Boundary-layer control by electric elds: A

feasibility study
R. Vilela Mendes
Grupo de Fsica-Matematica, Complexo II, Universidade de Lisboa,
Av. Gama Pinto 2, 1699 Lisboa Codex, Portugal
J. A. Dente
Laboratorio de Mecatronica, DEEC, Instituto Superior Tecnico,
Av. Rovisco Pais, 1096 Lisboa Codex, Portugal
May 31, 1998
Abstract
A problem of great concern in aviation and submarine propulsion
is the control of the boundary layer and, in particular, the methods
to extend the laminar region as a means to decrease noise and fuel
consumption. In this paper we study the ow of air along an air-
foil when a layer of ionized gas and a longitudinal electric eld are
created in the boundary layer region. By deriving scaling solutions
and more accurate numerical solutions we discuss the possibility of
achieving signicant boundary layer control for realistic physical pa-
rameters. Practical design formulas and criteria are obtained. We also
discuss the perspectives for active control of the laminar-to-turbulent
transition uctuations by electromagnetic eld modulation.

1 Boundary layers and boundary layer con-

trol
Whether a ow is laminar or turbulent, the eects of the viscosity of the
uid are greatest in regions close to solid boundaries. The region close to
1
the boundary, in which the velocity varies from zero, relative to the surface,
up to its full value, is called the boundary layer. The concept of boundary
layer introduced by Prandtl in 1904, was a most signicant advance in uid
dynamics, in the sense that it simplied the study by separating the ow in
two parts: (1) the region where velocity gradients are large enough to produce
appreciable viscous forces - the boundary layer itself - and (2) the external
region where viscous forces are negligible compared to other forces. From
the computational point of view the concept of boundary layer also plays
a signicant role because, rather than having to deal with time-consuming
general purpose nite-element codes, results of comparable precision may be
obtained by fast and relatively simple nite-dierence implicit algorithms.
When a uid ows past a solid body, an airfoil for example, a laminar
boundary layer develops, in general only for a very small distance near the
leading edge, followed by a transition to a turbulent boundary layer. Never-
theless, because near the solid wall velocity uctuations must die out, below
the turbulent region there always is a laminar sub-layer which in general is
very small (of the order of a micrometer). The transition from the laminar
to the turbulent region is controlled by the local Reynolds number, dened
in terms of the eective thickness of the boundary layer. It also depends on
the smoothness of the surface and on the external perturbations. The skin
friction drag is proportional to the gradient of the longitudinal velocity at
the solid boundary. Because of the mixing properties of the turbulent layer,
the gradient in the laminar sub-layer is much greater than the gradient at a
fully laminar layer. Therefore transition from a laminar to a turbulent layer
greatly increases the skin friction drag. Another eect to be taken into ac-
count is the separation of the boundary layer, which occurs at points where
the pressure gradient along the surface reverses sign. The eddies, generated
by the resulting reverse ow, disturb the ow and form a wake where energy
dissipation decreases the pressure, thereby increasing the pressure drag.
Because of the very large ratio between laminar and turbulent skin fric-
tion drag, much eort has been devoted to develop techniques to delay the
transition as a means of decreasing fuel consumption and noise. Care should
however be taken because, in general, a turbulent boundary layer is more
stable towards separation than a laminar boundary layer. Some of the active
control techniques that have been proposed include suction of slow-moving
uid through slots or a porous surface, use of compliant walls and wall cool-
ing (or wall heating for liquids). Injection of fast-moving uid, on the other
2
hand, is eective in avoiding separation but it increases turbulence. Most of
these ideas are fairly old (see for example [1],[2]) however, in view of their
interest for the applications and to obtain a more accurate characterization
of the physical mechanisms, studies of boundary layer control using these
aerodynamic methods are still, at present, being vigorously pursued (see for
example [3] [4] [5] [6] and papers in [7])
Another class of techniques for active boundary layer control consists in
acting on the ow by means of electromagnetic forces. Here dierent tech-
niques should be envisaged according to whether the uid is weakly conduct-
ing (an electrolyte like seawater or an ionized gas) or a good conductor (like a
liquid metal). Proposals for boundary layer control by electromagnetic forces
are also relatively old and trace its origin at least to the papers of Gailitis
and Lielausis[8], Tsinober and Shtern[9] and Moat[10] in the sixties. Inter-
est in these techniques has revived in recent years and some more accurate
calculations and experimental verications have been carried out, mostly in
the context of electrolyte uids[11] [12].
In this paper we will be concerned with the ow of air along an airfoil
when a layer of ionized gas is created on the boundary layer region. Local
ionization of the air along the airfoil is not practical from the technological
point of view, therefore we will assume that a stream of ionized air (or some
other ionized gas) is injected through a backwards facing slot placed slightly
behind the stagnation point (Fig.1). The body force that we consider to
be acting in the ionized uid is a longitudinal (along the ow) electric eld
created by a series of plate electrodes transversal to the ow and placed inside
the airfoil with the edges on the airfoil surface.
The emphasis of our study is on nding physically reasonable ranges of
parameters and analytic approximations that might lead to simple designing
procedures. For this purpose, before the numerical calculation of Section
3, we dedicate some time to the study of scaling solutions and analytical
approximations.
The provisional conclusions of our study are that it is possible to use
this technique to control the prole of the boundary layer laminar region.
With the rates of ionization that are needed and the injection method, it
is probably unrealistic to expect that the laminar region may be extended
over all the airfoil in normal (aviation) working conditions. Therefore this
method should be used in conjunction with methods for control of turbulent
boundary layers (riblets, large-eddy breakups, additives, etc.) in the rear part
3
of the airfoil. Also the injection of the stream of ionized gas in the leading
edge may, by increasing the velocity component normal to the airfoil, create
turbulence. Therefore it seems advisable to have a compensating suction
region after the injection slot. The ionization rate will also improve if the
gas extracted through the suction region is recycled through the ionizer.
Notice also that, once the uid in the boundary layer is ionized, large
scale velocity uctuations may be detected by a few local probes. This raises
the possibility of obtaining a negative feedback eect by an appropriate time-
dependent modulation of the electric eld. By controlling the growth of the
velocity uctuation in the transition region, a further extension of the laminar
region may be obtained. This is brie y discussed in the last section of the
paper.
The overall conclusion is that, when used in conjunction with other tech-
niques, as explained above, the method of boundary layer control by electric
elds might be interesting from the fuel consumption point of view. This
study was carried out as a preparation for an experiment being set up in our
Mechatronics Laboratory.

