ICARUS: Neutrinos Travel At Light Speed. Period.
    By Tommaso Dorigo | August 14th 2012 12:06 PM | 25 comments | Print | E-mail | Track Comments
    About Tommaso

    I am an experimental particle physicist working with the CMS experiment at CERN. In my spare time I play chess, abuse the piano, and aim my dobson...

    View Tommaso's Profile
    Little less than one year ago the world of fundamental physics was shaken by the bold claim of the OPERA collaboration, which produced a measurement of the time of flight of neutrinos traveling underground from Geneva to the Gran Sasso mine in central Italy. The beam of neutrinos, produced by the CERN SpS proton synchrotron, was observed to produce interactions in the large mass of the OPERA detector with about 60 nanosecond anticipation with respect to what would be expected for a particle traveling at exactly the speed of light (2439096.1+-0.3 nanoseconds, since the flight path is of 731221.95+-0.09 meters).

    Following the CERN announcement of the OPERA result physicists around the world busied themselves with pointing out several issues with the measurement, among which the statistical analysis of the data, the precise measurement of several delays in the detection process, and other hardware components, and remained generally quite cold. Of course the experimental proof that neutrinos traveled at a speed faster than c, the speed of light in vacuum, would be a datum impossible to reconcile with known physics. The result had a statistical significance of six standard deviations if one believed all uncertainties, but in fact this was the center of the discussion. Could the total timing error be believed ? I pointed out in this blog that I did not believe the precision of the time delay introduced by a light fiber (40000 nanoseconds, with a claimed accuracy of 1 nanosecond!).

    In contrast, the general public received a flood of information from newspapers and blogs that over-hyped the result, claiming Einstein had been proven wrong, and casting under bad light the whole "establishment" of research in fundamental physics.

    After the announcement of last September new experimental results were produced; one, again by OPERA, confirmed the original find with the analysis of some additional neutrino interactions obtained from a short-bunched CERN beam designed expressly for the test. The competitor experiment ICARUS, another neutrino detector in the Gran Sasso mine exposed like OPERA to the CNGS beam, found instead results in conflict with the claimed superluminal velocity of neutrinos. Theoretical arguments implied that, unless one assumed further anomalies in the behaviour of neutrinos, these particles should lose energy if superluminal, but ICARUS did not see such an effect. Also a reanalysis of the data from NOMAD found an inconsistent picture in this respect.

    It took a few more months before OPERA traced the source of the 60 ns difference between expected and observed signal timing to a loose cable in one of its electronic crates. In the meantime, ICARUS produced independent measurements of the time of flight, finding results in complete accord with neutrinos traveling at light speed, and not faster.

    Now ICARUS has published a new result, much more precise than the former one, obtained from another dedicated run in May this year. They collected 25 neutrino interactions, and due to the better designed experimental conditions and the availability of four different synchronization methods, the uncertainty in the time difference (arrival time of neutrinos minus expected arrival time for particles traveling at light speed) is now really very small. Of course the result is consistent with zero: -0.18+-0.69+-2.17 ns. I would say this is a rather definitive proof. An upper limit on the speed of neutrinos is also set at 1.0000016, at 95% confidence level. On the right you can see the distribution of time differences for the 25 neutrino interactions, which is well-centered around zero.

    So, neutrinos are not superluminal, as many of us had trusted all along. Science proceeds by trial and error, and this time error preceded trial, but otherwise the scientific process was disciplined: a claim is put forth, it is put to the test, and finally disproven; the original claim itself is then found to arise from human error. This time we also had the positive result that the person responsible of the too bold claim paid in person by having to leave the place of spokesperson of the experiment. The only thing that did not work well in this context is the fact that the result received too much exposure by the media... But it is better to have overhyping of false results than censorship!


    The headline is misleading. If oscillating neutrinos traveled at light speed, it would be nearly as revolutionary as if they traveled superluminally.

    In fact, ICARUS's results are almost equally consistent with (slightly) sub-luminal and (slightly) super-luminal neutrinos.

    Agreed Blaise, but simplification is necessary when doing outreach, please keep that in mind.
    Putting neutrinos at lightspeed into the headline is not simplification but totally wrong. Perpetuating confusion is not outreach.
    Yeah, from this mutual cocksucking I get the impression that a "loose" cable caused a 6 sigma 60 ns deviation from c. And that loose cable caused that deviation at more than 6 sigma, if we're still friends? So we're now at 6 sigma c, no?
    So the conclusion is we're now quite sure that neutrinos travel at c? And have no mass?
    I just wanna know the common physisist's opinion.
    Worse problem now is that neutrinos seem to interact with beta-decay on earth. Surprise? Here's a big problem for geologists and their K40-dating.

