Low and intermediate mass dimuons in NA60

1
Low and intermediate mass dimuons in NA60
G. Usai INFN and University of Cagliari (Italy)
2
General question of QCD
Origin of the masses of light hadrons?
Expectation approximate chiral SU(nf)L x SU(nf)R
symmetry ? chiral doublets, degenerate in mass,
with
However, we observe

• spontaneous chiral symmetry breaking ? ltqqgt ? 0

3
Several theoretical approaches including lattice
QCD still in development
Lattice QCD(for mB0 andquenched approx.) two
phase transitions at the same critical
temperature Tc
cL
-
cm
qq
L
1.0 T/Tc
1.0 T/Tc
deconfinement chiral
symmetry transition
restoration
hadron spectral functions on the lattice only now
under study
explicit connection between spectral properties
of hadrons (masses,widths) and the value of the
chiral condensate ltqqgt ?
Use r as a probe for the restoration of chiral
symmetry (Pisarski, 1982)
4
Why focus (mainly) on the r?
Gtot MeV r (770)
150 (1.3fm/c) w(782)
8.6 (23fm/c)
f(1020) 4.4 (44fm/c)

In-medium radiation dominated by the ?

1. r life time t 1.3 fm/c ltlt tcollision gt 10 fm/c
2. continuous regeneration by ??

main difficulty Properties of r in hot and dense
matter unknown (related to the mechanism of mass
generation) Properties of hot and dense medium
unknow (general goal of studying nuclear
collisions)
5
Standard dimuon detection NA50, PHENIX, ALICE,
magnetic field
Thick hadron absorber to reject hadronic
background Trigger system based on fast
detectors to select muon candidates (1 in 10-4
PbPb collisions at SPS energy) Muon tracks
reconstructed by a spectrometer (tracking
detectorsmagnetic field) Extrapolate muon
tracks back to the target taking into account
multiple scattering and energy loss, but -
poor reconstruction of interaction vertex (sz 10
cm) - poor mass resolution (80 MeV at the f)
6
Measuring dimuons in heavy ion collisions the
NA60 case
2.5 T dipole magnet
beam tracker
vertex tracker
targets
Matching of muon tracks

• Origin of muons can be accurately determined
• Improved dimuon mass resolution

7
The NA60 pixel vertex detector
DIPOLE MAGNET 2.5 T
HADRON ABSORBER
TARGETS
40 cm

• 12 tracking points with good acceptance
• 8 small 4-chip planes, plus
• 8 big 8-chip planes (4 tracking stations)
• 3 X0 per plane
• 750 µm Si read-out chip
• 300 µm Si sensor
• ceramic hybrid
• 800000 R/O channels - 96 pixel assemblies

1 cm
8
Vertexing
Y
X
Resolution 10 - 20 ?m in the transverse plane
?z 200 ?m along the beam direction Good vertex
identification with ? 4 tracks
Extremely clean target identification (Log scale!)
9
Contributions to dimuon mass resolution
The dimuon invariant mass resolution has two
components Multiple scattering in the hadron
absorber dominates the resolution for low
momentum muons The variance ?s of the angle

distribution is
proportional to 1/p At high momenta the
resolution is
dominated
by the tracking accuracy

(dp/p proportional to p)
? at mmm 3 GeV the resolution is

dominated by this component Track
matching not so important ? at mmm 1 GeV track
matching

is very effective to increase the

momentum
resolution
10
Muon track matching
The muon spectrometer and the pixel telescope
determine the track parameters in two reference
planes z1 and z2. A choice of the track
parameters at each plane is
Muon spectrometer
Pixel telescope
Absorber
Measured points
Measured points
p1,r and its covariance matrix are propagated to
z2
11
muon spectrometer track parameters with errors
muon spectrometer surface z1
Contributions to multiple scattering between z1
and z2 are added to C1
hadron absorber
muon spectrometer track parameters propagated to
pixel telescope surface
pixel telescope surface z2
pixel telescope track parameters with errors
weigthed mean
joint least square ansatz
M(p2,fit) distributed as a c2 with 5 dof
12
M(p2,fit) distributed as a c2 with 5 degrees of
freedom
The pixel telescope improves drastically the
angular resolution 10 mrad (muon spectrometer
only) ? 1 mrad (adding pixel telescope)
The momentum resolution is comparable in the two
detectors. However, the use of the momentum
information in a high multiplicity environment is
fundamental to achieve the matching ? the pixel
telescope must be a spectrometer
13
Improvement in mass resolution

