Aerodynamics Particle Level: at The

Aerodynamics
at the
Particle Level
v.12
Charles A. Crummer, PhD

University of California, Santa Cruz (ret.)
arXiv:nlin/0507032v12 [nlin.CD] 24 Apr 2018
charlie.crummer@gmail.com
May 17, 2021
Contents
1 Preface 1
2 Introduction 2
3 Total force on the surface of the airfoil 5
3.1 Physical parameters affecting the pressure on the airfoil . . . . . . 6
4 Mechanics of fluid interaction 7
4.1 Fluid flow over a flat surface . . . . . . . . . . . . . . . . . . . . 9
4.2 Fluid flow over a curved surface . . . . . . . . . . . . . . . . . . 10
4.3 Static buoyant lift: the aerostat . . . . . . . . . . . . . . . . . . . 14
4.3.1 Light gas Balloon . . . . . . . . . . . . . . . . . . . . . . 15
4.3.2 The Hot-air Balloon . . . . . . . . . . . . . . . . . . . . 17
4.3.3 Submarines and Fish . . . . . . . . . . . . . . . . . . . . 18
4.4 Vortex fluid motion . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.5 Finite wings and wingtip vortices . . . . . . . . . . . . . . . . . . 24
4.6 Leading Edge Extensions . . . . . . . . . . . . . . . . . . . . . . 25
4.7 Birds in flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
5 Bernoulli flow and Coandǎ flow 26
5.1 Bernoulli’s equation . . . . . . . . . . . . . . . . . . . . . . . . . 26
i
5.2 Bernoulli at the particle level . . . . . . . . . . . . . . . . . . . . 29
5.2.1 Venturi’s tube . . . . . . . . . . . . . . . . . . . . . . . . 30
5.2.2 The two-fluid atomizer . . . . . . . . . . . . . . . . . . . 32
5.2.3 Conventional atomizer . . . . . . . . . . . . . . . . . . . 35
5.2.4 Flit gun . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.2.5 Flow into an expansion chamber . . . . . . . . . . . . . . 37
5.3 The Joule-Thomson effect . . . . . . . . . . . . . . . . . . . . . 39
6 The Coandǎ effect 40
6.1 Organ pipe beard . . . . . . . . . . . . . . . . . . . . . . . . . . 41
6.2 The Bunsen burner . . . . . . . . . . . . . . . . . . . . . . . . . 42
6.3 The Coandǎ propelling device . . . . . . . . . . . . . . . . . . . 43
7 Calculation of lift 44
7.1 Using Newton’s Third Law: Effects on the air caused by the pres-
ence of the airfoil . . . . . . . . . . . . . . . . . . . . . . . . . . 46
7.2 Using Newton’s Second Law: Effects on the airfoil caused di-
rectly by air pressure . . . . . . . . . . . . . . . . . . . . . . . . 51
8 Stalling wing 54
9 Rocket engine diffuser 55
9.1 The diffuser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
ii
10 The high-bypass turbofan jet engine 57
11 The vortex refrigerator 58
TM
12 The Dyson Air Multiplier fan 61
13 Spinning objects in the flow: the Magnus effect 62
14 Gurney and Fowler flaps 62
15 Slots and slats 64
16 Slurries 65
17 Summary 67
18 Conclusion 68
A On the consideration of fluids at the particle level 71
B Henri Coandǎ’s Propelling Device 82
iii
List of Figures
1 Velocity profile in the boundary layer for laminar flow . . . . . . . 4
2 Interaction between fluid particles and a real surface . . . . . . . . 8
3 Fluid flow over curved and flat surfaces . . . . . . . . . . . . . . 10
4 Behavior of particle flow over a curved surface . . . . . . . . . . 12
5 Flows over a Wing . . . . . . . . . . . . . . . . . . . . . . . . . 14
6 Hot-air Balloon . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
7 Fish’s Swim-Bladder . . . . . . . . . . . . . . . . . . . . . . . . 18
8 Submarine Ballast System . . . . . . . . . . . . . . . . . . . . . 19
9 The vortex process . . . . . . . . . . . . . . . . . . . . . . . . . 23
10 Downwash and wingtip vortices . . . . . . . . . . . . . . . . . . 24
11 Pitot tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
12 Venturi tube . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
13 Theoretical atomizer . . . . . . . . . . . . . . . . . . . . . . . . 33
14 Air velocity in the tube . . . . . . . . . . . . . . . . . . . . . . . 34
15 dV /dt to achieve the correct velocity . . . . . . . . . . . . . . . . 35
16 Real atomizer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
17 Nozzle Detail . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
18 Flit Gun . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
19 Restricted exit orifice . . . . . . . . . . . . . . . . . . . . . . . . 38
iv
20 Joule-Thomson apparatus . . . . . . . . . . . . . . . . . . . . . . 39
21 Beard on an organ flue pipe . . . . . . . . . . . . . . . . . . . . . 42
22 Bunsen Burner . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
23 Typical force configuration on an airfoil in an air flow . . . . . . . 45
24 Geometry outside the airfoil . . . . . . . . . . . . . . . . . . . . 48
25 Illustration of the covariant derivative. . . . . . . . . . . . . . . . 49
26 Bending of the airflow by an airfoil. . . . . . . . . . . . . . . . . 51
27 Coandǎ effect geometry. . . . . . . . . . . . . . . . . . . . . . . 53
28 Detail at the Diffuser Wall . . . . . . . . . . . . . . . . . . . . . 56
29 Early turbojet . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
30 The entrance cowl for an Airbus A380 turbofan engine . . . . . . 59
31 Ducted fan tailrotor . . . . . . . . . . . . . . . . . . . . . . . . . 59
32 Vortex tube schematic . . . . . . . . . . . . . . . . . . . . . . . . 60
33 Vortex tube flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

TM
34 The Dyson Air Multiplier fan . . . . . . . . . . . . . . . . . . 61
TM
35 Operation of the Dyson Air Multiplier fan . . . . . . . . . . . . 61
TM
36 Cross section of the Dyson Air Multiplier fan . . . . . . . . . . 62
37 The Magnus effect . . . . . . . . . . . . . . . . . . . . . . . . . 63
38 Gurney flap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
39 High-lift wing devices . . . . . . . . . . . . . . . . . . . . . . . 64
v
40 Bus burned by the pyroclastic flow from the World Trade Center
collapse on September 11, 2001. . . . . . . . . . . . . . . . . . . 67
vi
Abstract
All aerodynamic forces on a surface are caused by collisions of fluid

particles with the surface. Upwash, downwash, lift, drag, the starting vortex,
the bow wave, and any other phenomena that would not occur without the
surface are caused by its presence as it interacts with the air flow. While
the standard approach to fluid dynamics, which is founded on the “fluid
approximation,” is effective in providing a means of calculating a wide range
of fluid behavior, it falters in its ability to account for the effects of complex
interactions of the fluid either with itself, other fluids, or with solid bodies.
One of the conditions required to justify the fluid approximation is that the
flow be steady [21], i.e. that the particles of the fluid not be interacting with
each other or with any surface. It is these very interactions, however, that
are the causes of aerodynamic effects on solid bodies in the flow. This is
not to say, of course, that the fluid approximation is never useful, but that
some well-known and important effects such as the Coandǎ effect are not
explained by that model.
1
1 Preface
The purpose of this paper is to set the stage for a close examination of fluid phe-
nomena, an examination at the particle level. Most fluid phenomena of interest
are the result of its behavior in interaction with surfaces, other fluids or, indeed,
with itself. The eddies and turbulence attendant fluid shear are extremely com-
plex. As one fluid is injected into another, the shear effects depend further on the
different attributes of the fluids. If a fluid is flowing, it is doing so with respect to
something, a surface for instance.
A dimensionless quantity used to characterize the nature of fluid flow is Reynolds’

number:
ρvL
R=
η
where
ρ is the density of the fluid,

v is its velocity,
η is the fluid’s viscosity and
L is called “a characteristic length.”
What does “characteristic length” mean? L is a length that is defined only in terms
of the boundaries of the flow such as the diameter of a tube or the chord length
of an airfoil. What length is it and why? In fact, Reynolds’ number is only well-
defined in discussions of model scaling of fluid flows in interaction with solid
surfaces. For example the characteristics of a flow around a boat with a beam of
4 meters in an ocean current of 10 knots will be the same for a scale model of the
boat in the same ocean water whose beam is 0.4 meters and where the current is
100 knots.
What meaning can references to Reynolds’ number have?
Bernoulli’s relation involves the fluid velocity. In a Venturi tube, it is the velocity
with respect to the wall of the tube. If a high fluid velocity implies a low pressure,
1
how can the pressure readings in different parts of the tube be different since the
sensors are in the boundary layer of the fluid at the surface of the wall of the tube?
The boundary layer is stationary, or nearly so (see Figure 1 below).
It is these and other baffling questions that have launched the author into these
investigations.
Even though aerodynamics engineers are masters at designing airframes, they are
refining known technology. Without understanding from first principles, light-
ing engineers would just be refining incandescent lamps and we would not have
fluorescent lights or LEDs.
2 Introduction
The behavior of real fluids, i.e., compressible and viscous, is to this day baffling
in many ways. Part of the reason is that explanations of fluid behavior are hold-
overs from the pre-twentieth century belief that a fluid is a fundamental entity, not
composed of anything else.[2] The trouble with this approach is that it provides
only viscosity and pressure as ways of understanding how the fluid interacts with
itself or with solid bodies. Both are intensive variables but what do they mean for
in cases where the fluid approximation is not valid?
Pressure, p, (stress normal to a surface) can be understood as that fluid property

which causes a normal force on a surface in the flow,
dFn = p(s) dA.
The shear force provides part of the drag on a surface. It is derived from the shear
stress, τ, tangential to the surface.
dFs = τ ⊗ dA,
where
2
τ(s) ≡ µS(r, s) r=0 . (1)
Here,
s is the location on the surface of the airfoil,
r is a length in the direction normal to the surface,
v(s, r) is the velocity of the flow relative to the surface,
µ is the dynamic viscosity of the fluid,F
S(r, s) = ∂ v(r, s)/∂ r is the shear and
τ is the resulting shear stress on the surface.
Figure 1 shows qualitatively the velocity profile in the boundary layer during lam-
inar flow. The curve is differentiable and indicates that there is slip at the surface.
Admitting the possibility of slip at the airfoil surface is contrary to the no-slip as-
sumption of Ludwig Prandtl[8] but in view of the development in Section 4 below
and the work of Johan Hoffman and Claes Johnson,[17] there is reason to suspect
the reality of the no-slip assumption. At the surface, because of the interaction of
the particles in the flow with each other and with the (possibly submicroscopic)
features of the surface, the behavior is very complex but for laminar flow this
structure is smoothed out as the disturbance recedes into the flow.
A common example of this is the bow-wave of a slowly moving boat. Close

inspection of the behavior of the water at the bow reveals great complexity but far
from the boat the wave is very regular and smooth.
In order that the viscosity, τ(s), as defined in Equation (1) have meaning, the
function v(s, r) must be smooth and differentiable. However, as the flow velocity
increases, there is an onset of turbulence. The boundary layer develops eddies near
the surface [8] and v(s, r) becomes non-differentiable and so the partial derivative
in Equation (1) ceases to exist. The behavior of the fluid becomes very complex
and the flow becomes unsteady; the fluid approximation becomes invalid.
Since the work of Boltzmann [3] and Einstein [7], i.e., theory based on the pos-
tulate, and supporting evidence that fluids are composed of tiny particles, deeper
3
Figure 1: Velocity profile in the boundary layer for laminar flow
insight is possible by considering in detail the interactions of these particles with

each other, those of other fluids, and those of solid bodies in the flow. In fact it
may be helpful to remember that the only interactions a fluid can have, according
to this model, are through momentum transfer or Van der Waals forces between
its particles and between the particles and the surface.1 The molecules of a gas
at standard pressure are only within Van der Waals distance about 1/100th of the
time they are apart so these forces only play a part in particle-particle scattering.
The notion, therefore, that a streamline in a gas flow is “attracted” by a surface

is not correct. If a stream of gas, as in Coandǎ flow,[9] seems attracted to a solid
object it is due to its self-interaction, interaction with gas outside the flow, and
the forces its particles exert on the surface as they strike it, not due to an attractive
force between the particles and the surface. Evidently, then, the “attraction” is due
to the fact that a low pressure is created in the boundary layer and it is the higher
ambient pressure that forces the boundary layer to stay in contact with the surface.
In contrast to the work of Bernoulli, there is no “Coandǎ equation” because, other
than Newton’s laws, we have no physical model for the behavior of the particles
in the boundary layer. Henri Marie Coandǎ was an engineer and observed effects
that are widely incorporated into modern aerodynamic design but physicists have
not developed a tractable mathematics to describe the behavior of such a large
number, ∼ 1023 , of simple interactions without the fluid approximation. In any
case, to explore a mathematical model is not the same as to explore the physical
world (See Appendix A). One goal of theoretical physics is the calculation of the
results of experiments, another is to understand why the world works as it does.
1 We do not consider plasmas, which are affected by long-range electromagnetic forces.
4
The miracle is that mathematics is as useful as it is in describing and predicting
physical effects.
The statements made below about fluid flow are conclusions and hypotheses com-
ing from a consideration of particles obeying Newton’s laws. The author’s inten-
tion is to stimulate the reader’s thoughts about the behavior of fluids in regimes
where the flow is not steady, and hence the fluid approximation is invalid. Another
aim of this paper is to discern causes of phenomena. A mathematical equation
does not contain causal information. For example, the thrust of a rocket is not
caused by the velocity of the exiting gases but by the pressure difference between
the throat and the projection along the axis of the motor of the throat area onto the
back wall of the motor. Bernoulli’s equation relates the the exit velocity and the
pressure difference but conveys no information as to which is the cause and which
is the effect. It is only from experience with the physical world and abstractions of
that experience that one knows that in Newton’s second law it is force that causes
acceleration, not the reverse.
It is hoped that an understanding of the true causes of aerodynamic effects will

lead to new aerodynamic designs and the rethinking of designs already created.
Imagine, for a moment, that in the absence of a tractable mathematical model,
non-mathematical understanding is possible.
3 Total force on the surface of the airfoil
For perfectly elastic collisions the effect on a surface over an area ∆A results in a
force, ∆F with components normal and transverse to the area.
∆F = m ∑ ai
∆A
where m is the mass of one particle and the ai are the accelerations of the parti-
cles hitting the surface area ∆A and the summation is over the area. The normal
components of the ai ’s are due to pressure and the transverse components are due
to the viscous interaction of the fluid with the surface and with itself.
5
As the particles move over the surface, they are affected by the molecular protu-
berances on the surface and by Van der Waals forces between the particles and the
surface. This friction, i.e., viscosity, force is proportional to the area ∆A as well.
The total force on the airfoil, then, is the vector sum of the normal and tangential
force components over the total airfoil area:
Ftotal = − ∑ (∆Fn + ∆Fs ) .

air f oil
The integral form of this equation is
I
Ftotal = − (dFn + dFs ) . (2)
air f oil
The minus sign indicates that the force on the particles is opposite to the force on
the surface of the airfoil. There is sometimes confusion about the above surface
integral. It does not mean that there is a net forward flow under the airfoil. Just
that the flow is the vector sum of a translation and a rotation. This rotation is
a consequence of the presence of the airfoil in the flow. The surface integral,
reduced to a line integral in the case of an infinite wing, just indicates the integral
of all forces on all sides of the wing. (See also Section 7 below.) There is a
formal derivation of the Kutta-Joukowski theorem at [10] and [11]. Notice that the
circulation integral is conducted in a region of potential flow, where Bernoulli’s
equation would hold, i.e., far from the surface. The circulation is an integral
of the velocity field around the airfoil. But the Kutta-Joukowski theorem begs
the question as to how this velocity field is created and how it could be derived
mathematically from first principles.
3.1 Physical parameters affecting the pressure on the airfoil
We will ignore Van der Waals forces and assume that the fluid is approximately
an ideal gas. The equation of state for an ideal gas is
pV = NkT, (3)
6
where
p is the pressure,
V is volume,
T is the Kelvin temperature.
N is the number of particles in V and

