LHCb: Stealth 4th Red LHC workshop (Virtual) (4 – 6 November 2020)

Carlos V´azquezSierra

CERN

November 5, 2020

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 1 / 18 The LHCb detector [IJMP A30 (2015) 1530022]

LHCb is a 20 5 m GPD in the forward region (upgrade during LS2 – see backup). × Single-arm forward spectrometer (2 < η < 5) along the beamline (z). Tracking and vertexing systems identification systems Side view

π± e±

VELO µ± TT RICH1 RICH2 Magnet ↓ T1 T2 T3 Excellent vertexing resolution (fast mixing). M1 SPD HCAL Excellent mass resolution (low masses). PS/ECAL M2 M3 M4 M5 Excellent jet reconstruction and tagging capabilities – see backup.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 2 / 18 The LHCb detector [IJMP A30 (2015) 1530022]

LHCb is a 20 5 m GPD in the forward region (upgrade during LS2 – see backup). × Single-arm forward spectrometer (2 < η < 5) along the beamline (z). Tracking and vertexing systems Particle identification systems Side view

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 3 / 18 The LHCb detector [IJMP A30 (2015) 1530022]

Hardware level L0: removed for Upgrade Ia (UIa). → benefit for low mass searches. → Software level HLT: Topological triggers on DV. → Down to pT 80 MeV/c (µ). → ∼ UIa: full event reco (30 MHz). → benefit for exotic searches. → Turbo (since 2015) lines: Any event part can be saved. → Can work directly on them. → Online µ-ID and jets in turbo. → GPU-based HLT1 (Allen project) from UIa [Comp Soft Big Sci (2020) 4 7]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 4 / 18 STEALTH physics @ LHCb

[STEALTH physics @ LHCb] workshop (J. Zurita):

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 5 / 18 STEALTH physics @ LHCb

[STEALTH physics @ LHCb] workshop (J. Zurita):

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 6 / 18 Dark [PRL (2020) 124 041801]

Search for dark photons decaying into a pair of :

Kinetic mixing of the dark (A0) with off-shell photon (γ∗) by a factor ε:

1 A0 inherits the production mode mechanisms from γ∗. + + 2 A0 µ µ− can be normalised to γ∗ µ µ−. → → 3 No use of MC no systematics from MC. → Separate γ∗ signal from background and measure its fraction. Prompt-like search (up to 70 GeV/c2) displaced search (214 – 350 MeV/c2). → A0 is long-lived only if the mixing factor is really small. 1 Used 5.5 fb− of Run 2 LHCb data (13 TeV).

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 7 / 18 Dark Photons [PRL (2020) 124 041801]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 8 / 18 Dark Photons [PRL (2020) 124 041801]

No significant excess found – great sensitivity (especially m > 10GeV and m < 0.5 GeV). Displaced search is also done (small displaced region is excluded) – see backup. 2 ε − 10 5

− 10 6

− 10 7 LHCb LHCb prompt-like A0 search LHCb (2016 data) BaBar+KLOE+CMS − 10 8 1 10 m(A0) [ GeV ] 10−2 A1 HPS APEX (g 2)e LHCb

ε − KLOE CMS NA48/2 BaBar KLOE −3 BaBar 10 E774 LHCb

NA64 10−4 E141

LHCb KEK − 10 5 Orsay PS191 NOMAD ν-CAL I CHARM E137 10−6 10−2 10−1 1 10 m(A0) [ GeV ]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 9 / 18 Dark Photons – non-minimal searches [JHEP 10 (2020) 156]

Signature sensitive to other models model-independent search: →

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 10 / 18 Dark Photons – non-minimal searches [JHEP 10 (2020) 156]

UL @ 90% C.L. on σ(X µµ) (top: inclusive, bottom: b-associated): →

2HDM Higgs θH world best limits: LHCb R1 [JHEP→ 09 (2018) 147] → CMS R1 [PRL 109 (2012) 121801] → CMS R2 [PRL 124, 131802 (2020)] → Belle Y X γ [PRD 87 (2013) 031102] → → Other scenarios covered too (i.e. HV).

