GENERAL I ARTICLE

Ludwig Prandtl and Boundary Layers in Fluid

Flow
How a Small Viscosity can Cause large Effects

Jaywant H Arakeri and P N Shankar

In 1904, Prandtl proposed the concept of bound-

ary layers that revolutionised the study of fluid
mechanics. In this article we present the basic
ideas of boundary layers and boundary-layer sep-
aration, a phenomenon that distinguishes stream-
lined from bluff bodies.
Jaywant Arakeri is in the
Mechanical Engineering A Long-Standing Paradox
Department and in the
Centre for Product It is a matter of common experience that when we stand
Development and in a breeze or wade in water we feel a force which is
Manufacture at the Indian called the drag force. We now know that the drag is
Institute of Science, caused by fluid friction or viscosity. However, for long it
Bangalore. His research is
mainly in instability and
was believed that the viscosity shouldn't enter the pic-
turbulence in fluid motion. ture at all since it was so small in value for both water
He is also interested in and air. Assuming no viscosity, the finest mathematical
issues related to environ- physicists of the 19th century constructed a large body
ment and development.
of elegant results which predicted that the drag on a
body in steady flow would be zero. This discrepancy
between ideal fluid theory or hydrodynamics and com-
mon experience was known as 'd' Alembert's paradox'
The paradox was only resolved in a revolutionary 1904
paper by L Prandtl who showed that viscous effects, no
matter how'small the viscosity, can never be neglected.
P N Shankar does
research in fluid mechan- More precisely, it is the Reynolds number Re, a dimen-
ics at the National sionless measure of the relative importance of inertial
Aerospace Laboratories. to viscous forces in the flow, which is the determining
His other interests factor. Prandtl postulated that for certain kinds of high
include music, sports,
Reynolds number or nearly frictionless flows, for exam-
science and reading of an
omnivorous character. ple the flow past a streamlined body like an airfoil, the
viscous effects would be confined to thin regions called
boundary layers. For certain other kinds of high Re

--------~--------
48 RESONANCE I December 2000
 GENERAL I ARTICLE

flows, such as the flow past a bluff body like a sphere, d'Alembert's
viscous effects need not be confined to such thin layers;
paradox was
viscosit.y t.hen has a more dramatic effect than what its
resolved by
low value might suggest. The key concept of boundary
Prandtl who
layers has now spread to many other fields; boundary
showed that
layers often arise in what are known as singular pertur-
viscous effects, no
bation problems.
matter how small
In this article we illustrate the boundary-layer concept the viscosity, can
by considering flows around three representative bodies, never be
namely a thin plate aligned with the flow, an airfoil and neglected.
a circular cylinder.
Viscous Stress and the No-slip Condition
It is useful when studying fluid motion to consider the
motion of a fluid particle or a small element of fluid.
Although forces such as that due to gravity are at times
important we ignore them here. For the purposes of this
article the only forces that we will consider are those due
to pressure and viscosity. These forces can accelerate or
decelerate a fluid particle.

For common fluids the viscous force is proportional to

viscosity x rate of deformation of fluid element, or vis-
cosity x spatial gradient of velocity.

Another important point relevant to the boundary layer

is the no-slip condition at a solid wall: the fluid right
next to a solid wall has the same velocity as the wall
(Figure 1).

(a) (b)
line of dye at Figure 1. Illustration of the
initial time line of dye at
flow flow short time later no-slip condition. Fluid is
flowing past a stationary
solid surface. The figure
shows how an initially ver-
tical/ine ofdye is displaced
at a later time by the flow.
The fluid and the dye next
wall wall to the wall do not move.

