Long-Lived Particles at the LHC Zhen Liu (Fermilab) LPC Topic Of The Week (TOTW) seminar Jun. 19th , 2018 neutral BSM charged CMS Public Results lepton any charge quark displaced HSCP photon dilepton anything signature-basedprogram! disappearing displaced track lepton
displaced displaced dijet photon
displaced displaced Not pictured: vertex conversion stopped particles 2 J. Antonelli EXO Higgs meeting, Nov 7 3 General LLP Map
3 6/19/18 Zhen Liu LLP @ LHC LPC TOTW MATHUSLA physics case, D. Curtin et al, appearing tonight Outline
ZL (chapter editor) et al, Simplified Models • LLP Theory (chapter 2 of LLP community report; appearing soon) • LLP Coverage ZL, B. Tweedie, 1503.05923 • Timing for LLP Jia Liu, ZL, L.-T. Wang, 1805.05957
4 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Outline
ZL (chapter editor) et al, Simplified Models • LLP Theory (chapter 2 of LLP community report; appearing soon) • LLP Coverage ZL, B. Tweedie, 1503.05923 • Timing for LLP Jia Liu, ZL, L.-T. Wang, 1805.05957
5 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Easily long-lived: SUSY • RPV–small B/L-violating couplings
• Gauge mediation—suppressed couplings via SUSY breaking scale
• Mini-split spectrum—suppressed couplings through ”decoupled” heavy particles
• Pure Wino/Higgsino–nearly degenerated, disappearing track
6 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Easily long-lived: hidden sector Hidden sector feeble couplings to SM via various portals, suppressed by the smallness of the couplings (e.g., Strassler, Zurek, et al) Fig. credit: B. Shuve
Can be related to big questions: • Dark matter; • Neutrino mass; • Baryogenesis; Taking the neutral naturalness example: …etc. Fig. credit: B. Shuve
7 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Classification: Production
Simplified Models (chapter 2 of LLP community report)
• Factorize production and decay; • Production affects kinematics of LLP and trigger consideration (except for LLP triggers, which are rare currently); • Decay affects search strategy in picking up the LLPs, convoluting with lab frame geometries;
8 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Classification: Production
• Factorize production and decay; • Production affects kinematics of LLP and trigger consideration (except for LLP triggers, which are rare currently); • Decay affects search strategy in picking up the LLPs, convoluting with lab frame geometries; resonant non-resonant
9 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Classification: Production, Decay and Models
Neutral Long-lived particles LLP decay modes
Canonical production Mapping to UV Models modes: DPP, HP, HIG, RES, CC X represents the LLP *model definitely include missing energy; +signature not appeared in the minimal/simplest model setup;
10 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Classification: Production, Decay and Models
Charged Long-lived particles Colored Long-lived particles LLP decay modes
LLP decay modes
Canonical Mapping to UV Models Canonical Mapping to UV Models production modes production modes
and there are many more exotic signals: Stopped particles, dashed tracks, coplanar tracks, fireballs, dark showers, etc;
A LLP model file library also under construction, many modes already tested
11 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Outline
ZL (chapter editor) et al, Simplified Models • LLP Theory (chapter 2 of LLP community report; appearing soon) • LLP Coverage ZL, B. Tweedie, 1503.05923 • Timing for LLP Jia Liu, ZL, L.-T. Wang, 1805.05957
12 6/19/18 Zhen Liu LLP @ LHC LPC TOTW neutral BSM charged CMS Public Results lepton any charge quark displaced HSCP photon dilepton anything signature-basedprogram! disappearing displaced track lepton
displaced displaced dijet photon
displaced displaced Not pictured: vertex conversion stopped particles 13 J. Antonelli EXO Higgs meeting, Nov 7 3 Enlarging the coverage: CMS Dijet
14 Fig. credit: E. Kuflik Enlarging the coverage: CMS Dijet
• LLP searches usually are also sensitive to other decay topologies without/with little efficiency loss. • We emphasis this point and show the power to all SUSY LLPs in our study.
15 Fig. credit: E. Kuflik Overview of our study
Hadronic R-parity violation
gauge mediation
mini-split
Our selection of signals (naïve naturalness driven, light Higgsino, stop and gluino) covers a large range of displace decay topologies, including 1j+MET, 2j+MET, 3j+MET, 2j, 3j, as well as heavy flavors, making it easy for theorists to estimate exclusions for their own models in concern.
16 Overview of our study
Applied to all models Hadronic • CMS displaced dijets (tracker) R-parity • ATLAS low-EM jets (HCAL) violation • ATLAS muon spectrometer vertices* • CMS charged stable particles gauge Applied to models with mediation leptonic decays
• CMS displaced dileptons • CMS displaced electron & muon mini-split • ATLAS displaced muon + tracks Our selection of signals (naïve naturalness driven, light Higgsino, stop and gluino) covers a large range of displace decay topologies, including 1j+MET, 2j+MET, 3j+MET, 2j, 3j, as well as heavy flavors, making it easy for theorists to estimate exclusions for their own models in concern.
-1 17 * 7 TeV, 2 fb ** All via direct pair-production Overview of our study
Applied to all models Hadronic • CMS displaced dijets (tracker) R-parity • ATLAS low-EM jets (HCAL) violation • ATLAS muon spectrometer vertices* • CMS charged stable particles gauge Applied to models with mediation leptonic decays
• CMS displaced dileptons • CMS displaced electron & muon mini-split • ATLAS displaced muon + tracks Our selection of signals (naïve naturalness driven, light Many results, great boost to LLP Higgsino, stop and gluino) covers a large range of displace and SUSY LLP; to save time, I decay topologies, including 1j+MET, 2j+MET, 3j+MET, 2j, 3j, as well as heavy flavors, making it easy for theorists to will only go through two results estimate exclusions for their own models in concern.
-1 18 * 7 TeV, 2 fb ** All via direct pair-production A Typical Efficiency MAP
With detailed simulation and our own modeling of the displacement, after carefully calibrating with existing searches, we can derive the limits from many search of our simplified models.
Efficiency map for RPV stop decays into light jet pairs in the CMS displaced dijet analysis. • Lines at increase of 100 GeV • Low mass suffers more for cuts on jet energy • High mass approaches constant efficiency shape • Low efficiency at low lifetime (cut to remove SM) • (Shift in peak due to Lorentz Factor)
19 Gauge Mediation SUSY Breaking (top squark)
GMSB Stop ➞ Top (*) + Gravitino • Displaced searches (dijet, !+tracks, With detailed simulation " + !, HCAL, ! and our own modeling of spectrometer) the displacement, after covers mid-lifetime Heavy charges carefully calibrating with • stable particle existing searches, we can searches (pink; derive the limits from many lifetime CHAMP/HSCP) FULL coverage up to covers long lifetime search of our simplified 500 GeV! • Prompt (gray) models. covers short lifetime
Dijet search has very good sensitivity reach, lepton plus tracks searches also sensitive to leptonic top- and b-decays. HCAL and muon spectrometer searches sensitive to higher lifetimes but so far suffers large efficiency cost. Optimization may provide additional information, e.g., heavy neutral displaced particles.
20 Gauge Mediation SUSY Breaking (Higgsino)
GMSB Higgsino ➞ Higgs + Gravitino • Displaced searches 8 TeV Result (dijet, !+tracks, " + !, HCAL, dilepton, ! spectrometer) covers mid-lifetime Heavy charges • No stable particle searches (pink; CHAMP/HSCP) to cover long lifetime—as there is no charged LLP; CMS 13 TeV prompt search, 1709.04896, 1801.03957 • No prompt searches to covers short lifetime due to large background; Dijet search has very good sensitivity reach, lepton plus tracks searches also sensitive to leptonic b-decays. HCAL and muon spectrometer searches sensitive to higher lifetimes but so far suffers large efficiency cost.
