Exploring von Neumann's View of Quantum Measurement

[Nicky]

I have a question about quantum entanglement experiments, such as the two-photon "delayed choice" experiment performed by Aspect et al. http://prola.aps.org/abstract/PRL/v49/i25/p1804_1 . Can anyone estimate how much time elapses between the arrival of a single photon at the detector, and the initiation of a current or voltage that could be considered "macroscopic"? I know too little about photomultipliers and the like to have any idea.

In other words, does the experimental setup rule out the possibility that detector states become entangled with detected photon states, so that the "wavefunction collapse" is actually much later than the arrival of the photon at the detector? I am wondering if the time it takes for the detector to settle is long enough that a timelike signal could pass from one detector to another.

Thanks in advance if anyone can shed light on this ... er .. so to speak.

 

[vanesch]

I have a question about quantum entanglement experiments, such as the two-photon "delayed choice" experiment performed by Aspect et al. http://prola.aps.org/abstract/PRL/v49/i25/p1804_1 . Can anyone estimate how much time elapses between the arrival of a single photon at the detector, and the initiation of a current or voltage that could be considered "macroscopic"? I know too little about photomultipliers and the like to have any idea.

Photomultipliers usually have response times in the order of a few nanoseconds...

In other words, does the experimental setup rule out the possibility that detector states become entangled with detected photon states, so that the "wavefunction collapse" is actually much later than the arrival of the photon at the detector?

Ha, that's a nice idea :-))) Especially if the "detector" is the person looking at the results of the correlations ;-)

cheers,
Patrick.

 

[Nicky]

photomultiplier

Photomultipliers usually have response times in the order of a few nanoseconds...

Hmm ... how much current does the photomultiplier produce at the end of that response time? Is it just a few electrons per millisecond or Avagadro's number of them?

 

[James Jackson]

Depends how many initial photons go in and what the photosensitive bit's made of!

 

[Nicky]

Depends how many initial photons go in and what the photosensitive bit's made of!

Only one initial photon goes in. Let's say the photosensor is made of silicon, then how many excited electrons are flowing per second after the response time has elapsed?

 

[DrChinese]

In other words, does the experimental setup rule out the possibility that detector states become entangled with detected photon states, so that the "wavefunction collapse" is actually much later than the arrival of the photon at the detector? I am wondering if the time it takes for the detector to settle is long enough that a timelike signal could pass from one detector to another.

The Aspect experiment was intended to compensate for this. Over time, many improvements have been made to the process. The current state of the art is much more rigorous. Using fiber optics, distances are much longer and the time varying elements are more sophisticated. Thus the locality issue you describe is ruled out. Please reference:

The 1998 Innsbruck Experiment (EPR with 1 kilometer of separation):
http://arxiv.org/PS_cache/quant-ph/pdf/9810/9810080.pdf
by Weihs, Jennewein, Simon, Weinfurter and Zeilinger

 

[vanesch]

Hmm ... how much current does the photomultiplier produce at the end of that response time? Is it just a few electrons per millisecond or Avagadro's number of them?

You get a short current pulse which integrates, to say, a few tens of femtocoulomb (say, 100000 electrons). That's good enough to be seen with a charge-sensitive amplifier. The pulse itself takes a few nanoseconds, and during that time, currents of the order of a few microamperes flow from the last anode.

cheers,
Patrick.

 

[Nicky]

The Aspect experiment was intended to compensate for this. Over time, many improvements have been made to the process. The current state of the art is much more rigorous. Using fiber optics, distances are much longer and the time varying elements are more sophisticated. Thus the locality issue you describe is ruled out. Please reference:

The 1998 Innsbruck Experiment (EPR with 1 kilometer of separation):
http://arxiv.org/PS_cache/quant-ph/pdf/9810/9810080.pdf
by Weihs, Jennewein, Simon, Weinfurter and Zeilinger

Thanks for the reference. Yes, that does seem to settle the locality issue.

 

[Sherlock]

Thanks for the reference. Yes, that does seem to settle the locality issue.

It solves part of the issue. That is, *if* A and B are causally
affecting each other, then these causal influences must be
travelling faster than light. We can be pretty sure of that.

What we can't be sure of yet is whether or not A and B
are causally affecting each other.

Bell-type analyses show that the current state of the
art of descriptive physics is quantitatively inadequate.
Quantum theory isn't descriptive physics. So, there's
no qualitative understanding of how the correlations are
produced. They might be due to local interactions or
they might be due to superluminal interactions. Nobody
knows.

The question of whether or not nonlocal causality
is a fact of nature remains unanswered.

