Size dependent electrocatalytic activities of h-BN for oxygen reduction reaction to water

In a future sustainable society based on renewable energies, hydrogen is considered to play a central role as the most important energy carrier.¹ Hydrogen can be generated by electrolysis using renewable energy based electricity and converted to electricity by a fuel cell without emitting CO₂. In both electrolysis of water and fuel cell, electrocatalysts to reduce overpotential for hydrogen evolution (HER)/oxidation (HOR) and oxygen evolution (OER)/reduction (ORR) reactions are required.² In a fuel cell, the large overpotential at a cathode for ORR, which is due to the fact that electrochemical ORR is a complex reaction with multiple electron transfer steps of slow kinetics,³ is the major cause of energy loss, and many efforts have been made to find efficient electrocatalysts for ORR.⁴ Currently, Pt based ORR electrocatalysts are widely used in practical fuel cells, including one in a commercial fuel cell vehicle by Toyota, Mirai,⁵ because of their high reactivity. However, due to the high cost, less abundance, and poor stability in the electrochemical environment of Pt based electrocatalysts, many research groups are working hard to find non-precious metal electrocatalysts, such as non-precious metals,^6–9 metal oxides,^10–13 metal oxynitrides,^14–19 metal carbides,^20–23 and N- and B-doped carbon materials.^24–28 

Recently, we have proposed theoretically that the combinations of insulating h-BN and electrocatalytically inert metal substrates could work as efficient electrocatalysts for ORR^29–32 and proved experimentally that various forms of h-BN on the gold substrate indeed acted as electrocatalysts not only for ORR^31,33 but also for HER.³⁴ Although the highest ORR activity was achieved by the modification with BNNS among the various h-BN, the overpotential was still very high compared with that at the Pt electrode and oxygen was reduced almost 100% to H₂O₂ via the two-electron process.

Strong substrate dependencies^31,33 and theoretical evaluations³¹ suggest that the interaction between BNNS and Au is very important in achieving higher electrocatalytic activities for ORR. The ORR activity was improved indeed has not only a higher ORR rate but also a higher fraction of oxygen to H₂O was achieved by increasing the BNNS/Au interaction by decorating BNNS with Au nanoparticles (NPs).^35,36 When size controlled small (∼5 nm) Au NPs were used for BNNS decoration,³⁶ overpotential for ORR was more than 300 mV less than that at the Au electrode and only 100 mV more than that at the Pt electrode, and more importantly, more than 80% of oxygen was reduced to water by the four-electron process. The importance of the interaction at the Au-BNNS interface was suggested by theoretical calculation.³⁷ The HER overpotential was also reduced by the decoration with Au- and Au-alloy NPs.³⁸ 

Another way to increase the number of Au-BNNS interfaces is to use small sized BNNS. We have shown already that the smaller the size of BNNS, the higher the HER activity.³⁴ In the present paper, we have investigated the effect of the size of BNNS used for Au modification on ORR activity in O₂ saturated 0.5M H₂SO₄ solution. The smaller the size of BNNS, the smaller the overpotential for ORR as was the case for HER and the higher the fraction of O₂ reduction to water by the four-electron process. Density functional theory (DFT) calculations on the free energy change along the ORR pathway with the OH-terminated 3 × 3 island of h-BN as a model of BNNS showed that the active sites for four-electron reduction of O₂ are located at the edges of BNNS adsorbed on Au(111), supporting higher activity of smaller BNNS for ORR.

Au and Pt working electrodes were prepared on n-Si(111) (P doped, resistivity of 1–10 Ω cm, Shin-Etsu Chemical), which was pre-cleaned by sonication in acetone and Milli-Q water for about 5 min each and then thoroughly washed with conc. sulfuric acid and Milli-Q water by depositing 20 nm thick Ti as an adhesion layer followed by 150 nm Au and Pt, respectively, using radio frequency magnetron sputtering (JSP-8000, ULVAC, 450 W RF power). A gold single crystal prepared by the Clavilier method³⁹ using a gold wire (99.999% pure, ϕ = 1 mm, Tanaka Precious Metal) was used as a substrate for atomic force microscopic (AFM) measurements.