21 Outline
ZL (chapter editor) et al, Simplified Models • LLP Theory (chapter 2 of LLP community report; appearing soon) • LLP Coverage ZL, B. Tweedie, 1503.05923 • Timing for LLP Jia Liu, ZL, L.-T. Wang, 1805.05957
22 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Realizing the great potential of the LHC What’s the best place to >XX) look for LLPs (short- - lifetime-limit, and long lifetime-limit)? scale in reach in model in model in scale reach
Log Log scale in proper lifetime parameters (e.g.,parameters Br H
23 6/19/18 Zhen Liu LLP @ LHC LPC TOTW *except for forward physics, e.g., LHCb, FASER Realizing the great potential of the LHC
Line with increasing signal probability due to long What’s the best place to lifetime >XX) look for LLPs (short- - lifetime-limit, and long lifetime-limit)?
Line with decreasing signal probability due to long lifetime scale in reach in model in model in scale reach
Log Log scale in proper lifetime parameters (e.g.,parameters Br H
24 6/19/18 Zhen Liu LLP @ LHC LPC TOTW *except for forward physics, e.g., LHCb, FASER Realizing the great potential of the LHC
Line with increasing signal probability due to long What’s the best place to lifetime >XX) look for LLPs (short- - lifetime-limit, and long lifetime-limit)?
Line with decreasing signal probability due to long lifetime scale in reach in model in model in scale reach
Log Log scale in proper lifetime L2 (e.g.,parameters Br H 1 1 L/d P = d⌦ dL e in 4⇡ d Z ⌦ ZL1 L2 '2-34/5 ⌦ 1 L/d dL e !"#$ = &'()*×,#-×./(#$×."#$×. ⇡ 4⇡ d 01$ ZL1
⌦ L /d L /d d
25 6/19/18 Zhen Liu LLP @ LHC LPC TOTW *except for forward physics, e.g., LHCb, FASER Precision timing--a new dimension Precision timing information now compliments spatial information, and its bears great potential to fully realize LHC’s physics reach in LLP. For long-lived particles (whose lifetime is macroscopic >~ mm), they generically move slower and their long-lived nature substantiates their slowness in motion at colliders. ATLAS has similar endcap/forward timing proposal
• 30 picosecond timing resolution at CMS after phase2 upgrade (in front of ECal, 1.2 m from beam); • Proposed to enable 4d construction of vertices: • reducing the pile up level; • Reducing pile-up track mis- association in to the primary interaction;
26 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Timing BSM
signal arrival time
SM reference particles arrival time
27 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Timing BSM
signal arrival time
SM reference particles arrival time
For CMS timing layer (1.2 m, t0=4 nanoseconds)*, 30 picosecond timing resolution indicates sensitivity to BSM signal having >1% velocity (boost factor ! < 7) /path difference w.r.t. SM particles!
LLP (with mass > 10s of GeV) typically all have much slower motion!
*for pseudorapidity 0; higher rapidity enlarges the timing difference; *SM particles essentially all travel at speed of light;
28 6/19/18 Zhen Liu LLP @ LHC LPC TOTW LLPs arrive (very) late ctau =10 m [Lt1,Lt2]=[0.2,1.2]m
We also consider a possible timing layer outside Muon spectrometer, making use of the large LHC detector volume. Signals: • Red: Higgs decaying into glueballs (neutral naturalness) • Blue: Higgsinos (GMSB SUSY)
29 6/19/18 Zhen Liu LLP @ LHC LPC TOTW LLPs arrive (very) late ctau =10 m [Lt1,Lt2]=[0.2,1.2]m
We also consider a possible timing layer outside Muon spectrometer, making use of the large LHC detector volume. Signals: • Red: Higgs decaying into glueballs (neutral naturalness) • Blue: Higgsinos (GMSB SUSY) Backgrounds: • Gray (Dashed): Pile-up with natural spread of 190 ps (beam property) • Gray (Solid): Hard collision spread due to uncertainties in timing 30 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 2 lay since the mass of the new particle can be compara- searches in this letter.3 ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- signal search strategy of using the timing information, onds (ns), for entering EC, to tens of ns, for exiting the and more importantly, the corresponding consideration MS. As a result, with tens of picosecond (ps) timing for the background. A typical signal event of LLP is resolution, we have a sensitivity to percent level time delay caused by slow LLP motion, e.g., 1 > 0.01 X a with boost factor < 7. In Fig. 2, we show typical time b delay t for a hypothetical timing layer at the outer Timing layer part of the ATLAS MS system for benchmark signals `a and the background, and the distributions for EC are `SM put in appendix. The two benchmark signals considered here are the glueballs from Higgs boson decays, and the LT2 electroweakino pair production in the Gauge Mediated `X SUSY Breaking (GMSB) scenario. Both the glueballs LT1 SM and lightest neutralino proper lifetimes are set to be c⌧ = 10 m. The 10 GeV glueballs (red dashed line) have X larger average boost comparing to the 50 GeV glueballs (solid red line), and hence have a sizable fraction of the signals with delay time less than one nanosecond. For FIG. 1. An event topology with an LLP X decaying to two the electroweakinos pair production, the signals are not light SM particles a and b. A timing layer, at a transverse boosted and hence significantly delayed compared to the distance L away from the beam axis (horizontal gray dotted T2 backgrounds, with 99% of the signal with t>1 ns. line), is placed at the end of the detector volume (shaded region). The trajectory of a potential SM background particle is also shown (blue dashed line). The gray polygon indicates Search strategy.— We consider the signal with an ISR the primary vertex. jet timestamping the primary vertex and another SM ob- ject from the LLP decay (e.g., jet for this study) which shown in Fig. 1. An LLP, denoted as X,travelsadis- has large time delay t.TostudythesensitivitytoBSM signals with timing, we propose two searches using such tance `X into a detector volume and decays into two light SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- ing layer located at the beginning of EC, and one with a transverse distance LT2 away from the beam axis. In a typical hard collision, the SM particles generally travel ATLASLate geometry comers for awill precision be spotted timing layer easily: located at close to the speed of light. The trajectories of charged SM the end of MS. They are tabulated as following: particles can be curved, which increase the path length j LT2 LT1 Trigger ✏trig ✏sig ✏fake Ref. in comparison with neutral SM particles. For simplicity, 3 EC 1.17 m 0.2 m DelayJet 0.5 0.5 10 [12] CMS timing module we only consider neutral LLP signals where background 9 MS 10.6 m 4.2 m MS RoI 0.25, 0.5 0.25 5 10 [24] ATLAS MS LLP search from such charged particles can be vetoed using particle ⇥ (without timing) 2 identification and isolation. Hence, the decay products ForDesigned both searches, 2 generic we assume search: similar performance of tim- of X, taking particle a for example, arrives at the timing ingno resolutionrestriction of on 30 the ps. signal, For the as MS search, because of layer with a time delay of thelong larger as they time can delay deposit and much less background due to “shielding” by inner detectors compared to the EC case, ` ` ` energy (30 GeV pT min)* t = X + a SM , (1) a less precise timing (e.g. 150 ps) could also achieve j X a SM similar physics reach. The ✏trig, ✏sig and ✏fake are the e - cienciesMultijet forand trigger, pile- signalup selection and a QCDOther jet backgrounds: faking with a SM 1. It is necessary to have prompt ' ' thebackground delayed jet can signal be witheffectivelypT > 30 GeV in• ECInteraction or MS, with material decay products or Initial State Radiation (ISR) which respectively. arriving at timing layer with the speed of light to derive rejected use timing* • Cosmic rays For the EC search, we assume a new trigger• strategy the time of the hard collision at the primary vertex (to Beam halo of a delayed jet using the CMS upgrade… timing layer. “timestamp” the hard collision). ISR jets could easily be This can be realized by comparing the prompt jet with present for all processes, and we use this generic feature All have mature veto mechanism; need to pT > 30 GeV that reconstructs the four-dimensionalrevisit to see pri- the impact of timing to “timestamp” the hard collision for the proposed new mary vertex (PV4d) with the arrival time of another jet
31 6/19/18 Zhen Liu LLP @ LHC LPC TOTW
2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles, the ma- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles in a jet still travel with nearly searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. 3 delay at MS from LHC gion of Interest (MS RoI) trigger for very similar search 100 from ATLAS [27] as reference, with an e ciency of ✏trig =0.25 and 0.5 for the two benchmark BSM signals, -1 ) 10 and a signal selection e ciency of ✏sig =0.25. The back- bin
/ grounds are mainly from the punch-through jets, and its 2 j,MS 9
t fake e ciency can be inferred to be ✏ =5.2 10 , 10-2 fake
