Short answer: The resistance of a Josephson junction sets a transmon's frequency, and it changes with time and with heating, which makes frequency targeting hard. A 2026 arXiv preprint reports that junction resistance ages on a logarithmic curve, faster in ambient air than in nitrogen or vacuum. A 2023 paper reports larger junctions with 1% frequency spread across 32 qubits, and a 2024 preprint reports tuning by alternating-bias assisted annealing with a precision of 7.7 MHz. All three are laboratory studies.
Why resistance matters
The 2023 abstract explains that reproducibility of qubit parameters is a challenge for scaling superconducting processors, because signal crosstalk constrains frequency separation between neighboring qubits. The frequency uncertainty of transmons comes from fabrication variation in junction area, tunnel barrier thickness and capacitor. The 2026 abstract adds that understanding junction aging, and the effect of annealing on resistance, is important for large-scale processors.
2026 preprint: aging and annealing
The authors study junctions stored under different conditions from immediately after fabrication up to two to three months. The aging curve follows a logarithmic curve. The aging amplitude is mainly set by fabrication conditions and the aging speed by storage conditions. Junctions stored at ambient laboratory conditions aged faster than those stored in nitrogen or vacuum, and the speed changed appreciably when storage conditions changed.
They also compare thermal annealing under nitrogen with annealing under ambient conditions up to 250 degrees Celsius. Under nitrogen, resistances decreased at all temperatures tested. Under ambient conditions, resistance increased at 200 degrees and decreased at 250 degrees. The authors could not decrease resistance below its initial value, which they say suggests a lower limit on the tuning range. This is a preprint dated February 2026 and has not been reviewed.
2023: reducing sensitivity
The authors fabricate larger Josephson junctions and report reducing the wafer-level standard deviation of resistance down to 2%. They characterize 32 identical transmons and report a frequency standard deviation of 40 MHz (1%), with quality factors above 2 million. Two-level-system spectroscopy shows no significant increase in the number of defects causing relaxation. By simulation, for their parametric-gate architecture and counting only errors from frequency uncertainty, they say scaling to 100 qubits would give an average of only 3 collisions between gate transition frequencies, assuming 2% crosstalk and 99.9% target gate fidelity. That is a simulation and not a built device.
2024: alternating-bias assisted annealing
The 2024 preprint reports precision tuning of tunable transmons by performing alternating-bias assisted annealing at room temperature with commercially available test equipment. They report a frequency-equivalent tuning precision of 7.7 MHz (0.17%) from targeted resistance tuning on hundreds of qubits, with a resistance tuning range up to 18.5%. Cryogenic measurements show evidence of improved coherence after annealing. On multi-qubit processors, they report an empirical frequency tuning precision of 18.4 MHz, and iSWAP gate fidelities as high as 99.51% (uncertainty 0.20) on two tuned 9-qubit processors. Their yield model predicts high yield beyond 1000 qubits, a modelled projection.
Limits
The three studies use different junction designs and tuning methods, so the numbers should not be treated as one comparison. Two are preprints.
FAQ
Why does junction resistance age?
The 2026 authors report a logarithmic aging curve that depends on storage conditions.
Does storage atmosphere matter?
Yes. Ambient storage aged faster than nitrogen or vacuum.
Can resistance be tuned down?
Under nitrogen annealing it fell, but the authors could not go below the initial value.
How precise is the 2024 tuning?
7.7 MHz from resistance tuning, and 18.4 MHz on multi-qubit processors.
Are these production methods?
These abstracts describe laboratory studies.
All field notes