2 Ionized boundary layers with electric elds

2.1 The boundary layer equations

We use orthogonal curvilinear coordinates with xe parallel to the surface along

the ow and ye normal to the surface. If  is small ( denoting the curva-
ture and  the boundary layer thickness) the conservation and momentum
equations in the incompressible uid approximation may be written
@ ue + @ ve = 0 (1)
@ xe @ ye
!
@ ue + ue @ ue + ve @ ue = 1 @ pe + e @ 2 ue + @ 2 ue + 1 e (xe; ye)Ee (xe; ye) (2)
@ te @ xe @ ye em @ xe @ xe2 @ ye2 em e x
!
@ ve + ue @ ve + ve @ ve = 1 @ pe + e @ 2 ve + @ 2 ve + 1 e (xe; ye)Ee (xe; ye) (3)
@ te @ xe @ ye em @ ye @ xe2 @ ye2 em e y

ue and ve are the components of the uid velocity eld along the xe and ye
directions. em is the mass density, ee the electric charge density and Ee an
4
applied electric eld. The tilde denotes quantities in physical dimensions to
be distinguished from the adimensional quantities dened below. We consider
typical values Lr , r , Ur , r ,r ,r , Er as reference values for, respectively, the
airfoil width, the boundary layer thickness, the uid velocity, the uid mass
density, the kinematic viscosity, the uid charge density and the electric eld.
Then we dene the adimensional quantities
t = teULr ; x = Lx ; y = y ; u = Uu ; v = UvLr
e e e e
(4)
r r r r r r

