Quantum Batteries

Speculative If the collective charging speed-up measured in organic microcavities survives when the quantity being measured is extractable work rather than absorbed photon energy — which no published experiment has tested, and 7 says exactly what would — the prize is an on-chip energy store that fills in roughly a tenth of a picosecond and fills faster, per cell, the more cells it has. That last clause is the whole claim. It is a statement about rate, not about capacity, and it does not compete with a battery.

Established The measured numbers are small and the direction of the small is informative. The first Dicke-model quantum battery charged in 0.094 to 0.120 picoseconds and stored between 0.011 and 0.184 electron-volts per molecule; a later microcavity of the same lineage reports a stored-energy density of about 2.63 × 10¹⁸ eV per cubic centimetre, which is 0.42 joules per cubic centimetre, or 0.12 watt-hours per litre. A commercial lithium-ion cell passed 700 watt-hours per litre before 2018. The quantum device is roughly six thousand times less dense than the thing it is named after, so the upside cannot be storage and any brief that says otherwise is selling something.

Frontier The gain, if it is real, accrues to the inside of a quantum computer rather than to a grid or a car. A 2026 Physical Review X proposal puts a quantum battery on the cryogenic side of the wiring and powers gates from it, claiming the removal of per-qubit drive lines could quadruple the number of qubits that fit in a dilution refrigerator. That is a specific, checkable number about a specific bottleneck, and it is a different business from energy storage: the customer is a machine that already lives at millikelvin and already counts every watt crossing its wiring.

Speculative The condition is extraction, and it has a number. The best measured discharge from one of these devices is 10 to 40 microwatts per square centimetre, so delivering a single watt would take between 2.5 and 10 square metres of device — this brief’s arithmetic on the published power density, not a figure any group has claimed. Until extractable work is measured on the same sample series as charging power, the upside stated here rests on a quantity nobody reports, and the bottleneck list in 5 is the honest reason why.

Established A quantum battery is a quantum system with a non-degenerate internal Hamiltonian, used to hold energy and give it back. The 2024 Colloquium in Reviews of Modern Physics defines it as a d-dimensional system whose internal Hamiltonian has non-zero bandwidth, charged by preparing it away from its ground state. Nothing about that definition requires the device to be small, cold, or entangled; it requires only that the accounting be done quantum-mechanically.

Established The quantity that matters is ergotropy, not energy. Ergotropy is the maximum work extractable from a state by cyclic unitary operations. The gap between the energy a system holds and the work it gives back is locked energy, and it is a property of the state rather than a loss term better engineering removes. A quantum battery can be full of energy and worth nothing, which is why a paper reporting absorbed energy has not yet reported a battery.

Established Superextensive charging means the per-cell charging power rises with the number of cells. Charge N identical cells in parallel, each with its own charger, and total power scales as N: extensive, and what every classical battery pack does. Charge them collectively and total power can scale faster than N. The Colloquium gives the Dicke battery a mean power going as N to the three-halves, an advantage of the square root of N over parallel charging. The organic-microcavity claim is the same statement in intensive form: energy and power per molecule rise as molecules are added.

Established The physical mechanism on the experimental side is superabsorption, the time-reverse of superradiance. N emitters sharing one optical mode couple to it with a strength that goes as the square root of N, so the collective bright state absorbs faster than N independent molecules would.

Established This is a Dicke cooperative effect, known on the emission side since 1954. It is a statement about light-matter coupling before it is a statement about thermodynamics, and nothing about it is in dispute.

Frontier Whether that cooperative enhancement is a quantum advantage is a separate question with a separate answer, and the answer currently on record is no. The Colloquium states that for the Dicke battery the relevant advantage ratio “was found not to scale with N”, and concludes that “no quantum advantage R scaling with N can be found in the charging dynamics of the Dicke battery”. The collective speed-up is real and measured; attributing it to entanglement is a separate claim that its own field’s review declines to make.

Established There is a hard ceiling on how much any quantum battery can win. Gyhm, Šafránek and Rosa proved in Physical Review Letters in 2022 that the maximal charging speed-up is extensive — at most quadratic scaling of power against the classical linear case — and that reaching it requires global operations touching every cell at once. Local k-body interactions buy an advantage bounded by k, not by N. So the headline is not open-ended: the theory says the prize is a factor, the factor is bounded, and the bound tightens the more realistic the control model becomes.

Frontier Three timescales have to separate before any of this is a device. The Colloquium requires relaxation time much greater than charging and extraction times. In the microcavities charging is picoseconds and relaxation was, at first, nanoseconds; the whole experimental programme since 2022 has been an attempt to raise the third number without raising the first with it.

Established The anchor result is real and the commission’s description of it is right in outline and wrong in two details. Quach and colleagues, working with Lumogen-F Orange dye dispersed in polystyrene inside a Fabry-Perot microcavity with an active layer near 185 nanometres, measured charging of the collective excited state by transient absorption and reported superextensive behaviour. The work appeared as a preprint in December 2020 and in Science Advances in 2022. That is the first Dicke-quantum-battery signature, as commissioned.

Established The stored energies are not femtojoule-scale; they are nanojoule-scale in the 2022 device and tens of microjoules in the 2024 one. The 2022 paper’s table gives a peak stored energy per molecule between 0.011 and 0.184 eV over molecule counts from 0.21 × 10¹⁰ to 16.0 × 10¹⁰; the largest per-molecule figure times the largest molecule count is about 3 × 10¹⁰ eV, roughly 5 nanojoules for the whole cavity — this brief’s arithmetic, not the paper’s. The 2024 triplet device states its own capacity as 1.23 × 10¹⁴ eV, or 2.0 × 10⁻⁵ joules. Per molecule the figure is far smaller still: 0.184 eV is 3 × 10⁻²⁰ J, ten thousand times below a femtojoule.