-RE-S-O-N-A-N-C-E--I-D-e-ce-m-b-e-r-2-0-0-0---------~~----------------------------4-9
 GENERAL I ARTICLE

Figure 2. The figure shows

t
the linear velocity variation
in a fluid between a station-
ary bottom plate and a top 71l
plate moving with velocity
_7
Uo' The force per unit area
y 7
/u(y) h
to be applied on the plates
is'L
17 o
1

On the other hand, in ideal or non-viscous flow the fluid

next to a solid surface can 'slip' past it.
To illustrate these two points, consider a fluid between
parallel plates, with the lower plate stationary and the
upper plate moving with velocity Uo (Figure 2). By
the no-slip condition, the fluid next to the lower plate
has zero velo.city and the fluid next to the upper plate
has velocity Uo; if the gap between the plates is small
enough, the fluid velocity 7J,(y) varies linearly from zero
to Uo, 7l.(y) = UoY / h.
The viscous stress on either plate is /kd7J,/ dy = /kUo/ h.
The larger the shear Uo/ h, the larger the force required
to move the plate.
Prandtl's Resolution of the Paradox: Flow past
a Thin Plate
The key concept of
boundary layers Now consider our first example of a thin flat plate placed
has now spread to in a steady uniform flow, with velocity Uoe and pressure
many other fields; Poe; the plate is aligned with the flow. If the fluid were
boundary layers ideal, i.e., frictionless or without viscosity, and the plate
often arise in what was of negligible thickness, the flow would be undis-
are known as turbed (Figure 3a). By Bernoulli's equation the pres-
singular sure on both sides of the plate would be Poe. The drag
perturbation
on the plate would be zero.
problems. When the fluid has even a small viscosity the no-slip

-so--------------------------------------------~-------------------------------------------
RESONANCE I December 2000
 GENERAL I ARTICLE

a) Figure 3. Uniform flow with

UOC,P CXl
velocity Uco past a thin flat
plate. (a) The flow is undis-
turbed if the fluid is non-
viscous. (b) For a fluid with
small viscosity thin bound-
ary layers develop on ei-
A B ther side of the plate. The
boundary layers become
turbulent beyond a certain
distance from the front
edge. The fluid velocity var-
b) ies sharply from zero at the
Ux-
surface to UCO across the
boundary layer.

Note: Boundary layer on only

U~,PCl'; the upper surface is shown
and the vertical scale is highly
exaggerated.

.. laminar .!
'
transi tion l'
zone
. turbulent ..

boundary condition changes the situation totally. Imag-

ine a fluid particle which impinges on, say, the top sur-
face. The part of the fluid particle that touches the
surface gets 'stuck' to the surface while the rest of the
particle keeps moving forward. The particle is highly
sheared and the viscous forces are large. Just past the
leading edge, only the fluid next to the wall is slowed
down. Further down the plate the random motion of the
molecules slow down adj acent fluid layers away from the
wall - the retarding effect of the wall is spread outward
by viscous diffusion of momentum.
For a given length of plate the region over which vis-
cous action is felt is thin if the viscosity is small, that
is, when the Reynolds number is large. This thin viscos-
ity affected region is called the boundary layer; outside
the boundary layer the flow is practically non-viscous.

-RE-S-O-N-A-N-C-E--I-D-e-c-em---be-r--2-00-0----------~------------------------------------~
 GENERAL I ARTICLE

Within the boundary layer the velocity gradient normal

to the wall is large; the large velocity gradient multi-
plying a small viscosity gives a non-negligible viscous
force.
Since the viscosity is small and the boundary layer is
thin, Prandtl drew three important conclusions:
l. The flow outside the boundary layer is practically
unaffected and is almost the same as that predicted by
ideal flow theory without the boundary layer.
2. There is negligible variation of pressure across the
boundary layer, i.e., pressure on the surface ~ pressure
at the edge of the boundary layer.

3. The flow in the thin boundary layer could be dealt

with by simplifed boundary-layer equations.
Thus in the case of the flow over a thin plate the velocity
just outside the boundary layer is U 00 and the pressure
both in and outside the boundary layer is constant and
equal to Poo . There is a great simplification in what
needs to be computed. Instead of having to solve the
complicated N avier-Stokes equations, one has to now
solve the simpler boundary-layer equations (see Box 1).
What has happened is that whereas, when J1 ---+ 0, we
can treat the major portion of the flow as inviscid, we
still need to account for the crucial role of viscosity in
bringing the fluid to rest at the boundaries; moreover the
inviscid outer flow and the boundary layer must be cor-
rectly matched. For the mathematically minded reader
a simple model problem which illustrates this procedure
Since the is discussed in Box 2.
boundary layers It is now possible with the simplified procedure to esti-
are thin the Inate the way the boundary layer grows along the plate.
pressure transmits The boundary layer thickness, 6, grows with the down-
unchanged stream distance by viscous diffusion; the larger the vis-
through the cosity, the faster the diffusion. We observe that fluid is
boundary layer. carried downstream by the flow with velocity approx-

