Photon "Wave Collapse" Experiment (Yeah sure; AJP Sep 2004, Thorn...) There was a recent paper claiming to demonstrate the indivisibility of photons in a beam splitter experiment (the remote "wave collapse" upon "measurement" of "photon" in one of the detectors). 1. J.J. Thorn, M.S. Neel, V.W. Donato, G.S. Bergreen, R.E. Davies, M. Beck http://marcus.whitman.edu/~beckmk/QM/grangier/Thorn_ajp.pdf [Broken] Am. J. Phys., Vol. 72, No. 9, 1210-1219 (2004). http://marcus.whitman.edu/~beckmk/QM/ [Broken] The authors claim to violate "classicality" by 377 standard deviations, which is by far the largest violation ever for this type of experiment. The setup is an archetype of quantum mystery: A single photon arrives at a 50:50 beam splitter. One could verify that the two photon wave packet branches (after the beam splitter) interfere nearly perfectly, yet if one places a photodetector in each path, only one of the two detectors will trigger in each try. As Feynman put it - "In reality, it contains the only mystery." How does "it" do it? The answer is -- "it" doesn't do "it" and the mysterious appearance is no more than a magic trick. Unlike the earlier well known variants of this experiment (,), the present one describes the setup in sufficient detail that the sleight of hand can be spotted. The setup is sketched below, but you should get the paper since I will refer to figure and formula numbers there. Code (Text): G photon Source TR photon PBS T DG <---------- [PDC] ----------------\----------> DT | | R V DR The PDC Source generates two photons, G and TR. The G photon is used as a "gate" photon, meaning that the trigger of its detector DG defines the time windows (of 2.5ns centered around the DG trigger) in which to count the events on the detectors DT and DG, which detect the photons in the Transmitted and Reflected beams (wave packets). The "quantum" effect they wish to show is that after a detector, say, DT triggers, the other detector DR will not trigger. That would demonstrate the "indivisibility" of the photon and a "collapse" of the remote wave packet at DR location as soon as the photon was "found" at the location DT. In order to quantify the violation of classicality,  defines a coefficient g2 which is a normalized probability of joint trigger of DT and DR (within the windows defined by DG trigger) and is given via: .... g2 = P(GTR) / [P(GT)*P(GR)] ... (1) or in terms of the (ideal) counts as: .... g2 = N(GTR)*N(G) / [N(GT)*N(GR)] ... (2) where the N(GTR) is count of triple trigger, the N(GT) of double triggers DG and DT, etc. The classical prediction is that g2>=1 (the equality g2=1 would hold for a perfectly steady laser source, the "coherent light"). This inequality is eq (AJP.3). The quantum "prediction" (eq AJP.8,13) is that for a single photon state TR, the g2=0. The paper claims they obtained g2=0.0177 +/- 0.0026. The accidental (background) coincidences alone would yield g2a=0.0164, so that the g2-g2a is just 0.0013, well within the std deviation 0.0026 from the quantum prediction. Perfection. The two tiny, little clouds in this paradise of experimental and theoretical perfection are: a) while there is a QED prediction of g2=0, it is not for this kind of detection setup (that's a separate story which we could [post=529314]pursue later[/post]), and b) the experiment doesn't show that (2) yields g2=0, since they didn't measure at all the actual triple coincidence N(GTR) but just a small part of it. Let's see what was the sleight of hand in (b). We can look at the coincident detections scheme as sampling of the EM fields T and R where the sampling time windows are defined by the triggers of gate DG. Here we had sampling window of 2.5ns and they measured around 100,000 counts/sec on DG (and 8000 c/s on DT+DR). Thus the sampled EM field represents just 0.25 ms out of each second. The classical prediction g2>=1 applies for either continuous or sampled measurements, provided the samples of GT and GR are taken from the same position in the EM stream. For the coincidences GT and GR,  does seem to select the properly matching sampling windows since they tuned the GT and GR coincidence units (TAC/SCA, see AJP p.1215-1216, sect C, Fig 5) to maximize each rate (they don't give unfortunately any actual counts used for computing via (2) their final results in AJP, Table I, but we'll grant them this). Now, one would expect, that obtaining the sequence of properly aligned GT and GR samples (say a sequence of length N(G) of 0's and 1's), one would extract the triple coincidence count N(GTR) by simply adding 1 to N(GTR), whenever both GT and GR contain 1 for the same position in the bit array. But no, that's not what  does. For "mysterious" reasons they add a third, separate coincidence unit (AJP.Fig 5; GTR) which they tune on its own to extract its own separate sample of EM fields. That alone is a gigantic loophole, a secret pocket in the magicians sleeve. If the sampling windows GT and GR for the new GTR unit are different enough from the earlier