HL-LHC/FCC-HH DIHIGGS SEARCHES - JAHRED ADELMAN - JAHRED ADELMAN - INDICO-FNAL

1

 HL-LHC/FCC-hh
 DiHiggs searches
 Jahred Adelman

Jahred Adelman NIU EF Snowmass

Key long-term LHC (and beyond!) goal

Observe the Higgs boson self-coupling, crucial
 to testing if the Higgs potential is the one
 predicted in the Standard Model (SM)

 2

Reminder: Why do we care so much about this?
• Study the exact dynamics of
 electroweak symmetry
 breaking, map out the Higgs
 boson potential
• Further our understanding of
 potential solutions to the
 hierarchy problem
• As a window into new
 physics beyond the SM
• Study EWSB to understand
 matter-antimatter
 imbalance and EW
 baryogenesis 3

Where might you study this?
 SM hh production dominated by box diagram, not
 hh self-coupling of triangle diagram, with
 destructive interference between the two
 Reminder: It took 40 years to observe the Higgs boson.
 We’ve good great machines and we’re clever, but unless
 BSM physics completely surprises us, we’ll need lots of
luminosity and √s (HL-LHC and FCC-hh are great options)
 LHC: hh
 Need production
 differential
 measurements 3 orders of
 to understand magnitude
 λ more rare than
 single h
1910.00012 production! 4

Recapping the HL-LHC
 ATL-PHYS-PUB-2018-053 and

 Reminder: HL- CMS-PAS FTR-18-019

LHC is needed by
 the LHC to
 observe SM hh
 production (and
even then, it won’t
 be a piece of
 cake) 5

Recapping the HL-LHC

 Quick recap of
 what we know
about the HL-LHC
 searches …

 6

Studying HL-LHC prospects

 4b channel is an
 obvious one, with
 the highest hh BR.
 Challenge: Large
 backgrounds
 (multijet and ttbar)

 7

hh(4b)
• Multi-jet and ttbar Signal

 2j [GeV]

 Events / 25 GeV2
 ATLAS Simulation
 background critical to
 200 0.16
 s = 13 TeV, 24.3 fb-1

 msubl
 Resolved, 2016
 180 0.14

 understand 160 0.12

 SR
 0.1

• Increased b-tagging
 140

 120 0.08

 efficiency important, CRs 100 0.06

 0.04

 boosted topology for BSM
 80

 0.02
 60

• Trigger is tricky and 60 80 100 120 140 160 180
 mlead
 2j
 200
 [GeV]
 0

 CMS Phase-2 3000 fb-1 (14 TeV)
 crucial to understand 10 9

 Events / bin width
 Simulation Preliminary Multijet
 HH → bbbb tt

 (combination of multijet 108
 Single Higgs
 Bkg. uncertainty
 SM HH (× 100)

 and b-jet triggers?) 107

• Roughly 1.0-1.5σ at 3 106

 ab-1, but systematics 105

 degrade performance 104

 significantly −0.2 0 0.2 0.4 0.6 0.8
 BDT output 8

hh(4b)
 CMS Phase-2 3000 fb-1 (14 TeV)
 Loss in signal significance [%]

 30 Simulation Preliminary

 95% CL exclusion limit on σHH/ σSM
 HH
 3.4
 ATLAS Preliminary
 3.2
 25
 HH → bbbb Projection from Run 2 data
 3
 s = 14 TeV, 3000 fb-1
 2.8 HH→bbbb
 20 2.6 No systematic uncertainties
 2.4
 15 2.2
 2
 1.8
 10
 1.6
 1.4
 5
 1.2
 30 40 50 60 70 80 90 100 110
 Minimum offline jet p [GeV]
 0 T
 45 50 55 60 65 70 75 80

 Minimum jet p threshold [GeV]
 T

• Improving (worsening) background modeling helps
 (hurts) significantly
• Raising jet pT thresholds due to trigger reduces
 sensitivity quite a bit

