Qubits and quantum devices
Can a laser fix mistuned qubits on a quantum chip?
Open access · cc by · source: Europe PMC
Briefly heating individual qubit junctions with a laser sets their frequencies precisely enough to avoid clashes between neighbours, without harming how long the qubits stay coherent.
Study at a glance
- Design
- Other — Laboratory process study: Josephson junctions on multiqubit chips were laser-annealed to target resistances, then chips were cooled and qubit frequencies, coherence and gate errors measured; Monte Carlo and gate simulations supplement the data.
- N
- No single N: 390 qubits were tuned in the precision experiment, 241 qubits from seven Falcon and two Hummingbird processors were used for frequency-assignment analysis, and 221 qubits on four partially tuned chips were used for the coherence comparison.
- Population
- Fixed-frequency superconducting transmon qubits on IBM 27-qubit Falcon and 65-qubit Hummingbird processors
- Outcome
- Frequency-equivalent tuning precision, tuning success rate, qubit coherence times (T1, T2), and two-qubit gate fidelity
Structured fields used in claim comparison tables when every cited study has a complete layer.
Key findings
About 89.5% of qubits reached their target, and the tuning itself was precise to about 4.7 MHz in frequency terms. Once chips were cleaned, bonded and cooled, the practical frequency precision was about 18.5 MHz, dominated by imperfect prediction of cold frequency from room-temperature resistance rather than by the laser step. Tuned and untuned qubits had essentially the same coherence (aggregate T1 about 79 microseconds), and a tuned 65-qubit processor had all 72 two-qubit gates working with a median fidelity of 98.7%.
Methodology
The team used a focused green laser to anneal the tiny tunnel junctions of fixed-frequency transmon qubits, nudging each junction's resistance (and so its frequency) down toward a planned target. They tuned 390 qubits to test how precise the process is, then cooled whole processors to measure actual qubit frequencies, coherence times, and two-qubit gate errors. They compared tuned and untuned qubits on the same chips and used simulations to estimate yield and gate error before and after tuning.
Limitations
The work comes from one company's fabrication line and processor designs, so it does not show how well the method transfers to other qubit architectures such as tunable-frequency qubits. The yield gains and the predicted drop in gate error rely on Monte Carlo and gate-error models rather than on a direct before-and-after measurement of the same chip. The method can only lower frequencies (resistance only increases, up to about 14%), and it does not address errors from next-nearest-neighbour or spectator qubits, which the authors flag as future work.
How this study connects
Role on claims
Each row is a claim on a concept or method page where this paper supports, challenges, or qualifies the statement. Roles are hand-checked — not a model guess.
Post-fabrication tuning helps frequency crowding.
Transmon frequencies can be tuned after fabrication: laser annealing hit 89.5% of targets with about 4.7 MHz tuning precision, did not hurt coherence (T1 about 79 µs), and gave a 65-qubit processor with a median two-qubit fidelity of 98.7%.
Evidence for the claim as stated.
Platform differences: spin-qubit studies measure dephasing and single-qubit fidelity on single devices, whereas transmon work reports processor-scale two-qubit fidelities; the numbers are not directly comparable across platforms.
Same question, contrary or null result.
Open questions
Tensions this paper is part of
From concept pages' “where studies disagree.” Disagreement means the same question; scope means different assays, populations, or outcomes.
Platform differences: spin-qubit studies measure dephasing and single-qubit fidelity on single devices, whereas transmon work reports processor-scale two-qubit fidelities; the numbers are not directly comparable across platforms.
Related papers in this topic
Same topic cluster — not a recommendation engine.