EPR Debate: Nature Agrees with Einstein

[DrChinese]

Here are Dehmelt's own words:
"As the man who first found a way to catch an electron, bring it to rest in free space for a year and measure its magnetism in order to study its structure, I naturally would like to tell you a little bit about it. At the University of Gottingen, my teacher Richard Becker, in one of his lectures, drew a dot on the blackboard declaring 'Here is an electron ...' or rather 'Hier ist ein Elektron ...'. This appeared to be drastically at odds with the famous Physicist Heisenberg's claim that an electron truly at rest could not be localized and could be found anywhere in space."

All the best
John B.

In the presence of a troll? That would be someone stirs up the pot with no intention of producing any meaningful discussion.

The title of this thread="A fresh look at EPR". Above, we have a quote by one Nobel winner commenting on another. Is there any point being made here? What, that science marches forward? How zzzzzzzzzzzzzzzzzzzzz interesting.

 

[RandallB]

REF post #22

Dr Chinese

Allow me to modify my analogy to make it as fair a comparison to testing spins, polarizations, etc. at diff angles etc. Let me know if my analogy makes any unfair or ‘unrealistic’ comparisons or assumptions.

Instead of having Alice and Bob ask questions they will only test for male or female coming in their area on their side of the resort. We have very high confidence that only couples of male and female depart the resort and that they separate in opposite directions. We also have very high confidence that only females wear perfume and males wear cologne (useful only to Chris and Dan as only they can test for smell). Chris test locale is near Alice while Dan’s locale is near Bob. During testing Alice and Chris are Space-Like Separated from Bob and Dan. A constraint on testing is that only one test can be performed in a locale.
Testing reports are always given as a “+ “ for Female going by Alice’s area, whether observed directly or implied by the local observation. Therefore Dan directly detecting perfume would imply that Bob would have seen a Female, implying Carol would see a Male - therefore Dan report’s the result as a “-“. Thus correlation is when same signs are reported. Non-correlation is when conflicting signs are reported.

Now to your items:
1) Yes very simple 4 permutations for Alice (A) and Bob (B)
Only one question here: For 2 and 3 (A+B- & A-B+) the actual results are showing near Zero results for those permutations correct? How many non-correlation results are acceptable at this level of testing to consider the tests themselves viable?

2) LR ‘local realistic’ was this introduced by Bell ? No I’m not sure I see this as “imminently reasonable”. One issue of concern, C cannot conduct all his own tests, when A is also testing D must stand in. Is running a test as in #1 and getting Zero non-correlations good enough to consider this substitution reasonable? Just looking to know in general how Bell or the testers – no need to explain the real tests.

3) The 8 permutations are clear and that statically they must total 100%. From a testing view I’m a little unclear as to how to interpret a “non-correlation”.
Even on an individual result, is there a level of these that we should start to question the viability of out tests being used?

4) Proceeding with the above results being “acceptable”
………………………CORR NON-CORR
A&B [1,2,7,8] [3,4,5,6] (C unknown)
A&C [1,3,6,8] [2,4,5,7] (B unknown)
B&C [1,4,5,8] [2,3,6,7] (A unknown)

A&C(Corr) + A&B(Non) – B&C(Corr) = [1,3,6,8] + [3,4,5,6] - [1,4,5,8]
Result = [3] [3] [6] [6]

Yes I agree with the statistical math here, it only requires that there are acceptable “Non-correlations”.

5) and 6) Yes, valid sampling of valid tests should be able to have their stat % calculated in this manner. And should give %’s that are Zero or greater.

7) Now maybe this is where part of my confusion comes in. The use of an LR is based on a Non-QM assumption that Alice or Bob in their testing are NOT affecting the results. Is this why LR should be useful? And when tests are performed and give a Negative % of probability for ANY permutation provides us with some evidence that LR is not valid and implies QM is correct because the testing is influencing the results??
If this is the point – I guess I can see the point, IF non-correlation results are useful.
What do testers consider is going on when the get non-correlation results back at item one when there is only 4 permutations? Are they considered valid?


ON TO QM.

8) Sure QM says once an out come is seen it is known in both test areas “At the same time”. Although I’d expect QM to agree that it cannot define which test was first or ‘controlling’ since the tests were space-like separated. Thus even if the test are not space-like separated but only time or distance separated, nether test would be “First” nor “Controlling” according to QM. Do I understand this part of QM correctly?

9) I don’t know what this means – since QM is rejecting the use of a LR, how could it be able to predict anything based on an LR?
Or are you saying that in this particular type of optical EPR test, that the new angle test is flawed based on QM. Flawed such that QM already knows that the test by Chris for perfume won’t work because QM knows that in fact some women actually do wear cologne while the Non-QM view would not recognize that?

I hope this is detailed enough to show what areas I need help in.
Appreciate your time in helping me.

Thanks RB

 

[DrChinese]

Dr Chinese

Allow me to modify my analogy to make it as fair a comparison to testing spins, polarizations, etc. at diff angles etc. Let me know if my analogy makes any unfair or ‘unrealistic’ comparisons or assumptions.

Instead of having Alice and Bob ask questions they will only test for male or female coming in their area on their side of the resort. We have very high confidence that only couples of male and female depart the resort and that they separate in opposite directions. We also have very high confidence that only females wear perfume and males wear cologne (useful only to Chris and Dan as only they can test for smell). Chris test locale is near Alice while Dan’s locale is near Bob. During testing Alice and Chris are Space-Like Separated from Bob and Dan. A constraint on testing is that only one test can be performed in a locale.
Testing reports are always given as a “+ “ for Female going by Alice’s area, whether observed directly or implied by the local observation. Therefore Dan directly detecting perfume would imply that Bob would have seen a Female, implying Carol would see a Male - therefore Dan report’s the result as a “-“. Thus correlation is when same signs are reported. Non-correlation is when conflicting signs are reported.

