ATLAS L upgrade strategy for HLLHC W Kozanecki

ATLAS L- upgrade strategy for HL-LHC W. Kozanecki, CEA-Saclay � Motivation � Lessons from LHC Runs 1 & 2 � General guidelines & key issues for luminometer design at HL-LHC � ATLAS luminosity-upgrade projects for Run 4 � Beam-Conditions Monitor (BCM') LUCID-3 High-granularity timing detector (HGTD) In closing… W. Kozanecki Slide 1 FCAL Collaboration meeting, Kraków, 10 May 2018

Upgrade motivation: both physics & instrumental � The luminosity precision achieved so far � ATLAS best: 1. 8% in 2011 (7 Te. V) [LHCb: 1. 2% in 2011, combining vd. M & BGI] already is the dominant uncertainty in (some) SM measurements “Precision measurement and interpretation of inclusive W+, W− and Z/γ∗ production cross sections with the ATLAS detector”, ATLAS Collaboration, Eur. Phys. J. C (2017) 77: 367" � Theorists are pushing for a ~1% L measurement at HL-LHC � see e. g. G. Salam @ ECFA 2016 (https: //indico. cern. ch/event/524795/contributions/ 2235443/attachments/1347759/2034269/HL-LHC-SMHiggs-theory. pdf) � None of the present bunch-by-bunch (bbb) luminometers can handle HL-LHC conditions, except (presumably) � track counting (ATLAS): proved crucial in both Run 1 & Run 2 � pixel cluster counting (CMS): provides reference luminosity for physics analyses In both cases W. Kozanecki � unavailable online � ~ weeks delay for processing & offline analysis, labor-intensive Slide 2 FCAL Collaboration meeting, Kraków, 10 May 2018

ATLAS luminosity determination in Run 2: luminometers Note: all luminometers are independent of TDAQ (exc. trk-, vtx- & Z-counting) LUCID – Luminosity measurement using a Cherenkov Integrating Detector (bbb) LUCID-2 “ATLAS-preferred” for 13 Te. V pp data + Z counting (relative-L checks) MPX/TPX MBTS (HI, v. low-L pp) + Vertex counting + Track counting (both bbb, but stats!) BCM – Beam Conditions Monitor (bbb) W. Kozanecki Slide 3 “ATLAS-preferred” for (most) 7 & 8 Te. V pp data FCAL Collaboration meeting, Kraków, 10 May 2018

ATLAS L-monitoring algorithms used in routine operation � Event- (or zero-) counting algorithms: bunch-by-bunch (bbb) count the fraction of BX with zero “events” m from Poisson probability �L � is a monotonic (but highly non-linear) function of the “event” rate 2009 -2015 if m gets too large, no empty events “zero starvation” or “saturation” Hit-counting algorithms (bbb) count the fraction of channels hit in a given bunch crossing (BX) � Poisson formalism, very similar to that of event counting � linearity vs. m depends on technology, granularity, thresholds, . . . � more � sensitive to instrumental drifts than event counting 2012 - conceptually similar to hit-counting. Examples: ATLAS, LHCb Flux-counting algorithms: summed over all bunches W. Kozanecki 2016 - Track-counting algorithms: bbb, but TDAQ-limited � ATLAS: # LUCID hits. CMS: # pixel clusters. 2011 - example: current in TILE PMTs, current in LAr gaps Slide 4 FCAL Collaboration meeting, Kraków, 10 May 2018

Luminosity for physics: ATLAS analysis flow 1. Calibrate the absolute L scale by the van der Meer (vd. M) method dedicated beam conditions: m ~ 0. 5, Lbunch ~ 5 1028, nb ~ 30, Ltot ~ 1 -2 1030 � instrumental reqrmts: isolated bunches, low m, desqueezed optics (b* = 19 m) � accelerator-related reqrmts: few isolated bunches, moderate bunch currents, Qc = 0 2 distinct measurements � beam-separation scans These requirements/procedures are sufficiently fundamental that they are likely to remain valid/used at HL-LHC absolute luminosity scale based on factorizable vd. M analysis non-factorization correction from luminous-region evolution during scans (aka beam imaging) 2. Transfer calibration from vd. M (L ~ 1030) to physics (L ~ 1034) regime correct m-, nb- (& Ltot? ) -dependent biases in bbb luminometers � requires � best 3. 4. ≥ 1 luminometer “known” to be linear wrt Ltot & m if close in time to vd. M scans • Not understood • Dominant systematic since 2012 • Crucial to master bef. HL-LHC Characterize & correct instrumental drifts over the running year flux or rad. aging: PMT [LUCID, TILE], diamonds [BCM], scintillators [TILE] drifts (gain, timing, ID conditions, tracking efficiency. . . ), residual m-dep. , . . . Quantify the relative long-term consistency & stability of as many independent L msmts as possible W. Kozanecki Slide 5 FCAL Collaboration meeting, Kraków, 10 May 2018

