Main

Food demand is growing globally, but food production is ultimately constrained by the energy conversion efficiency of photosynthesis. Most crop plants can convert sunlight and CO2 into plant biomass at an energy conversion efficiency of only ~1% or less1. Large tracts of land are thus required for crop cultivation to capture the requisite solar energy to provide food for humanity. Recent breeding and genetic engineering efforts to increase photosynthetic efficiency have yielded only select gains in a limited number of food crops2,3,4. Increasing the energy efficiency of food production (solar-to-biomass conversion) would allow for more food to be produced using less resources.

Artificial photosynthesis seeks to overcome the limitations of biological photosynthesis, including low efficiency of solar energy capture and poor carbon dioxide reduction, and could provide an alternative route for food production. Recent studies have demonstrated systems that convert CO2 and H2O into reduced species, such as CO, formate, methanol and H2, through electrolysis processes. CO2, CO and H2 can be upgraded to fuels and chemicals through gas-phase fermentation by select bacteria5,6,7; however, gas–liquid mass transfer limits the volumetric efficiency and results in uneconomic fermentation systems. The use of formate or methanol as a carbon source for fermentation is limited because formaldehyde, a toxic intermediate, is formed during biological metabolism of these substrates8,9,10. To date, electrochemically derived substrates cannot support the growth of most food-producing organisms11. However, acetate is a soluble, two-carbon substrate that can be electrochemically produced12 and is more readily metabolized by a broad range of organisms. The use of acetate produced from CO2 electrolysis to cultivate food-producing organisms could allow food production independent of biological photosynthesis but has not yet been demonstrated.

Here we describe the development of a hybrid inorganic–biological system for food production. A two-step electrochemical process converts CO2 to acetate, which serves as a carbon and energy source for algae, yeast, mushroom-producing fungus, lettuce, rice, cowpea, green pea, canola, tomato, pepper, tobacco and Arabidopsis (A. thaliana) (Fig. 1). Coupling this system of carbon fixation to photovoltaics offers an alternative, more energy-efficient approach to food production.

Results

Acetate production from CO2 electrolysis

To provide a carbon and energy source independent of biological photosynthesis that can sustain the growth of food-producing organisms, we developed an electrocatalytic process to produce acetate (as either sodium or potassium acetate depending on the electrolyte salt) from CO2. Acetate produced directly from electrochemical CO2 reduction using a copper catalyst has a less than 15% carbon selectivity, which is defined as the amount of carbon in the end product(s) divided by the total amount of carbon reduced in the system13,14,15,16. However, recent studies on CO reduction have demonstrated that acetate can be produced at industrially relevant reaction rates with a carbon selectivity of greater than 50% using a nanostructured copper catalyst12,17,18. To achieve maximum selectivity and production of acetate from a direct CO2 feed, a two-step electrolyser system was demonstrated to convert CO2 to CO and then CO to acetate through a tandem process (Fig. 2a and Extended Data Fig. 1a). More specifically, CO2 is fed to the cathode of the first electrolyser, which utilizes a commercial silver catalyst supported on a gas diffusion layer (that is, a carbon paper) and produces a gaseous product stream containing CO, H2 and trace CO2. The gas diffusion electrode improves the gaseous CO2 transport to the electrocatalyst, achieving higher currents towards CO2 reduction products compared with dissolved CO2 in a typical batch reactor16,19,20,21. A solution of 1 M KHCO3 in deionized water was used as the electrolyte on the anode side (that is, anolyte) for the CO2 electrolyser and recirculated through the anode compartment to maintain the ionic conductivity. The presence of an aqueous electrolyte has been shown to reduce the total cell potential for anion exchange membrane-based CO2 electrolysers13,22. An IrO2 anode was utilized for CO2 reduction, due to its stability in neutral pH. The gas product stream was then fed to the cathode chamber of the CO electrolyser, which contains a commercial copper catalyst for CO reduction, 1 M KOH as the anolyte and a NiFeOx anode. This design is similar to a tandem system previously reported by Romero Cuellar et al.23. By maximizing the conversion of the first and second electrolysers, as well as specifically targeting acetate over other multi-carbon products, this system was able to achieve a single-pass conversion of CO2 to acetate of 25%, a large improvement over the <1% conversion previously reported (Supplementary Table 1). Effluents, the liquid products of electrolysis containing acetate and other by-products (Extended Data Fig. 1c,d), were evaluated as carbon and energy sources for the cultivation of food-producing organisms. Early experiments found that effluents with an acetate-to-electrolyte ratio below 0.4 did not support the growth of algae (Supplementary Note). Maximizing the ratio of acetate to electrolyte was therefore crucial for integrating these carbon products with biological food production.

Operating parameters of the tandem CO2 electrolysis system were identified that maximized the conversion of CO2 feed to acetate. For the production of CO, the operating current density of the CO2 electrolyser was 100 mA cm−2 at an inlet flow rate of 7 ml min−1 CO2, which maximized CO2 conversion to CO at 43% and maintained a high level of performance (Extended Data Fig. 1e–h). For the CO electrolyser, the operating current density of 150 mA cm−2 was selected to achieve greater than 80% conversion of electrochemically produced CO into C2+ products over the entire duration of electrolysis (Extended Data Fig. 1g,h). Initial experiments found that trace amounts of CO2 negatively affected acetate selectivity in the CO electrolyser and led to a rapid increase in CO electrolyser voltage (Fig. 2b,c). When a 5 M NaOH scrubber was introduced between the two electrolysers, acetate selectivity was increased by over a factor of 3, because the scrubber prevented the unreacted CO2 from the first reactor from reaching the CO electrolyser (Fig. 2d–f and Extended Data Fig. 1b). Overall, 57% of reacted CO2 formed acetate at a production rate of 0.7 g d−1 cm−2, representing the highest conversion of CO2 feed to acetate reported to date (Fig. 2e and Supplementary Table 1).

The tandem CO2 electrolysis system operated stably while producing effluent with high acetate-to-electrolyte ratios. The electrolyser voltage for the CO2 electrolyser remained constant near 2.95 V (Fig. 2f) with less than a 60 mV increase in voltage over the six-hour experiment. The CO reactor operated at 2.22 V with a voltage increase of 160 mV over the course of the experiment (Fig. 2f), which is attributed to the pH shift in the electrolyte from 13.7 to 13.4 as acetate accumulated. Both catholyte (through the use of a cold-trap) and anolyte were collected and analysed for product quantification. As >99% of the produced acetate was collected in the anolyte, this was the primary effluent used for food production. The final effluent produced contained 0.75 M acetate and an acetate-to-electrolyte salt ratio of 0.75, representing the highest recorded acetate-to-electrolyte product stream to date (Supplementary Table 1).

Using CO2 electrolysis to grow food heterotrophically

We used the photosynthetic alga Chlamydomonas (C. reinhardtii), which can grow heterotrophically on acetate in the dark, as a model to determine whether the improved effluent produced by the electrolysers could support the growth of food-producing organisms. Chlamydomonas is added to processed foods for protein fortification and has been shown to have a positive effect on human gastrointestinal health24,25,26,27,28. Algae can produce large amounts of starch, protein and oil. Chlamydomonas grown heterotrophically can produce over 1 g of starch, 1 g of protein and 0.7 g of lipid per litre per day28,29. To evaluate effluent as a carbon source, Chlamydomonas was grown heterotrophically in the dark with effluent diluted to match the acetate concentration of a typical liquid heterotrophic growth medium (17.5 mM acetate in Tris–acetate–phosphate (TAP) medium) and adjusted to pH 7.2. Effluent-derived acetate served as the sole carbon and energy source. Effluents from different electrolysis experiments used throughout this manuscript are noted by the acetate-to-electrolyte concentration ratio (acetate concentration: electrolyte concentration), with more details provided in Supplementary Table 2. All effluents evaluated with an acetate-to-electrolyte ratio of greater than 0.4 M acetate: 1 M electrolyte enabled the growth of Chlamydomonas in the dark (Supplementary Note, Fig. 3a–c, Extended Data Figs. 2 and 3a–e, and Supplementary Table 3).

