Particle identification with the ALICE TOF detector at very high particle multiplicity

13 Pages·2004·1.4 MB·English

Eur Phys J C 32, s02, s165–s177 (2004) EPJ C direct
Digital Object Identifier (DOI) 10.1140/epjcd/s2003-01-0013-5
electronic only
Particle identification with the ALICE TOF detector
at very high particle multiplicity
ALICE TOF Group (Bologna-CERN-ITEP-Salerno):
A.N. Akindinov4, A. Alici1,2, F. Anselmo2, P. Antonioli2, Y.W. Baek3,6, M. Basile1,2, G. Cara Romeo2,
E. Cerron-Zeballos3,6, L. Cifarelli1,2, F. Cindolo2, F. Cosenza5, A. De Caro5, S. De Pasquale5, A. Di Bartolomeo5,
M. Fusco Girard5, V. Golovine4, M. Guida5, D. Hatzifotiadou2, A.B. Kaidalov4, S.M. Kiselev4, G. Laurenti2,
E. Lioublev4, M.L. Luvisetto2, A. Margotti2, A.N. Martemiyanov4, S. Morozov3,6, R. Nania2, P. Otiougova3,6,
A. Pesci2, F. Pierella1,2, P.A. Polozov4, E. Scapparone2, G. Scioli1,2, S. Sellitto5, A.V. Smirnitski4,
M.M. Tchoumakov4, G P. Vacca2, G. Valenti2, G. Venturi1,2, D. Vicinanza5, K.G. Voloshin4, M.C.S. Williams2,
S. Witoszynskyj3,6, B.V. Zagreev4, C. Zampolli1,2, and A. Zichichi1,2
1 Dipartimento di Fisica dell’Universit`a, Bologna, Italy
2 INFN, Bologna, Italy
3 CERN, Geneva, Switzerland
4 Institute for Theoretical and Experimental Physics, Moscow, Russia
5 Dipartimento di Fisica “E.R.Caianiello” dell’Universit`a and INFN, Salerno, Italy
6 World Laboratory, Lausanne, Switzerland
Received: 4 Aug 2003 / Accepted: 13 Sep 2003 /
Published Online: 13 Oct 2003 – (cid:1)c Springer-Verlag / Societa` Italiana di Fisica 2003
Abstract. A procedure developed to achieve particle identification in very high multiplicity conditions
using a complex time-of-flight system is illustrated in detail by simulating and studying the performance
of the ALICE TOF detector in a realistic scenario of Pb-Pb and p-p interactions at LHC.
1 Introduction – jets
– fluctuations
ALICE(ALargeIonColliderExperiment)[1]isanexperi- – direct photons
mentattheCERNLargeHadronCollider(LHC),optimi- – heavy-quarks and quarkonia
zed for the study of Pb-Pb collisions, at a centre-of-mass – coherent ultraperipheral phenomena.
energy of 5.5 TeV per nucleon pair. The prime aim of the
experiment is to study in detail the behaviour of nuclear
TheLHC(Pb-Pb)runningprogrammewillenabletheex-
matter at extreme densities and temperatures. The most
plorationofthisnewnuclearandsubnuclearphysicsfield,
striking expected phenomenon is a QCD phase transition
withrespecttothepresentlyongoingRHIC(Au-Au)run-
of nuclear matter into a deconfined state of quarks and
ning, with a factor of 28 jump in centre-of-mass energy
gluons, i.e. the QGP (Quark-Gluon-Plasma) state, with
and a factor of 4.5-12 jump in multiplicity and energy
chiral symmetry restoration and quark masses reduced to
densities.
thesmallbareones.ThisQGPwillpresumablyundergoa
Theparticleidentification(PID)powerofaverylarge
fastdynamicalevolutionintoadilutefinalhadronicstate.
time-of-flight (TOF) system covering the central rapidity
Tounderstandhowcollectivephenomenaandmacroscopic region (|y| ≤ 1) is of crucial importance in the ALICE
propertieswithmanydegreesoffreedominthedomainof
experiment. Pion/kaon/proton separation for the bulk of
strong(nuclear)interactionsemergefromthemicroscopic
charged particles in the intermediate momentum range
lawsofsubnuclearphysicsisthechallengeofALICE.This
(up to a few GeV/c) will provide an essential tool of in-
relies on a large number of observables [2], in particular:
vestigation for most of the relevant physics observables
– particle multiplicities listed above. Some of the basic physics motivations for
– particle spectra (chemical and kinetic freeze-out, tem- the ALICE TOF detector have been outlined elsewhere
perature and collective flow, hadron yields and chemi- [1,3,4,5].
