Utilizing abundant solar energy to drive carbon dioxide (CO₂) reduction is a promising method to convert renewable energy to value-added chemicals [1, 2]. Integrating molecular elecrocatalysts with semiconductor photoabsorbers is an attractive approach to obtain efficient photoelectrochemical (PEC) systems for CO₂ conversion. Over the past three decades, molecular electrocatalysts based on precious metals have been commonly investigated because of their high selectivity and efficiency [3-11]. Only a few earth-abundant metal-based molecular electrocatalysts, such as iron, cobalt (Co), nickel, and manganese (Mn) complexes, have been investigated [12-19]. Semiconductor materials based on metal oxides, chalcogenides, and silicon (Si) are able to efficiently separate photogenerated charges, thus enhancing photoelectric conversion efficiency and reaction kinetics [3, 7, 11, 16, 17, 20]. Incorporation of noble metal electrocatalysts into semiconductor PEC systems has been mostly investigated for homogeneous and heterogeneous CO₂ catalytic reduction systems [3, 5, 7, 11]. In contrast, PEC systems composed of non-precious metal complexes and semiconductor photoabsorbers have rarely been reported. Very recently, Reisner et al. [21] constructed mesoporous titanium dioxide functionalized with Mn complex for highly efficient CO₂ reduction in acetonitrile (MeCN); using this system, a turnover number of 112 was obtained at an overpotential of 420 mV. Fabre et al. [16] reported that a Si nanowire photocathode could effectively enhance the energy conversion efficiency of Mn carbonyl bipyridyl electrocatalysts for selective CO₂ reduction. Wang and co-workers [17] constructed a conjugated Co complex and Si photoelectrocatalytic system and investigated the effects of water and material morphology on its performance.

Herein, we demonstrate a hybrid PEC system combining a Si photoelectrode and Co complex electrocatalysts [22] for CO₂ reduction to carbon monoxide (CO). Adding a small amount of a proton source can greatly improve catalytic selectivity, which may shed light on enhancing the performance of CO₂ catalytic reduction systems.

To examine the electrochemical properties of the Co complex electrocatalysts (Scheme 1), cyclic voltammetry (CV) was carried out in dry MeCN solution using a glassy carbon working electrode. For complex 1, two reduction waves were observed at peak potentials of –0.58 and –1.59 V vs. NHE under argon (Ar), as shown in Fig. 1. Wave Ⅰ at peak potential of –0.58 V is assigned to reduction of Co^Ⅱ to Co^Ⅰ, and wave Ⅱ at peak potential of –1.59 V is assigned to ligand-based reduction. When the electrolyte was saturated with CO₂, wave Ⅰ remained the same, but the current of wave Ⅱ increased by ca. three-fold, suggesting the onset of electrocatalytic activity of 1 toward CO₂. Similarly, complex 2 displayed two reduction waves at potential of –0.67 and –1.49 V under Ar, and the current of the latter wave was enhanced ca. two-fold under CO₂.

The effect of addition of water as a proton source on the electrocatalytic activity of the catalysts was investigated (Fig. 2). When 0.8%–1.6% water was added under Ar, wave Ⅰ and Ⅱ of 1 were essentially unchanged. Conversely, under CO₂, the catalytic current of wave Ⅱ increased by 50% compared with that in dry MeCN, thus indicating that adding water is advantageous for electrocatalysis. For catalyst 2 under CO₂, the catalytic current of wave Ⅱ also increased with water addition, although enhancement was not as large as that of catalyst 1, indicating that catalyst 1 is a more efficient electrocatalyst than 2.

The electrochemical properties of the Co complexes were further investigated by varying the scan rate of CVs. CVs of 1 at various scan rates were plotted with the current normalized by dividing the original current by the square root of scan rate. Under Ar, the normalized currents of both wave Ⅰ and Ⅱ overlapped at various scan rates, suggesting that the currents of wave Ⅰ and Ⅱ are diffusional with no catalysis (Fig. 3(a)). Under CO₂, normalized currents of wave Ⅱ at lower scan rates are higher than those at higher scan rates, confirming that the process under wave Ⅱ is electrocatalytically active toward CO₂ (Fig. 3(b)).

When 0.8% water was added to 1 under Ar, the normalized currents still overlapped with each other at different rates, suggesting that 1 did not actively catalyze hydrogen (H₂) evolution (Fig. 3(c), (e)). Under CO₂, adding water increased the catalytic current, suggesting that the catalysis is accelerated in the presence of water (Fig. 3(d), (f)). Catalyst 2 demonstrated similar behavior under both dry and wet conditions (Fig. S2). The above results suggest that adding a small quantity of water is advantageous to enhance catalytic current. Although the mechanism of this effect of water is unclear, it is possible that water may protonate the transient metal carboxylate species and thus promote CO formation.

