On factors limiting the performance of photoelectrochemical CO₂ reduction

Artificial photosynthetic systems are considered very promising to reform the current energy supply system.¹ A fully operational artificial photosynthetic system enables the scalable utilization of solar energy, as well as provides sustainable chemical fuels for our societies. Ever since chemical fuels were first generated from sun light,² this topic has attracted the attention of thousands of researchers worldwide.^3–7 The chemical fuels can be hydrogen, hydrocarbon, or multi-carbon oxygenates, and a variety of designs have already been created as viable artificial photosynthetic systems.^8–11 Among them, solar hydrogen generation does provide an entirely new strategy for an energy consumption system with zero-carbon emissions. As the current carbon-based energy supply system cannot be changed immediately, due to the demands for economic growth and development, the solar-driven CO₂ reduction reaction (CO2RR) is considered a promising artificial photosynthesis system that holds great potential for promoting solar energy utilization¹² and closing the anthropogenic carbon cycle.^13,14

This can be achieved by coupling a photovoltaic (PV) cell to an electrochemical cell (EC) or in a photoelectrochemical (PEC) CO2RR system.¹⁵ For a PV-EC-CO2RR, in addition to PV, researchers focus more on the chemical reactions within EC-CO2RR and the circuit integration.^16,17 In general, the PV-EC-CO2RR can easily reach a high solar-to-fuel efficiency by using the well-developed PV cell, but requires additional physical hardware and relies on the economic performance of PV cells.¹⁸ The perspective of PV-EC-CO2RR is beyond the scope of this present study. Compared to the PV-EC-CO2RR, the PEC-CO2RR is a more integrated system for the solar-driven CO2RR.¹⁹ However, current studies on the PEC-CO2RR are still unpractical because of the high energy loss during CO₂ fixation and the low purity of the final products.

In a typical PEC-CO2RR system,²⁰ (photo)electrodes, electrolyte, membrane, product separation, reactant feeding, and external circuits are needed, as shown in Fig. 1(a). The CO2RR happens at the surface of the (photo)cathode, while the oxygen evolution reaction (OER) happens at the surface of the (photo)anode. The ion exchange membrane is used to prevent the short-circuiting transfer of the products to the opposite electrode and deliver charges between cathode and anode chambers. As the concentration of the dissolved CO₂ in the aqueous solution is usually very low, CO₂ feeding is required for a continuous supply of reactants. The photoelectrode (mostly the photocathode) plays a key role because it dominates charge generation, separation, and injection. Therefore, most of the researchers in this area are working on designing new photoelectrodes by using new catalysts and co-catalysts.²¹  Figure 1(b) presents the mechanism of energy conversion for a photocathode-based PEC-CO2RR. Unlike the PV-EC-CO2RR, there is a solid–liquid junction in PEC-CO2RR. The photovoltage is generated through the illumination of the electrode–electrolyte interface. The photogenerated electrons participate in the CO2RR at the photocathode surface, and the OER happens at the anode surface. At the surface of the (photo)cathode, the electric potential of the photogenerated electrons should be more negative than the standard reaction potentials (E⁰) of the products. Table I shows the main reactions and E⁰ for identified products. Among them, liquid products with high energy density are preferred compared to gaseous products due to easier storage and transportation.

Studies of the PEC-CO2RR date back to the late 1970s. In the early stage, the studies mostly focused on p-type III–V semiconductors-based photocathodes owing to their superior performance on photoelectric conversion. In 1978, Halmann proposed a p-GaP photocathode, which successfully produced formic acid, formaldehyde, and methanol with part (or all) of the required energy supplied by light.²⁴ In 1984, formate dehydrogenase enzyme, connected to a p-InP photocathode, was used to broaden light absorption and improve the selectivity to formic acid.²⁵ In 1988, Yoneyama and co-workers studied the effects of electrolytes on the PEC-CO2RR with p-type CdTe and p-type InP electrodes and found that the ions within the electrolyte can change the selectivity of the products.²⁶ In 1993, Shoichiro et al. found that surface treatment with etchants [such as aqua regia, concentrated HCl, concentrated HNO₃, and alkaline potassium hexacyanoferrate (III)] can also change the selectivity of the products.²⁷ In 1998, the PEC-CO2RR under high-pressure (40 atm) was studied using p-type semiconductor electrodes and achieved a very high selectivity of over 90% for CO production due to the high coverage of the adsorbed $(CO2)2⋅−$ radical anion complex.^28,29 Nakato et al. used small metal (Cu, Ag, or Au) particles in combination with a p-Si photocathode to adjust the selectivity of the products for the PEC-CO2RR.^30,31

