A comparative measurement of phi -> K+K- and phi -> mu(+)mu(-) in In-In collisions at the CERN SPS
ABSTRACT The NA60 experiment at the CERN SPS has studied $\phi$ meson production in
In-In collisions at 158A GeV via both the $K^+K^-$ and the $\mu^+\mu^-$ decay
channels. The yields and inverse slope parameters of the $m_T$ spectra observed
in the two channels are compatible within errors, different from the large
discrepancies seen in Pb-Pb collisions between the hadronic (NA49) and dimuon
(NA50) decay channels. Possible physics implications are discussed.
arXiv:1104.4060v1 [nucl-ex] 20 Apr 2011
A comparative measurement of φ → K+K−and φ → µ+µ−in In-In
collisions at the CERN SPS
R. Arnaldia, K. Baniczb,c, J. Castord, B. Chaurande, W. Chenf, C. Cical` og, A. Collah, P. Corteseh,
S. Damjanovicb,c, A. Davidb,i, A. de Falcoj,∗, A. Devauxd, L. Ducrouxk, H. En’yol, J. Fargeixd,
A. Ferrettih, M. Florisj, A. F¨ orsterb, P. Forced, N. Guettetb,d, A. Guichardk, H. Gulkanianm,
J. M. Heuserl, M. Keilb,i, Z.Lif, C. Louren¸ cob, J. Lozanoi, F. Mansod, P. Martinsb,i, A. Masonig, A. Nevesi,
H. Ohnishil, C. Oppedisanoa, P. Parrachob,i, P. Pillotk, T. Poghosyanm, G. Pudduj, E. Radermacherb,
P. Ramalheteb, P. Rosinskyb, E. Scomparina, J. Seixasi, S. Sercij, R. Shahoyanb,i, P. Sondereggeri,
H. J. Spechtc, R. Tieulentk, A. Urasj, G. Usaij, R. Veenhofb, H. K. W¨ ohrij,i
aINFN sezione di Torino, Italy
bCERN, 1211 Geneva 23, Switzerland
cPhysikalisches Institut der Universit¨ at Heidelberg, Germany
dLPC, Universit´ e Blaise Pascal and CNRS-IN2P3, Clermont-Ferrand, France
eLLR, Ecole Polytechnique and CNRS-IN2P3, Palaiseau, France
fBNL, Upton, NY, USA;
gINFN sezione di Cagliari, Italy
hUniversit` a di Torino and INFN, Italy
iInstituto Superior T´ ecnico, Lisbon, Portugal
jUniversit` a di Cagliari and INFN, Cagliari, Italy
kIPN-Lyon, Universit´ e Claude Bernard Lyon-I and CNRS-IN2P3, Lyon, France
lRIKEN, Wako, Saitama, Japan
mYerPhI, Yerevan Physics Institute, Yerevan, Armenia
The NA60 experiment at the CERN SPS has studied φ meson production in In-In collisions at 158A GeV
via both the K+K−and the µ+µ−decay channels. The yields and inverse slope parameters of the mT
spectra observed in the two channels are compatible within errors, different from the large discrepancies
seen in Pb-Pb collisions between the hadronic (NA49) and dimuon (NA50) decay channels. Possible physics
implications are discussed.
Heavy Ion Collisions, φ puzzle
The theory of Quantum Chromo-Dynamics (QCD)
predicts that matter under extreme conditions of
temperature and energy density undergoes a phase
∗Corresponding author at:
INFN, Cagliari, Italy
Email address: email@example.com
(A. de Falco)
Universit` a di Cagliari and
transition from hadronic matter to a plasma of de-
confined quarks and gluons (QGP). The occurrence of
this phase can be studied in the laboratory by means
of high energy heavy-ion collisions. Strangeness en-
hancement has been suggested as a signature of QGP
formation . The φ meson, being composed of an
ss pair, is an ideal probe for the study of strangeness
production. Moreover, the φ mass and branching ra-
Preprint submitted to Physics Letters BApril 21, 2011
Table 1: Centrality bins, measured number of φ, ratio between background and signal, measured φ mass and width.
