Advances in Clean Fuel Ethanol Production from Electro-, Photo- and Photoelectro-Catalytic CO2 Reduction

Using renewable energy to convert CO2 to a clean fuel ethanol can not only reduce carbon emission through the utilization of CO2 as feedstock, but also store renewable energy as the widely used chemical and high-energy-density fuel, being considered as a perfect strategy to address current environment and energy issues. Developing efficient electrocatalysts, photocatalysts, and photoelectrocatalysts for CO2 reduction is the most crucial keystone for achieving this goal. Considerable progresses in CO2-based ethanol production have been made over the past decades. This review provides the general principles and summarizes the latest advancements in electrocatalytic, photocatalytic and photoelectrocatalytic CO2 conversion to ethanol. Furthermore, the main challenges and proposed future prospects are illustrated for further developments in clean fuel ethanol production.


Introduction
With the fast development of the economy and society, the ever-increasing demand for energy all over the world while the limited fossil fuel resources lead to an aggravated energy crisis [1,2]. The huge consumption of fossil fuels causes the constantly accumulating of CO 2 in the atmosphere. By May 2020, the concentration of atmospheric CO 2 reached another record of 412.69 parts per million (ppm) [3], far exceeding the upper safety limit of 350 ppm, which may cause disastrous environmental consequences such as global warming, polar glacier melting, rising sea level, etc. [4]. On the other hand, the renewable energy sources from wind, sun, etc., have been rapidly developed in recent years. Unfortunately, the power from these renewable energy sources cannot be integrated into the electric grid well due to the intrinsic inferiorities of instability and anti-peak-load regulating, resulting in the huge waste and development limitation [5].
An ideal strategy to solve the energy and environmental problems is to convert CO 2 into fuels and value-added chemicals using renewable electricity and/or solar energy. Such a strategy can not only reduce the concentration of atmospheric CO 2 through the utilization of CO 2 as feedstock, but also store renewable energy as fuels and useful chemicals, thus relieving our dependency on fossil fuels [6][7][8]. Powered by renewable electricity and/or solar energy, CO 2 can be reduced to clean fuels, such as carbon monoxide (CO), methane, formic acid, methanol, ethanol, etc. [9,10]. By contrast, ethanol, a kind of clean and renewable liquid fuel with a higher heating value of −1366.8 kJ·mol −1 , is a preferred product. With a higher energy density, easier to store and transport than that of gas products, ethanol has also According to the variety of renewable solar energy assistance, CO 2 -to-ethanol conversion can be divided into three major categories: electrocatalytic reduction by an electrolyzer powered by commercial photovoltaic (PV) devices, photocatalytic reduction by an efficient photocatalyst, photoelectrocatalytic reduction by a semiconducting photocathode and an electrolyzer [12]. Over the past decades, numerous efforts have been devoted to researching the three kinds of CO 2 reduction techniques for the production of clean fuel ethanol [13][14][15][16]. Different from C 1 products (CO, CH 4 , formate, methanol, etc.), the multiple electron-proton transfers involved with ethanol production from CO 2 have been reported with low efficiency due to the kinetic barriers. Typically, multiple electron-proton transfer steps must be orchestrated with their own associated activation energies, thus presenting kinetic barriers to the forward reaction [12]. Therefore, efficient and robust electrocatalysts, photocatalysts and photoelectrocatalysts are required to promote this kinetically sluggish reduction process.
This review will focus on the most-studied catalysts and their corresponding catalytic systems for the reduction of CO 2 to ethanol in the categories of electrochemical, photochemical and photoelectrochemical approaches. Since the catalytic activity and selectivity are mainly determined by the structures and surface states of catalysts as well as the reaction conditions, the second section of this review provides the general principles of electroreduction, photoreduction and photoelectroreduction of CO 2 , as well as the theoretical foundation for ethanol production. Lastly, a short prospect is given of the challenges and new directions in the development of efficient CO 2 reduction to solar fuel ethanol.

CO 2 Electroreduction
Electroreduction of CO 2 is commonly carried out in a gas-tight, two-compartment electrolysis cell equipped with a working electrode, a counter electrode and a proton exchange membrane as the separator (Scheme 2A). The membrane was employed to restrict the transport of liquid phase products from the working electrode to the counter electrode where they can be oxidized [17]. Prior to the experiments, the applied gas-tight electrolysis cell should be vacuumed and then purged with CO 2 for 30 min to reach a constant pH value of the electrolyte. The reduction reaction of CO 2 is conducted and measured by cyclic voltammetry and potentiostatic electrolysis at fixed potentials. The gaseous and liquid products are generally quantified by a gas chromatograph and a nuclear magnetic resonance (NMR) spectrometer or a liquid chromatography, respectively. The faradaic efficiency (FE), which is defined as the percentage of electrons consumed for the formation of a given product, can be calculated as follows [18] where α is the quantity of transferred electrons for CO 2 reduction to a given product; n is the number of moles for a desired product; F is the Faraday's constant (96485 C·mol −1 ) and Q (A·s) is the total quantity of charge passed.
Catalysts 2020, 10, x FOR PEER REVIEW 3 of 25 Electroreduction of CO2 is commonly carried out in a gas-tight, two-compartment electrolysis cell equipped with a working electrode, a counter electrode and a proton exchange membrane as the separator (Scheme 2A). The membrane was employed to restrict the transport of liquid phase products from the working electrode to the counter electrode where they can be oxidized [17]. Prior to the experiments, the applied gas-tight electrolysis cell should be vacuumed and then purged with CO2 for 30 min to reach a constant pH value of the electrolyte. The reduction reaction of CO2 is conducted and measured by cyclic voltammetry and potentiostatic electrolysis at fixed potentials. The gaseous and liquid products are generally quantified by a gas chromatograph and a nuclear magnetic resonance (NMR) spectrometer or a liquid chromatography, respectively. The faradaic efficiency (FE), which is defined as the percentage of electrons consumed for the formation of a given product, can be calculated as follows [18] = where α is the quantity of transferred electrons for CO2 reduction to a given product; n is the number of moles for a desired product; F is the Faraday's constant (96485 C·mol −1 ) and Q (A·s) is the total quantity of charge passed. Scheme 2. Illustrations of (A) H-cell configuration with a catalyst deposited on a solid substrate, (B) flow cell configuration with a catalyst deposited on gas-diffusion electrode (GDE) and a flowing catholyte channel, (C) GDE architecture and its 3-phase interface mechanism, (D) the working electrode architecture in H-cell and its 2-phase interface mechanism for electrocatalytic CO2 reduction.. Electroreduction of CO2 is a multi-step reaction process involving multiple electron transfer, and generally takes place at the electrode/electrolyte interface for the heterogeneous electrocatalysts [19]. It experiences such a process involving three major steps of chemical adsorption of CO2 on the surface of electrocatalysts, activation of CO2 to cleave C=O bonds and form C-O or C-H bonds through electron and/or proton transfer, and desorption of products from electrocatalysts surface after configuration rearrangement [18]. The applied electrocatalysts and electrolysis potentials significantly affect the final reduction products that may vary in the carbon compounds of CO, methane, formic acid, ethylene, methanol, ethanol, etc., or a mixture of them. This kind of dependence on electrocatalysts and electrolysis potentials is ascribed to the different thermodynamic equilibrium potentials of these products from CO2 reduction, as displayed in Table  1 [20]. From the view of thermodynamics, the equilibrium potentials around −0.2 to −0.6 V (versus normal hydrogen electrode (NHE), pH = 7.0) of CO2 reduction are comparable to that of hydrogen evolution reaction (HER) (−0.41 V, Table 1) [4]. That is why H2 is the major side-product during CO2 electroreduction in aqueous electrolytes. Additionally, the very small differences between the thermodynamic potentials for CO2 reduction products bring about a challenge to selectively produce the desirable carbon compounds. Actually, the required potentials to drive CO2 reductions are more Scheme 2. Illustrations of (A) H-cell configuration with a catalyst deposited on a solid substrate, (B) flow cell configuration with a catalyst deposited on gas-diffusion electrode (GDE) and a flowing catholyte channel, (C) GDE architecture and its 3-phase interface mechanism, (D) the working electrode architecture in H-cell and its 2-phase interface mechanism for electrocatalytic CO 2 reduction.. Electroreduction of CO 2 is a multi-step reaction process involving multiple electron transfer, and generally takes place at the electrode/electrolyte interface for the heterogeneous electrocatalysts [19]. It experiences such a process involving three major steps of chemical adsorption of CO 2 on the surface of electrocatalysts, activation of CO 2 to cleave C=O bonds and form C-O or C-H bonds through electron and/or proton transfer, and desorption of products from electrocatalysts surface after configuration rearrangement [18]. The applied electrocatalysts and electrolysis potentials significantly affect the final reduction products that may vary in the carbon compounds of CO, methane, formic acid, ethylene, methanol, ethanol, etc., or a mixture of them. This kind of dependence on electrocatalysts and electrolysis potentials is ascribed to the different thermodynamic equilibrium potentials of these products from CO 2 reduction, as displayed in Table 1 [20]. From the view of thermodynamics, the equilibrium potentials around −0.2 to −0.6 V (versus normal hydrogen electrode (NHE), pH = 7.0) of CO 2 reduction are comparable to that of hydrogen evolution reaction (HER) (−0.41 V, Table 1) [4]. That is why H 2 is the major side-product during CO 2 electroreduction in aqueous Catalysts 2020, 10, 1287 4 of 25 electrolytes. Additionally, the very small differences between the thermodynamic potentials for CO 2 reduction products bring about a challenge to selectively produce the desirable carbon compounds. Actually, the required potentials to drive CO 2 reductions are more negative than the equilibrium ones, leading to the overpotentials [21]. The presence of overpotential originates from the difficult rearrangement of the linear CO 2 molecule to a bent radical anion, which requires enormous energy to occur [22]. Therefore, reducing or even eliminating the overpotentials will facilitate the electroreduction of CO 2 . Table 1. Half-reactions and the corresponding thermodynamic redox potentials of CO 2 reduction to the main products in aqueous solutions [20].