Δ ⇥ 3 normalized to 1300 fake MS barrel events at 8 TeV [27]. lay since the mass of the new/ particle can be compara- searches in this letter. Our Reference ATLAS MS displaced vertex search [24], ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- 10-3 signal search strategy of using/ the timing information, onds (ns), for entering EC, todue tens to of the ns, for vertex exiting reconstruction the requirement, can only 1 and more importantly, the corresponding consideration MS. As a result, with tens ofe↵ectively picosecond select (ps) signal timing events decaying in the 4-7 m for the background. A typical signal event of LLP is resolution, we have a sensitivityrange, to reducingpercent level the time derived search sensitivity with the 10-4 delay caused by slow LLP motion,full MS e.g., volume 1 X approximately> 0.01 by a factor of two. We 0. 0.5 1. 2 5 10 20 50 100 200 a b with boost factor < 7. In Fig.expect2, we that show with typical the time help of the timing layer and a Δt (nsdelay) t for a hypotheticalrelaxed timing layer vertex at reconstruction the outer requirement, the e↵ective Timing layer part of the ATLAS MS system for benchmark signals `a decay range could be extended to the full MS while FIG. 2. The di↵erential t distributionand for the typical background, signals andand themaintaining distributions the for same EC are signal e ciency. In comparison backgrounds at 13 TeV LHC. The plot is normalized to the `SM put in appendix. The two benchmarkwith LLP signals decay considered in the 7-10 m range of the MS, there is fraction of events per bin with a varying bin size, where for here are the glueballs from Higgsno detector boson decays, activities and the in the layers prior to that. Hence, LT2 t less than 1 ns are shown in linearelectroweakino scale and then pair in production loga- in the Gauge Mediated rithmic scale otherwise. Two representative signal models are the dominant background from punch can still be vetoed `X SUSY Breaking (GMSB) scenario. Both the glueballs shown, the delay time for the glueballs from the Higgs decay e↵ectively. LT1 SM and lightest neutralino proper lifetimes are set to be (red curves) and the GMSB neutralinosc⌧ = 10 from m. Drell-Yan The 10 GeV pair glueballs (red dashed line) have production (blue curves), with a light and a heavy benchmark X larger average boost comparingBackground to the 50 GeV consideration. glueballs — The main sources of the mass shown in dashed and solid curves, respectively. For all (solid red line), and hence haveSM a background sizable fraction faking of the such delayed and displaced signal signal events, the proper lifetime is set 10 m, and the dis- signals with delay time less thanare from one nanosecond. jets or similar For hadronic activities. The origin of tribution only counts for events decayed within [LT1 ,LT2 ] FIG. 1. An event topologyof with [4.2, an 10.6] LLP mX indecaying the transverse to two direction,the electroweakinos which follows pair the production,background the signals can be are classified not into same-vertex (SV) hard light SM particles a and b. A timing layer, at a transverse geometry of ATLAS MS in the barrelboosted region. and For hence the significantly back- collision delayed and compared pile-up to (PU). the For this study, we assume the distance L away from the beam axis (horizontal gray dotted T2 backgrounds, with 99% of the signal with t>1 ns. 4 line), is placed at the endground of the detector distribution volume shown (shaded in gray curves, we assume bunch time-spread distributions follow Gaussian distribution. spacing of 25 ns. The solid and dashed gray curves represent region). The trajectory of a potential SM background particle t2 backgrounds from a same hard collisionSearch vertex strategy. and hence— Wewith consider the signal withd an( ISRt) 1 2 is also shown (blue dashed line). The gray polygon indicates P = E 2 t , (2) PTjet timestamping the primary vertex and another SM ob- the primary vertex. a precision timing uncertainty of t = 30 ps and from the d t p2 t pile-up with a spread of t = 190 ps,ject respectively, from the LLP in units decay of (e.g., jet for this study) which fraction per 0.1 ns. The corresponding distribution for EC where the time spreads di↵er for di↵erent sources of shown in Fig. 1. An LLP, denoted as X,travelsadis- has large time delay t.TostudythesensitivitytoBSM t can be obtained approximately by scaling the horizontal axis backgrounds. tance ` into a detector volume and decays into two light signals with timing, we propose two searches using such X according to the ratio of size of the detector volume. The SV background for the signals mainly comes from SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- ing layer located at the beginningSM multi-jet of EC, and process. one with At least one prompt jet is required a transverse distance LT2 away from the beam axis. In a typical hard collision, the SM particles generally travel ATLAS geometry for a precisionto reconstruct timing layer PV4dlocated and at provide the timestamp, while at the timing layer. The delayedLate and displaced comers jet sig-will be spotted easily: close to the speed of light. The trajectories of charged SM the end of MS. They are tabulatedanother as following: trackless jet from the same hard collision faking nal, after requiring minimal decay transverse distance of particles can be curved, which increase the path length long-lived signals.j The fake jet has an intrinsic time delay LT LT Trigger ✏trig ✏sig ✏ Ref. 0.2 m (LT ), will not have good tracks associated2 1 with it. fake in comparison with neutral SM particles.1 For simplicity, t = 0. However,3 dueCMS to limited timing timing module resolution in Consequently, the major SM backgroundEC 1.17 is m from0.2 trackless m DelayJet 0.5 0.5 10 [12] we only consider neutral LLP signals where background reconstructing the9 PV4d, it could have a time spread. j,EC MS 310.6 m 4.2 m MS RoI 0.25, 0.5 0.25 5 10 [24] ATLAS MS LLP search from such charged particlesjets. can The be vetoed jet fake using rate particle of ✏fake = 10 is calculated us- The time resolution⇥ with(without the timing) planned precision timing identification and isolation.ing2 PythiaHence, the[25] decay by simulating products the trackless jets, where all PT For both searches, we assume similarupgrade performanceSame from-vertex CMS of tim- ishard t scattering= 30 ps. At 13 TeV with of X, taking particle a forcharged example, constituent arrives at the hadrons timing are too soft to be observed 1 ing resolution of 30 ps. For theintegrated MS search,background, luminosity because time ofint =spread 3 ab 30, the ps total number of layer with a time delay ofor missed due to trakcing ine ciencythe larger. The time trackless delay and jet muchsuch less background background events dueL to can be estimated, fraction is measured in the validation“shielding” data by for inner the detectors low- compared(precision to the EC timing) case, `X `a `SM SV EC j,EC 11 t = electromagnetism+ , jet search(1) at thea less ALTAS precise [26], timing and it (e.g. is 150 ps) couldEC : alsoNbkg achieve= j int✏trig✏fake 1 10 j X a SM 2 L ⇡ ⇥ found to be 10 . They also foundsimilar a huge physics additional reach. sup- The ✏trig, ✏sig and ✏fake areSV the e - MS j,MS 5 MS : N = j int✏ ✏ 4 10 , (3) pression through the energy depositionciencies ratio for trigger, between signal EC selection and a QCD jetbkg fakingL trig fake ⇡ ⇥ with a SM 1. It is necessary to have prompt ' ' the delayed jet signal with pT > 30 GeV in EC8 or MS, decay products or Initialand State hadronic Radiation calorimeter. (ISR) whichMoreover, due to the actual where j 1 10 pb is the multi-jets cross-section with decays within the tracking volumn,respectively. the signal contains j ' ⇥ j arriving at timing layer with the speed of light to derive pT > 30 GeV, ✏trig and ✏fake are the e ciencies for the low quality tracks in constrast to theFor truly the EC neutral search, jets, we assume a new trigger strategy the time of the hard collision at the primary vertex (to of a delayed jet using the CMS upgrade timing layer. and the energy deposition in the EC for the singal will be “timestamp” the hard collision). ISR jets could easily be This can be realized by comparing the prompt jet with present for all processes, andmore we than use this neutraljets generic feature as well, we hence consider our jet pT > 30 GeV that reconstructs4 the four-dimensional pri- to “timestamp” the hardfake collision rate for assignment the proposed of 10 new3 to be reasonabe reasonably The validity of these description should be scrutinized by ex- mary vertex (PV4d) with the arrivalperimental time of measurement, another jet e.g. from Zero-Bias events. From conservative. 32 6/19/18 Zhen Liu LLP @ LHC Refs. [ 12LPC, TOTW15, 19], our description is appropriate up to proba- 4 6 For the MS search, we consider a new timing layer at bility of 10 to 10 level–the limit where the distribution are the outer layer of the MS of ATLAS. We take the MS Re- shown. 2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles, the ma- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles in a jet still travel with nearly searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. 3 delay at MS from LHC gion of Interest (MS RoI) trigger for very similar search 100 from ATLAS [27] as reference, with an e ciency of ✏trig =0.25 and 0.5 for the two benchmark BSM signals, -1 ) 10 and a signal selection e ciency of ✏sig =0.25. The back- bin
