A photon enters a chain of N very weak beam splitters that would rotate its state from the "safe" rail to the "test" rail by a total angle of 90° if nothing disturbs it — each stage contributes a small rotation θ = π/(2N).
θ = π / (2N)
No bomb / dud bomb: photon always ends up fully rotated
into the TEST rail → normal detector.
Live bomb present: each stage is a which-path measurement.
P(no explosion after N stages) = cos^(2N)(θ)
P(explode) = 1 − cos^(2N)(θ)
A surviving photon is projected back to the SAFE rail at
every stage (quantum Zeno effect), so if it survives all N
stages it is found at the IFM detector — proof a bomb sits
in the device, without a single photon ever touching it.
- Bomb type — a live bomb absorbs any photon that reaches it (which-path information ⇒ collapse); a dud is transparent and reveals nothing, so it behaves exactly like "no bomb".
- N stages — more, weaker stages exploit the quantum Zeno effect: the explosion probability falls toward 0% while the interaction-free success probability climbs toward 100% as N → ∞.
- Fire one photon — plays a single trial through the ring, stage by stage, with the real outcome probabilities above.
- Run 300 trials — samples many trials instantly to check the empirical rates against the analytic prediction.
This is the Elitzur–Vaidman bomb tester (1993), extended to high efficiency by Kwiat, Weinfurter, Herzog, Zeilinger & Kasevich (1995) — the first scheme to show that quantum mechanics lets you learn the *existence* of an object through counterfactual, interaction-free measurement.