 

[vanesch]

It solves part of the issue. That is, *if* A and B are causally
affecting each other, then these causal influences must be
travelling faster than light. We can be pretty sure of that.

What we can't be sure of yet is whether or not A and B
are causally affecting each other.

Bell-type analyses show that the current state of the
art of descriptive physics is quantitatively inadequate.
Quantum theory isn't descriptive physics. So, there's
no qualitative understanding of how the correlations are
produced. They might be due to local interactions or
they might be due to superluminal interactions. Nobody
knows.

The question of whether or not nonlocal causality
is a fact of nature remains unanswered.

This is correct. The problem with these situations is that they are in a "twilight zone": on one hand the Bell conditions are violated. But on the other hand, there is no way to have an information transfer about the *choice* of polarizer of A by B. If there were such a transfer (that means, that B, by purely looking at his data, can find out what was the polarizer setting of A) then for sure there was a faster-than-light causal influence. But there is no such information transfer (B cannot find out what was the polarizer setting at A), and it can be shown that such transfer is impossible in quantum theory. You can only find out that there was a peculiar correlation by *bringing together* the data from both sides. And that leaves the possibility for locality to be still valid, depending on what is your view on quantum theory.

As Sherlock said, the safest attitude is to say that nobody knows if locality holds or not as a fundamental principle.

cheers,
Patrick.

 

[Nicky]

It solves part of the issue. That is, *if* A and B are causally
affecting each other, then these causal influences must be
travelling faster than light. ...

Has there been any attempt to measure the speed of such faster-than-light influences? It should at least be possible to find a lower bound for that speed, one would think.

 

[DrChinese]

Has there been any attempt to measure the speed of such faster-than-light influences? It should at least be possible to find a lower bound for that speed, one would think.

As I read the reference (Weihs et al), Figure 1, the lower bound was about 10c. That is assuming there is a non-local effect. The usual interpretation is that distance is not a factor, it is always "instantaneous".

 

[vanesch]

Has there been any attempt to measure the speed of such faster-than-light influences? It should at least be possible to find a lower bound for that speed, one would think.

You can't, really. As I said, the reason is that you only see the correlations, once you've brought together the data from both sides, using classical, slower-than-light communication.

If you mean, can you measure "the speed with which one can do measurements at both sides", then this leaves me wondering what you are talking about. For instance, let us place ourselves in a non-Bell context, with classical correlations. Imagine the usual game: I have a white ball and a red ball, and put randomly one in a grey bag, and the other one in a green bag. The grey bag is sent to Tokyo, the green bag is sent to London. We decide that at 12 AM GMT, both bags will be opened and looked at. So now we have a "correlation" between both results. What is the speed at which this correlation is established ?
In a similar way, what does it mean for the "correlation to propagate between both measurements" ? I can do (in a certain reference frame) the measurement at A slightly before, or slightly after, the measurement at B. In another reference frame (using relativity), I can inverse the order in which the measurements occur. In all these cases, the results are the same. So how are you going to attach a "speed" to this "propagation of correlation" ??

cheers,
Patrick.

 

[vanesch]

As I read the reference (Weihs et al), Figure 1, the lower bound was about 10c. That is assuming there is a non-local effect.

What exactly was that ? I wonder what it can mean...

cheers,
Patrick.

 

[Nicky]

You can't, really. As I said, the reason is that you only see the correlations, once you've brought together the data from both sides, using classical, slower-than-light communication.

If you mean, can you measure "the speed with which one can do measurements at both sides", then this leaves me wondering what you are talking about. ... what does it mean for the "correlation to propagate between both measurements" ? I can do (in a certain reference frame) the measurement at A slightly before, or slightly after, the measurement at B. In another reference frame (using relativity), I can inverse the order in which the measurements occur. In all these cases, the results are the same. So how are you going to attach a "speed" to this "propagation of correlation" ??

In these two-photon experiments, the polarizers and detectors are at rest with respect to one another, or nearly so. That is the reference frame I'm referring to when talking about the speed at which correlation hypothetically propagates. Of course, that speed will be measured differently by observers moving with respect to the apparatus, as you indicated. If this "propagation" point of view is valid, correlations of the EPR type would decrease or disappear as the distance between measurements increases.

 

[DrChinese]

What exactly was that ? I wonder what it can mean...

cheers,
Patrick.

I think the idea is: IF there were a causal effect that simply transmitted from A to B telling the polarization to comply with... then at what speed does that causal effect travel?

In Bohmian mechanics, which attempts to insert non-locality explicitly, I don't think there is any limit to the speed of propagation of the correlation effects. If there WERE some such effect, we know that it must be able to propagate at 10 times the speed of light or more. That is per my reading of Weihs, since they specify that the Einstein light cone could have been a tenth the actual size and locality would have still been respected.