Chlamydomonas grown on an effluent with an improved acetate-to-electrolyte salt ratio (0.75 M acetate: 1 M KOH) had a yield of 0.28 g algae per g acetate (Fig. 3b,c), which is comparable to yields reported from non-effluent medias29. Chlamydomonas utilized over 99% of acetate in the media (Extended Data Fig. 3f). No products of photosynthesis (such as carbohydrates) or ancient photosynthesis (such as petroleum-derived carbon sources) were required for growth. Hence, our cultivation of a photosynthetic organism using carbon fixed through electrolysis is fully decoupled from biological photosynthesis.

The nutritional yeast Saccharomyces cerevisiae is used as a food source in single-cell protein spreads and in the production of breads and fermented beverages30,31. Yeast is a heterotrophic organism most commonly grown with glucose derived from photosynthesis (that is, starch-derived) as a carbon and energy source. To cultivate yeast without inputs derived from photosynthesis, we replaced the primary carbon source, glucose, in the yeast–peptone–dextrose (YPD) media with effluent from electrolysis. All media were adjusted to pH 6.0. Electrolyser-produced effluent (0.75 M acetate: 1 M KOH) supported the growth of yeast, enabling a yield of 0.19 g yeast biomass per g acetate as well as a ninefold increase in OD600 and a twofold increase in dry weight compared with yeast grown without acetate or glucose (Fig. 3d,e). After successfully decoupling the production of yeast from photosynthesis-derived carbon and energy, we sought to achieve the same results in mushroom-producing fungus. Mushrooms are widely consumed as a food, and fungal mycelium has emerged as a high-protein meat analogue32. These fungi are typically cultivated on solid substrates composed of photosynthesis-derived carbohydrates, such as cellulose or rice flour. To cultivate fungal mycelium without these carbohydrates, we developed a solid-state fermentation approach that used effluent as the primary carbon and energy source. Due to the amounts needed, simulated effluents were used here, made up of the same components as electrolyser-produced effluents. Simulated effluent (0.691 M acetate: 1 M KOH) was added to YPD media without glucose to reach 0.5% (w/w) acetate; this substrate supported the growth of pearl oyster, blue oyster, elm oyster, coral tooth and enokitake mycelium (Fig. 3f and Extended Data Fig. 3g–i). The majority of the substrate was fully colonized by all five species, indicating that effluent can serve as a carbon source for the cultivation of a variety of mushroom-producing fungus mycelia. Higher levels of effluent inhibited mycelium growth (Fig. 3g and Extended Data Fig. 3g), probably due to the higher levels of effluent by-products, specifically propionate, which is used as an antifungal food preservative33. The successful growth of mycelium indicates the ability to produce mycelium-based foods and potentially mushrooms independent of carbon and energy derived from photosynthesis.

Acetate can be metabolized by crop plants

To further evaluate whether carbon fixed through CO2 electrolysis could be used to produce plant-based food, we examined the potential of acetate to serve as a carbon and energy source for crops. We tracked acetate incorporation into plant biomass using heavy-isotope 13C-acetate labelled at both carbon atoms to evaluate whether exogenous acetate can be metabolized by crops. Previous acetate incorporation studies have primarily focused on metabolites involved in lipid biosynthesis in only a few plant species. In these studies, acetate was used in low concentrations as a tracer, not as a carbon source (Supplementary Table 4)34,35,36. To investigate acetate utilization as a carbon source, we grew undifferentiated lettuce tissue (callus) (to avoid the carbon and energy stored in seeds) in the dark in liquid media containing effluent (0.691 M acetate: 1 M KOH) and added in 13C-acetate to enable tracking of carbon incorporation. Calli showed extensive 13C-labelling in multiple pathways, confirming that lettuce tissue metabolizes acetate as a carbon and energy source (Fig. 4a). Labelling of citrate/isocitrate, succinate, α-ketoglutarate and malate indicated that exogenous acetate was converted to biologically active acetyl-CoA and entered the tricarboxylic acid cycle (TCA) cycle to produce energy (GTP, NADH and FADH2) (Fig. 4a). Amino acids were also labelled with 13C, indicating that carbon from acetate can be used to build proteins. Products and intermediates of gluconeogenesis were also labelled, indicating that carbon from acetate can be used for carbohydrate biosynthesis (Fig. 4a and Extended Data Fig. 4). Carbon incorporation of 13C-acetate into amino acids and sugars through the TCA cycle, glycolysis and gluconeogenesis provides strong evidence that exogenous acetate can be readily incorporated into the bulk biomass of lettuce tissue grown in dark conditions and could be a carbon and energy source for other crops.

To test acetate uptake and utilization by whole plants, we grew lettuce plants in the light with 13C-acetate added to the plant growth agar medium at 2 mM. In vegetative leaf tissue, we observed that amino acids, sugars and intermediates from the TCA cycle, glycolysis and gluconeogenesis were labelled with 13C at multiple carbon positions (Fig. 4a–c). Finding labelled metabolites in leaf tissue indicates that carbon from acetate assimilated by the roots can be distributed throughout the plant. In addition to lettuce, we discovered that acetate can be incorporated into a broad variety of crops. Rice, green pea, jalapeño pepper, canola, tomato, cowpea, tobacco and Arabidopsis seedlings grown in the light on solid agar containing 13C-acetate all showed similar 13C-labelling of amino acids, carbohydrates and TCA cycle intermediates (Fig. 4a and Extended Data Fig. 5) as was observed in lettuce and lettuce callus. This metabolite labelling pattern suggests a conserved mechanism of acetate utilization that exists across diverse plant species.

Lettuce seeds germinated normally on all concentrations of acetate evaluated (up to 10 mM) (Extended Data Fig. 6a,b). However, plant growth was largely inhibited by acetate at concentrations that would have measurably increased plant biomass, although some growth parameters such as roots showed increased growth (Extended Data Figs. 6 and 7). Lettuce plants grown with electrolyser-produced effluent (0.691 M acetate: 1 M KOH or 0.648 M acetate: 1 M KHCO3) added to reach a final media concentration of 1.0 mM acetate did not show additional growth inhibition in plant weight or leaf number from secondary electrolysis products (Extended Data Fig. 7b,d). Plant tolerance and consumption of acetate as a heterotrophic energy source will need to be increased to fully decouple plants from biological photosynthesis. Together, these analyses explore acetate as a heterotrophic carbon source for crops and illustrate that many crop species can process acetate into a usable biological form and incorporate it into major energy-production cycles through TCA and biomass in the form of sugars and amino acids.

Discussion

We demonstrated a hybrid inorganic–biological system that can produce food from carbon dioxide and electricity, independent of biological photosynthesis. CO2 electrolysis for acetate production was significantly improved for the purpose of biological integration. Using a two-step process, we achieved electrochemical reduction of CO2 to acetate with a 57% carbon selectivity towards acetate, the highest published value to date. The electrolysis system was further engineered to produce improved effluents with acetate-to-electrolyte salt ratios as high as 0.75, well above the ratio determined necessary to support biological growth. Electrochemically derived acetate was incorporated into a diverse variety of organisms grown for food, including algae, fungi and crop plants. This includes the cultivation of a photosynthetic organism, Chlamydomonas, utilizing carbon fixed through electrolysis, independent of biological photosynthesis. 13C-labelling experiments showed that a broad range of crops can utilize exogenous acetate for energy and biomass production, suggesting that acetate has the capacity to support crop growth with further optimization.

