Murray Gell-Mann on Entanglement

[Ken G]

If it's just a matter of Alice updating her knowledge of Bob's situation, then I would think that would mean that Bob had 0% chance before Alice's measurement, even if Alice didn't know that. Which to me implies that Bob's result was predetermined, at least for that particular measurement choice, which is sort of a hidden-variables conclusion.

For me, the resolution of this is to get away from the idea that "there is a probability" of something happening. Instead, simply treat the purpose of a probability to be an assessment based on your knowledge. I cannot think of a single physical situation where there actually "is a probability" of something happening in some absolute sense (that isn't trivially 1 or 0)-- can you? I wager that any example you give there, I could show how you are simply connecting a set of assumptions with a set of possible outcomes based on those assumptions-- in short, you will always be talking about information. I think this is an important point, even in classical situations like playing with a deck of cards-- there never is any such thing as "the probability I will get a straight flush", there is only what I know about that deck (or think I know), and how I assess my chances in the long run. It's true that a classical deck supports a concept of "how the cards lie" prior to the deal, but the fact that the player never uses that concept shows that's not what they need probabilities for. So Alice "has a probability," and Bob "has a probability," and that's it.

 

[secur]

So this argument never ends!

Fortunately it can end for you, or me: just ignore the whole thing, and do something productive instead.

Another approach is to remember that these are interpretations. That means we can't decide which, if any, is right. (If we could they wouldn't be interpretations any more: one would be "physics", the other "wrong".) The proper approach then is to use whichever suits your purpose for a given situation. Wait for new discoveries which will allow a decision. More pro-actively, try to think of experiments which could decide.

Either collapse (in its naive form) contradicts the theory itself or it is an at best empty but at worst usually misleading phrase. That's why I'd prefer not to use it at all when talking about QT.

The phrase "at best empty but at worst usually misleading" means precisely: I don't like that interpretation. But other people do. The resolution: don't use any phrase you don't like. When others do, just translate it to the interpretation you do like.

Consider a parallel situation: two people are credited with one theorem. This happened often during the Cold War: Soviets said their scientist ("A") discovered it, while the West said their guy, "B", did. So one side called it A's theorem, the other B's. Made no difference scientifically but a big difference politically. Now, we used to have conferences where the two sides met for co-operative discussions. The scientists didn't care, but couldn't comfortably call it by the other's name, because their politicians would send them to Siberia, or cut their DARPA funding. The resolution was easy. We agreed to let each side call it as they wished. There was no confusion, each knew exactly what the other meant. It became an in-joke, and actually enhanced collegiality.

Recommend you do the same with these interpretations.

So collapse is unnecessary because vanhees can do his calculations without it. Why do you insist on it then?

The calculations can always be done without any interpretation. But people like to have an intuitive picture to go along with their math. Few, if any, really "shut up and calculate". It's reasonable that atyy, or anyone, insist they're allowed their favorite interpretation. But don't insist the other guy has to use it too! Let each use whatever language they're comfortable with. I admit it might get a bit confusing, but surely it's better than endless argument, or Siberia.

For me, the resolution of this is to get away from the idea that "there is a probability" of something happening. Instead, simply treat the purpose of a probability to be an assessment based on your knowledge. I cannot think of a single physical situation where there actually "is a probability" of something happening in some absolute sense (that isn't trivially 1 or 0)-- can you? ... So Alice "has a probability," and Bob "has a probability," and that's it.

True in classical physics, but for QM it's not so clear. You're advocating the "minimal statistical interpretation", a.k.a. "minimal ensemble interpretation". Perhaps we can call it "minimal ensemble statistical interpretation" (MESI). Other interpretations of QM disagree. They say QM probabilities are absolute: not merely describing our limited information but truly inherent in just one instance.

Can QM really be interpreted the MESI way? I can think of a couple objections, and would like to hear what MESI proponents think of these. One, covalent bonds. It seems that superposition of orbits - in one single molecule - is essential. The other, quantum computing. When we have a bunch of qbits in the typical Bell state, the probabilities (50/50) must be essentially present in each one. You can't say half are "really" in one state, the others in the other state, we just don't know which are which. Quantum computing won't work at all with that model - it seems. Please let me know if I'm wrong about these objections.

