Where’s SUSY?

The electroweak SUSY landscape

after ATLAS Run 2 searches

1

Ben Hodkinson

Outline

• SUSY 101
• SUSY signal models
• The pMSSM
• Reinterpreting ATLAS searches
• External constraints
• Overall ATLAS exclusion of EW SUSY scenarios
• Dark matter complementarity
• Comparison with Run 1
• Surviving models
• Summary

2

Based on the recent ATLAS electroweak pMSSM scan CONF note (ATLAS-CONF-2023-055) and my PhD thesis (https://doi.org/10.17863/CAM.104677)

→ If a plot has an ATLAS label it’s from the CONF note, otherwise from my thesis!

SUSY 101

3

Hierarchy problem

​

• Planck scale O(10¹⁹) GeV
• How is the measured Higgs mass only 125 GeV?
• Perfect cancellations of loops through 38 orders of magnitude?
• A new symmetry between fermions and bosons would naturally produce such a cancellation

4

∝+m(s)²

∝ -m(f)²

Corrections to squared Higgs mass

Fermion loop

Scalar loop

Hierarchy problem

​

• Planck scale O(10¹⁹) GeV
• How is the measured Higgs mass only 125 GeV?
• Perfect cancellations of loops through 38 orders of magnitude?
• A new symmetry between fermions and bosons would naturally produce such a cancellation

5

∝+m(s)²

∝ -m(f)²

Corrections to squared Higgs mass

Fermion loop

Scalar loop

The SUSY particle zoo

6

The SUSY particle zoo

7

The SUSY particle zoo

8

Pre-EWSB:

U(1) gauge field:

B

​

SU(2) gauge fields:

W⁰, W⁺, W^-

The SUSY particle zoo

9

Partners of the SM pre-EWSB

B and W fields are the “Wino” and “Bino”

Pre-EWSB:

U(1) gauge field:

B

​

SU(2) gauge fields:

W⁰, W⁺, W^-

The SUSY particle zoo

10

Electroweak sector

Pre-EWSB:

U(1) gauge field:

B

​

SU(2) gauge fields:

W⁰, W⁺, W^-

The SUSY particle zoo

11

The MSSM

12

• None of these have been found……

​

• SUSY, if it exists, must be a broken symmetry

​

• Minimal Supersymmetric Standard Model (MSSM)
• Includes all these sparticles
• “SUSY breaking terms” parameterize our ignorance about the SUSY breaking mechanism
• > 100 unknown parameters

Electroweak SUSY

The rest of this seminar will focus on electroweak production, ie. neutralinos and charginos

​

Interesting because:

• Smallest production cross-sections → Limits are weakest.
• Lightest neutralino is expected to be the lightest SUSY particle (LSP) and can be a dark matter candidate.
• Dark matter constraints favour scenarios with an LSP mass at or below O(1 TeV).

Electroweak SUSY

14

If bino, wino and higgsino mass parameters are well separated you get…

Higgsinos: Two neutralinos and a chargino(s) close in mass

Winos: One neutralino and chargino(s) close in mass

Bino: One neutralino

“Lightest SUSY particle” = LSP

Electroweak SUSY

15

If bino, wino and higgsino mass parameters are well separated you get…

Higgsinos: Two neutralinos and one chargino close in mass

Winos: One neutralino and one chargino close in mass

Bino: One neutralino

“Lightest SUSY particle” = LSP

SUSY signal models

16

Minimally Supersymmetric Standard Model

17

SUSY model has

many parameters!”

MSSM

• Minimum number of new particles and interactions to realize supersymmetry.

​

• 100+ free parameters

​

• Very difficult / impossible to interpret analysis results in the MSSM.

Simplified models

18

• Focus on one SUSY production process and decay chain.
• All other particles decoupled
• BR(Decay of interest) = 100%
• Pure bino/wino/higgsino states
• Signature driven approach

ATL-CONF-2023-046

Simplified models

19

• Focus on one SUSY production process and decay chain.
• All other particles decoupled
• BR(Decay of interest) = 100%
• Pure bino/wino/higgsino states
• Signature driven approach

ATL-CONF-2023-046

Simplified models

20

• Focus on one SUSY production process and decay chain.
• All other particles decoupled
• BR(Decay of interest) = 100%
• Pure bino/wino/higgsino states
• Signature driven approach

ATL-CONF-2023-046

Simplified models

• 2D grid of signal models for optimising searches
• 2D exclusion plots

​

• Doesn’t capture more complex phenomenology
• How do we connect searches for different models?
• How do we build a coherent picture of the global ATLAS constraints on SUSY?

