Artificial Photosynthesis

Artificial photosynthesis uses sunlight to drive chemistry that stores energy in molecular bonds — splitting water into hydrogen and oxygen, or reducing carbon dioxide into carbon-based fuels. The appeal over photovoltaics plus batteries is that a fuel keeps, moves, and drops into existing infrastructure; the appeal over biology is efficiency, since natural photosynthesis converts a low single-digit percentage of incident light at best.

Three technologies wear the name and they are not close to each other. Photoelectrochemical devices immerse a semiconductor in electrolyte so that absorption, charge separation and catalysis happen in one integrated unit. Photocatalytic sheets and particle suspensions disperse the catalyst in water and collect the gas — the cheapest architecture and by far the least efficient. Photovoltaic–electrochemical systems wire a solar cell to an electrolyser; this is not usually called artificial photosynthesis, and it is the benchmark. Conflating the three is how the field's headline numbers get quoted against the wrong comparison, and the performance separation between them is more than an order of magnitude.

The framing under test contains two claims: that sunlight can be turned into fuel directly, and that it can be done at useful efficiency. The first is established and has been for decades. The second fails, and it fails against the wrong opponent. The literature's implicit comparison is against natural photosynthesis and against doing nothing; against those, 11.2% on a postage stamp is impressive. Against a photovoltaic cell wired to an electrolyser at 30% solar-to-hydrogen averaged over 48 hours, the best integrated device is behind by a factor of nearly three and the best large-area demonstration by a factor of forty. The word doing the damage is “directly”, because directness is an aesthetic property rather than a performance one.

Established State the benchmark first, because every number below is read against it. A triple-junction III–V cell driving PEM electrolysers reached 30% solar-to-hydrogen averaged over 48 hours, and a commercial-scale system using perovskite/silicon tandems with earth-abundant nickel catalysts reached 20%. One review describes photovoltaic–electrochemical systems as having “entered the stage of industrial application”; another concludes plainly that PV-based electrolysis is nearest to practical deployment while photoelectrochemistry is “still at a nascent stage”. Established Reported solar-to-hydrogen efficiencies for genuine PEC span roughly 2–10%, with tandem photoanode/photocathode cells at about 4.3%, and stability typically 10–100 hours. Frontier The devices that carry the name are between three and forty times less efficient than the unglamorous wired alternative that already exists commercially.

Established The best module-scale integrated result is real and it is 16 square centimetres. A perovskite-based artificial leaf — a four-by-four array of 1 cm² sub-cells using chlorine-doped formamidinium lead triiodide with nickel-iron-cobalt oxyhydroxide and cobalt-molybdenum sulfide catalysts — reached 11.2% solar-to-hydrogen, holding 99% of initial performance over 140 hours with negligible measured lead leaching. Established Its authors state the remaining problem directly: it “remains a challenge to achieve a practical-size solar water-splitting device” maintaining efficiency, stability and scalability together over a large area.

Established The highest photocatalytic figure was obtained under conditions that change what it means. A 9.2% solar-to-hydrogen result with indium-gallium-nitride nanowires and rhodium/chromia/cobalt-oxide co-catalysts was measured on a 0.8 by 0.8 cm wafer under concentrated light at about 38 suns, at around 70 °C, in pure deionised water. Under tap water and seawater it fell to about 7%; in a 257 W system under natural sunlight it was 6.2%. Frontier A record that requires 38-fold concentration and thermal management is not a leaf; it is a concentrating solar plant with a photocatalyst at the focus, and the balance-of-system it implies is exactly the cost that particle suspensions were supposed to avoid.

Established The scale-efficiency trade is quantified across architectures and it is the field's defining engineering fact. One square metre photocatalytic panel reactors reach 0.4% under natural sunlight; 1 m² hydrogen-farm configurations 1.8%; and the largest field demonstration — a 100 m² panel array operated from September to December 2020 — reached a maximum 0.76% solar-to-hydrogen, evolving gas at up to 3.7 litres per minute. Established Its own authors described the system as “inefficient and energy negative overall”, which is the most useful sentence in the subject and it was written by the people who built it.