Again, this is not standard interpretation of what is happening. You could just as easily say the purported causal correlation effect travels backward in time too - what speed is that?

 

[vanesch]

Again, this is not standard interpretation of what is happening. You could just as easily say the purported causal correlation effect travels backward in time too - what speed is that?

Yes, that is what I had in mind ! It is sufficient to cut 50 cm off one optical fiber or another, and you CHANGE THE ORDER in which things happen, so how can you reasonably define a speed ? Or do they simply take the duration of the two measurements (say, 3 ns) and the distance between the two measurements, and calculate a "speed" from the ratio ?

cheers,
Patrick.

 

[ppnl2]

What exactly was that ? I wonder what it can mean...

cheers,
Patrick.

I may be wrong but I think the idea is to assume there is a FTL signal causeing the correlation. Then you look at the timeing of the measurements and calculate how fast such a signal would have to be to cause the correlation. It does not change the fact that you still have to look at both ends to see the correlation in the first place. And as far as I know most physicists don't believe in any such signal anyway.

 

[Nicky]

I think the idea is: IF there were a causal effect that simply transmitted from A to B telling the polarization to comply with... then at what speed does that causal effect travel? ... You could just as easily say the purported causal correlation effect travels backward in time too - what speed is that?

It's true that spacelike-separated events don't have a definite time ordering. However, it may be that EPR-type correlations are only allowed across certain spacelike intervals, and not others. Suppose there exists a reference frame in which all allowed correlations appear to propagate forward in time. In that case, the width of the cone enclosing all the corresponding vectors in Minkowski space can be viewed as the "speed" of the correlation signals.

Of course, that implies the existence of a preferred reference frame, at least for quantum correlation phenomena, which is not the current philisophical fashion.

 

[vanesch]

It's true that spacelike-separated events don't have a definite time ordering. However, it may be that EPR-type correlations are only allowed across certain spacelike intervals, and not others. Suppose there exists a reference frame in which all allowed correlations appear to propagate forward in time. In that case, the width of the cone enclosing all the corresponding vectors in Minkowski space can be viewed as the "speed" of the correlation signals.

Yes, but even without relativistic considerations, I have the following problem with trying to define a speed of propagation of any influence. Imagine an EPR setup in which the two particles are sent over long optical fibers, one end arrives at Alice, and the other at Bob. Now, Bob's fiber is slightly longer, so Alice measures "first" and Bob measures on average say 0.5 ns later. So we could then define a "speed" of the distance D between Bob and Alice divided by 0.5 ns. But now Bob shifts his photomultiplier 10 cm (0.3 ns) closer, by removing some piece of optical fiber. So now the speed will be something like D / 0.2 ns. Bob moves again his photomultiplier 10 cm closer: this time, Bob clicks first on average... so the speed is then D / (-0.1 ns) ?? No, because now suddenly the influence goes from Bob to Alice... So the speed is D/0.1ns but in the other direction. Given the detection time in a PM (if that's considered the "measurement process" whatever that may mean), some events will be going Alice-> Bob, others will be going Bob-> Alice and some will be damn close to equal times. How do you define a speed in this situation ?

cheers,
Patrick.

 

[Nicky]

Yes, but even without relativistic considerations, I have the following problem with trying to define a speed of propagation of any influence. Imagine an EPR setup in which the two particles are sent over long optical fibers, one end arrives at Alice, and the other at Bob. Now, Bob's fiber is slightly longer, so Alice measures "first" and Bob measures on average say 0.5 ns later. So we could then define a "speed" of the distance D between Bob and Alice divided by 0.5 ns. But now Bob shifts his photomultiplier 10 cm (0.3 ns) closer, by removing some piece of optical fiber. So now the speed will be something like D / 0.2 ns. Bob moves again his photomultiplier 10 cm closer: this time, Bob clicks first on average... so the speed is then D / (-0.1 ns) ?? No, because now suddenly the influence goes from Bob to Alice... So the speed is D/0.1ns but in the other direction. Given the detection time in a PM (if that's considered the "measurement process" whatever that may mean), some events will be going Alice-> Bob, others will be going Bob-> Alice and some will be damn close to equal times. How do you define a speed in this situation ?

You have defined the speed already (D/T), though I would think of it as the absolute value |D/T|. Yes, you can change the direction of the supposed "signal propagation" by choosing whether Bob or Alice performs the measurement first; the fact that the Bob->Alice and Alice->Bob results are indistinguishable only shows the symmetry of Bob's and Alice's views of the experiment.

