Monte-Carlo simulations for Drell-Yan in COMPASS

Transcription

1 Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans, LIP-Lisbon on behalf of the COMPASS Collaboration 26 th February 214 COMPASS Co-financed by: Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.1

2 Overview Polarized Drell-Yan in COMPASS Drell-Yan process generation he COMPASS experiment Simulation of the detector Geometrical acceptances Experimental resolutions Expected event rates and asymmetries statistical errors UML extraction of asymmetries Summary Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.2

3 MD studies in COMPASS In CERN transverse momentum dependent PDFs (MDs) can be accessed either from semi-inclusive DIS (SIDIS), or from Drell-Yan processes, using a transversely polarized target: µ µ N SIDIS γ* Final State Interactions h X q q γ DY Initial State Interactions Sivers and Boer-Mulders MDs are -odd, thus the prediction: * µ µ + f 1 (DY ) = f 1 (SIDIS) h 1 (DY ) = h 1 (SIDIS) he -odd effect is a manifestation of non-zero quarks orbital angular momentum. he sign change observation is considered a crucial test of non-perturbative QCD and the MDs approach. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.3

4 Polarized Drell-Yan For a transversely polarized target, one can calculate the cross-section asymmetry between the 2 possible spin configurations. he single polarized Drell-Yan cross-section can be written as: dσ d 4 qdω = α2 em Fq 2 ˆσ U A j`1 + A 1 ± S» (D [1] A sin φ S U cos2 θ + D [sin 2θ] A cos φ U + D [cos 2 θ]ãsin φ S ) sin φ S 2φ cos φ + D [sin 2 θ] Acos U cos 2φ + D [sin 2θ] (A sin(φ+φ S) sin(φ + φ S ) + A sin(φ φ S) sin(φ φ S )) + D [sin 2 θ] (Asin(2φ+φ S) sin(2φ + φ S ) + A sin(2φ φ S) sin(2φ φ S )) ff A: azimuthal asymmetries D: depolarization factors S: target spin component F = 4 q (P a P b ) 2 M 2 am 2 b ˆσU : cross-section surviving integration over φ and φ S. A: acceptance function Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.4

5 Azimuthal asymmetries he parameters entering Lam-ung relation (1 λ 2ν = ) are: λ = A 1 U µ = A cos φ U ν = 2A cos 2φ U he collinearity hypothesis: λ = 1, µ = ν =, was seen to be violated by experiments NA1(CERN) and E615(Fermilab) in the 8 s. he remaining azimuthal asymmetries contain a convolution of 2 MDs: A cos 2φ U : h 1 (π) h 1 (p); A sin φ S : f 1 (π) f 1 (p); sin(2φ+φ A S ) : h 1 (π) h 1 (p); sin(2φ φ A S ) : h 1 (π) h 1(p). All are expected to be sizeable in the valence quark region. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.5

6 he MC chain Drell-Yan process generation: Pythia Particles crossing the COMPASS spectrometer simulation: GEAN3 Reconstruction as for real data Analysis: extraction of acceptances, efficiencies and resolutions Note: he simulations discussed here use Pythia as Drell-Yan generator, thus polarization effects are not included in this MC. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.6

7 Unpolarized Drell-Yan in Pythia Pythia is multi-purpose Lots of parameters Different versions have different parameter defaults Pythia is LO. NLO can be partially simulated via parton showers wo known problems with using Pythia for Drell-Yan: cross-section too low dilepton p spectrum must be adjusted by a tuning of the fragmentation process Drell-Yan data obtained over the past 3 years by different experiments can be used to tune Pythia for the COMPASS (unpolarized) case. When COMPASS data will be available, a more precise tuning may be done. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.7

8 Drell-Yan cross-section In these simulations: Keep parton showers OFF Use LO PDF sets In these conditions, the departure wrt experimentally measured cross-section is well-known (K DY factor): Kenyon, 1982 important to estimate event rates in COMPASS. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.8