Half-Reactions
E 0 (V vs. NHE) Recently, a kind of flow cell using the gas-diffusion electrode (GDE) and flowing catholyte has been developed to accelerate the technology of CO 2 electroreduction toward its envisioned application of neutralizing CO 2 emission on a global scale [23,24]. In this system, the electrolysis cell should consist of three compartments: cathode chamber, anode chamber and gas chamber (Scheme 2B). An ion exchange membrane between cathode chamber and anode chamber is necessary to separate the catholyte and the anolyte. The catalyst is incorporated into electrolyzers on a GDE commonly by drop-casting, spray-coating and in situ growing (self-supporting GDE) [25]. The coated catalyst side of GDE directly contacts the catholyte. Gaseous CO 2 with a certain flow rate passes through the gas chamber at the back side of GDE, and reacts at the catalyst/electrolyte interface, constituting a three-phase boundary (Scheme 2C). This configuration could overcome the mass transfer limitations of two-phase interface and the low solubility of CO 2 in aqueous solution (Scheme 2D). Thus, it enables rapid delivery of CO 2 to the catalyst, resulting in higher or even commercially relevant current densities. Furthermore, this GDE configuration allows the use of more alkaline electrolytes such as KOH, which results in an increased local pH, favoring HER suppression and C 2 (ethanol) production.

CO 2 Photoreduction
A CO 2 photoreduction system commonly consists of a light source, a photocatalyst, a sacrificial electron donor and a photoreactor that is used as the light absorber [12]. Generally, solar energy can be applied as the energy source for CO 2 conversion [9]. However, CO 2 is optically inert at visible and UV radiation in the wavelengths of 200-900 nm [26]. Thus, a photocatalyst with a suitable band structure is required to promote the CO 2 photoreduction so that the electrons can be excited by sunlight and transfer to CO 2 species adsorbed on the surface of catalysts. The band structure constitutes a conduction band (CB), a valence band (VB) and a bandgap between them with no electron configuration [27]. Accordingly, the CB bottom of the photocatalysts must be more negative than the reduction potentials of CO 2 and its reduced products, while the top of the VB must be more positive than the oxidation potential of sacrificial reagents [4,28]. Thus, the electrons would transfer from the CB of photocatalysts to surface-adsorbed CO 2 species and convert it to solar fuels. As a sacrificial reagent, H 2 O is an ideal electron donor and a hydrogen source for the CO 2 photoreduction [29]. However, it can compete with CO 2 photoreduction for the electrons to generate H 2 , as shown in HER in Table 1. Therefore, over the semiconductor photocatalysts, H 2 O can be not only oxidized to O 2 by trapping the photogenerated holes in the VB, but also reduced to H 2 by catching the photogenerated Catalysts 2020, 10, 1287 5 of 25 electrons in the CB. Commonly, photocatalytic CO 2 reduction is performed in fluidized bed reactor or optical fiber reactor [27]. In fluidized bed reactor [30], the photocatalysts are well dispersed in aqueous solution, thus promoting the contact and reaction between the photocatalysts and water soluble CO 2 , but it suffers from the hard separation of products and low light utilization efficiency. By contrast, the photocatalysts coating on the optical fibers in optical fiber reactor [31], considerably improve the illuminated surface area of photocatalysts and light utilization efficiency. Thus, it may be a promising technique to enhance photocatalytic CO 2 reduction efficiency.
The CO 2 photoreduction process generally undergo four major steps: (1) CO 2 molecules are chemically adsorbed on the surface of photocatalysts; (2) under light illumination, the electrons of semiconductor photocatalysts can be excited by photons from VB to CB, leaving an equal number of holes in the VB; (3) the photogenerated electrons are separated from holes and migrate to the photocatalyst surface; (4) the electrons are used to activate and reduce CO 2 into solar fuels, while the holes are consumed by the oxidation of H 2 O [32]. According to reactions in Table 1, the photocatalytic CO 2 reduction products are different over various photocatalysts with different CB and VB positions, which is related to the number of electrons and protons (e − /H + ) involved in reduction reactions. Actually, one electron involved reaction in the reduction of CO 2 is highly unfavorable thermodynamically due to the very negative redox potential of CO 2 + e − = CO 2 − (−1.90 V vs. NHE) [22]. Therefore, multiple electrons and a corresponding number of protons must be involved in the photocatalytic CO 2 reduction reactions. The clean fuel ethanol can be produced from the CO 2 photoreduction reaction involving twelve electrons and twelve protons, which requires a suitable photocatalyst with multiple electrons easily migrating from a photocatalyst to CO 2 . From the point of view of the four photocatalytic steps, a highly active photocatalyst should possess the following characteristics: (1) a large surface area for increasing the adsorption of CO 2 and the surface active sites; (2) a narrow bandgap and proper band positions for utilizing solar energy effectively; (3) a nanostructure favorable for electron transport and improving the separation of photogenerated electron-hole pairs; (4) abundant surface oxygen vacancies for changing the electronic and chemical properties of the semiconductor surfaces and facilitating CO 2 adsorption/activation. Additionally, the co-catalysts are usually attached on the surface of photocatalysts to promote the separation and migration of photo-induced carriers, and effectively lower the reaction energy barrier for CO 2 activation and reduction [33].

CO 2 Photoelectroreduction
Photoelectrocatalytic reduction of CO 2 is considered as an integration of photocatalytic and electrocatalytic CO 2 reduction, where the solar energy and electricity synergistically promotes the conversion of CO 2 to clean fuels. During the CO 2 photoelectroreduction process, the applied potential facilitates the separation of photogenerated electron-hole pairs in the photocatalytic step, and, in turn, the extra light irradiation could reduce the overpotential in the electrcatalytic step [14]. Photoelectrocatalytic CO 2 reduction system employs semiconductor materials as the photocathodes that can not only used as catalysts, but also as the light harvesting agents. Compared to photocatalysis, much more semiconductors even with a lower CB level than CO 2 redox potential could be function as the photocathodes. Figure 1 shows the CB, VB band edge positions versus an NHE and band gap energies for several common semiconductor photocathodes relative to CO 2 reduction potentials for different products at pH = 7. The CB levels of most of the semiconductors shown in the figure are below the single-electron reduction potential of CO 2 to CO 2 ·− , and only several of them are above the thermodynamic potentials of proton-assisted multi-electron reduction in CO 2 . The reported photoelectrocatalytic CO2 reduction over electrodes to date mainly contain pure semiconductors, metal-modified semiconductors and semiconductor/metal hybrid catalysts [34][35][36]. The incorporation of metal species into the semiconductor electrodes has intrigued considerable attention due to their large redox-active surface area, increased light-adsorbing surface area and enhanced electron transport ability. The employed electrode systems in CO2 photoelectroreduction process can be mainly categorized into two kinds of configurations, those are half-cells with sacrificial reagents donating electrons for oxidation in anode electrode and full-cells with semiconductor photoanodes to compensate electrons for oxidation [14]. In photoelectrochemical reaction cells, the semiconductor electrode is immersed in the electrolyte and is connected to a counter electrode (half-cell) or another semiconductor electrode (full-cell) via an external circuit. When the semiconductor electrode is illuminated under simulated solar light, its electrons can be excited from the VB to CB, leaving an equal number of holes in the VB. Simultaneously, with the aid of extra electric field, the charge accumulating at the interface between semiconductor electrode and electrolyte will give rise to a perturbation of the energy levels of the semiconductor [37]. Inspired by this kind of perturbation, the photogenerated electron-hole pairs are spatially separated and are injected into the electrolyte at the respective electrodes to produce electrochemical CO2 reduction and water oxidation reactions. Through the conversion of CO2 to chemicals or fuels, the abundant electricity and solar energy can thus be effectively converted into and stored as chemical energy.