 

[Jilang]

For me, the resolution of this is to get away from the idea that "there is a probability" of something happening. Instead, simply treat the purpose of a probability to be an assessment based on your knowledge. I cannot think of a single physical situation where there actually "is a probability" of something happening in some absolute sense (that isn't trivially 1 or 0)-- can you? I wager that any example you give there, I could show how you are simply connecting a set of assumptions with a set of possible outcomes based on those assumptions-- in short, you will always be talking about information. I think this is an important point, even in classical situations like playing with a deck of cards-- there never is any such thing as "the probability I will get a straight flush", there is only what I know about that deck (or think I know), and how I assess my chances in the long run. It's true that a classical deck supports a concept of "how the cards lie" prior to the deal, but the fact that the player never uses that concept shows that's not what they need probabilities for. So Alice "has a probability," and Bob "has a probability," and that's it.

Radioactive decay comes to mind. How would knowledge enter into that?

 

[Ken G]

Can QM really be interpreted the MESI way? I can think of a couple objections, and would like to hear what MESI proponents think of these. One, covalent bonds. It seems that superposition of orbits - in one single molecule - is essential.

I don't have any objection to the concept of superposition, it is a form of information too. We have information about the state, and that allows us to predict what will happen-- the information includes interference effects. A minimal ensemble interpretation does not require we say the state is either one or the other, and we just don't know which, it says we have some information and we do some mathematics and make a prediction that involves the concept of superposition.

 

[Ken G]

Radioactive decay comes to mind. How would knowledge enter into that?

Even with radioactive decay, information plays a role. Let's say the setup is at t=0 it has been established that an unstable nucleus has come into being. That's a form of information right there, but let's say we regard that as a fact of nature, and look at the probability a decay will occur between t=t_o and t=t_o+dt. Of course we agree that probability is e^-t_o/T dt/T, and we will say that is also the probability of a decay in that interval after time t_o has elapsed if we have no other information. But someone else who has the information that no decay has occurred for time t_o will reassess that probability as just dt/T. So there, even with radioactive decay, we have two different physicists with different information who will assess two different probabilities, and both will test their probabilities over many repetitions of the same situation, and both will find that their probability worked perfectly. So in both those situations, we see there is not "the probability" that the decay occurs in a given interval, there is the probability of that based on what you already know has or has not happened, and that's different for different people, but it works just like a probability for either one.

 

[secur]

I don't have any objection to the concept of superposition, it is a form of information too. We have information about the state, and that allows us to predict what will happen-- the information includes interference effects. A minimal ensemble interpretation does not require we say the state is either one or the other, and we just don't know which, it says we have some information and we do some mathematics and make a prediction that involves the concept of superposition.

We agree entirely on the physics, then - which is all that really matters. (Although I still wonder if other proponents of MEI, or MSI, would agree with your stance.) However I'm puzzled by your terminology. We know the covalent bond works by observing one single molecule. The fact that it doesn't fall apart requires the exchange interaction. This isn't an ionic bond which can be explained without recourse to superposed states. So - what does the word "ensemble" signify here?

 

[Ken G]

We agree entirely on the physics, then - which is all that really matters. (Although I still wonder if other proponents of MEI, or MSI, would agree with your stance.) However I'm puzzled by your terminology. We know the covalent bond works by observing one single molecule. The fact that it doesn't fall apart requires the exchange interaction. This isn't an ionic bond which can be explained without recourse to superposed states. So - what does the word "ensemble" signify here?

The concept of "ensemble" in a superposition is simply you can have a lot of copies of the superposition, and that information allows you to predict the behavior of the ensemble.

 

[secur]

The concept of "ensemble" in a superposition is simply you can have a lot of copies of the superposition, and that information allows you to predict the behavior of the ensemble.

We're predicting a single molecule's behavior. Statistical averaging over an ensemble of molecules is irrelevant, since the probability that the constituent atoms remain bonded is 1 (normal conditions). Nevertheless it's a "QM phenomenon", i.e., QM is required to explain it.

 

[Ken G]

We're predicting a single molecule's behavior. Statistical averaging over an ensemble of molecules is irrelevant, since the probability that the constituent atoms remain bonded is 1 (normal conditions). Nevertheless it's a "QM phenomenon", i.e., QM is required to explain it.

It doesn't matter what the probability is, that is only relevant to the size of the ensemble you will need to demonstrate the effectiveness of the approach. The key point is, in practice, physics works like this: information-->prediction-->testing. So if you just take that at face value, that's all you need-- you regard the probabilities you use in the "prediction" phase to be a simple function of the information you have and the laws you apply. Different information, different prediction, but it's all the same physics, and that's all we ever test. It's the minimal approach-- you simply never need to assert anything you don't have direct evidence for.