Now to your items:
1) Yes very simple 4 permutations for Alice (A) and Bob (B)
Only one question here: For 2 and 3 (A+B- & A-B+) the actual results are showing near Zero results for those permutations correct? How many non-correlation results are acceptable at this level of testing to consider the tests themselves viable?

2) LR ‘local realistic’ was this introduced by Bell ? No I’m not sure I see this as “imminently reasonable”. One issue of concern, C cannot conduct all his own tests, when A is also testing D must stand in. Is running a test as in #1 and getting Zero non-correlations good enough to consider this substitution reasonable? Just looking to know in general how Bell or the testers – no need to explain the real tests.

3) The 8 permutations are clear and that statically they must total 100%. From a testing view I’m a little unclear as to how to interpret a “non-correlation”.
Even on an individual result, is there a level of these that we should start to question the viability of out tests being used?

4) Proceeding with the above results being “acceptable”
………………………CORR NON-CORR
A&B [1,2,7,8] [3,4,5,6] (C unknown)
A&C [1,3,6,8] [2,4,5,7] (B unknown)
B&C [1,4,5,8] [2,3,6,7] (A unknown)

A&C(Corr) + A&B(Non) – B&C(Corr) = [1,3,6,8] + [3,4,5,6] - [1,4,5,8]
Result = [3] [3] [6] [6]

Yes I agree with the statistical math here, it only requires that there are acceptable “Non-correlations”.

5) and 6) Yes, valid sampling of valid tests should be able to have their stat % calculated in this manner. And should give %’s that are Zero or greater.

7) Now maybe this is where part of my confusion comes in. The use of an LR is based on a Non-QM assumption that Alice or Bob in their testing are NOT affecting the results. Is this why LR should be useful? And when tests are performed and give a Negative % of probability for ANY permutation provides us with some evidence that LR is not valid and implies QM is correct because the testing is influencing the results??
If this is the point – I guess I can see the point, IF non-correlation results are useful.
What do testers consider is going on when the get non-correlation results back at item one when there is only 4 permutations? Are they considered valid?


ON TO QM.

8) Sure QM says once an out come is seen it is known in both test areas “At the same time”. Although I’d expect QM to agree that it cannot define which test was first or ‘controlling’ since the tests were space-like separated. Thus even if the test are not space-like separated but only time or distance separated, nether test would be “First” nor “Controlling” according to QM. Do I understand this part of QM correctly?

9) I don’t know what this means – since QM is rejecting the use of a LR, how could it be able to predict anything based on an LR?
Or are you saying that in this particular type of optical EPR test, that the new angle test is flawed based on QM. Flawed such that QM already knows that the test by Chris for perfume won’t work because QM knows that in fact some women actually do wear cologne while the Non-QM view would not recognize that?

I hope this is detailed enough to show what areas I need help in.
Appreciate your time in helping me.

Thanks RB

The only thing about your example is that it is a specific case in which the LR and QM and experiment predictions/results all overlap anyway. So it makes it harder to distinguish the scenarios.

1) Not sure, but I would guess well less than 1% in experiments.

2) Not sure I follow the point.

3) Non-correlations - you have them pegged properly in 4). You don't expect any when your special case is considered.

4) We are OK on this too.

5, 6) Ditto.

7) "And when tests are performed and give a Negative % of probability for ANY permutation provides us with some evidence that LR is not valid and implies QM is correct because the testing is influencing the results??"

Not just that testing affects the results, the results at one place appear (when pulled together) to be dependent on the results at another.

Not sure I follow about the "testers consider..." comment.

8) No one knows if A causes B or B causes A, or something else.

9) QM is not based on LR, and classical optics does not match LR. The pre-Bell mistake was that people ASSUMED that classical optics and LR were the same, and Bell shows they cannot be the same. Classical optics being essentially classical experimental results, which match the predictions of QM.

a. If you accept Bell, then QM and LR must yield different predictions (at least at certain specific angles).
b. If QM is wrong, then classical optics is wrong too (since the predictions of the two are identical).
c. For LR to be right, then existing experiments (Aspect et all) must be flawed in some way.

 

[Caroline Thompson]

... 9) QM is not based on LR, and classical optics does not match LR. The pre-Bell mistake was that people ASSUMED that classical optics and LR were the same, and Bell shows they cannot be the same. Classical optics being essentially classical experimental results, which match the predictions of QM.

I'm not the only person in this forum to think you've got this wrong! I'm afraid I forget who else pointed out, with me, that the classical prediction does not support QM. Classical optics is not incompatible with local realism -- in fact, it would be very strange if it were!

The mistake you make is in thinking that the classical optics prediction for coincidences between plane polarised "photons" in an experiment such as Aspect's obey the same cos^2(angle) law as QM. The law in fact obeyed, if you make all the idealised classical optical assumptions, together with (a) equal probability for all angles and (b) detection probabilities proportional to beam intensities, is not a cos^2 law but cos^2 + constant. The constant makes all the difference! It means that whereas the QM prediction gives curves with visibility 1, classical optics gives (in the ideal case) ones with visibility less than 1 (in fact, as shown in http://arXiv.org/abs/quant-ph/9903066, 0.5).

This constant term is the reason that local realists do not find it easy to produce models to fit the actual results. In order to fit them, there is no escaping the necessity to depart from the ideal case and consider the likely behaviour of real instruments.

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[DrChinese]

I'm not the only person in this forum to think you've got this wrong! I'm afraid I forget who else pointed out, with me, that the classical prediction does not support QM. Classical optics is not incompatible with local realism -- in fact, it would be very strange if it were!