Mechanisms of Ethanol Production from CO2
CO2 reduction is actually a kind of multiple electron-proton-involved reaction whether by electrocatalysis, photocatalysis or photoelectrocatalysis routes. Since CO2 is a linear molecule and one of the most thermodynamic stable carbon compounds, it demands input energy to activate CO2. Additionally, CO2 reduction to clean fuel ethanol using water as the reducing agent to supply protons and electrons, is an uphill reaction with a highly positive change in Gibbs free energy: 2CO2 + 3H2O = C2H5OH + 3O2 (∆G 0 = 1325.3 KJ·mol -1 ). Thus, input energy such as electricity or solar energy is required to overcome this reaction barrier. The most energetically demanding and likely key step is the activation of CO2 with one electron transfer from the catalyst to the lowest unoccupied molecular orbital (LUMO) of CO2 to form a CO2 ·− species [38]. The adsorption of CO2 onto the catalyst surface can allow for both heterogeneous electron transfer and stabilization of CO2 ·− species due to the decreased LUMO level of CO2 as the molecule bends [39]. Depending on the operating conditions, the type of reductant, the number and potential of the charge carriers involved in the reduction reaction, the following multiple electron-proton transfer steps determine the distribution of products. As shown in Figure 2, CO is formed by two electron-proton transfers through the intermediates of *CO2 ·− and *COOH, and is considered as the key intermediate to produce a C2 product like ethanol [40]. Competing with desorption and loss, a C1 intermediate like CO should be adsorbed on the catalysts surface firmly enough to persist until a second C1 intermediate is available for C-C coupling [41].  [42,43]. The key step of ethanol formation The reported photoelectrocatalytic CO 2 reduction over electrodes to date mainly contain pure semiconductors, metal-modified semiconductors and semiconductor/metal hybrid catalysts [34][35][36]. The incorporation of metal species into the semiconductor electrodes has intrigued considerable attention due to their large redox-active surface area, increased light-adsorbing surface area and enhanced electron transport ability. The employed electrode systems in CO 2 photoelectroreduction process can be mainly categorized into two kinds of configurations, those are half-cells with sacrificial reagents donating electrons for oxidation in anode electrode and full-cells with semiconductor photoanodes to compensate electrons for oxidation [14]. In photoelectrochemical reaction cells, the semiconductor electrode is immersed in the electrolyte and is connected to a counter electrode (half-cell) or another semiconductor electrode (full-cell) via an external circuit. When the semiconductor electrode is illuminated under simulated solar light, its electrons can be excited from the VB to CB, leaving an equal number of holes in the VB. Simultaneously, with the aid of extra electric field, the charge accumulating at the interface between semiconductor electrode and electrolyte will give rise to a perturbation of the energy levels of the semiconductor [37]. Inspired by this kind of perturbation, the photogenerated electron-hole pairs are spatially separated and are injected into the electrolyte at the respective electrodes to produce electrochemical CO 2 reduction and water oxidation reactions. Through the conversion of CO 2 to chemicals or fuels, the abundant electricity and solar energy can thus be effectively converted into and stored as chemical energy.

Mechanisms of Ethanol Production from CO 2
CO 2 reduction is actually a kind of multiple electron-proton-involved reaction whether by electrocatalysis, photocatalysis or photoelectrocatalysis routes. Since CO 2 is a linear molecule and one of the most thermodynamic stable carbon compounds, it demands input energy to activate CO 2 . Additionally, CO 2 reduction to clean fuel ethanol using water as the reducing agent to supply protons and electrons, is an uphill reaction with a highly positive change in Gibbs free energy: . Thus, input energy such as electricity or solar energy is required to overcome this reaction barrier. The most energetically demanding and likely key step is the activation of CO 2 with one electron transfer from the catalyst to the lowest unoccupied molecular orbital (LUMO) of CO 2 to form a CO 2 ·− species [38]. The adsorption of CO 2 onto the catalyst surface can allow for both heterogeneous electron transfer and stabilization of CO 2 ·− species due to the decreased LUMO level of CO 2 as the molecule bends [39]. Depending on the operating conditions, the type of reductant, the number and potential of the charge carriers involved in the reduction reaction, the following multiple electron-proton transfer steps determine the distribution of products. As shown in Figure 2, CO is formed by two electron-proton transfers through the intermediates of *CO 2 ·− and *COOH, and is considered as the key intermediate to produce a C 2 product like ethanol [40]. Competing with desorption and loss, a C 1 intermediate like CO should be adsorbed on the catalysts surface firmly enough to persist until a second C 1 intermediate is available for C-C coupling [41].  [40]. A subsequent seven electron-proton transfers to *C 2 O 2 H will result in the final production of ethanol. However, moderately adsorbed *CO species can further hydrogenate to *CHO, *CH 2 O or *CH 2 . The C-C bond formation by coulping *CHO, *CH 2 O or *CH 2 with CO could be more kinetically favorable than CO dimerization for some catalysts [44][45][46]. Notably, Cu is the only reported metal electrocatalyst that presents the optimal CO binding to the surface, and thus is capable of facilitating these value-added reactions.
Catalysts 2020, 10, x FOR PEER REVIEW 7 of 25 is the C-C bond made at low overpotentials through CO dimerization mediated by electron transfer rendering a *C2O2 − intermediate. Indeed, the C-C coupling involves only electron transfer, and proton transfer occurs after the formation of *C2O2 − intermediate to stabilize this intermediate for avoiding decoupling [40]. A subsequent seven electron-proton transfers to *C2O2H will result in the final production of ethanol. However, moderately adsorbed *CO species can further hydrogenate to *CHO, *CH2O or *CH2. The C-C bond formation by coulping *CHO, *CH2O or *CH2 with CO could be more kinetically favorable than CO dimerization for some catalysts [44][45][46]. Notably, Cu is the only reported metal electrocatalyst that presents the optimal CO binding to the surface, and thus is capable of facilitating these value-added reactions. Moreover, a radical mechanism was also reported for the ethanol production in photoelectrocatalytic CO2 reduction process [47]. As shown in Figure 3, the formation of the adsorbed intermediates *CHOO by the first electron and hydrogen radical transfer reaction is the key initial step for CO2 reduction. The subsequent multiple electron-proton-radical transfer reactions give rise to the formation of *CH2O, an important intermediate for C-C coupling and also for *CH3OH generation. Thus, the *CH2O intermediates should be stabilized on the catalyst's surface to some extent, hindering the formation of methanol as a product. Then, the *CH3OH intermediates can be cleaved into CH3 • and OH • radicals. C-C coupling is accomplished though the CH3 • radicals attacking *CH2O intermediates in the formation of *CH3CH2O • , and ethanol can be finally produced by a further one H • radical.  Moreover, a radical mechanism was also reported for the ethanol production in photoelectrocatalytic CO 2 reduction process [47]. As shown in Figure 3, the formation of the adsorbed intermediates *CHOO by the first electron and hydrogen radical transfer reaction is the key initial step for CO 2 reduction. The subsequent multiple electron-proton-radical transfer reactions give rise to the formation of *CH 2 O, an important intermediate for C-C coupling and also for *CH 3 OH generation. Thus, the *CH 2 O intermediates should be stabilized on the catalyst's surface to some extent, hindering the formation of methanol as a product. Then, the *CH 3 OH intermediates can be cleaved into CH 3 • and OH • radicals. C-C coupling is accomplished though the CH 3 • radicals attacking *CH 2 O intermediates in the formation of *CH 3 CH 2 O • , and ethanol can be finally produced by a further one H • radical.  [40]. A subsequent seven electron-proton transfers to *C2O2H will result in the final production of ethanol. However, moderately adsorbed *CO species can further hydrogenate to *CHO, *CH2O or *CH2. The C-C bond formation by coulping *CHO, *CH2O or *CH2 with CO could be more kinetically favorable than CO dimerization for some catalysts [44][45][46]. Notably, Cu is the only reported metal electrocatalyst that presents the optimal CO binding to the surface, and thus is capable of facilitating these value-added reactions. Moreover, a radical mechanism was also reported for the ethanol production in photoelectrocatalytic CO2 reduction process [47]. As shown in Figure 3, the formation of the adsorbed intermediates *CHOO by the first electron and hydrogen radical transfer reaction is the key initial step for CO2 reduction. The subsequent multiple electron-proton-radical transfer reactions give rise to the formation of *CH2O, an important intermediate for C-C coupling and also for *CH3OH generation. Thus, the *CH2O intermediates should be stabilized on the catalyst's surface to some extent, hindering the formation of methanol as a product. Then, the *CH3OH intermediates can be cleaved into CH3 • and OH • radicals. C-C coupling is accomplished though the CH3 • radicals attacking *CH2O intermediates in the formation of *CH3CH2O • , and ethanol can be finally produced by a further one H • radical.   Although numerous efforts have already been devoted to discover the intermediates and products involved in the CO 2 reduction process, the various proposed mechanisms of ethanol production indicate that more definitive studies are required.

The Advances of CO 2 Reduction to Clean Fuel Ethanol
Considering the fact that the linear CO 2 molecule is fully oxidized and extremely stable, it is rather difficult to convert CO 2 into fuels, especially for ethanol production involving multiple electron-protons. Therefore, whether using electricity, sunlight or both of them as input energy, it demands the corresponding specific catalysts to accelerate the CO 2 reduction reaction. Essentially, the performance of CO 2 reduction depends on the properties of the applied catalysts. In the following part, recent important progress in material exploration for CO 2 conversion to ethanol will be discussed in three categories, namely electrocatalysts, photocatalysts and photoelectrocatalysts, as shown in Scheme 3.
Catalysts 2020, 10, x FOR PEER REVIEW 8 of 25 Although numerous efforts have already been devoted to discover the intermediates and products involved in the CO2 reduction process, the various proposed mechanisms of ethanol production indicate that more definitive studies are required.