Established Durability is what kills these devices and the gap is two orders of magnitude. A 2026 review states the position without hedging: state-of-the-art devices often degrade within tens of hours, against a commercial requirement of 1,000 to 10,000 hours, through thermodynamic instability, photocorrosion and electrolyte-driven dissolution. Established The paradox it names is structural rather than incidental: the optical properties that make a material a good light harvester are the ones that make it vulnerable in a corrosive aqueous environment. A semiconductor that absorbs visible light efficiently has band edges positioned where water and its ions can attack it. Frontier The best durability results therefore come from protection or from repair. The 140-hour perovskite module used nickel foil encapsulation; a concentrated-light photocathode held 1,500 hours at 640 mW/cm² only through repeated platinum co-catalyst redeposition — the electrode was being repaired during the test; and one photocatalyst, aluminium-doped strontium titanate, has held its efficiency for over 1,600 hours under simulated sunlight, but absorbs only ultraviolet light, which is why the 100 m² array made from it managed 0.76%. Frontier The field's stability record and its efficiency record are held by different materials for opposite reasons. Established PEC systems have reached over 1,000 hours of continuous operation under optimised conditions, so the target is not unreachable in principle; what no reported device does is reach it simultaneously with double-digit efficiency at practical area under unconcentrated sunlight.

Established The largest bet ever placed on this subject was a national hub, and its own renewal document contains the concession. The Joint Center for Artificial Photosynthesis was established in 2010 as a Department of Energy Innovation Hub, directed from Caltech with Lawrence Berkeley National Laboratory and university partners, funded at $122 million over five years, with the stated mission of finding a cost-effective method to produce fuels using only sunlight, water and carbon dioxide, and named in the 2011 State of the Union. Phase I aimed at full-system hydrogen prototypes; Phase II at systems producing carbon-based fuels from CO2. Its achievements are real: corrosion protection for semiconductors in aqueous solution, earth-abundant water-splitting catalysts competitive with rare-earth alternatives, high-throughput materials screening, integrated test beds and a public materials database. Established And the most consequential sentence in its own renewal overview is about the second goal: “there is no currently known catalyst” meeting the CO2 reduction requirements. Frontier A hub founded to make fuel from CO2 recorded at its midpoint review that the catalyst class its mission required did not exist.

Established JCAP's funding concluded in 2020 and its successors are smaller and narrower. The Liquid Sunlight Alliance, Caltech-led with Berkeley Lab, SLAC and NREL, runs at about $60 million over five years, alongside the Center for Hybrid Approaches in Solar Energy to Liquid Fuels at UNC Chapel Hill — together $100 million over five years announced in 2020. Frontier That is roughly half JCAP's rate per hub, and reframed: the successor's stated aim is “co-design” to streamline the conversion steps rather than to discover new catalyst classes. Frontier The recipient institution's own announcement credits JCAP with raising solar-to-chemical efficiency “from less than 1 percent to 19 percent” and with the world's largest materials library. That 19% figure needs handling: it is a laboratory integrated-device result under favourable conditions, achieved with expensive III–V semiconductors, and it is quoted without an area, a duration or a cost. It is a real achievement and it is not a technology. The relevant comparison is that wired PV-electrolysis reached 30% over 48 hours in the same era with components available for purchase.

Established The CO2 route is much harder than water splitting for reasons that are chemical rather than engineering. Reducing CO2 requires more overpotential, competes directly with hydrogen evolution at the same catalyst, and must select among many possible carbon products. Reviews name competition from unwanted side reactions as the reason product yields stay low, and the hydrogen evolution reaction is the competitor that wins by default at most catalysts in water. Speculative Two figures widely quoted for the efficiency hierarchy on this route are deliberately not printed here as measurements. A solar-to-carbon-monoxide efficiency above 6.5% for wired photovoltaic-plus-electrocatalyst systems, and a solar-to-fuel efficiency below one tenth of one per cent for monolithic photocatalyst sheets making formate, both circulate in the secondary literature and both reach this brief only through an earlier programme page, unverified against a primary source in the research pass behind this brief. They are recorded as claims in circulation, not as measurements, and naming them stops their absence reading as an oversight. Frontier What is independently supported is the shape rather than the digits: the same integrated-versus-wired gap the water-splitting data shows, applied to a harder reaction with worse selectivity.

Frontier The field's own methodologists have begun publishing reporting protocols, which is itself a finding about the literature. A 2025 best-practice paper names the specific failures: incomplete light-intensity documentation, missing reactor geometry and flow details, undefined testing duration, and inconsistent active-area reporting. Established Its exemplar figures show what disciplined reporting looks like — a platinum/gallium-nitride/silicon photocathode at over 240 mA/cm² under 4,000 mW/cm², hydrogen faradaic efficiency above 97%, and 1,500 hours of operation — with the qualification that the stability was maintained by repeated catalyst redeposition. Frontier A field that needs a protocols paper to establish that active area should be reported consistently is a field whose headline numbers cannot be compared across groups, and that is the honest reading of most efficiency tables in the subject.

