Solutions

7 Vector and tensor analysis:

1.

2.

3. (a) The easiest way to solve this problem is the following. Consider
the vector:
Ai = ǫijk ∂j ∂k φ (16)
Since ǫijk = −ǫikj ,

−Ai = ǫikj ∂j ∂k φ = ǫikj ∂k ∂j φ (17)

since the derivatives can be interchanged. But the second term on

the right side is also equal to Ai , since the j and k are summed
over. Therefore, we get an equation of the form Ai = −Ai , imply-
ing that Ai = 0.
The physical interpretation is that ∇φ is in the direction ⊥ to
lines of constant φ. However, ∇ × ∇φ involves derivatives in the
plane ⊥ to ∇φ, which is tangent to surfaces of constant φ, so these
derivatives are zero.
(b) First prove that ǫijk ǫklm = δil δjm − δim δjl . Since ǫijk ǫklm is a real
vector, and it is isotropic, it has to be of the form:

ǫijk ǫklm = Aδil δjm + Bδim δjl + Cδij δlm (18)

Multiplying by δil δjm , δim δjl and δij δml respectively, we get the
following equations:

ǫijk ǫkij = Aδii δjj + Bδij δij + Cδij δij

→ 6 = 9A + 3B + 3C
ǫijk ǫkji = Aδij δij + Bδii δjj + Cδij δij
(19)
→ −6 = 3A + 9B + 3C
ǫiik ǫkll = Aδjl δjl + Bδjm δjm + Cδii δll
→ 0 = 3A + 3B + 9C

17
 The solutions to these equations are A = 1, B = −1 and C = 0,
and hence we obtain the above identity.
This can be easily used to prove:

∇ × ∇ × u = ǫijk ǫklm ∂j ∂l um
= (δil δjm − δim δjl )∂j ∂l um
(20)
= ∂i ∂j uj − ∂j2 ui
= ∇(∇·u) − ∇2 u

(c) If B = Sij Aij , (B is a scalar), then we find that B = −Sji Aji (since
Sij = Sji and Aij = −Aji ). But since i and j are summed over, it
is also true that B = Sji Aji . Therefore, B = −B, implying that
B = 0.
(d) The antisymmetric tensor Aij has only three independent compo-
nents (since the diagonal terms are zero), and therefore they can
be expressed in terms of the three components of a vector ωk . The
only stipulation is that Aij = −Aji , and this is satisfied if:

Aij = ǫijk ωk → Aji = ǫjik ωk = −Aij (21)

The product ǫijk Ajk is given by:

ǫijk Ajk = ǫijk ǫjkl ωl

= (δil δjj − δij δlj )ωl (22)
= 2ωi

4. Z
Fij dV f (r) ri rj (23)
V
First note that the only vector Fij can depend on is ai . So Fij has to
be a product of ai , δij and ǫijk . The only permissible combination of
these is:
Fij = Aδij + Bai aj (24)
For evaluating A and B, we need two equations. The first of these can
be obtained using:
δij Fij = 3A + B = I1 (25)

18
where the integral I1 is:
Z
I1 = dV f (r) r 2
ZV2π Z θ0 Z 1
= dφ sin (θ)dθ dr r 4 f (r) (26)
φ=0 θ=0 r=0
Z 1
= 2π(1 − cos (θ0 )) dr f (r) r 4
0

The second equation is obtained from:

ai aj Fij = A + B = I2 (27)

where the integral I2 is:

Z
I2 = dV f (r)(riai )2
ZV
= dV f (r) r 2 cos (θ)2
ZV2π Z θ0 Z 1 (28)
= dφ cos (θ)2 sin (θ)dθ dr r 4 f (r)
φ=0 θ=0 r=0
Z 1
3 4
= (2π/3)(1 − cos (θ0 ) ) dr r f (r)
0

The constants A and B are then given by:

A = (I1 − I2 )/2 (29)

B = (3I2 /2) − (I1 /2) (30)

For a sphere, we would expect the result to be an isotropic tensor, and
therefore B = 0. Since cos (θ0 ) = −1 for a sphere, we find that:
Z 1
I1 = 4π dr r 4 f (r) (31)
0
Z 1
I2 = (4π/3) dr r 4 f (r) (32)
0
Using this, we get:
4π 1
Z
A= dr r 4 f (r) (33)
3 0
B=0 (34)

19
5. The integral Iij is:
Z
Iij = dSai bj ak bl xk xl
S (35)
= Aδij + Bai aj + Cbi bj + Dai bj + Eaj bi

Multiplying the above equation by δij , ai aj and bi bj we get the following

simultaneous equations for A, B and C:

3A + B + C = 0

A+B =0
A+C =0
These three can be solved to give A = 0, B = 0 and C = 0. Multiplying
1̊ by aj bi we get:
E=0
Multiplying 1̊ by ai bj we get:
Z
D = a2i b2j dSak bl xk xl
S

In order to calculate this integral, take a and b along the x1 and x2

directions. Then the equation becomes:
Z 2π Z π
D= dφ sin (θ)dθ[cos (θ) cos (φ) cos (θ) sin (φ)]
0 0

It can easily be verified that the right side of the above equation is
identically zero.

