Anisotropic flow from perturbations in Gubser solutions in relativistic heavy ion collisions

Tabish Shibli

Supervisor: Subrata Pal

Department of Nuclear and Atomic Physics
Tata Institute of Fundamental Research, Mumbai

We present formulation of linearized perturbations of the hydrodynamic fields on top of conformal Gubser
flow that has boost-invariant solution with rotational symmetry in the transverse plane. We solved numerically
the evolution of the perturbations for the equations of motion of the hydrodynamic fields. Using the Cooper-
Frye method to convert the fields to particles, we obtain the spectra of hadrons produced in relativistic heavy ion
collisions. We studied the azimuthal distribution of hadrons and anisotropic flow, in particular the elliptic flow,
from these perturbations which creates azimuthal asymmetry in central collisions. The results are compared
with the available data for Pb-Pb collisions at Large Hadron Collider energies.

I. INTRODUCTION

Hydrodynamics is perhaps the simplest description of the dynamics of a many-body system. This is because the complicated
short-distance and short-time interactions of a particle are averaged out. Therefore, the effective degrees of freedom to describe
the system reduce to a handful of conserved charges and their currents instead of some multiple of the number of particles in the
system. The hydrodynamics equations are just the conservation laws and an equation of state. In spite of their apparent simplicity,
they can explain a vast number of macroscopic phenomena. High energy heavy-ion collisions are studied both experimentally
and theoretically to obtain information about the properties of nuclear matter under extreme conditions of high baryon density
and/or temperature. Ultra-relativistic heavy ion collisions at the Relativistic Heavy-Ion Collider (RHIC) and the Large Hadron
Collider (LHC) at CERN have produced a new state of matter, the quark gluon plasma with a lifetime of τ = 5 − 7f m/c and
volume up to a few thousand f m3 depending on the centrality of the collision.
Analytic solutions of the Navier-Stokes equations are infrequent. Bjorken flow [3] is an attempt to describe the average
motion of partons after a collision. This framework assumes an approximate boost-invariance along the beam direction about
mid-rapidity, translation invariance in the transverse plane, and rotation invariance in that plane. These symmetries become
manifest in Milne coordinates (τ, x⊥ , φ, η) as the translation and rotation invariance in the transverse plane imply that none of
the fields can depend on x⊥ or φ, while boost-invariance (if exact) implies that these quantities also do not depend on the space-
time rapidity η. These symmetries along with invariance under reflections η → −η, imply that the local four-velocity vector is
uµ = (1, 0, 0, 0) in the Milne coordinate system. One does not need any information about the equation of state to determine the
four-velocity profile. Symmetry considerations alone fix it. The medium need not even be equilibrated for uµ = (1, 0, 0, 0) to be
a well-motivated choice for the average four-velocity in the mid-rapidity region. If the medium is equilibrated and has equation
of state p = /3, then the energy density in the local rest frame scales as 1/τ 4/3 .
A major problem with this framework is the translation invariance in the transverse plane which is unrealistic since the nucleus
is only about 13 fm across. Translation invariance also assumes that the medium has no average local velocity in the x⊥ direction
(radial flow) until after it thermalizes. This is probably wrong, even for perfectly central collisions, and one might worry that it
significantly distorts the subsequent hydrodynamical flows on which much of heavy-ion phenomenology depends. Indications
that the absence of initial radial flow could be problematic can be found, for example, in Refs. [4] [5].
Gubser obtained a new solution to the relativistic conformally invariant Navier-Stokes equation that has boost-invariance with
transverse fluid velocity [1]. Specifically, the solution is a generalization of Bjorken flow [3] and describes a boost-invariant
medium expanding both longitudinally and radially, which makes it an attractive toy model for heavy ion collisions. Here,
we will consider the system to be conformally invariant as the QCD processes well above the confinement scale are close to
conformally invariant, this happens because the coupling runs only logarithmically with scale. This generalized Bjorken flow,
aka Gubser flow, with the assumption that translations in the transverse plane can be replaced by certain special conformal
transformation.
The key point is that the initial state and the dynamics of the collision respect relativistic conformal invariance and also, this
method is also limited to perfectly central collisions. These assumptions are by no means exactly satisfied by real collisions.
Here, the symmetries are as follows,