Entering the new century, the studies on the PEC-CO2RR are more and more diversified. Besides p-type III–V semiconductors, p-type metal oxides were also used as photocathodes. For instance, Gu et al. fabricated a Mg-doped CuFeO₂ photocathode for the PEC-CO2RR that produced formate with 400 mV underpotential.³² Kamimura et al. fabricated a p-type Cu₃Nb₂O₈ photocathode and detected CO among the products.³³ Jang et al. decorated Au on a ZnTe/ZnO photocathode and acquired a very high current density of −16 mA/cm² under 1 sun illumination.³⁴ Sagara et al. fabricated a B-doped g-C₃N₄ photocathode, which yielded ethanol as the main product.³⁵ The Cu₂O photocathode protected by TiO₂ and Al:ZnO layers was also used as a photocathode for producing CO at large current densities and high selectivity.³⁶ Sahara et al. tested a molecular and semiconductor photocatalyst hybrid photoelectrochemical cell, consisting of a NiO–RuRe photocathode and a CoO_x/TaON photoanode.³⁷ CO was produced in this hybrid PEC cell with the assistance of a 0.3 V external electrical bias.³⁷ However, the activities of these new types of photocathodes remain lower than p-type III–V semiconductors-based photocathodes.^38–40 Mi et al. fabricated a series of efficient GaN/n⁺-p Si based photocathodes for the PEC-CO2RR.^41–44 Indeed, Sn/GaN/n⁺-p Si favors the reaction path to formic acid with a high selectivity of 76.9%.⁴⁴ Both of the products of Cu-ZnO/GaN/n⁺-p Si and Pt-TiO₂/GaN/n⁺-p Si are syngas (CO and H₂).^41–43 They reported a solar-to-syngas efficiency of 0.87%, as well as a significantly reasonable high stability of 10 h.⁴¹ 

Overall, in the published works, the study on p-type III–V semiconductors-based photocathodes is the current state-of-the-art for the PEC-CO2RR. Although many researchers attempted to broaden the material scope of the photocathodes for the PEC-CO2RR, limited light absorption, short lifetime of the photo-generated carriers, serious photo corrosion of the materials, and poor stability of the system have not been effectively resolved yet. The studies on this still have a long way to go, and much more work is needed. In our opinion, finding novel materials and architectures for the photocathodes should be the next frontiers for PEC-CO2RR research. Furthermore, some important indicators such as products, selectivity, energy efficiency, and onset potential are often used for comparing the various published results. However, in practice, the effect of reaction conditions should not be ignored. Establishing rigorous testing standards is necessary for the PEC-CO2RR. Beyond material design, complex intermediates and reaction pathways, obstructed mass transfer, and large photovoltage requirements still limit the development of PEC-CO2RR systems. We think that the PEC-CO2RR poses challenges to both science and engineering.

Understanding the complex reaction pathways of CO2RR is very difficult because of the large amount of intermediates.^45,46 Similar to the EC-CO2RR, most of the hydrocarbons and oxygenates from C1 to C3 can be produced in PEC-CO2RR.⁴⁷  Figure 2 shows the reaction roadmaps of CO2RR. In general, carbon monoxide is commonly considered as the main intermediate for producing other hydrocarbons and oxygenates.^48,49 In Fig. 2(a), the CO pathway was used as a representative example for producing C1 molecules. Figure 2(b) shows a possible reaction roadmap for producing various C2 molecules. The pathway can be various for different catalysts, and one can construct other similar reaction routes using the same concept.

