Padova, 18 dicembre 2003 Andrea Dainese 1 Charm production and energy loss in nucleus-nucleus...

Padova, 18 dicembre 2003 Andrea Dainese 1

Charm production and energy loss in nucleus-nucleus collisions

with ALICE

Andrea Dainese

Università degli Studi di PadovaDottorato di Ricerca in Fisica – Ciclo XVI

ccc

D0K-

+

c


OutlineIntroduction: physics of hot and dense QCD matter

Heavy-ion physics at the LHC

Parton energy loss in the quark-gluon plasma (QGP)

the case of heavy quarks (charm)

Experimental feasibility for exclusive D meson reconstruction in Pb-Pb (and pp) collisions

Can we measure D meson pt distribution in Pb-Pb? energy loss of c quarks in the QGP?


Physics of hot and dense QCD matter

QCD phase diagramhadronic matter exists in different states

at high energy density (high temperature and/or high density) nuclear matter undergoes a phase transition to a deconfined Quark-Gluon Plasma (QGP)

Tem

pera

ture

Baryonic density

U. Heinz

“Is it possible to de-confine quarks?”



Lattice QCD (B ~ 0) estimates the phase transition at:

Tc ~ 170 MeV and c ~ 1 GeV/fm3 ~ 5 nucleus

freedom of degrees4n

T

phase transition:

37 3 :d.o.f. n

piongas

QGP 3 massless flavours



Tem

pera

ture

Baryonic density

U. Heinz

10-

s-o

ldU

niv

erse

?

Neutronstars ?


The Little Bang in the lab

In high-energy heavy-ion collisions large energy densities (> 2–3 GeV/fm3) are reached over large volumes (> 1000 fm3)

QGP evidence at CERN-SPS (up to: Pb-Pb, )

BNL-RHIC extending these results at

Next step: LHC with Pb-Pb @

deep deconfinement: “ideal gas” of gluons and quarks

large production of hard partons and heavy quarks

excellent tools to study properties of QGP

GeV 17NNs

GeV 200NNs

TeV 5.5NNs


Deep deconfinement at the LHC

LHC: factor 30 jump in w.r.t. RHIC

much larger initial temperature

study of hotter, hotter, bigger,bigger, longer-livinglonger-living ‘drops’ of QGP

s

‘Deep de-confinement’ closer to ‘ideal’ QGP easier comp. with theory

SPS

17 GeV

RHIC

200 GeV

LHC

5.5 TeV

initial T ~ 200 MeV ~ 300 MeV > 600 MeV

volume 103 fm3 104 fm3 105 fm3

life-time < 2 fm/c 2-4 fm/c > 10 fm/c


Hard Processes in AA at the LHCMain novelty of the LHC: large hard cross section

Hard processes are extremely useful toolslarge virtuality Q happen at t = 0

small “formation time” t ~ 1/Q

(for charm: t < 1/2mc ~ 0.1 fm/c << QGP ~ 5–10 fm/c)

Initial yields and pt distributions in AA can be predicted using pp measurements + pQCD + collision geometry + “known” nuclear effects (e.g. PDF nuclear shadowing suppression of low-x

(low-pt) production)

Deviations from such predictions are due to the medium

medium formed in the collision

Q

Pb

Pb c

c

time


Hard partons probe the medium

Partons travel ~ 5 fm in the high colour-density medium

Energy loss by gluon bremsstrahlungmodifies momentum distributions

depends on medium properties

PROBE

medium

probe IN(known from

pp, pA+ pQCD)

probeOUT

IN vs. OUT


QCD process:emitted gluon itself radiates E L2

Parton Energy Loss

Due to medium-induced gluon emission

Average energy loss (BDMPS model):

hardparton

path length L

2 ˆ LqCE RsCasimir coupling factor:4/3 for quarks3 for gluons

Medium transport coefficient gluon density and momenta

R.Baier, Yu.L.Dokshitzer, A.H.Mueller, S.Peigne' and D.Schiff, (BDMPS), Nucl. Phys. B483 (1997) 291.C.A.Salgado and U.A.Wiedemann, Phys. Rev. D68 (2003) 014008 [arXiv:hep-ph/0302184].

cold matter

QGP@ LHC

LHC

E , L = 5 fm

Kinematic constraint E Esignificantly reduceseffective average E


Q

Heavy Quarks: dead cone

Heavy quarks with momenta < 20–30 GeV/c v << c

Gluons radiation is suppressed at angles < mQ/EQ

“dead cone” effectDue to destructive interference

Contributes to the harder fragmentation of heavy quarks

Yu.L.Dokschitzer and D.E.Kharzeev: dead cone implies lower energy loss

Yu.L.Dokshitzer and D.E.Kharzeev, Phys. Lett. B519 (2001) 199 [arXiv:hep-ph/0106202].