--------~--------
52 RESONANCE I December 2000
 GENERAL I ARTICLE

Box 1. The Equations Governing Fluid Flow and their Approximate Forms.

Here we summarize for the mathematically minded reader the relevant two-
dimensional field equations in cartesian coordinates. In order to simplify
matters we only consider the flow over a thin flat plate aligned with the
flow. Let p and f1 be the density and viscosity of an incompressible fluid
and let ('11" v) be the (x, y) velocity components and p be the pressure. The
non-linear Navier-Stokes equations, which govern viscous fluid motion then
take the form
'/J. x + Vy = 0 (la)
+ V11. y ) =
p('/J.11. x -Px + f1(V,xx + 11. yy ) (lb)
p(v,vx + vV y ) = -Py + f1(V xx + V yy ) (Ie)

where the subscripts indicate partial differentiation. The first of the above
equations is the continuity or the conservation of mass equation. The next
two are the x and y momentum equations, respectively. For a viscous fluid
1J, and v have to vanish on stationary solid boundaries.

The Euler equations which describe inviscid or frictionless motion can be

obtained by just setting f1 = 0 in the above equations. Note that the highest
derivatives in the second and third equations are lost. Correspondingly, now
only the velocity component normal to a stationary solid boundary has to
vanish.
Prandtl's boundary layer equations follow from a careful simplification of
(1)
U x + Vy = 0 (2a)
p(v.v. x + vU y ) = -Px + f11J,yy (2b)
Py = O. (2e)

The continuity equation remains unchanged, the highest x-derivative has

been dropped in the x-momentum equation while the y-momentum equation
has been considerably simplified to a statement that the pressure is constant
across the boundary layer and is determined by the external inviscid flow.
Note that the boundary layer equations remain non-linear.

-RE-S-O-N-A-N-C-E--I-D-e-ce-m-b-e-r-2-0-0-0---------~-----------------------------~-
 GENERAL I ARTICLE

Box 2. A Model Boundary Layer Problem

To illustrate to the boundary layer concept we consider a function u(y) defined

on the interval 0 ~ y ~ 1. Let us assume that u(y) satisfies the simple ordinary
differential equation

d 2 71. du
f -2 + (1 + f)- + u = 0, 71(0) = 0, u(l) = 1. (1)
dy dy

If we wish to draw an analogy with the fluid flow equations, we may consider
f, U and y to be analogous to the viscosity, strean1wise velocity and the direction

nonnal to the plate, respectively. Note that the E in the coefficient of the middle
term is there only to silnplify the solution. The reader must be warned however
that the analogy is only a very crude, qualitative one. In any case, since the
differential equation is a linear one with constant coefficients it can be solved
exactly by, for example, assun1ing exponential solutions. The exact solution
which satisfies the above boundary conditions is given by

e-Y - e- Y/ E
u(y)= e -1 - e -1/' (2)
E

Now note that when E -r 0 (i.e. the viscosity vanishes)-l/E -r -00 and for
y > 0 -yiE -r -00; as a consequence for every positive y, u(y) -r e 1- y . This
is the outer solution. As E -r 0 this solution is a very good approximation to
the exact solution over most of the field, 0 (e) < y ~ 1. But it does not satisfy
the boundary condition at y=O since it gives the value e for u instead of O. This
pathology immediately suggests that although the outer solution is valid almost
everywhere a boundary layer is required near y = o.

We now outline the boundary layer analysis of (1) as E -r O. First, to directly get
the outer solution we let f -r 0 in (1) and get the equation

duo
- +710 = O. (3)
dy

for the outer solution uo(Y). Note that the second derivative term has been
dropped as has the small term multiplying the first derivative. Equation (3) can
be easily integrated and made to satisfy the boundary condition at x = 1. We
then get the outer solution: 71 o (y) = e 1 - y , just as we had found from the exact
solution.
Box 2. continued...

