630 likes | 637 Views
This report details the simulation effort by the UCSC/SCIPP BeamCal Simulation Group at the ECFA Linear Collider Workshop held in Santander, Spain from May 30 to June 5, 2016. The group, led by Bruce Schumm, focused on parameters of the ILC IP, Bhabha events, SUSY studies, and more. Various BeamCal geometries were explored, along with Vertex Detector configurations and IR geometry occupancies. The study examined the impact of different IP beam parameters on beamstrahlung observables reconstructed in the BeamCal. Conclusions on BeamCal efficiency, anti-DiD field effects, and contributors are discussed.
E N D
Report on the UCSC/SCIPP BeamCal Simulation Effort ECFA Linear Collider Workshop Palacio de la Magdalena Santander, Cantabria, Spain May 30 – June 5, 2016 Bruce Schumm UC Santa Cruz Institute for Particle Physics
The SCIPP BeamCal Simulation Group • The group consists of UCSC undergraduate physics majors (and one engineering major) • Christopher Milke (Lead)* Heading to SMU’s doctoral program in fall • Jane Shtalenkova, Luc D’Hauthuille, Spenser Estrada, Benjamin Smithers, Summer Zuber, Cesar Ramirez • AlixFeinsod • Led by myself, with technical help and collaboration from Jan Strube, Anne Schuetz, Tim Barklow • VXD Occupancy / Anti-DiD Field • Determining ILC IP parameters with the BeamCal • Bhabha events and the two-photon physics veto • SUSY in the degenerate limit
IR Layout Low-Z Mask M1 Mask BeamCal L* 3
BeamCal Face Geometry Options • “plugged” • Wedge cutout • Circle cutout “wedge” cutout Plug Insert plug here “circle” cutout 4
Incidence of pair backgrounds on BeamCal with and without “anti-DiD” field BeamCal Face Without anti-DiD With anti-DiD Beam entrance and exit holes 5 Tom Markiewicz, SLAC
Configurations Explored Nominal: L* = 4.1m; no antiDiD; plug in place Then, relative to Nominal: Small L*: L* = 3.5m AntiDID: Include antiDiD field Small L* AntiDID: L* = 3.5m with antiDiD field Wedge: Remove BeamCal plug Circle: Remove additional BeamCal coverage as shown in prior slide. 6
Vertex Detector Configurations We have studied occupancy as a function of two aspects of the VXD readout architecture • Pixel size • 15 x 15 microns2 • 30 x 30 microns2 • Integration time • 1 beam crossing • 5 beam crossings 7
Nominal IR Geometry Occupancy Distributions (Barrel) Stacked histograms! 15 x 15 1 BX 30 x 30 5 BX x10-3 x10-3 8
Nominal IR Geometry Occupancy Distributions (Endcap) Stacked histograms! 15 x 15 1 BX 30 x 30 5 BX x10-3 x10-3 9
We note that: • Pulse-by-pulse variation is small • Occupancy only appreciable for largest pixel size (30x30) and greatest integration time (5 Bx) • Inner layer (0) dominates occupancy in barrel • Inner layer (0) characteristic of occupancy in endcap • Study IR configuration dependence with layer 0 (both endcap and barrel) for 30x30 pixel integrating over 5 Bx. In terms of: azimuthal dependence in barrel; radial dependence in endcap 10
Barrel: Mean Occupancy vs. Phi x10-4 30 x 30 5 BX Occupancy roughly constant in phi 11
Endcap: Mean Occupancy vs. R 30 x 30 5 BX Occupancy varies drammatically with radius; dominated by inner radii 12
BeamCal Efficiency L* Dependence Base Small L* larger L* consistently displays higher efficiency 13
BeamCal Efficiency L* Dependence Factorized Difference is largely geometric 14
BeamCal Efficiency and the Anti-DiD Field Noticeable but small effect 15
Contributors • Luc D’Hauthuille, UCSC Undergraduate (thesis) • Anne Schuetz, DESY Graduate Student • Christopher Milke, UCSC Undergraduate • With input from Glen White, Jan Strube, B.S. Goal Idea is to explore the sensitivity of various beamstrahlung observables, as reconstructed in the BeamCal, to variations in IP beam parameters. The sensitivity will be explored with various different BeamCal geometries.
