Surface pH Effects on Catalytic Behavior of Pyridinic Nitrogen on Nitrogen-doped Carbon Nanotube in CO₂ Electrochemical Reduction

Abstract

Nitrogen-doped carbon nanotubes (NCNTs) have been considered a promising catalyst for the electrochemical reduction of CO₂ (CO₂ER) to generate CO. Although pyridinic N sites have been suggested to be the active center of NCNTs, their behavior in the reaction remains unclear because of the lack of experimental evidence. Herein we focused on the pH dependence of CO₂ER activity of NCNT and investigated the effects of local pH at the electrode surface to estimate the catalytic role of the pyridinic N. The results of the in situ local pH measurements using surface-enhanced Raman spectroscopy (SERS) revealed that CO₂ER activity disappears in an acidic environment at pH below 4. SERS detected no CO species at the surface during the reaction in the acidic electrolyte, and ex situ X-ray photoelectron spectroscopy indicated the protonation of the pyridinic N. These results suggest the protonation of pyridinic N, the active site of NCNT, inhibits the CO₂ adsorption and the following reduction to define the catalytic activity.

1. Introduction

Nowadays, the carbon-neutral transition is considered one of the most important tasks for industry all over the world, because the increasing CO₂ emission is the primary cause of global climate change.^1,2 Among the proposed strategies for decreasing CO₂ emission, the electrochemical reduction of CO₂ (CO₂ER) into valuable feedstocks for chemicals or fuels is an attractive approach due to its high product selectivity at mild temperatures and pressures.^3–5 When powered by electricity from renewable sources such as solar or wind, CO₂ER can offer a production process with net-zero, or even net-negative CO₂ emissions. To date, several kinds of CO₂ER products have been reported such as CO, formate, methane, and ethylene.^4–7 Especially, CO is a useful feedstock to produce valuable chemicals and fuels such as methane, methanol, and olefins by combination with conventional catalytic reactions.^5,8,9

Noble metals such as Au and Ag exhibit high selectivity and over 90 % Faradaic efficiency (FE) when used as catalysts for CO production via CO₂ER.^10–13 Toward implementation of this technology on a commercial scale, materials composed of earth-abundant elements are more favorable as a catalyst from a perspective of price and availability. In this context, nitrogen-doped carbon materials have been attracting much attention recently for their high CO selectivity comparable to Au and Ag.^14,15 In particular, nitrogen-doped carbon nanotubes (NCNT) are promising because of their high conductivity and unique electronic properties derived from the geometric features.¹⁶ For example, NCNT prepared by the chemical vapor deposition (CVD) method displayed FE of 80 % for CO at a low overpotential of −0.26 V.¹⁷ Pyridine-grafted NCNT prepared through the oxidation followed by calcination with melamine exhibited FE over 90 %.¹⁸ Just a calcination of a multi-walled carbon nanotube with 1,10-phenanthroline gave NCNT exhibiting FE of nearly 100 % for CO in a gas phase reaction using a flow cell.¹⁹ However, those NCNT catalysts have an issue of low current density compared with noble metals, and further improvement of the activity is still necessary.

Although increasing the density of the active site is a general strategy to improve the catalyst activity, applying such an approach to NCNTs is currently difficult because of the unclear catalytic mechanism and unknown active site. In previous studies, the content of pyridinic nitrogen indicated a positive correlation with catalytic performance.^18,19 DFT calculations also suggested that the pyridinic nitrogen is the most active for CO₂ binding and reduction among the nitrogen defects in NCNT because of the less energy barrier for the electron transfer step.²⁰ Because CO₂ binding to the pyridinic nitrogen should be driven by its basicity derived from the lone pair, the local pH at the NCNT surface would significantly affect the interaction between CO₂ and pyridinic nitrogen defining the CO₂ER activity. The analysis of the local pH and the chemical state of the NCNT surface during the reaction will bring us crucial experimental evidence of the catalytic mechanism of CO₂ER which has not been obtained so far.

In this study, the correlation between the surface pH and the CO₂ER activity of NCNT with pyridinic N is investigated. As for the local pH analysis, we have previously developed the in situ measurement system utilizing surface-enhanced Raman spectroscopy (SERS).²¹ Employing this technique allows us to obtain information on pH at the NCNT electrode surface which is different from the bulk pH, in an environment quite close to the real condition. Combined with the evaluation of the electrochemical performance and the chemical state analyses, the catalytic mechanism such as the active site, and the determining factor of the CO₂ER activity of NCNT is estimated.

2. Experimental

2.1 Catalyst preparation

The NCNT catalyst was synthesized by modifying a previously published method.¹⁹ Pristine multi-walled carbon nanotube (MWCNT) with a diameter of 30–50 nm was purchased from Nanjing XFNano Materials Tech Co., Ltd., and soaked in 1.0 M nitric acid to remove metal contaminants followed by filtration and vacuum drying. All chemicals, including 1,10-phenanthroline monohydrate were purchased from FUJIFILM Wako Pure Chemical Corp. unless otherwise indicated. In 50 mL of a 50 vol% ethanol aqueous solution, 600 mg of 1,10-phenanthroline monohydrate was dissolved. Subsequently, 200 mg of MWCNTs were added and stirred for 12 h at 25 °C. The obtained dispersion was heated to evaporate the solvent at 100 °C for 3 h, and the resulting solid was ground into a powder in an agate mortar. Finally, the yielded precursor was calcinated under an N₂ atmosphere with a flow rate of 100 sccm at 700 °C for 3 h in an electric furnace. The temperature was programmed to rise at 5 °C min⁻¹ and fall at 3.3 °C min⁻¹, respectively.

