Retroactive-Collapse Signaling with Double-Slit / Quantum-Eraser Experiments

Nick Arnot 10-14-2025

TLDR

I propose that entangled measurement choices made “in the future” can retroactively determine (“collapse”) the present’s interference vs. which-path outcomes, and that by toggling this you could write a binary message (interference = 1, no-interference = 0) into the past.

Core hypothesis

• Premise: In delayed-choice quantum eraser (DCQE) setups, the later choice to erase or preserve which-path information appears to determine whether earlier detection records show interference.
• Claim: If that “retroactive collapse” can be read locally without any coincidence sorting or classical communication, then a sender (in the future) could transmit bits backwards in time by choosing “erase” (1) or “don’t erase” (0).
• Desired outcome: A binary chain encoded across many trials, recoverable from the past-side detector stream alone.

Background & what standard quantum theory says

Double-slit & which-path

• Interference requires indistinguishability of paths. Marking the path (even in principle) kills interference.

Entanglement-assisted “quantum eraser” (delayed choice)

• A signal photon hits a screen (near “now”); its entangled idler is routed through optics that either erase, or reveal which-path later.
• Key textbook point: Interference fringes only appear after you sort the signal hits into subsets using the idler outcomes (coincidence/conditional histograms).
  • The unsorted signal-only data shows no fringes (it’s a smooth clump).
  • This is why no signaling (and no usable retrocausality) occurs in standard QM.

The blockers that must be addressed

1. No-signaling theorem: Local outcome statistics (marginals) at the “past” cannot depend on remote “future” settings.
2. Post-selection trap: “Retro” patterns arise only after conditioning on the later record; without it, the pattern washes out.
3. Information causality / monogamy: Prevents packaging controllable information in entanglement alone.

Bottom line: Standard quantum mechanics forbids readable backward-in-time messaging, because the “message” shows up only after joint sorting with the future records.

Theoretical pathways that could permit your idea

If you want real retroactive signaling, you must relax at least one standard assumption:

• (A) Retrocausal hidden-variable models (e.g., two-state vector formalism, transactional ideas): permit influences that run boundary-to-boundary (past/future). You must still show how local marginals become controllable.
• (B) Superdeterminism / measurement dependence: the future setting and the past hidden variables are correlated “all-at-once,” bypassing measurement independence. This can, in principle, maintain locality while changing statistics. A recent line of work argues collapse could emerge locally if future measurement settings co-fit the overall evolution; if you lean this way, be explicit about how measurement settings co-determine earlier micro-states and why that changes the past marginals, not just the joint distribution.
• 2510.11037v1
• (C) Nonlinear / non-CPTP dynamics: GRW-type objective collapse, Deutsch-style CTCs, or exotic gravitational couplings. Any nonlinearity can open signaling channels—but then you must quantify how and avoid contradictions.

To proceed on any experimentation, we need to pick one path and own its assumptions.

Minimal experimental spec for a binary “retro-message”

Actors & timeline

• Alice (Past): Holds the signal arm; records only her detector coordinates/timestamps. No access to Bob’s data or choices.
• Bob (Future): Holds the idler arm; chooses per trial: Erase (E) vs. Mark (M). His choices are space-like or time-delayed relative to Alice’s detections.

Your required success criterion (strong form)

• The distribution of Alice’s hits alone changes between Bob’s E and M choices:
  • E (erase) ⇒ fringe-bearing distribution (or any statistically distinct distribution A_E)
  • M (mark) ⇒ no-fringe distribution (or distinct distribution A_M)
• Difference must be detectable by Alice without any classical information from Bob. This is what encodes the bit.

If you can’t meet this strong criterion, your scheme cannot transmit information to the past, even if pretty plots appear after coincidence sorting.

Practical checklist

• Space-time ordering: Enforce Bob’s choice strictly after Alice’s detection events (or at least space-like separated) to preclude ordinary causal influence.
• Choice mechanism: Use a fast, random, setting-independent switch (e.g., QRNG) for E vs. M to avoid device-drift artifacts.
• No-leakage guarantees: Isolate lab timing, avoid clock-sync backdoors, coincidence window “creep,” detector-efficiency bias, or drift that could fake a difference.
• Pre-registration: Fix the statistics, binning, and tests before data taking.