However, for an industrial application, the target product should be easily separated from the reactor, which would take up a large part of the cost of operation. Therefore, high selectivity on a specific product is highly necessary. There are many factors that influence the selectivity of products and CO2RR pathways, but the main ones are the surface materials on photoelectrodes,⁵⁰ the morphology of the surface catalysts,^51–53 and the composition of the electrolytes.^54–57 Metals,^34,58 semiconductors,^59,60 and molecular electrocatalysts^61,62 are commonly used as co-catalysts to build the semiconductor/catalyst interfaces for the PEC-CO2RR. Among them, Cu based materials appeared to be excellent for producing liquid products such as alcohols and aldehydes.^63,64 Schouten et al. proposed the reaction mechanism of the CO2RR on Cu and pointed out that CO and HCOOH are primary intermediates in the first step, and most of the C1, C2, and C3 chemical compounds would be subsequently generated after CO.⁶⁵ By the combination of a ruthenium-based catalyst and a semiconductor p–n junction that integrates GaN nanowire arrays on silicon, Shan et al. reported a stable reduction of CO₂ to formate with Faradaic efficiencies of up to 64%.⁶² Morphology of the surface catalysts also plays an important role during the CO2RR. Liu et al. reported that the nanoneedle structure can produce a high local electric field on the Au tip, which in turn leads to a high local concentration of CO₂ close to Au surface.⁶⁶ Moreover, the mass transfer of CO2RR heavily depends on the composition of the electrolytes,^57,67 which will be discussed later in Sec. IV.

At present, numerical simulation is widely used for the PEC-CO2RR.⁶⁸ Among the developed methods, density functional theory (DFT) can be used to compute the electronic and crystal structure of (photo)catalyst, as well as the chemical processes.^69–72 Nørskov et al. devoted a considerable amount of efforts^73,74 in DFT calculations for the CO2RR. Figure 3(a) shows a free energy diagram that illustrates the lowest energy pathway of CO₂ to CH₄ (as an example of a possible product) on the Cu(211) catalyst.⁷³ Nie et al. studied the activation barriers of CO2RR on various Cu facets by using DFT calculations, indicating that the key selectivity-determining step on Cu(111) is the reduction of CO.⁷⁵ Bagger et al. calculated the binding energies of the intermediates ΔE_CO^* and ΔE_H^*, thereby displaying that Cu binds CO^* while not having H_upd, indicating that Cu is the only metal that produces products beyond CO.^71,76 However, simulating the CO2RR conditions by DFT remains a problem due to the complex reaction pathways and numerous rate parameters. It is necessary to further develop computational methodologies for this purpose.⁷⁷ 

Aside from the aforementioned numerical simulation, in situ techniques are also important to study the CO2RR mechanism by directly probing the products or intermediates at surface of the catalysts.^48,78–82 Raman and infrared spectroscopy are commonly used as in situ techniques [Figs. 3(b) and 3(c)].^82,83 By using in situ Raman spectroscopy, Yeo et al. demonstrated that CO₂ was reduced more preferentially on the metal rather than on oxide surfaces.⁸⁴ From the in situ Raman spectra for oxide-derived Cu electrocatalysts, Mandal et al. discovered that surface Cu oxide reduction happened before the CO2RR.⁸⁵ Wuttig et al. studied the CO adsorption profile on metallic Cu at various applied potentials, CO concentrations, and pH values by in situ surface-enhanced infrared absorption spectroscopy.⁸⁶ In addition, some synchrotron radiation-based in situ techniques are also developed for the CO2RR.⁷⁸ Scott et al. precisely measured the phase conversion potential of the near-surface region on polycrystalline Cu during CO reduction by in situ grazing incidence x-ray diffraction.⁸³ Eilert et al. used in situ x-ray absorption spectroscopy to investigate the formation mechanism of Cu electrocatalyst, which was used for the CO2RR.⁸⁷ 

In situ Raman, x-ray diffraction, and absorption spectroscopies mainly focus on the crystal structure, while in situ infrared spectroscopy mainly focuses on the intermediates or the interaction between the products and the catalyst surface. However, the current in situ techniques are limited by materials and environment. Most of the in situ techniques are only suitable for the CO2RR under dark conditions. Therefore, it is vital to develop more in situ techniques to fit the newly designed PEC-CO2RR systems.

