A A stronger proof of nonlocality, or what?

[Demystifier]

TL;DR Summary

What's the meaning of the paper published in Nature Physics?

Half a year ago a group of authors published a paper in Nature Physics https://www.nature.com/articles/s41567-020-0990-x which seems to be a proof of nonlocality even stronger than Bell nonlocality. More precisely, according to a popular exposition by one of the authors https://theconversation.com/a-new-q...unBA8Lv0lVf8FchIm-tNumW2LfFs0ChlOlc48X9nut8_4
the theorem (and experiment) show that at least one of the 3 assumptions must be false:

1. When someone observes an event happening, it really happened.
2. It is possible to make free choices, or at least, statistically random choices.
3. A choice made in one place can’t instantly affect a distant event. (Physicists call this “locality”.)

In particular, I wonder how the minimal statistical interpretation would interpret it, which of the three assumptions would it deny?

 

[martinbn]

You forgot one more option he gives " ..or quantum mechanics itself must break down at some level. "

I haven't looked at the paper, but my guess is that all these option are everyday language statements that do not accurately represent the results in the paper.

 

[PeterDonis]

This appears to be the arxiv preprint of the paper:

https://arxiv.org/abs/1907.05607

 

[PeterDonis]

When someone observes an event happening, it really happened.

This assumption is obviously violated if we assume that, for example, Wigner's friend can perform arbitrarily quantum operations on Wigner. "Arbitrary quantum operations" includes reversing decoherence, and if decoherence can be reversed, then all statements of the form that anything "really happened" are now invalid; such statements rely on decoherence being irreversible.

 

[Demystifier]

This assumption is obviously violated if we assume that, for example, Wigner's friend can perform arbitrarily quantum operations on Wigner. "Arbitrary quantum operations" includes reversing decoherence, and if decoherence can be reversed, then all statements of the form that anything "really happened" are now invalid; such statements rely on decoherence being irreversible.

I don't understand how reversibility is related to reality. For analogy, many phenomena in classical physics are reversible, but it doesn't make them unreal.

 

[atyy]

the theorem (and experiment) show that at least one of the 3 assumptions must be false:

1. When someone observes an event happening, it really happened.
2. It is possible to make free choices, or at least, statistically random choices.
3. A choice made in one place can’t instantly affect a distant event. (Physicists call this “locality”.)

In particular, I wonder how the minimal statistical interpretation would interpret it, which of the three assumptions would it deny?

(1) is false in the minimal interpretation (thinking along the same lines as @PeterDonis). So I would say this isn't a stronger proof than the Bell inequalities. In the paper, there are real observers (Alice and Bob) and non-real observers (Charlie and Debbie).

 

[Morbert]

I don't understand how reversibility is related to reality. For analogy, many phenomena in classical physics are reversible, but it doesn't make them unreal.

Roughly speaking, classical physics is always decoherent, so reversing a classical system can never reverse decoherence.

As an aside: QM doesn't describe irreversible processes as giving rise to reality. Instead irreversibility of processes in reality are what let us successfully apply quantum theory to it.

 

[Demystifier]

See also a parody: https://www.physicsforums.com/threads/collection-of-science-jokes-p2.847743/post-6463166

 

[martinbn]

See also a parody: https://www.physicsforums.com/threads/collection-of-science-jokes-p2.847743/post-6463166

Now you are doing it again!

1. Planets exist even if no human observes them.

I thought you agreed to be more precise about this. What you mean is that at any given time the positions (or some other observables) of the planets have definite values.

 

[Demystifier]

Now you are doing it again!

I thought you agreed to be more precise about this. What you mean is that at any given time the positions (or some other observables) of the planets have definite values.

It's a parody.

 

[martinbn]

It's a parody.

I know! So?

 

[PeterDonis]

many phenomena in classical physics are reversible, but it doesn't make them unreal.