 

[bhobba]

That's a fair statement. But it really is interpretation dependent. And a lot of physicists don't really get tangled up in the question anyway.

Exactly.

As with so many things in QM its interpretation dependent.

Even knowing what the formalism says is very difficult which is why studying interpretations is quite interesting. You may think, for example, that at first sight QM is random, but we have interpretations in QM where it isn't (eg BM) so great care is needed in deciphering what QM says.

I have said it before, and will say it again, I think a much better starting point to understand QM is the following:
http://www.scottaaronson.com/democritus/lec9.html

The key issue of interpretations is exploring what those 'negative' probabilities are saying and what it means.

Strangely a lot of it is simply an aggrumet about the meaning of prpbrability:
http://math.ucr.edu/home/baez/bayes.html

Thanks
Bill

 

[bhobba]

In all due respect to a physics giant, I think that Gell-Mann's definitive statement that measurement of one particle in EPR has no effect on the other particle is going beyond what we understand about quantum mechanics.

Yes - but its for a lay audience. I think a bit of latitude is reasonable.

Thanks
Bill

 

[bhobba]

In response to Dr. Chinese, I thought that question that physicists don't want to get tangled up in is the most important question of entanglement, i.e.spooky action at a distance: How can measurement of for example spin of one particle affect instantaneously the spin of a very distant particle?

They don't want to get tangled up in a going nowhere philosophical analysis of it - they leave that up to philosophers.

What the great physicist Bell did was lift its beyond that - and that most definitely is where physicists come into it because it is subject to experimental testing.

Thanks
Bill

 

[Ken G]

Exactly.

As with so many things in QM its interpretation dependent.

Even knowing what the formalism says is very difficult which is why studying interpretations is quite interesting. You may think, for example, that at first sight QM is random, but we have interpretations in QM where it isn't (eg BM) so great care is needed in deciphering what QM says.

I have said it before, and will say it again, I think a much better starting point to understand QM is the following:
http://www.scottaaronson.com/democritus/lec9.html

The key issue of interpretations is exploring what those 'negative' probabilities are saying and what it means.

Strangely a lot of it is simply an aggrumet about the meaning of prpbrability:
http://math.ucr.edu/home/baez/bayes.html

Thanks
Bill

I finally bookmarked both of those, glad you reposted them.

 

[Mister T]

However, it would be equally right to say that measuring one photon does affect the other photon, since a measurement collapses the wave function of both photons.

Not necessarily. One could argue that you don't affect the other photon. Instead all you affect is the result of a measurement. According to the interpretation of QM that includes wave function collapse the property you measure is not a property that the particle possesses.

 

[Ken G]

The reason I agree with Gell-Mann is that I feel in physics we should have a standard for the word "effect" that is different from what a philosopher might use. To say we have an "effect", we must be able to demonstrate causation, not merely correlation. Entanglement is an example of the old adage "correlation is not causation", because causation requires an arrow that is not present in a correlation. So the physicist is always agnostic about causation until it is demonstrated as such-- it is never necessary to demonstrate the absence of causation, it is necessary to demonstrate its presence by means that go beyond correlation.

 

[secur]

... it's the minimal approach ...

Right. Not the minimal ensemble approach. Having dropped that unnecessary and misleading term, your approach becomes just shut up and calculate. No one can argue with that.

 

[secur]

I agree with Gell-Mann ... So the physicist is always agnostic about causation until it is demonstrated as such ...

You're right. the physicist is - should be - agnostic on this and similar issues. At this time we don't know if there's any causation in the Bell experiment. There's no point in talking about it, until some new data is available.

But Gell-Mann said: "People say loosely ,crudely, wrongly that when you measure one of the photons it does something to the other one. It doesn't."

He's not agnostic. He's strongly denying any causative link. Neither of us can agree with that statement.

BTW it doesn't matter what he did or didn't say, and whether we agree - except, after all, that's what the OP asked.

 

[Ken G]

Right. Not the minimal ensemble approach. Having dropped that unnecessary and misleading term, your approach becomes just shut up and calculate. No one can argue with that.

I don't actually like "shut up and calculate", because it suggests that all that matters is the outcome of the calculation. I think more than that matters, that the scientist must always keep careful track of what they are testing and what they are simply assuming. It's OK to make assumptions, but they go in a different box from what has been demonstrated by testing. So there is an important philosophical component, but it is the philosophy of scientific thinking.

