Radyadour Kh. Zeytounian Auth. Navier-Stokes-Fourier Equations A Rational Asymptotic Modelling Point of View PDF

NavierStokesFourier Equations
.
Radyadour Kh. Zeytounian
NavierStokesFourier
Equations
A Rational Asymptotic Modelling
Point of View
R. Kh. Zeytounian
Universite des Sciences et
Technologies de Lille
LML cite Scientifique
59655 Villeneuve dAscq Cedex
France
zeytounian@free.fr
ISBN 978-3-642-20745-7 e-ISBN 978-3-642-20746-4

DOI 10.1007/978-3-642-20746-4
Springer Heidelberg Dordrecht London New York
Library of Congress Control Number: 2011943087
# Springer-Verlag Berlin Heidelberg 2012

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting,
reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication
or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965,
in its current version, and permission for use must always be obtained from Springer. Violations are liable
to prosecution under the German Copyright Law.
The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply,
even in the absence of a specific statement, that such names are exempt from the relevant protective laws
and regulations and therefore free for general use.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

Preface
Rationality and Asymptotics are the two main concepts associated with the
Modelling in Fluid Dynamics, which have completely changed our look on the
Understanding of NavierStokesFourier (NS-F) equations, governing the viscous,
compressible and heat conducting Newtonian baroclinic and non-adiabatic fluid
flows.1
This Rational Asymptotic Modelling (RAM) Approach have raised, on the one
hand, further new interesting questions and potentialities for Applied
Mathematicians, in their quest of rigorous existence and uniqueness results for
the Fluid Flow problems.
On the other hand, this RAM Approach have opening up of new vistas for the
derivation, by Fluid Dynamicians, various consistent simplified models related with
real stiff fluid flow problems, as an assistance to Numericians embarked on a
computational simulations of complex problems of engineering interest with the
help of high speed computers.
In this book we touch (see, in particular, the Chap. 6) the crucial problem of a
practical (rather than formal, abstract) Mathematics for a consistent RAM
Approach, via a Postulate and, some key rules inspired from asymptotics.
This mathematics for the RAM is applied in a consistent way to modelling of
various stiff problems of the: aerodynamics (Chap. 7), Benard thermal convection
(Chap. 8) and atmospheric motions (Chap. 9).
The main lignes of the aims of this book are set out in the Prologue, and in the
Overview a brief outline of the events related with my rather long RAM
Adventure, during the years 19682009, is given.
The book is divided into nine Chapters, an Epilogue, a list of References, and a
Subject Index.
In Chap. 2, the Newtonian (Classic) Fluid Dynamics is considered as a
Mathematical-Physical Science and the reader can find in four Sections a concise
1
These NS-F equations are, in fact, the equations usually named NavierStokes Compressible
equations and assumed often barotropic equations, by the Mathematicians !
v
vi Preface
material concerning the main theoretical concepts and principles, equations and
associated initial and boundary conditions.
The Chap. 3 is devoted to a tentative description of a rational way for the
obtention, from NS-F equations, various main model equations and also to a
discussion concerning their nonuniform validity, near the initial time (where the
initial data are given) and in the vicinity of a solid wall limiting the fluid flow (where
the boundary conditions for the velocity vector and temperature are given).
The Chap. 4, is entirely concerned with the application of RAM Approach for a
justification of Boussinesq model equations, assuming that the Mach number is a
small parameter.
The Chap. 5 is an application of the RAM Approach to large Reynolds numbers
unsteady fluid flow, which leads to a complicated Five Regions Structure of
unsteady NS-F full equations.
The Chap. 6, is a central one and present a sketch of a Mathematical Theory for
the RAM Approach. As a basis for this practical Mathematics, in the realization
of our RAM Approach, the following Postulate is accepted as true, despite its
simplicity:
If a leading order an approximate simplified model is derived from a NS-F fluid flow
problem, then it is necessary that a RAM Approach be adopted to make sure that terms
neglected in a such NS-F stiff problem really are much smaller than those retained in
derived approximate simplified, no-stiff, leading-order consistent model problem.
The Chap. 7, is concerned with the two applications of the RAM Approach in
Aerodynamics. First, the derivation of a through-flow model problem, for a fluid
flow in an axial compressor, when the blades in a row are very closely spaced.
Secondly, the low Mach number flow of a gas within a cavity which is changing its
shape and volume with time.
In Chap. 8, The RAM Approach concerns the famous Benard convection prob-
lem for a liquid layer heated from below. In particular, the following alternative is
demonstrated:
Either the buoyancy is taken into account, and in this case the free-surface deformation
effect is negligible and we rediscover the classical leading-order Rayleigh-Benard shallow
convection, unless viscous dissipation, rigid-free, problem, or the free-surface deformation
effect is taken into account, and in this case at leading-order, for thin films, the buoyancy
and viscous dissipation effects does not play a significant role in the so-called Benard-
Marangoni thermocapillary instability problem.
But, if you have intend to take into account, in the case of a deep liquid layer, the
viscous dissipation effect according to Zeytounian in equation for the temperature, then
it is necessary to replace, the Rayleigh-Benard shallow convection equations, by a new set
of equations called deep convection equations with a depth parameter.
The last Chap. 9 is devoted to atmospheric motions. First, we derive for 2D

steady lee-waves problem, in a baroclinic, non-viscous and adiabatic atmosphere,
from the Euler atmospheric equations, a single, exact but rather, awkward equa-
tion. This equation, coupled with an exact relation for the density, prove to be very
convenient for a RAM Approach of lee-waves starting problem, when we consider
the low Mach number case. Secondly, the low-Kibel/Rossby number asymptotic
Preface vii
model is considered, and a global quasi-geostrophic (GQG) model is derived from

NS-F hydrostatic dissipative atmospheric equations. Namely: the QG single main
equation model, initial condition (at time 0) via an unsteady adjustment (Adj) and
matching, and boundary condition (at the flat ground) via the Ackerbloms problem
in a steady Ekman boundary layer (Ek) problem and matching.
In Epilogue some concluding remarks are sketched briefly.
A postgraduate Course may involve most of the contents of this book, assuming
perhaps a working knowledge of a classical university fluid dynamics Course.
Short Courses for training Applied Mathematicians and Numericiens and young
Scientists in Industry and Research Laboratories can also be based on most of the
contents of this book.
In fact, the material in this book, it seems me, is primarily suitable (maybe
indispensable!) for use by the Scientists and Research Engineers working in the
fields of Fluid Dynamics and having as a main motivation the numerical simulation
of very stiff complex real fluid flows.
Finally, I thank Dr. Christoph Baumann, Engineering Editor, and the members
of the Springer Engineering Editorial Department, where the camera-ready manu-
script was produced in LaTeX and my English type-script was reread by a native
English speaker.
Paris Radyadour Khatchig
Zeytounian
.
Prologue
In the middle of fifties years of the twentieth century, with the works, at Caltech
(California, USA), of Kaplun (1954 and 1957), Lagerstrom and Cole (1955),
Kaplun and Lagerstrom (1957) and also Proudman and Pearson (1957) in England,
asymptotics gave a new tremendous impetus on research in theoretical fluid
dynamics.
Concerning, as to, in France, the asymptotics was introduced by Paul Germain,
who was without (any) doubt the initiator of the application of asymptotics and
modelling in France; Germain write, in Paul Germains Anniversary Volume
(2000):
During the Istanbul 1952 International Congress, Paco Lagerstrom spoke to me about some
questions which he thought to be of the utmost importance for the understanding of fluid
mechanics and which might be ripe for solution at that time. One of them was the
mathematical basis of the boundary layer concept, discovered by Prandtl, nearly 50 years
before. Another one concerned the steady flow of an inviscid fluid as the limit of a class of
corresponding flows of the same fluid, involving a vanishingly small viscosity, so that, in
some sense, solutions of the Euler equations might be related to a class of solutions of
NavierStokes equations, through some limiting process.... I began to foresee a new link
between mathematics and fluid mechanics, provided by asymptotic techniques; as a matter
of fact, no simply a link, but a way of thinking at an enormous variety of problems.
Needless to say that all this was opening avenues without a clear vision of getting the
right way in.
Indeed, Paul Germain has had an extremely fruitful scientific career in which he
contributed, in particular, to modelling and asymptotics in fluid mechanics, by
discovering new problems, new ideas, new methods, new fields of applications.
But, as it is written (page 3, in 2000) by Germain:
My best contributions to fluids were due to discovering, at the end of 1955, Jean-Pierre
Guiraud in fact, Guiraud first applied (in 1958) the method of matched asymptotic
expansions to hypersonic flow past a blunt-nosed plate, with a matching in subtle fashion to
an outer layer that is governed by hypersonic small disturbance theory.
Ten years later, the Van Dykes book (1964), at a time when numerical fluid
dynamics was in its infancy, was at pains to demonstrate that perturbation, or
ix
x Prologue
asymptotic, techniques could be used to advantage to simplify problem to the extent

that they were amenable to analytical treatment; the fluid dynamical literature
abounds with testimony to the success of this approach (see, for instance, the
Annotated Edition of Van Dykes book (1975), by Van Dyke himself).
But, it is necessary to observe that, at the beginning of the sixties, in fact, idea of
a RAM Approach was first envisaged in famous two works.
The first work being related with the problem of the kinetic heating during
re-entry, which gave aeronautical engineers a strong impulse to improve the boundary
layer theory. For this, the extension of higher (at least to second) order boundary
layer theory to compressible and heat conducting flow has been carried out by Van
Dyke (1962). This rational asymptotic (consistent) modelling Van Dyke approach,
of the second order compressible and heat conducting boundary layer model ,
derived from NS-F full problem (formulated below, in Chap. 2, of the present book)
was an important scientific contribution in the NASA program for the re-entry, from
a space station, the space shuttle, when it is necessary to take into account two
complementary effects such as: slipping of the flow and a temperature jump at the
wall.
The second work, is linked with the fact that, if the viscosity and heat conduction
have to be taken into account outside the boundary layer, that means that the
Rankine-Hugoniot equations which rule the jump across the bow-shock wave to
be rewritten and one must take into account the thickness of the shock. Unfortu-
nately, the proposed preceding evaluation was not completely satisfactory!
Germain and Guiraud (1960), are convinced that the only way to get correct results
was to apply matched asymptotic expansion (MAE) and for this the conservative
form of the full NS-F equations has been taken into account. In such a case, viscous
and heat conducting terms appear to have been taken care of but one has to add a
contribution from the inner expansion, the leading term of which gives the internal
shock structure. We observe that to order Re-1, the jump conditions are very easily
written out when one knows the internal shock structure to leading order only.
Germain and Guiraud gives (in a paper published only in 1966) explicit formulas in
gas dynamics for the shock conditions up to order 1.
It is interesting to observe that, in relation to above work related with the shock
waves in gas dynamics, Germain, in (page 11, in 2000), write:
A solution of a mathematical particular schematisation S0 of a physical situation is not
acceptable if it cannot be obtained by the limit of a solution of a more refined
schematisation S when S tends to S0
This requirement being, in fact, a strong condition for a consistent application of

the RAM Approach.
Already, from these above two early examples, it might have been clear that
asymptotics were well suited to derive mathematical models amenable, via simula-
tion, to numerical treatment rather than to obtain closed form of solutions
however, we observe again: at the time, numerical fluid dynamics was almost
nonexistent due to the lack of high speed computers. But, in the course of time,
advances in computer technology have led to the development of increasing
Prologue xi
accurate numerical solutions and have thereby diminished the interest in approxi-
mate analytical results.
It is now evident, and for me, as early as the 1975 year, when I work on my
Survey Lecture for, XIIth Symposium on Advanced Problems and Methods in Fluid
Mechanics (Bialowieza, 813 September 1975, Poland published in 1976), that
asymptotic techniques provide very powerful tools in the process of constructing
mathematical consistent models for problems which are stiff, from the point of view
of numerical analysis and simulation.
Whilst numerical, computational, fluid dynamics is now a mature discipline:
For some time the growth in capabilities of numerical simulation will be depen-
dent on, or related to, the development of RAM Approach. The simple definition of
this RAM Approach being:
The art of modelling assisted, rationally, by the spirit of asymptotics.
Real flows (such as the turbomachinery flows, which are arguably among the
most complicated known to man and are of great technological importance) are
extremely complex and exhibit an enormous range of length and time scales whose
resolution (a well known example is the problem related with the weather
forecast:
of what the weather will be like tomorrow or for the next few days!
will probably remain well beyond the capabilities of any computer foreseeable
future.
Our RAM Approach provide a rational and systematic method for obtaining
the necessary simplified flow models, which, in most cases still have to be solved
numerically. A such approach is an extremely worthwhile objective because most
of the relevant engineering computations are based (up to now!) on relatively ad hoc
models that are rife with internal inconsistencies usually these ad hoc models turn
out to be nonuniform validity- i.e., they break down in certain regions of the flow.
The applications, in Chaps. 79, show that we are able to make significant progress,
via the RAM Approach, toward developing a consistent basis for some of the more
prominent engineering and geophysics/meteo models flows. It is necessary to
observe that:
At the twenty-first century massive computations are capable to bring so much even for
understanding, but there seems to be no indication that they are in competition with
asymptotics, both are useful and complementary.
Let me close this Prologue by a remark:

Quite often the modelling of stiff fluid flow problems may be found by various empirical
procedures or by an ad hoc approach. But it seems me obvious that, the ultimate goal is to
find the mathematical key which explains, not only the success of the modelling, but the
validity and consistency of these procedures in practice during the numerical simulation
The RAM Approach being obviously a well adequate consistent method for a such
realization!
xii Prologue
Cited References
Kaplun, S.: ZAMP. 5, 11136 (1954)

Kaplun, S.: J. Math. Mech. 6, 595603 (1957)
Lagerstrom, P. A., Cole, J. D.: J. Rat. Mech and Anal. 4, 817 (1955)
Kaplun, S., Lagerstrom, P. A.: J. Math. Mech. 6, 585593 (1957)
Proudman, I., Pearson, J.R.A.J.: J. Fluid Mech. 2, 237262 (1957)
Germain, P.: My discovery of mechanics. In: Maugin, G.A., Drouot, R., Sidoroff, F.
(eds.) Continuum Thermomechanics (Paul Germains Anniversary Volume),
Kluwer Academic Publishers, pp. 127 (2000)
Guiraud, J.-P.: C. R. Acad. Sci., Paris, 246, 2842 (1958)
Van Dyke, M.: Second-order compressible boundary-layer theory with application
to blunt bodies in hypersonic flow. In: Riddell, F.R. (ed.) Hypersonic Flow
Theory, vol. 7, pp. 3776. Academic Press, New York (1962)
Germain, P., Guiraud, J.-P.: Conditions de choc dans un fluide doue de coefficients
de viscosite et de conductibilite thermique faibles mais non nuls. C. R. Acad.
Sci., Paris, 250, 19651967 (1960)
Germain, P., Guiraud, J.-P.: Conditions de choc et structures des ondes de choc dans
un ecoulement non stationnaire de fluide dissipatif. Journal de Maths. Pures et
Appliquees. 45, 313358 (1966)
Zeytounian, R. Kh.:La Meteorologie du point de vue du Mecanicien des Fluides.
Fluid Dynamics Transactions. 8, 289352 (1976)
Contents
1 NSF Equations and Modelling: A French Touch . . . . . . . . . . . . . . 1

1.1 My First Contribution to the RAM Approach
in Fluid Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 My Collaboration with Jean-Pierre Guiraud in the Aerodynamics
Department of ONERA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 A Few Remarks Concerning My Preceding Books on Modelling
in Newtonian Fluid Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Part I NavierStokesFourier Equations: General Main Models
2 Newtonian Fluid Dynamics as a Mathematical Physical Science . . . 19

2.1 From Newton to Euler . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.1 Eulerian Elastic Fluid . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.1.2 From Adiabaticity to Isochoricity . . . . . . . . . . . . . . . . . . 23
2.2 Navier Viscous Incompressible, Constant Density Equations . . . . 25
2.3 NavierStokesFourier Equations for Viscous Compressible
and Heat-Conducting Fluid Flow . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.1 The Cauchy Stress Principle . . . . . . . . . . . . . . . . . . . . . . 28
2.3.2 NavierStokes Constitutive Equations:
The CauchyPoisson Law . . . . . . . . . . . . . . . . . . . . . . . 29
2.3.3 Thermodynamics and Energy Equation via Fourier
Constitutive Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.4 NavierStokesFourier (NSF) Equations . . . . . . . . . . . . 34
2.4 Initial and Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . 35
2.4.1 The Problem of Initial Conditions . . . . . . . . . . . . . . . . . . 36
2.4.2 Unsteady Adjustment Problems . . . . . . . . . . . . . . . . . . . 38
2.4.3 Boundary Conditions for the Velocity Vector
u and Temperature T . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.4.4 Other Types of Boundary Conditions . . . . . . . . . . . . . . . 41
xiii
xiv Contents
3 From NSF Equations to General Main Model Equations . . . . . . . 51

3.1 Non-dimensional Form of the NSF Typical Problem . . . . . . . . . 52
3.1.1 Non-dimensional NSF Equations and Reduced
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.1.2 Conditions for the Unsteady NSF Equations . . . . . . . . . 57
3.2 General Main Model Equations . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.2.1 Some General Main Models . . . . . . . . . . . . . . . . . . . . . . 59
3.2.2 A Sketch of the Various General Main Models . . . . . . . . 61
3.3 Non-uniform Validity of Main Model Equations and the Local
Limiting Processes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.1 Large Reynolds Case: A First Naive Approach . . . . . . . . 63
3.3.2 Low Mach Number Case: The Navier System of
Equations and Companion Acoustics Equations . . . . . . . . 65
3.3.3 OberbeckBoussinesq Model Equations for the
RayleighBenard Shallow Thermal Convection . . . . . . . . 69
3.3.4 StokesOseen Model Equations in the Case of a Low
Reynolds Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.3.5 The Case of Non-linear Acoustics . . . . . . . . . . . . . . . . . . 76
3.3.6 A Sketch of Consistent Models Derived from NSF
Equations Relative to the Considered Reference
Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4 A Typical RAM Approach: Boussinesq Model Equations . . . . . . . . 81

4.1 Introductory Remarks Concerning Atmospheric Flow . . . . . . . . . 81
4.2 Asymptotics of the Boussinesq Case . . . . . . . . . . . . . . . . . . . . . 84
4.2.1 Euler Dimensionless Equations for the Thermodynamic
Perturbations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.2.2 Rational Derivation of Boussinesq Equations . . . . . . . . . . 86
4.2.3 Two Particular Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.3 The Steady 2D Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
4.4 The Problem of Initial Conditions . . . . . . . . . . . . . . . . . . . . . . . 92
4.5 A Sketch of a RAM Approach for a Boussinesq Model . . . . . . . . 94
Part II A Sketch of a Mathematical Theory of the RAM Approach
5 The Structure of Unsteady NSF Equations at Large

Reynolds Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
5.2 The Emergence of the Four Regions as a Consequence
of the Singular Nature of BL Equations Near t 0 . . . . . . . . . . 98
5.3 Formulation of the Unsteady NSF Problem . . . . . . . . . . . . . . . . 100
5.4 Derivation of the Corresponding Four Model Problems . . . . . . . . 103
5.4.1 EulerPrandtl Regular Coupling . . . . . . . . . . . . . . . . . . . 103
Contents xv
5.4.2 Acoustic and Rayleigh Problems Near

the Initial Time t 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
5.5 Adjustment Processes Towards the Prandtl BL Evolution
Problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
5.5.1 Adjustment Process Via the Acoustics/Gas Dynamics
Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
5.5.2 Adjustment Process Via the Rayleigh Equations . . . . . . . 112
5.6 Some Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6 The Mathematics of the RAM Approach . . . . . . . . . . . . . . . . . . . . 117

6.1 Our Basic Postulate for the Realization of the RAM Approach . . 117
6.2 The Mathematical Nature of NSF Equations:
A Fluid-Dynamical Point of View . . . . . . . . . . . . . . . . . . . . . . . 120
6.3 Formulation of Dimensionless Equations for Applications
of the RAM Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
6.3.1 Turbomachinery Fluid Flows . . . . . . . . . . . . . . . . . . . . . 124
6.3.2 The GZ Rolled-up Vortex Sheet Theory: Vortex
Sheets and Concentrated Vorticity . . . . . . . . . . . . . . . . . 127
6.3.3 The GZ Asymptotic Approach to Non-linear
Hydrodynamic Stability . . . . . . . . . . . . . . . . . . . . . . . . . 130
6.3.4 A Local Atmospheric Thermal Problem: A Triple-Deck
Viewpoint . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.3.5 Miscellanea . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138
6.4 Some Key Steps for the Application of the Basic Postulate . . . . . 138
6.4.1 Similarity Rules: Small Mach and Large Reynolds
Numbers Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
6.4.2 The Matching Principle . . . . . . . . . . . . . . . . . . . . . . . . . 141
6.4.3 The Least-Degeneracy Principle and the Significant
Degeneracies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.4.4 The Method of Multiple Scales (MMS) . . . . . . . . . . . . . . 145
6.4.5 The Homogenization Analysis . . . . . . . . . . . . . . . . . . . . 149
6.4.6 Asymptotics of the Triple-Deck Theory . . . . . . . . . . . . . 154
Part III Applications of the RAM Approach to Aerodynamics, Thermal

Convection, and Atmospheric Motions
7 The RAM Approach in Aerodynamics . . . . . . . . . . . . . . . . . . . . . . 161

7.1 Derivation of a Through-Flow Model Problem for Fluid Flow
in an Axial Compressor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.1.1 The Veuillot Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.1.2 The GZ Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
7.1.3 Transmission Conditions, Local Solution at the
Leading/Trailing Edges, and Matching . . . . . . . . . . . . . . 168
7.1.4 Some Complements . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.2 The Flow Within a Cavity Which is Changing Its Shape and
Volume with Time: Low Mach Number Limiting Case . . . . . . . . 171
xvi Contents
7.2.1 Formulation of the Inviscid Problem . . . . . . . . . . . . . . . . 172

7.2.2 The Persistence of Acoustic Oscillations . . . . . . . . . . . . . 173
7.2.3 Derivation of an Average Continuity Equation . . . . . . . . . 174
7.2.4 Solution for the Fluctuations u0 and r1 . . . . . . . . . . . . . . 178
7.2.5 The Second-Order Approximation . . . . . . . . . . . . . . . . . 179
7.2.6 The Average System of Equations for the Slow
Variation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
7.2.7 The Long Time Evolution of the Fast Oscillations . . . . . . 184
7.2.8 Some Concluding Comments . . . . . . . . . . . . . . . . . . . . . 187
8 The RAM Approach in the Benard Convection Problem . . . . . . . . 193

8.1 An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
8.2 Some Unexpected Results for the Benard Problem of an
Expansible Liquid Layer Heated from Below . . . . . . . . . . . . . . . 194
8.2.1 The Questionable Davis (1987) Upper, Free Surface,
Temperature Condition, and the Problem
of Two Biot Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . 196
8.2.2 The Mystery of the Disappearance of Influence
of the Free Surface in the RB Leading-Order Shallow
Thermal Convection Model . . . . . . . . . . . . . . . . . . . . . . 199
8.2.3 Influence of the Viscous Dissipation . . . . . . . . . . . . . . . . 201
8.3 The Marangoni Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
8.3.1 The Long-Wave Approach . . . . . . . . . . . . . . . . . . . . . . . 205
8.3.2 Towards a Lubrication Equation . . . . . . . . . . . . . . . . . . . 208
8.4 From Deep to Shallow Thermal Convection Model Problems . . . 210
9 The RAM Approach in Atmospheric Motions . . . . . . . . . . . . . . . . . 213

9.1 Some Models for the Lee-Waves Problem . . . . . . . . . . . . . . . . . 213
9.1.1 First Integrals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
9.1.2 An Equation for the Vertical Deviation dx; z . . . . . . . . 215
9.1.3 Dimensionless Formulation . . . . . . . . . . . . . . . . . . . . . . 218
9.1.4 Four Distinguished Limiting Cases . . . . . . . . . . . . . . . . . 220
9.2 The Low Kibel Number Asymptotic Model . . . . . . . . . . . . . . . . 223
9.2.1 The Dissipative (Viscous and Non-adiabatic) NSF
atmospheric Equations . . . . . . . . . . . . . . . . . . . . . . . . . . 226
9.2.2 Hydrostatic Limiting Processes . . . . . . . . . . . . . . . . . . . . 229
9.2.3 The Dissipative Hydrostatic Equations in p-System . . . . . 233
9.2.4 The Leading-Order QG Model Problem . . . . . . . . . . . . . 237
9.2.5 The Second-Order Ageostrophic GZ Model . . . . . . . . . . 247
9.2.6 Kibel Primitive Equations and Lee-Waves Problems
as Inner and Outer Asymptotic Models . . . . . . . . . . . . . . 248
Epilogue . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
Part I
NavierStokesFourier Equations:
General Main Models
Chapter 2
Newtonian Fluid Dynamics as a Mathematical
Physical Science
There are two mathematicalphysical descriptions of fluid dynamics.1 The first of

them is a microscopic description, from the Boltzmann equation for the (one-
particle) distribution function f t; x; x:
@f =@t x:rf 1=KnQf ; f ;
where f t; x; x is precisely the density probability of finding a molecule at the

space-position x (at the time t), with the velocity x. The parameter Kn in the
Boltzmann equation is the Knudsen number, which is the ratio of the mean free
path (between collisions of molecules a microscopic reference length, l ) and a
typical (macroscopic) reference length, L , of the classical continuum theory,
which is the ratio of Mach (M) and Reynolds (Re) numbers:
Kn M=Re;
where M U =c is the constant Mach (dimensionless) number, based on the

reference macroscopic velocity U and the speed of sound c , that characterizes the
compressibility effect. The parameter Re U L =n is the constant Reynolds
(dimensionless) number that characterizes the viscosity (via the kinematic viscosity
coefficient n ) effect.
1
In my 2001 review paper [29] written by a fluid dynamicist for fluid dynamicists the curious
reader can find a contribution concerning the many theoretical mathematical investigations of
NavierStokesFourier problems. My intent was to extract from the huge literature the basic results,
ideas, and goals of this currently wide activity and to present the results to the readers of Applied
Mechanical Review. I am sure that rigorous mathematicians will find in this paper many
shortcomings, non-rigorous formulations, and so on. I think, however, that such a paper will
stimulate further thinking by engineers and applied scientists, including some exchange of opinions,
and so on, and that it is therefore needed. The distance between theoretical mathematicians and
applied mathematicians and engineers has become too large! I hope that both old and new
investigators interested in Newtonian fluid flow problems might learn much from it.
R. Kh. Zeytounian, NavierStokesFourier Equations, 19

DOI 10.1007/978-3-642-20746-4_2, # Springer-Verlag Berlin Heidelberg 2012
20 2 Newtonian Fluid Dynamics as a Mathematical Physical Science
The right-hand side of the above Boltzmann equation, the collision operator
Q, is typical of kinetic theory of gases in that it preserves mass, momentum, and,
namely:

cQdx; 0;
where c 1; x1 ; x2 ; x3 ; jxj2 is a five-component (the so-called collisional

invariants) vector. For a modern exposition of the kinetic theory of gases (dilute
gas) see Cercignani, Illner, and Pulvirenty, 1994 [30].
The second continuum description is linked with the macroscopic length scale
(which is the real scale for applications in fluid flows), and is governed by the three
conservation equations of classical continuum mechanics: principle of conservation
of mass (assuming that the fluid possesses a density function rt; x), principle of
conservation of linear momentum (adopting the stress principle of Cauchy; see
Sect. 2.3.1), and the conservation of energy (we postulate that the total energy of a
volume the sum of its kinetic energy and its internal energy is conserved; see
Sect. 2.3.3).
The resulting three equations of continuum mechanics, which proceeds on the
assumption that a fluid is practically continuous and homogeneous in structure (see,
Serrin, 1959 [31]) are:
Dr=Dt rr: u 0;
rDu=Dt rf r:T;
rDE=Dt T : D divq;
where u is the velocity vector, f is the extraneous force per unit mass (a known
function of position x and time t), T is the Cauchy stress tensor, E is the specific
internal energy, q is the heat flux vector, and the term T :D, in the energy equation
for E, is a dissipation term involving the interaction of stress and deformation
(second-order tensor D). We observe that T :D stands for the scalar product TijDij
of two second-order tensors (dyadics), and Tij and Dij are, respectively, the com-
ponents of T and D.
The problem of the derivation of the fluid dynamic equations (derived from
above three continuum mechanics equations, with Cauchy stress tensor T and the
heat flux vector q, due to NavierStokes and Fourier constitutive equations see
Sects. 2.3.2 and 2.3.3) from the Boltzman equation for small Knudsen numbers
(Kn # 0) is shortly expounded in Sect. 5 of our review paper [29], where the reader
can find various pertinent references concerning this fluid dynamics limit of kinetic
equations, initiated by Hilbert in 1912.
2.1 From Newton to Euler 21
2.1 From Newton to Euler
Sir Issac Newton, English mathematician and physicist, was the greatest single
influence on theoretical physics until Einstein. In his major treatise, Philosophia
Naturalis Principia Mathematica (1687) [32] he presented a mathematical descrip-
tion of the laws of mechanics and gravitation, and applied this theory to explain
planetary and lunar motions. In the Second Law we read: The body moves in such
a way that at each moment the product of its acceleration vector by the density is
equal to the sum of certain other vectors, called forces, which are determined by the
motion taking place. That is:
r Du=Dt rg internal force per unit volume: (2.1)
A second part of Newtons Principia is related to the conservation of mass: To

each small solid body can be assigned a positive number m, invariant in time, called
its mass. That is:

D=Dt rdV 0: (2.2)
V
In (2.1), u is the velocity vector, g is the gravitational force per unit mass, and r
is the density. The (Cartesian) components of the nabla, , operator, in material (or
substantial) derivative
D=Dt @=@t u:r;
are @=@xi ; i 1; 2; 3, where the time is denoted by t, and x x1 ; x2 ; x3 is the

position vector.
In (2.2), dV is a volume element in the neighbourhood of the point P, and to this
volume element will be assigned a mass rdV.
We observe that D/Dt is related to the Euler rule of differentiation, and t, x is the
Euler timespace variable. To express (2.2) in the form of a differential equation,
the differentiation indicated in this equation is carried out by transforming the
integral suitably. In this case we derive the so-called equation of continuity (this
derivation is, in fact, due to Euler in 1755 [33]):
Dr=Dt rr:u 0: (2.3)
This (compressible) equation of continuity (2.3) remains unaltered when viscos-

ity is admitted.
2.1.1 Eulerian Elastic Fluid
In reality, fluid dynamics was first envisaged as a systematic mathematical

physical science in Johann Bernoullis Hydrodraulics (1737) [34], in Daniel
Bernoullis Hydrodynamica (1738) [35], and also in DAlembert Traite de

lEquilibre et du mouvement des fluides (1744) [36]. However, the fundamental
ideas expounded in these books were formulated mathematically as partial differ-
ential equations in an epochal paper by Euler (1755) [33] which firmly established
him as the founder of rational fluid dynamics. Nevertheless, he considered only
non-viscous (inviscid) fluid flows with the pressure a function only of the density (a
so-called isentropic/barotropic fluid flow the fluid being an elastic fluid). We
observe that an inviscid fluid is one in which it is assumed that the internal force
acting on any surface element dS, at which two elements of the fluid are in contact,
acts in a direction normal to the surface element. At each point P (with coordinates
xi, i 1, 2, 3) the stress, or internal force per unit area, is independent of the
orientation (direction of the normal) of dS, and the value of this stress is called the
pressure, p, at the point P. Therefore, the internal force per unit volume, appearing
in Newtons equation (2.1), has xi component @p=@xi , i 1, 2, 3. As a con-
sequence, for an inviscid (non-viscousEulerian) fluid we determine, from (2.1),
the classical Euler equation of motion (momentum equation):
rDu=Dt rg rp: (2.4)
Equations 2.3 and 2.4, which express Newtons principles for the motion of
an inviscid fluid, are usually referred to as the Eulerian fluid flow (compressible)
equations, and include one vector equation (2.4) and one scalar equation (2.3) for u,
r and p (five unknowns).
It follows that one more equation is needed in order that a solution of the system
of Euler equations be uniquely determined for given initial and boundary condi-
tions. According to Euler, if we add to Eq. 2.3 and 2.4 the following specifying
equation:
p pr (2.5)
which gives the relation between the pressure and the density, we shall have five
equations (a closed system) which include all the theory of the motion of fluids.
By this formulation, Euler believed (255 years ago!) that he had reduced fluid
dynamics, in principle, to a mathematicalphysical science; but it is crucial to note
that, in fact, Eq. 2.5 is not an equation of state, but specifies only the particular type
of motion (so-called barotropic) under consideration, and in this case the fluid is just
called an elastic fluid.
In my book (Zeytounian, 2002) [37], the reader can find a theory and
applications of non-viscous fluid flows, and in the next chapter, devoted to a
discussion of various general models derived from NavierStokesFourier
equations, we obtain, for large Reynolds number Re 1 as a vanishing viscosity
limit the full unsteady Euler compressible non-viscous adiabatic and baroclinic
equations for a thermally perfect gas (a trivariate fluid). (Concerning the NSF
equations see Sect. 2.3.)
2.1 From Newton to Euler 23
These general models very often form the basis of various chapters in fluid
dynamics treatises. It is obvious, therefore, that these treatises may be organized
through some models which are best obtained by asymptotic modelling. As an
example we mention the case of inviscid flows which are often considered as a
model, used from the outset, and need to be embedded in the more general model of
slightly (vanishing) viscous (laminar) or with slight friction (turbulent) flow,
to which asymptotic modelling is applied. Incompressible flows are seldom con-
sidered as flow at small Mach numbers which may lead to almost nonsensical
conclusions, as when one deals with incompressible aerodynamics, because phe-
nomena such as sound produced by quite low-speed flow cannot be understood
other than by low-Mach-number (hyposonic) aerodynamics.
2.1.2 From Adiabaticity to Isochoricity
In many cases the specification of the type of flow is given in thermodynamic terms.
The most common (and rather naive) assumption in the study of compressible fluids
is that no heat output or input occurs for any particle. In this case, heat transfer by
radiation, chemical processes, and heat conduction between neighbouring particles
are excluded, and the fluid flow is called adiabatic.
In order to translate either assumption into a specifying equation, the First Law
of Thermodynamics must be used, which gives the relation between heat input and
the mechanical variables [J. R. von Mayer (1842)]. If the total heat input from all
sources, per unit of time and mass, is zero, the First Law for an inviscid fluid can be
written in the following form:
Cv DT=Dt pD=Dt1=r 0; (2.6)
where Cv is the specific heat of the fluid at constant volume. The first term in (2.6)
represents the part of the heat input expended for the increase in temperature T, and
the second term corresponds to the work done by expansion. It is well known, also
from thermodynamics, that for each type of matter a certain relation exists among
the three (thermodynamic) variables, pressure p, density r, and temperature T:
fp; r; T 0; (2.7)
Thus the temperature can be computed when p and r are known. Naturally, the
equation of state (2.7) is not a specifying equation, since it implies temperature as a
new variable. Finally, Eqs. 2.3, 2.4, and 2.6, together with (2.7), form a closed
system of six equations for the six unknowns: u, p, r, and T.
For a thermally perfect gas (naturally, a perfect gas is not necessarily inviscid),
the equation of state (2.7) is explicit:
p RrT; (2.8)
where R is a constant depending upon the particular perfect gas. From (2.8) it
follows that for a perfect gas the condition p/r const implies a fluid flow at con-
stant temperature, or isothermal flow.
The specific entropy S of a perfect gas is then given by:
S R=g 1logp=rg const; (2.9)
where g is a constant, having the value 1.40 for dry air. Thus the motion of a perfect
gas with the condition p/rg const, as a specifying equation, is isentropic (con-
stant entropy motion or, since g > 1, polytropic). The equation of state for a perfect
gas in equilibrium, connected with the names of Boyle (see, Birch (1744) [38]),
Mariotte, Gay Lussac, and Charles, has been widely known since 1800.
In precisely the modern form, it was used freely by Euler, but did not appear again
in the hydrodynamical literature until used by Kirchhoff in his paper of 1868 [39].
In some presentations, no distinction is made between the term perfect gas and
ideal gas. Here the term perfect gas is defined precisely by the equation of state
(2.8). The term Eulerian fluid flow is used for an inviscid (non-viscous) and non-
heat-conducting flow, governed by the system of Eqs. 2.3, 2.4, and 2.6, with (2.7).
According to (2.6) and (2.7) this Eulerian fluid flow is a baroclinic and adiabatic
fluid flow. In Eq. 2.6 an expression for Cv in term of the variables T, p, and r is
needed, but for a perfect gas, where the equation of state is (2.8), it is generally
assumed that Cv R/(g 1) is a constant R being the usual gas constant.
As a consequence, we derive, for such a perfect gas with constants Cv and Cp (
g Cv), specific heats, the following conservation equation for specific entropy in the
case of a thermally perfect gas:
D=Dt logp=rg 0 ) DS=Dt 0: (2.10)
Equation 2.10, however, holds only for an adiabatic flow of a perfect inviscid
gas, when the entropy is constant for each particle but varies from particle to
particle. Generally, a thermally perfect inviscid gas in adiabatic flow does not
necessarily behave like an elastic fluid.
If we assume that in (2.10),
g tends to infinity incompressible limit case; (2.11a)
such that R O(1), and that in such a case,
Cv tends to zero but Cp R O1; (2.11b)
then we derive, again, from (in place of) (2.10) the following evolution equation
(conservation law) for density (isochoricity):
Dr=Dt 0 ) r u 0; (2.11c)
2.2 Navier Viscous Incompressible, Constant Density Equations 25
As a consequence, for a Eulerian incompressible but non-homogeneous