1. Translation: ξ µ = aµ , aµ is a constant.

2. Rotation and Boost: ξ µ = ωνµ xν , where, ωµν is constant and antisymmetric.

 3. Scale transformation: ξ µ = xµ .

4. Special conformal transformation: ξ µ = xν xν bµ − 2bν xν xµ , where, bµ is a constant.

Using these symmetries, a simple derivation of the invariant flow is given. Next, we look at the perturbations in the flow, their
equation of motions which we try to solve using numerical methods. Then we will look how these perturbations gives rise to
azimuthal structure of transverse momentum of particles, usually parametrized by Fourier coefficients vn ,
" #
dN dN X
Ep 3 = × 1+2 vn (pt , Y ) cos [n(φp − Ψn )] (1)
d p 2πpt dpt dY n

where Ψn are reaction planes characterizing the event, φp the azimuthal angle and Y is the momentum rapidity of the observed
particles. Elliptic flow is quantified with the second harmonic of the above azimuthal distribution vn of produced particles [9],
Rπ
−π
dφp cos(2φp ) 2πptdNdpt dφp
v2 (pt ) = hcos(2φp )ipt ≡ Rπ dN
(2)
−π
dφ p 2πpt dpt dφp

One of the most exciting result is the observation of large elliptic flow on hadrons in non-central relativistic heavy ion collisions
at RHIC and LHCa large elliptic flow [10]-[14]. Hydrodynamic model simulation can reproduce quite well the magnitude and
pattern of the elliptic flow for bulk hadron with a small shear-viscosity to entropy density ratio. In the present work we study
the contribution of perturbation on top of Gubser flow on the elliptic flow by solving numerically the equation of motion of the
perturbations. Hatta et al. [7] analytically obtained the perturbation only in the very early time approximation τ  4 fm/c. This
method ignores the dynamics of the evolution about the moderate (and late times) when the spectra and the build-up of elliptic
flow is important. In this report, we consider the entire dynamics by numerically solving the equations till freeze-out.
To get hadrons spectra and v2 in this report we start from section II in which we define the hydrodynamic fields and the
energy-momentum tensor from which we get the equation of motions of the fields. In subsection A, we define the coordinate
system that we use to make the calculations simple then in the next subsection we see how the fields transform under the
change of coordinates. The next subsection details the free parameters in the theory and we fix the values that we will be taking
for the simulations and finally in the next subsection we give the solution of the temperature field and plot them. In section
III we define perturbations that we are taking and in the following two subsections we give the equation of motions of these
perturbations which are derived from the conservation equations. Section IV we plot the solution of the perturbations which
we get by solving the equation of motions defined in the previous section numerically using RK4 method. In section V we the
Cooper-Frye formula which relates the particle spectra attained in an experiment to the hydrodynamics that is happening before
freeze-out. The subsections under this section defines the viscous stress tensor and the constant temperature surface that are
needed for the Cooper-Frye formula. Then finally in section VI we give the results of our simulations. In the first subsection
we give the spectra which is the rate of particles coming out with transverse momentum for pions, kaons and protons for two
different perturbation strengths. Next subsection gives the elliptic flow coefficient v2 as a function of the transverse momentum
pt again for all individual particles and also the total elliptic flow. Then finally in the last subsection we give the rate of particles
coming out with the azimuthal angle in the transverse plane.

II. GUBSER SOLUTION

In hydrodynamics one defines the dynamics of the fluid variables with five different fields, namely the four velocity uµ (x),

 
γ(x)c
uµ (x) = (3)
γ(x)~v (x)
 −1/2
v 2 (x)
where, ~v (x) is the local three-velocity and γ(x) = 1 − c2 , the Lorentz factor and a thermodynamical quantity e.g. the
pressure p(x) or energy density (x). The dynamics of these fields are given by the equations of motion of the system which are
the conservation equations (∇µ T µν = 0). The stress energy tensor (T µν ) is given as follows,

T µν = uµ uν + 4µν − ησ µν (4)
3
where, the projection matrix 4µν = g µν + uµ uν and the viscous stress tensor is given as follows,

σµν = 2∇<µ uν> (5)

2
The brackets denote a symmetric traceless projection onto the space orthogonal to uµ ,
1 1
A<µν> = 4µα 4νβ (Aαβ + Aβα ) − 4µν 4αβ Aαβ (6)
2 3
In Bjorken flow, Milne Coordinates are used (τ, x⊥ , φ, η), gµν = diag(−1, 1, x2⊥ , τ 2 ). In this text we use dS3 ×R coordinates,
for that we first have to rescale the metric as,
ds2
dŝ2 = (7)
τ2
and introduce new coordinates,
1 − q 2 τ 2 + q 2 x2⊥ 2qx⊥
sinh ρ = − , tan θ = (8)
2qτ 1 + q 2 τ 2 − q 2 x2⊥
which gives the following form of the metric,
dŝ2 = −dρ2 + cosh2 ρ(dθ2 + sin2 θdφ2 ) + dη 2 (9)
The 4-velocity transforms as the following here, x̂µ = (ρ, θ, φ, η) and xµ = (τ, x⊥ , φ, η),
∂ x̂ν µ
uµ = τ û (10)
∂xµ
The energy density and temperature transform as,

