Using RHIC to figure out nature’s first liquid
outline
what we know about the quark gluon plasma at RHIC
how to experimentally extract QGP propertiesespecially from particle jets arising from
quarks/gluons traversing the plasma
the fate of the away side jetwhere does the energy go?tools to extract the jet contribution (<1%) from the
complex underlying eventthe speed of sound in the quark gluon plasma
Quarks, gluons, hadrons
6 quarks: 2 light (u,d), 1 sort of light (s)2 heavy (c,b), 1very heavy (t)flavor & color quantum numbers
Quarks are bound into hadronsBaryons (e.g. n, p) have 3Mesons (e.g. ): 2 (q + anti-q)
Colored quarks interact by exchange of gluons Quantum Chromo Dynamics (QCD)
Field theory of the strong interaction parallels Quantum Electrodynamics (QED)
EM interactions: exchanged photons electrically unchargedgluons carry color charge
QCD phase transition
Color charge of gluons gluons interact among themselves theory is non-abelian
curious properties at large distance: confinement of quarks in hadrons
+ +…
At high temperature and density: force is screened by produced color-chargesexpect transition to free gas of quarks and gluons
non-perturbative QCD – lattice gauge theory
T/Tc
Karsch, Laermann, Peikert ‘99
/T4
Tc ~ 170 ± 10 MeV (1012 °K)
~ 3 GeV/fm3
required conditions to study quark gluon plasma
~15% from ideal gas of weakly interacting quarks & gluons
42
30Tg
RHIC is the tool
Collide Au + Au for maximum volume & temperatures = 200 GeV/nucleon pair, p+p and d+A to compare
What we’ve learned at RHIC so far
PCM & clust. hadronization
NFD
NFD & hadronic TM
PCM & hadronic TM
CYM & LGT
string & hadronic TM
time0 10-22 sec
the initial state:gluon interactionlimits max. density“color glass condensate”?
quarks andgluonsinteract
nearly instantthermalization
quark gluonplasma : “perfect”liquid, veryopaque
final state hadrons freeze out
final state hadrons retain memory of conditions at freezeout→ analogous to CMB
thermalization and liquid properties
experimental observablesuse final state distributions
to probe dynamics at freezeout
hydrodynamicsconstrained by datayields information on properties
collective effects are a basic feature of plasmas!
search for particle final state patterns indicative of early dynamics – “elliptic flow”
dN/d ~ 1 + 2 v2(pT) cos (2) + …
“elliptic flow”
Almond shape overlap region in coordinate space
x
yz
momentum space
Hydro. CalculationsHuovinen, P. Kolb,U. Heinz
v2 reproduced by hydrodynamics
STARPRL 86 (2001) 402
• see large pressure buildup • anisotropy happens fast • early equilibration !
central
Hydrodynamics solves eqn. of motionEquation of state from lattice QCD
data say: Ti ~ 400 MeV, i ~ 15 GeV/fm ~ 0, i ~ 0.6 fm/c
Collective effect probes early phase
correct dependence of flow on mass requiressofter than hadronic EOS!!
Kolb, et al
magnitude scales with thenumber of quarksimplication: quarks are the relevant degrees of freedom when the pressure is built up.
measure transmission of colored probes
hadrons
q
q
hadronsleadingparticle
leading particle
schematic view of jet productionfirst chance hard scattering of q,g
QGP induces scattered quarks to radiate energy -> jet quenching
AA
AA
AA
ddpdT
ddpNdpR
TNN
AA
TAA
TAA /
/)(
2
2
nucleon-nucleon cross section<Nbinary>/inel
p+p
QGP very opaque to quarks & gluons
hadrons suppressedby factor of 5photons have no colorcharge no suppressionmechanism: gluon radiation
Pedestal&flow subtracted
test role of collisions using heavier charm quarks
electrons (0.5 MeV/c2 mass) stop in matter (bremsstrahlung) radiation dominant process
muons (mass = 106 MeV/c2) have long range radiation is suppressed by the large massdominant energy loss mechanism is via collisions
use heavy quarks as second kind of probe collisions should be important for c, b quarks masses are > 1, 4 GeV/c2
heavy quarks lose energy and flow too!
