You are on page 1of 7

Chapter 5

Conformal mapping
We are generally interested in the complex z-plane, which corresponds to the plane with Cartesian
coordinates (x, y) = (z, z) . Sometimes it is easier to consider a simpler plane ideal ow and
then transform it into the geometry of interest. This is accomplished with conformal mapping.
In addition to the z-plane, consider another with complex coordinate = + i = ||e
i
,
And transform or map the ow in the z-plane from the -plane.
If the mapping function
z = f() (5.1)
is analytic, then any analytic function of z gives an analytic function of by simple composition
of functions; e.g.
g(z) = g (f()) . (5.2)
The function in the -plane is analytic since it has the derivative
dg
d
=
dg
dz
dz
d
=
dg
dz
df()
d
. (5.3)
This is really just the chain rule; conformal mapping is the chain rule for complex functions.
On the other hand, given g(f()) as an analytic function of , g(z) is analytic too since
dg
dz
=
dg
d
d
dz
=
dg
d
_
dz
d
_
1
=
dg/d
dz/d
. (5.4)
This could have obtained more easily by dividing (5.3) by f

() .
If the point z
1
= f(
1
) is mapped from the point
1
, then the value of the function g at
z
1
in the z-plane is the same as the value of g at
1
in the -plane, since z
1
= f(
1
) implies
g(z
1
) = g (f(
1
)) .
This is particularly useful if for g we consider the complex potential W . If C is a stream-
line of the ow in the -plane, then = W is constant along it, and this also applies to the
imaginary part of W along the curve f(C) in the z-plane to which C is mapped by z = f() . It
follows that f(C) is a stream-line in the -plane.
The complex velocities in the - and z-planes are then given by applications of (5.3)(5.4):
w() =
dW
d
=
dW
dz
dz
d
= w(z)f

() (5.5)
w(z) =
dW
dz
=
dW/d
dz/d
=
w()
f

()
. (5.6)
References: Abbott and von Doenho (1959, p. 50), Milne-Thomson (1973, p. 66).
47
48 AERODYNAMICS I COURSE NOTES, 2005
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
0.5
1.0
1.5
0
(z)
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
0.5
1.0
1.5
0
()
Figure 5.1: Mapping the rst quadrant of the -plane () to the upper-half z-plane (z) with
z = f() =
2
, to get the corner ow in the -plane from a uniform horizontal ow in the
z-plane with the complex potential W = z .
5.1 Mapping with powers of
A simple family of useful transformations is z = f() =
k
, with k a real constant .
5.1.1 Square mapping
The mapping z = f() =
2
takes the rst quadrant of the -plane into the rst two quadrants
of the z-plane. Since
2
= ||
2
e
2i
, an eect of the mapping is to double the angle each point
makes with the positive real axis: arg z = 2 arg . Also, it is no surprise that the square of a
positive real number is also real and positive, so that the positive real -axis maps to the positive
real z-axis, while the square of an imaginary number is negative, so the positive imaginary -axis
maps to the negative real z-axis.
Consider then the complex potential W = z , which represents a simple uniform horizontal
velocity in the z-plane, with stream-function (z) = W(z) = z = y , and horizontal stream-
lines = y = const. The same complex potential represents the corner ow of 3.5.3 in the
-plane (apart from a scale factor of 2):
W = z =
2
(5.7)
= W =
2
(5.8)
= ( + i)
2
(5.9)
= 2 . (5.10)
Four stream-lines are plotted in gure 5.1: the same four in each case. The velocity in the -plane
is given by
w() =
dW
d
= 2 (5.11)
which (apart from the scale factor 2) is the same as the complex velocity in 3.5.3.
Conformal mapping 49
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
0.5
1.0
1.5
0
(z)
-2.0
-1.5
-1.0
-0.5
0.0
0.5
1.0
1.5
2.0
-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
0.5
1.0
1.5
0
()
Figure 5.2: Mapping the rst three quadrants of the -plane () to the upper-half z-plane (z)
with z = f() =
2/3
, to get the external corner ow in the -plane from a uniform horizontal
ow in the z-plane with the complex potential W = z .
5.1.2 f() =
2/3
Just as z = f() =
2
takes the rst -quadrant to the rst two z-quadrants, z = f() =
2/3
takes the rst three to the rst two. If we consider the same ow in the z-plane as in the last
example; i.e. W = z ; and apply z = f() =
2/3
, we get the ow around an external corner, as
in gure 5.2. The velocity in the -plane is
w() =
dW
dz
dz
d
= 1
2
3

