A solar cell is usually imagined as a region: a junction spread over an area, with a depletion zone, a built-in field, and carriers that drift and diffuse over distances far larger than any single atom. Shrink the device to two atomic layers and that picture strains. The junction becomes thinner than the lengths the textbook picture relies on, and the question is no longer only how a junction behaves at small scales, but whether the idea of a junction survives at all when the active region is comparable to an atom.
A paper published in Nature Nanotechnology on August 26, 2026, answers that question in a striking way, and it is the subject of this article because our regular scans missed it. This is not breaking news; it is a missed result, now read closely. Nam Thanh Trung Vu, Mingjun Chen, Yi Wei Ho and colleagues, led from the National University of Singapore with collaborators at the Singapore University of Technology and Design, A*STAR, and Japan’s National Institute for Materials Science, report that individual ionizing vanadium acceptors in bilayers of tungsten diselenide, WSe₂, act as point-like photovoltaic junctions. Using photoconductive atomic force microscopy, they map photocurrent hotspots confined to about a nanometre around single dopants, show that the current polarity flips depending on which layer holds the dopant, and find that device photocurrent scales linearly with dopant concentration. Their paper is titled “Point-like photovoltaic junction in 2D semiconductor homobilayer.”
The claim is deliberately narrow, and that is what makes it interesting. The authors do not report a record-efficiency solar cell, and no power-conversion efficiency figure should be attached to this work — none appears in the published abstract or in the summaries of the paper. What they establish is more basic: in an atomically thin homobilayer, a single charged impurity can supply the asymmetry that charge separation requires, and it can do so locally, at a scale set by the material itself rather than by a depletion region.
The textbook junction, and why it breaks down here
In a conventional photovoltaic device, charge separation is a collective, extended affair. Doping one side of a semiconductor p-type and the other n-type creates a space-charge region whose collective field bends the bands and separates photoexcited electrons from holes. The relevant lengths — the depletion width, the diffusion length — are typically hundreds of nanometres to micrometres, vastly larger than the spacing between dopant atoms. Individual dopants disappear into averages. That averaging is the continuum limit, and almost all of solar-cell engineering lives inside it.
Two-dimensional semiconductors strain that limit from both sides. Screening is reduced in an atomically thin crystal, because field lines between two charges spill into the surrounding vacuum or dielectric rather than staying within the material. Coulomb interactions are therefore unusually strong, and excitons — the bound electron-hole pairs created when light is absorbed — are bound far more tightly than in most bulk semiconductors. At the same time, the entire device is only a couple of atomic layers thick, thinner than any depletion region a bulk picture would predict. As the paper’s abstract puts it, it had remained unclear how individual dopants contribute to the photovoltaic response “when the junction thickness becomes comparable with the atomic length scales and smaller than conventional depletion or diffusion lengths.”
The WSe₂ homobilayer is a deliberately clean test case. A homobilayer is two sheets of the same material, so no chemical interface between different compounds does the separating work, as in a heterojunction. Stack two WSe₂ sheets, and light can create two kinds of exciton: an intralayer exciton, with electron and hole in the same sheet, tightly bound; and an interlayer exciton, with the carriers separated vertically between the sheets, carrying a permanent out-of-plane dipole. The interlayer form is the more useful one for charge extraction, because its carriers are already parted — but converting a tightly bound intralayer exciton into a charge-separated state requires a local driving force. In a symmetric homobilayer, there is no obvious source for that force. The vanadium dopant supplies it.
The camera: photocurrent at the scale of one atom
The measurement that carries the paper is photoconductive atomic force microscopy, or pcAFM. A conductive probe scans across the sample while light illuminates it, and at each point the probe records the tiny photocurrent generated there. Sweep the probe across the surface and the result is a map of the local optoelectronic response, registered against the atomic structure itself. It is the same family of technique that the Twente group used to image moiré-confined excitons in twisted bilayer MoS₂ at room temperature, reported in this publication in October — a needle that reads photocurrent, rather than a spectrum averaged over thousands of unit cells.