6. (a) Since we know that the unit vectors eξ and eη are in the directions
of ∇ξ and ∇η respectively, it is necessary to find these vectors.
It is not possible to invert the expressions to determine ξ and η
in terms of x and y, but we can find the expressions for the unit
vectors directly. The unit vectors in the x and y directions are
given by:

∇x = ex = (∇ξ) sinh (ξ) cos (η) − (∇η) cosh (ξ) sin (η)
(36)
∇y = ey = (∇ξ) cosh (ξ) sin (η) + (∇η) sinh (ξ) cos (η)

20
 The above simultaneous equations can be solved to determine ∇ξ
and ∇η:

sinh (ξ) cos (η)ex + cosh (ξ) sin (η)ey

∇ξ =
[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2
(37)
− cosh (ξ) sin (η)ex + sinh (ξ) cos (η)ey
∇η =
[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2

The magnitude of the gradients ∇ξ and ∇η are:

|∇ξ| = |∇η| = {[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2 }−1/2 (38)

and therefore the unit vectors eξ and eη are:

sinh (ξ) cos (η)ex + cosh (ξ) sin (η)ey

eξ =
{[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2 }1/2
− cosh (ξ) sin (η)ex + sinh (ξ) cos (η)ey (39)
eη =
{[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2 }1/2

(b) From the above equation, it can easily be verified that eξ ·eη = 0,
indicating that the coordinate system is an orthogonal one.
(c) The expressions for hξ and hη can be derived using:

dx = dxex + dyey (40)

From equation 1̊, dx and dy are given by:

dx = sinh (ξ) cos (η)dξ − cosh (ξ) sin (η)dη

(41)
dy = cosh (ξ) sin (η)dξ + sinh (ξ) cos (η)dη

In addition, the equation for ex and ey are given in terms of eξ

and eη in 1̊a. Using these and after some algebraic simplification,
we get the following expression for dx:

dx = {[sinh (ξ) cos (η)]2 +[cosh (ξ) sin (η)]2 }1/2 {dξeξ +dηeη } (42)

Therefore, we find that

hξ = hη = {[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2 }1/2 (43)

21
 The Laplacian in an orthogonal coordinate system in two dimensions,
using the extension of the formula derived in class is:
" ! !#
2 1 ∂ h2 ∂φ ∂ h1 ∂φ
∇ φ= + (44)
h1 h2 ∂y1 h1 ∂y1 ∂y2 h2 ∂y2
Using this formula, we find that:
∂2φ ∂2φ
!
2 1
∇ φ= + 2 (45)
[sinh (ξ) cos (η)]2 + [cosh (ξ) sin (η)]2 ∂ξ 2 ∂η

7. The definition of ∇2 φ is ∇·(∇φ) where ∇φ is:

1 ∂φ 1 ∂φ 1 ∂φ
∇φ = e1 + e2 + e3 (46)
h1 ∂y1 h2 ∂y2 h3 ∂y3
Using the equation for the divergence of a vector derived in class,
" ! ! !#
1 ∂ h2 h3 ∂φ ∂ h3 h1 ∂φ ∂ h1 h2 ∂φ
∇·(∇φ) = + +
h1 h2 h3 ∂y1 h1 ∂y1 ∂y2 h2 ∂y2 ∂y3 h3 ∂y3
(47)

8 Kinematics
1.
2. The velocity field is,
x2 + y 2
!
vz = V 1− (48)
R2
Therefore, the rate of deformation tensor is given by
 
0 0 0
∇v = 

0 0 0 
 (49)
(−2x/R) (−2y/R) 0
The isotropic part of this rate of deformation field is zero, while the
symmetric and anti-symmetric parts are,
 
0 0 (−x/R)
S=

0 0 (−y/R) 
 (50)
(−x/R) (−y/R) 0

22
  
0 0 (x/R)
A=

0 0 (y/R) 
 (51)
(−x/R) (−y/R) 0

3. The velocity field is given by vθ = (Ω/r). In cylindrical co-ordinates,

it can easily be seen that the rate of deformation tensor is,
!
(∂vr /∂r) (∂vθ /∂r)
∇v = (52)
(1/r)(∂vr /∂θ) − (vθ /r) (vr /r) + (1/r)(∂vθ /∂θ)
!
0 −(Ω/r 2 )
∇v = (53)
−(Ω/r 2 ) 0
The stress tensor is symmetric in this co-ordinate system, and the curl
of the velocity is zero. The velocities in a Cartesian co-ordinate system
are,
Ωy
vx = −vθ sin (θ) = − 2 (54)
x + y2
Ωx
vy = vθ cos (θ) = (55)
x2
+ y2
The rate of deformation tensor in a Cartesian co-ordinate system is
!
(2Ωxy)/(x2 + y 2 )2 Ω(y 2 − x2 )/(x2 + y 2 )2
∇v = 2 2 2 2 2 (56)
Ω(y − x )/(x + y ) (−2Ωxy)/(x2 + y 2 )2

The stress tensor is symmetric. The vorticity can be verified to be zero

in both the Cartesian and cylindrical co-ordinates.

4.

