Basic Fluid Dynamics: Solutions

Tarun Deep Saini

Indian Institute of Science, Bangalore, India
[email protected]

1 Ideal Fluids

1. Hydrostatic equilibrium: A plane-parallel atmosphere, composed of a

perfect gas, is in hydrostatic equilibrium in an external gravitational field,
−ẑg. (a) Derive an expression for the entropy gradient if the atmosphere is
isothermal.

Solution

Hydrostatic condition implies v = 0, therefore the Euler equations take the

form
1 dp
∇p + ∇Φ = 0 ⇒ = −ρg (1)
ρ dz
From the first law of thermodynamics we have T dS = dU + pdV , and upon
dividing by the mass of the gas, it takes the from
 
1
T ds = du + pd (2)
ρ
or  
p 1 1
T ds = d u + − dp = dh − dp (3)
ρ ρ ρ
where we have used the definition of specific enthalpy in the last step. Note
that for an ideal gas the specific internal energy u depends only on the
temperature T . Since the equation of state for an ideal gas is
ρkB T
p= , (4)
µmp

1
it is clear that enthalpy too is only a function of temperature of the gas.
Thus, for an isothermal atmosphere we obtain
1
T ds = − dp (5)
ρ
Now using the equation of hydrostatic equilibrium we obtain the gradient
of entropy as
ds g
= (6)
dz T

(c) Earth’s atmosphere: In the lower stratosphere, the air is isothermal. Use
the condition of hydrostatic equilibrium to show that:

p(z) ∝ exp(−z/H), (7)

where the scale height, H = kT /(µmp g). Estimate the scale height. (Use
mean molecular weight µ = 29 and T = 300 K).

Solution

From the equation of hydrostatic equilibrium

dp
= −ρg (8)
dz
Using the equation of state and the fact that atmosphere is isothermal we
obtain
1 dρ µmp g
=− , (9)
ρ dz kB T
which can be solved to obtain
 
µmp g
ρ = ρ0 exp − z = ρ0 exp (−z/H) (10)
kB T

Since pressure is proportional to the density, temperature being a constant,

pressure too obeys a similar equation

p = p0 exp (−z/H) (11)

2
The scale height can be calculated to be H = 8.7 km

(b) Assuming that the air is isentropic, show that:

 
dT γ − 1 gµmp
=− (12)
dz γ k

Here γ ' 1.4 is the ratio of specific heats for gases like Nitrogen and Oxygen.
Why does the above expression vanish for γ = 1? Obtain expressions for
p(z) and ρ(z) if the atmosphere is isentropic.

Solution

From the definition of enthalpy we have

1
dh = T ds + dp (13)
ρ
For an isentropic atmosphere in hydrostatic equilibrium we obtain
dh
= −g (14)
dz
As mentioned earlier, for an ideal gas, enthalpy is only a function of tem-
perature  
kB
dh = du + d(p/ρ) = cv + dT = cp dT (15)
µmp
Using this and the expression for enthalpy gradient we obtain
dT g
=− (16)
dz cp

Using the fact that cp − cv = kB /µmp and cp /cv = γ, we obtain

γ
cp = kB /µmp , (17)
γ−1
which gives  
dT γ−1 gµmp
= −g/cp = − (18)
dz γ kB

3
Solving for temperature we obtain
 
γ − 1 zgµmp
T = T0 − (19)
γ kB

We also have
dp dh dT µmp dT
=ρ = ρcp = cp p (20)
dz dz dz kB T dz
which gives
γ − 1 zgµmp γ/(γ−1)
   
p = p0 1 − (21)
γ kB To
For γ → 1 this gives
p = p0 exp (−z/H) (22)
as it should for an isothermal atmosphere (for an isothermal atmosphere
γ = 1.

2. Archimedes’ principle states that, when a solid body is totally or partially

immersed in a fluid, the total buoyant upward force of the liquid on the
body is equal to the weight of the displaced fluid. Prove the law assuming
conditions of hydrostatic equilibrium. Using this result, estimate how much
more would one weigh in vacuum.

Solution

The pressure force acts perpendicular to an area element. Let the area
element on the submerged surface be dS, pointing out of the submerged
volume. The total force on the volume element due to fluid pressure is given
by Z Z
F=− p dS = − ∇p dV (23)

Using the equation of hydrostatic equilibrium we obtain

Z
F = ρg dV = ρgVsub (24)

4
Incidentally, if the object is only partially submerged then a part of the
submerged surface lies inside the object. Treating the top part and the
bottom part separately, we find that the two contributions cancel and we
are left with the above result.