-54-----------------------------~------------------------------
RESONANCE I December 2000
 GENERAL I ARTICLE

The trickier part, the Prandtl part, is to get the boundary layer or inner solution.
We know from our discussion above that although the outer solution goes toe as
y -+ 0, the correct solution must actually go to zero very rapidly in a thin layer
of thickness Eo This suggests scaling or magnifying the region close to y = O. Let
'" = YIE be the inner variable; note that for fixed y , '" -+ 00 as € -+ 0 and so '"
tends to magnify y. Now we assume that the inner solution Ui(",) isa fUnction of
the inner variable", alone. Note that now -Jy =€-l!:ry and ~ ::::: E-2 £,.
Now if
Ui("') is assumed to satisfy (1) with", defined as above, and if the substitutions
are made and the leading terms in € alone are collected we find

d 2 u'
+ _duo
1 . = O.
(4)
-;;;;r
1,
d",

This can be imnlediately integrated twice to yield Ui(",) = Ao - Ale-71 where Ao

and A 1 are arbitrary constants. Since the inner solution has to satify the zero
. boundary condition at y = 0 (Le. '" = 0), Al = Ao and so Ui("') = Ao(l - e-71 ).
Now how do we connect the inner solution Ui("') to the outer solution uo(Y)?
This is done by matching the two solutions. Since U o is expected to hold in the
outer region (roughly f: < y S 1) while Ui is expected to hold in the inner region
( roughly 0 :::; y < f) and since both of them represent U we would expect ·them
to overlap or m,atch in some common region of validity. This suggests trying the
matching condition
lim uo(y) = lim u1,(",). (5)
y-,o 71-+00

1.00 l
Figure A. Comparison of the exact y ~
solution u(y) (--) with the inner and
outer solutions u, (y) (---) and u/y)
(- - - -) for two valu(!s of & • Note that
as & becomes smaller the perturba-
j
tion solution becomes better. Com-
pare with Figure 5. 0.00 1
0.0
i "

Box 2. continued...

-R-ES-O-N-A-N-C-E--I-o-e-c-em--be-r--2-00-0---------~~-------------------g-
 GENERAL I ARTICLE

This condition immediately leads to Ao = e. Thus we find the solution

(6a)

The two solutions can be combined to give a composite solution approximately

valid everywhere but we shall not do that here. We just note that when we
compare the approximate solution (6) with the exact solution (2), we find that
the boundary layer analysis correctly picks up the boundary layer near y = 0,
while the outer solution picks up the correct solution near y = 1 when E ~ O.
This is also shown very clearly in Figure A.

mately equal to Uoo . How fast the layer diffuses and

how fast the fluid is carried downstream together deter-
mine the boundary layer thickness. The boundary-layer
thickness 8 on a flat plate depends on the distance x
from the front or leading edge of the plate and on the
condition of the flow. If the flow is laminar or smooth,
steady and orderly, as it will initially be near the leading
edge, 8(x) is given by

The random where p is the fluid density and c is a constant. In this

motion in a situation, a measure of the relative importance of viscos-
turbulent boundary
ity is the Reynolds number Rex = pUoox/ /1; the smaller
the viscosity, or the higher the Reynolds number, the
layer enhances the
slower is the growth of the boundary layer. Beyond
momentum
a distance corresponding to Rex greater than 1 x 106
diffusion rate to
the boundary layer becomes turbulent. The fluid flow
many times more
within the boundary layer becomes chaotic; the fluid el-
than the viscous or
ements move about randomly in addition to their bulk
molecular diffusion
downstream motion. This random motion enhances the
rate obtained in the
momentum diffusion rate to many times more than the
laminar boundary
viscous or molecular diffusion rate obtained in the lam-
layer. inar boundary layer. The boundary layer grows more

-56-----------------------------~------------------------------
RESONANCE I December 2000
 GENERAL I ARTICLE

rapidly and nearly linearly with the distance x.