m = m ; p =  pU 2 ; RL = Ur Lr
e e
(5)
r r r r

 =  ;  =  ; E = EE
e e e
(6)
r r r

In general RL >> 1. Neglecting terms of order R1L and r2 we obtain, for
L2r
stationary solutions @u @v
@t = @t = 0
@u + @v = 0 (7)
@x @y
@u + v @u = 1 @p + ! @ 2 u + 1 (x; y)E (x; y)
u @x (8)
x
@y m @x @y2 m
@p = r (x; y)E (x; y) (9)
y
@y Lr
5
 2
where ! = r2LRrL = Lr2rUrr and = LUr r2rEr r . Unless the electric eld component
normal to the airfoil is very large, one has @y @p  0 and the pressure term in
the second equation may be expressed in terms of the uid velocity ue far
away from the airfoil
@u @u @u @ 2u
u @x + v @y = ue @x + ! @y2 + 1 (x; y)Ex(x; y)
e
(10)
m
To take into account turbulence 0
eects one 0 0
should 0also replace in (10) the
velocity elds u and v by u + u and v + v , u and v being uctuation elds
with zero mean, u0 = 0, v0 = 0. The eect of the turbulent eld on the mean
ow is now obtained by taking mean values. In a two-dimensional turbulent
boundary layer the dominant eddy stress is u0 v0 . Assuming the eddy shear
stress u0 v0 and the mean rate of strain @u @y to be linearly related

u0 v0 =  @u
@y (11)
one obtains nally
!
@u @u @u @ @u
u @x + v @y = ue @x + ! @y  @y + 1 (x; y)Ex(x; y) (12)
e
m
with
 = 1 + ! (13)
being, in general, a function of y through the dependence of the eddy viscosity
 on the local velocity eld.  should be obtained from a turbulence model.
To analyze the scaling solutions and for the numerical calculations in Sect.
3 we dene a stream function and make the following change of variables
 1
u e 2 y
 = !  (x) (14)
1
= (ue!) 2  (x)f (x; ) (15)
u = @@y ; v = @@x (16)

6
The continuity equation (7) is automatically satised by (16) and one is left
with
! !
@  @ 2 f + @ f @ 2 f +  2 @ue + 2 @ 2 f @f @f @ 2 f =  2(x)(x; )E (x; )
x
@ @2 @x @2 ue @x @2 @x @ @@x u2e m
(17)

2.2 Scaling solutions

We assume that the electric eld to be created by a series of plate electrodes

along the z-direction, that is transversal to the uid ow. For this electrode
geometry the mean electric eld in the x-direction may be parametrized by
Ex = g(x) l2 (xl()x+) y2 (18)

where x and y are the adimensional coordinates dened in (4). E0 = gl((xx))

is the eld at y = 0, controlled by the potential dierences between the
electrodes, and l(x) is of the order of the electrode spacing. For thin (laminar)
boundary layers the eld Ex may with good approximation be considered to
be independent of y throughout the boundary layer thickness, as long as the
appropriate charge density prole is chosen (see below).
For the main application we are addressing, ionized air would be injected
through a slot near the leading edge of the airfoil, being then carried along
the airfoil surface by the ow. The steady-state charge distribution in the
boundary layer is obtained from the continuity equation
@ u + @ v = j (19)
@x @y
j being the source of electric charge. For a point source at the position
(x0 ; y0), that is j = c(x0; y0)(x x0 )(y y0), the solution is
(x; y) = c(x0; y0) ( (x; y) (x0 ; y0)) (x x0 ) (20)
being the stream function dened before. Then, for a column of ionized air
injected at a backwards facing angle through a slot placed at x0 , behind the
stagnation point, each point acts as a point source of intensity proportional
to the local uid velocity. Furthermore the intensity of the eective source
7
is depleted up the column. Taking the depletion eect into account, one
obtains by integration of Eq.(20)
(x; y) = 0 (1 d1 (x; y))  (1 d1 (x; y)) (x x0 ) (21)
0 is the injection intensity and d1 characterizes the rate of depletion. The
conclusion is that the charge density is maximum at the airfoil surface, de-
creasing to zero at a distance that depends on the uid dynamics and the
injection regime. In numerical simulations one may easily use the fairly
accurate equation (21) for the charge density prole. Here however, the
dynamically-dependent charge density prole will be parametrized by the
simpler formula  
(x; y) = 0 1 u u (22)
e
We now look for scaling solutions of (17). A scaling solution is one for
which f is only a function of . Eq.(17) becomes
2 e
f +   ff 00 + u @u =  2(x)0 1 f 0 g(x) 2 l(x) 2 2
 00 0   
e @x uem uel (x) + ! (x)
(23)
with boundary conditions
f (0) = f 0 (0) = 0 f 0 (1) = 1 (24)


where, for simplicity, we have denoted f 0  @f @
@ and  = @x .
Let the pressure be approximately constant for length scales L of the
order of the airfoil, that is @u@xe  0: Let also  be a constant. This is the
case for the laminar part of the boundary layer. Then the factorized nature
of Eq.(23) implies that solutions exist only if