Similar to the EC-CO2RR, the PEC-CO2RR conducted in different cells can be very different from each other due to the different local environments for mass transfer. In aqueous electrolytes, the conductivity and the CO₂ concentration are very low, thus resulting in mass transfer limitations for the CO2RR.⁸⁸ Water provides plenty of H⁺ at surface of the cathode for the CO2RR. However, CO₂ has to overcome the mass transfer resistance before reaching the solid–liquid interface. Moreover, in order to make room for the following reactions, the products of CO2RR would better depart from the solid–liquid interface quickly. Once the mass transfer of carbonaceous species is blocked, the reactions would generate more hydrogen.

Generally, alkali metal bicarbonates, MHCO₃ (M = Li, Na, K, Rb, and Cs) are well suited for the CO2RR, as they can form a bicarbonate buffer system in the electrolytes, especially at the surface.⁸⁹ Resasco et al. found that the electrostatic interactions between hydrated cations in the Helmholtz layer are strongly associated with the cation size, thus resulting in cation-specific catalytic activities for the CO2RR.^57,67 Other than the metal cations, the species in the CO₂ saturated electrolyte also include protons (H⁺), hydroxides (OH⁻), dissolved CO₂, bicarbonate anions (HCO₃⁻), and carbonate anions (CO₃²⁻), as shown in Fig. 4. An excess of bicarbonates does reduce the polarization losses but would simultaneously increase the pH of the electrolyte and reduce the solubility of CO₂, while fewer bicarbonates would reduce the pH of the electrolyte, thus reducing the selectivity to the CO2RR compared to the hydrogen evolution reaction (HER). In more cases, appropriate bicarbonates (with a neutral pH condition) work better for the CO2RR due to the better balance between CO₂ concentration and hydrogen evolution reactions.⁹⁰ 

The mass transfer of the species in the electrolyte can be further enhanced by many ways.⁹¹ Lobaccaro et al. found that a smaller bubble size is better for maintaining the concentration of CO₂ in the electrolyte during the CO2RR than a large one [Fig. 5(a)].⁹² The use of gas diffusion electrodes (GDEs) is another effective approach.⁹³ Even for high pH conditions, GDEs can enable a high local concentration of CO₂ near the surface of the cathode.⁹³  Figure 5(b) illustrates one example of the GDEs based CO2RR. So far, many studies on the EC-CO2RR and PV-EC-CO2RR have been reported using GDEs,⁹⁴ but more efforts should be devoted for the PEC-CO2RR.

The polarization loss (Δϕ) can be represented as a sum of Δϕ_solution, Δϕ_Nernstian, and Δϕ_Membrane.⁹⁰ Actually, Δϕ comes from the voltage losses due to the transport of species and concentration gradients. As Δϕ would be released as thermal energy, the practicable PEC-CO2RR system must get this number as low as possible.

There are two ways to reduce large photovoltage requirements. One way is trying to reduce the polarization loss by adjusting the component of electrolyte,⁹⁵ or using new types of electrolyte such as a solid electrolyte and non-aqueous solvents.^96–98 The local reaction conditions on the electrode surface can be changed by adjusting the concentration, species, and pH value of the supported electrolyte.⁹⁵ With solid electrolytes, the ions are shuttled between the (photo)electrodes by ion-conducting solid polymers or ceramics,⁹⁹ similar to solid-state batteries. By conjunction with GDEs, the solid electrolyte-based CO2RR enables the continuous production of high-purity liquid fuel, as illustrated in Fig. 7.⁹⁶ Ionic liquids, a kind of non-aqueous solvent, can enhance the CO₂ solubility and conductivity simultaneously, but they are too expensive now. In addition, there are also some non-aqueous solvents such as acetonitrile,¹⁰⁰ dimethylformamide,¹⁰¹ dimethyl sulfoxide,⁹⁷ and methanol.¹⁰² However, for all non-aqueous solvents, even a small amount of water in non-aqueous solvents can lead to big changes in the selectivity and current density of CO2RR.¹⁰³ Another way to reduce the large photovoltage requirements is managing the light absorption by utilizing multi-junction or engineering the band alignments of the semiconductors within the photoelectrodes to supply a higher total photovoltage.^104,105 If only a single solid–liquid junction is used, the light absorption would be limited by the band alignment. As multi-junction devices have been proved to be a success in solar cells,¹⁰⁴ one must consider this result before implementing the PEC-CO2RR.

