LUMINOSITY MEASUREMENT AND CALIBRATION AT THE LHC W
LUMINOSITY MEASUREMENT AND CALIBRATION AT THE LHC W. Kozanecki, CEA-Saclay Lund, Sweden, 13 June 2016
Outline 2 Introduction: the basics Relative-luminosity monitoring strategies Absolute-luminosity calibration strategies – & their challenges Instrumental systematics in the high-L environment Achieved precision on L: Run 1 & early Run 2 Summary Selected bibliography W. Kozanecki 13 June 2016
3 Basics of L measurement: Rate = *L = number of inelastic pp collisions per bunch crossing nb = number of colliding bunch pairs fr = LHC revolution frequency (11245 Hz) inel = total inelastic pp cross-section (~80 mb at 13 Te. V) = acceptance x efficiency of luminosity detector eff = # visible (= detected) collisions per bunch crossing eff = effective cross-section = luminosity calibration W. Kozanecki 13 June 2016
The experimental environment 4 LHC fill 4560: 2244 bunches, 25 ns apart W. Kozanecki L-calibration sessions & other 13 June 2016
Pileup! 5 L-monitoring algorithms: rate = eff * L? Event- (or 0 -) counting algorithms: bunch-by-bunch (bbb) � � an “event” is a bunch crossing (BX) where a given condition is satisfied, e. g. : Event. OR = at least 1 hit in either the A arm of a luminometer, or the C arm, or both Event. AND (aka A. C) = at least 1 hit in the A-arm AND at least 1 hit in the C arm count the fraction of BX with zero events from Poisson probability � L is a monotonic (but non-linear) function of the “event” rate ( Appendix) examples: V 0 A. C (ALICE), LUCID_Bi_ORA (ATLAS), ≥ 2 VELO tracks (LHCb) Hit-counting algorithms (bbb) � count the fraction of channels hit in a given BX (e. g. CMS: # pixel clusters) W. Kozanecki linearity vs. depends on technology, granularity, thresholds, . . . 13 June 2016
6 L-monitoring: instrumental strategies ALICE ATLA S CMS / W. Kozanecki TOTE Preferred offline ( Lphys) luminometer Main addtn’l luminometers: offline corrections + systs. V 0 (scintillator arrays): A. C T 0 (Cherenkov arrays): A. C + DT cut ZDC (had. cal): Event. OR (Pb-Pb only) AD (“diffractive” scint. arrays): A. C LUCID-2 (quartz Cherenkov +PMTs): Event. OR [single-arm inclusive] Si tracker: track counting EM/Fwd calorimeter: current in LAr gaps TILE hadronic calorimeter: PMT currents Si tracker: pixel-cluster counting Pixel L telescope: evt cntg [3 -fold AND] 13 June 2016 Muon Drift Tubes : track-segment
7 Absolute-L calibration: actual strategy Principle: eff = Rcollisions /L (beam parameters) van der Meer scans: L = f (Sx, Sy, n 1, n 2) � Sx, y from currents S ~ ( 12+ 22)1/2 R vs. beam sep. (dx, dy); n 1, n 2 = bunch exploit luminous-region evolution in scan: (dx, dy) dependence of (x, y, z) position & width of luminous region (aka “beamspot”) + Beam-gas imaging: L = f( x 1, y 1, x 2, y 2, z, fc, n 1, n 2) � extract single-beam parameters from (x, y, z) distribution of reconstructed p-gas & pp evt vertices W. Kozanecki 13 June 2016 (stationary beams)
L calibration: van der Meer scans 8 Measure visible interaction rate μeff as a function of beam separation d The measured reference luminosity is given by effpeak with Sx, y = integral under the scan curve / peak This allows a direct calibration of the effective cross section σvis for each luminosity detector/algorithm scan effective crosssection width s bunch peak rate population s Hor. beam separation dx [mm] Key assumption: factorization of luminosity profile W. Kozanecki 13 June 2016