ˆ T̂
= , T = λ−1/4 (11)
τ4 τ

A. Parameters

The free parameters in the theory that have to be matched to experiments are q, λ, H0 and T̂0 and the equation of state(EoS).
’q’ is defined as inverse of the typical length scale of the system which would be the diameter of the colliding lead nucleus which
we have taken to be,
1
q= (12)
7.3f m
7.3 fm is a good approximation for the radius of lead nucleus. ’λ’ is defined from the relation between energy density of a system
and its temperature (Stefan-Boltzmann law). For a QGP is given by the following equation,
  2
7 π 4
 = NB + NF T (13)
8 30
where, NB and NF are the number of the Bosonic and Fermionic degrees of freedom. A QGP has 47.5 effective degrees of
freedom [6]. So for a QGP,
π2
λ = 47.5 × (14)
30
The entropy density is defined as s = Σ0 3/4 , where it is easy to show that Σ0 = 34 λ1/4 = 2.65098. H0 is a dimensionless
parameter related to the shear viscosity as follows,
η
H0 = Σ0 (15)
s
in this paper we will take multiple values of η/s = 0.05, 1/4π, 0.134, 0.2 ’T̂0 ’ comes as a constant of integration when we
calculate the temperature evolution from the conservation equations. It is defined a T̂ (ρ = 0). It defines the initial energy
density and hence the initial temperature. We estimate the value from entropy production per unit of rapidity which in turn is
connected to the production of charged particles as,
dS dNcharged
' 7.5 (16)
dηr dηr

3
One can get an analytical expression for the ideal case,
 1/3
1 3 dS
T̂0 = (17)
λ1/12 16π dηr
dNcharged
At the LHC, dηr ∼ 1600 which gives, T̂0 = 7.115. But, this gives a freeze-out time of 5.8f m/c at the center of the
fireball which is too small for freeze-out. Hence,  our the initial temperature to T̂0 = 10.1. It means we take an
 we increase
dNcharged dNcharged
increased particle production dηr = 2.86 dηr . Which gives a freeze-out time of about 8 − 9f m/c.
LHC
The equation of state (EoS) of a system defines what the constituents of the system are. Here since we are talking about
systems with very high energies, hence, it is an appropriate assumption to take the EoS of  = 3p.

B. Gubser solution

It is easy to see what the four velocity is in dS3 × R. From the SO(1, 1) × Z2 symmetry, it follows directly that ûη = 0 and
all the other components cannot depend on η. With the SO(3)q symmetry and uµ uµ = −1 the only possibility left over is then

ûρ = −1, ûθ = ûφ = ûη = 0 (18)

solving the conservation equation (∇µ T µν = 0) using the EoS  = 3p, one gets as in [2],
  
T̂0 H0 3 3 7 5 2
T̂ (ρ) = 1+ sinh ρ2 F1 , ; ; − sinh ρ (19)
(cosh ρ)2/3 9T̂0 2 6 2

where
ab a(a + 1)b(b + 1) z 2
2 F1 (a, b; c; z) =1+ z+ + ... (20)
c c(c + 1) 2

12
0.6 η/s=0
10 0.4 η/s=1/(4π)
0.2 η/s=0.2

8 0
-0.2
-9 -8 -7 -6 -5
6 ρ
T(ρ)

-2
-10 -5 0 5 10
ρ

FIG. 1: de-Sitter temperature T̂ (ρ) of the fireball is plotted on the y-axis with de-Sitter time ρ on the x-axis

The temperature of the fireball in MeV can be expressed using eq.(19) and (11) and then we can convert the above equation
in normal coordinates by using the transformation in eq. (8) and is plotted below,

4
 800
τ = 0.2fm/c
η/s=0
700
η/s=1/(4π)
η/s=0.2
600

Tempature (MeV)
500
τ = 1fm/c
400

300 τ = 3fm/c

200 τ = 5fm/c

100

0
0 2 4 6 8 10
xp(fm)

FIG. 2: The temperature of the fireball is plotted on the y-axis in MeV at different times (τ =0.2,1,3,5 fm/c) and on x-axis is the
radial distance from the center in the transverse plane.

III. LINEARIZED PERTURBATIONS ON GUBSER FLOW

Using the solution from [2], the temperature and velocity fields are perturbed as follows

T̂ = T̂b (1 + δ(ρ, θ, φ, η)),

ûµ = (−1, νi (ρ, θ, φ, η), νη (ρ, θ, φ, η)) (21)

where, we have decomposed the perturbations into scalers and vectors of the SO(3) isometry group of dS3 .

