Hossenfelder: Delayed Choice Quantum Eraser, Debunked?

[Halc]

TL;DR Summary

Questions on the validity of Hossenfelder video from 19 months ago

I have found serious errors in Sabine's videos before which has damaged my trust in her presenting peer-reviewed information.
I'm pretty sure I'm wrong about this one, but can't figure out where.

The video is here:
www.youtube.com/watch?v=RQv5CVELG3U

The gist is that individual photons get sent down and are logged by the coincidence counter connected to all five sensors.
Sensors D1 and D2 represent the which-path detection, in which case no interference pattern appears.
Sensors D3 and D4 have the 'which way' information erased (some time after the D0 pattern sensor has already been recorded) in which case the interference pattern appears.

Hossenfelder says that the two separate interference patterns (one from D3 events and the other from D4) are out of phase, so if added, they result in a normal curve without interference. This contradicts several things and hence makes me question the correctness of the video.

1: If D3 and D4 were the same sensor, the 'which slit' info would still not be there and a pattern should appear. She says not.
2: How does the pattern from one sensor get out of phase such that a different interference. pattern appears for half the photons? The two paths to D0 are the exact same length each time so the phase should be identical, regardless of which sensor (D3 or D4) picks up the idler photon. The peaks of the pattern should be the same regardless of how it is measured.

I'm missing something and I don't know what it is. I can accept #1 above as simply impossible, but #2 baffles me. I'm actually OK with quantum eraser being debunked like this because it seems pretty easy to send messages to the past with it if it didn't work the way she describes. It is implied that if one puts a Glan–Thompson prism right at the slits like that, splitting it into entangled beams, it counts as a which-slit measurement and there's no actual way to have one measurement device that detects the idler but doesn't count as a which-slit measurement.

At no point at D0 does a pattern appear. Only when the individual points are sorted (in the future, not the past) into D3 and D4-detected photons, does a pattern appear for each.

 

[vanhees71]

I'm not willing to watch videos. In the abstract Hossenfelder refers to a blogpost by Carroll. That's probably as "mediocre" as any pop-sci distortion of what's really done. Does she somewhere refer to a peer-reviewed source?

If it's the Walborn experiment than it's correct that the total ensemble doesn't show a double-slit interference pattern by construction, because 100% which-way information can in principle be gained due to the presence of the quarter-wave plates in the slits. For the partial ensembles, i.e., looking only at photon pairs where D3 registers one of the photons or looking only at photon pairs where only D4 registers one of the photons, the double-slit interference pattern is restored. That's precisely what's known as "delayed choice" and everything is in accordance with standard QED.

For a detailed (simplified) explanation, see

https://itp.uni-frankfurt.de/~hees/publ/habil-coll-talk-en.pdf

 

[Cthugha]

Hossenfelder says that the two separate interference patterns (one from D3 events and the other from D4) are out of phase, so if added, they result in a normal curve without interference. This contradicts several things and hence makes me question the correctness of the video.

Hossenfelder's explanation seems correct to me (at least at the popular science level).

1: If D3 and D4 were the same sensor, the 'which slit' info would still not be there and a pattern should appear. She says not.

I do not exactly understand what you mean by "the same sensor". If you intend to replace them by just one large detector that clicks if either D3 or D4 are hit, then this is not equivalent to which-slit info not being present. It is equivalent to ignoring this information which is not sufficient to get the interference pattern.

2: How does the pattern from one sensor get out of phase such that a different interference. pattern appears for half the photons? The two paths to D0 are the exact same length each time so the phase should be identical, regardless of which sensor (D3 or D4) picks up the idler photon. The peaks of the pattern should be the same regardless of how it is measured.

This is not how it works. The paths are not the same if you have a closer look at the setup. For D3, the first beam is reflected at the beam splitter, while the second is transmitted. For D4 the first beam is transmitted, while the second beam is transmitted. The phase shift between the transmitted and the reflected beams at a lossless 50-50 beam splitter is what shifts the patterns out of phase. The phase shift difference is also fixed (unless you make the beam splitter lossy or asymmetric).

 

[Halc]

For the partial ensembles, i.e., looking only at photon pairs where D3 registers one of the photons or looking only at photon pairs where only D4 registers one of the photons, the double-slit interference pattern is restored.

OK, I accept the accuracy, at least from a pop description, of what's presented. It seems to be a sort of paint-by number where each region on the page is assigned one of 4 values. A day later, somebody tells you to color the even numbers black and the odd numbered regions white, and a picture of boat appears. Alternatively one could choose to color the low two numbers black and it becomes a car. The choice in the future (how to color the page) affects the picture produced in the past about as much as that, hardly a case of reverse causality.

The paths are not the same if you have a closer look at the setup.

There are two signal paths from the photon emitter to to the D0 detector. These paths do not change in length from one run to the next. So in theory, if the photons (one photon taking both paths, no which-path collapse) are in the same phase each time, the interference patter can only appear one way, not the anti-pattern that cancels the first and adds to a non-interference pattern.
I know this is wrong, but the only way I can think of is that somehow the phase synchronization is lost somewhere. Sometimes perhaps one of the paths gets a reverse wave that's off by half a wavelength.

This in a way makes sense. If the waves are in sync, the idler photons interfere on one side of the 'erasing' splitter and add if they go to D3 and cancel if they go to D4. If one of the waves gets that half-wavelength shift, the reverse is true and D4 detects it. I'm ignoring the ones that go to D1 and D2. Let's assume those splitters are not even there for now.

This seems sort of a naive view, but I can't think of another way the out-of-phase pattern (the anti-one) can appear at D0. Something disrupts one or both paths (perhaps both by +/- quarter wavelength) somewhere. I suspect the G-T prism is involved since without that you have a normal double-slit setup with an interference pattern that doesn't require sorting the incidents into D3/D4 points.