29-42 411.5 · 104
42-95781.3 · 105
96-160 1336.5 · 105
161-250 1776.8 · 105
1019.9± 1.76.5 ± 1.2
7.5 ± 1.1
7.4 ± 2.3
Table 2: φ inverse slope and yield extracted from the analyses in the muon and kaon channels as a function of centrality.
eff(MeV)?φ?µµ(pT> 0.9 GeV)TKK
15209 ± 40.044± 0.002 ± 0.005
41 232 ± 40.197± 0.009 ± 0.021
78245 ± 40.48 ± 0.02 ± 0.05
133 250 ± 41.03 ± 0.03 ± 0.07253 ± 11 ± 5
177249 ± 51.65 ± 0.06 ± 0.07254 ± 13 ± 6
?φ?KK(pT > 0.9 GeV)
0.16 ± 0.05 ± 0.08
0.47 ± 0.06 ± 0.02
1.01 ± 0.07 ± 0.05
1.49 ± 0.11 ± 0.07
0.64 ± 0.17 ± 0.03
1.60 ± 0.18 ± 0.06
3.5 ± 0.2 ± 0.2
4.6 ± 0.3 ± 0.4
tios for the decay into kaon and lepton pairs may be
modified in the medium [2, 3].
φ meson production was first studied at the CERN
SPS by the NA49 [4, 5] and NA50 [6, 7] experiments.
NA49 detected the φ through its decay into K+K−
pairs, while NA50 studied the φ → µ+µ−channel.
The results obtained by the two experiments in Pb-
Pb collisions show discrepancies both in the yield and
in the inverse slope parameter Teffof the mTspectra.
The φ multiplicity in central Pb-Pb collisions mea-
sured by NA50 in the dimuon channel is higher by
about a factor of 4 with respect to the corresponding
NA49 measurement in the K+K−channel. The Teff
value found by NA50 is about 220-230 MeV, showing
a mild dependence on centrality, while NA49 mea-
sured an inverse mT slope that increases with cen-
trality and saturates at Teff∼ 300 MeV. It has to be
stressed that the NA50 acceptance is limited to high
pT, while NA49 is dominated by low pT.
The discrepancy between the NA49 and NA50 re-
sults, known as “φ puzzle”, triggered a considerable
effort to explain the observed differences [8, 9, 10, 11].
It was argued that in-medium effects may affect the
spectral function of the φ, causing a modification of
its mass and partial decay widths. Moreover, kaon
absorption and rescattering in the medium can re-
sult in a loss of signal in the region of the φ invariant
mass in the K+K−channel, thus reducing the ob-
served yield. This effect would be concentrated at
low pT, causing a hardening of the pT spectrum in
this channel [11, 9]. Nevertheless, according to those
calculations, the yield in lepton pairs is expected to
exceed the one in kaon pairs by about 50%, which
is much lower than the observed differences. More
recently, the CERES experiment at the SPS studied
φ production in Pb-Pb collisions both in the K+K−
and dielectron channels, finding an agreement with
the NA49 results . However, CERES’ measure-
ment of the Teffparameter is affected by a large sta-
tistical error that does not allow to draw firm conclu-
The NA60 experiment measured φ production in In-
In collisions at the CERN SPS. The φ meson is de-
tected through its decay both in muon and kaon pairs.
In this paper, results on the φ → K+K−channel will
be presented and compared to the already published
ones for the φ → µ+µ−decay mode  and to the
existing SPS measurements in different systems.
The detectors relevant for the present work are a
muon spectrometer (MS) inherited from NA50 and
a vertex telescope (VT). They are fully described
in [14, 15]. The muon spectrometer is composed of a
toroidal magnet, 8 multi-wire proportional chambers
for the tracking and a set of scintillator hodoscopes
that provides the main trigger, which selects muon
pairs coming from a common vertex. It is preceded
by a 12 λI thick hadron absorber.
high-granularity, radiation tolerant Si pixel detector,
placed between the target and the absorber in a 2.5 T
dipole field. It is used to determine the primary ver-
The VT is a
tex, the charged particle multiplicity and to measure
the momentum of the charged tracks.
The sample used for the results presented in this pa-
per was collected with a 158 A·GeV In beam im-
pinging on a 0.17 λI thick In target, composed of 7
subtargets placed in vacuum. The acquired statis-
tics consists of 230 million triggers, mainly dimuons.
