A mathematical reconstruction and an empirical distinction
Selecting a complex structure from real dynamical data can clarify the dependencies of a physical theory. If the reconstructed sector makes the same operational predictions as ordinary quantum mechanics, those predictions do not by themselves distinguish the new account. An independent physical restriction or deviation is needed for an empirical comparison. [P03] [B16]
This laboratory teaches that distinction with a standard two-output sinusoidal fringe. It fits two coefficients from two supplied probabilities and predicts another angle. Every value is a synthetic demonstration input. No experimental dataset is loaded or fitted.
The reconstruction papers give conditional mathematical results; the programme's next physical steps concern the source, its admissible dynamics, and how its records map to measured quantities. A useful test should identify those extra commitments and reserve observations that were not used to set them. [P01] [A02]
Fit two points; reserve a third
The default calibration probabilities are P(0°)=0.8 and P(90°)=0.6. They determine C=0.6 and S=0.2. At 180°, the predicted probability is 0.2. Change the held-out value to see the residual, or select a different held-out angle.
Explore · Standard quantum baseline
Change the calibration and comparison inputs to inspect the prediction.
The interactive values require JavaScript. The diagram and worked equations provide the same mathematical construction.
Try setting both calibration probabilities to 0.9. The fitted visibility exceeds one, so no admissible member of this normalized baseline fits both points exactly. The lab reports the inconsistency and keeps the supplied inputs. It does not clip them into a seemingly successful fit.
A constrained sinusoidal baseline
For V>0, define δ=atan2(S,C). Then the same model is P+(φ)=½[1+V cos(φ−δ)]. When V=0 the curve is flat and δ is undetermined; assigning a preferred offset in that case would add information not present in the fit.
The inequality V≤1 guarantees that every angle gives a probability between zero and one. Two individually allowed calibration probabilities can nevertheless violate this joint condition. Such a failure is informative: it concerns the assumed normalized sinusoidal family, not the existence of each individual probability.
Real measured frequencies have uncertainty, so an exact mismatch is not a significance test. A practical analysis would fit counts with a noise model, account for efficiencies and backgrounds, and propagate calibration uncertainties. The toy fit deliberately omits those complications to make its logical dependency clear.
Specify the additional physics
| Required ingredient | What to state |
|---|---|
| Source mechanism | Which microscopic process prepares and compares the phase-carrying records? |
| Baseline | Which ordinary quantum model and apparatus response are being compared? |
| New restriction | Which independently motivated equation or parameter constraint changes the prediction? |
| Calibration | Which observations fix nuisance parameters, clocks, and phase references? |
| Held-out observations | Which angles, preparations, or time scales remain unused during fitting? |
| Decision rule | How are uncertainty, alternative explanations, and statistical comparison handled? |
The export records the model version, synthetic inputs, validity check, coefficients, chosen held-out angle, prediction, and residual. A CSV download provides the admissible baseline over a full phase cycle. Invalid fits retain a JSON diagnostic; their CSV export is disabled because no physical baseline curve has been accepted.
A residual can motivate checking the model or apparatus. It is not evidence for SMM/DTT merely because it is nonzero. The next step is a specified alternative with consequences that cannot be absorbed by ordinary phase shifts, loss of coherence, or detector calibration.
Use Sources & model for the exact conventions and published reading. The research programme and open problems organize the wider questions.