(isochoric) fluid flow we obtain the following system of three equations for the
velocity u, pressure p, and density r:
r Du=Dt rg rp; (2.12a)
Dr=Dt 0; (2.12b)
r u 0: (2.12c)
This isochoric system of three Eqs. 2.12a2.12c is very well investigated in

Yihs 1980 book [4]. The above incompressible limiting process (2.11a,2.11b)
presents the possibility of taking into account some compressible (second-order)
effects of the order O(1/g) which is the case, for instance, in the theory of lee
waves downstream of a mountain!
2.2 Navier Viscous Incompressible, Constant Density Equations
The equation of motion of a viscous and incompressible homogeneous (with a

constant density) fluid flow was first obtained by Navier in 1821 [40] and later
by Poisson in 1831 [41]. The necessity of such a viscous equation (in place of
the above Euler equation (2.4) was strongly linked with the dAlembert theorem
(paradox?): An object moving with constant velocity U1 in a potential field (from
Bernoulli equation and Lagrange theorem in an incompressible fluid the velocity-
potential F must satisfy Laplaces equation) does not feel any force neither drag
nor lift.
Obviously, this result is in sharp contrast with experience! For instance, an
aircraft could not fly. Suppose that, initially, the aircraft and the fluid (air) are
both at rest, then the aircraft begins to move. Since vorticity cannot be produced
(Lagrange permanence of irrotational flow), the potential flow around the aircraft
cannot produce any lift, so that flight is impossible. Such a paradox can be avoided
if vorticity is present.
However, the problem remains of understanding how vorticity can be created in
the system. The conservation of vorticity in an inviscid (incompressible) fluid, while
reasonably far from the obstacle, is too drastic near the boundary of this obstacle!
A more accurate description of the interaction among the particles of the fluid
and the obstacle leads us to introduce the Navier (viscous and incompressible)
equation, which is a correction to the Euler (incompressible and non-viscous)
equation of motion.
Such a new (Navier) equation can explain the effects, such as vorticity produc-
tion, which are relevant near the boundary. This Navier equation has the following
rather simple form:
DuN =Dt 1=r rp g n DuN ; (2.13a)
where n is the constant kinematic viscosity, and D r2 is the Laplace operator

for the Navier velocity vector uN . The companion to (2.13a), equation of continuity
is simply:
r:uN O: (2.13b)
Because n multiplies the derivative of highest order in Navier equation (2.13a),

it cannot be inferred that the solutions of (2.13a), for very small values of n ,
reduces to Euler, below (2.14a), equation for an incompressible homogeneous and
non-viscous fluid flow, with uE as velocity vector:
DuE =Dt 1=r rp g O: (2.14a)
with
r:uE O: (2.14b)
It is important (in particular, in the framework of our RAMA) to observe that the
passage from compressible flow to incompressible flow, which filters the acoustic
fast waves, is a strongly singular limit.
The reader can find in our 2006 monograph [13], devoted to hyposonic flow
theory, various facets of the unsteady very slow flows at low Mach number, which
are strongly related to a category of fluid flow problems, called hyposonic, when
M << 1.
In the above Navier incompressible viscous equation (2.13a) and also in the
Euler incompressible non-viscous equation (2.14a), the term 1=r rp is not an
unknown quantity of the initial value problem. In fact, rp=r is the force term
acting on the particles of fluid allowing them to move as freely as possible, but in a
way compatible with the incompressibility constraint (2.13b) or (2.14b): r:uN 0.
Note particularly that for a Eulerian incompressible flow, DuE/Dt 0 admits
solutions violating the condition: . uE 0 at t > 0, even if the velocity diver-
gence vanishes at t 0. The pressure term in the above incompressible equations
(2.13a) and (2.14a) is not an unknown quantity, because it can be determined when
we have found the velocity field uN or uE for instance, taking the divergence of the
Euler equation (2.14a), we obtain a Poisson (elliptic) equation:
Dp r fr : uE : ruE r:gg;
and, knowing uE and external force g, we can find p by solving a Poisson equation
with a Neumann boundary condition,
@p=@n r fr : uE : ruE g : n;

2.2 Navier Viscous Incompressible, Constant Density Equations 27
in a domain with a boundary (after that the Euler equation (2.14a) is projected on
the outward unit normal n).
As a consequence of the above, it is sufficient to consider the Navier incom-
pressible equation in terms of vorticity oN r ^ uN assuming that g is
conservative:
DoN =Dt oN : ruN n DoN ; (2.15a)
with
r: uN 0: (2.15b)
Obviously, when we assume that g is conservative, any potential flow,
u rF;
trivially satisfies the Navier equation (2.13a) in term of the vorticity oN !

However, to obtain a well-set boundary value problem, for a fixed n > 0,
one must also (according to Stokes) replace the slip boundary condition on a
(stationary) boundary of a fluid flow domain, for uE in Euler non-viscous equation
(2.14a):
uE :n 0; (2.16a)
by the more stringent condition of no-slip boundary condition on a stationary

boundary:
uN 0; (2.16b)
for uN in Navier (2.13a).

Concerning this no-slip boundary condition (2.16b), it is interesting to note that
in his 1904 lecture to the ICM, Prandtl stated:
The physical processes in the boundary-layer (BL Grenzschicht) between
fluid and solid body can be calculated in a sufficiently satisfactory way if it is
assumed that the fluid adheres to the walls, so that the total velocity there is zero
or equal to the velocity of the body. If the viscosity is very small and the path of
the fluid along the wall not too long, the velocity will have again its usual value
very near to the wall (outside the thin transition layer). In the transition layer

(Ubergangsschicht) the sharp changes of velocity, in spite of the small viscosity
coefficient, produce noticeable effects.
Prandtl not only mentions the existence and nature of the thin boundary-layer
and its connection with frictional drag, but derives heuristically the boundary-layer
(so-called Prandtl) equations valid in a thin viscous layer close to the wall of the
solid body. These BL Prandtl equations, however, are not valid near the time t 0,
where the initial data are given in the case of an initial-boundary value problem.
Prandtl curiously did not have any idea concerning this singular nature of his
discovered BL equations in unsteady compressible case!
But it is also necessary to not overlook the important investigations of
Lanchester (1907) in England, concerning the nature of the boundary-layer and
explanation of separation (independently of Prandtl). (For a detailed discussion
concerning the initial and boundary conditions, see Sect. 2.4.)
In Chap. 5, as a consequence of the singular nature of BL compressible
equations, the unsteady full NSF equations for large Reynolds number are
analyzed in detail. For this, it is necessary to consider five regions and the related
matching conditions.
2.3 NavierStokesFourier Equations for Viscous

Compressible and Heat-Conducting Fluid Flow
According to Truesdell, in The Mechanical Foundations of Elasticity and Fluid

Dynamics (1966) [42, p. 2]:
Classical fluid dynamics describes the flow of media altogether without spring-
iness of form, so that when released from all deforming forces except a hydrostatic
pressure, they retain their present shapes; it is a partially linear theory, in which a
uniformly doubled rate of deformation if dynamically possible would lead to
doubled viscous forces.
2.3.1 The Cauchy Stress Principle
The derivation of the equation of motion for the velocity vector u, for real (viscous
compressible and heat conducting) fluids is based on the following stress principle
of Cauchy, 1828 [43]:
Upon any imagined closed surface S (with outward normal n to S) there exists a
distribution of stress vector Sx; t; n n:T, where T is the stress tensor, whose
resultant and moment are equivalent to those of the actual forces of material
continuity exerted by the material outside S upon that inside.
This statement of Cauchys principle is due to Truesdells paper of 1952 [44];
and as Truesdell remarks (1953) [45], the above well-known Cauchy principle
. . . has the simplicity of genius. Its profound originality can be grasped only when ones
realizes that a whole century of brilliant geometers had created very special elastic
problems in very complicated and sometimes incorrect ways without ever hitting upon
this basic idea, which immediately became the foundation of the mechanics of distributed
matter.
As a consequence of his stress principle, Cauchy obtained a general equation of

motion the simple and elegant Eq. 2.17 below which is valid for any fluid, and
2.3 NavierStokesFourier Equations for Viscous Compressible 29
indeed for any continuous medium, regardless of the form which the stress tensor
T may take.
Du=Dt rf r:T: (2.17)
We observe that the above equation of motion, discovered by Cauchy in 1828,

can be derived easily according to the principle of conservation of linear momen-
tum: The rate of change of linear momentum of a material volume V equals the
resultant force on the volume.
The necessity for a clear-cut statement of the postulates on which continuum
mechanics rests was pointed out by Felix Klein and David Hilbert, but the first
axiomatic presentation is due to G. Hamel (1908) [46]. With S(x, t; n) n .T, for
the stress vector S, the above principle is expressed by the statement:

D=Dt rDu=Dt dv rf dv divTdv; (2.18a)
v v v
applying the divergence theorem. Since v (a fixed volume) is arbitrary, we obtain

the Eq. 2.17.
We observe also that the stress forces are in local equilibrium, and it is postu-
lated that the stress tensor is symmetric:
T ij T ji : (2.18b)
2.3.2 NavierStokes Constitutive Equations:

The CauchyPoisson Law
In our 2004 book Theory and Applications of Viscous Fluid Flows [47], in Sect. 1.4
of Chap. 2, the reader can find a detailed account of the constitutive equation of a
viscous (a la NavierStokes) classical fluid, mainly inspired by Serrin (1959) [31].
Here we present only a short comment. A first important moment in the history
of NS constitutive equations is Stokes idea (1845) [48] of fluidity which can be
stated as four postulates:
1. T F(D) and D D(u).
2. T does not depend explicitly on the position vector x (spatial homogeneity).
3. There is no preferred direction in space (isotropy).
4. When D(u) 0, then T pI (Eulerian non-viscous fluid flow).
A medium whose constitutive equation (via stress tensor T, which define or
delimit the type of medium subject to study) satisfies these above four postulates is
called a Stokesian fluid.
With the above four postulates, according to matrix algebra, and if we add the
condition that the components of T be linear in the components of D(u), we deduce
(CauchyPoisson law):2
T p l divuI 2mDu: (2.19)
The coefficients l and m (of viscosity) being of scalar functions of the thermo-
dynamic state (considered in Sect. 2.3.3) and I is the unit tensor, with dij (the
so-called Kronecker symbol with dkk 1 and dij 0 if i j) as components.
Indeed, the fully general expression is Poissons (1831) relation [41] but the
name of Poisson is rarely quoted today. In Cauchy, (1828), [43], the term pI, in
(2.19), is absent.
Dynamical (Navier) equation (2.13a) equivalent to those resulting from (2.19) when
divu 0 and, m const, is due to Navier (1821) [40], and SaintVenant (1843) [49],
proposed (2.19) in the special case when m m0 const and mv l + (2/3)m 0,
which is the so-called (1845) Stokes relation [48].
This is the simple and elegant constitutive equation (2.19) for the viscous NS
motion, discovered by Cauchy in 1828. It is valid for any fluid, and indeed for any
continuous medium, regardless of the form which the stress tensor T may take.
The coefficients l and m (of viscosity) are scalar functions of the thermodynamic
state (considered in the next Section).
Concerning the long controversy regarding the Stokes relation:
3mv 0; (2.20a)
3mv being the bulk viscosity, in the classical theory of viscous fluids; see
Truesdell (1966) [42].
The viscosities coefficients (shear/dynamic and bulk) and the thermal conduc-
tivity k (see Sect. 2.3.3, Fouriers law (2.30)) are known functions, subject to the
thermodynamic restriction (ClausiusDuhem inequalities):
m 0; k 0 and mv 0: (2.20b)
2
For a perfect (absence of viscosity) fluid, the pressure has already appeared as a dynamical
variable in Euler Eq. 2.4. Characteristic of the discipline of gas dynamics is the postulate that the
thermodynamic pressure, introduced via functional relations among the state variables (see
Sect. 2.3.3), is equal to this dynamical pressure. When the deformation D(u) 0, for a perfect
fluid, p is the thermodynamic pressure when the fluid is compressible, while p is simply an
independent dynamical variable otherwise. For an incompressible perfect or viscous fluid (Navier,
see Sect. 2.2) p is not an unknown quantity, because it can be determined when we have found the
velocity field u. In some works (see [31]) a mean pressure, p* (1/3)Trace T is defined, and we
have the following relation: p p l 2=3mdivu.
The dynamical equation (in components form, with indices, i 1, 2 and 3)

resulting from (2.17) with (2.19) is the NavierStokes (compressible) equation for
the component ui of the velocity u:
r Dui =Dt @p=@xi @=@xj m@ui =@xj @uj =@xi @=@xi l@uk =@xk ;
(2.21)
since for the components of deformation tensor D(u) we have:
2Dij @ui =@xj @uj =@xi : (2.22)
A Stokesian fluid whose constitutive, NS compressible, equation is given by

(2.19) is called a Newtonian fluid.
In SaintVenant (1843) (and Stokes (1845)), the resulting dynamical equation,
for ui, in place of (2.21), is:
rDui =Dt @p=@xi m0 fDui 1=3@=@xi fDlogr=Dtg; (2.23a)
if we take into account the equation of continuity:
@uk =@xk Dlogr=Dt; (2.23b)
and the Stokes relation (2.20a).

For an incompressible homogeneous fluid, again we derive the Navier
dynamical equation (2.13a), with n0 m0 =r0 const.
Equation 2.23a (where m0 is a constant), with the specifying equation p p(r)
and continuity equation (2.23b), forms a closed system of equations the so-called
NavierStokes (compressible NS equations) for ui, p, and r, and governs a
barotropic(?) viscous and compressible fluid flow, without an energy equation for
the temperature T which unfortunately do not have any physical (fluid dynamics)
signification.
Slightly more complete general NavierStokes compressible NS equations, for
the unknowns ui, p, and r, are obtained from (2.21), with again the specifying
equation p p(r) and continuity equation (2.23b), if we assume that viscosities m
and l do not depend on temperature T and are known functions of the density
r only.
These above NS equations do not emerge via a RAM Approach from the full
unsteady NSF equations and, in fact, do not have in reality any interest for fluid
dynamicians!
We also obtain a simplified model of compressible viscous fluid flow if we
assume in addition, instead of p p(r), that the pressure is identically constant in
the fluid flow (isobaric fluid flow).
In this case, then, we arrive at the so-called Burgers model equations (as in
Kazhikov (1994) [50]).
Dlogr=Dt @uk =@xk 0; (2.24a)
r Dui =Dt @=@xj mr@ui =@xj @uj =@xi

(2.24b)
@= @xi lr@uk =@xk ;
which is a closed system of two equations for the velocity components ui and the
density r, when we assume that the viscosity coefficients, l and m, are a function
only of r.
A final important remark concerning the above NS compressible viscous
barotropic system equations, (2.23a) and (2.23b) with p p(r), or Eqs. 2.21 and
2.23b with p p(r), for the unknowns ui, p and r, which are mainly used by
applied mathematicians in their rigorous mathematical analyses (see, for instance,
P. L. Lions (1998) [51]) does not have any physical reality, mainly because just
viscosity always generates entropy (baroclinity).
For this, in particular, the various rigorous mathematical results concerning the
so called incompressible limit, related to the limiting process M # 0, in the
framework of above, compressible barotropic (p p(r)) and viscous, two systems,
seems (to me) very questionable.
In the short paper by Leray (1994) [52], this question is pertinently discussed.
2.3.3 Thermodynamics and Energy Equation via Fourier

Constitutive Equation
But in general (in reality) the coefficients of viscosity are assigned or empirical
functions of the positive variables r (density) and especially T (temperature), which
both are present also in equation of state (2.7) or (2.8) for a trivariate realistic fluid.
Indeed, Euler and Lagrange not only failed to include viscosity effects in their
equations of motion, forcing them adopt corresponding simplified (slip) boundary
condition (2.16a), but also oversimplified their equation of state.
In real fluids, the pressure, p, is a function of two variables, r and T (for a trivariate
fluid in a baroclinic motion, see the equation of state (2.7)). Obviously, again, it is
necessary to associate with NS compressible equations (2.21), (2.23b), and (2.7), for
ui, r, p and T, an energy equation, if we want to obtain a closed system of equations for
our six unknown functions. For this, some thermodynamical assumptions are required.
For the real fluid flows compressible, viscous, and heat-conducting the
mechanical energy is converted into heat by viscosity, and the heat of compression
is diffused by heat conduction.
Here we consider only an homogeneous fluid when the local equation of state is
according to the basic postulate of Gibbs (1875) [53]; see Truesdell (1952) [44], and
Serrin (1959) [31]:
E Er; S; (2.25)
where E is the specific internal energy and S is the specific entropy.

In this case the temperature T and the thermodynamic pressure p are defined by
the two relations:
T @E=@S and p @E=@1=r: (2.26)
Now, for any compressible fluid, by differentiating (2.25) along any curve on the
energy surface (characterizing the fluid) we obtain:
DE=Dt T DS=dt p D1=rDt: (2.27)
But, for any homogeneous medium in motion, the conservation of energy is

expressed by the equation of C. Neumann (1894):
r DE=Dt + p (@uk =@xk @qi =@xi 2m Dij l Dkk dij Dij ; (2.28)
where the qi are the Cartesian components of the heat flux vector q.
For the special case of a non-viscous incompressible fluid, the energy equation
was given by Fourier (1833) [54], and for small motions of a viscous perfect gas by
Kirchhoff (1868) [39], and in this case we also have p RrT, where R is the
constitutive constant of the viscous thermally perfect gas.
We observe also, that for a medium suffering deformation the two Eqs. 2.27 and
2.28 express different and independent assumptions: the former, the existence of an
energy surface, characterizing the fluid, and the latter, that mechanical and thermal
energy are interconvertible. Indeed the First Law, Second Law, and so on, of
thermodynamics is rather misleading terminology. (For a history of the origin of
thermodynamics, see H. Poincare (1892) [55], and Truesdell and Muncaster (1980)
[56].)
Finally, from (2.27) it follows that in place of (2.28) we can write the following
equation for the specific entropy:
rTDS=Dt @qi =@xi 2m Dij lDkk dij Dij : (2.29a)
In particular, if the heat flux (via the vector q) rises solely from thermal
conduction, then according to Fouriers law gives:
qi k @T=@xi ; (2.30)
where k is the thermal conductivity and qi the components of q.

With (2.30) we obtain from (2.29a) the usual form of the energy equation:
rTDS=Dt @k@T=@xk =@xk 2mDij lDkk dij Dij : (2.29b)

For an adiabatic (qi 0) and non-viscous (inviscid) fluid with m 0 and l 0:
DS=Dt 0: (2.31)
Therefore, if a non-viscous homogeneous fluid be in continuous motion devoid

of heat flux, then the entropy of each particle remains constant. In particular, if the
motion be steady, then the entropy is constant along each streamline.
But in general, the real motion is not isentropic S is different of a constant
in the each point of the flow and in time but constant for each particle (along the
trajectory) for a Eulerian fluid flow.
But even if the flow is isentropic, isentropicity in fluid remains valid only up to
the first shock front encountered by the particles, after which it may well fail its
isentropicity property.
2.3.4 NavierStokesFourier (NSF) Equations
The three equations continuity (2.23b), NS for compressible motion (2.21), and
energy (2.28) with the two state relations
p Rr T and E Cv T; (2.32a, b)
valid for a thermally perfect gas with constants specific heats, constitute the
so-called NavierStokesFourier (NSF) equations for a compressible, viscous
and heat-conducting Newtonian fluid.
In this case it is assumed that the three constitutive (dissipative) coefficients are
functions of r and T:
l lr; T; (2.33a)
and
m mr; T (2.33b)
in (2.21) and (2.28), and
k kr; T (2.33c)
in (2.30).
The compact form of these NSF equations is:
Dr=Dt r u 0; (2.34a)
r Du=Dt rp rgk r:P; (2.34b)

2.4 Initial and Boundary Conditions 35
rCv DT=Dt p r u r k rT F; (2.34c)
where F is the viscous dissipation function and gravity g gk acts in the
negative x3 direction. In (2.34b, 2.34c):
P lr uI 2mDu (2.35a)
and
F 2mTrace Du2 lr u2 ; (2.35b)
where:
Trace Du2 Du : Du 1=4@ui =@xj @uj =@xi 2 ; (2.36)
and D(u) is the rate-of-deformation tensor.
2.4 Initial and Boundary Conditions
Obviously, there has always been considerable interest in initial-boundary value

problems for various systems of partial differential equations arising in Newtonian
fluid dynamics.
This interest of fluid dynamicians stems primarily from efforts to create useful
computational models of various processes for the purposes of simulation, predic-
tion, and the detailed study of various fluid flow phenomena.
Naturally, the initial-boundary value problems for fluid dynamics equations
should have been carefully investigate but unfortunately, rigorous proof of the
existence and uniqueness of solutions of these well-posed fluid dynamics problems
requires very difficult mathematical investigation.
While initial-boundary value problems for these systems of equations are not
easy to analyze, mathematical tools useful for such problems can be found in the
works of Kreiss (1970, 1974) [57, 58], Belov and Yanenko (1971) [59], Oliger and
Sundstrom (1978) [60], Majda (1984) [61], and Kreiss and Lorenz (1989) [62].
The solveability of these problems (a fundamental problem in rigorous mathe-
matical theory of NSF equations) is discussed in Chap. 8 of our Theory and
Applications of Viscous Fluid Flows (2004) [47], and here we note only that the
baroclinic and barotropic Eulerian equations are both symmetrical, hyperbolic
systems, but isochoric and incompressible equations are not hyperbolic. This has
a profound influence on the well-posedness of initial-boundary value problems for
these systems of partial differential equations.
2.4.1 The Problem of Initial Conditions
The above NSF unsteady equations (2.34a2.34c) contain a total of five times
derivatives for the components ui of the velocity u, density r, and temperature T. As
a consequence, if we want to resolve a pure initial value, or Cauchy, problem (in
the L2 norm, for example), then it is necessary to have a complete set of initial
conditions (data) for u, r, and T:
t 0 : u u x; r r x; T T x; (2.37)
where r (x) > 0 and T (x) > 0.

Moreover, when we consider a free-boundary problem or an unsteady flow in a
bounded container, with a boundary depending on time, an initial condition for the
(moving) boundary @Ot has to be specified.
For an NS compressible but barotropic flow for the velocity u and density r,
governed by Eqs. 2.21 and 2.23b with p p(r), as initial conditions we assume:
t 0 : u ub x and r rb x: (2.38a)
For the isochoric Euler equations (2.21) it is necessary to impose also
t 0 : u ui x and r ri x: (2.38b)
If the flow is continuous, when r (x) const, we have an incompressible flow

(Eqs. 2.22 or 2.23b), and it is sufficient to assume only an initial condition for the
velocity u:
t 0 : u ui x: (2.38c)
It is important to note that for both isochoric and incompressible divergence, free
flows, it is necessary that

the boundary integral u n dO vanish (2.39a)
and
r ui 0: (2.39b)
Naturally, this last condition has no analogue for compressible (baroclinic or

barotropic) flows because of the occurrence of the term @r=@t in the continuity
equation (2.23b).
Obviously, for the Laplace equation, which governs an incompressible, irrota-
tional Eulerian unsteady flow (for example, waves on water), we do not have the
possibility of imposing any initial conditions! But this Laplace equation is very
appropriate for the investigation of waves on (incompressible) water, and in this
case it is necessary to consider a free-boundary problem; that is, a problem for
which the fluid (water) is not contained in a given domain but can move freely.
Usually, for this Laplace elliptic equation, one boundary condition is given
(on the contour line containing the fluid) but in the case when the boundary is
known!
Two unsteady, dynamic and kinematic, conditions are needed (and also two
initial conditions) at the free surface (interface) x3 Zt; x1 ; x2 , because the
surface position Zt; x1 ; x2 has to be determined as well as potential function
f (t, x1, x2, x3).
For the free surface problem (for the function f(t, x1, x2, x3) and Z(t, x1, x2))
governing the non-linear waves on water, we can consider two physical problems.
First is the so-called signalling (two-dimensional) problem, in which we have
as initial conditions, when the water is initially at rest in a semi-infinite channel, the
following conditions:
f0; x1 ; x3 0 and Z0; x1 0; when x1 >0; (2.40a)
and at initial time t 0 an idealized wave-maker, at x1 0, will generate a

horizontal velocity disturbance, such that the initial condition is:
@f=@x1 W Bt=t ; for x1 0 and t >0; (2.40b)
where W and t are the characteristic velocity and time scales associated with the
wave-maker idealized by the function B(t/t ).
A second category of the problem for water waves, in the infinite channel, is
obtained by specifying an initial surface shape but zero velocity:
for t 0 : Z a z x1 =l ; x2 =m and
f0; x1 ; x2 ; x3 0; (2.40c)
where l and m are the characteristic wavelengths (in the x1 and x2 directions) for
the three-dimensional water wave motion.
In (2.40c) the scalar a is a characteristic amplitude for the initial elevation of
the free surface characterized by the function z (x1/l , x2/m ). (Concerning the
boundary conditions (kinematic and dynamic) for this free surface problem, see the
next Sect. 2.4.2.)
For meteorological motions (considered in the Chap. 9), when we consider
various approximate model equations f -plane equations, primitive equations,
quasi-geostrophic equations, or Boussinesq equations it is necessary, in fact
(mainly because the filtering acoustic waves), to resolve associated unsteady
adjustment problems for the formulation of consistent initial conditions for these
simplified model equations.
2.4.2 Unsteady Adjustment Problems
For the study of a compressible fluid flow it is necessary to have for the determi-
nation of the solution of the corresponding evolution unsteady equations Euler,
Navier, and NSF a set of initial data for r, u, and T (the Cauchy problem).
However, when we consider, for example, the incompressible equations ((2.13a)
or (2.14a) with r:u 0) for u and p, one is allowed to specify a set of initial
conditions less in number than for the full compressible baroclinic equations.
This is due to the fact that the main low-Mach-number limiting process (in fact,
M tends to zero, with t and x both fixed), which leads to the approximate
incompressible (model) equations, filters out some time derivatives these cor-
responding to acoustic fast waves, because such waves are of no importance for
low-speed aerodynamics and various atmospheric and oceanic motions at least
for steady flows.
When Re"1 (large Reynolds number), from the Navier (incompressible and
viscous) equations we derive the Prandtl boundary layer equations, and accord-
ingly, for an unsteady flow the term @u3 =@t disappears in the limiting momentum
equation for the vertical (x3-direction) component (u3) of velocity! For low
Reynolds number (Re # 0) in the Stokes and Oseen limiting (steady) equations,
the unsteady terms also disappear! Due to this, one encounters the problem of
finding an answer to the following question: What is the initial condition that is
necessary to prescribe for u a solution of an incompressible equation, and in what
way is this condition related to the starting initial conditions (with given data)
associated with the exact, compressible equations?
It is important to note that the exact initial conditions for the full compressible
equations are not in general consistent with the estimates of basic orders of
magnitude implied by the approximate model (without acoustic waves) equations.
A physical process of time evolution is necessary to bring the initial set to a
consistent level as far as the orders of magnitude are concerned.
Such a process is called unsteady adjustment of the initial data set to the
approximate structure of incompressible equations under consideration. This
process of adjustment, which occurs in many fields of fluid mechanics besides
Boltzmann kinetic theory (first discussed by Hilbert in 1912), is short on the time-
scale of approximate simplified equations, and ultimately, in an asymptotic sense,
we obtain values for the consistent set of initial conditions suitable for the
simplified equations.
When we consider the set of approximate simplified model equations, usually
derived heuristically, with timespace fixed, then it is first necessary, for instance,
to elucidate various adjustment problems namely, concerning Prandtl boundary
layer, Stokes and Oseen steady, Navier incompressible viscous, and Boussinesq
equations.
A number of adjustment problems occur in meteorology for atmospheric
motions (adjustment to hydrostatic balance) and to geostrophy (as in the case, for
example, cited in Sect. 9.2), and the reader can find a detailed discussion of these
adjustment problems in Chap. 5 of our Meteorological Fluid Dynamics (1991) [19].
However, it is important to note that, depending on the physical nature of the
problems, we may have two kinds of behaviour when the rescaled (short) time goes
to infinity. Either one may have a tendency towards a limiting steady state, or an
undamped set of oscillations (as, for example, the inertial waves in the inviscid
problem of spin-up for a rotating fluid; see Greenspan (1968; }2.4) [63]). The
problem, considered in Sect. 7.2, is also very particular, and requires a special
approach due to the persistence of acoustic oscillations. (For the terminology of the
initial layer as adapted to this kind of singular perturbation problem, see Nayfeh
(1973; p. 23) [64].) Finally, we note that usually, the process of the unsteady
adjustment of the aerodynamical (or meteorological) fields is a result of the gene-
ration, dispersion and damping of the fast internal waves. According to method of
matching asymptotic expansions (MMAE), the initial conditions for the limiting
model equations are, in fact, matching conditions between the two asymptotic
representations the main one (with t fixed), and the local one (near t 0),
which is a necessary companion to main one!
In conclusion, we can say that the aim of the unsteady adjustment problem can
be stated as follows: Clarify just how a set of initial data associated with an exact
system of unsteady equations can be related to another set of initial data associated
with a simpler, approximate model system of equations which is a significant
degeneracy of the original system of exact equations considered at the start, but
with the less time derivatives in this approximate model system.
In order to solve such a problem it is necessary to introduce an initial layer in the
vicinity of t 0, characterized by a short fixed time t. Obviously such an unsteady
adjustment problem is very important in meteorology for the formulation of a well-
posed initial/Cauchy evolution in a time-prediction problem relative to, for exam-
ple: what the weather will be like tomorrow or for the next few days?
Concerning the rigorous mathematical results of the singular limits in com-
pressible fluid dynamics, see, for instance, the paper by Beiraro da Veiga (1994)
[65], and also the various references in this paper. More recent papers have been
published concerning the passage of compressible incompressible, by
Desjardins, Lions, Grenier, Masmoudi, Hagstrom, Lorenz, and Iguchi; and for
references see our Topics in Hyposonic Flow Theory (2006) [13]. Here we do not
consider these contributions, but instead discuss some of these singular-limit
problems (low-Mach asymptotics) which deserve a serious, consistent, fluid
dynamics investigations via a RAM Approach (as in [13]).
Concerning the low-Mach asymptotic, we observe also that in the case of a flow
affected by acoustic effects in a confined gas (internal flow within a bounded
domain D(t)), over a long time when the wall D(t) is started impulsively from
rest, a multiple-time-scale technique is necessary, because acoustic oscillations
remain undamped and the unsteady adjustment problem (with matching) does not
work (see Sect. 7.2 on applications in aerodynamics).
2.4.3 Boundary Conditions for the Velocity Vector u and

Temperature T
Several boundary conditions could be considered with respect to different physical

situations.
If we consider, as a simple example, the motion of a fluid in a rigid container O
(with a boundary O, independent of time), a bounded connected open subset of Rd
(where d > 1 is the physical dimension), the different structure of the equations
leads to the necessity of distinguishing between viscous (NSF, NS, or Navier) and
inviscid (Eulerian) fluids.
(a) For a viscous (NSF or Navier) fluid: m > 0 and mv l 2=3m > 0
In this case, the physical effects due to the presence of the dynamic (shear)
viscosity coefficient m yield the validity of the steady no-slip condition:
u 0 on @O; (2.41a)
(b) For a bulk-viscous fluids: m 0; mv >0

Since only the bulk viscosity coefficient mv is different from zero, in this
situation the slip boundary condition
u:n 0 on @O (2.41b)
where it is assumed here, and in what follows, that n n(x) denotes the unit
outward normal vector to @O.
(c) For an inviscid (Eulerian) fluid: m 0; mv 0
Also in this case, the slip boundary condition (2.41b) is assumed.
As concerns the (absolute) temperature T, the boundary condition takes differ-
ent forms in the two alternative cases, k > 0 and k 0.
(d) Conductive fluids: k > 0
Several boundary conditions have physical meaning. Limiting ourselves to the
most common cases, we can require :
T Tw on @ODirichlet (2.42a)
k@T=@n X on @ONeumann (2.42b)
k@T=@n hT T0 X on @Othird type; (2.42c)
where Tw > 0 and X are known functions, and h > 0 is a given constant.
(e) Non-conductive (adiabatic case) fluids: k 0
No boundary condition have to be imposed on temperature T if (2.41a) or
(2.41b) are satisfied, since in these cases the temperature is not subjected to trans-
port phenomena through the boundary.
According to Gresho (1992, pp. 4752) [66], if uN0(x) is the initial (t 0)

velocity field for the Navier equation (2.13a), then in the domain O it is necessary
to impose the above incompressibility constraint (2.13b):
r uN 0 x 0; (2.43a)
and on the boundary @O Gs:
n uN o s n ws; 0 n w0 s; (2.43b)
where w(s, t) is the specified boundary condition for the Navier velocity vector
which satisfies Eq. 2.13a.
2.4.4 Other Types of Boundary Conditions
In many situations (inflowoutflow problems) the velocity cannot be assumed to

vanish on O. This is the case, for instance, for the flow around an airfoil, where an
inflow region is naturally present upstream (and an outflow region appears in the
wake), or the flow near a rigid body, where the velocity can be assumed to vanish
only on the boundary of the body. In these cases, several different boundary
conditions may be prescribed.
Let us begin by considering the viscous case. Concerning the velocity-field, a
(non-zero) Dirichlet boundary condition can be imposed everywhere, or, alterna-
tively, only in the inflow region: that is, the subset of O where u n < 0, whereas,
on the remaining part of the boundary, the conditions
u n U > 0 and n D t 0; (2.44a)
have to be prescribed. (Here, t is a unit tangent vector on O, and D D(u) is the

rate of strain (deformation) tensor.)
Let us, moreover, remark that the condition
u n 0 and n D t 0; (2.44b)
could also be considered, on the whole, O. In this case, however, no inflow or

outflow regions would be present.
More important is to analyse the boundary condition for the density r, since now
it turns out that it is necessary to prescribe it on the inflow region. In fact, the first-
order hyperbolic continuity equation (2.13a) can be solved by means of the theory
of characteristics, and the boundary datum for r on the inflow region is indeed a
(necessary) Cauchy datum for the density on a non-characteristic surface.
Let us note also, that if the heat conductivity coefficient k is vanishing and the
fluid is inviscid, the same type of Dirichlet-inflow boundary condition has to
be imposed on the temperature T, since in such a case Eq. 2.6 is also of the hyper-
bolic type for T. More complicated is the situation when the inviscid (Euler) case
m l k 0 is considered.
In fact, in this case the Eulerian system is a first-order hyperbolic one, and the
number of boundary conditions, in the case of an open boundary (or a boundary
located in the interior of a body or fluid), is different depending on whether the
flow is
subsonic : juj < a;
or
supersonic : juj > a;
where
a g RT 1=2 ;
is the local sound speed for the perfect gas.

