Wheeler's Delayed-Choice Experiment

The basic idea, in plain language

They sent single photons — little packets of light — into a device called an interferometer, which splits each photon into two possible paths and then can recombine them. In one configuration, the device lets you see interference patterns (wave-like behavior); in another configuration, it lets you find out which path the photon took (particle-like behavior). The twist is that they decide very late — after the photon is already inside — whether the device will measure it as a wave or as a particle.

How it works in everyday language

• A photon enters the setup and meets the first beamsplitter, which sends it "both ways at once" in the quantum sense — there is no definite answer yet about which path it took.
• The two paths are separated and then guided to a second region where you might recombine them or you might just send each path directly to its own detector.
• A fast, quantum-based random number generator decides, for each photon, whether the second beamsplitter is "in place" (paths recombined) or "not in place" (paths kept separate). This decision happens only after the photon has already gone past the first splitter and is traveling inside the interferometer, and the timing is arranged so the photon cannot be "told" in advance by any normal signal.

Two possible situations result:

• Closed configuration (wave mode). The second beamsplitter is in place, and the two paths are recombined. Over many photons, the counts in the detectors form an interference pattern — exactly what you expect if the photon behaved like a wave that went through both paths and interfered with itself. In this mode, you cannot tell which path the photon took; the setup simply does not give you that information.
• Open configuration (particle mode). The second beamsplitter is effectively removed, and each path goes straight to its own detector. Now each detection tells you, with very high reliability, which path the photon traveled. There is no interference pattern: the counts are just evenly spread, as you would expect from a "particle" choosing one path or the other.

In the actual experiment, they switch between these two modes randomly, one photon at a time, with the choice made only after the photon is already on its way inside the device.

What is so puzzling about the delayed choice?

In a classical, story-like picture, you might want to say: "the photon must have decided at the first beamsplitter whether it is going to act like a wave (go both ways) or a particle (choose one path)." But the experiment shows that you only see:

• wave-like behavior when you later choose to recombine the paths, and
• particle-like behavior when you later choose not to recombine them.

And this choice is made so late that the photon could not have known which experiment you would run when it entered. That undermines the idea that the photon had already "settled" into being either a wave or a particle before your last-minute decision. The behavior you see depends on how you eventually decide to measure it, not on some earlier hidden choice the photon made at the entrance.

What the experiment really says — and doesn't

• It does not show that we can change the past or send signals backward in time.
• It does show that it is wrong to imagine the photon as having a definite, classical-style "story" (like a tiny bullet picking one path) that exists independent of what you ultimately measure.
• The safe takeaway, in plain terms: until the measurement setup is finalized, the photon's path is not a matter of simple fact in the way we are used to in everyday life.

Quantum mechanics handles all this cleanly: the photon is described as being in a kind of superposed "both paths at once" state, and the measurement choice (wave-type vs particle-type apparatus) determines which aspects of that state become actual, concrete outcomes.

Jacques and colleagues implemented Wheeler's thought experiment with real single photons, fast electronics, and genuine random choices, closing obvious loopholes like "maybe the photon somehow learned the setting in advance." The results strongly support the standard quantum view: what is real and definite at the end depends in a deep way on how you choose to look — and you can choose this even after the system has already entered the apparatus.

Setup and aim

Jacques and colleagues realize Wheeler's thought experiment by sending true single photons through a Mach–Zehnder interferometer whose output beamsplitter can be randomly inserted (the "closed" interferometer) or removed (the "open" interferometer) after the photon has entered the apparatus. The aim is to test whether a photon's behavior — wave-like interference vs particle-like which-path — can be thought of as fixed at entrance, or whether it is genuinely determined by the measurement choice made in a delayed, space-like separated manner.