A liquid exfoliation method was used to obtain BNNS.^31,33 BN powder (99% High Purity Chemicals, BBI03PB) was sonicated in isopropanol (IPA) (Regent grade, Wako Pure Chemicals) solution, with 3 mg/ml as the initial concentration in an ultrasonic bath for 96 h. The dispersions were centrifuged at 1500 rpm for 60 min after sonication, and 3/4 of the supernatant was collected to be used for further measurements.

Size controlled BNNS was obtained by filtration using MF-Millipore filter (Merck Millipore, VSWP type) of various pore sizes. The diluted BNNS dispersion mentioned above was filtered by a filter of 1 µm pore size filter followed by filtration using a filter of 0.45 µm pore.³⁴ The BNNS residue on the 0.45 µm filter was collected and dispersed in IPA to prepare the BNNS (0.45–1.0 µm)/Au electrode. The filtrate was further filtered by a filter of 0.22 µm pore. The BNNS residue on the 0.22 µm filter was collected and dispersed in IPA to prepare the BNNS (0.22–0.45 µm)/Au electrode. The filtrate was further filtered by a filter of 0.1 µm pore, and the BNNS residue on the 0.1 µm filter was dispersed in IPA to prepare the BNNS (0.1–0.22 µm)/Au electrode. The amount of the filtrate after the filtration using the filter of 0.1 µm pore was too small to proceed for further filtration or to use for surface modification.

Surface modification by BNNS was carried out by self-evaporation of IPA solution of a dispersion of BNNS on Au or Pt substrates.³⁴ The substrate was placed perpendicularly in a 10 ml glass beaker, which contained 5 ml of BNNS dispersed solution of 0.2 mg/ml, until IPA solution was evaporated completely at room temperature (∼24 h) and then dried at 100 °C for 2 h in a vacuum oven.

Size distribution of BNNS dispersed in IPA solution was measured by using the dynamic light scattering (DLS) method using a laser scattering particle size distribution analyzer (HORIBA-LA-950V2).

Scanning electron microscopy and high resolution transmission microscopy (HRTEM) were carried out by using a field emission scanning electron microscope (FE-SEM: S-4800, Hitachi) and JEOL-JEM-2100F with a power of 200 keV, respectively.

Raman scattering measurements were carried out using Horiba-Jobin-Yvon, model-T64000 with Ar–Kr laser (514.5 nm, 0.1 W) with an exposure time of 10 s. The resolution of the spectrometer was 1 cm⁻¹.

Electrochemical measurements were carried out in the three-electrode configuration. A Pt wire and Ag/AgCl (sat. NaCl) were used as the counter and reference electrodes, respectively. A dual potentiostat/function generator (Hokuto Denko, HR-101B/HB-111) and a rotation control unit (Hokuto Denko, HR-2002) were used to control the potential and the rotation speed of the rotating disk electrode (RDE) and rotating ring disk electrode (RRDE). All of the electrochemical measurements were carried out in a 0.5M H₂SO₄ electrolyte solution, which was prepared by using H₂SO₄ (Super special grade, Wako Pure Chemicals) and Milli-Q water at room temperature. The electrolyte solution was saturated with Ar or O₂ gas by passing the ultrapure Ar (99.999%) or ultrapure O₂ (99.999%) gas for at least 1 h before each electrochemical measurement.

Before electrochemical measurements, all the electrodes were pre-treated by cycling the potential between 1500 and −100 mV in an Ar saturated 0.5M H₂SO₄ electrolyte solution at a sweep rate of 100 mV/s for 100 cycles to remove any surface contaminants. All potentials shown in this paper are referred to the reversible hydrogen electrode (RHE).

