Delayed-Choice Experiments

Delayed-choice experiments test whether the measurement arrangement can be chosen after a quantum system has entered an interferometer, while still producing the statistics appropriate to the final arrangement. They are associated with John Wheeler’s proposal to decide late whether an interferometer is open, revealing path information, or closed, revealing interference.

The responsible conclusion is not that the future sends a controllable signal to the past. The conclusion is that quantum predictions are assigned to the complete experimental arrangement, including the measurement actually performed. A delayed choice prevents a simple story in which the photon first decided to behave as a classical wave or a classical particle.

For background on amplitudes and complementarity, see Probability Amplitudes, Double-Slit Experiment, and Complementarity.

Wheeler’s Idea

Wheeler’s delayed-choice thought experiment sharpened a question already present in two-path interference. If an experiment is arranged to reveal which path a system took, interference disappears. If the paths are recombined coherently, interference can appear. Wheeler asked what happens if the choice between these arrangements is made only after the system has already entered the interferometer.

The point was not to build a time machine. The point was to undermine a naive classical narrative:

The photon must have chosen in advance whether it was a wave or a particle.

Quantum mechanics does not need that narrative. It assigns amplitudes through the apparatus and probabilities to the measurement outcomes actually registered. The delayed choice makes it harder to pretend that the system carried a fixed classical description independent of the later measurement context.

Interferometer Setup

A Mach–Zehnder interferometer is the cleanest model. The first beamsplitter creates a coherent superposition of two path modes:

$$\lvert\psi\rangle = \frac{1}{\sqrt2} \left( \lvert u\rangle + e^{i\phi}\lvert \ell\rangle \right),$$

where $\lvert u\rangle$ and $\lvert \ell\rangle$ label upper and lower paths, and $\phi$ is a relative phase.

If the second beamsplitter is absent, the two paths are measured separately. This is the open interferometer. The two detectors answer a path question. In the ideal balanced case,

$$P(u)=P(\ell)=\frac12,$$

independent of $\phi$.

If the second beamsplitter is present, the paths are recombined before detection. This is the closed interferometer. The detectors answer an interference question. With a standard beamsplitter convention, the output probabilities have the form

$$P_0 = \frac12(1+\cos\phi), \qquad P_1 = \frac12(1-\cos\phi).$$

The same state before the final measurement can therefore lead to different observable statistics depending on which final measurement basis is implemented.

Choice of Measurement Arrangement

In a delayed-choice experiment, the decision to insert, remove, or effectively control the second beamsplitter is made after the system has passed the first beamsplitter. Modern realizations use fast electro-optic devices, random-number generation, or quantum-controlled beamsplitters.

The late choice is important because it blocks a simple classical explanation in which the system knew the entire fixed apparatus from the start and selected a matching behavior. If the final arrangement is chosen sufficiently late, then no ordinary signal from that choice can reach the system at the earlier beamsplitter event.

But the delayed choice does not change the formal rule. Quantum mechanics still computes probabilities for the final measurement actually performed. The open interferometer measures path-like information. The closed interferometer measures a phase-sensitive output basis.

This is a measurement-basis issue, not a license to combine incompatible stories. One cannot infer a definite earlier path from an open arrangement and also infer phase interference from a closed arrangement in the same run.

What Is and Is Not Retrocausal

Delayed-choice experiments are often described as if the future changes the past. That phrasing is usually misleading.

What the experiment shows:

• late choices of measurement arrangement produce the statistics predicted for that arrangement;
• path information and interference visibility are tied to incompatible measurement contexts;
• a classical wave-or-particle story assigned before the measurement arrangement is fixed is inadequate.

What it does not show:

• controllable signaling to the past;
• a detector choice rewriting a recorded event;
• a photon consciously deciding how to behave;
• simultaneous access to full which-path information and full interference in one run.

The careful language is that the experimental arrangement determines which observable is measured. If the arrangement is chosen late, the quantum state still supplies probabilities for that late-chosen measurement. No additional retrocausal mechanism is required by the standard formalism.

Modern Realizations

Early delayed-choice experiments used optical interferometers and fast switching. Later single-photon experiments closed more of the conceptual gap between Wheeler’s proposal and laboratory implementation. In particular, experiments with single photons and fast random choices showed that interference or path-like statistics appear according to the final setting even when that setting is chosen after the photon has entered the interferometer.

Quantum delayed-choice experiments go one step further by putting the presence or absence of the second beamsplitter into a quantum superposition or quantum-controlled operation. These experiments test a controlled interpolation between open and closed behavior. They are best understood as ordinary quantum circuits with a control system, not as evidence that a photon is literally both a classical wave and a classical particle.