3. Fluid equations as conservation laws: Using the continuity equation, the

Euler equation, and the first law of thermodynamics, derive conservation
laws for the momentum and energy of an ideal fluid. Hint: proving conser-
vation means writing equations in the form:
∂
(mom. or energy density) + ∇ · (mom. or energy current density) = 0
∂t
(25)

Solution

Conservation of momentum: We first write down the Euler equations in the

component form as follows
∂ 1
vi + vj ∂j vi = − ∂i p , (26)
∂t ρ
where summation over repeated indices is assumed. Multiplying both sides
by ρ yields
∂
ρ vi + ρvj ∂j vi = −∂i p (27)
∂t
Multiplying the continuity equation by vi
∂
vi ρ + vi ∂j (ρvj ) = 0 (28)
∂t
Adding the two equation we obtain
∂
(ρvi ) + vi ∂j (ρvj ) + ρvj ∂j vi = −∂i p (29)
∂t
Which can be written as
∂
(ρvi ) + ∂j (ρvi vj + pδij ) = 0 (30)
∂t

5
Defining the stress tensor

Tij = ρvi vj + pδij (31)

this can be written as

∂
(ρvi ) + ∂j Tij = 0 (32)
∂t

Conservation of energy: Total energy per unit volume is the sum of kinetic
energy per unit volume, 12 ρv 2 , and the internal energy per unit volume, ρu,
where u is the specific energy. We shall consider each separately.

Kinetic energy
 
∂ 1 2 1 ∂ρ ∂v
ρv = v2 + ρv · (33)
∂t 2 2 ∂t ∂t
Using equation of continuity and the equations of motion given in the form
 2
∂v v ∇p
+ω×v+∇ =− (34)
∂t 2 ρ

we obtain
 
∂ 1 2 1 1 ∇p
ρv = − v 2 ∇ · (ρv) − ρv · ∇( v 2 ) − ρv · (35)
∂t 2 2 2 ρ

From the first law of thermodynamics we have

∇p
= ∇h − T ∇s (36)
ρ
using which the time rate of change of kinetic energy takes the form
 
∂ 1 2 1 1
ρv = − v 2 ∇ · (ρv) − ρv · ∇( v 2 ) − ρv · ∇h + T ρv · ∇s (37)
∂t 2 2 2

Internal energy: Consider

d(ρu) = ρdu + udρ (38)

From first law of thermodynamics

1
du = T ds + dρ (39)
ρ2

6
Combining the above equations

d(ρu) = ρT ds + hdρ (40)

or
∂(ρu) ∂s ∂ρ
= ρT +h (41)
∂t ∂t ∂t
For an ideal fluid the flow is adiabatic, implying ds/dt = ∂s/∂t + v · ∇s = 0.
Using this and the continuity equation we get
∂(ρu)
= −ρT v · ∇s − h∇ · (ρv) (42)
∂t
Adding the time rate of change of kinetic and internal energy gives
 
∂ 1 2 1 1
ρv + ρu = − v 2 ∇ · (ρv) − ρv · ∇( v 2 ) − ρv · ∇h − h∇ · (ρv) (43)
∂t 2 2 2
which can be rearranged to give
   
∂ 1 2 1 2
ρv + ρu + ∇ · ( ρv + ρh)v = 0 (44)
∂t 2 2

4. Convective instability: When a fluid is disturbed and it settles back into

equilibrium it usually manages to reach mechanical equilibrium faster than
thermal equilibrium.

(a) Estimate these time scales for a parcel of air of size 1 m, and 1 km. The
coefficient of thermal conductivity, κ = 0.2 cm2 sec−1 and speed of soundcs =
350 m sec−1 .

Solution

The time scale of thermal diffusion is given by

l2
τthermal = (45)
κ
and that of pressure equilibration as
l
τsound = (46)
cs

7
Thus for a parcel of size 1 m

τthermal = 13.9 hr (47)

τsound = 3 ms (48)

Thus, for a parcel of size 1 km = 1000 m

τthermal = 1.39 × 105 hr (49)

τsound = 73 s (50)

(b) Earth’s atmosphere could be used as an example of this kind: it is in

approximate mechanical equilibrium but has temperature gradients in hy-
drostatic equilibrium. Derive the conditions (called Schwarzchild criterion)
under which this equilibrium is stable.

