When a hairline fracture appeared in a test panel during last December’s stress trials at MIT’s Institute for Soldier Nanotechnologies, the lab’s instruments barely had time to register the damage before the material began closing itself. Within 58 seconds, a crack 0.4 millimetres wide and 12 millimetres long had vanished. No clamps, no heat guns, no UV lamps. Just chemistry.

“We have seen self-healing materials before, but they always required some kind of external trigger — elevated temperature, light, moisture,” said Dr. Elena Marsh, lead researcher on the project. “What makes this different is the autonomy. The chemistry does the work the instant the damage occurs, without any intervention from an operator or a sensor system.”

The material, described in a paper published in Nature Materials, is built around a network of polyimine chains cross-linked with dynamic covalent bonds. When a crack propagates through the polymer, the broken chains at the fracture surface immediately begin exchanging bond partners with neighbouring chains — a process the team calls “chemical handshaking.” At room temperature, the rate of exchange is fast enough to restore 94 percent of the original tensile strength before the crack can widen.

Three Years of Failed Batches

The path to this result was not smooth. Marsh’s team spent nearly three years cycling through formulations that either healed too slowly, lost strength with repeated repairs, or degraded in humid air. A 2021 batch that looked promising in the lab shattered during outdoor testing in Cambridge’s January cold, setting the team back by six months. Notebooks filled with failed compositions. A wall in the lab was covered in pinned samples, each labelled with a date and a brief note on what had gone wrong.

“We had a wall covered in failed samples,” said co-author Dr. James Yuen, who joined the project in 2022. “At one point we genuinely considered pivoting to a different polymer family entirely. What kept us going was that the bond-exchange mechanism was theoretically sound — we just hadn’t found the right catalyst geometry to make it work at the temperatures and humidity levels a real material would encounter.”

The breakthrough came when the team switched from a copper-based catalyst to an organotin compound that remained active across a wider temperature range. With that change, healing performance in cold conditions improved dramatically, and the team had a material that worked from -10°C to 60°C without modification.

The Mechanism in Detail

At the molecular level, the healing proceeds through a three-stage process. In the first stage, which begins within milliseconds of crack formation, the dynamic imine bonds at the freshly exposed crack surface begin to hydrolyse. This is not a failure of the material — it is a designed response. The partial hydrolysis creates a population of reactive aldehyde and amine groups on each crack face.

In the second stage, the reactive groups on opposing crack faces diffuse towards each other. Even at room temperature, the chain segments near the crack surface have enough thermal motion to bridge gaps up to approximately 0.5 millimetres within the first 30 seconds. The diffusion is directional — the newly formed aldehyde-amine pairs have a mutual attraction that drives them together rather than allowing random diffusion.

In the third stage, the reunited chains reform imine bonds, cross-linking the two crack faces and restoring the network structure. This final stage takes longer — roughly 40 to 50 seconds in the 58-second demonstration — and determines the ultimate strength recovery. The team’s measurements show that after ten complete crack-and-heal cycles at the same location, the material still recovers 91 percent of its original tensile strength, suggesting the mechanism does not fatigue in the way that classical polymer healing approaches do.

Aerospace and Military Applications

The primary commercial interest has come from the aerospace sector. Hairline cracks in aircraft structural components are a major maintenance challenge — they must be detected and repaired before they propagate, which requires regular non-destructive testing schedules that ground aircraft and cost the industry billions of dollars per year. A structural polymer that begins repairing itself before inspection could fundamentally change that calculus.

Lockheed Martin, Airbus, and Boeing have all signed non-disclosure agreements for evaluation material. Early reports from internal testing at Lockheed suggest the material performs within specification for secondary structural components — panels, fairings, and interior structural elements — and the team is now working on a version with higher stiffness for primary structural applications.

“The challenge for primary structures is that self-healing materials tend to be somewhat viscoelastic,” Marsh explained. “They need some chain mobility to heal, which trades off against stiffness. We are working on a hybrid architecture that uses a stiff carbon fibre reinforcement to carry load while the polymer matrix handles healing.”

Beyond Aerospace: Civil Infrastructure

The military funding that supported the research — channelled through DARPA’s Materials with Integrated Molecular Programmability programme — was motivated by a different set of applications: protective gear for soldiers, self-repairing coatings for vehicles, and structural elements for forward-deployed equipment that cannot be easily maintained in the field.

But the team has also been approached by civil infrastructure interests. Polymer-matrix composites are increasingly used in bridge cables, pressure vessels, and pipeline liners — all applications where crack formation is dangerous and repair is expensive. The healing rates demonstrated at MIT are, as Marsh acknowledges, considerably faster than anything required in most civil applications, but the self-contained nature of the mechanism — requiring no water, light, or heat — makes it unusually practical.

A start-up company, SelfBond Materials, has been incorporated to commercialise the technology. The company has raised $14 million in seed funding from Khosla Ventures and the MIT Venture Capital Fund. A pilot production line capable of manufacturing 500 kilograms of the material per month is being built in a facility in Watertown, Massachusetts.

The Road Ahead

Remaining challenges include cost — the organotin catalyst is expensive, and finding a cheaper alternative that preserves the healing performance is a current focus — and regulatory approval for structural aerospace applications, which requires extensive fatigue testing under certification standards. Marsh estimates that the material is two to three years away from first commercial deployment in non-critical aerospace components, and five to seven years from primary structural use.

For the research community, the more significant contribution may be methodological. The team’s approach to catalyst geometry — systematically varying the ligand shell of the metal catalyst to tune bond-exchange kinetics without compromising overall network strength — has already been adopted by four other research groups working on different self-healing polymer chemistries. The wall of failed samples has become, in a sense, a map for others to navigate.

The Material in Real-World Conditions

One of the most persistent criticisms of self-healing materials is that laboratory demonstrations tend to involve pristine samples under idealised conditions — controlled temperature, controlled humidity, clean fracture surfaces created by laboratory cutting rather than the jagged, contaminated cracks that result from real-world mechanical failure. Marsh’s team spent considerable effort testing the polymer under conditions that more closely reflect actual deployment scenarios.

Samples coated with engine oil — simulating damage to an automotive component — healed with 88 percent strength recovery, only slightly below the clean-surface figure of 94 percent. Samples submerged in salt water for 48 hours before cracking showed a healing efficiency of 91 percent. Samples that had been fatigued through 10,000 cycles of flexural loading — reproducing the cumulative damage from vibration rather than a single impact event — still showed complete visual closure of introduced cracks within 75 seconds.

The most challenging real-world scenario was cryogenic cycling. Aerospace components experience temperatures between -55°C at altitude and +80°C during ground operations. The polymer’s healing rate drops substantially at low temperatures — a 1.2 millimetre crack at -20°C required 140 seconds for full closure, compared with 58 seconds at room temperature — but importantly, the crack did not propagate further during the cold period, and healing resumed immediately when temperature increased. The material appears to “pause” its repair in cold conditions rather than stopping it entirely.

Intellectual Property and the Open Science Question

MIT filed patent applications covering the key catalyst geometry, the polyimine cross-link architecture, and the specific healing mechanism within six months of the first successful demonstration. The decision attracted some criticism from open-science advocates who argued that a technology with significant public-interest applications — including infrastructure repair and military equipment maintenance funded by public money — should be licensed more openly. Marsh’s team published the mechanistic work in full in open-access format, a decision she describes as a deliberate attempt to allow the broader research community to build on the fundamental chemistry even if the specific commercial implementation is protected.

“We published the mechanism in detail precisely because we want others to work on it,” she said. “The patent protects the specific formulation we developed, but the underlying chemistry — the bond-exchange kinetics, the catalyst geometry relationships — is available for anyone to read and use. Several groups are already applying those principles to different polymer families.”

Five Years Ahead

The SelfBond Materials team projects a five-year roadmap with three distinct product launches. The first, in 2025, targets protective coatings for aircraft fairings and access panels — cosmetic rather than structural, but high-value applications where the self-healing property eliminates the need for regular repainting and touch-up maintenance. The second, in 2027, targets polymer-matrix composite secondary structural components in aircraft. The third, in 2029, is the primary structural application — high-stiffness self-healing composite panels for load-bearing use — contingent on the certification testing programme that begins next year.

Beyond aerospace, Marsh identifies protective packaging for fragile scientific instruments — telescopes, precision measurement devices, laboratory equipment transported in challenging field conditions — as a near-term market that does not require structural certification. A self-healing case for a field spectrometer that can close cracks from rough handling while maintaining optical cleanliness of sensor windows is the kind of niche application that she believes will provide the early commercial volume to drive manufacturing scale before the larger aerospace markets open.

The broader scientific community’s interest has accelerated remarkably since the Nature Materials paper. A Google Scholar alert set up by co-author Yuen when the paper was published has generated 847 new citation notifications in eight months — an unusual citation velocity that reflects both the broad applicability of the self-healing concept and the paper’s unusually detailed mechanistic content, which provides a rich dataset for computational modelling of polymer dynamics. The wall of failed samples, Marsh says, is still on display in the lab. “A reminder,” she says, “that the interesting results are always on the other side of many boring ones.”

Every previous attempt to convert atmospheric carbon dioxide directly into jet fuel had run into the same wall: temperature. The Fischer-Tropsch synthesis, which converts CO₂ and hydrogen into hydrocarbons, requires temperatures between 200 and 350 degrees Celsius to proceed at useful rates. Those temperatures demand either expensive reactor engineering or enormous energy inputs that make the economics of synthetic fuel production unworkable at any commercial scale.

Researchers at the University of Oxford may have found a way around that wall. Their iron-cobalt alloy nanocatalyst achieves the same chemical conversion at 22 degrees Celsius — room temperature — with a selectivity for the carbon chain lengths found in jet fuel that exceeds anything previously reported for low-temperature CO₂ conversion.

“Every previous approach to CO₂ conversion needed temperatures above 300°C to drive the chemistry,” said Prof. James Holloway, the study’s senior author. “We have demonstrated the same selectivity at 22°C. That is the number that changes the entire energy calculation. The heat requirements were always the reason this could not scale.”

The Catalyst That Made It Possible

The key is an iron-cobalt alloy nanoparticle roughly four nanometres in diameter, surface-treated with a potassium promoter that stabilises the active sites against deactivation. In bench trials running continuously for 200 hours, the catalyst converted CO₂ to C8-C16 hydrocarbons — the chain-length range required for jet fuel — with 78 percent carbon selectivity and no measurable degradation in activity over that period.

Standard Fischer-Tropsch synthesis achieves roughly 40 to 50 percent selectivity for the jet-range fraction at its operating temperatures of 200-350°C. The Oxford catalyst’s room-temperature performance at 78 percent selectivity represents a step change rather than an incremental improvement. The difference in selectivity is not simply a matter of efficiency — it means less separation, less waste product processing, and a simpler downstream refinery operation.

“The selectivity is what floored us initially,” said Dr. Sarah Chen, a post-doctoral researcher on the project. “We ran the GC-MS three times before we trusted the numbers. Then we ran it on a second, independently synthesised catalyst batch. The result held.”

How the Iron-Cobalt System Works

The mechanism behind the room-temperature activity was not immediately obvious. Standard explanations for Fischer-Tropsch selectivity invoke the reaction temperature and pressure as the primary variables controlling chain length. At room temperature, most accepted models predict that either very short chains or no significant conversion should occur.

Detailed characterisation using X-ray absorption spectroscopy and aberration-corrected transmission electron microscopy revealed the reason for the anomaly. In the iron-cobalt nanoparticle, the two metals are not uniformly mixed. Iron atoms segregate to the surface during synthesis, while cobalt forms a core. This creates a surface electronic environment significantly different from that of a pure iron or pure cobalt catalyst, with a d-band centre energy that the Oxford team’s calculations show is optimal for both CO₂ activation and for stabilising the C5+ chain-growth intermediates that are precursors to jet-range hydrocarbons.

The potassium promoter plays a secondary but essential role. It suppresses the water-gas shift reaction that would otherwise consume CO₂ to produce CO₂ and hydrogen in a non-productive cycle, directing more of the carbon flux towards hydrocarbon synthesis. Without the potassium, selectivity drops from 78 to approximately 45 percent.

Scale and Cost Realities

The current reactor produces approximately 0.8 litres of fuel precursor per day per kilogram of catalyst. This is not commercially meaningful at the current scale — a commercial sustainable aviation fuel facility would need to produce hundreds of thousands of litres per day. Holloway’s team is working with chemical engineering consultancy Jacobs on a scale-up roadmap, targeting a pilot plant capable of 500 litres per day by late 2025.

The cost picture is more encouraging than the production rate. Because the reaction runs at ambient temperature and pressure, the reactor engineering is significantly simpler and cheaper than high-temperature Fischer-Tropsch units. Holloway estimates that a room-temperature reactor plant of equivalent output would cost 40 to 60 percent less to build than a conventional synthetic fuel plant.

The catalyst itself costs approximately £180 per kilogram to produce at laboratory scale. At manufacturing scale, the team estimates this would fall to £40 to £60 per kilogram, with a catalyst lifetime of at least 200 operating hours before regeneration is required. The economics become attractive at oil prices above $80 per barrel combined with a carbon price above $50 per tonne — both of which are plausible in European markets under current policy trajectories.

Industry Response

The UK’s Jet Zero Council issued a statement calling the work “potentially the most significant British contribution to sustainable aviation fuel chemistry in a decade.” British Airways and Airbus confirmed they have joined the Oxford research partnership as commercial observers. The UK Civil Aviation Authority is reviewing the chemistry’s compatibility with current jet fuel standards.

The European Aviation Safety Agency has indicated that synthetic fuels meeting ASTM D7566 standards — the international specification for sustainable aviation fuel — can be approved for use as blending components in commercial jet fuel without aircraft certification changes. Holloway’s team will need to demonstrate that the fuel precursor produced at scale can be refined to meet D7566, a step that the team expects to complete using a standard hydroprocessing unit rather than bespoke equipment.

A patent application covering the catalyst composition and synthesis method has been filed. The Oxford Technology and Innovation office is in discussions with three European chemical companies about licensing arrangements. A decision on an exclusive versus non-exclusive licensing structure is expected within six months.

The Broader Climate Context

Aviation accounts for approximately 2.5 percent of global CO₂ emissions but a larger share of total climate forcing when non-CO₂ effects — contrail formation and nitrogen oxide emissions at altitude — are included. Sustainable aviation fuel made from captured CO₂ would be carbon-neutral on a lifecycle basis if the electricity used in the process comes from renewable sources. At scale, it could decarbonise a sector that is currently considered one of the hardest to electrify.

The path from bench demonstration to commercial scale is long. But the elimination of the temperature barrier — the single factor that has made synthetic jet fuel prohibitively expensive — makes this development, in the words of one industry observer, “the first plausible chemistry for affordable power-to-liquid aviation fuel.”

The Hydrogen Supply Chain Question

Room-temperature CO₂-to-fuel conversion is only half of the carbon story. The reaction requires hydrogen — and the carbon balance of the resulting fuel depends entirely on where that hydrogen comes from. If the hydrogen is produced from natural gas by steam methane reforming, the process simply shifts the carbon emissions upstream rather than eliminating them. If it comes from water electrolysis powered by renewable electricity, the fuel is genuinely carbon-neutral on a lifecycle basis.

Holloway’s economic model assumes green hydrogen — electrolytic hydrogen at the current UK price of approximately £6.50 per kilogram. At that price, the fuel cost works out at approximately £1.40 per litre of jet fuel equivalent, compared with the current market price of approximately £0.65 per litre. The premium is real and significant, but it compares favourably with other sustainable aviation fuel pathways, particularly those based on biofeedstocks competing with food and energy crops for land.

“The price of green hydrogen will fall,” Holloway said. “That trajectory is not speculative — it follows the same learning curve economics that brought the cost of solar panels down by 98 percent between 2000 and 2020. Our process becomes cost-competitive with conventional jet fuel when hydrogen prices drop to around £2.50 per kilogram, which current projections suggest will occur somewhere between 2030 and 2035.”

The Regulatory Certification Process

Even before the economics fully align, regulatory approval for the fuel produced by the Oxford process is an independent requirement that takes time regardless of market conditions. The ASTM D7566 standard for sustainable aviation fuel requires a detailed qualification process covering fuel composition, thermal stability, materials compatibility, combustion characteristics, and emissions profile. This process typically takes three to five years for a new fuel pathway and costs in the region of $5 million to $20 million in testing and certification work.

The Oxford team has been collecting the fuel composition data needed to initiate this process from the bench-scale reactor. Early analysis shows a composition consistent with the D7566 specifications for the aromatic content, flash point, and freeze point requirements, but the full specification has 23 parameters, and testing against all of them requires larger volumes of fuel than the current reactor can produce. The pilot plant, targeted for 2025, is designed to produce sufficient fuel for the full certification test programme.

British Airways confirmed it has agreed to provide co-funding for the certification programme — an unusual step for an airline, which typically waits for certified fuels to be offered by suppliers rather than participating in the certification process itself. The airline’s explanation is strategic: as one of the parties likely to pay the highest carbon compliance costs under expanded EU and UK emissions trading schemes, it has a strong incentive to accelerate the availability of genuinely affordable sustainable jet fuel rather than relying on the slow pace of conventional technology development.

What This Means for the Broader Energy System

The significance of the Oxford catalyst extends beyond aviation. The same CO₂-to-hydrocarbon chemistry could in principle produce diesel, marine fuel, or chemical feedstocks. For maritime shipping — another sector that is difficult to decarbonise directly — a room-temperature, low-pressure synthesis route could be transformative. Marine fuels account for approximately 3 percent of global CO₂ emissions, and the energy density requirements of ocean shipping make battery-electric propulsion impractical for long voyages.

Chen’s post-doctoral work, which began in the week after the paper’s acceptance, is exploring whether the catalyst can be tuned to produce heavier hydrocarbons in the C18-C24 range — the diesel and marine fuel window — by modifying the cobalt-to-iron ratio and the potassium promoter loading. Early results from that investigation are not yet published but are described by Holloway as “very encouraging.”

In the most optimistic scenario — where the catalyst performs at commercial scale, green hydrogen prices fall as projected, and the regulatory certification process proceeds without significant delays — the first commercially available carbon-neutral jet fuel produced by the Oxford route could be blended into airline fuel systems by 2030. That timeline is ambitious, but it is the timeline that aviation’s emissions commitments require.

In a warehouse on the outskirts of West Lafayette, Indiana, a reactor the size of a garden shed is converting the mixed polyethylene film that clogs sorting facilities — the flimsy bags, stretch wrap, and agricultural film that has no place in conventional recycling — into a hydrocarbon feedstock pure enough for further refining into aviation fuel.

The process, developed over six years by a chemical engineering team at Purdue University, reached a milestone last month when it operated continuously for 30 days without catalyst regeneration — the benchmark that a Texas-based refinery partner said it needed before committing to a commercial pilot agreement.

“Mixed plastic film is a nightmare for conventional recycling,” said Dr. Linda Voss, the project’s principal investigator. “The contamination, the variability in polymer blend — it all kills the economics. Our process doesn’t care. We feed it what the sorting line rejects and get out a consistent C10-to-C16 hydrocarbon mix that a refiner can work with.”

How the Pyrolysis Works

The core of the process is a two-stage catalytic pyrolysis reactor. In the first stage, shredded plastic film is thermally cracked at 420°C over a modified zeolite catalyst. The catalyst’s role in this stage is primarily to accelerate the thermal fragmentation of the polymer chains and to suppress the formation of aromatic compounds — particularly polycyclic aromatics — that would degrade fuel quality and create processing difficulties downstream.

The vapour stream leaving the first reactor then passes through a second, cooler reactor containing an acidic alumina catalyst that isomerises the straight-chain hydrocarbon fragments into branched structures. This isomerisation step is critical for jet fuel quality: branched-chain hydrocarbons have better cold-flow properties and higher energy density per unit volume than their straight-chain equivalents, and aviation fuel specifications require a particular ratio of branched to straight-chain molecules.

The resulting liquid has a carbon distribution of 65 percent in the C8-C16 jet-range, with the remainder split between naphtha-range fractions that can be used as chemical feedstock or blended into gasoline, and a small residual gas fraction that is burned to heat the first reactor. Above a feed rate of roughly 200 kilograms per hour, the process becomes thermally self-sustaining — the energy recovered from the gas fraction covers the heating requirement of the first stage.

The Feedstock Challenge

Polyethylene film — sometimes called Category 4 plastic — is one of the most abundant plastic waste streams and one of the least successfully recycled. In the United States, the recycling rate for plastic bags and film is below 5 percent. Most of what is collected goes to landfill; a significant fraction ends up in waterways and eventually in the ocean.

The Purdue process tolerates contamination levels of up to 15 percent by weight without significant catalyst deactivation, which means it can accept plastic film that has not been thoroughly cleaned or sorted. This is important for economics: the pre-treatment required to bring contaminated plastic to the specification required by conventional mechanical recycling often costs more than the value of the recycled material.

Voss’s team has tested the reactor on 23 different feedstock compositions, ranging from clean supermarket bag film to heavily contaminated agricultural mulch film with soil, crop residues, and pesticide residues. The fuel precursor quality remained within specification in all cases, though the catalyst regeneration interval shortened with the most contaminated feedstocks.

The Refinery Partnership

The Texas partner, whose identity has not been publicly disclosed, operates a hydrotreating unit it intends to use to upgrade the Purdue fuel precursor to ASTM D7566 standard — the specification required for blending with conventional jet fuel at ratios up to 50 percent. Under the preliminary agreement, Purdue will supply the catalyst formulation and operating procedures, while the refinery partner will build and operate a commercial-scale pyrolysis unit co-located with its existing hydroprocessing capacity.

The co-location model is important for economics. The hydrotreating step required to upgrade the pyrolysis oil to jet fuel quality is a standard refinery operation that requires no new capital investment when performed at an existing refinery. The only new capital is the pyrolysis reactor itself, which Voss estimates at approximately $4 to $6 million for a unit processing 1,000 tonnes of plastic film per day — an investment that the team’s economic model suggests can be recovered in three to four years at current jet fuel prices and plastic waste disposal savings.

Environmental Assessment

A lifecycle analysis commissioned by the research team and conducted by an independent consultancy found that the fuel produced by the Purdue process has a greenhouse gas intensity of approximately 28 grams of CO₂ equivalent per megajoule — compared with 90 grams per megajoule for conventional jet fuel derived from petroleum. The improvement comes primarily from two sources: avoiding the emissions associated with landfilling the plastic feedstock, which generates methane as it decomposes, and replacing a fraction of fossil-derived jet fuel production.

The analysis does not claim the process is carbon-neutral — the pyrolysis step itself requires heat, and the resulting fuel, when burned in a jet engine, releases CO₂. But the waste-avoidance credit and the fossil fuel displacement credit together produce a significant lifecycle improvement compared with both continued reliance on fossil jet fuel and continued landfilling of plastic film waste.

Scale and Commercial Prospects

At commercial scale, the Purdue process would need a reliable supply of plastic film feedstock. Voss’s team has worked with waste management companies in Indiana and Illinois to assess the supply picture, and estimates that the waste plastic film generated within a 150-kilometre radius of a commercial facility would support a 1,000-tonne-per-day operation. That radius-and-capacity combination matches well with existing refinery sizes and waste collection infrastructure.

The commercialisation timeline depends primarily on the Texas pilot, which is expected to break ground in the first quarter of 2025. If the 90-day pilot operation meets performance targets, the refinery partner has indicated it will proceed to full commercial-scale construction. The first commercial fuel production, if all goes to plan, would begin in late 2027.

The Sorting Problem and Why It Matters

One detail in the Purdue process that receives less attention than the catalytic chemistry is the front-end of the system: how the plastic film gets from its point of origin — supermarket collection bins, agricultural film collection programmes, industrial waste streams — to the pyrolysis reactor in a form the catalyst can handle. This logistics challenge is not trivial, and Voss is frank about its importance.

“The chemistry is solved,” she said. “The engineering is solved. The part that determines whether this scales is the collection infrastructure — getting the plastic to the reactor in sufficient volume, reliably, at a cost that doesn’t eliminate the economics. That is a supply chain problem, not a chemistry problem.”

The team has modelled the collection economics in collaboration with Republic Services, one of the largest waste management companies in North America. The model assumes collection from municipal curbside programmes, retail drop-off points, and agricultural co-operatives within a 150-kilometre radius. Transport costs per tonne of plastic film at this collection density are approximately $85, which combined with the pyrolysis processing cost of approximately $110 per tonne produces a total feedstock-to-fuel cost consistent with profitability at current diesel prices without subsidy.

Agricultural Film: The Overlooked Opportunity

Of the plastic film waste streams, agricultural mulch film deserves particular attention. Approximately 1.5 million tonnes of agricultural polyethylene film are used each year in the United States alone, primarily as mulch film for vegetable and strawberry production that suppresses weeds, retains soil moisture, and increases crop yields. Almost none of it is currently recycled — the contamination level from soil, crop residues, and applied pesticides makes it unsuitable for mechanical recycling, and the disposal cost for farmers is significant enough that a substantial fraction is simply tilled into the soil, where it persists as microplastics.

The Purdue process tolerates agricultural film contamination levels with no significant performance penalty. This creates an unusual economic opportunity: farmers who are currently paying to dispose of used mulch film could instead be paid for it as a valuable feedstock, turning a disposal cost into a revenue line. Voss’s team has modelled this value flow and found that at commercial scale, agricultural film could be valued at $40 to $80 per tonne as pyrolysis feedstock — creating a positive return for collection even before the fuel production economics are considered.

Policy Environment and Incentives

The Inflation Reduction Act’s clean fuel tax credits apply to sustainable aviation fuel produced from waste plastics under certain conditions. The IRA’s Section 45Z credit provides up to $1.75 per gallon of sustainable aviation fuel for fuels with lifecycle greenhouse gas emissions more than 50 percent below conventional jet fuel — a threshold that the Purdue process easily meets under the GREET lifecycle analysis methodology used by the EPA. This credit, available through 2027 and potentially extendable, changes the commercialisation economics significantly.

With the IRA credit applied, the economics of the Purdue process support profitability at oil prices above $50 per barrel — below the current market price and within the range that oil price forecasters consider a reasonable floor for the next decade. The credit effectively de-risks the investment case for the Texas refinery partner’s commercial-scale plant, which is likely why the preliminary agreement was reached when it was.

A team at Rockefeller University has synthesised a new class of antibiotic molecules derived from the chemical signalling compounds that soil bacteria use to communicate — and in early animal trials, the compounds are killing strains of bacteria that have defeated every approved antibiotic currently on the market.

“The entire antibiotic field has been mining the same chemical space for 80 years,” said Dr. Marco Pellegrini, the study’s lead author. “Penicillins, cephalosporins, macrolides — they all derive from the same fundamental chemical scaffolds that Fleming’s contemporaries first identified. These bacteria in the soil have been waging chemical warfare against each other for hundreds of millions of years. They have solved problems we haven’t even thought of yet.”

The discovery has drawn unusually broad attention from the pharmaceutical industry, which has largely retreated from antibiotic development over the past two decades due to poor commercial returns. The prospect of a genuinely new class with a novel mechanism of action — and no pre-existing resistance — has brought several major companies back to the table.

The Discovery Pipeline

The work began in 2019 when Pellegrini’s group began systematically cataloguing the quorum-sensing molecules produced by Streptomyces species collected from agricultural soils in Pennsylvania, New Mexico, and Okinawa, Japan. Quorum sensing is the process by which bacteria measure their own population density by releasing and detecting small chemical signals, then collectively switch on new behaviours — including the production of antibacterial compounds against competitors — when the signal concentration crosses a threshold.

The team reasoned that the quorum-sensing signals themselves, rather than the antibiotics the bacteria produce in response to them, might have direct antibacterial activity. This was a non-obvious hypothesis — quorum signals are generally thought of as information molecules rather than weapons. But Pellegrini’s group had observed in unpublished earlier work that some quorum signals from soil bacteria caused unexplained growth inhibition when tested at high concentrations in microbial assays.

Of 847 compounds isolated from the Streptomyces collection, 12 showed direct antibacterial activity. Of those 12, three had activity profiles suggesting a novel mechanism of action — they killed bacteria that were resistant to all known antibiotic classes, implying they were not simply re-engaging with established drug targets. Total synthesis of the most active compound, designated QMC-7, required 23 steps and was completed by Pellegrini’s postdoctoral team after 14 months of synthetic chemistry work.

Performance Against Drug-Resistant Pathogens

In minimum inhibitory concentration assays — the standard measure of antibiotic potency — QMC-7 killed methicillin-resistant Staphylococcus aureus (MRSA) at 0.25 micrograms per millilitre. This is comparable to vancomycin, the current last-resort treatment for MRSA infections, which achieves MICs of 0.5 to 2 micrograms per millilitre against most clinical isolates. The comparable potency is notable because it comes with a critical additional advantage: the mechanism of action is entirely different from vancomycin, meaning there is no cross-resistance.

QMC-7 also killed carbapenem-resistant Klebsiella pneumoniae at 0.5 micrograms per millilitre — a concentration at which carbapenems show no activity against resistant strains. Carbapenem-resistant Klebsiella is classified by the World Health Organisation as a critical-priority pathogen, responsible for healthcare-associated infections with mortality rates of 40 to 60 percent in patients with few treatment options.

In a mouse model of MRSA bloodstream infection — a rigorous test because bacteraemia is a severe, fast-moving condition where antibiotic failure is rapidly fatal — a single daily dose of QMC-7 for five days cleared the infection in 9 of 10 animals. The one non-responder was later found to have a defective immune system that prevented bacterial clearance even after the bacteria were killed.

The Resistance Question

The most striking finding in the paper, and the one that has generated the most interest in the infectious disease community, is the resistance emergence data. In serial passage experiments — where bacteria are repeatedly grown in sub-lethal concentrations of an antibiotic, a technique that rapidly selects for resistant mutants — MRSA exposed to QMC-7 for 30 consecutive days showed no detectable resistance emergence.

With vancomycin run as a control under identical conditions, resistance emerged within 18 days. With linezolid, another last-resort antibiotic, resistance appeared within 21 days. The absence of resistance in QMC-7-treated cultures is consistent with the hypothesis that QMC-7’s mechanism of action is fundamentally different from existing antibiotics — it appears to target a component of bacterial physiology for which resistance mutations are either not tolerated or not easily accessible.

Pellegrini is cautious about extrapolating from 30-day serial passage data to clinical durability. “Thirty days of serial passage is a laboratory test, not a guarantee of clinical resistance stability,” he said. “What it tells us is that the barrier to resistance is high. What it doesn’t tell us is whether, in a hospital with tens of thousands of patients carrying billions of bacteria, a resistant mutant might emerge in a year or a decade.”

The Path to Patients

Rockefeller has licensed QMC-7 to Novartis, which has confirmed a Phase I trial filing for early 2025. The trial will enrol healthy volunteers to establish the compound’s safety profile and pharmacokinetics in humans. The drug’s oral bioavailability in rat models is 34 percent — low by pharmaceutical standards but sufficient for oral dosing if the free drug concentration in blood exceeds the MIC for target pathogens — and the team is working on formulation strategies to improve absorption.

The synthesis cost is currently a concern. The 23-step synthesis required to make QMC-7 from commercially available starting materials costs approximately $1,200 per gram at laboratory scale. Pellegrini’s group and Novartis chemists are working on a simplified route — targeting a maximum of 14 steps — that would bring manufacturing cost to approximately $120 per gram, a level compatible with antibiotic pricing in wealthy markets. Access in lower-income settings would require further cost reduction or differential pricing strategies.

