Electroreduction of CO₂ From Flue Gas: Impurity Tolerance and Mechanistic Insights in Molecular Catalysis

Abstract

Direct utilization of diluted CO₂ from industrial flue gas containing SO₂, NO₂ and O₂ impurities is economically appealing and circumvents capture and purification prior to conversion. We present experimental data and density functional theory (DFT) calculations for understanding the effect of those impurities during CO₂RR catalyzed by a model Re molecular catalyst, [Re(bpy)(CO)₃Cl] (bpy = bipyridine). Under both 10% and 1% v/v CO₂ gas streams, high selectivity toward CO production was maintained, with faradaic efficiency (FE_CO) above 90% and 70%, respectively. O₂ reduction reaction (ORR) on the electrode surface represents a substantial competitive reaction that accounts for ≈80% of the charge consumed when a realistic CO₂ source mimicking an industrial waste‐incinerator stream (10% v/v CO₂, 10% v/v O₂, 100 ppm NO₂, and 50 ppm SO₂ in N₂ matrix) is used. Neither NO₂ nor SO₂ incorporated in the diluted CO₂ stream provoke a significant decrease in the CO production (FE_CO ≥ 82%). DFT calculations indicate low affinity of the catalyst for impurities, because they are in their reduced form at the CO₂RR potential. This is experimentally proven by cyclic voltammetry and ion chromatography. These findings call for new strategies that enhance the moderate O₂‐tolerance exhibited by Re complex on CO₂RR from flue gas.

Impurity‐tolerance exhibited by Re molecular complex for CO² conversion from flue gas containing 10% CO², 10% O², 100 ppm NO² and 50 ppm SO² in N² matrix. This tolerance is based on electrochemical reduction of gas impurities on the electrode surface under CO² electrolysis conditions.

1. Introduction

The electrochemical reduction reaction of carbon dioxide (CO₂RR) has emerged as a promising route to transform CO₂ into value‐added e‐fuels [1, 2, 3, 4]. The 2e⁻/2H⁺ reduction of CO₂ can generate CO, a product that holds industrial relevance as a component of synthetic gas (syngas), which serves as a feedstock in processes such as Fischer‐Tropsch and water‐gas‐shift (WGS) and as a building block for different industrial commodity chemicals such as methanol and acetic acid [5, 6]. Nowadays, the flue gases emitted by industrial exhausts (such as power plants, cement plants, and incinerators) represent point sources of CO₂ [7]. However, unlike the pure CO₂ sources commonly used for laboratory studies, these exhausts contain low share of CO₂ (4%–25% v/v) alongside other compontents such as water, N₂, O₂, and gaseous impurities such as NO _x and SO _x  [8]. In this case, catalysts are required to selectively convert CO₂ over other competitive reactions such as hydrogen evolution reaction (HER) and oxygen reduction reaction (ORR), as well as NO _x /SO _x reduction [9, 10, 11, 12, 13, 14]. Molecular electrocatalysts offer a high product selectivity and the possibility of tuning them via synthetic ligand modification. For this reason, they have been deeply investigated as catalysts for CO₂RR [15, 16]. Among them, [Re(bpy)(CO)₃Cl] (bpy = bipyridine), herein refered as Re complex, was introduced in the 80's by J. M. Lehn as a model molecular catalyst for the photoreduction and electrochemical reduction of CO₂ to CO [17]. Re(I) complex represents a model molecular catalyst for CO₂RR studies thanks to its well‐established mechanism [18, 19], high selectivity for CO production and almost negligible H₂ production from HER. Thus, the use of the model Re(I)‐molecular catalyst for CO₂RR under realistic industrial conditions by direct utilization of low‐concentrated CO₂ from a flue gas source is of great interest, since it avoids the energy intense and costly steps of CO₂ capture and concentration prior to conversion [7].

Ishitani and collaborators [20, 21, 22] have explored different variations in the functional groups within the Re complex, making the catalyst suitable for the reduction of low‐concentrated CO₂ gas streams (down to 1% CO₂), but the effects of O₂, NO₂, and SO₂ impurities, have not been addressed yet. Recently, the multiple scientific challenges present in the direct CO₂RR from industrial flue gases, such as those related to low concentration of CO₂ and the presence of impurities, have been reviewed [23]. However, the effects of impurities (NO₂ and SO₂) have been mainly studied in the context of heterogeneous systems with deleterious results towards CO₂RR [9, 24]; meanwhile, the use of homogeneous molecular complexes for the CO₂RR under conditions mimicking those of industrial gas effluents remains limited [25, 26, 27]. Particularly scarce are those dealing with the presence of O₂ in the gas stream during CO₂RR catalyzed by a molecular complex. A notable example stands out: using a Fe‐phorphyrin catalyst with four ferrocene groups, kinetic studies revealed that CO₂ reacts with the catalyst 500 times faster than O₂, leading to selective conversion of CO₂ to CO while any O₂ was reduced to benign H₂O [27]. Some heterogenized molecular catalysts for CO₂RR have been explored under flue gas conditions. As examples, a cobalt phthalocyanine heterogenized together with vinylpyridine polymer onto carbon black nanoparticles allowed a FE for CO of 90% under a diluted CO₂‐stream [28], as well as a cobalt phthalocyanine immobilized on carbon nanotubes with a microporous polymer on top [29]. Thus, it is of interest to extend such studies to a model molecular catalyst such as the Re complex, owing to its well defined CO₂RR mechanism. So far, only the catalytic reduction of pure nitrous oxide (N₂O) gas to N₂ has been deeply studied using the Re complex [30, 31].

In this work, we aim to evaluate the electrocatalytic performance of the Re complex for CO₂ to CO conversion under an environment which mimics the flue gas composition of an industrial waste‐incinerator [32, 33] containing diluted CO₂ (1%–10% v/v), O₂ (1%–10% v/v), and impurity traces of NO₂ (0–1000 ppm) and SO₂ (0–1000 ppm). To the best of our knowledge, this is the first electrochemical investigation of a Re‐based molecular catalyst operating under these conditions with controlled impurity levels, thereby addressing a gap in the use of homogeneous molecular systems for CO₂RR under realistic flue gas environments. Thus, we explore herein by cyclic voltammetry (CV) and controlled potential electrolysis (CPE) the behavior of the Re complex as a molecular platform during CO₂RR under a sequence of different gas mixtures by introducing each component, individually and in combination, and quantifying its effect on the faradaic efficiency and energy efficiency (EE) of CO production. Finally, density functional theory (DFT) calculations provide additional mechanistic insights regarding the exceptional tolerance of Re complex during CO₂RR in the presence of NO₂ and SO₂ impurities, as well as the role of O₂ and the contribution of ORR.

