bond due to the Rh insertion in the surface directly correlates to the enhanced HER activity. To fully understand improved OER activities, it is necessary to consider oxide formation on the Co surface in the presence of Rh, which could lead to Rh-doped oxides or mixed oxide phases. However, this issue remains open for now, while the high OER activities reported here certainly motivate further research.