eduKate Learning Manual
Science | Edge Cases Science | Physical World
Understand → Observe → Explain → Test → Transfer → Go Deeper

The Delayed-Choice Quantum Eraser

Why the Future Does Not Rewrite the Past

Quick Answer

In the famous 2000 experiment by Kim and colleagues, an entangled photon pair is created. One photon—the signal—is detected at a movable detector D0. Its partner—the idler—travels through a longer optical path and is later detected in one of several ways.

Some idler detections preserve which-path information. Condition D0 events on those idler outcomes and no interference appears. Other idler detections erase the ability to distinguish the two alternatives. Condition D0 events on one erasing outcome and a fringe pattern appears; condition on the complementary erasing outcome and an oppositely phased antifringe pattern appears.

If you ignore the idler result and look at all D0 detections together, the fringe and antifringe subsets add to a pattern with no usable interference. Therefore the later idler result cannot transmit a message to the earlier detector and cannot reveal any rewritten past.

Physical Review Letters — Delayed “Choice” Quantum Eraser →

Physical Review A — Causality, Memory Erasing and Delayed-Choice Experiments →

What You Will Learn

• What which-path information means.
• Why interference depends on indistinguishable alternatives.
• How entangled photon pairs encode path correlations.
• Why the signal detector alone shows no fringes.
• What coincidence counting does.
• Why one conditioned subset shows fringes and another antifringes.
• Why the two subsets cancel in the unsorted data.
• What “erasing” information means operationally.
• Why delayed sorting does not change an earlier event.
• Why no faster-than-light or backward-in-time signalling is possible.
• How complementarity is expressed through correlations.
• Why postselection must not be confused with retrocausality.

Part 1 — Start With Ordinary Two-Path Interference

Suppose a quantum object can reach a detector through two indistinguishable alternatives A and B.

The probability amplitude is the sum

ψ = ψ_A + ψ_B.

The detection probability then contains an interference cross-term depending on the relative phase.

If the alternatives become perfectly distinguishable in principle, their coherence in the observed subsystem is lost and the interference term disappears.

Part 2 — Which-Path Information Can Be Stored in Another Photon

In a quantum eraser, the path alternative of the signal photon becomes correlated with a partner photon.

Schematically, the entangled state can look like

|Ψ⟩ ∝ |A⟩_s|A⟩_i + |B⟩_s|B⟩_i.

The idler states carry information that can distinguish which alternative is correlated with the signal.

That entanglement means the signal subsystem by itself is no longer described by one coherent pure superposition across A and B.

Part 3 — The Raw D0 Data Do Not Show the Famous Fringes

This is the most important correction to many popular explanations.

If you collect every signal-photon detection at D0 and ignore what happened to the idler photon, there is no ordinary interference pattern from which you could learn the later idler setting.

the “wave pattern” appears only after correlated events are selected into particular coincidence subsets.

Part 4 — Some Idler Detectors Preserve Which-Path Information

In the Kim experiment, certain idler detection routes identify which source region or path alternative the photon came from.

Condition the D0 records on those which-path-preserving idler outcomes and the corresponding subset does not show interference fringes.

The alternatives remain distinguishable in the correlation record.

Part 5 — Other Idler Detectors Measure a Superposition Basis

Beam splitters can combine the idler alternatives so that a given output does not uniquely reveal A or B.

Detecting the idler in such a superposition basis projects the correlated signal subset into a coherent conditional state.

One output corresponds roughly to a plus combination of the alternatives; the complementary output corresponds roughly to a minus combination.

Part 6 — Fringe and Antifringe

Conditioned on one erasing idler outcome, the D0 positions form a fringe pattern.

Conditioned on the other complementary erasing outcome, the D0 positions form an interference pattern shifted by approximately half a fringe period—an antifringe.

Where one subset has a maximum, the other tends to have a minimum.

Add the two subsets without labels and their interference terms cancel.

Part 7 — Coincidence Sorting Creates the Visible Pattern

To reveal the conditional patterns, experimenters compare timestamps from D0 and the idler detectors and identify which detections belong to the same entangled pair.

They then sort the already-recorded D0 positions into bins based on the idler outcome.

This is called coincidence counting or postselection.