Density functional theory (DFT) calculations were performed using the gradient-corrected exchange–correlation functional of Wu and Cohen (WC)⁴⁰ as implemented in the SIESTA package.⁴¹ The Au(111) surface has been modeled by the 8 × 8 slab with four layers of Au atoms. The top two layers of Au are fully relaxed, while the bottom two layers were frozen. The BNNS was modeled by the OH-terminated 3 × 3 island of h-BN. Entropy, zero-point energy, and solvent effects were taken into account when calculating the change in free energy along the ORR pathway. Double-ζ plus polarization function (DZP) basis sets were used to treat the valence electrons of all atoms, while the core electrons were represented by Troullier–Martins norm-conserving pseudopotentials.⁴² The basis sets for hydrogen and oxygen atoms were optimized with the use of the Nelder–Mead simplex method^43,44 to fit the theoretical value of the free energy balance for the O₂(g) + 2H₂(g) → 2H₂O(l) reaction to the experimental value of ΔG_exp = −4.92 eV, where the free energy for solvation of a water molecule ΔG_solv(H₂O) has been set to its experimental value of ΔG_solv(H₂O) = −6.32 kcal mol⁻¹.⁴⁵ Other computational details are described in our previous works.^31,35,37,46

Figure 1(a) shows the size distribution of size selected BNNS, which were obtained as described in Sec II, determined by DLS. While the sizes of as prepared BNNS were not controlled and broadly distributed from less than 0.01 µm to more than 20 µm [Fig. 1(a-i)], BNNS remained on the 0.45, 0.22, and 0.1 µm filters had a very narrow size distribution of 0.45–1.0 µm [Fig. 1(a-ii)], 0.22–0.45 µm [Fig. 1(a-iii)], and 0.1–0.22 µm [Fig. 1(a-iv)], respectively, as controlled by the pore size of the filters.³⁴ 

Figure 1(b) shows a typical HR-TEM image of BNNS remained on a 0.22 µm pore sized filter. The size of the BNNS was close to the average size determined from the DLS measurement as shown in Fig. 1(a-iii), and the Fourier transform (FFT) of the image [Fig. 1(c)] revealed that the BNNS are of hexagonal atomic arrangement.

Figure 2 shows the SEM images of the Au substrate [Fig. 2(a)] and of the as prepared BNNS deposited on the Au substrate by self-evaporation of IPA solvent in the BNNS solution. Size selected BNNS deposited on Au electrodes are shown in Figs. 2(b)–2(d). The dark regions in Figs. 2(b)–2(d) correspond to BNNS deposited on the Au surface as confirmed by EDX measurements (see Fig. S1 in the supplementary material).

Raman measurements suggest that majority of BNNS on Au are of monolayer as previously reported.^31,34

The bottom panel of Fig. 3 shows disk currents (i_disk) at (a) bare Au, (b) BNNS(as prepared)/Au, (c) BNNS(1.0–0.45 µm)/Au, (d) BNNS(0.45–0.22 µm)/Au, (e) BNNS(0.1–0.22 µm)/Au, and (f) bare Pt electrode (bottom panel) in RRDE configuration with the rotation rate of 1500 rpm as a function of disk potential, which was scanned negatively from 1200 to 100 mV with the sweep rate of 10 mV/s, measured in O₂ saturated 0.5M H₂SO₄ solution.

