Friday, September 23, 2011

Yet another voice in the neutrino OPERA

I feel just a little prophetic, given that only a week ago I wrote,
... it is also useful to simply measure as many new things as possible. History is littered with moments where we didn't bother measuring something because we knew what the result would be, only to get a big surprise whenever someone finally did measure it. So nowadays if something is measurable, then someone, somewhere, is trying to measure it.”

Well, true to the claim, someone out there had the ridiculous idea that measuring the speed of neutrinos was a good idea, go figure. And, lo and behold, it seems they've found something! Maybe? Who knows!

Well, what sort of blog writer, who is a scientist (physicist even), would I be if I didn't join in with the neutrino cacophony? So, what follows is my attempt to say something interesting that hasn't already been said a million times already. I will focus on why this is such a surprising result and why this means nobody believes it.

This disbelief has nothing to do with the quality of the experiment. The measurement was recorded with impeccable accuracy and has been extremely carefully analysed. Neither is the disbelief because the result "proves Einstein wrong". It even has very little to do with the prospect of over-turning special relativity.

The disbelief is due to the fact that this observation brings into question the holy principle of causality – that is, the seemingly incontrovertible fact that cause must come before effect and not after

And, as my former PhD supervisor, said in the Guardian today:

“If we do not have causality, we are buggered."

For people living under rocks, what has happened?

Just briefly, in case anyone reading this isn't aware of what I'm talking about, an experimental group, OPERA, based in Italy, have been detecting neutrinos (a type of fundamental particle) that were generated at CERN (not in the LHC), in Geneva. OPERA have also calculated the speed of these neutrinos. What has made this front page news on the websites of the BBC, Guardian, Daily Telegraph, Nature, New Scientist, etc is that these neutrinos seem to be travelling faster than the speed of light. This is problematic and if true would be arguably the biggest discovery for 100 years, or more. The original article is here.

So why doesn't anyone believe it? (Would you believe the impossible?)

I think it is safe to make the following two claims about this result:

  1. Almost no physicist, including those at OPERA, believes (yet) that these neutrinos are actually travelling faster than the speed of light.
  2. Absolutely nobody can see any fault in OPERA's methodology, yet.

It might seem that this is an incredibly odd viewpoint for physicists to hold. How can you be completely convinced that something is wrong, despite the fact that the method used to arrive at the result appears to be 100% correct? It isn't an odd viewpoint though at all. I'm going to try to explain why through analogy.

Suppose you were a juror in a murder trial. Now suppose an expert witness explains how they can determine the time of death of the victim based on the body temperature of the victim's body when it was found. They explain how this has been used in many different trials and has always been accurate to within ten minutes. Based on this evidence, you decided that their method must work well and should be trusted.

If the expert witness was to then conclude that the victim died 30 minutes after they were last seen alive, you would probably choose to believe their testimony. It is trustworthy and the result also makes sense. However, suppose the expert witness instead concluded that the victim died six hours after the body was found. Then, despite the fact you previously thought the method was entirely sound and you have found no particular flaw in the methodology you would probably have serious doubts about the testimony. The point is, that while you wouldn't know exactly what went wrong in the analysis, you would presume that something had to have gone wrong because the outcome was too unbelievable.

The same thing is happening with these neutrinos. There doesn't seem to be any flaw in the methods OPERA have used, but the result is just so unbelievable that it is far more likely that something has gone wrong than that the result is correct, even though nobody has any idea what.

This concept was summed up by Carl Sagan (apparently) with the phrase “extraordinary claims require extraordinary evidence”. OPERA have really, really good evidence, but neutrinos travelling faster than the speed of light is extraordinary and they don't have extraordinary evidence.

Why is faster than light such a big deal?

One of the consequences of special relativity is something called “the relativity of simultaneity”. This is an effect that causes two people (observers) moving relative to each other to not consider the same set of events to be simultaneous. More simply, what this means is that if one of the observers sees two events happen at two different places at exactly the same time, the other one will see them happen at different times. It is important to realise that this isn't an illusion but a physical fact. Events that happen at exactly the same time for one observer do not happen at the same time for another. This is counter-intuitive, but it is the way the world works.