More than 99% of the dimuon events consist of un-
correlated pairs coming from the decay of pions and
kaons in muons, or non-muon tracks. Since the tracks
used in the present analysis are reconstructed using
the Si tracker alone, the dimuon trigger acts like a
minimum bias trigger. The distortion of the multi-
plicity distribution introduced by the dimuon trigger
drops out, since the results are obtained normalizing
the number of φ → K+K−obtained in each multi-
plicity class to the number of events in that class, as
discussed later in the text. Therefore, no bias is in-
troduced by the dimuon trigger.
The charged tracks associated to the vertices recon-
structed in the vertex tracker are used for this analy-
sis. In order to avoid events with reinteractions of nu-
clear fragments in the subsequent targets, only events
with one vertex in the target region are selected. The
vertices are required to lay inside one of the Indium
subtargets. In addition, this cut eliminates events
with ambiguous vertices and pile-up events, and re-
jects about 55% of the statistics.
The tracks are selected requiring that the reduced
χ2of the track fit is lower than 3.
centrality is determined through the measurement
of the charged particle multiplicity.
of produced charged particles Nchis extracted from
the raw charged track multiplicity, applying a cor-
rection that takes into account the detector accep-
tance, reconstruction efficiency and secondary par-
ticle production.The number of participants is
then obtained assuming Npart∝ dNch/dη. We find
Npart≈ dNch/dη|3.7, with a systematic error of
about 10% for peripheral collisions and 5% for cen-
The analysis in the K+K−channel is performed as-
suming that all the charged particles associated with
the primary vertex are kaons, and building all the
possible opposite sign pairs among the tracks of each
event (as discussed below, like sign pairs are used
for systematic checks and normalization purposes).
This results in a considerable combinatorial back-
ground, which is reduced by a factor 10 in the φ
mass range applying a cut on the pair’s opening an-
gle, 0.005 < θKK < 0.15 rad. Due to the low ac-
ceptance for the signal at low pT, and to the cor-
responding high background to signal ratio (about
three times higher than at high pT) a cut on the pair
transverse momentum pT > 0.9 GeV/c was applied.
Furthermore, in order to exclude the borders of the
detector acceptance, only tracks in the rapidity re-
gion 2.9 < y < 3.7 are selected, where the rapidity is
calculated assigning the kaon mass to the tracks. Af-
ter applying these cuts, the ratio between background
and signal ranges from ∼ 190 to ∼ 460, depending on
The residual background is subtracted with an event
mixing technique: events from runs taken in ho-
mogeneous conditions are grouped in pools accord-
ing to the position of the target associated to the
vertex, the centrality of the collision and the direc-
tion of the event plane .
imuthal angle is calculated by means of the flow vec-
tor.Its components for the second harmonic are
where the weight wi associated to the i − th track
is its transverse momentum and ϕi is its azimuthal
angle. Tracks from different events belonging to the
same pool are mixed. Each event is mixed with two
other events. Mixed events are subject to the same
cuts as the ones applied to the real data.
The obtained mixed spectra are normalized asking
that in the mass region 1.02 < m < 1.06 GeV/c2the
number of mixed like-sign pairs coincides with the
corresponding one in the real data. Changing the
mass range yields negligible differences in the results.
Alternatively, the normalization factor is chosen such
that the integral of the opposite sign mass spectra in
the whole mass range is equal for the real and mixed
samples. For both methods, the normalization fac-
tor is evaluated either for each subtarget separately,
or averaged over the targets. The differences in the
results coming from the choice of the normalization
criterion, causing variations of about 4% in the de-
termination of the number of φ mesons per centrality
bin, are taken into account in the systematic error.
The event plane az-
iwicos(2ϕi) and Qy
1 1.051.1 220.127.116.11
Signal (real - mixed)
Figure 1: Invariant mass spectrum of the opposite sign pairs af-
ter combinatorial background subtraction for pT> 0.9 GeV/c
integrated in centrality.
