Relational QM question from Rovelli's book Helgoland

[msumm21]

TL;DR Summary

The explanation of Bell violations in this book, according to relational QM, doesn't make sense to me

On pages 94-96 of Helgoland, Rovelli explains Bell violations according to relational QM. He speaks of spacelike separated measurements, one in Beijing and the other in Vienna. He says any talk of correlations between the 2 is meaningless, specifically:
"it makes no sense to ask whether the two results are the same or not."

He later says of the measurements:
"we cannot assume that both exist, because there is nothing with respect to which both can be determined."

He explains that the result in Vienna is not real in Beijing until a signal about it arrives in Beijing, and vice versa.

I'm not clear what he means about one of the measurements not being real or not existing, while the other is real and exists. Consider Alice in Beijing measuring one particle. So her particle's spin exists and is real, while Bob's in Vienna is not? Alice doesn't "directly" see the spin of her particle in Beijing, she only sees the result after that particle has entangled with many things to magnify the result to readable measurement. So I'm not clear why her particle & result is "real" after that chain of thousands (millions?) of entanglements, but then making just 1 more inference from entanglement (that the entangled particle in Vienna has the opposite spin) is somehow no good.

How does relational QM decide where to draw this line between which entanglements are valid to use for inference of reality and which are bogus?

 

[Nugatory]

He explains that the result in Vienna is not real in Beijing until a signal about it arrives in Beijing, and vice versa.

We have the same annoyance in ordinary boring special relativity: events A and B are spacelike-separated, therefore we have to be in the intersection of their future light cones to be able say that they both have happened.

 

[martinbn]

https://arxiv.org/abs/1806.08150

 

[msumm21]

We have the same annoyance in ordinary boring special relativity: events A and B are spacelike-separated, therefore we have to be in the intersection of their future light cones to be able say that they both have happened.

Understood, but to me this doesn't resolve the fact that there seems to be a nonlocal influence in these experiments. It appears to me that he thinks this mysterious action at a distance is resolved by RQM, see below...

https://arxiv.org/abs/1806.08150

In the abstract they say, referring to Bell experiments, "there is no need to appeal to a mysterious space-like influence to understand it."

If I understood, relational QM says (at least) one measurement was not real and therefore no spacelike influence, case closed. Is that it? I agree you could define the word "real" in such a way as to claim one measurement is not real. However, the fact that experiments have done the measurements with spacelike separation and have subsequently shown violations of Bell inequalities means there was some space-like influence. I don't see how RQMers claim this effect vanishes because they change their definition of "real" to exclude one of the measurements for some amount of time.

 

[DrChinese]

https://arxiv.org/abs/1806.08150

From the cited paper, "The notion of locality in relational quantum mechanics (2018)":

[From section I:]"One encounters at least five different notions of ‘locality’ in the literature:
1. No superluminal signalling: signals cannot propagate faster than light;
2. No superluminal causal influence: causes and effects of events are no further away than permitted
by the velocity of light;
3. No space-like influence: space-like separated quantum systems do not influence each other;
4. Point-like interaction: quantum systems (or fields of the Lagrangian in quantum field theory) interact only at the same point in spacetime;
5. Local commutativity: space-like separated local observables commute. [Note that this notion of locality is completely meaningless - nothing changes in quantum theoretical predictions for entangled systems regardless of the local/nonlocal split. I.e. there is nothing to test.]"

...

[From the conclusion section:] "With the relational interpretation, quantum mechanics is still ‘non-local’ in the sense of Bell, but it nonetheless remains true that ‘causes and effects of events are no further away than permitted by the velocity of light’. Indeed, if the observer O sees a light signal from A, he will think the radioactive particle in the region of the past cone Λ is the cause of the detection of an α particle in A. The same thing could be said symmetrically for B. Neither causes nor effects travel faster than light whatsoever. There are correlations between A and B because there is a common cause in their common past. Earlier claims in [5] that relational quantum mechanics was local were maybe misleading: the relational interpretation is not locally causal in the sense of Bell. However it should be clear now that this kind of non-locality cannot be interpreted as a superluminal interaction, and the relational interpretation is indeed local in all the senses listed in section I."

---

And yet again, we have a paper (this one by respected writers, in my opinion) that provides an "interpretation" denying superluminal/nonlocal influences. See statements in bold, for example. What's wrong with this picture?