Take, for example, d 3. An analysis of the sign of the eigenvalues of the
associated characteristic matrix yields the following conclusion: The number of
boundary conditions must be five or four on an inflow boundary, depending on
whether the flow is supersonic or subsonic, and zero or one on an outflow boundary,
again depending on whether the flow is supersonic or subsonic.
Obviously, in the case of an open boundary the normal velocity is non-zero on
the boundary, except at certain points. In both cases, no obvious physical boundary
conditions are known. We will not enter more deeply into this argument, and will
only briefly discuss the inviscid case subjected to the slip boundary condition
(2.41b), for which the boundary is a characteristic surface.
Further information on inflowoutflow boundary-value problems for compress-
ible NS and inviscid Euler equations can be found in two pertinent papers
produced by Gustafsson and Sundstr om (1978)[67] and Oliger and Sundstrom
(1978) [60]. Here, we note only that the solid-wall slip stationary boundary condi-
tion (2.41b) the normal velocity
un u n 0;
should vanish at the boundary, is consistent with the number of inward charac-
teristics (one).
The reader can find also in two papers by Viviand and Veuillot (1978)[68] and
Viviand (1983)[69], a discussion of boundary conditions for steady Euler flow,
considered as the limit (when time tends to infinity the pseudo-unsteady
method) of an unsteady flow (which does not have a precise physical meaning).
It is necessary to note that in numerical, computational, fluid dynamics, these

problems of boundary conditions are very thoroughly considered by often taking
into account the constraints related with the various particularity of the considered
fluid flow problem and associated with numerical algorithms.
Another interesting set of boundary conditions appears when we consider the
free-boundary problem; that is, a problem for which the fluid is not contained in a
given domain but can move freely. In this case the vector n T is prescribed on
(interface) S, where, moreover, u n is required to be zero (stationary case) or
equal to the normal velocity of the boundary itself (non-stationary case).
The value of n T can be zero (free expansion of a fluid in the vacuum), or
to (see, for example, the case of the well-known Benard problem considered in
Chap. 8):
pe n 2sKn rs s; on interface; (2.45a)
where pe is the external pressure, s sT is the surface tension (temperature-

dependent, when the fluid is an expansible liquid), K is the mean interfacial
curvature, and:
rs r nn r; (2.45b)
is the surface (projected) gradient at the interface, respectively.

But in this (viscous) case, it is necessary to also write a heat transfer condition
across the interface, for an expansible and thermally conducting fluid (liquid),
kT@T=@n hs T prescribed function; (2.45c)
which is a Newtons cooling law, where the heat-transfer (constant) coefficient hs

is sometimes called the Biot number. We observe that rigorously, in a Benard
convection problem (with a temperature dependent tension in a free-surface) it is
necessary to take into account two Biot numbers, respectively, for the conduction
(motionless/no convective motion) state and convection state (see Chap. 8).
But the problem of two Biot numbers asymptotics is actually widely open (see,
in Chap. 8, a discussion concerning this two Biot numbers problem in the frame-
work of the Benard problem for an expansible liquid layer on a solid horizontal flat
surface and heated from below.) Since ds=dT 6 o, then, for the film problem it is
necessary to take into account a Marangoni number proportional to the gradient:
ds=dTTTo ;
where T is a constant temperature.

In such a case we consider a thin film BenardMarangoni free-surface problem,
which is fundamentally different from the classical RayleighBenard thermal
instability problem. (See, for instance, in Velarde and Zeytounian (2002) [70],
CISM Courses and Lecture (N0 428), pp. 12390.)
Naturally, we are now imposing one more condition on the interface S, since it
is an unknown of the problem; in the non-stationary case an initial condition for the
interface has to be added too (see (2.40a) or (2.40c)).
For an inviscid incompressible fluid (water) when we consider the wave on the
water (in this case the problem for an irrotational flow is governed by the Laplace
equation), an obvious physical simple condition is (if we assume that the surface
tension is negligible), in place of (2.45a):
p pA ; on interface between water and air above; (2.46)
where pA denotes the air (constant, ambient) pressure on interface S and usually
this ambient air (above the interface) is assumed passive (at temperature TA
const, pressure pA const, with negligible viscosity and density).
In the case of a viscous liquid (thin film Marangoni problem, discussed in
Chap. 8) the above condition (2.46) is replaced by a rather complicated explicit
upper free surface condition (see Sect. 8.2.2).
Now, if the equation of the interface is x3 Z(t, x1, x2), in a Cartesian system
of coordinates (0, x1, x2, x3), then from the Bernoulli incompressible integral we
obtain the following dynamic condition on interface (for the wave on the water
the inviscid fluid problem) according to (2.46):
@f@t 1=2rf2 g Z 0; on x3 Zt; x1 ; x2 ; (2.47)
and since the interface is a material wave surface we have also a kinematic
condition:
@f=@x3 @Z=@t @f=@x1 @Z=@x1

(2.48)
@f=@x2 @Z=@x2 ; on x3 Zt; x1 ; x2 ;
Finally, if we assume that the water rests on a horizontal and impermeable

bottom of infinite extend at x3 h0, where h0 const is supposed finite, then
we have the following simple (flat) bottom boundary condition for the Laplace
equation:
@f=@x3 0; on x3 h0 ; (2.49)
The Laplace equation for the potential f, with (2.40a), (2.40b) or (2.40c) and
(2.47)(2.49), constitutes a well-posed problem for the investigation of the non-
linear unsteady waves on the water (see, for instance, Whitham (1974) [71], and the
review paper by Zeytounian (1995) [72]).
It is important to note that each physical problem has specific boundary con-
ditions related to the intrinsic nature of the problem. For example, in gas dynamics
problems the boundary conditions are different if the fluid flow is subsonic
(M < 1), supersonic (M > 1), transonic (M ~ 1) or hypersonic (M >> 1). If, for
instance, the undisturbed basic flow is in the x > 0 direction and the body in
question (in x, y plane) is located on the x-axis, with its leading edge at the origin
and its trailing edge at x 1 (with non-dimensional variables), then we can assume
that the body shape is described by:
y d hx; (2.50)
where the non-dimensional parameter d is the maximum value of y of the body.

The 2D steady velocity potential j(x, y) is a solution of the steady 2D Steichen
dimensionless equation:
h i h i
a2 M2 @=@x2 @ 2 =@x2 a2 M2 @=@y2 @ 2 =@y2
2M2 @=@x@=@y@ 2 =@x@y 0; (2.51)
with the following relation for the local sound speed:

a2 1 g 1=2 M2 f1 @=@x2 @=@y2 g: (2.52)
In this case the slip condition is:
@=@y d(dhx=dx@=@x;
when x 2 0; 1; on y d hx: (2.53)
Far away, upstream, from the body the flow should be undisturbed, which
requires:
@=@x ! 1 and @=@y ! o as x ! 1: (2.54)
In most applications, the bodies of interest are thin and streamlined, so that
generally d is a small non-dimensional parameter (d << 1). We note here only that
the classical linear, subsonic and supersonic theory is invalid when respectively:

M2 1 =d3=2 O1 transonic similarity
d M O1 hypersonic similarity
Zd O1 far field similarity;
where Z x M2 11=2 y is a characteristic coordinate, such that:
Z
1 with M fixed:
In the case of justification of the well-known Boussinesq (1903) [8] assertion

(see, for instance, Chap. 4) concerning the convection in fluids [13]: The deri-
vatives of the density can be neglected except when they intervene in the calcula-
tion of the force of Archimedes.
It is also necessary to consider the hyposonic (M << 1) case, and in such a case,
for the atmospheric motions it is necessary to take into account the following
constraint:
M=Bo O1 hyposonic similarity; (2.55a)
where
Bo L =RT =g (2.55b)
a ratio of two lengths, is the so-called Boussinesq number (see our (1990) [12], p. 15).
The lee waves problem (related mainly with the dynamic influence of a moun-
tain in a baroclinic, stratified, adiabatic atmosphere) is strongly influenced by the
relief slip condition and also by the upstream flow conditions. In an unbounded
atmosphere the radiation (in a simple Boussinesq model case) Sommerfeld condi-
tion for the Helmholtz equation at infinity (in altitude) plays an essential role. (See,
in [13], various typical examples considered by Guiraud and Zeytounian.)
In the low Rossby model for atmospheric flow, the effect of the solid (earth)
surface is taken into account (by matching) through the so-called viscous Ekman
layer. Indeed, the viscous coefficients are so small that we should expect the
boundary conditions to be close to those valid for the corresponding inviscid sys-
tem. The viscous equations do, however, require additional boundary conditions,
and as an effect, viscous boundary layers may occur at the boundaries.
Such boundary layers may sometimes be appropriate, as in the rigid wall
situation (for example, the Ekman boundary layer). However, at open boundaries
they are inappropriate.
In comparison to flows in interior or exterior domains, there are two new issues
when the boundary extends to infinity. First, in addition to the usual initial and
boundary conditions there needs to be some prescription of fluxes or pressure drops
when the flow domain has several exits to infinity (as in (2.54)). Second, the
solutions of interest often have infinite energy integrals, and recently a technique
of integral estimates to deal with this problem has been developed. These estimates
are called Saint Venants type, because the method was first used in the study of
Saint Venants principle in elasticity.
Concerning, more precisely, the behaviour of an incompressible fluid velocity
field at infinity, we note that in Dobrokhotov and Shafarevich (1996) [73], a simple
method is given which makes it possible to determine an upper bound for the decay
rate at infinity of an incompressible fluid velocity field of general form; that is, to
determine a lower bound for the field itself.
This method is based on the use of simple integral identities which are valid for
solutions of the Navier incompressible, viscous equations, in the external region
which decrease quickly enough. For the equations in entire space, some of these
identities were obtained by the two authors noted above.
The property of slow decay or spreading of localized fluid flow is a consequence
of incompressibility, and is not associated with viscosity alone (in contrast to the
case described by Serrin (1959) [31, 74]), so that it also holds good for an inviscid
Eulerian fluid flow (in this case the reasons for spreading are related with the non-
uniform external flow and non-linearity).
In fact, in order to compute in a bounded region a fluid flow modelled by a
problem formulated on an infinite domain, one often introduces an artificial bound-
ary S and tries to write on the domain O O, bounded by S, a new problem whose
solution is as close as possible to the original exact problem. When the solution of
this new problem in O coincides with the restriction of the original problem, the
boundary S is said to be transparent.
Here, we note also that the reader can find valuable information concerning this
approach with applications to both inviscid and viscous fluid flows in various
recently published papers in the leading journals devoted to numerical fluid dynam-
ics (see, for instance, the recent issues of Journal of Computational Physics).
The general slip condition in an unsteady case:
n u uP 0; (2.56)
is satisfied, in any case, for an impermeable solid wall, where uP is the velocity of
the moving wall. On the other hand, from the kinetic theory of gases, when the
Knudsen number, Kn is small, we obtain
n ^ u uP 0 on a moving wall: (2.57)
As a consequence of (2.56) and (2.57), we again deduce the no-slip condition

(but for a moving wall):
u uP ; on the moving wall: (2.58)
The above condition (2.57) is the so-called weak form of the no-slip condition on
the moving wall.
Concerning the boundary condition for the temperature T on the wall, from the
kinetic theory of gases, again when the Knudsen number Kn is small, we obtain:
T TP bq n; (2.59)
where b is a scalar function (related with the kinetic, Knudsen, sub-layer).

An interesting case of boundary condition is related to the so-called Prandtl
Batchelor condition (see, for instance, the papers by Batchelor (1956)[75] and
Wood (1957)[76]).
For a 2D incompressible, steady Eulerian fluid flow, from Eq. 2.13a, when n 0,
we derive the following equation for the 2D steady stream function cx; y:
r2 c Fc; (2.60)
where the function F(c) is arbitrary! But, if the domain O where the flow is
considered is a bounded connected open subset of R2, then we do not have the
possibility of utilizing the behaviour condition at infinity for the determination of
this function Fc.
The key of this indeterminacy is strongly related with the vanishing viscosity
problem. In fact, with the limiting process Re " 1or; n # 0) in the steady form of
the Navier Eq. 2.13a. Namely, if we assume that the limit streamlines are closed in
O, then according to Batchelor [75] we derive for the limit Euler stream line
g (which is the one of the stream lines c constant) the following
PrandtlBatchelor condition:
dFc =dc 0 and Fc F const: (2.61)
As a consequence, the Eulerian vorticity, o 1=2Fc , for a steady

incompressible 2D fluid flow, is constant in any region where the streamlines are
closed.
From the matching performed by Wood [76] with the corresponding Prandtl
boundary-layer in the vicinity of @O the value of this above constant, F , is well
determined.
On the other hand, in Guiraud and Zeytounians short paper (1984) [77] a
process for setting in motion a viscous incompressible liquid inside a 2D cavity is
considered, and it is shown that the basic process occurs for a time of the order of
t O(Re). Then a flow, a la PrandtlBatchelor, with constant Euler vorticity is
established after a time t >> Re.
In this same paper [77], a GZ functional equation is derived which governs the
distribution of the vorticity in the main stage of interest, and for the simple case of a
cylindrical cavity it is shown that the vorticity tends towards its own steady-state
value exponentially.
Finally, concerning the case of overspecified and underspecified boundary
conditions, it is important to note that when for a given problem the number of
boundary conditions is overspecified, the difference approximation (for a numerical
calculation) may well be stable. However, the effective boundary conditions which
influence the solution are, in general, difficult to determine, especially for problems
in several space dimensions.
They may well be a complicated function of the conditions given and bearing
little resemblance to them. An additional complication induced by over-specifica-
tion is that the underlying solution being approximated is not generally continuous.
In order to avoid the problems associated with the proper selection of boundary
conditions, the order and type of the differential equations is often raised to obtain a
problem that is easier to analyze and approximate.
For example, the Eulerian equations are usually modified by adding dissipative
terms so that the number of boundary conditions is appropriate. Unfortunately, this
idea seldom works. If a spurious boundary layer of appreciable size results, the
effects are not unlike those for discontinuities (for a system of equations, the errors
can propagate away from the discontinuity through other components of the
solution), and unless the dissipative terms are very large, the error introduced at
the boundary will again propagate into the interior.
Now, if the boundary conditions are underspecified there are no a priori
estimates for the differential equations. In order for an approximation to be com-
putable these must be a sufficient number of boundary conditions specified for the
approximation. This cannot be fewer than the number required for the differential
equation.
The well-posedness of the initial boundary value fluid flow problems follows,
to some extent, from properly formulated initial and boundary conditions, and
is strongly linked with the various facets (through the existence and uniqueness
results) of the solvability of these fluid flow problems.
We recognize that in large part what might be called mathematical topics
in fluid dynamics has remained closed to the mainstream of theoretical fluid
dynamics and mathematical physicists, due in large part as judiciously observed
in the book by Doering and Gibbon (1995) [78] devoted to applied analysis of the
Navier (NavierStokes incompressible) equations to the technical nature of
rigorous investigations, often phrased in the unfamiliar language of abstract (non-
linear) functional analysis.3
The above summary of Chap. 2 presents the main theoretical concepts and
principles, and also equations and associated initial and boundary conditions, of
classical/Newtonian fluid dynamics. Various theoretical concepts can be found in
our books devoted, respectively, to non-viscous (2002) [37] and viscous (2004) [47]
fluid flows. In our survey paper on the well-posedness of problems in fluid dynam-
ics (a fluid-dynamical point of view) (1999) [79] the problem is carefully consid-
ered, and an historical survey of some mathematical aspects of Newtonian fluid
flows can be found in our (2001) [29] surveys.
3
The curiour reader can find in Handbook of Mathematical Fluid Dynamics, vol. 1 to 4,
numerous papers related with rigorous mathematical results, of existence, unicity, regularity,
well-posedness and limiting processes for solution of fluid flow problems, mainly by
compactnesse-a very abstract functional approach!
Chapter 3
From NSF Equations to General Main Model
Equations
First of all, the main question is relative to the Rational Asymptotic Modelling
Approach (RAMA) of Newtonian fluid flow problems. Our task is as follows.
Starting from a real technological or geophysical problem, mathematically
formulated via the NSF equations and associated initial and boundary conditions,
we want to obtain a simplified and consistent model problem which is possibly
resolved via a numerical/computational simulation with the help of a super-high-
speed computer.
Our mathematical status (expounded, more precisely, in Chap. 6) is very nave,
and is based on a logical, non-contradictory, constructive process, according to a
main heuristic postulate and some associated key steps for its realization.
The desirable, but difficult, requirement concerning the rigorous proof that the
error admitted is really of the order suggested is simply abandoned, in consequence
of two goals: first, to consider, as much as possible, real complicated problems; and
second, to efficiently assist numericians in our numerical simulation.
In our survey paper with Guiraud (1986) [11], devoted to emphasizing the
considerable support that mastering asymptotic tools may afford to researchers
embarked on rational modelling of very difficult problems of fluid flows, the reader
can find a guide for classification of various models in a RAMA.
Concerning the level1 of general models, our thesis below is that very
often, various writings in fluid dynamics (especially in text-books for university
undergraduates) are organised through several particular fluid flows.
1
In addition to above-mentioned level of general models, it is necessary to consider also two
particular levels: local models, in order to elucidate the behaviour in some localized region, and
global specific models, which sometimes occur when the flow under consideration, as a whole,
may be determined via asymptotic modelling.

52 3 From NSF Equations to General Main Model Equations
3.1 Non-dimensional Form of the NSF Typical Problem
In this book we work mainly with dimensionless time t and position vector x, and
also with dimensionless velocity vector u and thermodynamic functions p; r, and T.
Respectively, the physical time-space, velocity vector and thermodynamic func-
tions (with dimensions) are reduced via the characteristic time t , the characteristic
length L , the characteristic constant velocity U , and characteristic constant
thermodynamic values p , r and T .
For simplicity I have used, for our above dimensionless quantities, u; p; r, and
T, as function of dimensionless (t, x), the same notation as in Chap. 2.
3.1.1 Non-dimensional NSF Equations and Reduced Parameters
First, in non-dimensional form, for a thermally perfect gas, we obtain as equation

of state:
p r T; (3.1)
because p R r T , according to (2.8), written for the reference values.

In non-dimensional form we have for the time derivative, with respect to
material motion (along trajectories):
S D=Dt S @=@t u r; (3.2)
where
S L =t U is the Strouhal number (3.3)
From (2.34ac) with (2.35a, b) we derive the following three unsteady full NSF
non-dimensional equations:
SDr=Dt r u 0; (3.4a)

r S Du=Dt 1=gM2 rp Bo=gM2 rk 1=Rer flr:uI 2mDug ;
(3.4b)
r S DT=Dt g 1 p r u g=Pr Refr krTg

g 1 gM2 =Ref2m TraceDu2 lr u2 g: (3.4c)
which form a closed set of three non-dimensional partial differential equations, for
four dimensionless quantities, u; p; r and T, as functions of dimensionless time
3.1 Non-dimensional Form of the NSF Typical Problem 53
space variables t; x, provided that one adds the non-dimensional equation of state
(3.1) for an ideal (thermally perfect) gas.
The reduced parameters in the above non-dimensional NSF equations (3.4ac)
are
Re L U =n the Reynolds number; (3.5a)
M U =gRT 1=2 the Mach number; (3.5b)
Pr Cp m =k the Prandtl number; (3.5c)
Bo gM=FrL 2 the Boussinesq number; (3.5d)
where
FrL 2 Uo2 =gL ; (3.6)
is the square of the Froude number based on the characteristic length L .

The Strouhal number takes into account the unsteady effects and the Reynolds
number the viscous effects. In particular, in Chap. 5, the consideration of the
limiting case, Re " 1, when S O(1), leads, in the framework of our RAMA, to
a complicated asymptotic structure of the above NSF system of equations (3.4ac)
which is very different from classical regular coupling:
Euler non viscous fluid flow , Prandtl boundary layer fluid flow
This complicated, four regions, singular coupling of the full unsteady NSF
equations arise mainly because of the singular nature of the Prandtl boundary layer
equations near t 0, where the initial data are given.
The compressibility is strongly related with the Mach number, and the acoustics
effects are, in fact, closely linked to the following similarity relation:
S M M O1; (3.7)
when
S 1
a strong unsteady-state effect, and
M 1
quasi-incompressibility.
In reality, only the low Mach number asymptotics presents the possibility of
deriving in a consistent form, as a limiting simplified incompressible model of
above NSF system (3.4ac), the associated Navier incompressible and viscous
system of equations.
These Navier equations are again singular near t 0, where the equations of
acoustics are significant but, unfortunately, this is really the case only if the fluid
flow domain is not a bounded, time dependent, domain (a cavity, as is the case in an
internal flow; see, for instance, Sect. 7.1).
In the Benard thermal convection problem a liquid layer heated from below,
when the fluid is an expansible liquid such that the relation:
r rT (3.8)
is often assumed, we can introduce the constant coefficient of cubical expansion:

dr
a 1=r : (3.9)
dT TT
Concerning the relation (3.8) it is necessary to observe that it is only consistent

in the framework of a RAM Approach as a leading-order approximation of
a trivariate liquid, r r T; p, when the above cubical expansion (3.9) of the
considered liquid is very small (more precisely, for a DT 1) in the Grashof
(Gr) number (3.10) below.
In this case, if DT is the difference between the two temperatures in a layer of
thickness d , then the following similarity relation:
a DT =Frd 2 Gr O1; (3.10)
plays a significant role for the derivation of classical OberbeckBoussinesq shallow

thermal convection equations (see, for instance, Chap. 8). In (3.10) the so-called
Grashof number is, in fact, a similarity parameter, and
Pr Gr Ra (3.11)
is the Rayleigh number. The reduced dimensional parameter
Pr Re Pe0 L U =k0 =r0 Cp (3.12)
is the Peclet number.

We observe that in the definition of Gr (3.10), in the numerator the parameter
e a DT (3.13)
is the expansibility parameter a small 5
103 parameter for many liquids,
which plays the role as a main small parameter in our derivation (in Chap. 8) of an
approximate limiting consistent model for shallow thermal convection (a la
RayleighBenard). But for this it is necessary to assume also that
Frd 2 1 such that Gr O1

In fact, in the framework of the full Benard convection problem, three significant
convection cases merit interest:
1. Shallow-thermal, Rayleigh-Benard convection, without viscous dissipation and
upper free surface deformation, when, Frd 2 1.
2. Deep-thermal, Zeytounian convection, with viscous dissipation, when,
1 Frd 2 a gd =Cv .
3. Thermocapillary, Marangoni convection, with a deformable upper free surface,
when, Frd 2 1.
A fourth case also deserves careful asymptotic investigation, linked with:
4. Ultra-thin film, when, Frd 2 1;
For this fourth case, many questions are still open and deserve serious discussion
and investigations.
For Pr 1 with Re 1, but Pe O(1), the fluid motion is quasi-non-viscous
but strongly thermally conducting, and the so-called high thermal conductivity
model equations are valid, with a specified boundary condition for the temperature
on the wall.
We observe that with Pr 1, and Pe O(1) we assume that
m k0 =Cp r L U : (3.14)
Strictly speaking, when Re 1 the viscosity effects remains important, mainly

in a boundary layer near the wall which has thickness of the order of
HCL L =Re1=2 ; (3.15a)
and in this case we have, again, a similarity relation, namely
e2 Re Re? O1; where HCL =L e; (3.15b)
when we assume that:
e # 0 and Re " 1: (3.15c)
In (3.15b, c) the parameter e is a long-wave parameter, and in the viscous fluid,

when the limiting process (3.15b, c) is performed, the so-called long-wave approx-
imation is very similar to classical (Prandtl, 1904) boundary-layer approximation.
For the atmospheric viscous and non-adiabatic motions (when, obviously, the
Mach number is very low), we take into account (see Sect. 9.2) the Coriolis force,
which is characterized by the Coriolis parameter,
f 2O sinf ; (3.16a)
where O jOj, and O is the angular velocity of the rotating earths frame, and f a
reference latitude (for f 45 , we have a 6,300 km for the earths radius).
In this case it is helpful to employ spherical coordinates l; f, and r, and for the
gradient operator r we can write:
r 1=rcosf@=@l i 1=r@=@f j @=@r k: (3.16b)
But it is more convenient to use a right-handed curvilinear coordinates system

(x, y, z) lying on the earths surface (for a flat ground we have r a ) at latitude
f and longitude l 0; namely:
x a cosf ; y a f f ; z r a : (3.16c)
Although x and y are, in principle, new longitude and latitude coordinates

in terms of which the basic NSF (written for the atmospheric motions with
the Coriolis force term characterized by the Rossby number Ro) equations may
be rewritten without approximation, they are obviously introduced with the expec-
tation that for a small sphericity parameter,
d L =a (3.16d)
they will be the Cartesian coordinates of the f -plane approximation.

For the atmospheric motions a fundamental parameter is the Kibel number,
such that
Ki 1=f t Ro 1=f =L =U ; (3.17)
when we assume that the Strouhal number S 1 or t L =U .

On the other hand, for atmospheric motions in the whole troposphere, a very
significant characteristic constant vertical altitude is
Hs RT =g (3.18a)
and usually (for the weather prediction, in the hydrostatic approximation)
es Hs=L 1: (3.18b)
Indeed, the viscous and non-adiabatic effect in the tropospheric motions are
limited within a thin layer near the ground the Ekman layer which is
characterized by the following Ekman number:
EkS Ki=Res n =f HS 2 ; (3.19a)
where, for Re 1,
ReS eS 2 Re O1: (3.19b)

The reader can find in my two books (1990) [12] and (1991) [19] various asymp-
totic models for the atmospheric motions. Section 9.2 presents an application of the
RAM Approach to atmospheric motions when the Kibel number is a small parameter.
Finally, when we consider the oscillatory viscous flows (when the Strouhal number
S 1; see, for instance, Riley (1967) [80]), when the flow is induced by a solid body
performing harmonic oscillations (assuming that the velocity of the body is U cos o)
in an unbounded viscous fluid which is otherwise at rest, with t 1=o being a typical
time. In this case, in dimensionless Navier equations for an incompressible and
viscous fluid we have two main parameters, a and b2 , such that:
1=S a U oL and Re=a b2 oL2=n : (3.20)
In fact, Riley (1967) considers solely the situation corresponding to a 1,

high Strouhal number, and for the case of b2 O1 it is assumed that the
oscillatory fluid flow is strongly viscous: Re 1.
3.1.2 Conditions for the Unsteady NSF Equations
Roughly speaking, we can expect that the equations of motion for a viscous fluid
are parabolic. However, a more detailed analysis of the structure of equations (3.4a)
and (3.4b) for the velocity vector u and the density r shows that this seems correct
concerning equation (3.4b) for the velocity vector u, but not quite correct for the
system of Eqs. 3.4a, b, because the continuity equation with respect to density r is
hyperbolic in the compressible case, even for non-trivial viscosity!
Thus, to be more exact, we can say that a system of two NavierStokes equations
(3.4a, b), with a specifying relation p Pr and viscosity coefficient
m 3=2 lr m r;
is a hyperbolicparabolic or incompletely parabolic system according to the defini-

tion suggested by Belov and Yanenko (1971) [59] and Strikwerda (1977) [81], in
the rigorous study of the mathematical properties of these equations.
The Fourier equation (3.4c), written in the dimensionless form, for a small
characteristic Mach number (hyposonic flow, M 1), with respect to temperature
T, is parabolic.
Kreiss and Lorenz (1989) [62] presents pertinent information concerning the
initial-boundary value problems for NavierStokes equations. On the other hand,
Oliger and Sundstr om (1978) [60] discuss initial-boundary value problem for
several systems of partial differential equations from fluid dynamics.
We observe also that both the viscosity and heat conduction terms (characterized
mainly by the inverse of the Reynolds number, 1/Re), in dimensionless equations
(3.4b) and (3.4c), are usually very small (Re 1), but since these terms change the
character of the NSF partial differential equations (from parabolic to hyperbolic in

the case of a Eulerian non-viscous adiabatic fluid flow), we are obliged to reinves-
tigate the boundary conditions.
Concerning the initial condition, at initial time t 0 the situation is unchanged
relative to hyperbolic Euler unsteady compressible non-viscous equations con-
sidered in Zeytounian (2002) [37], Chap. 8. In the NSF equations (3.4ac) we
have three derivation in time for the velocity vector u, density r, and temperature T.
As a consequence, if we want resolve a pure initial-value or Cauchy problem
(in the L2 -norm, for example), it is necessary to have a complete set of initial
conditions (data) for u, r, and T (as was mentioned in Chap. 2).
We see (according to Sects. 2.4.3 and 2.4.4) that several boundary conditions
could be considered with respect to different physical situations for the above NSF
system of Eqs. 3.4ac.
If we consider, as a simple example, the viscous and heat-conducting fluid flow
above a rigid, solid, steady wall, simulated by the equation
z 0;
then, when m > 0 and mv l 2=3m > 0, it is necessary to assume the

following no-slip condition:
u 0 on z 0: (3.21)
for the velocity vector in non-dimensional NSF equation (3.4b).