The Advances of CO2 Reduction to Clean Fuel Ethanol
Considering the fact that the linear CO2 molecule is fully oxidized and extremely stable, it is rather difficult to convert CO2 into fuels, especially for ethanol production involving multiple electron-protons. Therefore, whether using electricity, sunlight or both of them as input energy, it demands the corresponding specific catalysts to accelerate the CO2 reduction reaction. Essentially, the performance of CO2 reduction depends on the properties of the applied catalysts. In the following part, recent important progress in material exploration for CO2 conversion to ethanol will be discussed in three categories, namely electrocatalysts, photocatalysts and photoelectrocatalysts, as shown in Scheme 3. Scheme 3. Illustrations of the catalysts and the corresponding strategies for achieving elctrocatalytic, photocatalytic and photoelectrocatalytic CO2 conversion into clean fuel ethanol, respectively.

Electrocatalytic CO2 Reduction to Ethanol
Since the pioneering work on CO2 electroreduction to HCOOH over mercury cathodes was reported in 1954 [48], much research has been done on electrocatalytic CO2 conversion into fuels [10,15,18]. However, studies on the ethanol production from CO2 electroreduction have increased in the last five years. In this process, theoretically, electrons are released from water oxidation at the anode and travel through an external wire to the catalysts' surface at cathode to reduce CO2 to various products. The ethanol production is a combination of the oxidation reaction at anode and reduction reaction at cathode involving twelve electron-protons. As the catalysts for CO2 electroreduction, metals and metallic complexes have been extensively investigated [18,49]. Among these metal-containing catalysts, Cu-based catalysts have been reported as the most promising electrodes that are possibly capable of catalyzing the reduction of CO2 to clean fuel ethanol [50,51]. The unique catalytic property of Cu originates from its moderate binding energy for CO intermediates, as evidenced by Density functional theory (DFT) calculations. The currently identified Cu-based catalysts those can electroreduce CO2 to ethanol include modified Cu (morphology, size, facet, doping, organic additives, et al.), Cu-based alloys, Cu/carbon composites and Cu-based metal-organic porous materials (Table 2). Moreover, metal-free nitrogen-doped carbon materials have also been reported recently to be capable of ethanol production from CO2. Scheme 3. Illustrations of the catalysts and the corresponding strategies for achieving elctrocatalytic, photocatalytic and photoelectrocatalytic CO 2 conversion into clean fuel ethanol, respectively.

