Solar-powered electrochemical reduction of CO2 and H2O to syngas, followed by fermentation, could lead to sustainable production of useful chemicals. However, due to insufficient electric current densities and instabilities of current CO2-to-CO electrolysers, a practical, scalable artificial photosynthesis remains a major challenge. Here, we address these problems using a commercially available silver-based gas diffusion electrode (used in industrial-scale chlorine–alkaline electrolysis) as the cathode in the CO2 electrolyser. Electric current densities up to 300 mA cm–2 were demonstrated for more than 1,200 hours with continuous operation. This CO2 electrolyser was coupled to a fermentation module, where the out-coming syngas from the CO2 electrolyser was converted to butanol and hexanol with high carbon selectivity. Conversion of photovoltaic electricity, CO2 and H2O to the desired alcohols achieved close to 100% Faradaic efficiency. Industrial production of useful and high-value chemicals via artificial photosynthesis is closer than expected with the proposed scalable hybrid system.
High-value organic chemicals, such as butanol and hexanol, have multiple applications in coatings, chemical synthesis, and as solvents and fuels. Currently, about four million metric tonnes of 1-butanol are produced worldwide. In Germany, about 0.6 million metric tonnes of 1-butanol are produced from fossil fuels per year1. Until now, these organic chemicals have been produced largely from oil. This production is complex, costly and is not sustainable due to limited oil sources. The value that is addressed here is the ratio between the commercial price of butanol (1.2€ per kg) and its heat of combustion (10.80 kWh per kg) = 0.11€ per kWh relative to that of methane (≈0.15€ per kg):(15.42 kWh per kg) = 0.01€ per kWh.
At the end of their life cycle, most organic chemicals end up in the atmosphere as the greenhouse gas CO2, the concentration of which is rapidly and continuously rising. Attempts are therefore being made to generate organic chemicals from CO2 and water powered by renewable electricity as so-called ‘artificial photosynthesis’2,3,4. However, relative to fossil fuels, renewable energy has a low energy density (for example, in Germany, installed renewable electric power generating 100 GW from a combination of photovoltaics (PV), windmills and biomass combustion is distributed over almost 400,000 square kilometres; https://www.energy-charts.de/power_inst_de.html). Therefore, it is necessary to decentralize industrial production of organic chemicals from CO2 and water via electro-catalysis, to be smaller scale and to run with less chemical infrastructure5,6,7. By-product accumulation is typical, for example in the Fischer–Tropsch process8, and thus elaborate separation techniques must also be avoided. This favours the involvement of microorganisms as they are often more selective than chemical catalysts to produce desired organic compounds. The Faradaic efficiency (efficiency of electron transfer into products) must be high in terms of converting all electricity into only the desired products. This favours the involvement of anaerobic microorganisms. If aerobic microorganisms are involved, a significant portion of the electrons can end up in water, increasing the electricity demand. Further, anaerobic fermenters are less expensive (there is less corrosion and in many cases sterilization is not required) than aerobic ones.
Herein, we describe the production of butanol and hexanol from CO2, H2O and renewable energy and how such a decentralized industrial plant with a capacity of a few 10,000 tonnes of butanol and/or hexanol per year might operate. A conventional PV module providing electricity with an energy conversion efficiency of around 20% (ref. 9), a CO2 electrolyser constantly operating at high current densities and a bioprocess module for the anaerobic conversion of CO, H2 and CO2 (syngas) to the desired alcohols were employed (Fig. 1 and Supplementary Fig. 1).
The CO2 electrolyser in the hybrid system is the heart of this technical photosynthesis (Fig. 1) as it produces a mixture of mainly CO and H2 (syngas of various composition) from electricity and CO2 and H2O. The lower redox potential of the CO/CO2 couple (E0′ = −520 mV) as compared to that of the H+/H2 couple (E0′ = −414 mV) allows for anaerobic bacteria to perform a number of reduction reactions, such as reduction of carboxylic acids to their respective aldehydes (E0′ = −580 mV), or the reductive carboxylation of acetyl-CoA to pyruvate (E0′ = −500 mV). These reactions are thermodynamically and kinetically unfavourable; if H2 was the sole electron donor in conjunction with CO2 as the carbon source10, practically no alcohols or pyruvate-derived products would be formed11,12. Therefore, the presence of CO is essential to gain the desired product selectivity.
For an industrial-scale electrochemical reduction, CO2 concentrations higher than atmospheric concentrations are needed to produce larger concentrations of CO. Renewable sources of CO2 include breweries, anaerobic digestion plants and power plants that convert biomass into electricity. CO2 is also a side product of cement and steel production13. All of these CO2 production sites are widely dispersed. CO2 can also be concentrated from the atmosphere (0.04%) at a theoretical energy cost of at least 20 kJ mol–1. Carbon capture from the atmosphere cannot be centralized because scrubbing out 1 mol CO2 (44 g) requires at least 56,000 litres of air, which is another reason why industrial production of organic chemicals from CO2 and H2O should be decentralized.