Solution

If a parcel of air at height z rises by a height η, it would come into pres-

sure equilibrium with its surroundings much faster than it would come into
thermal equilibrium. Since the parcel is in equilibrium at height z, where
pressure and specific entropy are p and s respectively, then it has a specific
volume given by V (s, p). At height z + η it will have a specific volume given
by V (s, p(z + η)), since it would have ascended adiabatically. The atmo-
sphere, however, need not have constant entropy at all heights. Therefore,
the parcel of air it replaces would have a specific volume V (s(z +η), p(z +η)).
For stability we require the parcel to be denser than the air it replaces, that
is
V (s(z + η), p(z + η)) − V (s, p(z + η)) > 0 (51)
Expanding in the small displacement ψ, we obtain
 
∂V ds
>0 (52)
∂s p dz

From the first law of thermodynamics we have

T ds = du + pdV = d(u + pV ) + V dp = cp dT + V dp (53)

8
from where we deduce that
   
∂V T ∂V
= >0 (54)
∂s p cp ∂T p

Therefore, condition for stability reduces to

ds
>0 (55)
dz
Expressing the specific entropy in terms of pressure and temperature we
obtain    
∂s dp ∂s dT
+ >0 (56)
∂p T dz ∂T p dz
using the first law in the from

T ds = cp dT − V dp (57)

the derivatives of entropy be easily evaluated, and using the equation of

hydrostatic equilibrium dp/dz = −ρg = g/V we obtain
 
dT g
− < (58)
dz cp

5. Linear theory of sound: Assume the unperturbed medium is unbounded,

static, uniform in its properties: ρ = ρ0 , p = p0 , and v0 = 0. The medium
is then perturbed:

(i) Write down the linearized continuity and Euler equations satisfied by
perturbations ρ1 , p1 , and v1 .

(ii) What is the linearized equation satisfied by the perturbed vorticity?

(iii) Assume that the perturbation gives rise to a pure potential flow, v1 =
∇φ1 . Use this in the linearized Euler equation, and express p1 in terms
of φ1 .

(iv) Assume
p that the flow is barotropic, with sound speed defined by cs =
dp0 /dρ0 . Derive a wave equation for φ1 .

9
 (v) Write down the general solution for φ1 , corresponding to plane waves
traveling in the ±x directions.
(vi) What are the corresponding expressions for v1 and p1 ? Are the waves
transverse or longitudinal?

Solution

(i) Since v0 = 0, the nonlinear term in the convective derivative does not
contribute anything. The linearized Euler equations are
∂v1 ∇p1
=− , (59)
∂t ρ0
and the linearized continuity equation is
∂ρ1
+ ρ0 ∇ · v1 = 0 (60)
∂t

(ii) Since the fluid is compressible, we first derive the equation of motion for
vorticity. The Euler equations are
 2
∂v ∇p v
+ω×v =− −∇ (61)
∂t ρ 2
Taking the curl on both sides
∂ω ∇ρ × ∇p
+ ∇ × (ω × v) = (62)
∂t ρ2
The non-linear terms do not contribute in the first order since they are
zero for the unperturbed flow (the term on the right hand side is in fact
identically zero for a polytropic equation of state), therefore
∂ω
=0 (63)
∂t
and unless the perturbation has a non-zero vorticity, the vorticity remains
zero.

(iii) Inserting v1 = ∇φ1 in the Euler equation gives

 
∂φ p1 ∂φ1 p1
∇ + =0⇒ + =C (64)
∂t ρ0 ∂t ρ0

10
(iv) For an adiabatic gas p = p(ρ), which implies p1 = c2s ρ1 . The previous
equation then becomes
∂φ1 ρ1
+ c2s =0 (65)
∂t ρ0
Differentiating with respect to time we obtain

∂ 2 φ1 c2 ∂ρ1
2
+ s =0 (66)
∂t ρ0 ∂t
Using the continuity equation we we can write this as

∂ 2 φ1 2 ∂ 2 φ1
− cs ∇ · v1 = 0 ⇒ − c2s ∇2 φ1 = 0 (67)
∂t2 ∂t2

A general solution of this equation is

φ1 = φ0 ei(k.x±ωt) (68)

where ω/k = c2s . Therefore the solution for pressure is given by

p1 = ∇φ1 = φ0 kei(k.x±ωt) (69)

which is obviously longitudinal.