Frontier The 2022 scaling evidence is a set of pairwise exponents, not one power law fitted across a decade of N, and the exponents disagree with each other. The supplementary table reports, for three pairs of experiments, charging-power exponents of 0.94, 1.20 and 2.19 with matching stored-energy exponents of 0.52, 1.18 and 2.30. An exponent of 0.94 is not superextensive; 2.19 is strongly so.

Established The paper is honest about the spread; the coverage of it was not. It separates regimes where the cavity coupling dominates from regimes where it does not. But a reader who takes “superextensive charging was measured” as one clean scaling law across the whole sample set has been given the wrong picture, and the spread is the most important number in the paper.

Established Charging time fell as molecule count rose, in the regime where it should. The rise time, defined as the time to reach half the peak stored energy, ran from 0.094 to 0.120 picoseconds, and the paper reports the inverse-square-root dependence on N expected from collective coupling in the coupling-dominated case. Subextensive charging time is the cleanest part of the 2022 result and it is the part that least depends on interpretation.

Established Decoherence stabilised the stored energy rather than destroying it, which was the surprise of the 2022 paper. The authors report that the same environmental coupling that would be expected to bleed the state also suppressed the fast collective re-emission, and they name the practical obstacle plainly: “the key challenge for practical applications of this effect is its integration in devices where energy can be efficiently extracted and used”. They also note that concentration quenching limits performance at high dye loading, which is a direct constraint on making N large.

Established The storage problem was attacked in 2024 by moving the energy off the bright singlet and onto a dark triplet, and it worked by three orders of magnitude. Tibben and colleagues built a two-layer cavity: a Rhodamine 6G donor for fast superabsorbing charge and a palladium tetraphenylporphyrin acceptor whose triplet states are spin-forbidden from radiating.

Established Measured phosphorescent lifetime of the acceptor was 40.0 ± 0.4 microseconds outside the cavity and 57 ± 2 microseconds inside it. Against nanosecond-scale storage in bright-singlet Dicke batteries that is a thousandfold extension, and it is the largest single improvement anyone in this field has reported.

Frontier The same paper gives the first device-level energy density in the field, and its own two statements of capacity differ by a factor of a thousand. The preprint writes the total capacity as the triplet energy times the triplet count, “approximately 1.23 × 10¹⁴ eV”, and then renders the same quantity as 5.47 microwatt-hours. 1.23 × 10¹⁴ eV is 2.0 × 10⁻⁵ J, which is 5.47 nanowatt-hours. The electron-volt figure is the one consistent with the paper’s own quoted energy density of 2.63 × 10¹⁸ eV per cubic centimetre over a film of roughly a square centimetre and a few hundred nanometres, so this brief reads the watt-hour figure as a unit slip and uses the electron-volt value throughout. Anyone quoting 5.47 microwatt-hours is quoting a number a thousand times too large.

Frontier The triplet route costs energy density and the trade is explicit. At resonance only about half the population reached the triplet state, so measured capacity is roughly half the theoretical maximum for that film. Storage time went up a thousandfold and the extractable fraction went down by half, and no paper yet reports the product of the two as one figure of merit.

Frontier The 2026 result closed the cycle: charge, store, and deliver electrical power out of the device. Hymas and colleagues, publishing in Light: Science & Applications, built eight Fabry-Perot devices with copper phthalocyanine absorbers and fullerene charge-transport layers, spanning molecule counts from about 2.8 × 10¹⁴ to 7.9 × 10¹⁴.

Frontier They report superextensive charging power and energy density with subextensive charging time, excited-state populations persisting for tens of nanoseconds, and peak discharging power densities of about 10 to 40 microwatts per square centimetre. The ratio of collective to non-collective discharge power rises linearly in N, which is the first evidence that the collective effect survives all the way to the terminals rather than only to the excited state.

Frontier That 2026 series spans a factor of 2.8 in N, which is not a decade, and this is the central evidential weakness of the field. A factor of 2.8 in the independent variable cannot distinguish N to the first power from N to the three-halves with any confidence once device-to-device fabrication scatter is admitted, and eight devices is a small sample for a scaling claim. The 2022 work spans a factor of about 76 in N and does not report ergotropy; the 2026 work reports extractable electrical power and does not span the range. Nobody has done both at once.

Established A completely different platform — nuclear spins in a room-temperature NMR magnet — has measured the square-root-of-N advantage across a proper range of N, and reports storage times seven orders of magnitude longer than any microcavity. Joshi and Mahesh charged a single battery spin from star-topology charger ensembles with N = 3, 9, 12, 18 and 36, in five molecules including tetraphenylmethane and tetraethylammonium tetramethylborate, on a 500 MHz spectrometer at 298 K.

Established They observed “a significant quantum advantage ranging from about 1.5 to over 6” plotted against the theoretical square-root-of-N curve, estimated a battery storage time constant of about 200 seconds, and demonstrated charge held and then consumed by a load spin after up to two minutes, with better than 85% asymptotic charging efficiency. That is a complete charger-battery-load circuit, which no optical device had until 2026.

Frontier The NMR result is the strongest scaling evidence in the subject and the weakest device. The advantage tracks the predicted square root of N across a twelvefold span, which no optical experiment has matched, and it holds energy for minutes rather than microseconds. But the battery is one nuclear spin: the ensemble is on the charger side, the stored energy is a polarisation difference of no engineering consequence, and the apparatus is a superconducting magnet the size of a small car. It settles physics and offers nothing to build.

Established Quantum battery capacity has been verified experimentally on a photonic platform, at high precision, on two qubits. Yang and colleagues measured capacity — ergotropy plus antiergotropy — on two-photon states from a periodically-poled KTP Sagnac source at 810 nm, against entropy, coherence and entanglement, with tomographic fidelities above 0.98 and absolute errors at or below 0.02. It confirms the theoretical figure of merit is measurable. It involves two photons and no scaling.