Z
dkη
δ(ρ, θ, φ, η) = δ(ρ)S(θ, φ)eikη η , (22)
2π
Z
dkη
νs (ρ)∂i S(θ, φ)eikη η + νv (ρ)Vi (θ, φ)eikη η ,


νi (ρ, θ, φ, η) = (23)
2π
Z
dkη
νη (ρ, θ, φ, η) = νη (ρ)S(θ, φ)eikη η (24)
2π
The role of perturbations in this text is to generate harmonic flow (v2 ). For that purpose we will not be including the perturbations
along the boost direction (kη = 0). Also, we take l = 2, hence, the equations become,

5
 " r #
2,2 3 1
δ (ρ, θ, φ) = δ(ρ) − (Y2,2 (θ, φ) + Y2,−2 (θ, φ)) + Y2,0 (θ, φ)
8 2
r
1 5
−3 sin2 θ cos(2φ) + (3 cos2 θ − 1)

= δ(ρ)
8 π
" r #
3 1
νθ2,2 (ρ, θ, φ) = νs (ρ)∂θ − (Y2,2 (θ, φ) + Y2,−2 (θ, φ)) + Y2,0 (θ, φ)
8 2
r
2iνv (ρ) 3
+ [Y2,2 (θ, φ) − Y2,−2 (θ, φ)]
sin θ 8
r
3 5
sin (2θ) cos2 φνs (ρ) + sin (2φ) sin θνv (ρ)

= −
4 π
" r #
3 1
νφ2,2 (ρ, θ, φ) = νs (ρ)∂φ − (Y2,2 (θ, φ) + Y2,−2 (θ, φ)) + Y2,0 (θ, φ)
8 2
" r #
3 1
+νv (ρ) sin θ∂θ − (Y2,2 (θ, φ) + Y2,−2 (θ, φ)) + Y2,0 (θ, φ)
8 2
r
3 5
sin θ sin θ sin (2φ) νs (ρ) − sin (2θ) cos2 φνv (ρ)
 
= (25)
4 π
As usual we do most of our calculations in the dS3 ×R conformal frame. We use a subscript b to denote unperturbed background
quantities.

A. Equation of motion for the vector modes

The equation of motion for νv which follows from the conservation of energy-momentum tensor takes the form,

νv0 (ρ) = −Γv (ρ)νv (ρ) (26)

where,

4 Tb 2 1 H02 H0
Γv = 0 tanh ρ + 0 tanh4 ρ − [11 + 4 cosh 2ρ] tanh ρ sech2 ρ
9 Tb 3 Tb Tb 9Tb0

B. Equation of motion for the scaler modes

Considering the perturbations (25) with the vector perturbations νv set to zero. If we define,

w
~ = (δ, νs , νη )

then instead of a single equation, the energy-momentum conservation gives us the coupled equations,

~ 0 = −Γs w
w ~ (27)

where this Γs is a 3×3 matrix with all entries non-zero:

6
 H0 tanh2 ρ
Γs,11 = ,
3T̂b
6 sech2 ρ[H0 tanh ρ − T̂b ]
Γs,12 =
3T̂b
2H0 tanh ρ
Γs,21 = + 1,
H0 tanh ρ − 2T̂b
−8H0 T̂b [1 + 2 sech2 ρ] + 3H02 tanh3 ρ + 4T̂b2 tanh ρ
Γs,22 = ,
3T̂b [H0 tanh ρ − 2T̂b ]
h i
T̂b H0 (25 sech2 ρ + 8) − 4T̂b tanh ρ + 6H02 tanh3 ρ
Γs,33 =
6T̂b [H0 tanh ρ + T̂b ]

Notice that some of the coefficients diverge at certain ρ, firstly when T̂b = 0, as we will show that we don’t trust hydrodynamics
in this region. The second route is for H0 tanh ρ − 2T̂b = 0, note that,
1
T̂b0 = tanh ρ(H0 tanh ρ − 2T̂b ) (28)
3

IV. SOLUTION OF PERTURBATIONS

The above differential equations for perturbations is solved using RK4s for initial location ρ0 = −2.5 having the initial values,
δ0 = 0.592360, νs0 = −0.888754 and νv0 = 0.256715. The solutions are shown in the below graph on the left side for five
values of η/s = 0, 0.05, 1/4π, 0.134, 0.2. The right side is the solution for same initial location but half the value of the initial
perturbations.

0.5 η/s=0
0.2
η/s=0.05
η/s=1/4π
0 0
δ(ρ)

η/s=0.134
η/s=0.2
-0.5 -0.2
-0.4
-1
-2.5 -2 -1.5 -1 -0.5 0 -2.5 -2 -1.5 -1 -0.5 0
0 0
-0.2 -0.1
-0.2
νs(ρ)

-0.4
-0.6 -0.3
-0.4
-0.8
-0.5
-2.5 -2 -1.5 -1 -0.5 0 -2.5 -2 -1.5 -1 -0.5 0
0.2
0.25 0.16
νv(ρ)

0.2
0.12
0.15
0.08
0.1
-2.5 -2 -1.5 -1 -0.5 0 -2.5 -2 -1.5 -1 -0.5 0
ρ ρ

FIG. 3: The ideal solution (black line) is compared to the viscous cases for l=2 and initial location of the perturbations is
ρ0 = −2.5 the left side and the right side differ in the initial perturbations where the left side is double than the right side.

