Like @craiggidney.bsky.social points out, what they are doing in this paper is really pretty much the same as that: post selecting a weakly entangled state into a subspace where there is lots of entanglement.

This is the kind of post selected entanglement that has been used in the SPDC based demonstrations of loop hole free bell tests.

Before their measurement, they have 2 such states. So when you e.g. post select on seeing a photon pair across the two states, some of this entanglement persists.

There is definitely still entanglement left, even if they don't admit it. The entanglement here can be seen when you don't post select. SPDC makes states roughly, in the fock basis, like: a|0,0> + b|1,1> + c|2,2> +... For some a,b,c where typically |a|^2>>|b|^2>>|c|^2. This state is entangled.

They pass the photons through filters such that if you post select on measuring a photon pair, both photon's frequencies are well defined and so there is no longer this kind of entanglement (approximately, a small amount would still remain in practice).

If you post select on measuring a photon pair and you resolve this mode information e.g. time and/or frequency resolving measurement, you can witness entanglement in this DoF.

It can often happen in SPDC (how they create the photon pairs) that the photons created are not always in the same "mode" i.e. the same time/frequency wavepacket

In this rev. mod. phys. hal.sorbonne-universite.fr/hal-03016379... they call it "modal unitary". I also see people call it the "linear optical unitary" representation.

There are lots of different terms out there. For something more physics-y I'd say you could also call it the "transfer matrix representation" which comes from what people in classical optics call the same thing. You could add "electric field" or "single photon" as a prefix to be more specific.

Ah sorry, it's a bit blurry even in the footage on my phone. Here's a zoom in of the screenshot. Although the jist of my post only requires observing that it is an example of a crudge analogue randomness generator.

So I'd perhaps suggest splitting up theory for experiments and experiments into different categories. Protocols could go with theory for experiments? Algorithms and applications could be merged?

I wonder is this stems from Scirate. If you follow the voting there, it is easy to be misled that most work posted to quant-ph is theory work. But I think this is due to scirate users seemingly being mostly theorists. There's also lots of cool experimental work being posted on quant-ph too.

Splitting things up seems like a nice idea. Although I think having most categories focused on subdividing theory work, giving experiments only 1 category out of the 7, which is also shared with theory for experiments seems too lopsided.

another example

Whenever I see massive equations written out like this, I wonder who it's actually for. Is ever helpful to the reader? Perhaps a better way to communicate such expressions would be to stick whatever mathematica/sympy notebook that created the expression up on github...?