21

SUSY model

simple!”

Phenomenological MSSM

22

Project 19-dim pMSSM into 2D planes and evaluate our sensitivity:

Run-2 EWK scan result

• Middle ground between MSSM and simplified models.

​

• MSSM + a few simplifying assumptions

​

• Includes all sparticle production and decay modes

​

• 19 parameters

Phenomenological MSSM

23

• Middle ground between MSSM and simplified models.

​

• MSSM + a few simplifying assumptions

​

• Includes all sparticle production and decay modes

​

• 19 parameters

SUSY MODEL

could scan and analyse the parameter space on the timescale of her PhD.

pMSSM

pMSSM assumptions

​

1. No new sources of CP-violation (beyond CKM matrix)
2. No flavour-changing neutral currents (FCNCs)
3. Universality of 1st and 2nd generation sfermions
4. R-parity conserved
5. Lightest SUSY particle (LSP) is the lightest neutralino

24

Based on experimental constraints and general features of SUSY breaking mechanisms.

pMSSM assumptions

​

1. No new sources of CP-violation (beyond CKM matrix)
2. No flavour-changing neutral currents (FCNCs)
3. Universality of 1st and 2nd generation sfermions
4. R-parity conserved
5. Lightest SUSY particle (LSP) is the lightest neutralino

25

Based on experimental constraints and general features of SUSY breaking mechanisms.

R-parity conservation

• Stops proton decaying!
• SM particles have P_R= +1
• SUSY partners have P_R= -1
• R-parity is conserved multiplicatively

​

​

​

26

Note: R-parity violating models also exist but have very different collider signatures (not considered here).

R-parity conservation

• Stops proton decaying!
• SM particles have P_R= +1
• SUSY partners have P_R= -1
• R-parity is conserved multiplicatively

→ Sparticles must be produced in pairs at colliders

27

Note: R-parity violating models also exist but have very different collider signatures (not considered here).

R-parity conservation

• Stops proton decaying!
• SM particles have P_R= +1
• SUSY partners have P_R= -1
• R-parity is conserved multiplicatively

→ Sparticles must be produced in pairs at colliders

→ Lightest SUSY particle (LSP) cannot decay

28

Note: R-parity violating models also exist but have very different collider signatures (not considered here).

R-parity conservation

• Stops proton decaying!
• SM particles have P_R= +1
• SUSY partners have P_R= -1
• R-parity is conserved multiplicatively

→ Sparticles must be produced in pairs at colliders

→ Lightest SUSY particle (LSP) cannot decay

→ LSP is an ideal cold dark matter candidate!

29

Note: R-parity violating models also exist but have very different collider signatures (not considered here).

Phenomenological MSSM

= MSSM + assumptions

→ 19 parameters

30

Full MSSM

= 100+ parameters → Too many

MSSM

Simplified models are too simple

Simplified

model

pMSSM

Scanning the pMSSM

31

Workflow

• Scan the 19-dimensional pMSSM to produce sets of models
• Reinterpret Run 2 searches to determine which models are (not) excluded
• Produce global picture of ATLAS’ sensitivity to electroweak SUSY
• Identify scenarios we’ve missed due to non-simplified phenomenology

32

Random scan

Generate models & MC events

Evaluate searches with SimpleAnalysis and RECAST

EWK pMSSM scan

Random scan with uniform priors over parameters relevant to production of charginos and neutralinos:

​

​

​

​

​

​

​

33

Interlude:

Dark Matter in the pMSSM

34

Dark Matter: Thermal freeze-out

35

Planck mission: Observes a “density” of DM in the universe of: Ωh² = 0.12

→ Compared to 0.0224 for baryonic matter

​

The SUSY story of how we got there (also applies to non-SUSY cold dark matter models):

​

1. Early universe: Very hot, lots of sparticles + SM particles being created/annihilated.

​

• Universe cools: Heavier sparticles stop being produced and decay to LSP (which can pair-annihilate into SM particles)

​

• Cosmological expansion: LSP’s spread out so annihilation rate decreases.
• “Freeze out” = density of LSP’s remains constant.
• The “density” of DM observed today is determined by the LSP annihilation cross-section.