 9

Studying HL-LHC prospects

 bb channel has
 relatively large BR
 and is best current
 ATLAS channel. Split
 analysis based on
 tau decays (had vs
 lep)

 10

ATLAS hh(bb )

 Sub-leading τhad-vis offline p threshold [GeV]

 95% CL exclusion limit on σHH/ σSM
 HH
• Important question: What 60
 ATLAS Preliminary
 Projection from Run 2 data
 s=14 TeV, 3000 fb
 -1
 1.55 1.46
 1.5

 1.4

 will hadronic tau triggers 50 HH→bbτhadτhad 1.13 1.12 1.08 1.3

 T
 κ λ = 20 BDT

 look like? 40 0.97 0.96 0.98 0.92
 1.2

 1.1
• Dominant backgrounds 1

 rejected with MVAs: ttbar, 30 0.92 0.91 0.93 0.92
 0.9

 30 40 50 60 70
 QCD and Z+jets Leading τhad-vis offline p threshold [GeV]
 T

• Most background
 normalization data-driven
 and should scale with lumi
• 1.5-2.5σ evidence at end
 of HL-LHC

 11

Studying HL-LHC prospects

 bbγγ channel has
 smaller BR but
 good mass
 resolution and is
 best current CMS
 channel

 12

ATLAS hh(bbγγ)
• Take advantage of

 (1/N) dN / dx
 Signal (test sample)
 20
 Background (test sample)
 18

 diphoton mass resolution 16 ATLAS Simulation Preliminary
 14 s = 14 TeV, 3000 fb-1

• Look for two photons and 12

 two b-jets near Higgs 10

 8

 mass, use MVAs to reject 6

 4

 backgrounds (continuum/ 2

 0
 0.4 0.42 0.44 0.46 0.48 0.5 0.52 0.54 0.56 0.58
 tth dominate) (a) mgg , high mass categoryBDT response

• Split into bins of mHH to CMS Phase-2 3000 fb-1 (14 TeV)

 Events/(1.6 GeV)
 50 Simulation Preliminary Pseudo-data
 improve sensitivity to 40
 pp→HH→γ γ bb Nonresonant backgr.
 HP, 350 < MX < 480 GeV Full backgr.
 multiple couplings 30
 Sig. + Full backgr.

• Roughly 2.0σ 20

 significance at end of HL- 10

 LHC, systematics 0
 105 110 115 120 125 130 135 140 145

 negligible
 mγ γ [GeV]

 (c) mgg , medium mass category
 13

Studying HL-LHC prospects

 bb+4lepton
 channel has
 very few
 expected
 events but is
 quite clean

 14

CMS hh(bb+4L)
• Delphes parametric analysis
 CMS Phase-2 Simulation Preliminary 3000 fb -1 (14 TeV)
• Main irreducible

 Events/0.1 GeV
 0.3
 ttZ

 backgrounds: tth(ZZ), ggH 0.25
 WH
 VBF(H)
 ZH
 gg(H)
 and ZH 0.2
 ttH
 HH → bbZZ(4l)

• Reducible ttbar and DY 0.15

 backgrounds have much
 0.1
 larger cross sections, hard to
 0.05
 model with available Delphes
 samples, assumed to be 0
 120 121 122 123 124 125 126 127 128 129 130

 negligible
 3 ab-1 significance
• Go down low in pT (5/7 GeV
 0.37σ, systematic
 for ele/µ), form SF-OS Z
 uncertainties
 candidates and make
 negligible
 kinematic requirements
 15

Studying HL-LHC prospects

 Combinations
 obviously needed
 to study self-
 coupling

 16

ATLAS hh combination
 Second minimum due to signal yield 8

 -ln(Lκ / Lκ =1)
 ATLAS Preliminary bbbb

 λ
 similar to first minimum (mhh helps 7 Simulation and Projections from Run 2 data bbττ

 λ
 s = 14 TeV, 3000 fb-1, κλ = 1
 break degeneracy) 6 Systematics uncertainties included bbγ γ
 Combination
 5