In CO2 electrolysers, CO2 rather than HCO3– or CO32– is reduced at a gas diffusion cathode14. Furthermore, CO2 has a low solubility in salt-based electrolytes. Thus, the electric current density at the cathode’s surface is generally too low for carrying out any technical applications (only a few mA cm–2) when operating the electrolyser at a CO2 pressure of 1 bar and ambient temperature15,16,17,18,19,20,21, despite the fact that the Faradaic and energy conversion efficiencies are high22,23. We solved this problem by letting the gas diffusion cathode interact at its opposite side with 100% gaseous CO2 rather than dissolved CO2 (Fig. 2a). Covestro recently introduced such an operation mode in industrial chlorine–alkaline electrolysis called oxygen depolarization cathode (ODC) (cathode reaction: ½O2 + H2O + 2e– → 2OH−; anode reaction: 2Cl– – 2e– → Cl2; https://www.covestro.com/en/sustainability/lighthouse-projects/sauerstoffverzehrkathode), which can work at electric current densities of more than 400 mA cm–2 (ref. 24). There ODC was introduced to suppress the hydrogen evolution reaction (HER) and therefore decrease cell voltage and increase energy efficiency. Surprisingly, these commercially available gas diffusion electrodes also work in CO2 electrolysis.
At this silver-based cathode, CO2 was reduced to CO (CO2 + H2O + 2e– → CO + 2OH−) and at the iridium-oxide-coated titanium anode25,26, water was oxidized to O2 (H2O – 2e– → ½O2 + 2H+). In close proximity of the cathode, the pH becomes very basic (CO2 + H2O + 2e– → CO + 2OH−), which in the bulk catholyte is neutralized by excess of CO2 (2OH− + 2CO2 → 2HCO3−). Thus, the overall cathode chamber reaction is: 3CO2 + H2O + 2e− → CO + 2HCO3−. In other words, three CO2 molecules have to be supplied to the gas diffusion electrode to accomplish electrochemical reduction of one CO2.
The current–voltage characteristic of the CO2 electrolyser (Fig. 2a) is depicted in Fig. 2b. The electric current density increased with increasing voltage and temperature. Extrapolation to zero current density yielded a minimal cell voltage of 2.3 V. The increase in current density was due to an increase in electrolyte conductivity when the temperature was raised from 30 °C to 60 °C.
At the silver-based gas diffusion electrode of the gas-to-gas CO2 electrolyser (Fig. 2a), not all the CO2 supplied could be converted to CO, and H2 was competitively formed by proton reduction (HER). CO-Faradaic efficiency and H2-Faradaic efficiency still totalled around 100%, because no other gaseous products were formed. The formation of traces of formic acid27 can be neglected. The CO to H2 ratio depends on the CO2-supply rate (Fig. 2c and Table 1), because the rate of CO2 reduction to CO at a given current density is kinetically limited by the diffusion of CO2 to and CO from the cathode in the presence of a huge excess of H2O. In contrast to most literature, our electrolysers were continuously operated far away from the thermodynamic limit to yield an industrially applicable current density. The maximum volumetric CO formation rate (l s−1) is given by the total cell current Icell (A)
is the molar volume of CO (22.4 l mol−1), F is the Faraday constant (96,485 A s mol−1), and z is the number of electrons required to reduce CO2 to CO, which is two. In order to characterize the operation conditions, we introduced the experimental parameter λ as the ratio of the volumetric CO2 supply rate (l s−1) and
Usually, λ > 3 is required to sufficiently reduce the competing HER. λ increases with increasing CO2 flow rate and decreases with increasing electric current density. Furthermore, catholyte and anolyte are continuously mixed to ensure constant pH, K+ and HCO3– concentration during the course of the long-term experiment.
At 300 mA cm–2 and a CO2 flow rate of 90 sccm (λ = 4.78) the Faradaic efficiency for CO was near 70% and that for H2 near 30% (Fig. 3a), and both remained almost constant within experimental error for more than 1,200 h (Fig. 3b) due to anolyte and catholyte mixing. Such high CO-Faradaic efficiencies at high current densities have in the past only been achieved for minutes under specific conditions, where in the electrolyser the electrolyte was cycled only once through the electrolysis cell and then discarded18 or when electrocatalytic CO2 reduction to CO was performed at high CO2 pressure (>15 atm)21.
The minimum cell voltage of 2.3 V (Fig. 2b) corresponds to a theoretical maximum total (CO + H2) energy efficiency of 64%, assuming an adiabatic cell potential of 1.47 V (for calculation see Methods). With increasing current density the CO energy conversion efficiency decreased (Fig. 3a). At 300 mA cm–2, the total energy efficiency for CO and H2 dropped to nearly 20%. Therefore, when CO:H2 ratios lower than around 2:1 (Fig. 3b) are anticipated, it is more energy efficient to use in parallel a separate H2O electrolyser rather than to increase the electric current density in the CO2 electrolyser. H2O-electrolysers operate stably at electric current densities of, for instance, 1.5 A cm–2 at around 2 V with an energy efficiency of around 70% (ref. 28).