cal composition, elliptic flow and early pressure) The large-coverage, powerful TOF detector we envi-
– particle correlations (Hanbury Brown-Twiss radii, in- sage building (see Fig. 1) should efficiently operate un-
tercept parameter, space-time information from radii der severe multiplicity conditions. It should have an ex-
and dynamics, lifetime of the collision, final state in- cellent intrinsic response and a low overall occupancy at
teractions, phase-space density, spin densities) the highest envisaged charged particle density. Present-
166 ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector
Fig. 1. Transverse view of the ALICE detector, where the 18 sectors of the TOF system can be seen. The inner and the outer
walls of the TPC are indicated.
day extrapolations from recent charged-particle multipli-
city√measurementsperformedatRHICinAu-Aucollisions ALIC TOF Detector (Strips) 5/ 3/ 3
at s = 130 GeV per nucleon pair are model dependent
andaffectedbywideuncertainties.HencedNch/dy inPb-
Pb at LHC could range from 3000 up to 8000 [2].
The ALICE TOF will be a barrel detector with, as
basic element, the double-stack MRPC (Multigap Resi-
stive Plate Chamber) strip. Each strip is equipped with
a mosaique of individual readout pads, thus making the
TOF barrel a large-coverage, high-granularity hodoscope.
Thestripsaregroupedintomodulesandtiltedinspacein
order to minimize the traversal path for particles coming
from the interaction vertex (Fig. 2).
The outstanding features of such a detector in terms
of efficiency, time accuracy and rate capability have been
extensively described elsewhere [3,4,5]. The main ALICE
TOF parameters are:
– radius: rTOF = 3.7 m; length: lTOF = 7.4 m;
– overall thickness: 30 cm, i.e. 20% of X0;
– polar-angle coverage: 45◦ ≤θ ≤135◦; full φ coverage;
– totalnumberofmodules:90,i.e.18(alongφ)x5(along
θ);
– totalnumberofMRPCstrips:1638,with15-19strips/ Fig. 2. TOF detector layout in AliROOT where the tilted
module; MRPC strips are visible.
– total number of readout pads: 157248, with 2 x 48
readout pads/strip;
– total surface: ≈170 m2; total sensitive surface: ≈150 – intrinsicMRPCtimeresolution:50-60ps;overalltime
m2;sensitivearea/strip:7.4x122cm2;individualrea- resolution: 120 ps (including all other additional sour-
dout pad size: 2.5 x 3.5 cm2; ces of time jitter);
– dead space: ≈10% of total surface; – intrinsic MRPC efficiency: 99%.
ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector 167
TheALICETOFwilloperateinthepresenceofasolenoid- event with a number of fired pads. In order to simulate
al magnetic field directed along the beam axis, with B = the effect of track reconstruction and fitting in the TPC1,
0.2-0.5 T. A charged particle density dNch/dy = 8000 at the GEANT3 information about track position and mo-
B = 0.2 T [3,4,5] represents the most ”prohibitive” con- mentum at the outer layer of the TPC is smeared with
ditions.InasingleSHAKER-generatedevent[1,3],corre- the following errors: σφ = 2 mrad, σθ = 2 mrad, σp/p =
sponding to a central Pb-Pb collision, these lead to 12000 2.1%, σz = 0.77 mm, σr = 0.62 mm. The momentum
φ
charged primary particles produced in the TOF detector dependence of the TPC tracking errors has been also in-
angularacceptance,outofwhich45%onlyreachtheTOF cluded[5].Thereconstructionefficiencyhasbeenassumed
surface, due to magnetic field, decays and interactions. to be 100% (the actual value being ≈90%).
On the other hand, a lot of secondaries are produced, For each particle tracked/reconstructed in the TPC
resulting in a very high background level. The total num- (called “TPC track” in the following), a statistically sig-
ber of fired pads (i.e. producing signals) is about 25000 nificant sample of probe tracks is generated and extrapo-
whichcorrespondstoadetectoroccupancyof≈15%.Yet, lated in the magnetic field from the TPC to the TOF.
only 25% of these pads are fired by particles which can The origin of these probe tracks is set at the outermost
be reconstructed in the ALICE TPC (Time Projection end-pointofthe“mother”-particletrackintheTPC(cor-
Chamber) [6]. responding to a hit in the last TPC pad row). At the
Moreover, there is a significant spatial gap between TPCend-point,theprobetracksdeviatefromthemother
the TPC and the TOF (rTOF − rTPC = 1.2 m) which is track according to a Gaussian distribution with standard
filledbytheALICETRD(TransitionRadiationDetector) deviationgivenbyθrms ,themeandeflectionangledueto
plane
[7], thus causing a high track deviation due to multiple multiple scattering2 through the outer wall of the TPC,
scattering. This tangles the task of matching the TPC- the whole TRD and the inner wall of the TOF (corre-
reconstructed tracks with the corresponding pads of the sponding to ≈6%, 14% and 1% of X0, respectively). The
TOF detector. overall multiple scattering effect is attributed at once to
InthispaperwewilldescribeboththeTPC-TOFtrack each probe track and a unique tracking step is performed
matching and PID procedures developed so far in order from the TPC to the TOF. No energy loss is considered
to evaluate the performance of the ALICE TOF system. for the probe tracks, which is a valid approximation here
These methods, tested via Monte Carlo simulations, are since the multiple scattering effect is dominant.