GT/GR windows (e.g. shifted by just 1.25 ns in opposite directions), the classical prediction via (2) will also be g2=0 (just background). And as they're about to finish they pause at the door, with 'and by the way' look say "There is one last trick used in setting up this threefold coincidence unit." where they explain how they switch optical fibers, from DR to DG, then tune to G+T to stand in for R+T coincidences, because they say "we expect an absence of coincidences between T and R" (p 1216). Well, funny they should mention it, since after all this, somehow, I am also beginning to expect the absence of any GTR coincidences. Also, unlike the GT and GR units which operated in START/STOP TAC mode the GTR unit operated in START GATE mode, where a separate pulse from DG is required to enable the acceptance of DT and DR signals (here the DT was used as START and DR as STOP input to TAC/SCA, see fig 5, while in the other two units G was used for SATRT and T or R for STOP). It surely is getting curioser and curioser, all these seeming redundancies with all their little differences. The experiment  also had a 3rd GTR unit with its own tuning, but they didn't give any details at all. The AJP authors  give only the qualitative sketch, but no figures on the T and R sampling window positions for GTR unit (e.g. relative to those from GT and GR units) were available from the chief author. Since the classical prediction is sensitive to the sampling window positions, and can easily produce via (2) anything from g2=0 to g2>1, just by changing the GTR windows, this is a critical bit of data the experimenters should provide, at least mention how they checked it and what the windows were. Of course, after that section, I was about ready to drop it as yet another phony 'quantum magic show'. Then I noticed at the end they give part numbers for their TAC/SCA units, (p1218, Appendix C): http://www.ortec-online.com/electronics/tac/567.htm [Broken]. The data sheet for the model 567 lists the required delay of 10ns for the START (which was here DT signal, see AJP.Fig 5) from the START GATE signal (which was here DG signal) in order for START to get accepted. But the AJP.Fig 5, and the several places in the text give their delay line between DG and DT as 6 ns. That means when DG triggers at t0, 6ns later (+/- 1.25ns) the DT will trigger (if at all), but the TAC will ignore it since it won't be ready yet, and for another 4ns. Then, at t0+10ns the TAC is finally enabled, but without START no event will be registered. The "GTR" coincidence rate will be close to accidental background (slightly above since if the real T doesn't trip DT and the subsequent background DT hits at t0+10, then the DG trigger, which is now more likely than background, at t0+12ns will allow the registration). And that is exactly the g2 they claim (other papers claim much smaller violations and only on subtracted data, not the raw data, which is how it should be, the "nonclassicality" as the Quantum Otpics term of art, not anything nonclassical for real). So, as described in , the GTR unit settings will cut off almost all genuine GTR coincidences, yielding the "perfect" g2. I did ask the chief experimenter about the 10ns delay and the inconsistency. Oh, he knew it all along, and they have actually used the proper delay (and not the 6ns as stated the paper), but the paper was too long for such details. Lemme see, they wrote the paper and the AJP told them to shorten it to such and such length. Now, say, the six of them sweated for a day editing the text, and just had it at about the right length, except for 5 extra characters. Now, having cut out all they could think of, they sit at the table wondering, vot to do? Then suddenly a lightening strikes and they notice that if they were to replace the delay they actually used (say 15ns, or anything above 11.5, thus 2+ digit anyway) with 6 ns they could shorten the text by 10 characters. Great, let's do it, and so they edit the paper, replacing the "real" delay of say 15ns with the fake delay of 6ns. Yep, that's how the real delay which was actually and truly greater than 10ns must have become the 6ns reported in the paper in 10 places. Very plausible - it was the paper length that did it. And my correspondent ended his reply with: "If I say I measured triple coincidences (or lack thereof) then I did. Period. End of discussion." Yes, Sir. Jahwol, Herr Professor. --- Additional References: 2. J. F. Clauser, ``Experimental distinction between the quantum and classical field-theoretic predictions for the photoelectric effect,'' Phys. Rev. D 9, 853-860 (1974). 3. P. Grangier, G. Roger, and A. Aspect, ``Experimental evidence for a photon anticorrelation effect on a beam splitter: A new light on single-photon interferences,'' Europhys. Lett. 1, 173-179 (1986). 4. R. J. Glauber, "Optical coherence and photon statistics" in Quantum Optics and Electronics, ed. C. de Witt-Morett, A. Blandin, and C. Cohen-Tannoudji (Gordon and Breach, New York, 1965), pp. 63–185.