9 Conservation equations:
1.

D = 2µsij sji − (2/3)µs2kk

= 2µ(s211 + s222 + s233 + 2s12 s21 + 2s13 s31 + 2s23 s32 )
−(2/3)µ(s211 + s222 + s233 + 2s11 s22 + 2s11 s33 + 2s22 s33 )
= 2µ(s212 + s213 + s223 ) + (2µ/3)((s11 − s22 )2 + (s22 − s33 )2 + (s11 − s33 )2 )
(57)

23
10 Viscous flows:
1. (a) The problem can be taken as the superposition of two problems —
one with the force of gravity parallel to the line joining the sphere’s
centeres, and the other with the force of gravity perpendicular to
the line joining the centers. It is clear from reversibility that if
two spheres separated when they fell, they would come together
when the direction of gravity is reversed. This violates symmetry,
so the two spheres can neither come together not separate from
each other, and therefore the distance between their line of centers
remains a constant.
(b) When Fg , the force due to gravity, is to p, U = Uk gi . When Fg
is ⊥ to p, then U = 12 Uk gi . In the general case where Fg is at an
angle to p, then the components in the k and ⊥ directions are:

Fk = Fg cos (θ) F⊥ = Fg sin (θ) (58)

This induces a velocity Uk cos (θ) parallel to p and a velocity

1
U sin (θ) perpendicular to p. Therefore, the velocity in the ver-
2 k
tical direction is given by:
 
1
Uv = (Uk cos (θ)) cos (θ) + U
2 k
sin (θ) sin (θ)
(59)
= 21 Uk (1 + cos (θ)2 )

The velocity in the horizontal direction is:

Uh = (Uk cos (θ)) sin (θ) − ( 12 Uk sin (θ)) cos (θ)

(60)
= 12 Uk cos (θ) sin (θ)

2.

3. It is convenient to separate the fluid velocity ui = Gij xj + u′i, where

u′i → 0 for r → ∞. The velocity u′i is obtained by solving the equation:

µ∇2 u′i = ∇p ∇2 p = 0 (61)

The solution of this is given by:

u′i = u′ih + (1/2µ)xip (62)

24
where u′ih is the solution of the homogeneous equation ∇2 u′ih = 0.
The velocity u′i and the pressure p can depend only on the tensor Gij ,
and so the possible solutions are:
!
δjk 3xj xk
p = 2µλ1 Gjk − (63)
r3 r5
!
λ2 Gij xj 15xi xj xk 3(δij xk + δjk xi + δik xj )
u′ih = + λ3 Gjk − (64)
r r7 r5
The equation for the pressure is inserted into 2̊, and we can simplify
the above expressions by noting that Gii = 0:
!
−3xi xj xk Gij xj 15xi xj xk 3(xj δik + xk δij )
   
u′i = λ1 Gjk 5
+λ2 2
+λ3 Gjk −
r r r7 r5
(65)
One of the constants in the above equation can be determined from the
equation of continuity:
∂i u′i = 0 (66)
!
−3(δii xj xk + δij xi xk + δik xi xj ) 15xi xi xj xk
λ1 + Gjk
r5 ! r7
δij 3xi xj
+ λ2 Gij 3
− 5
r r
(67)
15(δii xj xk + δij xi xk + δik xi xj ) 105xi xi xj xk
+ λ3 Gjk −
r7 ! r
9

3(δij δik + δik δij ) 15(xi xj δik + xi xk δij )

− + =0
r5 r7
The above equation can be simplified using the conditions that Gii = 0
and xi xi = r 2 .
xi xj
−3λ2 Gij 5 = 0 → λ2 = 0 (68)
r
In addition, the center of the particle is at the origin and is stationary,
so we have the condition:
ui = 0 at r = a (69)
This gives us:
−3xi xj xk 15xi xj xk 6xk
   
λ1 Gjk + λ 3 G jk − 6 + Gij xi = 0 (70)
r5 a7 a

25
The above equation can be easily solved to give:

5a3 a5
λ1 = λ3 = (71)
6 5
Inserting these into the expression for the velocity 5̊, the final expression
for the velocity is:

5 a3 xi xj xk Gjk 5 a5 xi xj xk a5 (xj δik + xk δij )

!
ui = Gij xj − + G jk −
2 r5 2 r7 2r 5
3
5µa Gjk xj xk
p= −
r5
(72)
The stress acting on the particle is given by:
!
∂ui ∂ul
Til = −pδil + µ + (73)
∂xl ∂xi

At the surface of the particle, (∂ui /∂xl ) is:

∂ui 5 25
= Gil − 2 Gjk (δil xj xk + δjl xi xk + δkl xi xj ) + 4 Gjk (xi xj xk xl )
∂xl 2a 2a
5 35
+ 2 Gjk (δil xj xk + δjl xi xk + δkl xi xj ) − 4 xi xj xk xl Gjk
2a 2a
1 5
− (δjl δik + δkl δij )Gjk + 2 Gjk (xj xl δik + xk xl δij )
2 2a
−5 5
= xi xj xk xl Gjk + 2 Gik xk xl
a4 a
(74)
Using the above relation, the stress is:
5µ 10µ
Til = (δil xj xk + δij xk xl + δjl xi xk )G jk ) − (xi xj xk xl Gjk ) (75)
a2 a4
The product Til nl xm is given by:
xl
Til nl xm = Til xm
a
5µ 10µ
= 3 (2xi xj xk xm + δij xk x2l xm )Gjk − 4 (xi xj xk x2l xm )Gjk
a a
5µ
= 3 Gik xk xm
a
(76)

26
 The integral of this over the surface of a aphere is:
5µ 4 3
Z  
dA 3
Gik xk xm = 5µGim πa (77)
A a 3

4. (a) At low Reynolds number, the velocity can be separated into a

homogeneous and a particular solution:

ui = uhi + upi (78)

The homogeneous solution is obtained from the equation:

∂j2 uhi = 0 (79)

while the particular solution can be expressed as a function of the

pressure:
1
upi = pxi (80)
2µ
where the pressure is obtained by solving:

∂j2 p = 0 (81)

Note that the angular velocity vector Ω is a pseudo vector, while ui

and p are real vectors. Since Ω is the only vector in the system, ui
and p are linear functions of Ω. However, there is no way to make
a real scalar p from δij , ǫijk and the pseudo vector Ωi . Therefore,

p=0 (82)

Therefore, the particular solution for the velocity is also zero. The
general solution for the velocity has to be of the form:
A Bi xi
ui = + 3 + ... (83)
r r
The only way to make a real vector ui which decays as r → ∞
and is linear in Ωk is:
xk
ui = λǫijk Ω (84)
r3

27
 where λ is a constant. Note that the incompressibility condition
is automatically satisfied by the above equation:
!
δik 3xi xk
∂i ui = λǫijk Ωj − =0 (85)
r3 r5
The constant λ can be determined from the boundary condition
that at r = a, ui = ǫijk Ωj xk . This gives λ = a3 , and therefore the
solution for the velocity is:
ui = ǫijk Ωj xk a3 /r 3 (86)

(b) The force per unit area on the surface is fi = Til nl . Since the
pressure is zero, the stress tensor Til is given by:
Til = µ(∂l ui + ∂i ul ) (87)
The strain rate is:
!
3 δkl 3xk xl
∂l ui = ǫijk Ωj a −
r3 r5 !
(88)
3 δik 3xi xk
∂i ul = ǫljk Ωj a −
r3 r5
Therefore, the force fi is given by:
xl
fi = Til
r" !
3 δkl xl 3xk xl xl
= µa ǫijk Ωj − (89)
r4 r 6!#
δik xl 3xi xk xl
+ǫljk Ωj a3 −
r4 r5
The fourth term on the right side of the above equation contains
x × x = 0. Further, it can be easily verified that the first and
third terms cancel, and so the force is given by:
−3µa3 ǫijk Ωj xk
fi = (90)
r4
The torque on the sphere is:
Z
Lm = dAǫmin xj (−3µ/a)ǫijk Ωj xk
Z (91)
= ǫmin ǫijk Ωk (−3µ/a) dAxn xk

28
 In order to calculate the integral in the above equation, use direc-
tional symmetries:
Z
dAxn xk = λδnk
Z
dAr 2 = 3λ (92)
4πa4
λ=
3
Therefore, we get
Lm = ǫmin ǫijk Ωj (−4πµa3 )
(93)
= −8πµa3 Ωm

5. Consider a cylindrical coordinate system with the origin located at the

wall perpendicular to the center of the disk. It is appropriate to non -
dimensionalise the length, time and velocity scales as follows:
z∗ r∗ u∗z u∗r ǫ t∗ ǫU
z= r= uz = ur = t= (94)
aǫ a U U a
where the variables with the superscript ∗ are dimensional, while those
without the superscript are dimensional. The scaling for ur in the above
equation was determined from the mass conservation equation:
1 ∂(r ∗ u∗r ) ∂u∗z
+ ∗ =0 (95)
r ∗ ∂r ∗ ∂z
With the above scaling, the momentum equations in the r direction is:
a ∂p∗ 1 ∂ 2 ur 1 1 ∂ 1 u2r
! ! !
Re ∂ur ∂ur ∂ur ∂ur
+ ur + uz =− + 3 + r −
ǫ2 ∂t ∂r ∂z µU ∂r ǫ ∂z 2 ǫ r ∂r ∂r ǫ r
(96)
In the leading approximation, this equation reduces to:
∂ 2 ur ǫ3 a ∂p∗
2
= + O(Reǫ) + O(ǫ2 ) (97)
∂z µU ∂r
From this equation, it is appropriate to scale p∗ by (µU/aǫ3 ), and the
leading order equation is:
∂ 2 ur ∂p
= (98)
∂z 2 ∂r
29
The momentum conservation equation in the z direction is:

1 ∂ 2 uz 1 ∂
! !
Re ∂uz ∂uz ∂uz 1 ∂p ∂uz uz
+ ur + uz =− 4 + 2 2
+ r −
ǫ ∂t ∂r ∂z ǫ ∂z ǫ ∂z r ∂r ∂r r2
(99)
In the limit ǫ → 0, this equation is:
∂p
= 0 + O(ǫ2 ) + O(ǫ3 Re) (100)
∂z
The boundary conditions are:

ur = 0 uz = 0 at z = 0
(101)
ur = 0 uz = −1 at z = 1

Equation 7̊ for the pressure field implies that p is only a function of r

(p = p(r)). Using this information, equation 5̊ for the radial velocity
can be solved:
1 ∂p 2
ur = (z − z) (102)
2 ∂r
This solution satisfies the boundary condition 8̊ ur = 0 at z = 0 and
z = 1. The value of the pressure gradient can be obtained by examining
the continuity equation:

∂uz 1 ∂(rur )
=− (103)
∂z r ∂r
Integrating this between z = 0 and 1, we get

1 ∂(rur )
Z 1
uz |1z=0 = − dz
0 ! r ∂r !1
−1 ∂ ∂p z3 z2
−1 = r − (104)
r ∂r ∂r 6 ! 4 z=0
1 1 ∂ ∂p
−1 = r
12 r ∂r ∂r

The above equation for the pressure can be solved to obtain:

p = −3r 2 + C1 log (r) + C2 (105)