Electrocatalytic CO 2 Reduction to Ethanol
Since the pioneering work on CO 2 electroreduction to HCOOH over mercury cathodes was reported in 1954 [48], much research has been done on electrocatalytic CO 2 conversion into fuels [10,15,18]. However, studies on the ethanol production from CO 2 electroreduction have increased in the last five years. In this process, theoretically, electrons are released from water oxidation at the anode and travel through an external wire to the catalysts' surface at cathode to reduce CO 2 to various products. The ethanol production is a combination of the oxidation reaction at anode and reduction reaction at cathode involving twelve electron-protons. As the catalysts for CO 2 electroreduction, metals and metallic complexes have been extensively investigated [18,49]. Among these metal-containing catalysts, Cu-based catalysts have been reported as the most promising electrodes that are possibly capable of catalyzing the reduction of CO 2 to clean fuel ethanol [50,51]. The unique catalytic property of Cu originates from its moderate binding energy for CO intermediates, as evidenced by Density functional theory (DFT) calculations. The currently identified Cu-based catalysts those can electroreduce CO 2 to ethanol include modified Cu (morphology, size, facet, doping, organic additives, et al.), Cu-based alloys, Cu/carbon composites and Cu-based metal-organic porous materials (Table 2).
Moreover, metal-free nitrogen-doped carbon materials have also been reported recently to be capable of ethanol production from CO 2 . The morphology and structure of metallic Cu catalyst significantly affect the product distribution and selectivity of CO 2 reduction. For example, the in situ deposited Cu nanodendrites exhibited increased selectivity toward the formation of ethylene compared to the polycrystalline Cu [84]. It is known that the Cu (111) surfaces preferentially catalyze the methane formation, while the Cu (100) surfaces favor the formation of ethylene [40]. Hence, several strategies, such as controlling morphology, size or the exposed facet of the Cu catalyst, could be employed to attempt to produce the desired ethanol from CO 2 reduction. For instance, Smith and his co-workers prepared Cu nanowire arrays by electroreduction of Cu(OH) 2 and CuO nanowire arrays on Cu foil substrates [52]. On these electrocatalysts, the selectivity of hydrocarbon products at a fixed potential can be tuned by altering the length and density of Cu nanowire which is linked to the increased local pH within the nanowire arrays. Ethanol with a very low FE nearly 4% was produced at −1.1 V vs. RHE on Cu catalyst when the nanowire length increased to 7.3 µm or more. Jiao's group fabricated a nanoporous Cu catalyst through the annealing of Cu(OH) 2 nanorods and the electrochemical reduction of the nanoporous CuO [53]. When the porous Cu was integrated into a CO 2 flow cell electrolyzer with 1 M KOH as the electrolyte, it exhibits a high FE of 17% towards ethanol at the current density of 653 mA·cm −2 and the potential of −0.67 V vs. RHE. This kind of porous structure facilitates rapid gas transport across the electrode-electrolyte interface especially at high current densities. Similarly, the Cu catalysts synthesized by electrochemical oxidation-reduction cycling of Cu foil can electroreduce CO 2 to ethanol with an increased FE up to 18% at −1.0 V vs. RHE in 0.1 M CsHCO 3 [54]. Using the sol-gel Cu 2 (OH) 3 Cl as the precursor, Sargent's group presented an electro-redeposition method to prepare the Cu catalysts with controlled morphologies and oxidation states [55]. At −1.1 V vs. RHE, the electro-redeposited Cu catalyst exhibited a FE of 12% for ethanol product. Loiudice et al. reported the highest FE for ethanol (around 10%) achieved on Cu nanocrystal cubes with 24 nm edge length by tuning nanocrystal spheres (7.5 nm and 27 nm) to nanocrystal cubes (24 nm, 44 nm, and 63 nm) [56]. Overall, the cube-shaped Cu was more intrinsically active than the spheres, and smaller nanocrystals showed higher activity for the same morphology. Additionally, Jiang and co-workers tuned the facet exposure on Cu foil by the metal ion battery cycling method [57]. The 100-cycled Cu nanocube catalyst with exposed (100) facets exhibits a six-fold improvement in C 2+ to C 1 product ratio compared with the polished Cu foil and an ethanol FE of 13% at −0.95 V vs. RHE.
Additionally, the inclusion of a grain boundary into active sites of Cu-based electrocatalysts has been considered to improve the selectivity of electrocatalytic CO 2 reduction towards multi-carbon products. In a recent report [58], the grain boundary can be controllably grown and enriched in electrodeposited Cu by using the poly (vinylpyrrolidone) additive. The obtained grain-boundary-rich metallic Cu was able to convert CO 2 to ethanol with a high FE of 32% and a partial current density of −45 mA·cm −2 at −1.3 V vs. RHE in a flow cell, which is superior to the electrodeposited Cu without grain boundary.
Oxide-derived copper (OD-Cu) has been discovered as a simple method to improve the intrinsic catalytic properties towards C 2+ formation owing to the introduction of Cu + species on the surface [45,59,85]. A recent report of thick Cu 2 O-film-derived Cu catalysts achieved a higher FE of ethanol at lower overpotential than that on thin OD-Cu films, which can be attributed to the higher content of Cu + species [86]. Yeo's group systematically tuned the FE of ethanol by changing the thickness of the deposited Cu 2 O overlayers. The highest FE of 16% for ethanol formation was achieved on 3.6 µm film in 0.1 M KHCO 3 electrolyte at −0.99 V vs. RHE [45]. These systems have verified the promotion of ethanol production by Cu + species. However, the resultant Cu + species are prone to being reduced to Cu 0 under CO 2 reduction conditions, especially at the high applied reducing potentials required to produce ethanol [87]. Therefore, research efforts have been done to stabilize the Cu + species during CO 2 redution. For instance, a 3D dendritic Cu-Cu 2 O oxide composite was developed by in situ reduction in an electrodeposited copper complex on Cu substrate to keep the Cu + /Cu 0 ratio unchanged during CO 2 reduction reaction, which resulted in a high FE of 32% for ethanol formation [59]. Yu's group recently reported that the nanocavities in the multihollow Cu 2 O can confine carbon intermediates formed in situ, which, in turn, covers the local catalyst surface and thereby stabilized Cu + species [60]. At the potential of −0.61 V vs. RHE in 2 M KOH, this catalyst yields a maximum ethanol FE of 27% and delivers a high current density of −320 mA·cm −2 in a flow cell system (Figure 4). intermediates formed in situ, which, in turn, covers the local catalyst surface and thereby stabilized Cu + species [60]. At the potential of −0.61 V vs. RHE in 2 M KOH, this catalyst yields a maximum ethanol FE of 27% and delivers a high current density of −320 mA·cm -2 in a flow cell system ( Figure  4). Besides, the incorporation of heteroatoms into catalysts is another efficient approach to stabilized Cu + species and promote CO2 electroreduction. Introducing N into Cu to form Cu3N, when as the support of Cu catalyst, showed the enhanced FE (around 19%) for ethanol, which results from the stabilized Cu + by N in the Cu3N structure [61]. A B-doped oxide-derived-Cu has been reported to promote C2 formation with a higher Faradaic effiencicy (20%) than that of OD-Cu (12%), due to the Cu + species stabilized by the introduction of B [62]. It has also been reported that B can be used to tune the local electronic structure of Cu with positive valence sites, which results in boosting the ethanol formation with a high FE of 27% at −1.1 V vs. RHE in 0.1 M KCl [63]. By incorporating sulfur atoms in the catalyst core, and Cu vacancies in its shell, Sargent and his co-workers realized Cu2S-Cu-V core-shell nanoparticles that enhance CO2 reduction to ethanol with a high FE of 25% in a flow cell [64]. In a recent report, F atoms in the F-modified Cu catalyst facilitate the increase in Cu + sites and keeps them unchanged during long-term CO2 reduction [65]. Thus, a FE of 16% towards ethanol was achieved at −800 mA·cm -2 (−0.54 V vs. RHE) in a flow cell system. By increasing surface Cu + sites, the modification of F also promotes H2O activation to *H species, CO adsorption and the hydrogenation of *CO to a *CHO intermediate that can readily undergo coupling.
In order to accelerate H2O dissociation to *H species and change the H adsorption energy on Cu, Sargent's group reported a complementary approach of hydroxide doping to tune the *H species on Cu [66]. The enhanced *H coverage easily attacks the *HCCOH, forming *HCCHOH, the key intermediate towards ethanol. Hence, the most efficient Ce(OH)x-doped-Cu catalyst exhibits a high ethanol FE of 43% and a partial current density of −128 mA·cm −2 in a flow cell.
Bridging homogeneous molecular systems to tune heterogeneous catalysts has been considered a promising approach for the development of new electrodes, combining the advantages of both approaches [88]. When organic molecules or metal complexes are adjacent to heterogeneous active sites, the binding interactions may tune the stability of intermediates, and improve catalytic performance by increasing ethanol FE as well as decreasing overpotential. An good example of this bridge is N-substituted pyridinium additives, which are able to form a deposited film on Besides, the incorporation of heteroatoms into catalysts is another efficient approach to stabilized Cu + species and promote CO 2 electroreduction. Introducing N into Cu to form Cu 3 N, when as the support of Cu catalyst, showed the enhanced FE (around 19%) for ethanol, which results from the stabilized Cu + by N in the Cu 3 N structure [61]. A B-doped oxide-derived-Cu has been reported to promote C 2 formation with a higher Faradaic effiencicy (20%) than that of OD-Cu (12%), due to the Cu + species stabilized by the introduction of B [62]. It has also been reported that B can be used to tune the local electronic structure of Cu with positive valence sites, which results in boosting the ethanol formation with a high FE of 27% at −1.1 V vs. RHE in 0.1 M KCl [63]. By incorporating sulfur atoms in the catalyst core, and Cu vacancies in its shell, Sargent and his co-workers realized Cu 2 S-Cu-V core-shell nanoparticles that enhance CO 2 reduction to ethanol with a high FE of 25% in a flow cell [64]. In a recent report, F atoms in the F-modified Cu catalyst facilitate the increase in Cu + sites and keeps them unchanged during long-term CO 2 reduction [65]. Thus, a FE of 16% towards ethanol was achieved at −800 mA·cm −2 (−0.54 V vs. RHE) in a flow cell system. By increasing surface Cu + sites, the modification of F also promotes H 2 O activation to *H species, CO adsorption and the hydrogenation of *CO to a *CHO intermediate that can readily undergo coupling.
In order to accelerate H 2 O dissociation to *H species and change the H adsorption energy on Cu, Sargent's group reported a complementary approach of hydroxide doping to tune the *H species on Cu [66]. The enhanced *H coverage easily attacks the *HCCOH, forming *HCCHOH, the key intermediate towards ethanol. Hence, the most efficient Ce(OH) x -doped-Cu catalyst exhibits a high ethanol FE of 43% and a partial current density of −128 mA·cm −2 in a flow cell.
Bridging homogeneous molecular systems to tune heterogeneous catalysts has been considered a promising approach for the development of new electrodes, combining the advantages of both approaches [88]. When organic molecules or metal complexes are adjacent to heterogeneous active sites, the binding interactions may tune the stability of intermediates, and improve catalytic performance by increasing ethanol FE as well as decreasing overpotential. An good example of this bridge is N-substituted pyridinium additives, which are able to form a deposited film on polycrystalline Cu electrodes upon reduction, tuning the selectivity of ethanol formation [67]. A maximum ethanol FE of 31% was achieved on a polycrystalline Cu electrode with an N-tolylpyridinium chloride additive in a CO 2 -saturated 0.1 M KHCO 3 electrolyte at −1.1 V vs. RHE. Besides this, a nanostructured Cu electrode using N,N'-ethylene-phenanthrolinium dibromide as a molecular additive is capable of forming ethanol with a FE of 15% during CO 2 reduction at −1.07 V vs. RHE in 0.1 M KHCO 3 [68]. The organic molecule such as 1-octadecanethiol can also be used to modify the dentritic Cu electrode with hydrophobicity [69]. By surpressing HER, this kind of hydropobic electrode attains 17% FE of ethanol at −30 mA·cm −2 in 0.1 M CsHCO 3 compared to 4% on a hydrophilic equivalent. Another exciting example is the porphyrin-based metallic complex (5,10,15,20-tetraphenyl-21H,23H-porphine iron(III) chloride, FeTPP[Cl]) functionalizing Cu surface, which can provide intermediate-CO-rich local environment that facilitates C-C coupling and steers the reaction pathway towards ethanol [70]. By intergrating it into a flow cell system, the FeTPP[Cl]-functionalized Cu electrode exhibits a CO 2 -to-ethanol FE of 41% and a partial current density of −124 mA·cm −2 at −0.82 V vs. RHE ( Figure 5).
Catalysts 2020, 10, x FOR PEER REVIEW 12 of 25 polycrystalline Cu electrodes upon reduction, tuning the selectivity of ethanol formation [67]. A maximum ethanol FE of 31% was achieved on a polycrystalline Cu electrode with an N-tolylpyridinium chloride additive in a CO2-saturated 0.1 M KHCO3 electrolyte at −1.1 V vs. RHE. Besides this, a nanostructured Cu electrode using N,N'-ethylene-phenanthrolinium dibromide as a molecular additive is capable of forming ethanol with a FE of 15% during CO2 reduction at −1.07 V vs. RHE in 0.1 M KHCO3 [68]. The organic molecule such as 1-octadecanethiol can also be used to modify the dentritic Cu electrode with hydrophobicity [69]. By surpressing HER, this kind of hydropobic electrode attains 17% FE of ethanol at −30 mA·cm -2 in 0.1 M CsHCO3 compared to 4% on a hydrophilic equivalent. Another exciting example is the porphyrin-based metallic complex (5,10,15,20-tetraphenyl-21H,23H-porphine iron(III) chloride, FeTPP[Cl]) functionalizing Cu surface, which can provide intermediate-CO-rich local environment that facilitates C-C coupling and steers the reaction pathway towards ethanol [70]. By intergrating it into a flow cell system, the FeTPP[Cl]-functionalized Cu electrode exhibits a CO2-to-ethanol FE of 41% and a partial current density of −124 mA·cm −2 at −0.82 V vs. RHE ( Figure 5).

Cu Alloy
Coupling another metal with Cu, as a form of interface engieering has been suggested as an effective strategy to break the conventional scaling relationships and tune the binding energy of targered intermediates on Cu surface, thus enhancing the reaction kinetics and selectivity for CO2 reduction [89,90]. It is promising to design Cu bimetallic electrocatalysts, which will possess intriguing catalytic behavior with respect to that of single-metal electrocatalysts. Those metals (such as Au, Ag, Zn and Pd) with CO as the main product could provide abundant CO to couple with the key intermediate *CO or *CHO on Cu sites for further ethanol formation. For example, Cu63.9Au36.1 alloy electrode, which was prepared through electrochemical deposition with a nanoporous Cu film as the template, produced ethanol with an FE of 12% at −0.41 V vs. RHE in 0.5 M KHCO3 [71]. By pulsed electroreduction, ethanol was formed with a maximum FE of 25.5% over Cu55Ag45 alloy electrode among the Cu-Ag alloys with different atomic ratios [72]. The key factors for the selective ethanol production from CO2 are the formation of an oxide layer on Cu and desorption of intermediates on Ag under anodic bias. A kind of high-surface-area CuAg alloy wire was developed by electrodeposition method with 3,5-diamino-1,2,4-triazole (DAT) as an inhibitor [73]. The alloy film containing 6% Ag shows higher activity and selectivity for the electroreduction of CO2 to ethanol with FE of 25% in comparison to the CuAg poly (20%) without adding the DAT inhibitor, at a cathode potential of −0.7 V vs. RHE and a total current density of −300 mA·cm -2 . The origin of the selective ethanol formation is suggested to be the stabilization of Cu2O overlayer by CuAg wire and the optimal availability of the CO intermediate due to the Ag incorporated in the alloy. Another kind

