the ATLAS detector Recent searches for light scalars with

(1)

Adrian Salvador Salas (IFAE-BIST Barcelona) 21st March 2022

XIII CPAN Days (Huelva)

Recent searches for light scalars with

the ATLAS detector

(2)

Introduction

2 A. Salvador | XIII CPAN Days | 21 March 2022

➔Several BSM theories include an extended scalar sector:

◆Extended Higgs sector models (2HDM, Higgs triplets…) with at least one pair of charged Higgs bosons that can be lighter than the top quark

◆Models addressing the “flavour problem” (flavon models) or dark matter result in new particles, including light neutral scalars

➔Recent ATLAS analyses searching for light scalars:

◆Light charged Higgs bosons in t→bH^±(cb): ATLAS-CONF-2021-037

●Complementary to other light H^±ATLAS searches:

○H^±→cs Eur. Phys. J. C 73, 2465 (2013)

○H^±→τν JHEP09(2018)139

◆Light neutral scalar t→qX(bb) (q=c/u): Ongoing

●Complementary to the t→qH ATLAS search JHEP05(2019)123

➔

Search for H

⁺

produced in association with top and bottom quarks in 200 - 2000 GeV mass range

◆Full Run-2 dataset used (139 fb^-1

), previous publication based on 2015+2016 (36 fb

^-1

) 10.1007/JHEP11(2018)085

◆

Focused on the single lepton channel since it provides the best significance

(3)

H ➝ cb analysis: overview

➔Analysis using the full LHC Run-2 139 fb^-1

pp collisions recorded with the ATLAS detector

◆ATLAS-CONF-2021-037

◆First time this search is performed in ATLAS!

➔

Search for H

^±

→cb in the low H

^±

mass range of 60 - 160 GeV in tt̄ events

➔

Focused on the single lepton channel since provides the best significance

➔

Model independent upper limits on

ℬ

=

ℬ

(t→H

^±

b)⨯

ℬ

(H

^±

→cb)

➔

Search for H

⁺

), previous publication based on 2015+2016 (36 fb

^-1

) 10.1007/JHEP11(2018)085

◆

Based on 1908.00826, PhysRevD.101.035021

(4)

➔

Select events with:

◆Exactly one lepton: e^± or µ^±

◆MET ≥ 20 GeV and MET + m_T^W > 60 GeV

◆≥4 anti-kt4 jets

◆≥2 b-tagged jets at 60% efficiency + 1 loose b-tagged jet at 70%

●Tagging based on Deep Neural Network

➔

Main background from SM tt̄+jets

◆Smaller background than for H⁺→cs

due to the small contribution of BR(W→cb)~10⁻⁴

◆The tt̄+light reducible background contribution decreases with tightening the b-tagging

(x3 better rejection at 60% than 70%)

H ➝ cb analysis: Selection

https://link.springer.com/article/10.1

140/epjc/s10052-019-7450-8

(5)

➔

Classify events according to jet and b-jet multiplicities:

◆Either 4j, 5j and 6j combined with 3b, ≥4b or 2b+1bl

◆The categorisation constraints background uncertainties

➔

Categorisation of the 9 regions:

◆Main signal regions: 4j3b, 5j3b and 6j3b

◆tt̄+≥1b background control regions: 4j4b, 5j≥4b and 6j≥4b

◆Data-based corrections for tt̄+jets: 4j2b+1bl, 5j2b+1bl and 6j,2b+1bl

H ➝ cb analysis: Event categorisation

ATLAS-CONF-2021-037

2b+1bl 3b

tt̄ RW SR CR

≥4b

4j

5j

6j

(6)

6

H ➝ cb analysis: tt̄+jets MC correction

➔

Mitigate differences observed in data/MC distributions due to tt̄+jets mismodelling

➔

Data/MC-based factors extracted from 4j2b+1bl, 5j2b+1bl and 6j2b+1bl regions

1.Correct the 2b+1bl jet multiplicity distribution

2.Correct the 2b+1bl H_T^all = p_T distribution for each jet multiplicity

➢Corrections parameterised with a 1/H_T^all rational function

➔Factors applied appropriately to the SRs improve prefit agreement

A. Salvador | XIII CPAN Days | 21 March 2022

ATLAS-CONF-2021-037

l,jet,MET ∑

(7)

7

H ➝ cb analysis: Neural Network

➔

Use of a multivariate technique to separate signal and background in the signal regions

➔

Neural network description:

◆

Training performed inclusively on the ≥3b signal regions

◆

Based on low-level object information:

●p_T, η and φ of first six leading jets ordered in b-tagging

●Lepton p_T, η , φ, MET and φ_MET

●Di-jet mass variables of the fourth leading jet in b-tagging (expected to originate from the c-quark) and a selected b-jet:

●m_jbleading b-jet,m_jb^maxΔR,m_jb^minΔR

◆

H

⁺

mass is used as input parameter

◆

All H

⁺

mass samples included in a single training

●Simplifies training, more signal statistics and benefits from kinematics continuity

ATLAS-CONF-2021-037

4j3b 5j3b

(8)

8

H ➝ cb analysis: Fit results

➔

Simultaneous binned profile likelihood fit

◆One fit for each H^± mass hypothesis

◆Use binned NN output in the 3b regions and normalisation in the ≥4b regions

◆Systematic and statistical uncertainties included as nuisance parameters (NP)

●Included normalisation of tt̄+≥1b and tt̄+≥1c backgrounds as NP

➔

Model independent limits

ℬ

=

ℬ

(t→H

^±

b)⨯

ℬ

(H

^±

→cb)

◆Predictions from three 3HDM benchmark models: 1810.05430

➔

Broad ≥1σ excess

◆Largest excess: ~3σ at m_H±= 130 GeV

➔Five times better sensitivity compared to √s = 8 TeV CMS result

(1808.06575) and extended the m

_H±

range

ATLAS-CONF-2021-037

(9)

t ➝ qX analysis: motivation

➔Theoretical models:

◆Flavon models predict light spin-0 particles, mainly

coupling to third-generation fermions Phys.Rev.D94,056003

◆Lepton-flavour anomalies introduced by flavour violating ALP interactions Phys.Rev. Lett.124.211803

➔Signal search:

◆Top decays into up/charm quark and a new scalar/pseudo-scalar/ALP

◆Light scalar particles, m_X < m_t predicted to mostly decay to bb̄

◆Wide range of masses to explore:

●At high masses (m_X > 150 GeV):

Low signal acceptance due to missing q in t→qX

●At low masses (m_X < 50 GeV):

Low signal acceptance due to merging of the b-tagged jets in X→bb̄

➔

Search for H

⁺

), previous publication based on 2015+2016 (36 fb

^-1

) 10.1007/JHEP11(2018)085

◆

Phys.Rev.D94,056003

BR t➝q+ﬂavon

BR ﬂavon

(10)

➔

H

^±

→cb and t→cX(bb) have the same final state!

➔

Preliminary, the same selection can be defined:

◆Exactly one lepton: e^± or µ^±

◆MET ≥ 20 GeV and MET + m_T^W > 60 GeV

◆≥4 jets

◆≥2 b-tagged jets using DL1r at 60% + 1 b-tagged at 70%

➔

Event categorisation in 9 possible regions

◆Main signal regions: 4j3b, 5j3b and 6j3b

◆tt̄+≥1b background control regions: 4j4b, 5j≥4b and 6j≥4b

◆Data-based corrections for tt̄+jets: 4j2b+1bl, 5j2b+1bl and 6j,2b+1bl

➔

Differences:

◆t→uX expected to yield less events in ≥4b regions than t→cX

◆Other selections and regions could be defined for very light X with merged bb̄

t ➝ qX analysis: Event selection and categorisation

ATLAS-CONF-2021-037

t➝qX

(11)

➔

Background modelling and tt̄+jets corrections from H

^±

→cb:

◆Expected to be valid in a large fraction of the m

X range unless the 2b+1bl region has more signal or new regions are defined

➔Multivariate techniques:

◆Revisit the H^±→cb parameterised NN to train t→cX and t→uX channels

●Profiting from flavour tagging and basic kinematic information

●Use the m_bb resonance instead of m_jb and other kinematic variables

◆Study other tools like the low p_T b-tagger used in ATLAS h→aa→4b 2005.12236

➔

Profile likelihood fit and

ℬ

(t→qX)⨯

ℬ

(X→bb̄) limits

◆Fit strategy expected to be valid for most m

X values

◆Results to be compared with the 36 fb^-1 ATLAS t→qH search (JHEP05(2019)123)

t ➝ qX analysis: Strategy

Mass and ΔR of the pair of leading b-jets

(12)

Summary

➔Presented the full Run 2 ATLAS search for H^±

→cb in top-quark decays

◆

H

^±

→cb search never performed in ATLAS

◆

Search profits from tight b-jet selection and a parameterised NN to separate signal and background

◆

Model independent exclusion limits at 95% CL on

ℬ

(t→H

^±

b)⨯

ℬ

(H

^±

→cb)

●Largest excess of ~3σ at m_H±

= 130 GeV

◆

Improved sensitivity (x5) compared to CMS √s = 8 TeV result (1808.06575) and extended the m

_H±

range

➔

Introduced the ongoing t→qX(bb) ATLAS search:

◆Never performed in ATLAS before

◆

H

^±

→cb has the same final state and a similar strategy can be used as baseline

(13)

13

Backup

(14)

ATLAS t ➝ qH analysis limits

14

JHEP05(2019)123