Consistent Histories and Locality

[PeterDonis]

the probabilities for outcomes of measurements on 1&4 are independent of what is done on 2&3

One has to be very careful phrasing this. As you state it, it could be true or it could be false, depending on how your ambiguous wording is interpreted.

It is impossible to send signals to observers measuring photons 1 & 4 by choosing whether or not to allow a swap operation to take place on photons 2 & 3 at the BSM. In that sense your statement is true.

However, if you do two experiments, one in which the swap operation is done and one in which it is not, and you hand the two sets of data (the measurement results by run for all four photons) to someone, without telling them which set is the "swap" set and which is the "no swap" set, they can tell which is which by looking at the photon 1 & 4 correlations in each subset picked out by the four possible combinations of photon 2 & 3 results. In the "swap" set, there will photon 1 & 4 correlations in each subset, and those correlations can violate the Bell inequalities, whereas in the "no swap" set there will be no correlations even by subset. In that sense your statement is false.

 

[martinbn]

One has to be very careful phrasing this. As you state it, it could be true or it could be false, depending on how your ambiguous wording is interpreted.

It is impossible to send signals to observers measuring photons 1 & 4 by choosing whether or not to allow a swap operation to take place on photons 2 & 3 at the BSM. In that sense your statement is true.

However, if you do two experiments, one in which the swap operation is done and one in which it is not, and you hand the two sets of data (the measurement results by run for all four photons) to someone, without telling them which set is the "swap" set and which is the "no swap" set, they can tell which is which by looking at the photon 1 & 4 correlations in each subset picked out by the four possible combinations of photon 2 & 3 results. In the "swap" set, there will photon 1 & 4 correlations in each subset, and those correlations can violate the Bell inequalities, whereas in the "no swap" set there will be no correlations even by subset. In that sense your statement is false.

My point is that if you give them the two sets of the measurement result only of photons 1&4, then they cannot tell which is which.

 

[PeterDonis]

My point is that if you give them the two sets of the measurement result only of photons 1&4, then they cannot tell which is which.

Yes, but why would you do that? Photons 2 & 3 are part of the total system; photon 2 starts out entangled with photon 1, and photon 3 starts out entangled with photon 4, and those entanglements are stipulated in the preparation of the system. If you leave out those results, you're leaving out relevant information.

It is of course true that a reduced density matrix for a quantum system that traces over other systems with which that system might be entangled, will not show correlations with those other systems. That's a simple mathematical fact that is stated in pretty much every QM textbook. But that doesn't mean those correlations do not exist, or that they are not physically meaningful.

 

[martinbn]

Yes, but why would you do that? Photons 2 & 3 are part of the total system; photon 2 starts out entangled with photon 1, and photon 3 starts out entangled with photon 4, and those entanglements are stipulated in the preparation of the system. If you leave out those results, you're leaving out relevant information.

It is of course true that a reduced density matrix for a quantum system that traces over other systems with which that system might be entangled, will not show correlations with those other systems. That's a simple mathematical fact that is stated in pretty much every QM textbook. But that doesn't mean those correlations do not exist, or that they are not physically meaningful.

But the claim is that measurment on 2&3 affects 1&4. What is the meaning of that if it cannot be observed?

 

[PeterDonis]

the claim is that measurment on 2&3 affects 1&4. What is the meaning of that if it cannot be observed?

It can be observed: when you look at the presence or absence of correlations in the 1&4 subsets corresponding to the four possible combinations of 2&3 results, depending on whether the experimenter made a choice to have a swap take place, you are observing that the experimenter's choice of whether or not to make a swap at 2&3 affects 1&4. In any other branch of science, that would be a commonplace claim: experimenter makes an intervention and the presence vs. the absence of that intervention shows up in a predictable way in the data. Why it somehow becomes problematic when we're talking about QM and entangled systems is not clear to me.