    I am not sure exactly what is your question Paul. The wrong result was of a delta t of about 60 nanoseconds. The new result is a delta t compatible with zero, once accounting for the loose cable and other small adjustments. In any case, when an experimental physicist says "the speed of neutrinos is measured to be the speed of light" you should know that she is approximating things. It should be taken to mean "the speed of neutrinos is too close to the speed of light for our measurement method to discern a difference from it".

    As for beta decay, I don't know what surprise that is, nor what relevance it has. Neutrinos can exchange a W boson with a nucleon, so what ?

    > Perpetuating confusion is not outreach.

    I dare say, good sir, that this is a bit reach coming from a person who blogs at 100% certainty level about how above-c neutrinos can be "explained" by higher dimensions and similar stuff minted from the fantasy aether. Really.

    This whole incident sheds some light (plus a few neutrinos to boot) on the larger problem of how electronic information is disseminated these days. At the time OPERA's test pointed to either superluminal velocity or a discrepancy, someone evidently saw the story and "aggregated" it. Since so much online reporting these days is aggregation, as opposed to journalism, the aggregated version got distorted like sound in an echo chamber, with the result being that the world was turned upside down and the speed of light was shattered.

    What's more, when the test did turn out to be a discrepancy, in an ironic way, that fact got more publicity than the original test itself - this time with words like "fraud" appended to it. When online facts turn out to be wrong, the culprit can't possibly be the aggregator but rather the source of the information the aggregator copied.

    The sad fact is, were it not for aggregators looking for low-hanging fruit to propagate through the blogs, this everyday error in measurement caused by, of all things, a faulty cable would have been a non-story - just part of the scientific process. My guess is, if anyone within the next few years postulates a reasonable, new concept of superluminal particles, that person will be branded as part of the global conspiracy started by the Opera folks who, someone will inevitably suggest, should have stuck to making browsers.


    Nice end of the story Tommaso, isn't it?
    Even the "bad guys" got "shot" in the end! Only some Morricone music is missing to wrap this up :)

    Is it true that on the "loose connector"-side of the beam, people not only didn't resign or "paid in person" but also got promoted this year?

    -- Bo

    Hi Bo,

    I do not know the related gossip, sorry - but people do move up once in a while even in HEP, fortunately ;-)

    Whilst various media outlets and those whose 'science' is nothing more than trying to prove science wrong went off the rails I'm not sure it caused any damage to reputations generally. Several friends asked me at the time to explain what it meant. They were quite happy to accept that it was an observation that required a lot of corroboration. Many bloggers and commentors saw it as an opportunity to explain how science works. Others were impressed that the team themselves found the flaw.

    It may even be a salutary lesson to those who try to gain publicity by releasing sensational reports. Over the same time frame NASA publicised a paper that claimed a bacteria could utilise arsenic instead of phosphate in nucleic acid. In biology and biochemistry this pretty epochal stuff. Not only was the paper flawed but work performed by other groups just found it plain wrong. There has been no response from the original authors of any substance. This has left severe bruising in various areas. A badly handled case.

    So congrats to the Opera and now the ICARUS teams. And as ever, thank you Dr Dorigo for explaining thses experiments well enough to allow a lay person to get a glimmer of understanding.

    If Neutrinos travelled at c, then they would be massless. If they have no mass, how can they change flavor?

    Hi NHM,

    see comments above - to be pedagogical I decided to avoid discussing that detail here. In any case we will never be able to detect a difference in travel speed from the mass of neutrinos with experiments on Earth and home-made neutrinos. Oh well, never is a strong word. Let's just say the ICARUS measurement is well compatible with massive neutrinos, that move at a speed not experimentally distinguishable from c.

    Little less than one year ago the world of fundamental physics was shaken by the bold claim of the OPERA collaboration...
    This is a distortion of the truth. No genuine fundamental physicist was shaken by this claim because all of them knew that it would soon be proven wrong. Only armchair physicists, superficial science journalists, and people like the owner of this blog were "shaken" and dreaming about the fall of relativity. No competent physicists have joined them.
    That's not fair to Tommaso. Science2.0 indeed saw one prolific armchair physicist who got carried away by the hype, but that wasn't Tommaso. Judge for yourself (the titles are most revealing): .