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 11 / 18 Dark Photons – the future

0 0 Cover di- final states in D∗ D A0(ee) decays: Hardwareless trigger is required (softer→ final state than in the di- mode), → High statistics get 3 1011D0 per inverse fb! → → × Prospected reach for Run III and beyond: [arXiv:1812.07831]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 12 / 18 Light dark bosons decaying into µµ [JHEP 09 (2018) 147]

Light -0 copiously produced in -gluon fusion: Many models: NMSSM, 2HDM+S, etc. Recent review on LHC searches: [arXiv:1802.02156] 1 Search using LHCb Run 1 (3 fb− ) published in JHEP. Look for a di-muon resonance from 5.5 to 15 GeV/c2 (also between Υ peaks): Mass-interpolated efficiencies in bins of pT , η (model independent results also given). Production x-section (8 TeV) limits for a scalar (vector) on the left (right). First scalar limits between 8.7 and 11.5 GeV/c2 and competitive with CMS elsewhere. No excess observed for more details ask me during the coffee break → 20 / 100 , LHCb Υ(1S) 95% CL upper limits LHCb Υ(1S) 95% CL upper limits ) [pb] ) [pb] Υ(2S) Median expected Υ(2S) Median expected

√s = 8 TeV µµ √s = 8 TeV µµ Υ(3S) 68% expected 80 Υ(3S) 68% expected 15 → →

95% expected 0 95% expected φ A ( Observed ( Observed CMS expected 60 × B × B ) )

CMS observed 0 φ

10 A → → ¯

q 40 gg q ( ( σ σ 5 20

0 0 6 8 10 12 14 6 8 10 12 14 m(φ) [GeV] m(A0) [GeV]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 13 / 18 Dark decaying into jet pairs [EPJC (2017) 77 812]

0 HV πv decaying to bb¯ – especially with SM-like H πv πv production. → In most of the cases only one of the two πv decays into the LHCb acceptance. Experimental signature is a single displaced vertex with two associated jets. 1 Limits with partial LHCb Run I (2 fb− ) dataset published (95% C.L. below):

3 ) 10 v

π 2 2 v mπ = 25 GeV/c mπ = 50 GeV/c Reconstruct the displaced vertex and v v 2 2

→ π mπ = 35 GeV/c mπ = 35 GeV/c , πv → cc̄ v v LHCb 0 2 2 H 2 m = 43 GeV/c m = 35 GeV/c , π → ss̄ find two associated jets. ( 10 πv πv v ⋅ ℬ ) 0

Use πv detachment to discriminate →H SM gg 10 σ/σ between signal and background. (

Background dominated by bb¯ events 1 and material interactions. 10−1 No excess found – next steps:

Analyse LHCb Run 2 data for lower 10−2 → 1 10 102 103 masses/jet substructure. Lifetime [ps] Confining dark showers [PRD 97 → (2018) 095033]. Exclusive displaced hadronic → vertices [JHEP 01 (2020) 115].

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 14 / 18 ( ) + Hidden-sector bosons in B K ∗ χ(µ µ−) → 0 0 + + B K ∗ χ [PRL 115 (2015) 161802] / B K χ [PRD 95 (2017) 071101 (R)] → + → Search for hidden-sector bosons χ µ µ− in b s penguin decays: → → Axial-vector portal (χ as ) [LNP 741 (2008) 3] Scalar (Higgs) portal (χ as inflaton) [JHEP 05 (2010) 10]

0 First dedicated search (K ∗ χ) over such a large mass range: 0 + Pro: K ∗ K π− vertex leads to better τ(χ) resolution and less background. Con: B0 →K 0χ has smaller branching fraction than the B+ K +χ mode. → ∗ → Allow for prompt and detached di-muon candidates – up to 1000 ps ( 30 cm). ∼

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 15 / 18 ( ) + Hidden-sector bosons in B K ∗ χ(µ µ−) → 1 Full LHCb Run I dataset (3 fb− ) used for both searches. Look for a narrow di-muon peak (mass resolution between 2 and 9 MeV/c2). Exclude narrow QCD resonances [PRL 115 (2015) 161802] MVA selection almost independent of χ mass and decay time (uBoost). BR normalised to (B+ K +J/ψ)( 10 4) or (B0 K 0µ+µ )( 10 7). B → ∼ − B → ∗ − ∼ − Constraints on τ(χ) between 0.1 and 1000 ps (left), [PRD 95 (2017) 071101 (R)] Constraints on mixing angle θ2 between the Higgs and χ in the inflaton model (right):