2. Results

2.1. Molecular Catalyst Characterization Under Diluted CO₂ and O₂ by Cyclic Voltammetry

Figure 1 shows the cyclic voltammograms of a 0.5 M solution of [TBA][PF₆] in CH₃CN containing 1 mM of [Re(bpy)(CO)₃Cl] (Re complex) on bare glassy carbon electrode (GCE) under inert conditions (argon‐saturated, black plot). Previous reports describing the mechanism of Re complex under inert conditions attribute the first reversible wave centered at −1.72 V versus Fc⁺/Fc to the one‐electron reduction on the bpy ligand (bpy/bpy^•−) and the second wave located at −2.04 V versus Fc⁺/Fc to the quasireversible metal center reduction (Re^I/Re⁰) [5, 18, 19]. Since proton‐coupled electron transfer is required for CO₂ reduction [19], 1.5 M of trifluoroethanol (TFE), a weak Brønsted acid (pKa = 35.4) [19], was added to the solution as a proton source. Upon TFE addition, the catalytic response of Re complex under different concentrations of CO₂ with or without the simultaneous presence of O₂ was studied. Under 100% CO₂ (red plot), a catalytic wave for CO₂ reduction with an onset potential at −1.76 V versus Fc⁺/Fc appeared [5]. Upon dilution of CO₂ to 10% (blue plot) and 1% (purple plot), the intensity of the catalytic current densities (j _cat) decreases, as expected because of the limited reactant availability. In addition to this, Figure 1 also includes the effect of diluted CO₂ in the presence of O₂, both gases at 10% v/v, (green plot). This CV exhibits two well separated reduction processes: (i) a new irreversible reduction peak at −1.35 V versus Fc⁺/Fc associated with ORR and (ii) the catalytic peak previously displayed under 10% CO₂ with an onset potential at −1.76 V versus Fc⁺/Fc and associated to the CO₂ reduction to CO. The different catalytic current density reached by those two peaks (0.75 mA cm⁻² and 3.5 mA cm⁻² for ORR and CO₂RR, respectively) is mainly related to the different type of process taking place: direct conversion on the GCE surface for ORR and mediated conversion in solution for the CO₂RR. Moreover, 35 times lower solubility displayed by O₂ than CO₂ in acetonitrile might also play a role (Table S1) [34, 35].

FIGURE 1.

Cyclic voltammograms on a bare GCE disc of 1 mM of Re complex and 0.5 M [TBA][PF₆] either in CH₃CN solution under argon (black plot) or in CH₃CN solution containing 1.5 M TFE as a proton source under 100% CO₂ (red plot), 10% CO₂ (blue plot), 1% CO₂ (purple plot), and 10% CO₂ + 10% O₂ (green plot), the latter three in a N₂ matrix. Scan rate 0.1 V s⁻¹.

Table 1 (entries a–c) reports the j _cat/j _p ratio values derived from the CVs shown in Figure 1, with j _cat being the maximum catalytic current density and j _p the current density under Ar gas. As it can be observed from the different ratios, 30.7, 13.7, and 2.2 for 100, 10, and 1% CO₂ gas stream, respectively, the decrease of j _cat/j _p is not a linear function of the CO₂ concentration, showing that even at 1% CO₂, the catalyst still produces a current only 15 times lower than that obtained with 100% CO₂. In addition to this, the j _cat/j _p ratio using 10% v/v CO₂ in the presence of 10% v/v O₂ (green plot) is only slightly decreased as compared to the ratio obtained in the absence of O₂ (entries d with 12.5 vs. b with 13.7).

TABLE 1.

Comparison of the catalytic parameters from CV of 1 mM Re complex in CH₃CN solution containing 1.5 M TFE as a proton source under different CO₂ atmospheres and flue gas compositions. Potentials are referred versus Fc⁺/Fc. Catalytic potential (E _cat/2) corresponds to the potential value at half of the catalytic current. The overpotential (η) was determined from the difference between E _cat/2 and E ⁰CO₂/CO (CH₃CN, TFE) = −1.36 V versus Fc⁺/Fc [19].

Entry	CO2 (%) in N2 matrix	O2 (%) in N2 matrix	Impurities (ppm)	E cat/ 2 (V)	η (V)	j cat/j p
a	100	—	—	−1.96	0.60	30.7
b	10	—	—	−1.87	0.51	13.7
c	1	—	—	−1.76	0.40	2.2
d	10	10	—	−1.90	0.54	12.5
e	10	—	NO2, 100	−1.87	0.51	13.7
f	10	—	SO2, 50	−1.93	0.57	12.2
g	10	10	NO2, 100	−1.94	0.58	12.7
h	10	10	SO2, 50	−1.92	0.56	11.4

Figure S1 presents different voltammograms for studying the ORR associated peak in the presence and the absence of the Re complex under 10% (v/v) O₂ in a N₂ matrix. Figure S1 shows an almost identical reduction wave in potential and current density with an onset of −1.1 V versus Fc⁺/Fc, in the presence and the absence of the Re complex. Therefore, these results showed that this reduction peak originates from the ORR on the GCE surface and is not mediated by the Re complex. In addition, the effect of varying CO₂ concentration in the presence of 1 mM Re complex and 10% v/v O₂ in a N₂ matrix was studied via CV on a bare GCE. The CVs in Figure S2 show that neither the onset potential for the CO₂‐related catalytic peak (onset at −1.76 V vs. Fc⁺/Fc) nor the one corresponding to ORR (onset at −1.1 V vs. Fc⁺/Fc) are shifted with higher CO₂ concentration in the gas stream. The only evident change in CV corresponds to the increase of current density of the CO₂RR catalytic peak, consistent with greater reactant availability. However, the O₂ reduction signal remains unchanged, indicating that ORR proceeds under these conditions, independently of the concentration of CO₂ available.