Delayed-choice entanglement swapping and delayed-choice quantum eraser experiments extend the same theme to entangled systems and correlation measurements. They require additional care because postselection and coincidence sorting can make the story sound more dramatic than the operational facts.

Modern Interpretation

The modern interpretation is compact:

Quantum mechanics predicts probabilities for outcomes of the measurement actually performed.
Delayed choice removes a naive pre-existing wave-or-particle narrative.
It does not add controllable retrocausality to the formalism.

The experiment is therefore a lesson in measurement context. In an open interferometer, the detector basis correlates with paths. In a closed interferometer, the detector basis is sensitive to phase. The same preparation can be interrogated in incompatible ways, but one run realizes one final arrangement.

This is why delayed-choice experiments are historically tied to complementarity. They dramatize the fact that quantum phenomena cannot always be described by one fixed classical picture independent of the measurement conditions.

Common Misconceptions

• Delayed choice does not prove that the future sends messages to the past.
• The photon does not need to decide in advance to be a classical wave or a classical particle.
• Open and closed interferometers measure different observables.
• The disappearance of interference in a path measurement is not caused by human knowledge; it follows from distinguishable alternatives or path records.
• Quantum delayed-choice experiments are quantum-controlled measurement circuits, not literal half-wave half-particle photographs.
• Delayed-choice quantum eraser experiments require postselection; they do not allow faster-than-light communication.

Cross-Links

• Foundations Experiments and Quantum Reality
• What These Experiments Do and Do Not Prove
• Quantum Eraser Experiments
• Complementarity
• Double-Slit Experiment
• Interference With Matter
• Probability Amplitudes
• Born Rule
• Measurement in a Chosen Basis
• Projective Measurement
• State Update Rule
• What Measurement Formalism Does Not Settle
• Common Historical Misconceptions

References

• J. A. Wheeler, “The ‘Past’ and the ‘Delayed-Choice’ Double-Slit Experiment,” in A. R. Marlow, ed., Mathematical Foundations of Quantum Theory, Academic Press, 1978.
• J. A. Wheeler and W. H. Zurek, eds., Quantum Theory and Measurement, Princeton University Press, 1983.
• T. Hellmuth, H. Walther, A. Zajonc, and W. Schleich, “Delayed-choice experiments in quantum interference,” Physical Review A 35, 2532-2541, 1987, DOI: 10.1103/PhysRevA.35.2532.
• V. Jacques et al., “Experimental Realization of Wheeler’s Delayed-Choice Gedanken Experiment,” Science 315, 966-968, 2007, DOI: 10.1126/science.1136303.
• R. Ionicioiu and D. R. Terno, “Proposal for a Quantum Delayed-Choice Experiment,” Physical Review Letters 107, 230406, 2011, DOI: 10.1103/PhysRevLett.107.230406.
• A. Peruzzo, P. Shadbolt, N. Brunner, S. Popescu, and J. L. O’Brien, “A Quantum Delayed-Choice Experiment,” Science 338, 634-637, 2012, DOI: 10.1126/science.1226719.
• X.-S. Ma, J. Kofler, and A. Zeilinger, “Delayed-choice gedanken experiments and their realizations,” Reviews of Modern Physics 88, 015005, 2016, DOI: 10.1103/RevModPhys.88.015005.

Exercises

1. In the closed interferometer model on this page, what are $P_0$ and $P_1$ when $\phi=0$?

Solution

Use

$$P_0=\frac12(1+\cos\phi), \qquad P_1=\frac12(1-\cos\phi).$$

For $\phi=0$, $\cos\phi=1$, so

$$P_0=1, \qquad P_1=0.$$

2. Why does the open interferometer not show phase-dependent output probabilities?

Solution

With the second beamsplitter absent, the two paths are detected separately. The detector basis distinguishes the paths rather than recombining their amplitudes. Without recombination, the relative phase between paths does not appear as an interference term in the output probabilities.

3. What does the delayed choice rule out in a naive classical story?

Solution

It rules out the simple story that the photon decided at the first beamsplitter whether it was a classical wave or a classical particle according to a final apparatus arrangement already fixed in advance. The late choice shows that the statistics correspond to the measurement arrangement actually implemented, not to an earlier classical label.

4. Why does delayed choice not allow signaling to the past?

Solution

The experimenter can choose which measurement arrangement is implemented, but the individual outcome remains probabilistic and no already recorded event is controllably changed. The correlations and probabilities are explained by the quantum state and the final measurement context; they do not provide a channel for sending usable messages into the past.