An atomically thin sheet of material should not be able to block light. That is the intuition the new result violates, and the violation is the point. In a paper now published in Physical Review Letters, Haowei Xu of City University of Hong Kong and MIT, Wang Yao of the University of Hong Kong, and Ju Li of MIT calculate that a moiré exciton superlattice — a two-dimensional crystal of trapped excitons, one atom thick — can have a transmittance of nearly zero at a collective resonance, and can be switched to nearly full transparency by a 1° adjustment of the twist angle or less than 1% heterostrain. The same collective physics, the authors argue, gives the lattice long-lived subradiant states that could hold a photon, and a gate-controlled way to put it in and take it out.
Every verb in that summary needs its qualifier, and this desk will keep applying it. This is a calculation, not a measurement: no moiré lattice has been switched from opaque to transparent in a laboratory, no subradiant state has been loaded with a photon and read back. What the paper offers is a prediction — specific, quantitative, and testable. The trade press picked it up in mid-October; the analysis here is about what the predicted effect would mean if it is real.
Why a single layer blocking light is surprising
A few-layer two-dimensional material is normally almost transparent. Monolayer graphene, the standard comparison, transmits about 97% of incident light. There is simply not much matter there to absorb or reflect, and even an exciton resonance in a single layer removes only a small fraction of the light passing through.
The paper’s claim is that the moiré lattice defeats that accounting — not by adding more matter, but by making the matter radiate cooperatively. Each exciton is a point-like dipole, localized to a few nanometers by the moiré potential, sitting on a regular lattice whose spacing is a substantial fraction of the resonant wavelength. At the collective resonance, the field each exciton scatters interferes with the fields scattered by all the others, and the scattered field cancels the transmitted field almost exactly on the far side of the sheet. In the paper’s worked example — a twist angle of 0.2°, a non-radiative linewidth of 100 GHz — the transmittance minimum appears at a detuning of about −161 GHz from the exciton resonance.
The mechanism is a collective resonance, the same physics that lets a two-dimensional array of cold atoms act as a nearly perfect mirror. What is new is the claim that it arises naturally, without any assembly, in a condensed-matter system — because a moiré superlattice is an ordered array of emitters, formed by the stacking itself, with one exciton per site.
The real-space point the paper is making
The paper’s deeper argument is not about the switch but about what controls it. Most treatments of optical response in condensed matter work in reciprocal space: define a susceptibility per unit cell, multiply by the number of cells, and the real-space arrangement enters only through scaling factors. That works when emitters are packed far below the wavelength, so their relative positions do not matter. Moiré excitons break that assumption. The moiré superlattice constant is 10 to 100 nanometers when the twist angle is a degree or less, against a resonant wavelength of hundreds of nanometers, so excitons a few sites apart radiate with phases that differ by a meaningful amount, and the interference between them reshapes the total response. The real-space structure is not a detail; it is the physics.
The collective shift of the exciton resonance depends strongly on the lattice spacing, and the lattice spacing is set by the twist. Change the twist by less than a degree, and the collective resonance moves far enough that a fixed incident frequency goes from sitting on the resonance — opaque — to sitting far from it — transparent. Heterostrain does the same thing by a different route: stretching one layer relative to the other changes the lattice constant in the same way. The abstract’s headline number is heterostrain of about 1% or less; the body of the paper phrases the same result as under 2%. Either way, a structural change of order one part in a hundred flips an optical property from one extreme to the other — and heterostrain, unlike twist angle, can in principle be applied in situ and reversibly.
Superradiant and subradiant: the two lives of a collective exciton
The transmittance switch is the headline, but the more consequential claim for quantum technology concerns the lifetimes of the collective states. The paper models the lattice in the single-excitation subspace — one exciton shared coherently across the array — and diagonalizes the effective Hamiltonian for a finite lattice of about 2,874 excitons, roughly 50 by 50 sites spanning about 10 by 10 micrometers. The eigenstates are collective excitons labelled by their in-plane wavevector, and their radiative decay rates split sharply at the light cone.
For an infinite lattice the split is absolute: states with in-plane wavevector inside the light cone radiate faster than a single exciton — superradiant — while states outside it have a radiative decay rate of exactly zero. They are completely dark, because a collective exciton can only emit a photon whose in-plane momentum matches its own. In the finite 50-by-50 lattice the transition softens, but the contrast survives.