• Opposite-sign dimuon mass distributions before
  quality cuts
• No muon track matching
• (two magnet settings)

4000 A
?(1020)
f(1020)
sM(f) ? 20 MeV
sM(f) ? 80 MeV
dN/dMmm (Events/50 MeV)
sM(J/?) ? 100 MeV
sM(J/?) ? 70 MeV
Vertex selection and muon track matching
Drastic improvement in mass resolution Narrow
vector mesons clearly resolved But still sitting
on a large unphysical background
(80 of collected statistics)
4000 A
(100 of collected statistics)
6500 A
14
Problems with the matching fake matches
Fake match muon matched to a wrong track in the
vertex telescope Can be important in high
multiplicity events (negligible in pA or
peripheral AA)
Simple technique the match with the smallest c2
is retained. But is it correct or fake? Fake
matches can be studied and subtracted using an
overlay Monte Carlo - Monte Carlo muons are
superimposed to real events (in the vertex
telescope) - Reconstructed as real events, fake
matches can be tagged and the fraction relative
....to correct matched muons is then evaluated
15
The Monte Carlo provides also the kinematic
distribution (mass, pT, ...) of the fake
matches Here is the example for the f meson
The fake-match contribution appears localized in
mass (and pT) space as a broad peak correct
matches s? 23 MeV


wrong matches ?fake 110 MeV
16
Background sources (dimuons)
Main source of background (uncorrelated) decays
of p and K
? the hadron absorber should be as close as
possible to the interaction point If we have N
pions, the average number which decays within 1m
is m10-3N
We have the probabilities (
)
? number of detected muon pairs is
A (A--) acceptance for a like sign muon
pair A- acceptance for a opposite sign muon pair
17
In NA50 acceptance was independent of charge
In NA60 acceptance is different for and

Cuts to equalize it (image
cut in NA50) impossible ? Event Mixing
Define a pool of m and m- tracks out of a sample
of like sign events ( and --) . Pick m and m-
from these like sign pools corresponding to
different events.

The m and
m- are picked in a fraction which reflects the
probabilities to detect them in the experimental
apparatus
Combine them to form artificial pairs of all sign
combinations. If N(mixed) and N--(mixed)
reproduce the corresponding data samples N and
N--, then N-(mixed) should give the
combinatiorial background of the - sample.
18
Accuracy of background subtraction
Estimated estimated through the comparison of
N/--(mixed) to N/--(real)
In NA60 the accuracy is 1 all over the dimuon
mass range. Is that good or bad? It depends on
the signal to background ratio ...
19
The signal to background ratio depends on the
matching c2 cut. Tight cut more precise matching
helps to reject tracks not precisely connected
to primary vertex
1

• The worst case happens in the continuum region
  around the w, where the bkg/signal can reach 25
  in the most central collisions
• ssignal/signal 25 in low mass continuum region
  at most
• In more peripheral collisions ssignal/signal is
  much better

20
The quest of the correct background normalization
CLAS experiment photoproduction of vector mesons
off nuclei ee- combinatorial background
determined by event mixing
Background normalization found directly from fit
best fit prefers r meson with mass shift (in
medium effect)
Background normalization following prescription
for P P- best fit prefers r meson with no in
medium effect
21
The final mass spectra (mmmlt2GeV/c2)
Red distribution final spectrum after getting
rid of fake and combinatorial background. The
net data sample consists of 420000 events! (?
50 of total statistics)
?
?
f
For the first time ? and ? peaks are clearly
visible in dilepton channel (23 MeV/c2 mass
resolution at the f) ???? is also visible