By powering electrolysis with photovoltaics, the conversion of sunlight and CO2 to food in our system (photovoltaics to electrolysis to acetate to yeast) is almost 18 times more solar-to-biomass energy-conversion-efficient than typical food production, which relies on biological photosynthesis (photosynthesis to glucose to yeast) (Extended Data Fig. 8). For algae production, our process (photovoltaics to electrolysis to acetate to algae) is almost four times more solar-to-biomass energy-conversion-efficient than biological photosynthesis of crop plants (photosynthesis to crop plants) and is equivalent to or higher than the instantaneous energy efficiency of outdoor algae production37 (Fig. 5 and Extended Data Fig. 8). There is potential for continued improvement of the system through advances in photovoltaics, electrolysis and acetate utilization in food-producing organisms. Our approach to food production is ideally suited for applications where high energy efficiencies and low physical space usage are desired, such as on space flight missions or in controlled environments on Earth. Widespread adoption of this approach in conjunction with readily available solar energy could allow for the production of more food or animal feed for a given solar footprint, which will help meet the rising demand for food without the expansion of agricultural lands.

Methods

Electrolysis methods

Cathode preparations

For the CO2 reduction electrodes, commercial silver (Ag) catalysts (nanopowder with <100 nm particle size, 99.5%) were purchased from Sigma Aldrich. Catalyst ink was prepared by suspending 100 mg of Ag nanopowder in 20 ml of 2:1 (v:v) isopropanol/deionized water. The suspension was sonicated for five minutes to allow complete dispersion of Ag particles. Afterwards, Sustanion XA-9 ionomer (5% (w/w) in ethanol, Dioxide Materials) was added until the ionomer reached 10% (w/w) in solution (metal basis). The resulting mixture was further sonicated for an additional 30 minutes to allow complete dispersion. To prepare the electrodes, a 25 cm2 piece of Sigracet 39BB gas carbon paper (Fuel Cell Store), used as a gas diffusion layer, was heated on a hot plate to 100 °C. Catalyst ink was airbrushed onto the carbon paper until a loading of 1.4 mg cm−2 was reached. The electrode was dried at 70 °C overnight to ensure the complete evaporation of solvents.

For the CO reduction electrodes, commercial copper (Cu) catalyst (nanopowder with 25 nm particle size (TEM)) was purchased from Sigma Aldrich. Catalyst ink was prepared by suspending 100 mg of Cu nanopowder in 20 ml of 2:1 (v:v) isopropanol/deionized water. The suspension was sonicated for five minutes to allow complete dispersion of Cu particles. Afterwards, Sustanion XA-9 ionomer (5% (w/w) in ethanol, Dioxide Materials) was added until the ionomer reached 20% (w/w) in solution (metal basis). The resulting mixture was further sonicated for an additional 30 minutes to allow complete dispersion. To prepare the electrodes, a 5 cm2 piece of Sigracet 39BB carbon paper (Fuel Cell Store) was heated on a hot plate to 100 °C. Catalyst ink was drop-casted onto the carbon paper until a loading of 2.0 mg cm−2 was reached. The electrode was dried at 70 °C overnight to ensure the complete evaporation of solvents.

Anode preparations

For experiments using 1 M KHCO3, IrO2 catalysts were prepared via a previously reported method38. In a typical preparation, titanium felt (Fuel Cell Store) was degreased using acetone and etched in boiling 0.5 M oxalic acid (98%, Sigma Aldrich). The titanium felt was then dip-coated in a solution of 10 ml of isopropanol, 10 vol% concentrated HCl (ACS reagent, Sigma Aldrich), containing 30 mg dissolved IrCl3·xH2O (99.8%, Alfa Aesar). This was followed by drying at 100 °C for ten minutes and calcination at 500 °C for ten minutes. This procedure was repeated until a loading of 3 mg cm−2 was achieved.

For experiments conducted in alkaline conditions, NiFeOx anodes prepared by a previously reported method39 were used. In a typical preparation, Ni foam (>99.99%, MTI Corporation) was sonicated in a 5 M HCl solution for 30 minutes to remove the NiOx layer on the surface. The foam was then rinsed using deionized water and ethanol and dried in air. Electrodeposition was carried out in a standard three-electrode electrochemical reactor containing the nickel foam as the working electrode, a platinum wire counter electrode and a Ag/AgCl (Pine Research) reference electrode. The electrolyte bath contained 3 mM Ni(NO3)2·6H2O and 3 mM Fe(NO3)3·9H2O. A constant potential of −1.0 V versus Ag/AgCl was applied for 300 seconds. The electrode was then removed and rinsed with ethanol and deionized water, followed by drying overnight at 70 °C to fully remove solvents. A fresh cathode was used for each experiment, whereas the anodes were reused multiple times.

Flow electrolyser

Both the CO2 and CO electrolysers were constructed as 5 cm2 membrane electrode assemblies with serpentine flow patterns. The cathode end plates and the CO2 electrolyser anode end plate were gold-plated stainless steel. The CO electrolyser anode end plate was not gold coated, as stainless steel is stable in alkaline conditions. The CO2 gas flow rate into the electrolysers was controlled via a Brooks GF040 and held at 7 ml min−1 for the two-step experiments. The back pressure was controlled using a back-pressure controller (Cole-Palmer). The anolyte flow rates for both electrolysers were controlled via a peristaltic pump, with flow rates ranging between 0.5 and 1 ml min−1. A cold trap chilled using ice was placed between the gas outlet stream of the CO electrolyser and the back-pressure controller to capture any vaporized liquid products that might exit through the gas stream.

For the CO2 electrolyser, the Ag cathode and the IrO2 anode were pressed into a Sustanion 37-50 Anion Exchange Membrane (Dioxide Materials) at 20 lb-in torque. Laser-cut Teflon (McMaster Carr) was used as a gasket to ensure that the electrolyser was gas-tight. 1 M KHCO3 was used as the anolyte for all CO2 reduction experiments. A constant current was applied using a power source (National Instruments). For the CO electrolyser, the Cu cathode and the IrO2 or NiFeOx anode were pressed against an FAA-3-50 anion exchange membrane (Fumatech) at 20 lb-in torque. 1 M KHCO3 or 2 M KOH was recirculated through the anode chamber. Liquid products were accumulated in the anolyte of the CO electrolyser until a target concentration was reached. The quantification of liquid products was conducted at the end of the experiment. For the CO electrolyser, the applied current was controlled using an Autolab PG128N potentiostat (Metrohm).

Gas products were quantified using a Multiple Gas Analyzer No. 5 gas chromatography system (SRI Instruments) equipped with a HaySep D and Molsieve 5 A column connected to a thermal conductivity detector. All gas products were quantified using the thermal conductivity detector data. Liquid products were quantified using a Bruker AVIII 600 MHz NMR spectrometer. Typically, 20 to 100 μl of collected electrolyte exiting the reactor was diluted to 500 μl, and then 100 μl D2O containing 20 ppm (m/m) dimethyl sulfoxide (≥99.9%, Alfa Aesar) was added. One-dimensional 1H spectrum was measured with water suppression using a pre-saturation method.

Effluents containing all liquid products were used for the growth of food-producing organisms. Electrolyser-produced effluents were used when possible, but in some cases simulated effluents of the same makeup were used.

