Blasius' Theorem

Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

PHYSICS 200B : CLASSICAL MECHANICS

SOLUTION SET #4

[1] Blasius’ theorem says that the force per unit length of a body of constant cross-sectional
profile Σ is given by
dW 2
I  
i
F̄ = Fx − iFy = 2 ρ dz ,
dz
C
where C = ∂Σ is a closed curve which traces the boundary of Σ, and W (z) is the complex
potential.

Consider a 2D flow with stream function ψ(x, y) = A(x − c)y, where A and c are real
constants. A circular cylinder of radius a is introduced into this flow, with its center at the
origin. Find W (z) for the resulting flow. Use Blasius’ theorem to calculate the force per
unit length exerted on the cylinder.

We first find the conjugate harmonic function φ(x, y) satisfying


∂φ ∂ψ ∂φ ∂ψ
= = A(x − c) , =− = −Ay .
∂x ∂y ∂y ∂x
We conclude
φ(x, y) = 12 Ax2 − Acx − 21 Ay 2 ,
and thus
w(z) = φ(x, y) + iψ(x, y) = 12 Az 2 − Acz .
Now we introduce a cylinder of radius a. The boundary of the circle must be a streamline,
but as |z| → ∞ we have v = ∇φ where φ(x, y) is given above. To make this so, we invert
w(z) in the circle |z| = a and write
W (z) = w(z) + w(a2 /z̄)
Aa2 Aca2
= 12 Az 2 − Acz + − .
2z 2 z
Using Cauchy’s theorem, we then find
dW 2
I  
i
F̄ = 2 ρ dz
dz
C
2
Aca2 Aa2
I 
i
= 2 ρ dz Az − Ac + 2 − 3
z 2z
|z|=a

= i
2ρ · 2πi · 2A2 ca2 = −2πρA2 ca2 .
Thus, Fx = −2πρA2 ca2 and Fy = 0.

[2] Show that the Joukowski transformation Z = z + a2 /z can be written in the form
z−a 2
 
Z − 2a
= ,
Z + 2a z+a

1
so that n o
arg(Z − 2a) − arg(Z + 2a) = 2 arg(z − a) − arg(z + a) . (1)

Consider the circle in the (x, y) plane which passes through z = −a and a with its center
at z0 = ia ctn β. Show that the above transformation takes this circle into a circular arc
between Z = −2a and Z = +2a, with subtended angle 2β (see figure). Obtain an expression
for the complex potential in the Z plane when the flow is uniform at speed V and parallel
to the real axis. Show that the velocity will be finite at both the leading and tailing edges
if Γ − −4πV a ctn β.

Figure 1: Geometry of the circle and its image in problem 2.

We have
Z − 2a z + a2 z −1 − 2a (z − a)2 /z (z − a)2
= 2 −1
= 2
= .
Z + 2a z + a z + 2a (z + a) /z (z + a)2
Taking the argument and using arg(z1 /z2 ) = arg(z1 ) − arg(z2 ) the desired result follows
immediately.

Next let z0 = ia ctn β. The radius of the circle in the z-plane is b, where

b2 = a2 + (a ctn β)2 = a2 csc2 β ,

so b = a/ sin β. The locus of points on this circle may be written as z(θ) = z0 − ib eiθ , where
θ ∈ [0, 2π). Thus,
z ± a = e∓iβ − eiθ · ia csc β ,


and we have 2
e−iβ − eiθ

Z + 2a
= .
Z − 2a eiβ − eiθ

2
Now
e−iβ − eiθ −iβ e
i(β+θ) − 1
= −e ·
eiβ − eiθ eiβ − eiθ
ei(β+θ)/2 − e−i(β+θ)/2 ei(β+θ)/2
= −e−iβ · ·
ei(β−θ)/2 − e−i(β−θ)/2 ei(β+θ)/2
sin 21 (β + θ)
 
= −e−iβ ·  .
sin 21 (β − θ)


Thus,
arg(Z + 2a) − arg(Z − 2a) = 2π − 2β ,
which says that the circle in the z-plane maps to an arc in the Z-plane as shown in fig. 1.

Now consider the complex potential


V b2 Γ
W (z) = V (z − z0 ) + + log(z − z0 ) ,
z − z0 2πi
corresponding to uniform flow at infinity with a streamline along |z| = b. Then the complex
potential in the Z-plane is W(Z) = W (F (Z)) where
 p 
F (Z) = z = 12 Z ± Z 2 − 4a2 .

