LHC-iTools fileLHC-iTools Methods and tools for the interpretation of the LHC results Sabine Kraml - conseil scientifique - 23 Jan 2018

LHC-iTools

Methods and tools for the interpretation of the LHC results

Sabine Kraml - conseil scientifique - 23 Jan 2018

Motivation• The search for new phenomena beyond the SM (BSM)

is one of the top priorities of the LHC program.

• To this end, the LHC collaborations are pursuing - precision measurements of “known” processes

(jets, EW boson, top quark, Higgs, etc. prod.) - searches for new physics in a vast variety of channels.

• Results are typically interpreted by the experiments in terms of the SM, popular minimal BSM scenarios, simplified models, EFT fits, … often on an analysis-by-analysis basis.

• For a full understanding of the implications for physics at the TeV scale physics we need, however, to be able to confront all kinds of theoretical models against the LHC results.

• Close theory-experiment interaction

• Sophisticated public tools for a comprehensive, global view of what the data tell us about TeV scale physics.

\

Why build tools for (re)interpretation?

vanilla new physics

non-minimal models

1. Avoid the streetlight effect

not sexy

not mainstream

new theories nobody has though of yet

‘weird’ signatures

soft stuff

2. Ensure long-term impact of results, use in global analyses, etc.

We want to know what all(!) the LHC and other data tell us about the TeV scale and beyond

Recasting based on MC event simulation

Interpretation of Higgs measurements

Reinterpretation of Simplified Model results

Activities at the LPSC

Higgs constraints on new physics: Lilith and beyond

• From end of 2011 onwards, we pursued detailed studies of the implications of the 125 GeV Higgs boson for new physics

‣ 5 topCite100+, 4 topCite50+, 1 PRL‣ “editor’s suggestion” and “pick of the month” in PRD

• The computer code developed for this purpose was turned into a public program (Lilith) by two of our students:

‣ J. Bernon, B. Dumont, Eur.Phys.J. C75 (2015) no.9 ‣ Springer Thesis Award for B. Dumont

• Lilith is a Python library which assess the compatibility of a non-standard Higgs sector with all available signal strength measurements of the observed state at 125 GeV.

• Easily extensible and very fast, which is important for large scans. The results of Lilith can be used to constrain a wide class of new physics scenarios.

2HDM Type I Run1+Run2

2HDM Type II Run1+Run2

Higgs constraints on new physics: Lilith and beyond

• If the kinematic distribution of the 125 GeV Higgs signal depends on model parameters, simple scaling of production cross sections and decay branching ratios (relative to the SM) is invalid

➡ must account for the change in the signal selection efficiency.

• This can arise from new tensor structures or the presence of new Higgs production modes, e.g., from decays of heavier new states.

➡ particle-level differential measurements

• ATLAS and CMS are providing total and differential fiducial cross section measurements for several Higgs decay modes, as well as `simplified template cross sections’ for specific production modes.

Future: We want to develop the relevant machinery for making use of these data.

Using simplified model results: SModelS

• It has become standard that ATLAS and CMS present the results of their BSM searches in terms of “simplified model” constraints.

• Simplified models (SMS) reduce full models with a plethora of particles and parameters to subsets with just 2-3 new states and a simple decay pattern.

• Concept used by SUSY, Exotics, DM searches

• Very convenient for optimising analyses that look for a particular final state, as well as for comparing the reach of different strategies.

• Understanding how SMS results constrain a realistic model with a multitude of parameters, relevant production channels and decay modes is, however, a non-trivial task.

• Automated tool:

[GeV]b~m400 500 600 700 800 900 1000 1100 1200 1300

[GeV

]10 χ∼

m

0

100

200

300

400

500

600

700

800

900

CMS Preliminary

10χ∼ b→ b~, b~b~ →pp Moriond 2017

(13 TeV)-135.9 fbExpectedObservedSUS-16-032, 0-lep sbottom

)missTSUS-16-033, 0-lep (H

)T2SUS-16-036, 0-lep (M

[GeV]q~m400 600 800 1000 1200 1400 1600 1800

[GeV

]10 χ∼

m

0

200

400

600

800

1000

1200

CMS Preliminary

10χ∼ q→ q~, q~q~ →pp Moriond 2017

(13 TeV)-135.9 fb

q~one light

)c~,s~,d~,u~ (Rq~+L

q~

ExpectedObserved)miss

TSUS-16-033, 0-lep (H)T2SUS-16-036, 0-lep (M

[GeV]t~m200 400 600 800 1000 1200

[GeV

]10 χ∼

m

0

100

200

300

400

500

600

700

800

900

CMS Preliminary

10χ∼ t→ t~, t~t~ →pp Moriond 2017

(13 TeV)-135.9 fbExpectedObserved

)missTSUS-16-033, 0-lep (H

)T2SUS-16-036, 0-lep (MSUS-16-049, 0-lep stopSUS-16-051, 1-lep stopSUS-17-001, 2-lep stopComb. 0-, 1- and 2-lep stop

01χ∼

+ m

t

= m

t~m

[GeV]g~m800 1000 1200 1400 1600 1800 2000 2200

[GeV

]10 χ∼

m

0

200

400

600

800

1000

1200

1400

1600

1800

2000CMS Preliminary

10χ∼t t→ g~, g~g~ →pp Moriond 2017

(13 TeV)-135.9 fbExpectedObserved

)missTSUS-16-033, 0-lep (H

)T2SUS-16-036, 0-lep (M)JSUS-16-037, 1-lep (M)φ∆SUS-16-042, 1-lep (

2-lep (SS)≥SUS-16-035, 3-lep≥SUS-16-041,

Using simplified model results: SModelS

Working principle of SModelS

collaboration with Santo Andre (A. Lessa) and Vienna (W. Waltenberger)

• Since the first public release in 2014 (v1.0), the code base has undergone significant structural changes.

• Version 1.1 published in 2017 comes with many new features; most important: use of efficiency maps.

• Extensive database: 186 results from 21 ATLAS and 23 CMS SUSY searches, covering 37 topologies.

• Update to 35/fb results from CMS in progress

9

(ATLAS did not yet provide 13 TeV SMS results which can be used)Gluino mass [GeV]

500 1000 1500 2000 2500 3000 3500 4000

Frac

tion

of e

xclu

ded

poin

ts

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1Excluded by EM

Excluded by UL

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Fraction of Bino LSP ATLAS excluded points excluded by SModelS

[GeV] g~m0 200 400 600 800 1000 1200 1400 1600 1800 2000

[GeV

]0 1χ ~

m

0100

200

300

400

500

600

700

800900

1000

12 90 86 81 90 89 103 107 111 97 111 95 114 111 101 82 124 97

14 67 75 55 74 82 83 95 77 104 104 71 78 67 110 107 111 102

13 38 48 53 43 57 48 60 53 61 55 53 46 53 65 36 37 47

20 71 81 70 85 98 94 105 114 105 114 104 87 64 86 98 82 82

88 117 100 111 102 105 112 106 107 128 132 103 134 125 108 107 82 85

393 607 137 137 112 124 148 115 109 107 127 105 96 100 85 85 85 63

1152 131 123 140 149 157 140 133 106 109 109 88 90 90 63 57 84

518 773 147 140 133 181 162 152 154 128 94 92 109 84 64 49 63

1133 139 149 143 139 153 167 135 148 144 129 101 85 78 74 49

368 706 139 131 153 131 174 143 136 141 139 138 119 88 62 47

944 131 129 114 123 149 118 138 116 122 103 103 73 48 29

312 618 111 80 106 97 102 83 97 90 63 57 38 29 16

691 119 62 89 71 71 63 62 59 41 29 15 12 8

180 319 58 55 65 48 64 41 33 21 11 12 4 1

289 60 23 26 25 22 18 13 5 1

62 58 14 9 9 8 8 3 1

35 8 6 1 3 1

3 3 3 1 1

1 1

1

Fraction of Bino LSP ATLAS excluded points excluded by SModelS

• Variety of phenomenological studies, e.g. constraints on sneutrino LSP

• Extensive study of the coverage of the pMSSM by simplified model results

• Identified important missing topologies

• Many talks, e.g., CHEP, EPS-HEP

Thesis of U. Laa, 2017

Using simplified model results: SModelSPostdoc: S. Kulkarni 2012-2014,

now at HEPHY Vienna.