No evidence for signal observed. Large fraction of allowed inflaton parameter space ruled out.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 16 / 18 Conclusions

LHCb proved to be very competitive for stealth physics: Excellent vertexing, tracking and soft trigger. Especially competitive for low masses and lifetimes. Rich variety of models and signatures can be approached. Bright prospects for the future: Removal of hardware trigger access softer kinematics. Better vertex resolution and tracking→ capabilities. New techniques under development for ideas on new signatures. Many other existing results and proposed studies not covered: LFV Higgs decays [EPJC (2018) 78 1008] Searches for HNLs (produced from B and D decays, or from prompt W decays) (existing). LLP µ+jets (existing), ALPs γγ (on-going), True at LHCb (proposed)→ see backup. → → White paper on Stealth physics at LHCb in preparation stay tuned! → We are looking forward to ideas for new signatures and techniques: Do not hesitate to contact us if interested!

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 17 / 18 Thanks for your attention!

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 18 / 18 Backup

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 19 / 18 The future of LHCb

Physics case for an LHCb Upgrade II: Opportunities in flavour physics, and beyond, in the HL-LHC era [CERN-LHCC-2018-027]

Challenging conditions – higher rate, pile-up, occupancy and fluence. 1 Expect to collect 300 fb− by the end of Upgrade II. Detector sub-systems have to be able to cope with such conditions. In particular – trigger and tracking systems are crucial for exotic searches.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 20 / 18 The upgraded LHCb VELO

Upgrade II VErtex LOcator: [CERN-LHCC-2017-003] Probably based on Upgrade Ia VELO (silicon pixels). Access to shorter lifetimes, better PV and IP resolution, and real-time alignment. But – 10x multiplicity, pile-up and radiation damage w.r.t. Upgrade Ia(b). Possibility of removing RF-foil for Upgrade II: better IP resolution + no material interactions. →

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 21 / 18 VELO material map [JINST 13 (2018) P06008]

Background dominated by material interactions for displaced searches at LHCb. Mandatory to keep control of material interactions – veto them in an efficient way:

Background mainly due to γ conversions (left plot). A new VELO material map has been developed: Model in great detail both sensors & envelope. Assign a p-value to material interaction hypothesis. Sensitivity improvement by (10) to (100). Based on data from beam-gasO collisionsO (plot below).

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 22 / 18 The LHCb reconstruction

Tracks with tracking stations & VELO hits (a.k.a. long tracks): Excellent spatial and momentum resolution. Reconstruction of particles decaying within VELO. Presence of a VELO envelope (RF-foil) at 5 mm from beam: Background dominated by heavy flavour∼ below 5 mm. → Background dominated by material interactions above 5 mm. →Having a precise model of material interactions is crucial. A detailed VELO material veto map is used: [JINST 13 (2018) P06008] Sensitivity improvement by one to two orders of magnitude. → See backup for more details on the material veto map. →

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 23 / 18 The LHCb reconstruction

Downstream tracks: Reconstruction of particles decaying beyond VELO. Tracks with worse vertex and momentum resolution. Trigger on downstream tracks better for LLP (6 2 m) signatures. Optimisation studies on-going →[LHCb-PUB-2017-005] Upstream tracks: Reconstruction of soft charged particles bending out of the acceptance. New tracker (UT) – high granularity, closer to beam pipe. Proposal to add magnet stations (MS) inside the magnet improve low p resolution. →

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 24 / 18 Jet reconstruction and identification at LHCb

Jet reconstruction: [JHEP (2014) 01 033] Particle flow algorithm (including neutral recovery) jet input. → Anti-kT algorithm for clustering (R = 0.5) efficiency > 95% for pT > 20 GeV. Jet energy scale calibrated on data (using Z→ µµ + jets), → Energy resolution from 10 to 15% for a pT range between 10 and 100 GeV. Secondary Vertex (SV) identification and jet tagging: [JINST 10 (2015) P06013] Reconstruct SV from displaced tracks kinematic and quality requirements on both, → Train two Boosted Decision Trees (BDTs) for a two-step jet flavour tagging: SV displacement from PV, kinematics, charge and multiplicity; p 2 2 2 SV corrected mass, defined as Mcorr(SV) = M + p sin θ + psinθ. BDT(bc udsg) to separate light and heavy flavour jets, BDT(b c) to separate b from c-jets. Tagging| efficiency of b(c)-jets of 65% (25%) with 0.3% contamination| from light jets.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 25 / 18 ( ) + Hidden-sector bosons in B K ∗ χ(µ µ−) →

1 Full LHCb Run I dataset (3 fb− ) used for both searches. Look for a narrow di-muon peak (mass resolution between 2 and 9 MeV/c2). Exclude narrow QCD resonances - mass distribution: [PRL 115 (2015) 161802]

MVA selection almost independent of χ mass and decay time (uBoost).