Also important to realise is that, although the relativity of simultaneity was first predicted by special relativity, it has subsequently been observed many, many times in many different circumstances. Therefore, even if special relativity turns out to be wrong for some reason, the relativity of simultaneity will still be true.

So, why is relativity of simultaneity a problem if neutrinos are travelling faster than light? The problematic situation arises when you consider two separate events, A and B, that happen at locations far enough apart that a light ray could not have passed from one event to the other. Because of relativity of simultaneity there will be potential observers that will see event A happen before event B, and other potential observers who see event B happen before event A. Note that the events need to be separated like this for this statement about simultaneity to be true.

If neutrino beams can travel faster than light, there is an immediate problem. A neutrino beam that is emitted at event A could actually be received at event B, and will, according to some observers, arrive before it left!?!

Even worse, what if someone at event B decides that whenever he receives a neutrino beam from event A, that he will send a neutrino beam of his own back to the location of event A? Because of the relativity of simultaneity, according to some observers, this neutrino beam from the location of event B to the location of event A will arrive at A before A has sent her beam. But what if A then chooses because of this not to send her beam at all!? Well, then B will never receive the neutrino beam and so will never send his beam. But A is there, looking at B's beam, so B must send his beam. But he can't, but he must, but he can't, but he must.

And causality has evaporated in a puff of faster than light neutrinos.

However, if instead nothing can travel faster than light, then no information can ever be sent between these events, A and B. And thus causality survives.

Now, one last time for emphasis, while this effect of relativity of simultaneity was initially predicted by special relativity, it has now been seen and measured. It is real. It is fact. So while faster than light neutrinos would indeed cause special relativity to need to be corrected, relativity of simultaneity would survive. That is why this neutrino observation is so very, very shocking. An observation that simply overturned special relativity would certainly be interesting and would generate a lot of interest. However, overturning causality would result in a paradigm shift of philosophical proportions.

So forget the neutrinos proving Einstein wrong and forget them overturning modern physics. These neutrinos seem to have a problem with cause and effect itself.

I couldn't have said it better myself...

So that is why nobody believes these results. OPERA themselves have done a stellar job and if they'd measured anything else with this much care and precision people would be jumping on this result. But yeah...

“If we do not have causality, we are buggered."


  1. In case someone does raise this point...

    Yes it is possible that neutrinos might travel faster than light, but still information can not be sent fast enough for causality to become an issue. In fact, if these neutrinos *are* travelling faster than light that is almost certainly what is happening (how though, I have no idea).

    The point is that without changing something else (I don't know what could be changed - maybe light doesn't travel at the speed of information... although that raises many problems of its own relating to photon masses) causality does become an issue.

  2. You've assumed Lorentz-invariance, have you not? If Lorentz-invariance is broken there is a preferred frame of reference, so then you could in principle specify that things travelling faster than light only propagate forward in time as seen in that preferred frame. Then closed causal curves would not be possible.

    It's a bit of a pedantic objection, I admit.

  3. Thanks for the comment, Sesh. Maybe you're right. And if these neutrinos were being recorded to travel faster than the speed of light over tiny, tiny, particle physics type distances then I wouldn't object to your point at all. (Neither however do I think this result would be as shocking if that were the case)

    But, they're not. They travelled from Geneva to Gran Sasso, near Rome. You can catch a train over that distance. Surely we know what happens when cosmic rays or laser beams travel those distances and there isn't a preferred frame for them.

    I've tried really hard to think about the implications assuming nothing about Lorentz-invariance, or relativity, and just thinking, "what else have we actually measured in the past?". And, if light travels at the speed of information, then it seems inevitable that I can make electrons, light, muons, etc... that will think these neutrinos are travelling backwards in time.

    *Seems* is a pretty crucial word though in that paragraph. I guess my conclusion (one you probably agree with) is that even if there is Lorentz-violation then it will be of a particularly dramatic type that affects macroscopic scales in a very subtle and profound way.

