Work with us
Crosstalk capping your array?
Integrate the dark qubit — measure the gains on your process, then license.
ethan@admissibletech.com
Flux-tunable superconducting chips don't scale cleanly, and the reason is mundane: every flux-bias line is a little antenna that talks to its neighbors. Add qubits and it compounds — each new bias line can disturb every qubit already on the chip, so the calibration burden grows faster than the chip itself. You end up tuning the whole chip against itself.
Our control wiring confines the field instead of broadcasting it. A coaxial solenoid delivers the flux a qubit needs while a matched return cancels the stray field by geometry — an exact, first-order cancellation — so the bias stays local. The crosstalk matrix goes near-diagonal: each qubit feels its own line and almost nothing from anyone else, which turns calibration from a chip-wide fight into N independent jobs you can run in parallel.
Against a conventional bias wire that's a ≥300× reduction in crosstalk, validated in field solvers. It's also a manufacturing non-event — the solenoid is wound in the standard trilayer process with no extra mask steps. To be precise about scope: this confines uniform, laboratory flux noise; it isn't a fix for surface-spin, dielectric, or quasiparticle noise, and it's validated in simulation, not yet silicon.
Provisional filed (App. 64/022,566), eight related inventions. EM/field-solver validated; not yet silicon-proven.
The two panels show the same control task done two ways. A conventional flux line broadcasts its field across the chip, so one bias knob nudges many qubits at once. The VPT solenoid traps that field inside itself — full control authority on its own qubit, almost none on the rest of the array. Confinement is only the first defense; symmetry cancels whatever field is left.
First, a multi-turn coaxial solenoid confines the bias field to its own interior, so the field a neighbor feels falls to a few parts in ten thousand of the control field. Second, a gradiometric geometry gives the qubit two equal, opposite loop areas, so a uniform external field cancels to first order — the qubit is "magnetically dark." Add a Z₂-symmetric operating point and you have three independent layers of field protection acting at once.
The field-confinement and rejection numbers come from full electromagnetic field solvers (FastHenry and OpenEMS), giving a control coupling kctrl ≈ 0.68 and a stray-to-control ratio of order 10⁻³. The mechanism addresses uniform/neighbor flux noise; it is a design to validate on a partner's hardware, not yet silicon-proven.