• Fakes/CB lt 10

22
Phase phase coverage (mass-pT)
The dimuon kinematics can be specified by
(m,y,pT,cos?) The probability that a dimuon with
certain kinematic values is detected depends
on Thickness of the muon filter, position of the
target relative to the detecting elements,
magnetic fields (both in the muon spectrometer
and in the vertex telescope), ...
Drop with no vertex magnet
The dipole magnetic field in the vertex region
improves significantly the acceptance for low
mass and low pTopposite sign dimuons
23
Phase phase coverage (y-pT)
A fixed target experiment usually covers the
forward rapidity emisphere. NA60 (and its
predecessors) are optimized to cover the range
3-4 in the lab system (the target rapidity is
zero, the beam rapidity is 6) corresponding to
0-1 in the CMS system
Example of phase space coverage for a few
processes (Monte Carlo)
??mmg
r?mm
f?mm

• Dimuon rapidity coverage in the lab frame
• roughly between 3.3 and 4.3 for low masses
• between 3 and 4 for the J/y dimuons
• (mid rapidity is at 2.9)

24
Measuring the collision centrality
The collision centrality can be measured via the
charged particle multiplicity as measured by the
pixel vertex telescope
Track multiplicity of charged tracks for
triggered dimuons for
opposite-sign pairs combinatorial background
signal pairs
4 multiplicity windows
Centrality bin multiplicity ltdNch/d?gt3.8
Peripheral 428 17
SemiPeripheral 2892 63
Semi-Central 92160 133
Central gt 160 193
25
Which processes populate the dimuon mass spectrum
below 1 GeV?
26
Vector meson dominance
??mm-
Anomaly in the form-factor VMD predicts a
(significantly) smaller value
Dalitz decay ??mm-p0
27
Vector meson dominance
Previous data (Landsberg et al.) fitted with a
pole formula
??mm-
Dalitz decay ??mm-g
? Dalitz form factor
28
Isolate possible excess by subtracting cocktail
(without r) from the data
How to fit in the presence of an unknown source?
? Try to find excess above cocktail (if it
exists) without fit constraints

• ? and ? fix yields such as to get, after
  subtraction, a smooth underlying continuum
• ?
• (?) set upper limit, defined by saturating the
  measured yield in the mass region close to 0.2
  GeV (lower limit for excess).
• (?) use yield measured for pT gt 1.4 GeV/c

29
Evolution of the excess shape with centrality
The evolution of the excess with centrality can
be studied with precision with a rather fine
binning in multiplicity
data cocktail (all pT)

• Clear excess above the cocktail ?, centered at
  the nominal r pole and rising with centrality
• Excess even more pronounced at low pT

cocktail ?/? 1.2
30
Sensitivity of the difference procedure
Change yields of ?, ? and ? by 10 ? enormous
sensitivity, on the level of 1-2, to mistakes
in the particle yields.
The difference spectrum is robust to mistakes
even on the 10 level, since the consequences of
such mistakes are highly localized.
31
Systematics
The largest source of systematic error comes from
the subtraction of combinatorial and fake matches
background. In principle there are other
uncertainty sources as the form factors, but
these are negligible compared to the background.
Illustration of sensitivity ? to correct
subtraction of combinatorial background
and fake matches ? to variation of the ? yield
The systematic errors of continuum 0.4ltMlt0.6 and
0.8ltMlt1GeV are 25 (at most) in the most central
collisions
The structure in ? region looks rather robust
32
Evolution of the excess shape as a function of
centrality
Quantify the peak and the broad symmetric
continuum with a mass interval C around the peak
(0.64 ltMlt0.84 GeV) and two equal side bins L, U
continuum 3/2(LU) peak C-1/2(LU)
Fine analysis in 12 centrality bins
continuum/r
Peak/cocktail r drops by a factor ?2 from
peripheral to central the peak seen is not the
cocktail r
peak/r
nontrivial changes of all three variables at
dNch/dygt100 ?
peak/continuum
33
p-r, and electromagnetic interactions in vacuum
Free r Lagrangian (vector meson)
Free pions Lagrangian
self interactions are neglected
p-r and e.m. interactions introduced via gauge
couplings
g grp pion-r coupling constant
Direct g-r coupling
The r couples only to conserved currents, so that
34
If g grg ? electromagnetic field equations
? The r meson is the only hadronic source of the
electromagnetic field
The hadronic part of the electromagnetic current
is then proportional to the r meson field
What does it mean? Hadron matter couples to a
qqbar pair which propagates as a vector meson
which then materializes as a photon All QCD
complexity, gluon self interactions and

confinement are incorporated in the
physical vector
meson which forms the
intermediate state
m m-
Hadron medium
g
r
r
35
The r self-energy
The free rm field describes a bare meson which
we can interpret as the qqbar component of the
physical r meson. The bare r propagator is given
by
However, r is strong coupled to pions