L calibration: beam-gas imaging (BGI) 9 Extract p-density distributions r 1, 2 (x, y, z) from simultaneous fit to 3 D distributions of B 1 -gas, B 2 -gas & pp collision vertices Each beam modelled by non-factorizable sum of 3 D Most critical: gaussians vertexing L = 2 c fr n 1 n 2 cos f/2 ∫ r 1(x, y, z, t) r 2(x, y, efficiency z, t) dx dy dz dt & resolution LHCb only! Typical L W. Kozanecki 13 June 2016
10 L calibration: lessons from LHC run 1 The central role of beam dynamics �L calibs: widely-spaced low-I bunches, no high- trains! injected-beam quality, parasitic beam-beam (+ - dependence) � orbit drifts can cost 2 -3% of bias &/or systematics � beam-beam deflections & dyn. b must be corrected for � non-factorization: an often dominant uncertainty Luminosity instrumentation: redundancy essential! W. Kozanecki 13 June 2016
Evidence for non-factorization ? 11 Sx, y in on- vs. off-axis scans � Σx , Σy in offset scans larger than onaxis [in this example: 10 -20%] � varies from fill to fill empirical tailoring of beams in L, xyinjectors (x, y luminous size) in on-axis � scans W. Kozanecki � Vertical luminous width depends on horizontal separation (and vice-versa) [in this example: ~20%] � correct usingle-beam parameters from combined fit to beam -separation dependence of L and of luminous-region observables: 13 <x, y, z> June 2016 L
! Non-factorization: impact eff diff, (w/o – with) non-factorztn [%] 12 LHCb: factorization bias LHCb ± 1% beam imaging LHCb: vis consistency (non-factorizable beam model) ATLAS vd. M scans Ref. W. Kozanecki without/with nonfactorization correction Magnitude of factorization bias: • fill-dependent • time-dependent within a fill
Non-factorization corrections ! 13 Accounting for non-factorization • Direct measurement: beam-gas imaging (LHCb) • Factorizable (= “naïve”) vd. M analysis + non-factorization correction from spatial dependence of Nvtx(x, y, z) [~L (x, y, z)] Ø luminous-region evolution analysis (ALICE, ATLAS) Ø beam-beam imaging (CMS) • Non-factorizable vd. M analysis (ALICE, ATLAS, CMS, LHCb) Ø coupled vd. M fits to L (dx), L (dy) Consistency btwn methods: filldependent Associated systematic: 0. 5 – 3%
14 Aging pains: a price to pay for high L Long-term drift correction 2012 pp ± 1% 2012 pp ATLAS (2012, 8 Te. V pp): • BCM (diamond) response degraded by ~ 2% over the year • corrected using either calorimeteror track-based L (systematic: 0. 3%) • resulting relative stability < 0. 5% across 5 independent luminometers syst = 0. 5 % 2012 pp
15 Long-term consistency of L measurements syst = 1. 0% LPCC / LDT In-situ Bi 2015 pp, crucial! 13 calibration Te. V ( Appendix) LT 0 / LV 0 2015 pp, 13 Te. V syst = 1. 0% syst = 0. 12% Lcalo / Ltrack 2013 p. Pb, 5 Te. V 2012 pp, 8 Te. V
16 Total L systematics: vd. M or BGI - & more Appdx Adapted from ref. [17], Table 14 W. Kozanecki 13 June 2016
L performance summary (new since 2014) 17 LHCb ATLA S CMS LHCb ALICE ATLA S CMS LHCb Running period 2011 pp 2012 pp 2015 pp √s [Te. V] 7 8 8 8 13 13 L /L [%] 1. 7 1. 9 2. 5 1. 2 3. 4 Prelim. 2. 1 ALICE ATLA S CMS LHCb Running period 2010/2 011 Pb. Pb 2013 p-Pb / Pb-p 2013 pp √s. NN [Te. V] 5 5 5 2. 76 3. 1 3. 7 L /L [%] W. Kozanecki 5. 8/4. 2 3. 7/3. 4 2. 7 3. 6/3. 4 2. 3/2. 5 2. 7 3. 9 Prelim. 2. 2 2016 13 June
Why does L matter? some examples… 18 Physics measurement ALIC E √s. NN [Te. V] ssystt ssyst. L ot [%] Ref. Total inelastic pp cross-section 7 +4. 5 -7. 2 3. 6 [1] EM dissociation cross section in Pb-Pb collisions 5 6. 5 5. 8 [3] Top-quark pair production cross-section 7 3. 5 2. 0 [7] Top-quark pair production ratio, 8 Te. V / 7 8/7 3. 9 3. 7 [7] Top-quark pair production cross-section 13 5. 5 2. 6 [11] Fiducial inclusive Z cross-section 13 3. 3 2. 7 [12] Forward Z+jet production 8 4. 8 1. 2 [15] Prompt D 0 production cross-section 13 5. 3 3. 9 [16] ATLA S Te. V CMS LHCb W. Kozanecki 13 June 2016