 2 (x) = 2  (x) = 1 g (x) = c4 l 2 (x) (25)
c1  (x) c3  (x)
c1, c3 and c4 being constants. Therefore
p
 (x) = c1 x + c2
g(x) = c(x3 ) (26)
l(x) = pc4  (x)
8
There are two physically interesting situations. The one with c1 6= 0 c2 = 0
and the one with c1 = 0 c2 6= 0. The rst one corresponds to a boundary
layer starting at x = 0 and growing with x 12 and the second to a constant
thickness boundary layer. The rst one corresponds to an equation
f 000 () + 21 f ()f 00 () + 1 f 0 () '2 '+1 2 = 0
 
(27)
2
p q
with c1 =  , c2 = 0 , '1 =  0 c3 c4 , ' = ue c4 and the second to
ue m ! 2 !

f 000 () + 1 f 0 () b2 +a 2 = 0

 
(28)
p q
with c1 = 0 , c2 6= 0 , a =  0 c3 c4 , b = ue c4 .
ue m ! !
In the rst case one chooses c2 = 0 to obtain a boundary layer starting at
x = 0. The scaling hypothesis requires then an electric eld that is singular
at x = 0, y = 0 (Ex  x 1 ). In any case this electric eld solution is not very
interesting
1
for our purposes because it leads to a boundary layer growth of
x , as in the free force Blasius solution. Therefore it will be more interesting
2
to consider a small eld free region in the leading edge of the airfoil and match
the Blasius solution there with the constant thickness solution of Eq.(28).
Gailitis and Lielausis[8] have also obtained a theoretical solution of con-
stant thickness. However they consider a dierent force eld distribution and
no dependence of the uid charge density on the boundary layer dynamics.
Therefore their boundary layer prole has a very dierent behavior.
The solution of Eq.(28) is easily obtained by numerical integration (see
below). Notice however that with the replacement
( ) = 1 f 0 ( ) (29)
and choosing c4 = ! ue , which is a simple rescaling of  , Eq.(28) becomes the
zero-eigenvalue problem for a Schrodinger equation in the potential a=(1+2 ),
00 () + 1 +a 2 () = 0 (30)
One may use the well-known WKB approximation to obtain
2 4 1
f () = 1  (1p+  ) pa
0
(31)
 + 1 + 2
9
Eq.(31) is a very good approximation to the exact solution for a  1 (see
Fig. 2). Fig.3 shows the eective boundary layer thickness as a function of
a. The eective boundary layer thickness  is dened here as the value of 
at which the velocity u reaches 0.95 of its asymptotic value ue. A very fast
thinning of the boundary layer is obtained (several orders of magnitude) for
a relatively short range of the a parameter. Fig.3 shows the variation of 
for small a. For large a (and small ) one has the asymptotic formula
 ' 2:p
9957
a
which is obtained from Eq.(31).
If the longitudinal electric eld Ex is assumed to be a constant (E0 )
throughout the boundary layer thickness, with the same charge prole, the
solution is even simpler, namely
p
f 0 ( ) = 1 e  h (32)
with  = pc2 and
h = cu220 E0 (33)
e m
Again, since  is a constant, this is not fully realistic because it leads to a
constant thickness boundary layer.
For reference values of the physical quantities in Eqs.(4-6) we take
Ur = 100 m s 1
Lr = 1 m
r = 10 3 m
r = 1:2 Kg m 3 (34)
Er = 500 V cm 1
r = 15 C cm 3
r = 1:5  10 5 m2 textnormals 1
For these reference values, the adimensional constants ! and dened after
Eq.(9) are
! = 0:15 (35)
= 62:499