Frontier The open questions are best stated as a hypothesis space, because the architectures compete rather than compose, and several live positions are minority ones.

Frontier Integrated photoelectrochemical tandem water splitting is the mainstream position: 2–10% typical, 4.3% for photoanode/photocathode tandems, over 1,000 hours under optimised conditions. What would settle it is double-digit efficiency at a square metre or more, under one sun, for 1,000 hours, without repair — four conditions no reported device meets together. Frontier Perovskite artificial leaf modules are the most promising integrated result in the literature, at 11.2% on 16 cm² with 99% retention over 140 hours; what would settle them is the same performance at 1 m² for 1,000 hours, plus a lead-leaching measurement over a service life rather than a week. Frontier Photocatalytic particle sheets have the only months-long outdoor operation in the field and the lowest efficiency; the science is frontier and the economics speculative, and what would settle the science is a visible-light photocatalyst with quantum yield comparable to what strontium titanate achieves in the ultraviolet — because the ultraviolet limitation is precisely why the efficiency is 0.76%.

Frontier Concentrated photoelectrochemistry buys efficiency and reintroduces the costs flat photocatalysis existed to avoid. The 9.2% record needed about 38 suns; the 1,500-hour durability run needed 640 mW/cm² and periodic catalyst redeposition; the exemplar photocathode ran at 4,000 mW/cm². Tracking, optics and cooling are then in the cost stack, and at that point the comparison is against concentrated PV rather than against a panel. Frontier Solar-thermochemical routes — photothermal reverse water-gas shift, metal-oxide redox cycles — are a legitimate and promising solar-fuels family, and they should be labelled as thermochemistry driven by concentrated heat rather than photochemistry driven by photons exciting a semiconductor. Speculative A frequently cited outdoor demonstration of about 103 m² producing carbon monoxide at over 16% solar-to-chemical reaches this brief only through an earlier programme page and is not independently verified here; it is recorded as a claim in circulation. Its more important property is categorical rather than numerical: reporting a thermochemical result under the artificial-photosynthesis heading inflates what the photocatalytic and photoelectrochemical literatures have achieved.

Speculative Three routes sit further out and each has a specific reason. Bio-hybrid systems — semiconductor-bacteria and enzyme-electrode pairings — achieve selectivity for multi-carbon products that abiotic catalysts struggle with, at current densities and robustness far below industrial relevance; what would settle them is a continuous multi-week run at industrially relevant current density. Speculative Photoelectrochemical CO2 reduction to fuels awaits a catalyst with high selectivity for a single multi-carbon product at industrial current density, and the largest programme ever funded to find one recorded that none was known. Speculative Direct solar ammonia by photocatalytic nitrogen reduction has a persistent and troubled literature beset by a well-documented artefact: trace nitrogen contamination produces false positives at the low yields typical of the field, and isotope-labelled control experiments have only recently become expected practice. That artefact, not the chemistry, is why the route is flagged where it is.

Frontier The most honest thing the field has done recently is change its objective, and it should be recorded as that rather than as progress. The current review literature increasingly frames photoelectrochemistry's prospects as expanding beyond water splitting to hydrogen peroxide, ammonia and chlorine — higher-value, lower-volume products. That is a sound commercial instinct and it is a retreat from the framing under test. Handwave Making hydrogen peroxide photoelectrochemically is chemistry, not energy; it does not turn sunlight into fuel at scale and it is not measured against photovoltaics plus electrolysis. Speculative Two further positions belong on the list for completeness: floating and marine deployment on seawater, where 1 m² platforms have operated and efficiency falls relative to pure water; and solar fuels as a seasonal store rather than a fuel supply, where nobody has published a comparison against the storage requirement Energy Storage Revolutions quantifies, at the efficiencies actually demonstrated. The absence of that analysis is itself the finding.

Established The first bottleneck is the one that gives the architecture its name. Putting the semiconductor in the electrolyte is what makes a device “artificial photosynthesis” rather than a solar panel plumbed to a box, and it is the source of most of the disadvantage. It causes photocorrosion, which limits lifetimes to tens of hours. It produces a mixed hydrogen–oxygen stream that must be separated at a cost and a hazard. And it makes efficiency fall with area, because resistive losses, illumination non-uniformity and bubble management all worsen when the light-absorbing surface is also the reacting surface. Frontier Separating those functions solves all three problems at once, which is what a wire and an electrolyser do, and it is why they win.