Dark matter in the pMSSM

36

• Recall the SUSY DM candidate is the neutralino (𝜒̃⁰₁ )

​

• Neutralino is a mix of the wino, higgsino and bino fields.

→“Bino-like”, “Higgsino-like” or “Wino-like” depending on dominant component…

​

• DM phenomenology is different for each case…

Dark matter relic density of models

We allow LSP to be a sub-dominant DM component

• Require Ωh² ≤ 0.12

​

Higgsino/Wino-like LSP:

• Mass near to chargino / 2nd neutralino
• Enhanced co-annihilation with chargino / 2nd neutralino
• Underestimates relic density unless m(LSP)~TeV

37

Dark matter relic density of models

We allow LSP to be a sub-dominant DM component

• Require Ωh² ≤ 0.12

​

Higgsino/Wino-like LSP:

• Mass near to chargino / 2nd neutralino
• Enhanced co-annihilation with chargino / 2nd neutralino
• Underestimates relic density unless m(LSP)~TeV

​

Bino LSP:

• In general overestimates relic density
• Flat scanning strategy doesn’t sample many models with satisfactory relic density

38

Bino-LSP models: DM relic density

Regions with satisfactory DM relic density for bino-LSP models:

• Z/h/A funnel

​

​

​

• Enhanced co-annihilation with 2nd neutralino or chargino
• Wino-like C1/N2 close in mass
• Significant higgsino component

​

Targeted scan performed to oversample these regions

39

Bino-LSP models: DM relic density

Regions with satisfactory DM relic density for bino-LSP models:

• Z/h/A funnel

​

​

​

• Enhanced co-annihilation with 2nd neutralino or chargino
• Wino-like C1/N2 close in mass
• Significant higgsino component

​

Targeted scan performed to oversample these regions

40

Final pMSSM model samples

EWKino scan

• All LSP types
• Don’t impose dark matter constraint
• 12,280 models

Bino-DM scan

• Bino-like LSP only
• Require Ωh² ≤ 0.12
• 8,897 models

41

Workflow

42

Workflow

43

Initial constraints applied

Workflow

44

Only simulate models where we expect some sensitivity

Initial constraints applied

Workflow

45

Only simulate models where we expect some sensitivity

Particle-level evaluation first to check if model is likely to be excluded or not

Initial constraints applied

Workflow

46

Only simulate models where we expect some sensitivity

Detector simulation for models where particle-level evaluation is insufficient

→ This is what separates this from non-ATLAS pheno studies

Particle-level evaluation first to check if model is likely to be excluded or not

Initial constraints applied

Excluding pMSSM models with ATLAS Run 2 analyses

47

Phenomenologically viable SUSY models

The ATLAS Run 2 dataset

Searches included

48

All key signature Run 2 EW searches are included, covering each final-state lepton multiplicity:

​

• Fully hadronic
• 1 lepton + 2 bjets (1Lbb)
• 2 leptons + 0 jets (2L0J)
• 2 leptons + 2 jets (2L2J
• Compressed 2 lepton
• 3 lepton (3L)
• 4 lepton (4L)
• Disappearing track

1Lbb

49

Simplified model limits

pMSSM exclusion

1Lbb simplified model:

• Pure-Bino LSP
• Pure-Wino chargino and 2nd neutralino
• BR(𝜒̃⁰₂→ 𝜒̃⁰₁ h) = 100%
• SM-like Higgs decays

1Lbb

50

1Lbb simplified model:

• Pure-Bino LSP
• Pure-Wino chargino and 2nd neutralino
• BR(𝜒̃⁰₂→ 𝜒̃⁰₁ h) = 100%
• SM-like Higgs decays

3L

51

Preference for Z decay mode

3L simplified model:

• Pure-Bino 𝜒̃⁰₁
• Pure-Wino 𝜒̃⁰₂ and 𝜒̃^±₁
• BR(𝜒̃⁰₂→ 𝜒̃⁰₁ Z^*)= 100%
• BR(𝜒̃^±₁→ 𝜒̃⁰₁ W^*)= 100%

