Probing new physics with pairs of Higgs bosons

28th July 2021 | By

Life after the Higgs boson discovery has been filled with many interesting questions about how to fill in the gaps of the Standard Model. What if there are more Higgs bosons? Could the Higgs boson be involved in how gravity works? A variety of extensions to the Standard Model predict new particles that could be observed via processes involving the Higgs boson.

The ATLAS Collaboration has released a new result searching for pairs of Higgs bosons (HH) produced by new particles. The Higgs bosons would then each decay into pairs of bottom (b) quarks – known as the '4b decay channel' – which accounts for 34% of all HH events produced in LHC collisions. These b-quarks create collimated spray of particles (known as jets) that can be recorded in the ATLAS detector.

Figure 1: Distribution of the background estimate for the resolved channel compared to observed data. A new particle would show up as a “bump” in the observed data (black) relative to the background estimate (turquoise), and the expected shapes of these bumps are shown for a variety of different signal masses in the various colours. (Image: ATLAS Collaboration/CERN)

Though the 4b channel may have the most data, nothing is easy in the search for HH. First, physicists need to find the b-quarks – a challenge in itself, as other, more common quark and gluon processes also produce a big muddle of jets. ATLAS physicists have developed powerful algorithms, called b-taggers, to sort through this mess and pick out jets created from b-quark decays out of the proverbial haystack. With four b-quarks in the final state, the 4b decay channel takes full advantage of these b-taggers!

However, these powerful algorithms can only do so much; simulations are unable to accurately predict a large portion of this difficult background. Physicists thus use sophisticated machine learning techniques to learn what these events look like directly from LHC data.

Despite these challenges, the HH→4b analysis allows physicists to search for new particles across an incredibly broad mass range, spanning from the lowest (250 GeV) to the highest (3000 GeV) masses, and offers leading sensitivity across most of this range. This is done by considering two separate analysis strategies, one for the resolved (low masses, up to 900 GeV) and another for boosted (high masses, above 1500 GeV) regimes. At intermediate masses (900–1500 GeV), the two strategies are combined for optimal sensitivity.

The HH→4b analysis allows physicists to search for new particles across an incredibly broad mass range, spanning from the lowest (250 GeV) to the highest (3000 GeV) masses.

Figure 2: Limits on the production of a spin-0 particle which decays to HH as a function of new particle's mass m(X). The resolved channel (dashed pink) sets limits from 251 GeV to 1.5 TeV, and the boosted channel (dashed blue) sets limits from 900 GeV to 3 TeV. The dashed black shows the expected limit from the combination of both channels, while the solid black shows the corresponding observed limit. The expected and observed limits agree well within uncertainties, corresponding to no significant excesses observed. (Image: ATLAS Collaboration/CERN)

In the resolved analysis, physicists identify four, small-radius b-jets to take to the village matchmaker. In this case, it's a Boosted Decision Tree (BDT) that pairs up the jets to create the most “HH-like” topology. Then, after accounting for background processes that can leave a similar signature in the detector, ATLAS physicists search for the tell-tale "bump" in the HH mass distribution that would indicate the presence of a new particle (see Figure 1)

As this mass increases, the Higgs-boson decay products get closer and closer together, such that the small-radius b-jets used for the resolved analysis start to overlap. This is where the boosted analysis takes over, reconstructing individual Higgs bosons in the ATLAS detector as single large-radius jets. Events are then separated based on the number of b-tags within these jets. While both jets are expected to contain two b-quarks, sometimes only one of these is identified by the tagger. Physicists make separate predictions for events with two, three or four total b-tags, maximising the sensitivity of the analysis in this high mass regime.

In the end, no sign of new physics is seen and limits are set for the production of HH via the decay of a hypothetical new spin-0 or spin-2 particle (see Figure 2). These limits are stronger than previous results due to ATLAS improvements in b-tagging, the introduction of machine learning techniques, and the use of the full Run-2 dataset. As more data are collected and physicists further refine and improve their methods, ATLAS' sensitivity will only grow – extending towards new physics processes that currently remain just out of reach.

About the event display: A data event from 2018 that passes the resolved signal region event selection. The value of the corrected m(HH) is 629 GeV, the masses of the two Higgs boson candidates are 111 GeV and 116 GeV.