Leo1233783 said:
The projection concept is not mainstream.
I don't agree with you here. In practice, the projection 'postulate' is used extensively when calculating stuff (certainly in quantum optics/information anyway). Whether one views this as representative of a physical 'collapse' or merely as a convenient mathematical tool for getting the right answers is irrelevant here.
One can avoid using this postulate if you want - personally I find the calculations harder doing it that way - I have to mangle my thoughts into a place that doesn't work so well for me.
So for the entanglement swapping the basic idea is this. Start off with (1,2) and (3,4) being entangled pairs of qubits. In principle we could have the pair (1,2) in London and pair (3,4) in New York. Now let's courier particles 2 and 3 to Manila. Clive, on holiday in the Philippines, performs a Bell measurement on the particle pair (2,3) - and we suppose it is possible to do an ideal measurement here (some people on here call it a filter measurement, I've always known this as an ideal von Neumann measurement).
Using the projection 'picture' - the state of the particles (1,4) will be projected into one of the 4 entangled Bell states after Clive's measurement/filter. Thinking that way makes the calculation really easy. Yes, I agree that we don't have to do it this way, but why make life difficult for oneself?
We could imagine having large numbers of such pairs. If Clive does the measurement on these pairs and some short time later (we'll imagine them all in the same frame of reference with synchronised clocks) Alice and Bob (who have qubits 1 and 4, respectively) now do a standard set of measurements that they would do if they were going to test for violations of Bell's inequality.
When they subsequently get the results of Clive's measurements on each particle pair (2,3) they will be able to identify 4 sub-ensembles, one for each of Clive's 4 possible results - and they will find a violation of the Bell inequality for each sub-ensemble, but no violation for the entire data set.
Whether you use the projection postulate or not when you're working out the prediction for the measurements on the sub-ensembles you'll get the same prediction - each sub-ensemble shows a violation of the BI. It's up to you how you interpret that result - for me it's just a whole lot easier to imagine that each particle in a sub-ensemble has been projected into a definite Bell state by Clive's measurement. But of course there are known 'philosophical' problems with thinking this way (you won't predict the wrong results for experiments but thinking in 'projection' terms doesn't sit too well with relativity).
Yer pays yer money and you takes yer pick
You can go nuts and use the POVM formalism to calculate the results and their probabilities for the ideal entanglement swapping case - throw the whole 'no-projection' shebang at it, but the POVM formalism contains all of the ideal filter measurements and the 'projection' as a special case anyway. The utility of the POVM formalism is that it allows us to consider non-ideal measurements in a general theoretical setting (generalized measurements) so we can consider the case of entanglement swapping if we had a non-ideal measurement, for example. Furthermore every measurement we can dream up (ideal or otherwise), can be represented as a POVM. This then allows us to optimise results over ##all## possible measurements - a useful thing to do in QKD, for example, where one might be interested in the lowest possible error rate caused by a measurement. But every POVM can also (theoretically) be represented as a system coupled to an ancilla upon which ideal 'projective' measurements are made - so I would argue that the notion of 'projection' is sort of really hiding in there anyway
So the notion of 'projection' still gets used quite a lot - even if ultimately we must consider it as just a cute mathematical device for getting to the right answers quicker (results and their probabilities and states after measurement so that we can predict results and probabilities of subsequent measurements).