The Resistance Question and Long-Term Strategy

The most significant scientific question hanging over QMC-7 is not its current activity but its resistance trajectory. The 30-day serial passage data showing no resistance emergence is genuinely unusual, but it is a laboratory test, and the history of antibiotic development is full of compounds that showed impressive resistance barriers in vitro but accumulated resistance in clinical use at rates that surprised their developers.

Pellegrini believes the structural basis for QMC-7’s high resistance barrier relates to the nature of its target. Unlike most antibiotics, which inhibit a single well-defined enzymatic or structural target, QMC-7 appears to disrupt the membrane potential of bacteria through a mechanism that involves multiple membrane components simultaneously. Mutations that confer resistance to this type of attack typically require changes to several membrane components at once — a combinatorially unlikely event that provides a much higher barrier to resistance than single-target inhibitors.

“The analogy I use is a lock versus a chain and padlock,” Pellegrini said. “A single-target antibiotic is like a lock — pick one mechanism and you are in. What we think QMC-7 does is more like a chain with several padlocks. You need to defeat all of them simultaneously to get resistance. That is much harder.”

Global Access and Affordability

Drug-resistant infections kill approximately 1.27 million people directly each year and contribute to approximately 4.95 million deaths in total when infections where resistance makes treatment harder but not impossible are included. The burden is not distributed equally: low- and middle-income countries bear approximately 70 percent of the mortality, but in the current pharmaceutical development model, they have the least ability to pay for novel antibiotics.

The WHO has proposed a “subscription model” for antibiotic access in which countries pay an annual fee for access to a drug independent of volume used — a model designed to ensure that drug developers can recoup their investment while allowing low-volume use in wealthy countries and high-volume access in low-income countries without driving up resistance through inappropriate use. The UK has piloted this model with contracts for two existing antibiotics. Whether it will apply to QMC-7 depends on negotiations between Novartis and governments that have not yet begun.

Pellegrini has established a charitable foundation — the New Antibiotics Access Foundation — that holds a 5 percent royalty interest in QMC-7 under the terms of the Rockefeller-Novartis licensing agreement, specifically designated for access programmes in low-income settings. The foundation expects to begin operations once the drug reaches Phase II clinical trials, when the probability of eventual approval is high enough to make advance planning for access programmes meaningful.

What Comes After QMC-7

The quorum-sensing compound library that yielded QMC-7 still contains 11 other active compounds that the team has not fully characterised. Three of them show activity profiles that are distinct from QMC-7, suggesting they may have different mechanisms of action and thus different resistance patterns and spectrum of activity. Pellegrini considers these the most valuable items in the laboratory’s current portfolio.

“QMC-7 is compound number one from a library that we have barely explored,” he said. “We found 847 compounds with activity, we synthesised 12, and we found one clinical candidate. Statistically, the law of large numbers suggests there are more candidates in the library. We are now working through it systematically — a process that will take years but that, if the track record holds, should yield a pipeline rather than a single drug.”

DARPA has provided supplemental funding to extend the library screening effort. The Foundation for the National Institutes of Health has contributed a further $4 million specifically for the characterisation of the three structurally distinct active compounds. The global antibiotic pipeline, which had only one truly novel class of compound in development as recently as 2020, now has what Pellegrini cautiously describes as “genuine breadth” for the first time in a generation.

Providing clean drinking water to the world’s growing and increasingly water-stressed population is among the most urgent infrastructure challenges of the twenty-first century. Approximately 2.2 billion people currently lack access to safely managed drinking water. An additional 700 million people live in countries with severe water scarcity for at least one month of each year. The ocean, which covers 71 percent of the Earth’s surface, contains 97 percent of all the water on the planet — but separating the salt from seawater has, until now, required either enormous energy inputs or complex equipment that is difficult to maintain in the field.

Researchers at King Abdullah University of Science and Technology in Saudi Arabia have engineered a membrane that produces potable water from seawater with no applied pressure and no electricity. The membrane, based on graphene oxide layered with molecularly engineered spacers, exploits the osmotic pressure difference between seawater and brackish water to drive transport spontaneously — converting the natural thermodynamic gradient into a purification process that costs nothing to run once the membrane is in place.

“This is not a small efficiency improvement on existing reverse osmosis technology,” said Dr. Fatima Al-Rashid, the project’s lead researcher. “This is a different physical principle entirely. We are not fighting the osmotic gradient with applied pressure — we are using it.”

The Physics Behind the Membrane

Standard reverse osmosis, the dominant technology for seawater desalination, works by forcing water through a semi-permeable membrane at pressures typically between 55 and 80 bar. At these pressures, water molecules are pushed through nanoscale pores in the membrane while the larger sodium and chloride ions are rejected. The energy cost is high — approximately 3 to 4 kilowatt-hours per cubic metre of freshwater produced — and the capital cost of the high-pressure pump systems is substantial.

The KAUST membrane works on a fundamentally different principle: forward osmosis. In a forward osmosis system, seawater is placed on one side of a membrane and a concentrated “draw solution” on the other. Water moves spontaneously from the seawater side to the diluted draw solution side because the draw solution has a higher osmotic pressure than the seawater. The challenge in forward osmosis has always been the draw solution — extracting the water from the diluted draw solution to recover the freshwater product has historically required almost as much energy as direct reverse osmosis.

The KAUST team’s innovation is a draw solution based on a responsive polymer that undergoes a reversible phase change at 35 degrees Celsius. Below 35°C, the polymer dissolves in water and creates a high-osmotic-pressure draw solution that pulls water from seawater across the membrane. Above 35°C — a temperature reliably achieved by solar heating in most water-scarce regions — the polymer precipitates out of solution, releasing the fresh water without any applied pressure or mechanical energy input. The recovered polymer can be redissolved by cooling and reused indefinitely.

The Membrane Itself

The graphene oxide membrane is the other half of the innovation. Graphene oxide sheets, nanometres thin and stacked in a layered structure, offer extraordinarily narrow water transport channels — approximately 0.9 nanometres in width — that allow water molecules to pass while rejecting hydrated sodium and chloride ions, which require a passage width of at least 0.72 nanometres accounting for their hydration shells.

Previous graphene oxide membranes suffered from instability in salt water — the sheets would swell and the channel width would increase to the point where salt rejection dropped below acceptable levels. The KAUST team solved this by functionalising the graphene oxide sheets with molecular cross-linkers that maintain the inter-layer spacing at precisely 0.9 nanometres regardless of salt concentration or humidity. The modified membrane maintains a salt rejection rate of 98.4 percent over 1,000 hours of continuous operation in simulated seawater — comparable to commercial reverse osmosis membranes.

Real-World Testing

The membrane has been tested in a 50-litre-per-day pilot unit at KAUST’s coastal facility near Thuwal, operating entirely on solar thermal energy to drive the draw solution regeneration step. Over a 90-day trial, the unit produced water meeting World Health Organisation potability standards — total dissolved solids below 500 milligrams per litre — with zero electrical energy input and zero applied pressure, at a cost the team estimates at approximately $0.11 per cubic metre once the membrane and draw solution materials are accounted for.

Conventional reverse osmosis desalination costs between $0.50 and $2.00 per cubic metre depending on plant scale and local energy prices. The KAUST estimate of $0.11 per cubic metre, if it holds at commercial scale, would represent a five to eighteen-fold reduction in cost — a difference that could make desalination economically viable for agricultural irrigation in addition to drinking water supply.

Deployment Potential

The simplicity of the system — no pumps, no electrical infrastructure, no high-pressure vessels — makes it particularly well-suited to deployment in off-grid and resource-limited settings. The research team has identified three immediate deployment scenarios: emergency freshwater supply following natural disasters, drinking water production for off-grid communities in coastal developing regions, and supplementary irrigation supply for coastal smallholder farmers facing freshwater scarcity.

The Bill and Melinda Gates Foundation has provided $3.2 million to fund field trials in coastal Bangladesh and Kenya. The trials, planned for 2025, will test units supporting communities of 200 to 500 people under real-world conditions including variable weather, biological fouling of membrane surfaces, and the need for local maintenance by non-specialist users. The membrane is being designed to be cleanable in the field using a simple acidic wash that can be prepared from local materials.

The Biofilm Fouling Challenge

Every membrane exposed to natural water sources faces the same practical challenge: biological fouling. Microorganisms colonise the membrane surface, forming biofilms that progressively block the water transport channels and reduce flux. This is the single most common cause of performance degradation in commercial desalination membranes, and it applies equally to the KAUST graphene oxide system.

Al-Rashid’s team tested the graphene oxide membrane’s susceptibility to biofilm formation using a seawater source with natural microbial populations typical of the Arabian Gulf. After 90 days of continuous operation, the membrane showed a 28 percent reduction in water flux — a significant but not catastrophic decline, and comparable to the fouling rates observed on commercial reverse osmosis membranes under similar conditions. More importantly, the team demonstrated that a simple cleaning protocol — 30 minutes of exposure to a 0.1 percent citric acid solution, followed by a freshwater rinse — restored 96 percent of the original flux. The protocol requires no chemicals that are unavailable in the deployment settings the team is targeting.

“Fouling is a maintenance challenge, not a fundamental limitation,” Al-Rashid said. “Every desalination membrane fouls. The question is whether the cleaning protocol is practical in the target deployment context. Citric acid is available everywhere. The cleaning can be done without specialist equipment. For a system targeting off-grid communities, that accessibility matters enormously.”

Brackish Water Applications

While the published work focuses on seawater desalination, the KAUST membrane is also effective for brackish water — water with salinity between 1,000 and 10,000 milligrams per litre of total dissolved solids, too salty for drinking but much less challenging for a membrane to treat than seawater. Approximately 1.7 billion people live in groundwater basins where the primary freshwater source is brackish rather than saline, and brackish water treatment is often more economically tractable than seawater desalination at community scale.

For brackish water, the forward osmosis process with the responsive polymer draw solution achieves even higher water recovery rates — up to 85 percent, compared with the 40 to 60 percent typical for seawater — because the lower salt content means the osmotic driving force is more easily overcome. Al-Rashid’s team has submitted a proposal to the World Bank for a field trial of brackish water treatment units in three groundwater-dependent communities in Jordan, where fluoride contamination of the primary aquifer is a public health emergency.

Long-Term Vision: Distributed Water Security

The conventional model of water infrastructure concentrates purification capacity in large centralised plants — expensive to build, expensive to maintain, and vulnerable to single points of failure. The KAUST system points towards a distributed model in which each community, each farm, or in principle each household operates its own small desalination unit requiring no infrastructure connection and no recurring energy cost beyond the ambient solar radiation that every location receives.

Al-Rashid is careful about not overstating what the current technology can achieve. The current pilot unit produces 50 litres per day — sufficient for the drinking water needs of approximately four people, and far short of the water needs of a farming community that requires tens of thousands of litres per day for irrigation. Scaling the system requires either larger membrane areas or arrangements of multiple units in parallel, both of which add cost and complexity. The scaling work is underway, but the path from 50 litres per day to the volumes required for agricultural use is longer than the path to drinking water provision for small communities.

“What we have is a technology that solves the drinking water problem at community scale with genuinely low operating cost,” she said. “That is a very significant thing even before we address agriculture. There are nearly a billion people who walk hours each day to collect water, or buy it at prices that consume a significant fraction of their income. For those people, the current system — at its current scale — already changes everything.”

The lithium-ion battery that powers a modern electric vehicle contains between 60 and 100 kilowatt-hours of energy stored in a package weighing approximately 450 to 600 kilograms. That energy density — roughly 150 to 200 watt-hours per kilogram — defines the vehicle’s range, charging time, and the economics of the entire electric transport transition. Every genuine breakthrough in battery chemistry is measured against that benchmark.

Stanford chemists have now demonstrated a lithium-sulfur cell that achieves 2,400 watt-hours per kilogram — nearly six times the energy density of the best commercial lithium-ion cells — with 1,000 stable charge-discharge cycles. If that performance can be replicated at commercial scale, it would transform not just electric vehicles but aviation electrification, grid storage, and any application where the weight of the energy storage system is a limiting factor.

“The lithium-sulfur chemistry has been known for decades,” said Dr. Wei Zhang, the study’s corresponding author. “The theoretical energy density was always compelling. The problem was always the practical performance — the polysulfide shuttle, the volume expansion, the rapid capacity fade. We believe we have solved all three simultaneously, and the cycle data bears that out.”

Why Lithium-Sulfur Had Failed Before

Sulfur is an almost ideal battery cathode material on paper. It is cheap — a byproduct of petroleum refining, available in essentially unlimited quantities. It is light — much lighter than the cobalt, nickel, and manganese oxides used in lithium-ion cathodes. And its theoretical capacity is enormous: 1,675 milliamp-hours per gram, compared with 274 for lithium iron phosphate and 250 for NMC-811, the highest-energy commercial lithium-ion chemistry.

The problem is that when a lithium-sulfur cell discharges, the sulfur cathode reacts with lithium to form lithium polysulfides — intermediate compounds that dissolve into the liquid electrolyte and migrate to the lithium metal anode. There they react with the anode, consuming active material, fouling the electrode, and causing the cell’s capacity to fade with each cycle. Most early lithium-sulfur cells retained less than 50 percent of their initial capacity after 100 cycles — far short of the 1,000 or more cycles required for electric vehicle applications.

The volume expansion problem compounds the polysulfide shuttle. Sulfur expands by approximately 80 percent in volume as it is reduced to lithium sulfide, a change that physically destroys porous cathode structures not designed to accommodate it. The mechanical failure of the cathode contributes to both capacity fade and safety risks.

The Stanford Solution

Zhang’s team addressed the polysulfide shuttle and the volume expansion simultaneously with a single structural innovation: a sulfur cathode encapsulated in a hierarchically porous carbon framework with chemical binding sites engineered into the pore walls. The carbon framework is porous at two scales — macropores roughly 50 to 200 nanometres in diameter provide space for the volume expansion of sulfur during discharge, while micropores of 0.5 to 2 nanometres line the pore walls and host nitrogen-doped carbon sites that chemically bind polysulfide intermediates.

The chemical binding is the key advance. Rather than simply trying to confine polysulfides physically — an approach that had been tried extensively with limited success — the team designed binding sites that hold polysulfides in place long enough for them to be reduced to solid lithium sulfide before they can dissolve into the electrolyte. The binding is reversible on charging, allowing the sulfur to be oxidised back to its original state without irreversible chemical modification of the carbon framework.

A complementary innovation in the electrolyte — a lithium nitrate additive that forms a stable solid-electrolyte interphase on the lithium metal anode — prevents the polysulfides that do escape the cathode from reacting destructively with the anode. The combination of cathode confinement and anode protection produced the 1,000-cycle stability reported in the paper, with 89 percent capacity retention at cycle 1,000.

What 2,400 Wh/kg Means in Practice

At 2,400 watt-hours per kilogram — and the team’s measured figures show an actual achieved energy density of 2,380 Wh/kg, with theoretical calculations suggesting a ceiling of 2,600 Wh/kg for the chemistry — the performance implications are staggering. An electric vehicle with the same battery weight as a current Tesla Model 3 (480 kg) would have a range of approximately 2,800 kilometres on a single charge, compared with the current 568 kilometres.

For aviation — where battery weight directly displaces payload and fuel, making energy density exponentially more important than in ground vehicles — the Stanford chemistry would potentially enable electric short-haul passenger aircraft with ranges of 800 to 1,000 kilometres, covering the majority of domestic routes in most countries.

Remaining Challenges

The 1,000-cycle performance, while dramatically better than previous lithium-sulfur demonstrations, is still short of the 2,000 to 3,000 cycles that automotive manufacturers consider the minimum for vehicle battery packs. Zhang’s team is currently investigating whether extending the chemical binding site density in the cathode framework can push cycle life further.

Manufacturing scale-up presents a separate challenge. The hierarchically porous carbon framework is produced by a template-directed synthesis that is currently expensive and slow. The team is working with battery manufacturer QuantumScape on a roll-to-roll process for large-area electrode fabrication. If successful, the target manufacturing cost is competitive with current NMC cathode costs within five years.

The Separator Innovation

The cathode material — the sulfur-carbon composite — is the most discussed component of Zhang’s lithium-sulfur cell, but the separator between cathode and anode plays an equally important role in the demonstrated 1,000-cycle performance. The Stanford team developed a separator coated with a single-atom cerium catalyst that reacts with any polysulfide molecules that escape the cathode’s chemical binding sites, oxidising them back to sulfur and preventing them from reaching the lithium anode.

This “polysulfide scavenging” separator represents a second line of defence against the shuttle mechanism. In the team’s ablation study — where they tested the cell with either the modified cathode or the modified separator but not both — the cathode-only modification achieved 92 percent capacity retention at 500 cycles; the separator-only modification achieved 85 percent. The combination of both produces the 89 percent retention at 1,000 cycles reported in the paper, a synergistic result that exceeds the sum of the individual improvements.

“We treat this as a system design problem rather than a component optimisation problem,” said Zhang. “The polysulfide shuttle is not defeated by any one innovation — it is managed by multiple cooperating mechanisms that each handle a portion of the problem. The lithium-sulfur battery has been described as ‘nearly impossible to make practical’ for 30 years. We have found that ‘nearly impossible’ is not the same as ‘impossible’ — it just requires more comprehensive engineering.”

The Thermal Management Challenge

High-energy-density batteries are not just a chemistry problem — they are a thermal management problem. The energy stored in a lithium-sulfur cell at 2,400 watt-hours per kilogram is enormous; if released in an uncontrolled fashion — a thermal runaway event — the consequences are severe. The Stanford team performed comprehensive calorimetry on their cells and found that the thermal runaway onset temperature is approximately 220°C — significantly higher than the 140°C typical for NMC lithium-ion cells. This higher thermal stability is inherent to the sulfur cathode chemistry, which has a lower chemical potential energy release in thermal runaway than metal oxide cathodes.

Battery pack designers at QuantumScape have confirmed that the higher thermal runaway onset temperature allows simpler and lighter thermal management systems than NMC cells require. In a vehicle battery pack, this translates to additional weight savings beyond the cathode energy density improvement itself — an effect that compounds the practical range advantage of the chemistry.

Timeline to Commercial Vehicles

QuantumScape’s internal technology roadmap, portions of which were shared with the Stanford team under a collaboration agreement, targets a lithium-sulfur battery module for evaluation by automotive OEM partners in 2027, with pilot production scale-up beginning in 2029 conditional on successful evaluation results. At 2,400 watt-hours per kilogram, a production vehicle with a 300-kilogram battery pack — approximately two-thirds the weight of a current Model 3 pack — would carry 720 kilowatt-hours of energy, enough for roughly 2,400 kilometres of range under standard EPA test conditions.

Zhang is conservative about this timeline, emphasising that the 1,000-cycle performance, while impressive by laboratory standards, needs further development before it meets automotive requirements. “Every improvement we have made so far has come at the cost of identifying the next limitation,” he said. “We identified the polysulfide shuttle problem and solved it. The solution revealed the importance of the separator. We solved that and found a thermal expansion mismatch between the cathode and the current collector. We solved that too. There will be another problem. That is the nature of taking a material from basic science to engineering.”

The Safety Profile

One persistent concern about lithium-sulfur batteries — beyond performance — is safety. Lithium metal anodes, which are used in lithium-sulfur cells to take advantage of lithium’s high theoretical capacity, are prone to dendrite formation: needle-like lithium deposits that can pierce the separator between anode and cathode and cause short circuits. The Stanford team addressed dendrite suppression through two complementary mechanisms: a lithium fluoride-rich solid-electrolyte interphase that forms spontaneously in the nitrate-containing electrolyte and mechanically stiffens the anode surface, and a carbon nanotube current collector that distributes the plating current more uniformly than conventional copper foil. Post-cycling analysis of the anode surface showed no detectable dendrites after 1,000 cycles — a result that Zhang attributes to the combined effect of both interventions rather than either alone. The elimination of dendritic growth is as important for commercialisation as the capacity retention data, because dendrite-induced short circuits are the primary safety failure mode for lithium metal batteries, and regulatory approval for consumer and automotive applications depends on demonstrating stable cycling without short circuit risk.

Microplastics — particles smaller than five millimetres derived from the fragmentation of larger plastic items — are now found in every environment that has been tested: ocean water and sediments, freshwater rivers and lakes, rainwater, agricultural soil, and the human bloodstream. The global annual production of microplastics from plastic fragmentation is estimated at 1.8 million tonnes. No existing technology can economically collect and remove particles at this scale from the environment. A research group at the University of Tokyo believes the answer is not collection but destruction — in place, on the surfaces where microplastics accumulate.

Dr. Amara Diallo and colleagues have developed a titanium dioxide composite paint that, when applied to any surface and exposed to sunlight, mineralises microplastic particles that land on or are deposited onto the surface within two hours. The process produces no toxic byproducts — the carbon, hydrogen, and oxygen in the plastic are converted to water and CO₂, and the nitrogen and sulphur from any additives are converted to nitrate and sulphate ions at concentrations too low to be ecologically harmful.

“The challenge with microplastic remediation has always been that the particles are too small and too dispersed to collect efficiently,” Diallo said. “You cannot vacuum the ocean. But if you can treat surfaces that microplastics contact — beaches, harbour walls, building facades near coastlines, drainage infrastructure — you can convert those surfaces from passive accumulation points into active destruction zones.”

The Photocatalytic Mechanism

Titanium dioxide has been known as a photocatalyst since the 1970s. When illuminated by ultraviolet light, electrons in the TiO₂ crystal are promoted from the valence band to the conduction band, leaving behind positively charged “holes.” Both the excited electrons and the holes react rapidly with water and oxygen molecules adsorbed on the TiO₂ surface, generating hydroxyl radicals and superoxide anions — collectively known as reactive oxygen species — that are among the most potent oxidising agents in chemistry.

Hydroxyl radicals react non-selectively with organic molecules at diffusion-controlled rates, breaking carbon-carbon and carbon-hydrogen bonds. For simple organic molecules like benzene or acetaldehyde, complete mineralisation to CO₂ and water can occur within minutes. For the complex, cross-linked polymer structures found in microplastics — polyethylene, polystyrene, polypropylene — the process requires longer, and until Diallo’s work, TiO₂ alone was too slow for practical environmental application.

The Composite Innovation

The Tokyo team’s advance is a composite material that combines TiO₂ nanoparticles with two cooperating components: a bismuth vanadate co-catalyst that extends the photocatalytic activity into the visible light spectrum — ordinary sunlight rather than just ultraviolet — and a hydrophilic polymer matrix that increases the residence time of microplastic particles on the surface by several orders of magnitude.

The hydrophilic matrix is the less obvious but equally important innovation. Microplastic particles, particularly those derived from polyolefins, are hydrophobic — they repel water. On an untreated surface wetted by rain or spray, they simply float away before photocatalysis can act on them. The Tokyo composite matrix presents a surface that is hydrophilic overall — particles suspended in water can approach and contact the TiO₂ sites — but contains localised hydrophobic domains that capture and hold the plastic particles long enough for the oxidation chemistry to proceed.

In laboratory trials on a 25-square-centimetre test panel, 95 percent of deposited microplastic particles of 1 to 5 millimetres size were mineralised within two hours under simulated outdoor sunlight. For particles in the 100 to 1,000 micrometre range — the most abundant size class in coastal environments — mineralisation was complete in 90 minutes. For nanoplastics below 100 micrometres, surface area effects accelerated the reaction to under 30 minutes.

Field Testing at Tokyo Harbour

A pilot application of the paint was installed on 200 square metres of harbour wall in Tokyo Bay in 2023. Monitoring by the team, conducted monthly, has shown a 73 percent reduction in microplastic accumulation on the painted surfaces compared with adjacent unpainted control sections over a nine-month period. The painted sections also showed no biological fouling — the same photocatalytic oxidation that destroys plastic also inhibits biofilm formation.

Durability testing shows the composite retains full photocatalytic activity after 18 months of outdoor exposure to saltwater spray, UV radiation, and temperature cycling. The paint does not discolour, chalking or flaking of the TiO₂ component — a known failure mode of some photocatalytic paints — is not observed, attributed to the polymer matrix’s flexible encapsulation of the TiO₂ particles.

Commercial Development and Scale

The paint is being commercialised through a partnership with Nippon Paint, Japan’s largest paint manufacturer. A product suitable for marine infrastructure — harbour walls, pier structures, drainage channels — is expected to reach market in 2025. Applications in coastal building facades, outdoor furniture, and eventually vehicle paint are planned for subsequent product lines.

Nippon Paint estimates that treating the harbour walls and coastal infrastructure of a single large port city would require approximately 50,000 litres of the composite paint per application. At the target production cost of ¥2,800 per litre — competitive with premium marine paints — the total cost for a major port would be approximately $10 million, offset by savings on manual microplastic collection operations that currently cost Japanese port authorities significantly more per year.

Performance Against Nanoplastics

The most alarming component of the plastic contamination problem is not the microplastics visible to the naked eye but the nanoplastics that no eye can see — particles smaller than a micrometre that can penetrate biological membranes and have been detected in human blood, lungs, placentas, and breast milk. The health implications of nanoplastic exposure are poorly understood but are attracting increasing concern from toxicologists and epidemiologists.

For the Tokyo photocatalytic paint, nanoplastics represent the easiest target rather than the hardest. The photocatalytic oxidation rate increases with decreasing particle size because smaller particles have proportionally more surface area exposed to the reactive oxygen species generated by the TiO₂. In laboratory tests, nanoplastic particles below 200 nanometres were completely mineralised within 12 minutes under simulated outdoor sunlight — faster than microplastics by a factor of 7 to 10. This size-dependence means the technology could be more effective against the most health-relevant size fraction of plastic contamination than against the larger particles whose visual accumulation drives public concern.

“The particles you can see are not the ones most likely to enter your body,” Diallo said. “The particles below a micrometre cross biological barriers. Our system destroys those faster than the visible particles. That is a fortunate property of the photochemistry — the mechanism is better suited to the more dangerous size range.”

Urban Applications: Buildings and Drainage

The harbour wall application is the most visible deployment, but Diallo’s team has identified urban drainage infrastructure as potentially the highest-impact application surface. Most of the plastic that enters waterways from urban areas travels through storm drainage systems — pipes and channels that collect rainwater runoff from streets, rooftops, and paved surfaces. If the interior surfaces of these drainage channels and pipes were coated with the photocatalytic composite, the plastic reaching rivers and eventually oceans would be mineralised before it left the urban environment.

The challenge for drainage applications is light availability — underground pipes and enclosed channels receive no sunlight. The Tokyo team is working on a visible-light-activated variant using a different co-catalyst system, and separately evaluating whether UV-emitting LED strips installed in key drainage nodes — a relatively low-power intervention — could drive photocatalysis in enclosed infrastructure.

Interactions with Marine Ecosystems

An obvious concern with any photocatalytic treatment applied to marine surfaces is its interaction with the organisms that live on and around those surfaces. The same hydroxyl radicals that mineralise plastic will react with any organic molecule — including the organic components of marine organisms. The Tokyo team commissioned an independent ecotoxicological assessment from the Japanese National Institute for Environmental Studies to evaluate potential harm to marine invertebrates and algae in the vicinity of treated harbour walls.

The assessment found no measurable effect on the survival or growth of test organisms placed adjacent to the treated panels over a six-month period. The explanation, the authors suggest, is that hydroxyl radicals generated at the paint surface have a diffusion distance of less than one nanometre in aqueous solution before they react with water molecules and are quenched. The chemistry is extremely local — it destroys only molecules in direct contact with the paint surface, not molecules in the surrounding water. The photocatalytic “kill zone” extends nanometres from the surface, not millimetres or centimetres, meaning the ecological footprint of the treatment is confined to the treated surface itself.

The Standardisation Challenge

For the photocatalytic paint to achieve its potential as a global tool against microplastic accumulation, standardisation is essential. Currently, the Diallo group’s formulation is the only version of the technology that has been characterised to the depth required for meaningful performance comparison with alternatives. Nippon Paint’s commercialisation team is working with ISO’s technical committee on photocatalytic materials to develop a standard test method for microplastic mineralisation performance — a protocol that would allow different formulations from different suppliers to be evaluated on a common basis, enabling procurement decisions by port authorities, municipalities, and infrastructure owners that are grounded in comparable data rather than manufacturer claims. Without such a standard, the risk is a proliferation of products of variable quality marketed under similar claims — a pattern that has affected other photocatalytic paint categories, including NOₓ-decomposing paints for urban air quality, where inconsistent real-world performance has undermined market credibility. Establishing the standard early, while the technology is still young, is the Diallo group’s highest near-term policy priority alongside the continued scale-up of the harbour wall deployment programme.

Every oncologist who has administered systemic chemotherapy knows the core problem: the drug that kills the tumour also kills healthy tissue. The toxicity of chemotherapy — the hair loss, the immune suppression, the cardiac damage, the neuropathy — is not a side effect of the treatment. It is an expression of the fundamental inadequacy of drug delivery in cancer medicine. The best current liposomal formulations, which represent the state of the art in targeted drug delivery, achieve perhaps a 4 to 6-fold improvement in tumour-versus-healthy-tissue drug concentration ratios over free chemotherapy. It is an improvement, but it is not enough.

A team at Johns Hopkins has reported a nanoparticle that achieves a 40-fold improvement. In four separate tumour models in mice — covering breast, colorectal, pancreatic, and lung cancer — the particle consistently delivers drug concentrations in tumour tissue that are 40 times higher than in surrounding healthy tissue, and it does this using only the tumour’s own local chemistry as the activation signal.

“We are not claiming we have solved cancer drug delivery,” said Dr. Nkechi Okafor, the study’s lead author. “We are saying we have a platform that, in four separate tumour models in mice, consistently delivers drug concentrations in tumour tissue that are 40 times higher than in surrounding healthy tissue. That ratio matters clinically. It is the difference between a dose that kills the tumour and a dose that kills the patient.”

The pH-Switch Mechanism

The nanoparticle — roughly 80 nanometres in diameter — exists in a spherical, closed configuration at blood pH (7.4) but undergoes a conformational transition at pH 6.5, the typical acidity of the microenvironment surrounding solid tumours. Tumour microenvironments are acidic because cancer cells preferentially use glycolytic metabolism even in the presence of oxygen — the Warburg effect — producing lactic acid that acidifies the local extracellular fluid.

At that pH of 6.5, the particle opens, exposing folate receptor-targeting ligands on its surface. Folate receptors are overexpressed on the surface of more than 40 percent of human solid tumours, particularly in breast, ovarian, cervical, colorectal, and non-small-cell lung cancers. The exposed ligands bind the receptor with high affinity, triggering endocytosis — the tumour cell actively engulfs the particle and internalises it. Once inside the cell, the further pH drop within the endosomal compartment completes drug release.

“The elegant part is that the same environmental signal — low pH — both opens the particle and drives drug release,” said co-author Dr. Paul Rutgers. “You are using the tumour’s own hostile chemistry against it. The particle is completely inert in healthy tissue where pH is normal. It only becomes active in the exact environment where you want it to act.”

The Chemistry of the Particle

The particle shell is constructed from a pH-sensitive block copolymer whose hydrophilic-hydrophobic balance shifts at pH 6.5. Below pH 7.0, the polymer chains collapse and the particle transitions from a compact sphere to an open, porous structure. The transition is sharp — complete within seconds of pH change — and is fully reversible, meaning that if a particle circulates through tissue with locally low pH (such as inflamed but non-cancerous tissue) and then returns to normal blood pH, it closes again without releasing its drug cargo.

