Detailed solutions

Introductory fluid mechanics: solutions

Simon J.A. Malham

Simon J.A. Malham (22nd October 2014)

Maxwell Institute for Mathematical Sciences
and School of Mathematical and Computer Sciences
Heriot-Watt University, Edinburgh EH14 4AS, UK
Tel.: +44-131-4513200
Fax: +44-131-4513249
E-mail: [email protected]
2 Simon J.A. Malham

Solution (Trajectories and streamlines: expanding jet)

By definition the equations for the particle trajectories are
dx dy dz
= xe2t−z , = ye2t−z , and = 2e2t−z .
dt dt dt
We solve the third equation first. Separating variables we see it is equivalent to
Z z Z t
eζ dζ = 2e2τ dτ
z0 0
⇔ e − ez0 = e
z 2t
−1
z z0
⇔ e =e − 1 + e2t .

We substitute this expression for ez into the second trajectory equation, this gives

dy e2t
=y z
dt e − 1 + e2t
0

y t
e2τ
Z Z
dη
⇔ = dτ
y0 η 0 ez0 − 1 + e2τ
e2t
d(e2τ )
y Z
1
⇔ log =
y0 2 1 ez0 − 1 + e2τ
ez0 − 1 + e2t
y 1
 
⇔ log = log
y0 2 ez0
 ez0 − 1 + e2t 1/2
⇔ y = y0 .
ez0
Similarly we can show that
 ez0 − 1 + e2t 1/2
x = x0 .
ez0
If x0 = 1, y0 = 1 and z0 = 0, then the parametric equations for the particle trajectory
are x = et , y = et and z = 2t.
To find the streamlines, we fix t, and solve the system of equations
dx dy dz
= (e2t ) xe−z , = (e2t ) ye−z , and = (2e2t ) e−z .
ds ds ds
Solving the third equation we get ez = ez0 + 2e2t s. Substituting this into the second
streamline equation implies

dy e2t
= z y
ds e 0 + 2e2t s
Z y Z s
dη 1
⇔ = e2t z0 + 2e2t σ
dσ
y0 η 0 e
ez0 + 2e2t s
y 1
 
⇔ log = log
y0 2 ez0

1/2
⇔ y = y0 1 + 2e2t−z0 s) .

1/2
Similarly we find x = x0 1 + 2e2t−z0 s) .
Introductory fluid mechanics: solutions 3

Solution (Trajectories and streamlines: three dimensions)

The velocity field u(x, t) = (u, v, w)T is given for t > −1 by
x y
u= , v= and w = z.
1+t 1 + 12 t
To find the particle paths or trajectories, we must solve the system of equations
dx dy dz
= u, =v and = w,
dt dt dt
and then eliminate the time variable t between them. Hence for particle paths we have
dx x dy y dz
= , = and = z.
dt 1+t dt 1 + 12 t dt
Using the method of separation of variables and integrating in time from t0 to t, in
each of the three equations, we get
1 + 12 t
         
x 1+t y z
ln = ln , ln = 2 ln and ln = t − t0 ,
x0 1 + t0 y0 1 + 12 t0 z0

where we have assumed that at time t0 the particle is at position (x0 , y0 , z0 )T . Expo-
nentiating the first two equations and solving the last one for t, we get

x 1+t y (1 + 12 t)2
= , = and t = t0 + ln(z/z0 ).
x0 1 + t0 y0 (1 + 21 t0 )2
We can use the last equation to eliminate t so the particle path/trajectory through
(x0 , y0 , z0 )T is the curve in three dimensional space given by

2
1 + 21 t0 + 1


1 + t0 + ln(z/z0 ) 2 ln(z/z0 )
x = x0 · , and y = y0 · 1
.
(1 + t0 ) (1 + 2
2 t0 )
To find the streamlines, we fix time t. We must then solve the system of equations
dx dy dz
= u, =v and = w,
ds ds ds
with t fixed, and then eliminate s between them. Hence for streamlines we have
dx x dy y dz
= , = and = z.
ds 1+t ds 1 + 12 t ds
Assuming that we are interested in the streamline that passes through the point
(x0 , y0 , z0 )T , we again use the method of separation of variables and integrate with
respect to s from s0 to s, for each of the three equations. This gives
     
x s − s0 y s − s0 z
ln = , ln = and ln = s − s0 .
x0 1+t y0 1 + 12 t z0
Using the last equation, we can substitute for s − s0 into the first equations. If we then
multiply the first equation by 1 + t and the second by 1 + 12 t, and use the usual log
law ln ab = b ln a, then exponentiation reveals that
 1+t  1+ 12 t
x y z
= = ,
x0 y0 z0
which are the equations for the streamline through (x0 , y0 , z0 )T .
4 Simon J.A. Malham

Solution (Streamlines: plane/cylindrical polar coordinates)

We are concerned with describing the streamlines associated with the velocity field
(u, v, w)T = α(t) · (x − y, x + y, 0)T , where α = α(t) is an arbitrary function of t.
Using the hint given we change to cylindrical polar coordinates for which x = r cos θ,
y = r sin θ and z = z. Hence if the coordinates x = x(s), y = y(s) and z = z(s)
represent a streamline for a given fixed t for a particle starting at (x0 , y0 , z0 )T , then
in cylindrical coordinates the same streamline is represented by r = r(s), θ = θ(s) and
z = z(s). By direct computation we observe that

dx dr dθ
= cos θ − r sin θ ,
ds ds ds
dy dr dθ
= sin θ + r cos θ .
ds ds ds
Multiply the first equation above by −r sin θ and the second equation by r cos θ. Adding
the resulting two equations reveals that
dθ dx dy
r2 = −y +x .
ds ds ds
Now multiply the first equation above by −r cos θ and the second equation by r sin θ.
Adding these resulting two equations reveals that
dr dx dy
r =x +y .
ds ds ds
The equations for the streamlines are
dx
dy
dz
= α(t) x − y , = α(t) x + y and = 0.
ds ds ds
Hence z(s) = z0 for all s. The equation for r = r(s) is given by

dr dx dy
r =x +y
ds ds ds


= x · α(t) x − y + y · α(t) x + y
= α(t) x2 + y 2

= α(t)r2 .

Hence we see r = r(s) satisfies r0 = α(t)r which implies r(s) = r0 exp α(t)s . Similarly

the equation for θ = θ(s) is

dθ dx dy
r2 = −y +x
ds ds ds


= − y · α(t) x − y + x · α(t) x + y
= α(t) x2 + y 2

= α(t)r2 .

Hence we find θ = θ(s) satisfies θ0 = α(t), i.e. θ(s) = θ0 + α(t)s. Hence for any particle,
the θ = θ(s) coordinate increases linearly in time, while the r = r(s) coordinate
exponentially expands or contracts depending on the sign of α(t) and the z = z(s) is
constant. The streamlines thus look like horizontal exponential spirals.
Introductory fluid mechanics: solutions 5

Solution (Steady oscillating channel flow)

The streamlines for this problem by definition are given by
dx dy
= 1 + x sin y and = cos y.
ds ds
Using the chain rule, in other words using that

dy dy/ds dx dx/ds
= or =
dx dx/ds dy dy/ds

we can solve for the streamlines directly as follows.

dx 1 + x sin y dx 1 sin y
= ⇔ = + x.
dy cos y dy cos y cos y

We observe this is a first order linear differential equation for x = x(y), i.e. we are
treating y as the independent variable and x as the dependent variable and solution. We
can solve first order linear differential equations using the integrating factor technique,
where in this case the integrating factor is
 Z y   
sin η
exp − dη = exp log(cos y) = cos y.
0 cos η

Hence the integrating factor is cos y and we multiply the first order linear differential
equation above through by it. Hence we have
dx
cos y = 1 + (sin y) x
dy
d

⇔ (cos y) x = 1
dy
⇔ (cos y) x = y − y0 + (cos y0 ) x0
y − y0 + (cos y0 ) x0
⇔ x= .
cos y