The mistake you make is in thinking that the classical optics prediction for coincidences between plane polarised "photons" in an experiment such as Aspect's obey the same cos^2(angle) law as QM. The law in fact obeyed, if you make all the idealised classical optical assumptions, together with (a) equal probability for all angles and (b) detection probabilities proportional to beam intensities, is not a cos^2 law but cos^2 + constant. The constant makes all the difference! It means that whereas the QM prediction gives curves with visibility 1, classical optics gives (in the ideal case) ones with visibility less than 1 (in fact, as shown in http://arXiv.org/abs/quant-ph/9903066, 0.5).

This constant term is the reason that local realists do not find it easy to produce models to fit the actual results. In order to fit them, there is no escaping the necessity to depart from the ideal case and consider the likely behaviour of real instruments.

Caroline
http://freespace.virgin.net/ch.thompson1/

First, what is classical optics to QM anyway? It is like Newtonian gravity to Einsteinian gravity. It is ultimately not that critical that QM coincide with classical optics other than in cases in which correspondence would be expected.

Second, if you can get classical optics to yield a different prediction for Aspect-like correlation percentages, go for it.

None of the above affects the fact that all extant tests support the predictions of QM, or that the QM formula cos^2 is the same applied in classical optics in cases in which polarized light is passed through a second polarizer.

As to who originally pointed out that LR and classical optics were at odds, I see that as being Bell. Before Bell, the LR prediction was cos^2 matching classical optics. After Bell, the LR predicted value changed (to something that no LRist is willing to stand behind).

 

[Caroline Thompson]

... if you can get classical optics to yield a different prediction for Aspect-like correlation percentages, go for it.

See Appendix C of my paper at http://arxiv.org/abs/quant-ph/9903066. The basic LR (classical optics) prediction is in fact very close to Aspect's raw data, not adjusted by subtraction of accidentals, at least in his first experiment. It differs markedly from the quantum theoretical prediction and from the published (adjusted) data, but, as you may have realized, I and other local realists challenge the legitimacy of the subtraction. As far as I know, even quantum theorists (in particular, Wolfgang Tittel and his team at Geneva) now agree with me.

None of the above affects the fact that all extant tests support the predictions of QM, or that the QM formula cos^2 is the same applied in classical optics in cases in which polarized light is passed through a second polarizer.

Right, but this is for two polarisers "in series". In the Bell test experiments we have two polarisers "in parallel". It is only in the very special case when all the light is polarised in the same direction that classical optics and QM agree.

As to who originally pointed out that LR and classical optics were at odds, I see that as being Bell. Before Bell, the LR prediction was cos^2 matching classical optics. After Bell, the LR predicted value changed (to something that no LRist is willing to stand behind).

No DrChinese, I'm afraid you've got this all wrong! I don't know what makes you think classical optics is not a LR theory. Bell originally talked only of spin-1/2 particles, not light, so naturally the local realist model he considered was slightly different from the one applicable to light. He assumed, for instance, that when a particle is detected, it must be + or - (though later he added the possibility of zero). It could never be "both + and -". In classical optics, both + and - simultaneously is perfectly possible.

It might help if you checked out my wikipedia page on hidden variable theories: http://en.wikipedia.org/wiki/Local_hidden_variable_theory

The basic LR theory for Bell's spin-1/2 particles yields a zig-zag prediction, whilst the basic one for an experiment using polarised light yields a sine curve but shifted upwards, so that it no longer has full visibility.

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[DrChinese]

1. Right, but this is for two polarisers "in series". In the Bell test experiments we have two polarisers "in parallel". It is only in the very special case when all the light is polarised in the same direction that classical optics and QM agree.

2. No DrChinese, I'm afraid you've got this all wrong! I don't know what makes you think classical optics is not a LR theory. Bell originally talked only of spin-1/2 particles, not light, so naturally the local realist model he considered was slightly different from the one applicable to light. He assumed, for instance, that when a particle is detected, it must be + or - (though later he added the possibility of zero). It could never be "both + and -". In classical optics, both + and - simultaneously is perfectly possible.

1. In EPR, we use one entangled particle (say the Right) to gain knowledge of the other (say the Left \theta). A measurement of the Right tells us it is polarized at 0 or 90 degrees relative to an arbitrary angle we shall call 0 degrees. Using classical optics, I then use the knowledge I have to predict the likelihood of detection at the other polarizer when the polarization is known:

P(+,0)=cos^2(\theta)
P(-,0)=sin^(\theta)
P(+,90)=cos^2(90-\theta)
P(-,90)=sin^(90-\theta)

Assuming the light is randomly polarized initially so the Right side odds are 50-50, we get:

P(correlation)=P(+,0)+P(-,90)=.5cos^2(\theta)+.5sin^(90-\theta)=cos^2(\theta)

which is also the QM prediction. For \theta=22.5 degrees, P(correlated)=.8536.

If you care to make a case that this is not how the rest of the world looks at this problem, by all means, go ahead. To me, it looks like dismissive hand-waving on your part.

2. We know the classical correlation probability of .8536 is what the LR position USED to predict before Bell. After Bell, the new LR value for \theta=22.5 degrees, P(correlated)=.7500. I don't think you will find this value in a lot of old books on Optics. And for that matter, you won't see it in many places today either. In fact, I doubt there are very many people who know where that value even comes from! But the fact is that this is about the highest value you can have for an LR prediction of correlation at 22.5 degrees that is consistent with Bell.

Again, if you can use classical optics to yield ANY other predicted value than the one I have given of .8536, then by all means, go ahead. Just make sure the result is .7500 or less so it can match your LR prediction, and then your point will be made. I have seen your charts on possible values per LR. Simply tell us which one is your prediction and then derive it from classical optics.