2
(m kg)
k is Boltzmann’s constant, (∼ 1.38 × 10−23 (sec2 ◦ K) )
We now choose units so that k = 1. Then the particle density, ρ is
N p
ρ= = (4)
V T
Then,
p = ρT. (5)
Far away from the airfoil, the pressure, p, is approximately constant and uniform
except for the effect of gravity and the presence of the airfoil[11]. The com-
pressibility of air can be ignored. But on the surface of the airfoil, it is precisely
pressure differential that causes lift. Equation (5) reveals that ρ, and T , subject to
the laws of thermodynamics, are at the disposal of the aeronautical engineer for
creating a favorable pressure field on the airfoil’s surface.
4 Mechanics of fluid interaction
Aerodynamic forces affecting a rigid surface are always net forces produced by
differences in pressure between different parts of the surface. The absolute pres-
sure on a surface area element is the density of the normal components of the
forces acting on the surface there.
7
Figure 2: Interaction between fluid particles and a real surface
Aerodynamic forces on a body are caused only by collisions of fluid particles with
the body’s surface.2 At the molecular level, the flow particles encounter any sur-
face as a molecular structure which is rough, with protuberances whose size is
of the order of magnitude of the flow particles themselves (see Figure 2.). As
particles collide with the surface, their momentum components normal to the sur-
face there cause lift, positive or negative, and stagnation pressure and the parallel
components cause viscous drag and give rise to a boundary layer which is carried
along by the surface (see Ref. [8]). It is clear, then, that the microscopic struc-
ture of the surface and the properties of the fluid will affect drag and lift, even for
Φ = 0. 3 A perfectly smooth surface would have no viscous drag, there would be
no shear in the fluid near the surface and, it would appear, a wing made of this
material would have lift only if the air flow momentum density had components
normal to the bottom surface of the wing, i.e. due to the angle of attack, Φ.
Even though these momentum transfers occur only in the boundary layer that ap-
pears to be “dragged along” by the surface, they are responsible for the whole of
lift and drag. Actually, fluid particles can leave and enter the boundary layer by
moving normal to the surface. Dust on a surface in a flow is not disturbed later-
ally because the boundary layer is motionless, or nearly so, at the surface. The
boundary layer is created by the interaction of the main flow particles with parti-
cles bouncing off the surface. For the time being, we assume that all collisions,
particle-particle and particle-surface, are perfectly elastic and that the particles are
spheres.
2 The Coandǎ effect in liquid-surface flow, however, may be caused in large part by van der
Waals forces, which are attractive.
3 In the design of his racing plane, the H-1, Howard Hughes insisted that the rivets attaching
the aluminum skin be flush rather than projecting above the skin.
8
4.1 Fluid flow over a flat surface
Let us consider the flat surface in panel a) of Figure 3. The pressure on the surface
is due only to the normal components of the momenta of the impacting particles.
Flow along such a surface will not affect surface pressure. As particles are blown
away from the surface, other particles are drawn in from outside to replace them.4
Pressure on the surface is due to collisions of particles with the surface. Where the
flow has no normal component, the pressure is due only to the thermal motion and
density of the particles in the boundary layer, i.e. the static atmospheric pressure.
Hence this pressure will be a function only of the mass of a particle, the particle
density, and Kelvin temperature of the air at the surface.
In reality, the fluid particles in a layer around a surface boundary seem to be car-
ried along with the surface, i.e. the distribution of the components of their veloci-
ties parallel to the surface is nearly[17] circularly symmetric about a mean which
is the velocity of the surface relative to the free-stream velocity.5 Particles, as
large as dust particles or as small as the molecules making up the flow, experience
Van der Waals forces attracting them to the surface. Whether or not the particles
are fixed on the surface by these forces depends on the structure of the molecules
making up the flow and those making up the surface. These Van der Waals forces
are responsible for the “wetting” of the surface. In some cases, e.g. Teflon and
water, the fluid drains off the surface quite readily just under the force of gravity.
In other cases, e.g. modern motor oil on a bearing surface, the fluid may adhere
for months or even years.
In any case, however, particles continually leave the boundary layer and enter it
transversely from the flow due to heat energy or, at an angle of attack, because
they have velocity components normal to the surface.6 Beyond a mean free path7
or so but still in the boundary layer, the distribution of the normal components of
particles’ velocities moving toward or away from the surface will depend on the
temperature and density of the particles. Even though it is regularly driven at high
speed, a car will accumulate dust on its body. An air stream directed toward the
4 Place a sheet of paper flat on your hands. Blow over the top surface of the paper. This
experiment refutes the notion that the pressure in a free flow is less than the ambient static pressure.
Bernoulli flow, on the other hand, is (or could be) confined to a tube and is not free.
5 This property of fluid flow was utilized by Nicola Tesla [1] in his unique design of a rotary
pump.
6 It can be seen, then, that dust particles on a surface in an air flow are not disturbed not because
the fluid particles are necessarily entrained but that they come and go normal to the surface. Hence
they do not impart lateral forces to the dust particles.
7 ∼ 9 × 10−8 meters for N at standard pressure and temperature.
2
9
surface, however, will blow off some of that dust. As we will see, it is the mutual
interaction of flow particles and these “stagnant” boundary layer particles that is
responsible for a part of the lift on an airfoil at subsonic speeds.
An increase in the free-stream velocity means that the components of the veloc-
ities of the flow particles increase in the direction of the free-stream velocity and
parallel to the surface. The reason that the boundary layer remains quiescent, or
nearly so, is that the components of the colliding particles’ velocities parallel to
the surface reverse as they collide with microscopic irregularities. This is one of
the causes of aerodynamic drag and accounts for the fluid’s viscosity.8 If the col-
lisions are not perfectly elastic, the rebound speed is less than the incident speed
and the surface absorbs some of the particle’s energy, i.e. it heats up.
4.2 Fluid flow over a curved surface
Figure 3: Fluid flow over curved and flat surfaces
In a steady flow over a surface, stream particles have only thermal velocity com-
ponents normal to the surface. If the surface is flat, the particles that collide with
boundary layer particles are as likely to knock them out of the boundary layer as to
knock others in, i.e. the boundary layer population is not changed and the pressure
on the surface is the same as if there were no flow. If, however, the surface curves
away from the flow direction, the particles in the flow will tend to take directions
tangent to the surface, i.e. away from the surface, obeying Newton’s first law.
As these particles flow away from the surface, their collisions with the boundary
layer thermal particles tend to knock those particles away from the surface. What
this means is that if all impact parameters are equally likely, there are more ways
8 Though viscosity is supposed to be a property of the fluid, it is measured by the terminal
velocity of a ball in the fluid or the force it takes to slide two plates with the fluid between them.
Viscosity, then, has to do with the interaction of the fluid with itself as well as with solid bodies.
10
a collision can result in a depletion of the boundary layer than an increase in the
boundary layer population. The boundary layer will tend to increase in thickness
and to depopulate and, according to Equations (4) and (5), the pressure will reduce
there. This is why the flow is forced toward the surface, the Coandǎ effect with
the attendant “suction” that draws in fluid far from the surface.9 Those particles in
the flow that do interact with the stagnant boundary layer will give some of their
energy to particles there. As they are deflected back into the flow by collisions
with boundary layer particles, they are, in turn, struck by faster particles in the
flow and struck at positive impact parameters.
The following, Figure 4, consists of 4 frames taken from an animation[26] illus-

trating the behavior of flow particles as they interact with stagnant particles in the
boundary layer. The first panel shows the incoming particles in red approaching
from the right. In the second panel the incoming particles begin to interact with
the stagnant particles meant to approximate a boundary layer. The third panel
shows the boundary particles being blown away by the incoming set, thus reduc-
ing the pressure at the surface. As panel four shows, it is primarily the boundary
layer particles that make up the flow that clings to the curved surface.
These frames show, at least qualitatively, the Coandǎ effect. The figures are frames
taken from an animation made with Working ModelTM [25] software. In the video,
approximately 600 small circles are launched toward a fixed circle with stagnant
circles positioned around it, meant to simulate a boundary layer. All collisions are
perfectly elastic and the large circle has infinite mass. Of course, this model is
highly unrealistic because of the very small number of particles, their simple cir-
cular structure, the smoothness of the surface and the absence of thermal motion.
It does, however, show boundary layer depletion and the “wrapping” of the flow
around the surface. A more accurate simulation would have a continuous flow
impinging on a rough surface surrounded by particles. All the particles should be
interacting thermally with each other and with the surface. Notice also that the
wrapped flow contains almost none of the red incoming particles.
In reality, the flow shears past the surface (where the molecular motion is complex
and chaotic) and the fluid velocity as a function of the distance normal to the
surface is a smooth function of this distance. The velocity profile curve parameters
are constants depending on the fluid velocity, the physical characteristics of the
surface and the particles making up the fluid (See Figure 1). In any case, as the
9 This explanation suggests experiments exploring the structure at the edge of the main flow
that is away from the wall. The explanation of the mechanism by which the flow is “attracted” to
the wall implies how the flow should behave at its other edge too.
11
Figure 4: Behavior of particle flow over a curved surface
flow velocity increases, separation points will begin to appear[8]. These are the
points on the surface where the directional derivative of the fluid velocity normal
to the surface vanishes as does the shear (Equation (1)). As the fluid velocity
increases even further, the derivatives at the separation points actually reverse
sign, there is backward flow on the surface.[8] Vortices have formed downstream
from these stagnation points. All this is in the language of fluids. What is going
on at the particle level though?
The curved part of the surface acts as the obstruction mentioned in the explanation
of the vortex process (See Section 4.4.) because it presents stagnant particles to
the flow. The flowing particles as they approach the surface interact with these
particles and with the surface itself. Some populate the boundary layer and then
interact as stagnant particles with other particles in the flow. There is a constant
interchange of particles between the boundary layer and the flow. As these interac-
12
tions take place the process described above activates the boundary layer particles
like falling dominoes, causing the enveloping flow. When the surface curves away
from the flow, the flow particles, obeying Newton’s first law, tend to travel on tra-
jectories tangent to the surface and thus leave its vicinity, taking some boundary
layer particles with them. This reduced pressure in the boundary layer has two
effects. First, it causes the higher pressure in the main flow to force itself, and
smoke streamers, toward the surface, and second, it results in lift as the higher
pressure on the bottom of the wing has increased effect.
The Coandǎ effect is investigated in some detail in articles in Deutsche Luft- und
Raumfahrt[23]. H. Riedel’s paper has numerous diagrams of flow patterns and
distributions of pressure differentials on a wing surface in various positions with
respect to an air jet. Figure 5 is from this paper.
The first panel in Figure 5 shows the flow of and around a free stream in an at-
mosphere. Note the entrainment of air from outside the stream. Panels 2 and 3
show flow and distributions of pressure differentials on a wing in the flow. Note
that in panel 3 there is a sharp spike in downward pressure where the flow actually
impinges vertically on the surface. This downward pressure is the cause of wing
stall. Flow along a positively curved surface causes a lowering of the pressure on
the wing but for flow rates above a certain velocity a vortex will be created which
can cause an increase of pressure there.
The Coandǎ effect gives a hint at at what turns out to be the most important cause
of lift at zero or small angle of attack and subsonic conditions. While it may be
intuitive that to increase lift the pressure under the wing, and thus the angle of at-
tack, should be increased, in fact for a fixed wing it is more important to decrease
the pressure on the top of the wing. (See Reference [15], page 181.) Professor
Marco Colombini at the University of Genoa, Italy[18] has produced some inter-
esting animations illustrating the pressure distribution around a standard airfoil at
varying angles of attack.10 It is interesting to think of airfoil design as an exercise
in managing buoyancy.
10 These pressure distributions, however, do not show the bow wave the same as it is seen in
Figure 23.
13
Figure 5: Flows over a Wing
4.3 Static buoyant lift: the aerostat
Dirigibles, helium balloons and hot-air balloons utilize buoyant lift. They are
sometimes called aerostats because they achieve lift without movement, without a
main air flow. They generate neither upwash or downwash as a third-law reaction
to this lift. These devices rise due to the difference in the atmospheric pressure be-
tween top and bottom. This buoyant force acts naturally on everything immersed
14
in a fluid in a gravitational field. It acts on us but we don’t notice it because the
density of our bodies is so much greater than the density of air where we live.
Archimedes noticed this buoyant force and uttered the famous “Eυρηκα !” He
had discovered how to measure the density of the king’s crown and to test if it was
pure gold.
One might think that the buoyant force, which is due to the gravitational field,
would be negligible for an airplane because the airplane’s overall density is much
greater than air at standard conditions. However, see Section 6 below. Since the
buoyant force is due to the pressure differential between the top of a body and the
bottom as well as the overall density of the balloon, the buoyancy can be managed
by controlling these pressures and the density.
In France a balloon large enough to carry cargo is called a Montgolfière after the
brothers who in 1783 flew the first hot-air balloon to carry a living cargo.
A balloon, or aerostat, consists of an envelope filled with a gas that is lighter than
air, i.e. less dense. This gas can consist of molecules each of which is lighter than
the average weight of the molecules of air or air itself that is hotter and less dense
than the atmosphere surrounding the aerostat. The light molecules of Helium or
Hydrogen are used for the former type of aerostat.
4.3.1 Light gas Balloon
The ideal gas equation, Equation (4) above, is:
p N
ρ= = ,
T V
where ρ is the number density of the molecules, i.e., the number of molecules per
unit volume, and we’ve chosen units so that k = 1.
Archimedes’ equation for the buoyant force, B, on an object immersed in a fluid

is
B = ρ f m f gV, (6)
15
where
ρ f is the density of the particles of fluid in which the body is immersed,
m is the average mass of a fluid particle,
V is the volume of the immersed object and
g is the acceleration due to gravity.
This upward force is offset by the weight of the object immersed,
wballoon = ρg mg gV + wenvelope , (7)
where here ρg is the density of the gas in the balloon, mg is the mass of a gas
particle and the V 0 s are the same. The net upward force, then, is
Fnet = B − wballoon = (ρ f m f − ρg mg )gV. − wenvelope . (8)
Assuming that the temperatures of the gas inside the balloon and the outside the
balloon are equal, for a balloon with a loose, light, flexible and very voluminous
envelope, the atmospheric pressure and the pressure inside the balloon are the
same so, according to Equation (3),
ρ f = ρg = ρ, (9)
and the net upward force is proportional to m f − mg , (ignoring the weight of the
envelope), i.e.,
Fnet ∝ (m f − mg ). (10)
This is why a light gas like Hydrogen or Helium is used in high-altitude balloons.
If the envelope of the balloon is an elastic material like rubber, the pressure exerted
by the envelope on the gas within is an increasing function of the volume of the
16
balloon. The pressure inside the balloon is pg = p f + penvelope , which means that
in this case ρg > ρ f . A rubber balloon will only rise if ρg mg < ρ f m f , and then
only until ρ f = ρg (mg /m f ) or the elastic limit of the rubber is reached and the
balloon explodes.
4.3.2 The Hot-air Balloon
The latter form of aerostat mentioned above, a hot-air balloon, is lifted by an