This gives the equation for the streamline passing through (x0 , y0 )T .
Note that when x = 1 and y = −π/2 as well as when x = −1 and y = π/2, one
can check by direct inspection that dx/ds = 0 and dy/ds = 0 and thus both points are
stagnation points of the flow.
6 Simon J.A. Malham

Solution (Channel shear flow)

As stated, we consider the flow to be two-dimensional with
u = 0,
v = U (1 − x2 /a2 ),
and we ignore the z coordinate as well as the velocity component w which is zero.
In the hint we are told that we should assume the flow is incompressible, which in
two-dimensions corresponds to the condition
∂u ∂v
+ = 0.
∂x ∂y
Hence there exists a stream function ψ = ψ(x, y) such that
∂ψ ∂ψ
u= and v=− .
∂y ∂x
This choice is consistent with the incompressibility condition above. To find the stream
function we set
x2
 
∂ψ ∂ψ
=0 and = −U 1 − 2 .
∂y ∂x a
This is a pair of partial differential equations for the stream function ψ = ψ(x, y).
The first equation implies ψ = f (x) for some arbitrary function f = f (x) (i.e. it is a
function of x only). Substituting this into the second equation we find
x2
 
f 0 (x) = −U 1 −
a2
x3
 
⇔ f (x) = −U x − +C
3a2
x3
 
⇔ ψ(x, y) = −U x − + C,
3a2
where C is an arbitrary constant.
The total volume flux across y = y0 , |x| 6 a, is
+a x=+a
x2 x3
Z    
U 1− dx = U x −
−a a2 3a2 x=−a
4
= U a.
3
The volume flux across y = y0 , |x| 6 43 a, is
+3a/4 x=+3a/4
x2 x3
Z    
U 1− 2 dx = U x − 2
−3a/4 a 3a x=−3a/4
3 9

= 2U a −
4 64
39
= U a.
32
The ratio of the two volume fluxes is
39
32 U a
4
= 0.914 . . . .
3Ua
Introductory fluid mechanics: solutions 7

Solution (Flow inside and around a disc)

We are given a two dimensional flow in plane polar coordinates as follows
a2 a2 Γ
   
ur = U cos θ 1 − and uθ = −U sin θ 1 + − .
r2 r2 2πr
Here U , a and Γ are constants. Our goal is to compute the stream function for this
flow. First we check the incompressibility condition, which in polar coordinates is
1 ∂ 1 ∂uθ
(r ur ) + = 0.
r ∂r r ∂θ
Since by direct computation we find
∂ ∂ a2
 
(r ur ) = U cos θ r−
∂r ∂r r
a2
 
= U cos θ 1 + 2
r
and
∂uθ ∂ a2

 
= −U sin θ · 1 + 2 −0
∂θ ∂θ r
a2
 
= −U cos θ 1 + 2 ,
r
we see that the incompressibility condition is satisfied.
Second, the stream function ψ = ψ(r, θ) is the solution to the pair of partial differ-
ential equations
1 ∂ψ ∂ψ
= ur and − = uθ .
r ∂θ ∂r
Substituting the given forms for ur and uθ , these partial differential equations are
1 ∂ψ a2 ∂ψ a2 Γ
   
= U cos θ 1 − 2 and − = −U sin θ 1 + 2 − .
r ∂θ r ∂r r 2πr
Solving the first partial differential equation we find
a2
 
ψ = U sin θ r − + F (r),
r
where F = F (r) is an arbitrary function of r. Substituting this expression into the
second partial differential equation we find
a2 a2
 
∂ Γ
   
− U sin θ r − + F (r) = −U sin θ 1 + −
∂r r r2 2πr
a2 a2 Γ
   
⇔ −U sin θ 1 + − F 0 (r) = −U sin θ 1 + −
r2 r2 2πr
Γ
⇔ F 0 (r) =
2πr
Γ
⇔ F (r) = log r + C,
2π
where C is an arbitrary constant which we can chose to set to 0. Hence the stream
function is
a2
  Γ
ψ = U sin θ r − + log r.
r 2π
8 Simon J.A. Malham

Solution (Couette flow)

(a) We know that ur = uz = 0, ∂/∂t ≡ 0 (stationary flow), ∂/∂θ ≡ 0 for any of the
velocity components—the component uθ is independent of θ—and fr = fθ = fz = 0.
We are also told ρ ≡ 1. We have to show that for the given velocity field, we can indeed
find a pressure field such that the incompressible Euler equations hold. Substituting
all the above conditions into Euler’s equations in cylindrical polar coordinates implies
u2θ ∂p 1 ∂p ∂p
− =− , 0=− and 0=− .
r ∂r r ∂θ ∂z
Note
1 ∂(rur ) 1 ∂uθ ∂uz
∇·u= + + =0
r ∂r r ∂θ ∂z
trivially, and that u · ∇, which simplifies to (uθ /r)(∂/∂θ), also acts trivially on the
velocity field (which independent of θ). The second and third equations in the system
above imply p = p(r). The first equation implies
 2
∂p 1 A
= + Br
∂r r r
∂p A2 2AB
⇔ = 3 + + B2r
∂r r r
A2 1
⇔ p = − 2 + 2AB log r + B 2 r2 + K,
2r 2
where K is an arbitrary constant. Hence there is a pressure field for which the Euler
equations are satisfied by the given velocity field.
(b) The angular velocity is given in terms of the velocity component uθ = uθ (r) by
uθ /r. To compute the angular velocity at r = R1 and r = R2 we first directly compute
uθ (R1 ) and uθ (R2 ) as follows. First we have, using the formulae for A and B given,
A
uθ (R1 ) = + BR1
R1
R1 R22 (ω2 − ω1 ) R2 ω − R22 ω2
=− 2 2
− 1 21 · R1
R2 − R1 R2 − R12
R1 R22 ω1 − R13 ω1
=
R22 − R12
= R1 ω1 .
Second, similarly we have
A
uθ (R2 ) = + BR2
R2
R12 R2 (ω2 − ω1 ) R2 ω − R22 ω2
=− 2 2
− 1 21 · R2
R2 − R1 R2 − R12
= R2 ω2 .
Hence we deduce that uθ (R1 )/R1 = ω1 and uθ (R2 )/R2 = ω2 .
(c) Directly computing the vorticity we get
1 ∂
 T
ω = ∇ × u = 0, 0, (ruθ ) = (0, 0, 2B)T .
r ∂r
Introductory fluid mechanics: solutions 9

Solution (Venturi tube)

(a) The fluid is incompressible, so the volume flux (= cross-sectional area × uniform
velocity) through wide region equals that in narrow, i.e. we have

A1 u1 = A2 u2 .

Since A2 < A1 this implies

A1
u2 = u1 > u 1 .
A2

(b) Bernoulli’s Theorem implies the quantity H is the same in wide and narrow
regions on same streamline passing through the tube, i.e. we have
p1 p
1 2
2 u1 + = 12 u22 + 2 .
ρ ρ

(We assume the potential difference along any streamline is negligible.)

(c) The Bernoulli result above holds if and only if

p1 − p2 = 12 ρ(u22 − u21 ).