Future plans:

• Produce new efficiency maps for simplified models not considered by ATLAS and CMS to improve coverage of complex models

• Test constraints on new models like SUSY with Dirac gauginos

• Include lifetime information to be able to treat constraints from searches for long-lived particles (needs restructuring of database)

• Extend the model input from SLHA(-like) files to the Lagrangian level, in order to be able to link to, e.g., MadGraph implementations of new models.

• Finally, to go beyond the assumption of a Z2 symmetry we will completely revise the SModelS internal language used for the decomposition and the matching with the results in the database. The data description in the results database itself also has to be adapted. This will be SModelSv2.0.

10

Using simplified model results: SModelS

- New PhD student, H. Reyes Gonzalez, started in Oct. 2017- PRC project for collaboration with A. Lessa in Brasil (awaiting renewal for 2018)- Bilateral ANR-FWF project with HEPHY Vienna submitted in Jan 2018

NB: both Lilith and SModelS were interfaced to micrOMEGAS (Comput. Phys. Comm. 222, 2018.)

Recasting based on Monte Carlo simulation• Most direct way of (re)interpreting an experimental search:

reproduce it in a Monte Carlo simulation i.e. simulate the events that would be measured in an analysis if a particular model were true, and to compare this to the actually measured number of events and expected background.

• More general and more precise than using simplified model results, but very CPU-intensive

• Main difficulty: reliable emulation of detector effects.

• In 2014, together with B. Fuks and a number of students and postdocs at the LPSC, we started a “Public Analysis Database” (PAD) in MadAnalysis5, comprising several ATLAS and CMS new physics searches.

• The PAD has been growing since and is used by many people. Implementing and validating new analyses is however a very tedious and time consuming business.

Most of the time additional information and validation material is needed from the collaboration, which is not always available

50 100 150 200 250 30010

2

103

104

pT (b1) (GeV)

Nev

ents

/bin

t̃ → bχ̃±

1 (650/50/0.5) × 1000

t̃ → tχ̃01 (250/100) × 10

Very active field — here just some examples

Dilepton constraints on the Inert Doublet Model Belanger et al, 1503.07367

- Most important channel: pp -> AH, A -> Z(*)H- here, H ist the inert scalar, i.e. DM candidate. - Recasted 2 ATLAS analyses from Run 1:

dilepton SUSY search & the ZH, H>inv analysis- LHC just starts to probe Higgs funnel region

at mH~60 GeV, which is most interesting for DM.

- SM plus a real scalar DM field η with derivative pNGB interactions suppressed by powers of the scale f, plus a second singlet scalar mediator field s.

- Recasted ATLAS mono-jet search at 13 TeV (3.2 /fb)

Monojet searches for momentum-dependent dark matter interactions Barducci et al., 1609.07490

200 300 400 500 600 700 800

0.5

1

5

10

50

100

mη [GeV]

c sηorc ∂sη

Solid=MIDashed=MD

csg=10ms= 250 GeV

3.2fb-1

3.2 fb-1

300fb-1

300 fb-1

c∂sη Unitarity

• Used ATLAS and CMS SUSY searches in ttbar+MET final state at Run 1 to constrain scenarios with a fermionic top partner and a dark matter candidate.

• Efficiencies in all-hadronic, 1-lepton and 2-lepton channels are very similar for scalar and fermionic top partners.

• SMS results for stop–neutralino simplified models can also be applied to fermionic top-partner models, provided the narrow width approximation holds in the latter.

• Official eff. maps don’t extend to high enough masses, so we provide our own:

CM atlas_conf_2013_024

Benchmark (600,10)R

Contours and signal region markers

ATLAS Limit

SUSYSUSY with XQ

XQ to SDMXQ to VDM

List of Signal Regions

SR1SR2SR3

200 400 600 800 10000

200

400

600

800

Mt˜/ T [GeV]

MDM

[GeV]