9 Drell-Yan p spectrum p 2 is different for DY in proton or in pion induced collisions: proton π τ =.22 τ =.28 Kenyon (1982) parametrization for pions, at τ = xπ x p =.28, applied to the COMPASS s: p 2 = s = 1.6 (GeV/c)2 NA1, using π beam at 194 GeV on a 12 cm long W target obtained a sample of DY events in 4.5 M 8.5 GeV/c 2 with p = 1.1 GeV/c (NA1, CERN-EP/87-199). Since COMPASS uses NH3 target, expect very small p broadening effect. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.9

10 uning of Drell-Yan dimuon p in Pythia In the Pythia fragmentation, the hadrons primordial k distribution is assumed gaussian. he default value for the width of this gaussian is σ k = 2 GeV/c (PARP(91)), leading to a distribution too wide. Dimuon transverse momentum Entries 25 Mean Pythia default: PARP(91)=2 RMS Pythia tuned: PARP(91)=.9 Entries 25 Mean 1.13 RMS.5786 GeV/c + p 4 < M < 9 GeV/c p (GeV/c) By setting σ k =.9 GeV/c, one obtains an average dimuon p = 1.1 GeV/c, identical to the one reported by NA1. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.1

11 uning of other parameters in Pythia he minimum invariant mass of the remnant hadronic system (via PARP(111)) is also tuned. his parameter sets an upper limit to the momentum fraction of the interacting parton: x < 1 2 PARP(111)/E CM. x π 6 Entries 25 DY: 19 GeV/c + p Mean < M < 9 GeV/c2 RMS Entries 25 Mean RMS Pythia default: PARP(111)=2 Pythia tuned: PARP(111)= x π By changing this parameter, one extends to x π <.9. Another consequence is that the cross-section increases by 2%. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.11

12 PDF sets used PDF sets from LHAPDF are used. Here we choose LO PDF sets, for consistency. pion: GRVPI LO (set 252) nucleon: GRV98 LO (set 86) esting dependence of simulation on the pion PDF set (reaction GeV + p): PDF set σ 4<M µµ<9 DY (nb) p RMS p x π RMS xπ GRVPI OW-PI set OW-PI set esting dependence of simulation on the nucleon PDF set (reaction GeV + p): PDF set σ 4<M µµ<9 DY (nb) p RMS p x p RMS xp GRV MRS98 set Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.12

13 Pythia tuning summary parameter default tuned meaning MSP(61) 1 initial PS MSP(71) 1 final PS MSP(51) 86 PDF set for target MSP(53) 252 PDF set for beam PARP(111) 2 1 minimum mass of remnant PARP(91) 2.9 width of k gaussian In these conditions, the Drell-Yan cross-sections obtained in Pythia for the dimuon mass range generated 4 M µµ 9 GeV/c 2 is: 19 GeV + p:.145 nb 19 GeV + n:.74 nb Note: the beam momentum was implemented with a gaussian spread of 2% (as we have it in COMPASS) already at generator level (PARP(171)). Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.13

14 Generated distributions: proton vs neutron Dimuon mass Dimuon transverse momentum Dimuon longitudinal momentum 25 Entries 25 Mean p target n target RMS.994 Entries 25 Mean RMS Entries 25 Mean 1.15 RMS.5837 Entries 25 Mean 1.96 RMS Entries 25 Mean RMS Entries 25 Mean RMS M (GeV/c2) x1 (pion) Entries 25 Mean.3651 RMS.186 Entries 25 Mean.3943 RMS p (GeV/c) x2 (nucleon) Entries 25 Mean.252 RMS.138 Entries 25 Mean.2212 RMS p (GeV/c) L Entries 25 Mean.1131 RMS.2985 Entries 25 Mean.1731 RMS.2884 x F x x x F Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.14

15 he COMPASS spectrometer he COMPASS experiment at CERN uses a secondary beam produced from the SPS extracted protons in a production target. Polarized target: NH 3 π GeV/c π N u N d arget spin reversal every few days N u N d Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.15

