Experiment #8 · Scientific experiment

The Delayed-Choice Quantum Eraser

Retrocausality without retrocausation

John A. Wheeler (proposal); Kim, Yu, Kulik, Shih, Scully (experiment) · 1978 / 1999 · Quantum foundations

First published: J. A. Wheeler, "The 'Past' and the 'Delayed-Choice' Double-Slit Experiment", in *Mathematical Foundations of Quantum Theory* (1978). Kim et al., "A Delayed Choice Quantum Eraser", *Phys. Rev. Lett.* 84 (2000): 1–5.

Choosing whether to read which-path information *after* the photon has been detected still determines whether interference appears.

Wheeler's proposal asked what would happen if the decision whether to record which-path information were made after the particle had passed the slits. Kim et al. (1999) realised it experimentally using entangled photon pairs: a "signal" photon hits its detector before the experimenter decides whether to measure or erase the which-path information carried by its "idler" partner. The signal photon's detection pattern depends on what is later done to the idler — but only after the idler detection is correlated back. The result is paradoxical-looking but doesn't allow signalling, and is the crispest empirical pressure-point on which-direction time, measurement, and information actually flow.

Formulation

A pumped non-linear crystal produces entangled signal–idler pairs. Signal photons go to detector D0 (with no which-path information). Idler photons are routed through a beam splitter network where their which-path information is either preserved (detectors D3, D4) or erased (D1, D2). D0 hits are always recorded *before* the corresponding idler. Conditioning on idler outcomes after the fact reveals interference for erased-path subsamples and no interference for preserved-path subsamples.

Compare this experiment

Dimensions Engaged

⧖

Time

The most explicit pressure point on Time · Direction: either the past is in some sense settled only when its records are processed, or "delayed choice" is a story we tell about correlations that were already fixed at pair creation.

◎

Observer

Sharpens Observer · Metaphysical Agency: does the choice to erase or preserve which-path information *do* something, or only *sort* an already-determined ensemble?

⧉

Information

Targets Information · Ontological Status. The eraser suggests it is the *availability* of which-path information, not its actual recording, that destroys interference — pushing information towards a substantival role.

Responses — How Schools Engage

Affirms / takes the bait 2

Quantum Realism

Treat the joint wave function as the real entity: the pair is one quantum object and the "later" measurement is not later in any meaningful causal sense — it is a measurement on a single non-separable state.

Eternalism

Block-universe eternalism takes the eraser as natural: all events are equally real in the four-dimensional block, and "delayed choice" is just our slicing into it. There is nothing to *do* retroactively, because there was never an open past to begin with.

Denies / rejects the premise 2

Presentism

A challenge: if only the present exists, the signal detection at D0 was a fully determinate event when it happened, and the later eraser cannot change the pattern it was part of. Presentists typically retreat to a statistical / no-real-retrocausation reading, which the experiment puts under pressure.

Logical Positivism

No signalling is possible: the experiment's "retrocausal" appearance vanishes once you ask only about empirically accessible distributions. The verifiable content is exhausted by the Born rule on the joint state.

Reframes the question 2

Multiverse Theory

Everettian readings dissolve the paradox: nothing is "set" at D0 until decoherence selects a branch. There is no retrocausation, only branching correlations; the sorting after the fact picks out which branch the observer is on.

Simulation Theory

Compatible with a simulation in which detector records are written only when correlated readings demand it — a "lazy evaluation" reading of measurement. Not forced by the data, but harmonises with it more easily than substantival readings of the past.