2.2 Electrochemical measurements

Electrochemical measurements were carried out in two types of electrochemical systems: liquid-phase electrolysis using an H-cell (EC Frontier Co., Ltd.) separated by porous glass with a three-electrode system, and gas-phase electrolysis using a fuel-cell type reactor with membrane-electrode assembly (MEA) using a polymer electrolyte. In the H-cell measurements, a Pt wire with 0.2 mm diameter (The Nilaco Corporation) and an Ag/AgCl electrode (BAS Inc.) were used as counter electrode and reference electrode, respectively. The measured potentials were rescaled to the RHE according to the equation, E (vs. RHE) = E (vs. Ag/AgCl) + 0.199 V + 0.0591 V × pH. The working electrode was prepared by the drop-casting of catalyst ink onto the glassy carbon substrate. Specifically, 4 mg NCNT and 1 mg of 20 wt% Nafion^TM DE2020CS solution were dispersed in 200 µL n-propanol with sonication for 30 min to form homogeneous ink. The resulting ink was loaded onto a 4 cm² (2 cm × 2 cm) glassy carbon electrode and dried at room temperature. Before electrochemical measurements, the electrode was vacuum-dried at room temperature for 15 min to evaporate the residual solvent. Three kinds of different aqueous solutions were used as the electrolyte: 1.0 M KHCO₃, 1.0 M KCl, and 1.0 M KHSO₄. Each electrolyte (20 mL) was placed in both chambers of the H-cell, and CO₂ (>99.995 %, Taiyo Nippon Sanso Corp.) was bubbled over 30 min to remove air and obtain CO₂-saturated solution. Electrolysis was conducted under sealed conditions with continuous CO₂ bubbling at the rate of 50 sccm, and the potential was controlled by a potentio-/galvano-stat (HZ-7000, Hokuto Denko Corp.). The gaseous products released in the 10 min of reaction were sampled and analyzed by Agilent 990 Micro GC with a thermal conductivity detector which was calibrated with external standard gases.

For the experiments using the fuel-cell type reactor, the MEA was fabricated by combining catalyst-coated substrates (CCS) and an ion exchange membrane. The CCS was prepared by a similar drop-casting as described above. The cathode ink containing 20 mg NCNT and 10 mg ionomer in 1.0 mL n-propanol was loaded onto a carbon paper with a micro-porous layer (Pyrofil^TM MFK-A, Mitsubishi Chemical Corp.) to form a gas diffusion electrode (GDE). The anode was prepared in the same way with the ink made of 7.5 mg IrO₂ (Furuya Metal Co., Ltd.) and 2.8 mg ionomer in 400 µL n-propanol. The ionomers and the ion exchange membranes were selected from two types of combinations according to the required condition. The first combination is Nafion^TM DE2020CS and Nafion^TM NR212 (The Chemors Co.), the proton exchange type, which makes the catalyst surface acidic. Another was Sustainion^TM XA-9 and Sustainion^TM X37-50 (Dioxide Materials, Inc.), the anion exchange type, which gives the alkaline condition. The membrane was sandwiched between the CCSs for the cathode and the anode with PTFE gaskets for sealing and assembled using 5 cm² conventional fuel-cell hardware. Carbon and Ti plates were used as the current collectors for the cathode and anode, respectively. CO₂ (>99.995 %, Taiyo Nippon Sanso Corp.) was supplied directly at a rate of 150 sccm to the cathode, while 0.1 M of CO₂-saturated KHCO₃ was circulated at 20 mL/min to the anode. The gaseous products released in the 3 min of reaction were sampled and analyzed by Agilent 990 Micro GC.

2.3 In situ measurements of adsorbed species

The adsorbed species on the NCNT surface during CO₂ER were analyzed using SERS. Because NCNT does not exhibit any plasmonic enhancement effects that are observed on the roughened surface of metals such as Ag, Au, and Cu, the technique called “shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS)” was applied.^22,23 Au nanoparticles (NPs) with a diameter of ∼55 nm covered with a SiO₂ layer of ∼4 nm (Au@SiO₂) were synthesized using a previously described method,²³ loaded onto a 2 cm² NCNT electrode, and vacuum dried at room temperature for 15 min. The obtained Au@SiO₂-coated electrode was installed as a working electrode in the homemade measurement cell shown in Fig. 1a, and electrolysis was conducted with a continuous CO₂ flow at 10 sccm. This flow rate was employed to prevent the fluctuation of the liquid level from disturbing the Raman spectra, and other electrochemical conditions are the same as the measurements in H-cell. Raman spectra were measured through 3D laser Raman microspectroscopy by using a confocal optical system (Nanofinder 30, Tokyo Instruments, Inc.). The electrode surface was irradiated with He-Ne laser (633 nm) through a flat cover glass (Matsunami Glass Ind., Ltd.) to excite Raman scattering. The pinhole size of the confocal system was set to 50 µm, and the grating was 600 grooves/mm. The laser power and the data acquisition time per sample were 13.0 mW and 30 s, respectively, and an objective lens with a 50× magnification and NA (numerical aperture) of 0.5 (Olympus Corp.) was used.