Luminosity calibration by the van der Meer method W. Kozanecki Slide 6 FCAL Collaboration meeting, Kraków, 10 May 2018

On the importance of multiple, redundant L-measurement techniques Example: calibration transfer Example: long-term consistency (TILE anchored to tracks) m-dep. corresponds to a -8 % calibrationtransfer correction at m ~ 35 TILE -Tracks =1. 6 % W. Kozanecki All detectors suggest that LUCID L is underestimated by ~ 1 -4 % Slide 7 FCAL Collaboration meeting, Kraków, 10 May 2018

(25 ns spacing) N. B. : today <m> ~ 60, mpeak ~ 80 aim for m ≤ 200 @ HL-LHC W. Kozanecki Slide 8 FCAL Collaboration meeting, Kraków, 10 May 2018

General guidelines for luminometer design at HL-LHC � Dynamic range & linearity � maintain sub-percent accuracy from vd. M regime to high-L physics running �> 4 orders of magnitude in pile-up [0. 01 < m < 200] (single-bunch performance? ) � 3 orders of magnitude in # bunches [1 < nb < 2700] (total-rate-dependent biases? ) � 6 orders of magnitude in L [5 1028 (vd. M, 1 bunch) to 5 -7 1034 cm-2 s-1 (physics)] Bunch-by-bunch (bbb) luminosity information both for isolated bunches & for 25 ns bunch trains (out-of-time elx pileup? ) unbiased by physics/calibration trigger menu � even random triggers result in a biased bbb distribution if the m profile is not flat sampling every single bunch slot (including unfilled afterglow subtraction) � preferably � on every single turn, i. e. without deadtime (like LUCID & BCM) if impractical, need precision accounting of bbb deadtime available online with only moderately degraded precision (LUCID, BCM) � track-counting � (ATLAS) & PCC (CMS) L info currently takes weeks to be available Azimuthally- and z-symmetric coverage W. Kozanecki to minimize sensitivity to beamspot parameters (<z>L , sz. L ) + crossing angle Slide 9 FCAL Collaboration meeting, Kraków, 10 May 2018

Key issue # 1: redundancy � Redundancy across diverse luminometer technologies/algos: a must! Unclear that “one size fits all”, i. e. that one technology or L algorithm will perform satisfactorily under all beam conditions (we never managed so far) � Since 2016: event-counting algorithms all saturated (at m well below 60) � Among the established methods, only the hit-counting approach can handle the present m span (Run 2: 0. 01 < m < 100). Hence the interest in: � event counting over ≠ acceptance ranges simultaneously [ BCM': segmented diamond arrays] � charge-based (i. e. analog) L measurement [ LUCID-3 approach: quartz-fiber bundles + PMTs] � state-of-the-art pixel counting w/ dedicated readout [ATLAS HGTD; CMS fwd pixel detector] � Essential to maintain high-acceptance luminometers/algorithms at HL-LHC � precision vd. M calibrations at high m / high L are impractical (biases observed at m ~ 2 already) � must be able to handle very-low L (heavy ions) or moderate-m, moderate-energy pp running � Building-in “ladder-like” cross-calibration strategies (across m- & L- ranges) is crucial Such diversity/complementarity is the only way to disentangle/mitigate multiple sources of systematic uncertainty � need both: � multiple algorithms per luminometer (today: event- vs. hit-counting vs. charge measurement in LUCID, track-counting vs. PCC in ID, ≠ cell families in calorimeters) for internal consistency checks � a spectrum of luminometer technologies – that ≠ algorithms (e. g. LUCID) agree is a necessary, but not a sufficient condition for accuracy W. Kozanecki Slide 10 FCAL Collaboration meeting, Kraków, 10 May 2018