 4

 3
• Combine 3 dominant 2 2σ

 channels, 3.5σ evidence 1

 0
 1σ

 −1 0 1 2 3 4 5 6 7 8
 without (3.0σ with) κλ

 systematic 8

 Significance [σ]
 ATLAS Preliminary
 uncertainties 7 Simulation and Projections from Run 2 data
 s = 14 TeV, 3000 fb-1
 6
• As expected, sensitivity 5
 Systematic uncertainties included
 bbbb
 bbγ γ

 varies quite a bit if λ != 4

 3
 bbττ
 Combination

 1 (BSM!) 2

 1
• Critical to combine all 0
 −2 0 2 4 6 8
 channels κλ

 17

CMS hh combination
 CMS Phase-2 3000 fb-1 (14 TeV) CMS Phase-2 3000 fb-1 (14 TeV)
 10

 -2Δln(L)
 σ (gg→HH) [fb]

 104 Simulation Preliminary Assumes no HH signal Simulation Preliminary Assumes SM HH signal
 9
 95% CL upper limits - Median expected bbbb
 bbbb bbττ 8 bbττ
 bbVV(lνlν) bbγ γ bbVV(lνlν)
 bbZZ*(4l) Combination 7 bbγ γ
 3
 10 Theoretical prediction bbZZ*(4l)
 6
 Combination
 5

 4 95%
 2
 10 3

 2

 1 68%

 0
 −4 −2 0 2 4 6 8 10 −4 −2 0 2 4 6 8 10

 κλ κλ

• Combine 5 channels, 3.6σ evidence without (2.8σ
 with) systematic uncertainties
• As expected, sensitivity varies quite a bit to BSM
 physics
• Critical to combine all channels
 18

HL-LHC combination 1902.00134

• Combine results from both experiments, 4.5σ
 evidence without (4.0σ with) systematic
 uncertainties
• Even for the combination, SM will be tricky to study
• Critical to develop new ideas, to combine with
 single Higgs measurements 19

Recapping the HL-LHC

 Some
 thoughts and
 ideas and
 lessons now
 that we had
 a taste … 20

Some lessons / thoughts
 CERN-LHCC-2017-020

• Trigger thresholds clearly
 critical to understand for 4b
 and bb . Public results for HL-
 LHC upgrade: Interesting to
 examine? (b-jet triggers, tracks
 early on in trigger, asymmetric
 triggers, etc)
• What about the size of the 4b
 multijet background? Can
 more recent results be CMS-HIG-17-017
 illuminating for some of the HL-
 LHC and/or FCC-hh studies?
 21

Some lessons / thoughts 2001.05178
• What about the addition of
 extra channels? VBF hh (4b)
 helps to study extra couplings
 (what range of c2v are we
 sensitive to)? Can it improve
 sensitivity to overall hh? True not
 only for 4b analysis but perhaps
 also for other analyses?
• Can VBF channel be useful in
 certain BSM physics models?
• Do we have sensitivity to VHH
 at HL-LHC (or beyond) in, for
 example, 4b channel? And in
 BSM models?
 22

Some lessons / thoughts 1904.08549

• Do we gain anything from re-optimizing analyses
 using the latest MVA tools and techniques? Maybe
 for new channels, but also for boosted topologies?
• Can we gain from adding many small channels
 together such as 4L + bb?

 23

Some lessons / thoughts

 Difference between 4.5σ evidence without and (4.0σ
 with) systematic uncertainties is critical. Are we
 missing important systematics? Perhaps to be studied?
 (This is tricky, of course)
 24

Some lessons / thoughts
 3000 fb-1 (14 TeV) ATLAS-CONF-2019-049
 σ(pp → HH → bbbb) [fb]

 CMS Phase-2 95% CL upper limits
 5
 10 Simulation Preliminary Median expected
 68% expected
 104 95% expected

