For years, the moiré exciton has been a creature known by its shadow. Twist two atomically thin semiconductors by a degree or two, and the interference pattern between their lattices — the moiré superlattice — creates a periodic energy landscape that should trap excitons at regular, programmable positions. The evidence was strong but indirect: photoluminescence spectra split into sharp lines, absorption resonances shifted with twist angle, magnetic-field oscillations hinted at trapped dipoles. All of it was consistent with excitons sitting in a lattice of potential wells. None of it showed an exciton sitting in one.
A paper published in Nature Physics on August 26, 2026, changes that, and it is the subject of this article because our regular scans missed it. This is not breaking news; it is a missed milestone, now read closely. Laurens Westenberg and colleagues at the University of Twente, working with theorists at Utrecht University, crystal growers at Japan’s National Institute for Materials Science, and a collaborator at Brazil’s Centro Brasileiro de Pesquisas Físicas, report the first real-space imaging of moiré-confined excitons inside individual moiré unit cells. Using photocurrent atomic force microscopy at room temperature, they mapped where two kinds of exciton — direct and indirect — actually settle in a twisted bilayer of molybdenum disulfide, MoS₂. The answer is not what the simplest picture predicts: the two species choose different stacking sites within the same unit cell, and the map of where they appear is not a map of the deepest wells.
A landscape everyone believed in and nobody had seen
The moiré idea is simple to state and hard to verify. Take two sheets of MoS₂, each a semiconductor just three atoms thick, and rotate one relative to the other by about two degrees. The local alignment — the stacking registry — drifts slowly across the sample, cycling through several distinct arrangements before repeating. The result is a superlattice with a period far larger than the atomic lattice: in this experiment, about 9 nanometres, so roughly ten thousand repeats fit across the width of a human hair. Within each repeat, the changing registry modulates the local electronic environment, producing a periodic potential for excitons — the bound electron-hole pairs that carry light’s energy through a semiconductor.
That potential has been the field’s central promise. If excitons can be parked at chosen positions, spaced a few nanometres apart, simply by choosing a twist angle, then a twisted bilayer becomes a programmable lattice of light-matter excitations: a platform for ultrasmall light sources, optical sensors, and the study of interacting quantum particles in a controlled geometry. What it had not produced, until now, was a picture.
The reason is a mismatch between the tools available and the scale of the question. Optical spectroscopy — photoluminescence and absorption — measures energies with great precision but averages over thousands of moiré cells at once. It can show that a periodic potential exists and that excitons feel it, but it cannot say where, inside one cell, a given exciton sits. Sharper microscopies that resolve nanometre structure have generally required cryogenic temperatures, far from the conditions in which any device would operate. Two excitonic states with similar energies might live at entirely different stackings, and an optical signal’s brightness might reflect how efficiently excitons are created at a site rather than where they end up. The field has lived with that ambiguity for years. This site’s own moiré coverage reflects it: our explainer describes the landscape in general terms, and our recent report on the optical Aharonov-Bohm effect in twisted bilayer MoS₂ rests on magneto-photoluminescence oscillations — a phase signature of trapped dipoles, not an image of their positions. The Westenberg paper is a different kind of evidence.
The camera: a needle that reads photocurrent
The measurement is photocurrent atomic force microscopy, and its logic is direct. First author Laurens Westenberg scanned the twisted bilayer with an atomically sharp probe while light of a precisely tuned colour illuminated the sample from below. At each point, the probe recorded the tiny electric current the light generated — a current that depends on how efficiently photoexcited carriers are produced, separated, and collected at that exact position. Sweep the probe across the surface, and the result is a map of the local optoelectronic response with nanometre resolution, registered against the moiré pattern itself.
Two features of the experiment matter more than its cleverness. The first is that it works at room temperature. Earlier high-resolution approaches to moiré structure demanded extreme cold, which left open whether the spatial signatures they revealed would survive the thermal agitation of a real environment. The Twente maps show that they do: the site selectivity of moiré excitons is not a cryogenic curiosity. The second is that the method is performed on a real, working device geometry rather than a bare flake in a microscope: photocurrent is, by definition, a device-relevant quantity, the conversion of light into collected charge.
What the maps show: two species, two addresses
The headline result is site selectivity, and it comes in two parts. First, excitons in the twisted bilayer do not spread across the moiré cell or pool indiscriminately. Each stays confined to a spot of roughly two nanometres — small compared with the nine-nanometre moiré period, and small enough that a single unit cell holds distinct, resolvable addresses rather than a blur.
Second, and more surprising, different kinds of exciton choose different addresses. The paper distinguishes direct excitons, in which the electron and hole occupy configurations that let them recombine efficiently and emit light, from indirect excitons, in which the electron and hole are separated in layer, momentum, or both, recombining more slowly and living longer. That separation changes how each species experiences the moiré potential, and the maps show it: one group gathers at the crossing points of the moiré lattice, the other settles in the regions in between, where the first group is absent.
The distinction is not academic. Direct excitons are the natural candidates for light-emitting applications, because they recombine radiatively; indirect excitons, with longer lifetimes and susceptibility to electric fields, are the candidates for transport and storage. If each species can be directed to a chosen registry, then a single twisted bilayer could host light generation, energy transfer, and information storage at predetermined locations, separated by nanometres. The maps do not demonstrate such a circuit. They demonstrate the prerequisite: that the sorting exists and can be seen.