Established The theory side has hardened considerably and it has hardened against the optimistic reading. The 2022 Physical Review Letters bound caps the charging speed-up at quadratic and makes global operations the price of admission.

Frontier A 2024 analysis derives the entanglement cost directly: the entanglement depth required of the charging unitary is at least the ceiling of N times the square of the normalised charging rate, so maximum rate demands full N-party entanglement while parallel charging at rate one-over-root-N demands none at all. In the idealised model the advantage and the entanglement are the same quantity seen twice.

Frontier A 2025 analysis argues that the power bounds everyone quotes can be actively misleading. Working an all-to-all coupled spin chain, the authors obtain scaling near N to the 1.9 under the uncertainty-principle bound and then watch the advantage disappear under the tighter quantum-Fisher-information bound, because large energy variance can come from state mixedness rather than from entanglement useful for charging. Their summary is blunt: “While a superlinear power bound signals a possible quantum advantage in principle, it may not be tangible or even exist in practice.”

Frontier Superextensive scaling has also been shown to be a finite-size effect in at least one solvable model. An extended cluster-Ising chain, of a class believed to forbid the effect, exhibits superextensive charging power over “wide ranges of system sizes, reaching up to a thousand spins” — but the authors state that because it rides on superextensive growth of the stored energy it “cannot persist in the thermodynamic limit”. A scaling advantage that is real at 1,000 cells and gone at infinity is still useful if devices are small; it is not a law.

Frontier Control theory is closing part of the gap between idealised Dicke models and disordered solids. Dong, Li and Xiao used reinforcement learning to find charging ramps for a disordered Dicke battery and report ensemble-averaged normalised ergotropy above 0.97 with fluctuations below 0.02, and charging times going as N to the minus one-half, for sizes up to about N = 30. The learned protocols are simple few-segment sweeps resembling multilevel Landau-Zener driving, so they are implementable rather than merely optimal. This is simulation, not a device.

Established The field’s own review states that a fully operational proof of principle does not exist. The 2024 Colloquium surveys superconducting circuits and quantum dots as proposed but not demonstrated, organic microcavities and nuclear spins as realised in preliminary form, and says that “important but preliminary steps towards the experimental implementation of quantum batteries have also been made” while “a fully-operational proof of principle is yet to be demonstrated”. It names the ultimate limits on energy density, absolute power and storage lifetime as unexplored. Its five authors include the author of the anchor experiment, and it is more conservative than the press written about that experiment.

Frontier The single open question is whether the measured charging advantage survives translation into extractable work. Every optical experiment to date measures absorbed or stored energy through a transient optical signal. Ergotropy — the part that comes back as work — has been measured on two photons and computed in simulations, never measured on a scaling series. The gap is not pedantic: locked energy depends on the purity and structure of the state, and the states produced by fast collective charging are exactly the states most likely to carry correlations that lock energy away.

Frontier A related and sharper question: does charging faster make the stored energy less extractable, by construction? A 2026 Physical Review Letters result finds that the quantum correlations generated during charging typically reduce extractable work, and that continuously monitoring the environment can weaken those correlations and raise work extraction “beyond what is achievable in the ideal (closed system) limit”, for both cavity-mediated spin-spin and Dicke models. If correct, the field has been optimising a figure of merit that trades against the one that matters, and the fix requires continuous measurement hardware nobody has built into a microcavity.

Frontier What is the quantum resource, if it is not entanglement? The entanglement-depth result says maximum charging rate requires full N-party entanglement; the Colloquium says the Dicke battery’s advantage does not scale as a quantum advantage; the deceptive-bounds analysis says apparent advantage can come from mixedness. A 2025 study adds nonstabilizerness, or magic, to the list and finds the relationship conditional: a one-to-one correspondence between ergotropy and total magic emerges for charging Hamiltonians with U(1) symmetry and is lost for generic ones, ergotropy can be stored at zero magic cost through Clifford unitaries, and high initial magic is a bottleneck for maximum average power rather than a help.

Frontier The resource accounting for this field is not settled, which is a stronger statement than it sounds. A technology whose advocates cannot agree on which quantum property is being spent cannot say what would make a device better, which is why 7 asks for a measurement rather than a model.

Frontier Does the scaling survive to any size that matters, and is the thermodynamic limit the right question? The cluster-Ising result shows superextensive power at up to a thousand spins that provably vanishes at infinity. Organic microcavities operate at N near 10¹⁴, eleven orders of magnitude above that, which sounds decisive until one notices that the mechanism there is collective optical coupling rather than many-body interaction and the two are not the same physics. Nobody has written down what the crossover looks like in a real solid.

Speculative Whether room-temperature metastable states can hold charge for milliseconds without making it unextractable is the engineering frontier. A 2026 review chapter surveys dark triplets, singlet exciton fission into triplet pairs, and charge-separated states, noting charge-separated lifetimes from milliseconds to seconds in engineered donor-acceptor architectures and one report of triplets exceeding ten hours in carbon-germanium compounds at room temperature. Each route decouples the storage manifold from the absorbing one, and each therefore has to answer how the energy gets back out.

Speculative Can self-discharge be engineered down without new chemistry? Two 2025 theory papers say yes by different routes: one finds coherent ergotropy decays more slowly than incoherent ergotropy in a driven nitrogen-vacancy system, holding usable charge about 8 microseconds against about 1 microsecond unmitigated; another combines the Purcell and quantum Zeno effects, claiming four orders of magnitude of dissipation suppression and a projected 1.5-second half-life at a 1 kHz cavity decay rate. Both are simulations with no experimental counterpart.