 

[martinbn]

It can be observed: when you look at the presence or absence of correlations in the 1&4 subsets corresponding to the four possible combinations of 2&3 results, depending on whether the experimenter made a choice to have a swap take place, you are observing that the experimenter's choice of whether or not to make a swap at 2&3 affects 1&4. In any other branch of science, that would be a commonplace claim: experimenter makes an intervention and the presence vs. the absence of that intervention shows up in a predictable way in the data. Why it somehow becomes problematic when we're talking about QM and entangled systems is not clear to me.

I don't understand. If you don't make a BMS on 2&3 there is still going to be for subsets of the data for 1&4 with those correlations. We may not be able to tell which trials to look at but they are there. So I still don't understand why in one siruation they were caused by something romote and in the other not? Also in any branch of science if the probabilities of the outcomes do not change base on whether you do something over there or not suggests that the doing or not of something over there is not the cause of the outcoms.

 

[PeterDonis]

If you don't make a BMS on 2&3 there is still going to be for subsets of the data for 1&4 with those correlations.

No, there will not. That's the whole point. If no swap is done on 2&3, then the 1&4 subsets corresponding to each of the four possible combinations of 2&3 results (HH, HV, VH, VV) will not show any correlations, because photons 1&4 are not entangled if no swap is done. But if a swap is done on photons 2&3, then those four subsets of 1&4 results will show correlations, corresponding to the entangled Bell state that is indicated by each of the four 2&3 combinations.

These correlations do not allow signaling, since they are only detectable once the results are all collected and the subsets picked out. But they are a measurable difference between the swap and no swap cases.

 

[martinbn]

No, there will not. That's the whole point. If no swap is done on 2&3, then the 1&4 subsets corresponding to each of the four possible combinations of 2&3 results (HH, HV, VH, VV) will not show any correlations, because photons 1&4 are not entangled if no swap is done. But if a swap is done on photons 2&3, then those four subsets of 1&4 results will show correlations, corresponding to the entangled Bell state that is indicated by each of the four 2&3 combinations.

These correlations do not allow signaling, since they are only detectable once the results are all collected and the subsets picked out. But they are a measurable difference between the swap and no swap cases.

No, i am not saying that the subsets of 1&4 will correspond to the outcomes HH, HV, VH, and VV of the 2&3. I am saying that the data set of the 1&4 can be partitioned into such four subsets.

 

[PeterDonis]

i am not saying that the subsets of 1&4 will correspond to the outcomes HH, HV, VH, and VV of the 2&3.

This makes no sense. Of course you can pick any subsets out of the total 1&4 data set you like, if you don't care what they mean or don't mean. But the subsets that matter, physically, are the ones that correspond to the four possible combinations of 2&3 outcomes--because those are the ones that are predicted by the standard math of QM to show the Bell state correlations if the experimenter chooses to do a swap, but not if the experimenter doesn't. In other words, those are the relevant subsets for testing the theory against experiment.

I am saying that the data set of the 1&4 can be partitioned into such four subsets.

Of course it can. I'm saying the same thing. And I'm also saying that if you partition the 1&4 data that way, then you will see Bell state correlations in each subset if and only if the experimenter chose to do a swap.

If you dispute this, you are disputing both the standard math of QM and the experimental facts.

 

[DrChinese]

So I still don't understand why in one situation they were caused by something remote and in the other not? Also in any branch of science if the probabilities of the outcomes do not change base on whether you do something over there or not suggests that the doing or not of something over there is not the cause of the outcoms.

You have everything backwards. The 4 fold outcomes DO change depending on "whether you do something over there or not"! That's what the point of the experiment is!! To summarize (for the Nth time):

- 4 fold coincidences when experimenter selects SWAP=on (indistinguishable HH or VV for 2&3): Correlation high, as predicted by QM.
- 4 fold coincidences when experimenter selects SWAP=off: (distinguishable HH or VV for 2&3): Correlation negligible, as predicted by QM.