envelope containing heated air.
Figure 6: Hot-air Balloon
The burner that provides the heat is concentric with the opening at the bottom of
the balloon so that the hot air created by the burner rises and the cooler air that is
displaced exits from the annular space around it and, if it is opened, the parachute
valve at the top of the balloon. As the temperature increases, the pressure required
to keep the envelope inflated is provided by fewer and fewer molecules and thus
17
the weight of the balloon decreases. By the intermittent operation of the burner
and the parachute valve, the pilot can very precisely choose the height of the
balloon.
All the forces that result in the lift force, buoyant or dynamical, are due to colli-
sions of molecules with the envelope.
4.3.3 Submarines and Fish
Both fish and submarines must be able to control their densities so they can adjust
their depths in water. Fish do this by changing p in Equation (4). They do this by
contracting or relaxing the muscles around their swim-bladders.
Figure 7: Fish’s Swim-Bladder
A submarine’s density is changed basically by adjustments of the total weight of

the craft. It also has trim tanks, tanks that are the counterparts to “swim-bladders,”
and ballast tanks for gross depth control.
The trim tanks operate much like a fish’s swim-bladders. There is a more or less
fixed quantity of air in them. Their weight as well as the pressure of the air in
them is adjusted by injecting water or pumping it out.
To cause the submarine to submerge, the captain orders the ballast tanks, initially
filled with air, to be partially flooded with water as air is exhausted. In this way, N
is reduced and the weight is increased as water floods the ballast tanks. To surface
again, the captain orders that air from high-pressure tanks be let into the ballast
tanks to force the water out. The weight of the craft is increased and decreased
18
by the taking-on and exhausting of seawater. This control is achieved both by the
release of the high-pressure air into the ballast tanks and by pumps. The pumps
control both the ballast pressure and the overall weight of the vessel.
In order for the pumps to work, a certain amount of air must remain in the ballast
tanks at all times. By the same token, the high-pressure tanks must be maintained
at a pressure sufficient to overcome the water pressure at the deepest level the
submarine is allowed to go.
Figure 8: Submarine Ballast System
19
***********
In addition to its own weight a container of gas has weight due to the gravitational
force which affects each molecule of the gas inside. In the gas there is a density
gradient toward the earth. Newton’s law of gravitation is:
F = −rMm/r3 , (11)
where r is the vector from the centre of the earth to the particle.
As they collide with one another molecules, in general, will have momentum com-
ponents in any given direction due their temperature but they will always have a
downward momentum component due to their acceleration in the gravitational
field.
mv = m(vheat + vdown ). (12)
This downward component will always be a factor in collisions with other par-
ticles and will be transmitted in each collision according to Newton’s third law.
The momentum transfer to the envelope from the molecules on the bottom will be
greater than the molecules at the top just because of the greater particle density
there. The top molecules exert a force upward and the molecules on the bottom
exert a downward force. In a vacuum the difference between these forces is that
part of the weight of the object that is due to the gas inside.
If the container is in the earth’s atmosphere there is also a buoyant force. This
force is also due to the gravity force. The external atmosphere is denser at the
bottom of the envelope than at the top, hence the number of exterior atmospheric
molecular collisions per second at the bottom of the envelope is greater than at
the top. The difference between the number and force of the collisions per second
on the bottom (molecules with upward velocity components) and that on the top
(downward components) is the buoyant force. If the buoyant force is greater than
its total weight, the container rises.
The number density, ρ(y), of molecules inside the envelope is a monotonically

decreasing function of the height. This is due to the influence of gravity on each
molecule. If hot air molecules are injected into the envelope those molecules will
undergo collisions with the molecules they encounter.
20
Consider a hot molecule that enters with energy,
∆E = k∆T, (13)
where k is Boltzmann’s constant and ∆T is the temperature of the entering molecule.

Again, we choose units for which k = 1.
Consider two regions: upper and lower. In each, according to Equation (4),
p = ρT. (14)
Although the densities and temperatures of the two regions differ, there is only
one pressure, p, at the point of entry. Hence,
ρu × Tu = ρd × Td , (15)
and
ρd
Tu = × Td . (16)
ρu
This molecule’s energy will be shared among the molecules it collides with and
their added energy will be shared as they interact with others.
Assume the molecule is knocked upward. The resulting increase in temperature

in that region, u , compared with the lower region, d, will be,
ρd (y)
∆Tu = × ∆Td , (17)
ρu (y)
and similarly for the case where the entering molecule is knocked down. If the
entering molecule is knocked upward, there will be fewer molecules to share its
energy than if it were knocked downward since, due to the gravity field,
21
ρu (y) < ρd (y). (18)
The absorption of its energy there will cause a rise in temperature greater than if
it were knocked downward. If it is knocked downward, the denser aggregation of
molecules there will also absorb the molecule’s energy but the rise in temperature
will be less. It is not just that the hot air molecules themselves rise, it is also their
energy which is transmitted upward as they collide with molecules already there.
4.4 Vortex fluid motion
As a fluid stream passes through an opening in a barrier into stagnant fluid, eddies
appear. Consider the state of the fluid as the flow begins. Behind the barrier the
distribution of velocities of the particles of the fluid in a small volume is spheri-
cally symmetric (except for the effect of gravity) and the mean of the distribution
is a function of the Kelvin temperature.
Upstream, the pressure behind the barrier is higher than the pressure behind the
exit. A particle on a streamline just grazing the barrier encounters particles behind
that barrier whose mean velocities are zero. Downstream of the barrier, the result
of collisions with these stagnant particles is the slowing of a flow particle as well
as its deflection back into the flow. (See Figure 9.)
The greater the difference between the flow velocity and the thermal velocities of
the stagnant particles, the closer to 90◦ from the flow direction will be the direc-
tions of the stagnant particles after the collisions. Thus the interaction between the
stream and the stagnant region serves to sort out the colder stagnant particles and
force them away from the flow. The vortex heat pump described later in Section
11, Figures 32 and 33 uses this principle.
As the stream particles that have suffered collisions with stagnant particles are
hit by faster ones in the stream, they too are deflected with a velocity component
normal to the stream velocity. As they continue after being deflected away from
the stream, they hit other stagnant particles (Figure 9), forcing them toward the
same center. The result is that part of the flow is changed into a vortex. If the
obstruction is a hole in a plate, some of the energy of the flow is trapped in the
form of a vortex ring. If the flow is a pulse, this ring follows in its wake.
22
Let the lower half of the y-z plane be a barrier in the fluid. (See Figure 9.) The
velocity of the flow will be superimposed on the random motion of the molecules,
i.e. heat. As the flow begins, say from minus to plus in the x-direction, the mean of
the distribution of the velocities of those particles in the flow will be shifted toward
positive vx . As these particles pass the barrier, they collide with fluid particles that
have a velocity distribution with zero mean, i.e. the stagnant particles.
Call an impact parameter positive if the location of the impact point with a particle
in the flow is a positive distance in y from the center of one of these particles. If it
passes the edge of the barrier with sufficient speed, a flow particle is likely to hit
a particle behind the barrier with a positive impact parameter. This will result in
the stagnant particle being knocked back behind the barrier (See Figure 9a.). The
flow particle will be deflected up into the flow, with reduced momentum where it
will be deflected by other particles in the flow and eventually be knocked back,
away from the flow (Figure 9b). These particles still have an x-component of
velocity that is larger than their y-z velocities but as they interact with each other
and other particles in the flow in the way described above, they will participate
a circular flow and their energy will decrease. This process repeated statistically
with various impact parameters results in a vortex. We will call it the vortex
process.
Figure 9: The vortex process
Consider a velocity coordinate system local to a flow particle that passes very
close to the barrier and with its x-axis in the flow direction. Initially, that system
will be aligned with the coordinates mentioned above. As time goes on after the
particle has passed the y-z plane, the local system will, on average, rotate around
its y-axis. One can see in this way that the effects of this interaction with the
stagnant molecules will propagate into the flow on the downstream side of the
barrier. The result is a vortex.
23
Figure 10: Downwash and wingtip vortices
4.5 Finite wings and wingtip vortices
In flight, an airplane will generate a vortex at each wingtip. These vortices are
created as the higher pressure air under the wing leaks out from under the wing
and away from the fuselage. The vortex is formed as this air is drawn into the low
pressure region above the wing. The vortex is a nuisance11 and is a source of drag
and instability due to the vortex impinging on the top of the wing there. It can
actually reduce the overall lift. Figure 10 shows the vortices very clearly. Notice
that the axes of these vortices are parallel to the flight direction. The wingtip
vortices are an unwanted effect and are due to the necessity of finite-length wings.
Though the wingtip vortices are beautiful and spectacular, the concomitant to the
most important factor in producing lift is the huge trench left in the cloud by the
downwash off the trailing edges of the wings.
The French jet engine manufacturing company, Price Induction12 sells small high
bypass engines for small aircraft. One of their innovations is a turbofan using
11 See http://en.wikipedia.org/wiki/Wingtip_device#NASA_development for a description of
devices to control the wingtip vortex.
12 Price Induction, 2, Esplanade de l’Europe 64600 Anglet, FRANCE. www.price-
induction.com
24
composite, non-metallic blades. At speed, the fan blades elongate and actually
seal on the special bearing surface of the fan housing. The reason for this is to
eliminate vortices at the vane tips. This reduces power requirements, increases
the engine efficiency and increases thrust.
4.6 Leading Edge Extensions
Though wingtip vortices are unwanted, similar vortices are created on purpose by
so-called Leading Edge Extension (LEX) surfaces.[22] A LEX is a flat surface
extending a short distance from the fuselage and from near the cockpit aft to the
leading edge of the wing. At angle of attack vortices are created as the high-
pressure air flows from below the LEX to the lower pressure above. This causes
the vortices, clockwise on the left side and counter-clockwise on the right. These
vortices extend back over the wings and interrupt the stalling vortices that would
otherwise form over the wing. They blow away the particles that would cause
high pressure on the tops of the wings, especially near the roots. LEXs allow the
plane to operate at higher angles of attack than it otherwise could.
4.7 Birds in flight
The high-speed camera shows some very interesting aspects of birds taking flight.[6]
Perhaps the most interesting is that on take-off, when maximum lift is needed, a
bird’s power stroke is down and forward, not backward as it would do if it were
“swimming” in the air. This motion both pressurizes the air under the wing and
creates upwash13 on the leading edges of its wings. This upwash flows over the
leading edge and actually contributes to lowering the pressure on the top of the
wing.
On aircraft, the leading edge slots and slats are designed to control and make
use of upwash. Trailing edge flaps act like the big feathers on the trailing edges
of a bird’s wings. They help trap the flow and thus increase pressure under the
wing and they also extend the wing’s curved surface and hence the region of low
pressure on the top of the wing.
13 See Section 15
25
5 Bernoulli flow and Coandǎ flow
5.1 Bernoulli’s equation
For an incompressible14 fluid in steady[21] flow, a simple expression for the con-
servation of energy was derived by Daniel Bernoulli in 1737 in his book “Hydro-
dynamica”. In steady flow, the fluid can be enveloped in an actual or virtual tube.
That means that at any cross-section perpendicular to the tube’s walls, the fluid
has a uniform velocity across the tube, i.e. there can be no shear in the fluid. Fluid
neither leaves nor enters through the wall of the tube and the particles do not inter-
act with each other or with the wall of the tube. And finally, since the flow must
be laminar, the tubes themselves, actual or abstract, are restricted to a smooth,
gently varying shape. These assumptions preclude turbulence or eddy formation.
If these conditions hold to a good approximation, Bernoulli’s equation holds. If
such a tube cannot be drawn, the equation does not hold. Bernoulli’s equation
allows the calculation of general behavior but because of these assumptions the
theory is not able to predict other aspects of fluid dynamics, such as behavior in
the boundary layer of a surface in the flow.
Bernoulli’s equation is:
1
Energy Density = ρv2 + ρgh + p. (19)
2
where
p is the absolute pressure,
ρ is the density of the fluid,
g is the acceleration due to gravity,
h is the height in the gravitational field and
v is the velocity vector for a cell in the flow small enough so that the veloc-
ities of the particles in the cell are approximately equal.
14 See,
however http://www.efunda.com/formulae/fluids/bernoulli.cfm for a more general form
of the equation which describes the behavior of certain types of compressible fluids.
26
Note that the assumption that the flow consists of these cells amounts to the fluid
approximation. In the particle view the existence of these cells is not assumed
and the macroscopic fluid velocity is superimposed on thermal components of
the particles’ velocities. When the flow is incompressible and steady, the energy
density is conserved in the flow.
Bernoulli’s equation is an expression of the conservation of energy, a checksum

that is very useful in the calculation of the properties of a steady flow. It does
not speak to the question of cause and effect however. Fluid flow is caused by a
pressure differential or direct collisions of the fluid particles with surfaces, e.g. a
propellor, and is sustained according to Newton’s first law since the particles are
massive. In some circumstances a flow can also give rise to a pressure differen-
tial, the cause of the Coandǎ effect. Just because a fluid is flowing does not mean
that the pressure within the fluid has decreased. Velocity is relative to the inertial
frame where it is measured but pressure is a quantity independent of the inertial
frame where it is measured. The pressure in a fluid is measured as the momen-
tum transfer of the fluid particles striking some transducer that produces a pointer
reading. The force that moves the pointer is the integral over the (oriented) surface
area of the transducer of the rate the fluid particles transfer momentum to it.
dv
Pointer Reading ∝ F orce = m × ∑ ⊗ dA, (20)
dt
transducer
sur f ace
where m and v are the particle’s mass and velocity and dA is an area element.
We assume that the particle collisions with the surface are perfectly elastic, so the
tensor product, ⊗, gives a result normal to the surface element, dA.
The orientation of the transducer surface in the flow affects the pressure reading.15
The tensor product between the area tensor, A and the particle velocity v in Equa-
tion (20) is a force which the transducer converts to a pointer reading. A careful
examination of a common Pitot tube used to measure the speed of an airplane
(Figure 11) will show that the speed is measured as the difference in pressure be-
tween pressure sensor areas that are normal to one another in the same flow. (In
the figure, V is the velocity of the aircraft.)
15 A Michaelson interferometer with a vacuum chamber in one leg can be used to measure air
density from which the pressure can be calculated from thermodynamic principles. It does not
measure pressure directly however.
27
Figure 11: Pitot tube
A Pitot tube is a device to measure air speed, i.e., the velocity of the tube with
respect to the local ambient air. If the tube is correctly mounted on an aircraft
flying in air that is not moving with respect to the earth, it measures, after altitude
correction, the ground speed of the aircraft. It’s design makes use of Bernoulli’s
relation. It actually consists of two concentric tubes. The outer tube is welded
to the rim of the inner tube at one end and both tubes are sealed at the other
end except for a manometer or other relative pressure gauge which is connected
between the two tubes. A close examination of one design of a Pitot tube will
reveal small holes in the side of the exterior tube. These holes are exposed to
the air flow. In order for the device to work correctly, it is very important that
the surfaces of these holes be parallel to the flow so there is no ram pressure or
rarefaction of the air there. It is the pressure in the outer tube that is compared to
the ram pressure in the center tube. This pressure remains at ambient no matter
what the air speed, even zero.16 It is the ram pressure in the inner tube that changes
as the airspeed changes.
Although Bernoulli’s equation employs densities as factors in the potential and

kinetic energy terms, the equation in this form is only valid when the fluid can be
assumed incompressible and non-viscous because compression heating and vis-
cous interactions create heat energy. To account for this energy, thermodynamics
would have to enter the equation and a thermodynamic process be identified. This
process could vary in many different ways, depending in detail on the specific
case. It is for this reason that there is no heating term in Bernoulli’s equation. If
16 Ambient pressure is a function of altitude and so a correction must be made to the Pitot tube
reading.
28
compression is significant, Bernoulli’s equation in this form cannot be expected
to hold.17
5.2 Bernoulli at the particle level
Strictly speaking, Bernoulli’s equation does not apply over a real free surface
because particles will move lateral to the flow after striking protuberances on the
surface, violating a Bernoulli assumption.18
Think of a pressure vessel of a non-viscous gas feeding a Bernoulli tube (a real