In vanadium-doped WSe₂ bilayers, those maps do not show a uniform response. They show sharply confined photocurrent hotspots centred on individual vanadium dopants. That spatial coincidence is the first half of the argument: the dopants are not passive chemical defects that happen to sit in a photoactive material. They are the active photovoltaic sites. Excitons formed within the local region of a charged dopant have an enhanced probability of becoming charge-separated, and the photocurrent emerges as a collection of point-like sources, each determined by the position, charge state, and layer location of one impurity.
Vanadium enters WSe₂ as an acceptor: it can capture electronic charge and become ionized, leaving behind a localized electrostatic potential. In a bulk device, that potential would be one contribution among millions, smoothed into a space-charge region. At atomic thickness, with reduced screening, the potential of a single ionized acceptor remains highly localized and can interact directly with nearby excitons. The modelling reported in the paper places the effective exciton-dissociation region at approximately one nanometre — a length comparable to the thickness and interlayer spacing of the material, not to the depletion lengths of bulk electronics. Within that tiny volume, the dopant’s field is strong enough to compete with the exciton binding energy. Outside it, excitons behave much as they would in undoped material. The junction, in other words, has a radius rather than an area.
The polarity test: which layer holds the dopant
The most striking observation, and the one that most firmly establishes the point-like picture, is the layer dependence. Dopants in the top WSe₂ layer generate a photocurrent of opposite polarity to dopants in the bottom layer. Flip the dopant from one sheet to the other, and the measured current reverses direction.
That reversal is a direct signature of the vertical structure of the process, and it is hard to explain with any extended-junction picture. A charged acceptor in one sheet modifies the local energy landscape differently from an equivalent acceptor in the other sheet: under illumination, the resulting local field favours the movement of one carrier toward the probe or electrode while directing the opposite carrier away. Switching layers effectively reverses the orientation of the local junction, and therefore reverses the current. The polarity thus encodes depth information that a lateral map alone could not provide — not only where a dopant sits in the plane, but which side of the homobilayer contains it.
The microscopy does the identifying; the device measurements, described next, do the counting.
From hotspots to devices: photocurrent that counts dopants
Local imaging alone could be dismissed as a curiosity of one flake under one probe. The paper therefore pairs it with vertical WSe₂/V:WSe₂ homobilayer devices, in which an undoped WSe₂ layer is combined with a vanadium-doped layer, and measures the macroscopic photocurrent as a function of dopant concentration. The result is linear: doubling the number of ionizing acceptors approximately doubles the total photoresponse, over the measured range.
Linearity is the signature of independence. If the dopants acted through a shared, extended field — if adding more of them reshaped the depletion region, altered the built-in potential, or changed carrier transport in a coupled way, as happens in a bulk homojunction — the total response would be expected to bend, saturate, or otherwise depart from a simple sum. Instead, the data point to an additive collection of nanoscale junctions, each contributing roughly on its own. The device behaves like a sum of its atoms, not like an average over them.
The devices display a second, subtler signature of discreteness: a compensation voltage — the applied bias at which competing photocurrent contributions cancel — that remains independent of both illumination power and dopant density. Its independence from optical intensity suggests it is not set by how many photocarriers are generated; its independence from dopant concentration suggests the underlying energy offset is not substantially reshaped by the total dopant population. In the authors’ interpretation, that behaviour reflects the discrete nature of the junctions and the energetic alignment between the layers, rather than the broad, density-dependent electrostatics of bulk junctions.
Photocurrent spectroscopy and quasi-classical modelling complete the mechanism. The measurements indicate that charged dopants locally convert tightly bound intralayer excitons into charge-separated interlayer states: near an ionized acceptor, the local potential changes the relative energetic cost of keeping both carriers in the same layer, and the electron and hole lower their energy by separating between the two sheets. The result is an interlayer exciton, or a related charge-separated configuration, with a built-in dipole that can assist subsequent extraction — driven here not by a band alignment between different materials, but by a single impurity’s field inside one material.