"Reversibility" in classical physics does not erase all memories and records of what happened. "Reversibility" in QM, i.e., reversing decoherence, the kind of operation that is assumed to be used by Wigner on his friend, does erase all memories and records of what happened. And that invalidates the way we talk about such things in ordinary language. When we say someone, like Wigner's friend, "observes" a result, we mean that an irreversible record of the result has been formed. But Wigner's friend-type scenarios involve quantum operations that reverse the making of the record; so any description of such a scenario that says the friend "observed" a result (and some such claim has to be made in order to derive any purported contradiction or paradox) is simply wrong.

 

[Demystifier]

"Reversibility" in classical physics does not erase all memories and records of what happened. "Reversibility" in QM, i.e., reversing decoherence, the kind of operation that is assumed to be used by Wigner on his friend, does erase all memories and records of what happened. And that invalidates the way we talk about such things in ordinary language. When we say someone, like Wigner's friend, "observes" a result, we mean that an irreversible record of the result has been formed. But Wigner's friend-type scenarios involve quantum operations that reverse the making of the record; so any description of such a scenario that says the friend "observed" a result (and some such claim has to be made in order to derive any purported contradiction or paradox) is simply wrong.

OK, I see what you mean and it makes sense. But I think that the two kinds of "reversibility" have not much to do with the difference between classical and quantum physics. Instead, the relevant difference is the fact that in one case only a subsystem is reversed, while in another case everything (including the memories and records) is reversed. In principle, both kinds of reversion are possible in both classical and quantum physics. The classical Liouville equation and the quantum von Neumann equation are not that different.

But more to the point, if I forgot that something happened, it does not mean that it was not real.

 

[Morbert]

But more to the point, if I forgot that something happened, it does not mean that it was not real.

Roland Omnes remarks that a device capable of making you forget would either i) have to violate relativity or ii) be too bulky to not collapse into a black hole. So the point might be moot since you cannot forget what is real.

 

[Demystifier]

Roland Omnes remarks that a device capable of making you forget would either i) have to violate relativity or ii) be too bulky to not collapse into a black hole. So the point might be moot since you cannot forget what is real.

My brain forgets information from my short term memory all the time, often due to distractions by devices such as TV and computer screen. But my TV neither violates relativity nor collapses into a black hole. Of course, this forgetting is only due to coarsegraining, at the fundamental microscopic level there may be no forgetting at all. So I guess by forgetting you mean fundamental microscopic forgetting, am I right?

 

[PeterDonis]

I think that the two kinds of "reversibility" have not much to do with the difference between classical and quantum physics.

Yes, they do, because "time reversibility" means different things in classical and quantum physics. In classical physics, "reversibility" applies directly to observables. In QM, it applies to the wave function.

Instead, the relevant difference is the fact that in one case only a subsystem is reversed, while in another case everything (including the memories and records) is reversed.

No, the relevant difference is that in QM, since measurement involves entanglement, it is impossible to just reverse a subsystem; the "memories and records" are entangled with the subsystem so you can only reverse everything, not just one thing.

 

[stevendaryl]

My brain forgets information from my short term memory all the time, often due to distractions by devices such as TV and computer screen. But my TV neither violates relativity nor collapses into a black hole. Of course, this forgetting is only due to coarsegraining, at the fundamental microscopic level there may be no forgetting at all. So I guess by forgetting you mean fundamental microscopic forgetting, am I right?

Presumably, the state of your brain after experiencing something and then forgetting it in the normal way is macroscopically different than the state of your brain before you ever experienced it.

 

[Morbert]

My brain forgets information from my short term memory all the time, often due to distractions by devices such as TV and computer screen. But my TV neither violates relativity nor collapses into a black hole. Of course, this forgetting is only due to coarsegraining, at the fundamental microscopic level there may be no forgetting at all. So I guess by forgetting you mean fundamental microscopic forgetting, am I right?

This notion of "forget" isn't the same as reversing a record. E.g. If you make a measurement and then immediately bump your head such that you can't remember the result, your brain still records the result of the measurement, in the same way the surrounding environment also records the measurement. Bumping your head just makes the record inaccessible to you on a conscious level. Like smashing your computer before reading the data.