16 Beam and target he polarized Drell-Yan measurement in COMPASS will use a transversely polarized target. Due to the dipole field in the target (used to keep the transverse polarization), the beam must enter the hall with a tilt, in order to cross the target in all its extent. o avoid local heating of the target, that would destroy polarization, the beam spot must be wide (σ x,y = 1 cm). he target cells are cylinders with 2 cm radius, to intercept most of the beam (87%). he target is 11 cm of solid state ammonia in a bath of liquid He (.57 λ π int ). he non-interacted beam (in this target) will be stopped by the tungsten beam plug (or interact eventually with thin nuclear target placed upstream of it). he DY process has low cross-section, thus the need for high luminosity. But the location of the experimental hall at ground level, the beam line characteristics and the need to limit the target heating, restrict the beam intensity to I = 1 8 π/s. COMPASS beam comes in 9.6 s long spills (flat top), in a 33.6 s extraction cycle. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.16

17 Spectrometer simulation he spectrometer simulation (propagation of particles in matter, and detectors response) is done using different packages, depending on the simulation purpose: For Drell-Yan geometrical acceptances, data reconstruction studies, and experimental resolutions GEAN3 For detector occupancy and radiation conditions studies FLUKA For cross-checks (but using simplified geometry) GEAN4 arget Cells Hadron Absorber Concrete Shielding Muon SM1 RICH ECAL1 HCAL1 Filter SM2 DC1 DC4 RichWall MuonWall Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.17

18 Generated events spread along target A Pythia sample of (π, p) and (π, n), according to the proportions in NH 3 : 1 17 σdy πp σdy πn is prepared, generating Drell-Yan events with M gen > 3.5 GeV/c 2. he wider generation window is to be able to select 4 < M rec < 9 GeV/c 2 reconstructed events with margin for experimental smearing. he DY Pythia events are spread along the target according to the pion interaction length in NH 3. Z vertex distribution 4 35 Generated events Events in the acceptance Z vtx (cm) Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.18

19 COMPASS: DY and SIDIS measurements In COMPASS we have the unique opportunity to access, using the same spectrometer, MDs via the 2 different processes: DY and SIDIS. SIDIS and DY measurements have an overlapping region. Check the prediction for Sivers and Boer-Mulders sign change when accessed from these 2 processes. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.19

20 Acceptance he global geometrical acceptance of DY dimuons with 4 < M < 9 GeV/c 2 is 39%..6 Dimuon mass acceptance.6 Dimuon p acceptance.6 Z vtx acceptance M (GeV/c2) p (GeV/c) (cm) Z vtx 1 x π acceptance 1 x N acceptance 1 x F acceptance x π x N x F DY with π beam on fixed target: u-quark dominance with valence quarks interacting. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.2

21 Angular acceptances Collins-Soper frame ransverse target spin and φ S 1 cos θ CS acceptance.6 φ CS acceptance.6 φ S acceptance cos θ CS φ CS he spectrometer itself introduces a modulation in φ CS that must be taken into account. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.21 φ S

22 Acceptance per zone he acceptance can be analysed according to the zone where µ + and µ (respectively) are accepted in the spectrometer: at large polar angle (LAS) or at small polar angle (SAS)..3 φ CS acceptance: LAS-LAS.2 φ CS acceptance: LAS-SAS he geometrical acceptance is 39% 2 muons at Large Angle (LAS): φ CS φ CS acceptance: SAS-LAS φ CS φ CS acceptance: SAS-SAS 22% 2 muons at Small Angle (SAS): 2% one muon in LAS and another in SAS: 18% φ CS φ CS Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.22

23 Spectrometer related uncertainties he presence of an hadron absorber downstream of the polarized target is essential to keep detector occupancies at reasonable level (< 5%), and limit the combinatorial background of muons resulting from pion and kaon decays. Muon identification is also ensured in the spectrometer by requiring the tracks to cross heavy material walls (iron or concrete). he spectrometer includes 2 sets of calorimeters, also resulting in a significant material budget. A muon track in the acceptance of either spectrometer will have crossed at least 1 radiation lengths important multiple scattering. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.23