Cu Alloy
Coupling another metal with Cu, as a form of interface engieering has been suggested as an effective strategy to break the conventional scaling relationships and tune the binding energy of targered intermediates on Cu surface, thus enhancing the reaction kinetics and selectivity for CO 2 reduction [89,90]. It is promising to design Cu bimetallic electrocatalysts, which will possess intriguing catalytic behavior with respect to that of single-metal electrocatalysts. Those metals (such as Au, Ag, Zn and Pd) with CO as the main product could provide abundant CO to couple with the key intermediate *CO or *CHO on Cu sites for further ethanol formation. For example, Cu 63.9 Au 36.1 alloy electrode, which was prepared through electrochemical deposition with a nanoporous Cu film as the template, produced ethanol with an FE of 12% at −0.41 V vs. RHE in 0.5 M KHCO 3 [71]. By pulsed electroreduction, ethanol was formed with a maximum FE of 25.5% over Cu 55 Ag 45 alloy electrode among the Cu-Ag alloys with different atomic ratios [72]. The key factors for the selective ethanol production from CO 2 are the formation of an oxide layer on Cu and desorption of intermediates on Ag under anodic bias. A kind of high-surface-area CuAg alloy wire was developed by electrodeposition method with 3,5-diamino-1,2,4-triazole (DAT) as an inhibitor [73]. The alloy film containing 6% Ag shows higher activity and selectivity for the electroreduction of CO 2 to ethanol with FE of 25% in comparison to the CuAg poly (20%) without adding the DAT inhibitor, at a cathode potential of −0.7 V vs. RHE and a total current density of −300 mA·cm −2 . The origin of the selective ethanol formation is suggested to be the stabilization of Cu 2 O overlayer by CuAg wire and the optimal availability of the CO intermediate due to the Ag incorporated in the alloy. Another kind of bimetallic Cu 85 Ag 15 foam was synthesized by an additive (citrate)-assisted electrodeposition approach [74]. Such a foam structure enables the phase-segregation of Cu and Ag, and the well-despersed nano-sized Ag in the Cu matrix. After activation by Cu oxidation/reduction, the Cu 85 Ag 15 foam shows high selectivity towards ethanol with an FE of 33.7% at −1.0 V vs. RHE in 0.5 M KHCO 3 . Bimetallic CuPd catalyst with phase-separated atomic arrangements could achieve a FE of 15% for ethanol formation at −0.75 V vs. RHE in 1 M KOH [75]. While the ordered and disordered Cu-Pd nanoparticles primarily produce CO. This demonstrates that geometric and structural effects may played a more improtant role than electronic effects in determing catalytic performance for various Cu-Pd bimetallic materials. Notably, the FE of CO 2 electroreduction toward ethanol could be tuned by introducing different amounts of Zn to generate an in situ source of mobile CO reactant, and was maximized to 29.1% on Cu 4 Zn alloy electrode at −1.05 V vs. RHE in 0.1 M KHCO 3 ( Figure 6) [44]. Similarily, the bimetallic CuZn catalyst synthesized by in situ electrochemical reduction in ZnO-shell/CuO-core bimetal oxide also shows a preference towards ethanol production with a high FE of 41.4% at −200 mA·cm −2 (−0.68 V vs. RHE) in a flow cell, in comparison to 32% at −1.15 V vs. RHE (−31.8 mA·cm −2 ) in a H-cell [76]. The in-situ-generated CO on Zn sites is believed to combine the adsorbed *CH 3 on Cu sites and form a *COCH 3 intermediate, which is exclusively reduced to ethanol. These results indicate that incorporating foreign metals into a Cu matrix can promote or alter the reaction routes of CO 2 reduction and the FE of ethanol formation is greatly dependent on the nanostructures and compositions of Cu-based alloys.
Catalysts 2020, 10, x FOR PEER REVIEW 13 of 25 of bimetallic Cu85Ag15 foam was synthesized by an additive (citrate)-assisted electrodeposition approach [74]. Such a foam structure enables the phase-segregation of Cu and Ag, and the well-despersed nano-sized Ag in the Cu matrix. After activation by Cu oxidation/reduction, the Cu85Ag15 foam shows high selectivity towards ethanol with an FE of 33.7% at −1.0 V vs. RHE in 0.5 M KHCO3. Bimetallic CuPd catalyst with phase-separated atomic arrangements could achieve a FE of 15% for ethanol formation at −0.75 V vs. RHE in 1 M KOH [75]. While the ordered and disordered Cu-Pd nanoparticles primarily produce CO. This demonstrates that geometric and structural effects may played a more improtant role than electronic effects in determing catalytic performance for various Cu-Pd bimetallic materials. Notably, the FE of CO2 electroreduction toward ethanol could be tuned by introducing different amounts of Zn to generate an in situ source of mobile CO reactant, and was maximized to 29.1% on Cu4Zn alloy electrode at −1.05 V vs. RHE in 0.1 M KHCO3 ( Figure 6) [44]. Similarily, the bimetallic CuZn catalyst synthesized by in situ electrochemical reduction in ZnO-shell/CuO-core bimetal oxide also shows a preference towards ethanol production with a high FE of 41.4% at −200 mA·cm −2 (−0.68 V vs. RHE) in a flow cell, in comparison to 32% at −1.15 V vs. RHE (−31.8 mA·cm −2 ) in a H-cell [76]. The in-situ-generated CO on Zn sites is believed to combine the adsorbed *CH3 on Cu sites and form a *COCH3 intermediate, which is exclusively reduced to ethanol. These results indicate that incorporating foreign metals into a Cu matrix can promote or alter the reaction routes of CO2 reduction and the FE of ethanol formation is greatly dependent on the nanostructures and compositions of Cu-based alloys.

Cu/Carbon Composites
Another viable strategy to stabilize the reaction intermediates and promote the ethanol formation is to incorporate porous carbons into Cu catalysts. The large surface area and pore volume of porous carbons will drive the thorough distribution of CO2 molecules on the surface of catalysts and create abundant active sites for CO2 conversion. The Cu2O nanoparticles grown on a carbon support can be transformed into small fragmented nanoparticles during CO2 electroreduction, which were densely connected to each other [77]. Such a unique morphology is proposed to promote C-C coupling and ethanol formation with FE of 12%. In a recent report, a nitrogen-doped carbon nanospike electrode with electronucleated Cu nanoparticles is shown to acquire a fairly high FE of 63% at −1.2 V vs. RHE in 0.1 M KHCO3 for the electroreduction of CO2 to ethanol [41]. Subsequently, an oxide-derived Cu/carbon catalyst prepared by a facile carbonization of Cu-based MOF (HKUST-1) at 1100 °C was reported to exhibit highly selective CO2 reduction to ethanol with a FE of 35% at −0.5 V vs. RHE in 0.1 M KHCO3 [78]. Such intriguing catalytic behaviors originate from the intrinsic activity of Cu and the synergetic interaction between Cu and neighboring porous carbons. In a recent report of N-doped porous carbon-supported Cu nanoparticles [79], the pyridinic N-decorated porous carbon could in situ produce the reactive CO intermediate, which will diffuse to neighboring Cu sites and combine with the C1 intermediates formed on Cu sites by C-C coupling to produce ethanol. By optimizing the pyridinic N content up to 3.43%, the maximum ethanol FE of 64.6% was achieved at −1.05 V vs. RHE in 0.2 M KHCO3.

Cu/Carbon Composites
Another viable strategy to stabilize the reaction intermediates and promote the ethanol formation is to incorporate porous carbons into Cu catalysts. The large surface area and pore volume of porous carbons will drive the thorough distribution of CO 2 molecules on the surface of catalysts and create abundant active sites for CO 2 conversion. The Cu 2 O nanoparticles grown on a carbon support can be transformed into small fragmented nanoparticles during CO 2 electroreduction, which were densely connected to each other [77]. Such a unique morphology is proposed to promote C-C coupling and ethanol formation with FE of 12%. In a recent report, a nitrogen-doped carbon nanospike electrode with electronucleated Cu nanoparticles is shown to acquire a fairly high FE of 63% at −1.2 V vs. RHE in 0.1 M KHCO 3 for the electroreduction of CO 2 to ethanol [41]. Subsequently, an oxide-derived Cu/carbon catalyst prepared by a facile carbonization of Cu-based MOF (HKUST-1) at 1100 • C was reported to exhibit highly selective CO 2 reduction to ethanol with a FE of 35% at −0.5 V vs. RHE in 0.1 M KHCO 3 [78]. Such intriguing catalytic behaviors originate from the intrinsic activity of Cu and the synergetic interaction between Cu and neighboring porous carbons. In a recent report of N-doped porous carbon-supported Cu nanoparticles [79], the pyridinic N-decorated porous carbon could in situ produce the reactive CO intermediate, which will diffuse to neighboring Cu sites and combine with the C 1 intermediates formed on Cu sites by C-C coupling to produce ethanol. By optimizing the pyridinic N content up to 3.43%, the maximum ethanol FE of 64.6% was achieved at −1.05 V vs. RHE in 0.2 M KHCO 3 .

Cu MOF
Additionally, Cu-based metal-organic porous materials like HKUST-1, CuAdeAce, CuDTA and CuZnDTA were also reported to electrocatalytically convert CO 2 to ethanol with FEs of 6%, 1%, 3% and 4%, respectively [91]. These catalysts possessing a relatively high surface area, accessibility, and exposure of Cu active sites yield many opportunities for further performance improvements.