The sorting reveals correlations that were invisible in the marginal D0 data.

Part 8 — Delayed Choice Does Not Mean Delayed Physical Change

The idler optical path can be made longer so the D0 signal photon is detected before the idler reaches its final beam splitter and detector.

That timing makes the experiment psychologically striking. But the D0 event is not altered afterward.

The later idler outcome supplies a label used to reveal which conditional correlation the earlier record belonged to.

later information changes our conditional description of the joint record; it does not edit the earlier detector file.

Part 9 — Why No Message Can Be Sent Into the Past

Suppose an experimenter at the idler apparatus could choose whether to preserve or erase which-path information.

An observer at D0 who does not receive the idler results sees the same marginal statistics regardless of that choice.

Only after ordinary classical communication brings the idler labels can the D0 data be sorted into the relevant conditional subsets.

No usable signal travels backward in time or faster than light.

Part 10 — “Erasing” Does Not Destroy a Historical Fact

The word eraser is dangerous if interpreted literally.

The experiment does not delete a classical notebook entry saying “the photon definitely went through path A.” Instead, the idler is measured in a basis that either retains or removes distinguishability of the path alternatives in the selected quantum correlations.

What changes is the measurement basis and therefore which conditional interference can be observed.

Part 11 — Complementarity Is About Available Distinguishability

Interference visibility and path distinguishability trade against one another.

When the idler states corresponding to A and B are distinguishable, which-path information is available and signal interference disappears.

Measure instead in a basis that combines those idler alternatives and path distinguishability is absent in that conditioned subset, allowing interference correlations to appear.

The experiment therefore tests complementarity through correlations rather than forcing each photon to carry a classical “wave” or “particle” label.

Part 12 — Entanglement Makes Marginal and Conditional Statistics Different

A central mathematical lesson is that joint probability distributions contain more information than either marginal distribution alone.

The D0 marginal can be featureless while P(D0 position | idler outcome) contains fringes.

This is not uniquely quantum as a statement about conditional probability, though the entangled quantum state and complementary interference structure are quantum.

Part 13 — The Original Experiment’s “Choice” Was Random

In the 2000 delayed-choice quantum eraser, the relevant idler routing was determined randomly by beam splitters rather than by a human consciously deciding the photon’s past.

This matters because popular accounts sometimes turn the setup into a story about observation, consciousness or future intention.

No such ingredient is required by the experiment.

Part 14 — Why the Experiment Still Matters

Rejecting retrocausal storytelling does not make the experiment ordinary.

The setup beautifully demonstrates that quantum interference belongs to probability amplitudes for indistinguishable alternatives and that entanglement redistributes coherence into correlations between subsystems.

It also teaches an unusually important scientific skill: the dramatic interpretation must be tested against the raw marginal data and the exact data-processing procedure.

Failed Model → Better Model

Naive model	Why it fails	Better model
The future measurement changes what the earlier photon did.	The earlier D0 record never changes.	Use joint correlations and conditional probabilities.
D0 shows fringes before the idler is measured.	The unsorted marginal has no usable interference.	Separate raw data from coincidence-conditioned subsets.
Erasing destroys a recorded path fact.	The idler is measured in a basis that removes path distinguishability in a subset.	Describe measurement basis and entanglement.
Quantum eraser allows messages to the past.	The D0 marginal is independent of the later idler outcome/setting.	Apply no-signalling and require classical communication for sorting.

How Do We Know?

• Record every D0 position before sorting.
• Record every idler detector and timestamp.
• Plot the unsorted D0 marginal.
• Build coincidence subsets D0|D1, D0|D2 and which-path subsets.
• Compare fringe and antifringe phase.
• Add the erasing subsets back together and verify cancellation.
• Change optical delay so the signal is detected before or after the idler.
• Check that D0 marginal statistics remain unchanged.
• Require classical communication of idler labels to reconstruct conditional fringes.

Observation vs Inference

• Observation: coincidence-selected signal subsets can show interference or no interference depending on the correlated idler measurement.
• Observation: complementary eraser outputs produce fringe and antifringe subsets.
• Measurement: the unsorted signal marginal contains no message revealing the later idler result.
• Inference: entanglement and measurement basis control conditional path distinguishability.
• Boundary: the data do not require retrocausal signalling or rewriting of an already recorded event.

