Researchers demonstrate mobile qubits on silicon chips: what they achieved, why it matters, and what could come next

The core achievement in one paragraph

In May 2026, Nature published experimental results from a Dutch consortium centered on QuTech, TU Delft’s Kavli Institute of Nanoscience, and partners including TNO, demonstrating two-qubit quantum logic between single electron spins that are not parked in fixed dots during the interaction. Instead, each qubit rides in its own travelling electrostatic minimum—a kind of conveyor belt of voltage waves along a silicon–germanium stack—so the particles physically approach each other until their quantum wavefunctions overlap enough to turn on a controllable exchange interaction.

The paper also reports using that interaction to generate entanglement and perform conditional, post-selected quantum state teleportation across a larger spatial span. Together, the experiments argue that logic can happen on the move, not only after shuttling stops—an architectural degree of freedom that theorists have wanted for reconfigurable solid-state processors.

Vocabulary: what “mobile qubit” means here

A qubit is a quantum system with two levels, here the spin of an electron trapped in a nanoscale “quantum dot”—a puddle of confinement created by metal gate electrodes on a chip. Mobile means the dot’s potential well is dragged along a channel by time-varying voltages, so the electron relocates while engineers attempt to preserve its quantum phase.

This is distinct from everyday data buses: the goal is not merely transporting a classical bit, but keeping a coherent superposition—information stored in amplitude and phase—safe from noise caused by defects, vibrating nuclei, and charge jitter in the host crystal.

The specific numbers readers should remember

The Nature abstract and editorial summary highlight two headline benchmarks. First, when the two conveyor minima move symmetrically inward by about 120 nanometres each (240 nm combined approach), the team reports an average two-qubit gate fidelity near 99% for a shuttling-based conditional-Z-style interaction engineered via exchange. Second, they implement conditional post-selected quantum teleportation between spins whose relevant separation reaches about 320 nm, with an average gate fidelity near 87% in that protocol.

Those figures are lab metrics at cryogenic temperatures (about 200 millikelvin in the described setup) and high magnetic field (roughly 260 millitesla in-plane), not consumer-product specs. They nonetheless cross important psychological thresholds: two-qubit error near 1% is a regime where error correction discussions become concrete if surfaces or LDPC codes can be layered on with enough parallelism.

How the device is built (simplified)

The experiments use gate-defined quantum dots in an isotopically purified 28Si/SiGe heterostructure with a ~7 nm quantum well—silicon-28 is chosen to dilute magnetic nuclear spins that otherwise randomize electron phases. Titanium–palladium gates with aluminium oxide insulation sculpt double wells and barriers; a cobalt micromagnet supplies the magnetic gradient needed for electric-dipole spin resonance (EDSR), letting microwaves rotate qubits electrically.

The chip hosts a linear array of six quantum dots in the demonstration geometry, with sensing dots at the ends and ancilla spins used for parity readout via Pauli spin blockade variants. Phase-shifted sinusoidal signals on plunger and barrier lines create the travelling-wave potential that implements conveyor-mode shuttling—the same family of techniques that earlier achieved very high shuttling fidelities over micrometre-scale distances in related devices when only transport, not mid-flight gates, was the focus.

Why this is a breakthrough—not “just another shuttle”

Semiconductor labs had already shown that electron spins can be moved across meaningful distances with low loss of information—analogous to RAM that you can slide along a ruler. The harder problem is computing while sliding: bringing two mobile wavepackets close enough for a strong, tunable exchange coupling J without crashing into charge leakage, orbital excitation, or decoherence from electric noise amplified by field gradients.

The Nature work’s claim is that this middle step—high-fidelity two-qubit entangling gates between spins that remain in motion long enough to matter—is now demonstrated with interleaved randomized benchmarking-style evidence in the conveyor geometry. That upgrades mobile spins from a clever transport layer to a candidate universal primitive for large-scale layouts where wiring constraints would strangle static arrays.

Exchange coupling as the “volume knob”

Exchange is a quantum interaction that flips correlated spin states when electrons share orbital space. In double dots, J grows exponentially sensitive to barrier height and inter-dot distance; the paper describes tuning J up to tens of megahertz in some configurations while balancing dephasing times T2* that shrink as coupling speeds up.

