Hydrodynamics-some notes

HM∗

Theory Division, Physical Research Laboratory, Navrangpura, Ahmedabad 380 009, India
(Dated: November 21, 2023)
Here are some notes for hydrodynamics part of P304 course taught in July-Dec 2023 at NISER.
By nomeans it is original. Refs are Landau series Fluid mechanics and review article by Paul
Romatschke arXiv:0902.3663

PACS numbers: 12.38.Mh, 12.39.-x, 11.30.Rd, 11.30.Er

∗ hiranmayaniser.ac.in
 2

I. INTRODUCTION

Blah,blah,blah · · ·

II. NON RELATIVISTIC HYDRODYNAMICSS

Fluid dynamics is the motion of fluids i.e. liquids and gases.a

• The phenomana considered here are macroscopic and the fluid considered here is regarded as a continueous
medium
• Any small volume element ( small compared to the system size) large number of particles
• This means mean free path between collision ¡¡ infinitesimalvolume element ¡¡ total volume.
the degrees of freedom of for an ideal fluid are (i) the fluid velocity v(x, t) − (3components), the pressure P (x, t)(1 −
component) and the fluid mass density ρ(x, t)(1 − component). So we need 5 equations to study the dynamics. Let
us try to derive them.
Let us begin with the equation that describes conservation of mass. Consider a small volume V0 bounded by a
surface S. The mass of the fluid flowing out of the volume V0 per unit time is, using gauss divergence theorem,
Z Z
ρv · dS = ∇ · (ρv)dV
S V0

On the otherhand the decrease in mass of the fluid in the volume V0 per unit time is
Z Z
∂ ∂ρ
− ρdV = − dV
∂t V0 V0 ∂t

On equating, we have
Z  
∂ρ
+ (∇ · ρv) dV = 0
V0 ∂t

Since this holds good for any volume, the integrand must vanish. Thus
∂ρ
+ (ρ∇ · v + v · ∇ρ) = 0 (1)
∂t
Next, we shall consider the Euler equation which is equivalent of Newton’s second law . The force acting on a
volume V0 due to rest of the fluid in the volume is
I Z
P ds = − ∇P dV. (2)
S V0

The negative sign above is due to the inward force.. The acceleration of unit volume due to this force is
dv
ρ = −∇P
dt
dV ∂ ∂v dxi
. This can be rewritten as, using dt = v ∂t + ∂xi dt

∂v 1
+ (v · ∇)v + ∇P = 0 (3)
∂t ρ
which is the Euler equation for ideal fluid. This is because we have not taken into account any energy dissipation that
can occur due to internal friction (viscosity) in the fluid and no heat exchange between different parts of the fluid:
for the case where thermal conductivity and viscosity are not important.
 3

Alongwith the Euler equation eq.(3) ( 3 components), the continuity equation Eq.(1),another equation - the equation
of sate (EOS) i.e. the equation relating pressure P and mass density ρ completely specify the equations governing
the fluid dynamic evolution of ideal fluid.
The momentum of the fluid per unit volume is ρv. The rate of change of momentum is is
∂(ρvi ) ∂vi ∂ρ do later
=ρ + vi
∂t ∂t ∂t
. The above equation can be rewritten using the continuity Eq. (1) and the Euler equation Eq.(3) (home work)

∂(ρvi ) ∂Πij
+ =0 (4)
∂t ∂xj

where,

Πij = P δij + ρvi vj (5)

Using Greens theorem, one can show,

Z I
∂
ρvi dV = − Πij dSj (6)
∂t
which gives a meaning to the symmetric tensor Πij . It is the ith component of amount of momentum flowing per unit
time through unit area perpendicular to xj axis.
The momentum flux tensor given in Eq.(4) gives a reversible transfer of meomentum due to mechanical transportof
different particles from place to place and pressure forces acting on the fluid. The internal friction or the viscosity
causes irreversible transfer of momentum from places where velocities are large to places where the same is small.
Thus ingeneral, one can add to the momentum flux in eq.(4) which gives irreversible viscous of momentum and one
can write
0
Πij = P δij + ρvi vj − σij ≡ ρvi vj − σij (7)

where,σij is called the stress tensor which denotes the part of the momentum flux that is not due to transfer of the
0
momentum with the mass of the moving fluid. σij is the ’viscous’ part of the stress energy tensor which arises due to
internal friction when different fluid particles move with different velocities i.e. when there is a velocity gradient in
the fluid. When the velocity gradients are small one can neglect higher order derivatives of the velocity and can have
0
σij ∝ ∂vi /∂xj .The most general symmetric second rank tensor linear in velocity gradients can be written as
 
0 2
σij = η ∂i vj + ∂j vi − δij ∂k vk + ζδij ∂k vk (8)
3

Here, η and ζ are called the coefficients of shear and bulk viscosities respectively. The equation of motion of a viscous
0
fluid can be written down by adding the term ∂σik /∂xk term in the Euler Eq.(3) for the ideal fluid leading to (with
∂
∂k ≡ ∂xk ),
  
2
ρ (∂t vi + vj ∂j vi ) = −∂i P + ∂j η ∂j vi + ∂i vj − δij ∂k vk + ∂i (ζ∂k vk ) (9)
3

This equation is called Navier Stokes equation for non relativistic fluids. This equation is well tested and hence any
formulation of relativistic viscous hydrodynamics should reduce to this in the appropriate limit.