Speculative Is the right application inside a quantum computer rather than outside one? The 2026 Physical Review X proposal treats the battery as a coherent internal energy source for gates, keeping battery and qubits entangled through the operation so that the thermodynamic limit of zero dissipation for unitary gates becomes approachable, with multi-qubit gate times of a few multiples of pi over the qubit-resonator coupling and a universal gate set steered by one parameter per qubit. It is a proposal. But it is the first application argument in this field where the energy scale of the device and the energy scale of the job are the same.

Frontier How much of the apparent progress is the same three devices from one group? The 2022, 2024 and 2026 organic-microcavity results share authors and institutions. That is normal for a young field and is also the condition under which a systematic instrumental artefact would go uncaught. No independent group has published a superabsorption-scaling measurement with different dyes in a different laboratory: the field has no replication of its anchor.

Established Self-discharge is the binding constraint and the numbers state it exactly. Charging takes picoseconds; the first devices held charge for nanoseconds; the triplet device holds it for 57 microseconds.

Frontier By the review’s own test the triplet device passes against charging and has never been tested against extraction. The Colloquium requires relaxation time much greater than charging and extraction times; 57 microseconds beats picosecond charging by seven orders of magnitude, but extraction has not been timed on the same device, so the test is half run.

Frontier Extraction is the second constraint and it fights the first. Energy protected from the environment is by construction weakly coupled to everything, including the load: the triplet route buys storage by moving population into a state that cannot radiate, and then has to get energy out of a state that cannot radiate. The 2026 device solves this with conventional organic-photovoltaic engineering — a charge-transport layer and a fullerene acceptor — at 10 to 40 microwatts per square centimetre.

Speculative Area is the constraint nobody in the field states, because stating it ends the energy-storage conversation. At 10 to 40 microwatts per square centimetre, one watt requires 2.5 to 10 square metres of device and one kilowatt requires 2,500 to 10,000 square metres — this brief’s arithmetic on the published figure. That is also roughly the area over which a silicon photovoltaic would deliver a hundred to a thousand times more power in full sun, the comparison a reader makes unprompted and the field should make first.

Established Concentration quenching bounds N from above in the optical platform. The 2022 paper names it directly: quenching limits quantum-battery performance at high dye concentration. The superextensive claim wants N large; the chemistry punishes N large in the same film.

Frontier That chemistry is the likeliest reason the 2026 scaling range is so narrow. The devices varied cavity thickness and molecule count together across a factor of 2.8, and a quenching ceiling on usable concentration is exactly the constraint that would keep the range there.

Frontier The theory bottleneck is that the best available advantage needs global operations, and a solid-state film does not offer them. The quadratic ceiling on speed-up requires a charging unitary that acts on all cells at once. An optical cavity mode does couple to every molecule, which is why the microcavity is the platform of choice, but the coupling is a single collective mode rather than arbitrary global control, and the Colloquium is explicit that k-body interactions bound the advantage by k. The available global operation is the one that gives the Dicke advantage the review declines to call quantum.

Frontier Measuring N is itself a bottleneck. Every scaling exponent here is a fit of a measured optical quantity against an inferred molecule count derived from concentration, film thickness and mode volume, so a systematic error in N propagates straight into the exponent. No paper in this set reports an independent calibration of N, and a 2.8-fold range magnifies a 10% abscissa error far more than a 76-fold range does.

Frontier Ergotropy is not measured because measuring it is hard, not because it is unimportant. Extracting ergotropy means applying the optimal cyclic unitary, which for a mesoscopic exciton ensemble in a disordered film nobody knows how to do. The photonic verification worked because tomography on two photons is routine; there is no such route for 10¹⁴ coupled molecules, so the field’s central quantity has never been measured on a device.

Speculative Energy density is a bottleneck that is unlikely to move, because it is set by molecular excited-state energies and packing density. One exciton per molecule at roughly 1.8 eV, at realistic chromophore densities, lands near the 2.63 × 10¹⁸ eV per cubic centimetre already reported. There is no obvious factor of a thousand available in that arithmetic, which is what would be needed to reach even a tenth of lithium-ion volumetric density. Anyone proposing quantum batteries as grid or vehicle storage has to explain where that factor comes from, and nobody has.

Frontier Cryogenic platforms trade the density problem for a refrigeration problem. Superconducting-circuit quantum batteries remain proposals in the Colloquium’s survey, and a 2024 analysis of a two-qubit version derives ergotropy and instantaneous power analytically while reporting no measured energy, efficiency or decoherence figure. A device needing millikelvin is interesting only inside a machine that already runs at millikelvin, which is exactly the 2026 Physical Review X argument.

Frontier The field depends first on ultrafast metrology it already has, used for a quantity it has not yet chosen to measure. Transient absorption and differential reflectance at femtosecond resolution are mature; what is missing is a protocol converting those observables into an ergotropy bound rather than an energy estimate. That is an analysis dependency more than an instrument one, and cheaper to satisfy than anything else here.

Established It depends on organic photophysics for the storage manifold. Palladium tetraphenylporphyrin, Rhodamine 6G, copper phthalocyanine and C60 are standard organic-electronics materials, and the 2024 and 2026 devices are recognisably photovoltaic stacks with a cavity around them. The dependency runs the opposite way from the usual telling: quantum batteries need organic electronics, not the reverse.

Frontier It depends on singlet-fission and charge-separation chemistry to break the storage-versus-extraction trade. The 2026 review chapter frames all three storage routes as instances of one design principle — control the coupling between the absorbing manifold and the storage manifold — and every one of them is a materials-chemistry programme with its own decade-long literature that predates quantum batteries entirely.