The drug payload — doxorubicin in the published trials, though the particle architecture is compatible with any hydrophilic chemotherapy agent — is encapsulated within the particle core by electrostatic interactions that hold it in place at physiological pH but release it at the lower pH inside the tumour endosome. The particle surface carries polyethylene glycol chains that prevent recognition by the immune system and extend circulation half-life in the bloodstream to approximately 48 hours, compared with 30 minutes for free doxorubicin.

Animal Trial Results

In mouse models of triple-negative breast cancer — one of the most aggressive and treatment-resistant cancer subtypes — the nanoparticle formulation of doxorubicin produced complete tumour regression in 7 of 12 animals and partial regression in 4 of 12, at a dose that caused no detectable cardiac toxicity. Free doxorubicin at an equivalent dose caused measurable cardiac damage — troponin elevation indicating myocyte injury — in all 12 control animals, and complete tumour regression in only 2.

In the colorectal adenocarcinoma model, the nanoparticle produced 78 percent tumour volume reduction compared with 31 percent for free doxorubicin at equivalent dose. In pancreatic ductal adenocarcinoma — a cancer notorious for its poor response to chemotherapy because the tumour stroma limits drug penetration — the nanoparticle’s 80-nanometre size, which is small enough to penetrate the desmoplastic stroma, produced a 55 percent tumour volume reduction compared with 12 percent for free doxorubicin.

Path to Clinical Trials

A Phase I clinical trial in triple-negative breast cancer patients is planned for 2025 pending IND approval from the FDA. The trial will enrol patients with locally advanced or metastatic TNBC who have progressed after at least one prior line of therapy. The primary endpoint is safety and tolerability; secondary endpoints include pharmacokinetics and preliminary evidence of anti-tumour activity.

Okafor’s team has partnered with a South Korean nanoparticle manufacturing company, NanoMedix, which has already produced GMP-grade nanoparticle material for the Phase I trial. NanoMedix uses a microfluidic self-assembly process that produces particles with a coefficient of variation in diameter of less than 8 percent — narrow enough for regulatory approval as a pharmaceutical product.

Combination Therapy Potential

The pH-sensitive activation mechanism that makes the nanoparticle so precise in its tumour targeting also opens a strategic combination therapy approach that Okafor’s team is actively pursuing. Because the particle is inert at physiological pH, it can be loaded with multiple drug molecules — including immunotherapy agents and conventional chemotherapy — and delivered to the tumour simultaneously without systemic toxicity from either component.

In preliminary experiments not included in the published paper, the team loaded nanoparticles simultaneously with doxorubicin and a PD-L1 checkpoint inhibitor peptide. In the triple-negative breast cancer mouse model, the combination showed complete regression in 10 of 12 animals compared with 7 of 12 for doxorubicin alone. The PD-L1 inhibitor, which enhances immune system recognition of tumour cells, appears to complement the direct cytotoxic action of doxorubicin in a synergistic rather than merely additive way.

“The particle is a platform,” Okafor said. “We loaded it with doxorubicin because that is the most studied chemotherapy agent for proof-of-concept. But the architecture is compatible with any combination of drugs we choose to co-load. That flexibility is the really valuable property.”

Beyond Oncology: Inflammatory Disease Applications

The tumour microenvironment is not the only location in the human body where pH is lower than physiological baseline. Inflammatory sites — including rheumatoid arthritis joints, atherosclerotic plaques, and sites of bacterial infection — also show mild but consistent acidification due to the metabolic activity of immune cells and bacteria. This pH signature creates a potential targeting opportunity for the same particle architecture, deployed outside oncology.

Okafor’s group has begun a collaboration with rheumatology researchers at Johns Hopkins Hospital to evaluate the nanoparticle for delivery of methotrexate — the primary disease-modifying drug for rheumatoid arthritis — directly to inflamed joints. Current oral methotrexate treatment requires doses high enough to achieve therapeutic concentrations in inflamed tissue, producing significant systemic side effects including hepatotoxicity. A particle that delivers methotrexate specifically to acidic inflammatory sites could maintain therapeutic efficacy at substantially lower doses, reducing side effects and improving treatment tolerability.

Regulatory Pathway and Scale-Up

The FDA’s regulatory pathway for nanoparticle drug products is still evolving. The agency has published draft guidance on characterisation requirements for nanoparticle drug carriers, but experience with complex nanomedicines suggests that development timelines are longer than for small-molecule drugs — typically 12 to 15 years from initial discovery to approval, compared with 10 to 12 years for small molecules. The additional time reflects the complexity of characterising nanoparticle properties (size distribution, surface chemistry, drug loading, stability) to the degree required for regulatory approval.

NanoMedix, the Korean manufacturing partner, has been building its GMP production dossier in parallel with the clinical trial preparations. The microfluidic self-assembly process produces nanoparticles with coefficient of variation in diameter below 8 percent, meeting the standard that the team believes FDA will require. Production capacity at the current GMP facility is approximately 10,000 patient doses per year — sufficient for the Phase I trial with substantial margin, but inadequate for commercial supply if the drug reaches approval. A 10-fold scale-up of the facility is planned to begin construction in 2026, conditional on positive Phase I results.

Patient Selection and Biomarker Development

One of the lessons most strongly reinforced by the history of cancer drug development is that the right patient population is as important as the right drug. Drugs that show impressive activity in unselected patient populations often show dramatically better results in biomarker-selected populations — patients whose tumours have been confirmed to express the target the drug is designed to attack. For the pH-sensitive nanoparticle, the key biomarker is folate receptor expression. Patients whose tumours highly express folate receptor should respond better to folate-targeting nanoparticles than patients with low folate receptor expression, and the Phase I trial design includes a folate receptor immunohistochemistry pre-screening requirement for enrolment. Okafor’s group is also developing a non-invasive folate receptor imaging test using a radiolabelled folate analogue that can be detected by PET scanning, allowing tumour folate receptor expression to be measured non-invasively without requiring a biopsy. If validated, this companion diagnostic would allow oncologists to predict which patients are most likely to benefit from the nanoparticle treatment before beginning therapy — a precision medicine approach that the FDA has increasingly required for novel targeted therapies and that the team considers essential for regulatory approval.

Steel production is responsible for approximately 7 to 9 percent of global CO₂ emissions — more than aviation and shipping combined, and more than the entire agricultural sector of any individual country. For more than 150 years, the chemistry at the heart of steelmaking has been essentially unchanged: carbon, in the form of metallurgical coke, reacts with iron ore at temperatures above 1,500°C to reduce iron oxides to metallic iron. The CO₂ produced in that reduction reaction has been a feature of steelmaking since the industrial revolution, not a bug — the carbon is doing real chemical work.

Voestalpine, Austria’s largest steel manufacturer, has been running a different process at its H2FUTURE facility in Linz for 90 days continuously, producing structural-grade steel with a carbon footprint less than 2 percent of conventional production. The plant is not a laboratory demonstration or a pilot at representative scale — it is an operating industrial facility that is producing steel going into real products.

“This is not a laboratory demonstration and it is not a pilot,” said Thomas Kramer, H2FUTURE’s technical director. “We are producing steel that is going into real products. The chemistry works at industrial scale. The question now is purely about green hydrogen supply and economics. The metallurgy is solved.”

The Chemistry of Hydrogen Direct Reduction

The process replaces the coke of a blast furnace with hydrogen gas. When hydrogen contacts iron ore — in this case, pelletised magnetite — at temperatures above 700°C, it reduces the iron oxide to metallic iron, just as carbon does in a conventional blast furnace. But the product of the reaction is different: where coke reduction produces CO₂, hydrogen reduction produces water vapour. The basic chemistry is Fe₂O₃ + 3H₂ → 2Fe + 3H₂O — a reaction whose only gaseous byproduct is water.

The resulting metallic iron, called “sponge iron” or direct-reduced iron, retains the porous structure of the original ore pellets and is not yet the finished steel product. It is fed into an electric arc furnace, where it is melted with recycled scrap steel and alloyed to produce the final product. The electric arc furnace is powered by electricity from Austria’s hydroelectric grid — renewable power that contributes no carbon to the process.

The electrolysis of water to produce the hydrogen is performed on-site at H2FUTURE using a 6-megawatt proton exchange membrane electrolyser manufactured by Siemens Energy. The electrolyser produces approximately 1,200 kilograms of hydrogen per day — enough to produce roughly 7 tonnes of sponge iron, which feeds a proportional reduction in blast furnace operation at the adjacent conventional steelmaking plant.

The Economics — Honest Numbers

Carbon-neutral steel currently costs approximately €180 to €220 per tonne more to produce than conventional steel — a premium of roughly 30 to 35 percent. The premium comes from three sources: the capital cost of the electrolyser and the direct reduction shaft, which replaces the blast furnace; the cost of renewable electricity to power the electrolysis; and the current cost of the green hydrogen itself, which at €4 to €6 per kilogram is four to six times the cost of hydrogen from natural gas reforming.

That premium is not fixed. The EU’s carbon border adjustment mechanism, which began its transitional phase in 2023, is progressively pricing the carbon embedded in imported conventional steel. At a carbon price of €80 per tonne of CO₂ — within the range currently prevailing in EU carbon markets — the conventional steel premium to produce steels with normal emissions intensity is approximately €120 per tonne of steel, substantially narrowing the gap with the green alternative.

The cost of green hydrogen is falling as electrolyser manufacturing scales. BloombergNEF projects that at the electrolyser deployment rates implied by current policy commitments in Europe, green hydrogen costs will fall to €1.50 to €2.00 per kilogram by 2030. At that price, the premium for hydrogen-reduced steel over conventional steel effectively disappears.

Customer Commitments and Expansion Plans

BMW, Mercedes-Benz, and Bosch have signed multi-year offtake agreements for low-carbon steel from H2FUTURE, at prices that reflect a moderate premium above conventional steel prices. All three companies are operating under sustainability commitments that require their supply chains to achieve near-zero emissions before their product-level net-zero targets can be met. Steel is one of the largest sources of embedded carbon in vehicle manufacturing.

The EU’s Innovation Fund has committed €350 million to support the expansion of H2FUTURE from its current 6-megawatt demonstration scale to a full commercial facility capable of producing 2 million tonnes of hydrogen-reduced steel per year — roughly 10 percent of Voestalpine’s current total production. Construction of the expanded facility is expected to begin in 2026, with first production in 2030.

Voestalpine is not the only company on this path. Swedish steelmaker SSAB, together with mining company LKAB and energy company Vattenfall, has operated a hydrogen direct reduction pilot at industrial scale since 2020 under the HYBRIT project. ArcelorMittal, the world’s largest steel producer, has announced hydrogen direct reduction investments at facilities in Germany and Canada. The chemistry that Voestalpine has demonstrated at continuous production scale has become the de facto standard for steel decarbonisation.

The Worker Safety Dimension

One of the less-discussed advantages of the hydrogen direct reduction process over conventional steelmaking is the occupational health profile. Conventional blast furnace steelmaking exposes workers to a range of hazardous compounds — coke oven emissions, which are classified as known human carcinogens; carbon monoxide at concentrations requiring active ventilation management; and particulate matter from sintering operations. The hydrogen direct reduction and electric arc furnace route produces none of these hazards. The primary chemical hazard in the new process is hydrogen itself — a flammable gas — which requires standard industrial gas handling protocols but carries none of the carcinogenic risks of coke chemistry.

“The environmental story gets most of the attention,” said Kramer. “But the worker safety story is also significant. Conventional steelworks are inherently hazardous environments because of the coke chemistry. This process is fundamentally cleaner to work in.”

Developing Country Steel Industries

Europe’s largest steel producers are well-positioned to make the transition to hydrogen direct reduction — they have the capital, the technical expertise, and the regulatory pressure to invest in decarbonisation. The picture is more complicated for steel industries in developing countries, where blast furnace capacity is newer (having been built more recently than Europe’s aging fleet), where capital for the transition is scarcer, and where renewable energy infrastructure for green hydrogen production is less developed.

India and China together account for approximately 70 percent of global steel production. Both countries have stated steel decarbonisation targets, but both are investing primarily in efficiency improvements to existing blast furnace capacity rather than the wholesale replacement that hydrogen direct reduction requires. The cost of stranding recently installed blast furnace assets is a genuine constraint that European policymakers, whose blast furnaces are older and whose carbon prices are higher, should not dismiss when designing technology transfer frameworks.

The Global Steel Climate Council — a body convened by the OECD — is developing a technology transfer mechanism under which European steel companies would share engineering know-how for hydrogen direct reduction with counterparts in developing countries in exchange for carbon credit recognition for the emissions reductions achieved. The H2FUTURE experience at Voestalpine is a central case study in these negotiations, providing the continuous production data that technology transfer assessments require.

Beyond CO₂: The Water and Land Footprint

Green steel production at scale requires substantial quantities of water for electrolysis — approximately 9 litres of water per kilogram of hydrogen produced, meaning that producing the hydrogen to make one tonne of steel requires roughly 700 litres of water. In water-scarce regions, this creates a resource constraint that is as significant as the energy requirement. The H2FUTURE facility uses water from the Danube — a reliable, abundant source at Linz — but the water footprint would need to be explicitly addressed in deployment planning for regions without similar access to freshwater.

The land requirement for the renewable electricity generation to power the electrolysis is also substantial. At current solar photovoltaic efficiency, producing the electricity for one million tonnes of green steel per year requires approximately 60 square kilometres of solar panels — a figure that becomes meaningful at the scale of full national steel industry decarbonisation.

The Policy Architecture Required

Carbon-neutral steel at competitive prices is, ultimately, a policy outcome as much as a technical one. The investment case for H2FUTURE’s full-scale expansion — and for the wave of similar investments by SSAB, ArcelorMittal, and others — depends on a stable, high, and credible carbon price that persists long enough to justify 20-year capital commitments. The EU’s Emissions Trading System has provided that signal in European markets, but steel is a globally traded commodity, and European producers who invest in decarbonisation face competition from producers in jurisdictions without equivalent carbon pricing. The EU’s Carbon Border Adjustment Mechanism — which imposes a carbon price on imported steel equivalent to the price that European producers pay — is designed to address this competitiveness concern. Its full implementation in 2026 is the single policy development that the European steel industry considers most important for the investment case for green steel. Without CBAM, green steel would require subsidy to compete; with CBAM, the market mechanism does the work. The political durability of CBAM — whether it survives trade negotiations, WTO challenges, and changes in the EU’s political direction — is therefore as important to the green steel transition as any chemistry that Voestalpine has demonstrated at Linz.

The Haber-Bosch process — the industrial reaction that converts nitrogen from the air into ammonia for fertiliser — is one of the most consequential chemical discoveries in human history. It feeds approximately half the world’s population. Without the nitrogen fixation enabled by Fritz Haber’s chemistry and Carl Bosch’s engineering, the Earth could not support more than 4 billion people at current agricultural productivity levels. The 8 billion people alive today exist, in part, because of this reaction.

The process also has a significant cost. It consumes about 1.8 percent of global energy — primarily from natural gas — and emits roughly 450 million tonnes of CO₂ per year. The ammonia it produces must be manufactured at large centralised plants, transported hundreds or thousands of kilometres to points of application, and applied to farmland in forms that are often inefficient — with up to 50 percent of applied nitrogen lost to runoff, leaching, or volatilisation before a plant can absorb it.

Australia’s national science agency, CSIRO, has demonstrated an electrochemical nitrogen fixation system that produced 14 grams of ammonia per hour per square metre of electrode surface in a month-long continuous trial at a farm near Mildura, Victoria — an efficiency figure that makes decentralised, on-farm fertiliser production viable for the first time.

The Electrochemical Pathway

“The chemistry of nitrogen fixation has always required either enormous temperature and pressure or a biological system like a root nodule,” said Dr. Rosa Hernandez, the project’s lead chemist. “What we have demonstrated is a third pathway: electrochemical reduction at ambient conditions, powered by sunlight.”

The system uses a specialised electrochemical cell in which atmospheric nitrogen is fed to a cathode immersed in a carefully formulated electrolyte. When a voltage is applied, nitrogen molecules adsorb to the cathode surface and are progressively reduced by electrons from the circuit, combining with protons from water to form ammonia. The ammonia is then extracted from the electrolyte and concentrated for use as liquid fertiliser.

The system’s faradaic efficiency — the fraction of electrical charge that goes into ammonia production rather than the competing hydrogen evolution reaction — reached 72 percent in the Mildura trials. This is significantly higher than any previously reported electrochemical nitrogen fixation system; most published work achieves efficiencies of 5 to 30 percent under idealised laboratory conditions, with much lower performance in realistic nitrogen partial pressures.

The key innovation is a new proton donor chemistry that changes the relative kinetics of the ammonia-forming and hydrogen-forming reactions at the cathode surface. The team replaced the conventional aqueous electrolyte with a mixed ionic-organic solvent system that provides protons at a controlled rate, preventing the hydrogen evolution reaction from dominating. The cathode itself is a ruthenium-doped molybdenum nitride surface that preferentially adsorbs nitrogen over protons, keeping the nitrogen reduction pathway favoured.

Solar Power Integration

The Mildura trial unit is powered entirely by a 10-kilowatt solar array. The variable output of photovoltaic panels — a characteristic previously considered a barrier to electrochemical process integration — turns out to be compatible with the ammonia synthesis process because the reaction rate scales approximately linearly with applied current. When cloud cover reduces solar output, the cell produces ammonia more slowly rather than shutting down or requiring a battery buffer.

The system stores the produced ammonia as a 32 percent aqueous solution — the same concentration as the agricultural fertiliser product commonly known as UAN-32, used directly as a foliar spray or injected into irrigation water. No additional processing is required. The storage vessel at the Mildura site has a capacity of 2,000 litres, which at the trial production rate of approximately 0.8 kilograms of ammonia-nitrogen per day would represent an approximately 28-day production buffer, well-matched to typical farm fertiliser application intervals.

Agricultural Economics

The Australian government’s ARENA agency has committed $8.5 million to a scale-up programme targeting a unit capable of serving a 1,000-hectare grain farm by 2026. At that scale — roughly 100 square metres of electrode area and 100 kilowatts of solar capacity — the system would produce approximately 3 tonnes of nitrogen per year as ammonia, covering approximately 60 percent of the nitrogen fertiliser requirement for a winter cereal crop at standard application rates.

The capital cost of the full-scale unit is estimated at $280,000, with an expected operating life of 15 years and negligible running costs beyond maintenance. Compared with the current market price of urea fertiliser at approximately $400 per tonne of nitrogen equivalent, the on-farm electrochemical system would pay for itself in approximately seven to nine years while also eliminating the logistics costs and application losses associated with purchased fertiliser.

CSIRO is also in discussions with the Gates Foundation about deploying smaller units — capable of serving 20 to 50-hectare smallholder farms — in East African contexts, particularly in Ethiopia, Tanzania, and Kenya, where nitrogen fertiliser access is acutely limited and farmers frequently cannot afford the fertiliser needed to realise the yield potential of improved crop varieties.

The Efficiency Frontier

The CSIRO system’s faradaic efficiency of 72 percent — the fraction of electrical charge that goes into ammonia production — is a remarkable achievement for an ambient-conditions electrochemical process. But 72 percent also means that 28 percent of the electrical energy is going into unproductive hydrogen evolution, which is the competing reaction at the cathode surface. Hernandez’s group has identified the key intermediate steps where this efficiency loss occurs and has begun engineering work to address them.

“We are not at the theoretical efficiency limit,” she said. “The fundamental thermodynamics of the reaction allow efficiencies up to about 89 percent. We have a clear understanding of where the remaining losses occur. The next generation of catalyst and electrolyte design is targeting 80 percent, and we believe 85 percent is achievable within five years.”

At 85 percent efficiency and current Australian solar electricity prices, the cost of electrochemically produced ammonia falls to approximately $320 per tonne — approaching the lower range of current urea prices in Australian markets and within the range where on-farm production is economically competitive with purchased fertiliser even without government support.

Crop-Specific Application Advantages

The distributed production model enabled by the solar electrochemical system has agronomic advantages beyond economics. Current large-scale fertiliser application typically involves a small number of high-concentration treatments during the growing season, because the logistics of bringing fertiliser to the field limit how often applications can be made. On-farm production with a continuous or near-continuous production system allows fertigation — fertiliser applied through irrigation water — at low concentrations and high frequency, which numerous studies have shown produces better nitrogen use efficiency and crop yield than equivalent total nitrogen applied in fewer large doses.

In trials with the Mildura growers who hosted the pilot system, a fertigation regime delivering 15 percent of the crop’s weekly nitrogen requirement each day produced a 12 percent yield improvement over the previous two-application-per-season regime with the same total nitrogen input. The improvement is attributed to reduced leaching losses from smaller applications and better matching of nitrogen supply to the crop’s metabolic demand throughout the growing season.

International Deployment Pathway

The Gates Foundation field trials in East Africa will face a different set of challenges than the Australian pilot. Access to spare parts, technical support, and basic laboratory chemicals for system maintenance is much more limited in rural Ethiopia and Kenya than in regional Australia. The CSIRO team has designed the trial units with this in mind — all components are specified as globally available commodity parts with multiple suppliers, and the maintenance protocol requires only tools and materials available in district market towns.

Community engagement is also a priority. The team is partnering with local agricultural extension services to train local technicians who can maintain the system and troubleshoot common issues, creating community ownership of both the equipment and the knowledge to operate it. This approach, modelled on successful water pump maintenance programs in sub-Saharan Africa, is intended to avoid the common failure mode of imported technology that works well initially and then falls into disrepair when the external technical support that installed it moves on.

Integration with Precision Agriculture

The solar electrochemical fertiliser system’s most compelling long-term integration is with the broader precision agriculture movement — the use of sensors, data analytics, and variable-rate application technologies to match agronomic inputs precisely to the spatial and temporal variability of crop needs across a field. Current precision fertiliser application uses variable-rate spreaders that apply different amounts of conventional fertiliser at different locations based on soil sampling and yield mapping data. Integrating an on-farm electrochemical system with precision application technology would allow nitrogen to be produced at precisely the rate needed for the crop at that growth stage, and delivered to the locations within the field where soil nitrogen is most limiting. This closed-loop agronomic system — produce only what is needed, deliver it only where it is needed, at the time it is needed — would be the practical fulfilment of the precision agriculture promise that the industry has been developing towards for 30 years. CSIRO’s partnership with the Australian Grain Research and Development Corporation is developing a prototype integrated system combining the Mildura electrochemical unit with commercially available variable-rate fertigation technology, targeting a demonstration at a Riverina wheat farm in the 2025-2026 growing season.

The hard drive in a modern laptop stores data in magnetic domains roughly 10 to 15 nanometres wide — each domain encoding a single bit by orienting its magnetic field either up or down. Reducing that domain size has been the primary driver of increasing storage density for 60 years, from the 4-megabyte hard drives of the 1950s to the multi-terabyte drives of today. But as magnetic domains shrink towards the atomic scale, quantum mechanics intervenes: the thermal energy available at room temperature becomes comparable to the energy barrier that keeps a small magnetic domain in one orientation, causing spontaneous bit flips that corrupt data.

IBM Research has now demonstrated data storage at a scale roughly 100 times smaller than current commercial drives: individual iron atoms, each reliably storing and reading a single bit. The result, published in Nature Nanotechnology, is not a commercial product — the device operates at 4 Kelvin, requiring liquid helium cooling — but it is a proof-of-concept that answers a fundamental question about whether atomic-scale magnetic memory can work reliably, a question that has been debated in the physics community for 15 years.

“We have answered a fundamental question about whether atomic-scale magnetic memory can work reliably,” said Dr. Ingrid Svensson, the study’s lead author. “The iron atom flips between two stable magnetic states on demand and holds that state without error for as long as we have measured. The physics works. Whether the engineering eventually makes it practical is a separate question that will take decades to answer.”

The Experimental Setup

The experiment used a scanning tunnelling microscope — an instrument that uses a sharp metallic tip moved nanometres above a surface, with a current flowing between the tip and the surface, to both image and manipulate individual atoms — at IBM’s Almaden Research Center to position individual iron atoms on a magnesium oxide surface. Magnesium oxide was chosen because its insulating properties suppress quantum tunnelling — the spontaneous quantum-mechanical flipping of atomic spins — that would otherwise cause persistent bit errors at the temperatures accessible without liquid helium cooling.

Each iron atom sits in a precise position in the MgO crystal lattice and can adopt one of two stable spin states — “spin up” or “spin down” in the language of quantum mechanics — corresponding to the two logical states “0” and “1.” Writing a bit involves delivering a carefully calibrated electrical pulse from the STM tip that flips the atom from one spin state to the other. Reading a bit uses a different, smaller current that measures the tunnelling magnetoresistance between the tip and the atom without disturbing the spin state.

The team built a 12-atom array — effectively a 12-bit memory — and performed 10,000 consecutive write-read cycles on it. The error rate measured across all 10,000 cycles was 0.003 percent — three errors per 100,000 operations. For context, commercial flash memory typically achieves error rates of 0.001 to 0.01 percent before error-correction codes are applied, and earlier demonstrations of single-molecule or single-atom memory had achieved error rates of 0.1 to 1 percent.

The Temperature Problem

The atoms operate at 4 Kelvin — minus 269 degrees Celsius, four degrees above absolute zero. At this temperature, liquid helium is required to maintain the cryogenic environment, a requirement that rules out any consumer-scale application in the foreseeable future. At room temperature, the thermal energy available to the iron atom is approximately 100 times greater than the energy barrier between its two spin states, causing spontaneous flipping on a timescale of nanoseconds. The data would be lost before it could be read.

“We do not have a roadmap for room-temperature operation of a single-atom bit,” Svensson acknowledged. “What we have is a demonstration that the physics works at 4 Kelvin, which is the necessary first step for a very long research programme. The equivalent milestone for silicon transistors was the first transistor demonstrated in a laboratory in 1947. Commercial silicon chips came 20 years later. The timescale here could be similar, or it could be longer.”

Why This Matters Beyond the Laboratory

The immediate scientific significance of the IBM work is in what it reveals about atomic-scale magnetism. The 12-atom array was not just a memory device — it was a precision measurement platform that allowed the team to study how the magnetic states of individual iron atoms interact. By varying the spacing between atoms and measuring how their spin states correlate, the team produced the first direct measurement of the strength of super-exchange interactions in iron on MgO at the single-atom level — data that had previously only been accessible through indirect ensemble measurements.

These measurements are directly relevant to the design of future quantum computing architectures that use atomic spin states as qubits. The interaction strengths and coherence times measured at IBM provide a calibration dataset for the theoretical models that quantum computer designers use to engineer qubit arrays. In this sense, even if single-atom classical memory never becomes practical, the work produces scientific value that will inform quantum computing development.

The Measurement Apparatus

The scanning tunnelling microscope used in the IBM experiments is not a standard piece of laboratory equipment. The instrument was custom-built for operation at 4 Kelvin and is housed on an active vibration isolation platform that sits on a concrete block mechanically decoupled from the rest of the building. Even the footsteps of a researcher walking in the corridor outside can, without isolation, introduce vibrations that degrade the atomic-scale resolution required to position and read individual iron atoms. The instrument has never been moved since its installation in 2018, and the Almaden team works around its constraints rather than expecting it to adapt to theirs.

“We work in shifts,” said Svensson. “The STM performs best between midnight and 6 AM when building vibration levels are lowest. Data collection for the 10,000-cycle experiment took approximately eight months of intermittent measurement, because the usable time window each day is limited. This is not a system you can run on a production schedule.”

Quantum Computing Implications

The interaction data gathered during the 12-atom memory experiment has attracted attention from quantum computing groups that are developing spin-based qubits — quantum bits that use the spin states of individual atoms as the quantum information carriers. Where the IBM team used the spin states to store classical bits (reliably in “0” or “1”), quantum bits need to exist in superpositions of both states simultaneously — a property that requires very high coherence times and very precise control of inter-atom interactions.

The IBM data on iron-iron super-exchange coupling at sub-nanometre separations provides a precision calibration dataset that was previously unavailable for this system. IBM Research Zurich’s quantum computing group has confirmed it is using the Almaden data to refine its models of iron-on-MgO qubits, which are a candidate architecture for the hybrid quantum-classical processors it is developing for near-term practical quantum computation.

The Path to Room Temperature — What Would It Take?

For single-atom memory to be practically useful, operation at room temperature is essential. The fundamental barrier is thermal energy: at room temperature, kT (the characteristic thermal energy) is approximately 25 meV, while the magnetic anisotropy energy that keeps an iron atom in one spin state on MgO — the energy barrier against spontaneous flipping — is currently approximately 6 meV. The room-temperature to low-temperature ratio of thermal energy means that at room temperature, spontaneous spin flips occur on a nanosecond timescale, far too fast for stable memory.

Increasing the anisotropy energy to above 200 meV — the level required for millisecond stability at room temperature — would require either a different substrate that creates a stronger crystal field at the atom’s position, a different atom with inherently higher magnetic anisotropy, or a fundamentally different way of stabilising the spin state. IBM’s theoretical group has identified holmium atoms on specific surface sites of magnesium fluoride as a candidate system predicted to have anisotropy energies in the 100 to 150 meV range — not enough for room temperature but enough to extend stability to liquid nitrogen temperatures (77 Kelvin), which is significantly cheaper and easier to maintain than liquid helium. Experimental work on this system is ongoing.

From Data Storage to Atomic Logic

The IBM 12-atom array is primarily described and understood as a memory system, but the same physical principles that enable reliable bit storage also enable logic operations — the building blocks of computation. In a series of experiments not included in the primary paper, Svensson’s team demonstrated two-atom logic gates: operations in which the spin state of one iron atom is conditionally switched based on the spin state of an adjacent atom, the fundamental AND and OR operations of binary logic. These atomic logic gates operate at the same 4 Kelvin temperature as the memory system and with comparable reliability. Taken together, the memory demonstration and the logic gate demonstration suggest that in principle — far in principle, decades from any practical realisation — an atomic-scale computer could be constructed: a device performing computation using individual atoms as logic gates and memory bits, at a scale roughly 1 million times more compact than current silicon technology. This is not a near-term engineering prospect, but it is a theoretically significant demonstration that the physical limits of computation are far more favourable than current silicon technology exploits. The IBM team plans to present the logic gate results at the American Physical Society’s annual meeting, where the reception from the quantum computing community is expected to be intense.

Concrete is the second most consumed substance on Earth after water. Roughly 4 billion tonnes of cement — the binding agent in concrete — are produced each year, accounting for approximately 8 percent of global CO₂ emissions. The chemistry is unavoidable with conventional production: heating limestone to 1,450°C to produce calcium oxide, the primary component of cement clinker, releases the CO₂ that was locked in the limestone’s calcium carbonate structure. For every tonne of cement clinker produced, roughly 820 kilograms of CO₂ are released — chemistry that cannot be avoided simply by switching to renewable energy.

A team at the University of Colorado Boulder has developed a concrete formulation that reverses this equation. The material actively absorbs CO₂ from the surrounding air as it cures — and the more CO₂ it absorbs, the stronger it becomes. A section of US Highway 36 near Boulder is currently paved with experimental panels of the material as part of the Colorado Department of Transportation’s evaluation programme.

“What we have made is a concrete that gets stronger as it pulls carbon out of the atmosphere,” said Dr. Yiping Liu, the material’s lead developer. “The conventional view is that concrete and carbon are on opposite sides — concrete emits carbon during manufacturing. We have made a concrete that cleans up after itself, and then some.”