/ grounds are mainly from the punch-through jets, and its 2 j,MS 9
t fake e ciency can be inferred to be ✏ =5.2 10 , 10-2 fake
Δ ⇥ 3 normalized to 1300 fake MS barrel events at 8 TeV [27]. lay since the mass of the new/ particle can be compara- searches in this letter. Our Reference ATLAS MS displaced vertex search [24], ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- 10-3 signal search strategy of using/ the timing information, onds (ns), for entering EC, todue tens to of the ns, for vertex exiting reconstruction the requirement, can only 1 and more importantly, the corresponding consideration MS. As a result, with tens ofe↵ectively picosecond select (ps) signal timing events decaying in the 4-7 m for the background. A typical signal event of LLP is resolution, we have a sensitivityrange, to reducingpercent level the time derived search sensitivity with the 10-4 delay caused by slow LLP motion,full MS e.g., volume 1 X approximately> 0.01 by a factor of two. We 0. 0.5 1. 2 5 10 20 50 100 200 a b with boost factor < 7. In Fig.expect2, we that show with typical the time help of the timing layer and a Δt (nsdelay) t for a hypotheticalrelaxed timing layer vertex at reconstruction the outer requirement, the e↵ective Timing layer part of the ATLAS MS system for benchmark signals `a decay range could be extended to the full MS while FIG. 2. The di↵erential t distributionand for the typical background, signals andand themaintaining distributions the for same EC are signal e ciency. In comparison backgrounds at 13 TeV LHC. The plot is normalized to the `SM put in appendix. The two benchmarkwith LLP signals decay considered in the 7-10 m range of the MS, there is fraction of events per bin with a varying bin size, where for here are the glueballs from Higgsno detector boson decays, activities and the in the layers prior to that. Hence, LT2 t less than 1 ns are shown in linearelectroweakino scale and then pair in production loga- in the Gauge Mediated rithmic scale otherwise. Two representative signal models are the dominant background from punch can still be vetoed `X SUSY Breaking (GMSB) scenario. Both the glueballs shown, the delay time for the glueballs from the Higgs decay e↵ectively. LT1 SM and lightest neutralino proper lifetimes are set to be (red curves) and the GMSB neutralinosc⌧ = 10 from m. Drell-Yan The 10 GeV pair glueballs (red dashed line) have production (blue curves), with a light and a heavy benchmark X larger average boost comparingBackground to the 50 GeV consideration. glueballs — The main sources of the mass shown in dashed and solid curves, respectively. For all (solid red line), and hence haveSM a background sizable fraction faking of the such delayed and displaced signal signal events, the proper lifetime is set 10 m, and the dis- signals with delay time less thanare from one nanosecond. jets or similar For hadronic activities. The origin of tribution only counts for events decayed within [LT1 ,LT2 ] FIG. 1. An event topologyof with [4.2, an 10.6] LLP mX indecaying the transverse to two direction,the electroweakinos which follows pair the production,background the signals can be are classified not into same-vertex (SV) hard light SM particles a and b. A timing layer, at a transverse geometry of ATLAS MS in the barrelboosted region. and For hence the significantly back- collision delayed and compared pile-up to (PU). the For this study, we assume the distance L away from the beam axis (horizontal gray dotted T2 backgrounds, with 99% of the signal with t>1 ns. 4 line), is placed at the endground of the detector distribution volume shown (shaded in gray curves, we assume bunch time-spread distributions follow Gaussian distribution. spacing of 25 ns. The solid and dashed gray curves represent region). The trajectory of a potential SM background particle t2 backgrounds from a same hard collisionSearch vertex strategy. and hence— Wewith consider the signal withd an( ISRt) 1 2 is also shown (blue dashed line). The gray polygon indicates P = E 2 t , (2) PTjet timestamping the primary vertex and another SM ob- the primary vertex. a precision timing uncertainty of t = 30 ps and from the d t p2 t pile-up with a spread of t = 190 ps,ject respectively, from the LLP in units decay of (e.g., jet for this study) which fraction per 0.1 ns. The corresponding distribution for EC where the time spreads di↵er for di↵erent sources of shown in Fig. 1. An LLP, denoted as X,travelsadis- has large time delay t.TostudythesensitivitytoBSM t can be obtained approximately by scaling the horizontal axis backgrounds. tance ` into a detector volume and decays into two light signals with timing, we propose two searches using such X according to the ratio of size of the detector volume. The SV background for the signals mainly comes from SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- ing layer located at the beginningSM multi-jet of EC, and process. one with At least one prompt jet is required a transverse distance LT2 away from the beam axis. In a typical hard collision, the SM particles generally travel ATLAS geometry for a precisionto reconstruct timing layer PV4dlocated and at provide the timestamp, while at the timing layer. The delayedLate and displaced comers jet sig-will be spotted easily: close to the speed of light. The trajectories of charged SM the end of MS. They are tabulatedanother as following: trackless jet from the same hard collision faking nal, after requiring minimal decay transverse distance of particles can be curved, which increase the path length long-lived signals.j The fake jet has an intrinsic time delay LT LT Trigger ✏trig ✏sig ✏ Ref. 0.2 m (LT ), will not have good tracks associated2 1 with it. fake in comparison with neutral SM particles.1 For simplicity, t = 0. However,3 dueCMS to limited timing timing module resolution in Consequently, the major SM backgroundEC 1.17 is m from0.2 trackless m DelayJet 0.5 0.5 10 [12] we only consider neutral LLP signals where background reconstructing the9 PV4d, it could have a time spread. j,EC MS 310.6 m 4.2 m MS RoI 0.25, 0.5 0.25 5 10 [24] ATLAS MS LLP search from such charged particlesjets. can The be vetoed jet fake using rate particle of ✏fake = 10 is calculated us- The time resolution⇥ with(without the timing) planned precision timing identification and isolation.ing2 PythiaHence, the[25] decay by simulating products the trackless jets, where all PT For both searches, we assume similarupgrade performanceSame from-vertex CMS of tim- ishard t scattering= 30 ps. At 13 TeV with of X, taking particle a forcharged example, constituent arrives at the hadrons timing are too soft to be observed 1 ing resolution of 30 ps. For theintegrated MS search,background, luminosity because time ofint =spread 3 ab 30, the ps total number of layer with a time delay ofor missed due to trakcing ine ciencythe larger. The time trackless delay and jet muchsuch less background background events dueL to can be estimated, fraction is measured in the validation“shielding” data by for inner the detectors low- compared(precision to the EC timing) case, `X `a `SM SV EC j,EC 11 t = electromagnetism+ , jet search(1) at thea less ALTAS precise [26], timing and it (e.g. is 150 ps) couldEC : alsoNbkg achieve= j int✏trig✏fake 1 10 j X a SM 2 L ⇡ ⇥ found to be 10 . They also foundsimilar a huge physics additional reach. sup- The ✏trig, ✏sig and ✏fake areSV the e - MS j,MS 5 MS : N = j int✏ ✏ 4 10 , (3) pression through the energy depositionciencies ratio for trigger, between signal EC selection and a QCD jetbkg fakingL trig fake ⇡ ⇥ with a SM 1. It is necessary to have prompt ' ' the delayed jet signal with pT > 30 GeV in EC8 or MS, decay products or Initialand State hadronic Radiation calorimeter. (ISR) whichMoreover,Hard due collision to the BKG: actual detectorwhere time j The1 detector10 pb istime the resolution multi-jets cross-sectionfor MS with decays within the tracking volumn,respectively. the signal contains j ' ⇥ j arriving at timing layer with the speed of light to derive resolution ~30 ps pT > 30can GeV, be✏ downgradedtrig and ✏fake toare hundreds the e ciencies of ps for the low quality tracks in constrast to theFor truly the EC neutral search, jets, we assume a new trigger strategy the time of the hard collision at the primary vertex (to ofEC a (30ps) delayed cut: jet Δ usingt > 0.4 the ns CMS upgradeMS timing(200ps) layer. cut: and the energy deposition in the EC for the singal will be “timestamp” the hard collision). ISR jets could easily be ThisMS can(30ps) be realizedcut: Δt > by 1ns comparing theΔ promptt > 1ns jet with present for all processes, andmore we than use this neutraljets generic feature as well, we hence consider our jet pT > 30 GeV that reconstructs4 the four-dimensional pri- to “timestamp” the hardfake collision rate for assignment the proposed of 10 new3 to beBKG(SV)reasonabe << reasonably1 The validityBKG(MS of these-SV) description ~ 0.11 should be scrutinized by ex- mary vertex (PV4d) with the arrivalperimental time of measurement, another jet e.g. from Zero-Bias events. From conservative. 