 

[Demystifier]

Yes, they do, because "time reversibility" means different things in classical and quantum physics. In classical physics, "reversibility" applies directly to observables. In QM, it applies to the wave function.

No, the relevant difference is that in QM, since measurement involves entanglement, it is impossible to just reverse a subsystem; the "memories and records" are entangled with the subsystem so you can only reverse everything, not just one thing.

The comparison of classical and quantum physics can only make sense if we formulate both in similar formalisms. Hence, for the purpose of comparison, I propose to formulate classical and quantum mechanics in the language of Liouville and von Neumann equation, respectively. In this framework, none of the differences above sustain.

 

[PeterDonis]

The comparison of classical and quantum physics can only make sense if we formulate both in similar formalisms.

This is far too strong a requirement. The only thing we actually need to compare is predictions.

for the purpose of comparison, I propose to formulate classical and quantum mechanics in the language of Liouville and von Neumann equation, respectively

Doesn't this leave out measurements in the quantum case? If so, it's an incomplete comparison.

 

[Demystifier]

Doesn't this leave out measurements in the quantum case?

Not if you include the quantum state of the apparatus, e.g. as in the many-world interpretation.

 

[PeterDonis]

Not if you include the quantum state of the apparatus, e.g. as in the many-world interpretation.

So that means the validity of the comparison you describe is interpretation-dependent? That doesn't seem like a good choice of a method of comparison.

 

[Demystifier]

So that means the validity of the comparison you describe is interpretation-dependent?

Not necessarily. You can include the apparatus in a way that does not depend on interpretation. Many worlds are mentioned only because it is an example where the explicit quantum description of the apparatus is often used. Other examples are Bohmian interpretation and (interpretation independent!) theory of decoherence. All these approaches are based on the same theory of quantum measurements known as von Neumann theory. The only interpretation which is excluded by that approach is the Bohr's interpretation insisting that the apparatus must be described by classical physics.

 

[Morbert]

Just to add the usual consistent histories insight re/unitary evolution:

If we have closed system $|\Psi\rangle$ consisting of a measured system + apparatus + observer, with possible measurement outcomes $\{\lambda_i\}$, then the unitary description of the entire system is incompatible with the description of possible measurement outcomes, since the property $|\Psi\rangle\langle\Psi|$ that evolves unitarily does not commute with the measurement outcomes $|\lambda_i\rangle\langle\lambda_i|$Neither the unitary nor measurement description is more correct than the other, but only the latter is accessible to any observer who is a part of $\Psi$

 

[PeterDonis]

You can include the apparatus in a way that does not depend on interpretation.

But you can't have the entire system + apparatus obey the Von Neumann equation, which, as I understand it, is unitary, unless the interpretation you are using agrees with that, as the MWI does, correct?

 

[Demystifier]

But you can't have the entire system + apparatus obey the Von Neumann equation, which, as I understand it, is unitary, unless the interpretation you are using agrees with that, as the MWI does, correct?

Correct. But to avoid interpretation dependence, I can compare formalisms (von Neumann equation vs Liouville equation) without even talking about interpretations. And if one insists that I compare interpretations too, I can always choose the interpretation in which the two look most similar (which happens to be the Bohmian interpretation*).

*There is also an alternative, in which one modifies the interpretation of classical mechanics to make it more similar to standard "Copenhgen" QM. See my papers
https://arxiv.org/abs/quant-ph/0505143
https://arxiv.org/abs/0707.2319

 

[PeterDonis]

to avoid interpretation dependence, I can compare formalisms (von Neumann equation vs Liouville equation) without even talking about interpretations

But in the quantum case the formalism you are using in the comparison doesn't cover measurement unless you are adopting an interpretation like the MWI. So whether or not your comparison is complete is interpretation-dependent.

