The Signal
Innsbruck Team Reports Trapped-Ion Gate Errors Below 0.5%
Researchers report two-qubit trapped-ion gates with error rates below 0.5% in crystals containing up to 12 ions. Their structured-light method is designed to reduce unwanted coupling to nearby collective motion modes while preserving the ab
Quantum.world Team · Written with AI · fact-checked ·
“We experimentally realize high-fidelity two-qubit gates with error rates below 5 × 10−3 in ion crystals containing up to 12 ions confined within a single potential well.”
Tommaso Faorlin and coauthors, the paper’s authors, write in Physical Review Letters.
A light gradient targets the crowded modes
A team including researchers at the University of Innsbruck and Alpine Quantum Technologies reports a trapped-ion entangling gate designed to limit a common source of error: unwanted coupling between a gate and nearby collective motion modes. The study, published in Physical Review Letters on October 23, 2026, reports two-qubit gates with error rates below 0.5% in ion crystals containing as many as 12 ions.
Trapped-ion processors use the ions’ shared motion to mediate interactions between qubits. As chains grow, nearby motional frequencies can crowd together, making it harder to drive the intended mode without also disturbing spectator modes. The resulting unwanted motion can reduce gate fidelity.
The researchers shape the driving light to create a transverse, time-varying force on selected ions. That force produces a geometric phase—the basis of the entangling interaction—while the ions move along trajectories that return to their starting points. The authors say orienting the force in a plane orthogonal to the light’s propagation reduces the effects of spectral crowding while retaining single-ion addressing.
Structured beams drive the interaction
The main experiment uses a chain of barium-138 ions in a linear Paul trap. Two off-resonant laser beams, one with a standard Gaussian profile and one shaped into a TEM10-like mode, interfere at the ions to create a spatial light gradient. Adjusting the beams lets the team address selected ion pairs and generate the state-dependent optical force.
The gate uses axial motion, along the ion chain. The paper compares axial and radial implementations and identifies suppression of spectator-mode coupling as a principal advantage of the transverse-gradient gate.
The demonstrated gate takes about 220 microseconds and traces four loops in phase space. In the paper’s discussion of a two-ion crystal, the authors estimate axial spectator-mode infidelity at about 10−4 and give a radial estimate of 7.4 × 10−2. Table I presents a separate error budget for an eight-ion crystal: it lists spectator-mode infidelities of 2.1 × 10−4 for the axial implementation and 575.4 × 10−4 (5.754 × 10−2) for the radial implementation.
Below 0.5% across chains up to 12 ions
To estimate average gate error, the team applies gates repeatedly and tracks the decay in fidelity, a procedure intended to separate gate errors from state-preparation and measurement errors. For chains with more than two ions, the reported infidelity averages results from gates on the innermost and outermost ion pairs. The paper reports errors below 5 × 10−3 for crystals containing up to 12 ions. Its main data include measurements with barium-138 ions and a separate calcium-40 measurement; the supplemental material also provides an error budget for 12 calcium ions.
The result demonstrates performance in a single potential well for the tested ion chains, rather than a processor-scale demonstration of error-corrected computation. The paper makes compatibility conditional: the qubit states must experience a differential AC Stark shift, which allows the light gradient to exert different forces on the two states. With that condition met, the authors say the method can work with arbitrary qubit encodings.
The measured error is not the full error budget
The paper distinguishes experimental gate-error measurements from a detailed modeled error budget. That budget combines contributions from several sources with experimentally characterized terms; its axial spectator-mode estimate for the eight-ion crystal is at the 10−4 level, while other sources also contribute. The authors present reduced spectator-mode coupling—not the removal of every error source—as the main benefit of the transverse-gradient design.
The reported gate falls below the half-percent error mark in ion chains of up to 12 particles while preserving pair selectivity. The paper’s comparison of axial and radial motion highlights the reduced spectator coupling associated with the axial implementation, alongside a gate duration of roughly 220 microseconds.