alreadysatisfactoryalthoughtheydonotrepresentanul- Thenumberofprobetrackspermothertrack(Nprobe)
timatesolutionmainlybecausetheTRDissimplytreated must be adequate (the larger θrms , the larger this num-
plane
as passive material. This detector will have in fact some
ber should be). We have chosen here Nprobe = 20, as
tracking capability (not considered herein) which will li-
a good compromise between statistical requirements and
kely help improving the TPC-TOF track extrapolation.
CPU time. Therefore a shower of probe tracks is associa-
Howevertheseproceduresareofgeneralvalidityandcould
ted in this way to each TPC track.
be applied to other experimental contexts.
Every probe track can cross a TOF pad (or a TOF
The AliROOT package [8] (developed within the fra-
dead region) and be assigned a weighting contribution:
mework of the ALICE Software Project) was used in all
weight1 = (cid:5)eff if the sensitive pad hit has fired, weight1
thecalculationsdiscussedinthefollowing.Thepackageis =1−(cid:5)eff ifthesensitivepadhithasnotfired.Aschematic
based on ROOT [9] and GEANT3 [10]. It will allow using
view of the probe-track shower is shown in Fig. 3. This is
the same environment in the simulation, reconstruction
a simplified picture since, as already mentioned, the TOF
andanalysisstagesoftheexperiment.AliROOTisinterfa-
surface inside each module is not planar but made of an
cedwithseveraleventgenerators(JETSET[11],PYTHIA
arrangement of tilted strips.
[12], HIJING [13], SHAKER [1], [3], etc.), thus providing
A particle crossing a TOF pad can sometimes induce
a complete set of instruments to fully simulate nucleus-
a small signal also on the neighbouring pad(s) and this
nucleus (or proton-proton) collisions.
signal can be slightly delayed (the lower the charge indu-
ced,thelongerthedelay).Inordertominimizethiseffect,
to each fired pad we assign an additional weighting fac-
2 Track matching tor,weight2 =qi2nduced,whereqinduced isthe(normalized)
Matching the information acquired by the TOF with the
tracking data of the TPC can be done using the so-called 1 AKalmanfilteringprocedurehasbeenrecentlyimplemen-
“probe-track method” described herein.
ted for track finding and reconstruction in the ALICE TPC
In our simulations [3,4,5] all particles generated in
and ITS (Inner Tracking System) [14], the latter being a si-
the Pb-Pb final state are tracked in the detector using
licon vertex detector surrounding the beam pipe. In ref. [15]
GEANT3. A charged (primary or secondary) particle im-
some preliminary results have been obtained concerning the
pinging on the TOF sensitive surface across a given pad extension of this Kalman filter to the TOF, with an option to
produces, with a given efficiency (εeff), a time signal cor- include the TRD as a tracking device.
responding to its flight time. This time is smeared accor- 2 Here the pion mass is assumed for all tracks. By taking
ding to the overall TOF resolution (σTOF). The values advantage of the particle identification provided by the TPC
εeff = 99% and σTOF = 120 ps have been used in the (using dE/dx measurements), a more accurate value of θprlmanse
presentexercise[5].HencetheTOFdetectorissetineach could be derived, whenever possible, at low momenta.
168 ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector
fired pads
distancetothenearesthit(cm) hhTTOOFFRRmmiinn
250
NNeenntt==55886688
200 MMeeaann == 33..77
150 RRMMSS == 22..119988
100
50
0
0 1 2 3 4 5 6 7 8 9 10
Rmin(cm)
deviationoftheextrapolatedtrackfromthehit(cm) hhTTOOFFddRRpp22
350 NNeenntt==22999922
300
p>0.5GeV/c MMeeaann ==00..99996644
250
extrapolated RRMMSS == 11..220022
200
probe 150
tracks 100
50
0
0 1 2 3 4 5 6 7 8 9(cid:1)R(cm)10
deviationoftheextrapolatedtrackfromthehit(cm) hhTTOOFFddRRpp11
90
NNeenntt==22551100
80
end−point of a TPC−reconstructed track 70 p<0.5GeV/c MMeeaann == 22..661144
60 RRMMSS == 22..115511
50
Fig. 3. A schematic view of the probe tracks extrapolated to 40
30
the TOF surface. 20
10
0
0 1 2 3 4 5 6 7 8 9(cid:1)R(cm)10
charge induced on the pad, so that its resulting weight is:
weight=weight1×weight2. Fig. 4. Rmin and∆RdistributionsforoneSHAKEReventat
For a given TPC track, we have then a set of TOF B = 0.2 T.
pads and TOF dead regions crossed by the probe tracks.
