The goal of my project was to increase the significance of the semi-leptonic double higgs channel. In 2012 we were able to experimentally prove the existence of the higgs particle. Since then we have been trying to verify that its interactions with the other particles in the standard model match what our theory predicts. This has been very successful and we now have tight experimental bounds on many of the couplings between the higgs and other particles in the standard model. However, we do not have tight experimental bounds on the higgs self-coupling term. It is important to experimentally determine the higgs self-coupling both to test standard model predictions and also because this term is important to electroweak symmetry breaking. Electroweak symmetry breaking has important consequences for the origin of matter in the universe.

To experimentally determine the higgs self-coupling, we first have to be able to identify double higgs events at the Large Hadron Collider. This is difficult however because the primary decay channel of the double higgs suffers from an enormous background process originating from ttbar. We studied one of these decay branches where both ttbar and hh decay into the semi-leptonic branch with final decay products of two b-quarks, two light quarks (up, down, charm, strange), 1 lepton, and 1 neutrino.

Fig.1. On the left is a semi-leptonic double higgs decay. On the right is a ttbar semi-leptonic decay. One can see that the final products of the decay are identical for both decays.

Since all the decay products which are actually detected by CMS are the same for the signal and the background we must rely on a wide suite of kinematic variables to differentiate between the two processes. For example this might be the angle between the two b-quarks, which is generally small for hh and large for ttbar. The goal is to combine all of the differences in these kinematic variables together to genearlly determine what is a double higgs event and what is a ttbar event. This channel has historically been considered extremely difficult to work with and generally results in negligible significance when compared to other double higgs modes. We simulated the proton-proton collisions at the HL-LHC, then simulated the CMS detector's response to these events and extracted the kinematics variables. We then took this information and fed it into a neural network to separate the signal from the background.

Over the course of the research we identified seven new kinematic variables not used before in the literature that are useful for separating the signal from the background. We used these new variables in addition to ten low-level kinematic variables and four high-level kinematic variables that had been previously applied to the di-leptonic channel [2].

Fig. 2. The first two new kinematic variables dr_lj1 and dr_lj2 measure the angle between the lepton and the two light jets individually. We can see good separation between the signal and the background. The background generally has the lepton and light jet at 180 degrees, while in the signal they are generally travelling in the same direction.

Fig. 3. The next two new kinematic variables dr_lb1 and dr_lb2 measure the angle between the lepton and the two b-jets individually. We see that for the background one of the b-jets is travelling in the same direction as the lepton and the other is moving the opposite direction. For the signal we see that both b-jets are generally moving opposite the lepton. This also shows promising separation though not as much as dr_lj1 and dr_lj2.

Fig. 4: This variable m_ln shows the difference in invariant mass of the lepton-neutrino system between the signal and background if the z momentum of the neutrino is reconstructed on the hypothesis that it came from a double higgs.

Fig. 5: This variable Tm_ln shows the difference in invariant mass of the lepton-neutrino system between the signal and background if the z momentum of the neutrino is reconstructed on the hypothesis that it came from a ttbar.

Fig. 6: This variable m_lnjj shows the difference in invariant mass of the lepton, neutrino, and two light jet system between the signal and background if the z momentum of the neutrino is reconstructed on the hypothesis that it came from a double Higgs.

In a rough sense, the way we measure how well we were able to separate the signal from the background is by using statistical significance. The higher the significance, the better we did at differentiating what was a signal event and what was a background event. Previous studies had reached a significance of 0.13 sigma [1]. Using all 21 kinematic variables together, we were able to achieve a significance of 0.427 sigma.

References

[1] A. Adhikary, S. Banerjee, R. K. Barman, B. Bhattacherjee, and S. Niyogi. Revisiting the non-resonant higgs pair production at the HL-LHC. Journal of High Energy Physics, 2018(7), jul 2018.

[2] J. H. Kim, M. Kim, K. Kong, K. T. Matchev, and M. Park. Portraying double higgs at the large hadron collider. Journal of High Energy Physics, 2019(9):1–36, 2019.

Acknowledgments

Firstly I would like to thank the National Science Foundation for funding this research. I would also like to thank Dr. Ivanov for being my research advisor. I would like to thank Dr. Flanders, Dr. Greenman and Kim Coy for making this REU possible and fun. I would like to thank Dr. Jeong Han Kim and collaborators for sharing their code. Finally I would like to thank the rest of the REU for making the summer fantastic.

Final Presentation