Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

B Kim et al, 1999 experiment and causality

  1. Dec 5, 2017 #1
    Hi everyone! Sorry for my bad English!
    I read old posts in this forum, googled it and still can't figure out one thing:

    What other explanation could there be, other than a random event in the future determined if the photon behaved as a particle or wave in the first detector?

    Thanks a lot! I simply cant think of any other possible explanation... It seems so clear that the traditional order of cause and effects wasn't followed...

    Also, I saw the Truscott experiment and I loved It! Can you please recommend some other experiments on this matter?

    Thanks a lot!
  2. jcsd
  3. Dec 5, 2017 #2


    User Avatar
    2017 Award

    Staff: Mentor

    I'm not quite sure what you mean by random event in the future. The short answer is: a photon is neither a wave nor a particle. If you want to measure the particle behavior, then it is one. If you want to measure the wave structure, then you'll find this. I like to look at it this way: A tree is a plant full of leaves that transports incredible amounts of water from the ground upwards. So if you have actually a tree and look for the leaves, you will find them. However, it could have been another plant. If you have a tree and look for the amount of water coming from the soil, the you'll find it. However, it could also be a water pump. In order to work with a photon, you'll normally establish a wave equation. But that doesn't make it a wave. It simply allows us to speak of e.g. color.
  4. Dec 6, 2017 #3


    User Avatar
    Gold Member

    The way you wrote it you make it sound like particle and wave behaviors are mutually exclusive. But any photon produces "click" in detector regardless whether you recover interference pattern (looking at D0/D1 and D0/D2 coincidences) or not (looking at D0/D3 and D0/D4 coincidences). Actually if you look at summary pattern of D0/D1 plus D0/D2 coincidences and compare it with summary pattern of D0/D3 plus D0/D4 coincidences there is no difference. So how can you argue about particle or wave behavior when the only difference is how you split the same D0 detection pattern into two sub-patterns?
  5. Dec 6, 2017 #4
    Thanks! The tree was a great example, and I wasn't clear about what I meant!

    Let me put this way:

    How can the photon hits detector D0 in very specific spots based on which detector his entangled partner will hit in the future, and the detector his entangled partner will hit will be determined at random by 50% mirrors it will encounter at the future?
  6. Dec 6, 2017 #5


    User Avatar
    Gold Member

    Your question is like this: Let's assume future can affect past. How this can happen?
    There is no sensible answer to that question.

    On the other hand you can ask: Let's assume future can not affect the past. How can we understand results of Kim et al experiment?
    This question can be answered no problem. But then let's first agree on that second formulation of question.
  7. Dec 6, 2017 #6
    Thanks a lot Zonde! That s It! My bad English and limited knowledge didn't allowed me to express it correctly!

    That's what I was trying to ask all the time! =)
  8. Dec 6, 2017 #7


    User Avatar
    Gold Member

    Ok, no problem.
    So we have two mirrors BSa and BSb. Half of the photons after these 50% mirrors go to separate detectors and half are mixed together on third 50% mirror. Let's look at entangled partners of those two halves. The patterns they make in D0 detector are exactly the same. There is no difference whats so ever.
    So it turns out that the partners of photons-going-to-be-detected-separately and partners of photons-going-to-be-mixed can not be told apart just by looking at D0 data. So when the difference appears?
    Partners of photons detected at D3 and D4 make almost the same pattern (actually the average should be shifted by 0.7mm between them but it might be quite hard to detect). So nothing strange here.
    But partners of photons detected at D1 and D2 are more interesting as they make two complementary interference patterns. But if we calculate the phase difference from the two alternative paths they could travel, it the same for both photons in the pair. So the phase for both photons in the pair was the same while it's different from pair to pair.

    So the mystery is how the photons traveling only path A or only path B "learned" about the path difference when mostly only single photon is in the setup at a time. And here we come to some very old and silly idea that maybe photon at the double slit "looks" ahead of it, and if it can reach the end point by both paths it "decides" to travel as a wave by both paths and if not then it "decides" to travel as a particle by only one path. So if you will try to adopt that idea for Kim's experiment it will make you think that the photon at the moment of "decision" can not only look ahead of it in space but in the future of it as well. So the only conclusion is that silly old idea does not work unless you are ready to adopt in addition the idea that future affects past.
  9. Dec 6, 2017 #8
    Awesome! Thanks a lot!