The syngas collected was almost free of O2 (≪100 ppm), because the catholyte was separated from the site of CO2 reduction by the gas diffusion electrode. Any O2 diffusing into the cathode from the front side or dissolved in the electrolyte was reduced to OH− (ODC reaction), which reacted with CO2 to form HCO3– (Fig. 2a). The concentration of this O2 was very low, as can be deduced from the Faradaic efficiency of CO2 and H2O reduction to CO and H2, respectively, which was found to be nearly 100%. Very low O2 concentrations in the out-coming syngas are of importance, as higher concentrations are toxic for the bacteria converting the syngas to butanol and hexanol in the fermentation phase.
During development of the CO2 electrolyser described above we first experimented with a preliminary operational mode shown and described in Supplementary Fig. 2a. The main differences were independent electrolyte cycling and H2SO4 as anolyte. At current densities of 50 mA cm–2, this CO2 electrolyser also generated CO at high Faradaic efficiency for more than 1,000 h (Supplementary Fig. 3), but the CO Faradaic efficiency decreased more with increasing current density than in the setup shown in Fig. 2a. We coupled this preliminary CO2 electrolyser with a commercially available PV module and measured the power dependence on the voltage for the coupled system (Supplementary Fig. 4). The two curves intersected near 2.8 V and a power near 1.5 W, corresponding to a current density of about 0.5 A per 10 cm2. At the intersection point, solar module and CO2 electrolyser were optimally coupled and didn’t require extensive regulation.
Reference electrodes, which would have allowed determination of the rate-limiting electrode, were not included in our devices, because we wanted to work at high current densities and, thus, with a very high gas load. At high current densities, conductivity of the electrolyte becomes one major contribution to energy losses, which can be roughly calculated from the distance between anode and cathode and the corresponding cell voltage (Figs. 2 and 3). CO2 evolution (gas bubbles) on the anode side might be another loss channel (see Methods).
Conversion of syngas to acetate and ethanol
In a first experiment, syngas was produced by the PV-module-powered CO2 electrolyser at a rate of 16.52 sccm and a composition of 11.76 % CO (4.8 mmol h–1), 6.37 % H2 (2.6 mmol h–1) and 81.86 % CO2 (33.4 mmol h–1). This mixture was fed into two 1-litre fermenters, each with 0.5 litre culture at 36 °C, in which CO, H2 and CO2 were converted to acetate (reaction (1)) and ethanol (reaction (2)) by the activity of C. autoethanogenum11,29,30,31,32. The acetogen can, like the closely related C. ljungdahlii, grow mixotrophically33 and is amenable to genetic modification12,29.
∆G0′ = −94.6 kJ mol–1; ∆H0′ = −310.3 kJ mol–1
∆G0′ = −29.7 kJ mol–1; ∆H0′ = −34.3 kJ mol–1
After 50 h of cell growth, stationary conditions were reached, where the cell concentration, the rates of CO and H2 consumption and those of acetate and ethanol formation remained substantially constant for the next 45 h. The consumed CO and H2 (3.35 mmol h–1 in total) agreed well with the mol electron pairs (3.44 mmol h–1) required to reduce CO2 to acetate and ethanol (Table 2). The Faradaic efficiency (here the expression Faradaic efficiency is analogous to the characterization of the electrolyser) was almost 100%, indicating that essentially only acetate and ethanol were formed under the experimental conditions. The efficiency of energy conversion in syngas fermentation, calculated from the amount of ethanol produced (reactions (1) + (2)), was almost 80% (see Methods). Considering that the energy conversion efficiency of the CO2 electrolyser at a low electric current density of about 50 mA cm–2 was nearly 50% (Fig. 2c) and that of existing PV modules can be nearly 20%, the overall energy conversion efficiency of the described artificial photosynthesis system can be as high as 8% (Fig. 1). Similarly high conversion efficiencies have recently been described for a hybrid water splitting biosynthetic system involving Ralstonia eutropha for the reduction of CO2 with H2 to fuel alcohols34,35, albeit at much lower electric current densities. An integrated electromicrobial conversion of CO2 via formate to higher alcohols using recombinant R. eutropha strains also proceeded only at low current densities36.
Besides acetate and ethanol, C. autoethanogenum is known to produce small amounts of 2,3-butanediol and lactic acid29. These side products were, however, only formed in significant amounts when the CO concentration and the CO:H2 ratio in the syngas were much higher.
Conversion of syngas to butanol and hexanol
In a second experiment, the CO2 electrolyser was run at an electric current density of 150 mA cm–2 (Supplementary Fig. 3b), resulting in a constant syngas production at a flow rate of 16.23 sccm and composition of 10% CO (4 mmol h–1), 60% H2 (24.2 mmol h–1) and 30% CO2 (12.2 mmol h–1). Again the syngas was fed into two 1-litre fermenters each with 0.5 litre C. autoethanogenum culture. 22 h later, the fermenters were additionally inoculated with C. kluyveri cells. At the point of addition, the fermenter already contained acetate and ethanol, formed from CO, H2 and CO2 by the activity of C. autoethanogenum (reactions (1) and (2)). This inoculation led to the acetate and ethanol being converted to butyrate (reaction (3)) and hexanoate (reaction (4)) by C. kluyveri37,38 and then to butanol (reaction (5)) and hexanol (reaction (6)) by C. autoethanogenum10,39,40,41,42.
