Long Lived Particles

Sinéad M. Farrington

University of Edinburgh

Standard Model Shortcomings

• Standard Model: best-tested theory Fit residuals • Describes the fundamental particles and the interactions among them

• But... • 26 free parameters (compelling?) • Higgs mass appears to be unnaturally fine-tuned • Not possible to unify with gravity • Effective theory? (c.f. Classical mechanics è Special Relativity)

• Candidate overarching theories imply ‘New Physics’ • New fundamental particles • New fundamental interactions among them

2 S. Farrington, University of Edinburgh New Physics Search: Mass axis • Typically search for a mass resonance #

O1 Mass(O1,O2)

• Higher energy è Access higher mass states • Higher luminosity è Access rarer production processes • This has worked: ...J/ψ, ϒ, Z, Higgs, ... • But no evidence for resonances Beyond the Standard Model Ruled out phase-space: e.g. mass(Z’è τ+τ-) is > 2.4 TeV • This should not stop, it provides legacy constraints and there is a lot of discovery phase space still to explore. 3 S. Farrington, University of Edinburgh Alternative Axis: Lifetime LLP = Long Lived Single LLP production O1 Particle

LLP O2

4 S. Farrington, University of Edinburgh Alternative Axis: Lifetime LLP = Long Lived Single LLP production O1 Particle

LLP O2

OR O1

Double LLP production LLP invisible

LLP

invisible O2

5 S. Farrington, University of Edinburgh Why Long Lived Particles?

• Familiar case: b-hadrons • Long-lived because

• GF is small (= W mass is large) • Off-diagonal CKM matrix elements are small

• New Physics could be long-lived for the same reasons: • Massive propagators, small couplings • Lack of phase space • Inability to decay by a faster route e.g. because of mass difference to possible final states • It makes sense to search for these at the LHC to explore the full range of new physics that we can be sensitive to • It is part of a full exploitation of the LHC 6 S. Farrington, University of Edinburgh Standard Model Particle Lifetimes Prog.Part.Nucl.Phys. 106 (2019) 210-255

What else could be on this plane?

7 S. Farrington, University of Edinburgh Theory Motivations 1806.07396

(The short answer is that many theories motivate LLPs, this talk focuses on how the searches have been realised so far.) 8 S. Farrington, University of Edinburgh Experimental Considerations

• The hardware and software of ATLAS and CMS were designed with broad goals in mind and provide resolutions for many objects across large pt and mass ranges • But long lived particle sensitivity was not a major design criterion

9 S. Farrington, University of Edinburgh How do we detect LLPs at the LHC

• Bespoke reconstruction algorithms layered on top of standard ones, dedicated triggers in some cases • Will discuss several of these strategies • Sensitivity comes from a patchwork of methods

10 S. Farrington, University of Edinburgh Current Coverage (ATLAS)

11 S. Farrington, University of Edinburgh Current Coverage (CMS)

12 S. Farrington, University of Edinburgh LLP Link to 125 GeV Higgs Phys.Lett. B651 (2007) 374(379); Phys.Lett. B661 (2008) 263(267) • H(125) could mix with “dark sector Higgs”

• Or decay to long-lived scalars

• Or to dark fermions

(the fermions could be low momentum and hard to trigger on.) 13 S. Farrington, University of Edinburgh Current LHC Coverage: H to e/µ/j

• Large areas of phase space are being ruled out at the LHC • But small lifetimes are difficult for the LHC; large lifetimes capped by detector dimensions • Work is ongoing to maximise the sensitivity across LHC experiments and beyond 14 S. Farrington, University of Edinburgh This lecture

• Go through several types of searches, discuss the technical steps and results • Dileptons • Emerging jets • Displaced jet • Non-pointing photons • Heavy Stable Charged Particles • (not at all exhaustive, but intended to illustrate the breadth of experimental techniques)

… and what is next

15 S. Farrington, University of Edinburgh A few observations before we start • Coverage of the lifetime axis is patchy • Tends to be achieved by picking off parts of the phase space – very difficult to do a search across very broad lifetime spectrum because the technical methods are different across many of them • In most cases, the motivations are one or two theories, plus a model-independent desire to fully exploit the phase space accessible to our General Purpose Detectors (GPDs) • In many cases, non-standard identification is needed • Adaptations to existing code • Exciting area as it is challenging and is very active, demanding innovations across technical and analysis areas • Also an area where theory/experiment crossover can be influential • Reinterpretations of existing measurements can be powerful, especially for low lifetime regions

16 S. Farrington, University of Edinburgh Triggers (excerpt from ATLAS)

Sum <=1kHZ 17 ATL-DAQ-PUB-2019-001 S. Farrington, University of Edinburgh Di-lepton Displaced Vertex1905.09787