1/3
=
2
3
||
1/3
e
i/3
, (5.12)
so that the direction of ow at any point makes an angle =

3
one-third that of the line from
the origin to the point; e.g. for all points on the positive -axis, the ow is two oclock, and for
all points on the negative -axis, its one oclock.
5.1.3 Other powers
Other shapes can be obtained using other powers. Also, one can use a dierent complex potential.
One diculty with the above mapping functions is that they are not bijective. In the case
of z = f() =
2
, the third quadrant of the -plane also maps to the upper-half z-plane; it is
therefore not injective. For z = f() =
2/3
, the mapping itself is multivalued; for example,
should = 1 = e
i
go to e
2i/3
or should = 1 = e
i
go to e
2i/3
? In 5.1.2, we chose
the branch that maps = ||e
i
with 0
3
2
to z = re
i
= ||
2/3
e
2i/3
, since its continuous
in the rst three quadrants.
5.2 Joukowskys transformation
The mapping function f() = +a
2
/ for a real constant a is called Joukowskys transformation
(Milne-Thomson 1973, p. 97).
50 AERODYNAMICS I COURSE NOTES, 2005
5.2.1 Unit circle from a straight-line segment
A shifted special case is
z = f() = +
c
2
+
c
2
16
. (5.13)
Notice that both and
c
2
16
map to the same point: the mapping (5.13) is not injective. If
4|| = c then both these points are on the circle, otherwise one is inside and one outside, and
both the interior and exterior of the circle in the -plane map to the entire z-plane. This means
care must be taken in constructing the inverse transformation
= f
1
(z) =
c
4
_
2z
c
1 2
_
z
c
_
z
c
1
_
_
=
z
c
2

_
z(z c)
2
. (5.14)
Usually for aerodynamical applications were interested in the exterior, 4|| > c , so this condition
should be checked to determine the appropriate sign in (5.14).
The mapping (5.13) takes the circle 4|| = c in the -plane to the straight-line segment joining
zero and c in the z-plane (Abbott and von Doenho 1959, p. 50). To see this, put 4 = ce
i
and
let run from zero to 2:
z =
c
4
_
2 + e
i
+ e
i
_
(5.15)
=
c
2
(1 + cos ) , (5.16)
so that z is real and conned to 0 z c .
As runs from zero to , moves anticlockwise around the upper unit semicircle and z
moves left from c to zero, and then as increases to 2, moves anticlockwise around the lower
unit semicircle and z moves right from zero back to c .
In this double-tracing of the segment, it can help to think of the left tracing for 0 < <
as corresponding to the upper side of the segment and the right tracing for < < 2 as
corresponding to the lower surface. If the segment 0 < z < c is a stream-line, it can be replaced
by an impermeable thin at plate, and then dierences in the velocity or pressure between the
upper and lower surfaces can be tolerated, even though the plate thickness is neglected. Thus
in talking about the pressure at x =
c
2
, say, we should specify whether we mean the upper or
lower surface ; i.e. z =
c
2
i as 0 .
5.2.2 Uniform ow & ow over a circle
Consider the complex potential W = q

z in the z-plane (with q

a positive constant). Its


stream-function is (z) = W(z) = q

z = q

y ; which is constant along horizontal lines.


This means the real z axis is a stream-line, and so is the segment from z = 0 to z = c . This
further means that the corresponding complex potential in the -plane will have the circle 4|| = c
as a stream-line. That potential is
W = q

z = q

_
+
c
2
+
c
2
16
_
. (5.17)
Instead, lets shift some constants and consider instead
W = q

_
z
c
2
_
= q

_
+
c
2
16
_
. (5.18)
Conformal mapping 51
(z) ()
Figure 5.3: Flow over a thin plate at zero incidence (z), and ow over a circle (), being z- and
-plane representations of the complex potential W = q

z with the Joukowsky transformation


z = +
c
2
+
c
2
16
.
The corresponding stream-functions
= q

y = q

_
1
c
2
16||
2
_
, (5.19)
which are obviously constant along both y = 0 and 4|| = c , are plotted in gure 5.3 using the
Octave code listed in 5.3.
5.2.3 Thin at plate at nonzero incidence
Nontrivial ows over the thin at plate in the z-plane can now be obtained by modifying the
complex potential in the -plane, provided the circle || = a remains a stream-line. First, rotate
the whole stream-line pattern in the -plane by replacing with e
i
:
W = q