MA CMS_13_011

Benchmark (600,10)R

Contours and signal region markers

CMS Limit

SUSYSUSY with XQ

XQ to SDMXQ to VDM

List of Signal Regions

Stopbchargino_HighDeltaM_MET150Stopbchargino_HighDeltaM_MET200Stopbchargino_HighDeltaM_MET250Stopbchargino_LowDeltaM_MET100Stopbchargino_LowDeltaM_MET150Stopbchargino_LowDeltaM_MET200Stopbchargino_LowDeltaM_MET250StopTneutralino_HighDeltaM_MET300StopTneutralino_LowDeltaM_MET300

200 400 600 800 10000

200

400

600

800

Mt˜/ T [GeV]

MDM

[GeV]

MA ATLAS_1405_7875

Benchmark (600,10)R

Contours and signal region markers

SUSYSUSY with XQ

XQ to SDMXQ to VDM

List of Signal Regions

4jl4jlm4jm5j6jl6jm6jtp

200 400 600 800 10000

200

400

600

800

Mt˜/ T [GeV]

MDM

[GeV]

0-lepton stop search 1-lepton stop search 2-6 jets gluino/squark search

g

t̃χ0

t

t̃∗χ0

t̄

g

t̃ χ0

t

gt̃∗ χ0

t̄

t̃

g

g

t̃χ0

t

t̃∗χ0

t̄

g

TS0DM, V 0

DM

t

T̄S0DM, V 0

DM

t̄

gT S0

DM, V 0DM

t

gT̄ S0

DM, V 0DM

t̄

T

Scalar versus fermionic top-partner interpretation of ttbar + MET searches SK, Laa, Panizzi, Prager, 1607.02050

http://lpsc.in2p3.fr/projects-th/recasting/susy-vs-vlq/ttbarMET/

Generic gluino/squark searchcan also provide a limit on fermionic top partners, due to higher Meff than for stops.

official plots stop here

Person power

0

17,5

35

52,5

70

2012 2013 2014 2015 2016 2017 2018

Permanent Postdocs PhDPe

rson

pow

er in

mon

ths

We clearly profited from ample of person power in 2014–2016, which allowed us to make major advances on several lines and maximally exploit the LHC Run 1 results.

SWOT

\Strengths Competence, experience, motivation

Leading role in CERN Forum for BSM interpretation

Weakness Person power, mainly number of postdocs

Opportunity Wealth of Run 2 results to come out

Threat No more postdoc from Oct 2018 onwards

Decompose signatures of full model into SMS elements

Compare with experimental constraints in SModelS database

http://smodels.hephy.at

SModelS v1.1.1 now available, user manual: arXiv:1701.06586

Sign

al s

elec

tion

effic

ienc

y

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

[GeV]gluinom400 600 800 1000 1200 1400 1600

[GeV

]LS

Pm

0

200

400

600

800

1000

1200

1400

1600CMS (8 TeV)-119.3 fb simulation0χ∼bb→g~, g~g~→pp

Razor MultiJet box

arXiv:1312.4175

Efficiency maps correspond to a grid of simulated acceptance x efficiency values for a specific signal region for a specific simplified model.

Together with the observed and expected #events in each SR, this allows to compute a likelihood.

Upper Limit maps give the 95% CL upper limit on cross section x branching ratio for a specific SMS.

The UL values can be based on the best SR (for each point in parameter space), a combination of SRs or more involved limits from other methods.

[GeV]g~m400 600 800 1000 1200 1400

[GeV

]0 χ∼

m

0

200

400

600

800

1000

1200

3−10

2−10

1−10

1CMS (8 TeV)-119.3 fb

10χ∼t t→ g~, g~g~ →pp NLO+NLL exclusion

Razor 0L+1L+2L

theoryσ 1 ±Observed experimentσ 1 ±Expected

95%

CL

uppe

r lim

it on

cro

ss s

ectio

n (p

b)

Upper Limit (UL) maps

Sign

al s

elec

tion

effic

ienc

y

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

[GeV]gluinom400 600 800 1000 1200 1400 1600

[GeV

]LS

Pm

0

200

400

600

800

1000

1200

1400

1600CMS (8 TeV)-119.3 fb simulation0χ∼bb→g~, g~g~→pp

Razor MultiJet box

Efficiency maps (EM)

NB: the 95%CL exclusion curve is not used, cannot be re-interpreted

Great if these are available in

numerical form!

Limit on σ⨉BR Limit on Σε⨉σ⨉BR

Experimental constraints