• Search can be broadly motivated • Targets long lived Heavy Neutral Lepton

which has small mixing with the νµ • Uses large radius tracking algorithm • This cannot run on all data, instead filters are deployed, in this case 1 prompt + 1 displaced muon • Uses specialised displaced vertex reconstruction algorithm, 20% efficient • Signal region (eff~1-2%): • Tightly identified muon • Tight electron or additional muon • Displaced vertex • 2 opposite sign tracks • Mass > 4 GeV, 4

• ATLAS Large Radius Tracking algorithm increases upper d0 cut from 1cm to 30cm • Was a substantial software task, built on top of standard algorithms • Reconstructs about 95% of the “reconstructable” tracks from LLPs but is costly to run and to store, uses leftover hits

Primary vertex

• InDet • TRT • InDet d • SpacePoint 0 • SiSPSeeded • Ambiguity • Extension • Extensions Formation TrackFinder • Solver • Alg • Processor

19 S. Farrington, University of Edinburgh Dilepton

• Backgrounds • material interactions and metastable states (b/s hadrons) studied in control regions, negligible for vertex mass > 4 GeV • Random track crossings: modelled with ABCD method • Variables used: number of leptons in DV versus number of same sign/ opposite sign displaced vertex tracks

20 S. Farrington, University of Edinburgh Dilepton

21 S. Farrington, University of Edinburgh Emerging Jets 1810.10069 • Targeted at a heavy mediator between the SM and the “hidden sector” with a QCD-like confinement (“dark-QCD”) • The “dark” quark showers/ hadronises in the hidden sector and then slowly decays into a SM quark and a dark quark • Several displaced vertices plus missing energy due to dark quarks • Search • Trigger on sum of calo activity (900 GeV) • Use large impact parameter significance • Define 7 signal regions

22 S. Farrington, University of Edinburgh Emerging Jets • Backgrounds: b-jets or track mis-measurement • Measure mis-tag rates in gamma+jet samples • Define control regions to go with each signal region • Determine heavy flavour fraction by fitting to b-tagged samples

M(χDK) 23 S. Farrington, University of Edinburgh Displaced Jets (with tracks) • Signature based motivation, measured quantity is model independent cross section limit, but can be interpreted as search for R parity violating (RPV) stop quark decaying to quark+muon • Signature: displaced vertex within the tracking detector plus displaced muon • Trigger on normal muon trigger and missing transverse energy

ATL-CONF-2019-006

24 S. Farrington, University of Edinburgh Displaced Jets (with tracks) • Large radius tracking needed here as well, and displaced vertex reconstruction • Filter: muon (pt>60 GeV) or MET>180 GeV

• Rates of cosmic rays, fake muons, muons from heavy flavour are measured in control regions and checked in validation regions

25 S. Farrington, University of Edinburgh Displaced Jets (with tracks)

26 S. Farrington, University of Edinburgh Displaced Jets (Low EM fraction) • To be sensitive to longer lifetimes / higher boosts, look at decays that happen outside the EM calorimeter • Search motivated by long-lived scalars • LLP decays in the hadronic calorimeter • No associated tracks • Specialised trigger (shape and location of calo deposits) • Backgrounds • Jets of neutral hadrons • Beam-induced backgrounds (interactions of LHC beam gas, and beam-halo interactions with collimators upstream of ATLAS, results in muons travelling parallel to the beam pipe) • These muons then radiate photons (bremsstrahlung) in the hadronic calorimeter 27 1902.03094 S. Farrington, University of Edinburgh Displaced Jets (Low EM Fraction) • Multi layer perceptron trained to estimate LLP decay position

• Per-jet BDT (classify among QCD/signal, beam induced background) • Per-event BDT (eliminate beam induced background) • ABCD method to calculate remaining QCD background

28 S. Farrington, University of Edinburgh Displaced Jets (Low EM Fraction)

Scalar mass 125 GeV Scalar mass 600 GeV

29 S. Farrington, University of Edinburgh Displaced Jet (timing) • Use calorimeter timing to detect jets that arrive significantly after the bunch crossing, could be due to a heavy LLP for example • Backgrounds • Electromagnetic calorimeter timing resolution, electronic noise • Pile-up, satellite bunches, beam halo, cosmic muons

1906.06441

30 S. Farrington, University of Edinburgh Displaced jet (timing) • ABCD method used to predict backgrounds in three categories

cτ0(mm)