The invariant mass spectrum of the opposite
sign pairs after combinatorial background subtrac-
tion (normalized with the like-sign pairs) is shown in
Fig. 1, integrated in centrality. The function for the
φ peak is determined fitting the mass distribution ob-
tained with an overlay Monte Carlo simulation, that
consists in reconstructing a generated φ decaying into
a kaon pair on top of a real event. We use a gaussian
superimposed to an empirical function that takes into
account the mass tails and accounts for about 5%
of the total number of φ. Several functions and fit
mass ranges have been tested to describe the resid-
ual background, including polynomials of first and
second order and functions that are null at the KK
mass threshold. In order to check that the fit func-
tions correctly reproduce the signal and the residual
background, the fit procedure has been applied to
the like-sign invariant mass spectra in several pT in-
tervals, after the subtraction of the event-mixed spec-
tra. The like-sign pairs spectra have a shape which is
very similar to the corresponding opposite sign ones
for masses above the φ peak (m > 1.04 GeV/c2),
and show a smooth trend down to m = 1 GeV/c2.
Since these pairs do not contain any signal, then a fit
applied to them should result in a null signal com-
ponent, while the presence of a fictitious peak would
indicate a bias in the fitting procedure. The fit ap-
plied to the like-sign pairs gives a signal component
compatible with zero in all the pT bins, showing that
no artifact is introduced by the choice of the fit func-
It has to be noted that if the φ peak position and
width are left as free parameters of the fit, the corre-
sponding values are mφ= 1019.5±0.3±1.2 MeV/c2
and σm = 7.8 ± 0.3 ± 1.2 MeV/c2, in good agree-
ment with the simulations. Corresponding results as
a function of centrality are reported in Table 1, to-
gether with the number of reconstructed φ mesons,
obtained by integrating the function describing the
φ peak, and the ratio between background and sig-
nal in the mass range 1.005 < m < 1.035 GeV/c2.
Statistical and systematic errors for mφand σmare
added in quadrature. The values show no dependence
on centrality, indicating that no modification of the
spectral function is visible within the sensitivity of
the measurement. A similar result is obtained in the
dimuon channel . The results shown in the follow-
ing are obtained fixing mφand σmto 1.019 GeV/c2
and 7.8 MeV/c2, according to the Monte Carlo val-
In order to extract the mT distributions, the fit to
the invariant mass spectra is performed in several pT
intervals having a size of 200 MeV/c. It was checked
that results do not depend on the choice of the bin
size. The mT distributions are then corrected for the
geometrical acceptance and reconstruction efficiency
with an overlay Monte Carlo simulation.
pidity and decay angle distributions used for the φ
generation are tuned to the ones measured in muons.
The former is a gaussian with σy = 1.13, while the
dN/dcosθ distribution is assumed to be flat .
The acceptance in multiplied by reconstruction ef-
ficiency is almost flat for pT> 0.9 GeV/c, about 15%
in full rapidity. A reliable mT distribution could be
extracted only for semicentral (?Npart? = 133) and
central (?Npart? = 177) collisions, while for semipe-
ripheral collisions (?Npart? = 78), due to statistics
limitations, the resulting inverse slope parameter was
not stable when varying the analysis criteria. Results
are reported in Fig. 2, where the distributions in the
0 0.51 1.52
KK Central (x 2) KK Central (x 2)
Central (x 2) Central (x 2)
KK Semi-Central KK Semi-Central
Figure 2: Normalized transverse mass spectra for semi-central
(triangles) and central (circles) collisions for the φ → K+K−
channel (full symbols) compared to the corresponding ones in
muon pairs (open symbols) . Results for central collisions
are scaled by a factor 2 in order to improve readibility. Colour
φ → µµ channel in the same centrality intervals are
reported for comparison.
malized to the φ multiplicity as shown in Table 2
(details on the method used to measure the φ yield
are reported below). The mT spectra are fitted with
the function 1/mTdN/dmT ∝ e−mT/Teff.
duced χ2is ∼ 1 in both cases. Results are reported
in Table 2. The systematic error is dominated by the
choice of the function used for the background. The
systematic errors due to the choice of the normaliza-
tion criterion, to the variation of the analysis cuts and
of the starting point of the fit to the mass spectra are
also taken into account. Results are in good agree-
ment with the ones obtained in dimuons in the full
pT range, also reported in Table 2. Since in presence
of radial flow the Teff value may depend on the pT
range, the fit to the dimuon spectra was restricted to
the range pT> 0.9 GeV/c, giving Teff= 252±4 and
247 ± 3 MeV for semicentral and central collisions,
still in agreement with the measurement in kaons.