33 6/19/18 Zhen Liu LLP @ LHC Refs. [ 12LPC, TOTW15, 19], our description is appropriate up to proba- 4 6 For the MS search, we consider a new timing layer at bility of 10 to 10 level–the limit where the distribution are the outer layer of the MS of ATLAS. We take the MS Re- shown. 2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles, the ma- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles in a jet still travel with nearly searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. 2 lay since the mass of the new particle can be compara- searches in this letter.3 ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- signal search strategy of using the timing information, onds (ns), for entering EC, to tens of ns, for exiting the and more importantly, the corresponding consideration MS. As a result, with tens of picosecond (ps) timing for the background. A typical signal event of LLP is resolution, we have a sensitivity to percent level time delay caused by slow LLP motion, e.g., 1 > 0.01 X a with boost factor < 7. In Fig. 2, we show typical time 4 b delay t for a hypothetical timing layer at the outer Timing layer part of the ATLAS MS system for benchmark signals `a background to fake the signal in triggering and signature The collision time for two bunches of protons has a PU and the background, and the distributionswithout timing for information. EC are The background di↵erential typical temporal spread of t = 190 ps [12]. The dif- `SM put in appendix. The two benchmarkdistribution signals with considered respect to apparent delay time ( t) can ferential background from pile-up can be estimated as, here are the glueballs from Higgsbe boson estimated decays, as, and the LT2 electroweakino pair production in the Gauge Mediated PU SV @Nbkg( t) PU PU @Nbkg(t) SV PT N ( t; ). (6) `X SUSY Breaking (GMSB) scenario. Both the glueballs= N ( t; ). (4) bkg t @ t bkgP t @ t ' P LT1 SM and lightest neutralino proper lifetimes are set to be c⌧ = 10 m. The 10 GeV glueballs (red dashed line) have The key di↵erence between the background from the pile- The time delay cut on t reduces such background up and the same hard collision is that the typical time X larger average boost comparing to the 50 GeV glueballs PT PT through the tiny factor of ( t; t )if t/ t is greater spread is determined by the beam property for the for- than a few. The LLP signalP pays a much smaller penalty (solid red line), and hence have a sizable fraction of the mer, and by the timing resolution for the latter. They factor than the background due to its intrinsic delay, as signals with delay time less than one nanosecond. For typically di↵er by a factor of a few, e.g., 190 ps versus shown in Fig. 2. FIG. 1. An event topology with an LLP X decaying to two the electroweakinos pair production, the signals are not 30 ps for CMS with the current upgrade plan. For the EC light SM particles a and b. A timing layer, at a transverse boosted and hence significantly delayedThe background compared from to the the pile-up contains two hard (MS) search, if we apply cut t>1 (0.4) ns, the total distance L away from the beam axis (horizontal gray dotted collisions within the same bunch crossing but does not T2 backgrounds, with 99% of the signal with t>1 ns. estimated events from SM background including SV and line), is placed at the end of the detector volume (shaded occur at the same time. The majority of such back- PU is 1.3 (0.86), where SV backgrounds become com- region). The trajectory of a potential SM background particle ground can be eliminated by the standard isolation re- Search strategy.— We consider the signal with an ISR pletely negligible. is also shown (blue dashed line). The gray polygon indicates quirement, jet grooming procedure, etc. The background Backgrounds which do not from the hard collision are the primary vertex. jet timestamping the primary vertexfrom theand pile-up another requires SM ob- the coincidence of a triggered ject from the LLP decay (e.g., jet for this study) which hard to simulate, such as cosmic ray, beam halo, mis- hard event and fake signal events from pile-up (hard) col- connected tracks, interaction with detector material, etc. has large time delay t.TostudythesensitivitytoBSM shown in Fig. 1. An LLP, denoted as X,travelsadis- lision whose PV4d fails to be reconstructed. Since pile-up Thanks to the rich studies searching for LLP in all sub- tance ` into a detector volume and decays into two light signals with timing, we proposeevents two searches also have using spatial such spread, the interaction point in- detectors at the LHC, their properties are well measured X formation z would also enter the estimation of such back- SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- and can be vetoed e↵ectively. Furthermore, with the hard ing layer located at the beginningground. of EC, Therefore, and one given with that the typical spread is few signature (large energy deposition of more than 30 GeV) a transverse distance LT2 away from the beam axis. In cm, it can induce a time shift at most (100) ps [12], a typical hard collision, the SM particles generally travel ATLAS geometry for a precision timing layer located at ⇡ O and high track multiplicities with sizable time-delay, the Late comers will betypically spotted with aneasily: addition suppression of a geometrical signal can be well separated from these backgrounds. In close to the speed of light. The trajectories of charged SM the end of MS. They are tabulated as following: factor. Adding in quadrature, this will at most give an the future, the object reconstruction with separation not j PU particles can be curved, which increase the path length insignificant increase the spread in time t 60 ps. It LT2 LT1 Trigger ✏trig ✏sig ✏fake Ref. ⇡ only in spatial but also in time should help discriminate in comparison with neutral SM particles. For simplicity, has even less impact for MS, the pile-up background is EC 1.17 m 0.2 m DelayJet 0.5 0.5 10 3 [12] CMS timing module the various backgrounds. already small before timing cut. Thus, it can be safely we only consider neutral LLP signals where background 9 ATLAS MS LLP search In addition, in specific searches, signal typically has MS 10.6 m 4.2 m MS RoI 0.25,neglected 0.5 0.25 here.5 10 [24] from such charged particles can be vetoed using particle ⇥ (without timing) additional feature. For example, in our case, we actu- 2 At the HL-LHC, the total number of background identification and isolation. Hence, the decay products For both searches, we assume similar performance of tim- ally have two visible objects with di↵erent time delays. events canPile be-Up estimated,background, time spread of X, taking particle a for example, arrives at the timing ing resolution of 30 ps. For the MS search, because of Taking advantage of such characteristics, we expect the 190 ps (beam property) background can be further suppressed. layer with a time delay of the larger time delay and much less background due to EC : N PU = ✏EC n¯ j ✏j,ECf j 2 107, As a side note, triggering on delayed signals concern- “shielding” by inner detectors comparedbkg to thej ECLint case,trig PU fake nt ⇡ ⇥ ` ` ` ✓ inc ◆ ing the primary interaction vertex could become a very t = X + a SM , (1) a less precise timing (e.g. 150 ps) could also achieve PUj MS j j,MS j interesting and important application for the general X a SM similar physics reach. The ✏trig, ✏MSsig and : N✏bkg =are j theint✏trig e - n¯PU ✏fake fnt 50, (5) fake L inc ⇡ class of long-lived particle signals [30–32]. Triggers with ciencies for trigger, signal selection and a QCD jet faking✓ ◆ additional timing information (such as sizable delay) with a SM 1. It is necessary to have prompt the delayed jet signal with pT >where30 GeV = in 80 EC mb or is theMS, inelastic proton-proton cross- would complement current trigger system that focuses on decay products' or' Initial State Radiation (ISR) which inc respectively. section at 13 TeV [28].n ¯PU 100 is the average number very hard events, using HT , pT of jets, leptons, photons, arriving at timing layer with the speed of light to derive ⇡ For the EC search, we assumeof ainelastic new trigger interactions strategy per bunch crossing using instan- and missing ET [33, 34]. A much softer threshold could 34 2 1 the time of the hard collision at the primary vertex (to taneous luminosity 2 10 cm s [29]. In Eq. (5), be achieved with sizable time delays as an additional of a delayed jet using the CMS upgrade timing layer.