At the bare Au electrode, the cathodic current due to ORR started to flow at around 450 mV, which increased as the potential became more negative, but no limiting current was observed in the potential region used in the present study. In the case of BNNS deposited Au electrodes, the cathodic current started to flow from more positive potential than the bare Au electrode and limiting currents were observed in all the cases, as we previously reported.^31,33 Interestingly, current–potential relations were dependent on the size of BNNS. The potentials at which ORR current of −0.14 mA/cm² flowed at BNNS(as prepared)/Au, BNNS(1.0–0.45 µm)/Au, BNNS(0.45–0.22 µm)/Au, and BNNS(0.22–0.1 µm)/Au were 680, 697, 764, and 784 mV, respectively, showing that the smaller the BNNS size, the smaller the ORR overpotential. Furthermore, limiting currents were also dependent on the size of BNNS as −2.5, −3.2, −4.1, and −4.5 mA/cm² at BNNS(as prepared)/Au, BNNS(1.0–0.45 µm)/Au, BNNS(0.45–0.22 µm)/Au, and BNNS(0.22–0.1 µm)/Au, respectively, showing that the smaller the BNNS size, the larger the limiting current. The higher limiting current corresponds to the higher number of electrons involved in the ORR, i.e., the higher fraction of oxygen is reduced to water. Thus, the higher fraction of O₂ was reduced to H₂O via the four-electron process at an Au electrode modified with smaller sized BNNS.

The top panel of Fig. 3 shows ring currents (i_ring) at a Pt ring electrode held at 1100 mV, which were recorded when the disk potential was scanned negatively, corresponding to the disk currents shown in the bottom panel as a function of disk current. The ring current corresponds to the oxidation of H₂O₂ generated at the disk electrode, and therefore, the ratio between the disk current and ring current reflects the fraction of oxygen reduced to H₂O₂ and H₂O.^35,36 When the potential was scanned negatively, the ring current started to flow as soon as the cathodic disk current due to ORR started to flow and increased with the increase in the disk current at all the electrodes except at the bare Pt disk electrode where the cathodic disk current due to ORR started to flow at ∼800 mV and increased as the potential became more negative but almost no ring current was detected, showing no H₂O₂ was formed and oxygen was reduced to H₂O via direct four-electron reduction at the Pt disk electrode. We have reported already that oxygen is mainly reduced to H₂O₂ via the two-electron process both at the bare Au and the BNNS/Au electrodes.^35,36 Although disk currents at 0.2 V, where limiting currents were observed at all the electrodes except at the bare Au disk electrode, were in the order of Pt > BNNS(0.1–0.22 µm)/Au > BNNS(0.45–0.22 µm)/Au > BNNS(1.0–0.45 µm)/Au > BNNS(as prepared)/Au > bare Au, ring currents at 0.2 V were in the order of BNNS(as prepared)/Au > BNNS(1.0–0.45 µm)/Au > bare Au > BNNS(0.45–0.22 µm)/Au > BNNS(0.1–0.22 µm)/Au > Pt. Thus, as far as Au electrodes modified with the size selected BNNS are concerned, the smaller the BNNS, the higher the fraction of H₂O formation via the four-electron reduction process.

Thus, one expects a linear relation between 1/i and 1/ω^1/2 (K–L) plot, and n can be determined from the slope of linear relations.

Figure 4 shows the rotation rate dependent linear sweep voltammograms (LSVs) of (a) bare Au, (b) BNNS(as prepared)/Au, (c) BNNS(1.0–0.45 µm)/Au, (d) BNNS(0.45–0.22 µm)/Au, (e) BNNS(0.1–0.22 µm)/Au, and (f) bare Pt electrode in O₂ saturated 0.5M H₂SO₄ solution with the rotation rate between 0 and 2500 rpm when potential was scanned from 800 mV (for bare Pt, 1000 mV) to 100 mV with the scan rate of 10 mV/s.

Figure 5 shows the K–L plots of (a) bare Au, (b) BNNS(as prepared)/Au, (c) BNNS(1.0–0.45 µm)/Au, (d) BNNS(0.45–0.22 µm)/Au, (e) BNNS(0.1–0.22 µm)/Au, and (f) bare Pt electrode at various potentials obtained by using the results of Fig. 4. Linear relations were observed in all the cases, and the number of electrons transferred at each electrode at each potential was calculated from the plots.

Important parameters for the evaluation of ORR catalysts, such as onset potential potentials at 0.14 and at 0.5 mA/cm², half-wavelength potential, and number of electrons involved in ORR determined from the K–L plots, are listed in Table S1 in the supplementary material.