The quantity on the right is positive since u2 > u1 . Hence the pressure in the narrow
region is less!
(d) Examples are: a device for measuring flow speeds (measure p1 − p2 , A1 /A2
known =⇒ u1 , u2 ); a carburetor—the pressure drop draws in another fluid from a
side channel; or the lift of an aircraft wing.
10 Simon J.A. Malham

Solution (Clepsydra or water clock)

(a) Use Bernoulli’s Theorem for a streamline starting at the free surface and going
through the orifice. The potential energy of any fluid particle at the orifice outlet
compared to that at the free surface is −ρg. This we have
 dz 2 P0 P
1
2 + = 12 U 2 + 0 − ρg
dt ρ ρ
 dz 2
⇔ 1
2 = 12 U 2 − ρg.
dt

(Note that inside the container at the level at which water is pushed through the
orifice, the pressure is P = P0 + ρgz. The difference P − P0 accelerates the water
through the orifice.)
(b) Using incompressibility, we equate the volume flux through the orifice to the
rate of volume drop at the free surface (as it descends). We know S is the cross-sectional
area of the orifice and U is the velocity of the fluid through the orifice. The volume flux
through the orifice is equal to the cross section area S times the fluid velocity and is
thus SU . Similarly the cross-sectional area of the free surface at height z is say A, while
its velocity is dz/dt. Hence the drop in volume per unit time at the free surface—i.e.
the volume flux at the free surface—is A dz/dt. Hence incompressibility implies
dz
SU = A .
dt

(c) Rearranging the result from part (b) we see that

1 dz S
= .
U dt A
Thus using the result from part (a) we find that
  dz 2 
gz = 1
2 U2 −
dt
  1 dz 2 
1 2
= 2U 1−
U dt
  S 2 
1 2
= 2U 1− .
A
We assume S  A so that S/A  1. Hence with an error of order (S/A)2 we have
p
U∼ 2gz.

(d) We want dz/dt to be constant, so combining the results of parts (b) and (c),
and using that A = πr2 as the Clepsydra is a volume of revolution, implies
dz
2gz = πr2
p
S
dt
 π dz 2
⇔ z= √ r4 ,
2g S dt
as required.
Introductory fluid mechanics: solutions 11

Solution (Coffee mug)

(a) We are given ur = uz = 0, ∂/∂t ≡ 0, ∂/∂θ ≡ 0 and fr = fθ = 0 with fz = −g.
Substituting these into Euler’s equations in cylindrical polar coordinates implies

u2θ 1 ∂p
− =− ,
r ρ ∂r
1 ∂p
0=− ,
ρr ∂θ
1 ∂p
0=− − g.
ρ ∂z
Note we have
1 ∂(rur ) 1 ∂uθ ∂uz
∇·u= + + =0
r ∂r r ∂θ ∂z
trivially. And u · ∇ = (uθ /r)(∂/∂θ) thus also acts trivially on the velocity field which
is independent of θ. The second equation in the system implies p = p(r, z).
(b) Using that uθ = Ωr—this is the rigid body rotation ansatz—implies

∂p
= ρΩ 2 r ⇔ p(r, z) = 12 ρΩ 2 r2 + G(z),
∂r
where G an arbitrary function. Substitute this into the other equation
1 ∂p
= −g ⇔ G0 (z) = −ρg.
ρ ∂z

Hence G(z) = −ρgz + C where C is an arbitrary constant. This implies we have

p(r, z) = 12 ρΩ 2 r2 − ρgz + C.

At the free surface p = P0 :

P0 = 12 ρΩ 2 r2 − ρgz + C
⇔ z = (Ω 2 /2g) r2 − (C − P0 )/ρg.

(c) Choose C = P0 ⇔ z = (Ω 2 /2g) r2 . Relative to the chosen origin O, the bottom

of the mug is at z = −z0 .
(i) Initially the coffee depth is d. This implies the total volume is πa2 d. The incom-
pressibility volume constraint implies
a
Ω 2 r2
Z
2
πa z0 + · 2πr dr = πa2 d.
0 2g
2
The term πa z0 represents the cylindrical volume of coffee lying under the origin O.
The integral term represents the volume of coffee above the origin O but under the
free surface.
(ii) The constraint in part (i) is equivalent to the condition

Ω 2 a2
z0 + = d.
4g
12 Simon J.A. Malham

The coffee at the edge of the mug is at height

Ω 2 a2 Ω 2 a2 Ω 2 a2 Ω 2 a2
z = z0 + =d− + =d+ .
2g 4g 2g 4g
Spillage occurs when this exceeds h, i.e. when

4g(h − d)
z>h ⇔ Ω2 > .
a2

If the mug is initially less than half full then d < 12 h. Spillage occurs when

4g(h − d) Ω 2 a2 Ω 2 a2
Ω2 > ⇔ > (h − d) ⇔ − 6 −(h − d).
a2 4g 4g
Thus we observe that
Ω 2 a2
z0 = d − 6 d − (h − d) = 2d − h < 0.
4g

In other words the centre part of the free surface (the dip in the middle) will have hit
the bottom of the mug before spillage.
Introductory fluid mechanics: solutions 13

Solution (Channel flow: Froude number)

Recall the general theory from the notes. We saw there that in the case when the
actual maximum value of the river bed undulation ymax does not reach the maximum
possible bound y0 , ymax < y0 , that when the Froude number F was such that
d
F<1 then (h + y) < 0,
dy
while when
d
F>1 then (h + y) > 0.
dy
In the case when ymax = y0 , these last two properties still hold upstream until y
reaches it maximum ymax = y0 —all the arguments in the text for the case ymax < y0
apply until y actually reaches its maximum. Recall the upper bound y0 occurred when
h = h0 where
(U H)2/3
h0 := .
g 1/3
Recall that due to incompressibility U H = uh and in particular at the maximum
y = y0 we have U H = uh0 . Hence we see that
(uh0 )2/3 (uh0 )2 u2 u2
h0 = ⇔ h30 = ⇔ h0 = ⇔ =1
g 1/3 g g gh0
at y = y0 . Since we have (see the general theory in the notes)
−1
u2

dh
=− 1− ,
dy gh
we see that as y → y0 then dh/dy → ∞.
We know dy/dx = 0 at ymax = y0 so what happens to the free surface at this
point, and in particular, what is dh/dx? To answer this we need to expand locally and
look at higher order terms. Suppose that x = 0 at the maximum value for y = y(x) so
that y(0) = ymax = y0 . We expand y = y(x) and h = h(x) about x = 0 as follows:
y = y0 − Kx2 + · · · ,
h = h0 1 + a1 x + a2 x2 + · · ·

where we will ignore terms of degree x3 and higher. The quantities K, a1 and a2 are all
constants. Naturally since x = 0 is a local maximum for y = y(x) we know K > 0. We
substitute these expansions into the Bernoulli constraint (again see the general theory
in the notes)
U2 (U H)2
y= +H −h− .
2g 2gh2
−2
Note to expand h we use the Binomial expansion with X := a1 x + a2 x2 + · · · so

−2
h−2 = h−2
0 1 + a1 x + a2 x2 + · · ·

−2
= h−2
0 1+X
 
(−2)(−3) 2
= h−2
0 1 + (−2X) + X + ···
2!
= h−2 1 − 2(a1 x + a2 x2 + · · · ) + 3(a1 x + a2 x2 + · · · )2 + · · ·


0
= h−2 1 − 2a1 x + (3a21 − 2a2 )x2 + · · · .


0
14 Simon J.A. Malham

F<1 sub a 1 >0 F>1 sub a 1 >0

sub a 1<0

super a 1 >0 super super a 1 <0

a 1<0

y0 y0

x=0 x x=0 x

Fig. 1 The cases for the two incident (upstream) Froude numbers are shown. The flow is
subcritical if a1 x > 0 and supercritical if a1 x < 0. For a real flow friction will force the
supercritical solution upstream in each case. Thus for example, in the case F > 1 upstream,
there will be a discontinuity in the gradient of h = h(x) at x = 0.

Substituting this into the Bernoulli constraint we find

U2
(U H)2
y0 −Kx2 +· · · = +H−h0 1+a1 x+a2 x2 +· · · − 1−2a1 x+(3a21 −2a2 )x2 +· · · .