⇥ “timestamp” the hard collision). ISR jets could easily be This can be realized by comparingone hard the prompt collision jet needs with to timestamp the event, while criterion, which would be extremely beneficial for LLP, present for all processes, and we use this generic feature the other hard collision contains at least two jets, all of especially for compressed signal searches. pT > 30 GeV that reconstructs the four-dimensional pri- to “timestamp” the hard collision for the proposed new which have to be neutral to miss the primary vertex re- mary vertex (PV4d) with the arrivalconstruction. time of Otherwise, another jet this second hard collision will Augmented sensitivity on LLP through precision 34 6/19/18 Zhen Liu LLP @ LHC leave tracks LPC TOTW and reconstructed as another vertex in the Timing.— Our first example is Higgs decaying to LLP PU ¯ tracker, thus get vetoed. Therefore, the background Nbkg with subsequent decays into bb pairs. This occurs in is suppressed by at least one additional factor of neutral model [10] where the Higgs is the portal to a dark QCD 2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles,j the3 ma- j jet fraction fnt 10 . This additional factor fnt, more sector whose lightest states are the glueballs. The de- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles instrictly a jet still speaking, travel' with should nearly be the probability for a multijet cays of the 0++ glueballs are long-lived. This benchmark searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. hard process whose PV4d is failed to be reconstructed has been studied without exploiting the timing informa- and mis-assigned to the triggered PV4d, which need to tion [35, 36]. Typical energy of the glueball is set by be estimated through full detector simulation and cali- the Higgs mass, and the time delay depends on glueball brated with data. mass. The signal of LLPs produced through the decay of 2 lay since the mass of the new particle can be compara- searches in this letter.3 ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- signal search strategy of using the timing information, onds (ns), for entering EC, to tens of ns, for exiting the and more importantly, the corresponding consideration MS. As a result, with tens of picosecond (ps) timing for the background. A typical signal event of LLP is resolution, we have a sensitivity to percent level time delay caused by slow LLP motion, e.g., 1 > 0.01 X a with boost factor < 7. In Fig. 2, we show typical time 4 b delay t for a hypothetical timing layer at the outer Timing layer part of the ATLAS MS system for benchmark signals `a background to fake the signal in triggering and signature The collision time for two bunches of protons has a PU and the background, and the distributionswithout timing for information. EC are The background di↵erential typical temporal spread of t = 190 ps [12]. The dif- `SM put in appendix. The two benchmarkdistribution signals with considered respect to apparent delay time ( t) can ferential background from pile-up can be estimated as, here are the glueballs from Higgsbe boson estimated decays, as, and the LT2 electroweakino pair production in the Gauge Mediated PU SV @Nbkg( t) PU PU @Nbkg(t) SV PT N ( t; ). (6) `X SUSY Breaking (GMSB) scenario. Both the glueballs= N ( t; ). (4) bkg t @ t bkgP t @ t ' P LT1 SM and lightest neutralino proper lifetimes are set to be c⌧ = 10 m. The 10 GeV glueballs (red dashed line) have The key di↵erence between the background from the pile- The time delay cut on t reduces such background up and the same hard collision is that the typical time X larger average boost comparing to the 50 GeV glueballs PT PT through the tiny factor of ( t; t )if t/ t is greater spread is determined by the beam property for the for- than a few. The LLP signalP pays a much smaller penalty (solid red line), and hence have a sizable fraction of the mer, and by the timing resolution for the latter. They factor than the background due to its intrinsic delay, as signals with delay time less than one nanosecond. For typically di↵er by a factor of a few, e.g., 190 ps versus shown in Fig. 2. FIG. 1. An event topology with an LLP X decaying to two the electroweakinos pair production, the signals are not 30 ps for CMS with the current upgrade plan. For the EC light SM particles a and b. A timing layer, at a transverse boosted and hence significantly delayedThe background compared from to the the pile-up contains two hard (MS) search, if we apply cut t>1 (0.4) ns, the total distance L away from the beam axis (horizontal gray dotted collisions within the same bunch crossing but does not T2 backgrounds, with 99% of the signal with t>1 ns. estimated events from SM background including SV and line), is placed at the end of the detector volume (shaded occur at the same time. The majority of such back- PU is 1.3 (0.86), where SV backgrounds become com- region). The trajectory of a potential SM background particle ground can be eliminated by the standard isolation re- Search strategy.— We consider the signal with an ISR pletely negligible. is also shown (blue dashed line). The gray polygon indicates quirement, jet grooming procedure, etc. The background Backgrounds which do not from the hard collision are the primary vertex. jet timestamping the primary vertexfrom theand pile-up another requires SM ob- the coincidence of a triggered ject from the LLP decay (e.g., jet for this study) which hard to simulate, such as cosmic ray, beam halo, mis- hard event and fake signal events from pile-up (hard) col- connected tracks, interaction with detector material, etc. has large time delay t.TostudythesensitivitytoBSM shown in Fig. 1. An LLP, denoted as X,travelsadis- lision whose PV4d fails to be reconstructed. Since pile-up Thanks to the rich studies searching for LLP in all sub- tance ` into a detector volume and decays into two light signals with timing, we proposeevents two searches also have using spatial such spread, the interaction point in- detectors at the LHC, their properties are well measured X formation z would also enter the estimation of such back- SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- and can be vetoed e↵ectively. Furthermore, with the hard ing layer located at the beginningground. of EC, Therefore, and one given with that the typical spread is few signature (large energy deposition of more than 30 GeV) a transverse distance LT2 away from the beam axis. In cm, it can induce a time shift at most (100) ps [12], a typical hard collision, the SM particles generally travel ATLAS geometry for a precision timing layer located at ⇡ O and high track multiplicities with sizable time-delay, the Late comers will betypically spotted with aneasily: addition suppression of a geometrical signal can be well separated from these backgrounds. In close to the speed of light. The trajectories of charged SM the end of MS. They are tabulated as following: factor. Adding in quadrature, this will at most give an the future, the object reconstruction with separation not j PU particles can be curved, which increase the path length insignificant increase the spread in time t 60 ps. It LT2 LT1 Trigger ✏trig ✏sig ✏fake Ref. ⇡ only in spatial but also in time should help discriminate in comparison with neutral SM particles. For simplicity, has even less impact for MS, the pile-up background is EC 1.17 m 0.2 m DelayJet 0.5 0.5 10 3 [12] CMS timing module the various backgrounds. already small before timing cut. Thus, it can be safely we only consider neutral LLP signals where background 9 ATLAS MS LLP search In addition, in specific searches, signal typically has MS 10.6 m 4.2 m MS RoI 0.25,neglected 0.5 0.25 here.5 10 [24] from such charged particles can be vetoed using particle ⇥ (without timing) additional feature. For example, in our case, we actu- 2 At the HL-LHC, the total number of background identification and isolation. Hence, the decay products For both searches, we assume similar performance of tim- ally have two visible objects with di↵erent time delays. events canPile be-Up estimated,background, time spread of X, taking particle a for example, arrives at the timing ing resolution of 30 ps. For the MS search, because of Taking advantage of such characteristics, we expect the 190 ps (beam property) background can be further suppressed. layer with a time delay of the larger time delay and much less background due to EC : N PU = ✏EC n¯ j ✏j,ECf j 2 107, As a side note, triggering on delayed signals concern- “shielding” by inner detectors comparedbkg to thej ECLint case,trig PU fake nt ⇡ ⇥ ` ` ` ✓ inc ◆ ing the primary interaction vertex could become a very t = X + a SM , (1) a less precise timing (e.g. 150 ps) could also achieve PUj MS j j,MS j interesting and important application for the general X a SM similar physics reach. The ✏trig, ✏MSsig and : N✏bkg =are j theint✏trig e - n¯PU ✏fake fnt 50, (5) fake L inc ⇡ class of long-lived particle signals [30–32]. Triggers with ciencies for trigger, signal selection and a QCD jet faking✓ ◆ additional timing information (such as sizable delay) with a SM 1. It is necessary to have prompt Pile-up BKG: intrinsic resolution The detector time resolution for the delayed jet signal with pT >where30 GeV = in 80 EC mb or is theMS, inelastic proton-proton cross- would complement current trigger system that focuses on decay products' or' Initial State Radiation (ISR) which inc respectively.