10
 For comparison we mention that in the classical force-free Blasius solu-
tion, and for these reference parameters, the y coordinate y corresponding
to  (that is, the point at which uue = 0:95) is
p
y = 1:55  10 3 x (36)
Stability of a laminar boundary layer cannot safely be guaranteed for local
Reynold numbers greater than about 103. Therefore requiring
f e
RS = ueey ' 103 (37)
one obtains, for the reference parameters, ye ' 0:15 mm. Using (34) the
conclusion is that, for these parameters, the laminar part of a force-free
boundary layer is only of the order of 1 cm, just a tiny portion of a typical
wing.
Now we use the scaling solutions (32) and (31) to obtain an estimate of
the eects of a longitudinal electric eld. For the constant eld case (32)
from p
f 0 ( ) = 1 e  h = 0:95
and s
y =  c2 !
  p
e u = 0:15
using (33) one obtains
0 = 0:957
That is, to insure a constant thickness boundary layer with local Reynolds
number RS = 103 (at the point where uue = 0:95), one needs a charge density
e0 at y = 0, in physical units (and for the reference values of the kinematical
parameters)
e0 = 0 r = 14:36 C cm 3
For the variable eld case (31) the estimate depends on the separation of
the electrodes. Taking l(x) = 10, that is an electrode separation of the order
of one centimeter, and the references values for all quantities except for the
charge density (namely E0 = gl((xx)) = 1, ue = 1, etc.) one obtains c4 = 0:15,
g(x) = 10, c3 = g(x) (x) = 258:2, pc2 =  (x) = 25:8, and requiring
y = pc2 !
q
ue = 0:15
0 
f ( ) = 0:95
11
one nally obtains a = 39887:77 leading to
0 = 0:957
the same estimate as above. The large value of a that is obtained shows
that the WKB expression (31) is a good approximation for physically inter-
esting parameter values. On the other hand the fact that the same charge
density estimate is obtained both in the constant-eld and the variable-eld
cases, shows that it is realistic to consider the eld as approximately con-
stant throughout the laminar boundary layer thickness, as long as a variable
charge prole (21) or (22) is used.
The above estimates were obtained using the reference values for the
kinematic variables. For other values we have the following designing formula
(in normalized units)
3
0 E0 = 0:957 10ue6Rm2  (38)
S

3 Numerical results
For the numerical solution of Eq.(17), with  given by Eq.(22), we use an
implicit nite-dierence technique ([14] - [16]). Dene F (x; ) by
F (x; ) = @f @ (39)
and  
a1 = 1 @ @ +  @ f +  2 @f
@x @x
a2 = u2e m  2Ex0
(40)
a3 = 2 F
a4 = u2e  m   2Ex0 + ue2 @u@xe
Then Eq.(17) becomes
@ 2 F + a @F + a F + a @F + a = 0 (41)
@2 1 @ 2 3 @x 4

The derivatives are replaced by nite-dierence quotients with a variable grid
spacing concentrated near  = 0, where F changes more rapidly. Let k > 1
12
be the ratio between two successive grid spacings in the  direction.
k = i+1  i
i i 1
Then  2  Fi+1;j+1 +kFi+1;j 1 (1+k)Fi+1;j
@F
@ 2 i+1;j = 2
2
  F k 2 Fi+1;j 1 (1 k2 )Fi+1;j
@F
@ i+1;j = 
i+1 ;j +1

1
@F F ;j Fi;j
@x i+1;j =
i+1

x

1 = j+1 j + k2 (j j 1)

2 = (j+1 j )2 + k(j j 1)2
Substitution in Eq.(41) yields
Aj Fi+1;j+1 + Bj Fi+1;j + Dj Fi+1;j 1 + Gj = 0 (42)
with
Aj =
22 +
a11
Bj = 2(1+ k) a1 (1 k2 ) + a + a3