Algae methods

Chlamydomonas reinhardtii (strain 21gr+, CC-1690 from the Chlamydomonas Resource Center) was cultivated on TAP media40 where the source of acetate was a commercial supplier (Sigma 64-19-7), simulated effluent, electrolyser-produced effluent or none at all. When effluent (electrolyser produced or simulated) was the acetate source, it was added until the desired final acetate concentration was reached (17.5 mM). Ethanol (Koptec 64-17-5), propionic acid (Sigma 79-09-4) and n-propanol (Sigma 71-23-8) were added as indicated in Supplementary Table 3. All media were adjusted to pH 7.2 with 5 M HCl. Effluents and media concentrations are shown in Supplementary Tables 2 and 3.

Chlamydomonas was grown in 50 ml of media in 250 ml flasks in the dark. The flasks were stationary at 22 °C except for the experiment using 0.75 M acetate: 1 M KOH (data in Fig. 3b,c), where the flasks were at 30 °C with a shaking speed of 150 r.p.m.29. Aliquots were taken in a darkened biosafety cabinet and used for OD measurements at wavelength 750 nm (QuickDrop Spectrophotometer, Molecular Devices), chlorophyll extraction and quantification41, and cell counts using a haemocytometer or the Bio-Rad TC20 Automated Cell Counter. For dry cell weights, the entire culture was centrifuged, washed two times with deionized water to remove residual salts, dried overnight at 100 °C and then weighed. Images were taken with a Nikon 7500 DSLR camera.

Yeast and mushroom methods

Saccharomyces cerevisiae was cultured in a base medium of yeast extract (10 g l−1) and peptone (20 g l−1) with glucose, effluent or no additional primary carbon source42. Standardized comparisons of acetate and glucose were made on the basis of energy content, and a variety of glucose and effluent concentrations were tested, as shown in Supplementary Table 5. The highest concentration is based on 2% sodium acetate42. Media with a carbon source with the energetic equivalent of 53.36 kJ l−1 (0.061 M glucose or 0.019 M acetate) were determined to be the most efficient and thus were used in the yeast experiments shown in Fig. 3d,e. All media were adjusted to pH 6.0 with 5 M HCl. Cells were grown in 5 ml of media in culture tubes at 30 °C and 251 r.p.m. Growth was monitored by measuring OD at 600 nm and dry cell weight at 96 hours. For dry weights, the pellet was washed with deionized water to remove residual salts, dried overnight at 100 °C and then weighed. Effluents and media concentrations are shown in Supplementary Tables 2 and 3.

For the cultivation experiments with mushroom-producing fungi, the following five species were used: pearl oyster (Pleurotus ostreatus), elm oyster (Hypsizygus ulmarius), blue oyster (Pleurotus ostreatus var. columbinus), enokitake (Flammulina velutipes) and coral tooth (Hericium coralloides). The strains were purchased from liquidfungi.com and maintained on a liquid medium of glucose, yeast extract and peptone. Fungal mycelia were grown in a solid-state fermentation approach that roughly followed the PF-Tek methodology43; however, the carbon sources typically added, such as starch (rice flour), peat moss and coconut fibre, were omitted. A solid substrate of fine-grade vermiculite (30 g) was mixed with gypsum (0.5 g) and added to a 10 oz wide-mouth mason jar (Kamota). Liquid growth media (90 ml) were added to each jar, which soaked into the vermiculite. Media composition was the same as the maintenance media except that the carbon source was glucose (20 g l−1), acetate (as indicated), effluent (as indicated), no carbon source or a combination of these. Media pH was adjusted to 6.0 with HCl. The jars were closed with lids with four drilled holes (12.7 mm in diameter) covered with a synthetic filter disk (0.3 µm pore size) to allow gas exchange. The jars were autoclaved for 45 minutes at 121 °C. Liquid mycelium cultures were centrifuged, washed with sterilized media, resuspended and then used to inoculate the sterilized jars.

Plant methods

Plant material, media and growth conditions

The following nine plants were used: lettuce (Lactuca sativa L. cv. ‘Black Seeded Simpson’), rice (Oryza sativa ssp. japonica cv. ‘Kitaake’), green pea (Pisum sativum), tomato (Solanum lycopersicum cv. ‘Micro-Tom’), jalapeño pepper (Capsicum annuum cv. ‘Jalapeño’), canola (Brassica napus), cowpea (Vigna unguiculata L. cv. CB46), thale cress (A. thaliana var. ‘Columbia’) and tobacco (Nicotiana tabacum cv. Xanthi).

Plants were cultivated on 50 ml of basal medium of 1/2 Murashige and Skoog salts44 (Caisson Labs) and Gamborg’s vitamins45 (Phytotech Labs), unless otherwise noted. Solid media were made with 0.7% in vitro growth-grade agarose (Caisson Labs). The following were added as indicated: sucrose (2%), acetic acid, 13C2-acetate (Sigma Aldrich) or electrolyser-produced effluent. All media were adjusted to pH 5.8 using 1 M KOH or HCl. The plants were grown in 16:8-hour light–dark cycles under fluorescent lighting (100 µmol m−2 s−1) at 22 °C, unless otherwise noted.

Acetate feeding experiment

Lettuce was germinated in soil for 11 days. The plants were clipped at the base of the stem and were further trimmed so that all lettuce stems were the same length for the start of the experiment. Each plant was transferred to an 8 ml glass vial containing deionized water with acetate dissolved at various concentrations (nine plantlets per treatment: 0, 0.1, 0.3, 0.6, 1, 3, 6 and 10 mM acetate). After 29 days, the plantlets were imaged (using a Nikon 7500 DSLR with an AF-S VR Micro-Nikkor 105 mm f/2.8 G IF-ED lens) and removed from the solution for measurement of leaf number, stem length, root length and fresh weight. Stem length was measured from the first lateral root to the tip of the stem, and roots were measured on the basis of the longest root on the plantlet.

Experimental setup for plant germination and 13C exposure

For the lettuce germination experiments, seeds were sterilized using a 15% bleach solution and a drop of Tween for every 50 ml of sterilizing solution. The seeds were incubated in the sterilizing solution for 15 minutes while shaking and then washed five times with sterile water, five minutes for each wash while shaking. Ten lettuce seeds were placed on agar basal media (25 ml) supplemented with acetate (0.1, 0.3, 0.6, 1, 2, 3, 6 or 10 mM), with 2 mM labelled 13C2-acetate or with no additions to the base medium. They were allowed to germinate and grow for 28 days and were then imaged and removed to measure stem length, leaf number and fresh weight (not including the roots). Tissue samples for the 13C-labelling experiments and the untreated controls were frozen in liquid nitrogen and stored at −80 °C until they were used for metabolomic analysis. For the other plant species, the same sterilization procedure was used for the seeds, and the plants were grown until there was at least a few hundred milligrams of tissue for metabolomic analysis (canola, 14 days; rice, 14 days; Arabidopsis, 18 days; Micro-Tom tomato, 22 days; green pea, 22 days; Nicotiana benthamiana, 32 days; cowpea, 32 days).

Lettuce callus in liquid culture

Undifferentiated lettuce callus was generated by plating pieces of lettuce leaf tissue on callus-inducing media (1/2 Murashige and Skoog salts, 0.05 mg l−1 α-naphthaleneacetic acid, 0.4 mg l−1 6-benzyl aminopurine and Gamborg’s vitamins45) (Phytotech Labs). The calli were cut to the desired size and weighed to ensure that an equal amount of tissue was added to each flask. Each flask of calli was incubated in liquid 1/2 Murashige and Skoog media supplemented with acetate, 13C2-acetate or sucrose (2%) as indicated and to the specified concentrations. All cultures were grown at 22 °C in the dark at 100 r.p.m. Tissue samples for the 13C-labelling experiment and control samples were frozen in liquid nitrogen and stored at −80 °C until they were used for metabolomic analysis.