    So, please, assuming future can affect the past, how do we solve this mystery ?
  10. Dec 6, 2017 #9
    It is not a correct presupposition that after we know the position where the 1st photon hit at D0, the other photon hits D1 and D2 at random with 50% probability.
    If you look at Mach-Zehnger Interferometer experiments, you'll see that when a photon arrives to a beam splitter via two paths from the both sides, it is not detected after the reflection/passing the splitter with the same probabilities on both sides.
    Last edited: Dec 6, 2017
  11. Dec 6, 2017 #10


    User Avatar
    Science Advisor
    Gold Member

    There are a lot of experiments that call into question our views on the [causal] direction of time.

    Delayed-choice gedanken experiments and their realizations

    You can entangle particles *after* they are detected, for example. See Section F and Figure 9.
  12. Dec 6, 2017 #11


    User Avatar
    Science Advisor

    It only seems clear because explanations of the DCQE are basically terrible on purpose. They explain things backwards, and then confuse the explanation being backwards with causality being backwards.

    Here's what happens when I simulate the DCQE with my quantum circuit simulator Quirk:


    In the above circuit, the top line is akin to the idler photon and the next three lines are akin to the screen-hitting photon. The green box in the middle is showing how often the screen photon hits various parts of the screen, before we condition on the idler photon. Notice that it's evenly spread out (there's no wavy interference pattern). This matches what is recorded in the experiments.

    Now we measure the idler photon and focus on just the runs where that measurement result occurs. In other words, we look at our data and throw out the 50% of runs where the opposite idler measurement occurred and look at where the screen photon landed in the remaining half of runs we kept. This is why the diagram has a dashed lined from the detector to a note of "omits 50%". (This is where typical terrible articles start talking about "backwards in time effects", as if you throwing out half of your data corresponded to rewriting history instead of just... looking at a specific half of it.)

    If we measure the idler without rotating it first, i.e. we measure it along the Z axis, this doesn't do anything interesting. Both possible remaining halves are also evenly spread out. But if we measure it along the Y axis (by rotating it 90 degrees around the X axis before measuring along Z), then the two possible remaining halves do differ from each other. They make complementary wavy-looking patterns. (The total original data is still flat without waves.)


    The green expectation displays I'm showing in the diagram can be a bit misleading, because in a real experiment you don't get direct access to the expectations. You have to reconstruct them after the fact. So it's a bit helpful to reverse the situation and also consider how the screen landing position affects the idler.

    If we focus on the experimental runs where a particular screen landing position occurs, and figure out the corresponding state of the idler, this is what we get:


    Now the tall green rectangle is showing the screen landing position. Notice that the green box in the top right, which is showing how likely the idler measurement is to return ON instead of OFF, stays rock-solid at 50% when the X^1/2 gate is not present but flips out when the gate is present. But the green sphere things are always flipping out.

    The green spheres are showing the idler's state as a vector on the Bloch sphere. The important point is that a) the screen landing position affects which way that vector points (due to entanglement) but b) it's always pointing along the XY plane, and c) Z-axis measurements only depend on how up-or-down the vector points. If we measure along Z, all that crazy conditional spinning in the XY plane has no effect on the on-vs-off probability. But if we measure along X or along Y (or rotate before measuring along Z), then the crazy conditional spinning does affect our expectations.

    So instead of telling a backwards-in-time story about the idler photon measurement causing the screen position, we can tell a forwards-in-time story about the screen position determining the idler photon's state and thereby affecting its measurement. We simply choose whether to measure in a way that shows this effect, or in a way that is independent of it.

    Though... really we shouldn't be telling either the forward-in-time or backward-in-time stories. The thing to realize here is that the idler photon measurement and the screen photon landing position are correlated. If I tell you one, I've also told you something about the other. But remember: correlation is not causation. This age-old statistical advice is even more important to remember when it comes to quantum mechanics, because QM has a surprisingly counter-intuitive kind of correlation (i.e. entanglement). Is the screen photon causing the idler photon's rotation? Is the idler photon causing the screen photon to be in a wavy pattern? No. Both of those ways of thinking are misleading. The photons are merely correlated.

    All that being said, it helps to take a step back, stop focusing on the flickery noise of the individual runs, and just look over the conditional expectations:


    This isn't some magic time-defying experiment. It's not even inherently quantum; you can easily create analogous classical experiments. The quantum part is just there to mask the fact that we're confusing correlation for causation.
    Last edited: Dec 6, 2017
  13. Dec 6, 2017 #12
    Hi everyone! Thanks for the responses! Just one thing, i've misspelled a word in my answer:

    I intended to mean "can't ", not can! Damm autocorrect! =p sorry!