2
1 2 2
x
2 k
Dj =
2 a
1 k1
Gj = a4 a3 F
i;jx
The boundary conditions at  = 0 and  ! 1 are known
f (i; 1) = F (i; 1) = 0
F (i; N ) = 1
where N is the largest label of the grid, in the  coordinate, chosen to be
suciently large.
Because of the tridiagonal nature of (42) the solution in the line i + 1 is
obtained by the two-sweep method, the recursion relations being
Fi+1;j = j Fi+1;j+1 + j
j = Bj +DAjjj 1
j = BGjj++DDjjjj 11
with 1 = 0 and 1 = 0.
To start the integration process there are basically two methods. In the
rst the integration is performed from left to right in the x coordinate with
13
the grid extended to the left of the airfoil, where the ow is known. With
the solution known in the line i, the coecients Aj to Gj for Eq.(42) are
computed at the point (i; j ). Notice that f (i; j ) is obtained by integration of
the solution F . Z
f (i; ) = F (i;  )d
0
The integration now proceeds along the lines, from left to right. After a
complete pass the process is restarted using now for the calculation of the
coecients Aj , Bj , Dj and Gj the old values of F at (i + 1; j ). The process
is repeated several times until the solution stabilizes.
In the second method, which is the one we actually use, the integration
process starts from an approximate solution. The scaling solutions derived
in Sect.2 are particularly useful for this purpose.
For our calculations we considered an electric eld parametrized as in
Eq.(18), namely
ue l2
Ex = E0 ue l2 !2 2
! +  (x)
with ue!l2 = 666:66 which corresponds to l = 10, ue = 1 and ! = 0:15.
Notice that for these parameters, as pointed out in Sect.2, the electric eld
has only a small variation throughoutp the boundary layer region. For the
scaling function we take  (x) = x and consider  = 1. Then all results
depend only on the variable S
S = 62:1499 u2  0 E0
e m
(S = 1 when all quantities take the reference values).
In Fig.4 we show a contour plot of the numerical solution for f 0 (x; )
(= uue ) when S = 0:6. From the x dependence of the numerical solutions we
may compute the eect of the electric eld in extending the laminar part of
the boundary layer. By dening, as in Sect.2, the length of the laminar part
as the x coordinate corresponding to a local Reynolds number of 103 and
denoting by x0 ( uue = 0:95) the force-free value we have obtained for the ratio
R = xx
0
the results shown in Fig.5. For S = 0 we obtain the Blasius solution and
as we approach S = 0:957, corresponding to the scaling solution, the ratio
14
diverges. The matching of the results in the force-free and scaling limits is a
good check of the numerical algorithm. A clear indication of the results in
Fig.5 is that not much improvement is obtained unless one is able to obtain
ionization charge densities of the order of the reference value r .

4 Discussion and conclusions

# In this paper we have concentrated on controlling the prole of the bound-
ary layer. The prole has a direct eect on the laminar or turbulent nature
of the ow which, in a simplied manner, we estimated by a local Reynolds
number (37) dened as a function of the eective thickness. Another relevant
aspect, of course, is the active control of the transition instabilities that can
be achieved by electromagnetic body forces on the charged uid.
Turbulence and transition to turbulence are three-dimensional phenom-
ena. However, for the large scale small amplitude (Tollmien-Schlichting)
uctuations, that rst appear in the transition region, a two-dimensional
model is a reasonable approximation. In Eqs.(1-3) we2 make as before the
change of variables (4-6), neglect terms of order R1L , Lr2r and Lrr and split the
velocity and electric elds into
u = u + u00
v =v+v 0
Ex = Ex + Ex
where u , v , Ex are the steady-state solutions and u0 , v0 , Ex0 the time-
dependent components. Because of the continuity equation
@u0 + @v0 = 0
@x @y
we may dene a uctuation stream function 
u0 = @
@y ; v 0
= @
@x

Now we assume the uctuation to be a (small-amplitude) wave-like function

of x, y and t
(x; y; t) = F (y)ei(x t)
15
The imaginary parts of  and  control, respectively, the growth rates of
temporal and spatial uctuations. The (modulation) electric eld is assumed
to have a similar form
Ex0 = Eei(x t)
One now obtains
!
@u
iu + @x i @F i @u F + v @ 2 F = ! @ 3 F + E
@y @y @y2 @y3 m
The conclusion is that a space-time modulation of the electric eld, with
the appropriate phase, is equivalent to an eective viscous damping eect
which delays the growth of the transition region instability. For this to be
eective one needs to detect the phase of the wave instabilities by electromag-
netic probes. Absolute synchronization of the feedback electric modulation
is however not so critical as in acoustic noise cancelation, because here the
objective is only to obtain an eective damping eect. The simplied treat-
ment of the transition instabilities is justied by the fact that it is only for
the small amplitude large scale uctuations that one may hope to be able to
detect the phase with some reasonable accuracy.
# The kinematic reference parameters dened in (34) correspond to typ-
ical aviation conditions. The conclusion, both from the scaling solutions in
Sect.2 and the numerical results in Sect.3, is that, to obtain a signicant
controlling eect on the boundary layer by this method, the charge den-
sity 0 (at y = 0) should be of the order of the reference charge density
(r = 15C cm 3). This charge density corresponds to about 50 times the
ion concentration a few centimeters away from the emitter of a commercial
table-top negative corona discharge air purier with a power of less than 6
watts. Therefore, it seems technically feasible to achieve a signicant bound-
ary layer control by this method. Another possibility would be to use, instead
of air, some other easier to ionize gas. This could then be partially recovered
and recycled by suction.
# As explained in the introduction and because of the perturbation in-
duced by the injection method, it seems advisable to use this method in
conjunction with suction and passive control in the rear part of the airfoil.
Even if a fully laminar boundary layer may never be completely achieved,
just remember that any small improvement becomes, in the long run, quite
signicant in terms of fuel consumption.
16
 # The formula (38), derived from the scaling solutions, provides rough
design estimates. Better control over design parameters we hope to obtain
from the experimental work.