2.2. Molecular Catalyst Characterization Under Simulated Flue Gas by Cyclic Voltammetry

Figure 2a,b compare the CVs under a diluted CO₂ atmosphere (10% v/v in a N₂ matrix), in the absence and the presence of either NO₂ introduced at 100 and 500 ppm (Figure 2a) or SO₂ at 50 and 500 ppm (Figure 2b), two common flue gas impurities. As previously stated, this specific CO₂ concentration, together with a low presence of NO₂ and SO₂ (100 ppm and 50 ppm, respectively), reflect realistic conditions typically found in gas effluents of industrial waste‐incinerators. Higher NO₂ and SO₂ impurity levels were introduced to test the tolerance limit of the catalyst. The results from these CVs show that in all tested conditions, the catalytic current is about the same and thus CO₂RR operates without interference from NO₂ and SO₂ impurities under the tested conditions. Figure 2c,d present the CVs under a more representative flue gas composition (10% v/v CO₂ + 10% v/v O₂ in a N₂ matrix with the addition of either NO₂ or SO₂, respectively). The reduction of diluted CO₂ remains unchanged, and ORR process appears mainly insensitive to the presence of NO₂ (Figure 2c) and SO₂ (Figure 2d). Table 1 (entries e–h) reports the catalytic parameters (E _cat/2, η and j _cat/j _p ) calculated from CVs displayed in Figure 2. It shows that in both low‐concentrated CO₂ streams (10% v/v CO₂ and 10% v/v CO₂ + 10% v/v O₂), the presence of either SO₂ or NO₂ impurities at low concentrations (50–100 ppm range) has a minimal impact in all three catalytic parameters.

FIGURE 2.

Cyclic voltammogram of 1 mM of Re complex in a solution of 0.5 M [TBA][PF₆] in CH₃CN containing 1.5 M TFE as a proton source, recorded under 10% v/v CO₂ in a N₂ matrix with varying impurity levels: a) NO₂: 0 ppm (blue plot), 100 ppm (green), and 500 ppm (red); b) SO₂: 0 ppm (blue), 50 ppm (green) and 500 ppm (red); c) 10% v/v CO₂, 10% v/v O₂ and different concentration of NO₂, following the same color code than in (a); and d) 10% v/v CO₂, 10% v/v O₂ and different concentration of SO₂, the same color code than in (b). Scan rate 0.1 V s⁻¹.

The electroreductions of NO₂ (Figure S3a) and SO₂ (Figure S3b) in the presence of the Re complex, but in the absence of CO₂ and O₂, are also studied. These catalytic processes are analogous to the CO₂ to CO conversion and could be of interest in flue gas streams. Moreover, the effect of increasing the NO₂ and SO₂ concentration beyond the usual concentration limit found in typical waste‐incinerator flue gas is also reported. For the electroreduction of 100 ppm NO₂ (Figure S3a) and 50 ppm SO₂ (Figure S3b), a poor catalytic current from Re complex is observed (onset potential −1.75 V vs. Fc⁺/Fc). Increasing the SO₂ concentration beyond 500 ppm (Figure S3b) provokes the appearance of an additional reduction peak with an onset of −1.0 V versus Fc⁺/Fc, which might be associated to direct reduction of SO₂ on GCE. To support that, additional CVs studying the reduction of NO₂ and SO₂ directly on the GCE are also included (Figure S4a, b). In particular, Figure S4b exhibits a clear reduction peak at an onset of −1.0 V versus Fc⁺/Fc that fits well the second reduction peak found in Figure S3b.

2.3. Density Functional Theory Calculations on the Interaction of Different Flue Gas Components and [Re(bpy)(CO)₃Cl]

It is well established in the literature that NO _x and SO _x gas species might compete with CO₂ in reduction processes catalyzed by molecular complexes [9], which in some cases might reduce catalytically gas impurities such as NO₂ and SO₂. However, the cyclic voltammetry results reported in the previous section suggest that Re complex catalytic intermediate does not significantly react neither with NO₂ nor SO₂, and no competition is established with CO₂. Therefore, density functional theory (DFT) calculations were set up for reaching a deeper mechanistic insight. Table 2 shows the calculated binding energies of all flue gas components (CO₂, O₂, SO₂, and NO₂), as well as the binding energies of some of their corresponding reduction products (NO₂ ⁻, SO₂ ^•−, O₂ ^•− and H₂O₂), with the catalytic active form of the Re complex. Those calculations show that the Re⁰ center in the active form of the catalyst exhibits lower affinity (less favorable binding energy) for CO₂ compared to NO₂ and SO₂, which should come associated with a significant decrease in CO production from CO₂RR in the presence of either NO₂ or SO₂ impurities. However, the CV data shown in Figure S4 and the calculated reduction potentials for these molecules on the electrode surface (0.0 and −1.0 V vs. Fc⁺/Fc, respectively) indicate that both NO₂ and SO₂ molecules are electrochemically reduced on the electrode surface at the required potential for CO₂RR mediated by the Re complex (ca. −1.8 to −1.9 V vs. Fc⁺/Fc). The one electron reduction product of NO₂ and SO₂ produces new ionic species NO₂ ⁻ and SO₂ ^•−, respectively. Given their anionic nature, these species exhibit less favorable binding energy (Table 2) than CO₂ and are less prone to bind to the Re⁰ center in the anionic active form of the catalyst, which should explain why NO₂ and SO₂ molecules do not compete with CO₂ binding in flue gas under electrocatalytic conditions. Data from Table 2 are summarized graphically in Figure S5.

TABLE 2.

Calculated DFT thermodynamic data for the interaction of selected flue gas components and their corresponding reduction products with the active form of the Re complex.