Frontier It depends on quantum-control theory for anything beyond a passive cavity. The reinforcement-learned ramps reaching normalised ergotropy above 0.97 under disorder are ahead of the hardware: no organic microcavity has the fast tunable drive they assume, and closing that gap means borrowing electro-optic and polaritonic control techniques built for other reasons.

Frontier It depends on quantum thermodynamics settling what counts as a resource. Ergotropy, capacity, antiergotropy, coherent versus incoherent ergotropy, entanglement depth and nonstabilizerness are all live accounting variables in the 2024 to 2026 literature, and a 2026 PRX Energy paper adds ergotropy-preserving operations and isoergotropic states, which reshuffle the internal structure of extractable work while leaving its total unchanged. Until the field agrees which number is the product, two devices cannot be compared.

Speculative It depends, for the quantum-computing application, on fault-tolerant machines existing at a scale where wiring is the binding constraint. The claim that removing per-qubit drive lines could quadruple qubits per cryostat only matters if cryostat wiring is what is stopping the machine, which is a contested reading of the roadmap; the state of that argument is in Fault-Tolerant Quantum Computing.

Frontier The decisive experiment is a single sample series in one apparatus in which charging power, ergotropy and self-discharge time are all measured at each of at least ten well-separated values of N. Not three quantities in three papers on three device generations — three quantities on one axis. Everything this brief assesses turns on whether the collective speed-up is still there when the ordinate is work rather than absorbed energy, and whether buying storage time costs it.

Established No published experiment does this, and the reason is that each paper was optimised for one figure. The 2022 work spans about 76-fold in N and reports energy and power but no work extraction and no storage time. The 2024 work reports 57 microseconds of storage at essentially one configuration. The 2026 work reports electrical output across a 2.8-fold span of N. Each is a good paper; the union of them is not the measurement.

Frontier The experiment needs no new physics and no new instrument class, which is why its absence is a choice rather than a limit. The transient-absorption rig behind the 2022 scaling series, the phosphorescence-lifetime measurement behind the 57-microsecond figure and the photocurrent measurement behind the 10-to-40-microwatt figure have all run in the same laboratory group on the same device architecture. Nobody has announced a plan to run them on one sample set and no funder has been named as paying for it, so this brief treats the horizon as unscheduled rather than near.

Frontier A decade of N is the minimum, and the reason is arithmetic rather than taste. Distinguishing power going as N from power going as N to the three-halves means distinguishing exponents 1.0 and 1.5. Across a 2.8-fold range those predictions differ by a factor of 1.67 in the dependent variable, comparable to device-to-device scatter in solution-processed organic films. Across a tenfold range they differ by a factor of 3.16, and across the 2022 work’s 76-fold range by a factor of 8.7.

Established The uncertainty on a fitted exponent falls roughly as one over the logarithm of the range of the independent variable. That is ordinary regression rather than a claim about quantum batteries, and it is why the range of N is not a detail of the decisive experiment but the experiment itself.

Frontier The second experiment is an independent replication of superabsorption scaling by a group with no authors in common with the 2022 paper. Different dye, different cavity fabrication, different transient-absorption rig, same claim. The field has three landmark papers and one lineage, and a systematic error in the inference of N from concentration and mode volume would show up in all three identically.

Speculative The third experiment is a direct test of whether faster charging costs extractable work. The 2026 continuous-monitoring result predicts that correlations built during fast collective charging reduce ergotropy and that monitoring the environment recovers it. In a microcavity this would mean charging the same film at two different collective coupling strengths and measuring the photocurrent yield per absorbed photon at each. If the yield falls as the coupling rises, the central promise of the field is a trade rather than a gain.

Frontier A fourth, cheap experiment: calibrate N independently. Measure the molecule count in a finished cavity by a route that does not pass through the same concentration-and-thickness model used to build it — quantitative absorption against a standard, elemental analysis, or mass loss on ablation. Every exponent in this literature inherits the uncertainty of that number and none of them report it as an error bar on the abscissa.

Frontier A fifth: run the NMR platform’s scaling protocol on a battery ensemble rather than a single battery spin. Joshi and Mahesh already have advantage measured against the square root of N across N = 3 to 36 with 200-second storage; the missing piece is that the battery is one spin, so no statement about a battery of N cells follows. A star-topology system with several battery spins would give the field a scaling series with a directly measurable energy and storage long enough to run extraction at leisure.

Speculative A sixth, for the quantum-computing claim: run one gate off a quantum battery and account for every joule. The test of the Physical Review X proposal is a single two-qubit gate at stated fidelity with the battery’s energy budget measured before and after, against the same gate driven conventionally. Nothing about that needs a large machine.

Handwave The experiment most often implied in coverage — scaling a quantum battery until it charges a consumer device — is not an experiment, because no proposed route closes the six-thousandfold energy-density gap. Stating it as a test gives it a status it has not earned. The scalable version of the field’s own ambition, stated by its principal experimentalist, is coupling many microscopic quantum batteries into an integrated array, and that is an engineering programme whose first milestone is still the three-quantity measurement above.

Frontier A negative result that would matter: a careful scaling series showing exponent 1.0 within error. The 2022 supplementary table already contains one pairwise charging-power exponent of 0.94. If a full series came back flat, the collective-absorption physics would survive untouched, but the quantum-battery framing of it would not, and the field would properly become a branch of polariton photophysics.

Established The device is a Fabry-Perot microcavity containing an organic active layer, and its engineering vocabulary is organic photovoltaics. The 2026 stack is a copper-phthalocyanine absorber with fullerene charge-transport layers between mirrors, and the 2022 device was Lumogen-F Orange in polystyrene at an active-layer thickness near 185 nanometres. Every fabrication problem in Perovskite Tandem Photovoltaics — solution-processing uniformity, interface quality, encapsulation — arrives unchanged.

