Hello! Last week I was at the workshop hh searches with CMS, which was organized in Lyon by the IPNL (Institute de Physique Nucleaire de Lyon).

lyon_with_ipnl_logo
Vieux Lyon across the Saone River with the IPNL logo superposed.

It was an internal CMS meeting aimed at reviewing the status of ongoing Higgs pair production and easing future combinations. There, I had a short talk to summarize what we have been working on lately, which was well-received. I cannot disclose details of the analysis which are not yet published by the CMS Collaboration, but I will just say that great progress is being made and provide some general concepts on Higgs studies at the LHC, which I will refine in future posts.

One of the objectives of the AMVA4NewPhysics network is to develop and use new statistical techniques for Higgs boson studies at the LHC. With Higgs boson studies, we are referring to data analyses which look for processes where a Higgs boson has been produced in order to measure its properties and look for deviations from the Standard Model (SM) predictions.

Even when only SM processes are considered, the production of Higgs bosons can assume many different forms. The by far most probable way to produce a single Higgs boson at a proton-proton collider at TeV energies is through a process called gluon fusion, which was used for the discovery in 2012. While their rate is lower, other processes like vector boson fusion (VBF) where a Higgs is produced with two additional quarks and associated production with W, Z and top quark pairs (i.e. ttH),  are currently being studied in detail at the LHC.

higgs_boson_mech
Leading order diagrams for different modes of production of a single Higgs boson at a hadron collider: a) gluon fusion,  b) vector boson fusion, c) associated production with W/Z and d) associated production with top quark pairs. Figure from ETZ-Zurich Grab Group.

But that is not the whole story, because we do not detect Higgs bosons, but their decay products (or sometimes decay products of those). As a consequence, even when only a single Higgs boson is produced, we can search for it in many final states. The decay modes of a 125.09 GeV Higgs boson predicted by the SM are listed in the table below.

h_decays
Branching ratio for the different decays of the Higgs boson in the SM. Table from ATLAS/CMS couplings combination.

Nevertheless, the branching ratio (relative probability of a certain decay) is not the only thing to take into account, but also the expected backgrounds of your analysis (other processes which look like what you are looking for, as I explained in my previous post). The fact that the Higgs boson was discovered using two channels with very low branching ratio (~0.23% for H → γγ and ~0.013% for H → ZZ → l+ll+l) but with low background contributions and high mass resolution clearly supports the previous point made.

The study of alternative Higgs production mechanisms can provide final state topologies with lower backgrounds and/or be used to measure important Higgs properties. To exemplify the first case, take for example the search for H → bb, which is completely dominated by QCD multijet background in the gluon fusion channel, but which is it traceable when you consider VBF processes (i.e. two extra quarks in the event) or associated W/Z production.

For the second case, the most relevant example for the time being is ttH production, which can be used to measure directly the top-Higgs Yukawa coupling (which might be sensible to BSM physics).  For the most accurate results for Run I (not enough Run II to be competitive yet) on Higgs properties, check the CMS and ATLAS combination.

But you might say, what does Higgs pair production have to do with any of this? It allows to directly measure the coupling of the Higgs boson to itself λHHH (also referred as Higgs trilinear coupling). Any deviations of λHHH with respect to the SM prediction would mean that the Higgs potential is not as expected and give clear signs of New Physics.

However, Higgs pair production processes are extremely rare at the LHC (i.e. tiny cross sections), about 1000 times less likely than single Higgs production (which was already very rare). Because now two Higgs bosons are produced, the number of possible final states is half of the square of those for single Higgs production. Therefore, apart from having a very small cross section, the events are distributed over many possible final states, so having a large branching ratio is now crucial.

Many feasibility studies have been done, studying how well the different final states could be measured at the LHC, but most of them conclude that we would have to wait for the full luminosity of the HL-LHC for valuable measurements.

The main focus of my project within the AMVA4NewPhysics network is to develop the most performant statistical techniques for pushing the state-of-the-art in Higgs analysis which uses the H → bb channel. In the most promising Higgs pair production channels, one of the Higgs bosons always decays to bottom quarks to keep large branching ratios, while the other decays to cleaner final states as W+W, τ+τ or γγ .

While the products of my project might help in those channels, we have chosen the HH → bbbb channel as the benchmark and main playground for the techniques we will develop. The main advantage of HH → bbbb is that it accounts for the largest branching ratio, namely about 33% (0.575*0.575), of the total Higgs pair events produced. However, jet combinatorics and QCD backgrounds are very large, making this channel quite challenging, and even regarded as not useful at the HL-LHC by some. Other feasibility studies, as those carried out by members of this network and described in a previous blog post show MVA techniques for distinguishing signal and background can make this channel competitive.

It is also worth mentioning that the production rate of Higgs pairs might be greatly enhanced in some New Physics theories with respect to the SM, which might ease our task. To conclude, we are optimistic and have some ideas to further increase the sensibility of the HH → bbbb channel by countering its main drawbacks with statistical techniques, about which you will be able to read here soon. Comments are welcome!