Checkpoint Questions

1. What makes two alternatives interfere?
2. How can which-path information be stored in the idler?
3. What does D0 show before coincidence sorting?
4. What is coincidence counting?
5. Why do two erasing outputs give fringe and antifringe?
6. Why do those patterns disappear when added together?
7. What exactly is “erased”?
8. Why does a longer idler path not rewrite the D0 event?
9. Why can no message be sent into the past?
10. What evidence would expose a false retrocausal interpretation?

Answer Key

Open after attempting the questions

1. Coherent indistinguishable alternatives whose amplitudes retain relative phase.
2. Entanglement correlates path alternatives with distinguishable idler states.
3. A non-interfering marginal distribution.
4. Matching detections from the same pair and sorting signal records by idler outcome.
5. They correspond to complementary superposition phases of the idler alternatives.
6. The interference cross-terms have opposite phase and cancel.
7. Path distinguishability in the selected measurement basis, not an earlier historical record.
8. The event and its stored position are already fixed; only later correlation labels are added.
9. The signal marginal does not depend on the later choice, and coincidence sorting requires ordinary classical communication.
10. Show that the raw D0 distribution is unchanged and that only postselected subsets reveal fringes.

Primary Science Bridge

• waves can interfere when alternatives cannot be told apart;
• two particles can carry correlated information;
• sorting data can reveal patterns hidden in the full set;
• a later label does not change an earlier measurement;
• dramatic claims should be checked against raw data.

Secondary and JC Bridge

Core idea	Higher-resolution route
Interference	Probability amplitudes
Which-path information	Entanglement and reduced density matrices
Measurement	Choice of basis
Data analysis	Marginal vs conditional probability
Causality	No-signalling theorem
Complementarity	Visibility–distinguishability trade-off

Unfamiliar Transfer Challenge

Two laboratories share entangled photons. Lab A sees a featureless detector distribution. Lab B measures its photons in one of two bases. After the experiment, the laboratories compare timestamps and discover interference in selected subsets.

Could Lab A have known Lab B’s later basis before receiving the labels? Test the unsorted Lab A marginal. If it is basis-independent, no signalling or retrocausal message is available.

Deep Science Window — Reduced Density Matrix

For an entangled state, the signal subsystem is described by tracing over the idler. If the idler states correlated with paths A and B are orthogonal, the off-diagonal coherence terms disappear from the signal’s reduced density matrix. Conditioning on a particular idler superposition selects a different pure-state slice in which those coherence terms reappear.

Deep Science Window — No-Signalling Is Visible in the Algebra

Any complete measurement performed locally on the idler, when its outcome is ignored, leaves the remote signal density matrix unchanged. Different idler outcomes can define different conditional ensembles, but the weighted sum returns the same marginal state. This is why entanglement changes correlations without becoming a communication channel.

Evidence Boundaries

• Delayed choice ≠ future rewrites past.
• Quantum eraser ≠ destruction of a classical historical record.
• Conditional fringe ≠ fringe in the raw signal data.
• Entanglement correlation ≠ faster-than-light signalling.
• Postselection ≠ causal influence backward in time.
• Wave/particle complementarity ≠ photon consciously choosing an identity.

Manual Summary — KNOW → CONNECT → EXPLAIN → APPLY → CHECK

KNOW: entanglement, path information, conditional probability, coincidence counting, fringe/antifringe, no-signalling.

CONNECT: path marking to lost marginal coherence, idler-basis choice to conditioned subsets, and complementary subsets to cancellation in the raw data.

EXPLAIN: why delayed idler measurements reveal different correlations without changing an earlier detection.

APPLY: detect retrocausal overclaim in an unfamiliar entanglement experiment.

CHECK: inspect the marginal data before postselection and ask whether any message is available without classical comparison.

Research Sources and Further Reading

• Physical Review Letters — Delayed “Choice” Quantum Eraser
• Physical Review A — Causality, Memory Erasing and Delayed-Choice Experiments
• Physical Review Letters — Delayed-Choice Quantum Eraser With Thermal Light