31 S. Farrington, University of Edinburgh Non-pointing Photon 1909.06166 • Neutralino decay to photon+ gravitino

• Delayed signal in calorimeter following bunch crossing • Out-of-time photon reconstruction is required • Looks for a so-called elliptical shower – shape of deposit in the calorimeter is altered with respect to normal cases • Uses both normal and dedicated triggers with modified elliptical showers • Analysis requires >2 jets together with 2 displaced photons or 1

displaced photon with HT

32 S. Farrington, University of Edinburgh Non-pointing Photon • Backgrounds evaluated with ABCD method • The object (jet/photon) requirements largely eliminate backgrounds from beam effects etc

33 S. Farrington, University of Edinburgh Heavy Stable Charged Particle • Motivated by squark, gluino R-hadron, chargino, stau • Heavy implies slow-moving • Observed events slightly higher than predicted in some categories in CMS analysis

1902.01636 34 S. Farrington, University of Edinburgh Heavy Stable Charged Particle • Uses standard triggers: muon and MET (calo-based) • The pixel detector can provide dE/dx measurements which indicates βγ • Calibrate with low-momentum hadrons • Results in resolution on βγ of 14% • Then use Hadronic calorimeter and muon system to calculate time of flight and thus β • Calculate candidate mass using dE/dx and time of flight (m=p/βγ)

35 S. Farrington, University of Edinburgh Heavy Stable Charged Particle • Background estimation • Probability distribution functions determined from data sidebands

for momentum, βTOF and βγdE/dx

• Sample from these to predict the mTOF shape • Normalise to data in the low mass control region

36 S. Farrington, University of Edinburgh Current Coverage for LLP to tau

• Current efforts leave huge unexplored phase space (plot a couple of years old but the picture has changed little for taus) (cm) τ c J. Evans & J. Shelton JHEP 1604 (2016) 056 Lifetime, Mass(LLP) (GeV)

• So I spend my time on figuring out how we probe the rest of the phase space for taus (new hadronic tau ID, new triggers) 37 S. Farrington, University of Edinburgh Tau ID e/μ

+ Tau hadronic decays τ υe/μ LLP • BR(τh = 65%) υτ τ− h+ Tau ID h− h+ (K/π) • Tau ID method (Boosted Decision Tree) • Currently trained on Z/H decaying promptly

38 S. Farrington, University of Edinburgh Tau Trigger

υτ To maximise the sensitivity to τ+ LLP h+ LLP decays to third generation − τ υτ need to use fully hadronic decay modes h+ h− h+ (K/π)

• Trigger currently allows only up to |d0| < 2 (then 4 later in run 2) mm • Train for long-lived taus

Combine new τh trigger and ID

39 S. Farrington, University of Edinburgh Workshop on LLP to 3rd generation https://indico.ph.ed.ac.uk/event/59/timetable/#20191120

40 S. Farrington, University of Edinburgh Issues with complex variables & algorithms

• Can they be referred back to a model independent particle- level object? e.g. • Efficiency = (truth pass) AND (reco pass) ------(truth pass) - How do we (can we) evaluate “truth pass”? - How model independently can we do that? - If your reco selection intimately involves details of the detector, that’s fine so long as what you measure (truth) can be defined independently of this. - If your reco selection intimately involves details of the model its reinterpretability will obviously be limited, but still possible if a particle-level object can be defined. - If the there is no detector-independent truth level object defined, you mingle model dependence and detector dependence; only way to “reinterpret” is a full reanalysis model-by-model (e.g. RECAST?). Slide from Jon Butterworth

S. Farrington, University of Edinburgh Factorising effiencies?

• Event selection efficiency • Can be independent of LLP decay • Will be dependent on LLP production model

• Tau efficiency (vs d0, pT etc) • Can be independent of LLP production model (within some fiducial kinematic region) • Final-state particle efficiency (vs d0, pT etc) • Won’t tell you about event efficiency • Can be used to build efficiencies for many models; ideal for reinterpretation?

• arXiv: 1903.04497 Slide from Jon Butterworth Nov S. Farrington, University of Edinburgh 2019 Reinterpretations (one recast example)

43 S. Farrington, University of Edinburgh Simplified Models

1903.04497

44 S. Farrington, University of Edinburgh What can e+e- colliders add?

45 S. Farrington, University of Edinburgh e+e-

• Cannot probe higher on the mass axis, but e+e- target Higgs production, so focus LLP searches on those that couple to H(125) • Exellent vertex resolution è sensitive to the short ctau range (cτ~10-13s) that can be tricky for LHC experiments

What lies here? (p

…e-)

