4.

2 The Falkner-Skan equation

For the case of a at plate, we have just shown that the boundary layer solution is
self-similar in the sense that two proles u(x, y) at dierent values of x dier only by
a scale factor in y. This simplied the analysis considerably, because the momentum
equation reduced to an ordinary dierential equation in the single similarity variable
. What about more complicated geometries? In this section, we derive the general
conditions under which self-similar solutions exist. As usual we work in terms of the
stream function so that the continuity equation is automatically satised.
Consider again the boundary layer transformations 65, which tell us that the basic
scales of the y coordinate and the stream function are
y

= Re
1/2
y
L
and (x, y) =
LU
Re
1/2

(x

, y

). (106)
For a at plate, we argued that we must actually replace the overall length of the object
L by the present distance travelled along it, x. This gave (in an unusual format)
=
Re
1/2
_
2x/L
y
L
and =
LU
_
2x/L
Re
1/2
f(), (107)
in which the scaled stream function depends on space only through the single similarity
variable .
To extend this analysis to more general geometries, we recall that the only way the
geometry enters the BL equations is through its inuence on u
e
(x). For a at plate
just considered, u
e
(x) = U = constant. More generally, u
e
= u
e
(x). In what follows,
we aim to nd the most general form of u
e
(x) that still admits self-similar solutions.
To do so, we postulate a generalised similarity transformation
=
Re
1/2
h(x)
y
L
and =
Lu
e
(x)h(x)
Re
1/2
f(). (108)
where h(x) is an unknown, non-dimensional function of x representing the dependence
of on x. For h(x) =
_
2x/L and u
e
(x) = U we recover the at-plate variables 107.
(Notice that in this more general case we are rescaling both y and u by functions of x.
For the at plate, with u
e
constant, we only needed to rescale y.) For convenience, we
dene a second non-dimensional variable,
=
x
L
. (109)
We now change coordinates from (x, y) to (, ) in the momentum BL equation
u
u
x
+ v
u
y
= u
e
du
e
dx
+

2
u
y
2
. (110)
To do so, we start by calculating transformed expressions for /x and /y:

x
=
1
L

(x)
h(x)

, and

y
=
Re
1/2
Lh(x)

. (111)
The velocity components u and v are then
u =

y
= u
e
(x)f

(), (112)
24
v =

x
=
1
Re
1/2
_
d
d
(u
e
(x)h(x))f() Lu
e
(x)h

(x)f

()
_
. (113)
Using these, we now calculate all the component terms of the momentum equation 110.
The viscous term is

2
u
y
2
=
Re
1/2
Lh

_
Re
1/2
Lh

(u
e
f

)
_
=
U
L
u
e
h
2
f

.
(114)
The advection terms are
u
u
x
= u
e
f

_
1
L

_
(u
e
f

)
=
u
e
L
du
e
d
f
2
u
2
e
h

h
f

,
and
v
u
y
=
_

1
Re
1/2
d
d
(u
e
h)f +
L
Re
1/2
u
e
h

_
Re
1/2
Lh

(u
e
f

)
=
u
e
Lh
d
d
(u
e
h)ff

+ u
2
e
h

h
f

.
The pressure term, expressed as usual in terms of the exterior slipping velocity,
is
u
e
(x)
du
e
(x)
dx
=
u
e
(x)
L
du
e
(x)
d
. (115)
Putting all these together, we get the transformed momentum equation
U
L
u
e
h
2
f

+
u
e
Lh
d
d
(u
e
h)ff

+
u
e
L
du
e
d
{1 f
2
} = 0. (116)
Multiplying across by Lh
2
/Uu
e
we get
f

+
h
U
d
d
(u
e
h)ff

+
h
2
U
du
e
d
{1 f
2
} = 0. (117)
We rewrite this as
f

+ ff

+ (1 f
2
) = 0, (118)
in which (using d/d = Ld/dx to revert back x)
=
Lh
U
d
dx
(u
e
h) and =
Lh
2
U
du
e
dx
. (119)
The boundary conditions are (as usual)
f(0) = f

(0) = 0 (120)
and
f

() = 1. (121)
25
Recall now the arguments of the various functions in Eqn. 117: f = f(), u
e
=
u
e
(x), h = h(x) and so (a priori at least) = (x) and = (x). For a self-similar
solution to exist, and must actually be constants, independent of x (or equivalently
of ). By forming appropriate linear combinations of and , and integrating the
expressions thus obtained, we can calculate the forms that u
e
(x) and h(x) must take
in order for this to be the case. (We do not give the details here: they can be found in
Schlichting, Boundary Layer Theory, McGraw Hill, chapter VIII b.) These are
u
e
(x)
U
= K
2
2
_
(2 )
x
L
_

2
, (122)
and
h(x) =

(2 )
x
L
U
u
e
, (123)
where K is an integration constant. The requirement 2 = 0 is imposed. We do
not consider the case 2 = 0. As can be seen, the overall result is independent of
any common factor in and , as this can simply be absorbed into the denition of
h. As long as = 0, therefore, we are free to choose = 1. For convenience we also
introduce a new constant
m =

2
, so that =
2m
m + 1
. (124)
Using 122, 123 and 124 we get
u
e
(x)
U
= K
m+1
_
2
m + 1
x
L
_
m
, (125)
and
h(x) =

2
m + 1
x
L
U
u
e
. (126)
Substituting h(x) into Eqn. 108, we obtain the nal form of the similarity variable
= y
_
m + 1
2
u
e
x
= Re
1/2
x
y
x
_
m + 1
2
with Re
x
=
u
e
x

. (127)
Summary
For an exterior ow u
e
(x) x
m
we have the similarity variable given by Eqn. 127.
The scaled stream function f() then obeys the ordinary dierential equation
f

+ ff

+ (1 f
2
) = 0 (128)
with constant. The boundary conditions are
f

(0) = f(0) = 0, f

() = 1. (129)
This equation is known as the Falkner-Skan equation.
26
4.2.1 Physical interpretation
Looking back at Eqn. 125, we see that self-similar solutions of the 2D boundary layer
equations exist if the exterior inviscid ow is a power law
u
e
(x) = Cx
m
, C constant. (130)
It turns out (not shown here) that this is the prole of ow past a wedge of angle .

>0
<0
4.2.2 Special cases
m = 0: Blasius ow over a at plate, Sec. 4.1.
m = 1: 2D stagnation ow, e.g., near the nose of a cylinder (problem sheet 3).
Blasius flow
m = 0
U U
m = 1
2d stagnation flow
4.2.3 Numerical results
In accordance with the discussion of Sec. 4.1, the Falkner-Skan equation must be solved
numerically. The velocity prole is shown in Fig. 10 for dierent values of m. Note
that it develops an inexion point as m (and hence also ) becomes negative. This
is closely related to the phenomenon of boundary layer separation, to which we now
turn.
Figure 10: Falkner-Skan proles. (Schlichting, Boundary Layer Theory, McGraw Hill).
27