? the physical r meson appears as a
broad resonance. properties accounted for by the
second order self-energy diagrams
36
The r dressed propagator
Without loss of generality
where
r polarization scalar
The full (dressed) propagator comes from an
infinite sum of diagrams with self-energy
insertions
1PI
1PI
1PI
This infinite series can be easily summed ...
The r field is always coupled to conserved
currents (qmJm 0) and so the terms proportional
to qmqn can be dropped
37
The imaginary part of Pr
r
r
scattering r?r

1PI
According to the optical theorem
general expression of the decay width
In this specific case the final state is
(dominantly) pp. Thus we come to the result
mass dependent width
38
The real part of Pr
Determines the mass shift due to the self-energy
Needed to keep the photon massless
Fixes c1
Regularization. Cut-off or dispersion relations
m0r can be fixed from the comparison to the
measured pp- ? pp- elastic cross section
The mass shift induced by the r?pp self-energy is
small
39
? is dressed with free pions
vacuum spectral function
(like ALEPH data V(t? 2pnt ))
40
Dilepton Rate in a strongly interacting medium
dileptons produced by annihilation of thermally
excited particles ??- in hadronic phase
qq in QGP phase
at SPS energies ? ? - ???µµ- dominant
hadron basis
photon selfenergy
spectral function
Vector-Dominance Model
41
Physics objective in heavy ion collisions
Study the properties of the r spectral function
Im Dr in a hot and dense medium
42
r spectral function in hot and dense hadronic
matter
Hadronic many-body approach Rapp/Wambach et
al., Weise et al.
hot matter
hot and baryon-rich matter
? is dressed with hot pions Prpp , baryons
Pr B (N,D ..) mesons Pr M (K,a1..)

• melts in hot and dense matter
• - pole position roughly unchanged - broadening
  mostly through baryon interactions

43
r spectral function in hot and dense hadronic
matter
Dropping mass scenario Brown/Rho et al.,
Hatsuda/Lee
explicit connection between hadron masses and
chiral condensate
universal scaling law

continuous evolution of pole mass with T and r
broadening at fixed T,r ignored
44
Final mass spectrum
continuous emission of thermal radiation during
life time of expanding fireball
integration of rate equation over
space-time and momenta required
example broadening scenario
45
Thus, the spectral function accessible through
rate equation, integrated over space-time and
momenta LimitationContinuously varying
values of temperature T and baryon density rB,
46
Comparison of predictions to data
Two possibilities, in principle 1) Use the
prediction for
Generate Monte Carlo events of g decays into
muon pairs Propagate through the acceptance
filter and compare to uncorrected data Done
presently for invariant mass (work in progress
for acceptance correction)
2) Correct the data for acceptance in 3-dim space
M-pT-y and compare them directly to predictions
Done for pT distributions
47
Acceptance filtering of theoretical prediction in
NA60
Input (example) thermal radiation based on RW
spectral function
all pT
Output spectral shape much distorted relative
to input, but somehow reminiscent of the spectral
function underlying the input by chance?
48
Comparison to the main models that appeared in
the 90s
Rapp-Wambach hadronic model predicting strong
broadening/no mass shift Brown/Rho scaling
dropping mass due to dropping of chiral condensate
Predictions for In-In by Rapp et al (2003) for
dNch/d? 140, covering all scenarios
Theoretical yields normalized to data in mass
interval lt 0.9 GeV
After acceptance filtering, data and predictions
display spectral functions, averaged over
space-time and momenta
Only broadening of ? (RW) observed, no mass
shift (BR)
49
Comparison to the main models that appeared in
the 90s

