A quantum-dot micropillar is one of the best single-photon sources ever built. Put an indium arsenide dot inside a cavity of alternating semiconductor mirrors and the cavity does two favours at once: it hurries the dot’s emission through the Purcell effect, and it funnels the photons into a clean, collectable mode. But the photons that come out are all the same shape: a near-Gaussian beam, linearly polarized, carrying no orbital angular momentum. For the high-dimensional quantum information tasks that motivate much of the field — encoding more than one bit per photon, entangling a photon’s polarization with its spatial structure — that sameness is a limitation, and fixing it has always meant adding bulk optics downstream.
A paper in eLight from Jiantao Ma, Dong Liu, Shunfa Liu and colleagues, published 23 April 2026, moves the beam-shaping onto the chip itself. The team mounted a metalens — an ultra-thin metasurface that steers light by imposing a designed phase profile — on the back of the very semiconductor chip that carries the micropillar source. The two components are fabricated on opposite sides of a 350-micrometre gallium arsenide substrate and aligned to within a micrometre. The photons leave the chip already shaped: focused or deflected, carrying orbital angular momentum, entangled in polarization and orbital angular momentum, or carrying a topological spin texture. The source metrics survive the integration. The trade press picked the paper up this week, seven months after publication; what makes it worth covering now is not novelty of date but what the combination demonstrates — and what it does not.
The conflict the device resolves
The reason nobody had done this cleanly before is a genuine incompatibility, and the paper states it plainly. A microcavity works because it is sharply resonant: a high quality factor and a wavelength-scale mode volume concentrate the optical field at one frequency. A metasurface works because it is broadband: subwavelength structures impose phase shifts across a wide range of frequencies. Put a metasurface inside or directly on top of a cavity and the two requirements fight — previous attempts at metalens–emitter integration produced weak light–matter interaction and, crucially, no indistinguishable photons.
The design here sidesteps the fight by separating the two functions in space. The micropillar and the metalens are decoupled in the near field: they sit on opposite faces of the chip, and the 350-micrometre substrate between them is not dead space but the mode-matching element. At the micropillar’s output the beam is about 2 micrometres across with a divergence angle of 8.25 degrees; propagating through the substrate, it expands to about 103 micrometres — the right size to fill the 200-micrometre-diameter metalens on the backside. Generation and shaping are independently optimised; the substrate does the coupling.
That geometry also explains the emission direction, which is the first thing to get right about this device. The top mirror has 30 pairs of aluminium gallium arsenide / gallium arsenide distributed Bragg reflectors; the bottom has 20. Because the top mirror is more reflective, photons are guided downward, into the substrate, which is transparent at the emission wavelength of about 920 nanometres. The source emits through the back of the chip, not the front — the opposite of the usual micropillar, and the reason the metalens can sit where it does.
The fabrication, and why alignment is the paper
Fabricating this device is a double-sided exercise. The planar cavity — 20 bottom DBR pairs, a gallium arsenide cavity layer containing a single layer of low-density indium arsenide dots, 30 top DBR pairs — is grown by molecular beam epitaxy on a double-side-polished wafer, with metallic alignment marks on the front by electron-beam lithography and matching marks on the back by double-sided photolithography, aligned within 1 micrometre.
Positioning the dots is the hard part. Conventional fluorescence imaging in reflection does not work here, because the dots must be located relative to marks on the opposite face. The team developed a transmission-mode imaging technique: the dots are excited from the front while the front-side marks are illuminated from the back with a 1050-nanometre LED, and both images are combined to fix each dot’s position. The 3-micrometre micropillars are then defined by aligned lithography and chlorine-based dry etching, the front surface planarised with SU-8 photoresist, and the metalens fabricated on the backside by a third aligned lithography and etch. Simulations show a 1-micrometre alignment offset has negligible effect on performance — fortunate, since that is the accuracy the process delivers.
The source still works
The first question any cavity physicist asks is whether the metalens costs anything. The paper answers it with a collimation metalens — the simplest possible shaping function — and a full characterisation of the source behind it.
The measured quality factor of the investigated cavity is 2823. Time-resolved measurements on the negatively charged exciton show the lifetime falling from 675.6(7) picoseconds off resonance to 205.5(2) picoseconds on resonance: a spontaneous emission rate enhancement of about 3.288(5), below the calculated Purcell factor of 5.37 — a shortfall the authors attribute to the reduced accuracy of positioning dots and alignment marks when imaging through the 350-micrometre substrate. Under pulsed resonant excitation the dot shows clear Rabi oscillations, the signature of a coherently driven two-level system.