Species in solution	Calculated binding energy to the active form of Re complex (2e−) ΔG (kcal mol−1)
CO2	+3.0
O2	−20.5
SO2	−25.6
NO2	−46.2
NO2 −	+14.9
O2 •−	+9.3
SO2 •−	+6.3
H2O2	+3.5

The binding energy of O₂ with the active form of the Re complex is also included in Table 2. Similarly, DFT calculations show that the metal center in the catalyst active form exhibits higher affinity for O₂ than to CO₂. However, as happened with NO₂ and SO₂, the O₂ does not occupy the Re⁰ active site, because ORR is favored directly at the GCE surface and depletes all available O₂ prior to the reduction of the Re complex, as experimentally shown by CV (Figure 2 and Figure S1). In addition, the O₂ reduction products formed from one or two electron transfers, O₂ ^•− and H₂O₂, respectively, exhibit less favorable binding energy to the active form of the Re complex than CO₂ (Table 2).

2.4. Effect of Diluted CO₂ on CO₂RR by Controlled Potential Electrolysis

The catalytic performance of the Re complex was further evaluated under flue gas conditions by bulk electrolysis at constant potential (controlled potential electrolysis (CPE)) in batch conditions (solution saturated with the corresponding CO₂ stream at the beginning of the electrolysis). A total charge of 14 C was circulated in all CPEs, leading to a different electrolysis duration depending on the current displayed at the applied electrolysis potential in each case. Different gas streams, with different concentrations of CO₂ and with or without the addition of O₂, NO₂, and/or SO₂, were evaluated on GCE, as summarized in Figure 3 and Table 3. CO and H₂ were the only reduction products detected in the gas phase after CO₂ electrolysis, with the absence of any additional product from CO₂RR in solution. Under both 100% and 10% v/v CO₂ gas streams, high selectivity toward CO production was obtained after CPE at −1.95 and −2.05 V versus Fc⁺/Fc, with FE_CO above 90% (Figure 3, red and blue bars, respectively) and minimal HER. When the CO₂ concentration was reduced to 1% (Figure 3, purple bar, CPE at −1.8 V vs. Fc⁺/Fc), high CO selectivity was still obtained (FE_CO = 72 ± 5%). However, a poor total FE ≤ 75% was achieved in this case, which seems to indicate that a portion of the total charge is being consumed by processes not accounted by our product analysis.

FIGURE 3.

Selection of FE for H₂ (black bars) and CO (color bars) production obtained after controlled potential electrolysis (CPE) by applying the corresponding electrolysis potentials indicated in Table 3 and using 1 mM Re complex in CH₃CN + 1.5 M TFE as proton source under various gas streams as indicated in the top X‐axis. Error bars for FE_CO are indicated in black, and for FE_H2 in blue. The presence of NO₂ and/or SO₂ is indicated in the bottom X‐axis. The corresponding EE_CO (%, red squares) is indicated on the secondary right Y‐axis. Stirring rate 300 rpm.

TABLE 3.

CPE at different potentials versus Fc⁺/Fc of 1 mM Re complex and [TBA][PF₆] in CH₃CN solution containing 1.5 M TFE in a two compartments H‐type cell under different CO₂ concentration in a N₂ matrix, with the presence of O₂ (10% v/v) and/or NO₂ and SO₂ gas impurities when indicated. Total circulated charge 14 C. Solution stirring rate 300 rpm.

CO2 (%) in N2 matrix	O2 (%) in N2 matrix	Presence of impurities, (ppm)	E (V)	FECO (%)a	FEH2 (%)b	EECO (%)c
100	—	—	−1.95	91 ± 5	2 ± 1	62 ± 5
10	—	—	−1.95	92 ± 4	1 ± 1	63 ± 5
10	—	—	−2.05	92 ± 2	1 ± 1	61 ± 1
1	—	—	−1.80	72 ± 5	2 ± 1	54 ± 4
10	—	NO2, 100	−2.05	90 ± 2	2 ± 1	60 ± 1
10	—	NO2, 500	−2.05	84 ± 5	1 ± 1	56 ± 4
10	—	NO2, 1000	−2.05	88 ± 3	1 ± 1	58 ± 2
10	—	SO2, 50	−2.05	91 ± 3	2 ± 1	60 ± 2
10	—	SO2, 500	−2.05	85 ± 4	1 ± 1	56 ± 3
10	—	SO2, 1000	−2.05	82 ± 1	1 ± 1	54 ± 1
10	—	NO2, 100 SO2, 50	−2.05	94 ± 2	1 ± 1	62 ± 1
10	10	—	−2.05	22 ± 3	1 ± 1	14 ± 2
10	10	NO2, 100 SO2, 50	−2.05	20 ± 3	2 ± 1	13 ± 2
30	10	—	−2.05	21 ± 3	1 ± 1	14 ± 2
50	10	—	−2.05	20 ± 2	1 ± 1	14 ± 1

^a,b Experimental faradaic efficiency for CO (FE_CO) and H₂ (FE_H2) determined via gas chromatography after a total circulated charge of 14 C

^c Energy efficiency for CO (EE_CO) calculated as EE (%) = E _T /E × FE_CO where E _T is the thermodynamic potential for CO₂ reduction to CO (E ⁰ _CO2/CO (CH₃CN, TFE) = −1.36 V vs. Fc⁺/Fc), E (V) is the experimental cathode potential during electrolysis, and FE_CO is the faradaic efficiency for CO production.

Figure S6 shows the evolution of current density over time during CPE under different CO₂ concentrations: 100%, 10%, and 1% v/v CO₂ in N₂ matrix. Moreover, two different applied potentials are reported for the 10% v/v CO₂ gas stream. These current profiles show a continued diminution of current density all along the electrolysis, which is expected for electrolysis in batch conditions (solution saturated with the corresponding CO₂ stream at the beginning of the electrolysis) due to the progressive consumption and decrease in reactant available over time. As anticipated from the CV data in Figure 1, the current density signal displayed in Figure S6 shows an intensity that depends on the initial CO₂ concentration, with the highest values observed for 100% v/v CO₂ and decreasing progressively for 10% and 1% v/v CO₂. This behavior confirms the concentration‐dependent catalytic response relaying on the CO₂ availability. Using 10% v/v CO₂ gas stream a lower current density was obtained at −1.95 V compared to that at −2.05 V versus Fc⁺/Fc (blue and brown plot in Figure S6, respectively). Thus, all additional CPE experiments shown in Figure 3 and Table 3 were performed at −2.05 V versus Fc⁺/Fc in order to evaluate the effects of flue gas components without any additional effect associated with the applied potential.