Figure 1.

Experimental setup for the in situ SERS measurements. (a) Overview and cross-sectional view of measurement system for surface species detection using SHINERS method. (b) Overview and cross-sectional view of local pH measurement system with p-MBA-modified Ag NPs sensor.

2.4 In situ local pH measurements

The in situ local pH measurement was performed by modifying our original system described previously.²¹ In the homemade measurement cell shown in Fig. 1b, a 2 cm² NCNT electrode is installed as a working electrode, vertically to the pH sensor. This geometry allows us to adjust the location of the measurement spot accurately. Because CO₂ER involves a CO₂ supply and the generation of gaseous products, the top of the measurement cell in the current study was designed with a slope that prevents gas bubbles from accumulating under the pH sensor. The SERS-active pH sensor was fabricated by the electroless deposition of Ag NPs and the modification of p-mercaptobenzoic acid (p-MBA) as described in the previous paper.²¹ The substrate of the sensor was a flat cover glass (18 mm × 18 mm with a thickness of ∼170 µm) purchased from Matsunami Glass Ind., Ltd. The back surface of the pH sensor was irradiated with the 532 nm line from He-Ne laser to obtain Raman spectra of the modified p-MBA molecules on the Ag NPs. The location of the laser spot was adjusted at the interface between the electrolyte and the working electrode as shown in the cross-sectional view of Fig. 1b. The laser power and the data acquisition time per plot were 5.0 mW and 5 s, respectively, and other settings for the Raman measurements were the same as described before.

To obtain the pH calibration curve, the Raman measurements with this setup filled with standard pH solutions were conducted. For standard solutions with pH ranging from pH 2.0 to 12.0, Britton–Robinson buffer solutions were prepared via previously described method.²⁴ The standard solution with pH 1.0 was prepared by diluting the H₂SO₄ aqueous solution.

2.5 Characterization of the N species of NCNT

For the analysis of the chemical state of N defects of NCNT catalyst, X-ray photoelectron spectroscopy (XPS) was performed by PHI VersaProbe III (Ulvac-Phi, Inc.), using a monochromatic Al-Kα radiation. For C 1s and N 1s scans, the pass energy and step size were set to 55 eV and 0.1 eV respectively, and a dwell time was 200 ms. The binding energy for all peaks was referenced to the C 1s peak at 284.6 eV. The N 1s spectrum was fitted by the combination of the peak components represented by the Gaussian function after smoothing by the Savitzky-Golay method.

3. Results and Discussion

3.1 pH effects on the CO generation activity of NCNT

To evaluate the pH effects on the CO generation activity of NCNT, CO₂ER using the H-cell was performed. The pH was adjusted by using different electrolytes, 1.0 M KHCO₃, 1.0 M KCl, and 1.0 M KHSO₄. The bulk pH of each electrolyte with CO₂-saturation before the electrolysis was 7.66, 3.81, and 0.55 respectively. Cation species of electrolytes have been reported to affect the current density in CO₂ER due to the difference in their buffering capacities of local pH,²⁵ or the hydrated sizes that influence the interaction between reaction intermediates and the electrode.²⁶ Therefore, the cation of the electrolyte in this study was fixed as K⁺ to exclude the effects on the CO generation activity. The linear sweep voltammograms (LSV) taken at the scanning rate of 50 mV s⁻¹ (Figs. 2a and 2b) indicate that for every electrolyte, onset potential was near 0 V (vs. RHE), and the reduction current increased when the potential became negative. However, the curve of the electrolysis in KHCO₃ has an increasing inflection at −0.5 V, whereas the others not (Fig. 2b). According to the transitions in FE in each electrolyte described in Figs. 2c to 2e, only the electrolysis in KHCO₃ involves CO generation at −0.5 V, so the inflection observed in Fig. 2b should be attributed to CO₂ER. This onset potential is almost identical to that reported in 0.5 M NaHCO₃ electrolyte.¹⁹ Although the LSV curve for the case with KCl does not show any distinct inflection, CO generation was observed in KCl at negative potentials below −0.7 V (Fig. 2d). Interestingly, CO (with a FE over 80 %) was detected in both KHCO₃ and KCl when negative potentials were applied, whereas only H₂ was detected and the CO yield was less than detection limit at any potential in KHSO₄ (Fig. 2e). These results suggest that the proceeding reaction at the NCNT electrode is completely switched from CO₂ER to the hydrogen evolution reaction (HER) in KHSO₄. This phenomenon could be attributed to multiple factors. Lower pH means a higher H⁺ concentration, which promotes HER rather than CO₂ER as previously reported.²⁷ The bicarbonate concentration also matters, because CO_2(aq), the electrochemically active substrate in the liquid-phase CO₂ER, is supplied from the rapid equilibrium between bicarbonate and CO_2(aq).²⁸ Though the solubility of CO₂ in the aqueous solution depends on temperature and pressure but not on pH, the bicarbonate concentration decreases when the pH decreases, and almost no bicarbonate is in the solution with pH < 4, according to the equilibrium between CO_2(aq) and bicarbonate.²⁹ It will decrease the CO₂ER activity under extremely low pH because of the lack of substrate supply. In addition, low pH could cause some changes in the chemical state of the active site to switch the dominant reaction. Of course, the counter anion of each electrolyte would affect the interaction between CO₂ and the catalyst by acting as the H⁺ donor or poisoning the surface.³⁰