The raw φ multiplicity is determined fitting the
mass spectra for pT > 0.9 GeV/c and dividing the
All the spectra are nor-
0 50 100150200
Figure 3: ?φ?/Npart as a function of the number of partici-
pants in In-In collisions in the φ → K+K−(full circles) and
φ → µµ (open crosses) channels for pT > 0.9 GeV/c. Boxes:
systematic errors. Colour online.
number of φ mesons obtained by the total number
of events selected for this analysis.
then corrected for the branching ratio in kaon pairs,
(49.2 ± 0.6)%  and for the acceptance, evaluated
through a Monte Carlo simulation. The inverse mT
slopes used for the calculation of the acceptance are
the ones measured in the kaon channel, where avail-
able; otherwise the values obtained in dimuons are
used. Table 2 reports the results for the φ multiplicity
in the dimuon and kaon channels for pT> 0.9 GeV/c.
The main contributions to the systematic error are
given by the uncertainty in the choice of the fit func-
tion for the residual background component in the fit
of the invariant mass spectra and the normalization
criterion. The systematic errors due to the variation
of the analysis cuts and of the starting point of the
fit to the mass spectra are also taken into account.
In Fig. 3 the ratio ?φ?/Npart as a function of Npart
in the φ → K+K−channel for pT > 0.9 GeV/c is
compared to the corresponding one in dimuons. The
additional contribution due to the systematic error
in Npart, affecting both channels in the same way,
is not displayed. It can be seen that the yield in
This value is
the hadronic channel and the one in the dileptonic
channel are in agreement within the errors. A ratio
between the φ yields in dimuons and in kaons larger
than 1.18 in the common pTrange is excluded at 95%
The φ multiplicity in kaons was also calculated in
the full pT range. Results are reported in the last
column of Table 2. The systematic error due to the
uncertainty in the extrapolation to full pT caused
by the error in the inverse slope was taken into ac-
count. It has to be stressed that the extrapolation
to pT = 0 was done under the hypothesis that the
inverse slope does not change. Models like AMPT 
predict that kaon rescattering causes a depletion in
the number of reconstructed φ → KK which is con-
centrated at low pT, causing an increase of Teff in
that region. In Pb-Pb central collisions, the effect
would be already visible for mT− m0 ≥ 0.34 GeV,
corresponding to the region covered in this analysis.
In that region, the difference between Teff in kaon
and muon pairs would range from 30 to 50 MeV, and
the fractional loss in kaons at mT− m0∼ 0.34 GeV
would range from 35% to 50%. Such effects are not
seen in our data, suggesting that if any rescattering
effect is present, it would be concentrated at lower
pT. In the range 0 < mT − m0 < 0.34 GeV the
AMPT model predicts an inverse mT slope of about
330 MeV, close to the value observed by NA49 in
central Pb-Pb collisions. Since no theoretical pre-
dictions for In-In collisions are available, we esti-
mated the effect of the change of slope in the ex-
trapolation to pT = 0 under the extreme hypothesis
Teff= 330 MeV for 0 < mT− m0< 0.34 GeV, while
for mT−m0≥ 0.34 GeV the measured value is kept.
This variation would cause a reduction of the φ yield
in kaons of about 12%. This quantity can be consid-
ered as a conservative estimation of the uncertainty
in the extrapolation due to a possible suppression
mechanism of the hadronic channel.
In order to compare to other collision systems, the
inverse slope and the enhancement, quantified as the
ratio ?φ?/Npart in the full pT range, are plotted in
Fig. 4 as a function of the number of participants for
central C-C, Si-Si, In-In and Pb-Pb collisions [4, 5].
