Fluxus-Q

Served

Resonator frequency

The resonant frequency of a coplanar-waveguide readout resonator, from its layout geometry, with a fractional error band on every answer.

For: Readout-resonator designers who would otherwise run an eigenmode solve to place each resonator frequency.

0.48%

median error vs HFSS, claw given

1.09% p95 · grouped split · n = 1,160

Status. Served on the hosted Fluxus API.

Engine. Data-trained surrogate of an electromagnetic eigenmode solve.

Reference. SQuADDS HFSS eigenmode simulations — electromagnetic simulation, not measured silicon.

API

Endpoints

POST/v1/fluxus/resonator

one resonator → frequency and error band

Base URL https://api.cogitan.ai, bearer key from an approved account. Read GET /v1/fluxus/capabilities first: it states the same fidelity, validation, envelope and refusal conditions as this page. Access model.

Inputs

What you send

length_um

µm — total resonator length

kind

lambda4 | lambda2 — quarter- or half-wave

center_width_um

µm — CPW centre-conductor width

gap_um

µm — CPW gap

coupling_length_um

µm — optional; defaults to the trained value for the kind

claw_length_um

µm — optional; supplying it selects the claw-aware model

Outputs

What comes back

fr_ghz

GHz — predicted resonator frequency

rel_interval

fractional error band: the measured held-out p95, constant across inputs

source

engine identity

model_variant

which of the two models answered (claw-aware or claw-blind)

Validation

Measured against what, and how

Reference. SQuADDS HFSS eigenmode simulations — electromagnetic simulation, not measured silicon.

Sample. 1,160 simulated resonators over 19 distinct lengths.

Split. Grouped 5-fold cross-validation holding out whole resonator lengths, so every test resonator has a length the model never saw.

Shanto et al., SQuADDS: a validated design database and simulation workflow for superconducting qubit design, Quantum 8, 1465 (2024)

Claw length given

0.48% median

1.09% p95

Claw length not given

2.73% median

10.52% p95

Analytic CPW model (baseline)

20.03% median

what the surrogate replaced

Error-band coverage

0.983

constant band, 95% target

Envelope

Where the answer is trusted

λ/4 length

2,700 – 4,700 µm

λ/4 coupling length

100 – 500 µm

λ/2 length

2,700 – 8,700 µm

λ/2 coupling length

0 only

center_width_um

11.7 µm (fixed)

gap_um

5.1 µm (fixed)

claw_length_um

70 – 577.5 µm

Refusals

When you get no number

Any input outside the envelope returns HTTP 422 with code resonator_outside_envelope, naming the field, the value received and the allowed range, so the call can be repaired and retried.
The cross-section is pinned because the reference data holds it at one value. Answering off it was measured at 39% from HFSS — worse than the analytic formula — so it refuses instead.

Known limits

What it does not do

Without the claw length the model is a scoping tool: 2.73% median and 10.52% p95 is not enough to place a readout frequency. Supply the claw length when you know it.
The error band is a single constant (the measured p95), not a per-point interval. Per-point intervals were built and rejected on measurement: they covered less (0.893 against the 95% target) and were wider.
Only the CPW cross-section in the reference data is covered. A different centre width or gap is outside the envelope, however close.

Where something else wins

Outside the envelope the analytic CPW formula is the right tool, and the surrogate says so by refusing rather than extrapolating.

Corrected 2026-10-02

Until 2026-10-02 the endpoint did not check resonator length, and the envelope was applied as if λ/2 and λ/4 shared one. Default λ/2 requests were answered 13–25% high, outside the band served with them. The envelope is now per kind and the defaults are kind-aware.

Every figure on this page compares against a simulator or a published device measurement. None of it is a prediction of your fabricated hardware. For one certification-grade answer, run the full reference; this is for screening many designs cheaply and spending solver time on the survivors.