LHCb Versus CDF: Two Punches In The Face!
    By Tommaso Dorigo | July 27th 2011 05:48 AM | 10 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
    As a 20-year-long (ok well, 19) member of the CDF collaboration, I am very proud of this wonderful experiments' accomplishments in all areas of high-energy physics, from exotic searches to Higgs searches, from top quark measurements to b-physics measurements, and what not. CDF is a landmark in experimental physics, and the longest-lasting physics experiment ever. But it is not foulproof - nobody is in this wild world of statistical flukes and impossible-to-unearth systematic effects.

    Recent analyses in the B-physics sector by CDF produced a couple of important claims. I will not bother to tell you that I was extremely cold on both of them -I can offer circumstantiated proof in both cases, but the topic of this article is not my own (very good) nose for fake signals. Instead, I will just flash the two results and proceed to show how LHCb dismantled them.

    The first is the tentative observation ofa new hadron, called Y(4140), a bump observed in the invariant mass of pairs of muons and pairs of kaons. I will spare you the details, which you can find in the original preprint by CDF. The relevant figure is the one on the right, which shows the mass difference between the four-body system and the dimuon system. The signal is in red, the phase space is in blue.

    Being unable to fit a reasonable phase-space model to the endpoint spectrum, CDF interpreted it as the signal of a new particle sitting near threshold, and called it Y(4140). The signal was allegedly evaluated as a 5-standard-deviation thing after more data was added to a previous analysis which had found a first 4-sigmaish evidence. I wrote about the matter in some detail here, keeping a rather sceptical attitude; however, since my opinion at the time could offend some of my collaborators, I wrote it in Greek in my greek blog... I think you can still make sense of my deep scepticism on the Y in the post if you dump it into Here is the relevant part:

    Δεδομένου ότι τα αποτελέσματα και των δύο αναλύσεων έχουν δημοσιευτεί, οποιοσδήποτε μπορεί να αφαιρέσει τα παλιά δεδομένα (αυτά που φαίνονται στο πλάι) από τα νέα, βρίσκοντας έτσι προς έκπληξή του ένα ενδιαφέρον αποτέλεσμα: το σήμα δεν είναι τόσο στιβαρό στα νέα δεδομένα.

    Όπως πάντα, οι πιο ανήσυχοι ανάμεσά μας θα πρέπει να περιμένουμε περισσότερα δεδομένα για να πιστούμε γι’ αυτό το σήμα

    Have fun with google translate..

    Also, in another post which was more didactical in nature I touched the subject tangentially, (and this time in English), but if you are interested in the issue please visit that article as well.

    Now LHCb looked for the same signature in the large sample of data they collected in 2011 (376 inverse picobarns, to be precise). What they see is shown on the left. They can only fit 7+-5 signal events -a null observation- and their data shows a difference of 2.4 standard deviations with the CDF result. Note that, as reported in the figure, if the CDF signal were as estimated by CDF, LHCb would have been able to fit 39+-9+-6 events. The Y(4140) is on very shaky ground at the moment, and the new PDG will likely change its status in the particle zoo... This is punch number 1.

    Punch number two came not just from LHCb, but from a combination of LHCb and CMS data. This concerns the never-ending search for decays of B_s mesons to muon pairs, a very rare decay mode which may be enhanced in Supersymmetric models. CDF very recently reported a signal and tentatively measured the branching fraction as 1.8E-8 -five times the SM prediction, a result strongly smelling of SUSY particles contributing in virtual loops!

    Now LHCb and CMS report that their combined limit on the branching fraction yield a probability that the BR is indeed what measured by CDF as 0.3% (the CDF measurement would be off the page in this graph, at "18" in 10^-9 units). The PDF of the branching fraction obtained in a combination of their results is shown on the right. Punch number 2.

    I think it is rather unfair to pick two negative results to counter a positive one -this is a post-data practice which I cannot recommend. However, if we take every experimental input into account (not to mention my personal bias that the true BR of the B_s in dimuons is what the Standard Model predicts, and not five times more as some SUSY models would have it), we have to conclude that the CDF evidence of B_s decays is fading out.