III. RELATIVISTIC IDEAL FLUIDS

For motions close to the speed of light, the mass density ρ(x, t) may not be an appropriate degree of freedom-
instead, one can replce it by total energy density (x, t). Similarly, the fluid velocity v(x, t) is not a good degree of
freedom as it it does not have the correct transformation properties under Lorentz transformation. At each space
time point xµ , there can be a temperature T (x), chemical potential µ(x) and the Lorentz four vector velocity field
 4
µ
uµ (x) ≡ dx
dτ . Here, the Greek indices µ denote Minkowski space i.e. µ = 0, 1, 2, 3 with metric gµν = diag((+, −, −, −).
The proper time

dτ 2 = gµν dxµ dxν = (dt)2 − (dx)2 = (dt)2 [1 − v2 ]

( we use the natural units c=1 etc).This also means

dt dxµ
 
1 c=1, otherwise same
uµ = = γ(v)
dτ dt v

This also reduces to uµ = (1, v) in teh non relativistic limit. Further, uµ is normalised as uµ uµ = uµ gµν uν =
γ 2 (1 − v2 ) = 1.
Many physical system, perhaps including the universe, the QGP resulting from heavy ion collision may be ap-
proximately regarded as perfect fluid. A perfect fluid is defined as having at each point a velocity v , such that an
observer moving with this velocity sees the fluid around him as isotropic. This will be the case, if the mean free path
between collisions of the particles is small compared to the scale of lengths used by the observer. We can translate
the definition of the perfect fluid into a statement about the energy-momentum tensor. Suppose we are in a frame of
reference in which the fluid is at rest at some space time point i.e. uµ = (1, 0)- the socalled local rest frame (LRF). At
this point the perfect fluid hypothesis tells that the energy momentum tensor takes the for chracteristic of spherical
symmetry
ij
TLRF = P δ ij i0
TLRF 0i
= TLRF 00
= 0 TLRF = (10)

The coefficients P and  are called the pressure and energy density respectively.
µν
For an ideal fluid, in a general frame, the general form of the ideal energy momentum tensor T(0) can be constructed
out of the hydro dynamic degrees of freedom, namel
µν
y velocity vector uµ , the metric tensor gµν . We can express the tensor structure of T(0) as
µν
T(0) = c1 uµ uν + c2 g µν (11)

In the local rest frame uµ = (1, 0),

c1 + c2 0 0 0
 
µν  0 −c2 0 0
T(0)LRF = (12)
0 0 −c2 0
0 0 0 − c2

Comparing with eq.(10), we have c1 + c2 =  and −c2 = P so that

T µν = uµ uν − P ∆µν ∆µν = g µν − uµ uν (13)

Equivalently

T µν = ( + P )uµ uν − P g µν

In Eq.(13) we have introdued the operator ∆µν . Let us note that

uµ ∆µν = uµ (g µν − uµ uν ) = uν − uν = 0

Further,

∆µρ ∆ρν = (gρµ − uµ uρ )(g ρν − uρ uν )

= g µν − uµ uν − uµ uν + uµ uν = g µν − uµ uν = ∆µν

Therefore, ∆µν is a projection operator orthogonal to uµ . Using this we can define the projections of derivatives
along and orthogonal to uµ : D = uµ ∂µ and ∇µ = ∆ρµ ∂r ho sothat ∂µ = uµ D + ∇µ .
 5

The dynamical description of ideal fluid can be obtained by using conservation equations-e.g the energy momentum
conservation equation is given by
µν
∂µ T(0) = ∂µ (uµ uν − P ∆µν ) = 0 (14)

Taking the projection of the above equation along uν leads to

µν
uν ∂µ T(0) = ∂µ (uµ uν − P ∆µν ) = 0

Using the condition ∂µ (uα uα ) = 0, one has

D( + P ) + ( + P )∂µ uµ − uν g µν ∂µ P = 0

i.e.

D + DP + ( + P )θ − DP = 0

or,

D + ( + P )θ = 0 (15)

where, we have defined the expansion scalar θ as θ = ∂µ uµ . Similarly, taking the perpedicular projection
µν
∆α µ α ν
ν ∂µ T(0) = ( + P )u ∆ν ∂µ u − (∂µ P )∆
µα

where, we have used uν ∆α

ν = 0.
Further, note that

uµ ∆α ν µ α µ α ν µ α α
ν ∂µ u = u ∂µ u − u u uν ∂µ u = u ∂µ u = Du .