• Without baryons
• Not enough broadening
• Lack of strength below the r peak

• Improved model
• Fireball dynamics
• 4 p processes
• spectrum described in absolute terms

50
Semicentral collisions low vs high pT
Rapp-Hees
Rapp-Hees
Something is missing at high pT. What?
51
The vacuum r (and other) contributions
Ruppert-Renk
At high pT there is an important contribution
from the vacuum r r decays at kinetic
freeze-out Additional contribution Primordial r
(Rapp-Hees)
Rapp-Hees
52
The mass region above 1 GeV vector-axial vector
mixing
Above 1 GeV we can have contributions from 4p
processes. The spectral shape can be found for
instance from ee-?4p or studying (ALEPH) t?(2np)?
3p, 5p
2p, 4p, 6p
In addition, because of the pion heat bath, it
is possible also to have processes in which an
axial vector particle interacts with a pion, as
pa1?mm-. This effectively introduces a mixing
between vector and axial-vector states (at the
correlator level). This mixing depends on the
amount of chiral symmetry restoration
53
The mass region above 1 GeV models vs data
Ruppert / Renk, Phys.Rev.C (2005)
Rapp/Hees
Mass region above 1 GeV described dominantly in
terms of partonic processes, dominated by qqbar
annihilation
Mass region above 1 GeV described dominantly in
terms of hadronic processes, 4 p
? Hadron-parton duality
54
s(ee-?hadrons) in vacuum
e e-
p - p
r I 1
r
2p 4p ...
pp
e e-
h1 h2
r w f
KK
q q
_
qq

_
s sdual(1.5GeV)2 pQCD
continuum s lt sdual Vector-Meson
Dominance
55
Disentangling the signal sources in the IMR
The dileptons from charm decay can be identified
by tagging their production point with respect
to the primary interaction vertex

• Identify the typical offset of
• D-meson decay (100 µm)

• Need a very good vertexing accuracy
• (20-30 µm, in the transverse plane)

56
Measuring the muon offset
OffsetsdX, dY between the vertex and the track
impact point in the transverse plane at
Zvertex Resolution depends on track momentum use
offset weighted by the covariance matrices of the
vertex and of the muon track
For dimuons
57
Is the excess enhanced charm?
Procedure Fix the prompt contribution to the
expected DY yield and see if the offset
distribution can be described with enhanced Charm
dN/d?
New alignment
Answer No, Charm cant fill the small offset
region ? more prompts are needed
58
How many prompts are needed?
Procedure Leave both contributions free and see
if we can describe the offset distribution for
1.2 lt Mµµ lt 2.7
dN/d?
New alignment
Answer The best fit requires 2.6 times more
prompts than the expected Drell-Yan yield
59
Transverse momentum spectra
60
Spectra from a static fireball
In a static fireball at temperature T the
differential particle momentum distribution is
Lorentz invariant phase space element
Assume a thermal Boltzmann shape
? transverse mass spectra (integrated over
rapidity)
mT scaling all particle spectra have the same T
slope
61
An expanding fireball
Thermalized matter starts to expand because of
the pressure gradient with respect to the
surrounding vacuum. A collective motion (flow)
develops.
Flow velocity of a volume element of thermalized
matter in a spacetime point x
Sum all the particles 3-momenta and energies The
ratio gives the collective velocity
for completely random thermal motions
62
Fluid 4-velocity
Bjorken scaling At very high energies the
physics of secondary particle production should
be the same as described in different frames
moving along the z axis.
longitudinal flow field
Superimposed tranverse expansion ?
transverse flow
radial flow field
63
Dimuon emission
Excess dimuons continuum emission during all the
fireball lifetime (4-dim volume) we see not
only the emission at freeze-out!
In fluid local rest frame thermal pT
m-
g
Since the r is strongly coupled to the pions, the
thermal pT is boosted by flow in the lab frame
64
T vT anticorrelation
Example of hydrodynamic evolution (specific for
In-In Dusling et al.)
Monotonic decrease of T from early times to
late times medium center to edge
Monotonic increase of vT from early times to
late times medium center to edge
Potentially could permits to distinguish between
hadronic and partonic nature
? emission of dileptons sensing - Large T and
small vT at early times - Large VT and small T at
later time
65
Dilepton transverse momentum spectra
Obtained integrating dR/dq4 over fireball
space-time history
Superposition of spectra at different T weighted
by - Thermal factor exp(-E/T) (pmum dilepton
energy in local fluid rest frame) - Invariant
mass shape of spectral function - Volume
increase In addition, resonance decays determine
an overpopulation of pions ? Non zero chemical
potential mp(T) ? Fugacity exp(mp(T)/T)
66
Hadron pT spectra
When the temperature of a fluid element drops
below a certain value Tf , the mean free path
exceeds the dimesions of the system ?Thermal
equilibrium is broken and particles stream out
free to the detectors The isotherm T(r,t)Tf
defines a 3-dim hypersurface S in the space-time
? last-scattering surface
3-dim hypersurface ? divide S in infinitesimal
elements d3s
outward-pointing 4-vector perpendicular to S(x)
current of particles through x
number of particles passing through d3s
Total number of particles crossing S ? sum over
d3s
67
Cooper-Frye formula
With some mathematics one can show that
Integrated over f
Transverse flow-field
Once the mass is fixed (the particle is
specified), the function has only three
parameters vT, Tf and a normalization
68