D mesons quenching reducedRatio D/hadrons (or D/0) enhanced and sensitive to medium properties


Experimental study of energy loss

Quenching can be studied by comparing pt distributions of leading particles in pp and AA

Nuclear modification factor:

Energy loss RAA < 1

tpp

tAA

collAA dpdN

dpdN

NR

/

/1

PHENIX @ RHICs

NN = 130 GeV

dead cone?

PYTHIA pp +binary scaling

PHENIX


D0 K-+ in ALICEExclusive reconstruction direct measurement of the pt distribution ideal tool to study RAA

Weak decay with mean proper length c = 124 m

STRATEGY: invariant-mass analysis of fully-reconstructed topologies originating from (displaced) secondary vertices

Measurement of Impact Parameters

Measurement of Momenta

Particle identification to tag the two decay products


Background multiplicity in Pb-Pb

What is the background to hadronic D decays? combinatorial background given by pairs of uncorrelated tracks

with large impact parameter 2/ dydNB ch

2500/ dydNch

in central Pb-Pb at LHC

huge combinatorial background!huge combinatorial background!

Simulations performed using 6000/ dydNch

need excellent detector response and good selection strategyneed excellent detector response and good selection strategy


D0 K-+: Detection strategy with ALICE

Inner Tracking SystemMeasurement of Impact Parameters & Momentum Res. Improvement

Time Projection Chamber Track Finding & Measurement of Momenta

Time Of FlightParticle Identification (K/ separation)

Magnetic field (B = 0.4 Tesla in this study)


ALICE Barrel

||<0.9:B = 0.4 TTOFTPCITS with: - Si pixels- Si drifts- Si strips

PIXEL CELL

z: 425 m

r: 50 m

Two layers:r = 4 cmr = 7 cm

9.8 M


TrackingTracking efficiency ~70% with dNch/dy=6000

pionskaons

pt resolution = 1% at 1 GeV/c

D0 invariant mass resolution:

MeV 13)( ,%7.0)(

2

1)( M

p

p

M

M


Impact parameter resolution

< 60 mfor pt > 1 GeV/c

Crucial for heavy-quark IDSystematic study of resolution was carried out


TOF PID

TOF

Pb-Pb, dNch/dy=6000

Optimization for hadronic charmdecays was studied:minimize probability to tag K as


D0 K-+: Signal and backgroundSignal:

charm cross section from NLO pQCD (MNR program), average of results given by MRS98 and CTEQ5M PDFs (with EKS98 in Pb-Pb)

signal generated using PYTHIA, tuned to reproduce pt distr. given by NLO pQCD

contribution from bBD0 (~5%) also included

Background:Pb-Pb: HIJING (dNch/dy=6000 ! we expect 2500 !); pp: PYTHIA;

system shadowing

pp 14 TeV 11.2 1 0.16 0.0007

Pb-Pb 5.5 TeV (5% cent) 6.6 0.65 115 0.5

[mb] ccNN cc

totN dyKDdN /)( 0

MNR Program: M.L.Mangano, P.Nason and G.Ridolfi, Nucl. Phys. B373 (1992) 295.