~190 ps section at 13 TeVMS [can28]. n ¯bePU downgraded100 is the average to number very hard events, using HT , pT of jets, leptons, photons, arriving at timing layer with the speed of light to derive ⇡ ECFor (30ps) the EC cut: search, Δt > we1 ns assumeof ainelastic new trigger interactions strategy per bunch crossing using instan- and missing ET [33, 34]. A much softer threshold could hundreds of34 ps 2 1 the time of the hard collision at the primary vertex (to taneous luminosity 2 10 cm s [29]. In Eq. (5), be achieved with sizable time delays as an additional ofBKG(EC a delayed-PU) jet ~ 1.3 using the CMS upgrade timingMS (200ps) layer.⇥ cut: Δt > 1ns “timestamp” the hard collision). ISR jets could easily be This can be realized by comparingone hard the prompt collision jet needs with to timestamp the event, while criterion, which would be extremely beneficial for LLP, present for all processes, and we use this generic feature MS (30ps) cut: Δt > 0.4 ns the other hardBKG(MS collision- containsPU) << at1 least two jets, all of especially for compressed signal searches. pT > 30 GeV that reconstructs the four-dimensional pri- to “timestamp” the hard collision for the proposed new BKG(MS-PU) ~ 0.86 which have to be neutral to miss the primary vertex re- mary vertex (PV4d) with the arrivalconstruction. time of Otherwise, another jet this second hard collision will Augmented sensitivity on LLP through precision 35 6/19/18 Zhen Liu LLP @ LHC leave tracks LPC TOTW and reconstructed as another vertex in the Timing.— Our first example is Higgs decaying to LLP PU ¯ tracker, thus get vetoed. Therefore, the background Nbkg with subsequent decays into bb pairs. This occurs in is suppressed by at least one additional factor of neutral model [10] where the Higgs is the portal to a dark QCD 2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles,j the3 ma- j jet fraction fnt 10 . This additional factor fnt, more sector whose lightest states are the glueballs. The de- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles instrictly a jet still speaking, travel' with should nearly be the probability for a multijet cays of the 0++ glueballs are long-lived. This benchmark searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. hard process whose PV4d is failed to be reconstructed has been studied without exploiting the timing informa- and mis-assigned to the triggered PV4d, which need to tion [35, 36]. Typical energy of the glueball is set by be estimated through full detector simulation and cali- the Higgs mass, and the time delay depends on glueball brated with data. mass. The signal of LLPs produced through the decay of 2 lay since the mass of the new particle can be compara- searches in this letter.3 ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- signal search strategy of using the timing information, onds (ns), for entering EC, to tens of ns, for exiting the and more importantly, the corresponding consideration MS. As a result, with tens of picosecond (ps) timing for the background. A typical signal event of LLP is resolution, we have a sensitivity to percent level time delay caused by slow LLP motion, e.g., 1 > 0.01 X a with boost factor < 7. In Fig. 2, we show typical time b delay t for a hypothetical timing layer at the outer Timing layer part of the ATLAS MS system for benchmark signals `a and the background, and the distributions for EC are `SM put in appendix. The two benchmark signals considered here are the glueballs from Higgs boson decays, and the LT2 electroweakino pair production in the Gauge Mediated `X SUSY Breaking (GMSB) scenario. Both the glueballs LT1 SM and lightest neutralino proper lifetimes are set to be c⌧ = 10 m. The 10 GeV glueballs (red dashed line) have X larger average boost comparing to the 50 GeV glueballs (solid red line), and hence have a sizable fraction of the signals with delay time less than one nanosecond. For FIG. 1. An event topology with an LLP X decaying to two the electroweakinos pair production, the signals are not light SM particles a and b. A timing layer, at a transverse boosted and hence significantly delayed compared to the distance L away from the beam axis (horizontal gray dotted T2 backgrounds, with 99% of the signal with t>1 ns. line), is placed at the end of the detector volume (shaded region). The trajectory of a potential SM background particle is also shown (blue dashed line). The gray polygon indicates Search strategy.— We consider the signal with an ISR the primary vertex. jet timestamping the primary vertex and another SM ob- ject from the LLP decay (e.g., jet for this study) which shown in Fig. 1. An LLP, denoted as X,travelsadis- has large time delay t.TostudythesensitivitytoBSM signals with timing, we propose two searches using such tance `X into a detector volume and decays into two light SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- ing layer located at the beginning of EC, and one with a transverse distance LT2 away from the beam axis. In a typical hard collision, the SM particles generally travel ATLASLate geometry comers for awill precision be spotted timing layer easily: located at close to the speed of light. The trajectories of charged SM the end of MS. They are tabulated as following: particles can be curved, which increase the path length j LT2 LT1 Trigger ✏trig ✏sig ✏fake Ref. in comparison with neutral SM particles. For simplicity, 3 EC 1.17 m 0.2 m DelayJet 0.5 0.5 10 [12] CMS timing module we only consider neutral LLP signals where background 9 MS 10.6 m 4.2 m MS RoI 0.25, 0.5 0.25 5 10 [24] ATLAS MS LLP search from such charged particles can be vetoed using particle ⇥ (without timing) 2 identification and isolation. Hence, the decay products For both searches, we assume similar performance of tim- of X, taking particle a for example, arrives at the timing ing• resolutionEC: >0.8 ns of or 30 >1.2 ps. ns For the MS search, because of layer with a time delay of the largertiming time cut (<25 delay ns and always much less background due to “shielding” by inner detectors compared to the EC case, ` ` ` there) t = X + a SM , (1) a less• precise timing (e.g. 150 ps) could also achieve MS: 1 ns or 10 ns timing j X a SM similarcut physics (0.2 ns reach. or 2 ns The ✏trig, ✏sig and ✏fake are the e - ciencies for trigger, signal selection and a QCD jet faking with a SM 1. It is necessary to have prompt resolution sufficient) the delayed jet signal with pT > 30 GeV in EC or MS, decay products' or' Initial State Radiation (ISR) which • Significant improvement! respectively.• Little difference for signal arriving at timing layer with the speed of light to derive For the EC search, we assume a new trigger strategy the time of the hard collision at the primary vertex (to as they are very slow of a• delayedlarge tolerance jet using room the if CMS upgrade timing layer. “timestamp” the hard collision). ISR jets could easily be This can be realized by comparing the prompt jet with present for all processes, and we use this generic feature background non-gaussian; pT > 30 GeV that reconstructs the four-dimensional pri- to “timestamp” the hard collision for the proposed new mary vertex (PV4d) with the arrival time of another jet
36 6/19/18 Zhen Liu LLP @ LHC LPC TOTW