46 S. Farrington, University of Edinburgh Where can an e+e- collider do better?

• Firstly, where they cannot do better: • High mass, hence the focus on rare decays of the 125 GeV Higgs • Rates - the number of H(125) produced at HL-LHC will be larger than at e+e-(250) by two orders of magnitude • Rates – not just related to the Higgs, but in general, rare processes will benefit from the high luminosity of hadron collisions • Better: • Shorter lifetime LLPs (vertex resolution) • Lighter LLPs (no trigger thresholds, improved mass resolution) • Collimated decay products • Model independence (no need for triggering on associated objects)

47 S. Farrington, University of Edinburgh Where can an e+e- collider do better?

48 S. Farrington, University of Edinburgh e.g. CepC Design: d0 resolution CepC CDR arXiv:1811.10545 • Impact parameter resolution can reach 3µm for momentum above 30 GeV • For this reason, monolithic pixel sensors are considered, to achieve low material, high impact parameter resolution

GeV GeV 10 100

3µm

• Access low LLP lifetimes (<10-13 s), resolve LLP from prompt particles • LLP-objects are not listed as a design requirement • But studies of them could influence hardware choices 49 S. Farrington, University of Edinburgh e+e- Detector Environment

• No underlying event or pile-up

• Di-jet events at LEP/LHC • Tracking algorithms at e+e- can be more ambitious with their maximum impact parameter cuts, removing the need for bespoke tracking algorithms • Either fork the software and add LLP branch OR include as default 50 S. Farrington, University of Edinburgh Case Study: Higgs to displaced jets Phys. Rev. D 92 (2015) 012010

• Require two displaced vertices, either in inner detector or muon system

51 S. Farrington, University of Edinburgh Higgs to displaced jets at e+e- Alipour-Fard et al. arXiv:1812.05 • e+e- study • Assumes 3µm vertex resolution • Requires ZH, H to jj • The paper suggests further work: try following a CMS study on

distribution of track d0 in clusters to help sensitivity to low ctau

“long lifetime analysis”

“large mass analysis”

52 S. Farrington, University of Edinburgh Summary

• Substantial phase space still to cover, and the LLP community are supporting the use of simplified models to help ensure coverage across the phase space • Models can be an inspiration but ultimately the phase space we explore, and the limits set, must be made to some extent irrespective of the theories of the day • LHCb has sensitivities that are being exploited (previous talk) • Should plan for success and understand how we would characterise discoveries in this domain • Flavour universal couplings? CP? Spin? Multiple states?

53 S. Farrington, University of Edinburgh Extras

54 S. Farrington, University of Edinburgh Theoretical Motivations

• P.W. Graham, D. E. Kaplan, S. Rajendran, and P. Saraswat, “Displaced supersymmetry”, JHEP 07 (2012) 149, doi:10.1007/JHEP07(2012)149, arXiv:1204.6038. • R. Barbier et al., “R-parity violating supersymmetry”, Phys. Rept. 420 (2005) 1, doi:10.1016/j.physrep.2005.08.006, arXiv:hep-ph/0406039. • J. L. Hewett, B. Lillie, M. Masip, and T. G. Rizzo, “Signatures of long-lived gluinos in split supersymmetry”, JHEP 09 (2004) 070, doi:10.1088/1126-6708/2004/09/070,arXiv:hep-ph/ 0408248. • M. Fairbairn et al., “Stable massive particles at colliders”, Phys. Rept. 438 (2007) 1,doi: 10.1016/j.physrep.2006.10.002, arXiv:hep-ph/0611040. • G. F. Giudice and R. Rattazzi, “Theories with gauge mediated supersymmetry breaking”, Phys. Rept. 322 (1999) 419, doi:10.1016/S0370-1573(99)00042-3,arXiv:hep-ph/9801271. • T. Han, Z. Si, K. M. Zurek, and M. J. Strassler, “Phenomenology of hidden valleys at hadron colliders”, JHEP 07 (2008) 008, doi:10.1088/1126-6708/2008/07/008,arXiv:0712.2041. • L. Basso, A. Belyaev, S. Moretti, and C. H. Shepherd-Themistocleous, “Phenomenology of the minimal B-L extension of the Standard model: Z’ and neutrinos”, Phys. Rev. D 80(2009) 055030, doi:10.1103/PhysRevD.80.055030, arXiv:0812.4313. • M. J. Strassler and K. M. Zurek, “Discovering the Higgs through highly-displaced vertices”, Phys. Lett. B 661 (2008) 263, doi:10.1016/j.physletb.2008.02.008, arXiv:hep-ph/ 0605193.

55 S. Farrington, University of Edinburgh This talk

• Many plausible models predict LLPs • I will spend little time on theory scenarios • Will give an experimentalist’s view on which types of signature CepC may be able to compete or lead in. • My LHC examples are focused on ATLAS but CMS has a similar program.

56 S. Farrington, University of Edinburgh