2.5. Effect of NO₂ and SO₂ Impurities on CO₂RR by Controlled Potential Electrolysis

Under flue gas composition, the presence of 100 ppm NO₂ or 50 ppm SO₂ impurities in a 10% v/v CO₂ gas stream did not significantly affect either the FE_CO value, which remained at 90%, or the EE_CO value, which was kept at 60% (Figure 3 and Table 3). Notably, the combination of 10% v/v CO₂, 100 ppm NO₂ and 50 ppm SO₂ (Figure 3, navy blue bar and Table 3), which approaches the composition of an industrial flue gas effluent, preserved high CO selectivity (FE_CO = 94 ± 2% and EE_CO = 62 ± 1%), confirming the great tolerance of the Re catalyst under these impurity conditions. Figure S7a (NO₂),b (SO₂) show the current density profiles over time recorded during CPE under different gas compositions in 10% v/v CO₂. The presence of increasing concentrations of NO₂ (100, 500, and 1000 ppm) in the gas mixture did not significantly affect or alter the intensity of the current density compared to the impurity‐free conditions (0 ppm), indicating no observable impact of NO₂ impurity on the catalytic response over the electrolysis time. The presence of SO₂ in the gas mixture below 1000 ppm did not significantly impact the current density evolution over time during electrolysis. However, once 1000 ppm of SO₂ were present in the gas stream, a minimal deleterious effect on the current density was observed (orange plot in Figure S7b). These observations are consistent with both the CV results (Figure 2a,b) and the FE_CO data from Table 3, which also show minimal variation in the catalytic performance of the Re complex in the presence of these impurities.

Ion chromatography was also used to analyze any potential byproduct derived from either NO₂ or SO₂ electrochemical reduction on the GCE. In particular, nitrite anion (NO₂ ⁻) from NO₂ reduction during electrolysis was detected by ion chromatography. However, the one electron reduction product from SO₂ considered in the DFT calculations (SO₂ ^•−) is not a stable anion in solution and was not detected. Figure S8a shows the ion chromatography data collected from a blank experiment where 5% v/v NO₂ in N₂ matrix was bubbled in CH₃CN solution to demonstrate that neither NO₂ ⁻ (retention time 10.34 min) nor NO₃ ⁻ (retention time 15.75 min) were initially present in the solution, contrary to what has been reported in aqueous solution [13]. Figures S8b,c show the ion chromatography data from two independent electrolysis under a 10% v/v CO₂ gas stream containing NO₂ at 100 ppm (black plot) and 1000 ppm (red plot) from the anolyte and the catholyte, respectively. The presence of NO₂ ⁻ formed during the CPE is confirmed in both compartments, being more evident in the anolyte. Moreover, the total amount of NO₂ ⁻ detected is, as expected, larger in the electrolysis using 1000 ppm of NO₂ (red plot) than 100 ppm of NO₂ (black plot).

2.6. Effect of Simulated Flue Gas on CO₂RR by Controlled Potential Electrolysis

The introduction of 10% v/v O₂ in the gas matrix containing 10% v/v CO₂ provokes an increase in the current density during the electrolysis (Figure S9, green plot), but led to a sharp decline in CO selectivity during CO₂RR electrolysis (FE_CO = 22%), as observed in Table 3 and Figure 3 (green bar). This is attributed to the competitive ORR taking place at more anodic potential, as shown by CV in Figure 1 (green plot) and Figure S1 (green plot). However, the Re complex remains active for CO₂RR to CO, since ORR does not consume all the available electrons. This is probably because the applied potential during electrolysis in the presence of Re complex is negative enough to ensure the constant production of the Re active intermediate all along the electrolysis. Then, Re complex represents a suitable system to evaluate the effect of the complete flue gas composition by introducing both 50 ppm SO₂ and 100 ppm NO₂, together with 10% v/v O₂ and 10% v/v CO₂ (Figure S9, black plot). The use of this simulated flue gas composition presents no additional deleterious effect to the one observed upon 10% v/v O₂ incorporation in the gas stream. Therefore, the FE_CO and EE_CO remain unchanged at around 20% and 13%, respectively (Table 3 and Figure 3, olive bar). Comparing the current density evolution over time in the presence and absence of NO₂ and SO₂ impurities under the low‐concentrated CO₂ conditions containing O₂ shows no difference (Figure S9, green and black plots), in accordance with the CVs shown in Figure 2c,d. Finally, a strategy based on increasing CO₂:O₂ ratio was tested to improve the CO₂RR performance in the presence of O₂ at 10% (v/v). Additional electrolysis in Table 3 show the results of increasing the CO₂ concentration from 10% (v/v) to 30% and 50%, while keeping the concentration of O₂ at 10% (v/v) in the N₂ matrix, likewise fail to increase the selectivity towards CO (FE_CO = 20%, Table 3). Therefore, neither the increase of CO₂ availability nor the presence of NO₂ or SO₂ at the chosen concentrations affect the thermodynamically favored ORR.

3. Discussion

The CV and bulk electrolysis data presented here demonstrate that the catalytic performance of the Re complex for CO₂ conversion to CO (selectivity and efficiency) is not affected either by decreasing the CO₂ concentration in the gas stream to 10% v/v (FE_CO ≥ 90 and EE_CO ≥ 60%) or by adding traces of NO₂ (100 ppm) and SO₂ (50 ppm) in such a low‐concentrated CO₂ stream (10% v/v) (Figure 2 and Table 3). With such a complex gas mixture, we obtained performances comparable to that with pure CO₂ (FE_CO = 94% and EE_CO = 62%, Table 3). A very minor effect is only observed for SO₂ far beyond its typical flue gas concentration (1000 ppm) in the diluted CO₂ gas stream (Figure 2 and Table 3). This could be due to the competitive reduction of SO₂ directly on the GCE as experimentally observed (Figures S3 and S4) and theoretically modeled. In contrast, the Re complex catalytic performance is significantly impacted by the presence of O₂ in the gas stream (10% v/v) (Figure 1), and a substantial portion of the circulated charge during CPE is consumed for ORR taking place on the GCE (Table 3, FE_CO = 22%). ORR presents a reduction wave in CV at an onset of −1.1 V versus Fc⁺/Fc, in the presence and the absence (Figure S1) of the Re complex in solution, which means that ORR takes place on the GCE surface, and is not mediated by the Re complex. This peak in CV associated to ORR remains mainly insensitive to the simultaneous presence of NO₂ and SO₂ under flue gas composition (Figure 2). Thus, the presence of 10% v/v O₂ provokes a severe deleterious effect in FE_CO during flue gas electrolysis (FE_CO = 20% and EE_CO = 13%). Nevertheless, the Re complex exhibits a moderate O₂‐tolerance thanks to its ability to remain active for the CO₂RR keeping its CO selectivity in the presence of O₂ [26, 27]. Control electrolyses increasing the ratio of (CO₂:O₂) in the gas stream from (1:1) up to (5:1) do not overcome this limitation.