Established Gas separation is not incidental and it is rarely costed. The 100 m² array produced a moist, stoichiometric hydrogen–oxygen mixture and recovered hydrogen with a commercial polyimide membrane; the team ran deliberate ignition tests and the system survived. The developer's own account gives a hydrogen recovery rate of 73% at 94% purity. Frontier A 73% recovery applied to a 0.76% conversion is a system efficiency near 0.55% — and separating oxyhydrogen is a cost and a hazard that wired electrolysis structurally does not have, because an electrolyser separates the gases at the membrane by design.

Frontier The third bottleneck is a materials trade with no settled direction. The efficiency records depend on materials that do not scale gracefully: III–V multijunctions using gallium, indium and germanium; iridium and platinum-group catalysts for oxygen and hydrogen evolution; and lead in the best-performing perovskites. Frontier The counter-trend is real: the 11.2% module uses earth-abundant nickel-iron-cobalt oxyhydroxide and cobalt-molybdenum sulfide with only about 0.1 wt% platinum, and photocatalysis' whole appeal is that strontium titanate is cheap. The materials question is not settled in either direction and depends on which architecture wins, which is an unusual and honest position for a bottleneck section to end in.

Established The fourth is the economics of the cheap-panel wager, and every term in it is unproven. The field's own arithmetic makes the assumption explicit: at $100 per square metre, a 10-year service life and 4% annual depreciation, photocatalytic hydrogen costs about $3.50/kg — against natural-gas hydrogen at $0.50–1.70/kg. Speculative That figure assumes an efficiency the field has not demonstrated in the field, a lifetime it has not demonstrated at all, and a panel cost roughly at the level of installed utility solar — for a device that must additionally handle, separate and safely vent an explosive stoichiometric gas mixture. Handwave Any statement that the $3.50/kg cost target has been demonstrated is describing a model output as a measurement.

Established And the fifth is that the product enters a market that is demand-constrained rather than supply-constrained. Only about 4% of world hydrogen is made by electrolysis at all, with roughly 96% from fossil fuels — so solar fuels do not compete against grey hydrogen's price; they compete against green hydrogen's. Frontier Hydrogen Economies adjudicated that market as constrained by willingness to pay rather than by production cost, with a demand-side ceiling below roughly $2/kg for most sectors absent policy support. A photocatalytic panel producing hydrogen at a modelled $3.50/kg is entering that market and not a hypothetical one, and this brief does not smuggle in a demand case that brief refuted.

Established Nothing on this map produces a result this brief waits on. What it waits on are scientific results nobody is currently producing, recorded below as typed requirements rather than as edges to other briefs, and one market fact that Hydrogen Economies has already measured. No typed depends-on edge is claimed.

Established The missing results are specific and they are not restatements of the subject's title. First: a visible-light photocatalyst with quantum yield comparable to what aluminium-doped strontium titanate achieves in the ultraviolet, since the ultraviolet-only absorption of the best stable photocatalyst is the direct cause of the 0.76% field result. Second: a light absorber that resists photocorrosion in the electrolyte without encapsulation or in-run catalyst redeposition, since the two longest stability runs in the literature were achieved by protecting the electrode and by repairing it mid-experiment. Third: a CO2 reduction catalyst selective for a single multi-carbon product at industrial current density — the class a $122 million national hub recorded as not existing. Frontier None of these is a scale-up task, and none is on any funded programme's stated critical path, which is why they are recorded as scientific results nobody is producing rather than as engineering requirements.

Speculative And one comparison is missing from the literature rather than from this brief. Nobody has published a like-for-like land, water and capital comparison of a photocatalytic array against the same area of 20%-efficient photovoltaics feeding an electrolyser. The absence of that comparison, in a field whose entire case rests on it, is a finding about the field.

Established The highest-value experiment is the comparison the literature avoids, run properly. Take a fixed land area under identical irradiance; on one half deploy the best available photocatalytic or photoelectrochemical array, on the other a commercial photovoltaic field feeding a commercial electrolyser; run both for a year; publish hydrogen delivered, capital cost, water consumed, degradation and downtime. Frontier Every component exists, the cost is modest against a $100 million hub, and the result would settle the subject's central dispute in either direction. That it has not been run is the strongest available evidence about which answer the field expects.

Established Second: durability under a protocol that forbids repair. The longest stability results in the literature rely on encapsulation or on redepositing the catalyst during the run, and both are legitimate engineering that must be declared. A standing protocol — fixed illumination, declared area, declared duration, no intervention — would make efficiency tables comparable across groups for the first time, and the field's own methodologists have already written the specification for it.

Frontier Third: measure lead leaching over a service life. The most promising integrated module is a lead-halide perovskite that by design sits in water, and its reassuring leaching result covers 140 hours — two orders of magnitude short of the 1,000–10,000 hour commercial requirement. Speculative A negative result there would end the most promising line in the literature, and a positive one would be the single most valuable durability datum in the subject. Nobody has published either.

