Wednesday, November 14, 2012

ATLAS 1 lepton, 7 jets: a 4-sigma excess

Update: YK has pointed out (page 8: yess, passwords, I didn't have them, either) that the expectation was 6.5, not 4, so the confidence level is correspondingly weaker, about 3 sigma. Sorry.

Lots of new data were presented at the particle physics conference in Kyoto; lots of data (possibly much more important data) have been saved for the March 2013 Moriond conference, however.

Some papers by ATLAS and CMS continue to endorse the validity of the Standard Model at \(\sqrt{s}=8\TeV\). Even the branching ratios of the Higgs boson decay seem to return to their expected Standard Model intervals. In particular, ATLAS and CMS have published some studies with a similar outcome that used \(13/{\rm fb}\) and \(12.1/{\rm fb}\) of the 2012 data, respectively. The \(H\to\tau^+\tau^-\) decay seems to be compatible with the Standard Model now. CMS now sees a 4.5-sigma spin-0 scalar (not pseudoscalar) Higgs bump at \(126.2\GeV\) just in the ZZ channel. No new results on the diphoton channel. See Phil Gibbs on other Higgs news.

All this stuff is rather boring and we're used to it – it may be the not-so-new standard or it may be the silence before the thunderstorm. However, Matt Strassler has told us about the most interesting discrepancy unmasked in Japan.

The interesting observation comes from ATLAS' final states with 1 charged lepton (or antilepton) and 7 jets,\[

pp\to \ell^\pm jjjjjjj.

\] Note even Joseph John J. J. Thomson (the discoverer of the electron) had so many \(j\)'s. In this particular channel, 4 events are expected. Instead, they have gotten... now sit if you're not sitting yet... they have gotten 14 events.

Just for fun, let me offer you the particular numerical values of the Poisson distribution with mean \(\lambda=4\):\[

k & \frac{{\lambda^k}e^{-\lambda}}{k!}\\
0& 0.018 \\
1& 0.073 \\
2& 0.15 \\
3& 0.20 \\
4& 0.20 \\
5& 0.16 \\
6& 0.10 \\
7& 0.060 \\
8& 0.030 \\
9& 0.013 \\
10& 0.0053 \\
11&0.0019 \\
12& 0.00064 \\
13& 0.00020 \\
14& 0.000056 \\
15& 0.000015 \\
16& 0.0000038 \\ \hline

\] The probability that you get 14 or more events is just \(0.0076\) percent or 1 in 13,000. The LHC has to publish something like 13,000 similar results for an outcome that is this extreme or more extreme to be "really natural". I don't think it has happened. Of course, it doesn't mean that I think that the excess is real. Note that 4-sigma results have a 1-in-16,000 risk of a false positive so we're not far from a 4-sigma excess.

Prof Matt Strassler says that if the signal were real, there would also have to be an excess – possibly smaller excess – in the "somewhat smaller" final state with 1 charged lepton and 6 jets. Instead, there is a deficit in that channel.

Well, I don't think that Strassler's condition is necessarily true. It would be true if the excess were a continuous function of a continuous number of jets, \(k\). However, 6 or 7 may already look large, almost infinite, and one is tempted to say that physics is already messy at those "high" number of jets so the excesses have to be continuous.

But in reality, six or seven (now I prefer the words) aren't that infinite numbers. They're pretty small. There exist many kinds of reasons why the excess could be predicted for 1 lepton and 7 jets but not for a smaller number of jets. For example, the proton collisions may produce a pair of new particles and each of them may decay to 4 ordinary particles – either 4 jets or 3 jets plus a lepton.

I don't want to go into detailed models but of course, there are also some detailed models behind my tendency to pay attention to this particular excess. Search for multijet on this blog to review previous hints that the LHC may have seen other excesses in events with many jets in the past – and to be reminded of the "aromatic" models with flippons by Nanopoulos et al. who view multijet excesses as one of their favorite "smoking guns" of those \({\mathcal F}\)-\(SU(5)\) string-inspired models although their "real love" only starts at nine jets.

So we have another number to wait for: how many events CMS folks have seen in a similar 1-lepton-7-jet analysis?

Stay tuned.

Incidentally, many recent phenomenological papers start to prefer a light stop squark. The lighter stop could be \(375-500\GeV\) heavy while the heavier one could be close to \(1\TeV\). These figures are also desirable for the low \(B_s\to\mu^+\mu^-\) branching ratio that was recently bizarrely claimed by the mass media to "send SUSY to the hospital".

For your convenience, this is the Summer 2012 ATLAS exclusion chart for the LSP and stop squark mass. The combination \(150\GeV\) for the LSP and \(450\GeV\) for the stop seems to be more than compatible with the chart. If these figures are realized in Nature, ATLAS already has collected enough for a 5-sigma discovery of the stop by now.


  1. Thank you for the posts.

    Q: The odds are 1 in 13,000 if you have 1 experiment, but they are running "n" experiments. What is "n"?

    If you have a million questions, and then you sort the answers by the likelihood of the answer, you are likely to get an unlikely answer :-)

  2. Right, that was exactly my point. We say that the look-elsewhere effect hasn't been incorporated into the number "1 in 13,000". However, I still think that the number of similarly important figures or bins to this one that ATLAS and CMS have published is smaller than 13,000, so it's still rather unlikely that this should occur by chance.

  3. Prof Matt Strassler says that if the signal were real, there would also have to be an excess – possibly smaller excess – in the "somewhat smaller" final state with 1 charged lepton and 6 jets. Instead, there is a deficit in that channel.

    It would be true if the excess were a continuous function of a continuous number of jets, k.

    It would also be true if an important error mode of the detector were to under count the jets. The jets are counted by some sort of clustering algorithm, which would combine two that were too close to each other. That would be proportional to the square of the number of jets, but I don't know the constant.

  4. Dear Ralph, good point. But if you admit that the "real new physics" produces 7 jets (from seven quarks or gluons) plus a lepton, then such events could masquarade as 6 jets + 1 lepton events only if 2 jets happened to be too close to each other.

    Indeed, the probability that a jet pair is too close goes like k^2 where k is the number of jets. The other coefficient is roughly equal to the "solid angle of indistinguishability" over 4.pi (and over 2 because the ordering of the 2 jets is irrelevant). I am not sure about the numbers but I think that jets must have relative angle less than 15 degrees or so to be "merged". The solid angle in the circle 7.5 degrees of radius is about 0.05 steradians. That divided by 4.pi is about 0.004, so even if you multiply it by 7^2/2=24, you only get 0.1 excess event pretending to be 6 jets.

    I believe that the effect of "incorrectly identifying two jets" is still a small perturbation that may be neglected.

  5. Fascinating post! As a fellow science-lover, I thought you might be interested in theoretical physicist Sean Carroll’s new book about the Higgs boson, The Particle at the End of the Universe. I’m happy to share more
    info with you and send a review copy if you’re interested.

  6. I like this article a lot, because it most efficiently enhances the fun part of the story compared to other reports I have read :-)