Frontier The core engineering requirement is to run two contradictory optimisations in one film. The principal experimentalist states it as balancing optical performance and electrical functionality in one structure: the cavity wants a high-finesse mode and a dense chromophore population, the extraction circuit wants percolating charge-transport pathways and low series resistance, and both want the same cubic micron.

Frontier Scaling is to be done by arraying, not by enlarging. The stated route is coupling large numbers of microscopic quantum batteries into an integrated array rather than growing one device, which is the same architecture every real battery uses and is a concession that the single-device physics does not scale indefinitely.

Frontier Arraying raises a question the field has not answered. If the advantage comes from collective coupling within one cavity mode, an array of independent cavities recovers ordinary extensive scaling between cavities and keeps the superextensive part only inside each one — so the array’s aggregate behaviour is classical and the quantum claim applies to a subunit.

Speculative Thermal management is unaddressed and is not yet a problem. A device dissipating tens of microwatts per square centimetre needs no thermal engineering. If power densities ever rise the three or four orders of magnitude that would make the device interesting, thermal design becomes the constraint it is in every other energy technology, and nobody has done that calculation.

Frontier For the cryogenic application, the engineering requirement is wiring, not energy. The Physical Review X scheme delivers gates through a shared resonator with one control parameter per qubit — its resonant frequency — instead of an individual drive line, and the claimed payoff is heat load and channel density rather than stored joules. That is a credible engineering argument about a real constraint, and it belongs in the same conversation as the control-wiring problem described in Fault-Tolerant Quantum Computing.

Frontier Metrology is an engineering requirement in its own right here. A device whose headline claim is an exponent needs calibrated molecule counts, calibrated mode volumes and error bars on both axes. The existing literature reports exponents without abscissa uncertainties, the cheapest thing the field could fix, and it matters more the narrower the range of N.

Speculative If arraying works, the manufacturing question is whether roll-to-roll organic processing can hold cavity tolerances. A cavity resonant with a molecular transition needs few-nanometre thickness control across the array, and the storage layer needs oxygen exclusion because triplets are quenched by molecular oxygen. Neither is novel; both are what makes organic electronics expensive, and the cost discussion in Advanced Battery Technologies is the right prior for how slowly that improves.

Established The nearest neighbour is quantum energy teleportation, and the two are confused because both sound like free energy and neither is. Quantum Energy Teleportation establishes that the sender’s measurement always injects more energy than the receiver extracts. A 2024 proposal joins the subjects, arguing QET lets the energy density at a subsystem of a fully charged quantum battery exceed the classical per-subsystem maximum — a redistribution by local operations and classical communication, not a capacity increase.

Established Superabsorption is a light-harvesting idea before it is an energy-storage idea. The mechanism is the same cooperative coupling studied in photosynthetic antenna complexes, and the relationship to that literature is covered in Quantum Effects in Biology. The engineering target there is absorption cross-section per unit area, which is exactly the quantity superabsorption improves, and it is a better-posed problem than storage.

Frontier Polariton physics is the parent field and owns most of the materials science. Strong light-matter coupling, polariton condensation and cavity-modified chemistry all use these devices; Quantum Materials describes the reproducibility problem afflicting every claim that rests on a fabricated sample, which is the relevant prior for a three-paper, one-group literature.

Speculative The comparison that sets expectations is nuclear isomers, the other exotic storage idea with a real physics anchor and no release mechanism. Nuclear Isomer Energy Storage stores at densities no battery approaches and cannot get the energy out; quantum batteries get the energy out at microwatt densities and store almost none. The two fail at opposite ends of the same pipeline, and both are more useful as metrology than as energy.

Frontier Ergotropy and its measurement belong to quantum thermodynamics generally, which is where the accounting rules are still being written. The same formalism sets the work cost of computation described in Thermodynamics of Computation, and the quantum battery is in effect a testbed for whether those rules are operationally measurable in a mesoscopic solid rather than only in a two-qubit demonstration.

Frontier Precision measurement is where this physics has already paid. Collective enhancement of coupling is the same resource Precision Quantum Sensing converts into sensitivity, and a superabsorbing structure reads straightforwardly as a fast, sensitive photodetector — an application needing no storage at all.

Established The experimental centre of gravity is one Australian collaboration, with a second cluster in Italy and the United Kingdom. The organic-microcavity line runs through CSIRO with RMIT University and the University of Melbourne on the 2024 and 2026 devices; the 2022 anchor was a collaboration with Italian and British groups. No funder or named programme appears in the reporting this brief could reach, which is itself informative: this is university physics, not a funded technology programme.

Frontier The theory community is large, distributed and growing faster than the experiments. The 2024 to 2026 literature reached here includes groups in China, Spain, Italy, Poland, Austria, India, Israel, Japan and Australia in Physical Review Letters, Physical Review X, PRX Energy and Reviews of Modern Physics. The ratio of theory papers to devices is the institutional fact to carry away.

Frontier The field needs a reporting standard before it needs money. Papers report stored energy per molecule, capacity in electron-volts, capacity in watt-hours, normalised ergotropy, discharge power density and dimensionless advantage ratios, and no two of the three landmark experiments report the same quantity. A one-page convention — joules, watts per square metre, a measured half-life, N with an error bar, and ergotropy or an explicit statement that it was not measured — would do more than a grant.

Established The unit slip in the 2024 preprint is a small demonstration of the cost. A capacity written both as 1.23 × 10¹⁴ eV and as 5.47 microwatt-hours, quantities differing by a factor of a thousand, survived into a widely-cited preprint. A house convention on units would have caught it in review, and the fact that a reader must do the conversion themselves to notice is the argument for the convention.

Frontier Press amplification is running well ahead of the measurement, in a specific and checkable way. Trade coverage of the 2026 result describes a device that “charges faster as it scales up”, gives capacity as “a few billion electron-volts” and renders the storage ratio as a homely analogy — one minute of charging to two years of retention.

