Higgs Pair Production at the LHC and future colliders

Juan Rojo
VU Amsterdam & Nikhef Theory group
High Energy Experiment Seminar
Boston University Physics, 26/11/2018
Juan Rojo HEE Seminar, BU Physics, 26//2018
1
Higgs Pair Production at the LHC
and future colliders

2
Fingerprinting
the Higgs sector

3
Huge gap between Higgs and Plank scales?
Elementary or composite? More Higgs bosons?
Coupling to Dark Matter? Role in cosmological
phase transitions?
Is the vacuum state of the Universe stable?
1 GeV (Proton mass)
125 GeV (Higgs mass)
1019 GeV (Planck scale)
With radiative corrections,
the natural value of the
Higgs mass is Planck scale
Degrassi et al 12
3
The Higgs boson
Open questions in particle physics

4
The Higgs boson
4
Weakly interacting massive particles?
Neutrinos? Ultralight particles (axions)?
Interactions with SM particles? Self-
interactions?
Structure of the Dark Sector?
Dark matter
phase transitions?

5
The Higgs boson
5
Weakly interacting massive particles?
Neutrinos? Ultralight particles (axions)?
Interactions with SM particles? Self-
interactions?
Structure of the Dark Sector?
Dark matter
phase transitions?
Why 3 families? Origin of masses, mixings?
Origin of Matter-Antimatter asymmetry?
Are neutrinos Majorana or Dirac? CP
violation in the lepton sector?
Quarks and leptons

66
Crucial information on these fundamental questions will be provided by the LHC:
the exploration of the high-energy frontier has just started!

7Juan Rojo HEE Seminar, BU Physics, 26//2018
Fingerprinting the Higgs
In the SM the properties of the Higgs sector are uniquely determined
any deviation from the tight SM predictions smoking gun for physics beyond the SM
Required both in experimental data and in theory calculations
A precision of a few percent in Higgs couplings measurements is the goal
Snowmass 2013

EW symmetry breaking
Yukawa/couplings between Higgs
and SM bosons and fermions
proportional to mass
The Higgs boson is responsible to
break EW symmetry and give
particles mass
However, we lack understanding of
why and how EWS is broken:
what generates the weak scale?
What are the dynamics underlying
EW symmetry breaking?
What about first and second
generation fermions?

Higgs Pair Production
Double Higgs production allows accessing crucial components of the Higgs sector:
Reconstructing the full electroweak symmetry breaking potential
Probing the Higgs self-interaction
Asessing the doublet nature of the Higgs by means of the hhVV coupling
In the SM, hh rates are small: even in the gluon-fusion production mode, the cross-section at 14 TeV
is only 40 fb, and suppressed by branching fractions

xsecs in fb

Each possibility associated to completely different EWSB mechanism, with crucial implications for the
hierarchy problem, the structure of quantum field theory, and New Physics at the EW scale
Current measurements in single Higgs production probe Higgs potential close to minimum
Double Higgs production essential to reconstruct full potential and clarify EWSB mechanism
In the SM the Higgs potential is fully ad-hoc: many other EWSB mechanisms conceivable
Higgs mechanism Coleman-Weinberg mechanism
Arkani-Hamed, Han, Mangano, Wang, arxiv:1511.06495

Due to small rates in the SM, only final states where at least one of the Higgs bosons
decays into a bb pair are experimentally accessible at the LHC
∝ λ

Rates for double Higgs production generically enhanced in BSM scenarios, and LHC
searches in various final states have already started at Run I

19
via Vector-Boson Fusion
F. Bishara, R. Contino and J. Rojo, EPJC 77 (2017) 481, arXiv:1303.6636

The potential of the Higgs-like scalar can be then written as
Parametrize the couplings of the Higgs-like scalar h to the SM bosons and fermions (after
EW symmetry breaking) without assuming that it is part of an SU(2) doublet
Higgs Production at High Energies
All coefficients are in principle (model independent) free parameters to be constrained by data