    So in the end: two CDF effects do not get confirmed. Neither is officially dead yet. Stay tuned for more!


    It could be worse Tommaso, you've already punched us a couple of times before we entered the ring! We're about to unblind the single most precise measurement of phi_s (or -2\beta_s depending on your beliefs) and the Tevatron experiments have sent it from 2.7 -> 2 -> 1 sigma from the SM over the past years. It used to be a sexy analysis when I started on it, now it'll be nothing more than narrowing some contours. Stay tuned for it at lepton-photon....

    CMS and LHCb seem to have a pretty healthy excess in the plot you put above. Certainly not compatible with the central value of the CDF result but if you calculate the full probability I bet all the results are not that incompatible. Though I agree about new physics being unlikely. If the Bs -> mumu rate is high it is probably within the wiggle room of the SM prediction.

    Wow anon, thanks for your detailed opinion. I think I agree with most of what you say, but we differ in the conclusions strikingly. Because the prior of the SM being true is so close to 1.0 (I imagine a two-valued function here), and because any claim of new physics must be proven by a very sizable effect above stat and syst uncertainties, I take the 2-3 sigmaish excess of Bs as nothing more than a fluke.
    Anon, are you willing to bet 2:1 ? I will put 200$ where you put 100$ on the B_s.

    Let me know,
    longest lasting physics experiment ever
    Are you sure or is this mainly due to your definition of "physics"?
    Hi Sasha,

    my definition of physics does not include watching pitch flow down a tube. So yes, or no, depending on what can be agreed upon to be defined as a "physics experiment", one involving experimenters and not just a piece of material on a test stand.

    We are looking forward to updates at the Greek blog as well... ;)

    In both mass fit figures there is an excess of events around 1.2 GeV. Is this something interesting or just random fluctuations?

    In fact, I think there's where something is hiding!

    I think at least one of your punches is not quite fair. While it's true the central value of the CDF B->mm result is compatible at the <1% level with the CMS+LHCb combination, it's also true that the CDF 90% CL region falls in the middle of that plot, where their combination has an excess. Someone with a bias regarding the SM and the BF(B->mm) different from your own could well have written a different story.

    It's not that I'm convinced this signal will turn out to be true - although of course I hope it is. I'm not sure what I would put the odds at, but certainly larger than some of the other anomalies that have surfaced recently.

    You mention having been "cold" on this result. I would be interested to know your criticisms of the analysis. Or is this just a comment that you don't think the "signal" will turn-out to be real? The latter I understand, although I don't think your article makes that clear. You lump B-->mm in with the Y analysis. I think the B->mm result is in a different class from the Y analysis you describe. There are no phase-space effects, no complicated bgd shapes to understand, etc. It's a simple analysis, with lots of cross checks. Just curious.

    I don't often spend time reading blogs. From time-to-time I read yours. I like it. Keep up the good work.

    Full disclosure: I am a member of CDF (but perhaps that was obvious)

    I think I understand your point of view, and I cannot find anything to
    object to it. For sure the Bs->mm analysis is a different thing from
    the bump hunt for the Y. And for sure the incompatibility is at a
    level which might, one day, turn out to be a full agreement.

    I have been cold wrt the xs measurement of the Bs because I do not
    believe in SUSY. I think the BR is what the SM predicts, and that the
    one observed is a fluke. We have updated the Bs result many times and
    once in a while an upward fluke is bound to occur there too (we in CMS
    call this "look elsewhen effect" or "keep looking bias", as opposed to
    the more common "look elsewhere effect". Note that the LEW effect
    occurred in the Y search as well...).

    The thing is, a blog is not always a place where to be entirely fair.
    I am on purpose a bit provocative and flippant here at times, because
    I have learnt that it is a great way to foster discussion, force
    people to express their different points of view, and learn something in the process.

    So you understand that what I write there is not always an objective
    assessment of the situation. I have praised CDF results in my blog so
    much, I like to deviate a little sometimes if I can. The recent LHC
    results gave me a chance to do so, and I seized it. Sorry if I created
    some discomfort in my most long-time colleagues... But it's all great
    people who do understand the game and the rules. I am sure you do.