Established The correction is in the same article, after the headline. The ratio is right and the analogy invites the reader to imagine two years; the same piece quotes the principal investigator saying capacity is tiny and hold time is a few nanoseconds. This is not a field misrepresenting itself. It is a name doing work the measurements cannot.

Speculative Institutional risk here is the ordinary one for a small field with a marketable name. “Quantum battery” attracts attention that “superabsorbing polariton microcavity” would not, and the attention arrives attached to expectations about cars and phones the physics cannot meet. The risk is that a funding cycle keyed to the name outlives the patience keyed to it.

Frontier The live ethical question here is honest description, and it is unusually tractable. No one is harmed by a microcavity. The available harm is misallocation: a student choosing a doctorate, or a government choosing an energy-research line, on the strength of a name implying a technology the measurements do not support.

Established The field is largely doing this correctly at the primary-literature level. The 2024 Reviews of Modern Physics review states that a fully operational proof of principle does not exist and names energy density, absolute power and storage lifetime as unexplored. The 2022 paper names its own integration problem. The principal experimentalist is quoted in trade press saying capacity is tiny and hold times are nanoseconds. The gap opens between the papers and the coverage, not inside the papers.

Speculative There is a distributional point worth stating even though the sums are small. If the quantum battery’s real destination is the inside of a quantum computer, its benefits accrue to whoever owns quantum computers, which is a short list of states and firms. Framed as energy storage, the research reads as broadly beneficial; framed accurately, it is infrastructure for an already-concentrated capability, and the concentration argument in Compute Concentration applies to it.

Frontier The opportunity cost is measured in attention rather than money. A few university groups is not a budget that competes with grid-scale storage research. But a subject occupying the “next-generation battery” slot in a minister’s briefing without being a battery has a cost that appears in no ledger.

Speculative Against that, the case for continuing is sound and does not depend on the battery framing. Superabsorption is a genuine collective quantum-optical effect useful in detection and light harvesting; the ergotropy formalism is the right accounting for work in small quantum systems and needs experimental tests; the quantum-computing power-delivery application is well posed. A field can be worth funding and badly named at once, and saying so is not a criticism of the science.

Handwave The civilizational claim usually attached to quantum batteries — that they change how societies store energy — is not supported by any measurement and is not likely to become supported. A device at 0.12 watt-hours per litre against lithium-ion’s 700 and rising, with self-discharge measured in microseconds against a lithium cell’s months, is not a slower version of a storage technology. It is a different object with a similar name, and the honest civilizational implication of an energy technology that does not store energy is none.

Frontier The realistic stake is narrower and sits inside the energy budget of computing. If quantum processors scale, the energy and thermal cost of controlling them becomes a real constraint, and a coherent on-chip energy source that removes drive lines addresses it directly. The argument that computing energy is a civilizational-scale variable is made in Ultra Efficient Computing Energy Systems, which puts real hardware a hundred million times above the thermodynamic floor with a reachable ceiling near two hundred times. A quantum battery is a candidate contributor to that last factor, in one narrow class of machine.

Speculative A second stake is methodological and may outlast the devices. Quantum batteries are the first place where quantum thermodynamics has had to produce operationally measurable quantities in a room-temperature solid rather than in a thought experiment. Whether ergotropy can be measured on 10¹⁴ molecules is a question about the reach of the whole formalism, and it matters to the accounting in Thermodynamics of Computation whatever happens to the battery.

Frontier A third, smaller stake: superabsorbing detectors. A structure that absorbs faster than its molecule count would suggest is a photodetector argument before it is an energy argument, and detector sensitivity reaches astronomy, medical imaging and free-space optics in a way that microjoule storage does not. If this programme has a durable output it is likelier to be a detector than a battery, and that would still be a success.

Handwave Claims that quantum batteries make energy scarcity obsolete, or enable devices that never need charging, have no basis in the measured record at all. They are not weak versions of a true claim; they invert the finding. The measured device charges very fast and discharges very fast, which is the opposite of never needing charging, and it stores about a six-thousandth of what the cell in a phone stores per unit volume.

These horizons track when specific, stated quantities could be measured or built, not when a product ships. Each is anchored to a named gap in this brief.

• 10 yr: Frontier The three-quantity scaling series — charging power, ergotropy and self-discharge at ten or more values of N in one apparatus — is run, probably by the CSIRO-RMIT-Melbourne line or a European polariton group. An independent replication of superabsorption scaling appears or conspicuously does not. The single-gate energy-budget test of the Physical Review X proposal is attempted. Storage times reach milliseconds by singlet fission or charge separation. Discharge power density improves one to two orders of magnitude on the 10 to 40 microwatts per square centimetre now reported, or the extraction problem is conceded.
• 25 yr: Speculative If the scaling series comes back positive on ergotropy, superabsorbing structures appear as detectors and light-harvesting layers in commercial products, and as an on-chip energy source in at least one quantum-computing architecture. If it comes back flat, the subject is absorbed into polariton photophysics and the name is retired. Both outcomes are consistent with today’s evidence, which is what makes the measurement in 7 worth running.
• 50 yr: Speculative A quantum-coherent power-delivery layer inside cryogenic quantum processors is either standard or abandoned, and the deciding factor is whether wiring heat load rather than qubit quality is the binding constraint at scale — the argument running through Fault-Tolerant Quantum Computing. Nothing on this horizon involves storing energy for a grid or a vehicle.
• 100 / 250+ yr: Handwave Any horizon on which quantum batteries perform bulk energy storage requires a route to roughly three orders of magnitude more energy density than molecular excited states permit, and no such route has been proposed. Stating a century-scale date for it would be inventing one; the honest entry is that the physics as understood does not go there, and a brief such as Energy Storage Revolutions is where the long-horizon storage question actually lives.

