Another chapter in the saga of the search for the elusive, but dominant, decay mode of the Higgs boson has been reported by the CMS collaboration last month. This is one of those specific sub-fields of research where a hard competition arises on the answer to a relatively minor scientific question. That the Higgs boson couples to b-quarks is indirectly already well demonstrated by a number of other measurements - its coupling to (third generation) quarks being demonstrated by its production rate, for example. Yet, being the first ones to "observe" the H->bb decay is a coveted goal.
In this blog I reported recently about a CMS and an ATLAS search for H-->bb decays. CMS looked for  events where the Higgs boson is produced at large momentum by recoiling against an energetic gluon (such that the two b-quarks end up producing a single "boosted" jet); ATLAS looked for events where the Higgs boson is produced in association with vector bosons. While the size of the CMS "signal" was too small to call it a signal (significance of about 1.5 standard deviations), the very challenging events considered made the analysis quite interesting in my opinion; the ATLAS one was a much more solid 3.5 standard deviations one, and was indeed one of the results showcased by the collaboration in summary talks at the recent EPS conference, last July in Venice.

Now, CMS also publishes a preliminary result of the search for VH events, where "V" labels either a W or a Z boson. The Higgs boson can in fact be produced in association with vector bosons, when these "radiate" a Higgs boson thanks to the large coupling to it. The Higgs can then be seen to decay to b-quark pairs directly, above large but manageable backgrounds. To do that, you need to see two hadronic jets (collimated sprays of particles produced by the fragmentation of quarks or gluons) that both appear to originate from b-quarks.

Backgrounds, backgrounds: indeed, finding a vector-boson decay signal is "easy", when these decay to leptons. You can trigger the event data collection if you see the signal of an energetic electron or muon, or large missing energy produced by the two neutrinos into which a Z boson decays 20% of the time. By doing that you can "easily" acquire a sample of events where at least one W or Z is present. The problem is that the WH or ZH production rate is oh-so-much smaller than the production rate of events where the W or Z is created in association with two hadronic jets. The H->bb signal has to therefore be dug out of all those concurrent processes.

Processes that mimic the WH signature (with H->bb) are top pair production, W or Z plus jets production, single top quark production, diboson production, and QCD backgrounds where there is no W or Z boson in the first place, and the leptonic signal that triggered the event collection is spurious. Each of these processes needs to be reduced as much as possible, by exploiting their different kinematical properties with respect to those of the signal.

CMS uses a multivariate technique that uses many observable quantities in combination. This is a "Boosted Decision Trees" algorithm, a powerful discriminant. Each of the various decay modes of the W and Z boson to leptons is considered separately, as they have different background composition. One BDT distribution is shown below as an example: it is the category of events with two energetic muons, relics of a Z boson decay. As you can see, the ZH signal is orders of magnitude more rare than the sum of concurrent processes, but the discriminant "pushes" the signal to the right, and all backgrounds to the left, such that some excess from ZH events can be sought for.

As usual, above the data is represented by black points with error bars, and all backgrounds are shown by filled histograms. The signal expected for a 125 GeV Higgs boson is the red empty one. Since the graph is a "semi-log" plot, where the y axis is in logarithmic scale, it is not easy to realize which of the backgrounds has the largest effect. It is actually the "Z+bb" one, where indeed a Z boson is present as well as two b-quark jets - thus making the final state an "irreducible" background, as it contains the same objects expected from ZH production.

After all sub-channels are put together, the situation is the one shown below. Indeed, at the highest end of the BDT spectrum, some faint excess can be observed.

Above, all backgrounds are collectively painted in grey, and the fitted signal is in full red. THe bottom panel shows the background-subtracted distribution of the data, which does have an excess compatible with the Higgs component one expects (red empty histogram).

All in all, the excess has a statistical significance of 3.3 sigma, thus equivalent to the ATLAS one reported recently. Given these >3 sigma effects, and the rate consistent with expectations, hould we declare that the Higgs boson does indeed decay to b-quark pairs as expected ? If you ask me, yes. Unfortunately, though, it has become customary of HEP experiments to stick doggedly to the "five-sigma" rule, which requires that a signal be seen above that significance level before it is definitively claimed as observed. It is quite silly, in my opinion, because the five-sigma rule has been designed to address the "look-elsewhere effect" that plagues typical searches for new particles. In this case, there is no look-elsewhere effect at work. However, I have discussed this issue at length in a seminar I gave in various institutes and labs in the recent past, so I will not get into the details again here... 


Tommaso Dorigo is an experimental particle physicist, who works for the INFN at the University of Padova, and collaborates with the CMS experiment at the CERN LHC. He coordinates the European network AMVA4NewPhysics as well as research in accelerator-based physics for INFN-Padova, and is an editor of the journal Reviews in Physics. In 2016 Dorigo published the book “Anomaly! Collider physics and the quest for new phenomena at Fermilab”. You can purchase a copy of the book by clicking on the book cover in the column on the right.