2g 2
2gh0

Thus equation coefficients of powers of x we have

U2 (U H)2
x0 : y0 = + H − h0 − ,
2g 2gh20
(U H)2
x1 : 0 = −h0 a1 − (−2a1 ),
2gh20
(U H)2
x2 : − K = −h0 a2 − (3a21 − 2a2 ).
2gh20

Note that the relation between the coefficients at x0 simply reproduces the definition
for the maximum value y0 . Then for the relation between the coefficients at x1 , since
U H = uh0 and u2 = gh0 we see that the expression on the right is

(U H)2 u2 h20 u2
     
−a1 h0 − = −a1 h0 − = −a1 h0 − = 0,
gh20 gh20 g

generating no new information. The relation between the coefficients at x0 implies

u2 h20
K = h0 a2 + (3a21 − 2a2 )
2gh20
h
⇔ K = h0 a2 + 0 (3a21 − 2a2 )
2
 2K 1/2
⇔ a1 = ± .
3h0
Around x = 0 where y = y0 = ymax , using the Binomial expansion we see that

u2 (U H)2 (U H)2
= 3
= (1 − 3a1 + · · · ).
gh gh gh30

Hence locally the flow is subcritical so F < 1 if a1 x > 0, and supercritical so F > 1 if
a1 x < 0; see Fig. 1. Friction forces the supercritical solution downstream in each case.
Introductory fluid mechanics: solutions 15

Solution (Bernoulli’s Theorem for irrotational unsteady flow)

(a) Since the flow is incompressible and homogeneous, ρ is uniform and constant.
The flow is also irrotational so that u = ∇ϕ. The Euler equations imply
∂ p
 
(∇ϕ) + u · ∇u = −∇ − ∇Φ.
∂t ρ
Using the given identity we get
 ∂ϕ  p
+ 12 ∇ |u|2 − u × (∇ × u) = − ∇


∇ − ∇Φ
∂t ρ
 ∂ϕ p

⇔ ∇ + 12 |u|2 + + Φ = u × (∇ × u)
∂t ρ
⇔ ∇H = u × 0
⇔ ∇H = 0

for the quantity H given in the question.

(b) Hence H = f (t) for some function f = f (t). Using the suggested redefined
potential for u given by
Z t Z t
V =ϕ− f (τ ) dτ ⇔ ϕ=V + f (τ ) dτ
0 0

and substituting this into

∂H
= f 0 (t)
∂t
we get
∂ ∂ϕ p
 
+ 12 |u|2 + + Φ = f 0 (t)
∂t ∂t ρ
∂ ∂V p
 
⇔ + f (t) + 12 |u|2 + + Φ = f 0 (t)
∂t ∂t ρ
∂ ∂V p
 
1 2
⇔ + 2 |u| + + Φ = 0
∂t ∂t ρ
which gives the result.
16 Simon J.A. Malham

Solution (rigid body rotation)

Need to show that the rigid body rotation ansatz (u, v)T = (−Ωy, Ωx)T , where
Ω = Ω(t) represents an angular velocity, is a solution to the incompressible Euler
equations when a two-dimensional body force given by f = (αx + βy, γx + δy)T is
applied. We assume that the third velocity component w = 0. We first note that the
incompressibility condition ∇ · u = 0 is trivially satisfied by the rigid body rotation
ansatz. Second we note that
 ∂ ∂
 u 
u · ∇u = u +v
∂x ∂y v
 ∂ ∂
 −Ωy 
= −Ωy + Ωx
∂x ∂y Ωx
2
 
−Ω x
= .
−Ω 2 y

Hence if we substitute the rigid body rotation ansatz (u, v)T = (−Ωy, Ωx)T , into the
Euler equations we find
−Ω 2 x
       
−Ω̇y 1 ∂p/∂x αx + βy
+ =− + .
Ω̇x −Ω 2 y ρ ∂p/∂y γx + δy

Consider the curl of these equations—the curl of a vector u = (u, v)T which only has
two components is the scalar −∂u/∂y + ∂v/∂x. By direct computation this gives
1
Ω̇ = (γ − β).
2
Now let us separate the equations above, i.e. those resulting from substituting the
rigid body rotation ansatz into the Euler equations. We consider them as two partial
differential equations for the pressure p as follows
1 ∂p
= Ω̇y + Ω 2 x + αx + βy,
ρ ∂x
1 ∂p
= −Ω̇x + Ω 2 y + γx + δy.
ρ ∂y
Recall to show that the rigid body rotation ansatz is a solution to the incompressible
Euler equations we must find a pressure field p such that (u, v)T = (−Ωy, Ωx)T and p
satisfy them. We integrate the first equation above with respect to x. This generates
the relation
p 1 1
= Ω̇yx + Ω 2 x2 + αx2 + βxy + F (y),
ρ 2 2
where F = F (y) is an arbitrary function. If we now substitute this expression into the
second equation above and use that Ω̇ = 12 (γ − β) we find

F 0 (y) = Ω 2 y + δy.

Hence we deduce F (y) = 12 (Ω 2 + δ)y 2 + C, where C is an arbitrary constant. Thus a

solution to the two partial differential equations is
p 1 1 1
= (Ω 2 + α)x2 + (Ω 2 + δ)y 2 + (β + γ)xy + C.
ρ 2 2 2
Hence there exists a pressure p such that the rigid body rotation ansatz represents a
solution to the incompressible Euler equations.
Introductory fluid mechanics: solutions 17

Solution (Vorticity and streamlines)

Starting with the steady incompressible two-dimensional Navier–Stokes equations
for the velocity field u = (u, v)T our goal is to show they are equivalent to the system
of equations  
v
∇H = ω
−u

where ω is the scalar vorticity and H := p/ρ + 12 (u2 + v 2 ). Using the identity

1
∇ |u|2 = u · ∇u + u × (∇ × u),


2
we can write the steady Euler equations as (ρ is constant and we assume no body force)
1 p
 
∇ |u|2 − u × (∇ × u) = −∇


.
2 ρ

In our two-dimensional context |u|2 = u2 + v 2 , while by direct computation

   
u 0
u × (∇ × u) = v  ×  0 
0 ω
 
vω
= −uω  .
0

Substituting these expressions back into the equation above we find (ignoring the last
component)
1
 p   vω 
∇ u2 + v 2 + ∇


= ,
2 ρ −uω
thus demonstrating the equivalence.
To show the scalar vorticity ω is constant along streamlines we observe that
∂ ∂
∇ × ∇H ≡ 0 ⇔ (vω) − (−uω) ≡ 0
∂y ∂x
 ∂v ∂u  ∂ω ∂ω
⇔ + ω+u +v ≡ 0.
∂y ∂x ∂x ∂y
Incompressibility implies the first term on the left is zero. Hence the relation above is
equivalent to u · ∇ω ≡ 0 and thus ω is constant along streamlines.
18 Simon J.A. Malham

Solution (Poiseuille flow)

(a) Use the incompressible Navier–Stokes equations in cylindrical polar coordinates.
Assume that ur = 0, uθ = 0 and uz = uz (r) and ignore any possible body forces
fr = fθ = fz = 0. Note we have implicitly assumed the flow is steady. Only the third
equation for uz = uz (r) generates a non-trivial condition—the first two equations tell
us that the pressure is independent of r and θ. The third equation is
∂uz 1 ∂p 1 ∂ ∂uz
 
+ u · ∇uz = +ν r + fz .
∂t ρ ∂z r ∂r ∂r
Since the flow is steady ∂uz /∂t = 0, and as ur = uθ = 0 and uz = uz (r) only then we
see that u · ∇uz = 0. Further we assume there are no body forces, including gravity,
so fz = 0. Since we assume p = −Cz for some constant C, this equation becomes
ρν ∂ ∂uz
 
C=− r .
r ∂r ∂r

(b) We integrate the equation above with respect to r twice to find uz = uz (r) as
follows. Rearranging the equation and then integrating twice we find
∂ ∂uz C
 
r =− r
∂r ∂r ρν
∂uz C 2
⇔ r =− r +A
∂r 2ρν
∂uz C A
⇔ =− r+
∂r 2ρν r
C 2
⇔ uz = − r + A log r + B
4ρν
where A and B are arbitrary constants. We want the solution to be bounded for
0 6 r 6 a and therefore we must insist A = 0 because log r → −∞ as r → 0. The
no-slip boundary condition at the pipe wall r = a implies
C 2 Ca2
0=− a +B ⇔ B= .
4ρν 4ρν
Hence we see that
C 2 Ca2 C
uz = − r + = (a2 − r2 ).
4ρν 4ρν 4ρν