After steady state conditions were reached, the rates of CO and H2 consumption and those of acetate and ethanol formation remained substantially constant for 45 h. The consumed CO and H2 (14.88 mmol h–1 in total) agreed well again with the mol electron pairs (14.52 mmol h–1) required to reduce CO2 to the three acids and alcohols (Table 3). The Faradaic efficiency was close to 100%. The energy conversion efficiency of formation of butanol and hexanol from CO, H2 and CO2 was nearly 78% (see Methods).
As shown in Table 3, the conversion of CO, H2 and CO2 into butanol and hexanol can take place in a mixed culture of C. autoethanogenum and C. kluyveri under non-growth conditions. For an industrial process, however, it is important that the conversion takes place under chemostat (continuous) conditions and in two separate fermenters, as the optimum performance conditions of C. autoethanogenum and C. kluyveri differ. This is the basis for the design of an industrial process, forming butanol and hexanol at a scale of 10,000 tonnes per year (Supplementary Fig. 1). A prerequisite for a technical process is that the rate of carbon atoms being fixed in the gas fermenter is the same as the rate of carbon atoms being transformed to butanol and hexanol. For continuous fermentation, accumulation of the intermediates acetic acid and ethanol, as well as butyric acid and hexanoic acid, should be avoided. Here, we have shown experimentally that this prerequisite can be fulfilled (Supplementary Table 1).
To calculate the necessary current, pentanol is used as an example because it represents a 1:1 mixture of hexanol and butanol. Around 3.4 × 109 mol electrons are required to synthesize 10,000 tonnes pentanol. This means 328 × 1012 coulomb per year or 10.4 × 106 A (1 A = 1 C s–1). At a minimal CO to H2 consumption ratio of 1/5 in the fermentation (see reactions (9) and (10) in Methods), 16.7% of these electrons must be provided by the CO2 electrolyser. It would possess a power of 8.2 MW (4.7 V × 1.74 × 106 A), assuming a supply voltage of 4.7 V (Fig. 2b). The remaining 83.3% of the required electrons are provided by the H2O electrolyser, which at around 2.0 V, for example, would have to run at a power of 17.3 MW (2.0 V × 8.66 × 106 A). The overall plant would therefore consume 8.2 MW + 17.3 MW = 25.5 MW. This power could be provided by 14.6 × 104 m2 = 14.6 hectare PV modules with power density values of 175 W m–2. Taking into account night-time, location and weather of the hypothetical plant, an area 5–10 times larger would be required (https://en.wikipedia.org/wiki/List_of_photovoltaic_power_stations).
For the example of the production of 10,000 tonnes of the alcohols, 568 × 106 mol (that is, 25,000 tonnes) of CO2 are required. The CO2 has to be fed into the CO2 electrolyser at a rate of 1.1 × 103 mol min–1, which is equal to a flow rate of 24.6 × 106 sccm (100% CO2 at 105 Pa). At 4.7 V (30 °C) and a λ of 4.3, the input CO2 flow rate of our CO2 electrolyser was 90 sccm (Supplementary Fig. 3). Therefore, the CO2 electrolyser would have to be scaled up by a factor of 270,000. This can be achieved by increasing the electrode surface area from 10 cm2 to 1 m2 (factor 103) and by stacking 270 electrolysis cells. Both of which in principle are possible and are being done.
In the fermenters, butanol and hexanol were produced together at a rate of 0.6 × 10–3 mol per hour per litre of culture (Supplementary Table 1). For the production of 10,000 tonnes of butanol and hexanol per year (that is, 12.9 × 103 mol alcohol per hour), the fermenters would have to be scaled up by a factor of 21.6 × 106. This scale-up may be achieved by increasing the cell concentration in the fermenters by a factor of up to 30 and the volume from 1 litre to 700,000 litres, a volume not unusual in industrial fermentations43,44,45,46,47. At this large volume it is advantageous that C. autoethanogenum and C. kluyveri grow in the fermenters in media containing only vitamins and minerals. The highly selective conditions make sterilization of the media unnecessary. This is especially important because it significantly reduces investment costs.
Another advantage of this set-up is that the products, butanol and hexanol, are easily extracted, allowing the unused media components to be easily recycled, which reduces the costs of water and media components. Butanol and hexanol are separated in two steps from the aqueous broth. The first step is to extract the alcohols using oleyl alcohol as an extracting solvent. The second step is to separate the alcohols from the solvent by distillation and recycle the solvent. The main energy consumption of this separation is the distillation step. To vaporize hexanol, 0.16 kWh kg–1 is needed, because of the vaporization enthalpy of 60 kJ mol–1. In addition, the solvent has to be heated up to its vaporization temperature. This heat can only partly be recovered. Therefore, in sum roughly 0.6 kWh kg–1 hexanol energy is required to separate the products. This can be neglected compared to the energy consumed to synthesize the molecules, which is roughly 22 kWh kg–1 alcohol.