The headline source metrics, measured at a π-pulse with a 13-nanosecond delay in the Hong–Ou–Mandel interferometer: second-order correlation g²(0) = 0.070(1), photon indistinguishability visibility of 0.736(2), and an extraction efficiency of 35.7(5) percent at the first collection lens — a figure that already includes the loss introduced by the metalens. The detected count rate is 595 kilohertz, and the coupling efficiency into a single-mode fibre is about 0.65. The far-field divergence collapses from 28.30 degrees (25.22 degrees in the orthogonal cut) without the metalens to 1.44 degrees (1.06 degrees) with it.
These numbers are good, and they are not the best. The paper says so itself: the indistinguishability is still lower than the state of the art for microcavity-based single-photon sources, which approach near-unity visibility, and a companion News & Views piece in the same journal calls the device “a notable balance between brightness, purity, and indistinguishability” rather than a record on any single axis. The achievement is that the numbers are what they are with a metalens attached — the first report of quantum radiation from a single emitter integrated with a metasurface at this level of source performance. The metalens does not degrade the cavity source.
Four degrees of freedom, one photon
With the source characterised, the paper turns to what the metalens actually does. The demonstration device is a single metalens designed to focus and deflect right- and left-handed circularly polarised photons in different directions while encoding each with a different helical phase front. The meta-atoms are cuboid gallium arsenide structures, 900 nanometres tall on a 400-nanometre period, acting as truncated rectangular waveguides; their dimensions and orientation set independent phase shifts for the two circular polarisations, the orientation providing the Pancharatnam–Berry phase that decouples the two channels.
The measured far field matches the simulation: right-handed photons are focused and deflected one way with topological charge −1, left-handed photons the other way with topological charge +3. The orbital angular momentum purities, measured by projecting through vortex waveplates, are 82.2(2) percent for the −1 channel and 72.0(4) percent for the +3 channel. Divergence, directionality, polarisation, and orbital angular momentum are controlled simultaneously, on one photon, from one device.
The next demonstration is entanglement — not between two photons, but between two degrees of freedom of a single photon. Under longitudinal-acoustic-phonon-assisted excitation, the source emits horizontally polarised photons with zero orbital angular momentum. Passing through a metalens designed as a q-plate with topological charge 1/2, each photon becomes a vectorial beam in which polarisation and orbital angular momentum are entangled in the Bell state |Φ₊⟩ = (|L,−1⟩ + |R,+1⟩)/√2. Quantum state tomography gives an entanglement fidelity of 0.8914 ± 0.0023. Single-particle entanglement of this kind is a resource for high-dimensional encoding: one photon carries a two-qubit state.
Single-photon skyrmions
The most advanced demonstration is also the most conceptually interesting. An optical skyrmion is a beam whose polarisation vector, traced across the beam’s cross-section, wraps the full Poincaré sphere — flipping from down at the centre to up at the edge while swirling in between. The team synthesises it by passing horizontally polarised, zero-OAM photons through a metalens that imprints a helical phase front on the left-handed component only, giving the state (|R,0⟩ + |L,−s⟩)/√2, where s is the skyrmion number.
The measured skyrmion number for s = 1 is 0.99, against a theoretical value of 1, and the team also generates skyrmions with numbers from 2 to 5. The point is not the number but what it survives. The team propagates single-photon skyrmions through artificial atmospheric turbulence — Kolmogorov phase screens imposed by a spatial light modulator, 50 realisations at each strength. The local Stokes parameters distort and the skyrmion boundary wobbles, but the skyrmion number stays stable, while photons carrying only orbital angular momentum degrade faster. In simulation it falls only from 1.00 to 0.98 at turbulence strength D/r₀ = 0.5. This is a laboratory turbulence test, not a free-space link demonstration — but it is the first time the resilience claim for skyrmionic light has been tested at the single-photon level in an integrated device.
The honest boundary
Four qualifications locate this result precisely.
First, the indistinguishability. A visibility of 0.736 is far from the near-unity values of the best bare micropillar sources, and the paper does not hide it. The gap is attributed to the excitation method and the charge environment of the dots — the same factors that limit ordinary micropillar sources — rather than to the metalens. That is encouraging in one sense: the metalens is not the bottleneck, so the usual remedies should transfer. But it also means this device, as built, would not yet serve in an interference-based protocol that demands near-perfect photons.