Frontier Fourth: isotope-labelled controls as standard for nitrogen reduction. The photocatalytic ammonia literature is contaminated by trace-nitrogen false positives at exactly the yields it reports. Established Requiring labelled controls is not a new idea and the field has only recently begun to insist on it; retrospective application to the existing corpus would tell readers how much of it survives.

Established The engineering requirement that decides everything is area, and the scaling data are unambiguous. Efficiency falls from 11.2% at 16 cm² to 1.8% at a square metre in the best hydrogen-farm configuration, 0.4% at a square metre in a photocatalytic panel reactor, and 0.76% at 100 m². Frontier Those are different chemistries, so the trend is not one device degrading — it is the field's best result at each scale, and it falls. Resistive losses, illumination non-uniformity and bubble management all worsen when the absorbing surface is the reacting surface, and no reported architecture escapes it.

Established A photocatalytic plant is a chemical plant with an explosion hazard designed in. The 100 m² array's own team ran deliberate ignition tests because the reactor output is a moist stoichiometric hydrogen–oxygen mixture at ambient pressure across a large area. Recovery was 73% at 94% purity through a commercial polyimide membrane. Frontier Every square metre added is more surface across which an explosive mixture is generated and must be collected, which is a scaling property of the architecture rather than a solvable detail, and it does not appear in the cost model that gives $3.50/kg.

Established Water quality is an operating requirement, not an assumption. The 9.2% record used pure deionised water and fell to about 7% on tap water and seawater. Speculative A technology whose land-use case depends on deploying across hundreds of thousands of square kilometres cannot also depend on deionised feedwater at that area, and no published system analysis this brief located reconciles the two.

Established And the scale implication settles the ambition question with one figure. Supplying 18.9 TW of global primary energy by photocatalysis at 10% solar-to-hydrogen and 2,000 kWh/m²/yr irradiance would require about 800,000 km² — an area exceeding Japan. Frontier At the demonstrated 0.76%, the same arithmetic gives an area larger than any country except Russia. Established The field's own viability threshold is 5–10% solar-to-hydrogen, stated by the researcher who built the 100 m² array — an interested party naming a target his own system missed by a factor of seven to thirteen — and one review sets the industrial requirement at 10% flatly.

Within this map: Hydrogen Economies, which owns hydrogen production economics, electrolyser costs, demand-side willingness to pay, leakage and the carrier claim — this brief owns the photophysics, the device architectures, the efficiency and durability records, the CO2 route and the integrated-versus-wired comparison, and where the two meet on whether there is a buyer, that brief's answer stands; Energy Storage Revolutions, which established that the seasonal storage problem is real, large and unaddressed — solar fuels would in principle address it and at 0.76% conversion do not yet; Advanced Battery Technologies, the competitor that stores electricity rather than converting it; and Space-Based Solar Power, the other “more sunlight” answer, which loses to terrestrial photovoltaics on a similar structure of argument.

The boundary with Hydrogen Economies decides this brief's verdict and should be stated plainly. That brief concluded the hydrogen-as-carrier thesis fails on storage, heating and transport and holds only as industrial feedstock substitution, that electrolyser installed capital cost rose against forecast declines, and that the demand-side ceiling sits below roughly $2/kg for most sectors — and it adjudicated itself as depending on nothing on this map. Every one of those findings applies to solar fuels and makes this brief's position worse rather than better. Artificial photosynthesis produces hydrogen. If green hydrogen from grid-connected electrolysers cannot find buyers, a photocatalytic panel producing hydrogen at a modelled $3.50/kg — under assumptions of efficiency and lifetime it has not met — enters the same demand-constrained market.

This brief supersedes, and does not replace, the carbon programme's treatment, which reaches the same subject from the emissions side and remains the shorter route in. The quantified PV-electrolysis benchmark, the three-architecture taxonomy, the durability numbers and their mechanism, the conditions attached to the 9.2% record, JCAP and its catalyst concession, the gas-separation penalty and the land arithmetic are new to this brief.

Outside the map: semiconductor photophysics and electrocatalysis; concentrated solar thermochemistry, which owns the photothermal routes this brief insists on labelling separately; synthetic biology and plant science, which own engineered natural photosynthesis; and carbon capture, which owns where the CO2 for a CO2-reduction system would come from and whose costs a solar-fuels system inherits on top of its own.