A TPC track may be matched with a dead region when
or successive crossings) present on the TOF surface. ∆R
noneofthepadscrossedbytheprobetrackshasfiredand
is determined by the multiple scattering from the TPC to
whenthenumberofprobetrackscrossingthedeadregion
theTOFwhileRmin dependsontheTOFoccupancy.The
is larger than the number relative to the non firing pads.
lower the ratio, the better will be the matching efficiency.
Otherwise,thetrackismatchedwithapad.ForeachTOF
Figure 4 shows the Rmin and ∆R distributions rela-
pad a total weight is defined as the sum of the weighting
tive to one SHAKER event at B = 0.2 T. This figure
contributions from all the probe tracks that cross it. The
has been obtained with primary and secondary particles
padwiththelargestweightischosentobetheTOFmatch from the central region (|θ −90◦| < 45◦) tracked in the
for the TPC track. Notice that since tracks are curled in
TPC. The narrow peak at Rmin ≈ 0 corresponds to δ-
themagneticfieldandmaysometimesspiralthismatchis
ray background. The fraction of tracks with ∆R > Rmin
chosen as a first-crossing match on the TOF surface, for
is ≈25%. Without the TRD [7] in between the TPC and
each track.
theTOF,thisfractiondecreasesbyafactor≈0.8.Inprin-
It is interesting to consider, in a qualitative way, the
cipletrackswith∆R>Rmin cannotbecorrectlymatched
factors influencing the matching of tracks extrapolated
and will presumably have a wrong time assignment.
from the TPC to the TOF with the corresponding fired
The momentum acceptances of the various steps of
pads.Asalreadypointedout,thisisalongdistanceextra-
the TPC-TOF matching procedure are shown in Figs. 5
polation, from the TPC outer layer at 2.5 m radius up to and 6 for primary particles of different species3 generated
theTOFsurfaceat3.7m.Letusassumethatthehitpoint in the |θ − 90◦| < 45◦ region, together with the origi-
(first crossing) of a track on the TOF sensitive surface
nal particle momentum spectra. The areas between the
could be exactly determined, i.e. without multiple scatte-
dashed/dotted-line histograms of Figs. 5 and 6 represent
ring or any other source of extrapolation uncertainty. Let
the corresponding acceptance losses. The loss “without
us call this point an ”ideal” hit point. On the other hand,
TPC tracks” refers to particles not reconstructed in the
if we consider the realistic tracking of a particle from the
TPC,enteringtheTPCdeadspaceordecaying/interactin-
TPC to the TOF, as performed by GEANT3, we obtain
gbeforeorinsidetheTPC;theloss“outofgeometry”re-
a”real”hitpoint(morethanoneincaseofmultiplecros-
ferstoextrapolatedTPCtracksmissingtheTOFsurface;
sing). The success of the track matching depends then on
the loss “dead space” refers to extrapolated TPC tracks
the ratio <∆R>/<Rmin >, where ∆R is the deviation ending up in the TOF dead space (between neighbouring
of the real hit point (first crossing) of a track from the
corresponding ideal hit and Rmin the minimum distance 3 “Primary”electronsandpositronsareoriginatedfromthe
ofthisrealhitwithrespecttoalltheotherrealones(first decay chains of primary neutral pions.
ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector 169
π K
1 1
without TPC tracks
0.9 0.9
out of geometry
0.8 dead space 0.8
0.7 empty pad 0.7
0.6 0.6
wrong time
0.5 0.5
0.4 0.4
0.3 0.3
true time
0.2 0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5
p(GeV/c) p(GeV/c)
9756 primary π 854 primary K
104 103
103
102
102
10
10
1 1
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5
p(GeV/c) p(GeV/c)
Fig. 5. Acceptances (upper, dashed/dotted-line histograms) of primary π± (left) and K± (right), as functions of momentum,
for(fromtoptobottom):a)trackingintheTPC,b)extrapolationtotheTOFsurface,c)extrapolationtotheTOFsensitive-
pad surface, d) matching with the fired TOF pads, e) correct matching with the fired TOF pads. The solid-line histograms
superimposed refer to primary particles “really” hitting the TOF sensitive surface. All the acceptances are normalized to the
originalsamplesofprimaryparticlesgeneratedinthe|θ−90◦|<45◦ region(100HIJINGeventsatB=0.4T).Thesingle-event
momentum spectra (lower histograms) refer to these originally generated primaries.
strips,forexample);theloss“emptypads”referstoextra- derable as a consequence of their much softer momentum
polated TPC tracks impinging on TOF pads which have spectrum.
not fired (mostly because of particle decays and interac- Let us define the TPC-TOF matching efficiency, for
tions between the TPC and the TOF); the loss “wrong each type of particle, as:
time” refers to extrapolated TPC tracks matched with
pads fired by other tracks. The “true time” acceptance efficiency = Nmtatch(i)
in Figs. 5 and 6 refers to extrapolated TPC tracks well N(i)
matched with the pads they have actually fired.