  4. Yes, that's what I meant by saying it was a rather pedantic objection: Lorentz violation would have other consequences (which I haven't really thought about very carefully), but in principle superluminal velocities only cause a problem for causality if you also have Lorentz-invariance.

    There's some discussion of these issues in a review paper I read several years ago: When I've got a bit more time I might go back and look at it more closely!

    PS: I think they do mention that one could argue the cosmological frame defines a preferred frame, and therefore faster-than-light things only propagate forwards in cosmic time, but then they list a series of objections to that argument as well.

  5. I seem to remember a whole bunch of mumblings a couple of years ago about some strange results from LSND and MiniBoone and invoking extra dimensions to explain apparent faster-than-light neutrinos. So it seems this sort of stuff isn't new to the neutrino sector. If the result holds up I would have thought the prevailing interpretation will be that it is confirmation of some kind of braneworld scenario rather than the end of Lorentz invariance. I mean naively it would seem there would have to be significant lorentz non-invariance to explain this result which surely violates some bound already put on it from precision measurements in QED or something? Does anyone know if that is the case?

    If on the other hand the neutrino is special; say it can oscillate to a sterile neutrino and leave the brane whereas the other matter is confined to it, perhaps you can invade that bound. Of course I could be talking total shit, since it's not my area :)

  6. Thought you might be interested:

  7. Sean Carroll must have read your comment, Sesh. And he didn't even acknowledge it. I stand by what I wrote above though, if the neutrinos are traveling faster than light and the explanation is violation of Lorentz-invariance, then it must be a very subtle and profound violation. The neutrinos didn't violate Lorentz symmetry over tiny >TeV type scales so constraints must be very tight.

    Hi Max, thanks for the comment. It's like office 6.1 never died. Lots of people seem to be claiming that a possible explanation for this result would be neutrinos tunneling through extra-dimensions. I accept that this is entirely consistent with general and special relativity and that there are models where what was observed can happen. But, without some sort of extra Lorentz symmetry violation I don't see how these braneworld models don't still raise awkward causality issues. If special relativity hold on the brane where the light travels then surely I can still construct observers who will see event A happen before event B and others who will see B before A. It really doesn't matter how it happened, if special relativity holds the conclusion seems inescapable to me.

    We need to build a neutrino detector on the moon and fire a laser alongside a neutrino beam. Then we can see which one gets there first.

  8. Hi Shaun - a couple of questions from an interested non-physicist:
    What do you mean by Lorentz (symmetry?) invariance and things "propagating forwards in time"? And light travelling at the speed of information? And finally (something I might actually understand!), how do you measure the distance from Geneva to Gran Sasso to an accuracy of... a few cm?

  9. Hello Anonymous, thank you for the comment. Good questions. I was wondering whether I should add a quick note explaining some of the stuff we'd started discussing above. So, here goes (if any other physicists reading this wish to clarify anything I write, please go ahead),

    Lorentz invariance: This is basically the principle that the laws of physics are the same for every observer. The reason why this becomes an important issue if these neutrinos are travelling faster than light can kind of be seen in my example in the post itself. I wrote that there will be some observers that see event A happen before event B and some that will see event B before A. This relied on the assumption that someone travelling in the direction from A to B would describe the universe the same way as someone moving from B to A. If the laws of physics are different for these two possible observers, then "relativity of simultaneity" might also not hold for one of them. If Lorentz symmetry is broken in the right way (I don't know what that way would be) then it could be that no possible observer could see event B happen before event A and then causality survives. It is called a symmetry because if I change who I am or how I observe, the laws of physics don't change.

    Thing propagating forwards in time: Well I don't think I ever wrote this, so I'm going to have to speak for Sesh here and assume I know what he meant (I think I do). The idea is that if you are someone making an observation and you have with you a clock, the direction of forward time is governed by that clock. Something then propagates *forward* in time to you if the causes and effect for that thing follow the right order according to your clock. So a neutrino being made should come before a neutrino being detected. Therefore if they follow this order according to your clocks, they're propagating forward in time according to you.