N-doped Carbon Materials
Apart from Cu-based catalysts, metal-free nitrogen-doped carbon materials have been reported recently for electroredution of CO 2 to ethanol, and delivered comparable catalytic activities to Cu-based catalysts while possessing better durability. The electronegative nitrogen heteroatoms introduced into the carbon matrixes can increase charge density and convert the inert carbon structures to be highly active. Recently, Ajayan et al. developed nitrogen-doped graphene quantum dots with nanometre-size facilitating the production of ethanol with a FE of 16% at -0.75 V vs. RHE in 1 M KOH [80]. By doping mesoporous carbon with nitrogen, our group has explored a metal-free cylindrical mesoporous nitrogen-doped carbon as a robust catalyst for CO 2 electroreduction, enabling the efficient production of ethanol with an extremely high FE of 77% at −0.56 V vs. RHE in 0.1 M KHCO 3 (Figure 7) [81]. The superior electrocatalytic performance was abscribed to the synergy of nitrogen heteratoms and highly uniform cylindrical channel structures that can dramatically boost C-C bond formation in CO 2 electroreduction. Inspired by the potential of tuning the nanostructure of catalyst to acquire C 2 compounds, we further design a class of hierarchical porous N-doped carbon with medium micropores embedded in the channel walls of N-doped ordered mesoporous carbon by a pore-structure-engineering strategy [82]. The embedded medium micropores can not only enrich the exposed active sites (pyridinic and pyrrolic N), but also induce desolvation to accumulate electrolyte ions and enable high local electric potential. Both of them facilitate the activation of CO 2 molecules and the C-C coupling of key intermediates. Therefore, by scaling up the medium micropore content, the production rate of ethanol is increased to 2.3 mmol·g cat −1 ·h −1 , which is one order of magnitude higher than that of the counterpart without medium microproes (0.2 mmol·g cat −1 ·h −1 ). The FE towards ethanol generation could be maintained at a high value of 78% at −0.56 V vs. RHE. In another exciting example, boron and nitrogen co-doped nanodimond was reported for selective reduction of CO 2 to ethanol with a maximum FE of 93.2% at −1.0 V vs. RHE in 0.1 M NaHCO 3 [83]. The synegic effect of boron and nitrogen codoping and fine balance between nitrogen content and H 2 evolution potential drives the highly selective ethanol formation. These results open new insight into electrochemical conversion of CO 2 to clean fuel ethanol. microproes (0.2 mmol·gcat −1 ·h −1 ). The FE towards ethanol generation could be maintained at a high value of 78% at −0.56 V vs. RHE. In another exciting example, boron and nitrogen co-doped nanodimond was reported for selective reduction of CO2 to ethanol with a maximum FE of 93.2% at −1.0 V vs. RHE in 0.1 M NaHCO3 [83]. The synegic effect of boron and nitrogen codoping and fine balance between nitrogen content and H2 evolution potential drives the highly selective ethanol formation. These results open new insight into electrochemical conversion of CO2 to clean fuel ethanol.

Photocatalytic CO 2 Reduction to Ethanol
Photocatalytic CO 2 reduction has been paid consistent attention for several decades based on the utilization of solar energy and the concept of artificial photosynthesis [92][93][94]. During the reduction process of CO 2 , photocatalysts play a key role in lowering the potential of the electron-proton transfer reaction and the eventual catalytic performance. To date, many kinds of semiconductors have been employed as the photocatalysts for CO 2 reduction to solar fuels [95,96]. However, very few semiconductors like TiO 2 and graphitic carbon nitride (g-C 3 N 4 ) can photocatalyze the ethanol formation.

TiO 2
Actually, TiO 2 is considered the most appropriate candidate of photocatalysts due to its comparable conduction band energy (E cb ≈ −0.5 eV vs. NHE at pH = 7 as shown in Figure 1) to the reduction potentials of CO 2 (Reaction (1)-(7) in Table 1). However, it has yielded low CO 2 conversion rates to date, and mainly C 1 products of methane and methanol. For the sake of improving catalytic activity and producing ethanol, the incorporating strategies with metals, nonmetals and photosensitive materials have been adopted to modify TiO 2 . For example, Rh and Pd nanowires with high density of grain boundaries were in situ grown on TiO 2 nanosheets, acting as the cocatalysts to enhance photocatalytic CO 2 reduction performance [97]. The TiO 2 -Rh long nanowires and TiO 2 -Pd nanowires composites catalyzed CO 2 reduction to ethanol with an average production rate of 12.1 and 13 µmol· −1 ·h −1 , respectively, during the 4 h reaction under UV light (λ < 400 nm). Depositing Ni(OH) 2 nanosheets onto TiO 2 nanofibers could enhance charge separation efficiency and CO 2 capture capacity [98]. With 15 wt% Ni(OH) 2 loaded, 0.37 µmol·g −1 ·h −1 of ethanol was achieved over TiO 2 /Ni(OH) 2 hybrid catalyst. In another case, the incorporation of matrix facilitated the effective charge separation and CO 2 reduction, in which the average production rate of ethanol was maximized to 13.2 µmol·g −1 ·h −1 on 1.5wt%Ni 2+ -TiO 2 during 4 h of UV light irradiation [99]. Graphene quantum dots (GQDs) were combined with vanadium-doped TiO 2 (V-TiO 2 ) to effectively separate photogenerated electrons and holes, and 5%GQDs/V-TiO 2 exhibited the best photocatalytic activity with an ethanol production rate of 5.65 µmol g −1 h −1 under solar spectrum irradiation (Figure 8) [100]. The photosensitive AgBr with a narrow band gap was coupled with TiO 2 to improve the visible light activity, and the 23.2% AgBr/TiO 2 composite showed a relatively high ethanol yield of 13.28 µmol g −1 h −1 under visible-light irradiation for 5 h [101].
Ni(OH)2 nanosheets onto TiO2 nanofibers could enhance charge separation efficiency and CO2 capture capacity [98]. With 15 wt% Ni(OH)2 loaded, 0.37 μmol·g -1 ·h -1 of ethanol was achieved over TiO2/Ni(OH)2 hybrid catalyst. In another case, the incorporation of matrix facilitated the effective charge separation and CO2 reduction, in which the average production rate of ethanol was maximized to 13.2 μmol·g −1 ·h −1 on 1.5wt%Ni 2+ -TiO2 during 4 h of UV light irradiation [99]. Graphene quantum dots (GQDs) were combined with vanadium-doped TiO2 (V-TiO2) to effectively separate photogenerated electrons and holes, and 5%GQDs/V-TiO2 exhibited the best photocatalytic activity with an ethanol production rate of 5.65 μmol g −1 h −1 under solar spectrum irradiation (Figure 8) [100]. The photosensitive AgBr with a narrow band gap was coupled with TiO2 to improve the visible light activity, and the 23.2% AgBr/TiO2 composite showed a relatively high ethanol yield of 13.28 μmol g −1 h −1 under visible-light irradiation for 5 h [101].

G-C 3 N 4
As a novel nonmetallic semiconductor, g-C 3 N 4 with a moderate band gap (E g = 2.7 eV, as shown in Figure 1) has attracted significant attention in photocatalytic CO 2 reduction due to its high stability and responsiveness to visible light. It can be synthesized through the pyrolysis of some nitrogen-rich organic precursors, such as urea and melamine. The g-C 3 N 4 derived from urea (u-g-C 3 N 4 ) possessed a mesoporous flake-like structure with a larger surface area and photocatalyzed CO 2 reduction to ethanol in a yield of 4.5 µmol·g −1 ·h −1 with methanol as a co-product under visible-light irradiation for 12 h (Figure 9) [102], while the g-C 3 N 4 derived from melamine (m-g-C 3 N 4 ) without porous structure could exclusively yield ethanol at a lower rate of 3.6 µmol·g −1 ·h −1 . The above-mentioned different photocatalytic activities and selectivities for the formation of ethanol are possibly due to the differences in the crystallinity and microstructure of u-g-C 3 N 4 and m-g-C 3 N 4 . The non-porous structure of m-g-C 3 N 4 may not favor the fast exchange of the formed *OCH 3 or CH 3 OH, thus probably promoting the dimerization of *OCH 3 to form ethanol. Moreover, much effort has been devoted to improving the photocatalytic activity of g-C 3 N 4 via the combination with other semiconductors. For example, ZnO with a negative conduction band potential of −0.44 eV was coupled with g-C 3 N 4 by an impregnation method to generate ZnO/g-C 3 N 4 composite photocatalyst [103]. Although the CO 2 conversion rate was considerably enhanced over the optimal ZnO/g-C 3 N 4 composite, the ethanol yield was still as low as 1.5 µmol·g −1 ·h −1 under simulated sunlight irradiation. Meanwhile, the Ag 3 PO 4 /g-C 3 N 4 composite photocatalyst was also reported to significantly improve the CO 2 conversion rate, but exhibit a low ethanol yield of 1.3 µmol·g −1 ·h −1 under simulated sunlight irradiation [104]. When the two-dimensional g-C 3 N 4 nanosheets with few-layer thickness were used as the support of Pd to ensure equivalent charge migrations to various Pd facets, the selectivity of CO 2 photoreduction to ethanol strongly depends on the shapes of Pd nanocrystals on the C 3 N 4 nanosheets [105]. The optimal ethanol production rate was achieved on Pd nanotetrahedrons loaded on g-C 3 N 4 nanosheets with a Pd loading of 5.8 wt%, though the value only arrived at 2.18 µmol·g −1 ·h −1 . Therefore, it still remains challenging to selectively photocatalyze CO 2 reduction to ethanol over g-C 3 N 4 . μmol·g −1 ·h −1 under simulated sunlight irradiation [104]. When the two-dimensional g-C3N4 nanosheets with few-layer thickness were used as the support of Pd to ensure equivalent charge migrations to various Pd facets, the selectivity of CO2 photoreduction to ethanol strongly depends on the shapes of Pd nanocrystals on the C3N4 nanosheets [105]. The optimal ethanol production rate was achieved on Pd nanotetrahedrons loaded on g-C3N4 nanosheets with a Pd loading of 5.8 wt%, though the value only arrived at 2.18 μmol·g −1 ·h −1 . Therefore, it still remains challenging to selectively photocatalyze CO2 reduction to ethanol over g-C3N4.