Established The institutional record here is unusually legible because the money was concentrated in one place and then halved. A $122 million hub over a decade, ending in 2020, succeeded by $100 million over five years across two centres — roughly half the rate per hub — with a narrower “co-design” remit aimed at streamlining conversion steps rather than discovering catalyst classes. Frontier A funder that narrows a programme's remit after a decade is making a judgement, and it is legible even though nobody stated it. The judgement is not that the science failed; the outputs — corrosion protection, earth-abundant catalysts, high-throughput screening, a public materials database — are real and reusable. It is that the original goal was not reachable on the original schedule.

Frontier Claim discipline is the live governance issue and it is worse in this field than in most. “Solar fuel”, “artificial leaf” and “artificial photosynthesis” are applied interchangeably to devices differing by a factor of forty in efficiency, and the largest reported demonstration commonly cited under the heading is a thermochemical process. Established The field's own methodologists have written the remedy: report light intensity, reactor geometry, testing duration and active area, every time. Frontier That such a paper was necessary in 2025 is the strongest available statement about how comparable the field's efficiency tables are.

Established And several of the most useful figures here come from interested parties running against their own interest, which is why they carry weight. The developer of the 100 m² array published its 0.76%, its 73% recovery at 94% purity, its cost model and its own description of the system as “inefficient and energy negative overall”, and separately named a 5–10% viability threshold his system missed by a factor of seven to thirteen. Frontier The authors of the best module result stated that a practical-size device remains a challenge. A field whose leading practitioners publish the numbers that undercut them is being honest in the primary literature and misrepresented in the secondary one.

Established The direct ethical stakes are low, because this is early-stage research with no deployment footprint. Two questions become real at scale and neither is hypothetical.

Frontier Materials. The efficiency records rest on III–V multijunctions, iridium and platinum-group catalysts, and lead-halide perovskites. The 11.2% module measured negligible lead leaching over 140 hours — a reassuring result over a duration two orders of magnitude short of a service life, in a device that by design sits in water. Frontier The counter-trend is genuine, since the same module runs on earth-abundant nickel-iron-cobalt oxyhydroxide and cobalt-molybdenum sulfide with about 0.1 wt% platinum, and strontium titanate is cheap. The honest position is that the materials question is not settled in either direction and will be decided by which architecture wins.

Speculative Land and water. 800,000 km² at an efficiency nobody has demonstrated, using pure water where efficiency drops measurably with tap water and seawater, is a resource claim that has never been costed against the same land used for photovoltaics at above 20% module efficiency. Speculative Nobody has published that comparison in a form this brief could verify, and its absence is itself a finding — a field whose case depends on beating an alternative has not published the head-to-head.

Handwave And the claim-discipline duty is an ethical one rather than merely an editorial one. Any presentation of artificial photosynthesis as a near-term contribution to the carbon problem misrepresents a laboratory science to an audience allocating attention and money. Established The carbon programme's page on this subject says so, and this brief agrees with it.

Established The terminal position is that sunlight can be turned into fuel directly, at between one twentieth and one third of the efficiency of turning sunlight into electricity and then into fuel. Both halves of that sentence are measured. The first has been true for decades and is genuinely remarkable chemistry. The second is why nothing follows from it commercially.

Frontier The deeper result is structural and it generalises beyond this subject. Integration — the thing that earns the name — is the source of most of the disadvantage. Immersing the absorber causes the photocorrosion that limits lifetimes to tens of hours. Producing hydrogen and oxygen at the same surface creates a mixed stream that must be separated at cost and hazard. Making the light-absorbing surface also the reacting surface is why efficiency falls with area. Separating the functions solves all three at once. That is what a wire and an electrolyser do, and it is why elegance loses.

Frontier The one honest counter-argument is capital cost per square metre, and it deserves to be stated at its strongest. A photocatalytic sheet could in principle be so cheap that a factor of forty in efficiency does not matter — efficiency is a proxy for cost, not a terminal value, and a field that says so is reasoning correctly. The field states the wager explicitly at $100/m² and a ten-year life for $3.50/kg hydrogen. Speculative Every term in it is unproven, the panel cost target is roughly installed utility-solar cost, and the market it would enter is demand-constrained. The wager is coherent and it is losing on all three terms simultaneously.

Speculative What would change the picture is not incremental efficiency but a different absorber. The stability record and the efficiency record are held by different materials because good visible-light absorption and corrosion resistance are in tension at the level of band-edge positions. A stable visible-light absorber with high quantum yield would collapse that tension and reopen the entire subject, which is why it is recorded below as a missing scientific result rather than as an engineering target. Handwave Absent it, any projection of solar fuels as a material contribution to primary energy is assuming a material that does not exist, and the assumption is usually made silently.