Is this cancellation exact (as in SM, c2V= cV2) or only approximate (BSM, c2V ≠ cV2 )?
No model-independent information on c2V available so far at the LHC
Even for small deviation of the SM couplings, striking signals within the reach of Run II
In the absence of the Higgs, the amplitude for vector-boson scattering (VBS) grows with the partonic
center-of-mass energy, until eventually unitarity is violated
In the SM, the Higgs boson unitarizes the high-energy behaviour of VBS amplitudes
at high energies
∝ c2V ∝ c2
V
∝ cVc3

Is this cancellation exact (as in SM, c2V= cV2) or only approximate (BSM, c2V ≠ cV2 )?
No model-independent information on c2V available so far at the LHC
Even for small deviation of the SM couplings, striking signals within the reach of Run II
at high energies
ggF
VBF

Signal and background

Tiny SM rates at the LHC
QCD multijet background
overwhelming
O(1) deviations in Higgs
couplings lead to factor 20
increase in signal xsec

Gluon-fusion production
dominates total xsec
Can be efficiently
suppressed by VBF cuts
100 TeV required for
ultimate study of this chanel

Analysis strategy
In the presence of anomalous Higgs couplings, the final 4b state could be highly boosted
Define unified analysis strategy suitable both for SM (no boost) and bSM (large boost)

Analysis strategy
The huge QCD jet backgrounds can be reduced by exploiting the VBF topology
Require two forward jets, separated in rapidity, plus a veto in hadronic activity in central region
Additional cuts in the reconstructed Higgs invariant mass and the di-Higgs invariant mass mhh further
reduce the QCD multijet cross-sections

29
Analysis strategy
Forward coverage
instrumental at FCC for
Higgs physics
The huge QCD jet backgrounds can be reduced by exploiting the VBF topology
Require two forward jets, separated in rapidity, plus a veto in hadronic activity in central region
Additional cuts in the reconstructed Higgs invariant mass and the di-Higgs invariant mass mhh further
reduce the QCD multijet cross-sections

Results
After analysis cuts, backgrounds are still overwhelmingly large for mhh close to threshold
For BSM couplings, the ratio of cross-sections between BSM and SM increases as we probe the
regions of large mhh, eventually dominating over backgrounds
This is the key of the LHC/FCC sensitivity to c2V despite the tiny SM cross-sections

Results
The sensitivity to c2V improves the
higher the values of mhh accessible
A model-independent measurement
of c2V within 50% possible at RunII
Ongoing searches for VBF di-Higgs
from ATLAS and CMS

32
Boosting Higgs Pair Production
with Machine Learning
M. Gouzevitch, A. Oliveira, J. Rojo, R. Rosenfeld, G. P. Salam and V.
Sanz, JHEP 1307, 148 (2013), arXiv:1303.6636
J. K. Behr, D. Bortoletto, J. A. Frost, N. P. Hartland, C. Issever and J.
Rojo, EPJC76, no. 7, 386 (2016), arXiv:1512.08928

The 4b final state offers largest rates but overwhelming QCD multijet backgrounds
Made competitive requiring the di-Higgs system to be boosted and exploiting kinematic
differences between signal and QCD background with jet substructure
Boosted b-tagging by ghost-associating mass-drop tagged large-R jets with b-tagged
small-R jets
Tagging these events is challenging due to the very high rate of QCD multijets but doable
Scale-invariant tagging: event-by-event classification depending on final state topology
ggF di-Higgs in the 4b ﬁnal state

ggF di-Higgs in the 4b ﬁnal state
Similar contributions from the boosted, intermediate, and resolved
categories to the overall signal efficiency

Resolved vs boosted
Similar contributions from the boosted, intermediate, and resolved
categories to the overall signal efficiency
BoostedResolved

Analysis strategy

Results of cut-based analysis
After the (non-optimised) cut-based analysis the overall signal
significance is still very poor
b-jet mistags are dominant contribution to QCD backgrounds

Results of cut-based analysis
How to determine the optimal set of kinematic cuts
on differential distributions that maximise signal
significance? Exploit Machine Learning tools!