Second, the temperature and the wavelength. The device operates in a closed-cycle cryostat with a base temperature of 5.8 kelvin, and the dots emit at about 920 nanometres — near-infrared, but not a telecom wavelength. Neither is a flaw; both are the standard conditions of quantum-dot photonics. But they constrain any deployment story, and they are worth stating because coverage of “quantum light sources” has a habit of omitting them.
Third, the yield and the statistics. The fabricated micropillars have quality factors from 1750 to 3300, and for about 20 percent of them a single dot’s excitonic transition sits spectrally close to the cavity mode — one in five devices is the working fraction, a laboratory yield, not a manufacturing one. The reported metrics come from a single investigated dot, characterised thoroughly rather than many devices thinly. That is the right choice for a first demonstration, and it leaves the reproducibility question open.
Fourth, the scope. This is a source-engineering result, not a quantum computer and not a network demonstration. No two-photon interference between separate sources is shown, no entanglement is distributed, and no computation is performed. What is shown is a single-photon source whose output light can be shaped in multiple degrees of freedom at the point of generation.
What to watch next
The follow-ups that would matter most follow directly from the qualifications. The first is indistinguishability: whether the same integrated device can reach the near-unity visibilities of the best bare micropillars once the charge environment is stabilised, since the metalens itself is not the limiting factor. The second is yield: whether deterministic positioning and strain tuning raise the usable-device fraction from one in five toward something a foundry would recognise. The third is a second source: two micropillar-metalens devices interfering with each other would test whether shaped photons from separate chips are mutually indistinguishable — the precondition for any network use. The fourth is the skyrmion programme: whether the turbulence resilience holds in a real free-space link.
There is also a parallel thread worth noting. In July 2026, a Shanghai Jiao Tong team reported a related but distinct approach in Physical Review Letters: a 200-nanometre-thick “metacavity” in which the metasurface itself forms the cavity, with a Purcell factor of about 9.7. Ma and colleagues keep the cavity and the metasurface separate on one chip; the metacavity fuses them into one structure. Which scales better is an open question.
The conceptual point of this paper is narrower than its press coverage and more durable than its numbers. For a decade, the field has treated source quality and beam shaping as separate problems solved by separate components: a cavity makes the photon good, and a table full of optics makes it useful. This device shows the two can be one component without either being compromised — not perfectly, not yet at record performance, but measurably and on a single chip. The photon leaves the chip already shaped.
Sources and further reading
- Jiantao Ma, Dong Liu, Shunfa Liu, Jiawei Yang, Nilo Mata-Cervera, Bo Chen, Xueshi Li, Guixin Qiu, Kaixuan Chen, Hanqing Liu, Haiqiao Ni, Dunzhao Wei, Zhichuan Niu, Ying Yu, Yijie Shen, Liu Liu, Xuehua Wang & Jin Liu, “High-performance sources of multidimensionally engineered quantum light based on monolithic microcavity-metalens interfaces,” eLight 6, 16 (2026). DOI: 10.1186/s43593-026-00130-3. Published 23 April 2026. Primary source for all figures in this article. Preprint: arXiv:2603.15391 (submitted 16 March 2026).
- News & Views commentary: “Monolithic microcavity-metalens interfaces: bridging performance and complexity in quantum light sources,” eLight (2026). DOI: 10.1186/s43593-026-00134-z — used for the framing that the reported figures balance brightness, purity, and indistinguishability without yet matching the best cavity-only sources.
- Trade coverage: “Monolithic Microcavity-Metalens Interfaces Enable Multidimensional Quantum Light Control,” ScienMag, 18 November 2026 — the coverage that surfaced the paper this week; all figures cross-checked against the eLight paper.
- Parallel approach: “Metacavity Quantum Electrodynamics,” Physical Review Letters 137, 023601 (2026). DOI: 10.1103/j8gx-58hf. Published 7 July 2026 — the 200-nanometre metacavity with Purcell factor ~9.7, cited for comparison.
- Note on verification: all quantitative values here are taken directly from the eLight paper, cross-checked against the ScienMag coverage of 18 November 2026. The paper is from April 2026, not this week — the news this week is the coverage, and the article is framed accordingly. No market-size figure, commercial claim, or investment judgement appears in this article; the site’s Terms carry the not-investment-advice disclaimer.