Source
Catani, Leifer, Schmid & Spekkens (2021), LIT-019, read in full in NOTE-018, which re-derived the statistics.
What was actually shown
Scope (“TRAP” phenomenology, §2.1). Only the predictions that the standard arguments for three interpretational claims use: balanced (50-50) beamsplitters, phases ϕ ∈ {0, π}, a single excitation, the photon entering at port L. Unbalanced and continuous-phase predictions are excluded (§5.2.4).
The model (“toy field theory”). Each mode j has a physical state (N_j, Φ_j) ∈ Z₂ × Z₂. An agent may know at most one of N, Φ, N⊕Φ and is maximally ignorant otherwise (p. 12). Dynamics are local and deterministic. The beamsplitter is the Swap Rule: swap N_L and Φ_L⊕Φ_R, keep Φ_R and N_L⊕N_R (Eqs. 13, 31). A π phase shift is Φ ↦ Φ⊕1 (Eq. 34). A nondestructive measurement of N conditions on the outcome and randomizes that mode’s Φ (p. 14). NOTE-018 confirms that this is LIT-007’s quadrature epistricted theory of bits, with modes as systems and every map symplectic affine over Z₂.
Results.
- Phase shifter: ϕ = 0 gives output L with probability 1, ϕ = π gives R (Eqs. 26, 39).
- Which-way detector on R: it fires with probability ½, and either way it randomizes Φ_R, so the outputs become ½/½ (Eqs. 46–53).
- Bomb tester: a working bomb is certified without exploding with probability ¼, carried by a local phase flip of the unoccupied mode (§4.1).
- Delayed choice: N and the relative phase both have values throughout; the late choice fixes which one can be inferred (§4.2).
- Eraser: measuring one ancilla variable sorts runs by N_R (no interference in either subset), the other by the back-action on Φ_R (full interference in each); the choice changes nothing about modes L and R (App. A.3, Eq. 57).
- The account survives destructive detectors and a removed mirror (App. C.3–C.4), and recasts as a cellular automaton with kinematical locality (App. C.6).
The mechanism. The vacuum in the unused arm is a state of incomplete knowledge, not of reality. An empty mode still carries a phase bit, which a detector can flip without the arm being occupied. The flip propagates locally to the second beamsplitter, where the relative phase decides the output port. Dependence on later choices is inference, not influence.
What could have come out otherwise. Some prediction in the fragment could have resisted every local classical model. None did. The locality depends on reading the vacuum epistemically: the paper notes that every ψ-ontic model needs nonlocality for the eraser, because the eraser is an EPR-type scenario (App. B, citing Harrigan–Spekkens).
What this does not say
- That interference as such is classical. The paper’s title and §5.1.1 speak of “interference phenomena” and the “only mystery”. The body covers the balanced, ϕ ∈ {0, π} fragment. The paper itself concedes that unbalanced beamsplitters and the functional form of the wave–particle duality relation admit no generalized-noncontextual model (its ref. [24]), as do Hardy’s interferometers.
- That the model is generalized-noncontextual. That it is, and so classical by the authors’ preferred criterion, is asserted “for the same reason that other epistemically restricted … theories do” (p. 28). No argument shows that operationally equivalent quantum procedures in the fragment receive the same toy distribution (NOTE-018, C5). The conclusion that the three interpretational claims are not forced does not need it.
- That the fragment’s host theory is classical. The two-mode “qubit field theory” is the two-qubit stabilizer subtheory, which already contains the Peres–Mermin contextuality proof. The model is classical for the fragment, not for the theory the fragment is described in.
- That it works for fermions. Aharonov’s objection (fn. 42) stands unanswered.
Connections
- The toy bit theory whose kinematics this is, and how it differs from the stabilizer theory, is THEORY-474.
- No-broadcasting is another phenomenon LIT-007 lists as reproduced by epistricted theories; within quantum theory it is THEORY-695.
- Hance et al. (LIT-090) doubt in an appendix that nonnegative ontological models can reproduce the bomb tester. This model does, for the fragment (NOTE-072, C8); the account of what ψ-ontology theorems exclude is THEORY-586.
- THEORY-010 also finds an epistemically restricted class reproducing what was offered as a mark of quantum contextuality. The generalized and Kochen–Specker notions of noncontextuality are THEORY-011 and THEORY-015.