(c) Let S represent a disc cross-section of the pipe. Let dS represent a small patch
of area on S. Since ρ is the mass per unit volume and uz dS represents the volume of
fluid passing through dS per unit time, the mass flow rate across dS is ρuz dS. The
total mass flow rate across S is thus
Z Z a Z 2π
ρuz dS = ρuz (r) rdθdr
S 0
Z0 a
C
= 2πρ (a2 − r2 ) rdr
0 4ρν
πC 1 2 2 1 4 r=a
h i
= a r − r
2ν 2 4 r=0
πCa4
= .
8ν
Introductory fluid mechanics: solutions 19

Solution (Elliptical pipe flow)

(a) Using the Navier–Stokes equations in three dimensional Cartesian coordinates,
given u = 0, v = 0 and w = w(x, y) only, and assuming no body force, we are left with

∂p
0=− ,
∂x
∂p
0=− ,
∂y
∂p ∂2w ∂2w
 
0=− +ν 2
+ 2 ,
∂z ∂ x ∂ y

where note that u · ∇w = 0 as w is independent of z. Since we are given p = −Gz,

the first two equations are consistent and substituting this form for p into the final
equation gives the required result

∂2w ∂2w G
2
+ 2 =− .
∂ x ∂ y ν

(b) Substituting w = Ax2 + By 2 + C into the partial differential equation above,

we immediately deduce
G
A+B =− .
2ν

(c) Using the no-slip boundary condition, i.e. that w = 0 on

x2 y2 x2
 
2
+ 2 =1 ⇔ y 2 = b2 1 − ,
a b a2
generates the equation

x2 b2 2
   
Ax2 + Bb2 1 − +C =0 ⇔ A−B x + Bb2 + C = 0
a2 a2

which must hold for all x ∈ [−a, a]. Hence equating coefficients of x0 and x2 , we arrive
at the following three equations (including the result from part (b) above) for the three
unknowns A, B and C:

b2 G
A−B = 0, Bb2 + C = 0 and A+B =− .
a2 2ν
Solving the first equation for A in terms of B and substituting this into the third
equation we find

b2 G G a2
 
B 1+ =− ⇔ B=− · 2 .
a2 2ν 2ν a + b2

Substituting this into the first equation reveals

G a2 b2 G b2
A+ · 2 2
· 2 =0 ⇔ A=− · 2 .
2ν a + b a 2ν a + b2
20 Simon J.A. Malham

Finally substituting these two answers for A and B into the second equation implies

G a2 G a2 b2
− · 2 · b2 + C = 0 ⇔ C= · 2 .
2ν a + b2 2ν a + b2

(d) For a small patch of area dS of an elliptical cross section of the pipe, the volume
of fluid passing through dS per unit time is equivalent to the volume of the cylinder of
cross sectional area dS and length w (the orthogonal flow rate through dS), i.e. w · dS.
Summing over all such small patches of areas to make up the complete cross section
generates the integral Z
w dS

over the whole elliptical cross sectional area.

Computing the integral using the substitutions x = ar cos θ and y = br sin θ we see:
Z Z
w dS = Ax2 + By 2 + C dxdy
Z 2π Z 1
Aa2 r2 cos2 θ + Bb2 r2 sin2 θ + C abr dr dθ


=
0 0
2π 1
a3 b3
Z Z
G
= · 2 (1 − r2 ) r dr dθ
2ν a + b2 0 0
1
2πGa3 b3
Z
= r − r3 dr
2ν(a2 + b2 ) 0
πa3 b3 G
= .
4ν(a2 + b2 )

(e) Denote the flow rate from part (d) above by

πa3 b3 G
Q := .
4ν(a2 + b2 )
We wish to find the maximum of Q subject to the constraint that the cross sectional
area of the elliptic pipe is given, say by K, i.e. πab = K with K fixed. Simply substitute
that b = K/πa into Q and differentiate with respect to a implies

∂Q ∂ π(K/π)3 G
 
=
∂a ∂a 4ν(a + K 2 π −2 a−2 )
2

π(K/π)3 G (2a − 2K 2 π −2 a−3 )

=− · 2 .
4ν (a + K 2 π −2 a−2 )2
Thus ∂Q/∂a is zero and Q maximized if and only if

2a − 2K 2 π −2 a−3 = 0
p
⇔ a= K/π.
p
Note that when a = K/π, then

b = K/πa = K/(π · K 1/2 /π 1/2 ) = K 1/2 /π 1/2 = a.

Hence the flow rate Q is maximized when a = b.

Introductory fluid mechanics: solutions 21

Solution (Stokes flow: between hinged plates)

For the first part of the problem we can follow the Stokes flow section in the notes
together with the viscous corner flow example given in that section. Since we assume
we have an incompressible Stokes flow and it is stationary, we have

∇p = µ∆u and ∇·u=0

where p is the pressure field, u the velocity field and µ is the viscosity. Since the flow
is uniform in the direction aligned with the imaginary line of intersection between the
plates—denote this the z-axis—there exists a stream function ψ = ψ(r, θ) such that in
cylindrical polar coordinates
∆(∆ψ) = 0.
To deduce this we simply follow the argument in the viscous corner flow example in
the Stokes flow section. Note the velocity components ur and uθ and stream function
are related by
1 ∂ψ ∂ψ
ur = and uθ = − .
r ∂θ ∂r
For the second part of the problem let us determine the appropriate boundary condi-
tions. We will assume a lower plate that is stationary and Ω is the angular velocity of
the upper plate relative to the lower plate. Using no-slip boundary conditions on the
plate along θ = 0 which is stationary we see that

1 ∂ψ ∂ψ
=0 and =0 on θ = 0.
r ∂θ ∂r
Also using the no-slip boundary conditions on the moving plate which at time t lies
along θ = β(t) > 0, we have

1 ∂ψ ∂ψ
=0 and = −Ωr on θ = β.
r ∂θ ∂r
Hence our goal now is to solve the biharmonic equation ∆(∆ψ) = 0 subject to the four
boundary conditions above. We thus look for a solution of the form

ψ = Ωr2 f (θ),

for some arbitrary function f = f (θ). This solution ansatz is suggested by the boundary
condition on the moving upper plate. We solve the biharmonic equation here in two
stages. First we observe that
 ∂2 1 ∂ 1 ∂2

∆ Ωr2 f (θ) = Ω r2 f (θ)



+ +
∂r2 r ∂r r2 ∂θ2
= Ω 4f (θ) + f 00 (θ) .