In the above figure note that as the viscosity increases, the perturbations are suppressed. The temperature fluctuation starts
from a positive value at ρ = −2.5 then goes to zero and flips sign at around ρ = −1.5. This effect is due to the hydrodynamic
flow itself as eqs. (26) and (27) come from the evolution equations of the fluid which are the conservation equation. The above
equations are solved numerically using the RK4 method of solving differential equations as exact solutions to the modes equation

7
are not available. However, the ideal case can be solved exactly as shown in [2] which for l=2 is,
√
r r !
2 73
− 1 1 3 3
δ(ρ, θ, φ) = K sech 3 ρP 2 6 2 (− tanh ρ) Y2,0 (θ, φ) − Y2,−2 (θ, φ) − Y2,2 (θ, φ) (29)
3 2 8 8

where K is an integration constant. Below we have plotted the above solution in the real coordinates (x,y,τ ) for different τ . This
is to see how the perturbations evolve in time.

(a) τ = 1f m/c (b) τ = 3f m/c

(c) τ = 5f m/c (d) τ = 7f m/c

FIG. 4: The x and y axis define the real x and y axis in the transverse plane in fm and the height is temperature perturbation δ
which is unitless. These figures are at zero viscosity and initial value corresponding to the left side in fig. (7).

V. HADRON PRODUCTION

To go from the hydrodynamic picture to the particle picture we use the prescription given by Cooper-Frye in [8] which relates
the decoupling takes place on a hypersurface of constant temperature (Tf = 120M eV ) or constant energy density. At this
hypersurface, the momentum distribution of hadrons freezes such that,
Number of particles emitted = number of particles crossing the surface Σ, using this and some results from kinetic theory, it
can be seen that the particle spectra is related to the fluid parameters as follows,

pµ uµ
Z
dN
(2π)3 = −gi pµ dΣµ f ( ) (30)
pt dpt dφp dY Σ T

8
where, the distribution function is the solution of the Boltzmann equation in kinetic theory and is given by,
pµ uµ pµ pν π µν
 
pµ uµ /T
f( ) = feq + δf = e 1+ ,
T 2( + p)T 2
pµ pν π µν 3H0
= −pµ pν σ µν 3 (31)
2( + p)T 2 4T


where p̂µ = pρ , pθ , pφ , pη , the components are given by,

p̂ρ = q [mt cosh η(cosh ρ − sinh ρ cos θ) − pt sin θ cos(φ − φp )] τ 2 ,

p̂θ = q [−mt cosh η sin θ sech ρ + pt cos(φ − φp )(1 − cos θ tanh ρ)] τ 2 ,
p̂φ = q [pt sin(φ − φp )(cos(θ) − tanh(ρ)) csc θ] τ 2 ,
p̂η = q [mt sinh η(cos θ − tanh ρ) cosh ρ] τ 2 (32)
p
The transverse mass mt is defined as mt = m20 + p2t .

A. Viscous-Stress Tensor (σµν )

From eqns. (5) and (6) one can deduce, the σµν terms upto first order in perturbation, first we note that σµν is symmetric
under the exchange of µ and ν, the following terms of σµν are zero, σ00 = σ03 = σ13 = σ23 = σ33 = 0, the remaining non-zero
terms are,

1 1
σ01 = − νθ (ρ, θ, φ) tanh ρ, σ02 = − νφ (ρ, θ, φ) tanh ρ,
3 3
1
cosh ρ sinh ρ − cot θνθ (ρ, θ, φ) − csc2 θ∂φ νφ (ρ, θ, φ) + 2∂θ νθ (ρ, θ, φ) ,

σ11 =
3
1
σ12 = [∂φ νθ (ρ, θ, φ) + ∂θ νφ (ρ, θ, φ) − 2νφ (ρ, θ, φ)] ,
2
1
cosh ρ sinh ρ sin2 θ + sin2 θνθ (ρ, θ, φ) + 2∂φ νφ (ρ, θ, φ) − sin2 θ∂θ νθ (ρ, θ, φ) ,

σ22 =
3
1
σ33 = − sech2 ρ [sinh(2ρ) + cot θνθ (ρ, θ, φ) + 2∂φ νφ (ρ, θ, φ) + ∂θ νθ (ρ, θ, φ)] (33)
3
where, νθ and νφ are given in (25).