 

[Demystifier]

But in the quantum case the formalism you are using in the comparison doesn't cover measurement unless you are adopting an interpretation like the MWI. So whether or not your comparison is complete is interpretation-dependent.

I agree. No analysis is perfect. The analysis that you proposed is unperfect too, because, in a sense, it compares apples with oranges (classical with quantum). Mine is unperfect for the opposite reason, because it compares fruits with fruits (probability distribution with probability distribution).

 

[Kolmo]

It's a stronger no-go theorem than Bell's because it locates quantum theory more precisely in the landscape of general probability theories.

Bell like inequalities, combined with results like Fine's theorem, sketch out the region where probability theories are fundamentally Kolmogorov probability theories.

These new bounds, called the "Local Friendliness" bounds show a inequality obeyed by many non-Kolmogorov theories and yet still broken by quantum theory.

The bounds sketch out theories where it is generally difficult to demonstrate that their outcomes are randomly generated. Quantum Theory violates these bounds (still obeyed by many non-Kolmogorov theories) and thus is more accurate than Bell type results which simply tell you that Quantum Theory is not Kolmogorov.

If one prefers to think in terms of hidden variable theories, Bell's theorem rules out local hidden variable theories where as this result rules out a large class of nonlocal hidden variable theories. It can also be combined with Legett's earlier work ruling out a separate class of nonlocal theories.

 

[PeterDonis]

If one prefers to think in terms of hidden variable theories, Bell's theorem rules out local hidden variable theories where as this result rules out a large class of nonlocal hidden variable theories. It can also be combined with Legett's earlier work ruling out a separate class of nonlocal theories.

How does Bohmian mechanics, which is a nonlocal hidden variable theory whose predictions exactly match those of QM, avoid the limitations imposed by these results?

 

[Kolmo]

How does Bohmian mechanics, which is a nonlocal hidden variable theory whose predictions exactly match those of QM

Do you have a reference for where that is proved? I know it is equivalent for position measurements, but I've never seen the fully general proof. I'd just need to see it to make sure I'm answering correctly.

 

[PeterDonis]

Do you have a reference for where that is proved?

The math of Bohmian mechanics is equivalent to that of standard QM; that has been known since the 1950s. Since the math is equivalent, it must make the same predictions.

 

[Demystifier]

Do you have a reference for where that is proved? I know it is equivalent for position measurements, but I've never seen the fully general proof. I'd just need to see it to make sure I'm answering correctly.

The trick is that Bohmian mechanics uses the idea that all measurements can be reduced to position measurements, or more precisely, to positions of macroscopic pointers. See e.g. the paper linked in my signature below.

 

[Kolmo]

The math of Bohmian mechanics is equivalent to that of standard QM; that has been known since the 1950s. Since the math is equivalent, it must make the same predictions.

But where is the proof of this is what I am asking? There used to be arguments in the literature that it didn't give the same results for measurements of quantities outside of the basis chosen by the hidden variables. In standard Bohmian Mechanics that would be the position basis, but there are other "Bohmian" Mechanics if one chooses other quantities. Also that it had issues with multi-time correlations.

The mathematical equations are quite different so I'd like to see a proof of the equivalence, I don't think it's true that "it is known" since it was debated at the time and there is a long argument it is not equivalent in R.F. Streater's book "Lost Causes in and beyond Theoretical Physics". I'm not saying Streater is right, just asking for where it was proven.

 

[Demystifier]

Also that it had issues with multi-time correlations.

Multi-time correlations predicted by Bohmian mechanics are also in agreement with standard QM. See Appendix I of F. Laloe, Do We Really Understand QM?

 

[Demystifier]

But where is the proof of this is what I am asking?

See also
https://www.physicsforums.com/threa...es-from-bohmian-mechanics.977585/post-6235376
and
https://www.physicsforums.com/threa...es-from-bohmian-mechanics.977585/post-6236895

 

[Kolmo]

1) P. Holland, The Quantum Theory of Motion, Chapter 8.

Okay regardless of whether that is a full equivalence proof it does give what I need to answer the question.