Figure 2.

Electrochemical CO₂ reduction using an H-cell. (a), (b) LSV curves at 50 mV s⁻¹ for electrolysis in (i) 1.0 M KHCO₃, (ii) 1.0 M KCl, and (iii) 1.0 M KHSO₄; Faradaic efficiency of electrolysis at various potentials in (c) 1.0 M KHCO₃, (d) 1.0 M KCl, and (e) 1.0 M KHSO₄.

Among the possible causes for low CO₂ER activity in the electrolyte with low pH listed above, the influence of the substrate supply and the counter ions is excluded by the gas phase reaction with abundant active CO₂. Herein, a fuel cell-type reactor with an MEA composed of an ion exchange ionomer and membrane was used to supply CO₂(g) to the catalyst via a porous layer of carbon paper.³¹ By using a proton or anion exchange ionomer and membrane, the pH can be controlled at the electrode surface.³² As shown in Fig. 3, the reaction using the MEA with the anion exchange ionomer and membrane (AEM-type) exhibited a high FE of CO over a wide range of potentials. On the other hand, the MEA with the proton exchange ionomer and membrane (PEM-type) showed less FE of CO, up to 20 %. The amount of CO₂ supply in these experiments, 6.7 mmol min⁻¹, surpasses by far the converted amount up to 45 µmol min⁻¹ (the calculated value from the result of AEM-type cell with 2.8 V of the cell voltage applied, Fig. 3b). Comparing the FE in 1.0 M KHSO₄ (H-cell, Fig. 2e) and the PEM type cell (Fig. 3a), it is suggested that the abundant CO₂ supply contributed to the improvement in FE even in the acidic environment. However, the observed FE in the PEM-type cell is still lower than that in the AEM-type cell and HER is dominant. Such difference should be attributed to other factors derived from pH rather than the substrate supply or the anion effects.

Figure 3.

Faradaic efficiency and current density at various potentials in electrochemical CO₂ reduction using fuel-cell type reactors composed of (a) proton exchange membrane and ionomer, and (b) anion exchange membrane and ionomer.

3.2 In situ measurements of adsorbed species

The surface species on NCNT surface in each electrolyte during CO₂ER was confirmed by in situ SERS measurements with the SHINERS method. The SERS spectra of the NCNT surface in the electrolysis at −0.7 V (vs. RHE) using three types of electrolytes, 1.0 M KHCO₃, 1.0 M KCl, and 1.0 M KHSO₄ are shown in Fig. 4. A distinct peak at ∼1830 cm⁻¹ was observed during the electrolysis in KHCO₃ and KCl, while such a peak was not detected during the electrolysis in KHSO₄. This peak is attributed to the adsorbed CO on the NCNT surface based on previous Raman measurements of metal catalysts such as Au³³ or Ag^34,35 and stronger bonding between CO₂ and the active site, N or C. This indicates the absence of bonding or binding between the intermediate and active sites of the NCNT, confirming that CO₂ reduction does not occur in KHSO₄. One of the most probable causes of such change would be the local pH which affects the first formation of the coordination bond between CO₂ and the pyridinic N before the reduction to CO species. The CO₂ adsorption onto the pyridinic N depends on the acid-base reaction based on the basicity of the N atom with a lone pair, so the low pH at the electrode surface will inhibit the adsorption by changing the electronic state of N. Although direct evidence of the coordination bond of N–C has not been observed in the SERS measurement so far because of its weak Raman intensity, the DFT calculation in the previous study supports the assumption above by the result that the CO₂ adsorbs onto the pyridinic N in the NCNT surface.²⁰

Figure 4.

In situ detection of the surface adsorbed species on the NCNT electrode in CO₂ electrochemical reduction using (a) 1.0 M KHCO₃, (b) 1.0 M KCl, and (c) 1.0 M KHSO₄ as the electrolyte.