For the (absolute) temperature T, in the case of a heat conductive fluids, when
k > 0, in non-dimensional NSF equation (3.4c), just to limit ourselves to the most
common cases, we can require, in dimensionless form, the following temperature
condition:
T 1 wYt; P on z 0: (3.22)
where P is a position point on the wall z 0. In (3.22), w is a temperature reduced

constant parameter a rate for the wall temperature fluctuation, simulated by the
function Y(t, P) a measure for the given known function Y(t, P), which define the
temperature field on the wall.
This parameter w plays an important role during the RAM Approach via various
similarity rule with the above reduced parameters (3.5ad), and especially with the
Mach number, M.
We observe that in the compressible case, from the continuity Eq. 3.4a, with (see
(1.37)) initial condition r r x > 0 at t 0, we can affirm (see, for example,
Valli (1992) [82]) that one of the main mathematical problems (unlike the incom-
pressible case) is to find a priori estimates assuring the non-degeneracy of the
density r.
Today, 76 years later, despite tremendous progress in many aspects of the
mathematically rigorous Navier (NS incompressible) theory (see, for example,
3.2 General Main Model Equations 59
the book by Temam (2000) [83]), we have not yet answered the fundamental
questions raised by Lerays pioneer papers (1933/1934) [84]. That is, we have not
determined whether a solution that is initially smooth can develop a singularity at
some later time, or whether singularities are an important feature of turbulence.
This is also related to another question: Are singularities really necessary to
explain turbulence?. Lions book (1998) [51] is mainly devoted to new rigorous
mathematical results for NS compressible (but isentropic) fluid flows.
3.2 General Main Model Equations
Here I reiterate that our main goal is modelling a scientific activity which consists of
deriving (according to the RAM Approach), for various technological and geophys-
ical stiff fluid flow, consistent, approximate, simplified model problems, in a such
way that they become amenable, on the one hand, to mathematical analysis, and on
the other hand, to numerical simulation by a super-high-speed computers.
This is not an easy task especially when numerical computation involves
simultaneously, dominant and negligible effects in so-called stiff problems. In
such a complicated case, for the numerical simulation it is necessary to be able, via
the RAM Approach, to derive model consistent problems where the stiffness is
smoothed!
A such, the RAM Approach is remarkably illustrated (see Sect. 7.1) in the case
of an internal flow in a turbomachinery row with a large number of blades the
space between blades being very small when the stiffness is replaced by the
smoothness via a source term in a simplified model, consistent, through-flow
(without blades!), which occurs from redistribution of forces acted on the flow by
the blades of the row.
Below I present a number of fluid flow problems which are linked with general
main models, though we make no reference to the literature, as such reference is
both unnecessary and rather arbitrary.
In many cases of practical interest, the parameters (3.5ad) in NSF non-
dimensional equations make take on extreme values, either very large or very
small these extreme values of parameters being closely linked with various
particular fluid flows, which are of great interest in understanding the profound
nature of flow related with the full NSF problem.
3.2.1 Some General Main Models
Inviscid Euler fluid flows, which are often considered as models, used from the
outset, need to be embedded in the more general main model of slightly viscous
(laminar large Reynolds number flow) or slightly frictional (turbulent?) flow, to
which RAMA is applied when Re "/ .
Creeping flows, with numerous applications in thin films (lubrication), micro-

hydrodynamics, and so on, should be considered as flows at low Reynolds
Re 1 number. Entire books are devoted to creeping flows, but the role
of the low Reynolds number, as the main small parameter, related with Re # 0, is
ignored except in the very few expository pages.
Incompressible flows are seldom considered as flows at small Mach number.
This obviously can become almost nonsensical, as when one deals with incom-
pressible aerodynamics and acoustics. Phenomena such as sound produced by quite
low-speed flow cannot be understood other than as low-Mach-number (hyposonic,
M # o) aerodynamics.
Rotating flow, which dominates both industrial and geophysical nature appli-
cations, are indeed asymptotic models of flow at low Rossby number Ro # 0 or
low Kibel number Ki # 0.
Large-scale models of flows, of current use in simulations of meteorological
or oceanographic applications, are extracted from asymptotic modelling which
explains the role of the hydrostatic balance.
A number of models for the flows in porous media, flow with suspensions, or
turbulent flow, should be considered as models obtained through some kind of
homogenization (see Sect. 6.4).
The NavierFourier model, which take into account the effects of a slight
compressibility and the thermal effects, is derived from NSF equations (3.4ac),
with the condition (3.22), when we assume that the temperature parameter w is
small, such that for M 1 we assume the following similarity rule:
w l0 M2 with l0 O1:
The Stokes and Ossen compressible model equations, when Re and M both tend
to zero, such that
Kn M=Re 1:
This above constraint is realized if we take into account the following similarity
rule:
M L Re1a ; with L O1 and a >0; when Re # 0:
The model equations of non-linear acoustics when, in place of Re (given by

(3.5a)), it is necessary to introduce an acoustic Reynolds number Reac, such that:
Reac Re=M 1;
the characteristic length scale being
L 1=k ;
3.2 General Main Model Equations 61
where
p
k o= gRT
is the wave number, and o is a reference frequency determined by the main

frequency of the source signal.
The shallow-thermal, RayleighBenard, convection model is derived when
simultaneously the Froude number (Fr) and the expansibility parameter (e) tend
both to zero such that the Grashof number Gr O1.
Various meteo-fluid dynamics models linked with low Kibel number limiting
process (as in Sect. 9.2).
Discussion of the RAM Approach applied to some of the above problems can be
found in [12, 13, 17, 26], and [27]; and see also the above Overview, Sect. 1.3,
concerning my earlier books on modelling in Newtonian fluid flows. In Chaps. 79,
devoted to applications of the RAM Approach, we also consider some aerodynam-
ics, thermal convection and meteo/atmospheric stiff fluid flows, and derive the
associated consistent model problems.
3.2.2 A Sketch of the Various General Main Models
Below, the reader can find a sketch of the various consistent general main model
equations issues, with the help of a RAM Approach, from the NSF full unsteady
(and assumed exact) system of equations.
A first observation concerning these above-mentioned general main models is
relative to the fact that the later are derived from the RAMA when the space-time,
(x, t), fluid domain is (implicitly assumed) fixed. As a consequence (often, unfortu-
nately), these general main models equations turn out (in many cases) to be of non-
uniform validity, as they break down in certain regions of the fluid flow.
(Concerning this problem see Sect. 3.3.)
More precisely, these regions are usually strongly linked with the assumed initial
and boundary conditions. These general main models break down near the time, t 0,
where the initial data are given, and in the vicinity of the wall, z 0, where the
conditions (3.21) and (3.22), for u and T, are assigned. Below is an overview relative
to the above general main models arising from the full NSF equations (Fig. 3.1).
The experienced reader will be aware of the fact that the partial differential
equations of fluid dynamics (NSF system of Eqs. 3.4ac solely) are not sufficient
as is often the case, unfortunately, in mathematical fluid dynamics for discussing
fluid flow problems.
This seemingly anodyne remark not only has far-reaching consequences, but
also presents the possibility of discovering various new intrinsic structures from
NSF equations.
A good illustration of such a possibility (it seems to me) is our recent RAM
Approach, expounded in Chap. 5, applied to NSF full unsteady equations in the
framework of a large Reynolds number the discovered five regions structure being
a direct consequence of the singular nature of equations of the unsteady compress-
ible boundary layer near the initial time ( 0) where the initial data for the unsteady
full NSF equations are given.
Fig. 3.1 From NSF equations to some general main model equations
3.3 Non-uniform Validity of Main Model Equations

and the Local Limiting Processes
The regions of non-uniform validity, near t 0 and in the vicinity of z 0 on the

wall, of general main model equations, are mainly the results of the elimination of
some terms with partial derivatives in t and z, in dimensionless NSF full unsteady
3.3 Non-uniform Validity of Main Model Equations and the Local Limiting Processes 63
Eqs. 3.4ac during the main limiting processes, with t and x ( x, y, z) fixed,
mentioned in Sect. 3.2.1.
3.3.1 Large Reynolds Case: A First Naive Approach
When the Re tends to infinity, with t and x fixed, we first consider the non-
dissipative, Euler, main limiting process:
LimE Re " / with t and x fixed (3.23)
in NSF equations (3.4b) and (3.4c). The parameters Pr and M2 being fixed, the
dissipative terms in the right-hand side of both these equations disappears, and we
recover the Euler compressible equations:
S DrE =Dt r uE 0; (3.24a)
rE S DuE =Dt 1=gM2 rpE Bo=gM2 rE k 0; (3.24b)
r S DTE =Dt g 1 p r uE 0; (3.24c)
pE rE TE ; (3.24d)
where
uE ; rE ; TE LimE u; r; T:
Obviously, for these non-viscous equations it is necessary to assume, in place of

boundary wall condition (3.21), the slip condition:
uE :k wE 0 on z 0; (3.25)
This is the single boundary condition on the wall, for the above Euler unsteady
compressible equations, because the term @ 2 uE =@z2 is absent in the right-hand side
of the limit Eq. 3.24b.
Concerning the initial conditions, it is again necessary to specify three initial
conditions. However, the prescribed data in these three initial conditions, for Euler
equations (3.24ac), are all really the same than as for full unsteady NSF equations
(3.4ac)? This is an interesting, but disquieting, question.
The singular nature of the Euler main limiting process in the vicinity of the wall
z 0 is directly related with the fact that condition (3.21) for an NSF system of
equations is replaced by the slip condition (3.25). In addition, the Euler equations
(3.24ac) take (unless dissipative terms) the form of a hyperbolic system.
This leads inevitably to the necessity of a RAM Approach for a slightly viscous
and (when Pr O(1)) heat-conducting flow in a thin layer close to wall z 0, and
for this, according to a RAM Approach, it is necessary to consider a new limiting
process: the Prandtl local limiting process in NSF equations:
p
limPr Re " /with t and x; y; z z=1= Re fixed; (3.26a)
with
p
vPr ; wPr ; rPr ; TPr LimPr v; w=1= Re; r ; T: (3.26b)
From (3.26a, b), our small parameter, in a RAM Approach, is

p
e 1= Re 1 (3.27)
since Re 1.
In such a case in the leading-order we derive the well-known Prandtl boundary-
layer equations. A detailed derivation of these Prandtl boundary-layer equations is
presented in [26], Chap. 7. Here, in Chap. 5, we return to the case of a large
Reynolds number fluid flow in the framework of a four-regions structure of NSF
full unsteady equations at high Reynolds number.
Here, if we simply assume that the dissipative coefficients l, m, and k are
constant in dimensionless equations (3.4b, c) for u and T (in fact, equal to one),
then we obtain, with (3.26a, b), the following Prandtl BL unsteady equations (with
the horizontal gradient operator D @=@x; @=@y:
S @rPr =@t D:rPr vPr @rPr wPr =@z 0; (3.28a)
S rPr @vPr =@t vPr :DvPr wPr @wPr =@z

(3.28b)
1= gM2 DpPr @ 2 vPr =@ z2 ;
@pPr =@z 0; (3.28c)
S rPr @TPr =@t vPr :DTPr

wPr @TPr =@z g 1pPr D:vPr @wPr =@z
g=Pr@ 2 TPr =@z2 g g 1M2 j@vPr =@zj2; (3.28d)
pPr rPr TPr ; (3.28e)
where in (3.28b) and (3.28d) some dissipative terms are again present.
It is well known that the above two limiting systems (Euler and Prandtl) of
Eqs. 3.24ad and 3.28ae are related to the following matching condition (the
concept of matching will be discussed in more detail in Sect. 6.4.2):
limz"1 LimPr limz#o LimE ; (3.29)
As a first consequence of (3.29) we recover the slip condition (3.25), which is, in
our RAM Approach, a mathematically logical consequence of the passage from an
incompletely parabolic (hyperbolicparabolic) NSF system to a hyperbolic Euler
system.
In Chap. 5, devoted to a deconstruction2 of the full NSF unsteady equations,
when Re "/ , the consequence of the strong (BLPrandtl) degeneracy (3.28c) is
analyzed in the relation of initial conditions for a well-posed initial-boundary
unsteady problem for the Prandtl BL equations.
3.3.2 Low Mach Number Case: The Navier System of Equations

and Companion Acoustics Equations
The Navier equations govern incompressible viscous fluid flows, and are formally
derived from the full NSF equations (3.4ac), when M # 0.
Here, for aerodynamics, we assume that Bo 0. The case when Bo is not zero
(the gravity force being active) is considered in Chap. 4, in the framework of a
rational justification of well-known Boussinesq approximation.
More precisely, we consider the following Navier main limiting process:
LimM M # 0
(3.30)
with t and x fixed;
and we assume, again, that the dissipative coefficients l, m, and k are equal to one in
the non-dimensional equations (3.4b, c). During the above LimM (3.30) we assume
that S, g, Re, and Pr are all O(1).
First, according to (3.4b), in a very naive way we assume the following asymp-
totic expansion for the pressure p:
2
From the French deconstruction a word invented by Jacques Derrida (see the book by Charles
Ramond, Le vocabulaire de Derrida, Ellipses, Paris, 2001), which presents us with the possibility
of understanding the intrinsic structure of the NSF equations, and reveals the presence of a
profound unity in the puzzle of the partial models of fluid dynamics problems. To deconstruct
the NSF system of equations we have the possibility of unifying these partial models, used in
Newtonian fluid dynamics, according to our RAM Approch, using the various limiting processes
linked with the reference dimensionless parameters in equations, conditions, and geometry of
considered non-dimensional fluid flow problems. From a such process we re-establish a well-
ordered and unified family of partial fluid flows.
p p0 t M2 pN t; x :::; (3.31)
and we enquire how one can obtain information concerning the function p0(t).
Scrutinization of the starting NSF equations, and trying more sophisticated expan-
sion processes (in place of (3.31)) proves to be of no help. However, a way out
appears if (3.31) holds in the whole of a domain of space where p is known to
leading order, with respect to the smallness of M, as far as its dependence on x is
concerned. This, obviously, occurs when (3.31) holds in a neighbourhood of infinity
where the pressure is constant (especially in the case of a steady fluid flow past a
finite body external aerodynamic problem), and this leads, then, to p0(t) 1.
On the other hand, let us assume that the gas is contained in a container O
bounded by an impermeable but eventually deformable (with time t) wall, so that
the volume occupied by the gas is a given function of time: namely, V0(t). An
obvious way to proceed is to assume that the density and temperature go to definite
limits: r0 t; x and T0(t, x), according to (3.30), LimN. It is a very easy matter to
derive, from Eq. 3.4c, an equation satisfied by T0(t, x). Such an equation involves
the unknown function p0(t), and it has an obvious solution T0(t, x) T0(t), which
holds provided that the two unknown functions, r0 (t, x) and T0(t, x), meet the
requirement that
T0 =r0 g1 be independent of time 1: (3.32a)
It is then easy to reach the conclusion that
r0 t p0 t1=g ; (3.32b)
and, conservation of the mass for the whole of the gas contained within the
container gives
p0 t V0 tg ; (3.32c)
so that we have found our way out of the indeterminacy concerning the leading term
in (3.31). Of course, our argument relies on T0(t) being independent of space, and
we have to discuss the adequacy of that. It is obviously a matter of conduction of
heat within the gas. Such a phenomenon might have two origins one of which is
dissipation of energy within the gas; but consideration of Eq. 3.4c tell us that this
enters into account at a rate of O(M2) and is negligible as far as T0(t) is concerned.
A second origin for the conduction of heat is through the variations of temperature
on the wall or from heat transfer through it.
Let us return to Eqs. 3.4a, b and assume that u goes to uN(t, x) according to
(3.30). We put:
pN =gr0 t LimN fp p0 t=gr0 tM2 g p; (3.33)

This p is a fictitious pressure perturbation, and we derive the following quasi-

incompressible system of equations for Navier velocity vector uN and pressure
perturbation p:
r:uN S=r0 t dr0 t=dt; (3.34a)
S @=@t uN :ruN r p mT0 =r0 tRer2 uN ; (3.34b)
r0 t 1=V t; p0 t V tg ; T0 t Vt1g : (3.34c)
The above system of Eqs. 3.34a, b with 3.34c is a rather slight variant of the
Navier equations (see also Sect. 7.2) which consists of a fluid flow with time-
dependent viscosity and no-divergenceless velocity vector, rather than a
divergenceless motion with constant in space.
The usual set of Navier equations is obviously obtained for a constant-volume
container, and in this case (p0(t) and r0 t 1, in (3.33)) we can write, in place of
(3.34a, b), the following classical Navier (NS incompressible) equations for a
divergenceless velocity vector uN and fictitious pressure perturbation p:
r:uN 0; (3.35a)
S @=@t uN :r uN rp 1=Rer2 uN : (3.35b)
On the one hand, for the full unsteady NSF compressible and heat-conducting
equations (3.4ac), with Bo 0, aerodynamics case, if we want to resolve a
Cauchy problem it is necessary to impose three initial conditions for r, u, and T.
On the other hand, when considering the above Navier general main model
equations (3.35a, b), we must give only the initial value for uN, and this Navier
initial condition (uN) is such that:
r:uN 0: (3.36)
This shows that the Navier incompressible system (3.35a, b) where acoustic
waves are filtered is certainly not valid (singular) near t 0.
As a consequence, it is necessary to derive a new model of limiting equations
consistent in an initial (local) thin region with a short time. We therefore
introduce
t t=M O1; when M # 0; (3.37a)
and consider the following local acoustic limiting process:
limAc M # 0 with t and x fixed: (3.37b)

This short time t is well suited for studying the transient behaviour an unsteady
adjustment problem related to a process of matching between main (for instance,
(3.31)) and local,
p 1 Ma pac :::; (3.38a)
expansions.
In fact, for the case of an external aerodynamics when the matching is
possible we must deduce from NSF equations (3.4ac) a local consistent system
of acoustics equations which realize the above-mentioned matching.
For this, the following asymptotic expansions are considered, with (3.38a) for p,
u, r, and T, in the framework of unsteady NSF dimensionless equations (3.4ac):
u uac :::; (3.38b)
r 1 Mb rac :::; (3.38c)
T 1 Mc Tac :::; (3.38d)
with a, b, and c being three scalars to be determined.

We now consider the above (3.37a, b) local acoustic limiting process, when
again, S, g, Re, and Pr are all O(1).
Various detailed investigations by Zeytounian and Guiraud (see, for instance,
[85] and [86]) show that the above asymptotic local expansions (3.38ad) are
consistent only with the situation corresponding to an unbounded fluid flow, outside
a solid bounded body O, starting in motion impulsively mimicking a catapulting
process.
A poor justification for working this way is that this is a classical problem in
inviscid incompressible fluid dynamics (in fact, this transient behaviour is essen-
tially characterized by the weak compressibility) in the NSF equations, when we
consider the local (in time) limiting process (3.37a, b) with (3.38ad).
First, from (3.4a), at leading order we derive a least-degenerated (concerning the
least/significant degeneracy concept; see Sect. 6.4.3) continuity equation when
a 1 and b 1:
S @rac =@ t r:uac 0: (3.39a)
It is then easy to show that the more consistent limiting system of equations,
which is derived from the full dimensionless NSF equations (3.4ac) with (3.37a,
b) and (3.38ad), is the linear acoustics system of equations, if we assume that
a c 1. In such a case, with (3.39a) we obtain the following three equations:
S @uac =@t 1= grpac 0; (3.39b)

S @Tac =@t 1 gr:uac 0; (3.39c)
pac rac Tac : (3.39d)
If we consider the transient behaviour of an NSF fluid flow which is set into
motion from rest by the displacement of a solid body, an unbounded medium, as an
initial condition for above acoustics equations we write:
t 0 : uac 0; Tac 0; rac 0: (3.40)
For the determination of the unknown Navier initial condition (uN) it is

necessary to consider an unsteady adjustment problem (as described in Sect.
2.4.2) with a matching process, such that:
limt#1 uac limt0 uN uN : (3.41)
Here I note that in the book by Wilcox (1975) [87] the reader can find a
scattering theory which makes possible the analysis of the behaviour of the above
equations of acoustics (3.39ad), with (3.40), when t # 1, and in Zeytounian
(2000) [88] this matching has been realized.
Unfortunately, this is not the case when the gas is contained in a bounded
container O, with an impermeable but eventually deformable (with time t) wall,
so that the volume occupied by the gas is a given function of time, V0(t) a problem
first considered by Guiraud and Zeytounian in (1980) [88].
For instance, in a bounded fluid flow, inside a solid bounded body O; with a wall
St deformable with time t, the above matching relation (3.41) often does not work
(see Sect. 6.4.2, concerning matching). In Sect. 7.2, devoted to applied aerody-
namics problems, we return to the consideration of a such case. In particular, this
case may be related to compressible gas flow in the compression phase of an
internal combustion engine.
3.3.3 OberbeckBoussinesq Model Equations for the

RayleighBenard Shallow Thermal Convection
When we consider a weakly expansible (with a low, e, expansibility parameter

(3.13)) liquid, such that the following simplified equation of state is valid (at least,
in leading-order, when the expansibility parameter, e, defined by (3.13), tends to
zero):
r = r T; (3.42a)
then, in place of the NSF system of equations (2.34ac) with (2.35a, b), we have in
equation (2.34c) as Cv the coefficient (as a consequence of (3.42a)):
CT DEl T=Dt: (3.42b)
because the specific internal energy E El(T) is a function of temperature T only:
DEl =Dt CT DT=Dt: (3.42c)
For the derivation of a leading-order (relative to e 1) OberbeckBoussinesq

model as a simplified consistent RayleighBenard (RB) shallow thermal convec-
tion model problem we can assume (due to a low squared Froude number (see
(3.10)) hypothesis) that
rT rd 1 eY; (3.42d)
is consistent, with an error of Oe2 , when Y is given by relation (3.43a) and a

perturbation of pressure is introduced by its companion relation (3.43b).
Therefore, if we assume that Td is the constant temperature of the free surface in
the purely static motionless conduction basic state (which is the plane z d), such
that DT Tw Td ; in (3.13) and (3.43a), we can introduce the following
dimensionless temperature (Tw is the temperature at z 0):
Y T Td =DT ; (3.43a)
The companion dimensionless perturbation of pressure is:
p 1=Frd 2 fp pA =grd d z 1g; (3.43b)
where pA is the passive constant atmospheric pressure.

Thanks to the introduction of p by (3.43b), and when the limiting RB process
(3.45) is applied to the NSF system of equations for a weakly expansible liquid, we
can easily verify that relation (3.42a) is sufficient (and consistent according to
(3.42d)) for the density r, for the derivation of the dominant, leading-order,
Eqs. 3.44ac.
Indeed, if now, as the characteristic length scale we choose d , as characteristic
constant velocity nd /d , the characteristic time being d2 =nd , where nd is the
constant kinematic viscosity (at z d ), then we can derive, first, the following
dimensionless dominant equations (from an NSF system of equations written for
an expansible liquid) for our weakly expansible liquid:
r:uliq eDY=Dt; (3.44a)
1 e YDuliq =Dt rp GrY k r2 uliq Oe; (3.44b)

1 e Y 1 G DY=Dt 1=Prr2 Y Oe; (3.44c)
where
CT Cd 1 eG Y:
When (limiting O-B case)
LimOB e ! 0; with Gr O1; Pr O1 and G O1; (3.45)
is performed, from the dominant Eqs. 3.44ac, we obtain the OberbeckBoussinesq

model equations for the RayleighBenard shallow thermal convection in the fol-
lowing form:
r:uOB 0; (3.46a)
DuOB =Dt r pOB GrYOB k r2 uOB ; (3.46b)
DYOB =Dt 1=Prr2 YOB : (3.46c)
In Chap. 8 the reader can find some complementary consequences of the above
limiting O-B (3.45) case in the framework of the well-known Benard heated-from-
below thermal convection problem, for a weakly expansible liquid layer with a free
surface and temperature-dependent tension.
3.3.4 StokesOseen Model Equations in the Case

of a Low Reynolds Number
In the case of an incompressible viscous Navier fluid flow with S 1, from (3.35b)
we obtain:
Re DuN =Dt rpSt r2 uN ; (3.47a)
with
pst Re p; (3.47b)
the Stokes pressure, and for low Reynolds number, when the Stokes limiting
process is performed:
LimSt Re # 0 with pst and t; x fixed; (3.48)

we obtain, for LimStuN uSt(t, x) and pSt, the so-called Stokes equation:
r2 uSt rpSt : (3.49a)
with, from (3.35a),
r:uSt 0: (3.49b)
One would expect that the boundary conditions for the steady Stokes equations
(3.49a, b) would be same as those for the full Navier equations (3.35a, b); but it was
noted by Stokes (1851) [89] that solutions do not exist for stationary 2D flow past a
solid which satisfies both conditions:
at the solid wall; uSt 0 no slip; (3.49c)
as well as
at infinity; uSt ! U1 iuniform flow; (3.49d)
Indeed, the steady incompressible Stokes flow, governed by the above problem
(3.49ad), in an unbounded domain, exterior to a solid body, is only a local-
proximal flow valid mainly in the vicinity of the wall of this body (the Stokes paradox).
Far from the wall, near infinity, when Rejxj jx0s j O1, it is necessary to
derive another consistent local-distal model equation [90] for the Oseen (1910)
flow, in place of the above problem (3.49ad) for the Stokes flow.
Using Oseen timespace variables:
t0s Re t and x0s ; (3.50a)
and also the Oseen pressure,
p0s pSt =Re; (3.50b)
we derive, first, in place of the unsteady Navier equation (3.35b), the following
dimensionless Navier equation (when S 1) for u0 N (t0s, x0s):
Du0 N =Dt0s r0s p0s r0s 2 u0 N : (3.51)
But, near infinity, in the Oseen limiting process,
LimOs Re # 0 with p0s ; t0s and x0s fixed; (3.52)
a finite body shrinks to a point, which cannot cause a finite disturbance in the
viscous fluid flow.
Thus, in the outer, Oseen region, we can assume that for u0 N the asymptotic
expansion
u0 N i dReuOs t0s ; x0s :::; (3.53a)
is significant, when dRe 1, assuming that (with dimensionless quantities),
u0 N ! i at infinity: (3.53b)
With (3.52) and (3.53a, b), for uOs(t0s, x0s), we derive, from (3.51), the following
unsteady Oseen equation:
@=@t0s r0s :iuOs r0s p0s r0s 2 u0s : (3.54)
In the above Oseen leading-order equation for u0s, the gauge function,
dRe ! 0, with Re ! 0, is arbitrary, and only via a matching
local proximal , local distal
do we have the possibility of determining this gauge. (See Lagerstrom (1964) [91],
pp.1637, for a discussion of this matching process.)
We observe also that for a compressible fluid flow, when Re ! 0, it is also
necessary to specify the role of the Mach number, M. In fact, we must address the
problem of the behaviour of full NSF system of equations when simultaneously,
Re ! 0 and M ! 0; with t and x fixed: (3.55a)
Obviously, it must no forgotten that for the validity of these full NSF equations
(3.4ac) with (3.1), for a thermally perfect gas, under the above limiting low
Reynolds and Mach numbers (3.55a), it is necessary that the corresponding limit
fluid flow remains a low Knudsen, Kn, number flow (briefly discussed at the
beginning of Chap. 2).
As a consequence of the relation Kn M/Re, the above double limiting process
(3.55a) must be made with the following similarity relation:
M s Re1a ; s O1 fixed; a > 0; when Re ! 0: (3.55b)
But, with (3.55a, b) it is also necessary to take into account the following
asymptotic expansions in the framework of the NSF equations (3.4ac):
u uS :::; (3.55c)
p 1 Re12a pS :::; (3.55d)

T TS :::; r rS :::; (3.55e)
where the Stokes limit functions (with S as subscript) depend on x (fixed) only,
because the above expansions are singular near the time t 0. In fact, the time
t (also fixed in (3.55a)), in the compressible Stokes equations (3.56ac), plays the
role of a parameter!
These Stokes compressible (inner) equations are written in two parts:
r kS TS rTS 0; rS 1=TS ; (3.56a)
and
r uS uS rlogTS ; (3.56b)
rpS gso2 r flTS r:uS I 2mTS DuS g: (3.56c)
The first equation of (3.56a) with the associated boundary condition (see, for
instance, (3.22)) on the boundary G @S, determines TS, as soon as the tempera-
ture behaviour at infinity, or matching with the outer Oseen compressible equations,
is specified (see, for instance, in Lagerstrom (1964) [91], pp. 1912 and 2025) for
an unbounded fluid flow outside a solid bounded (by G) body S. The second
equation of (3.56a) is a relation (limit form of the equation of state (3.1)) between
TS and rS, and determines rS when TS is known.
Finally, the system of the two Eqs. 3.56b, c gives a closed system for the
determination of the velocity vector uS and the perturbation of pressure pS, when
the boundary condition (for uS) on the wall @S are used (for instance, (3.21)).
The scalar a > 0 in (3.55b) and in the second expansion for the pressure in
(3.55c) is determined only after the matching with the companion compressible
Oseen (outer and valid far at t 0) equations. (Concerning the singular (for
Re ! 0) region near t 0, see Sect. 9.3 of our (2002) [26].)
On the other hand, when the rate of temperature fluctuation (the scalar w in
(3.22)) on the wall @S tends to zero with Re ! 0, then we obtain a particular simple
solution for Eq. 3.56a: rS 1 and TS 1. In such a case, Eqs. 3.56b,c reduce to the
classical steady Stokes equation (3.49a) with (3.49b) for an incompressible low
Reynolds fluid flow if we assume that:
gso2 m1 1:
Curiously, in his first tentative attempts to derive the leading-order equations for
a compressible low Reynolds laminar fluid flow, Lagerstrom (in [91], pp. 1902),
does not take into account our above constraint (3.55b). He assumes that in the
NSF equation (3.4c) for the temperature, the ratio M2 to Re is unspecified, and
obtains a system of equations (where M2 is present). He also observes that this
derived (for Re ! 0) system is uniformly valid for any value of M, and considers
the case where
M2 tend to zero faster than Re; and M2 is of the same order as Re:
There is also the case in which
Re tends to zero faster than M2 ;
which (Lagerstrom [91], p. 192) presents some special difficulties and will not be
considered here!
We observe, also, that the steady Stokes equation (3.49a) for an incompressible fluid
flow may be obtained either by linearization or by letting Re tend to zero. But it is
fortuitous that these two procedures produce the same result in the incompressible case!
For compressible fluids the above, low Reynolds, Stokes equation for TS (the
first of Eq. 3.56a) is a non-linear equation. In Chap. 5 of our (2004) book [47], the
reader can find information concerning the unsteady case (the steady case being
singular near initial time) and the compressible case (influence of the temperature field).
Here, as a final remark concerning the compressible flow at low Reynolds
numbers, it seems me judicious to mention some remarks of Lagerstom ([91], p.
192), related to the validity of the NSF equations:
It may be objected that the NSF equations are no longer valid in the limit
considered above, especially not in the limit where (M2/Re) tends to infinity! It
must certainly be true that the NSF equations cease to be valid for extreme values
of certain parameters. However, a similar criticism would apply, say, to the
classical case of high Reynolds number and zero Mach number! As the Mach
number tends to zero, at constant free stream velocity, the velocity of sound, and
hence the temperature, tends to infinity! At sufficiently high temperatures any real
gas certainly has properties which are not accounted for in the NSF (and Navier?)
equations. The answer to the objection stated above is that letting a parameter tend
to zero, say, is a mathematical device for obtaining approximate models for small
values of the parameter (more generally one should consider complete expansions
of NSF equations for small values of this parameter!). The method [in fact, our
RAM Approach] is physically significant if there are values of the parameter which
are sufficiently large for the NSF equations to be adequate, and at the same time
sufficiently small for the mathematical method used to be approximately valid. As
yet, it has not been investigated carefully whether or not such values of the
parameters exist in the cases considered above.
In order to settle this question it is first necessary to carry out the matching
between the various local models (near the time, t 0, and in the vicinity of the
wall, z 0) with the main approximate model (an example is given in Sect. 9.2, on
the framework of low Kibel asymptotics). But the comparison with the
experiments, and especially with the results of associated (with derived consistent
model) computational simulation (by the numericians), is the more convincing test!
3.3.5 The Case of Non-linear Acoustics
We have already mentioned that in non-linear acoustics, in place of Re (given by

(3.5a)), it is necessary to introduce p
an acoustic Reynolds number Reac, such that
(Re L U =m=r and M U= gRT ):
Reac Re=M r =k m ; (3.57)

p
the characteristic length scale being L 1/k , with k oac = gRT the wave
number, oac a reference frequency determined by the main frequency oac =2 p of
the source-signal, and U as source velocity.
In fact, the acoustic Reynolds number Reac given by (3.57) compares the orders of
magnitude of the propagation and viscous dissipative terms in non-linear equations (a
la NSF) see Eqs. 3.60ad below and Reac 1 (see, for instance, Appendix 1 in
Coulouvrat (1992) [92]). On the other hand, in problems of non-linear acoustics it is
assumed that S 1 and necessarily M 1 such that (see [13]):
S M 1: (3.58a)
The reference acoustic time is then just:

p
tac L = gRT 1=oac : (3.58b)
When in place of thermodynamic functions p, r, T, we introduce the

corresponding perturbations p; o; y, such that:
p 1 Mp; T 1 My; r 1 Mo; (3.59)
and assume that the dissipative coefficients are constants, then with a constant bulk
viscosity mv and the help of the relations (3.57) and (3.58a) we can write the
following 3D system of unsteady dimensionless (a la NSF-dominant) acoustic
equations, for the acoustic velocity vector u and thermodynamic perturbations p; y,
and o:
@o=@t r:uac Mr:ouac ; (3.60a)
@uac =@t 1=grp 1=Reac fr2 uac 1=3 mv =m rr:uac g
Mo@uac =@t uac :ruac OM2 ;
(3.60b)
@ y=@t g 1r:uac g=Reac Prr2 y Mfgg 1=Reac 1=2Duac :Duac

1=3 mv = m r:uac 2
o@ y=@t uac :r y g 1 pr:uac g OM2 ; (3.60c)
with
p o + y Myo: (3.20d)
This above system of basic dimensionless equations (3.60ad) is the main

starting point for the derivation of various consistent and simplified (non-ad hoc)
non-linear models of acoustics via the RAM Approach.
Detailed information concerning such acoustic models is included in our (2006)
book [13].
In particular, the well-known KZK (Kuznetsov, Zabolotskaya, and Khokhlov)
equation (pp. 2316) was derived from the above system of Eqs. 3.60ad consistent
with a RAM Approach, when it is assumed that the 3D acoustic field is locally
plane, such that the non-linear wave propagates in the same way as a linear plane
wave over a few wavelengths the wave profile or amplitude being significantly
altered only at large distances away from the source (in the far field).
As a consequence, obviously, the so-called parabolic approximation, which
leads to the KZK model equation, may not be valid close to the source (in the
near field).
The condition for such an approximation to be valid is that the width of the
acoustic source, d, should be much larger than the wavelength (1/k), so that the ratio
a 1=k=d 1 and tends to zero as M # 0: (3.61a)
the transverse field variations being slow compared with longitudinal variations
along the acoustics axis.
A main hypothesis is also:
M2 1=Reac M (3.61b)
and first we introduce two transverse slow coordinates:
ay and z az; (3.61c)
and for u ac we write:
uac u i aV j W k: (3.61d)
For low Mach numbers, we consider the following asymptotic expansion:
U u; V; W; p, y, o U0 M U1 :::; (3.61e)
with M # 0 fixed and the following two similarity relations between the three small
parameters, 1/Reac, M and a:
1=Reac kM; a2 =M b; k and b are O1 and fixed: (3.61f)
First from the leading-order system for U0 (u0, V0, W0, p0, y0, o0) we derive,
first, a simple linear acoustic system for u0 and p0, y0, o0, and then for V0 and W0
the following two relations, which are valid in far-field, related with the function p0:
@V0 =@t 1=g@ p0 =@ and @W0 =@t 1=g@p0 =@ z; (3.61g)
or
@=@t @V0 =@z @W0 =@ 0; (3.61h)
which is a consistent consequence of our asymptotic approach.