We are simply saying that there is a published experiment by a top team, and we are asking for a description of how an QM interpretation can explain THAT experiment without recourse to nonlocality. Because to the naked eye, the results appear* to clearly demonstrate non-signaling nonlocality.

---

What you are saying is that the 2 fold correlated outcomes don't appear to change because we don't know which bin to place the "subsets" in. So what? That's a different experiment and has nothing to do with entanglement swapping. You can test that with any two sources and get the same results. Imagine you are testing the Earth's gravitational acceleration, 32 ft/sec^2. You let an apple fall 16 feet, but you don't record or report the elapsed time duration. That is essentially what you are describing, half an experiment.


*Of course there are Interpretations that are explicitly nonlocal. Other Interpretations may have features that can explain the apparent nonlocality in some manner that retains locality. Those are the explanations I am requesting.

 

[Morbert]

This makes no sense. Of course you can pick any subsets out of the total 1&4 data set you like, if you don't care what they mean or don't mean. But the subsets that matter, physically, are the ones that correspond to the four possible combinations of 2&3 outcomes--because those are the ones that are predicted by the standard math of QM to show the Bell state correlations if the experimenter chooses to do a swap, but not if the experimenter doesn't. In other words, those are the relevant subsets for testing the theory against experiment.


Of course it can. I'm saying the same thing. And I'm also saying that if you partition the 1&4 data that way, then you will see Bell state correlations in each subset if and only if the experimenter chose to do a swap.

If you dispute this, you are disputing both the standard math of QM and the experimental facts.

The standard math of QM says that Bell-inequality-violating correlations between the appropriate measurements on 1 and 4 will themselves be correlated with outcomes of a BSM on 2 and 3. This is all the math of QM commits us to. It does not commit us to nonlocal influence unless we insist on a hidden variable theory that reproduces these correlations QM predicts.

 

[DrChinese]

The standard math of QM says that Bell-inequality-violating correlations between 1 and 4 will themselves be correlated with outcomes of a BSM on 2 and 3. This is all the math of QM commits us to. It does not commit us to nonlocal influence unless we insist on a hidden variable theory that reproduces these correlations QM predicts.

That is most certainly not true. Any of the following are consistent with Bell:

a) Denial of locality via nonlocal hidden variables interpretations. (Bohmian, etc.)
b) Denial of hidden variables via various "local" interpretations (MWI, Time symmetric, etc.)
c) Denial of both local and realism/hidden variables. (I would personally place standard QM in this bucket.)

We should probably be discussing type c) in order to be consistent with all of the experiments I am aware of. The Heisenberg Uncertainty Principle - and experiments supporting it - strongly imply there is no realism or hidden variables. Swapping experiments strongly imply Einsteinian locality and causality fails.

 

[PeterDonis]

The standard math of QM says that Bell-inequality-violating correlations between the appropriate measurements on 1 and 4 will themselves be correlated with outcomes of a BSM on 2 and 3. This is all the math of QM commits us to. It does not commit us to nonlocal influence

The words "nonlocal influence" do not appear anywhere in my post. I am simply trying to make sure we are all clear about what the standard math of QM and the experimental facts say, because the posts by @martinbn appear to me to indicate that that is not clear to everyone in this thread.

 

[martinbn]

This makes no sense. Of course you can pick any subsets out of the total 1&4 data set you like, if you don't care what they mean or don't mean. But the subsets that matter, physically, are the ones that correspond to the four possible combinations of 2&3 outcomes--because those are the ones that are predicted by the standard math of QM to show the Bell state correlations if the experimenter chooses to do a swap, but not if the experimenter doesn't. In other words, those are the relevant subsets for testing the theory against experiment.


Of course it can. I'm saying the same thing. And I'm also saying that if you partition the 1&4 data that way, then you will see Bell state correlations in each subset if and only if the experimenter chose to do a swap.

If you dispute this, you are disputing both the standard math of QM and the experimental facts.

