Higgs Boson Mass: CMS On Top!
    By Tommaso Dorigo | July 3rd 2014 02:36 PM | 12 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
    A couple of weeks ago I reported here about the new measurement of the Higgs boson mass produced by the ATLAS experiment. That determination, which used the full dataset of Run 1 proton-proton collisions produced by the LHC in 2011-2012, became and remained for two weeks the most precise one of the Higgs mass. Alas, as I wrote the piece I already knew that CMS was going to beat that result very soon, but of course I could not say anything about it... It ached a bit!

    That's why I am quite happy today to report that CMS has again taken the lead in this LHC-internal competition. One which is less idle than it looks: the two detectors were designed, over ten years ago, with the explicit purpose of discovering and measuring the Higgs boson properties as precisely as possible. The hardware choices made by the two experiments were quite different, and so it is quite interesting to see who ended up having the upper hand in the precision of the Higgs boson mass measurement.

    Measuring the mass of a particle with high precision entails not only collecting as many of its decays as possible -something which requires a high-acceptance, redundant detection system. It of course also requires that the subdetector components devoted to measuring the energies and momenta of the decay products are as precise as possible, and that they enable a successful intercalibration.

    For the Higgs boson, which decays a few times in a thousand into a pair of energetic photons, determining the energy of those photons as precisely as possible became an obsession during the design and construction of the CMS detector. The electromagnetic calorimeter which does that job is a technological marvel, composed of thousands of crystals of lead tungstate, an optimal material due to its transparency, high atomic number, and response.

    Similarly, the reconstruction of electrons and muons from Z boson decays became the other crucial focus of the detector builders, as the Higgs may decay to two Z bosons and those two Z bosons will sometimes yield a signature of four energetic electrons or muons. To measure those particles with high precision, CMS chose an all-silicon inner tracker, and a very redundant muon detection system made up of three different types of muon detectors. For electrons, the same electromagnetic calorimeter optimized for photons was of course the right tool.

    So you can see how obtaining the best precision on the measurement of the mass of the Higgs boson can be seen as a proof of the winning strategy for detector design. And I am glad to see that once both LHC experiments have spelled their last word on the Higgs mass, the one coming out on top is indeed CMS.

    (Above: the CMS spectrum of the invariant mass of photon pairs passing the Higgs candidates selection (top), and the background-subtracted signal at 125 GeV (bottom). The events entering these graphs have been weighted by the expected signal over signal plus background fraction in the various categories from which the data comes. So, e.g., events in the vector-boson-fusion category have a larger weight, as there is less background there. This method allows one to see more clearly the Higgs signal, in a way not dissimilar to how the likelihood "sees" the data.)

    The measurement released today is MH=125.03 +- 0.30 GeV, a combination of the determinations (fully compatible, in the case of CMS) of the gamma-gamma and ZZ decay modes. If you compare it to the corresponding ATLAS number, MH=125.36+-0.41 GeV, you might at first sight say "what's the big deal?". Indeed, the two numbers are consistent with one another, and the uncertainties are both in the few tenths of a GeV region. However... 0.41 GeV is a lot larger than 0.30 GeV!

    The ATLAS Higgs mass uncertainty is 37% larger than the CMS one. This would occur for two systematics-free, equally precise experiments if the latter had collected 85% more data than the former. In other words, the better CMS measurement (a result allegedly due to a better detector and/or better software algorithms for energy and momentum measurements) can be "translated" into saying that for every Higgs event collected by ATLAS, it is like CMS collected 1.85 !

    The above is of course just a view on a small part of the whole set of measurements of Higgs boson properties that the two experiments have performed: cross section in the various production processes, branching ratios, couplings, spin, etcetera. In this post I do not have the time nor the energy to cover all of that, but I would not like to leave you with the impression that I am just picking the one bit where CMS is on top. In fact, the other measurements by the two experiments show a very good agreement -among each other and with Standard Model predictions- and in some cases ATLAS has a better precision; for instance, we know that ATLAS has since 2012 had a higher significance in its Higgs signals. Whether that was a statistical fluctuation, it is matter for another post.

    For more information on the CMS results on the Higgs boson, released for the ICHEP conference going on in Valencia (Spain), please see the CMS press release here, which contains a link to the preprint. I will devote a different article to a more careful examination of those results.