30
 The constant C1 is zero because the pressure has to be finite at r = 0,
while C2 = 3 because p = 0 at r = 1. Therefore, the pressure and
velocity are:

p = 3(1 − r 2 ) ur = 3r(z − z 2 ) uz = (2z 3 − 3z 2 ) (106)

The force on the disk per unit area is:

∂u∗z
fz = −nz Tij nz = p∗ − 2µ (107)
∂z ∗
There is a negative term in the above equation because the unit normal
to the disk is in the −z direction. The second term on the right side of
the above equation can be neglected, because it is O(ǫ2 ) smaller than
the first term. Therefore, the force per unit area is:
µU
fz = 3
3(1 − r 2 ) (108)
aǫ
The total force is determined by integrating the force over the area of
the disk:
r ∗2
!!
∗ ∗ µU
Z 2π Z a
Fz = dθ dr r 3 1− 2
0 0 aǫ3 a (109)
−3πµaU
=
2ǫ3
6. (a) It is appropriate to choose a two dimensional Cartesian coordinate
system with the origin on the channel wall. The equation for the
surface of the cylinder is:

(x∗c2 − x∗2 )2 + x∗1 2 = a2 (110)

where ∗ is used to denote dimensional quantities, and x∗c2 = a(1+ǫ)

is the position of the center of the cylinder. The above equation
can be reduced to:
q
x∗2 = a(1 + ǫ) − a2 + x∗1 2 (111)

The coordinate x∗2 is scaled by ǫa, and the coordinate x∗1 by ǫ1/2 a
to obtain the following equation for the surface:
x21
x2 = H(x1 ) = 1 + + ... (112)
2
31
 The Navier - Stokes mass and momentum equations are:
∂1∗ u∗1 + ∂2∗ u∗2 = 0 (113)
ρ(∂t∗ u∗1 + u∗1 ∂1∗ u∗1 + u∗2 ∂2∗ u∗1 ) = −∂1∗ p∗ + µ(∂1∗ 2 + ∂2∗ 2 )u∗1 (114)
ρ(∂t∗ u∗2 + u∗1 ∂1∗ u∗2 + u∗2 ∂2∗ u∗2 ) = −∂2∗ p∗ + µ(∂1∗ 2 + ∂2∗ 2 )u∗2 (115)
Scaling the velocity u∗1 by U, the velocity by ǫ U and theu∗2 1/2
3/2 2
pressure by (µU/ǫ a ), the dimensionless equations are:
∂1 u1 + ∂2 u2 = 0 (116)
(ρUaǫ3/2 /µ)(∂t + u1 ∂1 + u2 ∂2 )u1 = −∂1 p + (∂22 + ǫ∂12 )u1 (117)
(ρUaǫ5/2 )(∂t + u1 ∂1 + u2 ∂2 )u2 = −∂2 p + ǫ(∂22 + ǫ∂12 )u2 (118)
The inertial terms in the above equations can be neglected for
(ρUaǫ3/2 /µ) ≪ 1.
(b) The boundary conditions required for solving the above problem
are:
u1 = 0 at x2 = 0
u2 = 0 at x2 = 0
(119)
u1 = 1 at x2 = H(x1 )
u2 = 0 at x2 = H(x1 )
In addition, there should also be no net flow of fluid across any
surface, so:
Z H(x1 )
dx2 u2 = 0 (120)
0

(c) The Stokes equations are:

∂1 u1 + ∂2 u2 = 0 (121)
−∂1 p + ∂22 u1 = 0 (122)
−∂2 p = 0 (123)
The two momentum equations can be easily solved along with the
boundary conditions and the zero net flux condition to obtain:
3x22 2x2
!
u1 = − (124)
H2 H
6
∂1 p = (125)
H2

32
11 Potential flow:
1. The equation for the potential for a cylinder moving with a constant
velocity Ui is:
Ui xi
φ=λ 2 (126)
r
and the fluid velocity is:
!
δij 2xi xj
ui = λUj 2
− 4 (127)
r r

The constant λ can be determined from the condition that ui ni = Ui ni

at the surface of the cylinder r = a:

2x2i xj
!
xj Ui xi
λUj − = (128)
a3 a5 a

This can be solved to obtain λ = −(1/a2 ). Therefore, the equation for

the potential is:
Ui xi a2
φ=− 2 (129)
r
The pressure field is given by:

∂φ ρu2
p = p0 − ρ − ρUi ui − i (130)
∂t 2
The first, third and fourth terms in the above equation do not con-
tribute to the net force on the cylinder, and the only contribution is
from the second term which is given by:
Z
Fi = dApni (131)

The force along the direction of flow is:

dUj a2 xi xj
Z
Fi = dA 3
dt r (132)
dUj  2
= ρπa
dt
Therefore, the added mass of the cylinder is equal to the mass of fluid
displaced by it.