(As to your idea that the LR model predicts both + and - cases simultaneously... hey, you are the one that is talking to us about what is REASONABLE and INTUITIVE. And that doesn't sound so reasonable to me. In fact, it sounds like QM weirdness if you think about it )

 

[Caroline Thompson]

1. In EPR, we use one entangled particle (say the Right) to gain knowledge of the other (say the Left \theta). A measurement of the Right tells us it is polarized at 0 or 90 degrees relative to an arbitrary angle we shall call 0 degrees.

Sorry, but I'm afraid you've already gone wrong! A single measurement of polarisation does not, in classical theory, tell you what the input polarisation was.

The classical formula, given that the input polarisation was in a random direction, is:

P_{AB}(a, b) = <br /> \int^{\pi}_0 \frac{d\lambda}{\pi} \cos^2(\lambda - a)\cos^2(\lambda - b)

which evaluates to

\frac{1}{8} + \frac{1}{4} \cos^2 \theta

where \theta = b - a.

The logic behind this is well known, even to quantum theorists, and reproduced in many papers, including the one I keep referencing, http://arXiv.org/abs/quant-ph/9903066.

... Assuming the light is randomly polarized initially so the Right side odds are 50-50, we get:

P(correlation)=P(+,0)+P(-,90)=.5cos^2(\theta)+.5sin^(90-\theta)=cos^2(\theta)

which is also the QM prediction.

No, this is not how to deal with random polarisation directions in classical optics. It's some kind of compromise, influenced by QM thinking and the photon idea. In classical optics the initial beam has a definite polarisation direction, \lambda, i.e. it has the very "hidden variable" assumed in all (reasonable) local realist models for optical Bell tests. This is commonly known, reproduced in many texts other than mine.

If you care to make a case that this is not how the rest of the world looks at this problem, by all means, go ahead. To me, it looks like dismissive hand-waving on your part.

I'm afraid that, to me, you algebra looks like a quantum-theoretical fudge!

I have seen your charts on possible values per LR ...

But I have never given any! If you are using anything out of my Chaotic Ball papers then you are using it incorrectly. As I've said before, it illustrates the principle only. As I've also said before (and as you might now better understand, with the aid of the equation above), all that is needed to convert the basic, idealised, local realist (optical) Bell test prediction into one that can have higher "visibility" is to alter the assumed shapes of the functions involved, allowing for the true operating characteristics of the apparatus concerned. There is no need for the model to predict values anywhere near as high as 0.75. The high figure you quote is after "normalisation".

(As to your idea that the LR model predicts both + and - cases simultaneously... hey, you are the one that is talking to us about what is REASONABLE and INTUITIVE. And that doesn't sound so reasonable to me. In fact, it sounds like QM weirdness if you think about it )

Hmmm ... Well now, possibly (after seeing the equation) you will by now see for yourself why both + and - is perfectly natural, but just in case you don't, consider what happens if you input light polarised at 45 deg to the axes of a 2-channel polariser.

Suppose the initial intensity is I. The intensity in both outputs will then (under the usual assumption of a 50-50 split) be I/2. Assuming "perfect" detectors, giving a count every time for a signal of intensity I and for signals of lesser intensity a count with probability proportional to that intensity, the detector for the + channel has a probability 0.5 of a count and likewise, independently, the - channel. The probability of both at once is 0.25.

Is that so very weird?

You will have noticed, incidentally, that the above depends on a number of assumptions, none of which are absolutely necessary. You can get different answers under different conditions. As has been shown experimentally, when you have very weak light (in QM considered to be "single photon" level), the individual pulses making up the beam may not split 50-50. That, though, is a separate issue, not critical here.

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[RandallB]

DrChinese, Caroline

OK I think I’ve spotted where I needed the help.
My analogy does not fit the testing. At least not the testing you all are debating.

Just to be sure I understand the actually tests being do on EPR here.
The simple part A&B we expect the Non-Correlation to be near 0%
When we have testing set up for A, B, & C and of course can only test two at a time.
We still expect A&B Non-Correlation to be near 0% - correct?
But using the LR “C” we expect:
A&C tests and also B&C tests would both have a significant % of Non-Correlation! Plus QM and Non-QM both agree on this.

My analogy would not allow for this expectation of Non-coralation!
Test where A sees polarization of 0 or 90 degrees it would expect B tests correlating 100% when testing 90 degrees off of A’s view. And when testing polarization in B’s area at an angle at 22.5 degrees (I assume half of 45 is used for a reason) we should expect some non-correlation from anybodies view of it of what the LR will see.
Therefore the LR from QM and Non-QM views is NOT expected to be in 100 percent correlation with A. My analogy is not able to represent such an idea.

Maybe I can modify it so that it can – to do so; this is what I don’t know:
What are the predicted results for the amount of “Non-Correlation” both from QM and Non-QM views?
Why are the predictions different?

My guess is that using 45 degrees would predict the same amount of correlation (Also Non-correlation) from QM and Non QM views.
So using 22.5 is somehow important – True?

Thanks RB

 

[DrChinese]

...

Maybe I can modify it so that it can – to do so; this is what I don’t know:
What are the predicted results for the amount of “Non-Correlation” both from QM and Non-QM views?
Why are the predictions different?

My guess is that using 45 degrees would predict the same amount of correlation (Also Non-correlation) from QM and Non QM views.
So using 22.5 is somehow important – True?

Thanks RB

Yes, the 22.5 degrees is significant.

As you can see from your example, you are basically using 0 degrees which gives perfect correlations. Guess what? Both the QM and LR interpretations can handle this case equally well. Ditto for 90 degrees. Ditto for 45 degrees (as you guessed above).

For nearly 30 years, no one noticed there was a glaring hole in the logic. Then Bell came along. He noticed that if you use ANY other angles, you start getting significant problems with the LR predictions IF you expected them to yield similar results as QM - as happens for the 0, 45 and 90 degrees cases. It turns out that the deviation hits a maximum at 22.5 degrees and 67.5 degrees. The reason I don't bother with analyzing other cases is that the 22.5 degrees case leads to enough evidence that you can basically ignore all the others.