Conclusions 19 The absolute precision of the integrated L typically lies in the 2 -3 % (3 -6%) range for topenergy pp (HI) � main contributors to the uncertainty beam dynamics: phase-space control (non-factorization, satellites, ghosts), beam-beam calibration strategy instrumental linearity vs & Ltot (4 orders of magnitude!) instrumental stability & aging (more difficult for high-L expts) Run 2 already is a challenge; Run 3 is Terra Incognita W. Kozanecki the “ 2% wall” is unlikely to be broken - except for LHCb: 13 June 2016
Selected bibliography: ALICE 20 ALICE Collaboration [1] Measurement of inelastic, single- and double-diffraction cross sections in proton–proton collisions at the LHC with ALICE, Eur. Phys. J. C 73 (2013) 2456 [2] Performance of the ALICE Experiment at the CERN LHC, Int. J. Mod. Phys. A 29 (2014) 1430044 [3] Measurement of the Cross Section for Electromagnetic Dissociation with Neutron Emission in Pb-Pb Collisions at √s. NN = 2. 76 Te. V, PRL 109, 252302 (2012) [4] Measurement of visible cross sections in proton-lead collisions at √s. NN = 5. 02 Te. V in van der Meer scans with the ALICE detector, JINST 9 (2014) 1100 [5] ALICE luminosity determination for pp collisions at √s = 13 Te. V, ALICE-PUBLIC-2016 -002 W. Kozanecki 13 June 2016
Selected bibliography: ATLAS 21 ATLAS Collaboration [6] Improved luminosity determination in pp collisions at √s =7 Te. V using the ATLAS detector at the LHC, Eur. Phys. J. C 73 (2013) 2518 [7] Measurement of the ttbar production cross-section using e events with b-tagged jets in pp collisions at √s = 7 and 8 Te. V with the ATLAS detector, Eur. Phys. J. C 74 (2014) 3109 [8] Luminosity determination in pp collisions at √s =8 Te. V using the ATLAS detector at the LHC, to be submitted to Eur. Phys. J. C W. Kozanecki 13 June 2016
Selected bibliography: CMS 22 CMS Collaboration [9] CMS Luminosity Based on Pixel Cluster Counting - Summer 2013 Update, CMS-PAS-LUM-13 -001 (Sep. 2013) [10] Luminosity Calibration for the 2013 Proton-Lead and Proton Data Taking, CMS-PAS-LUM-13 -002 (Jan 2014) [11] Measurement of the top quark pair production cross section using e events in proton-proton collisions at √s = 13 Te. V with the CMS detector, CMS PAS TOP-16 -005 (March 2016) [12] Measurements of inclusive and differential Z boson production cross sections in pp collisions at √s = 13 Te. V, CMS PAS SMP 15 -011 (March 2016) [13] CMS Luminosity Measurement for the 2015 Data Taking Period, CMS-PAS-LUM-15 -001 (March 2016) W. Kozanecki 13 June 2016
23 Selected bibliography: LHCb & review LHCb Collaboration [14] Precision luminosity measurements at LHCb, JINST 9 (2014) P 12005 [15] Measurement of forward W and Z boson production in association with jets in proton-proton collisions at √s = 8 Te. V, JHEP 05 (2016) 131 [16] Measurements of prompt charm production cross-sections in pp collisions at √s = 13 Te. V, JHEP 03 (2016) 159 General review [17] P. Grafstrom & W. Kozanecki, Luminosity determination at proton colliders, Progr. Nucl. Part. Phys. 81 (2015) 97– 148 W. Kozanecki 13 June 2016
24 W. Kozanecki Supplementary Material 13 June 2016
25 W. Kozanecki 13 June 2016
26 ATLAS: redundancy many L msmts! Note: all luminometers are independent of TDAQ (exc. trk-, vtx- & Zcounting) LUCID – Luminosity measurement using a Cherenkov Integrating Detector (bbb) LUCID “ATLAS 2 preferred” for 13 Te. V pp data + Z counting (relative-L checks) MPX/TP X + Vertex counting + Track counting (both bbb) W. Kozanecki BCM – Beam Conditions Monitor (bbb) “ATLAS-preferred” for 7 & 8 Te. V pp data 13 June 2016