The subradiant side is where the photon-storage idea lives. The paper states that subradiant lifetimes can be extended by orders of magnitude relative to a single exciton — a large claim with an important qualification: it is a claim about the radiative lifetime. A real device also loses excitations non-radiatively, and the subradiant advantage only survives if non-radiative decay and inhomogeneous broadening stay below the collective radiative rate. The paper’s own numbers set the bar: transmittance below 0.2 requires inhomogeneous and non-radiative linewidths of about 100 GHz or less, and the system tolerates resonance-frequency disorder of order a few hundred GHz. Those are demanding but not implausible — moiré exciton linewidths as narrow as 20 GHz have been reported, and linewidths down to tens of GHz at a few kelvin are in the literature. The honest reading is that the effect is within reach of the best current samples, not of typical ones.
The gate switch: moving an excitation across the light cone
A long-lived dark state is a good place to store a photon, but a storage device also needs a way to write and read. A photon arriving from outside necessarily excites a superradiant state — only states inside the light cone couple to propagating light. So the device must move an excitation from inside the light cone to outside it, and back again, on demand.
The paper’s proposed mechanism is an electric field gradient, applied by gates. Interlayer excitons carry a permanent out-of-plane electric dipole, and a field gradient in the plane imprints a position-dependent phase on the exciton — the analogue, for a neutral dipole, of the way an electric field accelerates a charged particle. The collective exciton’s wavevector drifts with the duration of the applied gradient, at a rate set by the dipole moment times the gradient divided by Planck’s constant. With a gradient of order 10¹⁰ volts per square meter, the paper estimates the crossing can be done within nanoseconds. Reverse the gradient for the same duration, and the excitation comes back inside the light cone and radiates.
The sequence this enables is the paper’s central device proposal: a photon arrives and is absorbed into a superradiant collective state; the gate gradient pushes the excitation outside the light cone, where it sits dark and long-lived; later, the gradient is reversed, and the photon is re-emitted. That is a photon memory, built from the lattice’s own collective states rather than from an external cavity or an atomic ensemble.
One caveat on temperature: none of this is room-temperature physics as calculated. The linewidth numbers the paper relies on — tens of GHz — are cryogenic numbers. A photon memory that works at a few kelvin is still a photon memory, but the reader should not carry away an image of a room-temperature optical switch.
What would count as confirmation — and what would not
A prediction this specific invites a checklist, because the most likely misreading would be to treat any twist-dependent transmission change as confirmation.
First, the zero. A transmittance minimum near zero at a specific detuning, with the minimum moving as the twist angle changes, is the basic signature. But a transmission dip can also come from ordinary absorption or a cavity effect in the sample stack, so the dip alone is weak evidence. The discriminating feature is the pattern of the minimum with twist angle: the paper shows the transmission curves shifting systematically as the twist varies, with the near-zero minimum tracking the collective shift. A measurement mapping that pattern across several twist angles would be much harder to explain away.
Second, the switch. The stronger test is the toggle itself: at fixed frequency, transmission near zero at one twist angle and near one at another less than a degree away, or the same toggle driven by heterostrain in a single device. Heterostrain is the sharper version of the test, because it removes the confound of comparing different samples: one device, one frequency, strain on and off, transmission flipping.
Third, the lifetimes. The photon-memory claim needs its own evidence. The prediction is that excitations pushed outside the light cone live orders of magnitude longer than single excitons, and that a gate gradient can move an excitation across the boundary in nanoseconds. Demonstrating that requires a time-resolved experiment: excite, apply the gradient, wait, reverse it, and detect the re-emitted photon — with the storage time and the retrieval efficiency both reported. A long lifetime alone would not be enough; the gate-controlled transfer in and out is the fingerprint.
Fourth, the robustness. The paper claims the effects survive positional disorder, missing excitons, and variations in individual dipole strengths, but are sensitive to disorder in the exciton resonance frequencies. That asymmetry is itself testable — and it sets the practical bottleneck: samples with narrow, uniform exciton lines are the prerequisite.
How this fits with the rest of the field
The closest prior work is Kumlin, Srivastava, and Pohl’s 2025 Physical Review Letters study of superradiance in strongly interacting dipolar moiré excitons, which showed that dipole-dipole interactions between interlayer excitons can suppress superradiance in disordered states while enhancing it in ordered phases. Xu, Yao, and Li build on that line but make a different point: that the real-space lattice structure itself, through the collective shift, is a tuning knob for transmission — and that the knob is sensitive enough to function as a switch.