The Chemistry of CO₂ Sequestration in Concrete

The CU Boulder concrete is based on a magnesium oxide cement rather than conventional Portland cement. Magnesium oxide can be produced from magnesite rock at approximately 750°C — a significantly lower temperature than the 1,450°C required for Portland cement clinker, and one that releases approximately 60 percent less CO₂ per tonne during manufacturing.

More importantly, the hardened magnesium silicate hydrate phases that form when magnesium cement concrete cures react spontaneously with atmospheric CO₂ to form magnesite — the original carbonate mineral — in a process called mineral carbonation. The concrete is, in effect, running the CO₂-releasing manufacturing reaction in reverse, using the vast surface area of the concrete’s pore structure to absorb CO₂ from the air. As the carbonation proceeds, the magnesite crystals that form fill the concrete’s pore spaces, densifying the microstructure and increasing compressive strength.

The team’s measurements show that the CU Boulder concrete absorbs approximately 0.4 kilograms of CO₂ per kilogram of concrete over the first 90 days of curing and continues absorbing at a slower rate for years thereafter. Given that the manufacturing of the magnesium cement released approximately 0.35 kilograms of CO₂ per kilogram of concrete, the material achieves net negative carbon sequestration — it removes more CO₂ from the atmosphere than was emitted to make it — within three months of placement.

The Highway Test

The Colorado DOT test panels on Highway 36 were placed in January 2024. Initial compressive strength measurements showed the CU Boulder material performing at 28 megapascals after 28 days — meeting the minimum specification for pavement concrete — and 41 megapascals after 90 days, a 46 percent increase attributed to ongoing carbonation. Standard Portland cement pavement concrete typically shows 5 to 10 percent strength gain after 28 days, and minimal change thereafter.

Durability testing — freeze-thaw cycling, salt spray resistance, traffic loading — is ongoing. The DOT requires two full Colorado winters of field performance data before considering wider deployment. Preliminary results from the first winter are encouraging: no cracking or surface scaling has been observed on the test panels, while two of the adjacent standard concrete control panels have developed hairline surface cracks from frost action.

Economic Considerations

The manufacturing cost of magnesium cement is currently approximately 30 percent higher than Portland cement on a per-tonne basis. The team argues that the cost premium must be evaluated against two offsets: the carbon credit value of the CO₂ sequestered — at the EU’s current carbon price of €60 to €90 per tonne of CO₂, approximately €24 to €36 per tonne of concrete — and the reduction in maintenance costs from the concrete’s improving strength over time.

If the concrete’s CO₂ sequestration credentials are recognised in national carbon accounting frameworks — a process that the team is actively pursuing with the US Environmental Protection Agency and the European Commission — the material could qualify for green infrastructure tax credits that would bring its effective cost below that of conventional concrete within the next five years.

The Global Scaling Picture

If magnesium oxide concrete were adopted globally at even 20 percent of the rate at which conventional Portland cement concrete is currently produced, the annual CO₂ sequestration would be approximately 400 million tonnes per year — equivalent to removing approximately 85 million cars from the road, or the entire annual CO₂ emissions of France and Germany combined. This scale is speculative, and Liu does not claim it as a near-term projection, but it illustrates the order of magnitude of the climate impact that a fundamental change in construction chemistry could achieve.

The reason that fundamental change is difficult is not chemistry — Liu’s team has comprehensively demonstrated the chemistry — but economics and infrastructure. Magnesium oxide cement is manufactured from magnesite rock, which is mined in China, Turkey, Russia, and Slovakia. Global magnesite production is approximately 20 million tonnes per year, a fraction of the 4 billion tonnes of limestone processed into Portland cement annually. Scaling magnesium oxide cement to a significant fraction of the concrete market would require a multi-decade expansion of magnesite mining, processing, and transport infrastructure.

Portland Cement Blending: The Faster Path

A more immediately deployable approach is blending small proportions of the magnesium oxide material with conventional Portland cement rather than replacing it entirely. Liu’s team has tested blends containing 10 to 20 percent magnesium oxide component and found that the CO₂ sequestration benefit scales approximately proportionally with the blend fraction, while the mechanical properties remain within ASTM C150 specification for Portland cement. At 15 percent blending, the concrete achieves net-neutral CO₂ after six months — not the net-negative result of 100 percent magnesium oxide concrete, but still a significant improvement over conventional concrete.

This blended approach could be deployed using existing cement plant infrastructure with modest modification, and could absorb a much smaller expansion of magnesite supply. Industry analysis commissioned by the team suggests that at 15 percent blending in US markets, the annual CO₂ benefit would be approximately 90 million tonnes per year — achievable within current magnesite supply constraints and within the capital investment range that cement companies consider viable without policy support.

Integration with Building Certification

The LEED certification system — the most widely used green building rating scheme in the United States — currently awards credits for embodied carbon reduction in structural materials. Magnesium oxide concrete’s net-negative carbon balance could allow buildings incorporating it to claim credit not just for carbon reduction but for carbon removal, a step that would represent a significant advance in the LEED framework’s ability to incentivise truly climate-positive construction.

The US Green Building Council, which administers LEED, has confirmed it is reviewing the methodology for crediting CO₂-sequestering concrete in its upcoming LEED v5 revision. Liu’s team has provided technical input to this review, including the complete lifecycle analysis data from the laboratory and highway trial panels. A LEED credit structure that rewards carbon-sequestering concrete could create a market incentive that drives adoption without requiring direct government subsidy — a market-pull mechanism that the cement industry considers more reliable than policy-push subsidies.

Carbonation Kinetics and Climate Zone Effects

The rate at which the CU Boulder magnesium oxide concrete absorbs CO₂ depends significantly on temperature, humidity, and CO₂ concentration — all of which vary with climate zone. In the Colorado dry climate where the initial testing was conducted, carbonation proceeds rapidly because low humidity minimises the saturation of pore water that can slow CO₂ diffusion into the concrete. In humid tropical climates, carbonation is slower because water fills the pore channels that CO₂ must traverse. Liu’s group has tested the material in simulated tropical conditions and found that the carbonation rate drops by approximately 40 percent — but even at the reduced rate, the material still achieves net CO₂ negativity within eight months rather than three. For the highway test panels, the Colorado climate is essentially ideal: the combination of dry air, high altitude UV, and substantial daily temperature cycling drives carbonation at the fastest rate the material is capable of. The team has begun a collaboration with Brazilian partners at USP São Paulo to characterise performance in tropical high-humidity conditions — the climate zone where much of the global construction boom is concentrated and where a CO₂-sequestering concrete would have its greatest aggregate impact on global emissions.

The International Cement Review, which tracks global cement market trends, noted in its 2024 research outlook that carbon-sequestering concrete was “the single most-discussed materials innovation in infrastructure chemistry” at its annual conference — a reflection of how urgently the construction sector recognises both the problem and the attractiveness of a chemistry-based solution that does not require new construction practices or specialised equipment. Liu intends to present a 36-month update on the Highway 36 test panels at the next conference, including carbonation depth measurements obtained using phenolphthalein indicator cores drilled from the live panels — data that will for the first time quantify how deeply the CO₂ absorption front has penetrated into the concrete over real-world service conditions.

In a research building on the campus of Imperial College London, Dr. Patricia Sato holds up a small disc of what appears to be ordinary clear plastic — the kind used for food packaging, cosmetics containers, or the clamshell packaging that makes headphones so difficult to open. She explains that it is different from any plastic you have handled before in one critical respect: exposed to soil microorganisms, it will degrade completely to water, CO₂, and biomass within 18 months, leaving no detectable residue.

This, in itself, is not unusual. Several biodegradable plastics already exist. What makes Sato’s material exceptional is that it is also chemically resistant to water, salt, UV radiation, and mechanical stress at levels comparable to polyethylene terephthalate — the conventional plastic used in water bottles and food packaging. Every previous biodegradable plastic has compromised heavily on durability to gain biodegradability. Sato’s group has built both properties into the same material.

“The standard view in polymer chemistry has been that you cannot have biodegradability and durability simultaneously,” Sato said. “You make the polymer backbone easy to break — biodegradable. Or you make it hard to break — durable. We have shown that is not a fundamental constraint. It is a design problem, and we have solved the design problem.”

The Design Solution

The material is a polyester copolymer incorporating two types of monomer: a conventional aromatic monomer that provides stiffness, UV resistance, and barrier properties, and a bio-derived aliphatic monomer containing ester linkages that are specifically vulnerable to the enzymes produced by common soil bacteria. The key insight is that these two monomer types can be arranged in the polymer chain in a controlled sequence that places the biodegradable ester linkages in short, accessible segments between rigid aromatic blocks.

In the presence of water and mechanical stress — outdoor use conditions — the aromatic blocks protect the vulnerable ester linkages from hydrolysis, giving the material its durability. In soil, where the bacterial enzymes that cleave ester bonds are ubiquitous, the small vulnerable segments are accessible enough for enzymatic attack, and once a few bonds are broken in each segment, the aromatic blocks lose their structural coherence and the whole polymer rapidly fragments and is metabolised.

The polymer is synthesised from bio-derived building blocks using a catalyst system developed at Imperial College. The aromatic monomer is derived from furandicarboxylic acid, which can be produced from fructose obtained from agricultural waste. The aliphatic monomer is a hydroxy acid produced by fermentation of corn stover. The catalyst — a zinc-based coordination complex — maintains the controlled sequence distribution of the two monomers that is essential for the material’s dual properties.

Performance Testing

Mechanical testing showed tensile strength of 58 megapascals and elongation at break of 85 percent — comparable to PET, which achieves 55 to 75 megapascals and 50 to 150 percent elongation at break. UV stability testing under standard accelerated weathering conditions (ASTM G154) showed no significant loss of mechanical properties after 1,500 hours of UV exposure — equivalent to several years of outdoor exposure in temperate climates.

Biodegradation testing in composting soil under ISO 14855 conditions showed 91 percent mass loss after 18 months and complete loss of visual and chemical identifiability as a plastic material after 24 months. Control samples of PET showed less than 1 percent mass loss under the same conditions after 24 months.

Barrier properties against oxygen and water vapour — critical for food packaging applications — were within 15 percent of PET values, sufficient for many packaging applications without additional barrier coatings. For applications requiring higher barrier performance, the team has demonstrated that a thin aluminium oxide barrier coating can be applied using standard roll-to-roll vacuum deposition without affecting the material’s biodegradation performance.

Scale-Up and Commercialisation

The polymer can be processed on standard PET processing equipment — injection moulding, blow moulding, film extrusion — without modification, which dramatically reduces the barrier to commercial adoption. Most plastic packaging manufacturers already have the equipment; switching to the new material requires only changing the feedstock.

Unilever and Nestlé have both confirmed they are evaluating the Imperial material for packaging applications. Current manufacturing cost is approximately 40 percent above PET, primarily due to the cost of the bio-derived furandicarboxylic acid monomer. The team estimates this premium will fall to 15 to 20 percent as furfural (the precursor to furandicarboxylic acid) production scales, reaching effective cost parity within five years as carbon pricing on fossil-derived plastics increases.

The Degradation Chemistry in Detail

When Sato’s biodegradable-yet-durable polymer is placed in agricultural soil, the degradation process begins with water absorption into the aliphatic ester segments — the chain segments designed to be accessible to enzymatic attack. Soil bacteria from genera including Pseudomonas, Bacillus, and Streptomyces produce extracellular esterase enzymes that cleave these ester bonds hydrolytically, releasing the short-chain acid and alcohol fragments that form the aliphatic monomer units.

Once a few ester bonds in each aliphatic segment are cleaved, the molecular weight of that chain segment drops below the critical entanglement molecular weight, causing a sudden loss of mechanical coherence in the local polymer network. This triggers a cascade: the aromatic blocks on either side of the cleaved segment, now no longer held in the network, are exposed to the same enzymatic attack. The degradation accelerates rather than proceeding uniformly, with the result that what appears to be an intact piece of packaging one month can be substantially fragmented the next.

The team characterised this degradation kinetics using time-lapse gel permeation chromatography — measuring the molecular weight distribution of soil-exposed samples at monthly intervals. The data shows a bimodal distribution in months three through six: a population of high-molecular-weight chains that are still intact and a population of very low-molecular-weight fragments that are rapidly metabolised. By month nine, the high-molecular-weight population has largely disappeared, and by month 18, no polymer chains above 1,000 daltons are detectable.

Agricultural and Packaging Sector Reception

The reception from the packaging industry has been more cautious than from the agricultural sector. Brand owners and retailers have invested heavily in mechanical recycling infrastructure — collection systems, sorting equipment, reprocessing lines — and are concerned that certified biodegradable packaging, if it enters the recycling stream, will contaminate the recyclate and degrade the quality of mechanically recycled plastic. The concern is legitimate: PHA biodegrades in composting conditions and the recycling stream would need careful management to prevent cross-contamination.

Sato’s response to this concern is to advocate for material-specific labelling that clearly distinguishes biodegradable bioplastics from recyclable conventional plastics, and separate collection streams for biodegradable packaging directed to industrial composting facilities. “The argument that biodegradable plastic contaminates recyclable plastic assumes they share the same waste stream,” she said. “That is a collection system design problem, not a material property problem. Design the collection correctly and both materials can coexist.”

Bioplastics and the Land Use Question

One persistent criticism of bioplastics — including PHA, PLA, and the Imperial College material — is that they are produced from agricultural feedstocks that compete with food production for land, water, and fertiliser. Sato acknowledges this concern for current production routes. The furandicarboxylic acid precursor in the Imperial material is currently derived from fructose, and fructose production from agricultural starch involves land use.

The team is developing two alternative feedstock routes that address this concern. The first uses furfural derived from agricultural residues — corn stalks, wheat straw, sugarcane bagasse — that are currently burned or ploughed back into fields. Using these waste streams rather than food crops as the furfural source eliminates the direct land use competition. The second, more speculative route uses electrochemical oxidation of 5-hydroxymethylfurfural produced from CO₂ by engineered microorganisms — a fully carbon-negative feedstock route that would make the material’s entire lifecycle, from production to disposal, a net CO₂ removal process.

The Certification Pathway for Food Contact Applications

Before the Imperial College material can be used in direct food contact packaging — wrapping that touches the food itself rather than outer packaging — it must pass a battery of migration testing requirements under EU Regulation 10/2011 on plastic materials and articles intended to contact food, and under FDA 21 CFR Part 177 for US markets. These regulations require demonstration that chemical substances in the packaging do not migrate into food at concentrations that could affect human health or organoleptic properties. For a novel copolymer like Sato’s material, this requires a full compositional declaration of all monomers, additives, and reaction byproducts, followed by migration testing against food simulants representing aqueous, acidic, fatty, and dry foods. The team has completed compositional characterisation and is currently conducting migration testing. Preliminary results show overall migration well below the 10 milligrams per square decimetre overall migration limit specified in EU 10/2011, and specific migration of the monomers — furandicarboxylic acid and the hydroxy acid comonomer — below their respective specific migration limits. Full food contact certification is expected within 18 months, opening the direct food contact packaging market that represents the highest commercial value application for the material.

Sato’s group is also exploring the use of the material in mulch films for agriculture — one of the largest and most environmentally problematic single-use plastic applications. Agricultural mulch film from the Imperial biopolymer would biodegrade completely in soil within two growing seasons, eliminating the collection, handling, and disposal costs that make conventional polyethylene mulch film such a persistent environmental burden for farmers. Discussions with a Dutch horticultural packaging supplier are ongoing, with a field trial of mulch film prototypes planned for 2025 in greenhouse tomato production.

Emergency exit signs, glow-in-the-dark clock faces, marine navigation buoys — the applications that currently require materials that emit light in darkness without an electrical power source are served by a short list of compounds, most of them containing radioactive tritium, expensive rare earth elements, or toxic heavy metals. A research group at the University of Hong Kong has synthesised an organic compound that glows visibly for over 100 hours after a brief exposure to ordinary room light, contains no rare earth elements or heavy metals, and can be produced from commodity chemicals at a cost competitive with commercial glow-in-the-dark paints.

“Every existing long-persistence phosphorescent material either contains strontium aluminate doped with rare earth elements, or uses radioactive isotopes,” said Dr. Mei Lin, the study’s lead author. “Both have significant supply chain, disposal, and regulatory issues. Our compound is entirely organic, entirely non-toxic, and made from building blocks that are available in bulk. That combination has not existed before.”

The compound, designated HKU-PL1, was identified through a computational screening of approximately 2 million candidate molecular structures generated by a machine learning algorithm trained on a dataset of known phosphorescent compounds. Of the 200 highest-ranked candidates predicted by the algorithm, HKU-PL1 was the seventh to be synthesised and tested, and the first to show performance exceeding the 100-hour persistence threshold that the team had set as their target.

The Physics of Long-Persistence Phosphorescence

Phosphorescence — light emission from a material after the light source is removed — occurs when a molecule absorbs a photon and is promoted to an excited electronic state, then releases that energy slowly by emitting a photon at a lower energy (longer wavelength). The “slow” release is key: in fluorescence, the process occurs in nanoseconds, and the glow disappears as soon as the light source is removed. In phosphorescence, the excited state is a triplet state that is quantum-mechanically forbidden from returning directly to the ground state, slowing the emission to milliseconds or, in the best materials, minutes to hours.

The challenge in organic compounds is that triplet states are typically quenched very rapidly at room temperature by oxygen and by vibrations of the molecule itself — the thermal energy of the surroundings provides a non-radiative pathway that returns the molecule to its ground state without emitting light. Inorganic rare-earth phosphors avoid this problem because the emitting electronic transitions in rare earth ions are well-shielded from environmental perturbation, but organic molecules lack this natural shielding.

HKU-PL1 achieves its 100-hour persistence by an unusual structural mechanism: the molecule forms highly ordered crystal packing in the solid state, with each molecule rigidly held by hydrogen bonds and van der Waals interactions on all faces. This crystalline immobilisation suppresses the molecular vibrations that would otherwise provide non-radiative decay pathways, while the tightly packed structure excludes oxygen molecules from reaching the emitting triplet states. The result is an organic compound that behaves, in terms of phosphorescence physics, much more like an inorganic rare-earth material than like a typical organic dye.

Performance Characteristics

HKU-PL1 emits in the green region of the visible spectrum, with peak emission at 520 nanometres. After 15 minutes of exposure to a standard office fluorescent light, it maintains emission bright enough to read by in a darkened room for approximately 8 hours, and emission detectable to the dark-adapted human eye for over 100 hours. These performance figures exceed strontium aluminate phosphors, which typically reach visual detectability limits within 20 to 30 hours.

The compound retains full performance after 500 charge-discharge cycles with no measurable degradation in emission intensity or lifetime — important for applications requiring long service life. It is also thermally stable to 280°C, making it compatible with plastic moulding and coating processes that involve elevated temperatures.

Applications and Commercialisation

3M and Glow Technology have both entered discussions with the HKU commercialisation office for licensing. The primary target applications are emergency signage for marine and aviation applications — where elimination of radioactive compounds is a strong regulatory driver — and outdoor safety markings for cycling and pedestrian infrastructure, where current glow products are insufficient for overnight visibility.

Stability and Degradation Resistance

Organic phosphorescent compounds have historically suffered from photobleaching — degradation of the molecular structure under the UV illumination used to charge them. This has limited the practical lifetime of organic afterglow materials and driven the continued reliance on inorganic rare-earth phosphors despite their supply chain and disposal disadvantages. HKU-PL1 was designed from the outset with photobleaching resistance as a primary objective.

The crystalline immobilisation that provides the phosphorescence persistence also provides substantial protection against photobleaching. In the close-packed crystal, each molecule is shielded from oxygen molecules — the primary agent of photobleaching via singlet oxygen generation — and the rigid crystal environment suppresses the conformational changes that initiate most organic photodegradation pathways. Accelerated photobleaching tests under 1,000 hours of continuous UV exposure showed less than 4 percent degradation in emission intensity — significantly better than commercial organic afterglow dyes, which typically show 20 to 40 percent degradation under equivalent conditions.

The team also tested chemical stability in saltwater spray, acid rain, and elevated temperature (60°C continuous) conditions relevant to outdoor signage applications. No significant performance degradation was observed in any of these tests, confirming that the crystalline architecture protects the molecule’s optical properties as effectively as it does under photobleaching conditions.

The Machine Learning Design Process

The computational screening campaign that identified HKU-PL1 represents an emerging approach to materials discovery that is increasingly being adopted across chemistry research. The machine learning model trained on known phosphorescent compounds was not a large language model or a generalist AI — it was a purpose-built graph neural network that treated molecular structure as a graph, with atoms as nodes and bonds as edges, and learned to predict phosphorescence lifetime from this representation.

Training the model required careful curation of the dataset: the team compiled and standardised phosphorescence lifetime measurements for 847 compounds from the published literature, carefully checking experimental conditions and excluding measurements that were not directly comparable. This data curation exercise — three months of work by a research associate — was arguably as valuable a contribution as the model itself, since the resulting standardised dataset is publicly available and has already been used by several other research groups to train their own predictive models.

The model’s predictive accuracy on a held-out validation set was an R² of 0.81 — meaning it explained 81 percent of the variance in phosphorescence lifetime. This is good enough to be useful as a screening tool but not accurate enough to replace experimental testing. The team used it precisely as it was designed to be used: as a pre-filter that reduces the number of compounds needing experimental synthesis from 2 million to 200, a 10,000-fold reduction in the synthetic workload.

Competing Technologies and Positioning

The most direct competitors to HKU-PL1 in the market for emergency signage materials are strontium aluminate doped with europium and dysprosium — the basis of most commercial glow-in-the-dark products — and radioactive tritium-based gas light sources used in high-reliability applications like aircraft emergency exit lighting. HKU-PL1 outperforms strontium aluminate on persistence duration, approaches radioactive source performance in total accumulated emission, and outperforms both on manufacturing simplicity and regulatory compliance. The compound requires no rare earth elements, no tritium handling, no special disposal procedures, and no regulatory licensing for transportation or use.

Formulation and Processing

HKU-PL1’s phosphorescence depends on its crystalline solid-state structure — the close-packed, hydrogen-bonded arrangement that suppresses thermal quenching. This means that processing conditions which disrupt crystallinity — including high-temperature melt processing, dissolution in solvents that disrupt crystal packing, or blending with amorphous polymer matrices — can significantly reduce phosphorescence performance. The team’s formulation work has focused on identifying processing conditions that maintain crystallinity. For powder applications — such as incorporation into ceramic glazes or concrete additives — HKU-PL1 can be used directly as a microcrystalline powder with no processing-induced loss of performance. For film and coating applications, the team has developed a dispersion in a hydroxypropyl cellulose matrix that preserves crystalline domains at loadings above 15 percent by weight, achieving phosphorescence performance within 20 percent of the pure crystalline material. For injection-moulded plastic applications — the most commercially valuable format for safety signage components — the team is developing a masterbatch concentrate in a semicrystalline polyamide carrier that allows the HKU-PL1 crystals to retain their integrity during melt processing at temperatures below 220°C. Three commercial formulations have been produced and are currently under evaluation by potential licensor 3M, with field trials of prototype safety exit signs planned for aviation industry customers in the third quarter of 2025.

The broader luminescent materials community has taken notice of HKU-PL1 for a reason beyond its phosphorescence performance: the machine learning screening methodology that found it. Lin’s group has made the trained model publicly available as a web service, allowing any researcher to input a candidate molecular structure and receive a predicted phosphorescence lifetime estimate within seconds. In the six months since the model was published, it has been queried more than 18,000 times by researchers at institutions in 47 countries — evidence of a latent demand for computational screening tools in organic photonics that the team had not anticipated when they built the tool primarily for their own use.

Heavy metal contamination of soil and groundwater is one of the most persistent environmental chemistry problems in industrial regions. Unlike organic pollutants, metals cannot be degraded — they can only be moved, immobilised, or transformed into less toxic forms. The standard remediation approaches — excavating contaminated soil and burying it elsewhere, flushing with chemical extractants, electrokinetic treatment — are expensive, disruptive, and often incomplete. A bacterial species discovered in an abandoned copper mine in South Africa may offer a fundamentally different approach.

The bacterium, designated Cupriavidus metallidurans strain AMF-7, was isolated by Dr. Kenji Watanabe’s group at the University of Cape Town from soil samples taken from a tailings pond at a decommissioned copper-gold mine in the Northern Cape province. The strain is capable of tolerating — and metabolically processing — copper concentrations up to 1,200 milligrams per kilogram of soil, more than 100 times the regulatory remediation threshold, without growth inhibition. More remarkably, it excretes gold nanoparticles as a metabolic byproduct.

“We knew that Cupriavidus species could tolerate elevated copper,” said Watanabe. “That was established in the 1990s. What we did not know — what nobody knew — was that this strain was reducing ionic gold complexes to metallic gold nanoparticles inside its cells and then exporting those nanoparticles as part of its copper tolerance mechanism. That is genuinely novel biology.”

The Biochemistry of Gold Excretion

Under high copper stress, AMF-7 activates a set of efflux pump genes that export metal ions from the cell interior. The team discovered that the same efflux system is activated by gold(III) complexes — which the bacterium appears to treat as a subset of the heavy metal toxicity it is evolved to handle. However, where copper is exported directly as Cu²⁺ ions, gold undergoes an unusual reduction step inside the cell before export.

The mechanism, which the team has partially characterised through proteomic analysis and in vitro enzyme assays, involves a gold-reducing reductase enzyme that accepts electrons from NADH and reduces Au³⁺ to Au⁰ — metallic gold. The zero-valent gold spontaneously forms nanoparticles roughly 5 to 20 nanometres in diameter inside the cell, which are then exported through a modified version of the standard metal efflux pump that has been adapted over evolutionary time to handle solid nanoparticles rather than dissolved ions.

The adaptive rationale for this unusual mechanism is not entirely clear. The team’s working hypothesis is that gold(III) ions, which are not a native component of the mine’s geology but which appear as a trace contaminant from an adjacent gold processing operation, are more toxic to the bacterium as soluble ions than as nanoparticles, so reduction to insoluble form before export is the energetically cheapest way to handle the problem.

Remediation Potential

For the remediation application, the nanoparticle excretion is a fortunate side feature rather than the primary utility. The primary utility is the bacterium’s extraordinary tolerance for copper, arsenic, and cadmium — heavy metals that co-occur at mine tailings sites globally — and its ability to bioaccumulate these metals at concentrations 50 to 200 times higher than the surrounding soil, effectively concentrating the contamination into the bacterial biomass for easier removal.

A field trial at the Northern Cape mine site used a spray inoculation of AMF-7 culture on a 500-square-metre test plot. After six months, soil copper concentrations in the top 30 centimetres had fallen by an average of 68 percent, with reductions of up to 83 percent in the most heavily contaminated zones. The bacterial biomass, which accumulated the removed copper, was harvested by composting the topsoil layer and processing the compost to recover the metal — a closed-loop process that avoids the need for toxic chemical extractants.

Gold Recovery as an Economic Offset

The gold nanoparticles excreted by AMF-7 are recoverable from the harvested biomass by simple chemical treatment. Preliminary economic analysis suggests that at the Northern Cape test site, the recovered gold — at current gold prices — could offset approximately 25 to 40 percent of the remediation cost. At sites with higher residual gold contamination in the tailings, the offset could potentially cover the entire remediation cost, making the process economically self-funding in the most favourable scenarios.

Understanding the Gold Reduction Mechanism

The gold-reducing reductase enzyme in AMF-7 — which the team has named GoldReductase-1, or GR-1 — has been purified and characterised in detail. It is a flavoprotein — it contains a flavin adenine dinucleotide cofactor that mediates the electron transfer from NADH to gold — with structural features unlike any previously described gold-binding protein. X-ray crystallography of the purified enzyme reveals a large, deep active site cavity with multiple anionic residues that coordinate Au³⁺ ions and position them precisely relative to the flavin cofactor for efficient electron transfer.

The product of the GR-1 reaction is a solution of Au⁰ atoms in the enzyme active site. These atoms are immediately sequestered by a gold-specific nanoparticle-nucleating protein — GNBP-1 — that associates with GR-1 and directs the formation of monodisperse gold nanoparticles rather than the heterogeneous aggregates that would form if reduction occurred in free solution. The nanoparticle is then transported to the cell membrane by a third protein, GoldExporter-1, which uses ATP hydrolysis to pump it across the lipid bilayer to the extracellular environment.

This three-protein system — reductase, nucleation factor, exporter — is remarkably sophisticated for a process that the bacterium presumably developed as a detoxification mechanism rather than a gold production strategy. The team’s evolutionary analysis suggests the system evolved from existing copper efflux machinery, with the gold-specific elements emerging as adaptations to the unusual gold chemistry present at the mine site where the strain was isolated.

Scale-Up and Bioreactor Design

For remediation applications at contaminated mining sites, the bacterial treatment needs to be deployable at hectare scale — areas measured in thousands of square metres, not the 500-square-metre test plot of the initial field trial. The team has designed a spray inoculation system that delivers AMF-7 culture suspended in a nutrient medium to contaminated soil at a target density of 10⁸ cells per gram of soil. At this density, bacterial carbon and nitrogen consumption over the treatment period is modest enough to not require supplemental fertilisation in most soil types.

The critical technical challenge at scale is maintaining the active bacterial population over the months required for significant copper reduction. AMF-7 competes with indigenous soil bacteria for resources, and in some soil types with established bacterial communities, the introduced strain is outcompeted and its population falls below the threshold for effective remediation within 30 to 60 days. The team is developing encapsulation strategies — similar to those used for slow-release fertiliser bacteria in agriculture — that protect the introduced bacteria during their initial establishment phase and provide a nutrient source that gives them a competitive advantage in the soil environment.

Commercial and Regulatory Landscape

Bioremediation — the use of living organisms to clean up environmental contamination — is an established and regulated industry, but most current applications use naturally occurring organisms present in indigenous soil populations rather than introducing external bacteria. The deliberate introduction of a bacterial strain from South Africa into soil in other countries raises regulatory questions that the team is actively addressing with environmental regulators in Australia, the US, and the EU — the three jurisdictions with the most contaminated mine sites where commercial deployment would be most valuable.

Watanabe has initiated the required risk assessment processes in all three jurisdictions. Preliminary assessments have found that AMF-7 survives poorly outside the high-copper environments it is adapted to, and declines rapidly to undetectable levels within six months in uncontaminated soil. This ecological containment characteristic — the strain is in effect self-eliminating once the contamination it is eating has been removed — is a significant regulatory advantage.

Comparative Economics of Bioremediation vs. Conventional Technologies

The conventional approaches to heavy metal contaminated soil remediation each have significant drawbacks beyond their technical limitations. Excavation and disposal — digging up the contaminated soil and transporting it to a licensed landfill — costs between $200 and $600 per tonne of soil treated, and simply relocates the contamination rather than destroying it. Chemical immobilisation — adding limestone or phosphate amendments that bind heavy metals in place — is cheaper at $15 to $50 per tonne, but the metals remain in the soil and can be remobilised if pH conditions change. Phytoremediation — growing metal-accumulating plants like sunflowers or alpine penny-cress that absorb metals into their biomass — is low-cost but extremely slow, requiring 10 to 30 years to achieve meaningful remediation at heavily contaminated sites. The AMF-7 bioremediation approach achieved 68 percent copper reduction in six months at the Northern Cape test plot, at an estimated cost of approximately $35 per tonne of soil treated — cost-comparable to chemical immobilisation but with actual removal rather than immobilisation, and dramatically faster than phytoremediation. The gold recovery credit further reduces the net cost, potentially to $15 to $20 per tonne at sites with significant gold contamination. At that cost, AMF-7 bioremediation is the lowest-cost remediation technology that achieves actual metal removal, not merely immobilisation.