B. Constant temperature surface Σ

The constant temperature surface is defined as Σ̂µ = (ρ, θf (ρ), φ, η) in the gubser coordinates. What we need for eq. (30) is
the normal vector (dΣ̂µ ) to this constant temperature surface. Note that in calculating the constant temperature surface, we have
not taken into account the deformation effects of the perturbations. This is given by the following equation

√ ∂ Σˆν ∂ Σˆλ ∂ Σˆρ

dΣ̂µ = − −gµνλρ dρdφdη (34)
∂ρ ∂φ ∂η

√
 
∂θf
dΣ̂µ = −g − , 1, 0, 0 dρdφdη (35)
∂ρ
for θf which gives the constant temperature surface,

T̂ (ρ)
T = Tf =⇒ = Tf (36)
τ λ1/4
Tf λ1/4
cos θf = tanh ρ + sech ρ (37)
T̂ (ρ)

9
 where, Tf is the freeze-out temperature which is taken to be 120 MeV, for the ideal case the equation simplifies as,

Tf λ1/4
cos θf = tanh ρ + cosh ρ−1/3 (38)
q T̂0

2.5

2.5
2.0
2.0
1.5
θf (ρ)

θf (ρ)
1.5
1.0
1.0

0.5 0.5

0.0 0.0
-8 -6 -4 -2 0 -8 -6 -4 -2 0

ρ ρ
(a) Ideal case (b) Viscous case

FIG. 5: The constant temperature surface in ρ, θf coordinates the ideal case surface goes from ρ = −∞ to 0.129 and viscous
case (η/s = 0.134) surface goes from ρ = −6.6987 to 0.129.

for different viscosities, the constant temperature surface changes as is evident from Fig.5. At zero viscosity (ideal case) the
constant temperature surface is unbounded from the negative side and as we put in viscosity, the surface becomes bounded and
with increase in viscosity, the bound becomes smaller. The values of the bound that obtain and will be using in our calculations
and simulations are (η/s = 0.05, −8.19119)(η/s = 1/4π, −7.48721)(η/s = 0.134, −6.6987)(η/s = 0.2, −6.09493).
Following Fig.6 is a plot of the constant temperature surface in real coordinates. Here, the x and y axis are the space x and y
coordinates in the transverse plane and the z-axis is the freeze-out time(τf ) that we get when we put temperature as constant.

(a) Ideal case (b) Viscous case

FIG. 6: Constant temperature surface x and y axis are the transverse place x and y axis and the z axis is the freeze-out time for
freeze-out temperature of 120 MeV

to get the normal vector in eq. (35) we will also need the partial derivative of θf with ρ. This is given as,
" !#
∂θf 2 Tf λ1/4 sech ρ tanh ρT̂ (ρ) + T̂ 0 (ρ)
= − csc θf sech ρ − (39)
∂ρ q T̂ 2 (ρ)

where, T̂ 0 (ρ) is the ρ derivative of de-Sitter temperature which is given in eq. (28).

10
 VI. RESULTS

A. Hadron spectra

The eq. (30) relates the particle spectra to the kinetics in the system through the distribution function. Here we are interested
in understanding how the rate of particles coming out change with the transverse momentum (pt ). For that the particle rapidity
(Y ) is set to zero (mid-rapidity region), and both sides of the eq. (30) are integrated over φp which becomes,
Z π
pµ uµ
Z
dN g
=− 4
dφ p pµ dΣµ f ( ) (40)
2πpt dpt (2π) −π Σ T

The above equation is plotted vs pt Fig.7 for Pions(m0 = 0.7096; g = 1), Kaon(m0 = 2.5241; g = 1) and Protons(m0 =
4.7542; g = 2). The dashed line is the experimental data and the rest are simulation results for different viscosity (η/s =
0, 0.05, 1/4π, 0.134, 0.2) as is given in the legends of the Fig. (7). In the above figure we see the data (black dotted line) and
1/(2πpt)dN/dpt(GeV )
-2

10000 10000
π π
100 100

1 1

0.01 0.01
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
100 100
k k
1 1
Data
η/s=0.134
η/s=0
η/s=0.2
0.01 η/s=0.05 0.01 η/s=0 No Pert
η/s=1/(4π)

0.0001 0.0001
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
1/(2πpt)dN/dpt(GeV )
-2

10 10
1 Pr 1 Pr
0.1 0.1
0.01 0.01
0.001 0.001
0.0001 0.0001
0 1 2 3 0 1 2 3
pt(GeV) pt(GeV)

FIG. 7: The black dotted line is the data and rest of the lines are results of eq. (40) for pions(top), kaons(middle) and
protons(bottom), two different perturbation strength(left double of the right) and different viscosities. The red line is the ideal
no perturbation solution.