The DFT results show that O₂, NO₂, and SO₂ present high affinity for the Re complex and could favorably compete with CO₂ for binding to the catalytically active form of the complex (Table 2 and Figure S5). However, at the required potential for CO₂RR mediated by the Re complex (ca. −1.8 to −1.9 V vs. Fc⁺/Fc), O₂, NO₂, and SO₂ molecules are electrochemically reduced on the electrode surface, which produces new ionic species O₂ ^•−, NO₂ ⁻ and SO₂ ^•− presenting much less affinity for the Re complex than CO₂ as reported in Table 2. The electroreduction of NO₂ on GCE and the formation of NO₂ ⁻ has been confirmed by ion chromatography as shown in Figure S8.

4. Conclusions

Achieving CO₂RR directly using a gas stream released from industrial plants, without CO₂ purification, is a great challenge. The competing reduction reactions occurring during CO₂RR to CO electrolysis of simulated flue gas, where the CO₂ accounts for a small part of the gas stream and additional reducible gases such as O₂, NO₂, and SO₂ are also present in solution, might impact on CO₂RR selectivity and efficiency. We were able to demonstrate the ability of the model Re molecular catalyst to sustain CO₂RR to CO without degradation under a realistic CO₂ source by mimicking an industrial waste‐incinerator stream (10% v/v CO₂, 10% v/v O₂, 100 ppm NO₂, and 50 ppm SO₂ in N₂ matrix). The high NO₂ and SO₂ impurities tolerance demonstrated by Re complex (FE_CO = 94%) is contrasted by a moderated O₂‐tolerant CO₂RR catalytic performance (FE_CO = 20%). DFT calculations demonstrated the high affinity of impurities for the Re complex (NO₂ > SO₂ > O₂ > CO₂), as well as their instability under electrolysis conditions due to their reduction on the electrode surface before activating the Re complex.

In conclusion, the deep mechanistic understanding of the flue gas impurities role during CO₂RR provided here points molecular catalysts, such as the Re complex, as an attractive alternative to achieve efficient and selective CO₂RR and mitigate some of the deleterious impacts of direct usage of flue gas streams. Nevertheless, several challenges remain unsolved before allowing high conversion efficiency of gas stream input from practical/industrial CO₂ sources.

5. Experimental Section

5.1. Reactants and Characterization

All chemicals were purchased from commercial suppliers and used without further purification. Anhydrous acetonitrile of 99.8% purity (CH₃CN), tetrabutyl ammonium hexafluorophosphate ([TBA][PF₆]) of 99% purity, 2,2,2‐Trifuoroethanol (TFE, CF₃CH₂OH) of ≥99% purity, and ferrocene (98%) were purchased from Sigma‐Aldrich. The Re‐molecular catalyst Chlorotricarbonyl(2,2′‐bipyridine)rhenium(I), ([Re(bpy)(CO)₃Cl]), of 99% chemical purity was purchased from Strem Chemicals. Pure gases Ar (>99.99%), N₂ (>99.99%), CO₂ (>99.99%), and diluted gases 10% v/v CO₂ in a N₂ matrix, 10% v/v CO₂ + 10% v/v O₂ in N₂, 5% v/v NO₂ in N₂, and 5% v/v SO₂ in N₂ were custom‐prepared by Air Liquide.

Proton Nuclear Magnetic Resonance (NMR) was used to characterized the Re complex by using a Bruker AVANCE III 300 MHz spectrometer. The NMR spectrum of the complex is shown in Figure S10.

5.2. Simulated Incinerator Flue Gas Mixture

Prior to entering the electrochemical cell, CO₂ gas streams (100% CO₂ and 10% v/v CO₂) were passed through a gas purifier from Agilent (model CP17973) to remove residual oxygen and moisture. The complex gas mixtures used to purge the solutions were prepared by properly mixing gases at the appropriate ratio using mass flow controllers (Bronkhorst, EL‐FLOW Prestige, models FG‐201CV, 50 mL min⁻¹ or FG‐200CV, 1 mL min⁻¹). The gas flow rates were regulated online using FlowSuite 2 software. For all experiments involving the use of SO₂, a high‐pressure stainless‐steel regulator (DIM 200‐3‐5 T PURGE, AirLiquide) was used for the customized SO₂ cylinder to ensure safe gas delivery. The manometer includes a purge inlet, which helps to safely purge the system and avoid contamination and corrosion. The concentration range of the studied impurities was either 50, 500 or 1000 ppm of SO₂ or 100, 500 and 1000 ppm of NO₂, in 10% v/v CO₂ diluted in a N₂ matrix or 10% v/v CO₂ + 10% v/v O₂ in N₂.