5. Describe the flow of water in an idealized bathtub where water enters at

a large distance from the drain with a non-zero circulation, and the flow is
axisymmetric and stationary. What is the shape that the surface of water
acquires? You can neglect the viscosity of water.

Solution

We shall assume that the tub is very large and cylindrically symmetric. We
shall further assume that the flux of water is small, therefore, we shall ignore
the vertical motion of water while describing the flow. We also assume that
the flow is steady and incompressible.

11
Water enter at a large radius R0 , with a small radial velocity, and circulation
Γ0 I
Γ0 = v · dl (70)

For an ideal fluid, the net circulation remains constant with the flow, there-
fore, Γ(R) = Γ0 . Thus
I
Γ0
v · dl = 2πRvφ = Γ0 ⇒ vφ = (71)
2πR
Note that this is just a statement of conservation of angular momentum
since the angular momentum per unit volume ρRvφ is a constant for this
flow.

Continuity equation ∇ · v = 0, in cylindrical polar coordinates is

1 ∂(RvR ) 1 ∂vφ ∂vz 1
+ + = 0 ⇒ vR ∝ (72)
R ∂R R ∂φ ∂z R
Therefore, the speed of the fluid varies inversely with the radius. From
Bernoulli’s equation
p v2 K
+ + gz = 0 ⇒ 2 + z = z0 (73)
ρ 2 R
where z0 is the height at infinity. The constant pressure surface is thus
described by
K
(z − z0 ) = − 2 (74)
R
The vorticity of the flow is zero everywhere except at the origin, where it
becomes singular.
ω = k̂Γ0 δ 2 (R) (75)
And thus the flow is that of a line vortex.

2 Viscous Fluids

1. Molecular origin of shear viscosity: For most fluids, the shear viscosity
coefficient is only determined experimentally. For an ideal gas the shear

12
viscosity can be determined using kinetic theory of gases. Consider a flow
in x direction where the shear .i.e. ∂vx /∂y is non-zero. In this case, random
motion of the gas molecules moving with typical thermal velocities vt with
a mean free path ` (` = 1/(nσ), n is the number density and σ is the cross-
section of collision) deposit different amounts of x-component of momentum
across a plane y = const. Show that the (x-component of) momentum
deposited per unit time per unit volume of the fluid is given by:

 
∂ ∂ux
Fx ∼ η , (76)
∂y ∂y
with the coefficient of shear viscosity η = mvt /σ.

2. Using Navier-Stokes and continuity equations along with the first law of
thermodynamics, show that:

      
∂ 1 2 1 2
ρ v +u + ∇. ρv v + h − ζθv − 2ησ.v
∂t 2 2
ds
= ρT − ζθ2 − 2ησij σ ij (77)
dt

Solution

We have done part of this problem already while deriving the energy con-
servation equation. Let us cast the Navier-Stokes equation in the following
form
dvi ∂p ∂Tij
ρ =− − (78)
dt ∂xi ∂xj
where
 
1 ∂vi ∂vk 2 ∂vl ∂vl
Tij = −2ησij − ζθδij , σij = + − δij , θ= (79)
2 ∂xk ∂xi 3 ∂xl ∂xl
Recall the kinetic energy part from the previous problem
 
∂ 1 2 1 ∂ ∂
ρv = v 2 ρ + ρv · v (80)
∂t 2 2 ∂t ∂t

13
It is clear that the second term on the right hand side introduces the fol-
lowing additional terms
 
∂ 1 2
ρv = . . . + 2ηvi ∂j σij + ζvi ∂j θδij (81)
∂t 2
This can be written as
 
∂ 1 2
ρv = . . . + 2η∂j (vi ∂j σij ) + ζ∂j (vi θδij ) − 2η(∂j vi )σij − ζ(∂j vi )θδij
∂t 2
(82)
or  
∂ 1 2
ρv = . . . + ∇ · (2ηv · σ + ζvθ) − 2ησij σij − ζθ2 (83)
∂t 2
Apart from this we also cannot assume conservation of entropy (since dis-
sipative processes actually produce it), therefore, the ds/dt term that had
dropped out in the previous case remains on the right hand side, leading to
the result we are required to prove.

Q. Interpret different terms of this equation and show that it means that
both the coefficients of viscosity must be positive.