This new inequality rules out nonlocal theories where the particles can interact nonlocally but always possesses well defined values for observable quantities such as momentum and spin. In Bohmian Mechanics spin is a contextual result of the interaction of the device with the particle, not a property inherent to the particle.

So we might say the current result proves that particles do not intrinsically possesses most of the properties we measure them to have. Up until now this would have been "interpretational", now it is a proven fact.

 

[Demystifier]

Okay regardless of whether that is a full equivalence proof it does give what I need to answer the question.

This new inequality rules out nonlocal theories where the particles can interact nonlocally but always possesses well defined values for observable quantities such as momentum and spin. In Bohmian Mechanics spin is a contextual result of the interaction of the device with the particle, not a property inherent to the particle.

So we might say the current result proves that particles do not intrinsically possesses most of the properties we measure them to have. Up until now this would have been "interpretational", now it is a proven fact.

I don't think that's correct. In Bohmian mechanics, a particle always has both position and momentum. But momentum is not always distributed according to the Born rule, the Born rule for momentum is contextual. Similarly, in the Holland's book you will find Bohmian models in which a particle always has a well defined spin in all 3 directions, but again those values are not always distributed according to the Born rule.

You might revise your statement by saying that we now have a proof that the Born rule is contextual for most observables, but I think we already knew that from older theorems.

 

[Kolmo]

I don't think that's correct. In Bohmian mechanics, a particle always has both position and momentum

Sorry I meant our observed values are not those of the actual particle, i.e. observed spin and momentum are an emergent property of the particle-device set up, we're not just reading off the actual particle momentum for example.

You might revise your statement by saying that we now have a proof that the Born rule is contextual

It's not quite the same as Kochen-Specker contextuality so I wouldn't use that phrase. Also the paper is just looking at correlation-topes, it's not specific to QM so the Born rule as a QM specific thing is not as such a characterization of it.

 

[Kolmo]

Let me try to say it another way. This basically rules out theories were there is a nonlocal interaction between the particles but the device readouts are either directly reading out the properties of the particles or contextual to the device local to the particle.

In essence even nonlocal interactions between the particles combined with contextual values of the measurement results gives a bound which quantum theory violates. You need the values measured in each location to be contextual to both devices.

This is not the same Kochen Specker contextuality which is local and doesn't directly discuss the Born rule since it's about regions in correlation space in general, not about specific features of quantum mechanics.

 

[Sunil]

I don't think that's correct. In Bohmian mechanics, a particle always has both position and momentum.

No. It has both position and velocity. There is a trajectory $q(t)\in Q$ and so there is also a velocity $v(t)= \dot{q}(t)\in Q$. Momentum is contextual - the result of a momentum measurement depends on the position of the measurement device too, thus, is not a property of the particle. There is, in particular, no trajectory $p(t)$. One should be careful to distinguish momentum p and $mv$.

 

[Demystifier]

No. It has both position and velocity. There is a trajectory $q(t)\in Q$ and so there is also a velocity $v(t)= \dot{q}(t)\in Q$. Momentum is contextual - the result of a momentum measurement depends on the position of the measurement device too, thus, is not a property of the particle. There is, in particular, no trajectory $p(t)$. One should be careful to distinguish momentum p and $mv$.

In the sense in which momentum is contextual, in that sense the velocity $v\equiv m^{-1}p$ is also contextual. And by the way, even the measured position, that is the position of the macroscopic pointer of the apparatus, may differ from the Bohmian position of the measured microscopic particle; the authors who discovered it called it surreal Bohmian trajectories. So in that sense even position can be considered contextual in Bohmian mechanics.

 

[Sunil]

In the sense in which momentum is contextual, in that sense the velocity $v\equiv m^{-1}p$ is also contextual.