Others
Apart from TiO2 and g-C3N4, some other photocatalysts have also been reported. In a recent report of simultaneously loaded CuO and Pt nanoparticles on reduced HCa2Ta3O10 perovskite nanosheets for sunlight-driven conversion of CO2, ethanol was formed as a significant product at a rate of 113 μmol·g −1 ·h −1 [106]. This can be ascribed to their unique structure. Pt nanoparticles with good contact with perovskite nanosheets could serve as excellent trapping sites for photogenerated electrons with a high transfer rate. Meanwhile, the introduction of CuO nanoparticles not only

Others
Apart from TiO 2 and g-C 3 N 4 , some other photocatalysts have also been reported. In a recent report of simultaneously loaded CuO and Pt nanoparticles on reduced HCa 2 Ta 3 O 10 perovskite nanosheets for sunlight-driven conversion of CO 2 , ethanol was formed as a significant product at a rate of 113 µmol·g −1 ·h −1 [106]. This can be ascribed to their unique structure. Pt nanoparticles with good contact with perovskite nanosheets could serve as excellent trapping sites for photogenerated electrons with a high transfer rate. Meanwhile, the introduction of CuO nanoparticles not only significantly improves the electron-hole separation through the formation of a p-n junction, but also enhances the adsorption of CO 2 and stabilizes C 1 intermediates, thus favoring C-C coupling to form ethanol. In another work, Wang et al. reported the synthesis of BiVO 4 /RGO nanocomposites for CO 2 photoreduction, which exhibited improved ethanol formation (5.15 µmol·g −1 ·h −1 ) in comparison to pure BiVO 4 (3.61 µmol·g −1 ·h −1 ) [107]. This improvement was attributed to the effective charge transfer of photo-generated electron from BiVO 4 to RGO and improved light absorption. In a later study, porous TaON microspheres were synthesized for CO 2 photoreduction via facile nitridation of uniform amorphous Ta 2 O 5 sphere formed by hydrothermal treatment [108]. Under the visible light, the conversion of CO 2 to ethanol was improved with a rate of 2.03 µmol·g −1 ·h −1 , which is attributed to the porous spherical architecture of TaON that provided more active sites, enhanced trapping of incident illumination, and promoted charge transfer/separation. Besides this, some other semiconductors, such as Ag@AgBr/CNT [109], red Ag/AgCl [110], Sr 3 Ti (2−x−y) Fe x S y O (7−z) N z [111] and Zn 0.8 Cd 0.2 S [112], have been used to produce ethanol from photocatalytic CO 2 reduction (Table 3). More effort is still required to improve these photocatalytical systems.

Photoelectrocatalytic CO 2 Reduction to Ethanol
Photoelectrochemical reduction of CO 2 has been investigated following its first discovery by Halman in 1978 [113]. Employing semiconductors, such as GaP, silicon and CdTe, as the photocathodes [12,14], the conversion of CO 2 into hydrocarbons, especially ethanol, can be realized in the presence of water under illumination and bias potential. In spite of the increasing researches on photoelectroreduction of CO 2 in the last five years, the reports on clean fuel ethanol formation were extremely rare. For instance, a Cu/Cu 2 O electrode prepared by electrochemical deposition method catalytically reduces CO 2 to ethanol with the maximum yield of 5.0 ppm in 0.1 mol L −1 Na 2 CO 3 under the bias potential of 0.2 V vs. Ag/AgCl and UV-Vis irradiation [35]. Equipped with Pt-reduced graphene oxides (RGO)/Cu foam cathode and TiO 2 nanotube photoanode, the phoelectrochemical cell exhibited an ethanol production rate of 105 nmol h −1 cm −2 under the potential of 2 V and UV-Vis irradiation, which was even significantly higher than that of the simple sum of electrochemical and phochemical processes (82 nmol h −1 cm −2 ) [114], indicating the synergetic effect of electrochemical and phochemical reductions. Importantly, ethanol was observed as the main product over the boron-doped g-C 3 N 4 electrodes with or without coupling with Au, Rh or Ag [115]. The yield of ethanol was maximized on boron-doped g-C 3 N 4 /Au electrode with a value of around 150 nmol under the bias potential of −0.4 V vs. Ag/AgCl and simulated solar irradiation. Afterwards, ZIF-8 was incorporated into Ti/TiO 2 nanotubes electrode to increase the photocurrent, resulting in the ethanol formation of up to 10 mmol L −1 under the bias potential of 0.1 V vs. Ag/AgCl and UV-Vis irradiation for three hours ( Figure 10) [116]. According to the results mentioned above, highly efficient production of ethanol through photoelectrocatalytic route and even industrialization has a long way to go.
150 nmol under the bias potential of −0.4 V vs. Ag/AgCl and simulated solar irradiation. Afterwards, ZIF-8 was incorporated into Ti/TiO2 nanotubes electrode to increase the photocurrent, resulting in the ethanol formation of up to 10 mmol L −1 under the bias potential of 0.1 V vs. Ag/AgCl and UV-Vis irradiation for three hours (Figure 10) [116]. According to the results mentioned above, highly efficient production of ethanol through photoelectrocatalytic route and even industrialization has a long way to go.

Conclusions and Perspectives
In conclusion, recent research has indicated the feasibility of producing ethanol from CO2 by electrochemical, photochemical and photoelectrochemical processes using solar energy and/or renewable electricity over advanced catalysts. Despite the challenges ahead, it is promising to develop highly efficient and economical catalytic systems that use renewable energy to selectively convert CO2 into clean fuel ethanol over active catalysts in the near future, thus realizing the sustainable development of human beings.
In future studies, more effort should be directed towards the following strategies to boost the performance of electrocatalysts for CO2-to-ethanol conversion: (1) introducing edges by nanostructuring with cubes, quantum dots, etc., introducing defects by doping and making pores, or introducing grain boundaries by controlled electrochemical growth, into the catalyst surfaces to increase the active sites; (2) designing nanostructured catalysts with special morphologies, such as multi-hollow, core-shell and nanoporous structures, which can confine the CO intermediates for further C-C coupling and ethanol formation; (3) employing metal or nonmetal doping strategies to chemically modify the structures of catalysts; (4) exploring composite materials with synergetic effect as the potential catalysts for CO2 reduction to realize the cascade reaction; (5) using certain catalysts with high overpotentials towards HER to suppress HER, which would compete electrons

Conclusions and Perspectives
In conclusion, recent research has indicated the feasibility of producing ethanol from CO 2 by electrochemical, photochemical and photoelectrochemical processes using solar energy and/or renewable electricity over advanced catalysts. Despite the challenges ahead, it is promising to develop highly efficient and economical catalytic systems that use renewable energy to selectively convert CO 2 into clean fuel ethanol over active catalysts in the near future, thus realizing the sustainable development of human beings.
In future studies, more effort should be directed towards the following strategies to boost the performance of electrocatalysts for CO 2 -to-ethanol conversion: (1) introducing edges by nanostructuring with cubes, quantum dots, etc., introducing defects by doping and making pores, or introducing grain boundaries by controlled electrochemical growth, into the catalyst surfaces to increase the active sites; (2) designing nanostructured catalysts with special morphologies, such as multi-hollow, core-shell and nanoporous structures, which can confine the CO intermediates for further C-C coupling and ethanol formation; (3) employing metal or nonmetal doping strategies to chemically modify the structures of catalysts; (4) exploring composite materials with synergetic effect as the potential catalysts for CO 2 reduction to realize the cascade reaction; (5) using certain catalysts with high overpotentials towards HER to suppress HER, which would compete electrons with CO 2 electroreduction reaction in the result of low efficiency; (6) incorporating suitable molecular catalysts to reduce the overpotentials of CO 2 -to-ethanol conversion by stabilizing the intermediates; (7) designing and optimizing flow cell system using gas-diffusion-electrode to improve the current density to commercially relevant levels; (8) designing catalysts with typical structure models for density functional theory (DFT) calculation and using operando techniques to study the reaction mechanism of CO 2 reduction.
In spite of the great efforts, it still seems quite challenging to efficiently photoreduce CO 2 to desirable products. Although CO 2 could be reduced to ethanol using some certain semiconductor catalysts by photochemical route, the yield and selectivity of ethanol was extremely low and hard to practice on a commercial scale. In the following, several strategies that may promote the ethanol