I don't dispute any of this. May be I don't express myself well. What confuses me is that it seems that we say exactly the same thing except that you and @DrChinese add at the end "Therefore this proves that the measurements on 2&3 affect the outcomes for 1&4". I just don't see it.

 

[martinbn]

The words "nonlocal influence" do not appear anywhere in my post. I am simply trying to make sure we are all clear about what the standard math of QM and the experimental facts say, because the posts by @martinbn appear to me to indicate that that is not clear to everyone in this thread.

I am not questioning QM nor the experimental facts (which confirm QM, so in a theoretical discussion they are not really needed). I am questioning the conclusions @DrChinese makes from these experiments. He does use the words "nonlocal influence". And I was responding to his post that I quoted.

 

[PeterDonis]

What confuses me is that it seems that we say exactly the same thing except that you and @DrChinese add at the end "Therefore this proves that the measurements on 2&3 affect the outcomes for 1&4". I just don't see it.

"Affect" is a very general term. The experimenter's choice of whether or not to do a swap at 2&3 affects whether or not Bell state correlations appear in the measurements on 1&4. That's the experimental fact. What's wrong with using the word "affect" here?

He does use the words "nonlocal influence".

"Nonlocal influence" is a much more specific term than "affect". "Nonlocal influence" is a matter of interpretation; some QM interpretations claim it, some don't. But "affect", as above, is just the experimental fact; whether or not a swap is done at 2&3 makes a difference in whether or not Bell state correlations show up in the measurements on 1&4. If you agree with that last sentence, then I have no issue.

 

[DrChinese]

I am not questioning QM nor the experimental facts (which confirm QM, so in a theoretical discussion they are not really needed). I am questioning the conclusions @DrChinese makes from these experiments. He does use the words "nonlocal influence".

PetersDonis uses “affect” (as a verb). I use the term "nonlocal influence"; the authors of the paper(s) use the terms "quantum steering into the past" and "manifestation of the non-locality of quantum mechanics". These all describe what it looks like to the naked eye, considering that the experimenter can directly (causally?) influence the nature of the correlations observed - regardless of separation in space or time.

But these incredible experiments leave open some paths for interpretation, which is why we discuss this in the Interpretations subforum. My questions are simple: Does anyone know of an interpretations claiming locality that can explain these experiments? If so, I'd like to see the claim explained for the extent experiments. These experiments are roundly ignored by virtually all writers supporting local interpretations (MWI, QBism, Relational, etc.) - other than to merely claim that their (Bell defying) logic applies. These experiments don't need Bell. They are basically extensions of EPR's "elements of reality" updated for modern understanding.

 

[martinbn]

"Affect" is a very general term. The experimenter's choice of whether or not to do a swap at 2&3 affects whether or not Bell state correlations appear in the measurements on 1&4. That's the experimental fact. What's wrong with using the word "affect" here?

"Nonlocal influence" is a much more specific term than "affect". "Nonlocal influence" is a matter of interpretation; some QM interpretations claim it, some don't. But "affect", as above, is just the experimental fact; whether or not a swap is done at 2&3 makes a difference in whether or not Bell state correlations show up in the measurements on 1&4. If you agree with that last sentence, then I have no issue.

Fine, I am ok with the terms. But I don't understand the reasoning behind the experimental fact. So let me ask you this. Everyone agrees that the full set of measurements on 1&4 show no special correlations. And that there is a subset of 25% of them show correlations that violate Bell's inequality no matter what is done on the rest of the system. If you perform the experiment and do a BMS on 2&3 in 25% of the cases the result will be a projection to the phi minus state, and this subset of trials will match the subset of Bell inequality violating results of 1&4. So far I understand. But then the conclusion that if you don't do the BMS on 2&3 will change the outcome at 1&4 is unclear to me. Since we haven't done the BMS how do we know which subset of result at 1&4 to look at? If I understand you correctly you say that we measure the 2&3 in a different basis and look at the partition for the full set into the four subsets according to the four possible outcomes. But why? We cannot say that one of them would have been the subset with the phy minus state had we done the BMS on 2&3.