One can see from these figures that in the momen- whereN(i)isthetotalnumberofparticlesoftypei(with
tum range p> 0.5 GeV/c the fractions of tracks matched i = π, K, p) extrapolated to the TOF sensitive-pad sur-
with fired pads (“true time” + “wrong time” areas) cor- face and Nt (i) is the number of correctly matched
match
respond to the fractions of originally generated particles particles (with “true time”) of the same type. This effi-
(i.e. “real” particles) actually hitting the TOF sensitive ciency corresponds to the ratio of the areas below histo-
surface (solid-line histograms). It should be pointed out gramse)andc)inFigs.5and6,foreachkindofparticle.
that for electrons the loss “without TPC track” is consi- The results obtained with 100 HIJING-generated events
170 ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector
pp ee
1 1
0.9 0.9
0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
p(GeV/c) p(GeV/c)
557799 pprriimmaarryy pp 115577 pprriimmaarryy ee
102
102
10
10
1
1 10-1
0 0.5 1 1.5 2 2.5 3 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
p(GeV/c) p(GeV/c)
Fig. 6. Same as Fig. 5 for primary p(p) and e±.
aresummarizedinTable1fortwodifferentmagneticfield Table 1. TPC-TOFmatchingefficiencies(%)forprimaryha-
values: B = 0.2 T and B = 0.4 T. As one can see in this dronsextrapolatedtotheTOFsensitive-padsurfacewithB =
table, only the pion efficiency increases when B increases. 0.2 T and, in brackets, with B = 0.4 T (100 HIJING events).
This is due to the fact that only the TOF acceptance for
Primary hadron Efficiency
low-momentum pions (with p < 0.5 GeV/c) is sizeably
reduced when going from 0.2 to 0.4 T (see later on Ta- π± 38 (42)
ble 2). These soft pions, which mostly contribute to the K± 37 (36)
“wrongtime”sampleatB =0.2T,donotreachtheTOF
p(p¯) 35 (32)
at B = 0.4 T. As a consequence, the pion efficiency im-
proves. Using SHAKER, all the efficiencies in Table 1 are
a few percent higher due to the slightly harder particle
momentum spectra produced by this generator [3,4,5]. If to Nprobe = 20 (the present value), the matching effi-
the TRD is not included as passive material between the ciency improves by a factor ≈ 2. Meanwhile the percen-
TPC and the TOF, the pion and kaon efficiencies both tage of TPC tracks matched with “empty pads” decrea-
increase by a factor ≈1.3, and the proton efficiency by a ses by a factor ≈ 3 and the percentage corresponding to
factor ≈1.6 (thus becoming equal to the pion efficiency). “wrong time” pads increases up to a level which fits the
As anticipated, the number of probe tracks generated TOF occupancy caused by background hits, as it should
for each TPC “mother”-track plays a role in the TPC- be. Hence no significant improvement can be obtained for
TOF matching procedure. When going from Nprobe = 1 Nprobe >20.
ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector 171
3 Particle identification For the identification procedure we introduce specific
contour cuts, shown in Fig. 7, for kaons and protons. The
Usingthetrackingandtiminginformationparticleidenti- shapeofthecontoursisdictatedbytheconcerntoextend
fication (PID) may be performed. Different methods can the range of pure identification to the largest momenta,
be adopted for this PID task. In this section we will de- even at the expense of efficiency. Figure 7 shows two dif-
scribe a “contour method” (where the contour is formed ferent choices of the kaon contour cut. Unless differently
byanumberofstraightlinesinthemass-momentumplane specified, the kaon contour cut corresponding to the nar-
todefinethePIDselectionregion),anda“probabilityme- rower (m,p) region has been used as default cut in the
thod”. following.