Predicted outcomes table

AssumptionsAlice-only data (marginal)Can Bob send a bit to Alice’s past?Standard QM (unitary, CPTP, no-signaling)Identical across Bob’s E/M settings (no readable fringes)NoRetrocausal w/ unchanged marginalsIdentical marginals; differences only after sortingNoSuperdeterministic w/ measurement dependence that alters marginalsDifferent Alice-only distributions for E vs. MYes (in principle)Nonlinear dynamics enabling signalingDifferent Alice-only distributionsYes (but check consistency)

Your theory lands in the last two rows. You must specify the dynamical law that makes Alice’s marginal depend on Bob’s later setting.

How to formalize the model (sketch)

1. State the ontic variables λ\lambdaλ and where/when they’re set.
2. Specify dependence: ρ(λ∣Bob setting)≠ρ(λ)\rho(\lambda \mid \text{Bob setting}) \neq \rho(\lambda)ρ(λ∣Bob setting)=ρ(λ) (measurement independence violation) or an explicit retrocausal boundary condition linking future settings to present ontic states.
3. Give the map from (λ,Alice apparatus)→p(Alice outcome)(\lambda, \text{Alice apparatus}) \to p(\text{Alice outcome})(λ,Alice apparatus)→p(Alice outcome).
4. Derive distinct Alice-marginals PA(E)(x)P_A^{(E)}(x)PA(E)​(x) vs. PA(M)(x)P_A^{(M)}(x)PA(M)​(x).
5. Quantify effect size (e.g., total variation distance TVD(PA(E),PA(M))\mathrm{TVD}(P_A^{(E)}, P_A^{(M)})TVD(PA(E)​,PA(M)​)) to predict power curves for realistic counts.

Statistical plan (to claim a bit was sent)

• Primary endpoint: Reject H0 ⁣:PA(E)=PA(M)H_0\!: P_A^{(E)} = P_A^{(M)}H0​:PA(E)​=PA(M)​ with pre-registered α\alphaα.
• Tests: 2D binned χ² across the screen; or unbinned two-sample tests (e.g., energy distance).
• Effect size target: e.g., detect TVD≥0.02\mathrm{TVD} \ge 0.02TVD≥0.02 at 5σ with NNN counts—compute NNN in advance.
• Sanity checks: Shuffle Bob’s labels (placebo), vary coincidence windows (shouldn’t matter since you don’t use them), swap optic paths, change QRNG seed.

Common pitfalls to preempt

• Hidden post-selection: Any conditioning that smuggles in Bob’s info breaks the “Alice-only” condition.
• Coincidence window bias: Moving windows can fake fringes—don’t use coincidences at all for the primary endpoint.
• Detector nonuniformity / drift: Mimics pattern changes; randomize and calibrate frequently.
• Subtle classical cross-talk: Timing, EM leakage, or software pipelines can re-introduce Bob’s choice.

Consistency questions that should be addressed

• Causal paradoxes: If bits go to the past, how do you avoid “grandfather”-style loops? (Fixed-point / consistency constraints?)
• Thermodynamics / arrow of time: Where is entropy exported when a future choice reduces uncertainty about the past?
• Relativity: Ensure frame-independent description (no preferred foliation unless explicitly posited).
• Security: If retro-messages are possible, what prevents ubiquitous signaling via ambient entanglement?

Optional gravitational angle

A line of thought (superdeterministic + gravity-linked collapse) argues that outcomes and settings co-fit globally so that macroscopic outcomes are locally consistent; if you adopt this, specify how the future measurement setting participates in fixing earlier micro-states enough to alter local marginals (the crucial step your messaging requires).

2510.11037v1

Executive summary (for non-specialists)

• What’s new? You aim to turn the appearance of retrocausality in DCQE into an actual communication channel from future to past.
• What must change? You must break at least one standard assumption (no-signaling, measurement independence, or linearity).
• What would convince skeptics? A clean change in the past-side, unconditioned data driven solely by a later choice—with airtight controls.
• Why it matters: Demonstrating such an effect would rewrite the causal structure used in physics and information theory.

‍

Feel free to hit me up about this quantum@arnot.io