If we set F := f 00 + 4f then, second, we observe that



∆ ∆ Ωr2 f (θ)


= ∆ ΩF (θ)
 ∂2 1 ∂ 1 ∂2


=Ω + + 2 2 F (θ)
∂r2 r ∂r r ∂θ
= ΩF 00 (θ).
22 Simon J.A. Malham

Hence our solution guess ψ = Ωr2 f (θ) satisfies the biharmonic equation if and only if
F 00 (θ) = 0. Hence we find F = 4A + 4Bθ for some constants A and B—the prefactor
‘4’ is chosen for convenience. Hence we find, combining the complementary function
and particular integral for f 00 + 4f = F , we have

f = A + Bθ + C cos(2θ) + D sin(2θ),

where C and D are two further constants. For our solution ansatz ψ = Ωr2 f (θ) the
four boundary conditions become
1
f (0) = f 0 (0) = f 0 (β) = 0 and f (β) = − .
2
The four constants are determined by the four boundary conditions above, and indeed
we find
sin θ cos(β − θ) − θ cos β
f (θ) = .
2(β − tan β) cos β
We finally focus on the issue of the distance within O within which this solution is
self-consistent. Our solution ansatz was inferred from the boundary condition uθ = Ωr
on the upper plate. Hence the inertia terms u · ∇u scale like Ω 2 r2 /r while the viscous
terms ν∆u scale like ν(Ωr)/r2 . Our Stokes flow assumption was

Ω 2 r2 /r Ωr2
1 ⇔  1.
ν(Ωr)/r2 ν

Hence a necessary condition for the solution to be valid is r  (ν/Ω)1/2 . (What

happens when tan β = β?)
Introductory fluid mechanics: solutions 23

Solution (Stokes flow: rotating sphere)

We consider a Stokes flow around a rotating sphere, of radius a and angular velocity
Ω. Our goal is to compute the couple exerted on the fluid due to the sphere; the fluid
is at rest at infinity. Using the hint provided we use spherical polar coordinates (r, θ, ϕ)
and assume that the axis to which the latitude angle θ is measured is aligned with, and
in the same direction as, the angular velocity vector Ω. Given the geometry of the set
up, we assume a stationary flow and that ur = 0 and uθ = 0 and uϕ = uϕ (r, θ) only.
Hence we have  
0
u=0
uϕ
and uϕ → 0 as r → ∞. On the surface of the sphere r = a the velocity is uϕ = Ωa sin θ
where Ω = |Ω| the magnitude of the angular velocity Ω. This is because any point
on the surface of the sphere at latitude angle θ is a distance a sin θ from the axis of
rotation of the sphere. The form of the velocity field uϕ = uϕ (r, θ) at the sphere surface
r = a suggests we look for a solution of the form

uϕ = Ωaf (r) sin θ.

The steady Stokes flow equations involving the velocity field u and pressure p are

µ∆u = ∇p,

where µ is the first coefficient of viscosity. Since there is only one non-zero component
of velocity uϕ , the Stokes flow equations in spherical polar coordinates reduce to
1 ∂p
µ∆uϕ = .
r sin θ ∂ϕ
Since uϕ = uϕ (r, θ) only, we deduce that p = p(r, θ) can only be linear in ϕ. However
p must be periodic in ϕ hence p must be constant, i.e. invariant to ϕ. Hence we have
∂p
= 0.
∂ϕ
Hence the Stokes flow equations reduce to ∆uϕ = 0 or ∆u = 0. We employ the strategy
suggested in the notes for such Stokes flow equations and write

∆u = 0
 
0
⇔ ∆ 0  = 0
uϕ
 
0
⇔ ∇×∇× 0 =0
uϕ
 
0
⇔  0
= 0,

1 2
− r sin θ D (r sin θ) u ϕ

where
∂2 sin θ ∂ 1 ∂
 
D2 := + 2 .
∂r2 r ∂θ sin θ ∂θ
24 Simon J.A. Malham

Substituting the solution ansatz uϕ = Ωaf (r) sin θ into the equation above we find
1
D2 (r sin θ) uϕ = 0


−
 r sin θ


D2 r sin θ Ωaf (r) sin θ = 0


⇔

D2 rf (r) sin2 θ = 0


⇔
∂2 f (r) ∂ 1 ∂
 
sin2 θ (sin2 θ) = 0


⇔ 2
rf (r) + sin θ
∂r r ∂θ sin θ ∂θ
∂ f (r) ∂
sin2 θ f (r) + rf 0 (r) + sin θ


⇔ (2 cos θ) = 0
∂r r ∂θ
f (r)
sin2 θ 2f 0 (r) + rf 00 (r) − 2 sin2 θ


⇔ =0
r
f (r)
⇔ rf 00 (r) + 2f 0 (r) − 2 =0
r
⇔ r2 f 00 (r) + 2rf 0 (r) − 2f (r) = 0.

This is a linear second order Cauchy–Euler type of equation. If we set ρ := log r and
suppose F (ρ) = f (r) then we see using the chain rule that
dF dρ dF 1
f 0 (r) = = = F 0 (ρ)
dr dr dρ r
and
d 1 0 1 1 dρ d 1 1
 
f 00 (r) = F (ρ) = − 2 F 0 (ρ) + F 0 (ρ) = − 2 F 0 (ρ) + 2 F 00 (ρ).


dr r r r dr dρ r r
Substituting these expressions into the Cauchy–Euler equation above we find

F 00 + F 0 − 2F = 0,

for F = F (ρ). This is a second order constant coefficient differential equation for F
whose independent solutions are exp(ρ) = r and exp(−2ρ) = r12 . Hence the general
solution to the Cauchy–Euler equation for f = f (r) is
B
f = Ar + ,
r2
for arbitrary constants A and B. Using the boundary condition that uϕ → 0 as r → ∞
implies A = 0. Then on the boundary r = a of the sphere we must have

uϕ (a, θ) = Ωa sin θ
⇔ Ωaf (a) sin θ = Ωa sin θ
⇔ f (a) = 1
⇔ B = a2 .

Hence the solution to the Stokes flow equations and boundary conditions is

Ωa3 sin θ
uϕ = .
r2
Introductory fluid mechanics: solutions 25

Having computed the velocity field, we can now compute the coupling exerted
on the fluid due to the sphere. Recall that for an incompressible Newtonian fluid
the deviatoric stress σ̂ is related to the deformation matrix D through the formula
σ̂ = 2µD. The components of the deformation matrix D are given the appendix of the
notes. Since the only non-zero component of the velocity field is uϕ = Ωa3 sin θ/r2 the
only non-zero component of the deformation matrix D is
  −3Ωa3 sin θ
r ∂ uϕ
Drϕ = = .
2 ∂r r 2r3
The total couple exerted on the fluid at the sphere boundary is
Z π
a sin θR̂ × (σ̂n) 2πa2 sin θ dθ.
0

By analogy with the way we computed the force on a sphere in a uniform Stokes flow,
we split the sphere here into concentric rings about the axis of rotation and compute
the coupling at the sphere surface r = a. Hence in the above integral 2πa2 sin θ dθ
represents the area of the concentric ring at angle θ to the axis of rotation. Note we
have n = −r̂ and so σ̂n = −σ̂r̂ represents the stress at the angle θ on the surface
r = a of the sphere. The vector R̂ is a vector orthogonal to the axis of rotation and
a sin θ represents the distance from that axis of rotation to a point on the sphere. Thus
the coupling on the sphere at angle θ is −a sin θR̂ × (σ̂r̂). Now we compute the stress:

σ̂r̂ = 2µDr̂
  
Drr Drθ Drϕ 1
= 2µ  Drθ Dθθ Dθϕ  0
Drϕ Dθϕ Dϕϕ 0
 
0
= 2µ  0 
Drϕ
= 2µDrϕ ϕ̂.