D0 K-+: Selection of D0 candidates

Main selection: displaced-vertex selection

pair of tracks with large impact parameters

good pointing of reconstructed D0 momentum to the primary vertex

d0K d0

<<0cos pointing 1


S/B

initial

(M3)

S/evt

final

(M1)

S/B

final

(M1)

Significance

S/S+B

(M1)

Pb-Pb 5 10-6 1.3 10-3 11 % 37

(for 107 evts,

~1 month)

pp 2 10-3 1.9 10-5 11 % 44

(for 109 evts,

~1 year)

D0 K-+: Results

Note: with dNch/dy = 3000, S/B larger by 4 and significance larger by 2

0–14 GeV/c1–14 GeV/c

Pb-Pb, 5.5 TeV 107 events

pp, 14 TeV109 events


D0 K-+: d2(D0)/dptdy and d(D0)/dy

d(D0)/dy for|y| < 1 and pt > 1 GeV/c (65% of (pt > 0)) statistical error = 7 % systematic error = 19 % from b = 9 % MC correction = 10% B.R. = 2.4 % from AA to NN = 13 %

d(D0)/dy for|y| < 1 and pt > 0

statistical error = 3 % systematic error = 14 % from b = 8 % MC correction = 10% B.R. = 2.4 % inel = 5 %

inner bars: statisticalouter bars: systematic


Sensitivity on RAA for D0 mesons

‘High’ pt (6–15 GeV/c)here energy loss can be studied(it’s the only expected effect)

Low pt (< 6–7 GeV/c)Nuclear shadowing

tpp

tAA

collAA dpdN

dpdN

NR

/

/1


Energy-loss simulation

Energy loss simulated using one of the most

advanced models*

My contributions:

path length of partons in the medium

dead-cone for heavy quarks

estimate transport coefficient for

central Pb-Pb collisions at LHC,

on the basis of the hadron suppression

observed at RHIC*C.A.Salgado and U.A.Wiedemann, Phys. Rev. D68 (2003) 014008 [arXiv:hep-ph/0302184].

2 ˆ LqCE Rs

Q


Average relative energy loss

cold matter

QGP@ LHC


RAA with Quenching

RRAAAA ~ 0.4 ~ 0.4––0.50.5increasing at high increasing at high pptt

RRAAAA ~ 0.7 ~ 0.7––0.80.8decreasing at high decreasing at high pptt


D/hadrons ratio (1)Ratio expected to be enhanced because:

D comes from (c) quark, while , K, p come mainly (~80% in PYTHIA) from gluons, which lose 2 more energy w.r.t. quarks

dead cone for heavy quarks

Experimentally use double ratio: RAAD/RAA

h

almost all systematic errors of both Pb-Pb and pp cancel out!

D/h ratio: R

D/h = R

AAD / R

AAh

R

D/h ~ 2–3

in hot QGP

sensitive tomedium density


Summary D0 mesons can be exclusively reconstructed with the ALICE detector despite the high-multiplicity environment of central Pb-Pb coll. This will allow to: measure charm production in 0–15 GeV/c with

statistical uncertainty better than 10–15%systematic uncertainty better than 15–20%

study the mass and flavour dependenceof QCD energy loss


Back-up Slides


Heavy ion Physics at the LHC

LHC: factor 30 jump in w.r.t. RHICs

hotterhotterbiggerbigger

study of:

longer-livinglonger-living

‘drops’ of QGP

What are the properties of deconfined matter?

> 102 - 4< 2QGP (fm/c)

~ 105~ 104~ 103Vf [fm3]

100103 [GeV/fm3]

~ 2500650400dNch/dy

550020017s1/2 [GeV]

LHCRHICSPSCentral collisions

3

10

10


Initial- and final-state effectsIn absence of nuclear effects:

Hard processes, and charm production in particular,

allow to study and disentangle:

Initial-state (cold medium) effects, by comparing pA with pp and with pQCD predictions

Nuclear modification of PDFs

Final-state (hot medium) effects, by comparing AA, pA, pp (and pQCD)

Partonic Energy Loss

Thermal Charm Production (?)

pAin collN

AAin collN

collisionshardpp

hardpAAA NNN )(

at low pt

(also) at high pt


Initial-state effects: Shadowing

Bjorken-x: fraction of the momentum of the proton ( ) carried by the parton entering the hard scattering

At the LHC

Pb ion @ LHC ~ 105-106 partons

(mainly gluons)

q

qQ

2/1 sx

2/2 sx

2/s

sMsQx qq //

charm43 1010~ x

xa

xb

xa+xb

gPb(x)/gp(x)

Shadowing:• reduces initial hard yield at low pt

• scales trivially from pA to AA

“they are so close that they fuse”