one, glass). Before flow starts, the energy in the vessel is equally distributed
between the 3 degrees of freedom. When the fluid is vented into the tube, the
pressure in the tube is less than that in the vessel. If the venting is sudden, a
pressure wave will travel in the tube at the speed of sound and thus will precede
the air flow.
The reason that the pressure in the exit tube is less than in the vessel is that the
only particles that exit into the tube are those with velocity components in the exit
direction. Of course these particles exert a transverse pressure lower than that of
the vessel since they are selected for their momentum components being outward
into the tube. Because energy is conserved, these particles’ initial energy density
is now apportioned between pressure on the walls of the tube (the pressure read
by manometer) and the kinetic energy density of their velocity in the tube, 12 ρv2 .
This means that there will be a lower manometer reading in the exit pipe than in
the vessel. The pressure difference between the vessel and the end of the exit pipe
allows the flow of the exiting particles. At the particle level, Bernoulli’s equation,
where the exit tube is in the x-direction, is:
1 1
pvessel = ρ ∑(v2x + v2y + v2z ) = ptube + ρ ∑ v2x , (21)
2 2
where the sums are over the velocities of the all particles in the flow and
17 See http://www.efunda.com/formulae/fluids/bernoulli.cfm
18 Ifthe diameter of a real tube is much greater than the size of the wall’s microscopic protuber-
ances, the tube is a Bernoulli tube to a good approximation, however.
29
1
ptube = ρ ∑(v2y + v2z ) (22)
2
is the pressure at the tube wall.
At the exit orifice, it is just those particles that are moving toward the hole that
actually exit. The hole is a sorting mechanism hence the entropy decreases in the
exit flow. This sorting process at the exit selects particles that will give a lower
pressure when that pressure is measured at an orifice whose plane is parallel to the
flow, such as a manometer connection.
5.2.1 Venturi’s tube
Figure 12: Venturi tube
Consider a level (∆h = 0) Venturi tube (Figure 12) connected between two large
pressure chambers, one at pressure p0 and the other at pressure p1 where p1 < p0 .
The cross-sectional area of the tube varies from some A0 to a smaller area, A1 in
the throat. The flow velocities are v0 and v1 . Assume that both diameters are much
larger than the microscopic roughness of the tube wall. Say further that the fluid
flow is isothermal and inviscid, i.e. steady, and that all collisions, particle-particle
and particle-wall are perfectly elastic. This means that Bernoulli’s equation holds
approximately, i.e., energy density is conserved in the flow and the volumes V1
and V0 are equal since the mass flow rate conserved.
What does this mean at the particle level? A manometer reading is caused by the
transfer of momentum of particles impinging on its transducer, i.e. a column of
30
liquid, a diaphragm or some other object whose reaction is converted to a pointer
reading. These recording devices convert the transfer of the particles’ transverse
momenta to a force normal to the transducing surface.
When an orifice is opened in a pressure vessel, it sorts out and allows to exit those
particles which are at the orifice and which have velocity components in the direc-
tion of the plane of the orifice. The components of the exiting particles’ velocities
normal to the orifice will necessarily be smaller than those of particles which do
not exit. (See Equation (21)) If a manometer is fitted to the wall of the tube, the
transverse pressure can be measured at the entrance. As the tube’s diameter de-
creases, there is a further sorting process so that the pressure in that section is
lower still. Particles in the tube that are outside the imaginary projection of the
narrow tube back into the larger section, will strike the curving wall of the neck
and interact with other particles. They bounce off elastically with undiminished
energy and with a change of momentum. They will then energize the particles
near the small-diameter exit tube. The result of these interactions is the conserva-
tion of energy and the transfer of the energy in the annulus to the particles in the
smaller tube.
Rewriting Equation (19) with h = 0 and adding some more detail, we get
1m 1m
∑(v2x + v2y + v2z )0 = (v2 + v2y + v2z )1 , (23)
2 V0 2 V1 ∑ x
,
where the sums are over the particles in V0 and V1 respectively. With the main
fluid velocity in the x−direction, the conservation of mass yields
A1
(v̄x )0 = (v̄x )1 . (24)
A0
where v̄ is the average velocity component.
Further, since V0 = V1 = V we replace ∑ v2 by N v̄2 where N is the number of

particles in the volumes V0 and V1 , and put ρ = mN V . The pressures measured by
manometers in V0 and V1 are, respectively, p0 = 2 ρ(v̄2y + v̄2z )0 and p1 = 12 ρ(v̄2y +
1
v̄2z )1 so, with some algebra, we have the Venturi relation,
31
v
u 2 (P − P )
u 0 1
(v̄x )1 = u 2 . (25)
t
ρ 1 − AA10
It is clear from this development, then, that the higher velocity in the Venturi
throat is not the cause of the lower pressure there. The lower pressure and the
higher velocity are both due to the sorting effect of the narrowing tube and the
complex interactions of the particles as they enter the throat.
There is an interesting result of Bernoulli’s equation in the form of Equation (23)

and a result of Statistical Mechanics. According to Statistical Mechanics, the root
mean square of the molecular speed, vrms is related only to the temperature, T and
the molecular mass, not to the pressure, P.
1 2 3
mvrms = kB T,
2 2
where kB is Boltzmann’s constant, and thus no matter what p0 is, (v̄x )0 cannot
exceed the molecular speed, vrms , corresponding to the temperature T ! This is
why it is important and fortunate that the high pressure in a rocket motor is created
as an effect of high temperature. The beneficial effect of high temperature in
rocket propulsion will be seen below in Section 9.
5.2.2 The two-fluid atomizer
Atomizers are often cited as examples of devices that make use of the Bernoulli
principle. Figure 13 is an illustration of what is commonly understood to be a
two-fluid atomizer.
The vessel on the left represents a flexible bulb filled with air. The pressure p1
is created when someone squeezes the bulb. Although the molecules of air are
moving in the bulb, their motion is random in every direction so, at the macro-
scopic level and under the fluid approximation, the velocity v1 = 0. When the air
is flowing steadily in the horizontal tube, however, pressure p2 and velocity v2
develop.
32
Figure 13: Theoretical atomizer
The vertical tube is connected to the atomizer and its lower part is submerged
in the liquid meant to be atomized and mixed with the air. The air velocity in
the vertical tube is zero. Finally the mixture is expelled into the atmosphere at
pressure pa . Below are the parameters in the regions of the apparatus:
p1 is the pressure in the bulb,

p2 and v2 obtain in the horizontal tube,
pa is the atmospheric pressure and
h is the liquid height in the vertical tube.
If we assume that the air density, ρ, is constant then Bernoulli’s equations become
1
p0 = p2 + ρv22
2
and
p2 + ρw gh = pa ,
where ρw is the density of the liquid in the reservoir.
33
These relations result in
1
pgauge = ρv22 − ρw gh (26)
2
where pgauge = p1 − pa is the gauge pressure in the bulb and g is the acceleration
due to gravity. Equation (26) shows that the height of the liquid is a function of
both the air velocity, v2 , and the gauge pressure, pgauge , in the bulb. This relation
also predicts that if the bulb is squeezed suddenly, the height, h, will be negative.
The pressure pulse travels a the speed of sound but the atomizer can only function
correctly when v2 has had time to build up. Figures 14 and 15 illustrate this dual
dependence.
Figure 14: Air velocity in the tube
Figure 14 shows the air velocity in the tube, as a function of pgauge , required to
lift the atomizer liquid a distance of 5 cm. The next figure, Figure 15, shows the
volume rate of change in the bulb required as a function of D, the tube diameter.
Even for the smallest tube diameter, 1mm, the volume rate of change in the bulb
must be about 20cm3 /sec to raise the liquid 5cm. This is a clue as to why conven-
tional atomizers are not designed like this.
34
Figure 15: dV /dt to achieve the correct velocity
5.2.3 Conventional atomizer
In many physics books Figure 13, or its equivalent, is displayed as a schematic

of an actual two-fluid atomizer. A search of patent disclosures for two-fluid at-
omizers reveals, however, that none of them is designed like this. Figure 16 is a
diagram of a conventional two-fluid atomizer.
The pressure is applied from the ball on the left. High-pressure air exits the inner
tube on the right and sweeps first the air from the conical volume and then the
fluid from the chamber connected to the reservoir as it is “drawn up” by the low
pressure created in the dotted area. This process is detailed in Figure 17.
The particles with arrows exit the high pressure tube and collide directly with the
molecules of the other fluid, mixing with them and sweeping them away and out
of the nozzle. In order to achieve the best mixing and delivery, the detailed design
of this region is very important.
It is also important to note that, in this mixing and ejection region, Bernoulli’s
assumptions do not hold so Bernoulli’s equation cannot be used. Recall that
Bernoulli’s equation assumes that there is no interaction between the fluids nor
35
Figure 16: Real atomizer
Figure 17: Nozzle Detail
with the surfaces in the flow. In the case of a two-fluid atomizer, however, mixing
and delivery are the two important goals.
36
5.2.4 Flit gun
Figure 18 is a diagram of the famous “Flit Gun” bug sprayer.
Figure 18: Flit Gun
The air stream exiting the gun passes over the tube and interacts with the fluid in
it. First a partial vacuum is created in the tube then the liquid rises in the tube
and mixes with the air exiting the pump. The angle, θ < 90◦ , is crucial as are the
distances, s and d. If, for example, θ ≥ 90◦ , the liquid will be forced back into the
reservoir. Bernoulli’s equation does not hold for this extremely unsteady flow.
5.2.5 Flow into an expansion chamber
Figure 19 shows an apparatus in which high-velocity air from a low pressure re-
gion flows into a chamber at lower pressure. We refer to the sections in the figure
as Sections 1, 2, 3, 4 and a, the region under atmospheric pressure.
It is important to remember Equation (21) when thinking about the pressures in the
different sections. Bernoulli’s equation, Equation (19), does not rule out p3 < p4
in Figure 19. In fact that happens when the pressure, p1 , is high enough. It
may seem counterintuitive that the fluid can flow from a lower pressure into a
37
Figure 19: Restricted exit orifice
higher one but the forward momentum of the high velocity particles incoming
from Section 3 is greater than that of those already in Section 4, i.e.,
(mvx )3 > (mvx )4 ,
and thus the particles entering win in the contest of collisions with the particles
in Section 4 and actually enter that section. In contradistinction to the situation
upstream where the pressures are decreasing and the interparticle interaction is
negligible, there are now many collisions between particles. In the process, the
energy of forward motion of the entering gas is in part converted into internal
energy of the gas, i.e. heat, pressure and molecular energy of vibration and ro-
tation, depending on the structure of the gas molecules. We can call this partial
thermalization because some of the kinetic energy of the molecules in translation
into Section 4 is converted to heat and internal molecular excitation, while some
appears as macroscopic eddies and turbulence.
Later, in Section 9 of this paper, we will see how the temperature of the fluid
flowing into a region like Section 4 of Figure 19 but with an open end to the
atmosphere can cause a thrust enhancement in a rocket engine.
38
5.3 The Joule-Thomson effect
The above situation lies somewhere between the case of Bernoulli flow, i.e. no
particle interaction in a smoothly flowing fluid, and the behavior of a real, self-
interacting gas which produces the Joule-Thomson[4] effect. In this latter regime
the effects of particle collisions are of paramount importance.
Instead of Equation (19), we will use the form below for the specific energy den-
sity as the conservation law.
1 2
v + w = constant (27)
2
where w is the enthalpy,
w = ε + p/ρ (28)
and ε is the thermodynamic energy per unit mass of the gas. This is the energy of
interatomic oscillation (in case of polyatomic gases) and rotation of the molecules
as well as their potential energy due to the van der Waals forces between them.
In the case that the flow velocity vanishes, v = 0, equation (27) expresses the
conservation of Enthalpy. Such a case is obtained in the apparatus shown in Figure
20.
Figure 20: Joule-Thomson apparatus
39
In order that the thermodynamic process be adiabatic, i.e. no heat flowing in or
out, the apparatus is insulated. The piston on the left provides pressure, P1 at
temperature T1 , and the gas moves to the chamber on the right, at pressure P2
and temperature T2 by passing through a so-called “porous plug” made of packed
granules of a chemically non-interacting substance, for example frit made of sil-
ica. As the gas molecules bounce in chaotic fashion against each other and the
frit particles, the flow velocity is transformed into random motion and thermo-
dynamic energy. The following quote is taken from the Wikipedia article on the
Joule-Thomson effect:
“As a gas expands, the average distance between molecules grows. Because
of intermolecular attractive forces (Van der Waals forces), expansion causes
an increase in the potential energy of the gas. If no external work is ex-
tracted in the process and no heat is transferred, the total energy of the gas
remains the same because of the conservation of energy. The increase in
potential energy thus implies a decrease in kinetic energy and therefore in
temperature.
A second mechanism has the opposite effect. During gas molecule col-
lisions, kinetic energy is temporarily converted into potential energy. As the
average intermolecular distance increases, there is a drop in the number of
collisions per time unit, which causes a decrease in average potential en-
ergy. Again, total energy is conserved, so this leads to an increase in kinetic
energy (temperature). Below the Joule-Thomson inversion temperature, the
former effect (work done internally against intermolecular attractive forces)
dominates, and free expansion causes a decrease in temperature. Above the
inversion temperature, gas molecules move faster and so collide more often,
and the latter effect (reduced collisions causing a decrease in the average po-
tential energy) dominates: Joule-Thomson expansion causes a temperature
increase.”[4]
A device called a “heat pump,” using an appropriate gas, can be a heater or cooler
depending on how the pressures P1 and P2 are adjusted.
6 The Coandǎ effect
This effect, first investigated and employed by the Romanian aerodynamics en-
gineer Henri Marie Coandǎ (1886 – 1972), usually refers to the phenomenon in
which an air flow attaches to an adjacent wall which curves away from this flow.
40
(see [8] pp. 42, 664). Another aspect of this phenomenon is the entrainment of
molecules far from the jet (see the three panels in 5). The attachment effect is
taken for granted and it is the separation of the flow from an aerodynamic body
that is discussed as a precursor to the stalling of the surface ([8] p. 40). The effect
can be seen in some automobile advertisements. Streamers of smoke are seen to
hug the profile of a car in a wind tunnel even as the surface of the car curves away
from the flow. This behavior indicates a lower pressure that aerodynamicists call
suction at that part of the surface. This is puzzling since there are no long-distance
attractive forces acting in a gas or between the gas molecules and a surface under
normal19 conditions. Figure 3 in section 4.2 discusses the mechanism for suction.
6.1 Organ pipe beard
The Coandǎ effect is exploited in the design of large flue pipes in some pipe or-
gans.20 These pipes are like huge whistles and can, if they are overblown, sound
the octave rather than the fundamental tone. Anyone who has played an Irish
tinwhistle knows this effect. Much of the awesome power of the grand organ,
however, comes from the volume of the bass notes. The pipe sounds when a sheet
of air is blown over the mouth. Some of this air enters the pipe and of course it
must also exit. The only exit from these closed pipes is the mouth itself. The exit
path, then, starts at the top of the mouth of the pipe. The unwanted octave sounds
when air exiting from the mouth interferes with the wind sheet entering the pipe.
How, then, to avoid this interference?
Some organs utilize what are called beards to direct the air flowing out of the pipe
away from the air entering from the air chest. A beard is a circular dowel mounted
between the ears on each side of the mouth. As the air exits, it tends to flow in the
general direction of the beard but the beard is located so that the main flow passes
over it. As the surface of the beard curves away from the flow, a low pressure is
created on the top of the beard. This low pressure area attracts the flow and keeps
it from interfering with the flow entering the pipe. Figure 21 illustrates this. The
precise location and size of the beard also affect the timbre of the pipe’s sound.
19 A fluid in liquid form behaves differently.

20 Organ builder Bill Visscher, private communication.
41
Figure 21: Beard on an organ flue pipe
6.2 The Bunsen burner
A common application of the effect illustrated in Figure 5 is the Bunsen burner.