The PEC-CO2RR is very sensitive to cell design due to the additional energy losses on mass transfer and external circuits. Ideally, a practicable PEC-CO2RR system must have minimized polarization losses, broad light absorption, and be wireless. First, as mentioned before, the polarization losses mainly depend on the electrolyte. The species in electrolytes have to transfer within the surfaces of anode and cathode. Therefore, the mass transfer distance should be as short as possible. Second, the light absorption would be improved by using multi-junction devices. Third, the external circuit will generate much energy loss owing to the extra electric resistance, which includes ohmic drops for the electrical contacts and cables. Moreover, the wireless system saves hardware and is more integrated than the systems with external circuits.

Based on the above considerations, we propose a conceptually ideal PEC-CO2RR system compared to H-cells, which is the most widely used design, as shown in Fig. 8. In this wireless system, oxygen is produced within the anode chamber, while the high purity “carbon fuel” is produced within the cathode chamber. The anode and cathode are stacked in pairs with an ohmic contact. The mass transfer distance is very close to the overall thickness of electrodes. The ion-exchange membrane is put between the electrodes to balance the ions and separate the products within two chambers. In practice, the ion exchange membrane can be an anion-exchange membrane (high OH⁻ conductivity) or a cation-exchange membrane (high H⁺ conductivity). The anion-exchange membrane is used for basic and neutral conditions, while the cation-exchange membrane is used for acidic conditions. The anode absorbs the short wavelengths of light and the cathode absorbs longer wavelengths, enabling a no-bias PEC-CO2RR.

The PEC-CO2RR enables direct access to store solar energy into chemical fuels, and it provides a promising strategy to reduce current carbon emissions. In addition to challenges experienced in other similar photoelectrochemical systems, such as photoelectrochemical water splitting, the PEC-CO2RR is also faced with complex reaction pathways, obstructed mass transfer, and large photovoltage requirements.

In this perspective, we take a critical look at the progress in PEC-CO2RR and present the current state-of-the-art of current studies. First, the hybrid III–V semiconductors, such as GaN/n⁺-p Si, have a pretty good photoelectric conversion PEC-CO2RR, but still need to improve light absorption and long-time stability. The research on fabricating practicable electrodes is still at an early stage. Second, the mechanism of CO2RR is still unclear due to numerous intermediates. Although numerical simulation and in situ techniques are very useful, only several simple reaction pathways can be studied clearly. Third, in current PEC-CO2RR systems, the energy loss is very high, especially polarization losses. To solve this problem, many strategies have been proposed, such as using alternative electrolytes and multi-junction photoelectrodes. However, a large total voltage requirement is still one of the key obstacles of PEC-CO2RR. Furthermore, a highly integrated PEC-CO2RR system is proposed as an ideal system in which there are no external circuits, with an extremely short mass transfer distance and broad light absorption.

Overall, at present, PEC-CO2RR systems are very attractive to scientists whose research focus on scalable, stable, and highly efficient solar fuel generation. The research on the PEC-CO2RR still presents many challenges to both science and engineering. In the past few decades, although much progress has been made, industrial applications are still far away. In order to meet this goal, it is essential to develop new strategies and materials for the PEC-CO2RR. Moreover, it would be interesting to witness further development along with some emerging technologies, such as GDEs and solid electrolyte.

This work was supported by the Basic Science Center Program for Ordered Energy Conversion of the National Natural Science Foundation of China (Grant No. 51888103), the National Natural Science Foundation of China (Grant No. 51906199), the Natural Science Basic Research Program of Shaanxi Province (Grant No. 2019JQ-117), and the Fundamental Research Funds for the Central Universities (Grant No. xzy012019016). Y.L. acknowledges fellowship support from the Initiative Postdocs Supporting Program.