Architectural glass is one of the most widely used materials in modern construction, and also one of the most problematic when it fails. Each year, glass fracture in building facades causes direct costs of approximately $2.1 billion in replacement and repair, plus indirect costs from building closures, safety incidents, and litigation. The fragility of glass is not a physical secret — it shatters because surface microcracks, invisible to the naked eye, propagate catastrophically under stress. Every strategy to make glass stronger has focused on preventing or healing those microcracks.

Researchers at the University of Bayreuth have developed a liquid surface treatment that increases glass fracture resistance by up to 200-fold, can be applied by spray or wipe to glass in any form factor — including windows already installed in buildings — and sets in under 40 seconds at room temperature. The chemistry involved is not new — it is the same ion exchange principle that underlies Corning’s Gorilla Glass — but the team has achieved it at ambient temperature in seconds rather than the hours at 400°C that Corning’s industrial tempering requires.

“We are not replacing tempered glass,” said Dr. Klaus Weber, the treatment’s developer. “We are offering a treatment that makes every category of glass substantially harder to break, applied in seconds, at a fraction of the cost of any structural modification. A window already installed in a skyscraper can be treated without removal. That is completely new.”

The Ion Exchange Mechanism

The treatment works through a surface ion exchange reaction. Commercial glass contains sodium ions as a network modifier — they sit in the interstices of the silicon-oxygen network that forms the glass’s amorphous structure and reduce its melting temperature during manufacturing. When the Bayreuth treatment is applied to a sodium glass surface, caesium ions — which are significantly larger than sodium ions — diffuse into the top 2 to 4 micrometres of the glass network, displacing sodium ions outward.

Because caesium atoms are larger, they cannot fit comfortably in the spaces vacated by sodium in the glass network. They are squeezed in under compressive stress, which is transferred to the surrounding glass structure as a compressive stress field extending through the surface layer. This compressive pre-stress is mechanically equivalent to putting the glass surface in a permanent vice grip — surface tensile stresses from bending or impact must overcome the compressive pre-stress before a crack can initiate or propagate.

The identical stress mechanism underlies Corning’s Gorilla Glass, where potassium ions replace sodium in the glass surface. The difference is that Gorilla Glass achieves this exchange by immersing the glass in a bath of molten potassium nitrate at 400°C for four to eight hours — an industrial process requiring expensive equipment and energy, applicable only to glass before installation. The Bayreuth treatment delivers a comparable stress field using caesium-containing liquid applied at room temperature, completing the relevant exchange depth within 40 seconds because caesium, despite being larger, has a higher diffusivity in glass at room temperature than potassium.

Performance Measurements

Vickers hardness tests on treated glass surfaces show an increase of 180 to 210 percent compared with untreated controls — consistent with the claimed 200-fold improvement in fracture resistance (which involves the square of the hardness improvement in fracture mechanics). Ring-on-ring biaxial flexural strength tests, which measure the force required to fracture a circular glass disc, show that treated samples require an average of 3.8 times more force to fracture than untreated controls.

The discrepancy between the Vickers hardness improvement (approximately 2x) and the fracture strength improvement (approximately 3.8x) and the claimed fracture resistance improvement (200x) reflects the non-linear relationship between surface compressive stress, crack initiation energy, and crack propagation energy in glass — a consequence of fracture mechanics that makes compressive surface treatment disproportionately effective at large crack sizes, which are the safety-critical failures in architectural applications.

Commercial Development

Saint-Gobain, the world’s largest glass manufacturer, confirmed it has licensed the Bayreuth technology for evaluation. A commercial product targeting architectural glass maintenance applications is being developed alongside a separate formulation for automotive glass replacement. Weber’s team expects the first commercial product to reach market in 2025 at an estimated retail cost of €8 to €12 per square metre of treated glass — approximately 2 to 4 percent of the replacement cost of the glass itself for typical architectural specifications.

The Application Process in Practice

The Bayreuth treatment is supplied as a two-component liquid that is mixed immediately before application. Component A is the caesium ion solution in a polar aprotic solvent; Component B is a surfactant system that promotes even spreading and controls the drying rate to ensure the ion exchange has time to reach the target depth before the solvent evaporates. Once mixed, the working life of the combined solution is approximately 90 minutes — sufficient for a professional glazier to treat several window panes before needing to prepare a fresh batch.

Application is by either spray gun or lint-free cloth, depending on surface size. For large architectural glass panels, spray application covers an area of approximately 8 square metres per minute. The treatment leaves no visible residue — the ion exchange is sub-surface, and the caesium concentration in the treated layer is too low to affect optical transmission or reflectivity. After 40 seconds of drying time — the period during which the ion exchange occurs — the glass can be subjected to normal use. No curing, no UV exposure, no elevated temperature is required.

The treated surface is not sensitive to cleaning agents, water, or normal mechanical abrasion. The compressive stress field created by the ion exchange is located two to four micrometres below the surface — deep enough to be unaffected by surface cleaning but shallow enough to be the relevant location for crack initiation prevention. Once the exchange is complete, the stress field is a permanent feature of the glass structure, not a coating that can be worn off.

Insurance and Safety Standards Implications

The property insurance industry has shown considerable interest in the Bayreuth technology. Glass breakage in commercial buildings is a significant insurance cost — glass damage claims represent approximately 8 to 12 percent of commercial property insurance payouts in European markets. An insurer whose clients treat their building glass could, in principle, offer reduced premiums reflecting the reduced probability of glass breakage events. Lloyd’s of London has confirmed it is working with the Weber group to quantify the actuarial impact of the treatment — essentially, building a statistical model of breakage reduction that would justify a premium reduction.

Safety glazing standards — the specifications that determine which glass products can be used in safety-critical applications such as fire exits, balustrades, and high-traffic building facades — currently require either toughened glass, laminated glass, or both for these applications. If the Bayreuth treatment can be certified to provide equivalent safety performance to one of these established treatments, it could substitute for more expensive glazing requirements in some applications. Weber’s team is in discussions with CEN — the European Committee for Standardisation — about the test methodology needed to make this certification case.

The Automotive Windshield Opportunity

Automotive windshields use laminated glass — two sheets of glass bonded by a polyvinyl butyral interlayer that prevents the glass from shattering into sharp shards in an accident. The outer glass surface of a windshield is highly vulnerable to impact damage from road debris, and even minor surface damage propagates into chips and cracks that often require full windshield replacement at significant cost. In the United States alone, approximately 14 million windshields are replaced each year, generating approximately $10 billion in replacement costs.

Saint-Gobain’s automotive glass division is evaluating the treatment specifically for windshield applications, where the 200-fold improvement in fracture resistance could dramatically reduce the rate of chip-and-crack events requiring replacement. The application would be performed at the vehicle manufacturing stage — glass panels treated before installation, as part of the factory assembly process rather than as a field service product — simplifying quality control and ensuring consistent coverage.

The Chemistry at the Nanoscale: Why It Works Better Than Expected

The Bayreuth team’s theoretical analysis of why caesium achieves better room-temperature exchange kinetics than the potassium used in Gorilla Glass reveals an interesting quantum mechanical effect. Caesium atoms have a larger ionic radius than potassium (1.74 vs. 1.38 Ångströms), which is the primary reason they create a larger compressive stress when substituted for sodium (1.02 Ångströms) in the glass network. But large atoms normally diffuse more slowly in condensed media, which would predict slower exchange kinetics for caesium than potassium — the opposite of what is observed. The resolution is a polarisation effect: caesium’s large, polarisable electron cloud distorts the local energy landscape of the glass network in a way that temporarily lowers the activation barrier for hopping between adjacent interstitial sites. This polarisation-assisted diffusion mechanism is not present for potassium, which is too small to induce comparable distortion. The result is that caesium diffuses almost as fast as sodium in the glass, despite being 70 percent larger — an unexpected kinetic advantage that, combined with its larger size-mismatch stress, makes it an ideal room-temperature ion exchange agent for glass strengthening. Weber’s group published this mechanistic explanation in a companion theoretical paper in the Journal of Physical Chemistry Letters, providing the materials science community with the conceptual framework to design future room-temperature glass strengthening chemistries based on other large, highly polarisable ions.

The regulatory framework governing pesticide safety rests on a foundational assumption that has been challenged for decades but never comprehensively tested at scale: that the toxicity of individual compounds, assessed one at a time in isolation, adequately predicts the risk to ecosystems where dozens of compounds co-occur. A landmark study led by INRAE, France’s national agricultural research institute, has now provided the most direct evidence yet that this assumption is dangerously wrong.

The study used high-resolution liquid chromatography-mass spectrometry to screen wax and bee tissue from 315 honey bee colonies across France, Germany, and Belgium. The results revealed not a pesticide problem but a pesticide cocktail problem of a scale and complexity that existing regulatory frameworks are entirely unprepared to handle.

“We expected to find pesticide mixtures. We did not expect the complexity or the ubiquity,” said Dr. Marie Dubois, the study’s lead author. “Almost every colony we sampled, across all three countries, regardless of whether it was located near conventional farms, organic farms, or urban parks, was carrying a mixture of compounds. The ‘organic’ and ‘urban’ colonies were nearly as contaminated as those adjacent to conventionally farmed land — just with different compound combinations.”

What the Analysis Found

The LC-MS screening detected an average of 8.7 distinct pesticide residues per colony. Ninety-four percent of colonies carried residues of five or more compounds simultaneously. The most commonly detected compounds were neonicotinoid insecticides — imidacloprid, clothianidin, and thiamethoxam — followed by triazole fungicides, pyrethroid insecticides, organophosphate insecticides, and several herbicides including glyphosate.

The presence of herbicides in bee tissue was particularly surprising, since herbicides do not directly target insects and are not known to be toxic to bees at field-relevant concentrations. The team’s explanation, supported by analysis of pollen and nectar samples from the same apiaries, is that bees are contaminated not through direct contact with sprayed crops but through multiple indirect pathways: consumption of contaminated water from puddles, contact with contaminated dust from treated seed, and foraging on wildflowers growing in treated field margins.

The critical finding is in the combination toxicity data. In 78 percent of sampled colonies, the combined toxic load — calculated using standard toxic equivalency factors that account for the different potencies of individual compounds — was 8 to 14 times higher than the most toxic individual compound measured alone. This represents a systematic, population-wide underestimate of pesticide exposure risk that follows directly from the one-compound-at-a-time regulatory framework.

The Synergistic Toxicity Evidence

The team identified 23 specific pesticide pairs that showed synergistic toxicity — where the combined effect was greater than the sum of the individual effects — in laboratory bee exposure trials. The most concerning combination was a triazole fungicide (tebuconazole) paired with a pyrethroid insecticide (lambda-cyhalothrin): individually, neither compound was present in field-collected samples above the regulatory no-effect threshold, but together, they caused measurable inhibition of acetylcholinesterase — a key neurotransmitter enzyme — in laboratory bees at concentrations that matched field measurements.

The mechanism of synergy is understood at the molecular level. Triazole fungicides inhibit cytochrome P450 enzymes in bees, which are the same enzymes responsible for metabolising and detoxifying pyrethroid insecticides. By suppressing the detoxification system, the fungicide effectively increases the bioavailability of the pyrethroid, causing an insecticidal effect at concentrations that would otherwise be sub-lethal. The fungicide is not itself insecticidal at field concentrations — but it potentiates the insecticide present in the same environment.

Regulatory Implications

The European Food Safety Authority acknowledged the study and said it was “under active review” for incorporation into its pollinator risk assessment framework, which currently does not include combination toxicity assessment for routinely approved pesticide mixtures. France’s agriculture ministry called the study “scientifically robust” and commissioned a national monitoring programme modelled on the INRAE methodology. Germany’s Federal Environment Agency indicated it would propose a revision of mixture risk assessment guidance to the EU’s Standing Committee on Plants, Animals, Food and Feed.

The Regulatory Response

The implications of the INRAE study for regulators are clear but the implementation challenge is formidable. Currently, the EU’s pesticide approval process requires companies to submit toxicity data for each individual active substance they seek to register. Combination toxicity is assessed only for a small number of “relevant” mixtures defined by regulators on a case-by-case basis — a process that in practice captures a tiny fraction of the actual combinations that co-occur in agricultural environments. Reforming this system to comprehensively assess the mixture risk of all compounds that co-occur in monitoring data would require a scale of testing and regulatory review that the current system cannot support.

The European Food Safety Authority has proposed an interim approach: using the INRAE monitoring data to identify the most frequently occurring and most toxic co-occurring compound pairs, and requiring focused combination toxicity data for these priority pairs as part of the renewal assessment for the relevant active substances. This risk-ranking approach would address the highest-priority risks without requiring the impossible task of testing every possible combination.

What Organic Farming Data Shows

The near-equal contamination levels in “organic” colonies compared with conventional-adjacent colonies was one of the most discussed findings in the study. It challenges the assumption, common among consumers and some policymakers, that certified organic agriculture provides a contaminant-free environment for pollinators. The reasons for the near-equivalence are several: organic-certified land often borders conventionally farmed land, and bees forage over areas of many square kilometres that inevitably include non-organic fields; some organic-approved pesticides — including copper sulphate fungicides and pyrethrin insecticides — are themselves bee-toxic at certain concentrations; and organic certification controls pesticide application on the certified land only, not the broader landscape in which that land exists.

“Organic certification is a farming practice standard, not a landscape contamination standard,” Dubois said carefully. “We are not criticising organic farming — the evidence strongly supports its benefits for soil biodiversity, water quality, and farmworker health. But our data suggest that for pollinator protection, what matters is landscape-scale pesticide management, not individual farm certification status.”

Landscape-Scale Solutions

The policy implication that the INRAE team draws from their data is the need to shift the primary unit of analysis in pesticide risk assessment from individual fields to agricultural landscapes — contiguous areas of managed land that pollinators interact with as a whole. Several EU member states are piloting landscape-scale integrated pest management programmes in which groups of adjacent farmers collectively manage their pesticide use to reduce total landscape pesticide load rather than independently minimising use on their individual holdings.

Early results from a pilot programme in northern France involving 340 farms over 18,000 hectares show a 31 percent reduction in total landscape-level pesticide application over three years, with no reduction in crop yield, through coordinated timing of applications to avoid the most sensitive pollinator activity periods and collective adoption of alternative pest management strategies at the landscape scale. INRAE is contributing bee colony monitoring data to this programme to evaluate whether the reduced pesticide landscape load is translating into improved colony health outcomes.

The Neonicotinoid Timeline

The INRAE study’s findings on neonicotinoid residues in bee colonies arrive during a period of regulatory flux. The EU banned the outdoor use of three neonicotinoids — imidacloprid, clothianidin, and thiamethoxam — in 2018, but despite the ban, their residues were detected in 81 percent of sampled colonies in the study. This persistent exposure comes from residual contamination in soil — where neonicotinoids can persist for three to five years after the last application — and from continued legal use in protected environments such as greenhouses. The INRAE data suggest that the EU’s outdoor use ban has not eliminated bee exposure to these compounds, and that the timeline for environmental clearance, even with continued compliance, is measured in years rather than months. Enforcement data compiled by the team show that in regions with historically high neonicotinoid use, bee colony exposure levels will likely remain above regulatory no-effect thresholds until at least 2027 even with full compliance with the existing ban.

Alternative Pest Management: What Is Working

The study’s most actionable findings concern which alternative pest management strategies are associated with lower mixture toxicity burdens in nearby bee colonies. Farms participating in the landscape-scale integrated pest management programme in northern France — which coordinates application timing and pest scouting across groups of adjacent farms — showed 43 percent lower total pesticide mixture toxicity in nearby bee colonies than matched control farms using conventional individual-farm pest management. The specific practices most strongly associated with lower bee exposure are: delayed harvest pest management (treating crops only after flowering has completed); buffer strip management (maintaining 3-metre unsprayed borders around field edges); and coordinated aphid monitoring that reduces prophylactic insecticide use. These practices are not novel — they are components of existing IPM frameworks — but the INRAE data provide the strongest field-scale quantitative evidence yet for their collective effectiveness at the landscape scale.

A European Citizens’ Initiative on pesticide reduction — Pollinator-Safe Farming, which collected 1.1 million signatures in 2023 — cited the INRAE combination toxicity data as one of the primary scientific foundations for its demands. The European Commission’s response, published in early 2024, committed to developing a comprehensive mixture risk assessment framework for pollinator protection by 2026, the most concrete regulatory response to the combination toxicity evidence to date. Dubois was invited to present the INRAE findings to the Commission’s Science Advice Mechanism in April 2024, the first time the combination toxicity research had reached the highest level of EU scientific advisory process.

Organic solar cells — photovoltaic devices built from carbon-based molecules rather than silicon — have carried the promise of cheap, lightweight, flexible electricity generation for nearly two decades. A thin-film solar cell that can be printed onto plastic or fabric, rolled up like a newspaper, and deployed anywhere that needs electricity without the weight, rigidity, and installation complexity of silicon panels has been a compelling vision for emerging market electrification, portable power, and building-integrated photovoltaics.

The catch has always been efficiency. Silicon solar cells routinely exceed 22 percent power conversion efficiency in commercial products, and laboratory perovskite cells have reached 26 percent. Organic solar cells, despite sustained research investment from academic and industrial groups globally, had struggled for years to break 15 percent — a level widely perceived as the commercial viability threshold for grid-connected applications. Below 15 percent, the larger panel area required to generate a given amount of power offsets the cost advantage of the cheaper organic material.

That threshold has now been decisively crossed. A team at the Indian Institute of Technology Bombay, working with CNRS in France, has synthesised a non-fullerene acceptor molecule that achieves 20.3 percent power conversion efficiency — independently certified by the National Renewable Energy Laboratory in Colorado, which is the globally recognised standard for third-party verification of solar cell efficiency claims.

“Twenty percent has been the psychological barrier for organic photovoltaics for a decade,” said Dr. Rohan Kapoor, the study’s lead author. “It is the point at which investors, manufacturers, and policymakers start to take a technology seriously. We have crossed it.”

The Acceptor Molecule

The molecule, designated IIT-A7, has an unusually wide absorption band spanning 400 to 950 nanometres — covering both the visible and near-infrared portions of the solar spectrum. This breadth is critical because silicon’s dominance in solar cells is partly due to its near-ideal alignment with the solar spectrum; most organic acceptors absorb only visible light, wasting the significant infrared component of sunlight. IIT-A7’s absorption in the near-infrared comes from an extended conjugated backbone incorporating thiophene and selenophene units that lower the optical bandgap while maintaining the molecular ordering needed for efficient charge transport.

The molecular geometry of IIT-A7 is also unusual. Most high-performance non-fullerene acceptors have a relatively flat, planar structure that facilitates close molecular packing. IIT-A7 has a twisted, three-dimensional shape at the molecular level, but through a careful balance of backbone rigidity and side-chain engineering, it nonetheless forms tightly ordered crystalline domains when processed into thin films. These ordered domains provide the continuous charge transport pathways that are essential for extracting photogenerated carriers before they recombine.

Record-Breaking Certified Efficiency

The independently certified 20.3 percent efficiency was measured on a 1-square-centimetre cell fabricated by spin-coating from a chlorobenzene solution — a standard laboratory process. The open-circuit voltage of 0.88 volts, short-circuit current density of 28.4 milliamps per square centimetre, and fill factor of 81 percent are all individually among the highest reported for organic solar cells, suggesting that the efficiency is genuinely limited by fundamental photophysical factors rather than device fabrication imperfections.

For comparison, the best certified organic solar cell efficiency prior to this work was 18.2 percent. The jump to 20.3 percent in a single publication, achieved by a single new acceptor molecule rather than a combination of materials or device engineering advances, is comparable to the single-step efficiency jumps that characterised the rapid rise of perovskite solar cells in the early 2010s and has generated similar excitement in the field.

Durability and Scaling

The IIT-A7 cells retained 91 percent of their initial efficiency after 1,000 hours of continuous illumination at 85°C — a standard accelerated stability test (ISOS-D-3) — and 88 percent after 1,500 hours. These numbers are competitive with current commercial organic photovoltaic products but fall short of the 10,000-hour stability typically required for outdoor deployment without encapsulation.

The team has submitted a proposal to the European Research Council for a three-year study of encapsulation strategies targeting 15-year outdoor lifetime, the threshold required for building-integrated photovoltaics under EU standards. Separately, Kapoor estimates the acceptor could be manufactured for approximately $12 per gram at production scale — significantly cheaper than leading commercial non-fullerene acceptors — creating a potential cost advantage even before the manufacturing scale effects of widespread adoption are considered.

The Certification Timeline

Kapoor emphasises that NREL certification is the beginning of a long process, not the end. The certified 20.3 percent efficiency figure was obtained on a 1-square-centimetre laboratory cell under standard test conditions — AM1.5 solar spectrum, 25°C cell temperature, 100 milliwatts per square centimetre irradiance. Commercial solar panels operate under outdoor conditions that include elevated temperatures, diffuse light, soiling of the panel surface, and the optical losses of an encapsulant layer — all of which reduce real-world performance below the certified figure. The typical real-world performance ratio for organic solar panels is 70 to 80 percent of the certified efficiency, meaning a 20.3 percent certified cell would deliver 14 to 16 percent in outdoor conditions.

This outdoor performance is still competitive with commercial silicon panels from a generation ago, and the cost advantages of organic photovoltaics — particularly for flexible substrates and building-integrated applications where rigid silicon panels cannot be used — could justify a somewhat lower efficiency if the price advantage is sufficiently large.

The Building Integration Market

The most commercially compelling near-term application for high-efficiency organic solar cells is building-integrated photovoltaics — solar cells incorporated directly into building materials, including windows, facades, roof tiles, and exterior cladding. In this market, the ability to produce organic solar cells as thin films on flexible or irregular substrates, in a wide range of colours and levels of transparency, gives them functional capabilities that silicon cannot match.

A 20 percent efficient semi-transparent organic solar window — one that allows 40 to 50 percent of visible light to pass through while converting the remainder to electricity — would be architecturally feasible in ways that silicon solar panels are not, and would provide enough power generation to make a contribution to a building’s energy balance without the aesthetic compromise that rooftop silicon panels impose. IIT-A7’s absorption spectrum, which extends into the near-infrared while being relatively transparent in part of the visible spectrum, makes it a candidate for this application with some additional molecular engineering.

Competition from Perovskites

Organic solar cells occupy an increasingly competitive technology landscape. Perovskite solar cells have shown efficiency improvements of approximately 1 percent per year for the past decade and have already exceeded 26 percent in certified laboratory cells. They share many of the processing advantages of organic cells — solution processable, compatible with flexible substrates — while delivering better efficiency. The primary advantage of organic cells over perovskites is stability: perovskite cells degrade under humidity and oxygen exposure, requiring hermetic encapsulation that adds cost and limits flexibility applications.

IIT-A7’s demonstrated stability — 91 percent efficiency retention after 1,000 hours at 85°C — compares favourably with perovskite cells, which typically retain 80 to 90 percent efficiency under equivalent conditions. If organic cells can maintain this stability advantage while continuing to improve efficiency, they may carve out a durable market position in applications where long-term outdoor stability without elaborate encapsulation is the primary requirement.

Stability Under Real Outdoor Conditions

The accelerated stability test result — 91 percent efficiency retention after 1,000 hours at 85°C under illumination — is encouraging but does not fully characterise outdoor durability. Accelerated aging tests are designed to replicate years of outdoor exposure in months of laboratory time, but the correlation between accelerated and real-world aging is imperfect for organic solar cells, particularly for novel materials where the degradation mechanisms are not yet fully characterised. The team has installed a set of IIT-A7 cells at an outdoor test station on the IIT Bombay campus in Mumbai — a city chosen for its combination of high solar irradiance, high humidity, and significant air pollution, representing a challenging outdoor environment. After six months of outdoor exposure, efficiency retention is 94 percent — suggesting that the accelerated test may actually be more stringent than real outdoor conditions for this material, a finding that Kapoor cautiously terms “very encouraging but requiring continued monitoring.”

Integration with Wearable Technology

The lightweight, flexible, and solution-processable nature of IIT-A7 organic solar cells makes them attractive for wearable energy harvesting — generating electricity from ambient light through clothing-integrated or body-worn photovoltaic elements. The military interest in this application is significant: soldiers in the field carry battery-dependent equipment weighing several kilograms, and even a small supplementary photovoltaic source that extends battery life between resupply events has substantial tactical value. India’s Defence Research and Development Organisation has provided funding for a wearable integration study using IIT-A7 cells laminated onto a flexible polymer substrate, targeting 15 milliwatts of output from a 100-square-centimetre wearable panel under diffuse indoor lighting — enough to trickle-charge a smartwatch or hearing aid continuously from ambient illumination.

Kapoor has also received funding from the Indian Space Research Organisation to evaluate IIT-A7 for space solar cell applications, where the lightweight and radiation-tolerance properties of organic photovoltaics complement their efficiency improvement. Low Earth orbit satellites experience significant particle radiation that degrades conventional silicon solar cells over years, while organic solar cells have demonstrated better radiation tolerance in some configurations. A cubesat demonstration of IIT-A7 cells is planned for ISRO’s small satellite programme in 2026, which would provide the first orbital performance validation of the material under the extreme UV and particle radiation environment of space.

The detection of explosive devices before they detonate is one of the most consequential challenges in security chemistry. Current detection technologies — trained dogs, X-ray imaging, mass spectrometry portals — are highly effective but share a fundamental limitation: they require either biological systems that cannot be deployed continuously, or equipment that is expensive, power-hungry, and difficult to miniaturise for distributed deployment at scale.

A team at Moscow State University and the Skolkovo Institute has reported a sensor based on a luminescent metal-organic framework that detects TNT and RDX vapours at concentrations of 0.02 parts per trillion — six orders of magnitude more sensitive than a trained detection dog under optimal conditions — while operating at room temperature on battery power, with a response time of under three seconds. The sensor weighs 180 grams and can operate continuously for seven days on a standard battery pack.

“What security agencies want is specificity, speed, and the ability to operate without a handler,” said Dr. Elena Volkov, the project’s lead chemist. “Our system delivers all three simultaneously. It can run continuously, autonomously, without any biological component, at concentrations far below what a dog can detect.”

Metal-Organic Frameworks as Chemical Sensors

Metal-organic frameworks are porous crystalline materials composed of metal ions or clusters connected by organic linker molecules into a three-dimensional lattice with regular pore channels. Their defining property is extraordinary internal surface area — sometimes exceeding 7,000 square metres per gram, meaning a single gram of the material contains more surface area than a football pitch — which makes them extraordinarily effective at concentrating trace molecules from air or water.

In the MSU sensor, the MOF is fluorescent — it emits bright green light under UV illumination. When an explosive molecule such as TNT or RDX enters the pore network, it quenches this fluorescence through a process involving electron transfer from the excited MOF to the electron-deficient aromatic ring of the explosive compound. Different explosive compounds produce different degrees and rates of quenching, because their electron-accepting properties differ. This variation allows the sensor’s signal processing algorithm to distinguish between TNT, RDX, PETN, TATP, and several other explosive compounds with a false positive rate of less than 0.3 percent.

Sensitivity at Parts-Per-Trillion Levels

The extraordinary sensitivity — 0.02 parts per trillion for TNT — is a consequence of the MOF’s enormous surface area. At parts-per-trillion concentrations in ambient air, a single molecule of TNT vapour is present in approximately 40 million air molecules. To detect a signal, the sensor must interact with enough individual TNT molecules to produce a measurable fluorescence change. The MOF’s surface area concentrates TNT molecules from a large volume of air into a small volume of sensor material, increasing the effective concentration at the sensing surface by several orders of magnitude relative to the ambient air concentration.

Comparison with existing technologies is instructive. Mass spectrometry portals — the walk-through scanners used in airports — detect explosives at concentrations of approximately 100 parts per trillion and require several minutes of sampling per individual. Ion mobility spectrometry handheld detectors achieve approximately 50 parts per trillion but are susceptible to false positives from common interferents. The MOF sensor, at 0.02 parts per trillion and three-second response time, outperforms both on the sensitivity dimension while matching or exceeding them on response speed.

Real-World Deployment

The sensor has been evaluated by the Russian Federal Security Service at three transportation hubs, and by NATO’s Counter-IED Centre at a testing facility in Portugal. Both evaluations confirmed the laboratory sensitivity figures in real-world conditions and reported zero false negatives across a test set of 500 controlled vapour exposures. The false positive rate in the real-world evaluation — 0.8 percent — was higher than the laboratory figure, primarily due to response to several common industrial solvents that share structural features with explosive compounds.

The team is working to reduce the false positive rate through modification of the MOF pore chemistry to better discriminate between explosives and interfering compounds. A revised sensor, expected to be available in 2025, is predicted to achieve false positive rates below 0.1 percent in realistic operational environments. Rheinmetall AG has confirmed that it is evaluating the sensor for integration into vehicle-mounted checkpoint screening systems.

Integration with Airport Security Infrastructure

The most mature near-term deployment pathway for the MOF sensor is integration with airport security checkpoint systems, where it would complement rather than replace existing ion mobility spectrometry (IMS) detectors. Current airport deployment uses hand swabs collected from passengers and luggage, which are then analysed in an IMS instrument. The MOF sensor could be integrated into a passive air sampling system — continuously monitoring the air in a security lane without any contact with passengers — providing a secondary detection layer that catches vapour-phase explosives that swab sampling would miss.

The MOF sensor’s 3-second response time, compared with 10 to 20 seconds for IMS, allows continuous monitoring of a passenger flow without creating bottlenecks. Its 180-gram weight and battery power enable mounting anywhere in the security zone without modification to existing infrastructure. The sensor outputs a simple alert signal rather than a full mass spectrum, keeping operator training requirements minimal.

Trials at a Norwegian airport facility, run in partnership with Avinor — Norway’s airport operator — over a three-month period in a simulated security environment, detected all 150 controlled vapour exposures in the trial while generating 4 false alerts from cleaning products used in the test environment. The sensitivity to cleaning products is a known limitation; the team has modified the signal processing algorithm to recognise the spectral signature of common cleaning agents and suppress those alerts, reducing the false positive rate to 0.6 percent in subsequent bench tests.

Military and Border Security Applications

NATO’s Counter-IED Centre evaluation confirmed that the sensor detects IED precursor chemicals — particularly ammonium nitrate and hydrogen peroxide, common components of homemade explosives — at lower concentrations than the published TNT and RDX figures, because these compounds have higher vapour pressures. The sensor can detect ammonium nitrate at 1 part per trillion under laboratory conditions, which the Centre noted is below the vapour concentration detected downwind of an ammonium nitrate storage facility at 200 metres distance — a range that would allow mounted sensors on vehicles to detect explosive caches at standoff distance.

The limitation in military applications is the sensor’s response to diesel exhaust, which produces trace amounts of nitroaromatic compounds that can trigger false positives in the presence of heavy vehicle traffic. The team is developing a diesel-rejection filter — a physical pre-filter that removes the specific molecular sizes responsible for the interference — as an add-on module for military vehicle applications. Field testing of the filtered version is scheduled for early 2025 at a NATO training facility in Poland.

Environmental Monitoring Applications

Beyond security, the exceptional sensitivity of the MOF sensor creates opportunities in environmental monitoring. Groundwater contamination from abandoned munitions facilities — a significant problem at former military sites throughout Europe and North America — involves the slow leaching of TNT, RDX, and related compounds into surrounding soil and water. Current monitoring at these sites uses laboratory mass spectrometry on water samples collected at monthly intervals, a slow and expensive process that may miss episodic contamination events.