Hence the coupling is −2µa sin θDrϕ R̂ × ϕ̂. We observe that R̂ × ϕ̂ = Ω̂ as it should
be and −2µa sin θDrϕ is the component we are after. We can now compute the integral
to obtain the total coupling
π π
−3Ωa3 sin θ
Z Z 
(−2µa sin θDrϕ )(2πa2 sin θ) dθ = −4µπa3 sin2 θ dθ
0 2a3
Z0 π
3
= 6µπΩa sin3 θ dθ
0
Z π Z π 
3
= 6µπΩa sin θ dθ − cos2 θ sin θ dθ
0 0
 Z π 
3 2
= 6µπΩa 2+ cos θ d(cos θ)
0
2

= 6µπΩa3 2 −
3
= 8µπΩa3 .
26 Simon J.A. Malham

Solution (Lubrication theory: shear stress)

We assume lubrication theory holds for this scenario. Hence the flow in the narrow
layer is governed by the shallow layer equations. We also assume the flow is uniform
in one of the horizontal directions (the y-direction from the set up given) and so the
only relevant shallow layer equation is

∂p ∂2u
=µ 2,
∂x ∂z
where z represents the vertical coordinate, u = u(x, z) is the velocity in the x-direction
and µ is the viscosity. Integrating this equation with respect to z we find
1 ∂p
u=− z(h − z).
2µ ∂x
The volume flux through the layer is constant say Q, i.e. we have
Z h
u dz = Q.
0

Computing the integral on the left we find

1 ∂p h3 ∂p 12µQ
− =Q ⇔ =− 3 ,
2µ ∂x 6 ∂x h
as we required—so the constant A ≡ −12µQ.
Now suppose h = Ce−Bx ; this implies
∂p 12µQ 3Bx 4µQ 3Bx

=− e ⇔ p(x) = p(0) − e −1 .
∂x C3 BC 3

In particular we see that

4µQ 4µQ
1 − e3BL 1 − e3BL .



p(L) − p(0) = ⇔ −p̂ =
BC 3 BC 3
Hence we see that
∂p 3B p̂ e3Bx
= .
∂x 1 − e3BL
The shear stress on z = 0 is

∂u  1 ∂p −3B p̂
µ =− h=
Ce2Bx .
∂z z=0 2 ∂x 2 1 − e3BL

The maximum shear stress is at x = L and has the value

3B p̂

Ce2BL .
2 e3BL − 1
Introductory fluid mechanics: solutions 27

Solution (Lubrication theory: shallow layer)

(a) This whole part follows directly from the derivation of the shallow layer equa-
tions in the lubrication theory section of the notes once the replacement p = P − ρgz
is made, where P is the modified pressure. The explanation is as follows. With this
replacement since
∂p ∂P ∂p ∂P
= , and = ,
∂x ∂x ∂y ∂y
the first and second component Navier–Stokes equations are unchanged—the body
force −ρgẑ does not act on either of these components anyway. The body force −ρgẑ
does act on the third component Navier–Stokes equation and the pressure term and
body force term in combination are
1 ∂p 1 ∂
− − ρg = − (P − ρgz) − ρg
ρ ∂z ρ ∂z
1 ∂P
=− .
ρ ∂z
Hence making the replacement for the pressure in terms of the modified pressure in
the Navier–Stokes equations with a body force −ρgẑ, we obtain the Navier–Stokes
equations in terms of the modified pressure P without any body force. We can thus
proceed with the derivation of the shallow layer equations in the lubrication theory
section of the notes.
(b) Note from the shallow layer equations that the third equation implies
∂P ∂p
=0 ⇔ = −ρg
∂z ∂z
⇔ p = ρg(h − z)
⇔ P = ρgh.

where we assume that p = 0 at z = h(x, y, t) (the pressure is constant on the free

surface and we take it to be zero there). Substituting this expression for the pressure
into the first two shallow layer equations implies

∂2u ∂h
µ = ρg
∂z 2 ∂x
∂u ∂h ∂Γ
⇔ µ = ρg (z − h) +
∂z ∂x ∂x
1 ∂h ∂Γ
⇔ µu = 2 ρg z(z − 2h) + z ,
∂x ∂x
and similarly
∂h ∂Γ
µv = 12 ρg z(z − 2h) + z ,
∂y ∂y
where in the first integration from z to h, we used the given conditions for the applied
surfaces stresses at z = h(x, y, t) and in the second integration from 0 to z, we used
the no-slip boundary conditions at z = 0.
Let us now restrict ourselves to the surface z = h(x, y, t). The acceleration of any
particle at the surface is given by
∂h ∂h ∂h
+ u(x, y, h, t) + v(x, y, h, t) .
∂t ∂x ∂y
28 Simon J.A. Malham

From the incompressibility condition we have

Z h
∂h ∂h ∂h ∂u ∂v

+ u(x, y, h, t) + v(x, y, h, t) = − + dz
∂t ∂x ∂y 0 ∂x ∂y
Z h Z h
∂ ∂
= − u dz − v dz
∂x 0 ∂y 0
∂h ∂h
+ u(x, y, h, t) + v(x, y, h, t) .
∂x ∂y
Cancelling like terms on each side and subsituting our expressions for u and v above
into this last expression reveals that
∂h ∂ 3 ∂h ∂Γ ∂ 3 ∂h ∂Γ
   
2µ = 2
3 ρgh ∂x − h2 + 2
3 ρgh ∂y − h2 .
∂t ∂x ∂x ∂y ∂y
Simplifying the terms under the partial derivatives on the right-hand side reveals
   
∂h ∂ ∂ ∂ ∂
h2 Γ − 13 ρgh2 h2 Γ − 13 ρgh2



2µ + + = 0.
∂t ∂x ∂x ∂y ∂y
Introductory fluid mechanics: solutions 29

Solution (Lubrication theory: Hele–Shaw cell)

Part (a) is straightforward and part (b) follows from the standard theory underlying
the shallow layer equations in the notes.
(c) We are asked to show that the vertically averaged velocity field u = (u, v)T is
incompressible and irrotational. Recall that by definition
Z h

T 1
T
u(x, y), v(x, y) := u(x, y), v(x, y) dz.
h 0

By direct computation we find (note that h is constant)

 Z h   Z h 
∂ 1 ∂ 1
∇·u= u dz + v dz
∂x h 0 ∂y h 0
Z h
1 ∂u ∂v
= + dz
h 0 ∂x ∂y
Z h
1 ∂w

= − dz
h 0 ∂z
h w iz=h
= −
h z=0
= 0,

since w = 0 on z = 0 and z = h.
Integrating the shallow layer equations with respect to z twice we find
1 ∂p 1 ∂p
   
u=− z(h − z) and v=− z(h − z),
2µ ∂x 2µ ∂y
using the no-slip boundary conditions at z = 0 and z = h. Hence the scalar quantity
representing the curl of u = (u, v)T is given by
 Z h   Z h 
∂ 1 ∂ 1
∇×u= u dz − v dz
∂y h 0 ∂x h 0
Z h
1 ∂u ∂v
= − dz
h 0 ∂y ∂x
h
z(h − z) ∂ 2 p ∂2p
Z
1
 
=− − dz
h 0 2µ ∂x∂y ∂y∂x
= 0.

Hence the vertically averaged velocity field is both incompressible and irrotational.
(d) Within a distance O(h) of the cylinder the solution above is no longer rep-
resentative as no-slip boundary conditions must hold on the cylinder boundary. The
solution further away from the cylinder is least disturbed if we suppose the boundary
condition for u on the cylinder to be u · n = 0.
30 Simon J.A. Malham

Solution (Boundary layer theory: axisymmetric flow)

We consider an axisymmetric flow in cylindrical polar coordinates (r, θ, z) where
ur = −αr/2, uθ = uθ (r, t) only and uz = αz, where α is a constant. First we show this
flow is incompressible. In cylindrical polar coordinates we have
1 ∂ 1 ∂uθ ∂uz
∇·u= (rur ) + +
r ∂r r ∂θ ∂z
1 ∂
 α  ∂
2
= − r +0+ (αz)
r ∂r 2 ∂z
= 0.