Well, they start out with an example in which A and B share a common past. Note the circular reasoning? Naturally, you could always claim - as they do explicitly - that the common cause is in their common past. Not what I would personally call an insight. But why would you write a paper claiming that in 2018? There were theoretical papers about entanglement between an A and B that share no common past perhaps 20 years previously. More importantly, there were experimental realization of that theory at least 10 years earlier (2008).

"We have demonstrated high-fidelity entanglement swapping with time-synchronized independent sources. The swapped entanglement clearly violates a Bell-type inequality. These strong non-classical correlations between particles that do not share any common past..."

Wouldn't an advocate want to defend the Relational Interpretation against an experiment that maximally challenges their position? Because: To the eye (and not considering interpretations), it certainly looks like entangled particles ("space-like separated quantum systems") that have no common past DO have some kind of influence on each other. Those are denoted by experiments demonstrating: perfect correlations, violation of Bell inequalities, GHZ results, etc.

Admittedly, this is not a direct answer to the OP's question.

 

[martinbn]

And yet again, we have a paper (this one by respected writers, in my opinion) that provides an "interpretation" denying superluminal/nonlocal influences. See statements in bold, for example. What's wrong with this picture?

Nothing is wrong! They don't deny that other interpretations can have superluminal/nonlocal influences. They only explain that in this particular interpretation there are no such things. If someone insists that there are such influences in all interpretations, well, he has to prove it.

Well, they start out with an example in which A and B share a common past. Note the circular reasoning? Naturally, you could always claim - as they do explicitly - that the common cause is in their common past. Not what I would personally call an insight. But why would you write a paper claiming that in 2018? There were theoretical papers about entanglement between an A and B that share no common past perhaps 20 years previously. More importantly, there were experimental realization of that theory at least 10 years earlier (2008).

"We have demonstrated high-fidelity entanglement swapping with time-synchronized independent sources. The swapped entanglement clearly violates a Bell-type inequality. These strong non-classical correlations between particles that do not share any common past..."

Wouldn't an advocate want to defend the Relational Interpretation against an experiment that maximally challenges their position? Because: To the eye (and not considering interpretations), it certainly looks like entangled particles ("space-like separated quantum systems") that have no common past DO have some kind of influence on each other. Those are denoted by experiments demonstrating: perfect correlations, violation of Bell inequalities, GHZ results, etc.

The observer that analyses all the data has everything in his past.

 

[DrChinese]

The observer that analyses all the data has everything in his past.

This is always true with scientific experiments. It implies that the only possible type of non-locality or FTL influences are ones that include FTL signaling. Obviously this is circular logic.

There can (conceptually) be FTL influences that do not allow signaling. You must allow the proof of that to be brought to a central location, even if carried in a handwritten notebook on a horse-drawn buggy.

 

[msumm21]

Nothing is wrong! They don't deny that other interpretations can have superluminal/nonlocal influences. They only explain that in this particular interpretation there are no such things. If someone insists that there are such influences in all interpretations, well, he has to prove it.

I see Bell violations as an experimental fact that proves non-local influence (or determinism).

Regarding your last sentence, I view it the other way around: if RQM claims these correlations are not happening, I think it's on them to show/explain how Alice and Bob can perform the measurements without correlation, but then somehow end up with correlation once they compare their measurements. For example, are Alice's brain and measurement results scrambled as soon as soon as they are combined with Bob's (to make the correlation appear at that time)? If not, the measurement correlations already existed, right?

 

[martinbn]

I see Bell violations as an experimental fact that proves non-local influence (or determinism).

What influences do you see?

Regarding your last sentence, I view it the other way around: if RQM claims these correlations are not happening, I think it's on them to show/explain how Alice and Bob can perform the measurements without correlation, but then somehow end up with correlation once they compare their measurements. For example, are Alice's brain and measurement results scrambled as soon as soon as they are combined with Bob's (to make the correlation appear at that time)? If not, the measurement correlations already existed, right?

Presumably they do that, I have only glanced at RQM and cannot say what exactly they do or don't.

 

[PeroK]

I see Bell violations as an experimental fact that proves non-local influence.

The difficulty is that neither measurement can absolutely influence the other - precisely because the measurements are spacelike separated events. This means that there are frames of reference where the experiments take place in both orders. So, if you try to say that measurement A influenced measurement B, then what do you say in a frame of reference where measurement B was done first?