the results of the simulations for different viscosities and different perturbation strengths. One thing we clearly observe here is
that the particle production is lower than the data suggests. This is for the reason that not all of the particle production is a result
of the hydrodynamic flow we took into consideration. The remaining part is accounted for by using more realistic simulations
like 3+1 Hydro Models or the Glauber Model for the realistic initial conditions. When we scale up our data however, what
we see is that for the cases of kaons and protons, the larger perturbation simulations match better to the data, for kaon the
η/s = 0.134(Indigo) matches and for protons the η/s = 1/4π curve matches with the data. In the case of pions however, the
smaller perturbation matches with the low viscosity (∼ η/s = 0.05) curves at large pt (∼ 3GeV ) and high viscosity (∼ 0.2) at
low pt (∼ 0.5 − 2GeV ) where as the kaons and protons

11
 B. Hadron azimuthal distribution dN/dφp

In this section we are interested in knowing the particle distributions as a function of the azimuthal angle φp for different
values of viscosities and is plotted below.

∞
pµ uµ
Z Z
dN g
=− dpt pµ dΣµ f ( ) (41)
dφp (2π)3 0 Σ T

40 20 η/s=0
3
35 π η/s=0.05
η/s=1/(4π) K 2.5 Pr
15 η/s=0.134
30 η/s=0.2
2
dN/dφp

25 10 1.5
20 1
5
15 0.5
10 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
30 6
0.8
25
π 5 K Pr
4 0.6
20 3
0.4
2
15 0.2
1
10 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6

FIG. 8: Here are results of eq. (41) for pions(left), kaons(centre) and protons(right), two different perturbation strength(top
double of the bottom) and different viscosities. The black dotted line is the ideal solution

We see in Fig. (8) that the particle production is dominant along the x-axis (φp = 0, π). This should evidently be the case as
in Fig. (4) we see the perturbations form and move along the x-axis at those angles and it is the perturbations that are producing
anisotropy in the medium which is giving rise to the above particle distribution. We also note that with increase in viscosity, the
particle production is reduced as for high viscosity, the curve is shrunk. In the below graph we look for the particle distribution
with the azimuthal angle φp for one value of the transverse momentum, pt = 1GeV .

pµ uµ 
Z 
dN g µ
=− p dΣµ f ( ) (42)
dφp (2π)3 Σ T pt =1GeV

12
 0.7 0.3 0.05
0.6 π 0.25 K 0.04 Pr
0.5
0.2
0.03

dN/dφp
0.4
0.15
0.3 0.02
0.1
0.2
0.05 0.01
0.1
0 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6
0.4 0.2 η/s=0
0.02
η/s=0.05
0.35 π 0.15
η/s=1/4π K 0.015
Pr
η/s=0.134
0.3 η/s=0.2
dN/dφp

0.25 0.1 0.01

0.2
0.05 0.005
0.15
0.1 0 0
0 1 2 3 4 5 6 0 1 2 3 4 5 6 0 1 2 3 4 5 6

FIG. 9: Here are results of eq. (42) for pions(left), kaons(centre) and protons(right), two different perturbation strength(top
double of the bottom) and different viscosities. The black dotted line is the ideal solution.

C. Elliptic flow v2 (pt ) for hadrons

The elliptic flow is quantified with the second harmonic of the azimuthal distribution of produced particles as given in eq. (2),
Rπ
−π
dφp cos(2φp ) 2πptdN dpt dφp
v2 (pt ) = hcos(2φp )ipt ≡ Rπ dN
(43)
−π
dφ p 2πpt dpt dφp

In Fig.10 we have plotted the result of the simulation for pions, kaons and proton for different values of viscosity.
0.8 0.7
0.6
0.6 π 0.5 π
v2(pt)

η/s=0
0.4
0.4 η/s=0.05
η/s=1/(4π) 0.3
0.2 η/s=0.134 0.2
η/=0.2 0.1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
0.8 0.7
0.6
0.6 k 0.5 k
v2(pt)

0.4
0.4
0.3
0.2 0.2
0.1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
0.7
0.8 0.6
0.6 Pr 0.5 Pr
v2(pt)

0.4
0.4 0.3
0.2 0.2
0.1
0 0
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
pt(GeV) pt(GeV)

FIG. 10: Here we have plotted on the y-axis elliptic flow coefficient v2 (pt ) which is dimensionless as in eq. (2) and on the
x-axis transverse momentum pt for pions(top), kaons(middle) and protons(bottom), two different perturbation strength(left
double of the right) and different viscosities.

13
 The v2 that we get here is too large. Way larger than the experimentally observed value of 0.036 at around 3 GeV for central
collisions. However, this is to be expected because as can be seen from Fig. (4) the perturbations that we are taking travel along
hp2 −p2 i
 
the x-axis, hence creating a large elliptic flow as v2 is basically just the measure of variance in momentum space v2 ≡ hpx2 +p2y i .
x y
Also we see that the elliptic flow is sensitive to the viscosity and the strength of perturbation such that the flow is smaller for
larger viscosity. This seems to be alright as we would expect that for higher viscosity the perturbations are suppressed as can be
seen in Fig.3. The interesting thing that we note here is that even if we change the strength of the perturbation by half, the v2
decreases but not as much.
Total v2 (pt ) is the sum of v2 ’s for the pion, kaon and proton.