BDMPS model

2

0

2

2

ˆ 2

22

2

2

ˆ21

ˆ

LqCC

d

dIdE

C

d

dI

Lq

qq

Rsc

Rs

cRs

c

medium

c

kt

vacuum

medium


Dead cone effect

Dokshitzer-Kharzeev: energy distribution dN/d of radiated gluons suppressed by angle-dependent factor

high-energy part of gluon radiation more suppressed

strong reduction of quenching

effect vanishes for EQ >> mQ (high-pt charm = standard quenching)

,ˆ,,

ˆ1

11

/

23

2

2

2

2

qEmFd

dN

qE

m

d

dN

E

m

d

dN

d

dN

QQLHLIGHT

Q

Q

LIGHT

Q

Q

LIGHTHEAVY

Dokshitzer

D mesons quenching reducedRatio D/hadrons (or D/0) enhanced and sensitive to medium properties


Exclusive charm reconstruction

D0 K- decay allows the exclusive reconstruction of

open charm mesons and a direct measurement of the

c quark pt distribution

In order to probe:

additional cc production

energy loss of c quarks

Essential to measure:

total cc production cross-section

pt distribution of charm particles

channel D0 K- c D e

c low pt acceptance ~ 0 ~ 2-3 GeV/c

(electron ID)

c high pt

acceptance

~15 GeV/c

(statistics)

probably

> 15 GeV/c

S/B separation clean (inv. mass) non-trivial c/b disentangling

c D0 c D e


Layer Detector radius [cm] resolution r [m] resolution z [m]

1 Si Pixels 4 12 120

2 Si Pixels 7 12 120

3 Si Drifts 14.9 38 28

4 Si Drifts 23.8 38 28

5 Si Strips 39.1 20 830

6 Si Strips 43.6 20 830

ITS layers


%54)(/

)/()/(/

)(/

)(/0

000

0

0

DcdydN

XDBBBRBBbdydN

DcdydN

DbdydN

0DBb

Signal generationCharm cross section from NLO pQCD (MNR program), average of results given by MRST and CTEQ5M PDFs (with EKS98 in Pb-Pb)

Signal generated using PYTHIA, tuned to reproduce pt distr. given by NLO pQCD

Also the contribution from beauty was included:

system shadowing

pp 14 TeV 11.2 1 0.16 0.0007

Pb-Pb 5.5 TeV (5% cent) 6.6 0.65 115 0.5

[mb] ccNN cc

totN dyKDdN /)( 0


Classification of the processes

Hard scattering: LO graph

Processes classified w.r.t. # HVQs in hard scattering final state

No double counting because hard scattering is the process with largest virtuality

Q

Q

Q

Pair Creation (PC) Flavour Excitation (FE) Gluon Splitting (GS)

hardscattering

Q

QQ

partonshower


Hard cross section in pQCD

0

2 ))(()(ˆ

ˆ),(),(

k

kssQQij

QQijjpji

pi

QQpp

QQ

QxfQxf

inelpp

QQpp

QQppN /


Extrapolation to AA

Binary scaling:

Glauber model of collision geometry:

Shadowing: “a nucleon in the Pb nucleus contains less small-x partons than a free nucleon”

collisionsQQ

ppQQ

pAAA NNN )(

b1800collisionsN

Pb-Pb, b=0 at LHC:


Tuning of PYTHIA parameters

Comparison at the bare quark level

Heavy Quarks in PYTHIA:

MSEL = 4/5 Leading Order processessettings corresponding to MNR

good agreement with MNR LO

MSEL = 1 initial and final state Parton Shower processes describe contributions above LO

agreement with MNR NLO less good

parton shower processes NLO processes

massless Matrix Elements! cross section diverges at pthard 0

Tuning of parameters less “physics inspired”

Main parameter tuned: min. pthard (2.1 GeV/c for c, 2.75 GeV/c for b)


Charm NLO: (Pb-Pb)/Ncoll 5.5 TeV

MNR PythiaInclude

shadowing

pair creationflavour excitationgluon splittingTOTAL


Heavy quark fragmentation

M. Cacciari, CTEQ School 03


Fragmentation in PYTHIA

Use default parameters

Lund string fragmentation modellongitudinal fragmentation:Lund symmetric fragmentation functionModified to account for harder spectra in HVQ fragmentation

transverse momentum pick-up(px) = (py) = 230 MeV/c

Qq

q q Qq

z=E+pz

z'=f(z)(p

x, p

y)

(-px, -p

y)


