Adherents say quantum eraser tests show the present rewriting the past; physicists split on whether anything goes back.
The claim at the center is that a choice made now can change an event already recorded. Its adherents point to a family of laboratory tests, the delayed-choice experiments, in which the setting that decides what is learned about a particle is chosen after the particle has been detected or has entered the apparatus. The experiments themselves are not in dispute: they have been run and they match quantum theory. The fight is over what they mean, and a small group of physicists in quantum foundations argues that influences running backward in time are the best reading of quantum correlations.
The quantum eraser, 1999
Yoon-Ho Kim, R. Yu, S. P. Kulik, Y. H. Shih and Marlan O. Scully posted "A Delayed Choice Quantum Eraser" in March 1999, later published in Physical Review Letters in 2000. It reports an experiment proposed by Scully and Drühl in 1982. Entangled photon pairs were split: one photon went to detector D0, while its twin travelled about 2.5 meters further to detectors that either kept or erased which-path information, so anything learned from the twin came at least 8 nanoseconds after the first photon registered. Counted jointly with the twin through coincidence circuits, some subsets showed a double-slit interference pattern and others did not. The authors concluded that the which-path or both-path information of a quantum "can be erased or marked by its entangled twin even after the registration of the quantum."
Kim et al., "A Delayed Choice Quantum Eraser": https://arxiv.org/abs/quant-ph/9903047
Wheeler's test with single photons, 2007
Vincent Jacques, Alain Aspect, Jean-François Roch and colleagues reported in Science in 2007 an almost ideal version of John Wheeler's delayed-choice thought experiment. Single photons entered a two-path interferometer; a quantum random number generator 48 meters away then chose whether the paths would be recombined or left open, with the choice space-like separated from the photon's entry. Closed, the setup showed interference with 94 percent visibility; open, it gave the path with an error under 1 percent. They wrote that the photon's behavior "depends on the choice of the observable which is measured," and quoted Wheeler: "we have a strange inversion of the normal order of time." In the same passage they say nature behaves in agreement with the predictions of quantum mechanics.
Jacques et al., "Experimental realization of Wheeler's delayed-choice GedankenExperiment": https://arxiv.org/abs/quant-ph/0610241
Steering into the past, 2012
Xiao-song Ma, Anton Zeilinger and colleagues reported in Nature Physics the first realization of Asher Peres's delayed-choice entanglement swapping. Using four photons, they delayed the choice of measurement on two photons until after the other two had been registered; the choice was made by a high-speed tunable analyzer driven by a quantum random number generator, and it decided whether the first pair counted as entangled or separable. They describe it as "quantum steering into the past." They also write that if the quantum state is viewed as a real physical object, future actions appear to influence past recorded events, but that there is never a paradox if the state is no more than a "catalogue of our knowledge," in which case no physical influence into the past is needed.
Ma et al., "Experimental delayed-choice entanglement swapping": https://arxiv.org/abs/1203.4834
The case for retrocausality
Huw Price and Ken Wharton argued in Entropy in 2015 for what they call the retrocausal loophole in Bell's Theorem, traced to Olivier Costa de Beauregard's proposal of the late 1940s, a loophole they say receives little attention and remains poorly understood. On this view, Alice's measurement choice could affect her particle's past, and so the shared source, letting the influence follow a zigzag path rather than act at a distance. They write that Bell's result assumes measurement choices do not affect the particles' prior state, and that if such retrocausality is allowed, Bell's argument for action at a distance is undermined. They also argue that the zigzag lets no one signal, consistent with standard quantum mechanics, and avoids time-travel paradoxes.
Price and Wharton, "Disentangling the Quantum World": https://arxiv.org/abs/1508.01140
Time symmetry and the backward arrow
Matthew Leifer and Matthew Pusey, in Proceedings of the Royal Society A in 2017, took up an argument of Price's that any time-symmetric account of quantum reality must be retrocausal, involving influences that travel backward in time. They replaced one of his assumptions with a weaker one they call lambda-mediation and showed that their assumptions imply a timelike analogue of Bell's local causality. Their result, they write, shows "the impossibility of a (non-retrocausal) time-symmetric ontology."
Leifer and Pusey, "Is a time symmetric interpretation of quantum theory possible without retrocausality?": https://arxiv.org/abs/1607.07871
The reading without a backward arrow
Physicist Sean Carroll published on his blog in September 2019 a chapter cut from his book Something Deeply Hidden. Working through a version with electrons and a recording spin, he wrote that the eraser leaves the total pattern on the screen "smooth and interference-free"; interference shows only when hits are sorted afterward by how the partner was measured. He wrote that "a number of serious researchers in quantum foundations" take the delayed-choice eraser as evidence of retrocausality, and that Wheeler did not go that far. For his own Everettian view he wrote, "There's no need to invoke retrocausality," and said that someone who holds to collapsing wave functions and also wants time-symmetric laws can be led to see retrocausality as part of the story.
Sean Carroll, "The Notorious Delayed-Choice Quantum Eraser": https://preposterousuniverse.com/blog/2019/09/21/the-notorious-delayed-choice-quantum-eraser/
Where the record stands
The measured results are settled: delayed choices sort later-recorded data into subsets that do or do not interfere. Whether that sorting reaches back in time is an open question of interpretation, argued in print by physicists on both sides. Price and Wharton, who favor a backward influence, also write that it could not be used to send a signal.