Besidesthesegraphicalcuts,inthePIDprocedurede-
scribedherein,differentmatchingcutswereappliedtodif-
ferent particle species. The strongest matching cuts were
3.1 Contour method
applied to kaons (because of the large pion contamina-
tion): only unambiguous correspondences between tracks
ForthismethodallTPCreconstructedandTOFmatched
and TOF pads were selected, i.e. only pads matched with
primaryparticlesareenteredinamomentumversusmass
single tracks. For protons, pads matched with more than
plotasshowninFig.7forSHAKER“a`laRHIC”4 events.
one track were also considered since we expect one half of
Negative mass values correspond to negative square-root
these tracks to have “true time” and the other half to be
arguments in the mass determination formula:
outsidethePIDcontourcuts.Unlessdifferentlyidentified
(as kaons or protons), all the other TPC-reconstructed
m=p(t2/l2−1)1/2 (1)
primaries, matched or mismatched with the TOF, were
regarded as pions. This does not lead to a very large con-
p,tandlbeingtheparticlemomentum,time-of-flightand
tamination due to the overwhelming percentage of pions
track length. The mean flight-time for primaries from the
in the SHAKER and HIJING events [3]. Of course a pion
|θ−90◦| < 45◦ region reaching the TOF is ≈ 16 ns, and
contourcutmayalsobedefined,asforkaonsandprotons,
the mean flight-length ≈4.2 m, with B = 0.2-0.4 T.
whenever an increased purity of the pion sample is nee-
Notice that in this plot we have all the particles as-
ded. For Pb-Pb collisions we do not consider any contour
sociated with fired TOF pads. Tracks matched with mul-
cut for attempting electron identification (below say 0.5
tihit pads have been included here in order to check the
GeV/c)sincethepioncontaminationisprohibitiveinthis
full particle identification power of the detector, as we
case.
will see later on. We observe distinct clusters of points
corresponding to pions, kaons and protons as well as a
horizontal broad band which corresponds to mismatched 3.2 PID efficiency and contamination
tracks,withpadsgiving“wrongtimes”(seeSect.2).This
leads to contamination, even in the low momentum re- Let us call Nid(i), with i=π, K, p, the number of identi-
gionwherethemassresolution(∂m/m)isgood.Atlarger fied hadrons of different species according to the criteria
momenta,theparticleclustersbecomeobviouslywiderbe- specified in Sect. 3.1. This number is obviously Nid(i) =
causethemassresolution(∂m/m)increaseswiththepar- Nitd(i)+Niwd(i),whereNitd(i)isthenumberofwellidenti-
ticle momentum5 even for practically constant time reso- fiedparticlesoftypeiandNiwd(i)thenumberofnon-typei
lution (∂t/t). Moreover, at small momenta, the particle particlesmisidentifiedasparticlesoftypei.Theefficiency
clusters appear to be systematically distorted, i.e. curved and contamination of the PID procedure, for each type of
towards larger masses. The reason is that we calculate particle, are then:
the mass assuming that the particle momentum is con- Nt (i) Nw(i)
stant, i.e. the same as at the production vertex, whereas efficiency = id , contamination = id
itdecreasesduetoenergylosswhentheparticletravelsall N(i) Nid(i)
the way to the TOF. One can reduce this effect by using
where N(i) is the total number of particles of type i pre-
in the mass calculation some effective momentum corre-
sent in the (m, p) scatter-plot, i.e. of TPC-reconstructed
sponding to the middle-point of the track, instead of the
and TOF-matched primary particles. The momentum de-
momentum at the vertex.
pendence of the PID results is presented in Fig. 8 for
4 SinceitisobviousthattheoverallPIDefficiencyandconta- 100 HIJING events at two different magnetic field values:
B = 0.2 T and B = 0.4 T, separately for π, K and p.
mination will indeed depend on the particle momentum spec-
This figure characterizes the performance of the method
tra, we have made the exercise to modify the SHAKER ge-
itself. One can see that the π, K, p contaminations ob-
neration parameters in order to approximately reproduce the
recentexperimentaldataontheinclusiveπ,K,pdistributions viously increase with momentum when approaching the
from RHIC [16]. limits due to time resolution. For kaons and protons, the
5 The ∂m/m resolution has three main contributions: i) contaminations also increase at low momenta as a conse-
∂m/m = ∂p/p; ii) ∂m/m = (E/m)2∂t/t; iii) ∂m/m = quence of TPC-TOF track mismatching due to the severe
(E/m)2∂l/l. At relatively high momenta, it is mainly driven background conditions (see Sect. 3.1). Finally, the effect
bytheerrorsontime-of-flightandtracklengthratherthanby of the larger magnetic field intensity is almost negligible
the errors on the momentum determination. at this stage.
172 ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector
Fig. 7. Mass separation with the TOF detector (overall time resolution: 120 ps) as a function of momentum for 25 SHAKER
“a` la RHIC” events, with B=0.2 T.