    Light travelling at the speed of information: In special relativity any physical signal cannot move out from a point faster than the speed of light. Even the tiniest, subtlest effect ripples away from something at the speed of light (incidentally this is also the speed of gravity, but nobody knew this when SR was being discovered so the phrasing speed of light got stuck!). If instead, special relativity was still true, but somehow light was being impeded and actually travelled slightly slower than the speed of gravity, or some other physical effect - say neutrinos for example) then this might explain the results. I called it "the speed of information" because it would be the limiting speed in special relativity that any information signal could be sent at (which is the important speed when worrying about causality).

    The final one: Well, I'm glad you think you might e able to understand this, because I don't. This part of the measurement is the most impressive part of their measurement and I really don't know the details well. It involves GPS satellites and antennas attached to CERN and Gran Sasso. It involves a delicate measurement of the distances from the antenna to the collider and detector. They measured the location of Gran Sasso accurately enough that they recorded the continental drift of Italy's tectonic plate! And the discrete shift in location caused by an earthquake that occurred during the experiment. Sorry I couldn't say more about this bit. The OPERA guys had to learn a bunch of meteorology to do this bit of the experiment so I feel OK not knowing it myself, seeing as they didn't either three years ago.

    I hope that helps!

  10. OK, I think I might have over-simplified my description of Lorentz invariance. Dear anonymous, if you're happy with this description, great. If you wish me to be more, errm, honest, please ask and I'll try.

  11. A few interesting points:

    1. What precisely the "speed of information" means is a little ambiguous. Physicists often like to think of it as the group velocity of a wave packet, so the group velocity can't exceed the speed of light in a vacuum. However, that's actually not entirely true. In fact in materials with anomalous dispersion (i.e. a negative refractive index) group velocities can be >c. Or even infinite - which means pulse arrives at its destination at the same instant that it leaves the source. Some guys in California (Berkeley maybe?) have actually created such materials and measured such group velocities of light pulses. So light can travel faster than light - happily this doesn't violate SR because of some technical argument about exactly what kind of wave packet is not allowed to travel faster than light. Apparently it was shown by Somerfeld - a long long time ago - that it is actually wave packets with sharp edges that can't travel faster than light. Don't ask me about technical details ...

    2. The neutrino wave packets are clearly not ones with sharp edges, so perhaps there is a get-out clause in there somewhere. Again, don't ask me for details ...

    3. Because the neutrino wave packets don't have sharp edges, measuring exactly what time the pulse "arrived" is actually a little complicated. They do it by fitting profile shapes to the observed waveform. It's quite easy to believe that there are greater systematics in doing this than they have estimated.

  12. Sesh, this is all quantum mechanics stuff you're mentioning. Why is any of that relevant to the neutrinos? The distribution function of the neutrinos is a classical distribution based on the distribution of the proton bunch in SPS (not wave distribution of indiviudal protons, but classical distribution of the bunch). Given the energies of these neutrinos their de Broglie wavelengths will be negligible compared to the distances involved here, so any questions of phase vs group velocities isn't relevant, surely.

    Regarding point 3. They know exactly the profile shape they expect, they fit the observed one with one parameter and one parameter only (the shift in time) and get a reduced \chi squared of 1.06 and 1.12 for the two extractions. Not just this, but the two edges of the expected and observed distributions match extremely well. This would be incredibly odd if they were over or underestimating the width of the expected distribution.

  13. I do agree with point 1. though (I just don't think it is at all relevant). When quantum mechanics is involved, causality becomes very interesting. It seems that it survives almost through a highly convoluted conspiracy. Physical effects in quantum mechanics can (kind of) travel faster than light, you just can't ever use that to send information.

  14. Point 1 may not be relevant to the neutrino experiment per se, but I thought it was an interesting point to enlighten (or confuse) your anonymous non-physicist friend.