2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles, the ma- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles in a jet still travel with nearly searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. 2 lay since the mass of the new particle can be compara- searches in this letter.3 ble to its momentum. Here we outline a general BSM Typically, `SM/ SM range between several nanosec- signal search strategy of using the timing information, onds (ns), for entering EC, to tens of ns, for exiting the and more importantly, the corresponding consideration MS. As a result, with tens of picosecond (ps) timing for the background. A typical signal event of LLP is resolution, we have a sensitivity to percent level time delay caused by slow LLP motion, e.g., 1 > 0.01 X a with boost factor < 7. In Fig. 2, we show typical time b delay t for a hypothetical timing layer at the outer Timing layer part of the ATLAS MS system for benchmark signals `a and the background, and the distributions for EC are `SM put in appendix. The two benchmark signals considered here are the glueballs from Higgs boson decays, and the LT2 electroweakino pair production in the Gauge Mediated `X SUSY Breaking (GMSB) scenario. Both the glueballs LT1 SM and lightest neutralino proper lifetimes are set to be c⌧ = 10 m. The 10 GeV glueballs (red dashed line) have X larger average boost comparing to the 50 GeV glueballs (solid red line), and hence have a sizable fraction of the signals with delay time less than one nanosecond. For FIG. 1. An event topology with an LLP X decaying to two the electroweakinos pair production, the signals are not light SM particles a and b. A timing layer, at a transverse boosted and hence significantly delayed compared to the distance L away from the beam axis (horizontal gray dotted T2 backgrounds, with 99% of the signal with t>1 ns. line), is placed at the end of the detector volume (shaded region). The trajectory of a potential SM background particle is also shown (blue dashed line). The gray polygon indicates Search strategy.— We consider the signal with an ISR the primary vertex. jet timestamping the primary vertex and another SM ob- ject from the LLP decay (e.g., jet for this study) which shown in Fig. 1. An LLP, denoted as X,travelsadis- has large time delay t.TostudythesensitivitytoBSM signals with timing, we propose two searches using such tance `X into a detector volume and decays into two light SM particles a and b, which then reach timing layer at information, one with CMS geometry for a precision tim- ing layer located at the beginning of EC, and one with a transverse distance LT2 away from the beam axis. In a typical hard collision, the SM particles generally travel ATLASLate geometry comers for awill precision be spotted timing layer easily: located at close to the speed of light. The trajectories of charged SM the end of MS. They are tabulated as following: particles can be curved, which increase the path length j LT2 LT1 Trigger ✏trig ✏sig ✏fake Ref. in comparison with neutral SM particles. For simplicity, 3 EC 1.17 m 0.2 m DelayJet 0.5 0.5 10 [12] CMS timing module we only consider neutral LLP signals where background 9 MS 10.6 m 4.2 m MS RoI 0.25, 0.5 0.25 5 10 [24] ATLAS MS LLP search from such charged particles can be vetoed using particle ⇥ (without timing) identification and isolation.2 Hence, the decay products For both searches, we assume similar performancePrecision Timing of Enhanced tim- Search Limit (HL-LHC) of X, taking particle a for example, arrives at the timing 0 ing• resolutionEC: >0.8 ns of timing 30 ps. cut For (<25 the 10 MS search, because of layer with a time delay of the larger time delay and much less-1 background due to ns always there) 10
“shielding” by inner detectors compared to the EC case, h ` ` ` • MS: 0.2 ns or 1 ns timing -2 BRinv<3.5% t = X + a SM , (1) a less precise timing (e.g. 15010 ps) could also achieve cut (30 ps or 0.2 ns ) j X a SM similar physics reach. The ✏trig, ✏sig-3and ✏ are the e - resolution sufficient) XX 10 fake h → XX,X→ jj ciencies for trigger, signal selection→ and a QCD jet faking MS(30ps), Δt>0.4ns h
• Significant improvement! ( -4 with a SM 1. It is necessary to have prompt 10 MS(200ps), Δt>1ns ' ' the• delayed jet signal with pT > 30 GeV in EC or MS, decay products or Initial State Radiation (ISR) which 10 GeV benchmark point BR EC(30ps), Δt>1ns respectively. 10-5 arriving at timing layer with the speed of light to derive sensitive to the timing cut, MS2DV, noBKG For the EC search, we assume a new trigger strategy MS1DV, optimistic the time of the hard collision at the primary vertex (to as they are more boosted 10-6 of a delayedand having jet less using time the delay. CMS upgrade timing layer. “timestamp” the hard collision). ISR jets could easily be mX in [GeV] 10 40 50 This can be realized by comparing10-7 the prompt jet with 10-3 10-2 10-1 100 101 102 103 104 105 106 107 108 present for all processes, and we use this generic feature p > 30 GeV that reconstructs the four-dimensional pri- T cτ (m) to “timestamp” the hard collision for the proposed new mary vertex (PV4d) with the arrival time of another jet
37 6/19/18 8 TeV results from C. Csaki, E. Kuflik, S. Lombardo, O. 13 TeV MS search projection (w.o. timing), C. Coccaro, Slone, 1508.01522 D. Curtin, H. Lubatti, H. Russell, J. Shelton 1605.02742
2 Charged stable (at the scale of tracker or detector volume) par- 3 Although Jets contain soft (and hence slow) particles, the ma- ticles are highly constrained by the heavy stable charged particle jority of the constituent particles in a jet still travel with nearly searches by both ATLAS and CMS [14–16]. the speed of light [12, 21–23]. Challenges (opportunities) • Timing reducing background to 10^-10 level. Early measurement for HSCP (non-pointing photon) indicates the SM background behavior agree well with Gaussian up to 10^-6 (10^-4) level (experimental Monte Carlo went to 10^-9), where the plot ends (data insufficient); Would be an interesting SM property measurement; • For EC search, timing layer will be there. Delayed jet (anything) trigger would require non-trivial effort to realize, low+high level with jet ROI; Once realized, could be universal boost to LLPs at the LHC! • For MS search, a feasibility study on new timing layer options like this, balancing technology, design, cost, and physics goals would be a natural future step. As we have shown, except for the light LLP (~10 GeV), the large delay does not require 10s ps timing precision. Even (sub) nanosecond for the MS is sufficient; • There are many more handles on the signal selection and background rejection can be used;
38 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Summary and outlook
• LHC great detector for LLP searches, a rich program is still under development;
• Our recast study shows the broad coverage of LLP searches;
• All traditional LLP searches could be augmented by the timing information (re-optimization); • New searches can capture general features of the LLP in a very robust way by exploiting their delayed feature;
• Precision timing is a new dimension of particle physics information available for BSM searches
39 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Thank you! Backup
40 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 41 42 43 44 45 46 47 48 6/19/18 Zhen Liu LLP @ LHC LPC TOTW ATLAS non-pointing photon
49 6/19/18 Zhen Liu LLP @ LHC LPC TOTW CMS Heavy stable charged particle (HSCP) track+ToF
50 6/19/18 Zhen Liu LLP @ LHC LPC TOTW MS Volume
• Effective decay volume 4-7 m 4-10 m. • New layer and upgrades might relax/extend the MS Vertexing length. • We took the full volume in our study. • If stick to 4-7 m, the efficiency will reduce by roughly a factor of 2.
51 6/19/18 Zhen Liu LLP @ LHC LPC TOTW current status and challenges also Long-lived particles mentioned in many other talks in this workshop These nonconventional and rich BSM signatures receives a lot of attention as: • Theoretically well motivated: SUSY (RPV, GMSB, Split, compressed, etc.), neutral naturalness, hidden valley, dark shower… etc; • Experimentally challenging but bearing great potential for discovery: • New signatures could have been missed by conventional searches; • Low (zero) background analysis once carried out);
52 6/19/18 Zhen Liu LLP @ LHC LPC TOTW picture merit: Heather Russel Results on Mini-split SUSY
53 Mini-Split Gluino ➞ 2j + LSP
q q
g~ q*~ EM-neutral LSP
Arkani-Hamed & Dimopoulos (2004) Arvanitaki, Craig, Dimopoulos, Villadoro (2012) 54 Arkani-Hamed, Gupta, Kaplan, Weiner, Zorawksi (2012) Mini-Split Gluino ➞ 2j + LSP
FULL lifetime coverage up to >1 TeV!
• The dijet in the final boosts the efficiencies for displaced dijet searches. • The prompt jets+MET searches also covers a range of lifetime in the low mass, as fractions of long-lived particles decay promptly (boundary in dashed lines indicates possible extrapolation.
This figure shows one extreme case with large mass
55 splitting between the LSP and NLSP. How about a bit compressed? Mini-Split Gluino ➞ 2j + LSP
In case of compressed spectra (right panel) • Most searches rely on visible SM partcles greatly reduced due to energy cuts (necessary to cut away SM backgrounds from non- prompt decay and cosmic rays, etc) • Heavily charged stable particle search remains the same as no 56 decays are required. • Different displaced search channels are more complementary, more important. 57 58 59 Results on RPV SUSY
60 61 62 63 64 65 Leptonic decays
66 67 Displaced Leptons
J. Evans, J. Shelton, arXiv:1601.01326
68 69 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 70 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 71 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 72 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 73 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 74 6/19/18 Zhen Liu LLP @ LHC LPC TOTW 7512 6/19/18 Zhen Liu LLP @ LHC LPC TOTW Long-lived SM
76 6/19/18 Zhen Liu LLP @ LHC LPC TOTW