Key issue # 2: aging � Aging short-term (i. e. during a fill) observed or suspected in: � BCM (“anti-aging”, attributed to charge pumping) � TILE PMTs (suggested by time-evolution of laser correction) � LUCID long term (i. e. across a running period) observed in: � BCM diamonds (2012 drift correction) � LUCID PMTs (complex sequence of aging, partial recovery, partial stabilization) � TILE PMTs (partial recovery during technical stops) � TILE scintillators (time scale: days to weeks, depending on the cell family) � ATLAS PMTs (from comparison of Bi-calibration results bef. / after fill) IBL/pixel/strip tracker (time-dep. track-efficiency corrections) what monitoring/calibration/correction methods can be used? � BCM � / LUCID-1 [2009 -13]: lack of / disappointing performance of self-calibration strategy BCM' [HL-LHC] will calibrate the entire electronic chain. – however cannot track polarization/pumping/aging of diamond sensors themselves � LUCID-2 [2015 - ] & LUCID-3 [HL-LHC]: gain monitored by built-in 207 Bi source, stabilized by HV adjustmts – however Bi calibration does not account for observed time variation of m-dependence correction � W. Kozanecki track-finding efficiency monitored/corrected using Z mm tag & probe analysis Slide 11 FCAL Collaboration meeting, Kraków, 10 May 2018

Key issue # 3: afterglow 104 Bunch-by-bunch event rate per bunch crossing in ATLAS run 162882, as recorded by a LUCID algorithm (Event_OR) that requires at least one hit in either LUCID arm 7 Te. V 150 ns Observed luminosity averaged over the fill as a function of the bunch-crossing number for the LUCID_Event. OR and BCMV_Event. OR algorithms for a single LHC fill with 1042 colliding bunch pairs. On this scale the BCMV and LUCID luminosity values for colliding BCIDs are indistinguishable. The small “afterglow” luminosity comes in BCIDs where no bunches are colliding and is the result of induced activity seen in the detectors. Only 400 BCIDs are shown so that the details of the afterglow in the short and long gaps in the fill pattern can be seen more clearly 7 Te. V 50 ns 102 W. Kozanecki Slide 12 FCAL Collaboration meeting, Kraków, 10 May 2018

BCM TOF concept: • Collisions : in time • Background: out of time ATLAS L-upgrade: BCM’ (1) � � Diamond-based system, to provide: Fast, bbb safety system for ATLAS Background monitoring Luminosity measurement Location: within (removable) pixel section of ITK r ~ 100 mm, z ~ ± 1. 8 m, h ~ 3. 6 after 2 ab-1: � 200 Mrad, 2 x 1015 neq/cm 2 4 stations per side � functionalities: � mounted L sensor, fast-abort sensor, (slow) beam-loss monitor (BLM) on a pixel ring BCM’ stations Pixel module W. Kozanecki BCM’ ring Slide 13 FCAL Collaboration meeting, Kraków, 10 May 2018

ATLAS L-upgrade: BCM’ (2) � Sensor design p. CVD diamond substrate 300 -500 μm thick robustness of abort device Build dynamic range into sensor design 8 pads from 1 to 32 mm 2 (1 st design) � occupancy � 250 to 8000 MIP’s @ danger level (25 k/cm 2/BX) � option: � from 0. 06 to 2 at μ = 200 one of the 1 mm 2 pads 3 -D ? capacitance, for L only, dedicated HV line ? Pads bump-bonded to 2 chips � reduce C, L � accommodate � double � irregular topology bump rows for rigidity ? ASIC: for each channel, TDC (time stamp) + ADC (charge) for 1 st hit in each bunch slot analog-pulse injection for testing/calibrating entire elx chain W. Kozanecki � Slide 14 TDAQ 40 MHz synchronous readout over lp. GBT/VTR opto-link to custom FELIX pre-processor triggered events ATLAS data stream FCAL Collaboration meeting, Kraków, 10 May 2018

The present (Run-2) LUCID detector LUCID-2 can measure bunchby-bunch L both by • counting hits [digital; ref. algo] • ∫-ing charge [analog; dev. only] in every BX & every turn It has operated succesfully since 2015 • at high L (> 2 1034), without saturation up to m ~ 120 -130 • at low L (~ 1028) without significant statistical error IP LUCID-2 consists of • 16+16 photomultipliers monitored by radioactive Bi-207 sources (�� ) • 4+4 bundles of quartz fibers monitored by (�� ) LED signals W. Kozanecki Slide 15 ✫ (17 m away) Bi-207 run FCAL Collaboration meeting, Kraków, 10 May 2018

ATLAS L-upgrade: a possible future LUCID fiber detector � With m-values going up to 140 -200 the present type of quartz photomultiplier detector will saturate (hits in every BXs). � Present idea for LUCID-3: fiber detector with a Bi-207 source at the end of the fiber bundles that provides Cherenkov light for calibration � Prototyping started 2. 45 m long PUV 800 quartz fiber bundle with 35 fibers; each end epoxied in ferrules and polished One fiber end was connected to a Hamamatsu R 760 photomultiplier. At the other end, 50 ml of a Bi-207 solution was applied directly to the fibers & let to dry. W. Kozanecki Slide 16 FCAL Collaboration meeting, Kraków, 10 May 2018