Metabolomic analysis

Sample preparation

Plant tissue was freeze-dried, and then approximately 10 mg was weighed into a 2 ml tube and homogenized using a bead mill, using three 2.8 mm beads per tube. To each sample, 750 µl of 1:2 water:methanol was added, and the samples were then vortexed for 60 min at 4 °C. Then, 500 µl of chloroform was added, and the samples were vortexed at 4 °C for an additional 15 min. After centrifugation for 10 min (16,000 g at 4 °C), the top, polar layer was transferred to a glass vial and analysed by liquid chromatography–mass spectrometry (LC–MS).

LC–MS

LC–MS metabolomics analysis was performed at the University of California, Riverside Metabolomics Core Facility. The analysis was performed on a Synapt G2-Si quadrupole time-of-flight mass spectrometer (Waters). Metabolite separations were carried out on an I-class UPLC system (Waters) using a ZIC-pHILIC column (2.1 mm × 150 mm, 5 µM) (EMD Millipore). The two mobile phases used were (A) water with 15 mM ammonium bicarbonate adjusted to pH 9.6 with ammonium hydroxide and (B) acetonitrile. The column was held at 20 °C, and the flow rate was 150 µl min−1. The sample injection volume was 2 µl. The following gradient was performed: 0 min, 10% A, 90% B; 1.5 min, 10% A, 90% B; 16 min, 80% A, 20% B; 29 min, 80% A, 20% B; 31 min, 10% A, 90% B; 32 min, 10% A, 90% B.

The mass spectrometer was operated in negative ion mode (50 to 1200 m/z) with a 100 ms scan time. Tandem MS was acquired in a data-dependent fashion. The source temperature was 150 °C, and the desolvation temperature was 600 °C. Nitrogen was used as a desolvation gas (1,100 l h−1) and cone gas (150 l h−1). The collision gas used was argon. The capillary voltage was 2 kV. Leucine enkephalin was infused and used for mass correction.

Data processing and analysis

Data processing was performed with the open-source Skyline software46. Metabolites were identified by MS (less than 5 ppm) and tandem MS using the Metlin database47. Data for isocitrate and citrate and for 2-phosphoglycerate and 3-phosphoglycerate are included as cumulative values (iso/citrate and 2/3-phosphoglycerate) because they are not distinguishable through the LC–MS methodology used. The log2 13C enrichment was calculated for the heat maps using the equation:

$${\log}_2\left( {\frac{{{\mathrm{treatment}}\left( {\frac{{M + X}}{M}} \right)}}{{{\mathrm{control}}\left( {\frac{{M + X}}{M}} \right)}}} \right)$$

where M is the area under the curve measured by LC–MS of molecules made up of 12C atoms only, and M + X is the area under the curve measured by LC–MS of molecules with 13C-isotope atoms incorporated into the molecule, X being the number of 13C-isotope atoms incorporated. When multiple biological replicates were available, they were averaged before dividing treatment by control. The untreated control replicates shown are a single representative replicate normalized to the average of all replicates, which helps visualize any variation between controls.

Energy efficiency calculations

Efficiency calculations for electrocatalysis

The following equations were used to determine the efficiency of the electrolysis process. Faradaic efficiency (FE) from gas chromatography was calculated as

$${{{\mathrm{FE}}}}\left( {{{\mathrm{\% }}}} \right) = \frac{{{{n}} \times {{F}} \times {{x}} \times {{V}}}}{{{{j}}_{{{{\mathrm{Tot}}}}}}} \times 100$$

where n is the number of electrons transferred, F is Faraday’s constant, x is the mole fraction of the product, V is the total molar flow rate of the gas and jTot is the total current. Liquid Faradaic efficiency was calculated using quantitative 1H-NMR.

Carbon selectivity was calculated as

$${{{\mathrm{Carbon}}}}\;{{{\mathrm{selectivity}}}}\left( {{{\mathrm{\% }}}} \right) = \frac{{{{n}}_{{i}} \times {{F}} \times {{j}}_{{i}} \times {{C}}_{{i}}}}{{\mathop {\sum }\nolimits_{{i}} {{n}}_{{i}} \times {{F}} \times {{j}}_{{i}} \times {{C}}_{{i}}}} \times 100$$

where ni is the number of electrons transferred to product i, ji is the partial current towards product i and Ci is the number of carbons in product i. This value represents the percentage of CO2 reacted towards C2+ products found in a given product or the molar selectivity of a given product scaled to the number of carbons contained in it.

To determine the overall efficiency of the electrolysis process, we calculated the theoretical energy required to produce the C2+ products and divided that by the actual amount of energy it took to produce those C2+ products. The C2+ product distribution is shown in Supplementary Table 6, where we assume 1 g of the C2+ products.

Theoretical potentials were calculated using the following equation:

$${{E}}^0 = \frac{{\Delta {{G}}^0}}{{{{F}} \times {{n}}}}$$

where E0 is the theoretical potential, n is the number of electrons and ΔG0 is the Gibbs free energy of reaction.

Theoretical energy was calculated as follows:

$$\begin{array}{l}{{{\mathrm{Theoretical}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{C}}}}_{2 + }\;{{{\mathrm{products}}}} = \mathop {\sum }\limits_{{i}} {{E}}_{{i}}^0 \times {{F}} \times {{n}}_{{i}} \times \frac{{{{m}}_{{i}}}}{{{{{\mathrm{MW}}}}_{{i}}}}\\ = 21.86\;{{{\mathrm{kJ}}}}\end{array}$$

where $$E_i^0$$ is the theoretical cell potential for species i, mi is the mass produced of product i and MWi is the molecular weight of product i. The values for E0 and m can be found in Supplementary Table 6.

The actual energy input to produce 1 g of C2+ products is calculated using the following equation:

$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{C}}}}_{2 + }\;{{{\mathrm{products}}}}\\ = {{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}\;{{{\mathrm{electrolyser}}}}\\\quad + {{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}_2\;{{{\mathrm{electrolyser}}}}\\ = 31.22\;{{{\mathrm{kJ}}}} + 30.14\;{{{\mathrm{kJ}}}} = 61.36\;{{{\mathrm{kJ}}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}\;{{{\mathrm{electrolyser}}}} = \frac{{{{m}}_{{{{\mathrm{AcO}}}}} \times {{n}}_{{{{\mathrm{AcO}}}}} \times {{F}} \times {{E}}_{{{{\mathrm{CO}}}}}}}{{{{{\mathrm{MW}}}}_{{{{\mathrm{AcO}}}}} \times {{{\mathrm{FE}}}}_{{{{\mathrm{AcO}}}}}}}\\ = \frac{{0.716\;{{{\mathrm{g}}}} \times 4\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \times 2.3\;{{{\mathrm{V}}}}}}{{60.052\frac{{{{\mathrm{g}}}}}{{{{{\mathrm{mol}}}}}} \times 0.339}}\\ = 31.22\;{{{\mathrm{kJ}}}}\end{array}$$

where mAcO is the mass of acetate produced to produce a total of 1 g of C2+ products at the Faradaic efficiencies calculated and listed in Supplementary Table 6, nAcO is the number of moles of electrons passed per mole of acetate produced, ECO is the measured cell potential for the CO electrolyser, MWAcO is the molecular weight of acetate and FEAcO is the measured Faradaic efficiency of acetate. For the CO2 electrolyser:

$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}_2\;{{{\mathrm{electrolyser}}}} = \frac{{{{N}}_{{{{\mathrm{CO}}}}} \times {{n}}_{{{{\mathrm{CO}}}}} \times {{F}} \times {{E}}_{{{{\mathrm{CO}}}}_2}}}{{{{{\mathrm{FE}}}}_{{{{\mathrm{CO}}}}}}}\\ = \frac{{0.045\;{{{\mathrm{mol}}}} \times 2\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \times 2.95\;{{{\mathrm{V}}}}}}{{0.85}}\\ = 30.14\;{{{\mathrm{kJ}}}}\end{array}$$

where NCO is the moles of carbon monoxide necessary to produce a total of 1 g of C2+ products at the 92% (measured) conversion in the CO electrolyser, nCO is the number of electrons passed per mol of carbon monoxide produced, $$E_{{\mathrm{CO}}_2}$$ is the measured cell potential for the CO2 electrolyser and FECO is the measured Faradaic efficiency of carbon monoxide.