S ystem for M easuring the O verlap with G as Several relative lumi counters (monitors) ~0. 5 use zero counting Gas injected into beam pipe Beam Gas Imaging Ø Ø Our main detector for ref (calibrated cross section) • Vtx counter • Track counter Two calibration methods: BGI & VDM Van der Meer scans Si strip detector
A key issue: handling the pile-up 28 Event- (or zero-) counting: measure fraction of BX with no L counts If is the average # of inelastic pp collisions/BX, and NOR (NAND) the total # of OR (AND) “events” over Norbits, then (for 1 colliding bunch pair) the Poisson probability P to detect an “event”/BX is L ~ m = - ln(1 – POR) /e. OR ~ POR only when << 1 Hit-counting: measure fraction of empty (or occupied) channels similar formalism to event counting, but more sensitive to instrumental effects (drift, migration) highly linear in the limit of very large granularity (e. g. CMS: pixel-cluster counting) W. Kozanecki 13 June 2016
29 Absolute-L calibration: the initial plan Optical theorem + elastic pp scattering at low t � d. Rel/dt + Rinel (Rtot = Rel + Rinel) [TOTEM + ALFA] � d. Rel/dt in Coulomb-interference region [ALFA + TOTEM] d el/dt, el msrd at √s = 7 & 8 Te. V [ALFA, TOTEM] W. Kozanecki 13 June 2016
Beam-beam effects 30 Two beam-beam effects: beam-beam deflection and dynamic-β bias σvis if not corrected � < 0. 5% Pb. Pb, 1 - 2% for 8 Te. V pp and around 4% for 5. 4 Te. V pp � The interaction of the two beams during a scan causes distortion to the https: //indico. cern. ch/get. File. py/access? contrib. Id=10&session. Id=14&res. Id=1 scan curve � See details in ATLAS week talk Beam-beam deflection &material. Id=slides&conf. Id=206600 Dynamic-β size of effect exaggerated for demonstration (size of effects exaggerated for demonstration) True beam separation greater than nominal separation W. Kozanecki Beams focus/defocus each other by an amount that is a function of separation 13 June 2016
? Non-factorisation correction procedure ➜ estimate the true luminosity (i. e. unbiased by non-factorisation effects) ➜ estimate correction for non-factorisation, R, with an associated uncertainty - fit - data beam spot xposition Single beam profiles are parameterised by fitting the beamseparation dependence of the luminosity & of the beamspot displacement and width during a vd. M scan. This allows to: The [ATLAS/ALICE] procedure above is closely related to the “beam-beam imaging” scans [pioneered by LHCb & now used by CMS] in which one beam is scanned transversely as a probe across the other. W. Kozanecki - fit - data Beam separation (xscan) Specific L (arb. u. ) Beam spot xwidth 31 - fit - data Beam separation (x 13 June 2016
Non-factorization correction: beam-beam imaging 32 Pull distribution to cumulative event-vertex distributions for 2 single-beam models: factorizable non-factorizable Example of pull distributions of the fitted single-beam model of the single-gaussian (factorizable, left) and double-gaussian (non-factorizable, right) type to the vertex distribution accumulated during scan Y 3 of bunch pair 1631. (Caption adapted from Fig. 11 of CMS-PAS-LUM-2015 -001) W. Kozanecki 13 June 2016
Beam-conditions-dependent biases Calibration-transfer correction 33 D ( ~ time in Fall 2012) W. Kozanecki ATLAS (2012 & 2015): • luminometer response shifts by D = 24 % between vd. M (low L, bunches far apart) and physics (high L, 50 or 25 ns trains) • magnitude & sign ≠ for diamond- & PMT-based luminometers • track-counting & calo-based L crucial to “transfer” calibration vd. M high L • associated systematic: 1. 4 % (0. 9%) for 8 Te. V pp [2012] (13 Te. V pp [2015]) CMS (2012 & 2015) • qualitatively similar effects seen in CMS diamond detector – but no visible impact bec. main luminometer = Si pixel detector ALICE & LHCb: lower , L - less of an 2016 13 June