Thus the complex velocity


V(Z) = W ′ (Z) = W ′ (z) F ′ (Z)
Consider the case Z = 2a, corresponding to z = a. Since
1 Z z
F ′ (Z) = ± √ = ±√ ,
2 2 Z − 4a
2 2 Z − 4a2
2

we have that F (Z) diverges with an inverse square root singularity as Z approaches ±2a.
We now show that W ′ (z) vanishes when Z = ±2a, cancelling the singularity, provided
Γ = −4πV a ctn β. In this case,
( )
′ b2 /a2 Γ/aV 1
W (z = a) = V 1 − +
(1 − i ctn β)2 2πi 1 − i ctn β
( )
1 2i cos β
=V 1− +
(sin β − i cos β)2 sin β − i cos β
n o
= V 1 + e−2iβ − 2 cos βe−iβ = 0 ,

which vanishes! To find out the value of the velocity at the leading and trailing edges, set
z = a + δz. An intelligent parameterization here is to take δz = −iǫa csc β eiβ and see what
happens for complex ǫ. We then have
z − z0 = z − ia ctn β = a − ia ctn β − iǫa csc β eiβ
ia
=− (1 + ǫ) eiβ .
sin β

3
Then
( )
′ e−2iβ 2 cos β e−iβ
W (z) = V 1 + −
(1 + ǫ)2 1+ǫ
( )
e−2iβ ǫ
=V 1− · = 2iV sin β e−iβ ǫ + O(ǫ2 ) .
1+ǫ 1+ǫ

Next, we have F ′ (Z) = z/ Z 2 − 4a2 . We write Z 2 − 4a2 = (Z + 2a)(Z − 2a). For Z ≈ 2a
we may write Z + 2a = 4a + O(ǫ), and
a2 (z − a)2 2
Z − 2a = z + − 2a = = a −iǫ csc β eiβ .
z z
Thus,
z 1 a 1 i e−iβ
F ′ (Z) = √ ·√ = √ · √ = .
Z + 2a Z − 2a 2 a −iǫ a csc β eiβ 2 ǫ csc β
Thus we see that W ′ (z) vanishes as ǫ1 and F ′ (Z) diverges as ǫ−1 . Multiplying and taking
the limit ǫ → 0, we obtain the complex velocity at the edge Z = 2a to be

V = −V sin2 β e−2iβ .

[3] Show that an array of N identical point vortices of circulation Γ , placed equally about
a circle of radius a, will rotate at a constant angular frequency Ω. Find the value of Ω.

Let ω = e2πi/N . The locations of the vortices are taken to be zn = aω n where n ∈ {1, N };
note that zn+N = zn . The complex potential for a vortex located at the origin is W (z) =
(Γ/2πi) log z, and the corresponding complex velocity field is v̄(z) = Γ/2πiz. The complex
velocity of the j th vortex is a sum of contributions for all the others and is given by
N −1
Γ ω̄ j X 1
v̄j = .
2πia n=1 1 − ω n

Suppose N is odd. Then we pair the terms in the above sum: n with N − n. Note that
1 1 1 ωn
+ = + =1 ,
1 − ω n 1 − ω N −n 1 − ωn ωn − 1
since ω N = 1. There are (N − 1)/2 such pairs, so we conclude that
N −1
v̄j = Γ ω̄ j .
4πia
When N is even, we again pair n with N − n. The value n = N/2 is its own mate, and
there are (N − 2)/2 bona fide pairs. Thus,
N −1
X 1 N −2 1 N −1
= + = ,
1 − ωn 2 1 − ω N/2 2
n=1

4
since ω N/2 = −1. Thus once again we have v̄j = (N − 1)Γ ω j /4πia. Note that uniform
rotation in the (x, y) plane about the origin with angular frequency Ω means

v(r) = Ω ẑ × r = Ω (xŷ − y x̂) ,

and thus the complex velocity is v̄ = Ω (−y − ix) = −iΩz̄. For the j th vortex, zj = aω̄ j .
Thus, we conclude Ω = (N − 1)Γ/4πa2 .

[4] Consider a large circular disk of radius R executing a prescribed angular motion θ(t).
The disk is immersed in a fluid under conditions of constant pressure. Let the plane of the
disk lie at z = 0. Assume that the fluid velocity takes the form

vφ (r, φ, z, t) = r Ω(z, t) , (2)

with vr = vz = 0.

(a) Write down the Navier-Stokes equations for the fluid. Assume you can neglect the
(v · ∇) v term. (Under what conditions is this true?) Show that you obtain the diffusion
equation. What are the boundary conditions on the fluid motion?