The interesting result would be to find that for a large enough value of |D/T|, the correlation effect disappears, i.e. the Bell Inequality is no longer violated, or perhaps to find that the threshold value of |D/T| is anisotropic with respect to the direction of the Bob-Alice axis. This would tend to support a "superluminal signal" view of wavefunction collapse.

No such effect has been seen yet, but has it been positively ruled out?

 

[DrChinese]

... This would tend to support a "superluminal signal" view of wavefunction collapse.

No such effect has been seen yet, but has it been positively ruled out?

Technically, no. As noted, we are at 10c (or maybe -10c) and we can expect that to rise over time as distances increase. I don't expect us to ever see anything happen in this regard - i.e. a finite interval.

This does allow for some interesting speculation as to how to describe what is happening. In my opinion, there is something fundamental about the act of observation: if there are no hidden variables, then the photon polarization was not determinate prior to the observation. So what happens at the time of observation to explain the results?

a) We have Vanesch's favorite, Many Worlds. There is branching.
b) Perhaps in some rolled up unseen spatial dimension, the photons are actually not separated by any distance at all. Thus there is no superluminal effect because there is no distance to traverse.
c) Perhaps there is in fact a new superluminal force carrier - "chanceons".
d) Perhaps (my favorite speculation) the observation of one photon causes an effect to propagate to the past at the speed of light, and then change direction so it travels forward in time to the other photon. This would exactly trace a light cone that is consistent with observed results. Of course, I have no idea how or why such behavior occurs sometimes but not others, and this is a completely ad hoc idea. The combo of traveling both forward and backward in time allows arbitarily large distances to be traversed in arbitrary time intervals. Which is more or less what appears to happen anyway :)

 

[Nicky]

This does allow for some interesting speculation ...

a) We have Vanesch's favorite, Many Worlds. There is branching.

From what little I know about the EPR problem, I would tend to agree with Vanesch that Many Worlds is correct, but only if no evidence of superluminal propagation can be found.

b) Perhaps in some rolled up unseen spatial dimension, the photons are actually not separated by any distance at all. Thus there is no superluminal effect because there is no distance to traverse.

I would think (b) implies a superluminal effect after all. It's the effective speed of the signal in the lab frame that matters, not the speed in the particle's own frame.

c) Perhaps there is in fact a new superluminal force carrier - "chanceons".

This is my favorite explanation for EPR, except that the "chanceons" wouldn't carry any energy -- they can only select between energetically degenerate states. Otherwise you'd get superluminal transmission of energy and/or information, leading to causal paradoxes.

d) Perhaps (my favorite speculation) the observation of one photon causes an effect to propagate to the past at the speed of light, and then change direction so it travels forward in time to the other photon. ...

Interesting idea ... it does have some attributes of "Many Worlds" though, since the past which receives the backward-propagated effect isn't the same past that you started out with. Would it be "Many Histories"?

 

[straycat]

Has there been any attempt to measure the speed of such faster-than-light influences? It should at least be possible to find a lower bound for that speed, one would think.

There is a group headed by Gisin who have performed these types of experiments, and I have one reference [1] (4 years old now) that sets a lower bound at 2/3 * 10^7 c (!). Basically they did an EPR type experiment with entangled photons sent via optical fiber network to two villages near Geneva, with the source smack dab in the center. One detector is set spinning at some high angular velocity so that the frames of reference of each detector are not the same, and such that *each* detector, in its own frame, is the first to do the measurement! Pretty cool, eh?

I agree with some of the other comments that there is no actual "speed" because there is nothing actually being transferred from A to B. And I agree with vanesch that the MWI is conceptually the simplest way to understand what's going on here.

David


PS Patrick - wanted to make sure you saw my post on the Born and MWI thread ;)


[1] Zbinden et al. Experimental test of non-local quantum correlation in relativistic configurations. quant-ph/0007009
See also Gisin, Scarini, Tittel, and Zbinden quant-ph/0009055

 

[vanesch]

You have defined the speed already (D/T), though I would think of it as the absolute value |D/T|. Yes, you can change the direction of the supposed "signal propagation" by choosing whether Bob or Alice performs the measurement first; the fact that the Bob->Alice and Alice->Bob results are indistinguishable only shows the symmetry of Bob's and Alice's views of the experiment.

Yes, I understand that. What I wanted to say is that certain events in the sample will be almost perfectly synchronized, making the speed arbitrary high.

The interesting result would be to find that for a large enough value of |D/T|, the correlation effect disappears, i.e. the Bell Inequality is no longer violated, or perhaps to find that the threshold value of |D/T| is anisotropic with respect to the direction of the Bob-Alice axis. This would tend to support a "superluminal signal" view of wavefunction collapse.