    Of course that's unfair Anon, what do you expect from Lubos ? He is in purgatory AFAIAC, since he seriously misbehaved here not long ago. I have not written him off yet, because I do remember I did like to discuss with him a couple of times... It depends on how he chooses to act whether I will ever answer him again or not. The comment above is not insulting but neither is it a step in the right direction...

    "He is in purgatory AFAIAC"

    You need to study your Catholic theology. Lunatics do *not* go to purgatory, no matter what they do. Instead they go to Limbo, a place where everyone talks about the papers they might have written were it not for the satanic machinations of feminists....

    Dear Puccini,

    I have better things to do; anyway the "purgatory" is a better image than the limbo, because it is meant to explain that he will, if he behaves, one day see the light of God ;-)


    maybe it is worth mentioning that 4 experiments at LNGS have performed this measurment in 2012:
    Opera, Borexino, LVD, Icarus (sorted by the order of appearance of the pre-print on arxiv).

    So, it is not just something between Opera and Icarus.

    Hi Stefano,

    good to point that out. The Borexino and LVD measurements had escaped me, can you provide arxiv pointers for readers here ?

    Maybe I am missing something, but how on earth does one get a mean value of +0.18 ns from the data displayed in the histogram?

    Reading off the graph

    -6.0 ns: 1 count
    -5.5 ns: 2 counts
    -5.0 ns: 1 count
    -4.0 ns: 1 count
    -3.0 ns: 4 counts
    -2.5 ns: 3 counts
    -2.0 ns: 2 counts
    -1.5 ns: 1 count
    -1.0 ns: 1 count
    -0.5 ns: 1 count
    0.0 ns: 1 count
    1.5 ns: 1 count
    2.5 ns: 3 counts
    5.0 ns: 1 count
    6.0 ns: 2 counts

    I calculate a mean value of -1.06 ns (which within the error is still a null result but cetainly different from the quoted value)

    Some links to other recent neutrino velocity measurements:

    OPERA updated their paper on July 12:

    Borexino published their result on July 30:

    LVD published their result on August 7:

    and ICARUS published their result on August 13:

    See also

    Thank you Titus, quite helpful ! That also explain why I missed the Borexino and LVD results - I was on vacation !!

    OMG - Icarus people are *still* using PAW to make their plots ! ;-)

    [About three (business-) days ago I submitted to this blog a comment (along with an URL-style memo) whose receipt was acknowledged by the statement
    > Your comment has been queued for moderation by site administrators and will be published after approval.
    However, my comment has since then apparently neither been published, nor have I received information about its disapproval to the e_mail address provided.

    This is therefore a reminder to please publish my comment (and memo) that has been submitted, also noting the time of initial reception, if possible; or else to notify me of its possible disapproval -- thank you.

    Below, the already submitted comment text is repeated unchanged, only for reference, and I'll attach the same memo to the submission of this reminder. FW]

    Tommaso Dorigo wrote (August 14th 2012 12:06 PM):
    > Now ICARUS has published a new result [...] what would be expected for a particle traveling at exactly the speed of light (2439096.1+-0.3 nanoseconds, since the flight path [from BCT focal point to the ICARUS "reference entry point"] is of 731221.95+-0.09 meters).

    (The evaluation in terms of "meters" is surely based on the SI definition; cmp.

    Now, if in some particular trial the BCT focal point had stated some sufficiently short signal (regardless of any association with neutrinos; and if not associated with neutrinos then perhaps associated with other suitable particles),
    and if, for instance, 4878072 nanoseconds later (where 4878072 + 2 * 60 = 2 * 2439096) the BCT focal point observed that the ICARUS "reference entry point" had observed and echoed the initial signal,
    and if the deviation of the BCT focal point and the ICARUS "reference entry point" from having been mutually at rest during this trial had been negligible,
    would the corresponding "flight path" in this trial still be evaluated as 731221.95+-0.09 meters

    > [...] An upper limit on the speed of neutrinos [relative to the speed of light, c] is also set at 1.0000016, at 95% confidence level.

    If the upper limit on any speed value (relative to signal speed) is 1.0, by definition and with full confidence, what would be the proper procedure for evaluating the corresponding upper limit value at 95% confidence level, e.g. from the data used in the ICARUS publication ?

    Dear Frank,

    allow me to ignore the text above except for the meaningful question at the end.
    This is a very common problem - setting a confidence interval for a bounded parameter.
    There exist both frequentist and bayesian methods to address it. The problem is most
    brilliantly solved, in the context of frequentism, by Feldman and Cousins. They wrote
    a groundbreaking paper on the topic in 1999, and it has since collected over a thousand
    citations. You can easily get it by googling.