Search for charged heavy particles in the ATLAS detector

Search for charged heavy particles in the ATLAS detector

C. Gemme1, G.Gagliardi1, E. Guido1, A. Haas2, S. Mehlhase3, S. Passaggio1, L. Rossi1, C. Schiavi1

1INFN Genova, 2SLAC, 3Niels Bohr Institute Copenhagen

LLP workshop, Rome Tor Vergata, 3-4 December 2013

1

Outline:The Pixel dE/dx2011 dataset analysisOn-going 2012 analysisRun 2 perspectives

Motivations for the searches with dE/dx The ATLAS Pixel detector is able to measure the particle

ionization in the very first cm of the trajectory. • This is not the case for CMS.

and to tag identify tracks that do not behave as MIPs• High release due to low particles• Low release due to fractional charge particles

2

Track dE/dx is defined as the average of the 3 or more pixel clusters dE/dx, having rejected the one with the highest ionization.• Resolution ~ 12%

Motivations for the searches with dE/dx Bethe-Bloch empirical fit of

low mass particles allows to measure the particle mass from its energy release and momentum measurement

5-parameters function to describe the most probable value dependence on β:

• This calibration done in 2010 is very stable over the time.

Used to search for charged massive particles.

3

Proton mass

Theoretical motivations Stable Massive Particles (SMPs) predicted in a range of

SUSY and other BSM scenarios Within SUSY: SMPs with various electric charges, squarks

and gluinos could form bound states with a light quark system: R-hadrons

Some models foresee the electric charge can change due to nuclear scattering with the detector

4

Experimental techniques in ATLAS Generic signature: slow (β<1 and highly ionizing) and high-pT

particles: measurement done by different ATLAS subdetectors

04/11/2013 C. Gemme – INFN Genova

• Muon Spectrometer (β,µ-like) (Phys.Lett. B703 428-446 (2011), 37pb-1)

• Pixel + Tracker + Tile

Calorimeter (β) (Phys.Lett. B701 1-19 (2011), 34 pb-1 )

• Pixel Detector(dE/dx,p)+SCT(p) (ATLAS-CONF-2012-022 (2012), 2fb-1)

Pixel+Tile+Muon combined analysis(Phys.Lett. B720 277-308 (2013), 4.7 fb-1 )

-MDT

-Calo

-Pixel

Experimental techniques Tracker + Tile + MS

• Measurement of time-of-flight (Tile calorimeter), in the muon spectrometer and specific ionization energy loss (Pixel)

Muon Spectrometer• Complementary (looking

for R-hadrons even without an ID track)

• More sensitive to larger gluino-ball (neutral) fractions

• Signature similar to muons, except for timing

04/11/2013 C. Gemme – INFN Genova 6

Pixel-only analysis Pixel-only approach:

• R-hadrons have large pT and large ionization in Pixel, well above the MIP release

• Signal efficiency not affected by calorimeter or MS requirements

• Higher geometrical acceptance• Sensitive also to metastable

particles

04/11/2013 C. Gemme – INFN Genova 7

Main issues:• Background rejection high pT and isolation requirements;• Background estimation data-driven, properly reproducing

dE/dx dependence on p and η.• Pixel ionization is not available at the trigger level: other

signatures must be used

• The strong nature of gluino production mechanisms and the associated QCD radiation is exploited: gluon-gluon fusion, where ISR gives rise to jets. Jets + modest energy deposition of the heavy objects missing transverse energy (ET

miss)

Data and signal samples - 2011 Starting from the 2011 JetTauEtmiss stream, DESDM RPVLL

skimmed samples are generated based on:• an OR of several trigger requests

• in particular EF_xe70_noMu is the first unprescaled on period B-L, EF_xe60_verytight_noMu on period M

• an OR of different selections, among which TrackParticleFilter• Requires at least one track with nPix>=2, nSCT>=6 and pT>50

GeV From DESDM RPVLL are produced D3PD_RPVLL (release 17) Good Run Lists to require that tracking and ET

miss quality flags are fine.

Starting from the 2011 JetTauEtmiss stream, DESDM RPVLL skimmed samples are generated based on:• an OR of several trigger requests

• in particular EF_xe70_noMu is the first unprescaled on period B-L, EF_xe60_verytight_noMu on period M

• an OR of different selections, among which TrackParticleFilter• Requires at least one track with nPix>=2, nSCT>=6 and pT>50

GeV From DESDM RPVLL are produced D3PD_RPVLL (release 17) Good Run Lists to require that tracking and ET

miss quality flags are fine.

04/11/2013 C. Gemme – INFN Genova 8

Gluino R-hadron samples (different models, masses between 100 and 1500 GeV, varying probability of the R-hadron forming a gluon-ball)

sbottom & stop R-hadron samples (different models, masses between 100 and 1000 GeV)

gluino metastable R-hadron private samples (mass 400 GeV, ~ 𝜏2-10 ns)

Gluino R-hadron samples (different models, masses between 100 and 1500 GeV, varying probability of the R-hadron forming a gluon-ball)

sbottom & stop R-hadron samples (different models, masses between 100 and 1000 GeV)

gluino metastable R-hadron private samples (mass 400 GeV, ~ 𝜏2-10 ns)

MC

Data

Trigger efficiency

04/11/2013 C. Gemme – INFN Genova 9

Event Selection Trigger:• Trigger efficiency is ~ 15% on a wide range of masses.