• Common flow velocity in p,K,p and their
  anti-particles is seen at SPS and AGS energies

NA49/SPS results Common flow velocity seen for
very wideparticle species (Nucl.Phys A 715
61) Pion and deuteron are taken out from fit
procedure (many pions come from resonance decays
- deuterons are most likely produced with
proton-neutron coalescence) However, spectra
described are very well described with the
thermal parameter extracted with other particles
Common flow velocities are seen also in RHIC
Au-Au data (PHENIX and STAR)
69
Tf vT,f anticorrelation as a function of
centrality
Extracted with a two parameter fit to
experimental distributions
Evaluate c2 for fixed vT and Tf ?
Create a c2 map as a function of vT and Tf
Peripheral collisions shorter fireball lifetime
? less time to develop flow (smaller vT)
earlier decoupling at higher Tf Central
collisions bigger fireball lifetime ? more
time to develop flow (larger vT) later
decoupling at smaller Tf
Tf and vT,f are strongly anticorrelated
NA57 158 GeV Centrality classes 0 ? 40 to 53
most central 1 ? 23 to 40 most central 2 ? 11
to 23 most central 3 ? 4.5 to 11 most
central 4 ? 4.5 most central
70
Effect of radial flow on hadron pT spectra
Stable hadrons reflect the kinetic freeze-out
conditions. ? Fitting with exp(-mT/T) gives a T
dependent on the momentum range ? T from
exponential fit (call Tslope) is not anymore the
source temperature Tf. At high pT the spectra are
still exponential with a common slope which
reflects a freeze-out temperature blue-shifted by
the flow transverse velocity vT
At low pT, the pT spectra appear flattened and
mT scaling is broken. The T slope becomes mass
dependent (mT scaling is broken)
In principle allows to separate the thermal from
the collective motion
71
Mass ordering of hadronic slopes
? Flattening of spectra at low pT resulting in
higher Teff Pions softening at very low pT
because of resonance decays
Notice that for mi ? 0 we should see Ti,slope ?
Tf However, Ti,slope ? 170 MeV, while we know
that Tf 110-120 MeV for central Pb-Pb
collisions ? the linear approximation fails for
mi ? 0
72
f transverse momentum spectra
T slope extracted fitting
f pT spectra are corrected for acceptance after
background and side-window subtraction
73
T slope as a function of centrality
Fit with exp(-mT/Tslope) vs centrality increase
of Tslope (indication of radial flow)
NA60 (pT fit range 0-2.6 GeV)
NA49 (pT fit range 0-1.6 GeV) NA50 (pT fit range
1.2-2.6 GeV)
NA60 Preliminary
NA50 and NA49 differerences (f puzzle) Decay
channel (mm vs KK) pT fit range (high vs low)
The In-In measurement of NA60follows the NA49
systematics
74
Dimuon excess pT spectra
Divide the pT interval 0-2 GeV/c in 200 MeV
bins For each pT bin consider the mass projection
and determine the excess yield with the local
subtraction procedure
? pT spectrum of the excess
? Make this for 3 different mass windows
75
Strategy of acceptance correction
? reduce 3-dimensional acceptance correction
in M-pT-y to 2-dimensional correction in
M-pT, using measured y distribution as an
input ? use slices of ?m 0.1 GeV
and ?pT 0.2 GeV ? resum to three
extended mass windows 0.4ltMlt0.6
GeV 0.6ltMlt0.9 GeV
1.0ltMlt1.4 GeV
subtract charm from the data before acceptance
correction (based on IMR results we pospone
this discussion)
76
Dimuon excess pT spectra for three centrality bins
(spectra arbitrarily normalized)
hardly any centrality dependence ? integrate over
centrality
77
Centrality integrated excess pT spectra
(arbitrarily normalized at pT1GeV)