The main effect of viscosity, however small, is to cause

drag which is absent in ideal flow. The drag on the
plate is entirely due to the tangential stress at the plate
surface and is given by J.l( 8v,j 8y )1/=0. This tangential
stress at the surface is called wall shear stress and is
denoted by 1"'I}J. In the laminar boundary layer

Tw ~ 3/2jE!lP
0.332Uoo - ex /-LU oo / 8.
x

Tw is nlaximum near the leading edge and decreases as

x- 1 / 2with distance.

On the other hand, when the boundary layer becomes

turbulent the random motion tends to make the velocity
over more of the boundary layer closer to the free stream
speed U 00; that is the velocity profile is Inuch flatter. As
a consequence the decay to the no-slip condition occurs
over a slnaller distance, leading to a larger wall shear
stress compared to t.he laminar value. In the turbulent
boundary layer T w = C pU~/2; C depends only weakly
on x and normally lies between 0.002 and 0.004 for a
smooth surface.
Flow over an Airfoil

Next, as an example of a streamlined body consider

uniform steady flow past an airfoil. Figure 4 shows a The lift force on an
streamline picture of ideal flow; and the picture, would airfoil, which is
not look very different if the fluid had a small viscosity. mainly caused by
Consider the motion of a fluid particle along a stream- pressure
line just above the top surface of the airfoil. The fluid difference between
particle starting with the velocity Uoo ahead of the airfoil the top and bottom
slows down initially; its speed then increases rapidly be- surfaces, predicted
yond U 00 near the nose, reaches a maximum near about using ideal flow
the rnaximum thickness point, and slowly comes back theory is close to
to about U oo near the trailing edge of the airfoil (see the measured
Figure 4); call this velocity just above the airfoil surface force.

-RE-S-O-N-A-N-C-E-\--D-eC-e-m-b-er--2-00-0---------~------------------------------
 GENERAL I ARTICLE

b)

Figure 4(a) Streamlines in

non-viscous flow past an Ue . By Bernoulli's equation the pressure on the upper
airfoil. (b) The speed varia- surface is, Pe = Poo +l/2p(U!-U;); where p is the fluid
tion along a streamline that density. Since Ue > U00 over most of the upper surface,
goes just above the airfoil
the pressure there is less than the free stream pressure.
surface.
Since the opposite happens on the lower surface there is
a net upward or lift force on the airfoil.
When the fluid is viscous the velocity on the airfoil, by
the no-slip condition, has to vanish. There is one dif-
ference between the boundary layer on the flat plate
and the boundary layer on the airfoil: in the fiat plate
case, the velocity on the edge of the boundary layer is
a constant U00 , whereas in the airfoil case the veloc-
ity at the edge of the boundary layer is variable and
nearly equal to Ue . Figure 5 shows the velocity profiles
at one station on an airfoil for ideal and boundary layer
flow. Since the boundary layers are thin the pressure
transmits unchanged through the boundary layer. The

Figure 5. Velocity variation

near the surface in (a) a
non-viscous fluid and (b) a
fluid with low viscosity. The low-viscosity fluid
two velocities are different
only in the thin boundary
layer.

--------~--------
58 RESONANCE I December 2000
 GENERAL I ARTICLE

t~\,·--···.L.,,~~.~.~,:~~>:~~l
112 P Uoc2
'>"''"1
/'. . +
....."...,/".r

/'
,...,.;:i4- upper surface

Figure 6. A picture of flow past an airfoil. The

flow, which is from right to left, is made visible
by tiny particles in the fluid. The boundary
layers are too thin to be seen. The drag is
mainly caused by tangential viscous stress on
the surface.
(Reproduced from G K Batchelor. An Introduction to Assuming ideal flow
Fluid Dynamics, Cambridge University Press, 1967.) Measured

streanlline picture (Fig'uTe 6) and t.he surface pressure Figure 7. The pressure dis-
tributions on the upper and
(Figure 7) are very dose to those obtained in the ideal
lower surfaces of an airfoil.
flow case. That is why the lift. force, which is lllainly
There is negligible differ-
caused by pressure difference between thc top and bot- ence between the pressure
tOln surfaces. predicted using ideal flow theory is close distribution calculated as-
to the rneasured force. suming non-viscous flow
and the measured pressure
On the ot.her hand, the drag is caused by viscous stresses distribution in flow ofa low-
at the wall and canllot be predicted just based on an viscosity fluid. The higher
ideal flow calculation. But a calculation based on bound- pressure on the lower sur-
ary layer t.heory is effective. face and the lower pres-
sure on the upper surface
Flow around Bluff Bodies: Boundary Layer Sep- result in lift.
aration.