24 Resolutions (rec) x F 1.8 Smearing in x F (rec) x p Smearing in x p (rec) p 6 5 Smearing in p x F (gen) x p (gen) p (gen) x F resolution 8 x p resolution p resolution 5 4 Entries Mean RMS Entries Mean RMS Entries Mean.114 RMS x F (rec) - x (gen) F x p (rec) - x (gen) p p (rec) - p (gen) (GeV/c) Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.24

25 Vertex position resolution Z vtx (rec) (cm) Z_ (rec) vs Z (gen) vtx vtx 6 5 Z vtx resolution Entries Mean RMS (cm) Z vtx σ Resolution along target Z vtx (gen) (cm) Z vtx (rec) - Z (gen) (cm) vtx (cm) gen Z vtx Z vtx Generated events Accepted events Reconstructed events he resolution of the primary vertex position is important, since one must distinguish the events coming from each of the target cells, which are polarized in opposite directions Z vtx (cm) Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.25

26 Angular resolutions cos θ CS resolution φ CS resolution φ S resolution 35 3 Entries Mean Entries Mean Entries Mean RMS RMS RMS cos θ (rec) - cos θ (gen) CS CS φ (rec) - φ (gen) (rad) CS CS φ (rec) - φ (gen) (rad) S S Azimuthal resolutions in the order of 17 mrad are obtained from the simulation of DY events with 4 < M µµ < 9 GeV/c 2 (pure Drell-Yan MC, no pile-up included). Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.26

27 Expected event rates he expected events rates for Drell-Yan in the mass range 4 < M µµ < 9 GeV/c 2 are: With a beam intensity of Ibeam = particles/second, one expects 1/day. In 14 days, 14 events. 216 events in one year, if Ibeam = 1 8 particles/second. Note: Used K DY -factor=2. Assumed an SPS super-cycle of 33.6 s. Assumed 8% SPS efficiency (beam availability), and 85% efficiency in running of the spectrometer. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.27

28 Asymmetries statistical error he expected statistical errors in the asymmetries are given by: r cos 2φ 2 δ AU = 2 N ; δ Asin φ S = 1 fs r 2 N ; δ Asin(2φ±φ S) = 2 fs r 2 N Asymmetry Dimuon mass (GeV/c 2 ) uncertainties 4 < M µµ < 9 cos 2φ δ AU.56 δ A sin φ S.142 δ A sin(2φ+φ S).284 δ A sin(2φ φ S).284 Possibility to study the asymmetries in several x F or p bins. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.28

29 Extraction of azimuthal asymmetries he preferred method to extract the azimuthal asymmetries is the unbinned maximum likelihood method (UML): One maximizes the likelihood for each event characterized by (θ, φ, φ S ) to follow a probability density function: P ± cell = a± cell g± with a ± cell : acceptance function; g± : physics modulation function. In the extended maximum likelihood fit performed, the asymmetries are parameters of the fit. he method was used in COMPASS, in particular for a simultaneous extraction of Sivers and 7 other SIDIS transverse spin asymmetries (Phys.Lett.B 692 (21) 24). It is now adapted to the Drell-Yan case. A toy-mc injecting sizeable asymmetries has shown the ability of the UML to recover these without bias. More studies are ongoing. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.29

30 o conclude COMPASS polarized Drell-Yan measurement will start in a few months, in October 214. Physics data taking continue during 215 (full year). A second year of DY data-taking is planned, possibly in 218. After 1 year of data-taking: expected statistical error in the Sivers asymmetry 2% (systematic errors will be smaller). Unpolarized Drell-Yan MC was done using Pythia generator and a GEAN3 simulation of the spectrometer, to extract acceptances and resolutions in all the relevant variables. he method to extract asymmetries (UML) in the DY case was implemented, and tested with a toy-mc. he COMPASS measurements will contribute to the common effort of extracting the MDs, namely Sivers, Boer-Mulders and pretzelosity, as well as the transversity PDF. Monte-Carlo simulations for Drell-Yan in COMPASS C. Quintans p.3