This more or less forces you to the conclusion that relativity is wrong. So, we'd like to find a way to explain Bell violations without throwing away relativity. It's a heavy price to pay in order to accept that nature can manage Bell violations.

The question, therefore, is whether there can be non-locality without non-local influence - i.e. without a FTL message from A to B or vice versa. Then you can ask the question: who says nature must be local? Maybe things are simply correlated? Maybe there is no "mechanism" in the sense that you would like there to be? So, for the time being at least, we just have to accept that we don't know. But, we can reject the necessity for non-local influence. Science tells us what we can measure and in that respect QM is compatible with relativity. It's only if we insist on looking beyond the measurements that we run into trouble.

I haven't read Rovelli, but I don't understand why the fact that neither experimenter can know about the other result immediately changes very much. The correlations are still there as far as I understand the concept of correlation. I don't see how talk of correlations is meaningless. Perhaps if I read the whole text I would understand his argument better.

 

[DrChinese]

The difficulty is that neither measurement can absolutely influence the other - precisely because the measurements are spacelike separated events. This means that there are frames of reference where the experiments take place in both orders. So, if you try to say that measurement A influenced measurement B, then what do you say in a frame of reference where measurement B was done first?

You can certainly execute the ordering so that A occurs before B in all reference frames*. And as I am fond of repeating, you can make the decision (call that C) to entangle what's measured at A and B after those same measurements.

So Relational QM should be able to assign an absolute answer to the original question of @msumm21 - right? But in @martinbn's reference above, they assert: "There are correlations between A and B because there is a common cause in their common past." But in fact, there is no common past necessary for the A & B photons.**


*All 3 events can be executed at the same place, in the spacetime order A-B-C. There can be no question of ordering in this case. The A/B photons can be created elsewhere, mutually distant if desired, and entangled later at time C.

**And it can be arranged so that photons measured by A and B do not share a common past (see Peterdonis' definition in #26 here and per various experiments I cited there too).

 

[martinbn]

You can certainly execute the ordering so that A occurs before B in all reference frames*. And as I am fond of repeating, you can make the decision (call that C) to entangle what's measured at A and B after those same measurements.

So Relational QM should be able to assign an absolute answer to the original question of @msumm21 - right? But in @martinbn's reference above, they assert: "There are correlations between A and B because there is a common cause in their common past." But in fact, there is no common past necessary for the A & B photons.**


*All 3 events can be executed at the same place, in the spacetime order A-B-C. There can be no question of ordering in this case. The A/B photons can be created elsewhere, mutually distant if desired, and entangled later at time C.

**And it can be arranged so that photons measured by A and B do not share a common past (see Peterdonis' definition in #26 here and per various experiments I cited there too).

The case they consider there is a common past because the pair has a common sourse of creation.

In your favourite entanglement swapping you do not have a pair of entangled particles because only a subset of the measurments show the needed coraltations.

 

[DrChinese]

The case they consider there is a common past because the pair has a common sourse of creation.

In your favourite entanglement swapping you do not have a pair of entangled particles because only a subset of the measurments show the needed coraltations.

They don’t have a common source. They are created from different crystals.

You are confused about subsets. It is axiomatic that different entangled streams cannot be entangled, due to monogamy of entanglement. Please read the references and you will see that ONLY a swap leads to transfer of entanglement. Such swap requires an interaction and indistinguishability.

What is true is that the Bell state must be identified after a swap.

 

[martinbn]

They don’t have a common source. They are created from different crystals.

You are confused about subsets. It is axiomatic that different entangled streams cannot be entangled, due to monogamy of entanglement. Please read the references and you will see that ONLY a swap leads to transfer of entanglement. Such swap requires an interaction and indistinguishability.

What is true is that the Bell state must be identified after a swap.

See the other thread about the subsets. It is in the paper!

 

[msumm21]

What influences do you see?

That Alice's measurement result is a function of Bob's measurement as shown by Bell. I understand there's a loophole with determinism, but for the purposes here I'm ignoring that, because I think RQM is not using the determinism loophole, right?

The difficulty is that neither measurement can absolutely influence the other - precisely because the measurements are spacelike separated events. This means that there are frames of reference where the experiments take place in both orders. So, if you try to say that measurement A influenced measurement B, then what do you say in a frame of reference where measurement B was done first?