5.3. Electrochemical Studies

5.3.1. Cyclic Voltammetry

All electrochemical experiments were performed on Biologic Science Instruments SAS potentiostats (SP‐300 or VSP‐300) at room temperature (20 ± 2°C) in acetonitrile (CH₃CN). [TBA][PF₆] was used as a supporting electrolyte in solution (0.5 M) and TFE (1.5 M) was used as the proton donor. These catalytic conditions were used as the optimal conditions [19]. A three‐electrode setup was used with a glassy carbon (GC) disc electrode (diameter = 3 mm, area = 0.07 cm², BioLogic) as the working electrode (WE). Prior to use, the GC electrode was polished on a polishing cloth (DP‐Nap 200 mm, Struers) on a 1 μm diamond suspension (Struers), sonicated in water for 10 s and then dried before starting the experiments. The counter electrode (CE) was a platinum wire (diameter = 0.5 mm, Alfa Aesar, 99.5% purity), which was previously flame‐annealed. The reference electrode (RE) was a conventional Ag/AgCl/KCl_sat, separated from the solution via a salt bridge. All potentials were calibrated by adding the internal standard ferrocenium/ferrocene (Fc⁺/Fc) redox couple at the end of each experiment. CVs were run at 0.1 V s⁻¹ scan rate and only the third steady state cycle of all CVs is shown. E _cat/2 (V) corresponds to the potential at half of the catalytic peak current. The cathodic overpotential (η) was calculated from the difference between E _cat/2 and E ⁰ _CO2/CO (CH₃CN, TFE) = −1.36 V vs. Fc⁺/Fc in acetonitrile [19]. Catalytic response from CV was determined by calculating the current density ratio (j _cat/j _p ), where (j _cat) corresponds to the catalytic current density in the presence of CO₂ (at all different concentrations and conditions) and TFE, and (j _p ) corresponds to the reduction peak current density under inert conditions (Ar).

5.3.2. CO₂ Conversion by Bulk Electrolysis

Chronoamperometric studies (controlled potential electrolysis) were carried out in a gastight two‐compartment electrochemical H‐type glass cell with a glass frit to separate anolyte and catholyte chambers. 9.5 mL of 0.5 M [TBA][PF₆] + 1.5 M TFE in CH₃CN was used in the catholyte, while 3 mL of 0.5 M [TBA][PF₆] was used in the anolyte. The indicated volumes were constant in all electrolysis reported here. The working electrode was a 1 cm² GC plate (1 mm thick, type 2, from Alfa Aesar), the counter electrode was a 5 cm² GC rod (Alfa Aesar), and the reference electrode was a Ag/AgCl/KCl_sat electrode within a salt bridge. To minimize ohmic losses in the cell, a minimal distance between electrodes was kept, and both H‐cell chambers were stirred during the electrolysis. In addition, 85% of the electrolyte resistance was compensated using the potentiostat's Ohmic drop compensation module. Controlled potential electrolysis (CPE) were performed by gas saturating both anolyte and catholyte chambers (with 100%–1% CO₂ or the different gas mixtures) for 5 and 30 min, respectively. The gas was humidified in a solution containing the electrolyte before entering the H‐cell compartments. No continuous CO₂ gas was purged during the electrolysis. One millimolar of the Re complex was added only in the catholyte. An additional gas trap open to the atmosphere to avoid overpressure containing 0.1 M NaOH was added at the outlet of the H‐cell compartments to avoid the release of these gases in the open atmosphere. A schematic illustration describing the electrochemical setup is included in Figure S11. All CPE experiments were performed with two or three replicates to check the reproducibility of results.

5.4. Analytical Quantification of Products

Gas products were quantified by gas chromatography (Model 8610C, SRI Instruments) equipped with a thermal conductivity detector (TCD) to detect H₂ and a flame ionization detector (FIC) to detect CO. Gas aliquots of volumes varying from 25 to 100 µL were taken from the headspace of both compartments in the H type cell and injected into the GC. The GC was periodically calibrated using a custom standard gas mixture in CO₂ (0–75 nmol H₂). Calibration curves are shown in Figure S12. An ion exchange chromatograph (IC) (Metrohm 883 Basic IC), automatized by a 863 Compact IC autosampler and equipped with a Metrosep A Supp 5 (250 × 4 mm) column and a conductivity detector, was used to evaluate liquid products, with results indicating no appreciable quantities of liquid products. Faraday efficiency (FE, %) for a giving product is calculated from the ratio between the charge consumed to form each product and the total charge passed [36]. A constant total circulated charge (14 C) was used in all electrolysis in order to reach a fair comparison among all CPE results. A correction in the total circulated charge is made to account for the initial two electron reduction of the Re complex (1 mM in solution) to generate its active form. Catalyst activation charge = [number of electrons × Faraday constant × mol of catalyst] = [2 × 96 485 × 9.5 × 10⁻⁶] = 1.83 C. With this correction, an effective circulated charge of 12.2 C for all bulk electrolysis is calculated. The cathodic half reaction energy efficiency (EE, %) was calculated for the production of CO from CO₂ using the following Equation (1) [37].

$$\text{Cathodic energy efficiency},\text{EE}\left(\mathrm{\%}\right)=({E^0}_{{\mathrm{CO}}_2/\mathrm{CO}}/E)\times {\text{FE}}_{\mathrm{CO}}$$ (1)

where E ⁰CO₂/CO is the thermodynamic potential in volts required for the electrocatalytic reduction of CO₂ to CO (E ⁰CO₂/CO (CH₃CN, TFE) = −1.36 V vs. Fc⁺/Fc) [19], whereas E and FE_CO represent the actual cathodic potential applied in volts (vs. Fc⁺/Fc) and the CO Faradaic Efficiency (%), respectively.

5.5. Computational Methods and Modeling (Density Functional Theory Calculations)

DFT calculations of the binding energy were used to evaluate the affinity between incinerator flue‐gas components (CO₂, O₂, NO₂, and SO₂) and the catalytically active form of the Re complex. Moreover, the standard reduction potential of those molecules was also calculated by DFT. Calculations were carried out using the Gaussian 16 (rev. A.03) quantum chemistry software [38] at the B3LYP‐D3BJ [39, 40, 41, 42] level of theory. The LANL2DZ(f) basis set and associated pseudopotentials [43, 44] were used for the Re center, while remaining atoms were described using the all‐electron cc‐pVDZ basis set [45, 46, 47]. Implicit solvent effects (acetonitrile) were included both in geometry optimizations and energy calculations using the IEF‐PCM model [48] as implemented in Gaussian 16. Electronic energies were corrected by performing single‐point calculations on optimized geometries with a more extended basis set: LANL2TZ(f) [43, 44, 49] for Re and aug‐cc‐pVTZ [45, 46, 47, 50] for the other elements. Gibbs free energies were further corrected from the 1 atm reference state used in the Gaussian code to the 1.0 M solution standard state at 25°C using a +1.89 kcal mol⁻¹ correction [51]. Quasiharmonic entropy and enthalpy corrections (Grimme's [52] and Head‐Gordon's [53] methods, respectively) were applied as implemented in the GoodVibes code [54]. Final Gibbs free energy values are reported in kcal mol⁻¹, and standard redox potentials are reported in volts (V) versus Fc⁺/Fc. A dataset collection of the optimized structures is available in the ioChem‐BD repository [55] and can be accessed via https://doi.org/10.19061/iochem‐bd‐1‐404.