The advantages discussed above and the anticipated future low electricity costs (Fig. 1) all contribute to making this relatively small-scale approach competitive with traditional high-scale chemical productions of butanol and hexanol.
Under the experimental conditions described, butanol and hexanol were produced at almost the same rate (Supplementary Table 1). However, the ratio can be tuned, since it depends on the acetate to ethanol ratio48, which in turn depends on the CO/H2 ratio in the syngas29. This is important since at present the butanol market1 is larger than the hexanol market49, but this may vary in future. Hexanol, apart from its present direct use in perfumes and other cosmetic applications, can be chemically dehydrated to 1-hexene50 which is an important co-monomer for the production of polyethylene. Since hexanol is a liquid, it can also be easily transported to any polyethylene producer. Furthermore, hexanol could immediately substitute gasoline from fossil sources.
Other microorganisms can be combined with C. autoethanogenum instead of C. kluyveri, namely Pelobacter propionicus51, which ferments ethanol plus CO2 to propionate and acetate or oleaginous yeast52, which convert ethanol and acetate to lipids. The right combination allows the production of various platform chemicals via artificial photosynthesis involving CO2 electrolysis, as described here, at high Faradaic efficiencies.
The results reported above are examples. They are backed up, however, by many additional experiments performed under almost identical or very similar conditions leading to the same results.
Free energy and enthalpy changes are given under standard conditions: CO, H2 and CO2 as gas at 105 Pa and acetate, ethanol, butyrate, hexanoate, butanol and hexanol at 1 molal solution.
In our experiments the silicon PV modules were irradiated with a solar simulator with 1,000 W m–2, simulating conditions on a sunny day with an air mass coefficient of 1.5 (standard test conditions). The temperature of the PV module was kept constant at 25 °C by a Peltier element. PV module (12.8 × 12.8 cm2 = 164 cm2), from Conrad Electronics, consisted of 32 PV cells (1.7 × 2.6 cm2), each eight connected in series. At the maximum power point (1.8 W; Supplementary Fig. 4), the PV module delivered 482 mA per 10 cm2 at 3.7 V and operated at a solar energy conversion efficiency of 11% (1.8 W × 100% / 16.4 W).
Commonly, in CO2 electrolysers the anolyte and catholyte contain mostly KHCO3 at a pH around 7. The anolyte and catholyte are separated by a Nafion membrane. Due to the concentration difference of protons (10–7 M) and potassium cations (10–1 M), the Nafion membrane is mostly potassium conducting. This leads to the following catholyte reaction: 3CO2 + H2O + 2e– → CO + 2HCO3–. With the potassium ions from the anolyte, the catholyte concentration of KHCO3 increases proportional to the charge flow until precipitation. The anolyte reactions are as follows: H2O – 2e– → ½O2 + 2H+ and 2HCO3− + 2H+ → 2CO2 + 2H2O. The concentration of KHCO3 decreases proportional to the charge flow until all of the potassium cations have moved to the catholyte and only water is left as anolyte. To overcome this problem, two operation modes were tested: (i) in Supplementary Fig. 2a an almost continuous operation mode with sulfuric acid as the anolyte and KHCO3 as the catholyte separated by a Nafion membrane is shown. Both anolyte and catolyte are running in separated cycles. However, with each proton, about four molecules of water are pulled through the membrane, which leads to a concentration of sulfuric acid in the anolyte and a dilution of KHCO3 in the catholyte. (ii) The final solution is described schematically in Fig. 2a. In this set-up a mixture of KHCO3 and K2SO4 is used both as catholyte and anolyte, which are mixed after each cycle with the result that cations, anions, pH and water remain balanced. A Nafion membrane is no longer needed and is replaced by a ZrO2 diaphragm. Only the latter set-up is described here in more detail.
A commercial silver-based gas diffusion electrode (GDE) from Covestro, typically used as an oxygen depolarized cathode (ODC) in chloralkaline electrolysis24,54, was used as cathode. The silver-based cathode is sensitive to H2S (ref. 55) and other contaminants present in CO2 obtained from cement and steel production, alcohol formation in breweries, anaerobic digestion plants and power plants converting biomass into electricity. Therefore, the CO2 has to be freed of these contaminants before being used in the CO2 electrolyser.
As anode, an Ir-MMO (iridium-mixed metal oxide) coated titanium sheet was used25,26, which was purchased with the flow cell from ElectroCell. Such anodes are typically used in PEM type electrolysers (polymer / proton exchange membrane) for hydrogen production.
The aqueous catholyte and anolyte consisted of mixtures of KHCO3 and K2SO4 (see captions of Figs. 2 and 3) at pH 7 without further additives. The conductivity was about 0.13 S cm–1, which is low compared to alkaline hydrogen electrolyses (~1 S cm–1). In CO2 electrolysis, highly conductive catholytes like bases or acids cannot be used. While acids tend to form only H2 at the cathode due to the high proton concentration, bases like KOH lead to the formation of CO32– and HCO3−. Neither can be reduced at the cathode and may form the salts K2CO3 or KHCO3, which tend to crystallize and thus clog the GDE.