Using the values listed in Supplementary Table 6, the energetic efficiency was calculated by taking the ratio of the theoretical energy and the actual energy input per 1 g of product:

$$\begin{array}{l}{{{\mathrm{Energy}}}}\;{{{\mathrm{efficiency}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{C}}}}_{2 + }\;{{{\mathrm{product}}}}\;{{{\mathrm{production}}}}\\ = \frac{{{{{\mathrm{Theoretical}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{per}}}}\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{C}}}}_{2 + }\;{{{\mathrm{products}}}}}}{{{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{per}}}}\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{C}}}}_{2 + }\;{{{\mathrm{products}}}}}}\\ = \frac{{21.86\;{{{\mathrm{kJ}}}}}}{{61.36\;{{{\mathrm{kJ}}}}}} \times 100{{{\mathrm{\% }}}}\\ = 35.62{{{\mathrm{\% }}}}\end{array}$$

Our tandem electrolysis process has an energy efficiency of 35.62% for the production of all C2+ products, which accounts for losses due to selectivity, conversion and overpotential. If you consider ethylene as a complete loss and consider only products in the effluent, the energy efficiency of C2+ product production is 24.26%. We next calculated the energy efficiency of the production of just acetate, as this is what is used as the energy and carbon source for algal and yeast growth. To calculate acetate efficiency, we did not include losses due to selectivity because in a commercial setting, the other C2+ products (for example, ethylene) would be considered co-products of the process. The efficiency numbers including ethylene as a co-product were used to calculate the energy efficiency of the whole system. More details are shown below:

$$\begin{array}{l}{{{\mathrm{Theoretical}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}} = {{E}}_{{{{\mathrm{AcO}}}}}^0 \times {{F}} \times {{n}}_{{{{\mathrm{AcO}}}}} \times \frac{{{{m}}_{{{{\mathrm{AcO}}}}}}}{{{{{\mathrm{MW}}}}_{{{{\mathrm{AcO}}}}}}} + {{E}}_{{{{\mathrm{CO}}}}}^0\\ \times {{F}} \times {{n}}_{{{{\mathrm{CO}}}}} \times \left( {\frac{{{{{\mathrm{mol}}}}_{{{{\mathrm{CO}}}}}}}{{{{{\mathrm{mol}}}}_{{{{\mathrm{AcO}}}}}}} \times \frac{{{{m}}_{{{{\mathrm{AcO}}}}}}}{{{{{\mathrm{MW}}}}_{{{{\mathrm{AcO}}}}}}}} \right)\\ = 0.73\;{{{\mathrm{V}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \times 4\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times \frac{{1\;{{{\mathrm{g}}}}}}{{60.052\frac{{{{\mathrm{g}}}}}{{{{{\mathrm{mol}}}}}}}} + 1.333\;{{{\mathrm{V}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \\\times 2\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times \left( {2 \times \frac{{1\;{{{\mathrm{g}}}}}}{{60.052\frac{{{{\mathrm{g}}}}}{{{{{\mathrm{mol}}}}}}}}} \right)\\ = 4.70\;{{{\mathrm{kJ}}}} + 8.57\;{{{\mathrm{kJ}}}} = 13.27\;{{{\mathrm{kJ}}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}\;{{{\mathrm{electrolyser}}}} = \frac{{{{m}}_{{{{\mathrm{AcO}}}}} \times {{n}}_{{{{\mathrm{AcO}}}}} \times {{F}} \times {{E}}_{{{{\mathrm{CO}}}}}}}{{{{{\mathrm{MW}}}}_{{{{\mathrm{AcO}}}}}}}\\ = \frac{{1\;{{{\mathrm{g}}}} \times 4\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \times 2.3\;{{{\mathrm{V}}}}}}{{60.052\frac{{{{\mathrm{g}}}}}{{{{{\mathrm{mol}}}}}}}} = 14.78\;{{{\mathrm{kJ}}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}_2\;{{{\mathrm{electrolyser}}}} = {{N}}_{{{{\mathrm{CO}}}}} \times {{n}}_{{{{\mathrm{CO}}}}} \times {{F}} \times {{E}}_{{{{\mathrm{CO}}}}_2}\\ = \left( {2 \times \frac{{1\;{{{\mathrm{g}}}}\;{{{\mathrm{AcO}}}}}}{{60.052\frac{{{{\mathrm{g}}}}}{{{{{\mathrm{mol}}}}}}}}} \right) \times 2\frac{{{{{\mathrm{mol}}}}}}{{{{{\mathrm{mol}}}}}} \times 96{,}485\frac{{{{\mathrm{C}}}}}{{{{{\mathrm{mol}}}}}} \times 2.95\;{{{\mathrm{V}}}}\\ = 18.96\;{{{\mathrm{kJ}}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{acetate}}}}\left( {E_{{{{\mathrm{AcO}}}}}} \right)\\ = {{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}\;{{{\mathrm{electrolyser}}}}\\ + {{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{CO}}}}_2\;{{{\mathrm{electrolyser}}}}\\ = 14.78\;{{{\mathrm{kJ}}}} + 18.96\;{{{\mathrm{kJ}}}} = 33.71\;{{{\mathrm{kJ}}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Energy}}}}\;{{{\mathrm{efficiency}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{acetate}}}}\;{{{\mathrm{production}}}} = \frac{{{{{\mathrm{Theoretical}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{acetate}}}}}}{{{{{\mathrm{Actual}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{input}}}}\;{{{\mathrm{per}}}}\;1\;{{{\mathrm{g}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{acetate}}}}}}\\ = \frac{{13.27\;{{{\mathrm{kJ}}}}}}{{33.71\;{{{\mathrm{kJ}}}}}} \times 100{{{\mathrm{\% }}}} = 39.37{{{\mathrm{\% }}}}\end{array}$$

Efficiency calculations for food production

The efficiency values reported here are based on the statistical averages of at least three biological replicates. These calculations were conducted similarly to the approaches in Blankenship et al.1 and Nangle et al.48.