Understood, but spacelike separated experiments violate Bell inequalities, so the measurement result at one of the spacelike separated measurement events was a function of the choice & result at the other. If this is accepted (and determinism is denied), then this to me IS an influence.

Maybe someone could better explain their definition of "influence" and how it's not applicable to the case of Bell violations (without using determinism)?

 

[martinbn]

That Alice's measurement result is a function of Bob's measurement as shown by Bell. I understand there's a loophole with determinism, but for the purposes here I'm ignoring that, because I think RQM is not using the determinism loophole, right?


Understood, but spacelike separated experiments violate Bell inequalities, so the measurement result at one of the spacelike separated measurement events was a function of the choice & result at the other. If this is accepted (and determinism is denied), then this to me IS an influence.

Maybe someone could better explain their definition of "influence" and how it's not applicable to the case of Bell violations (without using determinism)?

Which one influences which? It is completely symmetric! Is Alice's measurment influencing Bob's or the other way around?

 

[PeroK]

Understood, but spacelike separated experiments violate Bell inequalities, so the measurement result at one of the spacelike separated measurement events was a function of the choice & result at the other. If this is accepted (and determinism is denied), then this to me IS an influence.

An influence is a physical thing that travels from one event to the other. What you really mean is that this implies an influence. But, logic and implications always depend on your hypotheses. I would claim that one of your hypotheses is that, in some shape or form, nature is local. So, you have a logical argument:

1. Nature is local
2. A and B are determined locally.
3. A and B are correlated

Conclusion: either A influences B or B influences A.

Note that 2. is needed otherwise there is nothing mysterious.

But, if we drop hypothesis 1, then the conclusion does not follow logically.

Maybe someone could better explain their definition of "influence" and how it's not applicable to the case of Bell violations (without using determinism)?

I would say there are three broad possibilities, which very crudely are:

A) The whole thing is predetermined in some way.
B) Relativity is wrong and FTL influences are at work.
C) Nature is non local.

Some people (Palmer, the superdeterminist) are convinced it must be A. You are convinced it must be B. I'm agnostic, but at long as C remains an option I would prefer that.

 

[msumm21]

An influence is a physical thing that travels from one event to the other.

Using this definition of influence, I would agree that there may be no influence (I can't imagine a physical thing travelling between Alice and Bob's measurement events). However, I've thought of influence as just one thing having an effect on another (if not for QM I would've thought that implied your definition).

I would say there are three broad possibilities, which very crudely are:

A) The whole thing is predetermined in some way.
B) Relativity is wrong and FTL influences are at work.
C) Nature is non local.

Our difference here is probably again just thinking of the word "influence" different. I would just have 2 options as (I think) my definition of influence is equivalent to non local (let me think on this some more). So I would have
A) determinism
B) non local

Just to clarify, I don't rule out A in general. Just trying to ignore A for this thread because I think RQM isn't using A to explain this anyway.

 

[martinbn]

Using this definition of influence, I would agree that there may be no influence (I can't imagine a physical thing travelling between Alice and Bob's measurement events). However, I've thought of influence as just one thing having an effect on another (if not for QM I would've thought that implied your definition).

Let me ask you again, which is the one thing that has an effect on the other? Alice's measurment on Bob's or the other way around?

 

[msumm21]

Which one influences which? It is completely symmetric! Is Alice's measurment influencing Bob's or the other way around?

As in Bell's paper, you can write it either way: A is a function of B or vice versa.

I'd assume it's symmetric as you say, but can we guarantee that? For example, maybe someone could make a theory which, upon creation of entangled partners, there is an assignment by nature: this one is the "alpha" and that one is the "beta" and .... I realize this is may be unlikely, but not sure it's provably impossible.

 

[PeroK]

Our difference here is probably again just thinking of the word "influence" different.

Okay, but for "influence" to be physically meaningful it must be something more than the measurement results themselves. Correlation is a mathematical/statistical characteristic of data. You must be careful not to use "influence" as just another word for "correlation".

In fact, if you think more deeply about the details for a Bell-type experiment, what exactly could be communicated by the particles? Consider an electron in a Stern-Gerlach experiment. It experiences a magnetic field. It's state evolves, but remains in a superposition of spin states. It is then absorbed by a detector. At what point did the electron realise it was measured as "spin-up" or "spin-down" about the z-axis? And, at what point could it generate its influence to a distant positron, with which it is entangled?