The electrolyte solutions were prepared with ultra-pure water (18.2 MOhm cm at 273 K, MilliQ Millipore system). Potassium hydrogen carbonate, potassium sulfate and sulfuric acid were used without further purification and were purchased from Alfa Aesar. All experiments were conducted by pre-saturating the electrolyte with CO2 for 30 min.
All electrochemical experiments were conducted in a flow cell, which was purchased by ElectroCell (Micro Flow Cell) and slightly modified. This flow cell is a sandwich of three compartments with two liquid channels containing the catholyte and anolyte, and one CO2 gas channel behind the cathode. The anode and the cathode compartments are separated by a ZrO2 diaphragm. The surface area of each electrode was 10 cm². CO2 gas (4.8 Linde) was used without further purification and continuously supplied at a constant flow rate controlled by a mass flow controller (Bronkhorst F-200CV-002-AAD-33-V) at atmospheric pressure. The flow rates are shown in the corresponding graphs.
A peristaltic pump (Ismatec ECOLINE VC-MS/CA8-6) was used to supply the electrolyte between the anode and GDE cathode at a flow rate of 200 ml per min. The flow rate had only minor effects on the CO Faradaic efficiency. The electrolyte reservoirs were magnetically stirred glass flasks held at a constant temperature (30 °C) using a hot plate in combination with a glass-covered contact thermometer in the electrolyte. Contact between the electrolyte and any metal components was avoided to exclude contamination and a possible deactivation of the silver-based catalyst. These glass flasks were used for gas separation from the electrolyte. Contaminated GDEs form predominantly H2.
The electrochemical flow reactor was operated using a potentiostat (Metrohm Autolab PGSTAT-30 with BOOSTER20A) under ambient pressure and temperature. Experiments were carried out in galvanostatic conditions, where a constant current is applied and the required voltage is monitored.
Conversion efficiency of the CO2 electrolyser
The theoretical minimal cell voltage for the electrolysis of CO2 to CO and O2 is 1.34 V when calculated from ΔG0 = −258.5 kJ mol–1 of the reaction: CO2(g) → CO(g) + ½O2(g) or from the redox potential difference E0′ = +0.82 V of the O2/2H2O couple and E0′ = −0.52 V of the CO2/CO couple56. Thus, at 3.65 V, the operation voltage of the CO2 electrolyser when coupled to the PV module, electrolysis theoretically proceeds at an over potential of 2.3 V with an energy conversion efficiency of 37%. However, it is better to consider ΔH0′ = −283.2 kJ mol–1 of the reaction, which corresponds to an adiabatic minimal cell potential of 1.47 V. Otherwise an external heat source is needed to compensate an entropy increase. This limits the efficiency of the setup to 40% when 3.65 V is applied.
In the CO2 electrolyser the distance between cathode and anode was 2 mm, which—depending on the electric current density—led to a significant ohmic loss. Ohmic losses were less at 60 °C than at 30 °C due to the higher conductivity of the electrolyte (Fig. 2b and Supplementary Fig. 2b).
Medium A for C. autoethanogenum contained per litre composition: 1 g NH4Cl, 0.1 g KCl, 0.2 g MgSO4·7H2O, 0.8 g NaCl, 0.1 g KH2PO4, 20 mg CaCl2·2H2O, 0.4 g l-cysteine-HCl, 0.4 g Na2S·9H2O, 20 mg nitrilotriacetic acid, 10 mg MnSO4·H2O, 8 mg (NH4)2Fe(SO4)2·6 H2O, 2 mg CoCl2·6 H2O, 2 mg ZnSO4·7H2O, 0.2 mg CuCl2·2H2O, 0.2 mg Na2MoO4·2H2O, 0.2 mg NiCl2·6H2O, 0.2 mg Na2SeO4, 0.2 mg Na2WO4·2H2O, 20 µg biotin, 20 µg folic acid, 100 µg pyridoxine-HCl, 50 µg thiamine-HCl · H2O, 50 µg riboflavin, 50 µg nicotinic acid, 50 µg Ca-pantothenoic acid, 1 µg vitamine B12, 50 µg p-aminobenzoic acid, 50 µg lipoic acid and 1 g yeast extract. The pH was adjusted to the pH indicated under continuous gassing.