(cid:3)
1 p 1 KK 1 pp
0.9 0.9 0.9
y 0.8 0.8 0.8
c 0.7 0.7 0.7
n
e 0.6 0.6 0.6
ci 0.5 0.5 0.5
fi 0.4 0.4 0.4
Ef 0.3 0.3 0.3
0.2 0.2 0.2
0.1 0.1 0.1
0 01 01
n 0.90 0.5 1 1.5 2 0.90 0.5 1 1.5 2 0.90 1 2 3 4
o 0.8 0.8 0.8
i
t 0.7 0.7 0.7
a
n 0.6 0.6 0.6
mi 0.5 0.5 0.5
a 0.4 0.4 0.4
t 0.3 0.3 0.3
n
o 0.2 0.2 0.2
C 0.1 0.1 0.1
0 0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 1 2 3 4
p (GeV/c)
Fig. 8. PIDefficienciesandcontaminationsvs.momentumforprimarychargedparticlesofdifferentspeciesmatchedwithfired
TOF pads (100 HIJING events), tracked with B =0.2 T (solid-line histograms) and B =0.4 T (dashed-line histograms).
ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector 173
2) Table 2. OverallPIDefficiencies(%)andcontaminations(%)
V/c104 50 HIJING events indifferentmomentum(GeV/c)ranges,relativetoallprimary
e hadrons generated in the |θ−90◦|<45◦ region (100 HIJING
M B=0.4T
events), tracked with B = 0.2T and, in brackets, with B =
0
1 0.4T.
(1/ (∗) [Identified pions, kaons and protons have respectively: p ≥
N m 0.15(0.2),0.25(0.3)and0.35(0.4)GeV/c,atB =0.2(0.4)T.]
dd
3
10 Primary Momentum Overall Contamination
π hadron range efficiency
π± (∗) p <2.5 63(48) 7(9)
K± (∗) p <2.5 14(13) 13(15)
102 p(p¯) (∗) p <4.5 27(24) 5(7)
K
p π± 0.5<p <2 74(72) 11(12)
K± 0.5<p <2 23(21) 11(12)
p(p¯) 0.5<p <2 37(33) 5(7)
0 0.2 0.4 0.6 0.8 1 1.2
2
m (GeV/c )
Fig. 9. Reconstructed mass with the TOF detector (overall tum range exceeding ≈2 GeV/c for kaons, and extending
TOF system resolution: 120 ps) in the momentum range 0.5 up to ≈4 GeV/c for protons.
≤ p ≤ 2.5 GeV/c, for 50 HIJING events at B = 0.4 T. The Proton-proton collisions are characterized by a much
grey- and dashed-area histograms superimposed show the in- lower multiplicity and a consequently much lower back-
dividualmassdistributionofthetruepions,kaonsandprotons ground with respect to typical heavy-ion collisions, thus
associated with the fired TOF pads. allowing us to be confident that a good particle identifi-
cation will be achieved with the TOF detector during the
LHC p-p running periods.
Figure 9 shows the reconstructed particle mass6 for Let us present some results obtained with a sample
TOF-identified primaries with momenta in the range p= of 6000 PYTHIA [12] events, corresponding to m√inimum
(0.5-2.5)GeV/c,wheretheachievedPIDqualityinterms bias p-p interactions at a centre-of-mass energy s = 14
of π/K/p separation clearly shows up. TeV. As for Pb-Pb collisions, all TPC-reconstructed and
Figure 10 shows the final PID power of the TOF de- TOF-matched primary particles are entered in a momen-
tector for pions, kaons and protons, expressed in terms of tum versus mass plot as shown in Fig. 11, where distinct
“overall efficiencies” which include all the acceptance and clusters corresponding to electrons, pions, kaons and pro-
efficiencyfactorsinvolvedintheanalysis.InthiscaseN(i) tons can be clearly seen. In Pb-Pb collisions, the identifi-
isthetotalnumberofprimaryhadronsoftypeioriginally cationofelectronswasnotconsideredbecauseofanexcee-
generatedintheangularregion|θ−90◦|<45◦.InFig.10, dingly high background level. In the present p-p analysis,
twodifferentcontourcutsforkaons(seeFig.7)havebeen we gain a significant PID efficiency for electrons in the
applied to the SHAKER “a` la RHIC” sample, thus sho- 0.1≤p≤0.5 GeV/c range.
wing how a better kaon efficiency is achievable, say for In Fig. 12 we show the overall PID efficiency and con-
p > 1.75 GeV/c, at the cost of a limited contamination tamination as a function of momentum, for e, π, K and p
increase. primary particles produced in the |θ−90◦|<45◦ central
Whenfoldingintheinclusiveparticlemomentumspec- region. The contour cuts defined for e, K and p in p-p
tra,theresults,integratedoverthewholePIDmomentum events are drawn in Fig. 11 and clearly differ from those
interval, are given in Table 2 for the HIJING event gene- previously adopted in the Pb-Pb case. The final PID po-
rator. These results obviously depend on the event gene- weroftheTOFdetectorturnsouttosignificantlyimprove
rator parameters (dNch/dy density, pt spectra, K/π and in the p-p case thanks to the much reduced background
p/π particle ratios, etc.) and the simulated experimental conditions. In the previously mentioned momentum ran-
conditions (in the presence of different materials, magne- ges (beyond to ≈ 2 GeV/c for K and up to ≈ 4 GeV/c
ticfieldvalues,etc.).Notwithstandingtheuncertaintiesat for p), the PID efficiencies are higher and the contami-
thepresentstage,wecanconfidentlyconcludethatineach nations lower, almost negligible for kaons and protons.