33
2. The equation for the stream function in the presence of a source at
(−d/2) and a sink at (d/2) and a uniform flow in the x direction is:
mθ1 −mθ2
ψ= + + Ur sin (θ) (133)
2π 2π
where θ1 is the angle subtended at the source and θ2 is the angle sub-
tended at the sink. The angles θ1 and θ2 can be related to θ and r as
follows:
r sin (θ) r sin (θ)
tan (θ1 ) = tan (θ2 ) = (134)
(d/2) + r cos (θ) r cos (θ) − (d/2)
These can be inserted into the above equation, and we can set ψ = 0
to obtain the equation for the surface of the body:
" ! !#
−1 r sin (θ) r sin (θ) 2πUr
tan − tan−1 + sin (θ) = 0
r cos (θ) − (d/2) r cos (θ) + (d/2) m
(135)
In the limit d → 0, the angles θ1 and θ2 are given by:
d sin (θ) d sin (θ)
θ1 = θ − θ2 = θ + (136)
2r 2r
Therefore, the equation for the streamline ψ = 0 becomes:
d sin (θ) 2πUr sin (θ)
= (137)
r m
which is the equation for an infinite cylinder of radius:
!1/2
md
r= (138)
2πU

3. (a) The fluid velocity field due to a sphere moving in potential flow
is:
−Uj a3 δij
!
3xi xj
ui = − 5 (139)
2 r3 r
The kinetic energy per unit mass is given by:
ρu2i ρUj Uk a6 δij
! !
3xi xj δik 3xi xk
= 3
− 5 3
−
2 8 r r !r r5
2 6 6
ρUi a 3ρUj Uk a xj xk
= + (140)
8r 6 8r 8
2 6
ρU a
= 6
(1 + 3 cos (θ)2 )
8r
34
 where θ is the angle made by the position vector with the direction
of the velocity vector.
The total kinetic energy is can be easily calculated from the above
expression:
ρu2i 2π π ∞ ρU 2 a6
Z Z Z Z
dV = dφ dθ sin (θ) dr r 2 (1 + 3 cos (θ)2 )
V 2 0 0 a 8r 6
πρU 2 a3
=
3
(141)
The added mass obtained is given by:
KE 2πa3 ρ
M= = (142)
(U 2 /2) 3
This is exactly half the mass of the fluid displaced by the sphere.
(b) The fluid velocity field for a viscous flow is given by:
a3
" #
3a xi xj 3xi xj
  
ui = Uj δij + 2 + 3 δij − 2 (143)
4r r r r
Since the velocity decays as (a/r) and the volume increases as
r 3 , the total kinetic energy is O(ρU 2 a2 r), which becomes infinite
for an infinite volume. The situation can be rectified, however,
by realising that the Stokes flow approximation is valid only for
(r/a) ≪ (1/Re), and beyond this the Oseen approximation has
to be used. This decays much faster. Therefore, we can estimate
the kinetic energy as O(ρU 2 a3 /Re). This is much larger than
the kinetic energy of O(ρU 2 a3 ) in potential flow for Re ≪ 1.
This might be expected from the minimum energy theorem, which
states that the potential flow has energy which is small compared
to any other flow.
4. (a) The condition for the flow to be irrotational is that the circulation
along any streamline should be a constant. This requires that:
Ω1 R12 = Ω2 R22 (144)
The fluid velocity as a function of radius is given by:
Ω1 R12
vθ = (145)
r

35
 (b) The momentum conservation equation for the fluid in the r and
z directions if viscous forces are neglected at steady state is given
by:
∂z p = ρg (146)
vθ2
∂r p = (147)
r
These equations can be integrated to give:

Ω1 R12
(p0 /ρ) + gz + =0 (148)
2r 2
This gives the equation for the surface. This is also identical to
the Bernoulli equation for the fluid at the surface. Therefore, we
find that z ∼ (1/r 2 ) for this interface.

5. The coordinates in the z and z ′ planes are related by

a2 x′
x = x′ + (149)
x′2 + y ′2

a2 y ′
y = y′ − (150)
x′2 + y ′2
A circle of radius a is transformed onto a line of length 2a along the
x axis centered at the origin. A flow around a cylinder in the z ′ plane
gets converted into a flow past a flat surface.
A circle of radius b in the z ′ plane gets converted onto an ellipse of
manor and minor axes (1 + a2 /b2 ) and (1 − a2 /b2 ) in the z plane.

6.

7. The velocity field due to the potential flow around a sphere with ve-
locity Ui is:
Uj a3 δij 3xi xj
!
ui = − − 5
2 r3 r
The rate of dissipation of energy due to viscous dissipation is:
1 Z
D = η dV (∂i uj + ∂j ui )(∂i uj + ∂j ui )
2 V

36
 where η is the viscosity. Since the flow is irrotational, the strain tensor
is symmetric and the equation for the dissipation rate becomes:
Z
D = 2η dV (∂i uj )(∂j ui )
V

The equation for the strain rate is:

Uk a3
!
−3(δik xj + δij xk + δjk xi ) 15xi xj xk
∂j ui = − +
2 r5 r7

The product (∂i uj )(∂j ui ) can be easily calculated from the above rela-
tion:
Uk Ul a6 18δkl x2i + 36xk xl
!
(∂i uj )(∂j ui ) =
4 r 10
The above equation can be easily integrated over the volume of the
fluid in spherical coordinates (using xk xl = r 2 cos (θ)2 and x2i = r 2 ) to
give:
D = 12πηaU 2
The drag force is given by:

FD = (D/U) = 12πηaU

This is twice the drag force due at zero Reynolds number (FD =
6πηaU). We would expect the drag in potential flow to be greater,
due to the mimimum dissipation theorem which states that the energy
dissipation in a zero Reynolds number flow is lower than that in any
other flow.