My point is that a classical structure of influences and messages doesn't fit well the with whole way QM works. If you naively think that the electron experiences a magnetic field, that its spin wave-function collapses to a definite state (in the z-direction) and that this is promulgated to the rest of the universe, then you may have something. But, if you look more deeply, that's not what happens. There is no explicit measurement of the electron's spin anywhere to be found. There is no definite wave-function collapse. The spin is inferred from where it interacted with the detector.

So, even if you want to believe in an influence in some meaningful way, it's difficult to see where and how these influences could possibly be produced.

 

[msumm21]

You must be careful not to use "influence" as just another word for "correlation".

I'm not saying influence=correlation. E.g. the determinism loophole where something C beforehand caused A and B to happen, I wouldn't say A influenced B or vice versa. In general, I'd not claim influence between spuriously correlated events.

 

[DrChinese]

Which one influences which? It is completely symmetric! Is Alice's measurement influencing Bob's or the other way around?

Exactly! Causality fails in QM, I think you've got it now...

 

[DrChinese]

As in Bell's paper, you can write it either way: A is a function of B or vice versa.

I'd assume it's symmetric as you say, but can we guarantee that? For example, maybe someone could make a theory which, upon creation of entangled partners, there is an assignment by nature: this one is the "alpha" and that one is the "beta" and .... I realize this is may be unlikely, but not sure it's provably impossible.

We have a theory, but it doesn't specify which causes which. It just says there is no relevant causal order, except by (unjustified and at this time unprovable) assumption. If you can experimentally entangle particles that no longer exist, why deny this point?

 

[msumm21]

We have a theory, but it doesn't specify which causes which.

Do you mean a theory with the property I was speaking of, i.e. one that's not symmetric between Alice and Bob? I realize I didn't explain what I meant by that...

I meant a theory that somehow treats two entangled particles in a fundamentally different way. For a (contrived) example, say one of the particles (Alice's) "chooses" its measurement outcome independent of Bob's, but Bob's "chooses" its measurement outcome as a function of Alice's measurement & result. So it's fundamentally not symmetric. I'm not saying I have such a theory, only that maybe someone could cook up such a thing. (would be awkward in the simpler case of time-like separated measurements.)

If you can experimentally entangle particles that no longer exist, why deny this point?

I'm not clear I understand here? Why deny what point?

 

[msumm21]

So I would have
A) determinism
B) non local

Regarding the list of possible ways to achieve Bell violations, I'd listed the 2 things above, but forgot to include the possibility discussed in post 8 above:

For example, are Alice's brain and measurement results scrambled as soon as soon as they are combined with Bob's (to make the correlation appear at that time)? If not, the measurement correlations already existed, right?

This 3rd option is what I'm guessing RQM proposes, but still unsure. The idea is that correlations change after the space-like separated measurements are brought together to compare. E.g. even if we start in $$|01+10\rangle$$ Alice & Bob can both measure $00$ , but later, when the results are brought together, one of their results (and their memory) changes to $1$ to make proper correlations appear. This is the only way I can see to make RQM work. If we name this "transient measurement results" the options become:

A) determinism
B) non locality
C) transient measurements

 

[martinbn]

Regarding the list of possible ways to achieve Bell violations, I'd listed the 2 things above, but forgot to include the possibility discussed in post 8 above:

This 3rd option is what I'm guessing RQM proposes, but still unsure. The idea is that correlations change after the space-like separated measurements are brought together to compare. E.g. even if we start in $|01+10\rangle$ Alice & Bob can both measure $00$ , but later, when the results are brought together, one of their results (and their memory) changes to $1$ to make proper correlations appear. This is the only way I can see to make RQM work. If we name this "transient measurement results" the options become:

A) determinism
B) non locality
C) transient measurements

No, this is not what they say at all.

 

[msumm21]

No, this is not what they say at all.

OK, so this is the main question behind this thread. I don't understand what they ARE saying. Alice and Bob make space-like separated measurements which they can physically record, but RQM says the correlations don't exist yet. Alice and Bob meet up and compare their recorded measurements and now the correlations appear and are "real."

In the example of Alice & Bob (A & B) measuring $\{0,1\}$ on $|00+11\rangle$, what exactly happens so that the results aren't correlated until they compare notes?