3.3 In situ local pH measurements

To investigate the effect of pH at the electrode surface in each electrolyte, the in situ local pH measurements were conducted by using our original SERS system described previously.²¹ This system detects pH change from the intensities of peaks corresponding to the deprotonated form of the carboxylic group in p-MBA modified to the surface of Ag NPs on flat glass. For a more precise quantification of the local pH, a calibration curve was constructed by evaluating the response toward Britton–Robinson buffer solutions as pH standards in the setup shown in Fig. 1a. As shown in Fig. 5a, the relative peak intensity at ∼1390 cm⁻¹ corresponding to COO⁻ $(I_{(\text{COO}^{ - })})$, the deprotonated form of the carboxylic group, increases along with the pH increase, while that at ∼1700 cm⁻¹ corresponding to C=O of COOH (I_(COOH)), the protonated form, decreases. According to the pKa equation of p-MBA below (1), pH is described as a linear function of the logarithm of the ratio between the protonated and deprotonated form, with the description of [A] as the deprotonated form.³⁶   

\begin{equation} \text{p$K$a} = -{\log}(\text{[A][H$^{+}$]/[AH$^{+}$]}) = -{\log}(\text{[A]/[AH$^{+}$]}) + \text{pH} \end{equation} (1)

Herein, if [A] and [AH⁺] are proportional to $I_{(\text{COO}^{ - })}$ and I_(COOH) respectively, the following Eq. 2 is described based on (1), with a constant X.   

\begin{equation} \text{pH} = \log(I_{\text{(COO${^{-}}$)}}/I_{\text{(COOH)}}) + \text{X} \end{equation} (2)

Based on the spectra in Fig. 5a, the actual relationship between pH and $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ was derived and shown in Fig. 5b. A part of the curve, from pH 4 to 7, showed the linear correlation following Eq. 2, while the other parts deviated from the linear. The main cause of such deviation in the region of low and high pH is supposed to be the broadening effect of the peaks nearby.³⁶ In Fig. 5a, the peaks at ∼1390 cm⁻¹ and ∼1700 cm⁻¹ are next to other peaks. Despite background subtraction, the intensity of the spectra cannot be completely reduced to zero, even at extremely low or high pH, owing to the effect of the neighboring peaks. This hinders the linear correlation between the peak intensities and corresponding concentrations. However, the value of $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ increases with increasing pH with few fluctuations, so the local pH was estimated by this calibration curve in this study.

Figure 5.

pH response of the p-MBA modified Ag NPs sensor employed in the in situ measurement setup. (a) SERS spectra of the p-MBA modified Ag NPs sensor contacted to Britton-Robinson buffer solutions with various pH, (b) pH calibration curve of the p-MBA modified Ag NPs sensor by plotting the peak intensity ratio $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ based on the spectra shown in (a).

The in situ local pH measurement system was applied to the CO₂ER by NCNT electrode in different electrolytes, 1.0 M KHCO₃, 1.0 M KCl and 1.0 M KHSO₄. The applied potential was gradually increased from −0.5 V to −1.0 V (vs. RHE), and the electrolysis was conducted at each potential for 200 s. The obtained spectra after application of each potential are shown in Figs. 6a to 6c, and the transitions in the $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ values are summarized in Fig. 6d. In every electrolyte, the local pH at 0 s was similar to the bulk pH. According to the XPS analysis, the N content in the NCNT surface was 1.6 at%. Although pyridinic N is a base, its content should be even smaller than that, which is supposed to have little effect on the local pH value near the NCNT electrode surface. In KHCO₃ and KCl, the clear increasing trends of $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ especially at the beginning of the electrolysis were observed. It indicates [H⁺] decrease due to the H⁺ consumption or OH⁻ production by CO₂ER or HER. The estimated local pH ranges for KHCO₃ and KCl were 6–11 and 3–5, respectively, and the rate of increasing of $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ in KHCO₃ was higher than that in KCl, as shown by the order of the current densities shown in Figs. 2c and 2d. The local pH increase in each electrolyte appears saturated below −0.8 V, while the current density transition in Fig. 2a suggests that H⁺ consumption or OH⁻ production still increases. Such difference could be due to the low response of the peaks to pH, especially in the strongly basic region away from the pKa of p-MBA. The increase in current density also accompanies an increase in bubble generation at the electrode surface, which could agitate the area near the electrode surface, mitigating and disturbing the pH increase. The buffering capacity of KHCO₃ based on the pKa of HCO₃⁻ also matters. HCO₃⁻ shows the following equilibrium in aqueous solutions.   

\begin{equation} \text{HCO$_{3}{}^{-}$} + \text{OH$^{-}$} \leftrightarrows \text{CO$_{3}{}^{2-}$} + \text{H$_{2}$O} \end{equation} (3)

Because the pKa value of this equilibrium is 10.33,³⁷ the local pH increase should be mitigated around 10.33 via the formation of CO₃²⁻ from HCO₃⁻ and the generated OH⁻. A similar difference between the current density and local pH change was also observed in the electrolysis in KHSO₄ where HER was the dominant reaction. Although higher current density should lead to greater H⁺ consumption to bring higher local pH at the electrode surface, the increase in the local pH looks saturated, which could be attributed to the similar reason as described in the results of KHCO₃ and KCl. In KHSO₄ solution, HSO₄⁻ releases H⁺ according to the surrounding pH as follows.   

\begin{equation} \text{HSO$_{4}{}^{-}$} \leftrightarrows \text{SO$_{4}{}^{2-}$} + \text{H$^{+}$} \end{equation} (4)