~ same E loss as u,d quarks energy loss not purely
radiativeneed collisions!
charm also flows thermalization with the light quarks??
what is the collision ?must exceed that for free qthermalization needs same!
plasma
ionized gas which is macroscopically neutralexhibits collective effects
interactions among charges of multiple particlesspreads charge out into characteristic (Debye) length, D
multiple particles inside this lengththey screen each other
plasma size > D
“normal” plasmas are electromagnetic (e + ions)quark-gluon plasma interacts via strong interaction
should expect screening and bound states should melt
lattice QCD can sort out
run
nin
g co
up
ling
coupling drops off for r > 0.3 fmso large bound states should melt
Karsch, et al.
Karsch, Kharzeev, Satz, hep-ph/0512239
so the larger bound states should melt
40% of J/ from and ’ decays
they are screened but direct J/ not?
obse
rved
/exp
ecte
d J
/
lattice says: larger than for free quarks
Lattice QCD shows ccresonant states at T > Tc, also implying high interaction cross sections
have multiple kinds of evidence for increased scatteringcross section & correlations which survive in QGPnatural to expect behavior as in other strongly coupled plasmas
generallya phenomenonin crystals butnot liquids
how does QGP transport energy from a jet?
g radiates energy, which gives akick totheQGP
1) can we detect evidence of asound (density) wave afterpassage of a jet?
2) does the induced correlationin q density distributionmanifest itself in particle production somehow?
experimental tool : 2 particle correlations
hadrons
q
q
hadronsleadingparticle
leading particle
schematic view of jet production
jet formation process in e+e- collisions :correlated hadrons
coneR
azimuth angle
Jet physics in PHENIX
Trigger:hadron with pT > 2.5 GeV/c
Count associated particles for each trigger at lower pT
(> 1 GeV/c) “conditional yield”
Near side yield: number of jet associated particles from same jet in specified pT bin
Away side yield: jet fragments from opposing jet
trigger“near side” < 90° jet partner
“away side” > 90° opposing jet
do a statistical analysis
CARTOON
flow
flow+jet dN
Ntrig d
includes ALL triggers(even those with no
associated particles inthe event)
jet Combinatorialbackground;collective flow causes additional correlation :
B(1+2v2(pTtrig)v2(pT
assoc)cos(2))
associated particles with non-flow angular
correlations -> jets!
1
combinatorial background large in Au+Au!
Detector acceptance induces correlations too
sources of uninteresting correlations: detector acceptance (square of singles acceptance), performance “holes”/dead channels
folding two 90° angular bites → pair acceptance
0 π
Acc
(Δφ
)
ΔΦ
Area = π
small probability tocatch pair of particleswith 90° opening angle
within each arm of PHENIXacceptance is nonuniformdetails vary slowly with time
complicated to correct analyticallypainful Monte Carlo simulation
MEASURE the acceptance correction from data
technique: “event mixing”select class of events with correlated pairspick one particle (trigger) from one event, partner
from a different eventensures no physics correlation
make exactly the same kinematic cuts on “fake” pairs from mixed eventsbin real and mixed events in the same way
collision location in detectorcollision impact parameter (centrality)
divide real by mixed event distribution
treat mixed events as “background”
divide foreground by this
to correct for acceptance
(mixed events)
dN
/d
resulting correlation function
normalize correction tofull acceptance
allows absolute count ofnumber of partners pertrigger
ncomb = <Ntrig/event><Npartner/event>
also remove the 3rd undesirable correlation
yesterday’s signal is today’s background!
ncomb counts probability of trigger-partner pairs in
underlying event (non-jet source) yet, we know these are not flat in !
elliptic flow → cos modulation
amplitude is measuredso we modulate ncomb by
1+2v2trigv2
partcos(2)
alas, we’re not done yet!
time dependent efficiency variationsonly mix events similar in time
or performancenormalize absolutely
for each group → residual multiplicity correlation
2 remaining non-jet correlations
real & mixed events differ in centralitymixed events sample this distribution twice, real once mixed have lower multiplicitycorrection depends on centrality bin width, mean & resolutiondetermined by simulating measured
probability distributions# NN collisions
N p
artn
er
Au+Au shows a sound-wave like pattern
peripheralcollisions(normal jet)
centralcollisions
system (in)dependence → medium property
g radiates energy, which gives akick totheQGP
+/-1.23=1.91,4.37 → cs ~ 0.33 (√0.33 in QGP, 0.2 in hadron gas)
How about the mysterious baryon excess?
