One molecule, made by chemists, works as a qubit you can read with light

A team mostly at the Ulm company NVision put a triplet-ground-state carbene molecule inside a host crystal of near-identical shape and addressed one molecule at a time with a laser: it initialised, controlled and read out the single molecule's spin, and reported millisecond coherence with dynamical decoupling and tens of milliseconds of spin relaxation at 4.5 K.

Research desk NVision, Ulm University and Hebrew University, arXiv preprint Read on 2 November 2026

A bar chart on a logarithmic time axis from 1 microsecond to 100 milliseconds for one carbene molecule at 4.5 kelvin: Hahn-echo coherence 12.2 microseconds, coherence under XY8 dynamical decoupling 2.2 milliseconds, and spin relaxation T1 21 milliseconds. Below, three panels: g2 at zero delay 0.14, showing light from one molecule; optical linewidth 38 megahertz; spectral stability over one hour. — Every number is the paper's own, at 4.5 K. The bar chart and its log scale are our drawing.

Qubits you can build atom by atom

A quantum network needs qubits that keep their state for a long time and can be read and linked with light. Today's best candidates in solids are defects in crystals, like the nitrogen-vacancy centre in diamond. They work, but every defect sits in a slightly different spot and no two are quite alike. Molecules offer a different deal: a chemist can make billions of identical copies and tune the structure on purpose. The catch has been that no molecule combined bright, stable light with a spin that lasts, so nobody could read a single molecular qubit optically.

What the NVision team reports

Researchers mostly at the Ulm company NVision, with Ulm University and the Hebrew University of Jerusalem, used a triplet-ground-state carbene, a molecule whose ground state carries a spin, set inside a host crystal of almost the same shape. Cooled to 4.5 K, single molecules could be picked out one at a time: the light passed the single-emitter test with g⁽²⁾(0) = 0.142, the optical line of one molecule was 38(2) MHz wide, and it stayed put to within about 2.5 MHz for more than an hour.

They then initialised, controlled and read out the spin of one molecule with light. Its coherence under a simple Hahn echo was 12.2(6) µs; under XY8 dynamical decoupling, a train of refocusing pulses, it reached 2.2(3) ms; spin relaxation, the ceiling on all of this, was 21(2) ms.

What this does not show

No spin-photon entanglement yet. The paper is titled a spin-photon interface and it does read the spin with light, but entangling the spin with an emitted photon, the step a network needs, is described as a future route.
Cold, not warm. Everything is at 4.5 K. The authors note the carbene becomes unstable around room temperature.
Milliseconds need help. The 2.2 ms is with thousands of decoupling pulses. Left alone, the spin loses its phase in about 12 µs.
A preprint, from a company. Posted on arXiv in May 2026, not yet peer-reviewed, and most authors work for NVision, which develops molecular qubits commercially. We found one inconsistency in the text, the doping level, given as 25 ppm in one place and 2.5 mol% in another.

Why it matters

It opens a third family of optically addressable spin qubits next to atoms and crystal defects, one that can in principle be designed on paper and made in bulk. Whether it beats diamond for networks depends on the entanglement step still to come. For how coherence times are measured and why decoupling pulses stretch them, see the basics and our note on a diamond qubit kept coherent by sound.

Sources

We read the sources above ourselves. Where a figure or number is ours, the article says so. If you find a mistake, tell us: a wrong sentence here gets logged like any other.