 

[martinbn]

PetersDonis uses “affect” (as a verb). I use the term "nonlocal influence"; the authors of the paper(s) use the terms "quantum steering into the past" and "manifestation of the non-locality of quantum mechanics". These all describe what it looks like to the naked eye, considering that the experimenter can directly (causally?) influence the nature of the correlations observed - regardless of separation in space or time.

I am not so hung up on the terminology. It just that the "directly (causally) influence" is not obvious to me. Not to mention that as you state it it implies influence in the past.

But these incredible experiments leave open some paths for interpretation, which is why we discuss this in the Interpretations subforum. My questions are simple: Does anyone know of an interpretations claiming locality that can explain these experiments? If so, I'd like to see the claim explained for the extent experiments. These experiments are roundly ignored by virtually all writers supporting local interpretations (MWI, QBism, Relational, etc.) - other than to merely claim that their (Bell defying) logic applies. These experiments don't need Bell. They are basically extensions of EPR's "elements of reality" updated for modern understanding.

I don't think that any interpretation claims that it is local in the sense of Bell inequality violations. The violations are a mathematical consequence of the theory, so no interpretation can avoid it. I think that all the claim is that they are local in the sense of lac of influence at a distance.

 

[Morbert]

PetersDonis uses “affect” (as a verb). I use the term "nonlocal influence"; the authors of the paper(s) use the terms "quantum steering into the past" and "manifestation of the non-locality of quantum mechanics". These all describe what it looks like to the naked eye, considering that the experimenter can directly (causally?) influence the nature of the correlations observed - regardless of separation in space or time.

The authors of the paper show that, given some sample of measurements on 1 and 4 over multiple runs, an experimenter can use a BSM to sort the sample into subsamples that exhibit Bell-inequality-violating correlations. This is different from showing that data in the sample is altered (immediately, retroactively or otherwise) by the BSM.

But these incredible experiments leave open some paths for interpretation, which is why we discuss this in the Interpretations subforum. My questions are simple: Does anyone know of an interpretations claiming locality that can explain these experiments? If so, I'd like to see the claim explained for the extent experiments. These experiments are roundly ignored by virtually all writers supporting local interpretations (MWI, QBism, Relational, etc.) - other than to merely claim that their (Bell defying) logic applies. These experiments don't need Bell. They are basically extensions of EPR's "elements of reality" updated for modern understanding.

Then entanglement swapping experiments are a red herring as your objections to these interpretations persist even in simpler experiments.

E.g. An experimenter can use CH to model an entanglement swapping experiment run with a set of possible histories and a probability space, and they can infer statements about photon 4 from measurements on photon 1 and photons 2,3 without nonlocal causal influence. Your objections to this approach (e.g. around the violation of the principle of unicity and the non-uniqueness of maximally fine-grained probability spaces) would be present even if we were considering a standard EPRB experiment.

Putting it another way: Entanglement swapping experiments show the establishment of entanglement between distant photons that do not have the same source, or never even coexisted. But none of these local interpretations rely on same-source particles to explain entanglement. They rely on a rejection of realism or on non-unicity or on spacetime state realism or on superdeterminism etc.

 

[PeterDonis]

I don't understand the reasoning behind the experimental fact

The experimental fact is what I described in post #71.

Everyone agrees that the full set of measurements on 1&4 show no special correlations.

Meaning, if you take the entire data set, not splitting it up into subsets based on the 2&3 results. Yes.

And that there is a subset of 25% of them show correlations that violate Bell's inequality no matter what is done on the rest of the system.

NO. Go read my post #71 again, carefully.

 

[PeterDonis]

the conclusion that if you don't do the BMS on 2&3 will change the outcome at 1&4 is unclear to me.