Fully hadronic

52

Targeted generic heavy electroweakino production

​

In the pMSSM this includes second chargino and third/fourth neutralino

2L0J

53

Preference for Wino-like charginos

→ higher production cross-section

2L0J simplified model:

• Pure-Bino LSP
• Pure-Wino chargino
• BR(𝜒̃^±₁→𝜒̃⁰₁ W^±)= 100%

2L0J

54

Higgsino 𝜒̃^±₁ preferred in Bino-DM scan

→ weaker sensitivity

2L0J simplified model:

• Pure-Bino LSP
• Pure-Wino chargino
• BR(𝜒̃^±₁→𝜒̃⁰₁ W^±)= 100%

Compressed

55

Compressed higgsino simplified model:

• Pure-higgsino (𝜒̃⁰₁,𝜒̃⁰_2,𝜒̃^±₁)
• BR(𝜒̃⁰₂→ 𝜒̃⁰₁ Z^*)= 100%
• BR(𝜒̃^±₁→ 𝜒̃⁰₁W^*)= 100%

At smaller mass-splittings, the radiative decay mode becomes dominant → reduced sensitivity

Disappearing track

Evaluated using cross-section upper limits only

56

bSM Higgs analyses

57

BR(h→invisible) gets contributions from h→𝜒̃⁰₁𝜒̃⁰₁

Constraints on the mass of the pseudoscalar A from Higgs measurements

External constraints

58

External constraints

59

EW precision constraints

Average W mass measurement

→ window widened by 6 MeV to account for uncertainty on top-quark mass

​

CDF result not used

→ would be strongly constraining if confirmed (~4% models lie in CDF W mass range)

​

Constraints on Δ⍴ and Γ(Z→invisible) also applied

60

Flavour constraints

61

Latest LHCb B_s→ μμ measurement

​

pMSSM models get loop contributions

​

Constraints on B_u→ 𝛕 𝝂 and b→s𝜸 also included

​

​

Overall exclusion

62

• For each model, the analysis with the best expected CLs is used
• If the observed CLs < 0.05 for that analysis, the model is considered “excluded”.
• No statistical combinations are performed.
• Only models which pass the flavour & precision EW external constraints are included
• Dark matter constraints considered separately

Overall ATLAS sensitivity

Sensitivity to compressed scenarios through heavier electroweakino decays

​

Even low mass bins don’t have 100% exclusion…

​

Important to improve depth of sensitivity as well as target new regions!

63

Overall exclusion

64

Bino-DM scan

​

When dark matter constraints are included, the region our searches target is not very well populated!

​

Mostly compressed scenarios remain

Grey bins = No models

Overall exclusion

65

Exclusion beyond the overlaid contour from FullHad, 2L2J and Disappearing Track

Overall exclusion

66

Disappearing track does a good job constraining wino-LSP scenarios (red)

​

Bino and Higgsino-LSP scenarios remain viable even at 100 GeV and below

Direct detection complementarity

67

ATLAS and LZ sensitivity is complementary

Bino-LSP dark matter

68

Z/h funnel regions are almost entirely excluded

​

𝜒̃^±₁/𝜒̃⁰₂ co-annihilation scenarios are compressed → hard to target

​

A/H funnel allows non-compressed models at higher mass

​

​

Higgsino/Wino-LSP dark matter

69

Higgsino-LSP models unexcluded right on the LEP limit!

​

Wino-LSP scenarios that saturate the relic density are entirely excluded (mostly by direct DM detection experiments)

Overall exclusion

70

ATLAS exclusion of each sparticle (after all external and dark matter constraints)

EWKino scan

Mainly wino/higgsino LSP

Bino-DM scan

Comparison with Run 1

71

A similar effort was published in 2015 using the Run 1 2L, 3L and 4L searches:

https://atlas.web.cern.ch/Atlas/GROUPS/PHYSICS/PAPERS/SUSY-2015-12/

​

This used a different model sample and scanning strategy, but we can still compare to get a qualitative idea of how the sensitivity has improved…

Run 1 vs. Run 2 EWK pMSSM scans

72

Run 1

Run 2

Run-1 vs. Run-2 EWK pMSSM scans

73

Run 1

Run 2

Uncovered scenarios

74

Benchmark models

75

Remaining Z/h funnel region model

​

Higgsino-like 𝜒̃^±₁/𝜒̃⁰₂

→ smaller production cross-section than typical simplified models

Benchmark models

76

A/H funnel region model with mixed 𝜒̃⁰₂ branching fractions

Benchmark models

77

Radiative 𝜒̃⁰₂ → 𝜒̃⁰₂ + 𝞬 decay

​

Heavier electroweakinos in reach

​

​

​

Summary

78

• Electroweak SUSY is alive and well
• Future searches shouldn’t forget about improving depth of sensitivity in “covered” regions.
• These results can provide benchmark models for future searches.
• Results and model files will be made public to allow more detailed phenomenological study of unexcluded models.
• Similar scans including sleptons and squarks are in the works!