 

[Ken G]

But Gell-Mann said: "People say loosely ,crudely, wrongly that when you measure one of the photons it does something to the other one. It doesn't."

Yes, you're right that Gell-Mann is going beyond agnosticism. I should have said I would have agreed with him had he simply said "people say that when you measure one of the photons it does something to the other, but in fact we have no need to imagine that is true, and no scientific evidence that it is true."

 

[ShayanJ]

Collapse is needed for the consistency of quantum mechanics (in the Schroedinger picture).

If you do calculations in one frame in which collapse is not needed, the collapse will be needed to achieve the same prediction in a different frame, assuming you use the Schroedinger picture.

So collapse preserves the principle of relativity: any frame is as good as any other.

But vanhees's description is frame independent and it doesn't need collapse. So I'm confused by your statement!
What criteria should a frame meet so that we don't need collapse to explain Bell type experiments in it?

 

[atyy]

But vanhees's description is frame independent and it doesn't need collapse. So I'm confused by your statement!
What criteria should a frame meet so that we don't need collapse to explain Bell type experiments in it?

In my understanding, for a Bell test, since the measurements are at spacelike separation, there is one frame in which A and B measure simultaneously. Since there is only one measurement in that frame, collapse is not required in that frame.

By relativity of simultaneity, if A and B measure simultaneously in one frame, they must measure sequentially in another frame. Thus in another frame one would have A measuring first, followed by B measuring. In that frame, the collapse is needed to specify the state of the system after A has measured.

 

[ShayanJ]

In my understanding, for a Bell test, since the measurements are at spacelike separation, there is one frame in which A and B measure simultaneously. Since there is only one measurement in that frame, collapse is not required in that frame.

By relativity of simultaneity, if A and B measure simultaneously in one frame, they must measure sequentially in another frame. Thus in another frame one would have A measuring first, followed by B measuring. In that frame, the collapse is needed to specify the state of the system after A has measured.

That makes sense for QFT. But that means NRQM doesn't need collapse because Galilean transformations preserve simultaneity of the experiments.

 

[atyy]

That makes sense for QFT. But that means NRQM doesn't need collapse because Galilean transformations preserve simultaneity of the experiments.

If you perform sequential measurements in NRQM, and use the Schroedinger picture, you will still need collapse.

 

[ddd123]

The reason I agree with Gell-Mann is that I feel in physics we should have a standard for the word "effect" that is different from what a philosopher might use. To say we have an "effect", we must be able to demonstrate causation, not merely correlation. Entanglement is an example of the old adage "correlation is not causation", because causation requires an arrow that is not present in a correlation. So the physicist is always agnostic about causation until it is demonstrated as such-- it is never necessary to demonstrate the absence of causation, it is necessary to demonstrate its presence by means that go beyond correlation.

How do you prove that there is causation?

 

[ShayanJ]

If you perform sequential measurements in NRQM, and use the Schroedinger picture, you will still need collapse.

If collapse is actually in the theory, its existence shouldn't depend on what picture we use. So if collapse is there in the Schrodinger picture, it should have a counterpart in the Heisenberg picture, some kind of an evolution for operators that doesn't satisfy the Heisenberg's equation of motion. Otherwise we can just stop using Schrodinger picture and then there is no collapse in the theory!

But otherwise, what you say makes sense to me!

 

[atyy]

If collapse is actually in the theory, its existence shouldn't depend on what picture we use. So if collapse is there in the Schrodinger picture, it should have a counterpart in the Heisenberg picture, some kind of an evolution for operators that doesn't satisfy the Heisenberg's equation of motion. Otherwise we can just stop using Schrodinger picture and then there is no collapse in the theory!

But otherwise, what you say makes sense to me!

Yes, you can hide the collapse by going in a sophisticated way to the Heisenberg picture - this requires a generalization of the Born rule. I have no problem with that.

There are other ways to avoid collapse, like insisting on never making sequential measurements (in principle it is possible, but almost impossible in practice).

Similarly, Bob can avoid nonlocality by insisting that Alice is not real at spacelike separation.

Many choices are possible, including accepting that the locality can be derived from nonlocality - concretely, the reduced density matrix of B (showing locality) is derived by tracing over the collapsed wave function of both A and B (showing nonlocality).

 

[ShayanJ]

this requires a generalization of the Born rule

Can you provide a reference?