Radiated gluons are collinear (inside jet cone)
Can also expect a jet “wake” effect,medium particles“kicked” alongside the jet by energy they absorb
And expect hard-soft combinations too C.M. Ko et al, Hwa & YangPRC68, 034904, 2003PRC67, 034902, 2003nucl-th/0401001 & 0403072
Fries, Bass & Muellernucl-th/0407102
Both increase probability of finding quarks near each other
again use 2 particle correlations
Select particles with pT= 2.5-4.0GeV/c
Identify them as mesons or baryons viatime-of-flight
Find second particle with pT = 1.7-2.5GeV/c
Plot distribution of the pair opening angles;integrate over 55°
Jet partner ~ equally likely for trigger baryons & mesons! Same side: slight decrease with centrality for baryonsDilution by combinations of 3 soft quarks
intermediate pT baryons ARE from jets
thermal quark combination
Dilutes jet partner yield
Conclusions
RHIC makes amazing new/old stuff behaves not like a weakly coupled gas, but like a strongly
coupled plasma we see structures in 2 particle correlations
away side like a sound wave in a shocked mediumnear side shows extra baryons compared to p+p
as expected if jet creates a bow wave have developed techniques to study < 1% level signals
on top of complex not-quite statistical backgroundsuse 2 particle correlationsmeasure the background in real datafold to determine “boring” correlationssuch techniques critical to extracting plasma properties
A word about the data handling
yearly raw data set size ~0.4 PB
large dataset size drives some limitations
cannot all be disk residentnot even post reconstructing detector info → particle tracks
insufficient server, disk, tape bandwidth for random access by ~200 analyzers
we run an “Analysis Train”single orderly pass over all data, utilizing many CPU’scopy a file from mass store to each CPUrun analysis modules from many users (current pass has 41)store histograms, small ntuples to diskcopy results onto shared diskaggregate multiple files per “run”
script driven by a single “conductor”
Data Archiving Rates
ATLASCMS
LHCb
ALICE
CDF
~25 ~40
~100 ~100
All in MB/sall approximate
~100
~150
~1250
~600
PHENIX Run-2
PHENIX Run-3
PHENIX Run-4
PHENIX Run-5
auxilliary slides
Is the energy density high enough?
5.5 GeV/fm3 (200 GeV Au+Au) well above predicted transition!
PRL87, 052301 (2001)
R2
2c
Colliding system expands:
dy
dE
cRT
Bj 22
11
02
Energy tobeam direction
per unitvelocity || to beam
value is lower limit: longitudinal expansion rate, formation time overestimated
Evolution of the Universe
Nucleosynthesis builds nuclei up to HeNuclear Force…Nuclear Physics
Universe too hot for electrons to bindE-M…Atomic (Plasma) Physics 104
K
E/M Plasma
Too hot for quarks to bind!!! Too hot for quarks to bind!!! 1012 KStandard Model (N/P) Physics
Quark-Gluon
Plasma??
Too hot for nuclei to bind 1010 KNuclear/Particle (N/P) Physics Hadron
Gas
SolidLiquidGas
Today’s Cold UniverseGravity…Newtonian/General
Relativity
Two Major Experiments to probe the Early Universe
With thanks to Tetsuo Hatsuda
WMAP
RHIC
4 complementary experiments
STAR
The Scope of the Tools (!)
STARspecialty: large acceptancemeasurement of hadrons
PHENIXspecialty: rare probes, leptons,
and photons
Benchmark the Probes in p+p collisions
calculable with perturbative QCD!
Produced photonsProduced photonsProduced pionsProduced pions
peripheralN
coll = 12.3 4.0
centralN
coll = 975 94
strongly interacting probe: a different story!
general explanation of jet quenching
Inelastic (radiative) energy loss in Au+Au
interaction of radiated gluons with gluons in
the plasma greatlyenhances the amount
of radiation
Is enough for fast equilibration & large v2 ?
Parton cascade using free q,g scattering cross sections underpredicts pressure must increase x50
Lattice QCD shows qqresonant states at T > Tc, also implying high interaction cross sections
Locate RHIC on phase diagram
Baryonic Potential B [MeV]
0
200
250
150
100
50
0 200 400 600 800 1000 1200
AGS
SIS
SPS
RHIC
quark-gluon plasma
hadron gas
neutron stars
early universe
thermal freeze-out
deconfinementchiral restauration
Lattice QCD
atomic nuclei
From fit of yields vs. mass (grand canonical ensemble):
Tch = 176 MeV B = 41 MeV
These are the conditions when hadrons stop interacting
T
Observed particles “freeze out” at/near the deconfinement boundary!