The pKa value of this reaction is 1.99,³⁷ which corresponds to the estimated range of the local pH at the NCNT electrode surface in KHSO₄. Thus, the released H⁺ would be supplied to the electrode surface to compensate for the H⁺ consumption by HER and mitigate the local pH increase. In addition, $\log (I_{(\text{COO}^{ - })}/I_{(\text{COOH})})$ in KHSO₄ fluctuated reflecting the spectral changes shown in Fig. 6c. Such a fluctuation mainly comes from the low resolution in the extremely low pH compared to the neutral region as described above. Especially in the case of the electrolysis at negative potentials, the end of the measurement, the stability of Ag NPs in the acidic solution with the continuous bubble formation could also matter. However, the maximum local pH was estimated not to exceed 4. Because pH 4 is within the high resolution region of this measurement system, the maximum local pH during electrolysis in KHSO₄ was considered to be accurately measured.

Figure 6.

In situ local pH measurements of electrochemical CO₂ reduction. SERS spectra during the electrolysis at various potentials in (a) 1.0 M KHCO₃, (b) 1.0 M KCl, and (c) 1.0 M KHSO₄, and (d) pH estimation based on the calibration curve in each measurement, (i) 1.0 M KHCO₃, (ii) 1.0 M KCl, and (iii) 1.0 M KHSO₄.

From the estimated local pH ranges of the electrolysis in KHSO₄, it is considered that CO generation does not occur below pH 4. Moreover, comparing the estimated local pH transition in KCl with the FE of CO shown in Fig. 2d, it is indicated that CO begins to generate at −0.7 V (vs. RHE) after the local pH increase saturated and completely exceeded pH 4 in Fig. 6d. Because of the short reaction time, the local pH at the electrode surface does not reach 4 and thus, no distinct reduction peak corresponding to CO₂ER is observed in the LSV for electrolysis in KCl (Fig. 2b). These results suggest that the reaction at the NCNT surface shifts to HER from CO₂ER below the local pH of approximately 4.

3.4 Characterization of the N species of NCNT

To examine whether the low pH caused any changes in the chemical state of the catalyst surface, an ex situ XPS analysis was performed on the NCNT powder soaked in each electrolyte. In this experiment, each electrolyte was used without CO₂ saturation to focus on the effect of the pH of the electrolyte rather than the interaction with CO₂. The pH value of each is as follows: 8.96 for 1.0 M KHCO₃, 7.00 for 1.0 M KCl, and 0.55 for KHSO₄. The obtained XPS spectra are shown in Fig. 7. According to the peak attribution based on the previous studies,^38,39 the N species in NCNT are composed of pyridinic N (398.7 eV) and pyridonic or pyrrolic N (400.1 eV), and the pyridinic N derived from 1,10-phenanthroline is proved to be retained after calcination. Although the peak for pyridonic N or pyrrolic N is dominant in each sample, their portion has been reported not to affect the CO₂ER activity.¹⁹ It is noteworthy that the protonated pyridinic N (401.7 eV) was detected only in the sample soaked in KHSO₄ (pH 0.58). It suggests that pyridinic N on the NCNT surface is protonated before CO₂ER under extremely low pH. Because protonated pyridinic N was detected even in the ex situ experiment, it may be observed even more in situ, where the sample is not neutralized during the measurement process.

Figure 7.

XPS spectra of N 1s for NCNT soaked in (a) 1.0 M KHCO₃, (b) 1.0 M KCl, (c) 1.0 M KHSO₄, and (d) H₂O.

3.5 Mechanism estimation

Based on the experimental findings in this study, the mechanism by which the local pH at the electrode surface defines the CO₂ER activity of the NCNT was proposed. As described above, it is indicated that the protonation of pyridinic N occurs in an acidic environment before the electrolysis. Whether the protonation occurs or not should be defined by the acid-base equilibrium. Because the pKa of 1,10-phenanthroline is 4.84,³⁷ pyridinic N on the NCNT surface would also be protonated at a pH of approximately 4.84. This pKa value is almost identical to the pH at the interface where CO₂ER activity switches to HER observed during in situ pH measurements, the local pH around 4. Therefore, it is suggested that the protonation or deprotonation of the pyridinic N should define the CO₂ER activity of NCNT.

A possible mechanism for determining reaction activity during NCNT catalysis is shown in Fig. 8. As discussed so far, two different reactions, CO₂ER and HER, compete during the electrolysis. In CO₂ER, CO₂ adsorbs on pyridinic N, driven by the basicity of N. Subsequently, the adsorbed CO₂ is reduced through the electron transfer steps, and CO is released finally. In HER, the pyridinic N is protonated according to the pKa followed by the electron transfer step to release H₂ finally. When the local pH at the electrode surface is higher than pKa, the equilibrium of pyridinic N protonation is shifted to the deprotonated form, and it decreases HER and increases CO₂ER. Therefore, in an electrolyte with higher pH like KHCO₃, the FE of CO increases. On the other hand, when the local pH at the surface is lower than pKa, the equilibrium of the pyridinic N is shifted largely to the protonation, and it is supposed to be almost irreversible. It inhibits CO₂ adsorption, because there is no lone pair on N to make use of, which makes NCNT inert to CO₂. This could be the reason for poor CO generation in the electrolysis using acidic electrolytes: KHSO₄ in H-cell and PEM in fuel-cell type reactor. Based on newly obtained experimental evidence in this study, the proposed catalysis mechanism supports the theory that the active site of NCNT is pyridinic N.