Second we compute the form of the vorticity for the flow, we have
 1 ∂uz ∂uθ 
r ∂θ − ∂z
∇×u= ∂ur ∂uz
∂z − ∂r
 
1 ∂ 1 ∂ur
r ∂r (ruθ ) − r ∂θ
 
0
= 0 
1 ∂
r ∂r (ru θ ) − 0
 
0
= 0 ,
ω
1 ∂
where we set ω = ω(r, t) to be ω = r ∂r (ruθ ). The equation for the evolution of
vorticity ω = ω(x, t) is given by
∂ω
+ u · ∇ω = ν∆ω + ω · ∇u.
∂t
For the vorticity form above, in cylindrical polar coordinates, using for any vector v,
we have
∂ v ∂ ∂
v · ∇ = vr + θ + vz ,
∂r r ∂θ ∂z
and using the form for the Laplacian in cylindrical polar coordinates from the formulae
in the Appendix, the equations above for the evolution of vorticity collapse to
∂ω ∂ω 1 ∂ ∂ω ∂uz
 
+ ur =ν r +ω .
∂t ∂r r ∂r ∂r ∂z

(a) For the moment we assume the flow is inviscid and at time t = 0 the initial
data has the form ω = ω0 f (r) for some function f = f (r) and some constant ω0 . We
are required to verify that
ω = ω0 eαt f reαt/2

is a solution for all t > 0. Note that this ansatz satisfies the initial condition at t = 0.
Looking at the evolution equation for ω = ω(r, t) above with ν = 0, we see that we
need to compute
∂ω
rα αt/2
= αω0 eαt f reαt/2 + ω0 eαt f 0 reαt/2 ·


e ,
∂t 2
∂ω
= ω0 eαt f 0 reαt/2 · eαt/2 .


∂r
Introductory fluid mechanics: solutions 31

Subsituting the ansatz for ω above and these two expressions for its partial derivatives
with respect to t and r into the acceleration terms in evolution equation for ω = ω(r, t),
we find
∂ω ∂ω
rα αt/2
= αω0 eαt f reαt/2 + ω0 eαt f 0 reαt/2 ·


+ ur e
∂t ∂r 2
αr
 
ω0 eαt f 0 reαt/2 · eαt/2


−
2
= αω0 eαt f reαt/2

∂uz
=ω .
∂z
Hence we deduce that the ansatz does indeed give a solution to the evolutionary vor-
ticity equation in the inviscid case when ν = 0.
We can interpret this solution ansatz as follows. Consider a fluid particle whose
radial position is prescribed by r = r(t). Suppose that at t = 0 its radial position was
r(0) = r0 . Its radial trajectory is thus given by
dr α
=− r ⇔ r = r0 e−αt/2 .
dt 2
Thus a particle that at time t has radial coordinate r started with radial coordinate
reαt/2 at time t = 0. Thus in the ansatz ω = ω0 eαt f reαt/2 we see that as time evolves

the vorticity intensifies exponentially fast due to: (i) Stretching via the exponential
prefactor eαt , and (ii) Inwards advection via the term reαt/2 in the argument of f .

(b) Now consider steady viscous flow. The equation for the vorticity above becomes
αr ∂ω 1 ∂ ∂ω
 
− =ν r + αω.
2 ∂r r ∂r ∂r
2
We look for a solution ω = ω(r) with the ansatz ω = ω0 e−αr /4ν , where ω0 is a
constant. We observe that
∂ω αr 2
= −ω0 e−αr /4ν ,
∂r 2ν
and
∂
 ∂ω  ω α ∂ 2 −αr2 /4ν

r =− 0 r e
∂r ∂r 2ν ∂r
ω α αr3 −αr2 /4ν
 
= − 0 2r − e .
2ν 2ν
Substituting these expressions into the vorticity equation above we find
1 ∂ ∂ω ω0 α αr2 −αr2 /4ν 2
   
ν r + αω = − 2− e + αω0 e−αr /4ν
r ∂r ∂r 2 2ν
ω0 α2 r2 −αr2 /4ν
= e
4ν
αr ω αr 2

=− − 0 e−αr /4ν
2 2ν
αr ∂ω
=− .
2 ∂r
2
Hence ω = ω0 e−αr /4ν is a solution to the steady viscous vorticity equation. A steady
solution is possible as there is a dominant balance between outwards diffusion, due to
the viscous term, and inwards advection and stretching.
32 Simon J.A. Malham

Solution (Boundary layer theory: rigid wall)

Our first goal is to seek a similarity solution in the form ψ = U (x)δ(x)f (η), where
η = y/δ(x), to the boundary layer equation
∂ψ ∂ 2 ψ ∂ψ ∂ 2 ψ ∂U ∂3ψ
− 2
=U +ν 3,
∂y ∂x∂y ∂x ∂y ∂x ∂y
and find the corresponding equation satisfied by f . Using the chain rule, we compute
the following partial derivatives:
∂ψ ∂2ψ U ∂3ψ U
= U f 0, = f 00 and = 2 f 000 ,
∂y ∂y 2 δ ∂y 3 δ
as well as
∂ψ ∂ ∂δ 0 ∂2ψ ∂U 0 U ∂δ 00
= (U δ)f − U ηf and = f − ηf .
∂x ∂x ∂x ∂x∂y ∂x δ ∂x
Substituting these into the boundary layer equations above we find
 ∂U U ∂δ 00
 ∂
 ∂δ 0 U 00
 ∂U U
Uf0 f0 − ηf − (U δ)f − U ηf f =U + ν 2 f 000
∂x δ ∂x ∂x ∂x δ ∂x δ
∂U 0 2 ∂ U ∂U U
⇔ U (f ) − (U δ) · f f 00 = U + ν 2 f 000 .
∂x ∂x δ ∂x δ
Rearranging this equation we find f 000 + αf f 00 + β 1 − (f 0 )2 = 0, where

δ(∂/∂x)(U δ) (∂U/∂x)δ 2
α := and β := .
ν ν
Since f is independent of x, this last relation implies α and β are independent of x
also and are thus constants.
The product rule implies
∂ 2 ∂δ ∂
(δ U ) = δU + δ (δU )
∂x ∂x ∂x
 ∂ ∂U

=δ (U δ) − δ + να
∂x ∂x
= να − β + να
= ν(2α − β),

where we have used the definitions of α and β above. Hence if 2α = β then δ 2 U = K a

constant. Then from the definition of β we have ∂U/∂x = νβ/δ 2 = (νβ/K)U . Thus we
have U = C exp(νβx/K), for some constant C, and δ = (K/C)1/2 exp −νβx/(2δ) .

However if 2α 6= β then δ 2 U = K ∗ + (2α − β)νx, where K ∗ is a constant. Again using

the definition of β we find
∂U νβ νβ
= 2 = ∗ U.
∂x δ K + (2α − β)νx

β/(2α−β)
Solving this first order differential equation implies U = C ∗ K ∗ +(2α−β)νx ,
where C ∗ is another constant. Hence U is algebraic generically and exponential in the
special case 2α = β.
If there is a rigid wall at y = 0 then no-slip boundary conditions imply u = 0 and
v = 0 at y = 0. Since ∂ψ/∂y = u and ∂ψ/∂y = −v we observe that these no-slip
boundary conditions translate to f (0) = f 0 (0) = 0. Also, since u → U (x) as y → ∞
and u = ∂ψ/∂y = U f 0 , we necessarily have f 0 → 1 as η → ∞.
Introductory fluid mechanics: solutions 33

Acknowledgements The lecture notes were to a large extent grown out of a merging of, lec-
tures on Ideal Fluid Mechanics given by Dr. Frank Berkshire [1] in the Spring of 1989, lectures
on Viscous Fluid Mechanics given by Prof. Trevor Stuart [5] in the Autumn of 1989 (both at
Imperial College) and the style and content of the excellent text by Chorin and Marsden [2].
They also benefited from lecture notes by Prof. Frank Leppington [3] on Electromagnetism.
Some of these solutions are derived from their solutions to the problems in their lecture notes.

References

1. Berkshire, F. 1989 Lecture notes on ideal fluid dynamics, Imperial College Mathematics
Department.
2. Chorin, A.J. and Marsden, J.E. 1990 A mathematical introduction to fluid mechanics,
Third edition, Springer–Verlag, New York.
3. Leppington, F. 1989 Electromagnetism, Imperial College Mathematics Department.
4. Majda, A.J. and Bertozzi, A.L. 2002 Vorticity and incompressible flow, Cambridge Texts
in Applied Mathematics, Cambridge University Press.
5. Stuart, J.T. 1989 Lecture notes on Viscous Fluid Mechanics, Imperial College Mathematics
Department.