As a matter of practice, Bell test do actually check a range of angles, more than just the 22.5 degrees case.

 

[RandallB]

OK I feel like I’m getting closer.

Testing at 45 will give us sin & cos .707 squared we get .5

22.5 and 67.5 degrees sin & cos gives .38 or .92 squared gives .146 or .854

So I’m jumping to some conclusions in using the squared.
Pythagoras feels right and more important the numbers look right that way.
At 45 degrees Pythagoras would predict 50-50 correlations vs. non-correlations. And at 22.5 degrees it will be .146 vs. .854 odds of correlations vs. non-correlations.

Just guessing again the above odds are Classic expectations (If that’s the correct way to refer to “Non-QM”). And the other expectation (QM?) agrees with it at 50-50.
BUT something about the way QM is defined forces a prediction that does not agree with the 14.6% vs. 85.4% distribution for 22.5 degrees.
This is the key I’d like to understand!
What is that prediction?
And how and why does it’s math come up with different numbers than above?

With luck I have some errors in thinking here, cuss correcting those is how I’ll learn something.

 

[Caroline Thompson]

... something about the way QM is defined forces a prediction that does not agree with the 14.6% vs. 85.4% distribution for 22.5 degrees.
This is the key I’d like to understand!
What is that prediction?
And how and why does it’s math come up with different numbers than above?

With luck I have some errors in thinking here, cuss correcting those is how I’ll learn something.

As far as I can tell, the QM formula is effectively a lucky guess, the LR one (see my latest message in the thread re negative probabilities) the only logical possibility. I recently copied out Shimony's derivation of the QM formula, for the benefit of wikipedia (see http://en.wikipedia.org/wiki/Quantum_mechanical_Bell_test_prediction). It seemed to me that Malus' Law was just "plucked out of thin air" and assumed to apply to the coincidences. It has only ever been derived experimentally using two polarisers in series, which is logically quite different. [Moreover, it applies to (classical) beam intensities, which will only be proportional to QM's counts if the detectors are ideal ones in that their response probabilities are exactly proportional to the input intensities.]

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[DrChinese]

As far as I can tell, the QM formula is effectively a lucky guess, ...

ROTFL

Funny, I can describe its applicability quite easily:

If the source is randomly polarized, and you know the polarity of any photon on the Right by means of observing its polarization, the probability of any particular polarity being seen on the Left is given by Malus' Law where the angle is the angle in-between.

I have long assumed that the applicability was so obvious that it did not need formal proof in most analyses, it was simply used as a starting point. Even most classical treatments of Bell use the same logic. Correlations are defined by Malus' Law.

 

[RandallB]

Caroline & Dr C
I appreciate any extra input – But before I can even evaluate the credibility of a website or see how Malus' Law defines correlations and evaluate that for myself. I’d be lost if I’m not on the right track. So for the following items:

1) Classical View or LC for Local Causes – Is it fair and recognized to use these terms to describe the Non-QM view? That is in principal local causes control or outcomes are predetermined before elements reach test areas. (LR used for Lorentz Relativity by some) Is there even a preference?

2) On 45 Degrees – Am I on track there?? Does both LC and QM predict 50% - 50% or not? It’s just that when I do look around at some of the formulas I see “(1/2)” being included with sin² – which would give me 25%?
Or is that 25% for each of two permutations out of four which will get me back to 50-50?

3) For 22.5 degrees or 67.5; Am I right is it (1/2) cos² times 2 permutations is giving a .146 vs. .854 odds of correlations vs. non-correlations.
AND WHO is making this prediction LC or QM (I guessed LC, but …..?)

4) Now what is the other prediction, not results the other prediction?
And How is it arrived at? And if it is QM that does not agree with .146 vs. .854; Is this where I will have no choice but to get familiar with Malus?

Can these be answered without needing me to become an optical, polarization expert to understand the answers?

 

[DrChinese]

Caroline & Dr C
I appreciate any extra input – But before I can even evaluate the credibility of a website or see how Malus' Law defines correlations and evaluate that for myself. I’d be lost if I’m not on the right track. So for the following items:

1) Classical View or LC for Local Causes – Is it fair and recognized to use these terms to describe the Non-QM view? That is in principal local causes control or outcomes are predetermined before elements reach test areas. (LR used for Lorentz Relativity by some) Is there even a preference?

2) On 45 Degrees – Am I on track there?? Does both LC and QM predict 50% - 50% or not? It’s just that when I do look around at some of the formulas I see “(1/2)” being included with sin² – which would give me 25%?
Or is that 25% for each of two permutations out of four which will get me back to 50-50?

3) For 22.5 degrees or 67.5; Am I right is it (1/2) cos² times 2 permutations is giving a .146 vs. .854 odds of correlations vs. non-correlations.
AND WHO is making this prediction LC or QM (I guessed LC, but …..?)

4) Now what is the other prediction, not results the other prediction?
And How is it arrived at? And if it is QM that does not agree with .146 vs. .854; Is this where I will have no choice but to get familiar with Malus?

Can these be answered without needing me to become an optical, polarization expert to understand the answers?

2 & 3 - Yes. As to who is making the prediction - it is definitely QM putting forth the .146 and .854 predictions for correlations at 67.5 and 22.5 respectively. Sometimes the LR side tags along, and sometimes it doesn't.

I don't think you have to be an expert to follow, think you are probably following pretty well. Some of it comes down to opinion or wording anyway.

 

[RandallB]

Sometimes the LR side tags along, and sometimes it doesn't.

Sometimes the Classical LR Local realist approach predicts the same thing and sometimes they DON'T! ? If we're talking about 22.5 degrees shouldn't that be a big problem!?

Especially when I made my own predications of what I'd think the results should be I thought I was using classical thinking! Based on using polarized light coming through filter at 22.5 degrees diagonal..