Then from a linear acoustic system we deduce as a solution for u0:
u0 Ft, , z; with t t x: (3.61i)
assuming that the fluid is unbounded and considering a solution for an outgoing
wave propagating towards x > 0.
However, the acoustic solution (3.61i) is in fact a good solution only for the
near-field close to acoustic source!
On the other hand, the KZK equation is a far-field equation, and to avoid various
cumulative effects it is necessary to consider the asymptotic expansion (3.61e) as a
non-secular two-scale expansion relative to variations along the acoustic axis and to
define the slow scale as:
x Mx; with @=@x @=@t M @=@x (3.61j)
and
@ 2 =@x2 @ 2 =@t2 0M: (3.61k)

In a such case, for u0 we write
u0 A t, x, , z; and p0 gu0; o0 u0; p0 g 1u0 ; (3.61l)
and as @=@t @=@t, from (3.61g) we obtain (an unexpected relation?)
@=@t@V0 =@ @W0 =@z r? 2 A; (3.61m)
where
r? 2 @ 2 =@2 @ 2 =@ z2 :
Now, from the linear system of equation for U1 in (3.61e), taking into account
(3.61jm), we obtain for u1 a single inhomogeneous acoustic equation in the
following rather awkward form:
@ 2 u1 =@t2 @ 2 u1 =@x2 @YA=@ t b=gr? 2 A; (3.61n)
with
YA G @ 2 A=@ t2 g 1A@A=@t 2@A=@ x; (3.61o)
where
G k=24=3 mv =m g 1=Pr: (3.61p)
Finally, according to the usual multiple-scale method (see, for instance,

Kevorkian and Cole [93]), the non-secularity validity condition for the asymp-
totic expansion (3.61e) leads to the following KZK equation for the amplitude
function At, x, , z:
@ 2 A=@t@x b=2gr? 2 A 1=2g 1@=@tA@A=@t

(3.61q)
G @ 3 A=@t3 :
This above KZK equation is a very representative leading-order equation in non-

linear acoutics. The above RAM Approach is a good example of an application of
the multiple scales method carried out in a consistent way!
3.3.6 A Sketch of Consistent Models Derived from NSF

Equations Relative to the Considered Reference Parameters
Re M O

Euler: t and x fixed Navier: t, x fixed
matching, slip condition p1
for uN and p =
for Eulerian flow gM2

Prandtl: t, x, y and z= z1/2 fixed matching, initial condition for uN at t = 0
wE Re
and wPr = 1/2
Re Acoustics : t = t , x fixed
M
Re O Re O
(incompressible case) (compressible case)

Stokes: t, x fixed from Navier Stokes: t, x fixed from NS-F
with pst = Re p with M = o Re1+a, a > 0
matching singularity at infinity P = 1+ Re1+2a [Ps+...]

matching
Oseen: tOs = Re and xOs = Re x fixed
pst
with pos = Re
Oseen: compressible
O Re
(expansibility parameter) Reac = >>1
M
R-B shallow thermal convection: Non-linear acoustics:

e SM = 1; P = 1 + Mp
with: Gr = fixed, Fr 2 O
Fr 2 d T = 1 + Mq, r = 1 + Mw
d
1/3 KZK model eq.
n
d>> d /2
g
Chapter 4
A Typical RAM Approach: Boussinesq Model
Equations
The following derivation-justification of Boussinesq equations via the RAMA for

atmospheric flows is a typical test example, in the sense that the various steps of the
derivation constitute a logical, well argued, non-contradictory train of thought,
which ensures the rational justification of these Boussinesq approximate equations,
even if we do not use any rigorous mathematically abstract tools!
An amazing success of our RAMA is the real possibility to derive, also, a
second-order consistent, well-balanced, model equation in the framework of the
well-known assertion of Boussinesq (1903), which take into account the various
non-Boussinesq effects. This, it seems to me, is not really possible if the RAMA
below is not applied.
In Monins book (1969) [94], the reader can find a very pertinent exposition
concerning scales of weather processes. The discussion below does not take into
account the influence of the Coriolis force, because the sphericity parameter defined
by (3.16d) is assumed very small. On the other hand, the parameter es, according to
(3.18b), is assumed equal to 1.
The ad hoc approach to justification of simplified Boussinesq equations was
developed throughout 19601980. Here we mention the papers by: Spiegel and
Veronis, Mihaljan, Malkus, Dutton and Fichtl, Perez Gordon and Velarde, and Mahrt,
and for precise references see our books [12, 13, 19], and also the survey [25].
4.1 Introductory Remarks Concerning Atmospheric Flow
Below, only the case of non-dissipative and adiabatic atmospheric flows, for
simplicity is considered, when the starting exact equations are the Euler
equations. See Chaps. 1 and 2 of our (1991) book [13] for a detailed derivation
directly from NSF equations. Namely, with the system of Euler equations (1.2ac)
with (1.3) it is necessary, above all, to assume the existence, also, of a hydrostatic
reference state.

82 4 A Typical RAM Approach: Boussinesq Model Equations
If p*(z*), r*(z*), T*(z*), the three thermodynamic functions are dependent only
on z*, the altitude of the motionless reference state, then we can work with the
following three basic equations (for a thermally perfect gas):
dT =dz Gz ; (4.1a)
dp =dz g r 0; (4.1b)
p R r T ; (4.1c)
where G(z*) is a given function of z*, in the considered adiabatic case. In the
troposphere, a good approximation is obtained when
Gz G const.
In order to make the governing exact Euler 3D unsteady, compressible, non-

viscous and adiabatic equations non-dimensional which is our first key step we
introduce the following characteristic constant quantities:
H ; U ; p 0; r 0; T 0 p 0=R r 0;
where the length scale H has been chosen to be representative of the vertical
motion of the lee-wave regime downstream of the mountain, since it is the effect of
gravity with which we are primarily concerned.
The characteristic velocity U is an average constant value of the velocity profile
which is assumed smaller than the speed of sound, defined by
c0 g p 0=r 01=2 ; (4.2a)
such that
Uo <<c0 :
Now, if we define the associated dimensionless quantities u0 , r0 , p0 , T0 , x0 , and t0 ,

respectively, linked with the above characteristic constant quantities U , r*(0),
p*(0), T*(0), H , and H /U , we rewrite the set of Euler equations, for atmosphere
in movement, relative to u0 , r0 , p0 , and T0 , as functions of t0 and x0 , in the following
(exact) dimensionless form (slightly different from (3.24ad)):
Du0 =Dt0 1=gM2 1=r0 r0 p0 1=FrH 2 k 0; (4.3a)
Dr0 =Dt0 r0 r0 :u0 0; (4.3b)
DT 0 =Dt0 g 1= g1=r0 Dp0 =Dt0 0; (4.3c)
p0 r0 T 0 : (4.3d)
4.1 Introductory Remarks Concerning Atmospheric Flow 83
In (4.3ad), r0 is now the gradient operator with respect to x0 , and the material
derivative (along the trajectories) is:
D=Dt0 @=@t0 u0 :r0 :
As main parameters, in Eqs. 4.3ad, we have for the Froude number the
following remarkable and very instructive relation:
1=FrH 2 Bo=gM2 ; (4.4)
where
M2 U =c0 2 (4.5)
is the square of the Mach number and
Bo H =H (4.6)
our Boussinesq number, where
H p 0=gr 0 HS : (4.7)
The length scale H* is (see (3.18a)) the altitude of the homogeneous hydrostatic
reference state a judicious characteristic vertical length scale for the altitude z*,
such that in dimensionless form we write the relation (Bo is defined by (3.5d)):
z0 Boz0; (4.8)
which plays an important role in the derivation of Boussinesq approximate

equations, and gives a deeper and more subtle meaning to the Boussinesq fluid
flow case!
Concerning the characteristic horizontal length scale L , we assume (see (2.18b)
and (2.16d)) that:
L HS ) eS 1 and as a consequence d<<1; (4.9)
which is a condition for a negligible influence of the Coriolis force, and also for a
derivation of Boussinesq model equations, without the hydrostatic approximation,
for the lee-waves regime flows downstream from the mountain.
As a consequence, the dimensionless reference hydrostatic state (p*0 , r*0 , T*0 )
is, in fact, only a function of Boz0 , and
@=@z0 1=Bo@=@z0 : (4.10)

Naturally, when for H we choose H*, then Bo 1 and z*0 z0 .

The above introduction and discussion of various dimensionless quantities and
reduced parameters form, in fact, our first key step towards the justification of
Boussinesq model equations in the framework of a Eulerian fluid flow.
4.2 Asymptotics of the Boussinesq Case
In the case of the Boussinesq approximation, if we want to obtain rationally the

corresponding Boussinesq system of simplified equations, a detailed (see below)
dimensional analysis shows that when M < < 1, then necessarily that must be also
Bo < < 1.
Below, our asymptotic rational approach confirms this observation. The emer-
gence of Bo (thanks to relation (4.8)) was our second (significant) key step, during
our 19691974 investigations relative to justification of this Boussinesq approxi-
mation, in the framework of our RAMA theory for a consistent derivation of the
leading-order model equations, a la Boussinesq.
Indeed, with two small parameters, M and Bo, it is necessary to consider, in
general, three limiting cases:
I Bo fixed; M # 0 and then Bo # 0 (4.11a)
II M fixed; Bo # 0 and then M # 0 (4.11b)
III M # 0 and Bo # 0; such that Bo Ma B (4.11c)
where B* O(1) and a > 0 is a scalar.

The Boussinesq equations are derived (see below) with the particular significant
choice a 1, under the Boussinesq limiting process, corresponding to case
(III)(4.11c), with B*, g, t0 , and x0 fixed during the Boussinesq limiting process
(see (4.18) below).
We observe that the relation Bo MaB* in (4.11c), when a 1, is the
Boussinesq similarity rule, and B* is the associated similarity parameter.
From (4.11c) we deduce an important (but unfortunate) feature derived below
(see (4.17ad)) for Boussinesq model equations. Namely, when a 1, then:
B 1 ) H Uo =gRT 0=g1=2 HB << H ; (4.12a)
and more precisely, for the usual meteorological values of Uo and T*(0), we obtain:
HB 103 m; while H 104 m! (4.12b)

4.2 Asymptotics of the Boussinesq Case 85
This is a very strong restriction for the various applications of the Boussinesq
equations in atmospheric motions.
For the prediction of lee waves around and downstream of a mountain in the
whole thickness of the troposphere (in troposphere), another set of model equations
are necessary the so-called deep (Zeytounian) convection equations and see
Chap. 10 in [12] and also the Sect. 6.2.3 of [13]. Here, in Sect. 9.1.4 we also
consider four distinguished limiting cases for the 2D steady lee-waves problem.
4.2.1 Euler Dimensionless Equations for the Thermodynamic

Perturbations
Now a third key step is necessary, which is linked with the existence of the
dimensionless hydrostatic reference basic state p*0 , r*0 , T*0 , function of Boz0 , and
for this it is necessary to introduce the thermodynamic perturbations, p, o, y, such
that:
p0 p0 Boz01 p; (4.13a)
r0 r0 Boz01 o; (4.13b)
T0 T0 Boz01 y: (4.13c)
With (4.13ac) the dimensionless Euler equations are rewritten in the following
new form, for the velocity vector u0 and thermodynamic perturbations p, o, and y:
D o=Dt0 1 or0 :u0

(4.14a)
1 oBo=T z0 f1 dT z0 =d z0 gu0 :k 0;
1 oDu0 =Dt0 T z0 =gM2 r0 p

(4.14b)
1 oBo=gM2 yk 0;
D y=Dt0 g 1=gDp=Dt0 1 pBoN 2 z0 u0 :k 0; (4.14c)
p o 1 oy; (4.14d)
where
N 2 z0 1=T z0 fg 1=g dT z0 =d z0 g; (4.15)
when (4.10) has been used. The relation (4.15) for N2(z*0 ) takes into account the
stratification of the hydrostatic reference-motionless state atmosphere (in the
Eulerian case (4.1ac)), and is related to the dimensionless Vaisala internal

frequency.
The consideration of the above full system of dimensionless exact Euler
equations, (4.14ad), for u0 and p, o, and y, was indeed our main third key step
in the quest for the justification of the Boussinesq model equations.
Obviously, for the atmospheric motions, p, o, and y are small, by comparison
with unity perturbations. As a consequence, we write in the framework of the
fourth key step the following asymptotic expansions:
p Mb pB Mc p :::; (4.16a)
o; y Md oB ; yB Md1 o ; y :::; (4.16b)
u0 uB Me u :::; (4.16c)
associated with the limiting process (4.11c). In (4.16ac), the exponents, b, c, d, and
e are real scalars which are determined consistently in Sect. 4.2.2.
4.2.2 Rational Derivation of Boussinesq Equations
The Boussinesq equations, as leading-order approximate equations, are derived

from the unbalanced system of Eqs. 4.17ad, which are obtained when the above
asymptotic expansions (4.16ac) are used in exact Euler dimensionless equations
(4.14ad).
First, we can write (with (4.8)), from Eq. 4.14c, the following unbalanced
equation for yB, when we take into account only the terms with the subscript B in
expansions (4.16ac):
DyB =Dt0 Md oB DyB =Dt0 Mbd g 1=gDpB =Dt0

Bo=Md N 2 Boz0 uB :k (4.17a)
0
Bo M bd
pB N Boz uB :k ::: 0;
2
and it is necessary to take into account also the similarity rule for Bo ( MaB*)
written in (4.11c), when the Boussinesq limiting process is performed:
M # 0 with t and x fixed; B O1

(4.18)
and N 2 0 O1 fixed:
Because g > 1, it is necessary, first, with (4.17a), that
a d; and it seems that with : d 1; a 1 and b 2?; (4.19a)

we derive not only a leading-order equation for yB:
DB yB =Dt0 B N 2 0uB :k 0: (4.20a)
but also we open the way for the derivation of a second-order non-homogeneous
equation for y*.
For a confirmation of the above (see (4.19a)) rather premature conclusion, it is
now necessary to consider the derivation of a leading-order equation for uB,
associated with (4.17a). From (4.14b) with (4.16ac), we obtain as a second
unbalanced equation for uB:
DuB =Dt0 Mb2 T0 Boz0=gr0 pB 1=g Bo Md2 yB k

(4.17b)
Md oB DuB =Dt0 1=gBoM2d1 oB yB k:
First the above (4.19a) choice of:
b 2 is confirmed and then also d 1; a 1! (4.19b)
Thus, we obtain, as a companion simplified (a la Boussinesq) leading-order

equation for uB, associated to equation for yB:
DB uB =Dt0 T0 0=gr0 pB B=g yB k 0 (4.20b)
The two terms in the right-hand side of (4.17b) are linked with the associated,
second-order equation for u*.
Now, from the unbalanced equation of continuity, obtained from (4.14a), with
(4.16ac),
r0 :uB Md DoB =Dt0 oB r0 :uB
(4.17c)
Bo 1=T0 Boz01 GBoz0uB :k :::;
and obviously our choice of (4.19a) simply produces, at the leading-order, the
following divergenceless constraint for the velocity vector uB:
r0 :uB 0: (4.20c)
Again, in the right-hand side of (4.17c), the two terms are linked with the
associated, second-order equation of continuity.
Finally, from (4.14d) we obtain as unbalanced equation:
oB yB Me oB yB Mbd pB ; (4.17d)
and as Boussinesq equation of state we derive:
o B yB : (4.20d)
which determines the perturbation of the density.

In Boussinesq equations (4.20a) and (4.20b) we have:
DB =Dt0 @=@t0 uB :r0
and we observe that, obviously, T*0 (0) 1, but in general N2(0) is different from
zero. The choice e 1, in (4.16c), indeed, follows when we want to derive the
second-order a la Boussinesq limit model equations for u*, o*, p* and y*, and we
have also d 1.
I invite the reader to derive these second-order Boussinesq equations as a
profitable and stimulating exercise! In particular, in the 2D steady case, starting
from the exact 2D, (x, z) steady @=@t 0 case, it is possible to derive, without
any approximations, an analogue to Eq. 4.26b but significantly complicated a
single equation (see Sect. 9.1).
So, with the following asymptotic expansions:
p M2 pB :::; o; y MoB ; yB :::; u0 uB :::
when the following Boussinesq limiting process is performed in an Euler system

of equations (written for the atmospheric flows):
M # 0 with t and x fixed;
B O1 fixed
and
N 2 0 O1 fixed;
then, at leading-order, we derive the following Boussinesq equations:
DB yB =Dt0 B N 2 0uB :k 0;
DB uB =Dt0 T0 0=gr0 pB B =g yB k 0:
r0 :uB 0;
oB yB :
with
N 2 0 g 1=g dT z0 =dz0 z 00
and
B Bo=M
This rather long, but instructive, derivation of the Boussinesq equations

(4.20ac) is the only RAMA way, in the framework of low Mach (hyposonic,
[13]) atmospheric flows, for a consistent rational derivation of not only the
Boussinesq system of equations, but also the second-order consistent equations
associated with these Boussinesq equations.
4.2.3 Two Particular Cases
It is instructive also to note two particular, less significant limit, cases of the above
Boussinesq system of equations. On the one hand, if
B "/, M << Bo; Bo fixed;
or
M # 0; and then Bo # 0 case I; (4.21a)
then we obtain the quasi-non-divergent limit equations for the velocity

components, vd (ud, vd), wd, and pd/g:
@pd =@z0 0; wd 0; yd 0; (4.22a)
D:vd 0; @vd =@t0 vd :Dvd Dpd =g 0: (4.22b)
where D @=@x0 ; @=@y0 .

If on the other hand
B # 0 , M >>Bo; M fixed;
or
Bo # 0; and then M # 0 case II; (4.21b)
then we obtain the classical incompressible Navier limit equations for ui, vi, wi,
and pi/g:
@ui =@x0 @vi =@y0 @wi =@z0 0: (4.23a)
Di ui =Dt0 r0 pi =g 0: (4.23b)
with
Di yi =Dt0 0; (4.23c)
o i yi : (4.23d)
where
Di =Dt0 @=@t0 ui :r0 :
In both cases we derive from the above dimensionless dominant full Euler
equations a less significant limit system of equations than the inviscid, adiabatic
Boussinesq equations.
4.3 The Steady 2D Case
In the steady 2D case, when the Boussinesq fluid flow is considered in the plan
(x0 , z0 ) with, as velocity components, uB and wB, we have the possibility of
introducing a 2D stream function cB(,x0 , z0 ) such that:
uB @cB =@z0 and wB @ cB =@x0 ; (4.24)
and, because the coefficient B N 2 0 constant, in Eq. 4.20a, we derive a first

integral:
yB B N 2 0z0 FB cB : (4.25a)
0
*
On the other hand, assuming T(0) 1 if in the two equations obtained from
(4.20b):
uB @uB =@x0 wB @uB =@x0 @ pB =g=@x0 ;
uB @wB =@x0 wB @wB =@x0 @pB =g=@z0 B=g yB 0;
we cancel out the terms with pB =g, then we derive the following equation for cB:
@ cB =@x0 @D2 cB =@z0 @ cB =@z0 @D2 cB =@x0 B =g@yB =@x0 ;
where D2 @ 2 =@x0 2 @ 2 =@z0 2 .

But
@yB =@x0 dFB cB =dcB @cB =@x0 ;
and as a consequence we can write:
@ cB =@z0 @=@x0 D2 cB B=gz0 dFB cB =dcB
@ cB =@x0 @=@z0 D2 cB B =g z0 dFB cB =dcB 0:

4.3 The Steady 2D Case 91
From above, we obtain a second first integral:
D2 cB B=gz0 dFB cB =dcB HB cB : (4.25b)
The functions FB cB and HB cB are two arbitrary functions of cB only.

In the particular case of an airflow over a mountain, if we consider a 2D steady,
uniform, constant flow in the direction of x0 > 0, with pB and yB both 0 at
infinity upstream x " / of the mountain, then, in place of (4.25b), we obtain for
the function
dB x0 ; z0 z0 cB x0 ; z0 (4.26a)
the following linear Helmholtz equation:
@ 2 dB =@x2 @ 2 dB =@z0 B2 =gN 2 0 dB 0:

2
(4.26b)
The dominant feature of Eq. 4.26b, for dB , from a mathematical point of view, is
that its linearity is not related to any particular hypothesis about the small
perturbations in the steady 2D Boussinesq equations.
However, from the exact slip-condition on the surface of the mountain,
simulated by the dimensionless equation,
z0 mh0 lx0 :
where m is an amplitude parameter and l>0, we must write the following bound-
ary, slip, non-linear, condition for dB on the surface of the mountain:
dB x0 ; z0 mh0 lx0 mh0 lx0 : (4.26c)
At infinity upstream, we have also a behaviour condition:
dB x0 " /; z0 0; (4.26d)
But, because of the emergence of steady lee-waves at infinity downstream of the

mountain, there is only the possibility of assuming one physically realistic bound-
ary condition, namely:
j dB x0 /; z0 j< /: (4.26e)
If we add, in particular, to the linear Helmholtz equation (4.26b), with slip

condition (4.26c), the following upper condition:
d B x0 ; z 0 1 0 (4.26f)
then we obtain the well-known Long (1953) problem [7].

However, in reality, the above, (4.26f), condition does not emerge consistently from
the exact dimensionless formulation of the lee-waves problem, considered in the
framework of the 2D steady Euler equations for the whole troposphere, and where the
upper boundary condition is assumed to be slip condition on the tropopause (simulated
simply, before the non-dimensionalization, by the equation z Hs 104 m).
In reality, in dimensionless form, for the above 2D steady Boussinesq equation
(4.26b), the consistent upper condition is:
dB x0 ; z0 " / / ! (4.27a)
The Boussinesq equation being a consistent equation only in a layer of the order
of 1 km in altitude, as a consequence of (4.27a) it is necessary to solve the lee-waves
problem in an unbounded atmosphere with an a la Sommerfeld radiation condition,
which expresses the condition that
no waves are radiated inwards. (4.27b)
This inner Boussinesq problem, in an unbounded atmosphere, has been consid-

ered by Miles (1969) [95], and also by Kozhevnikov (1963) [96].
In a paper by Guiraud and Zeytounian (1979) [97], the associated outer
Boussinesq problem is considered, with an upper slip condition on the rigid plane:
z 1=B (4.28)
with an outer vertical dimensionless coordinate z Mz0 .

In Guiraud and Zeytounians paper (1979) [97] it is shown that the upper and
lower boundaries of the troposphere alternately reflect internal short-wavelength
gravity waves excited by the lee waves of the inner (Boussinesq) approximate
problem, with a wavelength of the order of the Mach number (which is assumed
vanishing), on the scale of the outer region.
As a consequence, there is a double scale built into the solution, and we must
take care of it. In fact, the important point of our (1979) with Guiraud analysis is
that these short-wavelength gravity waves propagate downstream and that no
feedback occurs on the inner Boussinesq flow close to the mountain to the lowest
order at least.
Finally, we would understand the imposed upper boundary at the top of the
troposphere as an artificial one, having asymptotically no effect on the inner
Boussinesq (lee waves) flow, which is the only really interesting flow.
4.4 The Problem of Initial Conditions
In Boussinesq equations, as a consequence of the Boussinesq limiting process

(4.18), in Euler full unsteady Eqs. 4.14ad with three partial derivatives in time t,
we have only two partial derivatives in time t, for uB and yB! As a consequence, for
4.4 The Problem of Initial Conditions 93
a Boussinesq system of three equations, (4.20ac), for uB, pB, and yB we have only
the possibility of assuming the availability of two initial data, namely:
t0 0 : uB uB 0 x and yB yB 0 x; (4.29a)
with
r0 :uB 0 0: (4.29b)
But, on the other hand, for the Euler exact starting Eqs. 4.14ad, three initial data
are necessary for u0 , y, and o, at t0 0.
According to Sect. 2.4.2, we know that it is necessary to formulate an associated
unsteady adjustment problem to Boussinesq approximate equations by the intro-
duction, in Eqs. 4.14ac, of a local time,
t t0 =M; significant near t0 0 (4.30a)
and a local (acoustic) limiting process (in place of (4.18)) with local asymptotic
expansions. Such a local formulation can be found in Sect. 20 of Chap. 5 in my
(1991) book [19].
The initial data (in (4.29a)) at t0 0, for the Boussinesq equations, are derived
from a matching between two asymptotic representations:
the main one outer Boussinesq; relative to time t0 fixed;

at t0 0; and a local one inner acoustic; near t0 0; with (4.30b)
local short time t fixed when t " /:
with the matching:
LimBo t0 0 , LimAc t " /: (4.30c)
In reality (!) the result of our (1991) approach is valid only when we assume the
following initial conditions for the exact starting Euler system of Eqs. 4.14ad:
t0 0 : u u and p; o; y M p ; o ; y ; (4.31a)
with
u rf r ^ c : (4.31b)
In Sect. 20 of my [19] the reader can find a details derivation of the above initial
conditions (4.29a) via matching, at leading-order (4.30b), from the solution of the
following initial value problem of the classical acoustics:
@ 2 f=@ t2 Df 0: (4.32a)
t 0 : f f ; @f=@t p =g: (4.32b)

The solution of the above problem (4.32a, b) is straightforward and, from it we

obtain at infinity in time:
t !/ : jrfj ! 0 and j@f=@ tj ! 0: (4.33)
In such a case, for the derived (in Sect. 4.2) Boussinesq approximate equations
(4.20a, b) we obtain, for the unknown initial conditions at t0 0, yB o x and
uB0(x), the following relations:
uB o x r ^ c and yB o x p = g o : (4.34)
4.5 A Sketch of a RAM Approach for a Boussinesq Model
Fig. 4.1 Justification of Boussinesq equations and associated conditions
What happens if the initial values of p, o, and y the solutions of exact starting
Euler equations (4.14ad) are different from the assumed data M po , M o , and
M y , as is the case in (4.31a), are not known? Actually, this is an open problem
which deserves further investigation.
In Chap. 9, devoted to applications of our RAMA to some atmospheric flow
problems, the reader can find a discussion (in Sect. 9.1) concerning a few compan-
ion equations associated with the Boussinesq/Helmholtz equation (4.26b).
Chapter 1
NSF Equations and Modelling: A French Touch
This Overview is a brief outline of the events related to my rather long RAM
Adventure during the years 19682009. In 19681969 my discovery of asymp-
totics and rational modelling of fluid dynamics problems was, for me, a revelation,
and the Rational Asymptotics Modelling (RAM) Approach to these problems,
governed by the NavierStokesFourier (NSF) equations,1 has been my main
scientific activity during the last 40 years the systematic, logical and well argued
consistent approach via asymptotics, in perfect harmony with my idea about
mathematically applied, but not ad hoc, theoretical researches in fluid dynamics,
without any modern abstract, sophisticated, functional analysis!
This Overview first presents a short account of my first contribution to RAM in
fluid dynamics, related to a justification of Boussinesq equations used in Chap. 1 of
the original, version of my doctoral thesis, written in Moscow during 19651966. I
then relate various events concerning my collaboration with Jean-Pierre Guiraud,
working on asymptotic modelling of fluid flows at the Aerodynamics Department
of ONERA2 during the 16 years up to 1986, which resulted in the publication
of 26 joint papers in various scientific journals. Finally, a few remarks are
presented concerning my preceding seven books (three in French and four in
English), published during the years 19862009, on modelling in Newtonian fluid
flows.
Below we use Navier equations in place of NavierStokes incompressible
equations. In fact, as main fluid dynamics equations we have Euler, Navier, and
NSF equations. Concerning the so-called NavierStokes (isentropic) equations
often used by mathematicians in their rigorous investigations in reality these NS
equations are unable to describe any real fluid flows! Note also that in a RAM
1
Concerning the term NavierStokesFourier equations used in this book NSF equations,
governing classical, Newtonian, viscous, compressible and heat-conducting fluid flows it seems
to me that it is better adapted than the term commonly used (mainly by mathematicians),
NavierStokes compressible equations.
2
Office National dEtudes et de Recherches Aerospatiales, Chatillon-92320 (France).

2 1 NSF Equations and Modelling: A French Touch
Approach, the Euler (vanishing viscosity case) and Navier (low compressibility
case) equations are, in fact, derived consistently by limiting processes from NSF
full equations but this is not the case for the NS (isentropic) equations!
Concerning my Soviet Adventure of 19471966. . . In 1954 I graduated from
Yerevan State University with a Master of Sciences degree in pure mathematics (in
the class of Sergey Mergelyan3); after which, during 19551956, I worked in the
Institute of Water and Energy at the Armenian Academy of Sciences in Yerevan.
I then had the opportunity for serious study in theoretical fluid dynamics, and
in 1957 I chose dynamic meteorology as my main scientific research activity as
a Ph.D. student in the Kibel Department of the Hydro-Meteorological Centre in
Moscow.
Now, more than 50 years later, I am still proud to have been a student of Ilya
Afanasevich Kibel4 an outstanding hydrodynamicist of the twentieth century
who was active and creative throughout his entire career. Unfortunately, his life was
too short. He died suddenly, at the age of 66, on 5 September 1970.
Mainly on the basis of my various publications in mesometeorology (linked with
the lee waves downstream of a mountain in a baroclinic atmosphere and also with
the free atmospheric local circulations above the various Earth sites) during the
years 19571966, in the Kibel Department of the Hydro-Meteorological Centre in
Moscow, in 1968 I had the opportunity to publish my first course in mesometeorology
[1] for the engineering students at the Ecole de la Meteorologie in Paris.
In September 1966 I returned to Paris to write my thesis [2] on the basis of the
results of research (19611965) into the lee waves 2D (non-linear) and 3D (linear)
steady problems in non-viscous and adiabatic atmospheres, with the help of the
Boussinesq approximation. In 1969 I was awarded the degree of Docteur dEtat es
Sciences Physiques by the University of Paris, which added to my Russian Ph.D. of
1960, from the University of Moscow and my SSSR Academy of Sciences Chief
Scientific Research Worker degree in hydrodynamics and dynamic meteorology,
obtained in 1964.
3
Sergey Nikitovich Mergelyan (19282008) was an Armenian scientist an outstanding mathe-
matician, and the author of major contributions in Approximation Theory (including his well-
known theorem in 1951). The modern Complex Approximation Theory was based mainly on his
work (see, for instance, the book Real and Complex Analysis by W. Rudin; French edition,
Masson, Paris, 1978). He graduated from Yerevan State University in 1947, and in 1956 played
a leading role in establishing the Yerevan Scientific Research Institute of Mathematical Machines
(YerSRIMM). He became the first Director of this Institute, which today many refer to as the
Mergelyan Institute.
4
Ilya Afanasevich Kibel (19041970), Member of the SSSR Academy of Sciences, was one of
the leading Soviet scientist in the field of theoretical hydromechanics. He is famous as the founder
of the hydrodynamic method of weather forecasting, and for implementation of mathematical
methods in meteorology. See his pioneer monograph, An Introduction to the Hydrodynamical
Methods of Short Period Weather Forecasting, published in Russian in Moscow (1957), and
translated into English in 1963 (Macmillan, London). Some of his well-known works on the
meteo-fluid are published in Selected Works of I. A. Kibel on Dynamic Meteorology (in Russian,
GydrometeoIzdat, Leningrad, 1984).
1.1 My First Contribution to the RAM Approach in Fluid Dynamics 3
In the original version of my thesis (hand-written in Moscow during 19651966),

in Chap. 1 the Boussinesq approximate equations were derived in an ad hoc manner
(a la Landau as in [3, }56]). However, Paul Germain (future juryman during my
thesis defence in March 1969) was unfavourable towards this method of deriving
Boussinesq approximate equations for convection in fluids, and I was obliged to
completely rewrite that chapter! Germain considered that it is possible to derive these
Boussinesq equations by an asymptotic rational/consistent process (but by what
method?), and in a letter5 written in Paris and dated 8 March 1968, he wrote that
. . . I should understand the justification of our starting equations?
1.1 My First Contribution to the RAM Approach

in Fluid Dynamics
This justification problem was for me a difficult challenge 1 year before my

1969 thesis defence and I was in an awkward situation! For some time I did not
fully understand the question in Germains letter! Finally, however, I chose a
bastardized method via the so-called isochoric model equations, when the density
r is a conservative unknown function along the fluid flow trajectories in timespace
(t, x), such that
Dr=Dt 0; with D=Dt @=@t u r; (1.1)
where denotes the gradient vector and u the velocity vector this constraint
being often used in fluid dynamics when gravity plays an active role. This above
conservative condition on r is, in fact, an incompressible condition. In particular, it
is systematically considered in Yihs monograph [4]; and see also the book by
Batchelor, [5], p. 75.
For u, r, temperature T and thermodynamic pressure p RrT, where R is the
thermally perfect gas constant, when we consider a non-viscous, compressible and
adiabatic atmospheric motion, we have the following Euler non-dissipative system
of three equations:
5
Paul Germain wrote to me (in French!): Jai pu regarder les feuilles que vous mavez adressees
sur la mise en equation de votre probleme. Je prends note du fait que vous ne passez plus par la
forme intermediaire des equations de la convection qui figurait dans les documents que vous
maviez anterieurement donnes. Je ne suis neanmoins pas satisfait, car je ne vois toujours pas
comment est justifiee la coherence de vos approximations et pourquoi, alors que vous supposer les
perturbations de vitesses petites, en particulier la quantite: u2 + w2 U12, afin dobtenir des
equations lineaires, vous ne linearisez pas les conditions aux limites. Vous devez me trouver un
peu tatillon. Mais si je dois faire partie du jury de votre these, cest a titre de mecanicien des
fluides et comme tel, je souhaiterais comprendre le bien fonde des equations de depart. Or depuis
votre expose au seminaire, jeprouve toujours la meme difficulte et les variantes que vous mavez
proposees ne meclairent pas.
Dr=Dt rr:u 0; (1.2a)
DS=Dt 0; (1.2b)
r Du=Dt rp gr 0: (1.2c)
For the specific entropy we have the relation:
S Cv logp=rg ; with g Cp=Cv; (1.3)
the ratio of specific heats, and in (1.2c) the gravity force (with a measure g) is taken
into account
It is well known that an incompressible fluid motion is obtained (from a com-
pressible fluid motion) as a result of the following formal limiting incompressible
process:
Limis g ! 1 with Cp fixed: (1.4)
With (1.4), in place of Eq. 1.2b, according to (1.3), we recover the above
mentioned isochoricity condition (Dr/Dt 0) which leads, from the equation of
continuity (1.2a), to the usual incompressibility constraint:
r constant ) r:u 0:
Finally, in place of the Euler system of Eq. 1.2a1.2c, with (1.3) and as a
consequence of (1.4) we derive for the limit isochoric functions, uis, pis, and ris,
the following simplified isochoric system of inviscid equations:
Dris =Dt 0; r:uis 0; (1.5a, b)
ris Duis =Dt rpis gris 0: (1.5c)
From (1.5a, b), r:uis 0, we have the possibility of introducing two stream
functions, c and w (as in [6]), such that in a 3D, steady case, @uis =@t 0,
@Sis =@t 0, and @ris =@t 0, we obtain the following three relations:
uis rc ^ rw; (1.6a)
ris r c;w; (1.6b)
1=2uis 2 pis =ris g z I c;w; (1.6c)
where z ( x. k is directed above along the unit upward vector k) is the altitude,
and the two functions, r c;w and I c;w, are subject to a determination.
In particular, for the lee-waves problem, over and downstream of a mountain, this
determination is performed via the boundary conditions at upstream infinity where,
in a simple case, an uniform horizontal flow is assumed given.
From Eqs. 1.6a1.6c we derive (again, according to [6]) two scalar equations for
c and w :
r ^ uis :rc @I =@w pis =r2 @r =@w; (1.7a)
r ^ uis :rw @I =@c pis =r2 @r =@c: (1.7b)
If, now, uis 1 U1 z1 i is the speed (along the axis of x x. i) far upstream
of the mountain, which is simulated by the equation z mh(x), then at x ! 1,
with h 1 0, the conditions are:
uis 1 U1 z1 ; vis 1 wis 1 0 ris 1 r1 z1 ; (1.8a)

Z1
c U1 zdz c1 z1 ; (1.8b)
0
where uis 1 uis 1 ; vis 1 ; wis 1 , and z1 being, therefore, the altitude of a stream
line in the basic non-disturbed two-dimensional far flow. In this particular, simple
case, (1.8a,1.8b), the second stream function at infinity upstream is simply the plane
(x, z), and w1 y const.
We will suppose also, implicitly, that the solution of the considered lee-waves
problem ought to be uniformly bounded at all points of the infinite plane (x, z). We
assume also that c 0 determines the wall of the mountain, and in a such case:
I Bc and r Rc (1.9a, b)
and in place of two Eqs. 1.7a, 1.7b, with the conditions (1.8a,1.8b), we derive a
three-dimensional generalization of the 2D equation of Long, considered in his
well-known paper [7]:
r ^ rc ^ rw:rc 0; (1.10a)
r ^ rc ^ rw:rw U1 dU1 =dc1=2dlogR=dcrc ^ rw2

dlogR=dcf1=2U1 gz z1 g:
2
(1.10b)
In the case when (far upstream of the mountain):
U1 U1 0 const ) c1 z1 U1 0 z1 ; (1.11a)
r1 z1 r1 0expbz1 ; (1.11b)
and if we introduce (see (1.12)) the non-dimensional quantities (H is a characteristic

meso-length-scale)
x x=H; z z=H; C c=HU1 0 ; X w=H; (1.12)
we obtain, as in our thesis [2], in place of Eq. 1.10b, the following dimensionless
equation:
r ^ rC ^ rX:rX DC z lrC ^ r X2 1; (1.13)
where
D bH2 g=U1 02 and l bH=2; (1.14a)
and we observe that the following relation
l=D FrH 2 (1.14b)
is true, where FrH 2 U1 02 =gH is the square of a Froude number.

