Tag a photon's path and the interference vanishes
Ordinary double-slit interference requires that there be no way, even in principle, to know which slit a given particle went through -- the pattern is built from the interference of two indistinguishable possibilities. A quantum eraser experiment starts by deliberately breaking that indistinguishability: entangle each photon passing through the slits with a second, marker photon whose state records which-path information (for instance, a polarization state that differs depending on which slit the first photon took). The instant that which-path information exists anywhere in the universe, even if nobody has looked at it, the interference fringes at the detector screen disappear -- averaged over all photons, the pattern becomes a simple, fringe-free sum of two single-slit distributions.
Erasing the marker restores the fringes -- in a subset
The experiment's namesake move comes next: measure the marker photon in a basis that mixes the which-path information back together rather than reading it out directly -- for a polarization marker, this means measuring at a diagonal angle rather than along the axis that distinguishes the two paths. Sort the original photon's detector hits according to the outcome of that marker measurement, and within each sorted subset, interference fringes reappear, even though the raw, unsorted data (summing all subsets together) still shows no fringes at all. The eraser does not un-happen the earlier which-path tagging; it selects a measurement basis on the marker that no longer carries which-path information, and sorting by that measurement's outcome recovers coherence within each resulting subgroup.
Why this does not let you send a signal backward in time
Delayed-choice versions of this experiment -- where the decision to erase or not erase the marker is made after the original photon has already hit the detector screen -- have led to breathless claims about retrocausality, but the actual physics is more mundane and fully consistent with causality. The full, unsorted set of detector hits never shows interference, no matter what is later done to the marker photons; only after comparing each detector hit against its correctly paired marker measurement outcome does a fringe pattern emerge in the appropriately sorted subset. That comparison requires the ordinary, subluminal exchange of the marker measurement results, so no information can be extracted faster than light and no signal can genuinely be sent to the past.
What the marker actually needs to do
The critical requirement is not literally learning which path was taken -- it is that the two paths become distinguishable in principle through the entangled marker, whether or not anyone reads it out. A marker that only partially distinguishes the paths (say, a polarization rotation of less than 90 degrees rather than a full flip) produces partial which-path information and correspondingly partial fringe visibility -- interference and distinguishability trade off continuously against each other, a quantitative relationship formalized in wave-particle duality relations that bound how much of each you can simultaneously have.
What this reveals about complementarity
The quantum eraser is often presented as evidence against a naive picture in which particles simply travel a definite unobserved path through one slit or the other -- if that path were a hidden but pre-existing fact, no downstream choice about how to measure an entangled marker photon should be able to restore fringes for any subset of events. Instead, the pattern of results is exactly what standard quantum mechanics predicts from entanglement and basis choice alone, and the experiment is best understood as a clean illustration of complementarity: which-path information and interference visibility are two aspects of the same system that cannot both be fully present, and the eraser measurement simply chooses, after the fact, which aspect the data will be analysed to reveal.
Frequently asked questions
Does the quantum eraser let you send information back in time?
No. The full data set, without sorting by the marker's measurement outcome, never shows interference regardless of what is done to the marker photon later. Recovering fringes requires comparing each detection event against its marker's outcome, which needs an ordinary, light-speed-or-slower exchange of that outcome -- so causality is fully preserved.
What exactly gets erased in a quantum eraser experiment?
Not the earlier which-path tagging itself, but the ability to read out that tagging in a way that distinguishes the two paths. Measuring the marker photon in a different basis mixes the path information back together, so within each sorted outcome group the two paths are indistinguishable again and interference can reappear.
Can you get partial interference with a partial marker?
Yes -- if the marker only weakly distinguishes the two paths, you get partial which-path information and correspondingly partial fringe visibility rather than an all-or-nothing effect. The trade-off between distinguishability and visibility is a continuous, quantitatively bounded relationship in quantum optics.
Try it live
Everything above runs in your browser — open Quantum Eraser and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.
▶ Open Quantum Eraser simulation