An exploded view is shown in Figure 22 below.
The effect shown in the first panel of Figure 5 is utilized in the Bunsen burner and
any other burner where the fuel and air is mixed upstream of the flame. The needle
valve regulates the rate of fuel flow and the Coandǎ effect causes the mixing as it
draws in air from outside the burner.
42
Figure 22: Bunsen Burner
6.3 The Coandǎ propelling device
Henri Coandǎ held many patents but perhaps the most interesting for aerodynamic
design is his design of a propelling device [12]. The patent disclosure is Appendix
B. The device develops lift as an enhanced buoyant force produced by decreased
pressure on the top. This decreased pressure on the curved circular surface is
caused by a flow of gas at high-pressure exhausting tangentially to an annular
airfoil from an annular slit. In addition to enhancing the buoyant force, the device
would remove the bow wave that would hinder the motion of the device.
A bow wave is normally formed when an object moves through a fluid. It is easy
to see the bow wave of a ship or barge. As it is propelled in the water, a ship must
push water out of the way. Because the water has mass, force is required to move
43
it. By Newton’s third law, there is an equal and opposite force exerted on the ship.
This effect causes drag in addition to the viscous drag of the hull of the ship as it
moves through the water.
A toy helium balloon rises much more slowly than if it weren’t hindered by a bow
wave in the air. It is primarily the force of the bow wave that is responsible for
the phenomenon of terminal velocity. By extending his arms and legs, a skydiver
can control the terminal velocity, increasing or decreasing it. Figures 1 and 3 of
Appendix B illustrate the dissipation of the bow wave by the Coandǎ propelling
device.
7 Calculation of lift
Lift is caused by the collisions of fluid particles with the surface of the airfoil. By
Newton’s third law, this interaction of the particles with the surface results in an
equal and opposite reaction on the airflow itself; the particles bounce back. Say,
for example, that the lift force is in the “up" direction, then the third law force on
the air is “down."21 The lift can be represented in two ways: 1) as the summation
of all the forces on the surface or, according to Newton’s third law, 2) by the
negative of the force the surface exerts on the air. The latter is the approach that
led to the Kutta-Joukowski theorem.22
Figure 23 shows a typical force configuration on the surface of an airfoil.[14] The

air flow is from the left. Note that the primary contribution to the lift is from the
curved surface of the top of the wing. This lowered pressure, the so-called suction,
created there also causes the Coandǎ effect.
The flow particles far from the airfoil’s surface “feel” this suction as a sort of
reverse bow wave and, as a result, flow toward the surface. The low pressure,
maintained by the flow past the curved surface, results in a pressure gradient,
∂ P/∂ ξ , that decreases to zero as ξ increases.23 The pressure approaches the limit
p∞ , the ambient pressure. ξ is the normal distance from the airfoil’s surface.
21 Here,we are only concerned with the lift force on heavier than air craft. The situation is
different for aerostats.
22 See Reference [15] pages 236 and 237.
23 or more precisely the pressure gradient due to gravity
44
Another interesting aspect of this figure is the indication of a (conventional) bow
wave of positive pressure just below the leading edge. This bow wave results in an
upwash that moves against the main flow to join the flow above the leading edge
stagnation point. At high angles of attack this flow causes a vortex on the top
of the wing which becomes larger as the angle of attack and/or the flow velocity
increases. (See the third panel in Figure 5 above.)This vortex interferes with the
suction on top of the wing and if too large will eventually cause stall. Vortex
generators, sometimes mounted on wings and control surfaces,[5] in spite of their
name, inhibit the formation of this span-wise vortex. They do this by generating
small vortices emanating from their tips. These small vortices, for angles of attack
not too large, break up the larger span-wise vortex before it forms. The axes of
these vortices are in the direction of the flow.
A common stall warning device is a switch activated by a simple flap mounted on

the leading edge protruding forward, which, when it gets blown upward, causes
a horn in the cockpit to sound. All stall warning devices are activated, directly
or indirectly by the speed of the upwash.[13] Upwash is created by the viscous
interaction of the air with the lower surface of the wing. It can be seen as a stream
of water from a faucet strikes a plate held at an angle to the stream. Some of the
water flows upward before it finally turns and flows down the plate. If the plate
is held so the stream is near the top, the upwash will actually run up and over the
top of the plate. As we will see in Sections 15 and 4.7, this upwash can be turned
to advantage to increase lift.
Figure 23: Typical force configuration on an airfoil in an air flow
The total force on the wing, lift plus drag (the red arrow in Figure 23) is the vector
sum:
45
I
Ftotal = − (dFn + dFs ) . (2)
air f oil
where
Fn is the force normal to the surface and
Fs is the force tangent to the surface.
The minus sign is necessary because we are calculating force on the air and use
Newton’s third law to relate that to the total force, Ftotal , on the surface.
Newton’s second law is:
F = ma, (29)
or, componentwise,
Fi = mai , i = n, s. (30)
i = s denotes the component of the force and resulting acceleration along the
surface and i = n denotes the component normal to the surface.
7.1 Using Newton’s Third Law: Effects on the air caused by

the presence of the airfoil
At the surface of the airfoil, the pressure exerts a force equal in magnitude and
opposite in direction on the air. This pressure affects the air out to a distance of
∆y, often many airfoil chord lengths from the surface. Newton’s second law in
differential form is
ds dv
dFair f oil = −ρ · dA dr (31)
dt ds
46
where
ρ(s, r) is the air density in the volume dV = ds × dr × unit span.

ds/dt = v(s, r) = v(s, r) is the air speed,
v(s, r) is the velocity of the air,
dA is the differential surface area element,
r is the distance normal to the surface at ds.
s is the distance along the surface.
Again, the minus sign is required by Newton’s third law since we are interested in
the force on the airfoil.
The behavior of the air near the surface of the airfoil is very complex and chaotic
but because at angles of attack less than the stall angle, this layer, the boundary
layer, is very thin compared to ∆y, this complexity is not important. It is similar
to the behavior in the bow wave of a boat. The water is turbulent and moving in
a very complex way at the prow but some small distance away the water begins
to smooth out into regular waves that fan out as the boat passes. The information
as to the detailed behavior in the boundary layer has been lost to heat due to
the viscosity of the water. The only thing that has been propagated over a long
distance is the effect of the pressure in the boundary layer and even this vestige
of the behavior of the air near the surface will be lost at large distances from the
airfoil due to viscous heating. It is this fact that has to be ignored in the potential
flow[16] approximation, e.g., the Kutta-Joukowski theorem[10]. This theorem
requires the integral called the circulation
I
Γ≡ F · ds, (32)
be performed over a line encircling the airfoil but outside the boundary layer. Far
enough, indeed, so that the flow there is steady. In a real fluid there remains no
effect of the airfoil after much larger distances. The flow there is purely transla-
tional, with no rotation.
The presence of the surface causes shear in the boundary layer[18] so v 6= v∞ , the
flow speed far from the airfoil. (In fact on the top of the airfoil at angle of attack,
47
Figure 24: Geometry outside the airfoil
v > v∞ .) That means that the flow is not steady [21] there and Bernoulli’s equation
does not hold.
Integrating Equation 31 there results
Z ∆y I
dv
F per unit span =− dr ds ρ(s, r)v(s, r) . (33)
sur f ace C(r) ds
The contour C(r) follows the surface or outside the surface, the streamline con-
tour.
If the flow is not separated from the airfoil, the Coandǎ effect, the derivative of
v consists of two parts: ∂ v/∂ s and a geometric term that is the turning of the
velocity vector due to the curvature of the airfoil.24 Figure 25 illustrates this.
The dotted arrow in Figure 25 is the v(s + ∆s, r + ∆r) vector transported parallel
tail-to-tail with the v(s, r) vector so that ∆v can be calculated. Taking the limit as
∆s → 0, the covariant derivative of v is obtained:
∆v ∂ v v
∇s v = lim = + , (34)
∆s→0 ∆s ∂ s R(s)
24 The attached velocity field is a vector bundle over the surface of the airfoil. This surface is
assumed to be a differentiable manifold. More information about differentiable manifolds can be
found in any book on Differential Geometry.
48
Figure 25: Illustration of the covariant derivative.
where R(s) is the radius of curvature of the airfoil at ds.
We will now write the acceleration of the fluid at the surface as
ds
a = ∇s v · .
dt
We use the covariant derivative in Equation (35) below.
At the surface of the airfoil and due to its presence in the flow, the pressure causes
a force on the airfoil as well as on the air. Equation (33) then becomes
Z ∆r I
Ftotal = − dr ds ρ(s, r)v(s, r)∇s v. (35)
sur f ace C(r)
where
ρ(s, r) is the air density,
v(s, r) is the velocity of the air,
v(s, r) = ds/dt = v(s, r) is the air speed.
The presence of the surface causes shear in the air around it[18] so v(s) 6= v∞ , and
v∞ is the flow speed far from the airfoil. That means that the flow is not steady
49
[21] there and Bernoulli’s equation does not hold. This region is the boundary
layer and its thickness is δ .
Notice that the circulation,
I
Γ≡ F · ds, (32)
doesn’t arise in this derivation. In its place we have Equation (2).25 We are looking
at the effect on the air flow of the complex behavior of the air at the surface and in
the boundary layer. This effect exists as a reaction to the lift force. Equation (35)
replaces the Kutta-Joukowski theorem.
The difficulty is in evaluating the integrals in Equation (35). As has been noted
above, shear cannot be neglected. The pressure, even outside the boundary layer,
is not constant. The boundary layer is defined as that space, thickness δ , just
outside the airfoil surface where
∂ v(s, r)
'0
∂r r=δ
and r is in the direction normal to the airfoil.
The behavior of the air in the boundary layer may be complex but for laminar
flow over a non-stalling airfoil, its behavior results just in shear and a pressure
gradient. The density, ρ, is actually a function both of s and r. What value should
be assigned ρ? We are concerned with the cause of lift, i.e., the forces on the
surface of the airfoil. Our understanding is in terms of Newton’s laws.26
Figure 26 is a drawing of an airfoil that causes bending of the flow. Notice that
the bending is not just the deflection of the air by the lower surface of the airfoil.
The flow along the top is bent also. The cause of the bending of the flow over the
25 The circulation would be the first term in this equation, i.e. the integral of the forces normal
to the airfoil. Notice that this integral is a mathematical object and does not indicate that there is
a physical vortex around the airfoil. The resultant force on a buoyant object fixed in a quiescent
fluid is written the same as the circulation integral except that the integral is over the surface of the
3-dimensional object. In this case there is no flow at all.
26 When Isaac Newton was asked why the apple falls as it does, he is reported to have replied:
Hypothesis non fingo! That is, “I have no idea!” We don’t go any deeper than Newton’s laws.
50
Figure 26: Bending of the airflow by an airfoil.
top of the wing is also the cause of the Coandǎ effect. The behavior of the flow far
from the surface of the airfoil is affected by the complex interaction of the surface
of the airfoil with the molecules making up the flow.
The pressure at the surface of the airfoil, not the third law behavior of the flow far
from the surface, is what actually causes the lift and drag. If a method could be
developed to compute this pressure, then lift an drag could be computed from first
principles.
7.2 Using Newton’s Second Law: Effects on the airfoil caused

directly by air pressure
The lift force is due to the differential in pressure between the bottom and the top
of the airfoil. If we assume that the air is approximately an ideal gas, Equations
(3) and (4) show that the pressure, p, at a given temperature and the mass density,
ρ, are proportional.
First, assuming an ideal gas, notice that the mass density of the air is
nNm
ρ= ,
V
where
n is the number of moles of the gas in volume V ,
N ∼ 6.02 × 1023 is Avogadro’s number,
51
m is the mass of one particle.
This leads to,
ρ
p= × RT (36)
Nm
where
p is the pressure,
R is the gas constant and
T is the Kelvin temperature.
Some standard methods of increasing the pressure on the bottom of the wing are
angle of attack, of course, and trailing edge flaps.
But we have seen above that the most important contribution to lift, in the usual
range of angles of attack, is the suction at the top of the wing. This decrease in
pressure as the air density is reduced is the cause of the Coandǎ effect there. It is
due to the decrease in ρ, the decrease in the particle density in the boundary layer.
In a given constant volume, V, just above the surface of the wing, particles enter
and leave. Let η(s) be the particle density at s on the top of the wing. According
to Equation (3) we can write
p = ηkT, (37)
where η is the particle density. We submit the following model for the particle
density in the boundary layer of a surface curving away from the main flow,

v0
η(s) = η0 ℑ × Π(s). (38)
v(s)
where
52
Figure 27: Coandǎ effect geometry.
η(s) is the particle density at point s on the surface,
η0 is the ambient particle density,
ℑ is a dimensionless function,
v0 is the velocity of the main flow,
v(s) is the velocity of the flow just outside the boundary layer (bl) at point s
on the surface and
Π(s) = Pr(scatter into bl) is the probability that a particle in the boundary
layer will be scattered back into it.
Figure 27 depicts a portion of the curved part of the airfoil surface.
The main flow is from the left. Since they are stagnant or nearly so, the particles
in the boundary layer, thickness δ , will be scattered. We assume that all scattering
is in the direction of the main flow and all angles are equally probable. Those par-
ticles that are scattered into angle τ remain in the boundary layer. The probability
for a boundary layer particle to be scattered back into the boundary layer is
Z R+δ
1
Π(s) = τ(ξ , R(s))dξ . (39)
π R
Figure 27 yields,
53

