Should remain close to zero in the ideal model.
See the signal.
Understand the noise.
Explore how preparation and detector errors change a known quantum prediction. Keep the model fixed. Make the assumptions visible.
The experiment readout
Start with what should happen.
Run the ideal baseline, then introduce a preparation or detector fault. Compare the expected change with the variation from finite measurements.
The interacting contrast after subtracting the free control.
Three-particle clustering
Chance that the three model particles occupy consecutive sites. Values below use every shot, including invalid encoded outcomes.
| Setting | Ideal | With faults | Measured | Invalid |
|---|
— invalid outcomes retained in the denominator. No favorable outcome selection.
Uncertainty, definitions & run provenance
A synthetic experiment, with a fixed scope. Saved quantum states supply the prediction. This version changes preparation and endpoint readout only; it does not simulate control pulses, a refrigerator, gate noise, or a physical device.
Calibrate a virtual detector
Prepare known bit strings, take synthetic readings, and estimate the detector response from those readings.
Measure the detector before trusting it.
Known inputs let us separate detector response from the quantum prediction. Correlated errors require information beyond average bit error rates.
A fault you haven’t seen
Inspect a synthetic experiment with hidden assumptions. Record a diagnosis before revealing the injected fault.
Blind challenge
A seeded scenario from a fixed set of faults. The result may not uniquely identify its cause.
Hidden from this screen, not cryptographically secret. Local source and browser tools can expose the scenario.
Look at the evidence first.
Two contrast values rarely tell the whole story. Use the free control, clustering readings, and invalid outcomes to decide what remains uncertain.
What a real laboratory adds
This bench is one useful layer. A laboratory has separate preparation, control, environment, and measurement systems.
Fourteen model sites are not fourteen physical qubits here.
Our three-particle configurations are encoded into nine bits. A model-site detuning is not automatically a flux change on one hardware qubit. Physical controls need an explicit, tested mapping and separate calibration evidence.
Quantum device
Chip · Couplers · Readout resonators
A superconducting device provides physical energy levels and couplings. Their connectivity, leakage and drift need an apparatus-specific model.
Cold, quiet environment
Dilution refrigerator · Filters · Shielding
Cryogenic equipment and carefully engineered wiring reduce thermal and electromagnetic noise. Cooling alone does not establish quantum fidelity.
Control & timing
Waveform generators · FPGA · Flux sources
Controllers deliver pulses. Wiring can distort the intended waveform; a requested detuning reversal needs to be checked against the achieved one.
Detection & calibration
Amplifiers · Digitizers · Reference measurements
Weak signals become classified bits. This bench models declared classical bit errors and synthetic reference readings, without analog microwave signals.
Cold atoms & lattice microscopes
Vacuum chambers, laser cooling, lattice optics and high-resolution imaging can study spatial many-body physics directly. A demonstrated microscope may detect occupation parity, not exact particle number.
Our independent nearest-neighbor interaction, open boundaries and preparation need a new physical mapping. Cold atoms do not automatically reproduce the compact digital encoding.
A model helps ask a better question.
Published superconducting experiments have demonstrated bound states of microwave excitations. Those are distinct experiments with their own geometry, gates and outcome selection.
Agreement inside this bench tests software and declared assumptions. A new physical claim needs actual observations, calibrated controls, independent replication and a defensible comparison.
Primary research sources · Links open only when selected and require an internet connection. The experiment bench itself works offline.
Exports keep the settings used for that result, including the calibration's own seed and reference shots. Each new result gets a record ID; exporting it again keeps the same ID and filename. Local times and release labels are recorded declarations, not authenticated acquisition evidence. If secure randomness is unavailable, the record labels its ID as a best-effort fallback. Local file paths and URL queries are omitted.