• Depends on nothing on this map blocks the decisive measurement — the apparatus exists — but the applications depend on results elsewhere: the quantum-computing power-delivery case waits on the scaling argument in Fault-Tolerant Quantum Computing, the fabrication case on the sample-reproducibility problem in Quantum Materials, and the honest energy comparison on the cost and density baselines in Advanced Battery Technologies.
• Requires (not on this map) a single measurement in which charging power and extractable work are recorded together over a decade of N, which nobody has published; a storage state at room temperature holding charge for milliseconds without making it unextractable, where the best measured figure is 57 microseconds; a fabrication route that certifies molecule count independently of the concentration model used to build the film, so the scaling exponent has an error bar on both axes; a customer whose job is served by microwatts delivered in picoseconds, because that is the only performance envelope this device has; and an agreed set of reported figures of merit, absent which two devices cannot be compared and a thousandfold unit slip can survive into a cited preprint.
• Enables superabsorbing photodetectors and light-harvesting layers with absorption cross-sections above the independent-molecule limit; on-chip coherent power delivery for cryogenic quantum processors, removing per-qubit drive lines; and the first experimental tests of ergotropy as a measurable quantity in a mesoscopic room-temperature solid.
• Adjacent Quantum Energy Teleportation on what entanglement does and does not do with energy, Nuclear Isomer Energy Storage on the mirror-image failure mode, Quantum Effects in Biology on cooperative light harvesting, Thermodynamics of Computation on the work-accounting formalism, and Precision Quantum Sensing on the same collective coupling used for sensitivity instead of storage.

Handwave “Quantum batteries charge faster the bigger they get, so an electric car could charge in seconds.” The first clause is a fair summary of a measured effect; the second does not follow from it. The scaling applies within one collectively-coupled ensemble sharing a single optical mode, over molecule counts near 10¹⁴ in a film a few hundred nanometres thick storing on the order of 10⁻⁵ joules. A car battery stores about 10⁸ joules.

Handwave The advantage is a factor on the rate; the deficit is thirteen orders of magnitude on the quantity. A bounded factor does not close a gap of that size, and the theory bound says the factor is bounded.

Handwave “Quantum batteries store more energy than classical ones.” They store far less. The only device-level energy density in the literature reached by this brief is 2.63 × 10¹⁸ eV per cubic centimetre, which is 0.12 watt-hours per litre, against commercial lithium-ion above 700 watt-hours per litre before 2018. The confusion comes from theory papers about exceeding the classical per-subsystem energy density, which is a statement about how energy is distributed among subsystems of a fixed total, not about the total.

Frontier “The speed-up is caused by entanglement.” For the Dicke batteries actually built, its own field’s review says otherwise: the advantage ratio for the Dicke battery “was found not to scale with N”. The theory that does tie rate to entanglement — entanglement depth at least the ceiling of N times the square of the normalised charging rate — applies to the idealised global-unitary charging model, not to a cavity mode. Entanglement is the right currency for the theoretical maximum and is not demonstrated to be what the experiments are using.

Frontier “Superextensive charging violates thermodynamics, so it must be wrong.” It does not. Superabsorption is a rate effect from collective coupling to a shared mode, and it moves energy faster without creating any. The bound that does apply is the 2022 Physical Review Letters ceiling: quadratic speed-up at most, global operations required. Neither impossible nor unbounded.

Speculative “A thousandfold storage improvement means the storage problem is solved.” Nanoseconds to 57 microseconds is a real and large gain, and 57 microseconds is still eleven orders of magnitude short of the months a lithium cell holds charge. It is long enough to run extraction experiments and to power a gate; it is not long enough to store anything overnight, and the same paper reports that the route cost about half the achievable triplet population.

Frontier “The 2026 device has a capacity of 5.47 microwatt-hours.” That figure belongs to the 2024 triplet device and appears to be a unit slip. The same sentence gives the capacity as 1.23 × 10¹⁴ eV, which is 5.47 nanowatt-hours, and the electron-volt figure is the one consistent with the paper’s own energy density and film volume. Anyone quoting the microwatt-hour number is overstating the device by a factor of a thousand.

Speculative “Superextensive scaling has been demonstrated across a wide range of system sizes.” The 2026 electrical-output devices span a factor of 2.8 in molecule count. The 2022 optical work spans about 76-fold but reports pairwise exponents from 0.94 to 2.19, one of them not superextensive at all. The best-behaved scaling measurement in the field is an NMR experiment at N = 3 to 36 whose battery is a single nuclear spin. No one of the three is well described as a wide range.

Handwave “Quantum batteries are a near-term energy technology.” Nobody in this literature has proposed a route to bulk energy storage. The field’s own review says a fully operational proof of principle does not exist; its principal experimentalist says capacity is tiny and hold times short; the scaling strategy is arraying microscopic devices, with no announced first product. Treating this as an energy technology on a product timeline is a category error, not an optimistic forecast.

Frontier “The theory bounds prove a quantum advantage exists.” A bound is an upper limit, not an achievement. The 2025 analysis makes this concrete: a model reaching nearly N-squared scaling under the uncertainty-principle bound loses the advantage entirely under the tighter Fisher-information bound, because the variance driving the loose bound came from state mixedness rather than useful entanglement. Its conclusion — that a superlinear power bound “may not be tangible or even exist in practice” — is the correct reading of every bound in the field.

Speculative “It is only a matter of engineering now.” The central physical quantity — how much work these devices give back — has never been measured on one. Ergotropy has been verified on two photons and simulated up to about thirty spins; no tomography route exists for 10¹⁴ coupled molecules. A field whose defining figure of merit is unmeasured on its own hardware has a physics problem before an engineering one, and 7 states the fix.