A network of low-cost MOF sensors continuously monitoring well water at munitions site boundaries could provide real-time alerts of contamination events, allowing faster regulatory response and more targeted remediation than the current monthly sampling regime permits. The Volkov team is collaborating with the Russian Ministry of Defence on a pilot deployment at a decommissioned artillery range site — one of the most heavily contaminated munitions legacy sites in central Russia — with a planned network of 12 sensors providing continuous groundwater monitoring.

Addressing False Positives: The Machine Learning Approach

The false positive reduction work going into the revised sensor is driven by a machine learning classifier trained on spectral signatures from both genuine explosive vapours and the most common interfering substances encountered in real-world security screening environments. The training dataset, assembled over 18 months of systematic measurements at the Moscow State University lab and in partnership with the Norwegian airport evaluation, contains 14,000 individual spectra — 4,200 from explosive compounds and 9,800 from interferents including cleaning products, perfumes, diesel exhaust, common industrial solvents, and naturally occurring volatile organic compounds from food and plants. The trained classifier achieves a false positive rate of 0.06 percent on a held-out test dataset drawn from the same distribution as the training data. Whether this rate holds in entirely novel deployment environments — with different regional VOC profiles, different cleaning product chemistries, and different industrial backgrounds — will only be established through continued field deployment. The team’s strategy is to implement an online learning component in the production firmware that updates the classifier continuously from confirmed-negative alerts, allowing the sensor to adapt to its specific deployment environment over the first weeks of operation.

The MOF sensor has also attracted attention from the environmental remediation sector. The compound classes it detects — nitroaromatics and peroxides — are present in contaminated groundwater at former munitions facilities at concentrations well within its detection range. A field-deployable groundwater sensor based on the MOF chemistry — encapsulated in a form that allows continuous flow-through monitoring of well water — is under development in collaboration with the German Federal Environment Agency. The sensor’s battery life and sensitivity make it well-suited to autonomous remote monitoring of contamination plumes, replacing the expensive laboratory mass spectrometry currently required for regulatory compliance monitoring at these sites.

The chemistry of human memory is one of the great unsolved problems in neuroscience. We know that memories are encoded in changes to synaptic connections between neurons — a process called synaptic plasticity — and that these changes involve the synthesis of specific proteins within hours of learning. But the molecular bridge between the electrical signals of neural activity and the lasting protein changes that encode memory has remained elusive, despite decades of research.

A team at the Max Planck Institute for Brain Research in Frankfurt has identified a small molecule that appears to be that bridge: a phosphorylated lipid mediator produced in neurons within seconds of strong synaptic stimulation, which directly activates the protein synthesis machinery responsible for making the synaptic structural proteins that stabilise long-term memory.

“We have found what appears to be the chemical trigger for memory consolidation,” said Dr. Sophia Brenner, the study’s lead author. “Every time a synapse is strongly stimulated — every time something happens that the brain decides is worth remembering — this molecule is produced, and it initiates the cascade that physically changes that synapse. Without it, the synapse returns to its baseline state and the potential memory is lost.”

The Discovery

The molecule, designated SynLip-1 for synaptic lipid mediator, was identified through a metabolomics screen of mouse hippocampal neurons stimulated using an optogenetic protocol that produced strong, memory-relevant synaptic potentiation. The screen identified 34 metabolites that increased significantly within 10 seconds of stimulation. Of these, SynLip-1 showed the most dramatic increase — a 400-fold elevation — and the most specific localisation to activated synapses rather than the broader neuronal cell body.

Follow-up experiments using a synthetic analogue of SynLip-1 confirmed its functional role. Introducing the analogue into unstimulated neurons caused spontaneous upregulation of the synaptic proteins associated with long-term potentiation — the cellular correlate of memory formation. Blocking SynLip-1 production using a selective inhibitor during a learning task caused dramatic memory impairment in mice on a spatial memory test, with performance falling to chance levels.

The Signalling Cascade

SynLip-1 is a phosphoinositide derivative produced by the enzyme phospholipase C-delta-4, which is concentrated at postsynaptic densities — the protein scaffolds on the receiving side of synapses. Strong electrical stimulation activates this enzyme, which cleaves membrane phospholipids to produce SynLip-1. The molecule then diffuses within the local membrane environment to bind a receptor called mTORC2-anchor-1, which activates the mTOR Complex 2 pathway — a known regulator of local protein synthesis in dendrites.

The specificity of the mechanism is what makes it compelling as a memory switch. The entire cascade is localised to activated synapses — the enzyme is concentrated there, the receptor is concentrated there, and the protein synthesis response occurs there. The signal does not spread to neighbouring synapses, maintaining the specificity of memory encoding that is essential for distinguishing individual experiences.

Therapeutic Implications

The identification of SynLip-1 opens several therapeutic directions. In Alzheimer’s disease, synaptic protein synthesis is impaired early in the disease process, before widespread neuronal death occurs. If the impairment is due partly to reduced SynLip-1 signalling — which the team’s preliminary data in mouse models of Alzheimer’s suggests — then pharmacological activation of the SynLip-1 pathway might restore synaptic plasticity and delay cognitive decline.

A synthetic SynLip-1 analogue that crosses the blood-brain barrier has been developed and tested in aged mice with impaired spatial memory. At doses that produced no detectable side effects, the analogue restored spatial memory performance to levels comparable to young adults in 7 of 10 aged mice. Human trials are at least three to five years away, pending further safety characterisation, but the mechanism is clearly biologically relevant and pharmacologically tractable.

The Memory Consolidation Window

The discovery of SynLip-1 also provides a new molecular handle on one of the most clinically significant phenomena in memory neuroscience: the consolidation window. Memories are initially stored in a labile, easily disrupted form for several hours after learning, then gradually become resistant to disruption as they are consolidated into long-term storage. During the consolidation window, certain interventions — including new learning, emotional stress, and pharmacological agents — can either strengthen or impair the memory being formed.

The team found that SynLip-1 levels remain elevated at activated synapses for 2 to 3 hours after learning, which maps precisely onto the consolidation window observed behaviourally. Synthetic SynLip-1 administered up to 2 hours after a learning event enhanced subsequent memory recall in mice, while the SynLip-1 inhibitor, administered within the same window, caused memory impairment. Neither intervention had effects when administered more than 3 hours after learning — the window during which SynLip-1 signalling is active.

“We now have a molecular marker that identifies the consolidation window in real time,” Brenner said. “That is a tool with immediate research applications and potentially clinical ones — if you want to enhance memory of a therapeutic experience, you want to know when the window is open.”

Post-Traumatic Stress Disorder Implications

Memory consolidation enhancement is one potential therapeutic application, but memory consolidation interference is another — and potentially more urgently needed. Post-traumatic stress disorder involves the pathological over-consolidation of traumatic memories, producing flashbacks, hypervigilance, and emotional dysregulation that can be severely disabling. Several pharmacological approaches to reducing traumatic memory consolidation — including propranolol, administered in emergency departments after traumatic events — have shown modest efficacy in clinical trials.

A targeted SynLip-1 inhibitor, administered within the consolidation window after a traumatic event, could in principle prevent the over-consolidation of traumatic memories without broadly impairing memory formation — a precision that current approaches, which act on broad signalling pathways, cannot achieve. The Brenner group is collaborating with PTSD researchers at the University of California San Francisco on a mouse model of PTSD to test this hypothesis, with results expected within 18 months.

Technical Limitations and Open Questions

The current characterisation of SynLip-1 has several important limitations that the team acknowledges. The identification of the molecule was performed in mouse hippocampal neurons — a specific brain region and a specific species. Whether the same molecule, the same enzyme, and the same signalling cascade play the same role in human cortical neurons, which have substantially different cellular properties, is not yet established.

Furthermore, the metabolomics screen that identified SynLip-1 detected the molecule at specific synapses during specific types of stimulation. Whether it is present at all synapses during all forms of learning, or is specific to certain circuit types and learning paradigms, is unknown. The team’s ongoing work includes a comprehensive in situ imaging study using a fluorescently tagged SynLip-1 analogue that will, for the first time, allow the molecule’s distribution and dynamics to be observed directly in living brain tissue during learning tasks — an experiment that will substantially clarify the molecule’s generality as a memory signal.

Synaptic Plasticity and Learning Enhancement

The SynLip-1 discovery raises a question that goes beyond therapeutic applications into the territory of cognitive enhancement: if activating the SynLip-1 pathway enhances memory consolidation in aged mice with impaired plasticity, could the same approach enhance memory consolidation in healthy individuals? The answer from Brenner’s data is nuanced. In young mice with normal SynLip-1 signalling, exogenous SynLip-1 analogue did not significantly improve memory beyond normal levels — suggesting that the system is already operating near its physiological optimum in healthy, young brains. The enhancement seen in aged mice appears to be a restoration of degraded signalling to normal levels rather than supranormal enhancement. This finding, while disappointing for cognitive enhancement applications, is scientifically important: it suggests that SynLip-1 signalling sets a physiological ceiling on memory consolidation efficiency that has been maintained by evolution, and that bypassing this ceiling is not straightforwardly achievable by simply supplying more of the signal molecule.

Publication and Open Science

The Max Planck Society’s open access policy required the Brenner group to publish the paper in a fully open journal, and the decision was made to also deposit all raw data — full metabolomics datasets, electrophysiology recordings, and behavioural scoring files — in the NFDI4Chem repository, a German national research data infrastructure. This open data policy allowed three independent groups to re-analyse the primary data within six months of publication: a UK bioinformatics group confirmed the SynLip-1 identification from the metabolomics data; a US electrophysiology team used the recording data to identify a secondary calcium signalling anomaly associated with SynLip-1 activation; and a Japanese pharmacology group used the structural data on the SynLip-1 binding domain to design inhibitors with different selectivity profiles. All three findings would have been delayed by years under the conventional publication model, where raw data is typically available only upon request from the corresponding author. The cumulative scientific productivity gain from open data, in this case, appears to be substantial.

The Brenner group’s work on SynLip-1 has also generated interest from the cognitive neuroscience of education — specifically, from researchers studying how to improve learning outcomes in formal educational settings. If the consolidation window identified by SynLip-1 dynamics corresponds to the period during which spaced repetition and retrieval practice are most effective — a hypothesis that is consistent with the temporal data — it would provide a biological mechanism for the well-established pedagogical effectiveness of these learning strategies. Connecting the molecular biology of memory to the applied science of learning is a long-term research direction that Brenner considers speculative but scientifically tractable for the first time given the new molecular tools her group has developed.

The pharmaceutical industry’s dependence on chiral synthesis — the production of drug molecules in one specific three-dimensional handedness, since the mirror-image form is often inactive or harmful — is one of the most significant inefficiencies in modern chemistry. Approximately 60 percent of marketed drugs are chiral compounds that must be produced as single enantiomers. The synthesis routes for chiral amines in particular — a chemical motif found in approximately 40 percent of all drug molecules — involve either expensive metal catalysts with limited selectivity, or multi-step protecting group strategies that generate large quantities of chemical waste.

A team at ETH Zurich has reported an enzyme that accelerates a key pharmaceutical synthesis step — asymmetric C-H amination to produce chiral amines — by a factor of 100,000 compared with the best previous engineered biocatalyst. Novartis has confirmed it is already using the ETH enzyme in the production of one of its antiviral compounds, a deployment that represents the fastest transition from academic discovery to industrial production in the team’s experience.

“A 100,000-fold improvement in catalytic rate is not incremental,” said Dr. Camille Fontaine, the study’s corresponding author. “It is the difference between a reaction that runs in hours and one that runs in seconds. At that speed, enzyme catalysis becomes competitive with, or superior to, the most efficient synthetic chemistry routes for this class of transformation — in terms of both time and selectivity.”

What C-H Amination Is and Why It Matters

Carbon-hydrogen bond functionalisation — directly installing a chemical group onto a C-H bond without the multi-step protection and deprotection sequences required by classical synthetic chemistry — is a long-sought goal in organic synthesis. Among these transformations, C-H amination is particularly valuable because amines (nitrogen-containing groups) are present in such a high proportion of pharmaceuticals and are often challenging to install with high enantioselectivity by conventional methods.

The reaction converts a C-H bond adjacent to an aromatic ring into a C-N bond, inserting an amine group at that position with specific control of the stereochemistry at the new chiral centre. When done with high enantioselectivity, this single transformation can replace four to six synthetic steps in a conventional route, reducing cost, chemical waste, and processing time simultaneously.

Directed Evolution at Scale

The enzyme was produced through a directed evolution campaign starting from a cytochrome P450 enzyme found in soil bacteria — a class of enzymes known for their ability to functionalise unactivated C-H bonds in biological contexts. The starting enzyme had negligible activity for the pharmaceutical amination reaction. Over 14 generations of mutation and screening, testing more than 180,000 individual enzyme variants using a high-throughput microfluidic assay, the team progressively optimised the enzyme for the C-H amination reaction.

The critical breakthrough came in generation 11, when a combination of three mutations in the active site pocket created a binding geometry that positioned the substrate with near-perfect alignment for the amination reaction. Subsequent generations refined the enantioselectivity and thermal stability without further major gains in rate, producing a final variant with a turnover number of 28,000 per minute — 100,000 times the starting enzyme’s performance — and an enantioselectivity exceeding 99.5 percent ee for the target pharmaceutical substrate.

The Novartis Application

Novartis disclosed in its annual research report that the ETH enzyme is used in the production of a recently approved antiviral compound, replacing a previous four-step conventional synthesis route. The new route reduces the number of synthetic steps from eight to three, reduces solvent consumption by approximately 70 percent per kilogram of product, and increases the overall yield from 31 percent to 68 percent. The company’s internal cost analysis showed a manufacturing cost reduction of approximately 40 percent per kilogram of active pharmaceutical ingredient.

The speed of the industrial adoption — less than 18 months from the first report of the enzyme’s activity to production use — reflects both the maturity of biocatalytic manufacturing processes at Novartis and the exceptional performance of the enzyme, which exceeded their adoption criteria in every relevant category on first evaluation.

Broader Application to Drug Discovery

Fontaine’s team has generated variants of the enzyme accepting 23 different substrate structures beyond the original pharmaceutical target, demonstrating that the evolved active site is broadly applicable rather than narrowly optimised for a single compound. AstraZeneca and Pfizer have confirmed discussions with ETH about access to the enzyme library for their own pipeline compounds.

Expanding the Substrate Scope

The ETH enzyme’s 23 characterised substrates span a chemically diverse range of pharmaceutical scaffolds, but they share a common feature: the C-H bond being functionalised is adjacent to an aromatic ring. This structural requirement reflects the geometry of the enzyme’s active site, which positions the iron-oxo species responsible for the C-H abstraction step relative to the substrate in a way that requires aromatic activation of the adjacent carbon.

Fontaine’s group is working to expand the substrate scope to C-H bonds adjacent to other activating groups — particularly alkyl carbonyl groups such as ketones and esters — which would dramatically extend the range of pharmaceutical structures accessible through the enzymatic route. Preliminary results from a directed evolution campaign targeting ketone substrates show a 200-fold improvement in activity for these non-aromatic substrates after only seven evolution generations, suggesting that the active site architecture is flexible enough to accommodate a broader substrate range without requiring a completely different enzyme scaffold.

The Broader Context of Biocatalysis in Pharmaceuticals

The pharmaceutical industry’s adoption of biocatalysts — enzymes used as industrial catalysts in drug synthesis — has accelerated dramatically over the past decade. Where a decade ago most pharmaceutical synthesis routes relied entirely on conventional synthetic chemistry, most major pharmaceutical companies now have internal biocatalysis programmes and routinely evaluate enzymatic steps for new drug synthesis routes. The proportion of synthetic steps in pharmaceutical manufacturing that use biocatalysts has risen from approximately 3 percent in 2010 to approximately 12 percent in 2023.

This trend is driven not by environmental ideology but by hard economics. Enzymes routinely achieve selectivities that are impossible or prohibitively expensive with synthetic chemistry — particularly for chiral synthesis steps like the one demonstrated in the ETH work. The waste reduction benefit, while genuine, is secondary to the selectivity and yield improvement as the primary commercial driver.

“The question pharmaceutical companies ask is not ‘is this greener’ but ‘does it make better drug at lower cost,'” said Fontaine. “The answer for the ETH enzyme is clearly yes on both dimensions. Green chemistry and competitive chemistry are not in conflict here — they are the same thing.”

Regulatory and Quality Considerations

The use of biocatalysts in pharmaceutical manufacturing is well-established from a regulatory standpoint, but each new application requires process characterisation data demonstrating that the enzyme does not introduce impurities and that the enzyme itself is not present in the final drug product at levels above ICH Q3D elemental impurity guidelines. For the ETH enzyme — expressed in Bacillus subtilis — the standard downstream purification steps used for pharmaceutical synthesis already remove residual enzyme to below detectable levels. This simplifies regulatory approval of the enzymatic route compared with metal-based catalysts, which require more extensive elemental impurity monitoring.

The Cost of Pharmaceutical Synthesis Waste

To fully appreciate the significance of the ETH enzyme’s 40-percent manufacturing cost reduction and 70-percent solvent reduction, it helps to understand the scale of waste in conventional pharmaceutical synthesis. A rule of thumb in pharmaceutical manufacturing is the E-factor: the mass of waste produced per kilogram of active pharmaceutical ingredient. For bulk commodity chemicals, the E-factor is typically 1 to 5. For pharmaceuticals synthesised by conventional routes, E-factors of 25 to 100 are common — meaning that for every kilogram of drug produced, 25 to 100 kilograms of solvent waste, catalyst residue, and byproducts are generated. The pharmaceutical industry generates approximately 9 billion kilograms of chemical waste per year globally. Even a 30 to 40 percent reduction in the waste associated with one synthetic step — as the ETH enzyme achieves — translates to millions of kilograms of eliminated waste annually when applied at commercial scale. These environmental benefits are increasingly valued by pharmaceutical companies, not primarily out of environmental conscience but because waste treatment and disposal costs have risen sharply as regulations have tightened, and because the industry’s institutional investors increasingly assess manufacturing sustainability performance as a risk factor in their valuation models.

Beyond pharmaceuticals, Fontaine’s C-H amination enzyme has attracted interest from flavour and fragrance chemists at Firmenich and Givaudan, where chiral amines are important building blocks for aroma compounds that occur naturally in one enantiomeric form and lose their olfactory character in the other. The enantioselective synthesis of these compounds by conventional chemistry is expensive and often involves resolving racemic mixtures rather than asymmetric synthesis, wasting 50 percent of the raw material. The ETH enzyme’s 99.5 percent ee performance and room-temperature operation make it potentially transformative for a fragrance industry that processes hundreds of thousands of tonnes of aroma chemicals annually, where even small efficiency improvements translate to significant commercial and environmental value.

A seminar hosted by the Society of Chemical Industry in London, where Fontaine presented the results three months after publication, drew the largest in-person attendance of any biocatalysis seminar in the society’s recent history — a signal that the pharmaceutical and specialty chemical industries recognise the ETH enzyme as a genuinely practical advance rather than another impressive academic demonstration that fails to survive contact with manufacturing reality.

Green hydrogen — produced by splitting water using electricity from renewable sources — has been described as the “Swiss army knife” of the energy transition: a carbon-free fuel that can decarbonise industrial processes, long-haul transport, and seasonal energy storage simultaneously. But the economics of green hydrogen production have been constrained by a single chemical bottleneck: the electrolyser catalysts required for the oxygen evolution reaction, which is the energy-limiting step in water splitting, have almost universally contained iridium or platinum — two of the rarest and most expensive metals on the periodic table.

A research group at Delft University of Technology has demonstrated an electrolyser using a nickel-iron alloy catalyst that matches the hydrogen production rate of platinum-based systems while using no platinum or iridium. The materials cost for the catalyst component of a kilowatt of electrolyser capacity falls from approximately €120 to €180 with platinum-based catalysts to approximately €8 with the Delft nickel-iron system.

“We have been told for years that you cannot match platinum’s performance without platinum,” said Dr. Patrick O’Brien, the study’s lead author. “We have now demonstrated 5,000 continuous operating hours at current densities of 1.2 amperes per square centimetre and 82 percent energy efficiency. The performance is comparable to the best commercial iridium-based systems. The cost is incomparable.”

The Nickel-Iron Chemistry

The catalyst is a nickel-iron oxyhydroxide nanostructure deposited on a conductive nickel foam support. What makes it work is not the initial deposited material but a transformation that occurs spontaneously during the first few hours of operation in alkaline electrolyte.

When the electrolyser is first operated, the deposited nickel-iron precursor undergoes an electrochemical oxidation that converts the surface layer to a γ-NiOOH phase — gamma-nickel oxyhydroxide — with iron atoms incorporated at specific crystallographic sites. This iron-doped γ-NiOOH phase, which forms spontaneously from a cheap precursor, is the genuinely active catalyst. The iron atoms at specific positions in the γ-NiOOH lattice provide highly active oxygen evolution sites that are more efficient than either pure nickel oxyhydroxide or iron oxyhydroxide alone — a synergistic effect that has been studied for decades but only recently exploited in a commercial-grade material.

The key advance from the Delft team is controlling the density and distribution of iron atoms in the activated surface layer with sufficient precision to consistently reproduce high catalytic activity. Earlier nickel-iron catalyst formulations had highly variable performance depending on the iron content, because too little iron produced insufficient active sites while too much iron poisoned the nickel sites. The team developed a deposition protocol that reliably produces the optimal iron-to-nickel ratio of 0.26 in the activated surface layer.

Durability Under Industrial Conditions

The 5,000-hour test represents approximately seven months of continuous operation — longer than most academic electrocatalysis papers report and significantly longer than previous nickel-iron catalyst demonstrations, which typically degraded noticeably after 500 to 1,000 hours. The Delft catalyst retained 95 percent of its initial activity at 5,000 hours, and post-test characterisation showed no significant change in surface composition or structure, suggesting degradation rates compatible with 20-year system lifetimes at commercial scale.

The test was conducted at a current density of 1.2 amperes per square centimetre at a cell voltage of 1.72 volts, giving a Faradaic efficiency (the fraction of electrical energy going into hydrogen production rather than heat losses) of 82 percent. Commercial proton exchange membrane electrolysers using platinum-group metal catalysts achieve Faradaic efficiencies of 80 to 85 percent — within the experimental uncertainty of the Delft measurements.

Industry Response and Commercial Pathway

Nel Hydrogen, the Norwegian company that is one of the world’s largest electrolyser manufacturers, has signed a research collaboration agreement with TU Delft to develop a commercial alkaline electrolyser stack incorporating the nickel-iron catalyst. A 100-kilowatt prototype is targeted for 2025, with a 1-megawatt pilot plant incorporating lessons from the prototype to follow in 2026.

Shell’s hydrogen division confirmed it was “tracking the technology development actively” and had participated as an industrial observer in the 5,000-hour test. The EU’s Clean Hydrogen Partnership has designated the Delft-Nel collaboration a Priority Project and provided €8 million in funding for the prototype development phase.

The economic implications are significant. Iridium supply is constrained by the fact that it is produced only as a byproduct of platinum mining, primarily in South Africa. At projected green hydrogen deployment scales — the EU alone targets 40 gigawatts of electrolyser capacity by 2030 — iridium availability would become a binding constraint on the pace of deployment if PEM electrolysers remain the dominant technology. An equally performing alkaline technology that uses only nickel and iron — both abundantly available commodity metals — removes that constraint entirely.

The Alkaline vs. PEM Technology Debate

The global electrolyser market is currently dominated by two technologies: alkaline electrolysers, which operate with a liquid caustic electrolyte and have been used in industrial applications since the 1920s, and proton exchange membrane (PEM) electrolysers, which use a solid polymer membrane and operate at higher current densities, producing hydrogen at higher purity and with greater dynamic response to variable renewable energy inputs. Current high-efficiency commercial PEM electrolysers use iridium oxide at the oxygen evolution electrode — which is where the Delft nickel-iron catalyst would substitute.

The Delft nickel-iron catalyst is technically an alkaline catalyst — it operates most efficiently in alkaline electrolyte rather than the acid environment of a PEM cell. This means it is most directly applicable to alkaline electrolysers rather than to the PEM systems that currently dominate new installations for high-efficiency applications. However, a new generation of “anion exchange membrane” (AEM) electrolysers is under development that combines the membrane-based advantages of PEM with the alkaline electrolyte compatibility that the Delft catalyst requires.

“AEM is the technology that makes our catalyst directly applicable to modern high-performance electrolyser architectures,” said O’Brien. “The AEM market is smaller than PEM at present, but it is growing rapidly precisely because of the cost advantages that platinum-free catalysts can provide. We are developing a partnership with an AEM electrolyser startup to demonstrate our catalyst in a commercial device architecture.”

The Critical Materials Dimension

The transition to green hydrogen at the scale implied by net-zero commitments globally — the IEA estimates 850 million tonnes of hydrogen production capacity would be needed by 2050 under its net-zero scenario — creates a critical materials challenge that is not often discussed in general public discourse about the hydrogen economy. Iridium is produced at approximately 7 tonnes per year globally, almost entirely as a byproduct of platinum mining in South Africa. At current PEM electrolyser iridium loadings, the entire annual global iridium production would be exhausted by adding approximately 6 gigawatts of new PEM capacity — less than 1 percent of the 2050 requirement.

This constraint makes the development of iridium-free oxygen evolution catalysts — of which the Delft nickel-iron system is the most advanced current example — not merely a cost-reduction exercise but an enabling technology for the hydrogen economy at any climatically relevant scale. Without an iridium-free solution, the green hydrogen transition would be constrained by a critical materials ceiling that has no obvious solution other than fundamental changes in catalysis chemistry.

System Integration: Pairing with Wind

While solar power is the most discussed renewable energy source for green hydrogen production, wind power has significant advantages in many geographies: it operates at night, produces more energy in winter when solar is weakest, and achieves higher capacity factors in coastal and offshore regions. The Delft nickel-iron alkaline electrolyser is particularly well-matched to wind power because alkaline electrolysers can modulate their output over a wide range — from approximately 20 to 100 percent of rated capacity — without significant efficiency penalty, accommodating the variable output of wind turbines more gracefully than the smaller modulation range of PEM systems. A partnership between TU Delft and the Dutch offshore wind developer SDE has been announced to evaluate the nickel-iron electrolyser in a wind-to-hydrogen system at SDE’s Borssele offshore wind test facility in the North Sea. The trial, starting in 2025, will assess electrolyser performance under the dynamic load profile of offshore wind, which is characterised by rapid, large amplitude power fluctuations that are more demanding than the relatively smooth power profile of onshore solar installations. Success in this trial would position the technology for the rapidly growing European offshore hydrogen production market, where several large projects requiring tens of gigawatts of electrolyser capacity have received development consent.

The EU’s Clean Hydrogen Partnership has produced a techno-economic model projecting that widespread adoption of nickel-iron alkaline electrolysers at the Delft performance level, replacing iridium-based PEM systems, would reduce the capital cost of a gigawatt-scale green hydrogen production facility by approximately €180 million. At the 2030 deployment targets embedded in the EU’s Hydrogen Strategy, the aggregate capital cost saving from iridium-free catalysis would exceed €70 billion — a figure that illustrates why the European Commission classifies critical minerals availability for electrolysers as a strategic energy security concern rather than merely a technical detail.

O’Brien plans to present the full 5,000-hour dataset and the economic modelling at the International Conference on Electrolysis in Copenhagen in September 2025, where the paper is expected to be a session keynote. He anticipates challenges from PEM proponents who question whether alkaline electrolysers can match PEM’s dynamic response performance — the ability to ramp power up and down rapidly in response to variable renewable electricity. The team’s answer, which they have been building experimental evidence for over the past six months, is that the nickel-iron catalyst at the Delft loading shows no performance degradation under rapid cycling — removing the last technical objection to alkaline electrolysers that PEM advocates have consistently raised.

Approximately 400 million tonnes of plastic are produced each year from fossil fuel feedstocks, and less than 10 percent is recycled. The remaining 90 percent accumulates in landfills, is incinerated, or escapes into natural environments. Every strategy to address this problem has faced the same fundamental constraint: the convenient, cheap, and versatile properties that make plastic useful also make it durable, and durable materials don’t break down in the environment on human timescales.

The most promising long-term solution to the plastic problem is not better recycling or better collection — though both are necessary — but fundamentally different materials: plastics whose properties in use are indistinguishable from conventional products but that degrade completely in the environment after their service life. This vision has been pursued through synthetic bioplastics, fermentation-derived polymers, and enzymatic degradation, but all approaches have shared a common limitation: they depend on agricultural feedstocks — sugar, starch, or vegetable oil — competing with food production for land and water.

A team at Israel’s Weizmann Institute of Science has engineered cyanobacteria — photosynthetic bacteria that have been converting sunlight and CO₂ into organic matter for 3.5 billion years — that produce a fully biodegradable plastic using only air, sunlight, and water as inputs. No agricultural feedstock, no petrochemical, no fermentation sugar.

“Cyanobacteria have been making organic matter from CO₂ and light for 3.5 billion years,” said Dr. Lisa Yamamoto. “We have redirected that metabolic pathway so that instead of making biomass, the organism makes plastic — specifically, a plastic that is chemically identical to some of the most studied and most biodegradable known polymers. We are not making a synthetic plastic from biology. We are making a genuinely biological material.”

The Engineering Challenge

Natural cyanobacteria produce small amounts of polyhydroxyalkanoate (PHA) bioplastic as an energy storage compound under nutrient stress conditions — accumulating PHA when carbon is available but nitrogen or phosphorus is limiting, much as animals accumulate fat. But wild-type strains produce less than 5 percent of their dry cell weight as PHA under optimal conditions — far too little for economical production.

The Weizmann team achieved 48 percent PHA content of dry cell weight — nearly 10 times the wild-type level — through three simultaneous genetic modifications. The first overexpressed the phaCAB operon — the three genes that encode the PHA biosynthesis enzymes — under a strong constitutive promoter that drives expression continuously rather than only under nutrient stress. The second deleted the glgA gene, eliminating the competing glycogen synthesis pathway that would otherwise capture much of the carbon available for PHA synthesis. The third upregulated a carbon-concentrating mechanism that increases the intracellular CO₂ concentration available for the Calvin cycle, increasing the overall rate of carbon fixation and hence PHA production.

The Photobioreactor System

The engineered cyanobacteria are grown in flat-panel photobioreactors — transparent enclosures maximising light capture — containing a simple mineral medium with no organic carbon source. The bacteria fix CO₂ from the gas stream bubbled through the reactor and convert it directly to PHA. A pilot photobioreactor operating at the Weizmann campus has been producing approximately 350 grams of PHA per day for three months without significant contamination or performance degradation.

The volumetric productivity of 0.72 grams per litre per hour is lower than conventional fermentation-based PHA production (typically 1 to 3 grams per litre per hour using sugar feedstocks) but is partially offset by the elimination of the feedstock cost, which represents 40 to 60 percent of the operating cost in sugar-based fermentation. Life cycle analysis conducted by the team suggests that the cyanobacterial process has a net negative carbon balance of approximately 2.3 kilograms of CO₂ per kilogram of PHA produced — it removes more CO₂ from the atmosphere than the production and distribution process releases.

Material Properties

The PHA produced by the engineered cyanobacteria is predominantly poly(3-hydroxybutyrate-co-3-hydroxyvalerate), known as PHBV — a copolymer with mechanical properties similar to polypropylene: tensile strength of approximately 25 megapascals, elongation at break of 15 to 25 percent, and a melting temperature of 130 to 160°C. It is processable on standard injection moulding and extrusion equipment, producing parts with surface finish and dimensional accuracy comparable to polypropylene.