Medium B for C. autoethanogenum, which differed from medium A mainly in not containing yeast extract, contained per litre compostion: 0.5 g MgCl2·6H2O, 0.21 g NaCl, 0.135 g CaCl2·2H2O, 2.65 g NaH2PO4·2H2O, 0.5 g KCl, 2.5 g NH4Cl, 15 mg nitrilotriacetic acid, 30 mg MgSO4·7H2O, 5 mg MnSO4·H2O, 1 mg FeSO4·7H2O, 8 mg Fe(SO4)2(NH4)2·6H2O, 2 mg CoCl2·6H2O, 2 mg ZnSO4·7H2O, 200 µg CuCl2·2H2O, 200 µg KAl(SO4)2·12H2O, 3 mg H3BO3, 300 µg Na2MoO4·2H2O, 200 µg Na2SeO3, 200 µg NiCl2·6H2O, 200 µg Na2WO4·6H2O, 20 µg d-biotin, 20 µg folic acid, 10 µg pyridoxine-HCl, 50 µg thiamine-HCl, 50 µg riboflavin, 50 µg nicotinic acid, 50 µg Ca-pantothenate, 50 µg vitamin B12, 50 µg p-aminobenzoate, 50 µg lipoic acid, 10 mg FeCl3 with additional 500 mg l-cysteine-HCl. The pH was adjusted to 6 under continuous gassing.
Medium C for C. kluyveri contained per litre compostion: 0.25 g NH4Cl, 0.2 g MgSO4·7H2O, 0.31 g K2HPO4, 0.23 g KH2PO4, 2.5 g NaHCO3, 1 g yeast extract, 10 g K-acetate, 20 g ethanol, 0.25 g l-cysteine-HCl, 1.5 mg FeCl2·4H2O, 70 µg ZnCl2·7H2O, 100 µg MnCl2·4H2O, 6 µg boric acid, 190 µg CoCl2·6H2O, 2 µg CuCl2·6H2O, 24 µg NiCl2·6H2O, 36 µg Na2MoO4·2H2O, 3 µg Na2SeO3·5H2O, 4 µg Na2WO4·2H2O, 100 µg vitamine B12, 80 µg p-amino-benzoic acid, 20 µg biotin, 200 µg nicotinic acid, 100 µg Ca-pantothenoic acid, 300 µg pyridoxine-HCl, 200 µg thiamine-HCl · H2O. The pH was adjusted to 7.5.
Conversion of syngas to acetate and ethanol
C. autoethanogenum (DSMZ 10061, Braunschweig) was routinely grown at 37 °C under continuous gassing with 67% H2 and 33% CO2 on medium A containing additionally 20 g l–1 MES (adjusted to pH 6.0). The medium was then inoculated per litre with 100 ml of an active pre-culture grown to an OD600 of 0.5 on the same medium. At an OD600 of about 0.5, the cells were anaerobically harvested by centrifugation and the cell pellet from 100 ml culture re-suspended in 1 ml medium and the suspension then used to equally inoculate two 1 litre fermenters (run in series) each containing 500 ml of medium A and continuously gassed with syngas from the CO2 electrolyser at a rate of 16.52 sccm and a composition of 11.76% CO (4.8 mmol h–1), 6.37% H2 (2.6 mmol h–1) and 81.86% CO2 (33.4 mmol h–1).
In fermenter-1 the pH was kept constant at 5.5 and in fermenter-2 at 6.0 by the continuous addition of a base (anaerobic potassium hydroxide, 140 g l–1) with a peristaltic pump. After 50 h of growth, the OD600 in fermenter-1 increased from 0.05 to 1.3 and in fermenter-2 from 0.05 to 1.1 to then remain essentially constant for the next 45 h. In these 45 h, the flow rate of the gas coming out of the fermenters (16.03 sccm) and its composition of 6.03% CO (2.44 mmol h–1), 3.96% H2 (1.61 mmol h–1) and 90.00% CO2 (36.4 mmol h–1) remained substantially constant, indicating a constant conversion rate of syngas into acetate and ethanol (Table 2).
Conversion of syngas to butanol and hexanol
The experimental setup was identical to that described for Table 2, except for the fact that the CO2 electrolyser was powered at 5.4 V, resulting in an electric current density of 150 mA cm–2. At that time, cells of C. kluyveri (DSMZ 555, Braunschweig) were added and in medium A (see section 'Growth media') 20 g MES per litre was additionally present. The two 1-litre fermenters (run in series), each containing 500 ml medium, were gassed with 10% CO (4 mmol h–1), 60% H2 (24.2 mmol h–1) and 30% CO2 (12.2 mmol h–1) from the CO2 electrolyser at a flow rate of 16.23 sccm and then inoculated in equal parts with C. autoethanogenum cells harvested from 80 ml of an actively growing syngas culture (OD600 = 0.62). After 22 h, the fermenters were additionally inoculated in equal parts with C. kluyveri cells harvested from 60 ml of a culture (OD600 = 0.86) growing on ethanol and acetate. Six hours after the addition of the C. kluyveri cells, the rates of gas consumption and of product formation remained substantially constant for at least 24 h. In these 24 h, the flow rate of the outcoming gas (8.35 sccm) and its composition of 9.6% CO (2.0 mmol h–1), 54% H2 (11.3 mmol h–1) and 36.4% CO2 (7.6 mmol h–1) remained substantially constant.
C. kluyveri was routinely grown on ethanol and acetate at 36 °C in medium C. The pH was adjusted to 7.5 and the medium then inoculated to an OD600 of 0.1 with an actively growing pre-culture in the same medium. At OD600 = 0.86, the cells from 60 ml of culture were anaerobically harvested by centrifugation and the cell pellet re-suspended in 10 ml of anoxic medium C and equal parts of the suspension used to inoculate the two fermenters.