Solution

If there are no sources or sinks of energy, the left hand side and the right
hand side of Eq 77 vanish indentically. Note that the flux of energy now
contains additional terms dependent on shear. This is not surprising since
shear transports momentum through the fluid and transport of momentum
involves transport of energy as well.

Putting the right hand side to zero we obtain

ds
ρT = ζθ2 + 2ησij σ ij (84)
dt
since entropy of the fluid always increases due to the second law of thermo-
dynamics, it is clear that both coefficients of viscosity must be positive.

14
3. For an incompressible flow, show that the pressure is determined entirely
by the velocity field and derive a Poisson equation for the pressure.

Solution

For an incompressible fluid the Navier-Stokes equation can be written as

 
∂v p
+ (v · ∇)v = −∇ + ν∇2 v (85)
∂t ρ
where ω is the vorticity. Taking divergence on both sides and using the
incompressibility condition ∇ · v = 0, we obtain

∇2 p = −ρ∂i (vj ∂j vi ) (86)

which is the required Poisson equation

4.
Determine the flow of a viscous fluid of thickness h on an inclined plane due
to the force of gravity.

Solution

Let the direction of the flow be x and the direction perpendicular to the plane
by z, then the gravitational potential is given by Φ = gz cos(θ) − gx sin(θ),
where θ is the angle of the inclined plane.

The condition that the height of the flow remains a constant, along with the
continuity equation implies that vx is a constant. Moreover, the condition
that at the top of the fluid the pressure is given by the atmospheric pressure,
which is a constant along the direction of the flow, implies that the pressure
is only a function of z. The x component of the Navier-Stokes equation is
given by  
∂ p
+ gz cos θ − gx sin θ = ν∇2 vx (87)
∂x ρ
which gives
d2 vx g sin θ
2
=− (88)
dz ν

15
The general solution of this equation is given by
1
vx (z) = − g sin θ z 2 + Bz + C (89)
2ν
The flow satisfies the boundary condition vx (0) = 0. Furthermore, if we
ignore the viscosity of air, it is clear that the condition that ησij = 0 at
z = h, where h is the height of the flow, and that ησij be continuous,
implies that σij = 0. Note that continuity of ησij is a necessary condition
since its divergence occurs in the Navier-Stokes equations, and discontinuity
would lead to singular terms. The two constants can be determined from
these two boundary conditions, vx (0) = 0 and dvx /dz(h) = 0, to give the
solution
g sin θ
vx (z) = z(2h − z) (90)
2ν

5. Poiseuille flow: Consider fluid flow in a pipe of radius a. The flow is such
that the only component of velocity that is non-zero is along the pipe and
its variation is only along the cross-section .i.e. the flow can be determined
by a velocity field: vz (r).

a) Show that for a given pressure gradient along the tube, dp/dz, the velocity
field is given by:

dp 1 2
vz (r) = (a − r2 ). (91)
dz 4η

Solution

Since the flow is steady ∂t = 0. For a long pipe the flow is invariant along
the length of the tube, therefore, the gradient of the flow along the length
vanishes. Thus, the left hand side of the Navier-Stokes equation is zero.
Ignoring any gravitational field we obtain

p v2
 
∇ + = ν∇2 v (92)
ρ 2

16
Taking projection along the z-axis we obtain
dp
= η∇2 vz (93)
dz
Note that dp/dz is a constant. Expressing the Laplacian in the cylindrical
polar coordinates and noting that the only variation is along r, we obtain
 
1 d d 1 dp
r vz (r) = (94)
r dr dr η dz
Integrating twice we obtain
1 dp r2
vz (R) = + C log r + D (95)
η dz 4
where, C and D are constants. Since the flow is non singular at the origin,
C = 0. The other boundary condition is that vz (z) = 0, imposing which we
obtain  
 dp  1
vz (r) =   (a2 − r2 ). (96)
dz 4η

b) Compute the tangential force (per unit area) on the walls of the pipe.

Solution

The force per unit area is given by ηSrz , all other components of the stress
tensor vanish.  
 dp  a
Srz =   (97)
dz 2η
The total force on a piece of length l is given by
 
 dp 
F = 2πalη Srz =   πa2 l (98)
dz
which is not a surprising result since πa2 is the crossectional area of the tube
and the pressure gradient times the length is the pressure difference across
the segment of the tube. Clearly, the pressure forces are required to balance
the total viscous forces on the tube.

17