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 26 / 18 Massive LLPs decaying to µ + jets [EPJC (2017) 77:224]

Massive LLP into µ + two ( jets). → Signature sensitive to several benchmark models: mSUGRA RPV , Right-handed (Majorana) , Simplified MSSM production topologies:

One particular example: decay of a Higgs-like particle into two LLPs.

Look for a single displaced vertex with several tracks + high pT muon. Background dominated by bb¯ events and material interactions.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 27 / 18 Massive LLPs decaying to µ + jets [EPJC (2017) 77:224]

1 Search with full Run I (3 fb− ) LHCb data published last year. Results interpreted in H0 χ˜0χ˜0 benchmark model: → 1 1

Different constraints on (H0 χ˜0χ˜0) at 95% CL with Run 1 data at LHCb B → 1 1 ] 2 c 80 > 50% excl. B > 25% excl. [GeV/ 70 LHCb preliminary B 0 1 > 10% excl. χ B m > 5% excl. B 60

50

40

30

20

10 4 3 2 1 10− 10− 10− 10− 0 χ˜1 cτ [m]

Stringent limits – rejecting (H0 χχ) >10% down to 30 GeV/c2 (5 ps). B → No excess observed.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 28 / 18 Massive LLPs decaying to µ + jets [EPJC (2017) 77:224]

Prospects for Phase-II some naive extrapolations below: → Scale signal and background – consider increase of cross-sections, Conservative assumptions for jet reco, trigger, and material interactions, Optimistic assumptions for pile-up effect.

0 0 0 0 0 0 1 Regions where (H χ˜ χ˜ ) > 5% is excluded at 95% CL Different constraints on (H χ˜ χ˜ ) at 95% CL with 300 fb− at LHCb B → 1 1 B → 1 1 ] ] 2 2

c 80 1 c 80 LHCb 300 fb− > 5.0% excl. B LHCb published Run 1 > 1.0% excl. [GeV/ 70 [GeV/ 70 LHCb reachable B 0 1 0 1 > 0.5% excl. χ LHCb reachable χ B m m > 0.1% excl. B 60 60

50 50

40 40 LHCb preliminary

30 LHCb preliminary 30

20 20

LHCb jet substructure LHCb jet substructure

10 4 3 2 1 10 4 3 2 1 10− 10− 10− 10− 10− 10− 10− 10− 0 0 χ˜1 cτ [m] χ˜1 cτ [m] Our main aim is to reach lower masses and lower lifetimes. Removal of L0 trigger (Phase-I) much higher trigger efficiencies at the end! → Jet reconstruction efficiencies will be better for lower masses. Expected a better knowledge of material interactions (x3 less for UIa VELO).

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 29 / 18 Massive LLPs decaying to µ + jets (recast)

Limits recasted to look into sterile neutrinos [PLB (2017) 774 114-118]

Could we get best world-limit (5–10 GeV/c2) with same kind of search? Dedicated search with Run II data in preparation.

@ 95% C.L.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 30 / 18 Dark pions decaying into jet pairs [EPJC (2017) 77 812]

Naive prospects for Upgrade II (loose assumptions):

0 0 1 Regions where (H πV πV ) > 50% is excluded at 95% CL Different constraints on (H πV πV ) at 95% CL with 300 fb− at LHCb 70 B → 70 B → ] ]

2 1 2 c LHCb 300 fb− c > 50% excl. B LHCb published Run 1 > 5% excl. 60 60 B [GeV/ CMS dominated [GeV/ CMS dominated > 2% excl. V V

π π B > 1% excl. m m B 50 50

40 40

30 LHCb preliminary 30 LHCb preliminary ATLAS dominated ATLAS dominated 20 20 LHCb jet substructure LHCb jet substructure