R  R  R 
π dN π dN π dN
dφp cos(2φp ) pt dp + dφ cos(2φ ) + dφ cos(2φ )
  
−π t dφp
 −π p p pt dpt dφp  −π p p pt dpt dφp 
π k Pr
v2total (pt ) ≡ R
π
 R
π
 R
π
 (44)
dN dN dN
dφp pt dpt dφp  + −π dφp pt dpt dφp  + −π dφp pt dpt dφp 
  
−π π k Pr

In Fig.11 we have plotted the above equation for different viscosity values.
0.8
Total v2(pt)
0.6
v2(pt)

η/s=0
0.4 η/s=0.05
η/s=1/(4π)
0.2 η/s=0.134
η/s=0.2

0
0 0.5 1 1.5 2 2.5 3

0.6
0.5
0.4
v2(pt)

0.3
0.2
0.1
0
0 0.5 1 1.5 2 2.5 3
pt(GeV)

FIG. 11: Here we have plotted on y-axis the total elliptic flow coefficient v2total (pt ) as in eq. (44) and on the x-axis transverse
momentum pt for, two different perturbation strength(top double of the bottom) and different viscosities.

VII. CONCLUSION AND FUTURE DIRECTIONS

In conclusion, we have used conformal symmetry as an inherent symmetry of the QGP system formed in a heavy ion collision
using the framework given by Gubser in [1] and applied perturbations of the form in [2] to break the φ symmetry and hence
generate anisotropy in the spectra. This allows us to study the elliptical flow formed by the above mentioned symmetry breaking
which is quantified in the harmonic v2 as a function of transverse momentum. We note that the spectra that we get when we
introduce the perturbations somewhat matches the data but different for different sets of particles and what is the strength of
the perturbations. The elliptic flow produced by the said perturbations is quite large as the anisotropy that is introduced in the
system is also large and along both the x directions. This is corroborated from Figs. (8) and (9) which show a large particle
distribution along the 2 directions along the x-axis (φ = 0, π). This was done to show that the conformal symmetry is a
reasonable approximation of the HIC system at high energies which bears out.
Next we can look at the same system use the same codes for different kinds of initial conditions for the perturbations for
example a Gaussian initial distribution or a perturbation at a single point approximated as a delta function. Also as we have only
used first order hydrodynamics the next obvious step would be to do the same calculations for a higher order theories. Effects of
bulk viscosity can also be studied for the system. Also as we here are restricted to head on collisions, it will be interesting to see
effects for such a system in case of peripheral collisions.

14
 VIII. ACKNOWLEDGEMENT

I would like to take this opportunity to thank Dr. Subrata Pal for his guidance during my project. I would also like to thank
Chandrodoy Chattopadhyay and Sunil Jaiswal for the insightful discussions.

[1] Steven S. Gubser, Phys. Rev. D 82 (2010) 085027.

[2] S. S. Gubser and A. Yarom, Nucl. Phys. B846 (2011) 469.
[3] J. D. Bjorken: Phys. Rev. D27 (1983) 140151.
[4] P. F. Kolb and R. Rapp: Phys. Rev. C67 (2003) 044903, [hep-ph/0210222].
[5] U. W. Heinz: hep-ph/0407360
[6] S. Pal and S. Pratt: Physics Letters B 578 (2004) 310317.
[7] Y. Hatta: Phys. Rev. D90 (2014) 074026
[8] F. Cooper and G. Frye: Phys. Rev. D10, 186 (1974).
[9] D. Teaney: Phys. Rev. C68, 034913 (2013).
[10] K.H. Ackermann et al., STAR Collaboration, Phys. Rev. Lett. 86, 402 (2001)
[11] C. Adler et al., STAR Collaboration, Phys. Rev. Lett. 87, 182301 (2001).
[12] C. Adler et al., STAR Collaboration, Phys. Rev. C 66, 034904 (2002).
[13] K. Adcox et al., PHENIX Collaboration, Phys. Rev. Lett. 89, 212301 (2002).
[14] B.B. Back et al., PHOBOS Collaboration, Phys. Rev. Lett. 89, 222301 (2002).
[15] L.D. Landau and E.M. Lifshitz, Course of Theoretical Physics Volume 6, “Fluid Mechanics”, Elsevier, 2nd edition (1987).
[16] R. Baier, P. Romatschke, D. T. Son, A. O. Starinets, M. A. Stephanov: arXiv:0712.2451 [hep-th]

15