But, FrH 2 1 when H U1 02 =g, and this is indeed the case for the usual
meteo data.
The relation (1.14b) shows that the term proportional to l, in the main Eq. 1.13
must be small l 1, because it is necessary (in (1.13)) that
D l=FrH 2 O1; (1.15)
as a ratio of two small parameters.

The parameter D, being the main lee-waves parameter is the so-called Dorodnit-
synScorer parameter.
The relation (1.15) is, in fact, a similarity rule between two small parameters: l
and FrH2 the use of (1.15) being a key step in the derivation of our leading-order
consistent Eq. 1.17 below.
Rigorously, the term proportional to l can be neglected, in a first approximation,
relative to the term with D, which is assumed O(1), only when
b 2=H; (1.16)
and, in such a case, in leading-order approximate model Eq. 1.17, with subscript B:
r ^ rCB ^ rXB :rXB DCB z 0; (1.17)

where for XB we have as the equation (according to (1.10a)):
r ^ rCB ^ rXB :rCB 0; (1.18)
The effect of the compressibility is present only in the last term of (1.17) pro-
portional to D.
This above approximation is just the well-known Boussinesq approximation of
1903 [8]: The derivatives of r1 z1 can be neglected except when they intervene
in the calculation of the force of Archimedes.
In particular, if we assume (2D case) that:
XB Z y=H and CB cp x; z; (1.19)
we derive, from (1.17), a linear Helmholtz (a la Long [7]) equation:
@ 2 cp =@x2 @ 2 cp =@z2 Dcp z 0: (1.20)
But, if (1.20) is a linear equation (derived without any linearization!) the slip
boundary condition, along the wall of our mountain, remains non-linear the slip
condition being down a curvilinear surface of the mountain,
cp x; z k h x 0; (1.21a)
with
k m=H (1.21b)
and
h x hHx: (1.21c)
The above results are, in fact, the main part of my first theoretical contribution to
the RAM Approach in fluid dynamics, obtained during the rewriting of my Doctoral
thesis in Paris during 19681969.
I do not see, in reality, whether Paul Germain was completely satisfied with my
new derivation. But however that may be, my efforts in writing a new Chap. 2 for my
thesis were successful, and on 10 March 1969, after the defence of this thesis in the
Faculty of Sciences of the University of Paris, I obtained the degree of Docteur dEtat
es Sciences Physiques Paul Germain and Jean-Paul Guiraud being members of my
thesis jury, with, as President of the Jury, Paul Queney,6 Professor at the Sorbonne.
6
The first theoretical investigations concerning 3D lee-waves problems in linear approximation
was, in fact, carried out by Paul Queney. On the other hand, an excellent synthesis of theoretical
developments on relief (lee) waves will be found in WMO Technical Note: The Air flow over
mountains, N 34, Geneva, 1960, by P. Queney et al.
In the last chapter of this book (in Sect. 9.1) the reader can find a more elaborate
RAM Approach to fluid dynamics, for the 2D steady lee-waves problem, in the
framework of low-Mach-number fluid flow (hyposonic) theory, which leads to a
family of consistent, limiting leading-order model equations.
Concerning the full justification of the Boussinesq assertion and a satisfactory
answer to Paul Germains question this justification of the Boussinesq approxi-
mate equations was for 5 years a major challenge for me, and I devoted consider-
able effort to the resolution of this problem.
Only in 1973, in the framework of low-Mach-number asymptotics, taking into
account the existence of a hydrostatic reference state (function only of the altitude),
did I well understand the way for a consistent non-contradictory RAM Approach.
In 1974 [9] these Boussinesq approximate equations for a viscous and non-
adiabatic dissipative atmospheric motions were derived from the full unsteady
NSF dissipative equations.
To describe the atmospheric motions, which represent the departure from the
hydrostatic reference state, I have considered the perturbations of pressure, density
and temperature (these atmospheric perturbations being usually very small, relative
to the hydrostatic reference state) and have rewritten (without any simplifications)
the NSF atmospheric equations relative to these thermodynamic perturbations and
velocity vector.
This derived, very awkward, dimensionless system of equations is, in fact, a new
(exact) form of the NSF classical atmospheric equations well adapted for the
application of our RAMA theory. In Chap. 4 we discuss a detailed RAMA of these
Boussinesq approximate equations, inspired by my Boussinesqs Centenary Anni-
versary paper [10] of 2003, but for the sake of simplicity, only in the framework of
a Euler non-viscous, compressible and adiabatic system of Eq. 1.2a1.2c this
derivation being an instructive test problem for the formulation of our key steps in
Chap. 6, devoted to the mathematics of the RAMA.
1.2 My Collaboration with Jean-Pierre Guiraud

in the Aerodynamics Department of ONERA
In September 1967, thanks to the recommendation of Jean-Pierre Guiraud, I began a

new career as a research engineer in the Aerodynamics Department of the Office
National dEtudes et de Recherches Aerospatiales, in Chatillon, near Paris. After
working at ONERA for 5 years, in October 1972 I was thanks to my Doctoral
thesis (March 1969) appointed Titular Professor of Fluid Mechanics at the
University of Lille 1 a position which I held until 1996.
I continued part of my theoretical researches in fluid-flow modelling as a
Collaborateur Exterieur at ONERA, and during 16 years there, from 1970 to
1986, for a full day once a week I worked with Jean-Pierre Guiraud in exchanging
ideas and envisioning asymptotic modelling for various aerodynamics, stability/
1.2 My Collaboration with Jean-Pierre Guiraud in the Aerodynamics Department 9
turbulence and meteo problems. As a result of this collaboration, throughout

this period we jointly published 26 papers in various scientific journals. (See, for
instance, in the References, our 1986 paper [11], and references to other papers
published either jointly or separately.) These works (published during 19711986)
are devoted, with more or less success, to the application of the ideas that we
discussed concerning various fields in fluid dynamics all being motivated by the
need for solving or understanding the basis of the solution of technological and
geophysical problems involving fluid flows. These problems are related to:
Vortex flows in rotating machines (taking into account that the blades in a row are
usually very close).
Rolled vortex sheets (in a region where the contiguous branches of the rolled sheet
are so close to each other that they are very difficult to capture by numerical
simulation).
Hydrodynamic stability (in a weakly non-linear domain, through perturbation
techniques the underlying mathematical theory being the so-called bifurcation
theory).
Atmospheric flows (see Chap. 9 in this book, and our monograph, Asymptotic
Modeling of Atmospheric Flows [12], published in 1990).
Flow at low Mach numbers (see our Topics in Hyposonic Flow Theory, Lecture
Notes in Physics, vol. 672, 2006 [13]).
It was an extremely stimulating period of scientific research, for me. As far back
as at the end of 1970 years it is evident that asymptotic techniques provide very
powerful tools in the process of constructing working models for fluid dynamics
problems, which are stiff, from the point of view of numerical analysis, coupled
with a simulation via a powerful super-computer.
My approach differs from Van Dykes exposition in [14], in the sense that:
Computational fluid dynamics is now a quite mature discipline, and for some time
the growth in capabilities of numerical simulation will be dependent on, or related
to, the development of rational asymptotic modelling approach RAMA. If such
is the case, then a simple definition of our RAMA is: The art of a strongly
argumentative, consistent, non-ad hoc and non-contradictory modelling assisted
by the spirit of asymptotics.
It is my opinion that RAMA will remain for many years, or even decades a
quite powerful tool in deriving mathematically consistent models for numerical/
simulation fluid research. By mathematical, I mean that the models derived by
RAMA, the approximate consistent models, under consideration, should be
formulated as reasonably well-posed initial and/or boundary value problems, in
place of the starting full NSF extremely complex and stiff problems (as, for
example, turbomachinery flows, which are arguably among the most complicated
known to man and are of great technological importance see Sect. 7.1).
I again observe that our RAMA is an extremely worthwhile objective, because
most of the relevant engineering computations are based on relatively ad hoc
models that are rife with internal inconsistencies.
1.3 A Few Remarks Concerning My Preceding Books

on Modelling in Newtonian Fluid Flows
Concerning the joint efforts of Guiraud and myself in our tentative writing of a book
on the RAMA in fluid dynamics, I must say first that in 1977, after several attempts
to persuade him, we both worked intensively, up to 1982, on a hand-written (in
French) manuscript entitled The laminar flows at high Reynolds numbers: an essay
on the asymptotic modelling of Newtonian dynamics of fluids. The possibility of
publication, after a rewrite in English, became a reality at least for me!
Unfortunately, at that moment our opinions diverged concerning the opportunity
of such publication of our finished manuscript in its 1982 form. Guiraud wished to
pursue a deeper investigation of some delicate and difficult questions requiring time
and additional research. Contrary to Guiraud, I was of the opinion that further
investigations would be of no benefit and, in particular, would not provide anything
else to support our initial objective: to show the effectiveness of our RAMA!
Finally, in 1986 and 1987 I published alone (but by common consent) a course
in two volumes, in French, in the Springer series Lecture Notes in Physics (LNP):
Les Mode`les Asymptotiques de la Mecanique des Fluides, I [15] and II [16] more
or less inspired by the manuscript produced by myself and Guiraud in 1982.
As Titular Professor at the Universite de Lille 1, beginning in 1972, I systemati-
cally used, throughout almost 10 years, various parts of our manuscript in my
teaching of theoretical fluid mechanics as a first Course and second Course,
respectively, for final-year (M.Sc.) undergraduate students and post-graduate
research workers, and for students preparing a doctoral thesis.
In the beginning, in 1977, my goal was, in fact, a monograph devoted to RAMA
for Newtonian fluid flows, and I had in mind the derivation of various models
corresponding to parameters (not only to Reynolds number) characterizing various
(high or low) physical effects viscosity, compressibility, heat conduction, gravity,
Coriolis force, unsteadiness, geometrical constraints, and so on.
The above-mentioned two-volume Course was my first experience in opening a
new way into the difficult field of theoretical (analytical) fluid mechanics via the
NSF equations, offering fresh ideas together with a first systematic presentation of
asymptotic approach in fluid dynamics for both students and young researchers. In a
short critical review (J. Fluid Mech., 1991, vol. 231, p. 691), the following opinion
was expressed concerning this two-volume Course:
The text is in French. Equations are hand-written but very clearly done. In many of the areas
covered in these two volumes there is a conspicuous lack of suitable expository material
available elsewhere in the literature, and Professer Zeytounians notes are to be welcomed
for filling these gaps until fuller and more specialized accounts appear in book form.
In addition, the following appeared in Mathematical Reviews, 1988:

A reader having acquired a practical knowledge of the asymptotic methods which are
presented and used here may certainly benefit by the advanced material about
NavierStokes equations provided in the main body of these two volumes.
1.3 A Few Remarks Concerning My Preceding Books on Modelling 11
Later, in 1994, a third volume was published, also in French: Modelisation

asymptotique en mecanique des fluides newtoniens [17] and here it seems
judicious to quote several sentences extracted from a review (in Appl. Mech.
Rev., vol. 49(7), July 1996, p. 879) by J.-P. Guiraud:
The purpose of this book is to present, through extensive use of dimensional analysis and
asymptotic calculus, a unified view of a wide spectrum of mathematical models for fluid
mechanics . . . Usually, adequacy of a mathematical model is evaluated a priori through
physical insight, experience, and inquiries about the topic. Here, the reader is proposed to
become, on his own, an expert in adequacy, by systematic use of asymptotic approximation.
Although a rather large spectrum of books on fluid mechanics and asymptotic methods may
be found, it seems to this reviewer that the present one is rather exceptional by the extent
and logical organization of the material . . . A fascinating aspect is that the reader is led by
the hand through a jungle of very different mathematical models, including Euler and
Navier for incompressible fluid; Prandtl for boundary layer; Stokes, Oseen, and Rayleigh
for various viscous effects; the usual regimes of aerodynamics, Boussinesq for the atmo-
sphere and the ocean, primitive equations and quasi-geostrophic approximation for meteo-
rology, gravity waves, amplitude equations of KdV or Schrodinger type, and low Mach
number flows, including acoustics . . . Modelisation asymptotique en mecanique des fluides
newtoniens is a valuable book which is recommended both to individuals and libraries for
the precise purpose indicated in the second sentence of this review. In principle, it is self-
contained and might be a reference for students, engineers, and researchers who master
computational aspects but want to be able to assess what kind of approximations are
involved in the equations, as well as initial and boundary conditions with which they
struggle.
Concerning, more precisely, the application of our RAMA to atmospheric

motion, after my 1975 survey lecture,7 published in 1976 [18], I decided to write
a monograph devoted entirely to asymptotic modelling of atmospheric flows
(see [12]), and in 1985 a manuscript (written in French) was ready. This manuscript
was accepted by Prof. Dr. W. Beiglb ock, of Springer-Verlag, Heidelberg, for
publication in English. Unfortunately, the translation into English, by Lesly Bry,
is infelicitous!
I think that my Meteorological Fluid Dynamics [19] is good preparation for the
reading of [12], which was in fact published a little earlier than [19]. Here again,
I quote part of a review of [12] (SIAM Review, vol. 33(4), 1991, pp. 6723) by
Huijun Yang (University of Chicago):
The present work is not exactly a course, but rather is presented as a monograph in which
the author has set forth what are, for the most part, his own results; this is particularly true
of Chaps. 713. (quoted from the Preface of this book):In the book, the author viewed
meteorology as a fluid mechanics discipline. Therefore, he used singular perturbation
methods as his main tools in the entirety of the book . . . The book consists of the authors
more than 25 years work. In the 32 references to his own work, fewer than one third were
published in English, with the rest in Russian or French. Throughout the book, the reader
can strongly feel the influence of Soviet works on the author. However, the author does
7
Entitled La Meteorologie du point de vue du Mecanicien des Fluides, written for the XIIth
Symposium on Advanced Problems and Methods in Fluid Mechanics, Bialowieza (Poland), 813
September 1975.
have his own character. The issues raised in the book, such as the initialization (initial
layer) boundary layer treatment, and well-posedness and ill-posedness of the system, are
very important problems facing researchers today in atmospheric sciences and other related
sciences. The reader will find some valuable information on these issues . . . The mathe-
matically consistent treatment of the subject does give this book a unique place on shelves
of libraries and offices of researchers . . . This book is very different from recent books on
the market [for example, Holton [20], Gill [21], Haltiner and Williams [22], Pedlosky [23],
and Yang [24]]. I recommend that researchers in atmospheric dynamics and numerical
weather prediction read this book to have an alternative view of deriving atmospheric flow
models. Researchers in theoretical fluid mechanics might also be interested to see how
singular perturbation methods can be used in atmospheric sciences. The book may be used
as supplemental material for courses like numerical weather prediction or atmospheric
dynamics. However, I do not think it is a suitable textbook for a regular class: as the
author said in his Preface. I am well aware that this book is very personal one might even
say impassioned.
This review seems rather favourable, but this does not seem to be the case with
P. G. Drazins review (Journal of Fluid Mechanics, 1992, 242):
The author acknowledges that dynamic meteorology is too large a subject for him to
attempt to cover completely. He has set forth what are, for the most part, his own results
in accord with his conception of meteorology as a fluid mechanics discipline which is in
a privileged area for the application of singular perturbation techniques. He applies
the method of multiple scales or the method of matched asymptotic expansions to any
equations he can find, systematically reviewing his own and his associates extensive
research. So it is a very personal view of meteorology, covering some areas of geophysical
fluid dynamics with a formidable battery of notation. The nature of the subject demands a
large and complex notation in any rational treatment. But the notation will allow only a few
to benefit from this book. It is a barrier which I found hard to cross, not having the time and
will to work through the book line by line, as a result with which I am familiar was difficult
for me to follow. The approximations of dynamic meteorology are mostly singular.
Nonetheless, their essentials are well understood by meteorologists now, albeit in a rough
and ready way, and meteorologists are unlikely to be [influenced] by Zeytounians
approach. Yet the approximations of meteorology are subtle and deserve a more rational
development than is commonly understood. This is the achievement of the book. The
author derives many equations systematically, albeit not rigorously, from the primitive
equations, rather than solving equations governing particular problems. For all these
reasons, I feel that the book will be studied intensively by a few specialists but neglected
by others.
Obviously, the last two sentences of this critical review are very controversial
especially the assertion of albeit not rigorously. It seems clear that for Drazin, my
French touch is not to everyones taste! Indeed, the publication of this monograph
in 1990 was possibly premature, despite the publication, in 1985, of my survey
paper [25], Recent Advances in Asymptotic Modelling of Tangent Atmospheric
Motions, devoted to an asymptotic rational theory of the modelling of atmospheric
motions in a flat earth closely related with the so-called bplane approximation.
In [12], [19], and [25] my clear purpose was to initiate a process which does not
seem to have sufficiently attracted the attention of scientists. This process involves
the use of the RAMA for carrying out models; that is, for building approximate
simplified (but consistent) well-posed (at least from a fluid dynamicians point of
1.3 A Few Remarks Concerning My Preceding Books on Modelling 13
view) problems based on various physical situations and concerned with one or
several high or low parameters.
I do not, of course, affirm that this is the only method, nor even the most efficient
one, for deriving such problems in view of a numerical/computational simulation.
I do, however, feel that when such a procedure is feasible it should be undertaken.
As a matter of fact, the application of this approach makes it possible, in principle,
to improve, at least, a second-order model problem obtained from the NSF full
problem, used by advancement in the associated asymptotic expansion.
After 1996 having retired from the University of Lille 1, and with more time to
write I decided to return to my first (1977) idea concerning a monograph devoted
to RAMA for Newtonian fluid flows and the derivation of models corresponding to
various parameters (not only to Reynolds number) characterizing various (high or
low) physical effects. I (partly) realized this objective in 2002 with my monograph
[26] Asymptotic Modelling of Fluid Flow Phenomena the first book in English
devoted entirely to asymptotic modelling of fluid flow phenomena, dealing with the
art of asymptotic modelling of Newtonian laminar fluid flows. In Chaps. 212 of
this work I consider several important topics involved in the accomplishment of my
objective in determining how simplified rational consistent asymptotic simplified
models can be obtained for the most technologically important fluid mechanical
problems.
According to Marvin E. Goldstein (of NASA Glenn Research Center) in his
detailed, scrupulous, and rigorous review (SIAM Review, vol. 45(1), pp. 142146,
2003): There are enough of the selected topics that accomplish the authors objective
to make this book an important contribution to the literature. Goldstein also writes:
Applied mathematicians have always found fluid mechanics to be a rich and interesting
field, because the basic equations (i.e., the NavierStokes equations) have an almost
unlimited capacity for producing complex solutions that exhibit unbelievably interesting
properties, and because the dimensionless form of these equations contains a parameter
(called the Reynolds number) which is usually quite large in technologically and geophysi-
cally interesting flows. This means that asymptotic methods can be used to obtain appro-
ximate solutions to some very interesting and important flow problems. These solutions
usually turn out to be of non-uniform validity (i.e., they break down in certain regions of the
flow), and matched asymptotic expansions have to be used to construct physically mean-
ingful results. However, advances in computer technology have led to the development of
increasingly accurate numerical solutions, and have thereby diminished the interest in
approximate analytical results. But real flows (especially those that are of geophysical or
engineering interest) are extremely complex and exhibit an enormous range of length and
time scales whose resolution will probably remain well beyond the capabilities of any
computer that is likely to become available in the foreseeable future. So simplification and
modelling are still necessary, not only to meet the engineers requirement for generating
numbers but also for developing conceptual models that are simple enough to be analyzed
and understood. The asymptotic scaling techniques and the reduced forms of the general
equations that emerge from the matched asymptotic expansion process (as well as
from other singular perturbation techniques) provide a rational and systematic method for
obtaining the necessary simplified flows model, which, in most cases, still have to be solved
numerically. The author states that the goal of this book is to promote the use of asymptotic
methods for developing simplified but rational model for the NavierStokes equations
which can then be solved numerically to obtain appropriate descriptions of the flow.
This is an extremely worthwhile objective, because most of the relevant engineering

computations are based on relatively ad hoc models that are rife with internal incon-
sistencies. To my knowledge, this is the first book devoted to accomplishing the authors
stated objective, and it is, therefore, unfortunate that it is not as well executed as it could be.
To mitigate this last critical sentence, however, he continues:

However, there are enough of the selected topics that accomplish this objective to make this
book an important contribution to the literature. It contains an excellent chapter (Chap. 11) in
which the classical model equations for large-scale atmospheric motion are derived in a fairly
rational fashion. The author also devotes a full section (Sect. 6.6) to turbomachinery flows,
which are arguably among the most complicated known to man and are of great technological
importance. It is remarkable that he was able to make some progress toward developing an
asymptotic basis for some of the more prominent engineering models of these flows.
Goldstein concludes: It is this reviewers hope that the deficiencies in this work
will encourage others to write new and improved books with similar themes but
unfortunately, it seems that for the present this is not the case! On the other hand, in
2006 and 2009 Springer published my two monographs, Topics in Hyposonic Flow
Theory [13], devoted to hyposonic (low Mach numbers) flows, and Convection in
Fluids: A Rational Analysis and Asymptotic Modelling [27], mainly related to the
well-known Benard convection problem in a layer of weakly expansible liquid
heated from below.
Concerning the first of these, a considerable amount of stimulation and encour-
agement was derived from my collaboration with Jean-Pierre Guiraud, who, over a
period of 20 years, has played an important role in asymptotic modelling of the
various low-Mach-number flow problems presented in that book. The reader should
take into account that is the first book devoted to hyposonic flow theory, and it is the
authors hope that the various unavoidable deficiencies (noted by Goldstein in his
review of [13]) will persuade others to work on similar themes.
On the other hand, in [27] the main motivation was a rational analysis of various
aspects (in particular, the influence of the viscous dissipation, free surface and
surface tension) of the Benard convection (heated from below) problem. It presents
a careful investigation of three significant approximate models (see, for instance,
Chap. 8 in the present book) related to the Benard (1900) experiments, by which he
discovered his well-known Benard cells! It is evident that Professor Manuel G.
Velarde was influential when I wrote my book on convection in fluids [27], as I
benefited greatly from our collaboration and many discussions relating to
Marangoni thermocapillary convection during my sojourn at the Instituto Pluridis-
ciplinar UCM de Madrid in 20002004.
1.4 Conclusion
More than 50 years ago, with the works of Kaplun, Kaplun and Lagerstrom, and
Proudman and Pearson, asymptotic techniques provided a new impetus for research
in theoretical fluid dynamics. Twenty years later, a much more powerful revival
1.4 Conclusion 15
was possible due to the dramatic influence of high-speed computers and the
numerical analysis and simulation of fluid flow problems. The survey paper by
Birkhoff [28] includes, through a series of case studies, a detailed assessment of the
status, development, and future prospects of numerical/computational fluid
dynamics.
During the early times, asymptotic techniques were mostly used in order to
derive approximate solutions in closed form. Perhaps of more significance for the
progress of understanding and also of research, however, was the use of asymptotic
techniques in order to settle, on a rational basis, a number of approximate models
which much earlier were often derived by ad hoc non-rational procedures. One of
the most well-known examples is Kapluns celebrated paper on boundary-layer
theory, which provided a firm theoretical basis for some 50 years of boundary-layer
research.
From this early example it is clear that my 1977 idea, relative to asymptotic
techniques as a well-suited and invaluable tool for the derivation of mathematically
consistent models (from full fluid dynamics equations NSF equations) which are
amenable to numerical treatment rather than for obtaining closed form solutions,
was a perspective of scientific activity in interaction with numerical simulation
even though, in the 1970s, numerical fluid dynamics was almost non-existent due to
the lack of high-speed computers.
It is now evident that asymptotic techniques serve as very powerful tools in
the process of constructing rational consistent mathematical simplified models
for problems which are stiff, from the point of view of numerical analysis. Here,
Chap. 6 is devoted to the mathematics of the RAM Approach, which seems a good
basis for a practical use of this RAMA in simulation/computation via high-speed
computers.
As a matter of fact J. P. Guinaud, who read a large part of the Chaps. 1 to 6 of the
present book suggest me to quote what follows: While having been absent from
the Community of fluid mecanicians fifteen years from now, it was a pleasure for
me to read the report by Zeytounian, of a long coworking with him. It is mere
justice to mention that, during this active collaboration, a number of ideas were
initiated by Zeytounian. My main contribution was the result of ten years of
struggle with asymptotics before I had the good fortune to meet Zeytounian. A
significant contradiction is obvious in the scientific activity of J. P. Guinaud. He
never published any book whilst having written a number of documents,
corresponding to many courses he teached manuscripts which were much more
carefully written than simple notes to be distributed for the students. In particular,
the Guirand Notes (Topics in hypersonic flow theory, Department of Aeronautics
and Astronantics, Stanford University; SUDAER n 154, may 1963, Stanford,
California, USA) are Published in Russian by MiR, Moscow Editions, as a book
in 1965.
Part II
A Sketch of a Mathematical Theory of the
RAM Approach
Chapter 5
The Structure of Unsteady NSF Equations
at Large Reynolds Numbers
Numerous papers are devoted to investigations of NSF equations in the case of

large Reynolds numbers, and many references are cited in my Theory and
Applications of Viscous Fluid Flows [47]. Curiously, the case of unsteady NSF
equations for large Reynolds numbers, when Re tends to infinity, has been very
poorly considered. However, in Chap. 4 of [47] (Sect. 4.3) there is a presentation of
the asymptotic structure of unsteady state NSF equations.
Below, in this Chap. 5, I again give a presentation of this structure the reason
being that such unsteady NSF equations structure, at Re "/ , very well illustrates
the importance of the various limiting processes related to fixed time and space, and
forms a guideline in the elaboration of our mathematics for the RAMA.
On the other hand, such a structure illuminates the various consequences of the
singular nature of the Prandtl (1904) concept of the boundary layer. A strong
singularity for NSF equations opens (it seems to me) new perspectives for the
resolution of various paradoxes encountered in unsteady compressible fluid flow
theory. It is surprising that this close initial time singularity of the Prandtl boundary
layer, was for so long ignored.
5.1 Introduction
Fluid dynamicians and applied mathematicians have always found fluid dynamics
to be a rich and interesting field for investigations, because the basic system of
partial derivative equations for a Newtonian fluid the so-called NavierSto-
kesFourier equations (NSF) equations have a great capacity for producing
various particular fluid flow models.
In particular, a large class of such fluid flow models is closely linked with the
analysis of a dimensionless form of the NSF system, and more specifically with
the large Reynolds numbers (Re "/ ) for a compressible, weakly viscous, and heat-
conducting fluid flow.