    It's 85% more, not less! Otherwise that would mean that for 100 particles collected by CMS, Atlas has seen just 15! A bit to extreme...

    Yep, corrected.
    Tommaso, great post.
    An a bit off topic question: how your group decided long time ago to join CMS instead of ATLAS?
    Do you know in general how the various italian groups made such a decision?
    I know eg that in Bologna they have groups in both collaborations..

    Hi Anon,

    the CMS group in Padova is a large one (I believe it is the third largest group in CMS), and contains physicists who have designed and built the CMS drift tubes for muon detection (the ones that give CMS its exterior appearance), and physicists who have developed and built the silicon sensors for the inner tracker. The choice to join a starting experiment depends on the possibility to play a role in the design and construction of the detector, so that is what drove Padova physicists to CMS instead of ATLAS.

    As for me, I joined CMS in 2002, when it was no longer a question of design of hardware but of software algorithms....

    Assuming the Standard Model holds, does this improved value prove the vacuum is unstable ?


    I think it gives a meta-stable universe, in the sense that its lifetime may be quite long but not infinite. How long ? I do not know - there is some debate whether it is longer or shorter than the age of the universe.

    Thanks Tomasso !

    My term "prove" was a little unscientific -- any idea how many sigma the Higgs/Top quark masses are from stability ?

    I might want to take my family on an expensive holiday to Hawaii if the universe is about to end :-)


    Oh, the world is surely about to end. Only, we do not know what "about" means. I suggest you have that holiday anyway.

    Tommaso, Tommaso, tsk tsk tsk. Sorry, but I must defend my beloved detector.

    You should compare systematic uncertainty numbers if you are going to go on about a supposedly 'better detector' or 'better algorithms'.

    ATLAS syst: +/- 0.18 (
    CMS syst: +/- 0.15 (

    So you are going to go to town riding on a 0.03 GeV difference? You surely know that systematics are rather flaky at that level.

    I think this is very neat, that such complementary approaches (crystal with high E resolution, compared to LAr with very good pointing resolution for photons, and very different muon systems) give such compatible results. Well done to both collaborations! Especially to CMS colleagues who managed to eek out a signal with such inferior equipment!

    (that last jab was just for fun ... let the fireworks begin!)

    Hmm, you picked two different numbers of your choice, and still ATLAS is 20% worse (or like, CMS events count 1.44 times as much). I think it validates my point.
    In any case, I don't wish to argue. ATLAS is a great detector, too bad for it there's CMS around ;-)
    It is just like CDF versus D0 at the Tevatron. You guys are D0, if you didn't get it :)

    You don't wish to argue, yet you start the argument with provocative statements. That's a bit strange. I guess that's the prerogative of the blog owner.

    If you care to look a little closer, those aren't 'random numbers of my choice', they are the systematic components of the respective total uncertainties. So totally not random. Rather, the opposite of random.

    Would you prefer to compare statistical uncertainties as some sort of metric of excellence? Or maybe the observed p-values?

    Then you compare 0.15 to 0.18 and quote a 20% improvement! Tell me, what level of noise do you think is latent in these numbers? Do you think one can quote an uncertainty to 5 MeV? If it was 0.01 and 0.005 would you say that one detector was twice as precise as the other?

    But, if you want to go around quoting 30 MeV differences in Higgs mass systematic as definitive proof of superior design/implementation, well, I guess you are free to live in such a fantasy world where systematic uncertainties are rigorously defined to many significant figures ... where editorial choices don't exist ... and where CMS actually treats uncertainties rigorously. Its just a pity that this has to feature so prominently in your science reporting.

    Dr. Never Worked At the Tevatron


    sorry that you never worked at the tevatron - it was a lot of fun.

    Concerning random numbers: the combined mass measurement of CMS is 125.03 +0.26 -0.27 (stat) +0.13 -0.15 (syst). You pick 0.15 while it would have been more fair to take the average of negative and positive systematics, 0.14. Then you say that that number is practically coincident with the one of ATLAS, 0.18. And you do this by arguing that 30 MeV are too little difference to base speculations on. But it is not the absolute value of the difference what counts here: what those 30 MeV are is a full 20% of the estimated systematics.

    In any case, I think statistics also plays a role in a generic comparison of two results. If one experiment decides on a tighter selection to cut down its systematics, the statistical uncertainty will suffer from it. So that is a reason why in the end it is on total uncertainties that comparisons are made between experimental results.