Supporting Information

Additional supporting information can be found online in the Supporting Information section. Supporting Figure S1: Cyclic voltammograms recorded for a solution of 0.5 M [TBA][PF₆] in CH₃CN and 1.5 M TFE as a proton source under 10% (v/v) O₂ in an N₂ matrix (green plot) and under the same conditions, but adding 1 mM Re complex (red plot). CV containing 1 mM of Re complex and 0.5 M [TBA][PF₆] in CH₃CN solution under Argon plotted for reference (black plot). Scan rate 0.1 V s⁻¹. Supporting Figure S2: Cyclic voltammogram on a 3 mm diameter bare GC electrode disc of 1 mM of Re complex in a solution of 0.5 M [TBA][PF₆] in CH₃CN containing 1.5 M TFE as a proton source under different CO₂/O₂/N₂ (% vol.) atmospheres: 10/10/80 (green plot), 30/10/60 (blue plot) and 50/10/40 (orange plot). Scan rate 0.1 V s⁻¹. Supporting Figure S3: Cyclic voltammograms recorded on GCE in a solution of 1 mM of Re complex, 0.5 M [TBA][PF₆] and 1.5 M TFE in CH₃CN under different concentrations of gas impurities in N₂ matrix of (a) NO₂ and (b) SO₂. Scan rate 0.1 V s⁻¹. Supporting Figure S4: Cyclic voltammograms recorded on GCE for a solution of 0.5 M [TBA][PF₆] and 1.5 M TFE in CH₃CN under different concentrations of gas impurities in N₂ matrix of (a) NO₂ and (b) SO₂. Inset in (b) shows a magnified view of reduction region. Scan rate V s⁻¹. Supporting Figure S5: Schematic representation of the interaction between the active form of the Re complex and selected flue gas components. Computed binding energies (ΔG, kcal mol⁻¹) indicated next to each structure. Supporting Figure S6: Current density evolution during constant potential electrolysis (CPE) of 1 mM Re complex in a solution of 0.5 M [TBA][PF₆] in CH₃CN containing 1.5 M TFE on bare GCE under different gas atmospheres: 100% (v/v) CO₂ (red plot) at −1.95 V vs Fc⁺/Fc, 10% CO₂ at −1.95 V vs Fc⁺/Fc (blue plot) and at −2.05 V (brown plot), and 1% CO₂ (purple plot) at −1.8 V vs Fc⁺/Fc. Stirring rate for all CPE = 300 rpm. Supporting Figure S7: Current density evolution during constant potential electrolysis (CPE) at −2.05 V vs Fc⁺/Fc of 1 mM Re complex and 0.5 M [TBA][PF₆] in CH₃CN solution containing 1.5 M TFE on bare GCE under different gas atmospheres. (a) NO2 (0, 100, 500 and 1000 ppm) and (b) SO₂ impurities (0, 50, 500 and 1000 ppm), both in 10% v/v CO₂ in a N₂ matrix. In both cases, the same color code is used to indicate increasing impurity concentration: blue (0 ppm), brown (100 ppm NO₂ or 50 ppm SO₂), gray (500 ppm) and orange (1000 ppm). Stirring rate 300 rpm. Supporting Figure S8: Representative IC chromatograms recorded for blank and post‐electrolysis electrolyte samples. (a) Blank solution: anhydrous CH₃CN bubbled with 5% v/v NO₂ in N₂ matrix. (b) Anolyte samples taken after CPE (1 mM Re complex, 0.5 M [TBA][PF₆] in CH₃CN containing 1.5 M TFE; total circulated charge = 14 C) under a 10% v/v CO₂ atmosphere containing NO₂ at 100 ppm (black plot) and 1000 ppm (red plot). (c) Catholyte samples for the same experiments, with same color‐coding as in (b). The corresponding NO₂ ⁻ peak is labeled in panels (b) and (c) next to its retention time (10.34 min). Supporting Figure S9: Current density evolution during constant potential electrolysis (CPE) at −2.05 V vs Fc⁺/Fc of 1 mM Re complex in a solution of 0.5 M [TBA][PF₆] in CH₃CN containing 1.5 M TFE on bare GCE under different gas streams: 10% v/v CO₂ in a N₂ matrix (blue plot), 10% v/v CO₂ + 10% v/v O₂ in a N₂ matrix (green plot) and with the addition of 50 ppm SO₂ and 100 ppm NO₂ (flue gas composition, black plot). Stirring rate 300 rpm. Supporting Figure S10: ¹H NMR spectrum of [Re(CO)(bpy)Cl] (300 MHz, CD₃CN) δ/ppm, 1. 7.66 (dt, J = 6.5 Hz, 2H), 8.23 (dt, J = 8 Hz, 2H), 8.45 (d, J = 8.5 Hz, 2H), 9.05 (d, J = 5.7 Hz, 2H). Supporting Figure S11: Schematic illustration of the mass flow controller setup employed to create the different gas mixtures used to purge the H‐type cell, where constant potential electrolysis (CPE) were carried out. Supporting Figure S12: Calibration curves for (a) H₂ and (b) CO quantification using gas chromatography. The linear fit for both products follows the equation y = a + b*x, intercepting at 0 (a = 0), with b = 1.32 for H₂ and 277.97 for CO and R² = 0.999 for both cases. Supporting Table S1: Solubility of CO₂ and O₂ in different solvents at 298 K.

Conflict of Interest

The authors declare no conflicts of interest.

Supporting information

Supplementary Material