Potential dependencies of the fraction of H₂O production (four-electron route) obtained by K–L (open symbol) and RRDE (solid line) analyses are summarized in Fig. 6. Both results are agreed with each other reasonably well. Fraction of H₂O generation was almost ∼0% and 100% at the bare Au and bare Pt electrodes, respectively, as reported before. As the size of BNNS decreases, the fraction of H₂O generation increased and reached more than 80%, which is larger than that at the BNNS/Au electrode decorated with random sized Au NPs³⁵ and is close to that at the BNNS/Au electrode decorated with 5 nm size controlled Au NPs,³⁶ when BNNS(0.1–0.22 µm) was used.

We have already performed theoretical analysis of a free energy profile for intermediates of ORR on the model system consisting of a small Au₈ cluster supported on the monolayer BN on Au(111) to understand why four-electron reduction of oxygen to H₂O became possible at the BNNS/Au electrode by decorating BNNS by AuNP.³⁵ It was clarified that while the Au₈ cluster on BN/Au(111) mildly stabilizes O₂*, OOH*, and OH* intermediates by 0.5–0.7 eV, it drastically stabilizes O* by 1.6 eV in comparison with the adsorption at the BN/Au(111) surface, making the dissociation of OOH*, which is a uphill process at the BN/Au(111) surface, downhill, opening a four-electron reduction pathway of oxygen to H₂O. Here, we took a similar approach to clarify the ORR mechanism catalyzed by small BNNS on the Au(111) electrode.

We have used H-terminated 3 × 3 BN sheet on Au(111) to model the small BNNS island in our previous work on the BNNS size dependent HER electrocatalyst.³⁴ In the oxygen environment, however, oxygen readily adsorbs at the edge of H-terminated BNNS on Au(111) in a highly activated state and dissociates to form OH terminal groups at the edges of BNNS. These OH groups are strongly bonded to the BNNS edge forming stable OH termination, which cannot be reduced further to water due to energetic reasons. Thus, the small BNNS island was modeled by the OH-terminated 3 × 3 BN sheet on Au(111) in the present calculation.

Figure 7 shows the calculated free energy profiles along the ORR pathway for the pristine h-BN monolayer on the Au(111) surface (black) and OH terminated 3 × 3 BNNS on Au(111) (lines in different colors correspond to the reaction pathways calculated for the different non-equivalent reaction sites at the BNNS edges). The optimized structure of the OH-terminated 3 × 3 BNNS on the Au(111) surface is shown in the supplementary material (Fig. S2).

As we reported previously, since the dissociative ORR pathway for four-electron reduction of oxygen to H₂O, in which OOH* intermediate dissociates into O* and OH* fragments, is not allowed energetically due to the weak O* bonding (Fig. 7, black line) at the pristine h-BN monolayer on Au(111), ORR proceeds only via the two-electron path, leading to H₂O₂ formation.^31,35 On the other hand, at the small OH-terminated BNNS island on Au(111), while O₂*, OOH*, and OH* ORR intermediates are mildly stabilized, the dissociated O*…OH* intermediate is more strongly stabilized so that dissociation of OOH*, which is a uphill process at BN/Au(111) surface, becomes downhill (Fig. 7, lines in various colors), resulting in the opening of four-electron reduction pathway of oxygen to H₂O as was the case at the AuNP decorated BNNS/Au electrode. Optimized structures of all ORR intermediates for different non-equivalent reaction pathways presented in Fig. 7 are shown in Figs. S3 and S4(a)–(e).³⁵ 

Figure 8 shows optimized geometries of OOH* and O*…OH* intermediates adsorbed on the OH-terminated BNNS/Au(111) system for the most stable O*…OH* configuration [ΔG(O*…OH*) = −2.43 eV] before (a) and after (b) OOH* dissociation. The OOH* intermediate adsorbs on top of the B atom at the edge of the BNNS island [Fig. 8(a)]. After OOH* dissociation, O* remains adsorbed on top of the edge B, while OH* adsorbs on top of the nearest B atom as shown in Fig. 8(b). It is also important to note that the interaction of OOH* and O* with the OH-terminated B atom at the BNNS edge leads to the promotion of the interaction of the OH terminal group with the Au support. As a result, one OH terminal group bridges B and Au atoms (Fig. 8) at the interface between BNNS and Au(111). Thus, sites at the BNNS edges play a crucial role in OOH* stabilization and dissociation as we have been stressing.^31,35,36