"Don't do the BMS" means "don't allow the swap to take place at the BMS". But photons 2&3 go through the BMS and get measured afterwards in all the runs, not just the ones where a swap takes place.

 

[PeterDonis]

Since we haven't done the BMS how do we know which subset of result at 1&4 to look at?

You pick out the four subsets corresponding to the four possible combinations of photon 2&3 measurement results: HH, HV, VH, VV. Those measurement results are there on every run, whether a swap takes place or not.

 

[Morbert]

You pick out the four subsets corresponding to the four possible combinations of photon 2&3 measurement results: HH, HV, VH, VV. Those measurement results are there on every run, whether a swap takes place or not.

I think I see the point of disagreement: We have four subsets HH, HV, VH, VV when no swap occurs. We have four subsets HH', HV', VH', VV' when a swap occurs (or at least HH' and VV'). I don't think it's the case that HH = HH' and VV = VV'

From Ma:

The Bell-state measurement (BSM) corresponds to turning on the switchable quarter-wave plates. In this case, the phase of the interferometer is π/2 and the interferometer acts as a 50/50 beam splitter. Therefore, the two photons interfere and are projected onto a Bell state by polarization-resolving single-photon detections.

The operation "quarter wave plates off and then polarization measurement" will select a different four subsets from "quarter wave plates on and then polarization measurement"

As this is converging with the other thread I might give my responses there

 

[PeterDonis]

I don't think it's the case that HH = HH' and VV = VV'

I have not claimed that it is. Nor has @DrChinese. Nothing in what either of us has said requires this. You are attacking a straw man.

Again, if you know of an interpretation of QM where the claim you are stating here is relevant, what is it, and what reference do you have to support it?

 

[Morbert]

I have not claimed that it is. Nor has @DrChinese. Nothing in what either of us has said requires this. You are attacking a straw man.

Again, if you know of an interpretation of QM where the claim you are stating here is relevant, what is it, and what reference do you have to support it?

Right now, I am making an interpretation-free statement about the experiment described in the Ma paper

These two measurements [BSM and SSM] are mutually exclusive (complementary in the Bohrian sense) in the same way as measuring particle or wave properties in an interference experiment.

If the subsets identified by the BSM are different from the subsets identified by the SSM (which everyone agrees) then we cannot conclude the BSM has a nonlocal influence without some further interpretational commitment.

@DrChinese disagrees, and believes we are compelled by standard QM to conclude nonlocal influence.

 

[PeterDonis]

If the subsets identified by the BSM are different from the subsets identified by the SSM (which everyone agrees)

Of course they are, since the two subsets represent different sets of actual, physical runs of the experiment. It makes no sense to even ask whether they are "the same".

then we cannot conclude the BSM has a nonlocal influence without some further interpretational commitment.

Obviously you need some "interpretational commitment" since the basic math of QM, without any interpretation, makes no claim whatever about "nonlocal influence".

@DrChinese disagrees, and believes we are compelled by standard QM to conclude nonlocal influence.

@DrChinese is saying that on his preferred interpretation, there is "nonlocal influence"--because his preferred interpretation is basically a "nonlocal realist" interpretation, where each individual run is described by a quantum state, and the final state of the 4-photon system is different for runs where a swap occurs vs. runs where no swap occurs, and since that state involves spatially separated entangled subsystems, it is inherently nonlocal, and so is any change that is made to it by any operation done during the experiment.

There is at least one interpretation where no such claim can be made, namely, a statistical interpretation such as the one used by Ballentine, in which the quantum state never describes individual runs, it only describes ensembles. Such an interpretation makes no claim about "nonlocal influence" because it makes no claim at all about what is happening in individual runs. It only makes claims about statistics.

I am still not clear about exactly what the Consistent Histories interpretation claims--whether it makes claims about what is happening in each individual run or not, and if it does, whether its claims involve any kind of "nonlocal influence".