Figure 8.

Scheme showing the pH effects on the reaction involving the pyridinic nitrogen on the NCNT surface.

4. Conclusion

In this work, pH effects on the CO₂ER activity of NCNT were investigated for the estimation of the active site and the catalysis mechanism. Electrolysis with electrolytes of different pH revealed a pH dependence in both the liquid and gas phases. Poor CO₂ER activity was observed during electrolysis in acidic media, and the absence of adsorbed CO species suggested that a low local pH inhibits CO₂ adsorption. Moreover, in situ SERS measurements successfully detected the local pH at the NCNT electrode surface during CO₂ER, suggesting that the reaction shift occurred at a local pH of approximately 4. XPS analyses indicate the protonation of pyridinic N in an acidic condition, and the pKa of pyridinic N was almost identical to the pH at the interface where CO₂ER activity switches to HER observed in the in situ local pH measurements. Thus, it was estimated that the protonation/deprotonation of pyridinic N based on pKa defines the CO₂ER activity of NCNT. These findings also suggest the active site of NCNT should be pyridinic N with experimental evidence. This study provides a clear strategy for the improvement of N-doped carbon catalysts targeting pyridinic N, and the optimization of the reaction condition to bring high CO₂ER activity.

CRediT Authorship Contribution Statement

Kohei Ide: Conceptualization (Lead), Data curation (Lead), Investigation (Lead), Methodology (Lead), Writing – original draft (Lead)

Masahiro Kunimoto: Conceptualization (Supporting), Investigation (Supporting), Methodology (Supporting), Writing – review & editing (Lead)

Kota Miyoshi: Conceptualization (Supporting), Supervision (Supporting), Writing – review & editing (Supporting)

Kaori Takano: Conceptualization (Supporting), Supervision (Supporting), Writing – review & editing (Supporting)

Koji Matsuoka: Conceptualization (Supporting), Supervision (Lead), Writing – review & editing (Supporting)

Takayuki Homma: Conceptualization (Supporting), Project administration (Lead), Supervision (Lead), Writing – review & editing (Lead)

Conflict of Interest

The authors declare no conflict of interest in the manuscript.

