ATLAS And CMS Publish 2011 Higgs Results
    By Tommaso Dorigo | February 7th 2012 11:13 AM | 24 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
    You have seen it already two months ago, but those were "preliminary" results. Now both CMS and ATLAS have produced full-fledged documents (CMS here, ATLAS here) describing their respective combinations of different Higgs boson searches, using data collected in 2011 by the two experimental apparata at the CERN Large Hadron Collider.

    So what - you might retort - these are still "unpublished", in the sense that they have not seen the light of printed matter on scientific publication; they still only exist in the form of preprints. Let me clear the ground from misunderstandings here: between December 13th and today the two collaborations have worked hard at polishing off their results and documenting them as well as possible.  The results have not changed significantly, but this was not granted: it is what was meant by the word "preliminary" attached to every plot. Now, however, they are guaranteed to not change anymore.

    Today is therefore a good time to give a look at the results once again. I will not discuss the results, but maybe just paste below the p-value distributions produced by the experiments. These, I believe, are the most informative graphs for those who believe that the Higgs boson does exist. So here we go, Atlas first,

    and CMS second,

    A few remarks on these pics.
    1) the expected p-value in case of a Higgs boson in the data, in the two graphs, is shown by black hatched lines. Please observe that at 125 GeV ATLAS expected to see a p-value of the background-only hypothesis equal to 0.005, or a bit less than three sigma; CMS expected to see a p-value of exactly three sigma, so CMS had in principle more power (due to using more final states in their searches, and slightly more statistics in some cases IIRC). The data, however, chose to fluctuate the other way: ATLAS has an observed p-value of about 0.0002, CMS "only" 0.001, for the (a posteriori) most probable Higgs mass hypothesis.
    2) The various p-values for the individual searches are shown by different colored lines. The ATLAS plot has the expected p-value for each, while the CMS plot doesn't - a understandable choice given that CMS is displaying six independent analysis results against the three of ATLAS.
    3) Don't be deceived by observing a "disagreement" between expected and observed p-values in the ATLAS plot (for the 125 GeV hypothesis). What is plot is the "median" expected p-value, but the statistical fluctuations intrinsic of the searches naturally allow that quantity to be observed at very different values. Plus, it's a semi-logarithmic plot!
    4) I find the agreement of the two experimental searches impressive, and as I have already said here, I believe this is really the effect of Higgs boson decays in the ATLAS and CMS data we are looking at.

    To summarize, the two figures do talk of something going on in the 120 GeV region. For a Higgs believer, in fact, one might argue that the Higgs boson mass has already been measured: a mass of about 121+-5 GeV results from an eyeballing of the 95% CL limits. But that is too large a leap of faith. Let us then just say that there is a fat chance that the Higgs boson has indeed a mass of about 124-126 GeV. That being true, the LHC this year will definitely allow CMS and ATLAS to conclusively observe it - 8 TeV running conditions have been confirmed last week at the Chamonix workshop, and 20 inverse femtobarns of proton-proton collisions are a fair estimate for the amount of data that will be delivered next year.

    One word about the two publications: the documents are rather technical, so I do not advise newcomers to download them and try making sense of them. However, graduate students working in LHC experiments should, in my humble opinion, spend the better part of a weekend over them. I guarantee there is something to learn, and it is stuff that will be needed in order to really make sense of the next round of results! And don't be scared by the length of the papers upon downloading them: most of the pages just contains the authors names !!


    Is it just me or did the CMS gamma-gamma p-value go down by an order of magnitude? Is there a new CMS digamma mass plot available?

    The ATLAS plots have a pretty wide range of luminosities. Is more data expected to be added for the 2011 run, or is this just an artifact of the trigger?

    'Fat Chance' means little or no chance in English (it can only be used sarcastically)!
    Thanks for answering my query on the earlier thread.

    Thanks Brett  - my English is indeed perfectible. If I have a chance I'll correct.
    In slang, the opposite means the same thing.   So you can have fun with it.  Thus, if someone asks you about something you don't want to do, "Is there no chance you will do X?"  you can reply, "Actually, there are two chances; fat and slim."

    That is sure to get a laugh in America.
    Thanks Hank, I'll try to remember this!
    Thanks for the heads up.  Even though I am more of an astrophysicist I will certainly read these.  
    Science advances as much by mistakes as by plans.
    20 inverse femtobarns to be collected in 2012?
    Are you completely raving bonkers? ;)

    That's four times the 5/fb collected in 2011, and tentative estimates at an LHC workshop a few weeks back was suggesting 10/fb. Right now the LHC team are discussing the running of the LHC at the Chamonix 2012 workshop and I'll be amazed if they even come close to choosing 20/fb for 2012.

    20 inverse femtobarns to be collected in 2012?
    Are you completely raving bonkers? ;)

    That's four times the 5/fb collected in 2011, and tentative estimates at an LHC workshop a few weeks back was suggesting 10/fb. Right now the LHC team are discussing the running of the LHC at the Chamonix 2012 workshop and I'll be amazed if they even come close to choosing 20/fb for 2012.