34 vd. M-calibration systematics: pp examples Example breakdowns of the fractional systematic uncertainties affecting the determination of the visible pp cross-section σvis by the vd. M method at the LHC. Blank entries correspond to cases where the uncertainty is either not applicable to that particular experiment or scan session, is considered negligible by the authors, or is not mentioned in the listed reference. Source: Progr. Nucl. Part. Phys. 81 (2015) 97– 148, Table 12 W. Kozanecki 13 June 2016
35 BGI-calibration systematics: example Systematic uncertainties affecting the LHCb absolute luminosity calibration by the BGI method at √s = 8 Te. V [31, 127]. Source: Progr. Nucl. Part. Phys. 81 (2015) 97– 148, Table 13 W. Kozanecki 13 June 2016
36 vd. M-calibration systematics: p. Pb example Source: ALICE Collaboration, JINST 9 (2014) 1100, Table 3 W. Kozanecki 13 June 2016
37 LUCID-2 calibration using 207 Bi source Pulse-height distributions from a LUCID photomultiplier recorded in 13 Te. V runs on June 11 and 13, 2015 (blue) and in a calibration run recorded on June 25, 2015 (red). The physics runs were recorded using a random trigger, while the calibration run imposed a triggerthreshold requirement. The position of the peak created by Cherenkov photons produced in the quartz window of the photomultipliers is similar for highenergy particles from LHC collisions and low-energy electrons from the Bi-207 source. The vertical scale is set by the statistics of the low-μ run which has the smallest number of counts. The Bi-207 distribution has been arbitrary scaled down to a similar level. W. Kozanecki 13 June 2016
38 The hard path towards L stability: e. g. . Fractional difference in run-integrated luminosity between the LUCIDBi_Evt_ORA and track-counting algorithms. Each point corresponds to an ATLAS run recorded during 50 ns or 25 ns bunch -train running in 2015 at √s = 13 Te. V. Radioactive Bi-207 sources are used to monitor the gain of the PMTs in frequent calibration runs during the year. These pulse-height measurements are used to adjust the high voltage so that the gain remains constant throughout the year. In a second step, the Bi-207 calibrations are also used offline to correct the measured luminosity. The Figure shows the LUCID data before (red squares) and after the offline gain correction (black circles). Fractional difference in run-integrated luminosity between the LUCID_Bi_Evt_ORA and track-counting algorithms. By the end of the data-taking period, the cumulative increase in HV that had been applied during the year to keep the PMT gain constant, resulted in a significant decrease of the transit time. This, in turn, resulted in a loss of some events outside the timing window, and thereby in a decrease in detector efficiency. The impact of the transit time increase was different for different PMTs and was negligible for one of them. This PMT was used to correct the luminosity measured by the others. The Figure shows the LUCID data before (red squares) and after the transit-time correction (black 13 circles). June 2016
Total L systematics: ALICE example (pp, 13 Te. V) 39 Source: ALICE Collaboration, ALICE-PUBLIC-2016 -002, June 2016 W. Kozanecki 13 June 2016
Total L systematics: CMS example (pp, 13 Te. V) 40 W. Kozanecki Source: CMS PAS LUM-15 -001, March 2016 13 June 2016
- Slides: 40