Effect of fragm. on spectra

Average pt-reduction:

25% for charm

15% for beauty

Charm Beautybare quarkmeson


3D reconstruction with tracks

Track reconstruction in TPC+ITStrack seeding uses the position of the primary vertex: (x, y) from beam position (resolution ~ 150 m)z from pixels information (resolution ~ 150 m)

Vertex reconstruction in 2 steps:VERTEX FINDING: using DCA for track pairs

VERTEX FITTING:give optimal estimate of the position of the vertexgive vertex covariance matrix

give a


Expected resolutions

Average # rec. tracks = 7 (average on events with # > 1)

Average pT of rec. tracks = 0.6 GeV/c

Resolutions of track position parameters @ 0.6 GeV/c:(d0(r)) 100 m [ d0(r) is to the track! ]

(d0(z)) 240 m

mm

z zdvtx

907

240

tracksof#

)GeV/c 6.0()( )(0

mm

yx rdvtxvtx

555.3

100

2/ tracksof#

)GeV/c 6.0(),( )(0


Vertex Finding Algorithm

Aim: get a first estimate of the vertex position in (x,y) to be used as a starting point for vertex fitter

independent of beam size

improved w.r.t. beam size (hopefully)

Method:1. propagate tracks to vertex nominal position

2. calculate DCA (in space) for each possible pair of tracks (using straight line approximation)

3. get estimate of xvtx and yvtx from mean of results from all pairs


Vertex Fitting Algorithm

Tracks are propagated to the point given by the vertex finder (at the moment nominal position used)

A 2 is written as the sum of the single track 2s w.r.t. a generic vertex position rvtx:

where Wi is track “covariance matrix in global ref. frame”

The solution that minimizes this 2 is analytic:

Tivtxi

iivtx

ivtxi )rr( W)rr()r( 122

ii

ii

ivtx rWWr1

1

WV

iivertex

covariancematrix


Tuning of the algorithmCriterion used to reject mismeasured and secondary tracks from the fit: cut on the maximum contribution to the 2

i2 <

max

if max is too low too many tracks are rejected and we loose

resolution

if max is too high bad or secondary tracks enter the fit and we

loose resolution

This cut is tuned, as a function of event multiplicity, in order to optimize the resolution


Results: resolutions

(x) = 55 m (y) = 55 m (z) = 90 m


Results: pulls

The covariance matrix of the vertex describes correctly the resolutions


Resolutions VS multiplicity


Interaction vertex reconstructionPosition of the beam in (x,y) given by the machine with very high precision (stable for a long time)

“Nominal” size of the beam: = 15 m in Pb-Pb

= 15 m in pp (L ~ 1031 cm-2 s-1)

150 m in pp (if L is reduced at ALICE IP to ~ 1029 cm-2 s-1)

In pp the vertex position has to be reconstructed in 3D using tracks

vertextrackrd )(0

as in Pb-Pbpp with charmpp min. bias


Fast Simulation Techniques

The selection strategy of D0 mesons has to be optimized using 104-105 events for Pb-Pb and 106-107 for pp

1 Fully simulated Pb-Pb event takes 1.2 GB and ~12h!

A number of fast simulation techniques were used in order to achieve these statistics

fast response of the ITS (“fast points”, reduce ITS-time by factor 25)

TPC tracking parameterization, specifically developed for charm and beauty studies (reduces total time by factor 40)

Checks were done that these approximations don’t affect track resolutions (efficiencies were corrected)


Results for Pb-Pb (K,) Invariant Mass distribution (pT –integrated) (corresponding to 107 Pb-Pb events ~ 1 month of ALICE)

S/event = 0.0013 (1.3 104 D0 in 1 month)

B/event = 0.0116

Statistical Significance of the Signal:

after background

subtraction

%11/ BS

37/ BSS

The history of signal …

Total D0 / event 141decaying in K 5.4with K and in acc. 0.5after track rec. 0.14after (,) rejection 0.13after selection cuts 0.0013

… and of background

Selection cuts reduce background by a

factor ~10-7

10-2


What if multiplicity in Pb-Pb is lower?