Of these, we believe the following can be reconstructed in the BeamCal: • Total energy and its r, 1/r moment • Mean depth of shower • Thrust axis and value (relative to barycenter; could also use mode of distributions. What is wise choice though? Maybe just (0,0)?) • Mean x and y positions • Left-right, top-bottom, and diagonal asymmetries
IP Parameter Scenarios • Thanks to Anne Schuetz, GuneaPig expert • Relative to nominal: • Increase beam envelop at origin (via -function), for electron and positron beam independently, by 10%, 20%, and 30% • Move waist of electron and positron beam (independently) back by 100m, 200 m, 300 m. • Change targeting angle of electron and positron beam (independently) by 5 mrad and 20 mrad(in retrospect, isn’t this a bit much?) • Details at • https://wikis.bris.ac.uk/display/sid/GuineaPig+simulations+for+BeamCal+study
First (Early) Results • Luc has coded the following observables: • Deposited energy, mean depth of shower, L/R and up/down asymmetries, thrust (relative to barycenter) value. • He has taken eight beam crossings (working on larger sample soon!) and explored the following “trajectories”: • Beam envelope for electrons • Beam envelope for positrons • Electron waist position • Following are a core-dump of plots of these observables over those trajectories.
Beam Envelope Scan (Electrons and Positrons)
Total Deposited Energy e+ and e- beam envelope scan
Mean Depth e+ and e- beam envelope scan
L/R Asymmetry e+ and e- beam envelope scan
Up-Down Asymmetry e+ and e- beam envelope scan
Thrust Value e+ and e- beam envelope scan
Waist Scan (Electrons Only)
Total Deposited Energy e- waist scan
Mean Depth e- waist scan
L/R Asymmetry e- waist scan
Up-Down Asymmetry e- waist scan
Thrust Value e- waist scan
Summary and Conclusions • First look at BeamCal observables and IP parameter dependence • Need to finish coding observables (thrust definition question) • Need to increase statistics (~100 pulses generated; working on simulation) • Need to develop some more interesting IP parameter variations (discussion!) • Need to explore sensitivity to BeamCal geometry • But this should be a good foot in the door for now…
Degenerate SUSY and Electron Tagging • SUSY has a cosmologically-motivated corner where a weakly-coupled particle (stau) is nearly generate with the LSP (0) • We have generated events at Ecm= 500 GeV with • M~= (100, 150, 250) GeV • ~-0splittings of (20.0, 12.7, 8.0, 5.0, 3.2, 2.0) GeV • Concern: Two-photon events provide • greater and greater background as • splitting decreases. • Hope: We can tag the scattered electron or • Positron in the Beamcal and veto. • But: If photons are from Beamstrahlung, electron/positron do not get a pTkick (is this right Tim?)
Two-Photon Event Rate • Thanks to Tim Barklow, SLAC, we have ~107 generator-level two photons events, with electron/positron photon fluxes given by the Weizsacker-Williams approximation (W) and/or the Beamstrahlung distribution (B). • Events have been generated down to M = 300 MeV. • For this phase space, the ILC event rate is approximately 1.2 events/pulse. • 1 year of events corresponds to (1.2)x(2650)x(5)x(107) events, or about 1.6x1011 events per year. • How do we contend with such a large number of events in our simulation studies?