The product gases from the CO2 electrolyser were measured with a Thermo Scientific Trace 1310 gas chromatograph (GC) equipped with two thermal conductivity detector (TCD) channels. The GC is directly connected to the experiment via a heated transfer line. For separation of CO2, O2, CO and hydrocarbons, a micropacked GC column (shin carbon) was used with He as carrier. The determination of H2 was performed on a packed mol-sieve column using Ar as carrier. In the product gases from the cathode only CO and H2 were detected by GC when using Ag as the catalyst. No traces of hydrocarbons (GC calibrated for CH4, C2H4, C2H2, C2H4 down to 20 ppm) were detected. The cycle time of the GC was 12 min. The GC was calibrated biweekly. In relation to the fermenters, gas samples were taken before the first fermenter inlet and after the last fermenter gas outlet.
Liquid samples were taken from the cultures via the sampling port of the fermenters and concentrations of acetate, butyrate, hexanoate, ethanol, butanol and hexanol were determined by quantitative 1H NMR spectroscopy (Bruker AVANCE III HD, 600 MHz). Prior to quantitative analysis, samples were diluted with an adequate amount of phosphate buffer containing the internal standard trimethylsilyl propionate.
Calculation of energy conversion efficiencies
The energy conversion efficiency of the syngas fermentation to ethanol was calculated from the enthalpy changes (ΔH0') associated with reactions (1) and (2) and those associated with the combustion of H2 (reaction (7)) and CO (reaction (8)).
∆G0' = −474.3 kJ mol–1; ∆H0' = −571.7 kJ mol–1
∆G0' = −514.4 kJ mol–1; ∆H0' = −565.9 kJ mol–1
Forming ethanol from 5H2, CO and CO2 (reaction (1) + (2)) is exothermic by −344.6 kJ mol–1 to be compared with the enthalpy change of –1,711.2 kJ mol–1 associated with the combustion of 5H2 and 1CO (2.5 × reaction (7) + 0.5 × reaction (8)). Thus, when forming ethanol from syngas only (344.6/ 1,711.2) × 100 = 20.14% of the enthalpy change associated with the combustion of H2 and CO is liberated as heat. Therefore, the efficiency of energy conversion in syngas fermentation to ethanol (reaction (1) + (2)) is almost 80%.
The energy conversion efficiency of the formation of butanol (reaction (9)) and hexanol (reaction (10)) from CO, H2 and CO2 was calculated from the enthalpy changes associated with reactions (7)–(10) to be near 78%. Reaction (9) = reactions (1), (2), (3) and (5); and reaction (10) = reactions (1), (2), (3), (4) and (6).
∆G0' = −283.9 kJ mol–1; ∆H0' = −747.4 kJ mol–1
∆G0' = −453.4 kJ mol–1; ∆H0' = −1152.4 kJ mol–1
Calculation of the alcohol yield per kWh
The amount of butanol and hexanol obtained per kWh by electroreduction of CO2 was calculated as follows: butanol contains 24 mol electrons and hexanol 36 mol electrons per mol. Assuming butanol and hexanol are being formed at a molar ratio of 1:1 then per mol of the two alcohols 30 mol electrons are required from electricity. One mol of electrons equals 96,485.3 coulomb. One coulomb per second = 1 A. Per mol alcohols CO and H2 are required in a molar ratio of 1:5. When the CO2 electrolyser is run at 4.7 V (Fig. 2c) and the H2O electrolyser at 2.0 V, then on average both together run at 2.45 V. One V·A = 1 W = 2.45 V·0.41 coulomb s–1. In 1 h at 1 kW 0.41 coulomb add up to 1,475,000 coulomb = 15.29 mol electrons equivalent to 0.51 mol butanol/hexanol = 44.88 g butanol/hexanol.
Calculation of costs per kWh
The costs of butanol and hexanol obtained per kWh by electro-reduction of CO2 was calculated from a price of 1 kWh = 2.5 cents (http://cleantechies.com/2016/09/20/jinkosolar-marubeni-score-lowest-ever-solar-pv-at-us%C2%A22-42kwh-in-abu-dhabi)57 and from that of butanol = 1.2€ per kg (ref. 1). The costs of CO2 separation and purification are presently around 60€ per t = 0.276 cents per mol CO2 (ref. 58) but is expected to decrease significantly through improved separation and purification systems and carbon trading in the near future59. For the synthesis of butanol and hexanol in a molar ratio of 1:1, 5 mol of CO2 is required.
All data are available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The authors thank R. K. Thauer (Max Planck Institute for Terrestrial Microbiology, Marburg) for his help in preparing the manuscript. Evonik Creavis GmbH (T.H. and M.D.), Siemens AG (R.K. and G.S.) and Covestro AG (R.W.) thank the German Federal Ministry of Education and Research (BMBF) for funding part of this work within the Kopernikus Initiative ‘Power-to-X’ under contract number P2X-03SFK2J0.
Supplementary Figs. 1–4 and Supplementary Table 1