98 5 The Structure of Unsteady NSF Equations at Large Reynolds Numbers
Indeed, in technologically and geophysically relevant fluid flows, Re is usually

quite large. In 1904 Prandtl took into account this fact and derived (in an ad hoc
manner) his well-known Prandtl Boundary-Layer (BL) equations. Surprisingly, it
seems that the case of an unsteady fluid flow, when initial conditions are prescribed,
has not been carefully considered, and this BL Prandtl concept has become default
and become singular near t 0! Only in 1980, in a short note [98], was I the
(apparently) first to show that:
The limiting form (at Re " 1) of the unsteady NSF equations, near t 0 and in the
vicinity of a wall (z 0), bounding below the compressible, viscous and heat-conducting
fluid flow, is identified in place of Prandtl BL equations rather with the equations of the
Rayleigh compressible problem, considered first in 1951 [99] by Howarth.
In Chap. 5 I fully intend to present a more deep and careful investigation of this
singular problem, which not only has an obvious theoretical interest, but also
seemly a practical one. For the NSF unsteady full equations, we have considered,
as a typical working case, an emergency situation: namely, a sudden rise in
temperature locally on the wall at initial time t 0 which is a possible applica-
tion of our new four regions structure for the NSF unsteady equations!
5.2 The Emergence of the Four Regions as a Consequence

of the Singular Nature of BL Equations Near t 0
The main feature of the well-known Prandtl 1904 BL concept in a thin region near
the wall (z 0), in an aerodynamics problem for high Reynolds number, is linked
with the strong simplification of the equation of motion for the vertical component
w of the velocity u (v; w).
In steady and unsteady, in compressible viscous, in heat-conducting, and in
incompressible viscous fluid flows, this Prandtl BL concept in all cases produces
a very degenerate limit equation, when Re " 1, for w!
When we work with dimensionless quantities, then, for the variation of the
pressure in the direction normal to the (horizontal) wall, z 0, relative to vertical
coordinate, z, we obtain (when the gravity force is not taken into account):
@p=@z 0: (5.1)
and in particular, the partial derivative in time for the component w of the velocity
disappears in a BL system of equations!
If this failure seems not to have serious consequences in the usual case of steady
or unsteady incompressible viscous fluid flows, conversely, this is not the case for
an unsteady compressible viscous and heat-conducting fluid flow. In such a case, as
an unfortunate consequence of (5.1), we have a new four regions structure for
NSF equations governing these fluid flows at high Reynolds numbers.
5.2 The Emergence of the Four Regions as a Consequence 99
This new structure of unsteady NSF equations seems, in particular, very

significant for the various applications linked with the heat emergency situations
(sudden rise of a thermal source) explosions, fires, failures of oil and gas
pipelines, and so on in a local domain on the wall in contact with the fluid.
Such a four regions structure, at Re " 1, is linked with four limiting processes
in full unsteady NSF equations and replaces the two classical regions,
EulerPrandtl and regular coupling, linked with the following two limiting
processes (the horizontal coordinates x and y being fixed):
LimE e # 0; with t and z fixed: (5.2a)
and
LimPr e # 0; with t and z z=e fixed; (5.2b)
As a consequence of the strong degeneracy linked with (5.1), in the unsteady

case, the limiting process (5.2b) is singular near t 0. Therefore, because the
partial time derivative of the vertical component of the velocity is absent in BL
equations, we do not have the possibility of taking into account the corresponding
data which is prescribed for full unsteady NSF equations.
It is necessary to consider a third limiting process, inner in time a so-called
acoustic limiting process:
LimAc e # 0; with t t=e and z z=e fixed; (5.3)
An acoustic problem must be considered in a third acoustic region close to

initial time, and then (if possible) a matching with the boundary-layer (BL) region
far from the region of the acoustics. Indeed, the consideration, in this acoustic
region, of an unsteady adjustment problem, when t ! 1, presents the possibility
(in principle) of prescribing the correct initial conditions for unsteady BL equations
significant only far of the initial time in the Prandtl BL region.
But a new problem emerges because in this third non-viscous, near the initial
time acoustic region, we do not have the possibility of taking into account the
thermal condition on the wall z 0, prescribed for NSF equations in the frame-
work of the heat emergency problem.
As a consequence, it is necessary to consider a fourth limiting process simulta-
neously near t 0 and z 0:

LimRa e # 0; with y t e2 and Z z e2 fixed; (5.4)
In such a case, in this corner small fourth region, we derive the unsteady one-
dimensional NSF equations governing the compressible Rayleigh problem, with
the corresponding thermal condition on z 0 (Fig. 5.1).
Precisely, the above (related with the limiting process (5.4)) compressible
viscous and heat-conducting Rayleigh problem presents the possibility of
producing an answer to following question: What is the significant problem,

governing the unsteady compressible viscous and heat-conducting fluid flow,
emerging as a result of a suddenly rise in temperature locally on the wall, z 0,
at time, t 0?
Fig. 5.1 Four regions structure of NSF unsteady equations at large Reynolds number
In Sect. 5.3 the formulation of the starting unsteady NSF problem is given,
while Sect. 5.4 is devoted to the derivation of the corresponding four problems
linked with the above four limiting processes (5.2a, 5.2b), (5.3), and (5.4). Section
5.5 concerns the problem of the unsteady adjustment (via the Rayleigh problem
significant in the fourth region) to Prandtl BL evolution, in time, which is signifi-
cant in the second region close to wall z 0, but far from the third and fourth
regions near the initial time t 0. In Sect. 5.6 some conclusions are presented.
5.3 Formulation of the Unsteady NSF Problem
We consider the atmospheric dry air as a thermally perfect, viscous, and heat-
conducting gas, with constant dissipative coefficients. In such a case, the dimen-
sionless NSF equations are written (for dimensionless functions, v, w, p, r, and T)
in the following form all functions, variables, and coefficients being non-
dimensional:
@r=@t D rv @rw=@z 0; (5.5)
5.3 Formulation of the Unsteady NSF Problem 101

r@v=@t v : Dv w@v=@z 1 gM2 Dp e2 fDv
(5.6)
1=3 D D : v @w=@zg;

r@w=@t v : Dw w @w=@z 1 gM2 @p=@z e2 fDw
(5.7)
1=3@ =@zD : v @w=@z;
r@T=@t v : DT w @T=@z g 1p D : v
@w=@z g=Pr e2 DT gg 1e2 M2 fF (5.8)
2
2=3D : v @w=@z g;
with
p rT; (5.9)
where the horizontal (relative to coordinates (x, y)) velocity vector is v (u, v),
and viscous dissipation is written in the following form:
F @u=@z @w=@x2 @v=@z @w=@y2 @u=@y @v=@x2

h i (5.10)
2 @u=@x2 @v=@y2 @w=@z2 ;
and D D2 @ 2 =@z2 with D @=@x; @=@y.

For the above evolution equations (5.5)(5.8) we write. as initial conditions at
initial time t 0:
t 0 : v 0; w 0; r 1 and T 1: (5.11)
At the horizontal solid wall, z 0, we assume:
z0:vw0 (5.12a)
and
T Yt=b, P; when t 0; Yt=b, P 0; when t 0; (5.12b)
where

b t* t0 1; (5.13)
is a ratio of two time scales: t*, a short time scale, in comparison to characteristic
evolution time scale t , which appears in Strouhal number S in (3.3).
In condition (5.12a, 5.12b), the dimensionless function Yt=b; P is used to

simulate an emergency of a thermal spot at t+ 0, P being a point on a local
domain on the wall, P
D, for which the reference length scale L is a diameter for
this time-dependent domain D D(t).
In the above four dimensionless NSF equations (5.5)(5.8), we have assumed
that (S is the Strouhal number):

S 1 ) U0 L0 t0 ;
and in the fact (t0 is the time with dimension):
Yt=b; P Yt0 =t ; P:
The characteristic time t L /U is a long time and characterizes the

evolution of fluid flow after the thermal spot emergency, while the characteristic
short small time, t* t , is linked just with this emerging thermal spot short
(time) interval.
The above formulated NSF initial-boundary value problem, (5.5)(5.13), is a
very complicated mathematical problem, and the rigorous proof concerning its
well-posedness is obviously an intractable question!
In fact, our objective below is rather to analyze the specific structure of this
problem, when e tends to zero for large Reynolds number and consider the
relations, via matching, between the four particular fluid flows regions discussed in
Sect. 5.2.
The existence of these four fluid flows regions, at large Reynolds number, shows
that we have the possibility of considering this above NSF system of equations as a
puzzle! A challenging, but difficult, approach is to deconstruct1 this puzzle,
relative to limiting values vanishing or infinity of various reference
parameters, Re, M, or Pr. . ., in order to unify by a RAMA process the set of
various, ill-assorted, partial approximate system of equations, customarily used in
classical fluid dynamics.
We observe that from our above formulated mathematicalphysical
(5.11)(5.13) problem for the NSF equations (5.5)(5.8) with (5.9), it follows
that we have the possibility of first investigating not only the initial stage of the
motion, emerging as a cause and effect of a sudden rise in temperature locally on
the wall z 0 at t 0 in a corner fourth region, but also the evolution of this
thermal accident in second, Prandtl, and first, Euler, BL viscous and inviscid
Eulerian regions, by two matching processes.
1
In fact, the system of NSF equations, despite its dimensionless form, do not have a fixed
meaning, even if the various reference parameters, in NSF equations, give a good idea for an
investigation in this way! A real meaning is created, each time, in the act of the RAM Approach,
relative to a precise parameter (high or low), via the derivation of a consistent simplified model,
this process is just a deconstruction (a la Derrida) of NSF system of equations.
5.4 Derivation of the Corresponding Four Model Problems 103
The considered physical case is a typical free problem independent of any given
external flow. This case can be generalized for the atmosphere, where gravity plays
an obvious and important role in emerging convective local motion. For such an
atmospheric case, a typical example is a forest fire, which is actually a very bad
accident which causes a large amount of damage, and is considered to be an
environmental disaster!
5.4 Derivation of the Corresponding Four Model Problems
Below we consider the derivation of approximate leading-order equations for the

corresponding four limiting processes (5.2a, 5.2b), (5.3), and (5.4), discussed in
Sect. 5.2, and linked with the limiting process e # 0, from our full NSF problem:
(5.5)(5.9), with (5.10), and (5.11)(5.13) our main objective being a consistent
obtention of these four particular systems of equations related to the four limiting
processes (5.2a, 5.2b)(5.4).
5.4.1 EulerPrandtl Regular Coupling
When we consider the first, Euler, limiting process (5.2a), LimE, at e # 0, with t and
z fixed, from NSF, evolution equations (5.5)(5.8), with (5.9), we obtain a strong
degeneracy (close to the horizontal solid wall z 0), which leads to a Euler system
of equations for the leading-order functions, vE ; wE ; pE ; TE andrE , in the follow-
ing Euler asymptotic expansion associated with (5.2a):

v; w vE ; wE e vE 1 ; wE 1 :::; (5.14a)
p; T; r pE ; TE ; rE epE 1 ; TE 1 ; rE 1 :::; (5.14b)
where the Eulerian terms (with E as subscript) are dependent on t, x, y, and z. We

then obtain the following system of Euler inviscid compressible, adiabatic,
equations:
@rE =@t D:rE vE @rE wE =@z 0; (5.15a)
rE @vE =@t vE :DvE wE @vE =@z 1=gM2 DpE 0; (5.15b)
rE @wE =@t vE :DwE wE @wE =@z 1=gM2 @pE =@z 0; (5.15c)

rE @TE =@t vE :DTE wE @TE =@z g 1pE D:vE @wE =@z 0;

(5.15d)
with
p E rE T E : (5.15e)
Because all the dissipative terms, in the right-hand side of NSF equations
(5.6)(5.8), are absent in limiting, leading-order, Euler equations (5.15a5.15d)
(see, for instance [100], and [37], Chap. 9), these Euler equations cannot be valid in
the vicinity of the wall z 0.
The significant equations valid near z 0, which replace the above Euler
equations (5.15a5.15e), are derived when we introduce, in place of z, an inner
vertical coordinate, significant in the vicinity of z 0: namely,
z z=e; (5.16a)
when we consider the second, Prandtl, limiting process (5.2b), LimPr at e # 0, with
t and z fixed.
In such a case, again, from full NSF equations (5.5)(5.8), with (5.9), for the
leading-order functions vPr ; pPr ; TPr ; rPr , and secondorder vertical component of
the velocity, wPr 1 , we consider the following asymptotic, a la Prandtl, expansion
associated with (5.2b):

v; w vPr ; 0 e vPr 1 ; wPr 1 :::; (5.17a)
p; T; r pPr ; TPr ; rPr epPr 1 ; TPr 1 ; rPr 1 :::; (5.17b)
where the Prandtl (BL) terms (with Pr as subscript) are dependent on t, x, y, and z.
Then, with (5.2b) and (5.17a, 5.17b), we derive the Prandtl BL unsteady equations
(see Stewartson [101], and the more recent book by Oleinik and Samokhin [102]):
@rPr =@t D:rPr vPr @rPr wPr 1 =@z 0; (5.18a)
rPr @vPr =@t vPr :DvPr wPr 1 @vPr =@z

(5.18b)
1=gM2 DpPr @ 2 vPr =@z2 ;
@pPr =@z 0; (5.18c)

rPr @TPr =@t vPr :DTPr wPr 1 @TPr =@z
g 1pPr D:vPr @wPr 1 =@z
g=Pr@ 2 TPr =@z2 gg 1M2 j@vPr =@zj2 ; (5.18d)
with
pPr rPr TPr : (5.18e)
It is well known that both the above system of equations outer, Euler
(5.15a5.15e), and inner, Prandtl (5.18a5.18e) are related to the following
classical matching relation (discussed in Sect. 6.4.2):

lim z"1 LimPr limz#0 LimE ; (5.19a)
and, as a first consequence of (5.19a), we obtain for the Euler outer system of
Eqs. 5.15a5.15e the following single (slip!) condition:
wE 0 at z 0: (5.19b)
Then, as a second consequence of (5.19a), we see that from the strong

degenerated equation (5.18c), in the Pranddtl system of Eqs. 5.18a5.18e, we
have the possibility of relating the constant value of pPr , with respect to vertical
BL coordinate, z, with the value of pE at z 0:
pPr t; x; y pE t; x; y; 0 pE;0 t; x; y
and
DpPr gM2 rE;0 @wE;0 =@t vE;0 :DwE;0 : (5.19c)
Now, concerning the thermal condition (5.12), with the parameter b given by
(5.13), we are obliged to assume that (when t is fixed, in LimPr , far of the initial
time):
b be # 0 with e # 0; (5.20a)
and we have only the possibility of writing, for the BL equations (5.18a5.18e), the
following conditions:
on z 0 : vPr 0; wPr 1 0;
TPr Y1; P; at t >0 fixed: (5.20b)
Obviously, in the framework of the EulerPrandtl regular coupling we do not

have the possibility of taking into account the thermal accident linked with the
temperature emergency at initial time on the wall z 0.
On the other hand, the above Prandtl system of Eqs. 5.18a5.18d, with (5.18e),
due to the reduced BL equation (5.18c), must be considered as a system of two
equations for vPr and TPr , while the Prandtl vertical velocity wPr 1 must be computed
through the following relation:

z
wPr TPr =pE;0
1
f@=@t vPr :DrPr
0 (5.20c)
rPr D:vPr gd z;
since wPr 1 0 on z 0, according to the second condition in (5.20b), and the

(matching) relation:

limz"1 wPr 1 wE;0 1 ; (5.20d)
is, in fact, a regular coupling condition with the second-order linearized Euler
equations for the terms with E1 proportional to e, in Euler asymptotic expansion
(5.14a, 5.14b).
However, the problem of two initial conditions for the two unsteady Prandtl
equations, (5.18b) and (5.18d), for vPr and TPr , is more subtle, and is a direct
consequence of the change of the nature of Prantdl equations relative to the
incomplete parabolic character of NSF equations (see [59, 60, 67]).
In fact, for vPr and TPr we have a system of two hyperbolicparabolic equations
[102]:
rPr DPr vPr =Dt @ 2 vPr =@z2 F; (5.21a)
rPr DPr TPr =Dt g=Pr@ 2 TPr =@z2 G; (5.21b)
where DPr =Dt @=@t vPr :D , and the right-hand side F and G are a collection of
terms with the first-order derivatives relative to D and z.
The continuity equation in the Prandtl system (5.18a5.18d) is, in fact, an
equation determining wPr 1 , due to (5.20c), and in place of (5.18e) the relation:
rPr pE;0 =TPr ; (5.21c)
determines the density rPr .

Without loss, the generality the hyperbolicparabolic character of the system
of Eqs. 5.21a, 5.21b is related to the transmission of the information in the
planes z const, along the trajectories linked with the derivative operator
DPr =Dt @=@t vPr :D, supporting the hyperbolicity this information being
instantaneously diffused by vertical coordinate z on each normal direction to the
wall z 0, at each moment t (which just characterizes the parabolicity).
We observe also, that the domain of the dependence, for a fixed moment, of the
point on the wall has an angular form, but in the unsteady case the precise form of
this domain is not easy definable.
The above brief discussion shows explicitly that there is obviously a change in
the mathematical character of the fluid dynamics equations, when we pass from a
fourth order in time (four partial derivatives in time) unsteady NSF system of
Eqs. 5.55.8, to an unsteady Prandtl reduced system of two Eqs. 5.21a, 5.21b. This
strong modification leads to a singular nature of the system (5.21a, 5.21b) near the
initial time t 0 this singular nature of the Prandtl boundary-layer concept for
the unsteady case being (curiously) ignored up to 1980 (See, for instance, a recent
(1994) discussion by Van Dyke: Nineteenth-century roots of the boundary-layer
idea [103]).
More precisely, it is necessary to prescribe in the framework of Prandtl BL
equations (for instance for Eqs. 5.21a, 5.21b) only two initial conditions at t 0.
But unfortunately, the initial data for vPr and TPr , at time t 0 (designated by: v0 Pr
and T0 Pr ), are certainly different from the initial conditions (in particular, (5.11)) for
the full NSF system of Eqs. 5.55.8.
Indeed, the main question is the following. Since two, NSF, initial conditions
are lost during the Prandtl limiting process (5.2b), how are the (unknown?) initial
conditions:
vPr v0 Pr and TPr T0 Pr ; at t 0 (5.22)
for the unsteady Prandtl BL equations (5.18b) and (5.18d), and how are the data v0 Pr
and T0 Pr linked with the initial data in conditions prescribed at the start (at t 0)
for the unsteady NSF equations. The answer is strongly related to the obtention,
from the full unsteady NSF equations, of a particular system of equations valid
near initial time and written relative to a short time. In fact, we have three short
times:
t t=e; y t=e2 ; and s t=b; (5.23)
and the choice of data v0 Pr and T0 Pr , is realized via an unsteady adjustment problem,
when the adequate(?) short time tends to infinity in unsteady adjustment equations
valid near initial time!
5.4.2 Acoustic and Rayleigh Problems Near the Initial Time t 0
First, with the limiting process (5.3),
LimAc ; when e # 0; with t t=e and z z=e fixed;
from the NSF equations (5.5)(5.8), with (5.9), for the leading-order functions,
vAc ; wAc ; pAc ; TAc ; and rAc ; in the following asymptotic acoustic expansions,
associated with (5.3):

v; w vAc ; wAc e vAc 1 ; wAc 1 :::; (5.24a)
p; T; r pAc ; TAc ; rAc epAc 1 ; TAc 1 ; rAc 1 :::; (5.24b)
where the acoustic terms (with Ac as subscript) are dependent on t; x; y, and z,

we derive the following, compressible, non-viscous, adiabatic, unsteady, one-
dimensional, (t; z), equations:
@rAc =@ t @rAc wAc =@z 0; (5.25a)
rAc @wAc =@ t wAc @wAc =@z 1=gM2 @pAc =@z 0; (5.25b)
rAc @TAc =@t wAc @TAc =@z g 1pAc @wAc =@z 0; (5.25c)
pAc rAc TAc : (5.25d)
and also the following transport equation for vAc :
@vAc =@t wAc @vAc =@z 0: (5.26)
The system of Eqs. 5.25a5.25d valid simultaneously close to initial time,

t 0, and in a thin layer in the vicinity of the wall, z 0 are identical to the
usual equations for one-dimensional vertical unsteady motion in (non-viscous,
adiabatic) gas dynamics.
Once wAc has been obtained, through the solution of the system (5.25a5.25d)
with the proper initial conditions (that is, the initial conditions, (5.11), for starting
NSF equations), single (slip) boundary condition:
wAc 0 on z 0; t > 0; (5.27a)
and matching condition (in time)
Limt"1 wAc wPrjt0 0; (5.27b)
we may use the transport equation (5.26) in order to compute vAc .

Unfortunately, with the above non-viscous adiabatic system of
Eqs. (5.25a5.25d), and transport equation (5.26), we do not have the possibility
of taking into account our main emergency thermal effect (via the thermal spot
Yt=b; P on the wall), since the inviscid (non-viscous, adiabatic) system
(5.25a5.25d) and Eq. 5.26 are not valid close to the wall, where the conditions
(5.12a, 5.12b), with (5.13), are prescribed.
The Eq. 5.26 for vAc shows that
v0 Pr Limt"1 vAc ; (5.27c)

from matching, and it seems (in (5.22)) that we can assume (for our particular case),
as a value for v0 Pr , zero; but this is certainly not the case for T0 Pr .
Again, near initial time and close to the wall, where we have the conditions
(5.12a, 5.12b), according to Zeyounian [98], it is necessary to consider the Rayleigh
limiting process (see (5.4)):
LimRa ; when e # 0; with y t=e2 and Z z=e2 fixed; (5.28a)
with

v; w vRa ; wRa e vRa 1 ; wRa 1 :::; (5.28b)
p; T; r pRa ; TRa ; rRa epRa 1 ; TRa 1 ; rRa 1 :::; (5.28c)
where the Rayleigh terms, in (5.28a, 5.28b), with Ra as subscript, are dependent
on y; x; y, and Z.
In such a case, from full NSF unsteady equations (5.5)(5.8), (5.9), for leading-
order functions, vRa ; wRa ; pRa ; TRa ; and rRA , we derive below the Rayleigh
equations (5.29) and (5.30a5.30d) used in the compressible Rayleigh problem,
which are, in fact, the one-dimensional reduced form of the full NSF equations
valid in a corner region near initial time y 0, and close to the wall Z 0. Namely:
rRa @vRa =@y wRa @vRa =@Z @ 2 vRa =@Z2 ; (5.29)
@rRa =@y @rRa wRa =@Z 0; (5.30a)
rRa @wRa =@y wRa @wRa =@Z 1=gM2 @pRa =@Z

(5.30b)
4=3@ 2 wRa =@Z2 ;
rRa @TRa =@y wRa @TRa =@Z g 1pRa @wRa =@Z

g=Pr@ 2 TRa =@Z2 gg 1M2 fj@vRa =@Zj2 (5.30c)
2
4=3j@wRa =@Zj g;
pRa rRa TRa ; (5.30d)
These above Rayleigh equations, (5.29) and (5.30a5.30d), are applied in [99]
for the Rayleigh compressible problem by Howarth in 1951, but in the case of an
infinite flat horizontal plate (submerged in a viscous and heat-conducting and
originally quiescent fluid) which is impulsively started moving in its own plane
with a constant velocity.
In fact, from 0ur above RAM Approach I can now affirm that in a corner region
y; Z, which is significant for the small time near initial time and in thin layer close
to wall, at the leading-order for large Reynolds number, the above compressible
Rayleigh equations, (5.29) and (5.30a5.30d), for a viscous and heat-conducting
fluid flow, consistently replace the full unsteady NSF equations.
Both these unsteady systems, ((5.29) and (5.30a5.30d), and (5.25a5.25d) and
(5.26)), both valid near the initial time, are related amongst themselves by the
following matching relations:
LimZ"1 wRa wAcjzo ; (5.31a)
Limy"1 vRa ; wRa ; rRa ; TRa vAc ; wAc ; rAc ; TAc jt0 : (5.31b)
The reader can find in Antontsev et al. [104], Chap. 2, some mathematically
rigorous results concerning the above, a la Rayleigh, equations (5.29) and
(5.30a5.30d); and see also the review paper by Solonnikov and Kazhykhov [105].
Finally, if we assume that in a thermal spot
Yt=b; P; with b << 1;
b defined by (5.13), is equal to e2 , then, for the emergency of the temperature

accident we have the possibility of taking into account all starting initial (5.11) and
wall (5.12a, b) conditions, in the framework of an initial-boundary values Rayleigh
problem.
We therefore write, for Eqs. 5.295.30d, the following initial conditions:
y 0 : vRa 0; wRa 0; rRa 1 and TRa 1; (5.32a)
and, at the horizontal solid wall, Z 0, we assume:
Z 0 : vRa wRa 0 and TRa Yy; P; y 0: (5.32b)
The above starting problem, (5.29)(5.30a5.30d) with (5.32a, 5.32b), is a

typical problem for various emergencytemperatureaccident phenomena which
develop when y 0.
5.5 Adjustment Processes Towards the Prandtl BL Evolution

Problem
If we want to take into account the sudden heat emergency, at the time y+ 0, in a
local domain, P
D, on the wall Z 0, then it now seems justifiable that the main
working problem is just the above compressible, viscous, and heat-conducting
Rayleigh problem ((5.29), (5.30a5.30d), (5.32a, 5.32b)).
5.5 Adjustment Processes Towards the Prandtl BL Evolution Problem 111
This Rayleigh problem is valid, simultaneously, near the initial time and close to

wall in a small corner (fourth) region with a physical size of order (n /U 2)

relative to time and (n /U ) relative to vertical coordinate these time and length
scales being exactly those used by Howarth [99].
5.5.1 Adjustment Process Via the Acoustics/Gas Dynamics

Equations
Once the above Rayleigh problem (5.29), (5.30a5.30d), and (5.32a, 5.32b) is
solved (numerically), then we have the possibility (first) of prescribing, by
matching relations (5.31a, 5.31b) to Eqs. 5.25a5.25d, with (5.26), of gas dynamics
significant in the third inviscid region near the time, t 0, and characterized by
t and z the consistent conditions at t 0 and z 0.
As a consequence of this above matching, it seems that we can expect that in
conditions (at t 0 and z 0) for Eqs. 5.25a5.25d with (5.26) of gas dynamics,
the influence of the wall condition for the temperature is taken into account, but
only via the limit value Y/; P.
We observe that in the wall, z 0 at t > 0, condition (5.20b) for the Prandtl BL
equation (5.18a5.18e), this same function (independent of time?) Y/; P is also
present.
The acoustic/gas dynamics equations (5.25a5.25d) with (5.26), with these
initial conditions (5.31b) and single boundary condition (5.31a), for wAC at z 0,
which take into account the (partial?) influence of thermal spot (but independent of
time function Y/; P), present the possibility of considering, for t ! 1, an
unsteady adjustment inviscid problem for the initialization of the Prandtl BL
equations. As a typical example, see, for instance, our paper co-authored with
Guiraud [106], which determines, in particular, the initial data T0Pr.2
When both vPr0 and TPr0 are known, as a result of the above unsteady adjustment
inviscid problem, then later, via the initial-boundary value BL problem, significant
in the second Prandtl, t; z BL region, we have the opportunity to investigate the
quasi-steady evolution of the temperature accident arising from the Rayleigh
corner fourth region.
2
In [106], with Guiraud, we have formulated, for the primitive Kibel equations (see Sect. 9.2)
which are derived from the hydrostatic approximation to the Euler equations for non-viscous and
adiabatic motion a problem analogous to the one that was considered by Rossby (1938)
concerning the quasigeostrophic approximation (a problem which is now well known as the
adjustment to geostrophy). The major conclusion of our adjustment to hydrostatic balance is
that the initial conditions for the primitive equations may be derived from a full set of initial
conditions, for the full Euler equations, where in these Eulerian initial conditions the initial data
need not fit the hydrostatic balance. This obtention of initial conditions for primitive equations is
realized by solving the associated one-dimensional unsteady adjustment problem of vertical
motion to hydrostatic balance.
Curiously, however, this unsteady inviscid adjustment scenario, from gas

dynamics to BL, does not seem to be the only one possible. Indeed, a more detailed
analysis (see Sect. 5.5.2 below) shows the existence of a fifth matching region
between Rayleigh (fourth) and Prandtl (second) regions the existence of such an
intermediate fifth region ensuring matching between the Rayleigh corner and
Prandtl BL regions.
5.5.2 Adjustment Process Via the Rayleigh Equations
This fifth intermediate matching region appears when we investigate, first, the far
behaviour of the Rayleigh equations (5.29) and (5.30a5.30d), for large values of y
and Z. In the Rayleigh corner, fourth region, with Prandtl variables,
t e2 y and z eZ as e # 0; (5.33a)
we can introduce new intermediate variables, t* and z*:

p
y t =ke and Z z = ke; 0 < ke # 0 with e # 0; (5.33b)
where ke is an arbitrary gauge, and t* and z*, the new intermediate variables, are
fixed when e # 0, in such a way that both Rayleigh variables y and Z tend to infinity.
In this intermediate matching fifth region we have, as leading-order functions:
p
vInt ; wInt ; rInt ; TInt Limke#0 vRa ; wRa = ke; rRa ; TRa ; (5.33c)
where all intermediate functions (with subscript Int) are dependent on timespace
variables, t*, z*, and x, y.
When we take into account that the (unknown) gauge ke is certainly in an
order between e2 and e, such that we derive from the Rayleigh equations, (5.29) and
(5.30a5.30d), due to (5.33b, 5.33c), the following intermediate-matching model
equations for vInt, w*Int, rInt, pInt, and TInt:
@rInt =@t @rInt wInt =@z 0; (5.34a)
@pInt =@z 0; (5.34b)

rInt @TInt =@t wInt @TInt =@z g 1pInt @wInt =@z
(5.34c)
g=Pr@ 2 TInt =@z2 g g 1M2 j@vInt =@z j2 ;
rInt @vInt =@t w Int @vInt =@z @ 2 vInt =@z2 ; (5.34d)

5.5 Adjustment Processes Towards the Prandtl BL Evolution Problem 113
with
pInt rInt TInt : (5.34e)
Obviously, the above intermediate matching model equations, (5.34a5.34e),

first pointed out in our short note [98], are those derived when we carry out, on the
Rayleigh equations (5.29) and (5.30a5.30d), the usual approximations of the
classical Prandtl boundary-layer theory.
In particular, as in BL region, the pressure pInt is independent of z* and is
determined by matching:
limz "1 LimInt limZ#0 LimRa ; (5.35a)
where
p
LimInt e # 0; with t t=se; z z=e se; fixed; (5.35b)
the intermediate variables (t*, z*) being directly related to the starting (in NSF
equations) variables (t, z).
The intermediate gauge se, in LimInt, (5.35b), is linked with the above gauge
ke by the following relation:
ke se e2 : (5.36)
More precisely, this compatibility relation (5.36) is a direct consequence of the

investigation of the behaviour of the Prandtl BL equations (5.18a5.18e), when
t and z both tend to zero, towards the intermediate fifth region.
Indeed, if we write (again with t* and z* fixed):
p
t set ; z sez ; 0 < se # 0 with e # 0; (5.37a)
and if
p
vInt ; wInt ; rInt ; TInt Limse#0 vPr ; sewPr ; rPr ; TPr ; (5.37b)
then again we derive the same above intermediate matching model equations
(5.34a5.34d) with (5.34e), but from (5.18a5.18e). The relation (5.36) is, in fact,
a consequence of the compatibility between (5.33b, 5.33c) and (5.37a, 5.37b).
Unfortunately, the precise localization of this intermediate matching region
(characterized by the gauge se), between the Rayleigh and Prandtl regions,
does not seem possible at this stage of asymptotic analysis, and more careful
(second-order?) investigations are obviously necessary.
Finally, we observe that if on the one hand, when se e2 , then
t y and z z=e z=e2 Z, then we recover the Rayleigh region; and

if on the other hand, when se e0 1, then t* t and z z, and we recover

the Prandtl region.
The existence of such an intermediate region is a striking indication that it seems
possible (as a conjecture) to directly match the Rayleigh and Prandtl equations by
an adjustment problem, via the intermediate equations (5.34a5.34e), when the
intermediate time t #/ , using the matching condition:

limt #/ vInt ; TInt v0 Pr ; T0 Pr : (5.38)
This presents the possibility of obtaining, consistently, the associated initial data,
vPr0 and TPr0, in (5.22), for Prandtl unsteady equations (5.18b) and (5.18d) for vPr
and TPr.
5.6 Some Conclusions
First, it is clear that the above problem of matching, (5.38), deserves careful
consideration and may be an interesting numerical/computational problem.
We then observe that it is possible to considerably simplify the matching
problem between the Rayleigh and Prandtl equations, if we assume that the Mach
number M, in Rayleigh equations, is a small parameter, and assume, for this, that in
the wall the thermal condition (5.32b) can be written in the following form:
Yy; P 1 L0 M2 Sy; P; (5.39)
where L0 O1, and Sy; P replace thermal spot Yy; P.

In such a case, the solution of the Rayleigh problem is also expanded relative to a
low Mach number, M 1 (as in Howarths paper [99]). But here we do not
proceed further.
A third remark concerns the fact that further investigations are necessary for a
complete understanding of the above intriguing five-regions structure, which is very
interesting, because it is unusual and does not have an obvious clear interpretation!
But the above new five-regions (four regions plus the intermediate region) structure
of NSF equations, at large Reynolds number, as a consequence of the singular
nature of the unsteady Prandtl BL equations near the initial time, do not restrict
investigations to emergency phenomena, and have fundamental importance in the
RAM Approach of NSF equations.
I think that from this detailed further re-examination of boundary-layer Prandtl
theory, it is now possible to resolve some singularities arising in various unsteady
boundary-layer problems (see, for instance, Stewartson [101]).
A final remark concerns the pedagogical interest of such partition of NSF
equations, in five regions, for large Reynolds number fluid flows, and this RAM
Approach presents the possibility of deriving a new logical interpretation of Euler,
5.6 Some Conclusions 115
Prandtl, Rayleigh, acoustic/gas dynamics, and intermediate equations, as five

significant and particular models of full NSF unsteady equations for Newtonian
fluid flow at large Reynolds number.
Again we observe that not only Prandtl in 1904, but also (it seems) Blasius and
Schlichting (in Germany), Lagerstrom, Cole, and Kaplun (at Caltech), Van Dyke
(at Stanford), Stewartson and Smith (in England), and Germain (in France), did not
realize that indeed the concept of boundary-layer, which is an extension to long-
waves approximation in the case of a viscous fluid flow, is singular, in the case of an
unsteady fluid flow, near initial time, where initial data are prescribed in a well-
posed initial-boundary value problem.
Chapter 6
The Mathematics of the RAM Approach
6.1 Our Basic Postulate for the Realization of the RAM

Approach
As was determined at the outset, as a basis of our mathematics (below) for

realization of our RAM Approach, it is laid down that:
If a leading order, an approximate, model is derived from an NSF fluid flow
problem, then it is necessary that a RAMA be adopted to make sure that terms
neglected, in such a full unsteady NSF stiff problem, really are much smaller than
those retained in derived approximate simple (but consistent), no-stiff leading-
order, model problem.
This above basic postulate a statement that is accepted as true, despite its
simplicity is in fact a very good and suitable general rule a statement of what the
reader is advised to do in various modelling situations. (See, in particular, Chaps.
79 below, on applications of the RAM Approach.)
Until this postulate is applied to the derivation of approximate/rational and
consistent models for technologically or geophysically/atmospheric stiff problems,
it will be difficult to convince the (possibly sceptical) reader (as noted in Germains
The New Mechanics of Fluids of Ludwig Prandtl (2000) [107], p. 34) of its value
as a guide-line for our RAMA, on behalf of numerical/computational simulation.
Massive computations1 are actually capable of bringing so much to our under-
standing, but there seems to be no indication that they are in competition with our
RAMA, Both are useful and complementary, and, what is more, I think that the
RAMA, which is a new way of looking at the derivation (in place of a doubtful ad
hoc approach) of the approximate rational model problem on behalf of numerical
1
Recently, a new Intensive Computation Centre was inaugurated at Bruyeres-le-Chatel (Essone),
France, and in 2011 the new super-high-speed computer Curie capable of making 1.6 million
of billion operations per second should be fully operational!