These horizons track laboratory results, programme funding cycles and the benchmark this field is measured against rather than deployment:

• 10 yr: Frontier Expect module-scale integrated results to improve incrementally and area to remain the barrier: the credible near-term target is double-digit efficiency at a square metre for 1,000 hours without repair, which no device currently approaches. Frontier Expect the literature's centre of gravity to continue moving toward hydrogen peroxide, ammonia and chlorine, and expect that shift to be reported as progress when it is a change of objective. Speculative The most valuable thing that could happen in this window is not a record but a protocol: consistent reporting of light intensity, area, duration and intervention, which the field's own methodologists have already specified. Handwave Any claim of commercial solar fuels in this window is not supported by anything in the measured record.
• 25 yr: Speculative This is the window in which the subject either finds a stable visible-light absorber and becomes a technology, or consolidates as a chemistry programme producing high-value products. Frontier The competitive position is set from outside: photovoltaic efficiency and electrolyser cost are both moving, and every improvement in the wired benchmark raises the bar the integrated device must clear. Speculative A CO2 reduction catalyst selective for a single multi-carbon product at industrial current density would be a genuine surprise and would reopen the route a national hub declared catalyst-less.
• 50 yr: Speculative If solar fuels contribute materially to primary energy at this horizon, the most probable route on current evidence is not artificial photosynthesis but concentrated solar thermochemistry or wired photovoltaic electrolysis at scale — both of which already outperform the integrated devices by wide margins. Handwave Projections placing photocatalytic sheets in that role assume the cheap-panel wager was won, which requires an efficiency, a lifetime and a cost that have not been demonstrated individually, let alone together.
• 100 / 250+ yr: Handwave Beyond useful forecasting. The one durable observation is thermodynamic rather than technological: storing solar energy in chemical bonds is a permanent requirement of any civilisation that wants energy density and transportability, so some version of this problem is always worth working on. Handwave Which architecture solves it is not extrapolation from any measurement now on the record.

• Depends on Nothing on this map. No brief in this corpus produces a result this subject waits on: what it waits on are scientific results nobody is currently producing, recorded below, plus a hydrogen market whose demand ceiling Hydrogen Economies has already measured. No typed depends-on edge is claimed.
• Requires (not on this map) Three of the four are specific scientific results that no funded programme lists on its critical path, and stating them is more useful than restating the subject. The first: the most stable photocatalyst in the literature, aluminium-doped strontium titanate, has held its efficiency for over 1,600 hours and absorbs only ultraviolet light — which is the direct and sufficient explanation of why the 100 m² field array reached 0.76%. A visible-light absorber with comparable quantum yield is the single result that would change the subject's arithmetic. The second is structural rather than incidental, and the durability literature names it: the optical properties that make a material a good light harvester place its band edges where water and its ions attack it, so the two longest stability runs on record were achieved by encapsulating the electrode in nickel foil and by repeatedly redepositing the platinum co-catalyst during the experiment. An absorber stable in the electrolyte without protection or repair does not exist. The third is the one a $122 million national hub recorded as absent in its own renewal document: there is no currently known catalyst meeting its CO2 reduction requirements, and what is needed is high selectivity for a single multi-carbon product at industrial current density against a hydrogen evolution reaction that wins by default. The fourth is a market rather than a result: only about 4% of world hydrogen is made by electrolysis, so solar fuels compete against green hydrogen's price rather than grey hydrogen's, and the demand-side ceiling measured for that market sits below roughly $2 a kilogram for most sectors absent policy support — against a modelled $3.50/kg for photocatalytic hydrogen that assumes an efficiency not achieved outdoors and a lifetime not achieved anywhere.
• Enables In principle a solar fuel would address the seasonal storage requirement Energy Storage Revolutions quantifies, and would supply the carbon-neutral liquid fuels several transport and industrial pathways assume. No typed enabling edge is claimed, because at 0.76% conversion at the largest demonstrated area, and with the wired alternative at 30% over 48 hours, the enabling relation would be carried by photovoltaics plus electrolysis rather than by anything in this brief.
• Adjacent Semiconductor photophysics and electrocatalysis; concentrated solar thermochemistry, which owns the photothermal routes this brief labels separately; synthetic biology and plant science, which own engineered natural photosynthesis; carbon capture, which owns the CO2 source and whose costs a CO2-reduction system inherits; and within this map Hydrogen Economies, Energy Storage Revolutions and Advanced Battery Technologies.