8. The force acting on the object is:

Z
Fi = − dSpni (151)
S

where S is the surface of the sphere. From the Bernoulli equation, this
is:
1 2
Z  
Fi = dSρ uj − Uj uj ni (152)
S 2
The first term on the right can be simplified as follows:
1 1 1
Z Z Z Z
dS∞ u2j ni + dSw u2j ni − dS u2j ni = dV uj ∂i uj (153)
S∞ 2 Sw 2 S 2 V

37
 where Sw is the surface of the wall and S∞ is the surface at infinity. The
first term on the left is zero because the velocity decays as r 3 at large
r for a body in steady motion. The right side of the above equation in
an irrotational flow is:
Z Z
dV uj ∂i uj = dV uj ∂j ui
V ZV Z Z
= dS∞ uj nj ui + dSw uj nj ui − dSuj nj ui
S∞ Sw S
(154)
Once again neglecting the contribution from the surface at infinity, and
at the wall uj nj is zero because there is no flux through the wall. Also,
at the surface of the object, uj nj = Uj nj . With this simplification, we
get:
1 2
Z Z
Fi = ρ dSw uj ni + Uj dS(uinj − uj ni ) (155)
Sw 2 S

It can be show, using methods similar to that used in the class, that
Z Z Z
dS(uj ni − uinj ) = dSinf ty (uj ni − uinj ) + dSw (uj ni − uinj )
S S∞ Sw
(156)
The first integral on the right side is zero, and the final expression for
the force on the object is:
1 2
Z  
Fi = ρ dSw u ni + Uj (ui nj − uj ni ) (157)
Sw 2 j
At the wall, the unit normal ni is (0, −1, 0). Using this, the forces in
the two directions can easily be determined:

F1 = 0 (158)
1 2
Z  
F2 = dSw u + U1 u1 (159)
Sw 2 i
9. The fluid velocity due to surface displacements is O(ωξ0), while the
length scale is the wavelength of the fluctuations λ. Therefore, the
uj ∂j ui term is O(ω 2ξ02 /λ), while the ∂t ui term is O(ω 2ξ0 ). The former
can be neglected compared to the latter for (ξ0 /λ) ≪ 1, or when the
amplitude of the fluctautions is small compared to the wavelength.

38
 Consider a coordinate system where the z axis is in the vertical direc-
tion, and the fluid occupies the space z ≤ 0 in the absence of fluctua-
tions. The equation for the surface and the velocity potential can be
expressed as a function of the wave number and frequency as follows:

ξ = ξ0 exp (ikx + iωt) φ = f (z) exp (ikx + iωt)

The equation for the velocity potential is:

∂i2 φ = 0 =⇒ (∂z2 − k 2 )f (z) = 0

The above equation, along with the boundary condition that φ → 0 for
z → −∞, implies that:

f = C1 exp (kz)

The constant C1 is determined from the boundary condition that uz =

∂t ξ at z = 0.
C1 = k −1 iωξ0

The Bernoulli equation for the pressure at the surface is:

p + ρ(∂t φ + gξ) = p0

where p0 is the pressure above the surface. At equilibrium, in the

absence of any flow, we have:

p = p0

Therefore, the equation for the displacement field is:

∂t φ + gξ = 0

Inserting the expressions for φ and ξ, we find that the frequency is

given by: q
ω = (gk)

10. For the potential flow around the sphere, the velocity far from the
sphere is given by (∂φ/∂xi ) = Gij xj . Therefore, the potential is of the
form, !
Gjk xj xk δjk 3xj xk
φ= + BGjk − 3 + (160)
2 r r5

39
 Since the flow is incompressible, δjk Gkj = 0. At the surface, ui ni =
ni (∂φ/∂xi ) = 0. Using these two conditions, we obtain,

r3
!
Gjk xj xk
φ= 1+ (161)
2 3R3

12 Boundary layer theory:

1. For the outer potential flow solution, we have:

ρ−1 ∂1 p = U1 ∂1 U1 + U3 ∂3 U1 = 0 (162)

ρ−1 ∂3 p = U1 ∂1 U3 + U3 ∂3 U3 = 0 (163)
Therefore the pressure is a constant in the outer flow. It is also impor-
tant to note that there is no variation in the x3 direction for a plate
that is infinite in that direction. The equations for the fluid velocity in
the boundary layer are:

∂1 u1 + ∂2 u2 = 0 (164)

u1 ∂1 u1 + u2 ∂2 u1 = ν∂22 u1 (165)
u1 ∂1 u3 + u2 ∂2 u3 = ν∂22 u3 (166)
The mass conservation equation 3̊ and the momentum equation in the
x1 direction 4̊ are identical to those for the Blasius flow over a flat
plate, and the solution for the fluid velocity profile is identical to that
obtained in class.
u1 = U1 f ′ (η) (167)
q
where η = x2 / (νx1 /U1 ), and f (η) is the solution of the equation:

f ′′′ + (1/2)f f ′′ = 0 (168)

The equation for the velocity u3 5̊ is identical to that for u1 , therefore

the solution for u3 will be proportional to u1 . The condition u3 = U3
as x2 → ∞ requires that:

u3 = U3 u1 /U1 (169)

40
 In case the velocity U3 is a function of x1 , then the pressure gradient
in the x3 direction is not zero.

−ρ−1 ∂3 p = U1 ∂1 U3 (170)

The momentum equation for the x3 direction 5̊ now becomes:

u1 ∂1 u3 + u2 ∂2 u3 = U1 ∂1 U3 + ν∂22 u3 (171)

The above equation is identical to the equation we had solved earlier

5̊ in the absence of a pressure gradient, except for the additonal inho-
mogeneous term. The general solution is the same as 8̊, but there is an
additional particular solution. Since the velocity U3 is proportional to
x1 , we can try a solution of the form:

u3 = U30 u1 /U1 + U31 x1 g(η) (172)

Inserting this into 1̊0, we get the following equation for g(η):

g ′′ + (1/2)f g ′ − f ′ g − 1 = 0 (173)

This equation can be solved for the unknown function g(η), since the
function f (η) is known. The boundary conditions are g = 0 at η = 0,
and g = 0 at η → ∞.