Key experimental ingredients

• Single-photon source. Clock-triggered single-photon pulses emitted by an individual nitrogen–vacancy (NV) color center in a diamond nanocrystal, ensuring well-defined single-photon inputs (g²(0) well below 0.5).
• Mach–Zehnder interferometer. The first 50/50 beamsplitter creates two spatially separated paths with high interferometric visibility when the paths are later recombined.
• Variable output beamsplitter. The second beamsplitter is effectively variable with reflection coefficient R, allowing continuous tuning between fully open (no interference measurement) and fully closed (maximal interference measurement).
• Quantum random-number generator (QRNG). A QRNG chooses in real time whether the output beamsplitter is present (closed configuration) or absent (open configuration), independently for each photon.
• Relativistic separation. The timing and geometry are arranged so that the QRNG choice event is space-like separated from the event of the photon entering the interferometer — ruling out any subluminal "photon reads the setting" causal story.

Main empirical results

• In the closed configuration, detection statistics show high-visibility interference fringes (about 94% visibility) — the photon behaved as a wave during its passage through the interferometer.
• In the open configuration, the two detectors correspond to the two distinct paths, and the experiment can determine which path the photon took with an error probability below 1% — nearly perfect which-way information, no interference.
• When the choice between open and closed is made randomly on a shot-by-shot basis, each photon's detection statistics match exactly the prediction for the configuration chosen after the photon entered. There is no evidence of any intermediate "compromise" behavior.

Closed configuration: ~94% interference visibility. Open configuration: which-path determined with <1% error. The configuration chosen after the photon entered is the configuration the photon's behavior matches.

Conceptual implications

The experiment strongly supports the standard complementarity view: what is observed depends on the measurement context, and there is no need — nor room — to attribute a definite wave or particle trajectory to the photon independent of that context.

Because the choice is made in a space-like separated manner, retrocausal or "photon knew in advance" classical pictures become untenable unless one posits highly nonlocal or conspiratorial hidden variables. The data are fully consistent with ordinary quantum mechanics interpreted with care about when measurement context becomes physically meaningful.

In Jacques and colleagues' own framing, their results show that assigning a naïve story like "the photon decides at the first beamsplitter whether to behave as a wave or a particle" is incompatible with the observed dependence on the delayed choice. The "decision" about interference vs which-path information is encoded in the quantum state and only manifests when the measurement context is fixed — even if that context is chosen after the photon has already entered the interferometer.

Why this matters for the trilogy

The Wheeler delayed-choice result is one of the empirical anchors under Limen's observer chapters. The trilogy's claim that the universe is read in both directions of time — that the present is constructed at the interface between forward-propagating and backward-propagating constraints — is not the trilogy's invention. It is the picture that quantum mechanics under the two-state-vector formalism gives natively, and that the Jacques experiment renders empirically vivid.

The cleanest companion paper is Manning et al. (2015), which extended the result to single helium atoms — massive particles, not photons. The same delayed-choice dependence holds. The phenomenon is not a peculiarity of light. It is a feature of how quantum systems behave when their measurement context is fixed retroactively.

Read together with the two-state vector formalism explainer, this experiment is the empirical face of the same insight Aharonov and Vaidman gave the formal language for: at any intermediate moment, a quantum system is described by constraints from both its preparation and its eventual measurement. Wheeler's experimental realization makes the future-constraint part operational. The future genuinely participates in the constitution of the past, in the precise quantum-mechanical sense that the measurement context determines what the past looked like — whether it consisted of a wave passing through both paths or a particle passing through one.

This is the structural feature that lets the trilogy's symmetric 300-millisecond gaps — Libet's readiness potential on the past side, Lucía Reyes's cymatic pre-event window on the future side — sit naturally inside contemporary physics rather than as fictional flourishes. The field cosmology of Limen is one in which the rendering of "now" is jointly constrained by past preparation and future measurement, and Wheeler's delayed-choice result is the cleanest experimental window onto that structure.

For the original 2007 paper, see the PubMed entry. For the formal language that organizes this kind of result, see the two-state vector formalism explainer. For the single-atom extension (Manning et al., 2015), see the bibliography entry on the Reading page. For the broader picture, see What the Evidence Shows So Far.