Two-Photon Approach • Convenient data storage in 2016: ~5 TB • Tim Barklow: • 5 TB is about 109 generated (not simulated!) events • 1011 events requires 4000 2-day jobs • Jan Strube: Don’t worry about CPU (really?) • Proposed approach: • Do study at generator-level only. • Except: Full BeamCal simulation to determine electron-ID efficiency as a function of (E,r,) of electron. Parameterize with 3-D function and use in generator-level analysis • Devise “online cuts” applied at generation that reduce data sample by x100 (can this be done?) • Store resulting 109 events and complete analysis “offline”
In Search Of: “Online Selection” • For now, looking at three observables: • M: mass of system • S: Sum of magnitudes of pT for all particles in system • V: Magnitude of vector sum of pT for particles in system • Each of these is done both for McTruth as well as “reconstructed” detector proxy • Detector Proxy: Particles (charged ot neutral) detected if • No neutrinos • |cos()| < 0.9
ISO “Online Selection”: Mass (M) 0 Mass 2 GeV Splitting SUSY Signal; ~ Mass = 150 GeV Two-photon background Seems like a clean cut, but what is seen in the “detector”?
“Detected” Mass (M) 0 Mass SUSY Signal; ~ Mass = 150 GeV Two-photon background For 2 GeV splitting, even a cut of 0.5 MeV removes some signal
S Observable: “Detected” 0 Mass SUSY Signal; ~ Mass = 150 GeV Two-photon background Fairly promising as well; but is it independent of M?
V Observable: “Detected” 0 Mass Two-photon background SUSY Signal; ~ Mass = 150 GeV Not as promising; looks better for “true”, but even for Vtrue = 0, reconstructed V has significant overlap with SUSY signal (save for “offline” part of study?)
Cut flow for S, M Distributions • News is not the best: • S and M observables very correlated • 3% loss of signal (at 2 GeV!) reduces background by only ~2/3 • Other discriminating variables?
Bhabha Events • Issue: • Degenerate SUSY has background from two-photon events • Hope to reduce by detecting scattered primary e+/- in BeamCal and vetoing the event • If a SUSY event is overlain with a Bhabha event with an e+/- in the BeamCal, we will reject SUSY • What is the rate of Bhabha events with e+/- in the Beamcal? • Bhabhas with virtuality-Q2 > 1 GeV (~ 4 mrad scatter) available at • with cross section = 278 nb • Raw rate of 0.76 Bhabha events per beam crossing ftp://ftp-lcd.slac.stanford.edu/ilc4/DBD/ILC500/bhabha_inclusive/stdhep/bhabha_inclusive*.stdhep
Bhabha Event Classes Bhabha events fall into three classes Miss-Miss: Both e-/e+ miss the BeamCal; not problematic Hit-Hit: Both e-/e+ hit the BeamCal; should be identifiable with kinematics (need to demonstrate) Hit-Miss: One and only one of e-/e+ hit the BeamCal; background to two-photon rejection. Naively, 11% of SUSY events would be rejected due to Hit-Miss events, plus whatever fraction of the 48% of Hit-Hit crossings aren’t clearly identified based on e+/- kinematics.
Hit/Hit events: e+-e- angular correlation “Type a quote here.” –Johnny Appleseed
After cut of < 1.0 Mrad, 33% of Hit/Hit Bhabhas remain (16% of crossings). Can possibly eliminate with energy cut (need to balance against two-photon and SUSY events)
For Hit/Miss events, there may well be useful kinematic handles… but again, need to compare to two-photon and SUSY signal distributions
Vertex Occupancy Dependence on L* Configuration 30 x 30 5 BX x 10-4 3.5 m L* L* occupancy differences appear to depend on backscatter deflection angle 4.1 m L* 48
Vertex Occupancy Dependence on Anti-did Field Plug is in place! 30 x 30 5 BX x 10-4 Base Anti-did field generally improves occupancy in barrel and consistently improves occupancy in endcap Anti-did 49
Occupancy Dependence on Plug Geometry x 10-4 Base As expected, occupancy gets progressively lower as more of the BeamCal plug is cut away Circle Wedge 50