I have difficulties with that idea, in that brick walls, rivers, tunnels etc... don't seem to do anything.

No such effect has been seen yet, but has it been positively ruled out?

My personal theory is that beyond a spacelike separation of 70802 lightyears, the effect suddenly disappears But I can't get funding for an experimental verification...

cheers,
Patrick.

 

[Nicky]

Yes, I understand that. What I wanted to say is that certain events in the sample will be almost perfectly synchronized, making the speed arbitrary high.

Arbitrarily high speed does not contradict the superluminal hypothesis, unless it is also coupled with arbitrariness of direction. The suggestion has been made that all EPR-type signals are instantaneous relative to a single, preferred frame of reference (see quant-ph/0110160). If the laboratory is at rest relative to the preferred frame, or nearly so, measured speeds would be arbitrarily high.

I have difficulties with that idea [disappearance of EPR-like effects for some spacelike intervals], in that brick walls, rivers, tunnels etc... don't seem to do anything.

I don't understand the connection with macroscopic objects. What do you mean?

My personal theory is that beyond a spacelike separation of 70802 lightyears, the effect suddenly disappears But I can't get funding for an experimental verification...

Hey, if you are patient enough to wait 70802 years for the results to reach you, you deserve funding!

 

[vanesch]

I don't understand the connection with macroscopic objects. What do you mean?

I meant that no attenuation is seen in the effect, whether it is on the same optical table in the lab, or through fibers going to the other side of the town, crossing rivers etc...
If something had to propagate, you'd think that it matters somehow if that something is going to go through a lot of stuff or not. You can argue that neutrinos wouldn't be hampered either, but: 1) you'd have a 1/r^2 effect if the chanceons are emitted in a sphere and 2) don't forget that the chanceons have to interact with a photon to change its polarization !
So whatever it is that "propagates" it is going to be something real weird. A bit like the strange mechanical eather in which EM waves had to vibrate in the 19th century.

cheers,
Patrick.

 

[vanesch]

One detector is set spinning at some high angular velocity so that the frames of reference of each detector are not the same, and such that *each* detector, in its own frame, is the first to do the measurement! Pretty cool, eh?

Yup ! Didn't know about that one !

PS Patrick - wanted to make sure you saw my post on the Born and MWI thread ;)

Oops, must have overlooked it ! I'll have a look...

cheers,
Patrick.

 

[DrChinese]

I meant that no attenuation is seen in the effect, whether it is on the same optical table in the lab, or through fibers going to the other side of the town, crossing rivers etc...
If something had to propagate, you'd think that it matters somehow if that something is going to go through a lot of stuff or not. You can argue that neutrinos wouldn't be hampered either, but: 1) you'd have a 1/r^2 effect if the chanceons are emitted in a sphere and 2) don't forget that the chanceons have to interact with a photon to change its polarization !
So whatever it is that "propagates" it is going to be something real weird. A bit like the strange mechanical eather in which EM waves had to vibrate in the 19th century.

cheers,
Patrick.

I agree... IF there is some superluminal effect being broadcast... how does it know to only show up alongside the two entangled photons and nothing else? And it is not blocked by intervening objects?

Or maybe there isn't anything like this... :)

 

[Nicky]

I meant that no attenuation is seen in the effect, whether it is on the same optical table in the lab, or through fibers going to the other side of the town, crossing rivers etc...
If something had to propagate, you'd think that it matters somehow if that something is going to go through a lot of stuff or not.

Whatever it is that propagates faster-than-light (if anything does) must have zero scattering cross section with any state of any particle, except the one state it is meant to cancel. Otherwise there would be superluminal energy transfer with the particle from which it scatters, which is causally forbidden.

You can argue that neutrinos wouldn't be hampered either, but: 1) you'd have a 1/r^2 effect if the chanceons are emitted in a sphere and 2) don't forget that the chanceons have to interact with a photon to change its polarization !

It can't be neutrinos, since they carry energy and lepton quantum numbers with them, and hence classical information.

If there is a "chanceon", imagine its wavefunction. Presumably the chanceon sees very low potential near the photon detection events, and very high potential everywhere else, so the wavefunction is two small, dense "dots" with very tiny magnitude everywhere else. It's more like it's tunneling from one detector to the other, rather than propagating through space like a wave.

So whatever it is that "propagates" it is going to be something real weird. A bit like the strange mechanical eather in which EM waves had to vibrate in the 19th century.

Well we already know nature is weird ... it's just a question of what flavor of weirdness is out there. Hopefully there will be many more experiments that help to answer the question.