 

[atyy]

Can you provide a reference?

This is not the most general form, but it will give you the right idea: Eq 37 of http://arxiv.org/abs/quant-ph/0209123.

 

[Ken G]

How do you prove that there is causation?

The scientist never proves anything, they use models successfully. Causation is something that we say is present when we find that by using that concept, it gives us power over the situation. We control one thing, in order to control something else-- that's the value of the causation concept. But like all our concepts, we tend to take them too literally, and try to apply them in situations where we gain nothing by doing so. Like entanglement.

 

[vanhees71]

The collapse is nonlocal in the sense that the wave function is assigned to a spacelike surface of simultaneity, and the wavefunction on that hypersurface collapses instantaneously.

From the nonlocal collapse, the reduced density matrix of B can be derived, from which it can be seen that the collapse does not allow superluminal signalling.

So locality can be derived from nonlocality, and nonlocality does not contradict locality.

I disagree, and that's also not in accordance with what Peres writes in the here discussed article. Again, you have to distinguish between longranged-correlations ("nonlocality" realized by entanglement also in relativistic QFT) and local interactions (realized by microcausality of the local observables and locality of the interaction Hamiltonian, as also clearly specified by Peres; he gives even a stronger argument, why relativistic QT should be realized as local relativistic QFT, then Weinberg in QT of Fields vol. I!).

 

[vanhees71]

That makes sense for QFT. But that means NRQM doesn't need collapse because Galilean transformations preserve simultaneity of the experiments.

In other words: Nothing is wrong with even a naive collapse assumption for non-relativistic QT. There you can use it without contradicting the theory itself. That's not possible for relativistic local QFT (as applied in the formulation of the Standard Model). There a naive collapse assumption contradicts the very foundations of the theory. Since NRQT is just an approximation of relativistic QFT, one shouldn't use the collapse assumption there either, but at least it's not self-contradictory as if applied for relativistic QFT.

 

[vanhees71]

This is not the most general form, but it will give you the right idea: Eq 37 of http://arxiv.org/abs/quant-ph/0209123.

Why do you say there is an extension of the Born rule? Also note that the outcome of anything physical like the said probabilities are independent of the picture of time evolution since any two pictures of time evolution are connected by unitary transformations. It may be more or less convenient to use a specific picture, but there cannot be any difference concerning the physical outcomes of the formalism due to the change of the picture.

 

[Weddgyr]

"People say loosely ,crudely,wrongly that when you measure one of the photons it does something to the other one. It doesn't."
Do most physicists working in this field agree with the above statement ?

Most physicists are trained to avoid asking the question. It's been a hugely successful program.

There aren't any answers as of today. You can assume a non-local influence if you like, but that will be philosophically in conflict with relativity, etc. You can assume a conspiracy/super-determinism of detector settings, etc, which will be another philosophical muddle. You can just accept the quantum correlations as is, without need for a causal mechanism, but you will then be in conflict with the general philosophy of the scientific method. Above all you are a physicist so whichever option you pick you will of course NOT be doing any of that philosophical crap.

 

[Simon Phoenix]

That's not possible for relativistic local QFT (as applied in the formulation of the Standard Model). There a naive collapse assumption contradicts the very foundations of the theory

OK - that's fair enough - but does assuming a naïve collapse model actually lead one to derive any inconsistent experimental results?

 

[ShayanJ]

In other words: Nothing is wrong with even a naive collapse assumption for non-relativistic QT. There you can use it without contradicting the theory itself. That's not possible for relativistic local QFT (as applied in the formulation of the Standard Model). There a naive collapse assumption contradicts the very foundations of the theory. Since NRQT is just an approximation of relativistic QFT, one shouldn't use the collapse assumption there either, but at least it's not self-contradictory as if applied for relativistic QFT.

I'm trying to make sense of the way you think about this. But I seem to lack some essential knowledge about how people like you actually use QM.
The part I know is that you prepare a large number of identical systems in identical quantum states. Of course the preparation device is not perfect and there may be some deviations from the desired state, but I don't know whether you take that into account and use a mixed state for the ensemble or just make the approximation of a perfect device and use the desired pure state.
Anyway, the next step is you measure the probability distribution of a desired observable on this ensemble. What I don't understand is, what do you do if you want to measure the probability distribution of another observable on the same ensemble. Would you assume its in the same state before the first measurement? Or you do a Bayesian update? Is it even possible to do a second measurement on the same ensemble?