No. Because conceptually there is no such equivalence. Momentum is defined in dependence of the Lagrangian, $p_i = \frac{\delta S}{\delta \dot{q}}$. A Lagrangian depends on the context, it is contextual (in this sense) already in classical physics.
And, again, there is a Bohmian trajectory $q(t)$ and correspondingly $\dot{q}(t)$ but no such trajectory $p(t)$.

And by the way, even the measured position, that is the position of the macroscopic pointer of the apparatus, may differ from the Bohmian position of the measured microscopic particle; the authors who discovered it called it surreal Bohmian trajectories. So in that sense even position can be considered contextual in Bohmian mechanics.

First, the measured position is irrelevant, the trajectory $q(t)$ is the real position.

Then, "surreal Bohmian trajectories" are nothing but a subjective feeling of some people who expect the trajectories to look like classical trajectories. They simply failed to develop the correct Bohmian intuitions. That's also quite irrelevant.

The correct intuitions about Bohmian trajectories are not that strange. First, they look much less strange if considered only as average trajectories. In stable energy eigenstates there is, in the average, no motion. The classical motion is circular. The average velocity is zero, which is natural. That it should be exactly zero looks slightly strange. Then, let's look at superpositional states. There, the trajectory will be as if the position would have been measured. And that will be a quite classical trajectory. Only if in one part you make a measurement of something different from position, then even the far away other particle will do some surrealistic things too. If that measurement will be a measurement in the sense of the results made macroscopic or not does not matter.

 

[Demystifier]

I've just had a discussion with two authors (Wiseman and Cavalcanti) of the paper. They explained me that violation of their "local friendliness" would imply that the purely instrumental interpretation a'la "Quantum theory needs no interpretation" https://pubs.aip.org/physicstoday/article/53/3/70/411209/Quantum-Theory-Needs-No-Interpretation is wrong. In other words, the experiment could in principle rule out the interpretation that most practical physicists, including @vanhees71, seem to take for granted. But pure instrumentalists don't need to bother yet, because actual realization of such experiments, involving fine manipulations of quantum properties of the measuring apparatus, would be terribly difficult.

 

[vanhees71]

Which paper are you talking about? The Physics today article by Fuchs and Peres tells it all: All you need is the minimal interpretation. If you find an experiment which clearly and reproducibly contradicts the (probabilistic) prediction of QT, then QT is wrong and has to be substituted by a better theory, but not before that happens.

Whether or not QT or another "better" theory satisfies the one or other philosophical prejudice about how Nature must be because of some "ism" is not part of science.

 

[Demystifier]

Which paper are you talking about?

The one referred to in the first post of this thread. It discusses a thought experiment of the Wigner-friend type, which manipulates the quantum state of the measuring apparatus and concludes that, if quantum mechanics is applicable to the measuring apparatus as to any other quantum object, then the minimal purely instrumental interpretation of QM is wrong. But the actual experiment has not been performed(*), and probably will not be in a near future, so a pure instrumentalist, like you, does not need to worry.

(*) An experiment has been performed in which a true measuring apparatus is replaced with a single qubit, obviously because a single qubit is much easier to manipulate in a way needed for the experiment.

 

[Demystifier]

Whether or not QT or another "better" theory satisfies the one or other philosophical prejudice about how Nature must be because of some "ism" is not part of science.

How about minimal-ism and instrumental-ism, are they philosophical prejudices too?
Remarkably, the thought experiment is supposed to rule out precisely those ism's.

 

[martinbn]

@Demystifier why are you so sure of the outcome of the future experiment? May be it will not rule out but confirm the instrumentalist point of view?

 

[Demystifier]

@Demystifier why are you so sure of the outcome of the future experiment? May be it will not rule out but confirm the instrumentalist point of view?

Sure, it's also a possibility. In fact, this result would be much more interesting.

 

[Demystifier]

If you find an experiment which clearly and reproducibly contradicts the (probabilistic) prediction of QT, then QT is wrong and has to be substituted by a better theory, but not before that happens.

The experiment by itself would not say what a better theory is, but it would tell us that at least one of the 3 common sense assumptions in post #1 must be wrong.