    I will post soon my thoughts on what's been posted! I just wanted to thanks again everyone and correct my misspelling! =)
  14. Dec 6, 2017 #13
    So, after the idler hits D0, his entangled pair no more has 50% o chance to reach D1 or D2? Its like the 50% mirrors aren't 50% mirrors? Sorry if it's silly question! Thanks for your time!
  15. Dec 6, 2017 #14
    Thanks a lot! It's gonna be lots of fun to look into that! =)
  16. Dec 6, 2017 #15
    In the original work the photon hitting D0 is called "signal", not "idler".

    After the first photon hits D0, the other photon is not entangled with the first anymore. And yes, after that, the chances for the second photon to hit D1 or D2 are generally not 50%. The "50% mirrors" are still 50%, no doubt, but in the Kim's experiment you have an interference of two states, one which has passed the final beam splitter (50% mirror staying in the center between D1 and D2) and another state which was reflected from it. It's only after that interference that the photon arrives at D1 or D2 and that interference changes probabilities up to having them (in extreme cases) to 100% on the path to D1 and 0% on the path to D2 or vice verse depending where the first photon hit at D0. Again, you need to understand how a photon interferes in Mach-Zehnger Interferometer to see how this works in details and to see some resemblance between Kim's experiment and the M-Z interferometer.
    Last edited: Dec 6, 2017
  17. Dec 6, 2017 #16
    Thanks a lot my kind friend! So, if I understood correctly, we can't imply causation from correlation, that's fine! Isn't there some statistical tests to do that, like student's t test? I remember something about it from my statistics classes, but my memory isn't very sharp...

    Anyway, so, when the photon hits the detector D0 in certain points, it's like we made a choice to measure if the photon has came from one of the slits or both, so the photon behaves as he is supposed to behave if it was a particle or a wave?

    Like if it is a choice that we make to measure the wave property or particle property (and that choice is made by the place were the photon hits the detector D0), and so the entangled pair behaves accordingly to what is been measured?
  18. Dec 6, 2017 #17
    I will do that! Thanks!

    Just a quick question: is the interference between the states were the photon passed or was reflected like the interference patterns in the double slit experiment? Like, that's areas that no photons lands because of the destructive interferences between the probabilistic waves from each slit? Am I in the right track or I should stop wandering and go directly to study the interferometer?

    Thanks again!
  19. Dec 6, 2017 #18
    Depends what you call "like". In some sense yes. In some sense no.

    right. :biggrin: But practically, you need either to get some grasp of QM or to find a really good explanation for the beginners of this particular experiment of Kim. Otherwise I do not see how these things can be understood.
    Last edited: Dec 6, 2017
  20. Dec 6, 2017 #19


    User Avatar
    Science Advisor

    There are statistical tools for inferring causality (see: Causal inference in statistics: An overview), but they won't apply to this case. We aren't trying to figure out the mathematical model driving the samples we're seeing. We already know the underlying mathematical structure, and from that we know it's correlation-like instead of cause-like.

    Photons don't decide to act one way or another, they always act the same way. If you want to understand the DCQE, you have to stop trying to incorrectly pigeon-hole quantum mechancs into "it's a wave or else it's a particle" and just focus on what the math actually says. Anything you've read that says photons choose to act like particles sometimes and waves other times is oversimplified junk incapable of modelling experiments involving entanglement, such as the DCQE.
  21. Dec 7, 2017 #20


    User Avatar
    Gold Member

    Ok, I lied a bit about explaining all the things in this experiment. We do not explain interference but only describe it in phenomenological manner using vectors and wavefunctions and give statistical predictions. However there is no backward causation anywhere in this description.
    There are interpretations that are trying to explain things left unexplained by standard QM. The one that clearly has particles in it is Bohmian mechanics, but I can't say it explains entanglement very well. For myself I sort of expect that this mystery could be resolved by some kind of quantum memory within beamsplitter and thinking of photon as something between soliton and wave packet rather than particle. But there is no interpretation like this.

    There are couple of things on experimental side you can consider.
    Interference can be observed between two beams from independent sources even when mostly single photon is in the setup at a time: Interference of Independent Photon Beams
    So you might think that it's always wave while clicks in detectors are just probabilistic events. But if you detect downconverted light from SPDC process whenever you detect a photon in one arm you find second photon in the other arm at equal time delay from the source. So it turns out that this wave at least in this particular case demonstrates very particle like behavior.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook

Have something to add?
Draft saved Draft deleted