 103

 102

 1 2 3 4 5 6 7 8 9 10 11 12 SM
 Shape benchmark

 Useful to think about EFT models and benchmark
BSM scenarios, not just for individual analyses, but also
as a combination. How best to do this? Requires shape
 analyses for best sensitivity, likely a global fit of all
 channels at once. Single Higgs inclusion crucial, too 25

Some lessons / thoughts
Example from Christoph’s talk at joint EF01/EF02 meeting. My takeaway from
 that talk: Lots of great opportunities to explore BSM physics, but a few
 benchmarks are critical for comparisons between collider options

 26

HL-LHC vs FCC-hh 1905.03764
 Size of ~1 sigma uncertainty on self-coupling
 hh analysis only, hh analysis only, allow Single h analysis only, only Single h analysis only,
 single Higgs coupling
 only allow λ allow EFT variations allow all possible coupling
 variations too within
 variations corresponding to λ shifts variations
 uncertainties

 27

HL-LHC vs FCC-hh
 Clear that combinations with single Higgs
 measurements are critical, and also that FCC-hh will
 significantly improve constraints

 28

Another view ECFA Higgs Study Group 19

 Challenge for FCC-hh: 1% uncertainty on top quark
 Yukawa coupling leads to 5% uncertainty on Higgs
 self-coupling!
 29

Quartic couplings? hh White paper

 Quartic Higgs coupling also very interesting to study,
 but cross sections even smaller. Perhaps FCC-hh has
 an ability to set limits on this?
 30

Other couplings at FCC-hh hh White paper

 FCC-hh provides an opportunity to
 better study the hhVV coupling (not
 yet in the context of VBF for more
 general hh searches) and also tthh
 coupling

 31

Recap of non-exhaustive list of things to think about
• 4b/bb : Rethinking triggers?
• 4b/bb : Systematic uncertainties?
• 4b: Any updates on size of multi-jet backgrounds?
• 4b: Quartic coupling sensitivity?
• All: VBF? For BSM? Improving sensitivity? c2V?
• All: Vhh? tthh?
• All: Latest MVA tools? For boosted? Otherwise?
• New channel: Adding small, missing channels?
• All: Benchmarks for BSM? Models and parameters
• Combination: Updates on single Higgs inclusion?

 32

BACKUP

 33

Studying HL-LHC prospects

 bbll channel
 dominated by
 bbWW decays,
 but also
 includes bbZZ

 34

CMS hh(bbll)
 CMS Phase-2 Simulation Preliminary 3000 fb-1 (14 TeV)

 Events
 HH → bbWW, µµ channel
 107 Signal (x2000) tt

• Delphes parametric 106
 κλ = 10
 κλ = 4
 Drell-Yan
 ttH
 κλ = -5
 analysis 105
 SM (κλ = 1)

• Background dominated by 104

 ttbar and Z+jets (dilepton 103

 mass cuts to reject Z+jets 102

 and quarkonium decays) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
 NN output
 1

• Use a NN with kinematic CMS Phase-2 Simulation Preliminary 3000 fb-1 (14 TeV)

 Events
 HH → bbWW, µe + eµ channels
 quantities as input to 107 Signal (x2000)
 κλ = 10
 tt
 Drell-Yan

 separate signal and 106
 κλ = 4
 κλ = -5
 ttH

 SM (κλ = 1)

 background 105

• 95% CL upper limit of 3.5 x 104

 SM cross section (3.3 without 103

 102
 systematics) 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
 NN output 35

Scaling with luminosity?

 Michael Kagan

 We are getting smarter and more clever, and results
 scale better than naive expectation. Bodes well for the
 future?

 36

CMS hh(4b)
 3000 fb-1 (14 TeV)

 σ(pp → HH → bbbb) [fb]
 CMS Phase-2 95% CL upper limits
 5
 10 Simulation Preliminary Median expected
 68% expected
 104 95% expected

 103

 102

 1 2 3 4 5 6 7 8 9 10 11 12 SM
 Shape benchmark

 Significant differences in expected upper limits
 depending on the benchmark model chosen

 37