It must be noted that all adsorption sites at the edges of BNNS/Au(111) are not equivalent with the spread of the adsorption energies of ORR intermediates of about 0.8 eV (Fig. 7), and the reactivity of these sites depend on size, shape, and adsorbed configuration of BNNS, allowing many ORR pathways with slightly different energetics.

We examined electrocatalytic activities for the oxygen reduction reaction (ORR) of Au electrodes modified by as prepared and size selected (0.45–1.0, 0.22–0.45, and 0.1–0.22 µm) h-BN nanosheet (BNNS) in O₂ saturated 0.5M H₂SO₄ solution. It was demonstrated that the overpotential was reduced by all the BNNS modifications, but enhancement was dependent on the size of BNNS used for modification: the smaller the size, the smaller the overpotential for ORR, i.e., the larger the ORR activity, in this size range. Maximum reduction of the overpotential, ∼330 mV compared to a bare Au electrode, was achieved by modifying the Au surface by the BNNS of smallest size (0.1–0.22 µm) used in the present study, and more than 80% of oxygen is reduced to water via the four-electron process at this electrode as determined by both K–L analysis and RRDE measurements. Theoretical calculations showed that the four-electron reduction of oxygen to H₂O does not proceed both at bare Au and BNNS/Au because the dissociation of OOH* along the ORR pathway to H₂O is uphill, but it becomes downhill at the edges of BNNS islands adsorbed on Au(111) as OOH* is mildly stabilized, and the dissociated O*…OH* intermediate is more strongly stabilized, resulting in the opening of the four-electron reduction pathway of oxygen to H₂O. The decrease in the size of BNNS islands resulted in an effective increase in the number of the catalytically active sites and, hence, in the increase in the ORR electrocatalytic activity at the BNNS/Au electrode.

See the supplementary material for EDX analysis of deposited BNNS, size dependent electrochemical parameters, optimized structures for OH-terminated 3 × 3 BNNS on the Au(111) surface with and without adsorbed O2, and all relevant structures of the ORR intermediates along the five different reaction paths.

The present work was initiated with the support of the Elements Science and Technology Project on “Nano-hybridized Precious-metal-free Catalysts for Chemical Energy Conversion” and partially supported by the Development of Environmental Technology using Nanotechnology and World Premier International Research Center (WPI) Initiative on Materials Nanoarchitectonics (MANA) and Chemical Reaction Design and Discovery (ICReDD) from the Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan. Calculations were partially performed using computational resources of the Numerical Materials Simulator of NIMS and the MASAMUNE–IMR supercomputer at the Center for Computational Materials Science, Institute for Materials Research, Tohoku University (Project No. K2101). SEM and TEM measurements were performed at the NIMS Battery Research Platform with the help of Mr. K. Shinoda (TEM) and Ms. M. Oshida (SEM).

The authors have no conflicts to disclose.

Hung Cuong Dinh: Investigation (lead); Writing – original draft (equal). Ganesan Elumalai: Formal analysis (equal); Investigation (supporting). Hidenori Noguchi: Formal analysis (equal); Writing – original draft (equal); Writing – review & editing (equal). Andrey Lyalin: Formal analysis (equal); Investigation (equal); Writing – original draft (equal). Tetsuya Taketsugu: Formal analysis (supporting); Supervision (supporting). Kohei Uosaki: Conceptualization (equal); Project administration (lead); Supervision (lead); Validation (lead); Writing – review & editing (lead).

The data that support the findings of this study are available within the article and its supplementary material.