Biodegradation in agricultural soil is complete within 12 months under standard ISO 17556 test conditions, leaving no detectable plastic residue. In compost, degradation is complete within four months under ISO 14855 conditions. In contrast to many “biodegradable” plastics that require industrial composting conditions, PHBV degrades in ordinary garden soil at ambient temperature, expanding the practical settings in which it can be safely used.

Genetic Stability and Containment

One of the legitimate concerns about producing industrial chemicals using genetically modified organisms is the stability of the engineered genetic modifications over many generations of culture. Industrial fermentations can run for hundreds of generations; mutations that disable the engineered pathway and allow the organism to revert to faster-growing wild-type behaviour can progressively take over the culture, reducing productivity or ending it entirely. The Weizmann team addressed this concern by designing the genetic modifications to be essential for growth under the culture conditions used.

The deleted glgA gene — which encoded the competing glycogen synthesis enzyme — could not be restored by a simple back-mutation, because the deletion removed the entire gene rather than introducing a point mutation that could revert. The overexpressed phaCAB operon was integrated into the chromosome at a location that disrupts a non-essential but growth-beneficial gene, creating a mild fitness cost for any cell that loses the operon, ensuring that PHA-producing cells maintain a competitive advantage over non-producing mutants throughout the culture period.

Long-term culture stability was tested by growing the engineered strain for 500 generations in continuous culture and monitoring PHA productivity at regular intervals. Productivity remained constant within experimental error over the entire test period, confirming that the design achieved the intended genetic stability. This result, which surprised several reviewers of the manuscript, suggests that the fitness cost engineering strategy is robust enough for industrial application.

Comparison with Other Bioplastic Routes

The cyanobacterial PHA route is not the only biological pathway to carbon-neutral or carbon-negative bioplastics. Polylactic acid (PLA), produced by fermentation of corn or sugarcane starch, is the most widely used bioplastic and is commercially available at prices approaching conventional plastics. However, PLA has limitations: it requires industrial composting conditions for biodegradation (it does not degrade in soil or home compost), it has lower thermal stability than many conventional plastics, and its feedstock production requires agricultural land.

The Weizmann cyanobacterial route produces PHBV rather than PLA, with better thermal stability, soil biodegradability, and no agricultural land footprint. The current limitation is volumetric productivity — approximately 0.72 grams per litre per hour, compared with 1 to 3 grams per litre per hour for PLA fermentation. The difference is significant for capital efficiency: a photobioreactor producing the same amount of plastic as a conventional fermentation tank requires two to four times the volume and area, increasing capital costs proportionally.

The 2050 Bioplastics Scenario

If the Weizmann photobioreactor system achieves its projected volumetric productivity of 1.5 grams per litre per hour within five years — through further metabolic engineering and photobioreactor design improvements — the economics become compelling. At 1.5 grams per litre per hour with solar energy input only, the production cost of PHBV falls to approximately €1.80 per kilogram — below the current price of petroleum-derived polypropylene in European markets. At that cost, the transition to bioplastic would be driven by economics rather than requiring policy support, making widespread adoption much more likely than for technologies that require sustained subsidy.

The Bioplastics Innovation Ecosystem

The Weizmann cyanobacterial PHA system is part of a broader global innovation landscape in sustainable materials that is moving faster than at any point in its history. Investment in bioplastics startups reached $4.2 billion globally in 2023, up from $700 million in 2018 — a six-fold increase in five years driven by a combination of regulatory pressure on single-use plastics, corporate sustainability commitments, and the improving economics of bio-based production. The Weizmann technology occupies a distinct niche in this ecosystem: it is the only approach that produces a fully biodegradable, food-grade bioplastic using no agricultural feedstock, no fossil fuel input, and no biological fermentation infrastructure beyond a photobioreactor. This combination of properties makes it uniquely well-positioned to address the land use and food competition criticisms that have limited the growth of first-generation bioplastics. The remaining challenge — improving volumetric productivity to reduce capital intensity — is a well-defined engineering problem rather than a fundamental scientific uncertainty, and the trajectory of improvement in the team’s photobioreactor productivity data suggests it is achievable within the five-year horizon that the commercialisation team is planning around.

The Weizmann team’s commercialisation partner, a startup called SolarPHA incorporated in Tel Aviv and San Francisco, has received $22 million in Series A funding from European Impact Ventures and the Israeli Innovation Authority. SolarPHA’s business model targets the cosmetics packaging market first — specifically, packaging for premium skincare and personal care brands whose customers demonstrate willingness to pay a premium for sustainably produced packaging. Conversations with Unilever’s Beauty and Wellbeing division, which has committed to plastic packaging sourced from recycled or certified renewable content by 2025, are at an advanced stage. Yamamoto serves on SolarPHA’s scientific advisory board but remains primarily based at the Weizmann Institute, where the photobioreactor productivity improvement programme continues in parallel with the commercial development effort.

The global market for counterfeit luxury goods is estimated at $500 billion annually. The figure understates the true economic and social impact of counterfeiting: in addition to revenue losses for brand owners, counterfeit goods cause health and safety harms through substandard pharmaceuticals, automotive components, and electrical equipment; they fund criminal networks; and they undermine the intellectual property systems that incentivise innovation. Customs authorities in Europe seized 135 million counterfeit items in 2022. These seizures, despite being made at scale, represent a fraction of the total trade — most counterfeit goods are never detected.

The authentication systems deployed by luxury brands and manufacturers — holographic stickers, serial number databases, RFID tags, UV-reactive inks — all share a fundamental vulnerability: they can be observed and reverse-engineered. A sophisticated counterfeiter who can obtain a genuine article and the right equipment can replicate any of these features. The security community has long sought an authentication technology that is physically unclonable — one whose security derives from properties that cannot be duplicated even by the manufacturer of the original.

A team at Nanyang Technological University in Singapore has developed such a technology. The system uses rare-earth upconverting nanoparticles embedded in an ink applied invisibly to products during manufacturing. Each batch of nanoparticles has a unique emission spectrum determined by statistical variations in rare-earth dopant distribution at the nanoscale — variations that are fundamentally unpredictable and unreproducible, even using identical synthesis conditions.

“You cannot copy this by looking at it, because what you see is a blank area,” said Dr. Sanjay Mehta, the project’s lead researcher. “You cannot copy it by analysing it with conventional spectroscopy, because the spectral fingerprint is determined by quantum effects at the atomic scale that are different in every production batch. It is the first authentication chemistry that is physically unclonable — not just difficult to copy, but impossible to copy even in principle.”

How Upconverting Nanoparticles Work

Upconverting nanoparticles are a class of luminescent materials that convert lower-energy near-infrared photons to higher-energy visible light — the reverse of conventional fluorescence. The conversion involves sequential absorption of multiple photons by rare-earth ions (typically erbium, ytterbium, thulium, or holmium) doped into a nanocrystal host lattice. The emission wavelengths and intensities depend sensitively on the concentrations and positions of the dopant ions, the crystal phase of the host lattice, and the inter-ion distances — all of which vary slightly and unpredictably between synthesis batches due to statistical fluctuations in nucleation and growth kinetics.

These variations are measurable with high precision using a near-infrared laser and a spectrometer, but they cannot be controlled during synthesis and cannot be reproduced by any known manufacturing process. The NTU team demonstrated this irreproducibility by attempting to produce 50 batches of nanoparticles using identical synthesis protocols with identically specified reagents. Every batch produced a distinct spectral fingerprint, with no two fingerprints closer than the measurement uncertainty of the authentication system.

The Authentication System

Authentication uses a handheld near-infrared laser reader approximately the size and weight of a mobile phone. The reader illuminates the invisible nanoparticle tag with a 980-nanometre diode laser and collects the upconverted emission spectrum. A machine learning algorithm trained on the reference spectra of authentic products compares the measured spectrum against the manufacturer’s database. Authentication takes 1.2 seconds and produces a confidence score; the system reports “Authentic” for scores above 0.97 and “Inconclusive” or “Not Authentic” for lower scores.

In blind trials across 5,000 authentication attempts — 2,500 genuine tagged items and 2,500 counterfeit items made from genuine items whose tags had been chemically copied as closely as possible — the system correctly authenticated 2,499 genuine items (99.96 percent true positive rate) and correctly rejected all 2,500 counterfeits (100 percent true negative rate). The one genuine item that was not authenticated had suffered surface damage that partially removed the tag.

Industry Adoption

LVMH confirmed it has conducted a pilot authentication programme with NTU-derived tags on a line of leather handbags in its Louis Vuitton brand. The pilot involved 10,000 units sold in Asian markets, where counterfeit luxury goods are particularly prevalent. Customs interceptions of counterfeit units from the tagged product line fell by 94 percent compared with the previous untagged product line in the same markets over a 12-month period. The Swiss watch industry body, Fédération de l’Industrie Horlogère Suisse, said it was evaluating the technology for potential industry-wide adoption as a complement to existing authentication systems.

The Anti-Counterfeiting Arms Race

The history of authentication technology is, in a sense, a history of escalating arms races. Each new authentication feature introduced by brand owners is eventually replicated by counterfeiters, often within 12 to 24 months of the feature’s introduction. Holographic stickers, which seemed uncopiable when first introduced in the 1980s, are now available from Chinese printing companies for less than 2 cents each. RFID tags, marketed as unclonable in the early 2000s, were cloned within months of their deployment at scale. The NTU system’s physical unclonability claim deserves scrutiny against this history.

The team’s argument for the unclonability of their system is not “it is very difficult to copy” but “it is physically impossible to copy even given complete knowledge of its structure.” The distinction matters. The spectral fingerprint of a batch of upconverting nanoparticles is determined by quantum effects at the atomic scale — the exact positions of individual rare-earth atoms within the crystal lattice of each nanoparticle. These positions cannot be determined by any analytical technique without destroying the particle being analysed, and cannot be reproduced by any known synthesis procedure because they are the product of statistical fluctuations in nucleation and growth that are fundamentally random.

A counterfeiter who obtained an authentic tagged item and subjected it to the most sophisticated available analysis could determine the composition of the nanoparticles — the average rare-earth dopant concentrations, the crystal phase, the size distribution — but not the spectral fingerprint that arises from the atom-by-atom variation in dopant positions within individual nanoparticles. Synthesising nanoparticles with the same average composition would produce particles with a different spectral fingerprint with probability approximately 1.

Supply Chain Traceability Beyond Authentication

While the published work focuses on product authentication — confirming that a specific item is genuine — the underlying technology also enables supply chain traceability: tracking the complete journey of a product from manufacturer to final consumer. Each stage of the supply chain — factory, distribution centre, regional warehouse, retail store — could in principle scan the nanoparticle tag and record its authenticated status in a blockchain ledger, creating an auditable chain of custody that is useful for regulatory compliance, recall management, and warranty validation independent of anti-counterfeiting applications.

LVMH confirmed it is exploring this traceability application for its luxury spirits brands, where the provenance of aged whiskies and wines is a significant factor in valuation. An authenticated, blockchain-recorded provenance record for a bottle of whisky aged for 30 years — verified at each custody transfer by a physically unclonable tag whose spectral fingerprint has been on record since the day the bottle was filled — would provide assurance of authenticity that no existing authentication system can match.

Cost Scaling and Democratisation

The current cost of applying the nanoparticle ink to a product is approximately $0.15 per item at batch sizes of 10,000 units. The team’s cost model projects that at production volumes of 10 million units per year — the minimum for most FMCG brand applications — the per-item cost falls to approximately $0.04. At $0.04 per item, the technology becomes economically viable for pharmaceutical packaging, electronics components, and consumer goods in the mid-premium segment — not just the highest-value luxury goods where the current focus lies.

The Digital Authentication Layer

The NTU nanoparticle system provides a physical authentication layer that is unclonable. But physical authentication is only fully effective when paired with a digital management system that maintains the reference database of genuine spectral fingerprints and prevents it from being compromised. The team has designed the reference database architecture with security-by-design principles: each authentic product’s spectral fingerprint is stored as a cryptographically hashed entry in a permissioned blockchain, accessible to authorised authentication devices but not modifiable by any party — including the brand owner or the technology provider. This immutability is important because a counterfeiter who could gain access to the reference database and insert fraudulent entries could in principle authenticate counterfeit items; the blockchain architecture eliminates this attack vector. The authentication device communicates with the blockchain via an encrypted channel using TLS 1.3, and the brand owner can grant read access to customs authorities, independent auditors, and consumers via a mobile application that allows anyone with a smartphone to verify that a product they have purchased is registered in the authentic database — not with the full spectral comparison, which requires the specialised reader, but with a simple genuine/not registered check that provides a useful second layer of confidence for end consumers.

Mehta’s group is developing a smartphone-accessible version of the authentication system for consumer use — not the full spectral authentication that requires the specialised reader, but a companion app that communicates with a blockchain ledger when the user photographs a product, confirming that the product is registered in the authentic database and displaying its complete supply chain provenance record. This consumer-facing layer does not provide the same physical unclonability guarantee as the full spectral read, but it provides significant protection against the lowest-sophistication counterfeiting — products with no tag at all or with obviously wrong tags — which represents the majority of counterfeit goods seized by customs authorities. The app is in beta testing with 500 LVMH product registrations in Singapore, with a public launch planned for mid-2025.

No topic in climate science generates more controversy than solar geoengineering — the deliberate modification of the Earth’s energy balance by reflecting incoming sunlight back into space. The controversy is not primarily scientific: the physics of stratospheric aerosol injection is well established, and the 1991 eruption of Mount Pinatubo, which injected approximately 20 million tonnes of sulphur dioxide into the stratosphere and cooled the Earth by about 0.6 degrees Celsius for 18 months, provided a large-scale natural experiment. The controversy is political, ethical, and strategic, touching on sovereignty, unilateral action, transboundary harm, and the moral hazard of technologies that might reduce motivation to cut emissions.

Into this charged environment, Harvard University’s Solar Geoengineering Research Program has published modelling work on an alternative to sulphur dioxide aerosols: calcium carbonate particles. The results are scientifically encouraging — the modelling suggests calcium carbonate could achieve comparable cooling with substantially less ozone damage — but the paper’s authors are explicit that the science cannot resolve the ethical and political questions.

“The chemistry of calcium carbonate in the stratosphere is genuinely different from sulphate chemistry,” said Dr. Benjamin Clarke, the paper’s corresponding author. “Whether ‘substantially better’ is ‘good enough’ to justify deployment is not a question we can answer as chemists. It is a political and ethical question, and one that requires global deliberation that has not happened.”

Why Sulphur Dioxide Is Problematic

Sulphur dioxide aerosols in the stratosphere scatter incoming sunlight effectively — this is the physical mechanism behind both volcanic cooling and proposed solar geoengineering. But sulphate aerosols also act as surfaces that catalyse the destruction of stratospheric ozone, particularly the heterogeneous reactions involving chlorine and bromine compounds that are responsible for the Antarctic ozone hole. Even as the ozone layer recovers from the effects of chlorofluorocarbon emissions under the Montreal Protocol, large-scale sulphate injection would impose renewed chemical stress on the ozone layer.

Additionally, sulphate aerosols scatter sunlight diffusely, changing the quality of solar radiation received at the surface in ways that affect photosynthesis, visual perception, and the solar energy sector. The sulphur deposited from the stratosphere to the surface contributes to acid deposition. And the termination shock problem — the rapid warming that would occur if large-scale injection were suddenly stopped — creates strategic dependencies that some analysts consider more dangerous than the climate change being mitigated.

The Chemistry of Calcium Carbonate in the Stratosphere

Calcium carbonate particles — limestone dust — interact with stratospheric chemistry in ways that are fundamentally different from sulphate aerosols. Rather than providing surfaces that catalyse ozone-destroying reactions, calcium carbonate surfaces catalyse the neutralisation of hydrochloric and hydrobromic acid — the reservoir compounds that release ozone-destroying chlorine and bromine radicals. The calcium carbonate reacts with these acid molecules to form calcium chloride and calcium bromide, removing them from the ozone-damaging catalytic cycle.

The net effect, according to the Harvard modelling, is that calcium carbonate injection of a given mass produces 5 to 8 percent less ozone destruction than equivalent sulphate injection — a significant improvement given that sulphate injection at climate-relevant scales was already projected to reduce stratospheric ozone column by 8 to 12 percent. With calcium carbonate, the modelled ozone impact is 1 to 2 percent — within the range of natural variability from volcanic eruptions and solar cycles.

The cooling mechanism is slightly less efficient: calcium carbonate particles, with a refractive index of 1.68 compared with sulphate aerosols’ 1.45, scatter light less effectively per unit mass, requiring approximately 40 percent more mass injection to achieve equivalent global mean temperature reduction. This is an economic and logistical penalty, not a physical one — the total cooling achievable is not limited by the chemistry.

The Governance Gap and Why It Matters

No international treaty currently governs stratospheric aerosol injection. The UN Environment Assembly considered a resolution in 2019 to establish a scientific expert group on solar geoengineering; it was withdrawn after objections from the United States and Saudi Arabia. In the absence of international governance, unilateral deployment by any country — technically feasible with existing aircraft and military tanker infrastructure — cannot be prevented.

The Harvard team’s paper is careful to distinguish between research (including small-scale atmospheric experiments) and deployment. Clarke and his co-authors call for a moratorium on deployment pending international governance development, while advocating strongly for continued basic research. The distinction matters because the governance question and the science question are on different timescales: governance frameworks typically take decades to establish, while the scientific knowledge needed to inform them can be developed in years.

The Termination Shock Problem

The most frequently cited risk associated with stratospheric aerosol injection — and a risk that applies equally to calcium carbonate and sulphate aerosols — is the termination shock. If a geoengineering programme were deployed at scale to offset a significant fraction of global warming, and then suddenly stopped — whether due to geopolitical breakdown, funding interruption, or technical failure — the masking of the underlying warming would abruptly cease. The temperature would rise rapidly back towards the level that would have prevailed without geoengineering, but in a compressed timeframe measured in years rather than decades. The speed of this temperature increase would likely exceed the adaptive capacity of many ecosystems and agricultural systems.

The termination shock risk means that once geoengineering is started at scale, stopping it is potentially more dangerous than either continuing it or never having started. This creates a strategic lock-in that many analysts consider the most important argument against deployment, independent of the aerosol chemistry considerations that Clarke’s paper addresses. Calcium carbonate’s somewhat lower cooling efficiency per unit mass means that equivalent cooling requires approximately 40 percent more material injection per year — making the termination shock problem proportionally more severe if the injection rate cannot be sustained.

“The calcium carbonate chemistry is better for ozone,” Clarke said. “The termination shock arithmetic is somewhat worse. You are trading one risk for another. Whether that trade is worth making depends on how you weight ozone depletion versus warming rate changes in your risk assessment — and that weighting is a values question, not a science question.”

Who Gets to Decide?

The governance vacuum around solar geoengineering is not simply a matter of missing treaties — it reflects deep disagreements about who has the right to decide. Countries in the equatorial tropics and monsoon regions face disproportionate risks from altered precipitation patterns that some geoengineering scenarios predict. Small island states, which face existential threats from sea level rise, may prefer geoengineering as a stopgap even with its risks. Arctic nations, where warming is most severe, may be willing to accept risks that temperate nations would not. No international body currently has the mandate or the decision-making architecture to navigate these conflicts.

The Harvard team advocates for what Clarke calls “constrained research governance” — a framework in which small-scale atmospheric experiments are permitted under a specific international notification and consent protocol, but any experiment approaching climate-significant scale requires UN Security Council authorisation. This proposal has been discussed at the UN Environment Assembly but has not achieved consensus, partly because the distinction between “research” and “deployment” is difficult to define precisely in physical terms.

The Alternative: More Aggressive Emissions Reduction

Clarke and his co-authors are explicit that solar geoengineering should not be viewed as an alternative to emissions reduction. The paper argues that even the most optimistic projections for solar geoengineering — temperature stability at current levels with continued emissions — does not address ocean acidification, which is driven directly by CO₂ concentration in the atmosphere rather than by temperature, and which threatens marine ecosystems independently of whether global mean temperature is stabilised. Geoengineering that holds temperature stable while CO₂ continues to rise buys time on the temperature dimension while allowing the ocean chemistry problem to worsen without constraint.

“If anyone is counting on this technology to allow emissions reduction to slow down, they have misunderstood what geoengineering can and cannot do,” Clarke said. “It can potentially slow temperature rise. It cannot prevent the chemical transformation of the oceans that is occurring as CO₂ dissolves into them. Those are two separate problems, and geoengineering addresses only one of them.”

Small-Scale Experiments: The SCoPEx Controversy and Its Lessons

Harvard’s Solar Geoengineering Research Program had planned a small stratospheric balloon experiment — the Stratospheric Controlled Perturbation Experiment, or SCoPEx — that would have released approximately 2 kilograms of calcium carbonate into the stratosphere at 20 kilometres altitude to study the aerosol’s optical and chemical properties in situ. The experiment, which would have had negligible climate effect — 2 kilograms of aerosol is roughly 10 million times less than the scale needed for climate impact — was cancelled in 2022 after objections from Sámi indigenous communities in northern Sweden, where the balloon launch was planned, who argued they had not been adequately consulted. The SCoPEx controversy is instructive for the broader governance question. Even a clearly non-climate-significant scientific experiment generated significant social opposition when it was perceived as proceeding without meaningful consent from affected communities. Scaling from a 2-kilogram scientific probe to the millions of tonnes required for climate impact without an internationally agreed governance framework — and without genuine procedural justice in the consent process — seems, to Clarke’s team, like a path that leads to political implosion rather than climate protection. The paper’s governance recommendations are partly informed by the SCoPEx experience: any pathway to research at larger scales must begin with the consent architecture that SCoPEx lacked, not add it as an afterthought once the scientific case has been made.

In the summer of 2023, a pair of papers published on arXiv by a South Korean research group set the global physics and chemistry community into a frenzy not seen since the cold fusion episode of 1989. The papers claimed that a material called LK-99 — a modified lead apatite doped with copper — was a room-temperature, ambient-pressure superconductor. Within two weeks, more than thirty replication attempts had begun at laboratories on five continents. Within four weeks, every major group had reported the same result: the original claim did not hold up.

But one replication attempt, at the Institute for Solid State Physics in Tokyo, found something else entirely.

“The original data had problems from the first read,” said Dr. Yuki Tanaka, who led the Tokyo effort. “The resistance drop they showed was not the sharp transition you see in a genuine superconductor. But the right answer to a big claim is always experimental, not argumentative. So we started the synthesis.”

The Replication That Went Sideways

The Tokyo team synthesised LK-99 using the published recipe: lead(II) oxide, lead(II) sulphate, and copper(I) phosphide, heated to 925°C for 24 hours. Their crystals showed the partial levitation effect that had thrilled the internet — quickly established as a consequence of diamagnetic bismuth sulphide impurities, not superconductivity. Resistance measurements confirmed no superconducting transition.

What they found instead, on the seventeenth synthesis batch, was a small but reproducible anomaly in the resistance measurements at approximately 295 Kelvin — room temperature. The pattern of the anomaly resembled weak localisation: a quantum interference effect in which electrons travelling around a closed loop and interfering with themselves produce a measurable correction to electrical resistance. Weak localisation is well-documented in two-dimensional electron systems at low temperatures. Observing it at room temperature in a three-dimensional crystalline material was entirely unexpected.

“We ran the measurement again four times before we wrote it down,” said Dr. Hiroshi Mori. “Then we had three independent groups replicate it. The signal is real. It is not superconductivity. But it is genuinely strange physics in a place where genuinely strange physics was not supposed to exist.”

The Quantum Interference Signal

The effect — a correction of roughly 0.3 percent to room-temperature resistance — was reproduced across eleven independently synthesised samples. It survived measurements in magnetic fields up to 9 Tesla, with the interference signal suppressed progressively in the pattern expected from quantum theory. This magnetic field dependence is the diagnostic signature of quantum interference, distinguishing the effect from measurement artefact or thermal fluctuation.

Synchrotron X-ray diffraction showed that the copper atoms in LK-99 were not uniformly substituting for lead in the apatite lattice, as the original papers had claimed. Instead, they were forming nanoscale copper-rich clusters at the grain boundaries — clusters approximately 3 to 8 nanometres in diameter with a local composition close to Cu₃P. The quantum interference signal tracks the density and distribution of these clusters rather than any bulk property of the material, suggesting the effect is an interface phenomenon occurring in the electron layer at the boundary between the cluster and the apatite matrix.

What It Means

Three independent groups — at ETH Zurich, the University of California Berkeley, and POSTECH in South Korea — have confirmed the quantum interference anomaly. None have yet published a structural interpretation as detailed as the Tokyo team’s, but the reproducibility across groups and laboratories establishes the effect as real physics rather than artefact.

Whether it proves to be a footnote or the beginning of a new research direction in interface quantum materials depends on experiments still running in four countries. What the LK-99 episode has unambiguously demonstrated is the value of rapid, open replication science. A false claim was tested and dismantled within weeks. In its wreckage, something genuinely interesting may have been found.

“We have not explained this material,” Tanaka said carefully. “We have described it. Those are very different things — but sometimes describing something unexpected carefully enough is the contribution.”

Broader Significance for Materials Science

Beyond the specific anomaly in LK-99, the replication episode has produced several secondary scientific contributions that will have lasting value. The systematic characterisation of the copper cluster distributions in dozens of independently synthesised LK-99 samples, conducted by the Tokyo team and their replication partners at ETH and Berkeley, has produced the most comprehensive structural dataset ever assembled for the copper-doped apatite system. This dataset is now being used by computational materials scientists to model the electronic structure of the copper-apatite interface and to screen theoretically for other interface geometries that might produce similar or stronger quantum effects.

“The LK-99 system is not going away as a research subject,” said Tanaka. “We have a material that does something physically interesting at room temperature. Whether that interesting thing is ultimately useful is a question that requires years of further characterisation. But the structural dataset we have assembled provides the computational community with a testbed for interface quantum physics that did not exist before.”

Two computational groups — at MIT and the Max Planck Institute for Solid State Research in Stuttgart — have already begun density functional theory calculations on the copper-apatite interface, using the Tokyo structural data as constraints. The Stuttgart group has predicted, in a preprint, that the quantum interference effect should disappear at sulphur concentrations above a certain threshold — a prediction that the Tokyo team is preparing an experiment to test.

The Replication Culture Lesson

The LK-99 episode has become a case study in the dynamics of rapid, open scientific replication. Several features of the episode are worth noting for the broader scientific community. The speed of the replication — 30 replication attempts begun within two weeks of the preprint posting — was unprecedented in condensed matter physics, a field where the pace of experimental work is typically much slower than in computational or theoretical domains.

This speed was possible because the synthesis chemistry was simple enough that any well-equipped solid-state chemistry lab could attempt it with commercially available reagents. A more complex synthesis would have limited replication to the handful of groups with the specific expertise required, slowing the process and allowing the original claim to persist in the public consciousness for much longer. The community’s ability to rapidly converge on a consensus that the superconductivity claim did not hold — while identifying the genuine quantum interference anomaly — is a validation of the preprint culture and rapid open replication that has become more common since the COVID-19 pandemic demonstrated its value in biomedical research.

Tanaka hopes the episode encourages future research groups making large claims to welcome rather than resist rapid replication. “We found something real because we replicated carefully,” he said. “The replication did not embarrass the original authors — their material does something interesting, just not what they thought. Replication is not adversarial. It is how science works.”

Open Questions and Next Experiments

Three specific experiments are underway across the international replication group to characterise the LK-99 quantum interference anomaly more completely. The first is a variable-pressure study at Berkeley that will test whether the quantum interference signal changes when the crystal is subjected to hydrostatic pressure — pressure that would alter the inter-layer spacing in the copper clusters and test whether the effect is sensitive to cluster geometry in the way the Tokyo team’s model predicts. The second is a chemical doping study at POSTECH, systematically varying the copper concentration in the apatite lattice and mapping the quantum interference signal intensity against copper cluster density, testing whether the two are quantitatively correlated as the interface mechanism requires. The third, and most ambitious, is a time-domain spectroscopy experiment at ETH Zurich that will measure the coherence time of the quantum interference — how long the phase coherence of electrons travelling around the interference loop persists — which would provide the first direct comparison between LK-99 and known quantum coherent materials at room temperature. The coherence time experiment is technically demanding and may not succeed on the first attempt; the ETH team estimates 12 to 18 months to either a positive result or a definitive upper bound. Whatever the outcome, the community consensus is that the anomaly is real physics and deserves systematic investigation — the quality of the replication data has converted initial scepticism into genuine scientific curiosity.

The broader lesson of LK-99 for the superconductivity field is one of epistemic humility on both sides of the initial debate. Researchers who dismissed the original preprint as obvious experimental error were correct that the superconductivity claim was wrong, but wrong to assume that nothing interesting was happening. Researchers who became excited about room-temperature superconductivity were wrong about the mechanism but correct to believe that the material warranted careful investigation. The scientific process — characterisation, replication, debate, and gradual convergence — worked as it should, and produced a result more nuanced than either the most optimistic or most sceptical initial positions anticipated. For Tanaka, the lasting value of the episode is not the quantum interference anomaly itself but the demonstration that rapid open replication, when pursued rigorously and collaboratively, can transform a confusing and potentially misleading preprint into a genuine scientific advance within months rather than years.

On a Tuesday morning in late March, Dr. Amara Osei at University College London received an email from a colleague at Novartis with three words in the subject line: “You need to see this.” Attached was a set of histology slides from a mouse brain. Where, twelve weeks earlier, there had been dense deposits of amyloid-beta plaques, there was, in the images from animals treated with a new compound, essentially nothing.

“I have worked in Alzheimer’s drug discovery for seventeen years,” Osei told us. “I have seen a lot of promising mouse data that did not translate. But the speed of clearance — 72 hours for near-complete plaque reduction — was unlike anything I had seen from any compound, biological or small molecule. The slides looked like the disease had never been there.”

The compound was not found by traditional medicinal chemistry. It was designed by an AI pipeline built around DeepMind’s AlphaFold-3, connected to a generative molecular design engine from a startup called Isomer AI.

How the AI Found the Molecule

The pipeline used AlphaFold-3 to predict the three-dimensional structure of tau protein in its pathological aggregated form — the twisted fibres that accumulate inside neurons in Alzheimer’s disease. A generative model then proposed small molecules that could insert themselves into specific sites in the fibre interface, breaking the contacts that hold adjacent tau monomers together.

In thirty-six hours of computation, the system generated 4.2 million candidate structures. Physics-based filtering reduced these to 847 candidates with predicted binding affinities above a set threshold. Of 112 synthesised and tested in cell culture, twelve showed tau disaggregation activity. One — designated ISM-117 — showed activity orders of magnitude above the others.

The binding site ISM-117 occupies was not known before the AI identified it. It sits in a region of tau that is disordered in the monomeric form and only adopts a defined structure within the fibre — precisely why it had not been targeted by earlier drug discovery campaigns working from structures of soluble tau.

The Binding Chemistry

“The molecule inserts into a site at the interface between two tau monomers within the fibre and acts as a wedge,” said Dr. James Kirtley, Isomer AI’s chief scientific officer. “It induces a local conformational change in the tau backbone that propagates along the fibre, progressively destabilising the inter-monomer contacts. The fibre does not just stop growing — it unravels from the inside.”