Even a sillall value of viscosit:y has a dralnatic effect on

the flow around a bluff body such as a circular cylinder
or a sphere ,vhen conlpared to the ideal, zero viscos-
ity, flow. Fig'uTc 8 shows the streanlline picture of ideal
flow over a cylinder and the velocity variation along a
strealllline close to the surface. For bluff bodies, as in the
airfoil case, the \'clocity increases froill the leading edge
to the 111axilnU111 thickness point. For a circular cylinder
the velocity increases fronl t.he front of the cylinder to

--------~--------
RESONANCE I December 2000 59
 GENERAL I ARTICLE

a) (b)

-J-··--r·-.. .,---.--"I·-...........--·.,E
ABC D

a maximum of rv 2U00 at the maximum thickness point

Figure 8. (a) Streamlines in
non-viscous flow past a cir-
and reduces to near zero at the back of the cylinder.
cular cylinder (b) The ve/oc- In the case of flow of a fluid with low viscosity, bound-
ityvariation along a stream-
ary layers form on both the upper and lower surfaces
line that goes just above the
starting from the nose or front stagnation point of the
cylinder surface. The flow
accelerates from point B to body. The pressure accelerates the boundary layer flow
pointC and decelerates from in the front portion; viscous forces retard the flow.
point C to point D.
N ow we come to the main point. In the rear portion the
increasing pressure causes a rapid deceleration of the
flow. The slower moving boundary-layer fluid subjected
to this rapid deceleration reverses direction(Figure 9).

Figure 9. Streamline picture of the flow of a low-viscosity fluid past a circular cylinder. Except for
the boundary layers the flow in the front portion of the cylinder is nearly identical to the ideal flow
shown in Figure 8. Just after the maximum thickness point the increasing pressure with distance
causes the boundary-layer fluid to reverse direction and the boundary layer separates. In the rear
portion of the cylinder the flow, called the wake, is completely different from ideal flow; it is
unsteady and turbulent.

--------~--------
60 RESONANCE I December 2000
 GENERAL I ARTICLE

Figure 10. Visualization of

the flow over a circular cyl-
inder at a Reynolds num-
ber of 2000. The boundary
layers form at the nose and
separate at about the 9oo
points. The flow after the
separation is turbulent and
the fluid in the wake is
nearly stagnant. Method of
visualization is similar to
Then the boundary layer instead of remaining attached the one in Figure 6.
(Reproduced from An Album
to the body separates from it. Although boundary- of Fluid Motion, assembled by
layer separation, so characteristic of bluff bodies, is a Milton Van Dyke, The Parabolic
purely viscous effect in a narrow boundary layer, its Press, 1982).

consequences are global and far reaching. Consequently,

the observed streamline pattern and pressure distribu-
tion downstream of separation are totally different from
those predicted by ideal flow theory (Figures 10 and 11).

Figure 11. The pressure

c variations with angle over
a circular cylinder in non-
viscous flow and in flow of
0.0
a fluid with small viscosity.
In non-viscous flow equal
pressures in the front and
back results in zero drag. In
-2.0 viscous flow a higher pres-
sure in the front in relation
-3.0'-----'-----...:::w.."'------'--------J to the pressure in the back
0° 45° 90° 135° 180°
gives a drag - Yz pUocJ x
e frontal area.