ML discriminator
Artiﬁcial Neural Network
Trained on signal and background MC events
Higgs pT
Higgs m
di-Higgs m
ECF
𝛕12
Subjet pT
…..
Input
Output
Signal?
Background?

ML discriminator
Given a set of Nvar kinematic variables {ki} associated to event i, and a set of ANN weights {ω}, the
ANN output yi interpreted as probability that event originates from signal process
with y’i the true MC classification: y’i=1 for signal, y’i=0 for background
The general classification probability including background events is
Thus the error function to be minimised during the training is the cross-entropy:
ANN training performed with Genetic Algorithms using cross-validation stopping

ML discriminator
Background
rejected
Signal
accepted
Optimal signal/background discrimination from ML-driven combination of kinematical info

Results
Use of multivariate techniques allows to substantially improve the signal significance for this
process as compared to a traditional cut -based analysis
The total combined significance is enough to observe Higgs pair production in the 4b final
state at the HL-LHC. Substantial improvement if reducible backgrounds (fakes) can be eliminated

Results
The total combined significance is enough to observe Higgs pair production in the 4b final state
at the HL-LHC. Substantial improvement if reducible backgrounds (fakes) can be eliminated
Pre-MVA
result

Opening the black box
ML tools often criticised as black boxes, with little understanding of inner working
ANNs are simply a set of combined kinematical cuts, nothing mysterious in them!
Plot kin distributions after and before the ANN cut to determine the effective kinematic cuts
This info enough to perform a cut-based analysis with similar signal significance

Higgs-like peak sculpted in QCD background
Opening the black box
ML tools often criticised as black boxes, with little understanding of inner working
ANNs are simply a set of combined kinematical cuts, nothing mysterious in them!
Plot kin distributions after and before the ANN cut to determine the effective kinematic cuts
This info enough to perform a cut-based analysis with similar signal significance

What about @ 100 TeV?
It turns out that the bounds on the Higgs
self-coupling in the 4b final state become
worse at 100 TeV than at 14 TeV …
Reason is that QCD backgrounds grow
faster than signal with CoM energy
FCC Yellow Report 16

50
more about di-Higgs

Self-coupling from VBF
Arganda et al 18
Information on the Higgs
self-coupling can also be
obtained from VBF channel
Complementary (but not
competitive) with constraints
from gluon-fusion di-Higgs

High-precision di-Higgs calculations
Dreyer et al 18
hh VBF @ N3LO hh ggF @ NNLO with mtop effects
Grazzini et al 18

Self-coupling from single Higgs
Pagani et al 17
Information on the Higgs self-coupling
can also be obtained from single-
Higgs via radiative corrections
Complementary with constraints from
di-Higgs production within global fit

Higgs pair production holds unique potential
to uncover BSM dynamics at the LHC!
Higgs pair production is a cornerstone of the LHC program, providing direct information on
the EWSB potential and testing the nature of the EWSB mechanism
The 4b final state offers the highest yields, but requires clever analysis techniques for
taming the overwhelming QCD background both theoretically and experimentally
In the gluon-fusion channel, discovery is guaranteed at the HL-LHC, and competitive
constraints on the self-coupling can be imposed already at Run II provided systematic
uncertainties of the 4b cross-section can be kept under control
In the VBF channel, the steep rise of the cross-section with energy for anomalous
coupling is the key for the high sensitivity of this process to c2V
For such complex final states, the ultimate signal optimization of the extraction of the
Higgs couplings benefits from the use of machine learning methods
The di-Higgs frontier at the LHC

Higgs Pair Production at the LHC and future colliders

Recommended

Recommended

More Related Content

Similar to Higgs Pair Production at the LHC and future colliders

Similar to Higgs Pair Production at the LHC and future colliders (20)

More from juanrojochacon

More from juanrojochacon (20)

Recently uploaded

Recently uploaded (20)

Higgs Pair Production at the LHC and future colliders