5 Figure captions
Fig.1 Airfoil transversal cut showing ionized air injection, suction pump and
plate electrodes.
0
Fig.2 Exact ( ) and approximate (


) constant thickness scaling solution
f ().  
Fig.3 Eective boundary layer thickness  f 0 ( ) = uue = 0:95 for the
constant thickness scaling solution.
Fig.4 Contour plot of f 0 (x; ) for S = 0:6.
Fig.5 Ratio of boundary layer laminar regions with and without electric
eld control.

References
[1] H. Schlichting; "Boundary-Layer Theory" 6th. edition, MacGraw Hill,
New York 1968.
[2] A. D. Young; "Boundary Layers", BSP Professional Books, Blackwell,
Oxford 1989.
[3] N. M. El-Hady; "Eect of suction on controlling the secondary instability
of boundary layers", Phys. Fluids A3 (1991) 393.
[4] R. D. Joslin and P. J. Morris; "Eect of compliant walls on secondary
instabilities in boundary-layer transition", AIAA Journal 30 (1992) 332.
[5] R. A. Antonia and Y. Zhu; "Eect of concentrated wall suction on a
turbulent boundary layer", Phys. Fluids 7 (1995) 2465.
[6] A. D. Lucey and P. W. Carpenter; Boundary layer instability over com-
pliant walls: Comparison between theory and experiment", Phys. Fluids
7 (1995) 2355.

17
 [7] D. M. Bushnell and J. N. Hefner (Eds.); "Viscous drag reduction in
boundary layers", Progress in Astronautics and Aeronautics, vol. 123,
American Institute of Aeronautics and Astronautics, Washington 1990.
[8] A. K. Gailitis and O. A. Lielausis; "On the possibility of drag reduction
of a at plate in an electrolyte", Appl. Magnetohydrodynamics, Trudy
Inst. Fis. AN Latv. SSR 12 (1961) 143.
[9] A. B. Tsinober and A. G. Shtern; "Possibility of increasing the ow
stability in a boundary layer by means of crossed electric and magnetic
elds", Magnetohydrodynamics 3 (1967) 103.
[10] H. K. Moat; "On the suppression of turbulence by a uniform magnetic
eld", J. Fluid Mech. 28 (1967) 571.
[11] A. Tsinober; "MHD ow drag reduction", in [7], page 327.
[12] C. Henoch and J. Stace; "Experimental investigation of a salt water
turbulent boundary layer modied by an applied streamwise magneto-
hydrodynamic body force", Phys. Fluids 7 (1995) 1371.
[13] C. B. Reed and P. S. Lykoudis; "The eect of a transverse magnetic
eld on shear turbulence", J. Fluid Mech. 89 (1978) 147.
[14] F. G. Blottner; "Finite dierence methods of solution of the boundary-
layer equations", AIAA Journal 8 (1970) 193.
[15] R. T. Davis; "Numerical solution of the hypersonic viscous shock-layer
equations", AIAA Journal 8 (1970) 843.
[16] H. H. Hamilton II, D. R. Millman and R. B. Greendyke; "Finite-
dierence solution for laminar or turbulent boundary layer ow over
axisymmetric bodies with ideal gas, CF4 or equilibrium air chemistry",
NASA Langley Research Center report no. L-17102, 1992.

18