Artificial Photosynthesis
Turn sunlight, water and CO2 directly into fuel, efficiently, durably and cheaply. It works in the lab but is not practical.
Open in the interactive tree →Plants split water and fix CO2 but usually with under 1% efficiency. Artificial systems (semiconductors plus catalysts, photoelectrodes) reach more than 20% in the lab for splitting water into hydrogen, but catalysts corrode in water and light. The step from hydrogen to CO2-based fuels is harder still.
As of October 2026
As of 2026 lab prototypes reach a record of about 22.4% efficiency, but 'no practical system has been demonstrated' and the economics are not competitive; photovoltaics plus electrolysis is the current alternative. Unsolved: catalysts that convert CO2 from ordinary air (0.04%) efficiently and stay stable for years.
What is missing
- Stable, cheap catalysts without precious metals (iridium, platinum)
- Selective CO2 conversion to long-chain hydrocarbons at low concentration
- Durable light absorbers (perovskites are unstable in water)
- Scalable, inexpensive reactors that separate hydrogen from oxygen
Becomes possible once solved
- Synthetic fuels from sunlight for aviation and shipping
- A closed carbon cycle
- Local production of fertilizer and chemicals
Open steps
- Cheap catalysts without precious metals High AI leverageFind water-splitting and CO2-reduction catalysts from abundant elements that are active and stable in acid or neutral water for years.
- Selective CO2 to long-chain products Medium AI leverageConvert dilute CO2 selectively into C2+ fuels and chemicals rather than methane, carbon monoxide or formate, at useful rates.
- Durable light absorbers in water Medium AI leverageStabilize light absorbers, including perovskites, against corrosion in water and light for years.
- Scalable H2/O2-separating reactors Low AI leverageDesign reactors that separate hydrogen from oxygen safely and cheaply at scale while keeping absorber and catalyst efficiency.
Where AI could help
Medium AI leverage. AI speeds up catalyst and material screening, but durable, cheap catalysts and reactors must still be built and tested over long periods.
- Screen catalyst materials with machine-learned potentials instead of costly DFT
- Run closed-loop robotic labs that test catalyst and light-absorber mixtures around the clock
- Predict corrosion of absorbers and catalysts in water to find durable materials
- Optimize reactor flow and gas-separation designs by simulation
Shown so far
- In July 2020 a Liverpool mobile robot guided by a Bayesian search ran 688 experiments in eight days to optimize photocatalysts that make hydrogen from water and light. source
- In November 2024 Meta and Toronto researchers published OCx24, with 572 catalyst samples synthesized and tested for CO2 reduction and hydrogen evolution to check AI-model predictions against experiment. source
Prerequisites
- Photosynthesis discovered1779Artificial photosynthesis copies the plant process of making fuel from light, water and CO2
- Catalysis1835
- Quantum Chemistry & Bonding1927
- Photovoltaics1954
Unlocks
- Sunlight to Fuel2040s?