Frontier “Artificial photosynthesis beats natural photosynthesis, so it is the efficient route to solar fuel.” It does beat natural photosynthesis, and that is the wrong opponent. Established The benchmark is a photovoltaic cell wired to an electrolyser at 30% solar-to-hydrogen averaged over 48 hours, and 20% at commercial scale with perovskite/silicon tandems and nickel catalysts — modular, reliable and already deployed. Against that, the best integrated module is behind by a factor of nearly three and the largest field demonstration by a factor of forty.

Frontier “Efficiency records show the field is closing the gap.” Read the records with their conditions. The 9.2% photocatalytic figure needed 38 suns, about 70 °C, pure deionised water and a 0.64 cm² wafer, and the same system gave 6.2% outdoors. The 11.2% module is 16 cm². Established The field's best result at each scale falls as the scale rises — 11.2% at 16 cm², 1.8% at 1 m², 0.76% at 100 m² — and those are different chemistries, so this is not one device degrading.

Frontier “Durability is an engineering detail.” It is the structural consequence of the architecture. Devices degrade within tens of hours against a 1,000–10,000 hour commercial requirement, and the mechanism is that good visible-light absorbers have band edges positioned where water attacks them. Established The two longest stability runs in the literature were achieved by encapsulating the electrode and by repairing it during the experiment, and both are legitimate engineering that changes what the number means.

Handwave “The 100-square-metre outdoor result at 16% shows large-area artificial photosynthesis works.” The demonstration usually cited under that heading is photothermal reverse water-gas shift — concentrated solar thermochemistry, driven by heat rather than by photons exciting a semiconductor. Speculative Its numbers reach this brief only through an earlier programme page and are not independently verified here, so they are recorded as a claim in circulation rather than printed as measurements. The categorical point stands regardless: it is a legitimate and promising solar-fuels route, it is not artificial photosynthesis in the sense the term is normally used, and reporting it under that heading inflates what the photocatalytic and photoelectrochemical literatures have achieved.

Speculative “CO2-to-fuel efficiencies are converging on useful values.” The two figures most often quoted for that convergence — a wired solar-to-carbon-monoxide efficiency above 6.5% and a monolithic sheet-to-formate efficiency below one tenth of one per cent — reach this brief through an earlier programme page and were not independently re-verified in the research behind it. They are named here so their absence does not read as an oversight, and they are not printed as measured. Established What is independently supported is worse for the framing than either number: the largest programme ever funded to solve this route recorded in its own renewal document that “there is no currently known catalyst” meeting its CO2 reduction requirements.

Handwave “$3.50 a kilogram is the cost of photocatalytic hydrogen.” It is a model output at $100/m², a ten-year service life and 4% annual depreciation, assuming 10% solar-to-hydrogen. Established The 10% has not been achieved outdoors at any scale, the ten-year life has not been achieved at all, $100/m² is roughly installed utility-solar cost, and the device must additionally separate an explosive stoichiometric gas mixture that an electrolyser separates at the membrane by design. Frontier The wager that a cheap panel beats an efficient one is coherent and it is losing on all three terms at once.

Frontier “Photoelectrochemistry is pivoting to higher-value chemicals, which shows commercial maturity.” It shows commercial sense and it is a change of objective. Hydrogen peroxide, ammonia and chlorine are worth more per mole and are not measured against photovoltaics plus electrolysis. Handwave Making peroxide photoelectrochemically is chemistry, not energy, and describing the shift as progress toward the original goal is describing a retreat as an advance. It is nonetheless the most honest thing the field has done.

Speculative “Photocatalytic ammonia from air and sunlight is close.” The literature reporting it is beset by a specific and well-documented artefact: trace nitrogen contamination produces false positives at exactly the low yields the field reports. Established Isotope-labelled control experiments are the remedy and the field has only recently begun to require them, which means a substantial fraction of the existing corpus has not been tested against the artefact that would explain it.

Frontier “JCAP proved artificial photosynthesis works, raising efficiency from under 1% to 19%.” That claim comes from the recipient institution's own announcement of a successor award. The 19% is a laboratory integrated-device result under favourable conditions with expensive III–V semiconductors, quoted without an area, a duration or a cost. Established It is a real achievement and it is not a technology; wired PV-electrolysis reached 30% over 48 hours in the same era with purchasable components. Frontier And the programme's own renewal document is the more informative source, because it records the catalyst its second mission required as not existing.

Handwave And the framing itself: “directly” is doing work it cannot support. Directness is an aesthetic property. The measured consequence of making a device direct — putting the absorber in the electrolyte — is photocorrosion, a mixed explosive gas stream, and efficiency that falls with area. Established Separating the functions fixes all three, which is what the wired alternative does, and that is the whole result.