The team explores both barrier-controlled double-well approaches and a merged-elongated dot regime where interaction saturates—behaviour they relate to strong electron–electron correlations akin to Wigner molecule physics. Not every regime is good for gates; part of the scientific contribution is mapping where speed, adiabaticity, and noise trade-offs produce a sweet spot for ~99% CZ performance.

Teleportation in this context (not science fiction)

Quantum teleportation here is not Star Trek matter transport. It is a protocol that consumes a Bell pair—maximum entanglement between two qubits—to transfer an unknown single-qubit state from one location to another using classical measurement outcomes and conditional rotations. The paper implements a post-selected variant spanning five dots in the array narrative, illustrating non-local information flow compatible with modular semiconductor tiles.

The ~87% fidelity quoted for that teleportation block is lower than the best direct two-qubit gate headline, reflecting accumulated errors across preparation, shuttling, entanglement, and readout. Error budgeting—which stage dominates—is the kind of follow-on work experimentalists will now publish in supplementary analyses.

Coherence while moving: the companion research front

Separate but aligned work from overlapping institutions on the arXiv (“Coherence Protection for Mobile Spin Qubits in Silicon,” early 2026) systematizes how noise behaves during transport: passive gradient reduction, motional narrowing (the qubit averages spatial noise by moving fast enough), dynamical decoupling pulses between shuttle segments, and dressed-state shuttling with continuous drives.

That preprint reports coherence times on the order of tens of microseconds under some shuttling protocols over >200 nm, arguing that transport need not be the weakest link if control waveforms are co-designed with material quality. Readers should treat arXiv manuscripts as not yet identical to peer-reviewed journals, but the direction—co-movement of hardware and control theory—is clear.

Why connectivity is a strategic bottleneck

Many quantum error correction codes assume you can entangle qubits that are not nearest neighbours without prohibitive swap overhead. Superconducting grids often pay swap taxes; trapped ions and neutral atoms have demonstrated reconfigurable graphs using motion or optical rewiring. Semiconductors without mobility risk layout dead-ends as fab yield forces irregular arrays.

Mobile spins aim to import the flexibility of atomic architectures into a CMOS-adjacent platform that could, in the long run, leverage industrial process control—provided cryo control electronics and calibration automation scale.

What is still extraordinarily hard

Millikelvin dilution refrigerators and microwave lines per qubit do not scale like smartphone chips yet. Yield across thousands of dots, crosstalk during global conveyor phases, heterogeneous integration with classical cryo-CMOS, and runtime calibration under 1 K heat loads remain engineering cliffs.

On science, quasi-particle excitations, interface defects, and microscopic charge telegraph noise still vary wafer to wafer. Mobile operations add time-dependent Hamiltonians that stress benchmarks designed for static qubits—new characterization standards will emerge.

Most-cited factual anchors from the Nature paper summary

Platform: gate-defined electron spins in 28Si/SiGe. Temperature anchor: ~200 mK. Field anchor: ~260 mT in-plane. Shuttle distance anchor: ~120 nm per qubit inward for ~99% two-qubit fidelity regime. Teleport anchor: ~320 nm separation scale, ~87% average gate fidelity (post-selected protocol). Architecture: six-dot linear array with conveyor channels and parity readout ancillas.

Numbers may shift slightly in final author PDF tables; cite Nature directly for precision work.

A plausible ten-year arc (speculative but grounded)

Near term (1–3 years), expect dense process control data from multi-dot arrays, better cryo multiplexing, and demonstrations merging mobile gates with simple error detection rounds. Mid term (3–7 years), foundries may pilot 300 mm Si/SiGe lines with epitaxy tuned for dot uniformity if risk capital and defense contracts stay strong.

Long term (7+ years), if mobile semiconductor qubits survive benchmarking against superconducting surfaces and photonic networks, their killer angle could be factory throughput: lithographic precision plus graph rewiring without rewiring cryostats.

Bottom line

The Nature 2026 result is not a finished quantum computer; it is a pivotal proof that silicon spin qubits can meet, entangle, and teleport while in motion with fidelities that invite error-correction thinking. Combined with coherence protection research on the same material platform, mobile dots look less like a niche curiosity and more like a credible path for flexible, factory-style quantum processors—provided the cold control stack keeps pace with the nanoscale silicon it serves.