This leads to

( + P )Duα − ∇α P = 0 (16)

Equations Eq.(15) and Eq.(16) are the fundamental equations for a relativistic ideal fluid. It is illuminating to take
the nonrelativistic limit: i.e. for small velocities |v| << 1, one has

D = uµ ∂µ ' ∂t + (v · ∇) ∇α = ∆αµ ∂µ ' ∂ α

sothat D and ∆α essentially reduce to space and time derivatives respectively.Imposing further the nonrelativistic
equation of state i.e. P <<  and energy density  is dominated by mass density ρ, Eq.(15) reduces to the coninuity
equation Eq.(1).Similarly equation Eq.(16) reduces to Eq.(3).

IV. RELATIVISTIC NAVIER-STOKES EQUATION

If one is interested to include the dissipative effects like viscosity, one has to go beyond the ideal fluid limit. In this
case the fluid’s energy momentum tensor gets modified from the ideal fluid as in Eq.(13) and is given by
µν
T µν = T(0) + Πµν (17)

where, the extra termΠµν is the viscous stress tensor that inludes the contribution to the T µν arising from dissipation.
For simplicity, we shall consider a system without any conserved charges which is equivalent to taking chemical
potentials to zero. All momentum density is then due to flow of energy density .

uµ T muν = uν (18)

 6

which also means

uµ Πµν = 0

. Eq. (18) gives a choice of frame to define the fluid velocity. Suach a frame is refered to as the Landau frame. In
case of (non vanishing) conserved charges, one could have associated the fluid flow with a charge flow. These choices
reflect the freedom of defining the local rest frame (LRF) where the energy density is at rest (the Landau frame- as
the case here) or where the charge density is at rest ( such a frame is called Eckart frame). In the present case, for
further discussions we will continue with the Landau frame as deined in Eq.(18).
As we did in the ideal fluid case, the fundamental equations for viscos hyrodynamics are found by taking appropriate
projections of the energy momentum conservation equations.

uν ∂µ T µν = D + ( + P )∂µ uµ + uν ∂µ Πµν = 0 (19)

∆α
ν ∂µ T
µν
− ( + P )Duα − ∇α P + ∆α
ν ∂µ Π
µν
=0 (20)

The Eq.(7) can be futrther simplified by rewriting

∂µ Πµν = ∂µ (uν Πµν ) − Πµν ∂(µ uν) (21)

where, we have introduced the symmetrization notation (· · · ) as

1
A(µ Bν) = (Aµ Bν + Aν Bµ ).
2
Since Πµν is a symmetric tensor contraction with any second rank antisymmetric tensor will vanish. Further, in
Eq.(21), we can use the identity ∂µ = uµ D + ∇µ , and, the frame choice uµ Πµν = 0 leading to

D + ( + P )∂µ uµ − Πµν ∇(µ uν) = 0 (22)

sothat, Πµν does not have any derivative acting on it.The other equation we rewrite here is the Euler equation Eq.(20)

( + P )Duα − ∇α P + ∆α
ν ∂µ Π
µν
=0

In the above two equations the structure of Πµν is yet to be specified. An elegant way to detrmine the structure is
based on second law of thermodynamics i.e. i.e. positivity of entropy current ∂µ sµ ≥ 0 where, the entropy current is
sµ = suµ where s is the entropy density. Such an analysis leads to

Πµν = π µν + ∆µν Π

where, π µν is the trace less part built out of first order gradients of the velocity vector-
 
µν 1 α
π = η2 ∇(µ uν) − ∆µν ∇α u
3

and

Π = ζ∇α uα

In the non-relativistic limit the viscos tensor as written down above reduces to Navier Stokes equations with
identification of the coefficients η and ζ as coefficients of shear and bulk viscosities.
 7

The last class

In the last class there were some confusion regarding the thermodynamics that I told. Let me revise that. The
differential of the internal energy is given by the thermodynamic identity

dU = T ds − P dv + µdN

The first term is the heat transfer to the system, the second term is the workdone and the third term specifies the
constraint corresponding to a conserved quantum number with associated Lagrange paparameter which is the chemical
potential.
Further the internal energy is given by

U = T S − P V + µN

which leads to

dU = T dS + SdT − P dV − V dP + µdN + N dµ

Using the expression for dU we have V dP = N dµ + SdT . Dividing by the volume factor we have the relation

dP = ndP + sdT

, where n and s are the number density and entropy density respectively. Let us further note that the desnsities
satisfy the relation, with  = U/V ,

e = T s − P + µn

sothat d = T ds + sdT − P dV − V dP + µdn + ndµ. Using the expression for dP we have d = T ds + µdn. For vanishing
chemical potential, we have d = T ds, a relation we use often in hydrodynamics.