10 4 3 2 1 0 1 10 4 3 2 1 0 1 10− 10− 10− 10− 10 10 10− 10− 10− 10− 10 10 πV cτ [m] πV cτ [m] L0 removal (Upgrade Ia) highly beneficial access to lower jet masses. → Higher pile-up in Upgrade II: Impact of pile-up on jet reconstruction efficiencies needs to be studied in much detail. We have reasons to be optimistic – preliminary studies ongoing + ideas (see below). Some possible improvements to mitigate the effect of the increased pile-up: Remove neutrals (more pile-up dependent) from jet reco (only charged tracks). Consider ML techniques to seize pile-up contributions as in ATLAS and CMS.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 31 / 18 Dark Photons – displaced search [PRL (2018) 120 061801]

Looser requirements on muon tranverse momentum. Material background mainly from photon conversions [JINST 13 (2018) P06008] 0 + Isolation decision tree from Bs µ µ− search: [PRL (2018) 118 191801] Supress events with additional→ number of tracks, i.e. µ from b- decays. → Fit in bins of mass and lifetime – use consistency of decay topology χ2. Extract p-values and confidence intervals from the fit:

A0 2 A0 2 90% CL upper limit on nob[m(A0), ε ]/ nex [m(A0), ε ] −8 10 −8 10 n 10 2 ex A ε 9 2 0 [ ε m 8 ( A 7 0 ) −9 10−9 LHCb 10 ε , 10 LHCb 6 2 ] 5 4 3 −10 10 10−10 2 1 1 0 250 300 350 250 300 350

m(A0) [MeV] m(A0) [MeV]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 32 / 18 Dark Photons – combined prospects

Minimal scenario (LHCb) + Higgs portal (ATLAS/CMS):

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 33 / 18 Discovering True Muonium at LHCb [arXiv:1904.08458]

3 LHCb interested in vector ( S1) state: Direct production via kinetic mixing of TM with γ∗ `ala . Dissociation factor estimated to be 50% for Run 3 material budget. ∼ QED predicts τ 1.8 ps and mass 211 MeV: ∼ ∼ Very low mass off-shell γ production via η γγ∗ (small BR but high σ). Experimental signature→ resolvable displaced→ vertex decaying into e+e pairs. → − + Inclusive search – reconstruct only γ∗ e e− (simpler, smaller systematics). → Prospects (Run 3 & beyond) – different, conservative assumptions on reco ε: + Assume current LHCb e e− mass resolution ( 20 MeV at 211 MeV). 1 ∼ Discovery potential at 15 fb− (5σstat), inclusive search:

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 34 / 18 Discovering True Muonium at LHCb [arXiv:1904.08458]

+ True muonium (TM), or µ µ− (analogous to ): 1 paraTM: S0 pseudoscalar state, decays predominantly into γγ pairs. 3 + orthoTM: S1 vector state, decays predominantly into e e− pairs. Properties well predicted by QED: Provide precision tests of QED, Sensitive to BSM deviations in the TM decay rate. → However – TM is a fragile bound state: + TM could dissociate into µ µ− before decaying due to material interactions, LHCb has an aluminium (AlMg3) foil enveloping the VELO. Dissociation factor is calculated and included in these calculations.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 35 / 18 Discovering True Muonium at LHCb [arXiv:1904.08458]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 36 / 18 Discovering True Muonium at LHCb [arXiv:1904.08458]

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 37 / 18 ALPs decaying into pairs of photons

2 Constraints from LHC resonance searches above ma 60 GeV/c (a γγ, jj). ∼ → But – poor limits for low masses use γγ x-section measurements. [PLB (2018) 06 039] → LHCb could cover the region between 3 and 10 GeV/c2 (recast): [JHEP 1901 (2019) 113]

Trigger (MVA) for soft γγ searches. Two selections [LHCb-PUB-2018-006] [arXiv:1906.09058] B: cut around m(B0) (since 2015). → s ALP: up to 11 GeV/c2 (only 2018). → Planned search using 2018 LHCb data.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 38 / 18 ALPs decaying into pairs of photons

Trigger algorithm (MVA-based) for soft γγ searches: 1 Uses converted and calorimeter γ, 2 Pre-filters candidates by ET , 3 Combines two candidates to form γγ, 4 Filters again by ET , pT and γγ mass.

Carlos V´azquezSierra 4th Red LHC workshop (Virtual) November 5, 2020 39 / 18