Interesting observations about recent experiments.

Foundations of physics and/or philosophy of physics, and in particular, posts on unresolved or controversial issues

Interesting observations about recent experiments.

Postby minkwe » Sun May 18, 2014 5:39 pm

Two recent experiments from two different groups claim to have closed the detection loophole for photons:

http://arxiv.org/pdf/1311.4488v1.pdf
http://arxiv.org/pdf/1212.0533v2.pdf

However, evaluation of the results reveal that they both deviate from the QM prediction in exactly the same way. One of the correlations is completely off. Those of you familiar with the recent thread where I discussed Bell's illusion, degrees of freedom and the confusion of counterfactual probabilities with actual probabilities will recognize it.

You can find the two comments from Emilio Santos at:
http://arxiv.org/pdf/1307.1540.pdf
http://arxiv.org/pdf/1305.6876v1.pdf
minkwe
 
Posts: 1441
Joined: Sat Feb 08, 2014 10:22 am

Re: Interesting observations about recent experiments.

Postby gill1109 » Sun May 18, 2014 11:34 pm

Yes, very interesting. So maybe the state created was not the one which the experimenters claimed they have created.

But I think the problem is that these experiments are still far from loophole free. These experiments did not have rapid, random, local switches of detector settings. If detector settings are kept fixed for many runs, then changed to new settings, all kinds of things can happen. The experiment proves nothing. We are still waiting. They say it might take another five years.
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: Interesting observations about recent experiments.

Postby Heinera » Mon May 19, 2014 5:24 am

It is not at all clear that Santos' statistical reasoning is correct. In his analysis of the Christensen et al. data, he computes the following table:



He claims that the first three coincidence counts are close enough to the quantum predictions, but then he says "In contrast there is a dramatic difference in the latter correlation where the empirical result is almost four times the quantum prediction."

But the quantum prediction for this combination of settings is a ratio of 0.000933977, hence the small numbers in that column. It is thus not clear that the relative error is meaningful here; if the errors are mainly due to background noise independent of detector settings, then the absolute difference between empirical and predicted counts is what should be compared. And those differences are of the same magnitude for all four correlations.

And also, keep in mind that Christensen et al. did not design their experiment to verify quantum theory; the designed it to rule out the detection loophole for LHV models.
Heinera
 
Posts: 917
Joined: Thu Feb 06, 2014 1:50 am

Re: Interesting observations about recent experiments.

Postby gill1109 » Mon May 19, 2014 11:04 am

Heinera wrote:It is not at all clear that Santos' statistical reasoning is correct. In his analysis of the Christensen et al. data, he computes the following table:



He claims that the first three coincidence counts are close enough to the quantum predictions, but then he says "In contrast there is a dramatic difference in the latter correlation where the empirical result is almost four times the quantum prediction."

But the quantum prediction for this combination of settings is a ratio of 0.000933977, hence the small numbers in that column. It is thus not clear that the relative error is meaningful here; if the errors are mainly due to background noise independent of detector settings, then the absolute difference between empirical and predicted counts is what should be compared. And those differences are of the same magnitude for all four correlations.

And also, keep in mind that Christensen et al. did not design their experiment to verify quantum theory; the designed it to rule out the detection loophole for LHV models.

Indeed. If we would replace the "quantum predictions" (corresponding to perfect creation of a certain pure entangled state) with "realistic quantum predictions" (corresponding to say 98% the aimed for pure entangled state and 2% a completely mixed state, ie pure noise) then the observed coincidence counts would rather nicely match the predictions.

Indeed, the point is not that a particular quantum prediction is or is not more or less exactly reproduced, but that the CH inequality is violated by a large number of standard deviations.
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden

Re: Interesting observations about recent experiments.

Postby gill1109 » Tue Jun 03, 2014 1:55 am

Jan-Ake has the following comments:

I saw this but I don't think this is a problem. The qm predictions are noise-free but the experiment is not.

The dark counts, for example, will be more visible if there are less events to look for. They will be there anyway.

Also, there is one setting that is more well-behaved than the other in all Bell experiments, this is typically because laser beams are not perfectly circular. It might be that the experimenters choose to have the more well-behaved setting where there are more counts, to decrease total noise.

But this is beside the point. There is a violation, so what quantum state you are aiming for is beside the point. Bell, or CH, is not about QM. It is about what you can reach with a local model. A violation is a violation.

(Santos has a point that the setting counts are not the same, but I do not agree with his normalization procedure. I think the inequality should be amended instead, not the coincidences.)


Exactly. If the statistics are done a different way we have to rethink the derivation of the bound. Maybe a different bound is needed. The statistical analysis of data has to match the protocol (design) of experiment or observational procedure, otherwise it can easily be meaningless. This the problem when you give non-statisticians e.g. psychologists powerful tools like SPSS. Which allow them to do sophisticated statistical analyses by click and drool sorry menu driven user friendly graphical interface. They have a data set and it has some superficial features e.g. 20 rows and 2 columns, containing real numbers. Then you bung it in SPSS two-sample tests and ask for a paired (or not paired?) t-test. But maybe you had 20 independent samples of size 2!
gill1109
Mathematical Statistician
 
Posts: 2812
Joined: Tue Feb 04, 2014 10:39 pm
Location: Leiden


Return to Sci.Physics.Foundations

Who is online

Users browsing this forum: No registered users and 86 guests

cron
CodeCogs - An Open Source Scientific Library