The product selectivity of catalyst 1 (Fig. 4(a)) and 2 (Fig. 4(b)) was investigated under dry and wet conditions by controlled potential electrolysis measurements. After electrolysis at –1.5 V (vs. NHE), the Faradaic efficiency of CO increased from 79% under dry conditions to 90% with 2% added water for catalyst 1. In contrast, adding water to catalyst 2 increased H₂ evolution and lowered the CO production efficiency from 95% under dry conditions to 67% with 2% added water.

Semiconductor Si is capable of improving photoelectric conversion efficiency [16]. Therefore, PEC reduction of CO₂ was performed in MeCN solutions of 1 and 2 using a silicon planar photocathode (SiPL). Comparing the CVs under dark and light conditions revealed that a photovoltage of about 400 mV was generated and the photocatalytic current of wave Ⅰ for 1 and 2 was enhanced under saturated CO₂ (Fig. 5).

Vertically aligned Si nanowire arrays (Fig. S1) were also used to evaluate the photoelectrocatalytic performance of the catalysts because of their excellent charge separation and photoelectric conversion effect [16]. CVs of 1 mmol/L of catalyst 1 in MeCN + 0.1 mol/L Bu₄NPF₆ at the Si nanowire electrode under CO₂ indicated that a photovoltage of about 600 mV was generated (Fig. 6). Although the photovoltage of the Si nanowire electrode is larger than that of SiPL, which is preferable to lower the overpotential, non-specific H₂ evolution occurred markedly at the nanowire photoelectrode even in formally dry MeCN solution; thus, the Si nanowire electrode was not used for further study.

As was reported, protic additive could overcome the mass transfer limitation of interface between electrode and electrolyte [3]. To test this protic additive effect, 2%-10% methanol was added to catalyst solution. As was illustrated, after 10% methanol was added in catalyst solution, the photocurrent was obviously enhanced and the onset potential positively shifted for ca. 200mV, indicating catalytic enhancement effect of protic additive (Fig. 7).

Chronoamperometry and chopped light analysis were performed to evaluate the photoelectrode stability in this system. A solution of 1 mmol/L of catalyst 1 in MeCN + 0.1 mol/L Bu₄NPF₆ at SiPL under CO₂ was tested. During 14 h of electrolysis at –1.4 V, the current decayed by less than 30% (Fig. 8). Chopped light experiments indicated that the current increased linearly with the applied potential, and the photocurrent varied instantaneously with illumination, further confirming the outstanding photoelectric response and stability of the electrode (Fig. 9).

Methanol was also added to the PEC system as an alternative proton source to water. When 10% methanol was added to an MeCN solution of 1 during photoelectrolysis at –1.6 V, the steady-state current increased by ca. 2.3 fold relative to that in pure MeCN (Fig. 10(a)). Similar current enhancement was observed for 2 (Fig. 10(b)). The selectivity for CO increased with the addition of methanol (Fig. 11). After CPE at –1.5 V (vs. NHE) with added methanol under white light, the Faradaic efficiency of CO production by 1 rose from 49% to 83% (Fig. 11(a) and (b)) and that by 2 increased from 60% to 94% (Fig. 11(c) and (d), while the Faradaic efficiencies of H₂ and methane (CH₄) production decreased. When no catalyst was added with 10% methanol under CO₂, the Faradaic efficiency of CO formation after bulk electrolysis was 14%; the rest being H₂ and a trace amount of CH₄. Also, after adding 2% water, because of the competitive H₂ evolution reaction, Faradaic efficiencies for CO production lowered to 42% (Fig. S3). These results suggest that methanol is a superior proton source over water for PEC reduction of CO₂ to maximize CO selectivity.

In summary, two homogeneous photoelectrocatalytic systems composed of cobalt terpyridine and bipyridine complexes and a Si photoelectrode were investigated for efficient and selective CO₂ reduction. The system consisted of readily prepared complexes containing earth-abundant elements. Si as a photoabsorber was critical to lower the overpotential and increase catalytic current. Faradaic efficiencies for CO production reached 83% for the cobalt terpyridine complex and 94% for the cobalt bipyridine complex in MeCN solution with added methanol; adding methanol increased CO selectivity. A photovoltage of approximately 400 mV was generated and the photocurrent was stable during bulk electrolysis. This system is cost-effective and environmentally friendly, revealing a design to realize more efficient catalysts for CO₂ reduction.