-RE-S-O-N-A-N-C-E--I-D-e-c-e-m-b-e-r-2-0-0-0----------~-------------------------------6-1
 GENERAL I ARTICLE

Figure 12. Visualization of

flow around an airfoil at a
high angle of attack. The
boundary layer on the top
surface separates resulting
in a large drag caused by
pressure differential be-
tween the lower and upper
surfaces. Compare this pic-
turewiththatofflowaround
an airfoil at a small angle of
attack (Figure 6). Flow is
The drag on a bluff body arises mainly from the dif-
from right to left.
(Reproduced from G K Bat- ference between the higher pressure on the front face
chelor, An Introduction to Fluid and the lower pressure on the back face. The drag force
Dynamics, Cambridge Univer- per unit frontal area ~ pU!/2 is almost independent of
sity Press, 1967) viscosity!
Beyond an angle of attack around, say 12°, an airfoil
will stall: the lift drops and the drag sharply increases.
Stalling is due to boundary-layer separation (Figure 12).
An airfoil at a high angle of attack behaves like a bluff
body!
Conclusion
In most physical situations the notion that small causes
lead to small effects is valid: Prandtl showed, however,
that in the case of boundary layers in ft.uid ft.ow this
notion was at times invalid. We have brieft.y shown using
three types of flows how important boundary layers are
The slower moving in practically important flows. But they are also very
boundary-layer relevant to many natural flow fields such as those in
fluid subjected to a rivers, in oceans and in the atmosphere, where the scales
rapid deceleration, are very different. Although we are normally unaware of
reverses direction. it, boundary layers play an important role in everyday
The boundary life, for example when we stir our tea.
layer instead of
In order to get a feel for the actual numbers involved,
remaining attached
we conclude by estimating some Reynolds numbers and
to the body, boundary layer characterestics in some everyday situa-
separates from it. tions. We observe that the density and viscosity of water

-62------------------------------~-------------------------------
RESONANCE I December 2000
 GENERAL \ ARTICLE

are approximately 1000 kg/m 3 and 1 x 10- 3 kg/ms, re- Suggested Reading
spectively; the corresponding values for air are approx-
imately 1 kg/m 3 and 0.15 x 10- 4 kg/ms. This implies [1] S Goldstein (ed), Mod-
that if U is in km/hr and l is in m, Reair ~ 1.85 x 104Ul ern Development in
and Rewater ~ 2.8 x 10 5 Ul. We immediately see that un- Fluid Dynamics, Vol.l,
Dover Publ., 1965.
der most normal circumstances encountered in everyday
[2] D J Tritton, Physical
life the Reynolds numbers are likely to be very large; Fluid Dynamics, Van
thus boundary layers are likely to be present and the Nostrand, 1977
flows are likely to be turbulent.
The boundary layer on a thin lIn x 1m flat plate trav-
elling at 60 knlph (~17m/s) in air will remain larni-
nar nearly till the end of the plate, at which point the
boundary layer thickness is about 4mm. The drag on the
plate ~ 0.002 x 2 x (1/2pU 2 ) X 1 = 0.55N ~ 0.057kgf.
The same plate placed head-on will have a drag force
~ 1/2(pU 2 A) ~ 140N~14kgf. A person with a frontal
area ~ 0.317},2 travelling at 60kmph on say a motorcycle
will feel a drag force of about 4kgf.
It is close to a hundred years since Ludwig Prandtl in-
troduced the idea of a boundary layer. By doing so he
resolved a paradox that had been puzzling scientists for
almost two centuries. It took another 50 years to un-
derstand mathematically what Prandtl had intuitively,
through his own genius, seen to be true. Now his ideas
are not only routinely utilized in many flow situations
but are being applied to many other branches of the
physical and engineering sciences.
Acknow ledgements
Address for Correspondence
We thank M urali Cholemari and Shyama Prasad Das Jaywant H Arakeri
of Mechanical Engineering Departnlent, IISc, and V S Department of Mechanical
Engineering
Narasimha of NAL, for their help in the manuscript Indian Institute of Science
preparation. Bangalore 560 012, India.

PN Shankar
CTFD Division, NAL
Bangalore 560017, India.

-RE-S-O-N-A-N-C-E--I-D-ec-e-m-b-e-r-2-0-0-0---------~-----------------------------6-3