Possibility of plasma instability → anisotropy
small deBroglie wavelength q,g point sources for g fieldsgluon fields obey Maxwell’s equationsadd initial anisotropy and you’d expect Weibel instability
moving charged particles induce B fieldsB field traps soft particles moving in A directiontrapped particle’s current reinforces trapping B fieldcan get exponential growth
(e.g. causes filamentation of beams)doesn’t require strong coupling
could also happen to gluon fields early in Au+Au collisiontimescale short compared to QGP lifetimebut gluon-gluon interactions may cause instability to
saturate → drives system to isotropy & thermalization
Transport properties
transport of particles → diffusion
transport of energy by particles → thermal conductivity
transport of momentum by particles → viscosity
transport of charge by particles → electrical conductivityis transport of color charge an analogous question for us?
transport in plasmas is driven by collisions
Other strongly coupled plasmas
Inside white dwarfs, giant planets, and neutron stars (n star core may even contain QGP)
In ionized gases subjected to very high pressures, magnetic fields, or particle interactions
Dusty plasmas in interplanetary space & planetary rings Solids blasted by a laser We would like to know:
How do these plasmas transport energy?How quickly can they equilibrate?What is their viscosity? >10 can even be crystalline! How much are the charges screened? Is there evidence of plasma instabilities at RHIC? Can we detect waves in this new kind of plasma?
nove
l pla
sma
of
str
ong
inte
ract
ion
Strategy for conditional yields
Quality cutspT selection(PID cuts)
(Seed)trigger partner list
Pair cutsRaw correlation
Pair cutsEnsure sameacceptance
background
subtract
Subtracted correlation
acceptancenormalizeto or 2
divide PartnerConditional yield
(uncorrected)
Conditional Yield (true)
Partner efficiency correction:track efficiency, quality, etc.
norm by events
Mixed events
nB
perP.S.
others say maybe collisions not needed
BUT v2 is small…
diffusion = transport of particles by collisions
PHENIX preliminary
Moore & TeaneyPRC71, 064904, ‘05
D ~ 3/(2T) is small! → strong interaction of c quarks
larger D →less charm e loss fewer collisions, smaller v2
D = 1/3 <v> mfp = <v>/ 3D collision time → relaxation time
not an experimental artefact, part IIAu+Au Central 0-12% Triggered
Δ1
Δ2
d+Au
Δ1
J. Ulery
deposited energy doesn’t thermalize so fast
T. Renk
distribution +longitudinal expansion depopulate region & shift Mach peak
put together to get conditional Yields
Combinatoric background level determined by convolution of trigger and associated particle rate
v2 values taken from PRL 91 (2003) 182301 modulates combinatoric level by 1+2v2(pT
trig)v2(pTassoc)cos(2)
(solid lines in plot)
Trigger pT: 2.5-4.0GeV/c
Associated pT: 1.7-2.5GeV/c
QGP plasma properties known, so far
Extract from models, constrained by data
Energy loss <dE/dz> (GeV/fm) 7-10 0.5 in cold matter
Energy density (GeV/fm3) 14-20 >5.5 from ET data
above hadronic E density!
dN(gluon)/dy ~1000 From energy loss, hydro huge!
T (MeV) 380-400
Experimentally unknown as yet
Equilibration time0 (fm/c) 0.6 From hydro initial condition; cascade agrees very fast!
NB: plasma folks have same problem & use same technique
Opacity (L/mean free path) 3.5 Based on energy loss theory
viscosity ~0 hydro constrained by flow
Plasma properties we will measure at RHIC II
property measurement
T as fn. of
equation of state particle flows as fn. of critical point location
screening length onium spectroscopy
(x,v) jet tomography
diffusion open C, B spectra & flow
viscosity strange & charmed hadron flows
used to constrain 3d hydro
energy transport >2 particle correlations vs. T, pT
something new? follow up on surprises…
to explore at RHIC II ≥ 2014
property measurement challengequantify screening length
Y(2s), Y(3s)
c in Au+Au
statistics (acceptance) resolution? (~100 MeV)
S/B, granularity?
medium modified fragmentation fn.
-identified hadron correlations
>5 GeV/c h statistics (acc)
direct tag/decay subtract.
(granularity?? acceptance)
chiral symmetry chiral partners
(a1, )
? doable? granularity?
thermalization time flow of high pT non- 0
di-hadrons pT>20 GeV
acceptance, trigger, momentum resolution
plasma parton correlations
? something new?