ATLAS L-upgrade: (one of the applications of) the HGTD [High. Granularity Timing Detector] � � Original (& primary) motivation High vertex density & degraded z 0 resolution at high h ambiguous track-to-vertex association Spatially overlapping vertices can be resolved in the time dimension using accurate vertex timing measurements HGTD in a nushell two endcap disks at z = ± 3. 5 m Active area: 120 mm < R < 640 mm ⇒ 2. 4 < |h| < 4. 0 Si-based Low-Gain Avalanche Detector (LGAD) technology ⇒ st = 30 ps/track over the lifetime of HL-LHC 2 Si layers per disk RQ < 10% occupancy @ m = 200 ⇒ 1. 3 mm x 1. 3 mm pixels W. Kozanecki Slide 17 FCAL Collaboration meeting, Kraków, 10 May 2018

HGTD L measurement: Pixel Cluster Counting (PCC) � Use pads 2. 8 <|h| < 3. 1 � Excellent linearity… high granularity no saturation � 10% occupancy easily handled by Poisson formalism good statistical power over full m range Full detector simulation 0. 2 % vd. M scan � . . . in simulation! Deadtime-less, bbb readout � This effort would greatly benefit from acquiring real-life experience with PCC-based L determination using the forward-pixel disks in the present ATLAS detector � Hit count per ASIC (2 cm x 2 cm area) at 40 MHz (every BX on every turn) for � central time window, and, separately, for � sideband(s) for afterglow subtraction – take advantage of good time resolution W. Kozanecki Slide 18 FCAL Collaboration meeting, Kraków, 10 May 2018

Afterglow impact on PCC (CMS example) L signal 3 -6 % (50 ns spacing, expected larger @ 25 ns) Afterglow buildup Afterglow tail CMS afterglow subtraction: fit a single-bunch afterglow model (weighed by the L measured in the colliding bunches) to the afterglow-buildup & - tail data in the empty bunches W. Kozanecki Slide 19 FCAL Collaboration meeting, Kraków, 10 May 2018

HGTD: extensive design & R&D effort underway W. Kozanecki Slide 20 FCAL Collaboration meeting, Kraków, 10 May 2018

HGTD: documentation � Open LHCC presentations, 1 Dec 2017 ATLAS HGTD Overview, Motivation, Performance and Physics (D. Zerwas) � https: //indico. cern. ch/event/679087/contributions/2781884/subcontributions/242188/att achments/1568715/2473482/171201_LHCC. pdf ATLAS HGTD Design and R&D (L. Serin) � https: //indico. cern. ch/event/679087/contributions/2781884/subcontributions/242189/att achments/1568718/2473486/HGTD-LHCC-Final. pdf � Public plots (with detailed captions) � https: //twiki. cern. ch/twiki/bin/view/Atlas. Public/LAr. HGTDPublic. Plots#HGTD_ID R_figures Technical proposal W. Kozanecki https: //cds. cern. ch/record/2310228/ [submitted to LHCC for review, may not yet be accessible to non-ATLAS colleagues] Slide 21 FCAL Collaboration meeting, Kraków, 10 May 2018

In closing… possible contributions? � BCM' & LUCID The designs are still at a rather early stage… � with �. . first prototype(s) either planned for late 2018, or already under test but with an advantage: pull on 8 -10 y of experience with L determination at the LHC Both groups are small; both would welcome new collaborators � BCM' � (Ljubljana, Manchester, Ohio State, Toronto) contacts: Marko. Mikuz@cern. ch, william@physics. utoronto. ca � LUCID-3 � � (Alberta, Bologna, Lund) contacts: Vincent. Hedberg@cern. ch, giacobbe@bo. infn. it The HGTD is a much larger project (21+ Institutions) Design well advanced (Tech. Proposal @ https: //cds. cern. ch/record/2310228/ ) � however � no real-world experience with L determination by the PCC method attractive long-term investment with immediate pay-off (Run-2 analysis + fbk to HGTD designers): develop a PCC-based L determination using already existing ATLAS data Wide spectrum of potential contributions from new collaborators � e. g. sensor testing, electronics, module assembly, online monitoring/DAQ software, offline performance simulation & analysis, … � � contacts: Ana. Henriques@cern. ch, Laurent. Serin@cern. ch New ideas/technologies? all 3 strategies above face tough challenges. . . W. Kozanecki Slide 22 FCAL Collaboration meeting, Kraków, 10 May 2018