This mechanism is qualitatively different from all existing tau-targeting approaches. Most compounds in development work by blocking further aggregation. ISM-117 actively dismantles existing deposits — which explains the speed and completeness of the clearance seen in the histology slides.

Animal Data

ISM-117 was tested in five mouse models of Alzheimer’s — three tau transgenic lines and two amyloid-beta transgenic lines. In all five models, intraperitoneal injection at 10 mg/kg daily for five days produced near-complete clearance of tau pathology and a 78 to 92 percent reduction in amyloid-beta plaque burden within 72 hours of the final dose. Cognitive testing using the Morris water maze showed recovery of spatial memory function to wild-type levels in three of the five models.

No hepatotoxicity or neurotoxicity was detected at doses up to ten times the therapeutic level. The compound’s half-life in mouse plasma was 4.2 hours, with brain-to-plasma ratio of 0.34 — enough to achieve therapeutic concentrations in brain tissue at the tested doses.

The amyloid-beta clearance was particularly surprising since ISM-117 was not designed to target amyloid-beta. The current hypothesis is that the two pathologies are mutually reinforcing, and dismantling tau is sufficient to trigger microglial clearance of amyloid-beta. If confirmed in humans, this would represent a significant advantage over the current monoclonal antibody therapies that target amyloid-beta alone.

Clinical Development

Novartis has confirmed it licensed ISM-117 from Isomer AI and is preparing an IND application for a Phase I trial. The trial will enrol patients with early-stage Alzheimer’s. Osei is cautious but not without hope: “Mouse models have failed us before. What is different this time is the specificity of the target engagement and the mechanism. We are not hoping the molecule does something useful. We know exactly what it does and why.”

Second-Generation Compound Development

Even as the first-generation ISM-117 molecule moves toward Phase I trials, Osei’s group has already synthesised and begun testing a second-generation compound. The primary design objective for the second generation is improved blood-brain barrier penetration — ISM-117’s brain-to-plasma ratio of 0.34 is sufficient to achieve therapeutic concentrations at the doses tested, but a ratio closer to 1.0 would allow either smaller doses or lower dosing frequency, improving the tolerability profile for chronic administration.

The blood-brain barrier is a formidable pharmacological challenge. The tight junction proteins that line brain capillaries allow only small, lipophilic molecules to cross freely, while actively exporting many drug molecules via efflux pumps. ISM-117 is a mid-sized molecule with moderate lipophilicity — its blood-brain barrier penetration is satisfactory but improvable. The second-generation compound incorporates a bioisosteric replacement of one of ISM-117’s hydrogen bond donors — which impede passive transcellular diffusion — with a methyl ester group that is cleaved by brain esterases after crossing the barrier, releasing the active compound in situ.

This “pro-drug” strategy is well-established in CNS drug design. In preliminary rat pharmacokinetic studies, the second-generation compound showed a brain-to-plasma ratio of 0.81 — a 2.4-fold improvement over ISM-117 — with comparable in vitro tau disaggregation activity. Full animal efficacy testing is scheduled to begin in the third quarter of this year.

The OpenTau Consortium

The structural data that the AI pipeline generated — the binding site coordinates, the pharmacophore model, the conformational change mechanism — has been deposited in the Protein Data Bank and is publicly available. Osei and her colleagues at Isomer AI have used this availability as the foundation for an open research consortium — OpenTau — that invites academic groups globally to use the binding site data to develop their own tau-targeting molecules, with the condition that any compounds discovered through the programme are licensed back to a common pool for access in low- and middle-income countries.

Within six months of the consortium’s formation, 34 academic groups from 18 countries have joined. Three have already reported compounds with tau disaggregation activity in the OpenTau compound repository — none yet approaching ISM-117’s potency, but the diversity of chemical scaffolds represented is large, and the dataset is growing rapidly. The consortium’s approach inverts the usual dynamic of pharmaceutical development, where the identification of a compelling target triggers a proprietary race by multiple companies — here, the open data is being used to create a collaborative pipeline rather than a competitive one.

“Alzheimer’s is too important and too hard for any single group to solve alone,” Osei said. “If the binding site we identified is as valuable as we believe, the best thing we can do for the field — and for patients — is to put it in the public domain and let everyone work on it simultaneously. Competition can follow once there are candidates in the clinic and the commercial landscape is clearer.”

Competition and the Race for a First-in-Class Drug

ISM-117 is not the only AI-designed Alzheimer’s compound in development. The last three years have seen a wave of machine-learning-guided Alzheimer’s drug discovery campaigns, several of which are now in early clinical stages. Insilico Medicine filed an IND for an AI-designed GSK-3β inhibitor in 2023. Exscientia has a portfolio of neuroinflammation targets identified by its AI drug design platform. BioAge Labs is using machine learning to identify aging biology targets relevant to Alzheimer’s. The competitive landscape is unusually active, and the scientific community is watching to see whether the AI-design approach produces a different success rate in clinical development than conventional drug discovery — which achieves first-in-human Phase I success for central nervous system drugs approximately 12 to 15 percent of the time. The theoretical case for AI design producing better candidates is that physics-based binding affinity prediction should select for compounds with more specific, tighter target engagement — reducing off-target effects that contribute to attrition in human trials. Whether this theoretical advantage translates to better Phase II efficacy is the question that the next five years of clinical data will begin to answer. Osei’s group plans to publish a comprehensive comparison of ISM-117’s preclinical data package against the published preclinical data for all marketed Alzheimer’s therapies, to establish whether its profile differs in the ways that AI-guided design predicts it should.

The ISM-117 story has also attracted attention from the scientific community as a case study in the commercialisation of AI-derived scientific value. Isomer AI’s founding thesis — that AlphaFold’s protein structure prediction capability could be converted into a drug discovery platform by coupling it with a generative molecular design engine — was considered ambitious when the company was founded in 2021 with $3 million in seed funding. The ISM-117 licensing deal with Novartis, whose terms were not publicly disclosed but which venture capital analysts estimate at between $50 and $150 million in upfront and milestone payments, has validated the thesis dramatically and has triggered a wave of similar AI drug discovery company formations and funding rounds. Whether this enthusiasm will be vindicated by clinical success rates — the ultimate test of any drug discovery platform — remains to be seen. But the speed of the Isomer AI pipeline, which generated a clinical candidate in three years from company founding, compares favourably with the typical eight to twelve years required by conventional discovery programmes, and the quality of the mechanistic rationale for ISM-117 is widely acknowledged to be exceptional.

The numbers, when you first encounter them, seem too large to be real. One kilogram of a re-engineered enzyme, working continuously, could in principle process a metric tonne of PET plastic into its chemical building blocks within a single hour. Those building blocks — terephthalic acid and ethylene glycol — could be fed back into manufacturing to produce virgin-quality plastic, closing what is currently one of the most wasteful material loops in industrial history.

This is not a theoretical calculation. It is the performance benchmark that a team led by Dr. Sarah Nightingale at the University of Portsmouth has demonstrated in a bench-scale reactor, using an engineered variant of an enzyme called LCC-ICCG.

For context: the best commercial plastic recycling enzyme before this work had a turnover number of approximately 4 reactions per second. LCC-NV12, Nightingale’s variant, achieves 44,000 per second. The improvement is not incremental — it changes the economic calculus of enzymatic plastic recycling from interesting-but-marginal to potentially competitive with mechanical recycling.

The Starting Point: LCC and the Toulouse Milestone

Leaf-branch compost cutinase (LCC) was first isolated from compost samples in 2012. A landmark 2022 paper from the Toulouse Biotechnology Institute described a variant — LCC-ICCG, incorporating four amino acid substitutions — that could depolymerise 90 percent of a kilogram of PET into its constituent monomers in ten hours at 72°C. At the time, this was the most efficient biological PET deconstruction system ever reported and the benchmark against which all subsequent enzyme engineering has been measured.

Nightingale’s group aimed to push past that benchmark using a combination of directed evolution and computational protein design. “LCC-ICCG was limited by two things,” she explained. “The rate at which the enzyme could bind to the plastic surface, and the rate at which it could release the product monomers after cleaving the polymer chain. Both were bottlenecks. We targeted both simultaneously.”

Engineering the Acceleration

The team generated a library of 380,000 enzyme variants by introducing random mutations across the active site and the surface loops that control substrate access. High-throughput screening using a microfluidic fluorescence assay identified twelve variants with improved performance. The best — LCC-NV12 — incorporated seven amino acid changes from LCC-ICCG.

Cryo-electron microscopy revealed the structural reasons for the acceleration. Three of the seven mutations had opened a wider substrate channel, allowing the enzyme to processively cleave multiple ester bonds before dissociating from the plastic surface. Two others created new electrostatic interactions that accelerated product release by destabilising the enzyme-product complex. The combined effect was the 44,000 per second turnover number — a 10,000-fold improvement over LCC-ICCG.

In a bench-scale reactor operating at 72°C, LCC-NV12 depolymerised 1 kilogram of commercially sourced PET bottles to greater than 97 percent monomer yield in 3.2 hours — compared with 10 hours for LCC-ICCG under identical conditions. The monomers were recovered at 99.8 percent chemical purity by a standard crystallisation step, meeting the specification for use as feedstock in virgin PET manufacturing.

From Lab to Industry

Industrial partner Carbios — the French company that has been developing enzymatic PET recycling since 2011 and operates a demonstration plant in Clermont-Ferrand capable of processing 2 tonnes per day — is developing production fermentation processes for LCC-NV12. Current enzyme production cost is approximately €340 per kilogram; at manufacturing scale, the target is below €80 per kilogram within three years.

At €80/kg enzyme cost, the economics of enzymatic recycling of mixed PET waste become competitive with mechanical recycling — and superior to it for heavily contaminated or colour-mixed streams that mechanical recycling cannot handle economically.

L’Oréal, Nestlé, PepsiCo, and Solvay have all committed to taking enzymatically recycled PET from the Carbios facility. For the ocean plastic problem specifically, Nightingale notes that LCC-NV12 is best suited to the upstream challenge. “Our enzyme is best at intercepting plastic waste before it reaches the ocean. Remediation is always harder than prevention.”

What Happens to the Monomers

The terephthalic acid and ethylene glycol produced by LCC-NV12 are not waste — they are the same building blocks from which virgin PET is manufactured. Once purified, they can be fed directly into the polymerisation reactor at a standard PET manufacturing facility and converted back into PET indistinguishable from petroleum-derived virgin resin. This chemical equivalence is crucial for commercial adoption: brands that have committed to using recycled PET in their packaging need the recycled material to meet the same performance standards as virgin PET, including clarity, molecular weight distribution, and the absence of colouration or contamination.

Independent testing of PET produced from Carbios-processed enzymatic monomers by L’Oréal, which has published its results, shows performance metrics within 2 to 3 percent of petroleum-derived virgin PET across all relevant parameters — within the range that quality control systems at cosmetics packaging lines can accommodate without process adjustment. This “food-grade circular” PET — recycled to a quality equivalent to virgin — is the commercial proposition that distinguishes enzymatic recycling from mechanical recycling, which produces progressively lower-quality plastic with each recycling cycle.

PET vs. Mixed Plastics: Scope Clarity

A common misconception about enzymatic PET recycling is that it can address the mixed plastic problem — the broad category of unrecyclable, mixed-composition plastic waste. LCC-NV12, like all PET-ases, is specific to PET. It does not significantly degrade polyethylene, polypropylene, polystyrene, or the multi-layer plastic structures used in flexible food packaging, which are the most difficult-to-recycle categories by volume.

The global PET market is approximately 70 million tonnes per year — a large but finite fraction of total plastic production. Enzymatic recycling, even if it achieves 100 percent coverage of PET waste, does not solve the broader plastic problem. What it does is offer a solution for the specific category — bottles, polyester textiles, food trays — that currently dominates recycling discussion and for which the circular economy value proposition is most compelling. Nightingale is direct about this scoping: “We are solving the PET problem. The HDPE problem, the multi-layer film problem — those need different solutions. We cannot claim our enzyme does everything.”

Textile Applications: The Hidden PET Mountain

While PET bottles receive most of the public attention, polyester textiles — essentially PET in fibre form — are the largest and fastest-growing component of the global PET market. Approximately 50 million tonnes of polyester fibre are produced each year, and the vast majority is never recycled. The mechanical recycling infrastructure for polyester textiles is limited by contamination (dyes, finishes, blended fibres) and by the economics of collecting and sorting used clothing at scale.

Enzymatic depolymerisation is, in principle, agnostic to the form of the PET input — bottles and polyester fibres are chemically identical, and LCC-NV12 attacks both. The greater challenge with textiles is pre-processing: removing dyes and finishes that can inhibit the enzyme, and mechanically processing clothing items into a form suitable for reactor loading. Nightingale’s group has secured separate funding from the UK’s EPSRC to study the pre-treatment requirements for polyester textiles, with a view to extending the technology to clothing waste within the next product development cycle.

The Colour Problem in PET Recycling

One of the practical limitations of mechanical PET recycling is colour: mechanically recycled PET from coloured bottles produces a grey or mixed-colour recyclate that cannot be used for clear packaging — the highest-value PET application. Food-grade clear PET from recycled sources commands a price premium of 30 to 50 percent over coloured recyclate, and the colour contamination problem is the primary reason that the recycled PET market has been unable to fully displace virgin PET in clear packaging applications. Enzymatic depolymerisation solves this problem completely: when coloured PET is broken down to its monomers, the colour molecules — typically azo dyes or anthraquinone dyes — are removed as impurities during the monomer purification step. The terephthalic acid and ethylene glycol produced from enzymatic depolymerisation of coloured PET are colourless and indistinguishable from monomers produced from clear PET or from petroleum. Re-polymerisation produces crystal-clear PET with no colour residue. This colour-neutral property is a significant commercial advantage over mechanical recycling in the premium packaging market, and it is one of the reasons that L’Oréal — whose cosmetics packaging requires the highest optical clarity standards — was among the first brand owners to commit to the Carbios enzymatic recycling stream.

Nightingale has been invited to join the Ellen MacArthur Foundation’s advisory panel on chemical recycling — a signal of how central enzymatic depolymerisation has become to the circular economy movement. The Foundation, which has been the most influential advocate for circular economy principles in corporate sustainability commitments, is updating its guidance on plastic recycling to recognise enzymatic routes as a first-tier technology alongside mechanical recycling rather than an experimental alternative. This recognition matters for corporate sustainability reporting: companies that use the Ellen MacArthur framework to report their recycled content claims will increasingly be able to count enzymatically recycled content without the caveats that currently apply to chemical recycling technologies with less established performance records. At scale, this reporting recognition creates a market pull for enzymatic recycling capacity that operates independently of regulatory requirements — brand owners motivated by sustainability commitments will seek out the technology even before mandates require it.

The story begins with a batch of failed kombucha. In summer 2022, a postdoctoral researcher in Prof. Lena Fischer’s biochemistry lab at the Freie Universität Berlin was maintaining a bacterial culture of Gluconacetobacter diazotrophicus — a known nitrogen-fixing organism associated with sugarcane roots — that had been accidentally deprived of sucrose for four days.

The nitrogenase activity in that stressed culture was anomalously high — about three times the expected level. And nitrogen fixation was occurring at 25°C with no ATP supplement added to the assay. “That should not be possible,” Fischer said. “Standard nitrogenase consumes 16 ATP molecules per nitrogen molecule fixed. Without supplemented ATP, fixation at any significant rate at 25°C simply does not happen.”

What the lab had stumbled upon was an enzyme variant — subsequently isolated, purified, and designated GdNase-C — that appears to have solved one of the fundamental energy problems in biological nitrogen fixation. If the finding survives the scrutiny of full mechanistic characterisation, it could have implications not just for agricultural chemistry but for industrial ammonia production that consumes 1 to 2 percent of global energy annually.

Why Normal Nitrogen Fixation Is So Energy-Hungry

The N≡N triple bond, with a dissociation energy of 945 kJ/mol, is among the strongest chemical bonds in nature. Breaking it — the first step in any nitrogen fixation process — requires either the brutal conditions of the Haber-Bosch process (150-300 bar, 400-500°C) or the 16 ATP molecules consumed by biological nitrogenase enzymes. Each ATP molecule represents significant biological energy investment; no known living system had found a cheaper way to activate molecular nitrogen.

GdNase-C achieves nitrogen fixation at a rate of 340 nanomoles of ammonia per milligram of protein per minute at 25°C and atmospheric pressure. The comparable figure for the model Azotobacter nitrogenase is 580 nanomoles per milligram per minute at 30°C — so GdNase-C is not faster. What is extraordinary is the ATP consumption: 5.8 ATP per N₂ fixed, compared with 16 for every other nitrogenase ever characterised. This represents a 63 percent reduction in the energy cost of biological nitrogen fixation.

The Structural Explanation

Cryo-EM analysis revealed the reason. The FeMo-cofactor — the iron-molybdenum cluster at the enzyme’s active site — had an unusual coordination geometry not seen in any previously characterised nitrogenase. A cysteine residue conserved in all known nitrogenases is substituted by a selenocysteine in GdNase-C. This substitution appears to create a direct electron transfer pathway from the P-cluster to the FeMo-cofactor that bypasses one of the two standard ATP-consuming electron transfer steps in the normal mechanism.

“Bypassing that step is almost certainly why the ATP requirement is so much lower,” said Dr. Mattias Borg. “Each ATP-consuming step in normal nitrogenase is needed to drive electron transfer against an unfavourable gradient. If you create a more direct path, you need fewer steps. The selenocysteine does that — it acts as a molecular shortcut.”

How G. diazotrophicus acquired a selenocysteine at that position — when every other nitrogen-fixing organism uses cysteine — is not yet understood. Genomic analysis shows several unusual insertion sequences near the GdNase-C gene, suggesting a horizontal gene transfer event at some point in its evolutionary history.

Implications and Verification

The finding has been independently verified at three laboratories: the Max Planck Institute for Terrestrial Microbiology in Marburg, the John Innes Centre in Norwich, and the Institute of Plant and Microbial Biology at ETH Zurich. All three confirmed the anomalous ATP stoichiometry using ¹⁵N₂ isotope tracking and independent ATP quantification methods.

The European Research Council has fast-tracked a €12 million grant for the Fischer group to characterise the mechanism fully and attempt to transfer the selenocysteine substitution into standard Azotobacter nitrogenase. The Max Planck Institute for Coal Research has begun a parallel programme to use the enzyme’s structural features as a template for synthetic catalyst design — specifically, to understand whether the electronic structure of the selenocysteine electron transfer pathway can be replicated in a transition metal coordination complex.

“If this result stands up to full scrutiny,” Fischer said, “it means nature found a solution to cheap nitrogen fixation that we have been looking for industrially for over a century. The question is whether we can understand what evolution found well enough to replicate it.”

The Kombucha Origin: What It Tells Us

The detail that GdNase-C was found in a culture that had been inadvertently stressed by sucrose deprivation is more significant than it might initially appear. The stress condition — four days without the organism’s preferred carbon source — appears to have induced the expression of genetic loci that are normally silent in well-fed cultures. This is a familiar phenomenon in microbial biology: stress conditions activate secondary metabolic pathways and alternative physiological strategies that evolution has maintained in the genome but suppressed during normal growth.

The implication is that AMF-7’s unusual nitrogenase may not be unique to this strain, but rather a latent capability present in other G. diazotrophicus isolates that has simply not been expressed in standard culture conditions. Fischer’s group is now systematically screening 847 G. diazotrophicus strains from culture collections worldwide using the same sucrose-deprivation induction protocol, looking for other strains expressing similar anomalous nitrogenase activity. Preliminary results from 120 strains screened so far have found two others with elevated ATP efficiency — at 6.8 and 7.2 ATP per N₂, respectively, less efficient than GdNase-C’s 5.8 but still far better than the standard 16.

The Synthetic Biology Path Forward

The most direct route to exploiting GdNase-C’s low-energy nitrogen fixation in practical applications is genetic transfer — moving the modified nitrogenase gene into a nitrogen-fixing organism already used in agricultural settings. The obvious candidates are Rhizobium species that form symbiotic root nodules with leguminous crops, and free-living nitrogen fixers like Azotobacter vinelandii that can be inoculated into non-legume crop soils as biofertilisers.

Introducing a nitrogenase gene that uses 64 percent less ATP per N₂ fixed could, in principle, allow the host organism to fix more nitrogen per unit of carbon invested in the process — meaning that the same amount of organic matter metabolised by root-nodule bacteria could produce substantially more ammonia for the plant. This would increase the nitrogen nutrition delivered to crops by biologically fixed sources, potentially reducing dependence on synthetic fertiliser.

Fischer’s group has initiated collaborative work with a Wageningen University team specialising in Rhizobium genetics. The first challenge is gene expression — the selenocysteine codon UGA is normally read as a stop codon in most organisms, and expressing selenocysteine-containing proteins requires special machinery (the selenocysteine insertion system, or SECIS) that is not present in all organisms. The team has confirmed that Rhizobium leguminosarum contains the SECIS elements needed for selenocysteine incorporation, making it a tractable host for the transfer. The first transformants will be assessed for nitrogenase activity in early 2025.

A Cautionary Note on Industrial Displacement

The Haber-Bosch process employs approximately 150,000 people globally, supports the economy of major natural gas-producing countries that export ammonia as a commodity, and is deeply integrated into the financial structures of the global agricultural input industry. A biological nitrogen fixation system that matches or approaches Haber-Bosch efficiency would not automatically and immediately replace the industrial process — adoption in agriculture moves on the timescale of regulatory approval, farmer behaviour change, and seed company development programmes, all measured in decades.

Fischer is careful to frame the discovery as creating a possibility, not an immediate disruption. “We have found that the energy cost barrier to biological nitrogen fixation is lower than we thought,” she said. “That opens a research and development pathway that was not previously available. Whether and when that pathway leads to practical applications at agricultural scale depends on many factors beyond our control — factors that include policy, economics, and the time required for careful, rigorous development.”

Policy Context: Fertiliser Security and Food Systems

The Fischer group’s discovery arrives at a moment when the strategic importance of nitrogen fertiliser has been sharply reinforced by geopolitical events. Russia’s invasion of Ukraine in 2022 disrupted global natural gas markets — the primary feedstock for Haber-Bosch ammonia production — causing ammonia prices to spike by more than 300 percent between 2021 and 2022 peak. This price spike caused fertiliser shortages that reduced crop yields in multiple developing countries and contributed to food price inflation that affected hundreds of millions of people. The episode demonstrated that the global food system’s dependence on centralised Haber-Bosch production, concentrated in a small number of countries with access to cheap natural gas, creates systemic vulnerability that affects food security globally. A decentralised biological nitrogen fixation capability — whether through engineered organisms in root nodules, free-living soil bacteria with improved nitrogenase efficiency, or electrochemical systems like CSIRO’s — would distribute this critical agricultural capability more broadly and reduce the geopolitical leverage that natural gas-exporting countries currently exercise over global food production. This food security framing has elevated the Fischer group’s discovery from a biochemistry curiosity to a matter of strategic national interest in multiple countries, and has contributed to the rapid funding response from European research agencies.

Concrete is the most widely used construction material in human history — approximately 30 billion tonnes are produced each year, more than any other manufactured substance on Earth. And every year, a significant fraction of it cracks. Hairline fractures allow water to penetrate, corroding reinforcing steel and triggering deterioration that eventually requires expensive repair or replacement. The global cost of concrete infrastructure maintenance is estimated at $2.5 trillion annually.

Dr. Chelsea Rivera’s group at the University of Colorado Boulder spent six years working on a solution that draws not from synthetic chemistry but from the oldest construction material technology on Earth: organisms that make their own limestone. Their living concrete — embedding photosynthetic cyanobacteria directly into the concrete mix — has now passed the Colorado Department of Transportation’s standard evaluation protocol and is being laid in a 200-metre test section of a state highway near Fort Collins.

“What we have made is concrete that heals itself by growing new stone,” Rivera said. “The bacteria are not passengers in the concrete. They are doing structural work. And in three out of four test specimens, the healed material is stronger than it was before it cracked.”

The Biology Behind the Healing

The key organism is Synechococcus elongatus, a photosynthetic cyanobacterium chosen for three specific properties: its high calcium carbonate precipitation rate, its tolerance for the extreme alkalinity of concrete pore water (pH above 12), and its ability to survive encapsulated in an inert state for years before activation.

The precipitation mechanism is well understood. Cyanobacteria consume CO₂ during photosynthesis, raising local pH. At high pH, dissolved calcium ions precipitate as calcite or aragonite crystals. In living concrete, these crystals fill crack voids and bond to the existing cement matrix, restoring structural continuity — the same process that has built stromatolites and coral reefs for hundreds of millions of years, now put to engineering work.

Keeping Bacteria Alive in Concrete

The challenge nobody had previously solved is keeping cyanobacteria alive inside hardened concrete. The interior of a concrete structure is dark, highly alkaline, and has very low water activity — inhospitable conditions for a photosynthetic organism adapted to open water. Earlier bio-concrete approaches used spore-forming bacteria (Bacillus species) that could survive dormancy for years, but whose calcium carbonate precipitation rate is too slow for rapid structural healing.

Rivera’s solution was microencapsulation. The cyanobacteria are embedded in the concrete mix inside protective hydrogel microspheres — approximately 200 micrometres in diameter — made from sodium alginate cross-linked with a silica-organic hybrid material. The microspheres maintain internal humidity, buffer pH to the organism’s tolerance range, and rupture when crack propagation reaches them, releasing the bacteria directly into the crack void where water infiltration, and — in surface applications — light exposure, reactivates them within 12 to 18 hours.

The Strength Data

The headline result is the compressive strength recovery figures. Standard concrete that cracks and self-heals recovers at most 20 to 30 percent of its original strength through slow re-crystallisation. Rivera’s living concrete, in DOT-standard accelerated weathering tests, recovered 94 to 112 percent of original compressive strength across 48 test specimens. Three out of four specimens exceeded their original strength.

Electron microscopy shows why. The cyanobacteria precipitate aragonite crystals — a needle-like polymorph of calcium carbonate — growing perpendicular to the crack faces and interlocking with both the cement hydrate matrix and each other. The interlocking geometry creates a crack repair zone whose fracture toughness, measured by notched beam bending tests, exceeds that of the original uncracked concrete.

Highway Testing and Commercial Path

The Colorado DOT test section, placed in January 2024, has survived one full Colorado winter with no cracking events, while two adjacent standard concrete control panels have developed hairline surface cracks from frost action. The five-year monitoring programme required before wider DOT deployment will continue, but the first-year results have been described by DOT materials engineers as “exceeding expectations.”

Rivera’s spinout company, BioStruct Materials, has signed a licensing agreement with LafargeHolcim — the world’s largest cement producer — to develop a commercial-scale production process. The target is a living concrete mix at no more than 15 percent above standard concrete price, a threshold LafargeHolcim’s modelling suggests is achievable at production scales above 50,000 tonnes per year. The organisms themselves, when the concrete is eventually demolished, present no environmental concern: S. elongatus is a naturally occurring organism found globally in soil and freshwater.

Long-Term Structural Performance Modelling

One of the most important unanswered questions about living concrete is its long-term structural performance over the 50 to 100-year service life expected of infrastructure concrete. Laboratory testing and the first year of DOT highway data establish that the material works at short timescales, but the cumulative effect of decades of crack-heal-crack cycles on the concrete microstructure is not yet characterised. Rivera’s group has been working with structural engineering colleagues to develop physics-based models of the long-term performance based on the microstructural data from healed specimens.

The preliminary modelling results suggest that the long-term behaviour is likely to be better than conventional concrete rather than worse. Each healing cycle deposits new aragonite crystal material that is chemically bonded to the cement hydrate matrix and fills not only the crack void but also some of the original porosity of the surrounding concrete. Over many cycles, this progressive densification of the microstructure should increase both compressive strength and durability to water and chloride ingress — the primary mechanisms of long-term concrete deterioration in exposed infrastructure.

“The conventional model of concrete performance is that it slowly gets weaker over time from chemical attack and mechanical fatigue,” Rivera said. “Our preliminary modelling suggests living concrete may actually get stronger over time, as the bacteria progressively fill the microporosity that is the primary pathway for chemical ingress. We will not have experimental data to test that prediction for years, but it is a hypothesis we are designing the DOT test programme to eventually evaluate.”

Indoor Applications: The Light Problem and Its Solution

The greatest limitation of the CU Boulder living concrete for structural applications is light dependency. Synechococcus elongatus is a photosynthetic organism — it requires light to activate its calcium carbonate precipitation capacity, and in the interior of a building’s structural concrete, there is no light. The bacteria survive in dormancy in the microspheres, but they cannot heal internal cracks that are never exposed to light.

Rivera has a parallel research programme addressing this limitation using a non-photosynthetic calcium carbonate precipitating bacterium — Sporosarcina pasteurii — that can be activated by the pH change that occurs when water enters a crack, without requiring light. The trade-off is lower calcium carbonate precipitation rates: S. pasteurii operates at roughly 20 percent of the rate of S. elongatus under equivalent conditions. The team’s current data shows that this lower rate is still sufficient to heal cracks up to 0.25 millimetres in width over 60 days — adequate for interior structural concrete that is not exposed to the chloride and freeze-thaw conditions that make rapid healing critical in exposed applications.

Environmental Fate and Sustainability Credentials

The full lifecycle environmental assessment of living concrete, commissioned from Colorado State University’s engineering sustainability programme, found a net CO₂ benefit of approximately 120 kilograms of CO₂ per tonne of concrete over a 50-year service life — combining the reduction in maintenance interventions, the extended service life of the infrastructure, and the CO₂ captured in the aragonite crystals precipitated by bacterial healing activity. This figure excludes the manufacturing carbon footprint, which is comparable to conventional concrete, and focuses on the operational and end-of-life differences.

The embedded bacteria represent a negligible addition to the material’s mass — approximately 0.002 percent by weight — and consume only the organic carbon in the microsphere hydrogel and the dissolved CO₂ and bicarbonate already present in the concrete pore water. The organisms introduce no novel genetic material into the environment when the concrete is eventually demolished and crushed, since the cyanobacterial strain is identical to naturally occurring environmental isolates. Regulatory agencies in Colorado, California, and New York have all indicated that the material falls within existing concrete material specifications and does not require additional environmental permitting beyond standard concrete installation approvals.

Carbon Accounting and the Life Cycle Perspective

One detail of the living concrete’s environmental credentials that deserves careful examination is the question of whether the CO₂ sequestered by the cyanobacteria is additional — meaning CO₂ that would not have been sequestered otherwise — or merely displaced from where it would have naturally occurred. S. elongatus, being a photosynthetic organism, would under natural conditions fix CO₂ from the atmosphere and eventually release it back as the organism decomposes. The CO₂ it sequesters inside concrete is different: it is converted to calcium carbonate crystal, which is chemically stable for thousands of years under the alkaline conditions of the concrete matrix. The carbon fixed in these crystals is effectively sequestered on a geological timescale — not released when the organism dies, because the organism’s carbon ends up in the long-lived calcium carbonate structure rather than in its own biomass. This permanence argument is important for carbon accounting purposes: if the CU Boulder concrete is used in infrastructure with a 50-year design life, the CO₂ captured in aragonite crystals over that period is sequestered for the concrete’s entire service life and, if the demolished concrete is used as aggregate or fill rather than incinerated, potentially indefinitely. Rivera has engaged the US Environmental Protection Agency’s carbon accounting methodologies group to develop a protocol for crediting this permanent sequestration in national greenhouse gas inventory accounting — a protocol that, if adopted, would make living concrete one of the few construction materials eligible for carbon removal credits rather than merely carbon reduction credits.