News at work

Output of the neural network used to separate 125-GeV/c2 Higgs boson signals from the multi-jet background. The multi-jet background is very large, so the lower plot shows the output with the multi-jet contributions removed. The data tends to follow the background prediction, which implies no Higgs boson has been observed.

Searching for gold is a hard game—you spend so much time digging through the dirt. But if you persist, you may find a huge bounty of treasures … much like searching for the Higgs boson.

Scientists have been seeking the Higgs boson for almost 50 years and have recently found evidence for it from the Tevatron. They have also observed a 125-GeV/c² Higgs-like particle at the LHC.

The unique all-hadronic Higgs channel at CDF searches for Higgs decays into pairs of bottom quarks accompanied by two additional quarks. The search could uncover a huge treasure trove of Higgs particles, but you have to dig through a multi-jet background that’s a million times larger than the Higgs. Another difficulty is that the quarks are not directly observed in the CDF detector since they immediately fragment into jets.

CDF physicists took on the challenge of searching for these quark combinations and improved measurements of variables sensitive to Higgs boson production, such as the width of jets. They used a neural network (a computer "brain”) to separate the multi-jet background from the golden Higgs events.

The physicists improved the energy resolution of bottom quark jets by 18 percent, which increased the potential Higgs yield by 10 percent. They then used information from CDF’s trackers and calorimeters to measure a subtle difference in jet widths between those of quark jets and those of gluons, which look almost identical to quark jets. (True gluon jets tend to fragment over a slightly larger area.) This measurement helped to separate the golden Higgs events from the background. The improved measurements of the bottom quark energy, jet width and other information were fed to a special two-layer neural network to simultaneously identify Higgs bosons produced by three different processes.

Despite the enormous challenges, these innovations improved the search’s sensitivity by 40 percent compared to our previous search, which is the equivalent of gaining an additional 2.5 years of Tevatron data. The limits placed on the Higgs search are given in the lower plot, which shows that for a Higgs mass of 125 GeV/c², the expected (observed) limit is 11.0 (9.0) times the Standard Model prediction. The CDF all-hadronic Higgs search is as sensitive as CDF’s ttH and H → γγ searches and has not been attempted at the LHC. These innovations also have applications to future multi-jet searches.

Learn more

—edited by Andy Beretvas

Observed and expected 95 percent credibility level upper limits on (WH, ZH, vector boson fusion) Higgs cross section times the branching ratio for Higgs → bb divided by the Standard Model prediction, as a function of the Higgs mass.

These CDF physicists contributed to this data analysis. Top row from left: Yen-Chu Chen (Academia Sinca, Taiwan) and Francesco Devoto (University of Helsinki, Finland). Bottom row from left: Ankush Mitra and Song-Ming Wang, both from Academia Sinca, Taiwan.