The Navier-Stokes equations are


∂v
+ (v · ∇) v = −∇p + ν ∇2 v . (3)
∂t
If we neglect the nonlinear term, we have the diffusion equation,

∂Ω ∂ 2Ω
=ν . (4)
∂t ∂z 2
In deriving this, it is useful to write

v = r Ω(z, t) φ̂ = (xŷ − y x̂) Ω(z, t) . (5)

The nonlinear term is



(v · ∇) v = − vφ2 = −Ω 2 r . (6)
r
This may be neglected if
ν
Ω ≪ , (7)
R2
which is equivalent to Re ≪ 1, where the Reynolds number is Re = R vφ /ν.

(b) Our goal is next to find a complete solution to Ω(z, t) in terms of the function θ(t). To
this end, we perform the following analysis. Define the spatial Laplace transform,
Z∞
Ω̌L (κ, t) ≡ dz e−κz Ω(z, t) . (8)
0

5
You may assume in this problem that the fluid motion is symmetric about z = 0, i.e.
Ω(z, t) = Ω(−z, t), so we only have to consider the region z ≥ 0. The inverse Laplace
transform is
c+i∞
dκ +κz
Z
Ω(z, t) = e Ω̌L (κ, t) (9)
2πi
c−i∞
where the contour lies to the left of any branch cut or singularity on the line Im (κ) = 0.
Later on we will see that we can take c = 0, so the contour lies along the axis Re (κ) = 0.
Show directly that
∂t − νκ2 Ω̌L (κ, t) = Fκ (t) ,

(10)
where the function Fκ (t) on the RHS depends on Ω(0, t) and Ω ′ (0, t) (prime denotes differ-
entiation with respect to z). Find Fκ (t).

We have that
Z∞ ( )
−κz ∂Ω ∂ 2Ω
0 = dz e −ν
∂t ∂z 2
0
= ∂t − ν κ2 Ω̌L (κ, t) + ν Ω ′ (0, t) + κ Ω(0, t) .
  
(11)
Thus,
∂t − ν κ2 Ω̌L (κ, t) = −ν Ω ′ (0, t) + κ Ω(0, t) .
  
(12)

(c) Integrate the above first order equation from some arbitrary initial time t = t0 to
final time t and obtain Ω(z, t) in terms of the functions Ω(z, t0 ), Ω(0, t), and Ω ′ (0, t).
Show that the term involving Ω(z, t0 ) is a transient which decays to zero in the limit
t0 → −∞. Dropping the transient, performing the inverse Laplace transform, and rotating
the κ contour so that κ = ik, where k runs along the real axis, show that
Z∞ Zt
dk ikz 2 (t−t′ )
h i
Ω(z, t) = −ν e dt′ e−νk Ω ′ (0, t′ ) + ikΩ(0, t′ ) . (13)

−∞ −∞

Integrating, we obtain
Zt h i
νκ2 (t−t0 ) 2 ′
Ω̌L (κ, t) = e Ω̌L (κ, t0 ) − ν dt′ eνκ (t−t ) Ω ′ (0, t′ ) + ikΩ(0, t′ ) . (14)
t0

The first term is a transient which is negligible in the limit t0 → −∞. Remember that κ is
purely imaginary along its integration contour, so we can set κ ≡ ik with k real. Applying
the inverse Laplace transform, we recover the desired result.

(d) Find the total torque on the disk N (t). You will need to integrate r × f over the surface
of the disk, using the viscous stress tensor of the fluid. Show that
Nfluid (t) = πη R4 Ω ′ (0, t) , (15)

6
where η = ρ ν is the shear viscosity.

The viscous force per unit surface area is fi = σ̃ij nj , where nj is the surface normal and
 
∂vi ∂vj 2
σ̃ij = η + − δ ∇ · v + ζ δij ∇ · v (16)
∂xj ∂xi 3 ij

is the viscous stress tensor. For the flow v = r Ω(z, t) φ̂, the divergence vanishes. The
differential viscous torque dN = dN ẑ on the disk is then

dN = xfy − yfx dA
 
∂vy ∂vx ∂Ω
=η x −y dA = η r dA . (17)
∂z ∂z ∂z

Integrating, we find the total viscous torque:

ZR
∂vφ
N = 2 dr 2πr r η = πηR4 Ω ′ (0, t) . (18)
∂z
0

Note the factor of two, which arises from integration over both sides of the disk.