_

e
i
+
c
2
e
i
16
_
, (5.20)
as plotted in gure 5.4.
1
Notice that as || , W q

e
i
, which is the complex potential of a uniform stream
with speed q

inclined at an angle to the positive real axis. Because the Joukowsky transfor-
mation (5.13) far from the origin asymptotically approaches z +
c
2
, the velocity far from the
thin at plate
w(z) =
dW/d
dz/d
q

e
i
(|z| ) (5.21)
which is a uniform stream with the same speed q

and the same direction ; this is plausible


in gure 5.4.
Since the complex velocity for the thin at plate is recovered from (5.4), the conformal
transformation can introduce a singularity in w(z) if there are stationary points for which dz/d
1
A gure similar to 5.4 (z) is given by Lamb (1932, p. 86), following a slightly dierent derivation.
52 AERODYNAMICS I COURSE NOTES, 2005
(z) ()
Figure 5.4: Flow over a thin plate at nonzero incidence (z), and ow over a circle (), being z- and
-plane representations of the complex potential (5.20) with the Joukowsky transformation (5.13).
vanishes. This occurs for the Joukowsky mapping (5.13) at
d
d
_
+
c
2
+
c
2
16
_
= 1
c
2
16
2
= 0 ; (5.22)
i.e. 4 = c . These points map to z = c and z = 0 , the ends of the thin at plate.
5.2.4 Flow over the thin at plate with circulation
The ow can be further modied by adding a vortex at the origin in the -plane, since this still
preserves || = a as a stream-line. This changes the potential (5.20) to
W =
q

c
4
_
4
ce
i
+
ce
i
4
_
+
i
2
ln
4
ce
i
. (5.23)
Note that ln and ln(4/ce
i
) = ln (i+ln cln 4) dier only by a physically irrelevant complex
constant. The stream-lines are plotted in gure 5.5 for the choice =

6
and = q

c sin .
The latter choice happens to be exactly that which causes the ow in the z-plane to leave the
trailing edge (z = c ) of the plate smoothly, as can be seen in gure 5.5 (z); this signicance of
this choice will be explored in ch. 6.
5.2.5 Joukowsky aerofoils
The Joukowsky transformation (5.13) maps the circle 4|| = c to the thin at plate 0 < z < c ,
but it also maps circles not centred on = 0 to some very interesting shapes, including some
that look like aerofoils. This is not pursued here; see Abbott and von Doenho (1959, p. 51) or
Milne-Thomson (1973, pp. 99100).
5.3 Example GNU Octave code for conformal mapping
Note that the code deals with dimensionless quantities: all lengths in the - and z-planes are
normalized by c and all velocities are normalized by q

. This nondimensionalization reduces by


Conformal mapping 53
(z) ()
Figure 5.5: Flow over a thin plate at nonzero incidence with circulation (z), and ow over a circle
(), being z- and -plane representations of the complex potential (5.20) with the Joukowsky
transformation (5.13).
two the number of parameters that need to be set in the code.
function [Z, WZ, z, Wz] = flat_plate (x, y, alpha, Gamma)
[X, Y] = meshgrid (x, y);
z = Z = X + 1i*Y; # writing Z for zeta
Zz = joukowsky_inverse (z); # Z as a function of z
WZ = flat_plate_W (Z, alpha, Gamma);
Wz = flat_plate_W (Zz, alpha, Gamma);
endfunction
function Z = joukowsky_inverse (z)
Z = 0.5 * (z - 0.5 + sqrt (z .* (z-1)));
flip = abs (Z) < 0.25; # ensure its outside |Z|=0.25
Z(flip) = 1 ./ (16 * Z(flip));
endfunction
function W = flat_plate_W (Z, alpha, Gamma)
Z *= 4 * exp (-1i*alpha);
W = Z + 1 ./ Z + 1i*Gamma/2/pi * log (Z);
endfunction
%!demo
%! x = y = -2:0.1:2; a = pi/6; Gamma = 4*pi*sin(a);
%! [Z, WZ, z, Wz] = flat_plate (x, y, a, Gamma);
%! psi = -15:0.1:15;
%! axis ("image"), legend ("off")
%! contour (real (z), imag (z), imag (Wz), psi); pause
%! contour (real (Z), imag (Z), imag (WZ), psi)

You might also like