The subtle point: brightness is not depth
The paper’s most important insight may be the one that complicates its own pictures. It would be natural to read a photocurrent map as a map of the moiré potential — bright where the wells are deep, dark where they are shallow. The authors show that this reading is wrong, or at least incomplete. The contrast in their images reflects not just potential depth but the interplay of local generation, exciton motion, and trapping. A stacking site may generate excitons efficiently without being their final low-energy destination; excitons born elsewhere may migrate toward a different registry before they are detected. The photocurrent at a point is therefore a composite of where excitons are created, how they move, how their charge distributes between the layers, and where they get trapped.
This matters beyond one paper, because it explains a standing puzzle. Different experiments on moiré materials have sometimes produced apparently different spatial patterns, inviting the suspicion that one of them must be wrong. If the measured signal depends on the balance between generation and trapping, then different measurement geometries — different illumination, different contacts, different detection — can legitimately produce different maps of the same underlying physics. The Twente result does not just add a picture; it supplies a reading rule for the pictures that already exist, and for the ones other groups will now produce. Any future claim to have mapped a moiré potential will have to say how it separated trapping from generation.
The model that earns its keep
To interpret the images, the team built a Wannier-based effective moiré-exciton model, developed with the Utrecht theorists. Wannier functions are localized wavefunctions constructed from extended quantum states, and they provide a natural language for particles moving through a periodic potential: excitons become composite objects occupying effective sites of the moiré lattice, able to move between them by quantum tunnelling, with energies that vary according to the local stacking. The calculation reproduces both the measured exciton energies and their spatial localization — the direct species at one registry, the indirect species at another.
Agreement on both energies and positions is a demanding test, and it is the right one. A model that reproduces an averaged spectrum but predicts the wrong real-space pattern is, after this paper, visibly incomplete. The moiré field now has a benchmark of a new kind: theories of twisted van der Waals materials must get the addresses right, not just the energies. That constraint reaches into the details that matter most for design — interlayer coupling, dielectric screening, tunnelling rates, relaxation pathways — because each of those enters the calculation of where an exciton ends up. The model here passes that test for one material at one twist angle. Whether it generalizes is the open question.
What it is not
Three cautions, stated plainly because the framing invites overreading. First, this is not a device result. The experiment uses a device geometry and measures a device-relevant quantity, but it demonstrates imaging and sorting, not a working light source, sensor, or circuit. Westenberg’s own summary keeps the claim at the level of design: “Our results could help with the design of materials and devices that manipulate light absorption and electricity with much greater precision.” Could help with the design of — not “enables.”
Second, this is not a wavefunction image in the sense of our recent coverage of the Graz trPOT result, which reconstructed an exciton’s spatial wavefunction and phase in an organic film. Photocurrent microscopy maps where excitons are generated, move, and get trapped, as read out through collected charge; it does not recover the exciton’s internal quantum-mechanical phase. The Twente paper’s achievement is spatial: resolving species-specific addresses inside a moiré cell at room temperature. Its phase information comes from the model, not from the measurement.
Third, this is one material at one angle. Twisted bilayer MoS₂ at about two degrees is a deliberately clean test case — a homobilayer, where both sheets are the same material, so the moiré pattern comes purely from the twist rather than from a lattice mismatch between different compounds. Heterobilayers, different twist angles, and electrically gated structures will rearrange the landscape, and the site selectivity seen here need not transfer unchanged.
What to watch
The most interesting follow-ups are the ones the method itself makes possible. Apply photocurrent imaging to other twisted systems — heterobilayers such as MoSe₂/WSe₂, where interlayer excitons dominate; different twist angles, where the moiré period and the balance between competing states shift; gated devices, where an electric field can reshape the potential and, in principle, move confinement sites or change which species wins a given registry. Each of those experiments is now a direct test rather than an inference from averaged spectra.
There is also a quieter implication for the correlated-phases program. Moiré potentials are pursued partly as a route to collective excitonic states — ordered arrays of interacting excitons, and potentially condensates — because periodic confinement brings many interacting particles into a controlled geometry. Such phases depend sensitively on which sites the excitons actually occupy and how strongly neighbouring sites interact. A method that resolves site occupancy species by species gives that program an empirical footing it previously lacked.
The deepest reason this result matters is simpler, and it is the reason it belongs in a publication named for the exciton. Moiré physics has been, until now, a science of inference: beautiful spectra, sophisticated models, and a shared conviction about what must be happening at the nanoscale, assembled from signals that averaged over thousands of repeats. The Twente team has replaced a conviction with a picture — two species of exciton, sorted into complementary addresses inside a single nine-nanometre cell, at room temperature, on a working device. The moiré pattern is not merely a backdrop that shifts optical resonances. It actively determines where quasiparticles are born, where they move, and where they spend their time.
Sources and further reading
Westenberg, L. J. M.; Eek, L.; Verbakel, J. D.; Vonk, K.; Borggreve, S. J. H.; Watanabe, K.; Taniguchi, T.; de Boeij, P.; Arouca, R.; Morais Smith, C.; Bampoulis, P. “Real-space imaging of moiré-confined excitons in twisted bilayer MoS₂.” Nature Physics (2026), published August 26, 2026. DOI: 10.1038/s41567-026-03425-x. Preprint arXiv:2511.20398 (November 2025). University of Twente announcement via Science X, “A 2-degree twist creates a nanoscale grid that traps light energy at room temperature,” August 24, 2026. ScienMag summary, “Scientists directly image moiré-trapped excitons in twisted bilayer MoS₂,” August 2026.
For the wider moiré context, see this site’s explainer on moiré excitons and its report on the optical Aharonov-Bohm effect in twisted bilayer MoS₂ — a phase signature of trapped excitonic dipoles, distinct from the real-space imaging reported here. This article is analysis of published research, not investment advice.