    Dear Carla,

    I am only overestimating by a factor of two the "suggested" 10/fb. Last year, they "predicted" 1/fb and we got five...

    I am willing to bet that the final delivered 2012 luminosity will be closer to 20 than to 10.

    The Stand-Up Physicist
    A qualitative question. In the Atlas 2011 data, I see three valleys within valleys. Get another 10 inverse femtobarns, I can see them growing deeper together.

    The CMS data has a different feeling. The two big peaks are not valleys within valleys. Neither is a valley within the valley found by Atlas. A HUGE effort has gone into calibrating both machines. Do peaks shift around much after the amount of data already collected?
    Hi Doug,

    be careful, those are p-value plots, so their interpretation is complicated. They depend a lot on the vagaries of the collected data. I do not see a calibration problem here... For sure I would look somewhere else for it, not here. For instance, we have hundreds of thousands of Z->lepton pair events to use for this kind of business, as we have gamma-jet events, pizero and eta signals to gamma pairs, etcetera, etcetera... Not in a p-value Higgs search plot, no!

    Am I correct that results and analysis will be performed on 2012 data quite distinctly from the 2011 set, and that
    the results this summer will not be presented as a refinement of the 2011 results, but a completely new set?
    They may then produce another paper combining them, as they keep hinting they will do for Atlas and CMS.

    Hi Brett,

    I believe the 2012 data will be analyzed separately and then the
    results may be combined. One (strong) reason for thinking this is what
    will indeed happen is that in 2012 we will run at a different CM energy
    (8 TeV), which entails different background shapes, different rates,

    Oh boy, imagine all the combination plots I will need to do.

    @Carla - The plan for 2011 was only like one femtobarn (LHCb target was 0.75, achieved 1.04), its not crazy for the machine to deliver significantly more than anticipated

    @briv - I cant comment on this particular plot, but at least in some cases, the channels have less data than others (for example if you're looking in a mu mu channel, you dont need, say, ECAL information. So if the ECAL is down, you can still collect useable data, which is not useable for the ee channel)

    @anonymous well maybe they're hoping for 15-20/fb but officially 10/fb, just as for last year 2-3/fb was hoped for, but 1/fb was the official target. It'll be interesting what they decide at the end of friday, bearing in mind beam life time was significantly affected as they increased intensity last year.

    So...assuming tis is going to be confirmed at 5 sigma soon, what is the precision expected in the determination of the Higgs mass in the near future by the two collaborations? Will it be enough to see if the two central values of the two detectors differ in a significant way?

    Hi, with a combined per-experiment 5-sigma signal, the mass will be probably best known by the few H->ZZ candidates we get; but on the other hand the ZZ peak will by itself not be significant enough to guarantee that what we are measuring with it is the true Higgs mass. To explain better, assume you have five events at 125 GeV of mass, and one at each of the other mass bins (say 1 GeV wide bins). The signal by itself will be very insignificant, but the fit to a flat background plus a gaussian peak will have a very small uncertainty in the peak location (about half a GeV). The best result will come by combining the mass hypotheses of the various samples. I expect that a true 5-sigma result will produce a <1 GeV error on the mass, so we will indeed understand whether CMS and ATLAS are compatible or not with a good accuracy by then. Cheers, T.
    The CMS ZZ paper shows an small mZZ excess at ~320 GeV. That's exactly where two of the five CDF ZZ events were (both spot on 325). From the CMS paper it's too narrow to be a Higgs, and they take special pains to point out the ZZ system is significantly boosted in pT for these events. Were the CDF events at 325 boosted? Is anyone piqued about this?

    Hello Andrew,

    I do remember the CDF ZZ events at 325... And it is true that CMS has some events there:

    and also ATLAS does:

    However, while the CMS H->ZZ->4l limit does show a upward departure of the observed limit with respect to the expected one, strangely the ATLAS Higgs limit at 325 from the ZZ->4l channel alone seems exactly in line with the expected one, hmmm. Don't know why. See below:

    In any case, the thing requires to be watched, I agree. The CDF events were boosted for sure, because they fit to the ZZ hypothesis, so each dilepton pair had a significant combined momentum; but maybe I do not understand exactly your question.

    The CMS paper said that in those events near 325, the entire ZZ system was boosted more than 50 GeV in pT; i.e. as if they were recoiling against something else unreconstructed.

    Very interesting indeed. I looked at the DZERO searches for ZZ->4l by the way, and I did not see that kind of effect - they have two events with masses compatible with 325, but lousy mass resolutions; the 4-body Pt is however compatible with zero (but not really incompatible with 50 GeV, hehm):

    So the next question is: what is the Pt of the ATLAS events ? And what sort of twisted process could give rise to such a weird signature ? A Higgs of some kind produced against something escaping in another dimension ?