Time Machine

So, back to my crazy self, I've been thinking about physics. Mainly, quantum physics. I think I've found something wrong with either the theory, or my understanding of it. However if I'm right about this, then we should be able to setup a very simple experiment that violates causality. Still there? Okay, first, some background...

The double slit experiment is fairly simple and has been tested and confirmed. There're details here, and wikipedia does a much better job at explaining it than I could. There's something interesting to note, now. If you place something in front of each slit that sets the spin of the photon (don't worry if you don't understand spin, just accept that it can be set, and then determined later,) and make sure that the spin set by one slit is around a different axis than the spin set by the other slit. Have the detector modified so that not only can you determine where on the detector the photon hit, but also its spin. Doing this will get rid of the interference pattern. The photon can't have two different spins when it's detected, so is must have gone through one of the slits, but not both. Here's the weird part: If you place something after the double slits but before the detector that sets the spin of every photon to be the same, regardless of what it was before, the interference pattern will come back. The reasoning behind this is that since the detector can't determine which slit the photon passed through (the spin set by the slits has been erased by the new spin-setting filter in front of the detector), the photon actually passed through both, and therefore will produce an interference pattern. Oddly, it seems you can erase information, and in doing so you can erase the fact that the information ever existed and any affect that information had on a system.

There's a way to modify the experiment where rather than using a laser and two narrow slits in an otherwise opaque barrier, you split the laser into two beams using a double mirror, then recombine the two beams on the detector. I'll try to describe this process now, since it will make my next suggestion easier to follow:

Imagine you have a square. The top left corner has a laser beam directed at it from further left (the laser is pointing to the right, and is located to the left of the top-left corner). This top-left corner has a half-mirror slanted at a 45 degree angle, so its two sides are facing up-and-right and down-and-left. This mirror will split the laser so that half of it continues along it's original path heading towards the top-right corner of the square, and the other half will be reflected down towards the bottom-left corner of the square. Now, at the top-right and bottom-left corners you place mirrors that redirect the photons to the bottom-right corner of the square. To sum up, the laser starts at the top-left corner of the square and can either go right then down, or down then right, and ends up either way at the bottom-right corner of the square. You place the detector at the bottom-right corner, and you get an interference pattern. This is really the same experiment as the double-slit experiment, except that the paths that the photons can take are more widely separated.

Like in the first example I gave, with photon spin being used to detect which path the photon took, you can place a device called a downconverter along the two paths the photon could take. A down converter converts a photon into two photons with half the energy of the original. One of these photons (called the signal) is sent on the path the original would have taken, and the other (the idler) can be detected. Thus, we can listen for idlers coming from the two downconverters and determine which way around the square a photon went, and in doing so we destroy the interference pattern. That happens because information about which way around the square a photon went is recorded. The photon can't go both ways at the same time since it had to decide at the downconverters if it was either 100% on that path, or 100% on the other path. The downconverter makes the photon decide before the detector which way it went, so its wavefunction can't wait and recombine later at the detector. No wavefunction recombination = no interference pattern.

Here's the odd part: What if we take the two idlers and send them to a double mirror. You can set up the experiment so that once an idler passes through or is reflected off of the double mirror, you have no way of telling which downconverter produced the idler. When you do that, the information about which way the photon went is irrelevant. It could have gone either way around the square, and the results would be the same. When this happens, you get the interference pattern back. That's pretty cool. Even though you know the photon passed through one of the downconverters (earlier, we knew it had some definite spin), it doesn't matter since the path information held by the idlers is erased (earlier, the spin was set to be the same for all photons.) Erasing which-path information can affect the signal photons, even if you don't touch them after the downconverters and only look at the idlers. Somehow the idlers can tell the signal photons whether or not they can act like they took both paths and produce an interference pattern, or if they have to decide one way or the other, and produce two bright spots (one for each path.)

That's weird enough, but it's not that bizarre if you accept quantum entanglement or "spooky action at a distance" as Einstein called it. The real problem is when we start to look at this from a causality perspective. Imagine that we send the idlers along some path that takes a very long period of time, like a second or two. Lets setup a lab on the moon that can either accept the idlers from earth and pass them through a double mirror, or if it wants it can detect each idler independently. If it passes the photons through the double mirror we should see an interference pattern develop back on earth, and if it detects each idler (and the path it took) without erasing the path information, we should just get two bright spots on our detector here on earth. What's wrong here is that we made the observation (interference pattern or no) more than a second before the idlers are detected. Someone on the moon, a second from now, could thus send information back to earth using the photons the earth sent it. The information would travel backwards through time. Increasing the distance a light year or so, and sending a bunch of idlers, we could use Morse code (or any binary language) to send messages from the future when the idlers are detected back to the present when we conduct the experiment.

What if we set up another experiment where the people detecting the idlers have a separate interference-pattern detecting device that sends its idlers back to earth. It'll take a year to get here, and when the idlers do we can decide if we want to produce an interference pattern or not on the detector a light year away and a year in our past. Assuming the people a light year away are nice enough, they could in theory copy the message we sent them and send it back to us. They would do this by looking at the presence or absence of an interference pattern on their end, and combine the idlers we sent them when they detect an interference pattern at their end (thereby inducing an interference pattern at our end), and when they don't see an interference pattern on their end they could detect the idlers we sent them independently (thereby preventing an interference pattern at our end.) This would, in effect, allow us to send messages 2 years into the past.

This, obviously, violates causality. So what's the deal? What don't I understand here? Is it actually possible to send information backwards through time? I'm betting that this is like the perpetual motion machine I designed years ago: it looks good on paper to novices, but once you really understand what's going on it's obvious why it won't work.