−1 R
τ(ξ , R(s)) = sin . (40)
R+ξ
Combining Equations (37) through (40) there results
Z R+δ
η0 kT v0 −1 R(s)
p(s) = ×ℑ sin dξ (41)
π v(s) R R+ξ
for the pressure on the top of the curved wing. At or near zero angle of attack
the pressure on the bottom of a flat wing is approximately ambient so the total lift
force acting on the wing is
Z Z R+δ
η0 kT v0 −1 R(s)
Fdirect = S− ℑ sin dξ ds . (42)
π ↑ v(s) R R+ξ
where ↑ indicates the top surface of the wing and S is the area per unit span of
the bottom surface. Since we consider the simple case of a flat bottom surface
parallel to the flow, the force there is just the ambient pressure times the area. The
pressure on the upper surface, lowered by the action of the main flow shearing
past the curved surface, is responsible for most of the lift at zero angle of attack,
Φ = 0.
Figure 4 in Section 4.2 hints that the flow, v(s), over the curved surface consists
mainly of boundary layer particles activated by interaction with the main flow.
The function ℑ needs to be determined. Its form will have to do with the detailed
structure of the surface and the shape of the molecules of the flow.
8 Stalling wing
As the flow velocity increases, the secondary collisions that affect the flow par-
ticles slowed by collisions with boundary layer particles cause the slowed flow
particles to be more violently knocked back toward the surface by the main flow.
When they start to hit the surface itself, the forces on the surface there increase
(see panel 3 in Figure 5). The wing is stalling. As the velocity increases still
54
further, the flow near27 the surface reverses itself and flows back along the wing
and also increases the boundary layer population, and hence the pressure, there.
The flow is said to separate at the point where the backward flow rate equals the
forward fluid velocity. Downstream from this point, a vortex has formed.
9 Rocket engine diffuser
A rocket engine is composed of three basic parts: the pressure chamber, the throat
and the exit horn or diffuser. The following is Bernoulli’s equation applied to the
thrust of a rocket engine with no diffuser, just an exit orifice of area A:
1
pengine = pori f ice + ρvexhaust 2 (43)
2
The thrust, then, is
1
T hrust = (pengine − pori f ice ) × A = A × ( ρvexhaust 2 ). (44)
2
The thrust is caused by the pressure, i.e., particle collisions on the side of the
engine away from the throat, that is not offset by the throat pressure. It is not
caused by the mass flow rate in the exhaust. If it were confined with no exit
orifice, the gas would have provided pressure offsetting that at the other end of
the engine and the thrust would be zero in the above equation. In the absence of
a diffuser, thanks to Bernoulli’s equation, the thrust can be calculated using the
exhaust velocity and density but these are not what cause the thrust.
The thrust of a rocket engine without diffuser is easy to calculate but is much
enhanced by the non-Bernoulli flow in the diffuser.
27 Atthe surface the flow velocity is much less than the main flow velocity. The derivative of
the flow velocity in the direction normal to the surface vanishes at the so-called separation point
on the surface. Downstream of this point the particles just above the surface are flowing in reverse
of the flow.
55
Figure 28: Detail at the Diffuser Wall
9.1 The diffuser
The thrust of a rocket engine is substantially increased if the exhaust gases exit
into a diffuser[8]. Diffusers are prominent in films of rocket launches and may be
examined in aeronautical museums like Le Bourget outside Paris, France. As the
rocket exhaust exits into a parabolic chamber, it spreads to fill the entire volume
of the chamber[8] and, due to the fact that the gases in the diffuser, i.e., down-
stream from the orifice, are highly interacting, Bernoulli’s equation no longer
holds. A symptom of this breakdown of Bernoulli flow is that the exhaust ve-
locity is no longer uniform across the surfaces normal to the exhaust flow. The
exhaust spreads because of the large transverse thermal velocity components of
the hot gases. This can be seen as a sort of transverse pressure as is measured in a
Venturi tube. Combining Equations (21) and (43) we see that just at the orifice
1
pori f ice = ρ ∑(v2x + v2y ).
2
where the sum is over all the particles at the orifice. It is this transverse pressure
that drives the exhaust to the diffuser wall. As the hot gases collide with the
diffuser wall, they produce a force normal to that surface. The shape of the diffuser
is designed to provide force components in the direction of the rocket’s velocity.
(See Figure 28.)
In the absence of the diffuser, the extremely energetic exhaust gases would carry
much of their energy away. In fact, one can see this as the rocket plume spreads
56
beyond the diffuser at high altitude in a sort of umbrella shape. The diffuser, then,
extracts work from the gases before they exit beyond the rocket. The analogous
situation in an internal combustion engine is to delay the opening of the exhaust
valves in order to extract more work from the hot gas in the combustion chamber.
An interesting feature of some diffusers, like those of the French Ariane rocket,
is a sort of rifling on the diffuser wall. This rifling causes the gases to swirl, thus
increasing the path length for the exiting particles. This keeps them in contact with
the wall longer allowing more of their heat energy to be converted into thrust.
10 The high-bypass turbofan jet engine
By the 1950s the turbojet had largely replaced the piston engine driven propeller as
the main means of aircraft propulsion. The advance in the design of heat-seeking
missiles, however, was becoming a serious threat to military aircraft due to the
high temperature of the jet exhaust. The thought occurred to engine designers
that a sheath of cooler air would mask the heat signature of the jet exhaust so the
heat-seeking missiles could could not “see” it. This bypassing air, however, did
not only shield the exhaust but it increased dramatically the overall efficiency of
the jet engine in subsonic flight. Thus was born a true breakthrough in jet engine
design, a very useful spin-off of technology originally intended only for military
use.
Before the development of the modern high-bypass turbofan28 jet engine, the rim
of the entrance to the engine cowl was rather sharp. (See Figure 29.)
Modern turbofans (Figure 30) have a gently curved entrance duct. This may seem
trivial but it has the effect of greatly increasing the entrance aperture for the tur-
bofan due to the Coandǎ effect. This cowl creates a bow wave as it moves through
the air. The presence this bow wave at subsonic speeds29 creates a flow pattern
that precedes the aircraft. Examination of the third panel of Figure 5 shows how
the presence of a curved surface in an air flow causes what might be called Coandǎ
entrainment of air from outside. This entrainment compresses the air in front of
28 In high-bypass turbofan engines, most of the air entering the intake cowl bypasses the the jet
engine itself. This air is compressed somewhat by the shape of the entry cowl and its attendant
bow-wave. The compressed air is propelled by a fan driven by a conventional turbojet engine and
then exits the cowl in an annulus at the rear.
29 At supersonic speeds this effect disappears because the aircraft has outrun its bow wave.
57
Figure 29: Early turbojet
the fan so that the fan is moving air of an increased density, ρ. This increased den-
sity and the boost in velocity provided by the fan results in an increase in thrust
pressure p.
A similar Coandǎ entrainment enhances the performance of the shrouded tail ro-
tors (Figure 31) that are used on some helicopters.[24] Figure 31 is used with the
kind permission of Burkhard Domke.30
11 The vortex refrigerator
A device called a vortex refrigerator consists of a cylindrical chamber into one

end of which a gas is injected tangentially. Gas is then drawn off, cold, from the
axis of the cylinder and hot from its periphery. Along the length of the cylinder
the cold molecules are separated from the hot by the vortex process outlined in
Figure 9.
The temperature of the cold air exiting from a vortex generator can be substan-
tially below that of the compressed air at the inlet.31 Although the vortex process
described above would serve as a sort of “Maxwell’s dæmon” to separate the cold
particles from the hot there may be some other process that changes the distribu-
tion of the particles’ energies, e.g., increases it variance. It is easy to imagine that
30 http://www.b-domke.de/
31 See the websites of manufacturers of vortex refrigerators:
http://www.exair.com and also http://www.airtxinternational.com
58
Figure 30: The entrance cowl for an Airbus A380 turbofan engine
Figure 31: Ducted fan tailrotor
some of the translational energy of the inlet air would be converted into heat of the
exhausting air but it is more difficult to understand how the cold air exiting could
be as much as 28◦ − 50◦ centigrade below the inlet temperature. Unfortunately,
the measured temperature is a macroscopic quantity. It would be interesting to see
59
Figure 32: Vortex tube schematic
Figure 33: Vortex tube flow
if the inlet energy distribution, i.e. kinetic plus heat, is the sum of the cold and hot
exhaust distributions.
The centers of hurricanes are regions of low pressure. In the great hurricane of
1900 that struck Galveston, Texas, the pressure was the lowest ever recorded up to
that time, 936 millibars.32 The pressure recorded in the eye of hurricane Katrina
which hit the coast of the Gulf of Mexico in 2005 was even lower than this at 920
millibars. If the hurricane is over water, this low pressure causes storm surge. The
water in the center of the hurricane is pushed up because of the low pressure there
and the higher pressure outside the center. In the case of Galveston and hurricane
Katrina and most other hurricanes, this storm surge caused most of the damage to
the cities. The mechanism that causes the structure of hurricanes and cyclones is
not fully understood33 but it certainly is a result of complex particle interactions.
The flow is definitely not steady so Bernoulli’s equation will not hold.
32 Average standard pressure is 1013 millibars.
33 Seethe Louisiana Homeland Security website at:
http://www.ohsep.louisiana.gov/factsheets/FactsAboutHurricaneEye.htm
60
TM
12 The Dyson Air Multiplier fan
An interesting exploitation of the Coandǎ effect is the desk fan developed by

James Dyson.[20] In Figure 34 perhaps the most striking aspect is the lack of
visible fan blades. Air is drawn in by an impeller in the base and is expelled at
high speed from a slot around the large annulus.
TM
Figure 34: The Dyson Air Multiplier fan
This air is directed over a curved surface shaped like an airfoil, creating a low
pressure on the inside of the annulus as shown in Figure 23 of Section (7). Figure
35 shows the resultant Coandǎ flow.
TM
Figure 35: Operation of the Dyson Air Multiplier fan
Note that the volume rate of the air drawn in is much larger than the flow from the
impeller in the base. Figure 36 shows a cross section of the annulus with detail of
the annular slot.
61
TM
Figure 36: Cross section of the Dyson Air Multiplier fan
13 Spinning objects in the flow: the Magnus effect
The Magnus effect is well-known by players of tennis, ping-pong, baseball, soccer

and volleyball. It is illustrated in Figure 37. Discovered by Heinrich Gustav
Magnus (1802 - 1870), the Magnus effect has been used to power a ship across
the Atlantic.34
A rotating object in a flow will generate a differential pressure which will produce
a force on the object normal to its spin axis. The pressure is higher on the side
rotating into the flow than on the side rotating with the flow. The effect is caused
as the boundary layer dragged along by the object is pressurized by the main
flow. This pressurization is caused by the collisions of the flow particles with
the boundary layer particles and the surface structure of the object. The furry
surface of a new tennis ball enhances this effect. Another means of pressurizing
the boundary layer, used on some Formula One race cars, is the Gurney flap (See
Section 14 below).
14 Gurney and Fowler flaps
Although the Gurney flap was actually invented by Edward F. Zaparka in the
1930s[27], race driver and race car builder Dan Gurney accidentally rediscov-
ered it in 1971. The height of simplicity, it is a length of aluminum angle iron
bolted to the trailing edge of an airfoil. It causes an increased pressure on the side
34 See a picture of the ship that Anton Flettner built in the 1920s:
http://www.tecsoc.org/pubs/history/2002/may9.htm
62
Figure 37: The Magnus effect
of the airfoil from which it projects. It is easy to see that it functions as a dam,
trapping air from the flow, thus increasing the pressure on that side of the airfoil.
For a race car this is done to increase the force down on the wheels, decreasing
the chance for wheel spin and increasing traction for the rear wheels. Figure 38
details this simple but very effective device as used to increase lift on an airfoil.
Figure 38: Gurney flap
Ordinary Fowler flaps[28], common on commercial airliners, cause a similar damming

of the flow on takeoff and landing, where maximum lift is needed. The top sur-
63
face of a Fowler flap also utilizes the Coandǎ effect to enhance lift. “Slotted”
Fowler flaps direct some of the high-pressure air under the wing over the rear flap
sections, thus enhancing the Coandǎ effect there.
15 Slots and slats
A slot[28] is a gap between a slat and the leading edge of a wing or between the
sections of the Fowler flaps at the trailing edge.
Figure 39: High-lift wing devices
On takeoff and landing, some airliners, in addition to Fowler flaps, use leading
edge slats and slots[28]. Designs differ from manufacturer to manufacturer but a
leading edge mechanism of some kind is always incorporated into the design of
the wings of airliners. Leading edge slats and slots direct the upwash from the
bottom of the wing at angle of attack over the leading edge of the airfoil.
A simple experiment will demonstrate upwash. Hold a flat plate at a 10◦ or 20◦
angle with respect to a smooth stream of water. As the water strikes it, it climbs up
the plate against the direction of the water flow. This effect is due to the interaction
of the water molecules with the microscopic protuberances on the surface of the
plate. Another way of stating this is to say that the effect is due to the viscosity of
the water.
64
As the slat descends into the air flowing beneath the wing, it opens a slot. (See
the 4th panel of Figure (??.) This combination directs the upwash air smoothly
onto the top surface of the wing causing a Coandǎ effect there which increases lift.
Instead of slots and slats, some Airbus planes use vortex spoilers on the leading
edges of their wings. These spoilers, like slats, are deployed only on takeoff and
landing and serve to prevent a stalling vortex from forming along the leading edge
of the wing.
16 Slurries
A slurry is a suspension of insoluble particles. Normally one thinks of the particles

as suspended in water as in a mudslide or a cement slurry that can be pumped as
if it were a thick liquid. Slurries transport the sand in river bottoms and create the
sandbars that accrue in the rivers.
Quicksand is what could be called a bistable slurry (my term). It is a mixture of

water and fine sand or clay particles. Undisturbed, quicksand appears as though it
were solid ground (state #1). In this stage the slurry particles rest on one another
due to gravity and, since the particles are denser than water, they tend to settle
below the surface of the water. Even so, at every level there are water molecules
surrounding the heavier particles. When it is disturbed, e.g. by an earthquake
or by someone’s foot, the quicksand rapidly transforms into a liquid suspension
(state #2). In a sense what happens is a phase change, like the melting of ice. The
effective viscosity of the slurry reduces radically and quickly. This takes place
as the disturbance spreads through the stage #1 mixture at the speed of sound
there. This speed can be over 1400 m/s, depending on the temperature and the
shape, size and mass of the slurry particles. The sound wave, a compression-
rarefaction longitudinal wave, spreads through the quicksand, forcing the slurry
particles apart as the water molecules move between them.
These water liquid slurries are quite familiar but there are other kinds of slurries.
Smoke is a slurry of particles entrained in air. As in the case of the river slurries,
the soot particles deposit onto the surfaces with which the smoke comes in contact.
For example, in Paris the towers, churches and other monuments must be cleaned
periodically because of the deposits of smoke particles, in large part due to diesel
exhaust from cars. When the entraining air is still hot, the smoke is buoyed by the
65
air and as the hot air rises, so do the smoke particles. A very interesting kind of
gas slurry is the superheated cloud that accompanies a volcanic eruption.
The particles in the pyroclastic cloud that accompanies the violent eruption of a
volcano, buoyed and swept along by the hot gases, are created as the bubbles of
lava magma burst. In a private communication from Dr. Haraldur Sigurdsson
of the University of Rhode Island, he describes the pyroclastic phenomenon as it
occurred in the eruption of mount Vesuvius in 79AD:
The temperature of the erupting magma and gases [was] around

900 ◦C for Vesuvius. The gas-rich magma fragments on the way to
the surface, due to bubble expansion and bubble bursting about 5400
metres below the surface in the volcanic vent. The resulting mixture
of fragmented magma (volcanic ash) and gases has a relatively low
bulk density, less than that of air initially. Upon rising above the
volcano the mixture further mixes with or draws in atmospheric air
and a hot-air balloon effect occurs, with the entire eruption column
rising to levels where the eruption cloud is neutrally buoyant.
If mixing with air is inefficient, such as in the case when the erup-
tion rate [is high] (of the order of [a] billion kg/second) then the erup-
tion column is always more dense than air and collapses and flows as
a pyroclastic flow along the ground, hot, fast and deadly. That was
the case at Vesuvius.
An interesting aspect of pyroclastic clouds and pyroclastic flow is mentioned in

Dr. Sigurdsson’s last paragraph. When the mixing with the air and hot gasses is
inefficient, then, rather than rising like a smoke column, the slurry is too dense to
be buoyed by the atmosphere and the cloud flows along the ground. This cloud,
in pyroclastic flow, is still extremely hot, ∼ 900◦C. As it flows over water it will
cause the water to boil, or cause an automobile to burn, along with its contents, as
shown in Figure 40 below.
This happened in the pyroclastic flow that developed in New York on September
11, 2001 as the twin towers collapsed.
When a volcano erupts violently a slurry is formed consisting of fine magma parti-
cles suspended in the hot gases. This suspension develops as the Brownian motion
of the gas is transmitted to the magma particles. In addition to their kinetic energy
the particles themselves have been heated internally by conduction and by radia-
tion in the volcano. If the number density of these particles in the pyroclastic flow
66
Figure 40: Bus burned by the pyroclastic flow from the World Trade Center col-
lapse on September 11, 2001.
is high enough, the flow becomes more dense than the surrounding air and so it
hugs the surface of the ground or water, or whatever surface is low in the gravita-
tional field. It is the temperature of the particles themselves as well as the kinetic
energy of the slurry particles that comprises total energy density of the pyroclastic
flow. In the case of the collapse of the towers on September 11, 2001, the primary
repository for the energy in the flow was the iron and steel droplets in the slurry,
not the heat of the gas.
17 Summary
This paper begins an investigation into how a fluid consisting of particles interacts
with itself and with solid surfaces. Although the mathematics of fluid dynamics
is beautiful, it fails to explain many mysteries that appear in the behavior of com-
mon subsonic flows. Many Physics books, for example, attempt to explain the
phenomenon of subsonic lift by using Bernoulli’s equation. It is hoped that the
brief treatment above will disabuse the reader from such an explanation.
67
18 Conclusion
Anyone who has contemplated ocean waves in their magnificence or watched soli-
tons marching upstream in channel flow cannot but be amazed at the beauty ex-
hibited by the behavior of fluids, this in spite of difficulties that defy our ability to
calculate their detailed behavior. It is said that the first stage of understanding a
phenomenon is that of its careful observation. Perhaps the next is to attempt to for-
mulate the behavior of the constituents of the phenomenon. Prior to the turn of the
19th century, it was believed that fluids were fundamental entities, i.e., not com-
posed of anything. Since Einstein’s 1905 paper on Brownian motion, however, it
has been clear that fluids are composed of molecules. To assume, then, at least
for the sake of investigation, that all fluid behavior is caused by the interactions of
these molecules and those of solid surfaces immersed in the fluid, seems natural.
These assumptions are useful even in the absence of a tractable mathematics to de-
scribe behavior at this level. The current mathematical approaches, including the
Navier-Stokes equations, make the fluid approximation[21] fundamental. Some
of the most baffling behavior of fluids, however, takes place in regimes where the
fluid approximation is not valid.
Perhaps, as computers become faster and with more and more memory, some of
these behaviors will succumb to calculation. In the meantime we observe and
contemplate and we are amazed.
References
[1] See http://www.tfcbooks.com/articles/tdt7.htm for information on Nicola
Tesla.
[2] An interesting treatment of the pitfalls in confusing fluids and aggregates

of particles is found in: Woods, L. C. The Influence of the Philosophy of
Science on Research. Lecture given at the Auckland Institute of Technology.
(1998.) See Appendix A for this lecture in full.
[3] For readers not acquainted with Ludwig Boltzmann,