    Point 3 is definitely relevant. There are all sorts of things to take account of. The neutrino beam spreads as it travels to Gran Sasso, for instance, so in fact they won't see all the neutrinos that make up the profile. Does this affect the profile shape? Can you then fit it using only one parameter? I don't know. More importantly, some of the people in the OPERA team don't know either, and refused to sign the paper.

    By the way, these same points (or very similar) were made by John Butterworth in the Guardian ( and by John March-Russell, at coffee yesterday.

  15. Point 2. definitely made it seem like you thought point 1 was relevant. Even in point 3 you called the distribution function "neutrino wave packets".

    Anyway, John Butterworth only made similar points to your point 3. which is definitely relevant to consider (I only meant to claim points 1 and 2 weren't - sorry about that). I was giving my perspective on why I thought the statistical analysis must have been done correctly, despite these potential concerns. I've read a lot of "but these things need to be taken into account" statements since Friday and it seems to always turn out that "these things were taken into account". To directly address Butterworth's point in the Guardian, I don't see why the beam widening would change the shape of the expected distribution at Gran Sasso. And if it does, it is odd that they get such a good fit to the beam they claim to expect.

    These *are* all important things to look into and discuss, but only just in case. I think OPERA deserve the benefit of the doubt regarding possible simple mistakes. The way they presented the results shows they definitely don't have an agenda and have tried very hard to work out what mistakes they might have made. They don't expect this to be correct either.

    None of the people who took their names off of the OPERA paper (a tiny proportion of the OPERA collaboration) have anything they claim was at fault in the analysis, they simply think it was too "preliminary" to put forward. I don't see the fact that this happened as adding any extra information to what we already know. That is, that this result is shocking and hasn't been reproduced. Apparently people take their names off Fermilab, CMS and ATLAS papers all the time.

  16. Obviously if we had spotted the mistake already then there wouldn't be any need for this conversation. I'm just pointing out the most likely scenarios where a mistake might be - maybe it isn't there but somewhere else, maybe there isn't one at all.

    Also, on reflection, I don't think my point 1 above actually is a quantum effect. Rather, it may be a quantum effect that allows you to get group velocities >c through anomalous dispersion, but the interpretation of what constitutes signal velocity when you have a non-top-hat pulse is I think purely classical.

    And re the neutrino "wavepacket": I used the wrong word, I meant "pulse". It's a bunch of 16,111 neutrinos arriving at different times. The signal content is entirely in the edges of the profile shape, which are not top-hat in nature. It may be that there is wriggle room there in terms of the signal velocity. Or maybe not - I'm not an expert and I haven't even had the time to read the paper carefully! Only 10 days to go ...

  17. "The signal content is entirely in the edges of the profile shape, which are not top-hat in nature."

    I've seen this repeated many times and, though it isn't too relevant to our discussion, I don't agree. There are lots of hills and valleys in the plateau of the pdf and the observed distribution also seems to match them very well. There is a significant amount of the signal in the edges (perhaps/probably even more than 50% of it), but I don't think it is anywhere near the entire signal.

    Regarding point 1. It's a wave effect (hence why it becomes a quantum effect for particles). But this pdf *isn't* a wave. I'm really not following here (I'm starting to wonder whether I'm missing something)... group velocities, phase velocities, none of this makes sense in the context of this signal. It spreads horizontally, but it isn't composed of a bunch of Fourier modes that travel at different speeds and separate. The energy spread of these neutrinos is tiny (as in completely and utterly negligible) compared to the total width of this pulse. It is a bunch of particles, not a bunch of waves.

    So, unless I'm really lost here, there is no ambiguity in the velocity of the signal. There may be uncertainties in the predicted form of the signal, certainly - this was what I understood Butterworth to be postulating as a possibility. But I'm yet to read or hear a decent suggestion of what could have caused this. If I hear/read of one I'd be very interested. And I would then be very confused as to how they got such a good fit to the wrongly predicted signal.

  18. Oops, I meant to add, "good luck with the writing... see you on the other side" ... etc, etc.