2. The equations for the two components of the velocity are,

ux = Uf ′ (η)
1/2
1 νU

uy = (ηf ′(η) − f (η))
2 x
where η = (y/(νx/U)1/2 ). We use a similarity form for the equations,
(T −T0 )/(T∞ −T0 ) = h(η). Inserting this into the temperature equation,
we get,
d2 h dh
2
+ Prf (η) =0
dη dη
This is the equation that has to be solved for the similarity solution
h(η).
If the Prandtl number is large, we need to rescale η in the governing
equations. Note that the thermal boundary layer is small compared to

41
 the momentum boundary layer in this case, and so we are considering
the limit η ≪ 1. In this case, the leading approximation for f (η) is
f (η) = f ′′ (0)η 2 , since both f (η) and f ′ (η) are zero at η = 0. We
assume a form of the dimensionless variable η = Pra ξ, where ξ is the
new scaled co-ordinate, and a is a number that will be determined by
a balance between convection and diffusion. The governing equation
becomes,
d2 h dh
Pr−2a 2 + P r 1+a f ′′ (0)ξ 2 =0
dξ dξ
It is clear from the above that a = −1/3, and the dimensionless variable
ξ has to be defined as ξ = Pr−1/3 η. The governing equation for the
scaled temperature field now becomes,

d2 h ′′ 2 dh
+ f (0)ξ =0
dξ 2 dξ
This equation can be easily solved to obtain,

1 − 0ξ dξ exp (−f ′′ (0)ξ 3 /3)

R
h=
1 − 0∞ dξ exp (−f ′′ (0)ξ 3/3)
R

The flux at the surface is can be determined as,


dT 
q = − k 
dy y=0

1/2 1/3 k(T0 − T∞ ) dh 
= Re Pr (174)
L dξ ξ=0


Thus, the Nusselt number is proportional to Re1/2 Pr1/3 in the limit of

high Prandtl number.

3. The velocity scale, determined from balancing the inertial and buoy-
ancy terms, is (β(T1 − T0 )gh/ρ)1/2 , where the temperature is scaled by
(T1 − T0 ). The pressure scale is (β(T1 − T0 )gh). Scaled this way, the
momentum conservation equation becomes,
∂u∗ ∗ ∗ ∗ ∗ ν
+ u .∇u = −∇ p + q ∇∗2 u∗
∂t∗ h βgρ(T1 − T0 )

42
 In terms of the Gashof number (ν 2 /h3 βρg(T1 − T0 )), this equation can
be expressed as,
∂u∗
+ u∗ .∇u∗ = −∇∗ p∗ + Gr−1/2 ∇∗2 u∗
∂t∗
The temperature equation is,
∂T ∗
+ u.∇T ∗ = α∇2 T ∗
∂t
Introducing the scaled velocity and length, we obtain,
∂T ∗ α
∗
+ u∗ .∇∗ T ∗ = q ∇2 T ∗
∂t h hβg(T1 − T0 )/ρ

This can be expressed in terms of the Grashof number and Prandtl

number as,
∂T ∗ 1
∗
+ u∗ .∇∗ T ∗ = 1/2
∇2 T ∗
∂t PrGr
4. The boundary layer equations are,
∂ux ∂uy ∂uz
+ + =0
∂x ∂y ∂z
∂ux ∂ux ∂ux ∂p ∂ 2 ux
ux + uy + uz =− +ν 2
∂x ∂y ∂z ∂x ∂y
∂p
=0
∂y
∂uz ∂uz ∂uz ∂p ∂ 2 uz
ux + uy + uz =− +ν 2
∂x ∂y ∂z ∂z ∂y
For the outer flow, the pressure gradients are given by,
∂p
= −A2 x
∂x
∂p
= −B 2 z
∂z
From the mass conservation equation, we have,
∂uy
= −Af ′ (η) − Bg ′ (η)
∂y

43
This can be expressed in terms of the similarity variable to obtain
s
A ∂uy
= −Af ′ (η) − Bg ′ (η)
ν ∂η
This can be integrated once, with the boundary condition uy = 0 at
η = 0 to obtain, q
uy = − ν/A(Af (η) + Bg(η)
The above expressions are substituted into the x momentum conserva-
tion equation to obtain,
′
q q
A2 xf 2 − ν/A(Af + Bg)(Ax/ ν/A)f ′′ = A2 x + ν(Ax/(ν/A))f ′′′

Dividing throughout by A2 x, we obtain,

′
f ′′′ + f ′′ (f + (B/A)g) + (1 − f 2 ) = 0

The above expressions can be substituted into the g momentum con-

servation equations to obtain,
′
q q
B 2 zg 2 − ν/A(Af + Bg)(Bz/ ν/A)g ′′ = B 2 z + ν(Bz/(ν/A))g ′′′

Dividing throughout by B 2 z, we obtain,

′
g ′′′ + g ′′ (f + (B/A)g) + (B/A)(1 − g 2 ) = 0

Thus, a similarity solution can be obtained that depends only upon the
ratio (A/B).

44