See http://www-groups.dcs.st-and.ac.uk/
∼history/Mathematicians/Boltzmann.html
or any of many references to his work in Statistical Mechanics.
[4] See http://en.wikipedia.org/wiki/Joule−Thomson_effect
68
[5] Subsonic aerodynamics and performance of a smart vortex generator sys-
tem. Barrett, Ron, Auburn Univ., AL; Farokhi, Saeed, Kansas, Univ.,
Lawrence Journal of Aircraft 1996. 0021-8669 vol.33 no.2 (393-398). doi:
10.2514/3.46950
[6] http://www.paulnoll.com/Oregon/Birds/flight-slow-flapping-diagram.html
and
http://www.metacafe.com/watch/2167886/
birds_in_slow_motion_a_short/
[7] Collected Papers of Albert Einstein, vol. 2
[8] Schlichting, H. and K. Gersten, Boundary Layer Theory (8th ed.) Springer:
Berlin. (2000.)
[9] The “Coandǎ Effect”, Report No. 327, Office of the Publication Board, De-
partment of Commerce, Washington, DC (1944.)
[10] http://en.wikipedia.org/wiki/Kutta-Joukowski_theorem#Formal_derivation
[11] http://www.physics.thetangentbundle.net/wiki/Fluid_mechanics/Kutta-
Joukowski_theorem
[12] USP 2,108,652 Propelling Device. See Appendix B.
[13] Denker, John S. http://www.av8n.com/how/htm/airfoils.html section 3.5
[14] See the website of Professor Martin Hepperle: http://www.mh-

aerotools.de/airfoils/index.htm
[15] Anderson, John D. Jr. Introduction to Flight (3rd ed.) McGraw-Hill:New

York. (1989.)
[16] http://en.wikipedia.org/wiki/Potential_flow#Applicability_and_limitations
[17] Hoffman, Johan and Claes Johnson. Computational Turbulent Incompress-

ible Flow. Applied Mathematics: Body and Soul 4. Springer. (2007.) ISBN:
978-3-540-46531-7.
[18] Colombini, Marco. University of Genoa, Italy

http://www.av8n.com/irro/profilo4_e.html
[19] Vortex coolers are available for sale.

See http://www.newmantools.com/vortec/nema4.htm
for an assortment.
69
[20] http://www.dysonairmultiplier.com/
[21] Landau, L. D. and E. M. Lifshitz Fluid Dynamics (2nd ed.) (Volume 6:

Course of Theoretical Physics) U.K. Pergamon. (1984.)
[22] http://www.aerospaceweb.org/question/planes/q0176.shtml
[23] H. Riedel, ”Grundzüge der Strömungsforgänge beim Anlegen und Abreißen

eines Flachstrahls bei der Umströmung eines Flügelprofils (Coandǎ-Effekt)”
Deutsche Luft- und Raumfahrt Forschungsbericht 71-46 (1971.)
[24] Aerodynamic Performance Analysis of a Helicopter Shrouded Tail Rotor Us-

ing an Unstructured Mesh Flow Solver, Lee, H. D., O. J. Kwon, and J. Joo.
The 5th Asian Computational Fluid Dynamics [Conference], Busan, Korea,
(October 27 - 30, 2003.)
[25] Working Model allows the programming of 2D kinematic models. See

http://www.design-simulation.com/
[26] Senior thesis of Jacob Vianni, University of California, Santa Cruz. (2006.)
[27] See the interesting article:

http://www.allamericanracers.com/gurney_flap.html
[28] See http://www.centennialofflight.gov/essay/

Theories_of_Flight/Devices/TH17G5.htm
70
A On the consideration of fluids at the particle level
71
The Influence of the Philosophy
of Science on Research
L.C. Woods
(Balliol College, University of Oxford)
Lecture delivered at the Auckland Institute of Technology

on Friday, 7th August, 1998.
1
1. Introduction
When I was young, I read an article that claimed that the purpose of
scientific theory was to describe phenomena rather than to explain it. This
surprised and disappointed me. Apparently we could achieve no more than
an empirical account of the real world, and could not expect to understand
it. This conservative philosophy is known as instrumentalism, because it
maintains that a theory is no more than an ‘instrument’ for making predic-
tions. The opposing view is that theories relate to underlying mechanisms
and that these are responsible for the observed phenomena. To know the
mechanism is to understand the phenomenon. However this realist philoso-
phy usually depends on some metaphysical elements, introduced to enlarge
the fabric of the hidden world and thus to aid explanation. And it was the
liberal introduction of such unobservable elements that added force to the
instrumentalist position, an extreme form of which is known as positivism.
This holds that all statements other than those describing or predicting ob-
servations are meaningless (including this statement?). Knowledge is only
what can be verified directly. Logical positivism augments positivism by
admitting mathematical entities logically connected to observations, even
if they are not directly measurable.
An alternative title for this talk would be ‘The legacy of Logical Posi-
tivism’. This philosophy was greatly boosted in the first half of this century
by the difficulty of giving the formalism of quantum mechanics an objective
and realist interpretation. Hidden variables of one sort or another, even
including parallel universes, have been advanced as possible solutions. But
the experts have not been convinced. That debate continues and I expect
that it will ultimately be resolved in favour of a realistic explanation rather
than an instrumentalist description. In any case this unresolved difficulty
with realism at the smallest scales does not justify our rejecting it at the
deterministic mesoscopic and macroscopic scales. What I shall illustrate is
the way in which some scientists’ preference for mathematical description
over physical explanation has led to important differences in the way they
have pursued their research. Fields that are particularly vulnerable to the
legacy of positivism are thermodynamics and plasma physics. Very often
the researchers are unaware that they are labouring under the influence of
a largely discredited philosophy.
2. Some Philosophical Background

By a ‘mechanism’ I mean the representation of a real process in terms
involving familiar physical actions, e.g. we might say that the thermo-
electric transport of heat is due to the fact that higher energy electrons
have a smaller probability of colliding with ions than those of lower energy,
or that a solar prominence is supported high in the corona by magnetic
2
forces. Once we have the mechanism identified, fitting it out with suitable
mathematics is often the easier task. In specifying a mechanism, the first
objective is to try to identify those features essential to the phenomenon
under consideration. Elaboration of the model can follow when the min-
imalist position has proved itself. But it may be necessary to add some
unobservable structure to the mechanism from the outset, e.g. the mag-
netic field lying out-of-sight well below the photosphere when modelling a
sunspot structure.
Ockham was a 14th-century, Scholastic philosopher, who attacked the
supremacy of Papal power. His ‘razor’ was the statement that entities
are not to be multiplied beyond necessity. In a similar vein, Ernst Mach
(1838–1916) stated that ‘it is the aim of science to present the facts of
nature in the simplest and most economic conceptual formulations’. This
was a reaction against the metaphysical extravagances of the 17th and 18th
centuries, during which, inter alia, various fluids were adduced to ‘explain’
physical phenomena. A classical case was the chemists’ phlogiston that,
having negative weight, supposedly explained the increase in mass due to
burning—the heat drove off the phlogiston. Oxygen was yet to be discov-
ered. Heat had the properties of an indestructible fluid called ‘caloric’;
electricity was said to be composed of two fluids, with no more evidence
than this seemed to provide an explanation of some observations. But
nowadays we do talk of electron and ion fluids. Also Carnot managed to
establish the principle that later evolved into the second law of thermody-
namics, by employing the caloric concept. (It was the first law that later
destroyed the conservation of caloric.) So some metaphysical inventions
prove to be closer to the truth than at first imagined. Such elements evolve
from being metaphysical to eventually being considered to be ‘real’.
Atoms were ruled out by Mach and other anti-atomists of his day. They
could not be observed, so were not real. They could be admitted only as
a device to give economy of thought. Realists are much bolder, willing to
introduce unobserved elements and to take them as being real, in order to
provide ‘explanations’. A classic example is William Harvey’s (1578-1657)
explanation of the circulation of the blood and the function of the heart
as a pump. Although he had no microscope to see the capillary vessels
connecting the arterial and venous systems, he maintained from evidence
implying a circulation that they must exist. Pauli’s (1930) invention of
the neutrino to ensure the conservation of energy and momentum during
beta decay is another good example of a bold metaphysical creation. It
was not until powerful nuclear reactors were available that the existence of
neutrinos could be confirmed.
Analogy is a powerful, heuristic means of arriving at a possible descrip-
tion of a new phenomenon. It may be a mathematical likeness only, as with
the fact that the temperature in steady-state heat flux and the gravitation
potential both satisfy Laplace’s equation, or it can be a deeper physical
3
analogy, such as that between the transport of heat by colliding particles

and its transport by photons within the Sun. For example Einstein’s inter-
pretation of Brownian motion as being due to the uneven bombardment of
microscopic particles by molecules may have taken root in his mind from
an obvious macroscopic analogy, and its success led quickly to the full ac-
ceptance of the reality of molecules. Metaphysical inventions have a central
role in science, provided one always remembers that they are on ‘trial’ until
the indirect evidence is so strong that they can be considered to be ‘real’.
Maxwell was pre-eminent in his use of analogy. He deployed it in two
famous examples. First in his kinetic theory of gases he used a ‘billiard-
ball’ model to describe the trajectories of the molecules. This was excellent
for monatomic molecules, but partially failed with diatomic molecules for
reasons that are now obvious to us. His other great analogy was to represent
magnetic fields as vortex filaments in a ‘fluid’, and to separate them by
particles each revolving on its own axis in the opposite direction from that
of the vortices. These ‘idle-wheels’ (later identified as ‘electrons’) were to
allow the free rotation of the vortex filaments. To cap it all, he gave the
vortices elastic properties to represent the displacement current! In this
manner he arrived at his set of equations for the electromagnetic field. He
found that the velocity with which disturbances propagated through his
system of vortices and particles (70,843 leagues per sec.) agreed so closely
with Fizeau’s value for the speed of light that he remarked “we can scarcely
avoid the inference that light consists in the transverse modulations of the
same medium which is the cause of electric and magnetic phenomena”.
Then the mechanical description was abandoned, it being assumed that
the equations alone now represented the phenomenon. Maxwell’s analogy,
however absurd it seems today, led him to the greatest discovery of the
19th century, namely that light was electromagnetic in nature. The French
positivist, Pierre Duhem, observed sarcastically that Maxwell had cheated
by falsifying one of the equations of elasticity in order to obtain a result
that he already knew by other means.
However if one asserts the whole theory is simply the equations, one
is adopting a positivist point of view. It is Faraday’s great metaphysical
construction—the notion of an electromagnetic field permeating space—
that allows a return to a physical description of Maxwell’s equations. This
continuum picture is very helpful in trying to understand the interaction of
fields and particles. I shall use ‘mechanism’ in this extended sense in what
follows.
Since logical positivists eschew physical mechanisms, they are attracted
to mathematical treatments, especially when an axiomatic basis can be
adopted or devised to give the approach the gloss of pure mathematics.
The belief that, excepting blunders, mathematical proofs are absolutely
certain and therefore superior to physical arguments, should have been
dealt the coup de grâce by Gödel’s incompleteness theorem. This states
4
that if a set of rules of inference in a branch of mathematics is consistent,

then within that branch there must exist valid methods of proof that these
rules fail to identify as valid.
Of course the equations of mathematical physics follow from the as-
sumed mechanisms, but sometimes this dependence is inverted or forgotten
and the equations begin to assume an independent significance well beyond
their original range of validity. This does not seem to concern cosmologists,
whose big bang, in which the universe is created from nothing via a quan-
tum mechanical tunnelling process, is a wild, but apparently successful,
extrapolation of known physical laws.
The importance of imagining phenomena in terms of mechanisms rather
than the equations employed to represent them, is that mechanisms are of-
ten much more suggestive of modifications and extensions to more accurate
models of the processes, whereas equations, especially if they are compli-
cated as in the integro-differential equation of kinetic theory, or as in the
full set of MHD equations, are less helpful. Equations sometimes have
many terms, each of which usually represents a distinct physical process.
It is important to try to relate the terms individually to features of the
physical model and not simply to lump them together.
Phenomena in fields like biology, where the mechanisms are obscure or
unknown, and which rely on statistical data to suggest causal connections,
remain fertile for the positivists. To be true to their philosophy, they would
be content to rest the case for the dangers of cigarette smoking on the
correlation discovered between smoking and various types of illness. But
the merchants of death are realists. The cigarette manufacturers insisted
that, in the absence of proven biological mechanisms relating disease to
smoking, their product was innocent. It is true that mere correlation proves
nothing, a classical case being the noted correlation between the incidence
of prostitution in London just after the second World War and the salary
of Bishops.
3. How Positivists Confused the Basis of Plasma Physics

Whether or not the early plasma physicists knew any philosophy of
science I cannot tell, but it would appear from their mistakes that they
preferred formal mathematics to physical mechanisms. The most obvious
example is that plasma pressure was defined as momentum flux , which is
correct only if molecular collisions are sufficiently numerous. The classical
case of wall pressure being due to its bombardment by molecules should
have made the role of collisions obvious enough, but the momentum flux
definition, in which the collisions are only implicit, confused quite experi-
enced scientists into believing that there could be pressure gradients even
in a collisionless plasma. One wonders by what mechanism can purely
ballistic particles transmit a pressure force.
5
I was once challenged at a seminar I was giving at the UK Culham