The paper also connects, deliberately, to the atomic-physics literature on cooperative effects in ordered emitter arrays — the work of Shahmoon, Asenjo-García, and others on two-dimensional atomic arrays. The moiré claim is that nature provides the array for free: perfectly ordered, unit filling, one exciton per site, with a lattice constant tunable by twist and strain. Artificial atom arrays struggle to reach that combination. Whether the moiré lattice also solves the disorder problem — the resonance-frequency uniformity the effect is sensitive to — is the open question.
Why this matters, stated carefully
It is tempting to read “single-photon storage” and hear a quantum memory ready for deployment. The honest version is longer. A usable photon memory needs more than a long-lived dark state: efficient loading, high retrieval fidelity, a known output state, and a storage time that fits a real protocol. This paper establishes none of those. What it establishes is that the ingredients exist — a bright state to couple to, a dark state to hide in, and an electrical handle to move between them — in a system that is two-dimensional, gate-tunable, and fabricated by stacking rather than by trapping atoms one by one.
The transmittance switch is the more immediately testable half of the paper, and it may also be the more consequential in the near term: an atomically thin optical switch, toggled by strain or by a gate, is a component integrated photonics does not currently have in this form. But the same caveats apply: the switching is predicted at a collective resonance — a specific frequency, a narrow band, low temperature, a high-quality sample. A switch that works at one wavelength in one device at a few kelvin is a demonstration, not a technology.
The right summary, then: three theorists have calculated that a moiré exciton superlattice should behave as a collective optical element — opaque at its collective resonance, transparent a degree of twist away, with subradiant states that live orders of magnitude longer than single excitons and a gate gradient that can move excitations between the bright and dark sides of the light cone. Until a laboratory reports the switch flipping, or a photon coming back out of a dark state, the claims remain what they are: a precise, falsifiable prediction, awaiting its measurement — and the version of the story in which the device already exists should be resisted.
Sources and further reading
- Haowei Xu, Wang Yao, and Ju Li, “Cooperative Quantum Optical Effects of Moiré Exciton Superlattices,” Physical Review Letters, DOI: 10.1103/sp3h-pkqx; arXiv:2603.06998 (submitted March 7, 2026; revised October 2, 2026). All figures quoted above are taken from this paper. Note: the paper’s abstract gives the heterostrain figure as ≲1%; the body text phrases the same result as under 2%.
- Quantum Zeitgeist, “Quantum Moire Exciton Superlattices Switch From Opaque To Transparent,” October 2026 (trade-press coverage of the paper): https://quantumzeitgeist.com/city-hong-quantum-moire-exciton-superlattices/
- J. Kumlin, A. Srivastava, and T. Pohl, “Superradiance of Strongly Interacting Dipolar Excitons in Moiré Quantum Materials,” Physical Review Letters 134, 126901 (2025): the closest prior treatment of superradiance in moiré excitons.
- E. Shahmoon, D. S. Wild, M. D. Lukin, and V. Yelin, “Cooperative Resonances in Light Scattering from Two-Dimensional Atomic Arrays,” Physical Review Letters 118, 113601 (2017): the atomic-array cooperative optics this paper draws on.
- A. Asenjo-Garcia et al., “Exponential Improvement in Photon Storage Fidelities Using Subradiance and ‘Selective Radiance’ in Atomic Arrays,” Physical Review X 7, 031024 (2017): the subradiant photon-storage idea in atomic arrays.
- H. Yu and W. Yao, “Luminescence Anomaly of Dipolar Valley Excitons in Homobilayer Semiconductor Moiré Superlattices,” Physical Review X 11, 021042 (2021): the moiré exciton physics underlying the emitter model.
- Note on verification: this article reports a theoretical prediction. No moiré exciton superlattice has been experimentally switched between opaque and transparent states as of this writing, and no subradiant photon storage in such a lattice has been demonstrated. All figures attributed to the paper are results of model calculations, not measurements. The paper was submitted to arXiv on March 7, 2026 and revised October 2, 2026; it has since appeared in Physical Review Letters. No market-size or commercial figures are quoted, because the paper makes no commercial claims.