But if my classical prediction is simply the same as the QM prediction what good is the testing if it only confirms BOTH predictions. Unless my prediction is faulty and it can be shown that the LR MUST predicate something else – how is the testing useful at all??
Can it be shown the classical thinking MUST predict something else and not just “Tag along" ??

Reviewing my own ‘prediction’ - I can see in a classical sense I was using ‘Wave’ theory not making a prediction for an individual event(s). Is that enough to disqualify the approach I used to make the 'classical' prediction?? AND does (should) the LR predict??

ALSO, since the QM prediction gives the same 'negative' result in statistical math we talked of earlier - why is that acceptable from QM view?

RB

 

[Nacho]

DrChinese,

Thanks for your post #355

But I need to ask a question:

Aren't you taking the absense of a measurement and relating that to no photon present?

That's the problem I have with Bell Inequalities. It seems to me they are trying to prove/disprove QM, but then they don't hold to the underlying postivism philosophy when attempting to do that. That philosophy says you gain no information from the absense of a measurement, not that you can assign it a definate outcome .. the inequalities assign it a definate outcome .. in this case a negative/no particle found.

 

[DrChinese]

DrChinese,

Thanks for your post #355

But I need to ask a question:

Aren't you taking the absense of a measurement and relating that to no photon present?

That's the problem I have with Bell Inequalities. It seems to me they are trying to prove/disprove QM, but then they don't hold to the underlying postivism philosophy when attempting to do that. That philosophy says you gain no information from the absense of a measurement, not that you can assign it a definate outcome .. the inequalities assign it a definate outcome .. in this case a negative/no particle found.

You have it backwards. Bell sets a requirement for a local realistic theory. There is no requirement for QM, which does not assert that there is the possibility of a more complete specification of the system.

 

[DrChinese]

Sometimes the Classical LR Local realist approach predicts the same thing and sometimes they DON'T! ? If we're talking about 22.5 degrees shouldn't that be a big problem!?

Especially when I made my own predications of what I'd think the results should be I thought I was using classical thinking! Based on using polarized light coming through filter at 22.5 degrees diagonal..

But if my classical prediction is simply the same as the QM prediction what good is the testing if it only confirms BOTH predictions. Unless my prediction is faulty and it can be shown that the LR MUST predicate something else – how is the testing useful at all??
Can it be shown the classical thinking MUST predict something else and not just “Tag along" ??

Reviewing my own ‘prediction’ - I can see in a classical sense I was using ‘Wave’ theory not making a prediction for an individual event(s). Is that enough to disqualify the approach I used to make the 'classical' prediction?? AND does (should) the LR predict??

ALSO, since the QM prediction gives the same 'negative' result in statistical math we talked of earlier - why is that acceptable from QM view?

RB

Yes, sure, you can make any prediction you want to. It can make it match QM or not. It can match experiment or not. The norm is to have the local realistic (LR) predicition match QM so that it will not be ruled out by an experiment that yields experimental results compatible with QM. To date, all published tests of Bell's Inequalities have matched the QM predictions without exception (although there are those such as Caroline who dispute the validity of the results).

I will repeat: QM itself does NOT lead to negative probabilities. LR does if it matches the predictions of QM. These are two completely different scenarios even though they sound like they are the same. Local reality makes assumptions that QM does not.

 

[Nacho]

You have it backwards. Bell sets a requirement for a local realistic theory. There is no requirement for QM, which does not assert that there is the possibility of a more complete specification of the system.

Maybe I have it backwards .. probably I just didn't go far enough in my question/explanation. I'm really referring to conclusions based on them.

I do believe Bell Inequalties could prove or disprove realism. I believe both it and mathematics are based on the same underlying philosophy. Maybe the Aspect type experiments prove the quantum world doesn't follow realism, sans loopholes.

But, what I was trying to get to across was that some people (not necessarily you) take that conclusion and go too far, to say then that the Universe follows postivism rules and that the Universe is non-local. I don't see how that can be done as the formulization of the inequalities neglects tenants of QM (especially the Copenhagen Interpretation).

 

[RandallB]

Yes, sure, you can make any prediction you want to.

I will repeat: QM itself does NOT lead to negative probabilities. LR does if it matches the predictions of QM. These are two completely different scenarios even though they sound like they are the same. Local reality makes assumptions that QM does not.

OK I must be missing something on the assumtions: - before we run a test how can we accept a LR prediction of "anything it wants". We don't need to run a test to reject invalid predictions as we should be able to show a prediction to be invalid, based on the standards of the predicting theroy.
SO predictions are invalid or acceptable based on: -- First let's review:

[1] A+ B+ C+
[2] A+ B+ C-
[3] A+ B- C+
[4] A+ B- C-
[5] A- B+ C+
[6] A- B+ C-
[7] A- B- C+
[8] A- B- C-
Total probability 1.0 100% of 8 permutations

Three tests holding one Variable unknown (??)
Each test 4 permutations for 100%

Predictions: (C using 22.5 Degrees)
Test ?? - Corr - - NonCorr - Predict - BY:
A&B (C) [1,2,7,8] [3,4,5,6] 1.0 0.0 : QM & LR both
A&C (B) [1,3,6,8] [2,4,5,7] .14 .86 : QM ;; LR?
B&C (A) [1,4,5,8] [2,3,6,7] .14 .86 : QM ;; LR?

Statistical Math gives us:
A&B(NonCorr) + A&C(Corr) – B&C(NonCorr) = [1,3,6,8] + [3,4,5,6] - [1,4,5,8]
Result 0.0 + .14 - .86 = [1][3][6][8] + [3][4][5][6] - [1][4][5][8]
[3] [3] [6] [6] = -.72
[3][6]= -.36 or - 36% (Negative probability !)
(I can see where 45 degrees or 0.5 & 0.5 would not reveal this problem)

Before we even consider running a test, we know the LR (Local Realist) cannot predict .14 & .86 due to the negative probability that would predict for permutations [3][6] together! The LR position must agree with the validity of the Statistical Math as being a requirement part of the LR and a prediction of .14 & .86 on its face would be contrary to the LR view and assumtions BECAUSE: ...?