(e) By going to Fourier space in frequency, the k integral can be done. Show that

i eik+ z n ′ o
Ω̂(z, ω) = − Ω̂ (0, ω) + ik+ Ω̂(0, ω) , (19)
k+ − k−

where k± = ± eiπ/4 ω/ν. Thus, setting z → 0+ , we obtain


p

Ω̂ ′ (0, ω) = −ik− Ω̂(0, ω) . (20)

Taking the Fourier transform, we have

Z∞ Z∞ Zt
iωt dk ikz 2 ′
h i
Ω̂(z, ω) = −ν dt e e dt′ e−νk (t−t ) Ω ′ (0, t′ ) + ikΩ(0, t′ ) (21)

−∞ −∞ −∞
Z∞ Z∞ Z∞
dk ikz 2
h i
= −ν e ds e−νk s eiωs dt eiω(t−s) Ω ′ (0, t − s) + ikΩ(0, t − s)

−∞ 0 −∞
Z∞
dk eikz  ′  −i eik+ z h ′ i
=− Ω̂ (0, ω) + ik Ω̂(0, ω) = Ω̂ (0, ω) + ik+ Ω̂(0, ω) ,
2π k2 − iω
ν
k+ − k−
−∞

where we assume z > 0 in the last line. There is a subtlety here which is worth mentioning.
In the above
p derivation, we have assumed ω is real and positive. For general ω, the roots
are k = ± iω/ν and we define k+ to be the root with the positive imaginary part.

7
(f) Suppose the disk is suspended from a torsional fiber. Let the disk’s moment of inertia
be I and the restoring torque due to the fiber be Nfiber = −Kθ. Show that the equation
for the oscillation frequency of the disk is

ω 2 + eiπ/4 ων1/2 ω 3/2 − ω02 = 0 , (22)

where ω0 = (K/I)1/2 , and


π 2 ρ2 R8 ν
ων = . (23)
I2
Analyze this equation in the limits ω0 ≪ ων and ω0 ≫ ων , and find the frequency of damped
oscillations. Hint: The former case is easy – simply neglect the ω 2 term. For the latter
case, perturb about the ων = 0 solutions ω = ±ω0 . Find the real and imaginary parts of
the oscillation frequency ω in each case.

We have Ω(0, t) = θ̇(t), hence Ω̂(0, ω) = −i ω θ̂(ω). Then


r
′ iπ/4 ω
Ω̂ (0, ω) = i e Ω̂(0, ω)
ν
ω 3/2
= eiπ/4 1/2 θ̂(ω) . (24)
ν
The Fourier transform of the torque is then

N̂(ω) = πρR4 · eiπ/4 ν 1/2 ω 3/2 θ̂(ω) . (25)

Newton’s second law for the disk is then

−Iω 2 θ̂(ω) = −K θ̂(ω) + N̂ (ω) , (26)

from which we obtain the desired result of eqn. ??. To be perfectly correct, we should write
this as
ω 2 + eiπ/4 ων1/2 ω 3/2 sgn(Re ω) − ω02 = 0 , (27)

Suppose ω0 = 0. Then we have two solutions, ω = 0 and ω = −iων . For small ω0 , the latter
will continue to be highly overdamped. The former solution becomes finite, and neglecting
the O(ω 2 ) term (since ω is small), we find
4/3
ω = e−iπ/6 ω0 ων−1/3 . (28)
1 4/3 −1/3
The damping rate is then γ = −Im ω = 2 ω0 ων .

In the opposite limit, where ων ≪ ω0 , write ω = ω0 + δω and solve to first order in δω,
obtaining

δω = − 12 eiπ/4 ω0 ων . (29)
p
The viscous damping leads to a frequency shift and damping rate −∆ω = γ = ω0 ων /8.
Note that ∆ω < 0, as is the case with a simple damped harmonic oscillator.

8
Note: There is an easier way to solve this problem, if we use some intuition. The diffusion
equation Ωt = νΩzz and the boundary conditions are linear, which suggests we write our
solution as
Ω(z, t) = A(ω) e−Q|z| e−iωt . (30)
This is a solution to the diffusion equation if νQ2 = −iω. p Of the two roots for Q(ω), we
need the one with the positive real part, so Q = e −iπ/4 ω/ν. Setting z = 0 and using
Ω̇ = θ, we find A(ω) = −iω θ̂(ω). The Fourier component of the viscous torque on the disk
is then

N̂fluid (ω) = πρνR4 · (−Q)(−iω) θ̂(ω) (31)


= eiπ/4 πρR4 ν 1/2 ω 3/2 θ̂(ω) , (32)

which when plugged into the equation of motion for the disk yields the above equation for
the oscillation frequency.

You might also like