What it is not
Three cautions, stated plainly because the framing invites overreading. First, this is not a solar-cell result in the commercial sense. No efficiency is claimed, and the vertical devices are test structures designed to count dopants, not to compete with silicon or perovskite panels. Anyone quoting this paper as a step toward a practical atomically thin solar panel should say in the same breath that the paper demonstrates a mechanism at the single-dopant scale, and that scaling it into a useful device — controlling dopant number, position, and charge state across a wafer, extracting current efficiently, and doing so reproducibly — remains entirely ahead.
Second, this is not a demonstration of deterministic dopant placement. The paper visualizes dopants where they happen to sit in grown crystals, identified by scanning transmission electron microscopy and by their photocurrent signatures; it does not report placing vanadium atoms one by one at chosen coordinates. The vision the paper gestures toward — photovoltaic functionality engineered one microscopic junction at a time — is a direction, not a capability already in hand.
Third, the result is specific to its material and geometry. Vanadium in WSe₂ is a particular acceptor in a particular host, chosen in part because vanadium substitution in WSe₂ is an established route to p-type doping, with prior work on bound hole states around individual vanadium atoms in monolayer WSe₂. Whether other acceptors, other hosts, or monolayer rather than bilayer geometries produce equally clean point-like junctions is an open question. The paper’s claim is properly read as an existence proof in one well-characterized system, not as a general law of doped two-dimensional semiconductors.
Why it matters beyond one bilayer
The deeper reason this result belongs in a publication named for the exciton is that it changes the status of the dopant in exciton physics. Dopants and defects in two-dimensional semiconductors are usually cast in one of two roles: as sources of disorder, trapping, and unwanted recombination to be minimized, or as creators of localized emitters — single-photon sources at defect sites — to be exploited for quantum optics. The Vu paper adds a third role: the dopant as the functional centre of a photovoltaic process, the thing that splits the exciton rather than the thing that spoils it or traps it.
There are concrete follow-ups worth watching. One is placement and control: can growth or implantation techniques position ionizing acceptors precisely enough that a designed array of point junctions becomes fabricable, and does the linear scaling survive to the densities such an array would require? A second is generality: do other acceptor-host combinations show the same layer-dependent polarity and power- and density-independent compensation voltage? A third is extraction: the paper establishes dissociation within about a nanometre, but a device must also collect the separated carriers before they recombine, and collection at atomic thickness is a separate challenge from the dissociation problem this paper addresses.
None of those questions diminishes the August result. A field that has spent a decade treating dopants statistically now has a direct image of what one dopant does, alone, to light. The photovoltaic junction, reduced to its smallest possible form, turns out not to be a region at all. It is a point, about a nanometre wide, with a polarity set by which of two atomic sheets holds a single vanadium atom.
Sources and further reading
Vu, N. T. T.; Chen, M.; Ho, Y. W.; Wu, Q.; Wong, C. Q.; Chen, Y.; Budniak, A. K.; Loh, L.; Verzhbitskiy, I.; Watanabe, K.; Taniguchi, T.; Bosman, M.; Ang, L. K.; Trushin, M.; Eda, G. “Point-like photovoltaic junction in 2D semiconductor homobilayer.” Nature Nanotechnology (2026), published August 26, 2026. DOI: 10.1038/s41565-026-02251-9. Abstract and author, affiliation, contribution, and funding records via the Nature article page; summary via ScienMag, “Researchers create point-like photovoltaic junction in two-dimensional semiconductor homobilayer,” August 2026.
For context on vanadium doping of WSe₂, see Kozhakhmetov et al., “Controllable p-type doping of 2D WSe₂ via vanadium substitution,” Advanced Functional Materials 31, 2105252 (2021), and Mallet et al. on bound hole states around individual vanadium atoms in monolayer WSe₂, Physical Review Letters 125, 036802 (2020), both cited by the Vu paper. For the photocurrent-imaging context, see this site’s report on real-space imaging of moiré-confined excitons in twisted bilayer MoS₂ and its explainer on interlayer excitons. This article is analysis of published research, not investment advice.