Footnotes

K. Ide, M. Kunimoto, K. Miyoshi, K. Takano, and K. Matsuoka: ECSJ Active Members

T. Homma: ECSJ Fellow

References

• 1)  D. M. D’Alessandro, B. Smit, and J. R. Long, Angew. Chem., Int. Ed., 49, 6058 (2010).
• 2)  F. A. Rahman, M. M. A. Aziz, R. Saidur, W. A. Wan, A. Baker, M. R. Hainin, R. Putrajaya, and N. A. Hassan, Renewable Sustainable Energy Rev., 71, 112 (2017).
• 3)  A. Saravanan, P. Senthil kumar, D.-V. N. Vo, S. Jeevanantham, V. Bhuvaneswari, V. A. Narayanan, P. R. Yaashikaa, S. Swetha, and B. Reshma, Chem. Eng. Sci., 236, 116515 (2021).
• 4)  F. Li, D. R. MacFarlane, and J. Zhang, Nanoscale, 10, 6235 (2018).
• 5)  P. De Luna, C. Hahn, D. Higgins, S. A. Jaffer, T. F. Jaramillo, and E. H. Sargent, Science, 364, eaav3506 (2019).
• 6)  Y. Hori, Modern Aspects of Electrochemistry, Springer, New York, NY, p. 89 (2008).
• 7)  K. P. Kuhl, E. R. Cave, D. N. Abram, and T. F. Jaramillo, Energy Environ. Sci., 5, 7050 (2012).
• 8)  H. T. Luk, C. Mondelli, D. C. Ferré, J. A. Stewart, and J. Perez-Ramirez, Chem. Soc. Rev., 46, 1358 (2017).
• 9)  S. Rönsch, J. Schneider, S. Matthischke, M. Schlüter, M. Götz, J. Lefebvre, P. Prabhakaran, and S. Bajohr, Fuel, 166, 276 (2016).
• 10)  W. Zhu, R. Michalsky, Ö. Metin, H. Lv, S. Guo, C. J. Wright, X. Sun, A. A. Peterson, and S. Sun, J. Am. Chem. Soc., 135, 16833 (2013).
• 11)  M. Liu, Y. Pang, B. Zhang, P. De Luna, O. Voznyy, J. Xu, X. Zheng, C. T. Dinh, F. Fan, C. Cao, F. P. García de Arquer, T. S. Safaei, A. Mepham, A. Klinkova, E. Kumacheva, T. Filleter, D. Sinton, S. O. Kelley, and E. H. Sargent, Nature, 537, 382 (2016).
• 12)  S. Liu, H. Tao, L. Zeng, Q. Liu, Z. Xu, Q. Liu, and J.-L. Luo, J. Am. Chem. Soc., 139, 2160 (2017).
• 13)  M. Ma, K. Liu, J. Shen, R. Kas, and W. A. Smith, ACS Energy Lett., 3, 1301 (2018).
• 14)  D. M. Fernandes, A. F. Peixoto, and C. Freire, Dalton Trans., 48, 13508 (2019).
• 15)  T. Ma, Q. Fan, H. Tao, Z. Han, M. Jia, Y. Gao, W. Ma, and Z. Sun, Nanotechnology, 28, 472001 (2017).
• 16)  X. Wang, Y. Liu, D. Zhu, L. Zhang, H. Ma, N. Yao, and B. Zhang, J. Phys. Chem. B, 106, 2186 (2002).
• 17)  J. Wu, R. M. Yadav, M. Liu, P. P. Sharma, C. S. Tiwary, L. Ma, X. Zou, X.-D. Zhou, B. I. Yakobson, J. Lou, and P. M. Ajayan, ACS Nano, 9, 5364 (2015).
• 18)  Y. Zhang, H. Jiang, D. Niu, I. Manke, C. Yang, M. Zhu, X. Zhang, and R. Chen, J. Mater. Chem. A, 10, 1852 (2022).
• 19)  C. Ma, P. Hou, X. Wang, Z. Wang, W. Li, and P. Kang, Appl. Catal., B, 250, 347 (2019).
• 20)  P. P. Sharma, J. Wu, R. M. Yadav, M. Liu, C. J. Wright, C. S. Tiwary, B. I. Yakobson, J. Lou, P. M. Ajayan, and X.-D. Zhou, Angew. Chem., Int. Ed., 54, 13701 (2015).
• 21)  K. Ide, M. Kunimoto, S. Yoshida, M. Yanagisawa, and T. Homma, Electroanalysis, 34, 1682 (2022).
• 22)  J. F. Li, Y. F. Huang, Y. Ding, Z. L. Yang, S. B. Li, X. S. Zhou, F. R. Fan, W. Zhang, Z. Y. Zhou, D. Y. Wu, B. Ren, Z. L. Wang, and Z. Q. Tian, Nature, 464, 392 (2010).
• 23)  J. F. Li, X. D. Tian, S. B. Li, J. R. Anema, Z. L. Yang, Y. Ding, Y. F. Wu, Y. M. Zeng, Q. Z. Chen, B. Ren, Z. L. Wang, and Z. Q. Tian, Nat. Protoc., 8, 52 (2013).
• 24)  H. T. S. Britton and R. A. Robinson, J. Chem. Soc., 1456 (1931).
• 25)  F. Zhang and A. C. Co, Angew. Chem., Int. Ed., 59, 1674 (2020).
• 26)  S. Ringe, E. L. Clark, J. Resasco, A. Walton, B. Seger, A. T. Bell, and K. Chan, Energy Environ. Sci., 12, 3001 (2019).
• 27)  B. Kim, S. Ma, H.-R. M. Jhong, and P. J. A. Kenis, Electrochim. Acta, 166, 271 (2015).
• 28)  M. Dunwell, Q. Lu, J. M. Heyes, J. Rosen, J. G. Chen, Y. Yan, F. Jiao, and B. Xu, J. Am. Chem. Soc., 139, 3774 (2017).
• 29)  M. König, J. Vaes, E. Klemm, and D. Pant, iScience, 19, 135 (2019).
• 30)  A. Xu, N. Govindarajan, G. Kastlunger, S. Vijay, and K. Chan, Acc. Chem. Res., 55, 495 (2022).
• 31)  K. Yang, R. Kas, W. A. Smith, and T. Burdyny, ACS Energy Lett., 6, 33 (2021).
• 32)  M. Ünlü, J. Zhou, and P. A. Kohl, J. Phys. Chem. C, 113, 11416 (2009).
• 33)  J. L. Oberst, H.-R. M. Johng, P. J. A. Kenis, and A. A. Gewirth, J. Solid State Electrochem., 20, 1149 (2016).
• 34)  K. G. Schmitt and A. A. Gewirth, J. Phys. Chem. C, 118, 17567 (2014).
• 35)  T. H. Wood and M. V. Klein, J. Vac. Sci. Technol., 16, 459 (1979).
• 36)  S.-Q. Xiang, L. Zhang, S.-T. Gao, and L.-B. Zhao, J. Raman Spectrosc., 50, 1065 (2019).
• 37)  J. R. Rumble, CRC Handbook of Chemistry and Physics (102nd ed.), CRC Press, Inc., Boca Raton, FL (2021).
• 38)  D. Guo, R. Shibuya, C. Akiba, S. Saji, T. Kondo, and J. Nakamura, Science, 351, 361 (2016).
• 39)  K. Takeyasu, M. Furukawa, Y. Shimoyama, S. K. Singh, and J. Nakamura, Angew. Chem., Int. Ed., 60, 5121 (2021).