• significant mass dependence (also vs. mT, see
  below )
• possible origin
• different physics sources
• radial flow
• ? p-dependence of in-medium spectral
  function

78
Centrality integrated mT spectra
physics differences are better visible in mT-
than in pT
f mT spectrum nearly pure exponential Teff
nearly independent of fit range with some hint of
radial flow Excess spectra show an increase
(not flattening) at very low mT reminiscent of
pions Why?
79
Mass dependence of pT/mT spectra
differential fits to pT spectra, assuming locally
1-parameter mT scaling and using gliding windows
of ?pT0.8 GeV ? local slope Teff
at high pT, rho like region hardest, high-mass
region softest !
80
Systematics acceptance correction
pT spectrum of f at low pT much flatter (higher
Teff)
acceptance of f in betweenthat of the two mass
windows
? enhanced yield at low pT not due to incorrect
acceptance
81
mT spectrum of f nearly pure exponential
Teff of f nearly independent of fit range
? Different behaviour of excess not due to
incorrect acceptance
82
Uncertainty in combinatorial bkg subtraction
Estimated estimated through the comparison of
N/--(mixed) to N/--(real)
peripheral 1semiperipheral 0.8
semicentral 0.6central 0.8
corresponding fraction of CB for the four
centrality bins
83
enhanced yield at low-pT seen at all
centralities, including the peripheral bin
estimate of errors at low pT, due to subtraction
of combinatorial background peripheral
1semiperipheral 10 semicentral
20central 25
84
evolution of Teff vs mmm ?, ?, ?
Fit the spectra in the range 0.4-1.8 GeV/c
Linear rise the f seems to flow less
?
f
?
85
evolution of excess Teff vs mmm across the low
and intermediate mass
Fit the spectra in the range 0.4-1.8 GeV/c
Linear rise also for excess quite reminiscent of
radial flow of a hadronic source! But excess Teff
higher than hadron Teff. Why?
86
Evolution of the excess shape as a function of
centrality
Peak sitting on the continuum freeze-out r
without in-medium effects
The peak and the continuum can be disentangled in
the mass window 0.6-0.9 GeV with a simple shape
analysis by using side-windows
continuum 3/2(LU) peak C-1/2(LU)
Teff of continuum and peak can be measured
separately
87
evolution of excess Teff vs mmm across the low
and intermediate mass
Mass window 0.6-0.9 The peak Teff gets to 300
MeV! The continuum Teff drops to 230 MeV
88
evolution of excess Teff vs mmm across the low
and intermediate mass
Sudden drop at 1 GeV For Mgt1 GeV Teff is
roughly constant ? Seemingly non flow?
89
evolution of excess Teff vs mmm across the low
and intermediate mass
Summary In the region where 2p processes are
dominant (up to 1 GeV) there is strong evidence
for radial flow of dileptons. What is the
explanation for the drop? If the rise is truly
due to flow - the lack of flow above 1 GeV
could be naturally related to emission in an
early stage ? partonic processes - If the region
above 1 GeV is dominated by hadronic sources,
shouldnt Teff keep rising? How is the drop
explained in that case?