Event level: • offline confirmation of the MET trigger: Et

miss > 85 GeV• at least 1 primary vertex (PV) with at least 5 tracks

Track level (at least one track with)• High-pT: pT>50 GeV, primary, n.Pixel hits >=3, nBLayer hits

>=1 and n.SCT hits >=6 • Isolated: no primary with pT>1 GeV in a cone of radius

ΔR<0.25• LAr dead region veto, Electron veto, P>100 GeV & Δp/p<50%• High ionization: dE/dx > 1.800 - 0.055|η|+ 0.129|η|2 -0.037|η|3

MeV/g cm-2

• Jet cleaning04/11/2013 C. Gemme – INFN Genova 10

Signal efficiency and bkg rejection

04/11/2013 C. Gemme – INFN Genova 11

Data:JetTauEtmiss DESDM_RPVLL

MC: data flow table for gluino R-Hadron 400 GeV MC:

Efficiency for All R-Hadron samples

Background estimation Data-driven approach: we use data to parameterize the key

variables (p, η and dE/dx) and their inter-dependences, and then we generate a high-statistics random sample based on them, reproducing the data behaviour.• 2 data samples used (both selected in a way to reduce signal

contamination) requiring either low ionization or low momentum

04/11/2013 C. Gemme – INFN Genova 12

From p&dE/dx, the mass is then calculated (by Bethe-Bloch fit) • Normalization of the

background mass spectrum to data BEFORE the ionization cut (smaller uncertainty, 0.7%), and considering M<140 GeV (to reduce possible signal contaminations)

• Ionization cut is then applied

Background estimation (I)

13

✓ Data-driven approach: we use data to parameterize the key variables (p, η and dE/dx) and their inter-dependences, and then we generate a high-statistics random sample based on them, and reproducing data behaviour

✓ 2 data samples used (both selected in a way to reduce signal contamination)

Bkg 1 Bkg 2

✓ Used for p and η(p) distributions

✓ All the selection cuts, except the ionization cut, which is inverted (dE/dx<1.8MeVg-1cm2)

✓ Exploiting the fact that no signal is foreseen at low dE/dx

✓ It can’t be used for dE/dx (cutting on it to avoid signal-> questionable extrapolation to high tails)

✓ Used for dE/dx(η) distributions

✓ Tracks selected as in the nominal selection, but without cut on dE/dx and with pT>10 GeV and 40<p<100 GeV (covering the same η range as candidates)

✓ Exploiting the independence of dE/dx on p above the Fermi’s plateau & the fact that no signal is foreseen to have low p

E.Guido SUSY approval 08/06/2012

14E.Guido SUSY approval 08/06/2012

✓ Fraction of µ in data is 85.0%

✓ The two categories are reweighed according to this fraction

• muons

• non-leptons

Bkg 1

100<|p|<150 GeV 150<|p|<200 GeV

200<|p|<300 GeV |p|>300 GeV

15E.Guido SUSY approval 08/06/2012

✓ µ and non-lepton shapes are considered as compatible (a systematic error is associated)

|η|<0.5 0.5<|η|<1.0

1.0<|η|<1.25 1.25<|η|<1.5

1.5<|η|<1.75 1.75<|η|<2.0

2.0<|η|<2.25 2.25<|η|<2.5

Bkg 2

Background estimation (II)

16

✓ From p&dE/dx, the mass is calculated (by Bethe-Bloch)

✓ Normalization of the background mass spectrum to data BEFORE the ionization cut (smaller uncertainty, 0.7%), and considering M<140 GeV (to reduce possible signal contaminations)

✓ Ionization cut is then applied

Before dE/dx cut After dE/dx cut

E.Guido SUSY approval 08/06/2012

The peak at low mass is due to tracks with Pixel dE/dx < 1.26 MeVg−1cm2 for which the π mass is assumed

Systematics

Affecting Efficiency, background, other..

Some of them varying with mass (200-1500 GeV)

04/11/2013 C. Gemme – INFN Genova 17

(theoretical)

(experimental)

1 ÷ 4

-6 ÷ -0.3

Results Input: signal, background and data mass spectra. Frequentist

scan of the signal CL, as a function of R-hadron cross-section A gluino R-hadron with Mass<940 GeV, a stop R-hadron with

Mass<604 GeV and a sbottom R-hadron with Mass<576 GeV are excluded at the 95% CL. • Full detector analysis gives slightly higher limits, 985 GeV, 683

GeV, 612 GeV for gluino, stop and bottom R-hadrons respectively) at 95% CL.

C. Gemme – INFN Genova 18Gluino R-hadrons Squarks R-hadrons