We used dNch/dy = 6000, which is a pessimistic estimate

Recent analyses of RHIC results seem to suggest as a more realistic value dNch/dy = 3000 (or less)

Charm production cross section:estimate from NLO pQCD (only primary production, no collective effects)

average of theoretical uncertainties (choice of: mc, F, R, PDF)

BKG proportional to (dNch/dy)2

We can scale the results to the case of dNch/dy = 3000:

S/B = 44 %

SGNC = 74 (this only from scaling, obviously better with retuning of cuts)


Estimate of the errors

Statistical error on the selected signal = 1/Significance

Main systematic errors considered:correction for feed-down from beauty (B.R. B D0 is 65%!):

error of ~8% assuming present uncertainty (~80%) on @ LHC

Monte Carlo corrections: ~10%

B.R. D0 K: 2.4%

extrapolation from N(D0)/event to d(D0)/dy: pp: error on (~5%, will be measured by TOTEM) Pb-Pb: error on centrality selection (~8%) + error on TAB (~10%)

bb

inel

bb

Pb-Pbinel

pp


Comparison with pQCD for pp


Interpolation pp 14 5.5 TeV

In pQCD calculations the ratio of the differentialcross sections at 14 and 5.5 TeV is independent of the input parameters within 10% up to 20 GeV/c

pQCD can be safely used to extrapolate pp @ 14 TeVto 5.5 TeV

Necessary to compare Pb-Pb and pp by RAA

set PDF,,,t

F

t

Rc mm

m


Effect of shadowing


Transverse path length L (in Pb-Pb, b<3.5 fm i.e. 5% central)

Partons produced at mid-rapidity assumed to travel in the transverse plane

Parton production points (x0,y0) sampled according to density of collisions coll(x,y) and their azimuthal propagation directions (ux,uy) sampled uniformly

This definition of L is exact only for cylindric profile coll(r)=0 (R-r)

No exact definition for generic coll profile!

MC sampling varying b in [0,3.5 fm] according to dN/db b

0

00

0

00

) , ( 5.0

) , (

yxcoll

yxcoll

ulyulxdl

ulyulxldl

L


Energy-loss probability distributionFor given q and parton species, convolution of P(E|L) and L distribution

Energy loss can be sampled from this P(E) distribution


Inclusion of dead-cone effect

First approximation: dead-cone effect accounted for by folding the energy-loss probability P(E) with the heavy/light suppression factor FH/L

Folding done as function of the quark energy

This procedure yields same result as recalculating the quenching weights with

Single-gluon emission dominates

LHLIGHT

Fd

dN/


Transport coefficient choice

Require for LHC suppression of hadrons as observed at RHIC: RAA ~ 0.2-0.3 for 4<pt<10 GeV/c

pt distributions of hadrons at LHC:

partons (pt>5 GeV/c) generated with PYTHIA pp, 5.5 TeV

(average parton composition: 78% g + 22% q)

energy loss: pt’ = pt – E

(independent) fragmentation with KKP LO F.F.

RAA = (pt distr. w/ quenching) / (pt distr. w/o quenching)


L distribution vs. constant L


Quenching of charm quarks …

Procedure similar to that for light quarks and gluons

Differences:kinematic constraint: if E > pt then c quark is thermalized; assign pt’ according to thermal mt distribution with T = 300 MeV; in this way the total charm cross section is conserved

fragmentation to D mesons via PYTHIA string model


… and of D mesons


D/hadrons ratio (2)

RD/h is enhanced only by the dead-cone effect

Enhancement due to different quark/gluon loss not seen

It is compensated by the harder fragmentation of charm

pthadron = z pt

parton (ptparton)’ = pt

parton – E

(pthadron)’ = pt

hadron – z E

Energy loss observed in RAA is not E but zE

zcD 0.8; zgluonhadron 0.4 (for pt>5 GeV/c)

Ec = Egluon/2.25 (w/o dead cone)

zcD Ec 0.9 zgluonhadron Egluon

Without dead cone, RAAD RAA

h


B D “a` la CDF”

Impact parameter of D0 can be used to separate primary and secondary D0

Background shape from

side-bands in inv. mass

Background subtracted

ccbb /

ppPb-Pb

Padova, 18 dicembre 2003 Andrea Dainese 1 Charm production and energy loss in nucleus-nucleus...

Documents

Transcript of Padova, 18 dicembre 2003 Andrea Dainese 1 Charm production and energy loss in nucleus-nucleus...