For this comparison, the systematic error on Npart, of
about 5% in central collisions, is taken into account
NA49 KK (C / Si)
0 50 100 150 200 250 300 350 400
Figure 4: Inverse slope (top) and ?φ?/Npart in full pT range
(bottom) as a function of Npart for central collisions. Boxes:
systematic errors. Colour online.
in the calculation of the ratio ?φ?/Npartfor the NA60
The inverse slope shows an initial fast increase at
low Npart values, that becomes less pronounced go-
ing towards higher Npart. A lower value is observed
by NA50 as compared both to the CERES and NA49
measurements in Pb-Pb and to the NA60 In-In points
in the hadronic and dileptonic channels.
As stated above, in the presence of radial flow Teff
depends on pT. The NA60 analysis in dimuons ,
performed at low and high pT ranges, shows a dif-
ference limited to about 15 MeV in In-In. This fact,
complemented by the detailed blast wave analysis of
the dimuon data discussed in , makes it difficult
to ascribe only to radial flow the large differences in
Teff shown in Fig. 4. A further flattening caused by
kaon rescattering and absorption may lead to larger
Teffvalues in the hadronic channel in Pb-Pb.
Concerning the enhancement, the NA60 measure-
ments in the dilepton and hadron channels can differ
up to 18% (at 95% CL) for pT > 0.9 GeV/c and
22% in full pT considering the uncertainty in the
extrapolation arising from a possible suppression
of the hadronic channel at low pT.
NA50 result is extrapolated to full phase space
= 220 MeV, according to the NA50
measurement in peripheral collisions. Even assuming
as an extreme case Teff = 300 MeV, as obtained by
NA49 in central collisions, the NA50 enhancement
would exceed by a factor of ∼ 2 the central Pb-Pb
values measured in kaons.
The yield per participant measured in In-In exceeds
the one observed in Pb-Pb in the hadronic channel
(both NA49 and CERES) by about 30%.
might suggest a suppressing mechanism for the kaon
channel, below experimental sensitivity in In-In,
which shows up in Pb-Pb. The CERES measurement
in dielectrons is in agreement with the one in kaons
but, given the measurement errors, it cannot rule
out differences of the order of 40-50%, expected by
models including kaon rescattering. If one assumes
that no suppression mechanism affects the Pb-Pb
collisions, it is then difficult to understand why the
φ multiplicity in central Pb-Pb is smaller than in
In conclusion, the inverse slopes and yields measured
in the kaon and muon channels in In-In collisions
are in agreement, excluding for central collisions a
difference in the yields larger than 18 % (at 95 %
C.L.) in the common pT range.
modification of the φ mass and width is observed
as a function of centrality. When the comparison is
extended to other systems, it is difficult to reconcile
all of the observations into a coherent picture, albeit
there is some hint for a possible physics mechanism
leading to a difference in the two channels in Pb-Pb
collisions, while producing no remarkable difference
in In-In collisions.
In Fig. 4 the
In addition, no
 J. Rafelski, B. Muller, Phys. Rev. Lett. 48 (1982)
 D. Lissauer, E. V. Shuryak, Phys. Lett. B253
 F. Klingl, T. Waas, W. Weise, Phys. Lett. B431
 V. Friese, Nucl. Phys. A698 (2002) 487–490.
 C. Alt, et al., Phys. Rev. Lett. 94 (2005) 052301.
 B. Alessandro, et al., Phys. Lett. B555 (2003)
 D. Jouan, et al., J. Phys.s G35 (2008) 104163+.
 E. V. Shuryak, Nucl. Phys. A661 (1999) 119–
 S. Pal, C. M. Ko, Z.-w. Lin, Nucl. Phys. A707
 E. Santini, G. Burau, A. Faessler, C. Fuchs, Eur.
Phys. J. A28 (2006) 187–192.
 S. C. Johnson, B. V. Jacak, A. Drees, Eur. Phys.
J. C18 (2001) 645–649.
 D. Adamova, et al., Phys. Rev. Lett. 96 (2006)
 R. Arnaldi, et al., Eur. Phys. J. C64 (2009) 1–18.
 M. Keil, et al., Nucl. Instrum. Meth. A546
 G. Usai, et al., Eur. Phys. J. C43 (2005) 415–
 J. Adams, et al., Phys. Lett. B612 (2005) 181–
 A. M. Poskanzer, S. A. Voloshin, Phys. Rev. C
58 (1998) 1671–1678.
 W. M. Yao, et al., J. Phys. G33 (2006) 1–1232.