However from the QM view the Statistical Math is not even considered or referred to since the assumption and principal view of QM does not agree that the above Statistical Math would apply to this case at all BECAUSE: ...?

Therefore before running the test we know that the only realistic prediction that the LR based on LR assumtions & view (and such test results validate the LR view) WOULD BE: ...?

 

[DrChinese]

1. Before we even consider running a test, we know the LR (Local Realist) cannot predict .14 & .86 due to the negative probability that would predict for permutations [3][6] together!

2. The LR position must agree with the validity of the Statistical Math as being a requirement part of the LR and a prediction of .14 & .86 on its face would be contrary to the LR view and assumtions BECAUSE: ...?

3. However from the QM view the Statistical Math is not even considered or referred to since the assumption and principal view of QM does not agree that the above Statistical Math would apply to this case at all BECAUSE: ...?

4. Therefore before running the test we know that the only realistic prediction that the LR based on LR assumtions & view (and such test results validate the LR view) WOULD BE: ...?

1. Yes, exactly, this is what Bell pointed out.

2. Because, as you showed, negative probabilities result and that is an un-physical result.

3. Because QM already had the Heisenberg Uncertainty Relations (with its h), and had accepted that there were limits to the description of physical reality. (The realist position is that there is no absolute limit.)

4. LR must predict differing values from QM, since the .14 and .86 are the predictions of QM. So the LR advocate has a problem as now QM and LR must go their separate ways.

Keep in mind that we are in the world of "designer" theories in which theories can be constructed ad hoc to fit various scenarios. Of course, 99.99% of these are totally useless and most of those are a complete waste of time. QM is the .01% that is useful.

So if you (or someone else) want to construct a theory which can be disproven before testing, then there is no law against it other than it wastes time. There are also those who hold out hope that all existing published EPR tests will somehow be thrown out in the future.

 

[Caroline Thompson]

There are also those who hold out hope that all existing published EPR tests will somehow be thrown out in the future.

Yes, for example me, and not without good reason.

I've studied many reports of actual experiments and discovered alternative local realist explanations for all the results. In some experiments it's just a matter of the data having been adjusted in a manner that makes sense under QM but not under local realism. The adjustment increases the Bell test statistics, turning (in the specific instances for which I have the data) results that comply with the inequality into ones that do not. In other experiments, the "detection" or "fair sampling" loophole is open. This loophole crept into the experiments as a result of sleight of hand! It is related to the use of an inequality for which an unbiased estimate demands that each term be divided by the number of emitted pairs. The test used in practice involves division by the number of detected pairs and is simply not valid.

There are other loopholes, different ones (or combinations of them) being relevant in different experiments. See my website or a page I contributed to wikipedia:
http://en.wikipedia.org/wiki/Bell_test_loopholes

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[RandallB]

In some experiments it's just a matter of the data having been adjusted in a manner that makes sense under QM but not under local realism.

Caroline:
DrChinese's comment:
"LR must predict differing values from QM, since the .14 and .86 are the predictions of QM."
Makes no sense to me.
Just because one method/theory (QM) makes a prediction (A legit one based on that theory) cannot be justification for denying any other approach from making exactly the same prediction as long as it's legit within that other theory.
A credible argument as to why LR is wrong should at least define the prediction LR must make (Say for 22.5 degrees), within the definition of LR.
I guess Dr Chinise dosn't know or care as he's in a position to already "know" that even if it agrees with QM it's wrong.

You have at least implied that there are results that if reached during testing would support LR over QM. (Hence your concern over the apparent "adjusted" data favoring QM).

My main question then: in an example like 22.5 degrees, does LR make a prediction different than QM?
What is it ??
Are we talking about a tiny difference like .001% or less that is making this so hard to test?


Personally I wonder about the usefulness of Polarization testing, because it seems to me that a classical view, as I'd see it, would agree with 14% and 85% that QM predicts. But I suspect my "classical view" may not be consistent with the rules for LR. I can deal with that later.
For now I'd be impressed if someone could just explain what the LR prediction actually is for 22.5 degrees.

RB

 

[DrChinese]

Caroline:
DrChinese's comment:
"LR must predict differing values from QM, since the .14 and .86 are the predictions of QM."
Makes no sense to me.
Just because one method/theory (QM) makes a prediction (A legit one based on that theory) cannot be justification for denying any other approach from making exactly the same prediction as long as it's legit within that other theory.
A credible argument as to why LR is wrong should at least define the prediction LR must make (Say for 22.5 degrees), within the definition of LR.
I guess Dr Chinise dosn't know or care as he's in a position to already "know" that even if it agrees with QM it's wrong.

I will repeat: Bell's Theorem (and not DrChinese) precludes LR and QM from making the same predictions.

Also: As I have said before, some LR theories make predictions outside of the Bell Inequality. I have repeatedly asked for other values to discuss. In one of these provided by Caroline, the difference between LR and QM is large: 1.000 for QM vs. .7500 for LR at 0 degrees; .8536 for QM vs. .6768 for LR at 22.5 degrees (correlated cases). However, they agree at 45 degrees.

 

[Caroline Thompson]

I will repeat: Bell's Theorem (and not DrChinese) precludes LR and QM from making the same predictions.

Also: As I have said before, some LR theories make predictions outside of the Bell Inequality. I have repeatedly asked for other values to discuss. In one of these provided by Caroline, the difference between LR and QM is large: 1.000 for QM vs. .7500 for LR at 0 degrees; .8536 for QM vs. .6768 for LR at 22.5 degrees (correlated cases). However, they agree at 45 degrees.