Energy efficiency of the heterotrophic cultivation of Chlamydomonas without inputs from biological photosynthesis

The energy efficiency of biological photosynthesis is defined as the energy content of the biomass that can be harvested annually divided by the annual solar irradiance over the same area1. To be able to compare our process to biological photosynthesis, we calculated the conversion efficiency of sunlight to biomass using photovoltaics to power our process of the electrolytic production of acetate followed by the heterotrophic cultivation of Chlamydomonas in the dark. We define the energy efficiency as the increase in biomass energy content divided by the required solar energy input. The increase in biomass energy content is calculated as

$$\begin{array}{l}{{{\mathrm{Increase}}}}\;{{{\mathrm{in}}}}\;{{{\mathrm{algal}}}}\;{{{\mathrm{biomass}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{content}}}} = \Delta {{X}}_{{{{\mathrm{biomass}}}}} \times \Delta {{H}}_{{{{\mathrm{biomass}}}}}^{{{\circ}}}\\ = (0.3733\;{{{\mathrm{g}}}}\;{{{\mathrm{l}}}}^{ - 1} - 0.0776\;{{{\mathrm{g}}}}\;{{{\mathrm{l}}}}^{ - 1}) \times (21.5\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1})\\ = 6.36\;{{{\mathrm{kJ}}}}\;{{{\mathrm{l}}}}^{ - 1}\end{array}$$

where ΔXbiomass is the gain of algal biomass and $$\Delta H_{{\mathrm{biomass}}}^{{\circ}}$$ is the enthalpy of combustion of algal biomass, which was determined experimentally using an oxygen bomb calorimeter. The required solar energy input is calculated as

$$\begin{array}{l}{{{\mathrm{Solar}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{required}}}} = \frac{{{{{\mathrm{Acetate}}}}\;{{{\mathrm{used}}}} \times {{{\mathrm{Energy}}}}\;{{{\mathrm{for}}}}\;{{{\mathrm{acetate}}}}\;{{{\mathrm{electrocatalysis}}}}}}{{{{{\mathrm{Annual}}}}\;{{{\mathrm{photovoltaic}}}}\;{{{\mathrm{efficiency}}}}}}\\ = \frac{{({{C}}_{{{{\mathrm{AcO}}}}} \times {{{\mathrm{MW}}}}_{{{{\mathrm{AcO}}}}} \times w_{{{{\mathrm{consumed}}}}}) \times E_{{{{\mathrm{AcO}}}}}}}{{{\eta }}_{{{{\mathrm{AM}}}}\;1.5} \times {\eta}_{{{{\mathrm{annual}}}}}}\\ = \frac{{\left( {0.01748\;{{{\mathrm{M}}}}\;{{{\mathrm{acetate}}}} \times 60.052\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}}{{{{{\mathrm{mol}}}}}} \times 99.5{{{\mathrm{\% }}}}} \right) \times 33.71\frac{{{{{\mathrm{kJ}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}}}}{{22.8{{{\mathrm{\% }}}} \times 95{{{\mathrm{\% }}}}}}\\ = 162.55\;{{{\mathrm{kJ}}}}\;{{{\mathrm{l}}}}^{ - 1}\end{array}$$

where CAcO is the acetate concentration in the algal media, MWAcO is the molecular weight of acetate, wconsumed is the fraction of acetate consumed by the algae, EAcO is the energy required to generate acetate in the electrolyser (calculated above), ηAM1.5 is the maximum power conversion efficiency (peak solar intensity, AM1.5 spectral distribution) of a commercially available silicon solar cell (from Canadian Solar49,50) and ηannual is the photovoltaic annual efficiency, which is about 95% of the maximum power conversion efficiency value due to the changing solar zenith angle throughout the day and year1. The energy efficiency is then calculated as

$$\begin{array}{l}{{{\mathrm{Energy}}}}\;{{{\mathrm{efficiency}}}}\;{{{\mathrm{of}}}}\;{{{\mathrm{algae}}}}\;{{{\mathrm{production}}}} = \frac{{{{{\mathrm{Increase}}}}\;{{{\mathrm{in}}}}\;{{{\mathrm{algal}}}}\;{{{\mathrm{biomass}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{content}}}}}}{{{{{\mathrm{Solar}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{required}}}}}}\\ = \frac{{6.36\;{{{\mathrm{kJ}}}}\;{{{\mathrm{l}}}}^{ - 1}}}{{162.55\;{{{\mathrm{kJ}}}}\;{{{\mathrm{l}}}}^{ - 1}}} = 3.91{{{\mathrm{\% }}}}\end{array}$$

Comparison of this number with the ~1% annual efficiency for most crop plants1 shows that our approach can be almost four times more energy efficient than biological photosynthesis for the cultivation of photosynthetic organisms. Improvements in photovoltaic maximum power conversion efficiency would increase the efficiency of our approach. For example, multi-junction solar cells have been shown to reach efficiencies as high as 47.1%49, which coupled to our system would bring the overall energy efficiency of sunlight to food to ~9%.

For Fig. 5 and Extended Data Fig. 8, we calculated the energy efficiencies at each step within the system as follows:

$$\begin{array}{l}{{{\mathrm{Solar}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{electricity}}}}\;{{{\mathrm{efficiency}}}}\; = {\eta}_{{{{\mathrm{AM}}}}\;1.5} \times {\eta}_{{{{\mathrm{annual}}}}}\\ = 22.8{{{\mathrm{\% }}}} \times 95{{{\mathrm{\% }}}} = 21.7{{{\mathrm{\% }}}}\end{array}$$
$$\begin{array}{l}{{{\mathrm{Acetate}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{algal}}}}\;{{{\mathrm{biomass}}}}\;{{{\mathrm{efficiency}}}} = {{Y}}_{{{{\mathrm{algae}}}}/{{{\mathrm{acetate}}}}} \times \frac{{\Delta {{H}}_{{{{\mathrm{biomass}}}}}^{{{\circ}}}}}{{\Delta {{H}}_{{{{\mathrm{AcO}}}}}^{{{\circ}}}}}\\ = 0.2831\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{yeast}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}} \times \frac{{21.5\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}}{{14.58\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}} = 41.75\% \end{array}$$

where Yalgae/acetate is the yield of algae grown with acetate effluent media, and $$\Delta H_{\mathrm{AcO}}^{\circ}$$ is the enthalpy of combustion of acetate.

Energy efficiency of the cultivation of yeast

Nutritional yeast is heterotrophic and is typically cultivated with glucose derived from biological photosynthesis as the primary carbon and energy source. In our process, electrolysis-derived acetate is substituted for glucose. To compare these two ways to produce yeast, we calculated the amount of yeast that could be produced per area of land in Illinois, a large corn-producing state in the United States. To calculate the amount of yeast that could be produced by our process independent of biological photosynthesis, we used the annual average solar irradiance in Illinois to calculate the electricity that could be generated by photovoltaics to synthesize acetate in the CO2 electrolyser and subsequently used to cultivate yeast, as follows:

$$\begin{array}{l}{{{\mathrm{Yeast}}}}\;{{{\mathrm{produced}}}}\;{{{\mathrm{by}}}}\;{{{\mathrm{artificial}}}}\;{{{\mathrm{photosynthesis}}}} = {{{\mathrm{Available}}}}\;{{{\mathrm{energy}}}}\;{{{\mathrm{from}}}}\;{{{\mathrm{photovoltaics}}}}\\ \times {{{\mathrm{Acetate}}}}\;{{{\mathrm{electrocatalysis}}}} \times {{{\mathrm{Yeast}}}}\;{{{\mathrm{yield}}}}\\ = \left( {{{E}}_{{{\mathrm{e}}}} \times {\eta}_{{{{\mathrm{AM}}}}1.5} \times {\eta}_{{{{\mathrm{annual}}}}}} \right) \times \left( {E_{{{{\mathrm{AcO}}}}}} \right)^{ - 1} \times \left( {{{Y}}_{{{{\mathrm{yeast}}}}{/}{{{\mathrm{acetate}}}}}} \right)\\ = \left( {5{,}543{,}620\;{{{\mathrm{kJ}}}}\;{{{\mathrm{m}}}}^{ - 2}\;{{{\mathrm{yr}}}}^{ - 1} \times 22.8{{{\mathrm{\% }}}} \times 95{{{\mathrm{\% }}}}} \right) \times \left( {33.71\frac{{{{{\mathrm{kJ}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}}} \right)^{ - 1}\\ \times \left( {0.1889\;\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{yeast}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}}} \right)\\ = 6.73\;{{{\mathrm{kg}}}}\;{{{\mathrm{yeast}}}}\;{\mathrm{per}}\;{{{\mathrm{m}}}}^{2}\;{\mathrm{per}}\;{{{\mathrm{yr}}}}\end{array}$$

