The Observer Notebook · 2026-10-03

Quantum Zeno deep dive

If you keep checking whether a qubit has flipped, can it still flip?

Known, reproduced

A qubit is the simplest quantum object: a thing with two states, 0 and 1, that can sit partway between them. Left alone under a steady push, it swings from 0 to 1. Every check asks "still 0?", and a yes puts it right back at 0. Check ten times during one flip and it stays at 0 about 78% of the time. Check often enough and it barely moves at all.

78%unflipped, 10 checks per flipCLOSED FORM
1.5%flipped, steady leak γ = 100ΩEXACT SIM
4.2–8.4×faster decay, anti-ZenoEXACT SIM
97.6%flipped by dragging, 100 stepsEXACT SIM
01

Watch it, live

The panel runs the real rule. Between checks the qubit turns at a steady rate. At each check it stays at 0 with probability cos²(angle/2), otherwise it lands on 1 and the run is over.

Live · one qubit, N checks per flip DEMO SCALE · THE REAL RUN USED EXACT ODDS + 20,000 SHOTS PER N

…

left · ember arrow = the watched qubit; the faint dashed arrow is the same qubit with nobody watching · right · dots = this session's tally at each N you try · line = cos2N(π/2N), the chance of passing every check.

02

In plain words

Every quantum change starts slowly. Just after the push begins, the chance of having left 0 grows like the square of the time. Ten short waits therefore leak far less than one long one, and each check throws away the progress before it can speed up.

Nobody has to look. A steady leak of information into the surroundings works too: at a recording rate γ = 100Ω (Ω is the flip frequency), only 1.5% flips in one flip-time, and the rate falls like Ω²/γ. A state can only change if its record can.

Watching can also speed things up. For a decaying atom checked at the wrong pace, decay ran 4.2 to 8.4 times faster (anti-Zeno). Turn the question a little each time and the qubit follows: 100 small steps dragged it across 97.6% of the time. A vague check ("still inside the safe zone?") protects a superposition that a sharp one ("which level?") destroys. The sims also replayed the 1990 NIST ion experiment for 1 to 64 checks.

Prior artItano et al. 1990, Zabalo et al. 2020, Kofman and Kurizki 2000 (verified in v2).

Next clickThe measurement-induced phase transition post-selection problem, taken up by the toolkit.

This is textbook physics reproduced, not a discovery. The panel uses sharp, instant checks and a perfect two-level qubit; real checks take time and are never perfectly sharp. Its tally is a few hundred runs at most, so it wanders around the exact curve. The anti-Zeno, dragging and steady-leak numbers come from the exact sims in sims/zeno/, not from this panel.

Nearby on the bench