​

Thanks for listening!

See ATLAS-CONF-2023-055 for more details!

Backup

79

Stop limits

80

BUT simplified assumptions go into these limits too!

→ multiple complex decay chains can dilute signals

​

, arXiv:1710.11091

Slepton limits

81

g-2 hints at low-mass smuon

Strong production

82

How light can squarks and gluinos be?

83

Squarks and gluinos ~ 1 TeV if g-2 goes away

→ Bagnaschi, Sakurai, Ellis et al, arXiv:1710.11091

​

SUSY prediction of Higgs mass

84

From talk by John Ellis at ATLAS SUSY workshop 2022

Run 1 General pMSSM scan

85

Run 1 General pMSSM scan

86

Run 1 General pMSSM scan

87

EWK pMSSM scan range

88

89

Scan workflow

External constraints

90

91

RECAST

• RECAST: Complete preservation of your analysis software environment and workflow

​

​

​

​

• REANA: Infrastructure for running RECASTs on the cloud
• I’ve run over 5000 analysis jobs with REANA for 1000s models and multiple analyses!

92

https://recast.docs.cern.ch/

“EWKino” scan

93

• Flat scan of parameters relevant to EWKino production → 20,000 models sampled
• Sleptons and squarks decoupled
• Most bino-LSP models have 𝝮h² ≥ 0.12

“Bino-DM” scan

94

• Oversample bino-LSP models with satisfactory DM relic density
• 437,500 models sampled → 11,163 have bino-LSP and 𝝮h² ≤ 0.12

Analysis simplifications

• Simplified likelihood used for 3L offshell, Compressed and 4L
• Orders of magnitude faster
• Merges background samples and nuisance parameters
• Sets nominal background yield to the post-background-only-fit value
• Truth-smeared yield scaling for FullHad, 1Lbb, 2L0J, 2L2J
• Scale factor calculated via least-squares fit of truth-smeared vs reco-level yields for test sample of models
• Truth-level fits don’t include signal systematic or statistical uncertainties
• Statistical U/C dropped as it would kill sensitivity in very high cross-section models
• Disappearing track: Evaluated using upper limits on production cross-sections rather than SimpleAnalysis/RECAST

95

96

97

2L2J

98

2L2J simplified model:

• Pure-Bino 𝜒̃⁰₁
• Pure-Wino 𝜒̃⁰₂ and 𝜒̃^±₁
• BR(𝜒̃⁰₂→ 𝜒̃⁰₁ Z^*)= 100%
• BR(𝜒̃^±₁→ 𝜒̃⁰₁ W^*)= 100%

Preference for Z decay mode

​

Sensitivity to compressed scenarios when X₂^+- and X₄⁰ contribute

4L

99

4L targeted GGM and R-parity violating scenarios

Sensitivity to long decay chains involving heavier electroweakinos

EWKino scan unique sensitivity

100

Higgs mass

101

Search overlaps

102

Bino-DM scan

EWKino scan

Constraint overlaps

103

Bino-DM scan

EWKino scan

Flavour constraints

104

EW precision constraints

105

Overall exclusion

106

Overall exclusion

107

Overall exclusion

108

Overall exclusion

109

Bino-DM scan unique sensitivity

110

Branching fractions

111

For a Bino-LSP, searches are favouring the 𝜒̃⁰₂ → Z 𝜒̃⁰₁ decay mode

→ Partly due to the selection of searches used

→ The Z decay mode is also more versatile

Overall exclusion

112

Direct detection complementarity

113

Direct detection complementarity

114

ATLAS and Lux-Zeplin sensitivity is complementary

Run-1 vs. Run-2 EWK pMSSM scans

115

Run 1

Run 2