You're just not reading me, DrChinese! The figures are meaningless. The idea that LR predicts just one figure is wrong: LR represents the genuine scientific approach to modelling the real world, and it is absolutely right and proper that it should give a different result for each experiment. The exact experimental conditions vary.

That's one point. The other is that if the experimenters in practice pursue the aim of obtaining as high a visibility to their coincidence curves as they can (by using suitably low beam intensities, for instance, with the nominal aim of obtaining the "single photon level"), then they will land up, if they are lucky, with curves of almost 100% visibility. They will be very nearly exactly sinusoidal. Both LR and QM agree on this, so naturally there is no problem for LR theory in agreeing with your figures and with experiment, so long as we look only at the normalised data, where the normalisation is done with respect to the detected pairs! Had we normalised with respect to the emitted pairs, the figures would have looked very different. All coincidence rates would have been well below the QM prediction.

As far as I know, there has been only one exception to this -- the Rowe et al experiment using trapped ions. Here almost every emission was detected, so the two methods of normalisation coincide. As the whole world recognises, though, the experiment did not satisfy the basic requirements of a Bell test: the ions were very close together, the settings of the detectors may not have been independent, and the measurements made did not even pretend to be independent. [See M Rowe et al, “Experimental violation of a Bell’s inequality with efficient detection”, Nature 409, 791 (2001)]

Caroline
http://freespace.virgin.net/ch.thompson1/

 

[DrChinese]

You're just not reading me, DrChinese! The figures are meaningless. The idea that LR predicts just one figure is wrong: LR represents the genuine scientific approach to modelling the real world, and it is absolutely right and proper that it should give a different result for each experiment. The exact experimental conditions vary.

That's one point. The other is that if the experimenters in practice pursue the aim of obtaining as high a visibility to their coincidence curves as they can (by using suitably low beam intensities, for instance, with the nominal aim of obtaining the "single photon level"), then they will land up, if they are lucky, with curves of almost 100% visibility. They will be very nearly exactly sinusoidal. Both LR and QM agree on this, so naturally there is no problem for LR theory in agreeing with your figures and with experiment, so long as we look only at the normalised data, where the normalisation is done with respect to the detected pairs! Had we normalised with respect to the emitted pairs, the figures would have looked very different. All coincidence rates would have been well below the QM prediction.

I recognize that you postulate added parameters in experimental situations - one being beam intensity. So you should be able to put this in terms of a specific prediction, shouldn't you? I simply ask for a value to compare against the experimental results so we can rule a specific theory as "in" or "out". As best I see, all current LR theories are ruled out by experiment. But perhaps your added parameters would fix everything (although I personally don't see how that is possible).

Assume the added parameter "beam intensity" comes into the equation. You state that QM and your LR theories give identical predictions when this parameter is low. But this would then violate the Bell Inequality, no?

 

[Caroline Thompson]

I recognize that you postulate added parameters in experimental situations - one being beam intensity. So you should be able to put this in terms of a specific prediction, shouldn't you?

I'm afraid not. In the general formula for the local realist prediction we have a function that gives, effectively, the probability of detection as a function of intensity. The exact form of this function depends on the particular make of detector used. The only way I can think of to find this function is to do some supplementary tests using different input intensities and seeing what happens. You have to be careful, though, what physical mechanism you use to vary the intensity! You'd expect one answer if you vary the number of pulses per second, a different one if you vary the strength of each individual pulse. It is mainly the latter that is of interest for Bell tests. The polariser, in the "classical" model, is assumed to reduce the intensity of the indivual pulses.

I simply ask for a value to compare against the experimental results so we can rule a specific theory as "in" or "out". As best I see, all current LR theories are ruled out by experiment. But perhaps your added parameters would fix everything (although I personally don't see how that is possible).

You're absolutely right! Take account of all the messy details and your LR solution will reveal itself.

Assume the added parameter "beam intensity" comes into the equation. You state that QM and your LR theories give identical predictions when this parameter is low. But this would then violate the Bell Inequality, no?

I can't guarantee exact equality of the predicitions, only agreement with experiment. Yes, the results may very well violate an invalid Bell inequality, but they will never violate a valid one.

Caroline

 

[Nereid]

I've read through this thread, perhaps not quite carefully enough, and here's where my head is at at the moment:
- QM can predict the outcome of an experiment (accurately)
- LR can match the outcome of an experiment (accurately)
- QM's predictions apply across a range of experiments, of several different kinds
- LR cannot make 'before the experiment' predictions (the exact details are unknowable?)
- QM's predictions are general, with no ad hoc components
- LR's post-experiment analyses are ad hoc
- both QM and LR analyses of experimental results are consistent with the Bell inequality
- there have been no experiments conducted - to date - which proponents of LR regard as definitive tests of the Bell Inequality
- there have been many experiments conducted - to date - which proponents of QM regard as definitive test of the Bell Inequality
- proponents of LR have not (yet) proposed an experiment (or class of experiment) which, if conducted properly, would yield a good result which would unambiguously distinguish LR from QM
- proponents of QM have proposed many experiments which, if conducted properly, would yield good results which would unambiguously distinguish LR from QM.

 

[Caroline Thompson]

Hi Nereid

Hmmm ... Most of your points seem correct. However, I think the position re proposals for experiments that would in fact discriminate between QM and LR is that local realists (and there are, unfortunately, not many of these who have the courage to stand up and be counted) have from time to time suggested such experiments and been ignored. [The journals have not been helpful re publicising them.]

I think one could prove the matter one way or the other if only a sufficiently comprehensive set of tests were to be done, covering a wide range of parameters.

Caroline
http://freespace.virgin.net/ch.thompson1/