Laboratory for Fusion Research. A scientist claimed that ‘Collisions are not
essential for there to be a pressure gradient in a plasma!’ His argument
was that one could have a gradient in the number density, n, of ionised
particles, maintained by a strong magnetic field (which is correct) and if
the medium were isothermal, from the law p ∝ nT relating the pressure
to the temperature T , it follows that there would be a gradient in the
pressure. It seemed plausible to the audience. They had become so familiar
with the classical pressure/temperature law that they had forgotten that its
derivation required collisions, i.e. it is not true that p ∝ nT in a collisionless
plasma. One could take it to be a definition, but then it would have no
physical content.
Although this mistake seems harmless enough, it was compounded into
a more ridiculous and even dangerous notion for a plasma in a strong
magnetic field. If the magnetic field strength B say, has a gradient in a
direction orthogonal to the field vector B, as the charged particles gyrate
about the field lines with a radius inversely proportional to B, the variation
in the radius of gyration experienced by the particles causes them to drift
in a direction orthogonal to both B and its gradient. (The motion has
a similarity to that of a top on an inclined plane—the gravitational force
down the plane results in a motion of the top along the plane at right
angles to this force.) This is known as ‘grad B drift’ and depends on the
assumption that the average time interval τ between successive particle
collisions is much greater the gyration time, ωc−1 . The individual particles
have an average velocity uB across the field determined by the value of
grad B.
Now consider the whole collection of particles treated as being a fluid .
The equation of fluid motion does not have a term involving grad B, but
it does have a term proportional to the pressure gradient, which gives rise
to a fluid velocity v across the field lines depending on the magnitude of
grad p. Now the fun begins. The average drift velocity uB it would seem,
cannot possibly be the same as v, since the former depends only on grad
B and the latter only on grad p. This means that, with appropriate choice
of the two gradients, it is possible to send the ion mass qua particles in
the opposite direction to the ion mass qua fluid . But there can be only
one direction of mass motion. The amazing thing is that this reputed
‘paradox’ is acknowledged and accepted in the literature. That there must
be an error is not even appreciated. The simple mistake is to assume that
in the guiding centre description the particles are collisionless, but in the
fluid description they respond to pressure forces, i.e. to the impact of other
particles.
Why do I think this confusion over pressure is serious? Well the ex-
tremely expensive and unsuccessful fusion energy project, in which very
hot plasma was supposed to be confined by magnetic fields, was based on
6
theory in which guiding centre motion plays a role. And the (incorrect)
equations resulting from not understanding the nature of plasma pressure
failed to reveal that there would be a disastrous loss of plasma across the
tokamak fields. But the error goes deeper than that. The basic equation
on which all of the kinetic theory of plasmas was developed is Boltzmann’s
kinetic equation, which we shall next consider.
4. Why Boltzmann’s Equation is Incorrect

In 1872 Boltzmann published the paper “Further Studies on the Ther-
mal Equilibrium of Gas Molecules” that contains his famous integro-
differential equation for the evolution of the density of particles in phase
space. At a meeting in Vienna to commemorate the centenary of this pub-
lication, G.E.Uhlenbeck stated:
‘The Boltzmann equation has become such a generally accepted and central
part of statistical mechanics, that it almost seems blasphemy to question
its validity and to seek out its limitations. It is also almost a miracle how
the equation has withstood all criticisms...’
However, that there is an important limitation to the equation becomes

evident when terms second-order in the ratio of the microscopic (molecular
collisional) time-scale to the macroscopic time-scale, known as the Knudsen
number , are examined. (A typical second-order term involves two gradi-
ents, e.g. the heat flux includes a term q2 = −α∇v · ∇T , where α is a
constant and v is the fluid velocity. The first-order theory yields the clas-
sical transport equations of Fourier, Ohm and Newton and is correct.) For
example there are physically evident terms for the heat flux across mag-
netic fields that cannot be derived from Boltzmann’s equation. On the
other hand, his equation leads to second-order terms for heat transport in
an isothermal neutral gas, in which circumstances no such transport is pos-
sible. One such term has ∇p in place of ∇T in the (correct) second-order
expression for q2 just quoted. The physical nature of energy transport in
a non-conducting gas must depend on there being a temperature gradient.
The fault with the equation lies with the assumption that the collision
rate between molecules is proportional to the product of the distribution
functions of the colliding particles, regardless of anisotropies generated by
the presence of pressure gradients and fluid shear. Boltzmann’s main pur-
pose was to find a way of deducing the second law of thermodynamics from
mechanics and also to improve on Maxwell’s derivation of the equilibrium
distribution. In this he certainly succeeded. In fact only seven of the 96
pages of his paper deal with the calculation of the transport properties of
gases. The tacit assumption that the equation was valid over a wide range
of Knudsen numbers, made by Chapman, Enskog and many others since,
is where the fault really lies.
7
In the formulation of the kinetic equation, the distinction between con-

vection and diffusion is not correctly drawn. Diffusion is due to molecular
agitation superimposed on a reference frame that not only has the speed of
the fluid element, but which also accelerates and spins with it. This means
that the pressure gradient must be assumed known from the outset, since
the fluid element to which the frame is attached, is accelerated by this force.
The spurious terms in the heat transport mentioned above, arise because
of the neglect of these fluid accelerations. For example, scattering is taken
without comment to be isotropic in a frame that has the velocity of a fluid
element, but not its acceleration. Had the mechanism on which the equa-
tion was based, namely the collisional scattering of molecules in and out
of a fully convected element of phase-space, been clearly understood and
kept in mind, the error would have been soon discovered and the original
Boltzmann’s equation corrected for higher values of the Knudsen number.
Great sums and scientific effort have been invested into finding com-
puter solutions of Boltzmann’s equation in the regime of large Knudsen
numbers. Such a waste and just because of the apparently unshakeable
belief in equations rather than mechanisms. So far as research in fusion en-
ergy is concerned, the cost of the failed tokamak machine world-wide must
exceed ten billion dollars. At least in part this waste can be attributed
to the propensity of plasma physicists to adopt a positivistic view of their
science.
5. Understanding Entropy
My final example of the legacy of positivism comes from that will-o’-the-
wisp known as entropy. What is curious about this property of macroscopic
systems is that it is a purely defined quantity, not relating to a physical
property until precise details of the state of the system have been specified.
And the problem with ‘state’ is that it is observer-dependent, i.e. it depends
on what elements the observer wishes to include in his physical model of
the system under consideration. The greater the detail, the smaller the
resulting entropy of the system. In this case the ‘mechanism’ is simply the
chosen physical state. Unfortunately the positivist position seems to be
that entropy is a property of systems, independent of the observer—the
mechanism via state is quite ignored. Examples of this are to be found in
a subject optimistically termed ‘rational thermodynamics’, in which tem-
perature and entropy are taken to be ‘primitive’ quantities not requiring
definition.
This ignorance would be harmless enough, except that ‘the’ entropy,
or rather its rate of production, σ say, is made the basis for determining
the form of constitutive relations, namely the laws relating fluxes and the
thermodynamic forces driving them. The principle is that these fluxes and
forces appear as quadratic products in σ, which it is assumed must always
8
be positive. The second law of thermodynamics is claimed to support this

view. Fourier’s law relating the heat flux q to the temperature gradient
∇T , is a simple example. In the absence of other processes, the production
rate is σ = −q · ∇T > 0, which implies that q = −κ∇T , where κ is a pos-
itive constant termed the ‘thermal conductivity’. The physical mechanism
that generates these fluxes and that is responsible for σ generally being
positive is lost in a pseudo-mathematical haze, with the hope no doubt
that the pure mathematical appearance of the formalism will impress the
followers into believing that the reasoning is unassailable.
Of course this approach leads to gross errors, perhaps the simplest of
which is a theorem due to Coleman that asserts the entropy to be inde-
pendent of the gradients of temperature and fluid velocity. In a theory
correct to second-order in the Knudsen number, this is readily shown to
be wrong. Another evident error arises in the theory of heat flux across
strong magnetic fields. There is an interesting and very important term q∧
that happens to be orthogonal to both the temperature gradient and the
magnetic field. Since q∧ · ∇T = 0, this term does not generate entropy and
does not appear in the expression for σ. It therefore does not exist for the
rational thermodynamicist. In fact it is the second-order form of q∧ that is
responsible for the failure of tokamaks to retain their energy for more than
a few seconds at best, when minutes would be required for an economic
fusion reactor. And the same transport equation plays a dominant role in
coronal physics.
The entropy production rate is assumed to be always positive. This is
presented as an axiom, with the second law as justification. Unfortunately
the second law gives no guarantee that σ is always positive. In a strong
magnetic field it is in fact the case that σ may have either sign. The
argument is as follows. We expand σ as a power series in the Knudsen
number, ε say, ¡ ¢
σ ≈ σ1 + σ2 , σi = O(εi ) ,
where we have carried the expansion only to the second-order term. Of
the two terms, only σ1 is dissipative. It is therefore always positive. The
second-order term has three gradients and is consequently reversible, i.e.
it may have either sign. Moreover it can be much larger than σ1 , in which
case the total σ may be negative. This is not a failure of the second law,
which relates only to dissipative terms. The problem is that the rational
thermodynamicist has adopted an axiom, the physical content of which he
does not understand.
It can be shown that the stability of a continuum flow requires that
σ1 ≥ 0, σ1 + σ2 ≥ 0 ,
the inequalities in which have different roles. The first determines that the
thermal conductivity and the fluid viscosity be positive quantities, mak-
ing it a thermodynamic constraint; it is universal and independent of the
9
actual flows obtained. The coefficients of the second-order terms, being

determined by the same molecular behaviour as the first-order coefficients,
are closely related to them, leaving no freedom for further thermodynamic
constraints to be satisfied.
On the other hand the stability constraint ((σ1 + σ2 ) ≥ 0) can be
satisfied only by restricting the class of fluid flows. It sets a limit on the
gradients, or equivalently on the Knudsen number making it a macroscopic
or fluid constraint. Provided ε is sufficiently small—certainly less than
unity—σ1 will be larger than |σ2 | and the constraint will be satisfied. We
might imagine some initial state, with steep gradients, for which this does
not hold. The fluid flow may be momentarily unstable, but the resulting
fluctuation will quickly restore a new equilibrium state, in which σ2 is either
positive, or smaller in magnitude than σ1 .
Since the second law of thermodynamics is not normally associated with
the question of the stability or not of the flow field, it is restricted to the
thermodynamic constraint, σ1 ≥ 0, while the constraint on (σ1 + σ2 ) is
concerned solely with the macroscopic stability of the flow field. Certainly
with magnetoplasmas, even with convergent Knudsen number expansions,
the O(ε2 ) terms dominate the classical O(ε) terms by orders of magnitude
and instabilities in which (σ1 + σ2 ) changes sign periodically do occur. The
adoption of the inequality σ > 0 as an axiom is a serious mistake.
6. Conclusions
I have tried to show that the attitude of scientists to their research—
the way they go about it—is greatly influenced by the beliefs they have
adopted, consciously or otherwise, about the nature of the scientific enter-
prise. In this way, the philosophy of science does play an important, but
indirect role in research, a point of view that would have been quite obvious
to most 19th Century scientists. The importance of philosophical concerns
in cosmology and quantum physics, in which branches of science there is
no shortage of metaphysical invention, is obvious enough, but that this
extends into the classical realms of continuum physics is not sufficiently
appreciated.
University teaching is largely responsible for the inculcation of positivist
attitudes in the typical university graduate. Mathematical approaches to
physics and the mathematical sciences are easier to teach and easier to
learn. Mechanisms are not ignored, but they are given less attention, espe-
cially where it counts for the average student, namely in the examination
room. When I was an examiner for Oxford Finals in Mathematics, I al-
ways attempted to set some questions essay-type discussions of underlying
physical principles. But these were seldom answered—they were thought
to be too difficult and in any case how can one get high marks for a mere
essay!
B Henri Coandǎ’s Propelling Device
82

Aerodynamics Particle Level: at The

Uploaded by

Copyright:

Available Formats

Aerodynamics Particle Level: at The

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Aerodynamics Particle Level: at The

Uploaded by

Copyright:

Available Formats

Aerodynamics

Charles A. Crummer, PhD

3 Total force on the surface of the airfoil 5

3.1 Physical parameters affecting the pressure on the airfoil . . . . . . 6

4 Mechanics of fluid interaction 7

4.1 Fluid flow over a flat surface . . . . . . . . . . . . . . . . . . . . 9

4.2 Fluid flow over a curved surface . . . . . . . . . . . . . . . . . . 10

4.3 Static buoyant lift: the aerostat . . . . . . . . . . . . . . . . . . . 14

4.3.1 Light gas Balloon . . . . . . . . . . . . . . . . . . . . . . 15

4.3.2 The Hot-air Balloon . . . . . . . . . . . . . . . . . . . . 17

4.3.3 Submarines and Fish . . . . . . . . . . . . . . . . . . . . 18

4.4 Vortex fluid motion . . . . . . . . . . . . . . . . . . . . . . . . . 22

4.5 Finite wings and wingtip vortices . . . . . . . . . . . . . . . . . . 24

4.6 Leading Edge Extensions . . . . . . . . . . . . . . . . . . . . . . 25

4.7 Birds in flight . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

5 Bernoulli flow and Coandǎ flow 26

5.1 Bernoulli’s equation . . . . . . . . . . . . . . . . . . . . . . . . . 26

5.2.1 Venturi’s tube . . . . . . . . . . . . . . . . . . . . . . . . 30

5.2.2 The two-fluid atomizer . . . . . . . . . . . . . . . . . . . 32

5.2.3 Conventional atomizer . . . . . . . . . . . . . . . . . . . 35

5.2.4 Flit gun . . . . . . . . . . . . . . . . . . . . . . . . . . . 37

5.2.5 Flow into an expansion chamber . . . . . . . . . . . . . . 37

5.3 The Joule-Thomson effect . . . . . . . . . . . . . . . . . . . . . 39

6 The Coandǎ effect 40

6.1 Organ pipe beard . . . . . . . . . . . . . . . . . . . . . . . . . . 41

6.2 The Bunsen burner . . . . . . . . . . . . . . . . . . . . . . . . . 42

6.3 The Coandǎ propelling device . . . . . . . . . . . . . . . . . . . 43

9 Rocket engine diffuser 55

9.1 The diffuser . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56

11 The vortex refrigerator 58

13 Spinning objects in the flow: the Magnus effect 62

14 Gurney and Fowler flaps 62

15 Slots and slats 64

A On the consideration of fluids at the particle level 71

B Henri Coandǎ’s Propelling Device 82

1 Velocity profile in the boundary layer for laminar flow . . . . . . . 4

2 Interaction between fluid particles and a real surface . . . . . . . . 8

3 Fluid flow over curved and flat surfaces . . . . . . . . . . . . . . 10

4 Behavior of particle flow over a curved surface . . . . . . . . . . 12

5 Flows over a Wing . . . . . . . . . . . . . . . . . . . . . . . . . 14

8 Submarine Ballast System . . . . . . . . . . . . . . . . . . . . . 19

9 The vortex process . . . . . . . . . . . . . . . . . . . . . . . . . 23

10 Downwash and wingtip vortices . . . . . . . . . . . . . . . . . . 24

14 Air velocity in the tube . . . . . . . . . . . . . . . . . . . . . . . 34

15 dV /dt to achieve the correct velocity . . . . . . . . . . . . . . . . 35

19 Restricted exit orifice . . . . . . . . . . . . . . . . . . . . . . . . 38

21 Beard on an organ flue pipe . . . . . . . . . . . . . . . . . . . . . 42

23 Typical force configuration on an airfoil in an air flow . . . . . . . 45

24 Geometry outside the airfoil . . . . . . . . . . . . . . . . . . . . 48

25 Illustration of the covariant derivative. . . . . . . . . . . . . . . . 49

26 Bending of the airflow by an airfoil. . . . . . . . . . . . . . . . . 51

27 Coandǎ effect geometry. . . . . . . . . . . . . . . . . . . . . . . 53

28 Detail at the Diffuser Wall . . . . . . . . . . . . . . . . . . . . . 56

30 The entrance cowl for an Airbus A380 turbofan engine . . . . . . 59

31 Ducted fan tailrotor . . . . . . . . . . . . . . . . . . . . . . . . . 59

32 Vortex tube schematic . . . . . . . . . . . . . . . . . . . . . . . . 60

33 Vortex tube flow . . . . . . . . . . . . . . . . . . . . . . . . . . . 60

37 The Magnus effect . . . . . . . . . . . . . . . . . . . . . . . . . 63

39 High-lift wing devices . . . . . . . . . . . . . . . . . . . . . . . 64