Pb-Pb final state the ALICE TOF detector will be able Moreover some electron identification would indeed be
toidentifythousandsofpions,andhundredsofkaonsand possible (up to ≈ 0.5 GeV/c) with less than 10% con-
protons, with 10% or less contamination in the momen- tamination.
6 Trackswithnegativemassvalues(seeSect.3.1andFig.7)
areenteredinFig.9histogramatthecorresponding|m|values.
174 ALICE TOF Group (Bologna-CERN-ITEP-Salerno): Particle identification with the ALICE TOF detector
(cid:3)
1 p 1 KK 1 pp
0.9 0.9 0.9
0.8 0.8 0.8
y 0.7 0.7 0.7
c
n 0.6 0.6 0.6
e 0.5 0.5 0.5
ci 0.4 0.4 0.4
fi 0.3 0.3 0.3
f
E 0.2 0.2 0.2
0.1 0.1 0.1
0 01 01
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 1 2 3 4
n 0.9 0.9 0.9
o 0.8 0.8 0.8
ti 0.7 0.7 0.7
a
0.6 0.6 0.6
n
i 0.5 0.5 0.5
m
0.4 0.4 0.4
a
t 0.3 0.3 0.3
n
0.2 0.2 0.2
o
C 0.1 0.1 0.1
0 0 0
0 0.5 1 1.5 2 0 0.5 1 1.5 2 0 1 2 3 4
p (GeV/c)
Fig.10.OverallPIDefficienciesandcontaminationsvs.momentum,relativetoallprimarychargedparticlesofdifferentspecies
generated in the |θ−90◦| < 45◦ region (100 SHAKER “a` la RHIC” events), tracked with B = 0.2 T: applying the narrower
(solid-line histograms) and wider (dashed-line histograms) contour cuts shown in Fig. 7 for kaon identification.
3.3 Probability method
f (m)
Another promising approach appears to be the so-called P (m)= K
“probability method”. It has already been used in the 104 K fπ(m)+fK(m)+fp(m)
mmppiioonn
STARexperimentatRHIC[17].Foreachkindofparticle, fπ EEnnttrriieess 223355222222
MMeeaann 00..11446677
one can derive a probability density function (p.d.f.) ba- 103 f f RRMMSS 00..0044115544
π K p
sed on some separation parameter, for instance the mass,
by fitting the appropriate Monte Carlo simulated distri- 102 K p
bution. We can then compute the probabilities for each
TPC-TOF matched track, measured with a given mass
10
m, to be a pion, a kaon or a proton (neglecting the el-
ectron case) from the heights of the corresponding mass
p.d.f.’s (fπ, fK, fp) at this m-value. Each p.d.f. is norma- 1
lized to the corresponding particle yield. The probability
0 0.2 0.4 0.6 0.8 1
to be a particle of type i can be for instance defined as m Mass (GeV/c2)
Pi(m) = fi(m)/(fπ(m)+fK(m)+fp(m)). Examples of
these fi, for i = π, K, p, obtained as Monte Carlo mass Fig. 13. Examples of mass probability density functions for
distributions, are given in Fig. 13. Of course different sets
pions,kaonsandprotons,obtainedasmassspectraofunambi-
of mass distributions should be simulated, corresponding guouslyTPC-TOFmatchedprimarytrackswith0.5≤p≤2.5
todifferentappropriatemomentumbins,inordertoderive GeV/c(100HIJINGeventsatB=0.4T).Superimposed,some
more accurate p.d.f.’s.
Gaussian fits to guide the eye.
Howevermassisnotaverygoodseparationparameter
because its probability density function has not a proper
Gaussian shape as a consequence of the mass determi-
nation formula (1) given in Sect. 3.1. Time-of-flight is in- tracklengthandmomentummeasurementswithmasshy-
deedGaussianlydistributedandcanbemoreconveniently pothesis mi (see Fig. 14). Here a weighting procedure is
used for probability calculations. For a track measured applied to take into account the different particle yields.
with a given flight-time t, we can for instance calculate The time-based probabilities obtained in this way are
the probability to have a mass mi, for i = π, K, p, as showninFig.15asfunctionsofthemassderivedaccording
Pi(t) = g(ti)/(g(tπ)+g(tK)+g(tp)), where g(ti) is the to formula (1). One can see that the masses at which the
height of the Gaussian p.d.f. g (with mean value t and probability for a particle to be a pion becomes equal to
standard deviation σTOF) at the time ti derived from the the probability to be a kaon (≈ 0.4 GeV/c2), and the