where Ee is the average annual solar irradiance from the North America Land Data Assimilation System Daily Sunlight (kJ m−2) dataset for Illinois from 2000 to 201151; ηAM1.5 is the maximum power conversion efficiency (peak solar intensity, AM1.5 spectral distribution) of a commercially available silicon solar cell49; ηannual is the photovoltaic annual efficiency, which is about 95% of the maximum power conversion efficiency value due to the changing solar zenith angle throughout the day and year2; EAcO is the energy required to generate acetate in the electrolyser (calculated above); and Yyeast/acetate is the yield of yeast grown with acetate effluent media.

To calculate the amount of yeast that could be produced per area of land with biological photosynthesis, we used the annual corn harvest data from Illinois to determine the glucose that could be generated by photosynthesis and subsequently used as the primary carbon and energy source to cultivate yeast, as follows:

$$\begin{array}{l}{{{\mathrm{Yeast}}}}\;{{{\mathrm{produced}}}}\;{{{\mathrm{by}}}}\;{{{\mathrm{biological}}}}\;{{{\mathrm{photosynthesis}}}}\\ = {{{\mathrm{Glucose}}}}\;{{{\mathrm{produced}}}}\;{{{\mathrm{by}}}}\;{{{\mathrm{corn}}}} \times {{{\mathrm{Yeast}}}}\;{{{\mathrm{yield}}}}\\ = \left( {{{Y}}_{{{{\mathrm{corn}}}}\;{\mathrm{per}}\;{{{\mathrm{m}}}}^2} \times {{Y}}_{{{{\mathrm{glucose}}}}/{{{\mathrm{corn}}}}}} \right) \times \left( {{{Y}}_{{{{\mathrm{yeast}}}}/{{{\mathrm{glucose}}}}}} \right)\\ = \left( {0.0423\frac{{{{{\mathrm{bu}}}}\;{{{\mathrm{corn}}}}}}{{{{{\mathrm{m}}}}^2\;{{{\mathrm{yr}}}}}} \times 17.45\frac{{{{{\mathrm{kg}}}}\;{{{\mathrm{glucose}}}}}}{{{{{\mathrm{bu}}}}\;{{{\mathrm{corn}}}}}}} \right) \times \left( {0.51\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{yeast}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{glucose}}}}}}} \right)\\ = 0.376\;{{{\mathrm{kg}}}}\;{{{\mathrm{yeast}}}}\;{\mathrm{per}}\;{{{\mathrm{m}}}}^{2}\;{\mathrm{per}}\;{{{\mathrm{yr}}}}\end{array}$$

where $$Y_{{\mathrm{corn}}\;{\mathrm{per}}\;{\mathrm{m}}^2}$$ is the average corn kernel production in Illinois per square meter52, Yglucose/corn is the glucose produced from a bushel of corn kernels53 and Yyeast/glucose is the yield of yeast grown with glucose as the carbon source54. Using our artificial photosynthesis approach, almost 18 times more yeast could be produced per area of land.

For Extended Data Fig. 8, we calculated the energy efficiencies at each step within the systems as follows:

$$\begin{array}{l}{{{\mathrm{Acetate}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{yeast}}}}\;{{{\mathrm{biomass}}}}\;{{{\mathrm{efficiency}}}} = {{Y}}_{{{{\mathrm{yeast}}}}/{{{\mathrm{acetate}}}}} \times \frac{{\Delta {{H}}_{{{{\mathrm{biomass}}}}}^{{{\circ}}}}}{{\Delta {{H}}_{{{{\mathrm{AcO}}}}}^{{{\circ}}}}}\\ = 0.1889\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{yeast}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{acetate}}}}}} \times \frac{{21.39\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}}{{14.58\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}} = 27.7\% \end{array}$$
$$\begin{array}{l}{{{\mathrm{Sun}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{corn}}}}\;{{{\mathrm{kernel}}}}\;{{{\mathrm{efficiency}}}} = \frac{{{{Y}}_{{{{\mathrm{corn}}}}\;{\mathrm{per}}\;{{{\mathrm{m}}}}^2} \times \Delta {{H}}_{{{{\mathrm{corn}}}}}^{{{\circ}}}}}{{{{E}}_{{{\mathrm{e}}}}}}\\ = \frac{{1.073\;{{{\mathrm{kg}}}}\;{{{\mathrm{m}}}}^{ - 2}\;{{{\mathrm{yr}}}}^{ - 1} \times 16{,}200\;{{{\mathrm{kJ}}}}\;{{{\mathrm{kg}}}}^{ - 1}}}{{5{,}543{,}620\;{{{\mathrm{kJ}}}}\;{{{\mathrm{m}}}}^{ - 2}\;{{{\mathrm{yr}}}}^{ - 1}}} = 0.31\% \end{array}$$
$$\begin{array}{l}{{{\mathrm{Corn}}}}\;{{{\mathrm{kernel}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{glucose}}}}\;{{{\mathrm{efficiency}}}} = w_{{{{\mathrm{glucose}}}}/{{{\mathrm{corn}}}}} \times \frac{{\Delta {{H}}_{{{{\mathrm{glucose}}}}}^{{{\circ}}}}}{{\Delta {{H}}_{{{{\mathrm{corn}}}}}^{{{\circ}}}}}\\ = 0.688 \times \frac{{15.54\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}}{{16.2\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}} = 66.0\% \end{array}$$
$$\begin{array}{l}{{{\mathrm{Glucose}}}}\;{{{\mathrm{to}}}}\;{{{\mathrm{yeast}}}}\;{{{\mathrm{efficiency}}}} = {{Y}}_{{{{\mathrm{yeast}}}}/{{{\mathrm{glucose}}}}} \times \frac{{\Delta {{H}}_{{{{\mathrm{biomass}}}}}^{{{\circ}}}}}{{\Delta {{H}}_{{{{\mathrm{glucose}}}}}^{{{\circ}}}}}\\ = 0.51\frac{{{{{\mathrm{g}}}}\;{{{\mathrm{yeast}}}}}}{{{{{\mathrm{g}}}}\;{{{\mathrm{glucose}}}}}} \times \frac{{21.39\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}}{{15.57\;{{{\mathrm{kJ}}}}\;{{{\mathrm{g}}}}^{ - 1}}} = 70.1{{{\mathrm{\% }}}}\end{array}$$

where $$\Delta H_{\mathrm{biomass}}^{\circ}$$ is the enthalpy of combustion of yeast biomass55, $$\Delta H_{\mathrm{kernel}}^{\circ}$$ is the enthalpy of combustion of corn kernels56, wglucose/corn is the mass fraction of glucose found in a corn kernel53 and $$\Delta H_{\mathrm{glucose}}^{\circ}$$ is the enthalpy of combustion of glucose.

There are many ways in which the energy of biological photosynthesis can be improved on, such as increasing CO2 concentrations and metabolic engineering. We have included numbers for the theoretical maxima of both systems in Extended Data Fig. 8b (refs. 48,49,50,51,52,53,54,55,56,57,58,59).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.