118 6 The Mathematics of the RAM Approach
simulation, is the only one rational way to obtain a non-ad hoc consistent, simplified,
non-stiff, well balanced model problem from the full unsteady NSF stiff problem.
Obviously: the ultimate goal (especially for the numericians in the framework of
their simulation processes) is to find the mathematical key which explains the
success of the modelling used in industrial and technological applications.
In theoretical fluid dynamics, perhaps the most significance advance in research
and understanding during the last 50 years has been the use of asymptotic
techniques in order to settle, on a rational basis, a number of approximate models
which were often, much earlier, derived by the various and often very questionable
ad hoc procedures.
Actually, it might have been clear that asymptotic techniques were well suited
for deriving mathematical models amenable to numerical treatment, rather than for
obtaining approximate closed forms of solutions.
It is now also evident that asymptotic techniques are very powerful tools in the
process of constructing well argued, non-contradictory, consistent, and simplified
mathematical models for problems which are stiff, from the point of view of
numerical analysis.
In the framework of our RAMA, the main goal is the modelling, and not the
finding of solutions, via the derivation of rational consistent models from an initial-
boundary value problem formulated for a real fluid flow.
This RAMA has, as its main objective, the generation of simpler adequate
models amenable to numerical simulation. But in most cases an efficient method
of achieving this goal will be to retain more the spirit (the real meaning and pur-
pose) of asymptotic techniques rather than their complete and formal structure
nothing being said of the rigorous mathematics.
Concerning our mathematics for the realization of our RAMA, we have in
mind, first, the above formulated postulate and also several key steps which play a
decisive role during the rational derivation of consistent models (as clearly
indicated in Chap. 4, with the justification of Boussinesq equations, and also in
Chap. 5, in the framework of the five-regions structure of unsteady NSF equations
at large Reynolds numbers).
Concerning, precisely, the role of our basic postulate, we observe also that the
knowledge and skill that I acquired, with J.-P. Guiraud, during 19701980, working
within the framework of the RAMA, presents the possibility, thanks to this postu-
late, of deriving, in various cases, the dominant concepts and ingredients of a full
solution of the starting fluid flow stiff problem.
In fact, fluid dynamics inspired by this basic postulate is a new way of approaching
classic fluid flow problems via NSF full unsteady 3D equations which is very useful
in gas dynamics (from hyposonic to hypersonic), compressible aerodynamics, hydrody-
namic instability, technologically and geophysically (atmospheric) interesting (but very
stiff) flows, and various thermal and thermocapillary convection problems.
The necessity of such a basic postulate is obvious. Computerized numerical
simulation using a mathematical model created artificially in order to study what
could exist in a real and very complicated fluid flow is a very expensive activity
(costing a lot of money). Therefore, as a consequence, confidence is necessary
6.1 Our Basic Postulate for the Realization of the RAM Approach 119
concerning the consistency of the model used and the abilities of this model as a
valuable substitute for reality!
Unfortunately, the models derived by numericians and applied mathematicians
using ad hoc procedures obviously do not have the essential qualities mentioned
above! The cause and effects of such a situation are linked with the appearance of
internal inconsistencies in most of the relevant intensive engineering computations
based on relatively ad hoc models.2
On the other hand, applied mathematicians, who are interested in the application
of their rigorous results, based on abstract (modern) non-linear functional analysis,
have always found fluid mechanics to be a rich and interesting field, because the
basic equations (NSF equations) have an almost unlimited capacity for producing
complex solutions that exhibit unbelievably interesting properties.
Unfortunately, these rigorous investigations (proofs of the existence and unique-
ness of solutions) are based on relatively ad hoc and very simplified approximate
models which are rife with various internal inconsistencies. This is the case, as far I
know, in France especially, in works relative to the incompressible limit of com-
pressible fluid flows and in various papers devoted to models for oceanic and
atmospheric motions. Again, the question is: What is the scientific value of such
rigorous results based on inconsistent ad hoc models?
On the other hand, as observed by Germain in his Anniversary Volume (2000)
[108], p. 13:
Brilliant young physicists who began to be somewhat attracted by fluid dynamics and not
simply by hard physics, knew the usefulness of approximations and of non-dimensional
scaling. But they did not know that a systematic technique was available for building
approximate mathematical models and trying to measure quantitatively their validity . . .
I showed that the approximation is very often tied to the existence of a small parameter,
coming out from the non-dimensional form of the equations [concerning some facets of this
non-dimensionalization, see Sect. 6.3], and I intended to show that the process is sustained
by asymptotic singular expansions, and insisted on the methodology, in particular the
matching conditions and the concept of significant degeneracy [see Sect. 6.4.3] recently
created by Eckhaus (see [109]). I liked very much this last one, because it gives a systematic
2
I took the opportunity of expressing my (very critical) opinion during the conference debate,
organized by the Academie des Sciences, which took place in Paris on 29 June 2010. In his
presentation entitled Simulation by super-high-power computers: today and tomorrow, Olivier
Pironneau spoke mainly about the significance of numerical simulation and the increasing capacity
of super-computers, but did not mention, in any way, the role of fluid dynamics modelling for
simulation with the help of a consistent and rational model. The next three invited papers,
concerning numerical simulations of very interesting and difficult real problems (Earthquakes
at the planetary scale, Molecular structure of the living, and Advanced computations in
meteorology and climatology), are indeed of practical importance, and the results of simulations,
which were very attractively presented, gave a favourable impression but again without any
reference to the nature of the selected model and its relation with the starting real simulated
problem! It seems to me that Pironneau was not entirely satisfied by the lack of information
concerning the process of fluid dynamics modelling in the above-mentioned simulations. Eventu-
ally, he was asked the question one which requires serious thought How sure are you of your
results? All these numerical simulations will cost money.
way to find out what should be the various stretchings. I thought that this might be attractive
to physicists, because it is a quasi-systematic way of comparing the respective weights of
various terms in the equations, which measure the physical importance of phenomena they
are likely to describe.
A final comment concerns a quite curious recent aspect of the modelling for
hydro-aerodynamical problems conceived by applied mathematicians.
The latter to take into consideration initially a very simple single equation or
two coupled simple equations shows various numerical results (elegant and
illuminating) corresponding to a fluid flow over, for example, a simple hollow or
a rectangular obstacle. Then, in these single or coupled equations the numerician
adds, step by step, various terms which simulate (very naively) the influence of the
viscosity, compressibility, gravity or non-linearity, and possibly other physical
effects, numerically computing the corresponding fluid flows. As the final result
we have a family of fluid flows which possibly mimic a series of realistic flows.
Again, the question is: What is the scientific interest of such numerical
investigations in relation to the simulation of realistic fluid flows?
6.2 The Mathematical Nature of NSF Equations:

A Fluid-Dynamical Point of View
Rigorous proof of the existence, uniqueness, smoothness, and stability of solutions

of problems in fluid dynamics are needed to give meaning to the equations (NSF
equations) and corresponding initial and boundary conditions that govern these
problems.
Their formulation (see Sects. 2.3 and 2.4) shows conclusively that the NSF sys-
tem of equations is a closed system, and that with the assigned initial and boundary
conditions they produce certainty, relative to the well-posedness of problems in fluid
dynamics, that these equations are a numerical solution. Therefore, simulation via
the RAM Approach is required indeed, rigorously, as for any arbitrary reasonable
choice of a class of admissible initial data, a problem in fluid dynamics must be well-
posed (in the Hadamard sense). This means ([47], Chap. 8) that:
(a) The problem has a solution for any initial data in this class.
(b) The solution is unique for any initial conditions.
(c) The solution depends continuously on the initial data.
Obviously, this well-posedness is strongly linked with the mathematical
nature of NSF equations as a partial differential system of equations:
Roughly speaking, we can expect that, in particular, the equations for a viscous
fluid flow (in a simple, incompressible, case of the Navier equations) are parabolic,
and that the equations for an inviscid, non-viscous, adiabatic fluid flow (the Euler
equations) are hyperbolic.
However, a more detailed analysis of the structure of full NSF unsteady
equations shows that this above conclusion is not quite correct (at least for a
6.2 The Mathematical Nature of NSF Equations: A Fluid-Dynamical Point of View 121
compressible, viscous, and heat-conducting (NSF) fluid flow). Indeed, the equa-
tion of continuity with respect to density r is always, for a compressible fluid flow,
hyperbolic even for non-trivial viscosity or heat conduction!
Thus, to be more exact, we can say that a system of full unsteady
NavierStokesFourier equations is hyperbolicparabolic or incompletely para-
bolic, according to the definitions suggested in [59] and [62] in the study of the
mathematical properties of these equations. Concerning, more precisely, the linear
incompletely parabolic systems see the paper by Gustafsson and Sundstrom [67]. In
the linear case (when the coefficients of viscosity and heat conduction are constant)
the NSF system becomes:
@U=@t Ak @U=@xk Ljk @ 2 U=@xj @xk ; (6.1)
where Ak et Ljk are constant square matrices, and U (u1, u2, u3, T, r).
If we consider the following decomposition for U:
U u1 ; u2 ; u3 ; T; 0 0; 0; 0; 0; r UI UII (6.2a)
then in a such case we write, in place of (6.1), a system of two coupled equations:
@UI =@t Ak;I;I @UI =@xk Ak;I;II @UII =@xk

(6.2b)
Ljk;I;I @ 2 UI =@xj @xk ;
@UII =@t Ak;II;I @UI =@xk Ak;II;II @UII =@xk 0: (6.2c)
That is just the system (6.2b, c) of two coupled equations for UI and UII, which is
called incompletely parabolic if, on the one hand,
@UI =@t Ljk ;I;I @ 2 UI =@xj @xk (6.3a)
is parabolic, and if, on the other hand,
@UII =@t Ak;II;II @UII =@xk 0 (6.3b)
is hyperbolic.
In two of my survey papers (1999) [79] and (2001) [29] the reader can find
various theoretical results concerning the NSF system of equations. In addition, in
two books referring to non-viscous (2002 [37], Chap. 9) and viscous (2004 [47],
Chaps. 810) fluid flows, the reader can also find a fluid-dynamical point of view
relative to the well-posedness, existence, uniqueness, stability, turbulence, and
strange attractors for NSF equations.
Concerning, first, the rigorous mathematical results for the Navier incompress-
ible system of equations, see, in Temam (2000) [83], pp. 10491106, a discussion
of the development of NavierStokes (incompressible, in fact, Navier) equations in
the second half of the twentieth century, initiated by Leray in 1933 [84]. Lerays
first rigorous work, published in 1931, was only in fluid dynamics, and concerns
only the Navier equations. He proved the basic existence and uniqueness results
[84] for the Navier, viscous, and incompressible equations, and Temam, in [110],
has expressed the view that: No further significant rigorous work on Navier
equations was done until that of Hopf in 1951.
The interested reader can find in No. special de la Gazette des Mathematiciens
(2000) a tribute to Jean Leray (19061998) various papers by some well-known
mathematicians and, in particular, the paper by Chemin [111], pp. 7082, which
presents a remarkable summary via an illuminating analysis of chapters of the well-
known Leray paper of 1934 [84]: Sur le mouvement dun liquide visqueux
emplissant lespace, published in Acta Mathematika.
This magnificent paper by Leray was the principal stimulation (with two com-
panion papers published in Journ. de math. pures et appl., vol. 12, 182, 1933, and
vol. 13, 331418, 1934) for several of his own papers on the uniqueness, stability,
and regularity of viscous and incompressible fluid motions, and indeed was the
inspiration for a vast modern literature on the subject.
A particular place in the rigorous mathematical theory of viscous
imcompressible fluid flows is also undoubtedly deserved by Olga Aleksandrovna
Ladyzhenskaya (see, for instance, the paper devoted to her on her eightieth birth-
day, in Russian Math. Survey, vol. 58(2), 395425, 2003) who, with the numerous
papers and books published on the Navier equations problem during 19572002,
has played a particularly important role, having a considerable influence on the
development of the mathematical theory of Navier problems.
A typical a la Ladyzhenskaya result is the following major theorem: Suppose
that V is a generalized solution of the Navier unsteady problem:
Vt nDV V:rV rp f ; divV 0; V jt0 ; V j@OO;T0 (6.4)
in QT O [O, T] that belongs to the class Lm,n(QT) with m, s satisfying one of

the conditions:
n=m 2=s 1; m 2 n; /; s 2 2; / (6.5a)
and
m n e; e>0; s / : (6.5b)
If f 2 L2 QT ; V 2 HO W 1 2 and O is a bounded domain in 3 with

@O C2 , then V belongs to the space W22(QT), the corresponding pressure p is
such that rp 2 L2 QT , and (V, p) satisfy all the conditions of problem (6.4).
Thus, Ladyzhenskayas above theorem guarantees that the solution V has
generalized derivatives Vx, Vxx, and Vt in L2(QT), and that V and the corresponding
pressure, p, satisfy the Navier system for almost all x; t 2 QT :
6.2 The Mathematical Nature of NSF Equations: A Fluid-Dynamical Point of View 123
It is clear that under the given conditions on f and V the solution cannot be
smoother. However, if the smoothness of f and V increases, then V and p also
become smoother, as already follows from the results of the linear theory described
by Ladyzhenskaya (1961) [112] (English translation). Several mathematicians have
essentially repeated this theorem but unfortunately without any reference to
Ladyzhenskaya!
The significance of Ladyzhenskayas theorem is indisputable. It provides a good
guideline for those who wish to try themselves in solving the sixth problem of the
millennium, announced by the Clay Mathematics Institute of Cambridge.3
However that may be today, despite tremendous progress in many mathematical
aspects of fluid flow theory since Lerays pioneering thesis/paper of 1933, it is
necessary to be just a little more modest if we have in mind the (large) number of
problems that still remain open! For instance, we have not determined, until now (in
the framework of Navier problem), whether a solution that is initially smooth can
develop a singularity at some (finite) later time, or whether singularities are a
fundamental feature of turbulence.
In spite of the fact that Navier viscous and incompressible equations are a very
simplified form of full unsteady NSF equations, they are useful because they
describe the physics of many things of academic and economic interest. They
may be used to model the weather, ocean currents, water-flow in a pipe, air-flow
around a wing, the motions of stars in a galaxy, and so on. Despite their simplified
(relative to NSF equations for real compressible and heat-conducting fluid flow)
forms, they help with the design of aircraft and cars, the study of blood flow,
the design of power stations, the analysis of pollution, and many other things.
Coupled with Maxwells equations, they can be used to model and study
magnetohydrodynamics.
The Navier equations are also of great interest in a purely mathematical sense.
Somewhat surprisingly, given their wide range of practical uses, mathematicians
have not yet proven that (what are called Navier existence and smoothness
problems) in three dimensions solutions always exist (existence), or that if they
do exist, then they do not contain any singularity (smoothness).
The Clay Mathematics Institute has called this one of the seven most important
open problems in mathematics, and has offered a prize of $1 million for a solution
or a counter-example. The official problem description is provided by Ch. L.
Fefferman on the website at http://www.Claymath.org/millennium/.
As far as the numerical solution of the unsteady incompressible NavierStokes
(Navier) equations are concerned, Quartapelle [113] presents a unitary view of the
3
In Russian Mathematical Surveys, vol. 58(2) (2003), pp. 25186, there is a paper by O. A.
Ladyzhenskaya The sixth problem of the millennium: NavierStokes equations, existence, and
smoothness which presents the main results concerning solubility of the basic initial-boundary
value problem and the Cauchy problem for the three-dimensional non-stationary Navier (incom-
pressible) equations, together with a list of what to prove in order to solve the sixth problem of the
seven problems of the millennium proposed on the Internet, at http://www.Claymath.org/.
methods which reduce the equations for viscous incompressible flows to a system
of second-order equations of parabolic and elliptic type. Concerning the rigorous
mathematical results for NavierStokes compressible equations for viscous and
heat-conducting fluid flows (denoted, in this book, as NavierStokesFourier,
NSF, equations), the mathematically rigorous theory is actually still in its infancy.
In our (2004) [47], Chap. 8, the reader can find more recent references concerning
the published papers and books devoted to NSF equations. In 2004, two books
were published by Feireisl [114] (dynamics of viscous compressible fluids) and
Novotny and Straskraba [115] (introduction to the mathematical theory of com-
pressible flow).
Actually, it is firmly established that we are able to prove an existence theorem
for compressible viscous (but baroptropic NS equations) fluid flows, which is
global in time, without assuming smallness of the data, although uniqueness is an
open problem (the solution is weak). On the other hand, global existence of a
strong (each term in NS equations exists as an element of Hilbert space, at least)
and regular solution is proven, but assuming that the data are small enough and that
in such a case the uniqueness hold.
6.3 Formulation of Dimensionless Equations for Applications

of the RAM Approach
The non-dimensional approach of NSF equations, realized in Chap. 3, is a first key

step in our Mathematics for a possible application of the RAM Approach, thanks to
the presence in these dimensionless equations and conditions of the starting
problem, at least of a small (or high) parameter.
If this non-dimensionalization does not often an easy approach, the judicious
formulation of dimensionless equations is a very decisive step for the application
of the RAM Approach. Below I present a few examples of such distinctive
formulations for some basic fluid flow problems.
6.3.1 Turbomachinery Fluid Flows
First, one cannot escape the fact that in a turbomachine the blades of a row are
usually very closely spaced. As a consequence, as basic small parameter e, we
choose the reciprocal of the number,
N 2p=e>>1; (6.6)
of blades encountered along the periphery of a row. This small parameter is a

geometrical one, and does not appear in fluid dynamics equations. In a very simple
6.3 Formulation of Dimensionless Equations for Applications of the RAM Approach 125
case, for a 3D steady incompressible non-viscous fluid flow through a row in the
turbomachine (an axial compressor, assuming that the row is localized between two
infinitely long coaxial cylinders).
Obviously, when far upstream of the row, velocity is uniform and constant and
we have irrotationality, and the basic equations are very simple. For the velocity
vector u(u, v, w), we write:
r:u 0; (6.7)
and
r ^ u 0; (6.8)
where
r @=@rer 1=r@=@yey @=@zez ; (6.9)
in a system of (r, y, z) coordinates, such that the axis z is the axis of the
turbomachine. The row is between r R and r r . If we want to use the
smallness of the e 2p/N, in an asymptotic theory, obviously it is necessary to
make a change of variables, such that e<<1 appears in the above two Eqs 6.7 and
6.8. But, at once, we observe that Eq. 6.7 is integrated if we write for the velocity
vector u:
u rew ^ rc with u:rw 0 and u:rc 0; (6.10)
and the two stream functions w and c are of order one (the output in a channel
between two consecutives blades being of the order e).
We assume that the surfaces
y Yt; r; z (6.11)
are the blade skeletons in a row, when e # 0, and outside of the row are material
surfaces that are the extensions of blade skeletons.
By
Dr; z; (6.12)
we denote the breadth of the channel from blade to blade, with Dr; z 1 outside
the row. In the row, between two consecutives blades, we can write:
y = Yr; z e=2Dr; z; (6.13a)
and
y = Yr; z e=2Dr; z: (6.13b)

As a consequence of the above, with an error (at most) O(e2), we have the
possibility, for e<<1, to write in the blade-to-blade channel the following change
for y:
y Yt; r; z eDr; zw ) ew y Yt; r; z=Dr; z (6.14)
This relation allows us to write, in place of (6.10), for the velocity vector
u(u, v, w) :
u r y Yt; r; z=Dr; z ^ r cr; Y; z; (6.15)
the equation r:u 0, and the slip condition, for a non-viscous fluid on the blade,
being both automatically satisfied with (6.15).
Now, if we make the change from r; y; z tor; Yt; r; z eDr; zw; z,
then:
ur; y; z ) u r; w; z; e; cr; y; z ) c r; w; z; e; (6.16a)
and
@c=@r @c =@r 1=D@c =@w 1=e@Y/@r w@D=@r; (6.16b)
@c=@z @c =@z 1=D@c =@w1=e@ Y=@z + w@D=@z; (6.16c)
@c=@y 1=eD@c =@w (6.16d)
and also, from (6.15):
u 1=rD@c =@z; w 1=rD@c =@r (6.17a,b)
v ru @Y=@r ew@D=@r rw @Y=@z ew@D=@z: (6.17c)
Finally, it is necessary to use the equation (6.8), r ^ u 0, for a derivation of

well-adapted non-dimensional starting equations for our RAM Approach to turbo-
machinery flow.
Taking into account the above relations (we observe that r v G is the
circulation), we derive, in place of (6.8), as a starting non-dimensional system of
equations (unless any simplification, for an application, in Sect. 7.1), for u*, G* and
w*, the following system of non-dimensional equations (where the small parameter
e is present):
1=rD@w =@w @Y=@z@G =@w

(6.18a)
e=r@G =@z e=rw=D@D=@z@G =@w
1=rD@u =@w @Y=@r@G =@w

(6.18b)
e=r@G =@r e=rw=D@D=@r@G =@w
1=D@Y=@z@u =@w @Y=@r@w =@w

e@u =@z @w =@r (6.18c)

ew=D@D=@r@w =@w @D=@z@u =@w
(Concerning my work with Guiraud, related to various facets of an asymptotic

theory for the turbomachinery fluid flow, see some commentaries in [11], pp. 16
and 17.)
In Sect. 7.1 the above system of Eqs. 6.18ac is analyzed in the case when e
tends to zero, and a model problem is derived for the through-flow model.
6.3.2 The GZ Rolled-up Vortex Sheet Theory: Vortex Sheets

and Concentrated Vorticity
Vortex sheets are one of the basic ingredients of non-viscous fluid flows, which can
be viewed as vorticity concentrated on a core of small but finite diameter. They can
be considered as two-dimensional surfaces of zero thickness carrying a truly
concentrated vorticity.
We clearly see that as far as a compressible fluid is concerned, concentrated
vorticity is not the entire matter. We should add a concept of concentrated
baroclinity vector B rS, where S is entropy, by considering that along any
surface orthogonal to B, variation of p and r, at constant time, are related by a
barotropic relation. Then, by writing
B BS dS jSjn BS BC (6.19a)
we exhibit a concentrated baroclinicity BC (dS is the Dirac distribution uniformly

spread over S and |[S]| the jump in S when S is crossed in the sense of unit normal
n) which is orthogonal to the concentrated vorticity, and we may state that

j1=rj @ T=@pjSj (6.19b)
where T/p is calculated at p, and S a S + (1 a) S with 0< a<1.

The situation that we want to describe is one in which there are very many vortex
sheets closely spaced. We may mention two flow configurations in which this
occurs. The first one is the core of a highly rolled vortex sheet, while the second
concerns the set of trailing vortex sheets which are formed at the trailing edges of
the blades in a row of an axial turbomachine.
We assume that there is a small parameter built into the flow, which is the ratio
of the spacing between two consecutive sheets to the width of the region covered by
the sheets. We set C for this parameter, and following our G-Z paper (1977) [116]
we call it a closeness parameter.
The purpose of the RAM Approach, in this situation, is to derive a model which
avoids the stiffness of the problem, for a computation of the flow with a numerical
code capable of capturing many sheets.
The problem is one of multiple scaling, in the terminology of asymptotics.
A version of the multiple-scale technique (see Sect. 6.4.4) especially suited for this
kind of problem was devised with Guiraud with the purpose of describing rolled
vortex sheets, but its scope is more general, as may be seen from the analysis which
follows.
We start from the assumption that the velocity u, pressure p, density r, and
entropy S all suitably non-dimensional are functions of time t, position x, and of
one fast variable C1 wt; x, with rw approximately orthogonal to the sheets. We
use the notation:
U u; p; r; ST (6.20a)
and set
Ut; x U t; x; C1 wt; x (6.20b)
There are two ingredients in the technique used. The first one is a formal
expansion,
U U 0 CU 1 ::: (6.20c)
while the second one is the obvious observation that setting
C1 @w=@t y0 C y1 ::: (6.21a)
k rw k0 Ck1 ::: (6.21b)
we have
@U=@t C1 y@U=@w @U =@t (6.22a)
rU C1 @U=@wk rU (6.22b)
Substituting the above relations (6.20c)(6.22b) into the equations of motion for a
compressible non-viscous fluid, we obtain, at zeroth order, a set of equations from
which we conclude that, provided w const is not a Mach wave, we must have that
y0 k0 :u 0 does not depend on w:

On the other hand, by specifying that w is constant on the vortex sheets we

obtain
y0 k0 :u 0 0 (6.23a)
and as a consequence
k0 :@u 0 =@w @p 0 =@w (6.23b)
Then, for the vorticity we obtain
v C1 k ^ @u =@w r ^ u
so that if we ensure that v* is O(1), then, as was the case in the study of rolled-up
vortex sheets, we have the relation:

k0 ^ @u 0 =@w 0
which in turn enforces that (for density and specific entropy):
@r 0 =@w 0 and @S 0 =@ w 0 (6.23c)
Now, the model of rolled vortex sheets is recovered by assuming that U* is 2p-
periodic with respect to w. Then, by writing the equations of motion and for
vorticity at order one, it may be concluded that V k0 u*1 and S S*1 are
solutions of the following two equations:
@V=@t u 0 :rV V:ru 0 r:u 0 V

n o
k0 : r ^ u 0 =jk0 j2 k0 ^ V (6.24a)
1=gM2 fk0 ^ rp 0 =g r 0 g S 0
@S=@t u 0 :rS fk0 ^ rS 0 =jk0 j2 g:V 0 (6.24b)
The important point is that w does not occur in (6.24a, b). They are a set of
ordinary differential equations along the trajectories of the velocity field at zeroth
order approximation u*0. Via (6.24a, b), a number of configurations are allowed
with this description.
In our paper with Guiraud (1982) [117], two of them are presented. The first is an
extension of the Kaden problem, and the second concerns Mangler and Webers
solution.
We observe that the closeness parameter C is related often, but not always to
by C s2 , to a slenderness parameter (s) with respect to which u*0 may be
expanded. However, we emphasize here that obtaining a closed form solution should
not be considered as the ultimate goal of the analysis. In our opinion, the analysis
should rather be considered as a model allowing us to generate a flow with closely

spaced vortex sheets, rather than another one which does not involve such a feature!
Again concerning my work with Guiraud relating to various facets of an
asymptotics of rolled vortex sheets, see some commentaries in [11], pp. 1819.
6.3.3 The GZ Asymptotic Approach to Non-linear

Hydrodynamic Stability
The problem of hydrodynamic stability may be posed in the following terms.

Let u0 be a particular steady (independent of time t) solution of Navier incom-
pressible and viscous equations, written as:
Sdu=dt Fu and such that Fu0 0 (6.25a)
If u u0 + v, in another solution, then the problem of the stability of the

solution u0 is strongly related with: whether or not v remains small all the time,
provided that it is small initially?
Obviously, the stability has a direct relation with the third condition, (c), in
Hadamard well-posedness (mentioned at the beginning of Sect. 6.2). For the
investigation of such a stability, we consider the following evolution equation for v:
dv=dt Lu0 v Qv; v; with u u0 v (6.25b)
where L(u0) is a linear operator, depending on u0, and Q(v, v) is a quadratic

operator, such that Qav; av a2 Qv; v; a const. But for any v and w:
Qv w; v w Qv; v Mv; w Qw; w
where M(v, w) is bilinear and is a function of two variables.

We observe that if we take the inner product < .; . > of (6.25b) with v, we derive
from (6.25b):
@jvj2 =2=@t < Lu0 v; v> (6.26a)
because < Qv; v; v> 0, in the case of the Navier problem.

If we now let
h i
g inff< Lu0 v; v>= jvj2 =2 g; subject to the
(6.26b)
constraint; div v 0 and v 0 on @D
This above variational problem (6.26b) is similar to the classical variational

characterization of the first eigenvalue of the Laplacian. Consequently, from
(6.26a), we can write:
@jvj2 =2=@t g jvj2

0
Integrating this inequality, we obtain:
jvtj
jv0jexp2gt; for g >0
u0 is unconditionally stable
6.3.3.1 Confined Perturbations and the LandauStuart Equation
We assume that u is defined on a bounded spacial domain and, as a consequence,

L(u0) has (one may prove) a discrete spectrum; that is, there exists an infinite
sequence of eigenvalues, in fact complex:sn , n 1, 2,. . . with associated
eigenfunctions vn, such that
Lu0 vn sn ; and the equation Lu0 v s f (6.27)
with a given f, is uniquely solvable provided s is distinct from all sn .

Let sn ln imn , and we are interested here mainly by the unstable case but
restrict ourselves to weak instability. For this, we partition the indexes n into two
parts N and A (according to Guiraud (1980) [118]):
n 2 N ) ln Oeq and n 2 A ) ln <0 (6.28)
where e is a small parameter and q an exponent that is introduced here for later
convenience.
We must comment on the partition (6.28), which we have introduced as working
hypothesis to be checked in each particular case.
The flow u0 depends on parameters (for example, a Reynolds number Re),
and when Re < Rec, the flow configuration is stable, whereas it is unstable for
Re > Rec.
By continuity, this corresponds to the fact that when Re crosses Rec increasing,
some eigenvalues sn, those corresponding to n 2 N, cross the imaginary axis from
left to right.
When Re is close to Rec on either side, the
ln corresponding to indexes n 2 N are small
and this is expressed through the first relation of (6.28). The most frequent are those
when N has just one element with sn real or two elements with a complex
conjugate of sn , sn .
We now start from a normal-mode decomposition, and according to above

partition (6.28),
v Sn2N Cn vn Sn2A Cn vn X Y (6.29a)
and we set, X P*v, Y P**v, with P* and P** two projection operators. Then,
from (6.28), we see that
n 2 N ) sn eq ln0 in (6.29b)
and taking (6.29a) into account, we obtain the following system of two bifurca-
tion-type equations for X and Y in place of non-linear evolution equation (6.25b):
dX=dt eq L1 X L2 X P QX Y; X Y (6.30a)
dY=dt LA Y P QX Y; X Y: (6.30b)
It may be stated that
eq L1 corresponds to eq ln 0
L2 to in
both for n 2 N, and
LA corresponds to sn
for n 2 A.
Below, we shall concentrate only on the case when
P QX; X 0
and in such a case, when we ignore the transient unsteady-state phase t O(1), to
concentrate on the non-linear phase, for long time, and use a multiple-scale
technique with two times, t and
t eq t; X er X and Y e2r Y : (6.31)
The adequate choice being q 2 and r 1, from (6.30a, b) we obtain for X*

and Y* the following system (with two times):
@X =@t L2 X e2 @X =@t L1 X P MX ; Y Oe3 (6.32a)
@Y =@t LA Y P QX ; X Oe (6.32b)
with
X X 0 eX 1 e2 X 2 :::; Y Y 0 eY 1 ::: (6.33)
we obtain from (6.32a, b) the following leading-order equations:
@X 0 =@t L2 X 0 0; @X 1 =@t L2 X 1 0; (6.34a)

@Y 0 =@t LA Y 0 P Q X 0 ; X 0 (6.34b)

@X 2 =@t L2 X 2 @X 0 =@t L1 X 0 P M X 0 ; Y 0 (6.34c)
The solution for X*0 is written in the following form :
X 0 Sn2N X 0;n texpin t
and for Y*0 we obtain as a solution :
Y 0 Sp2N Sq2N Y 0;pq texpip q t
where Y*0, pq(t) is a function of X*0, p(t) and X*0, q(t) and LA.
It is now necessary to consider Eq. 6.34c for X*2, seeing that we want to
determine the dependence of X*0, n(t) relative to the long time t. With the above
relations, in place of (6.34c), we have the following equation for X*2:
@X 2 =@t L2 X 2 Sn2N @X 0;n =@t L1 X 0;n expiZn t

Sp2N Sq2N Sr2N P M X 0;p ; Y 0;qr expip q r t
0 (6.35)
In the rather complicated Eq. 6.35, according to the multiple-scale technique (see,
for instance, Sect. 6.4), it is necessary to eliminate secular terms in solution for X*2,
if we want the expansion (6.33) to be uniformly valid with respect to t and also t,
ignoring the transition phase. The result of this elimination is the system of ordinary
differential Eq. 6.36, which determines the coefficients X*0,n(t):

dX 0;n =dt L1 X 0;n SR n P M X 0;n ; Y 0;qr (6.36)
In (6.36), SRn represents the sum over resonant (p, q, r) triplets, where Rn is a
resonance condition:
Rn ) p q r n
The above derived Eq. 6.36 is, in fact, very similar to an a la Landau equation.
When N contains only two elements, the corresponding sn being complex

conjugates, X* is well determined by a single complex amplitude St, which
must be a solution of the LandauStuart equation:
dSt=dt sS KSjSj3 0: (6.37)
6.3.3.2 Unconfined Perturbations and the DHS Evolutionary Amplitude

Equation
We observe that when dealing with modes spread over a continuum, the extension
of the Stewartson and Stuart (1971) [119] technique to the whole of temporal
evolution is not at all obvious, and the difficulty has its root in the fact that, to
our knowledge, no standard mathematically rigorous technique exits for dealing
with bifurcation from a continuous spectrum!
In (1978) [120], Guiraud and Zeytounian succeeded in filling this gap, in a
purely formal way, for continuously spread TollmienSchlichting (TS) waves that
exhibit close similarity to the technique used above for discrete modes. For exam-
ple, in [120] we have, with Guiraud, elucidated that: The process by which, from a
fairly arbitrary initial perturbation, the wave packet is first organized and then
evolves, is related to four time scales of evolution.
The first one, of O(1) duration, is devoted to the decay of all but the amplified
modes, and the second phase, of much longer duration, O1=e, is a passive one
with respect to the organization of the amplitude of the perturbation. The wave
packet is dominated by the most amplified of the T-S wave, and is convected with
the group velocity associated with the packet and the amplitude which is Oe4 ,
practically unchanged. During the third phase of duration O1=e2 the amplitude is
modulated according to an exponential law which predicted by linear theory.
Finally, it is only during the last period of duration OjLnej=e2 that non-linear
effects come into play, leading eventually to bursting, and to the well-known
envelope, evolutionary, equation of Davey, Hocking, and Stewarston (DHS),
discovered in 1974. Thus we derive, for a leading-order amplitude A t, the
following DHS equation:
@A=@t gA 1=2a@ 2 A=@x2 + b@ 2 A=@2 FA (6.38a)
with
FA k1 A j Aj2 k2 AB; (6.38b)
where k1 and k2 are constants, and B is a function solution of the equation:
@ 2 B=@x2 @ 2 B=@2 @ 2 j Aj2 =@2 : (6.38c)

In our (2004) [47], Chap. 9, the reader can find some aspects of the stability
theory of viscous fluid flow and, in particular, a detailed presentation of our (with
Guiraud) weakly asymptotic non-linear stability theory of Navier incompressible
fluid flow.
In De Coninck, Guiraud, and Zeytounian (1983) [121], a kind of unified theory is
presented in which the purely discrete and the continuous cases are treated by two
facets of a somewhat unique technique. Our main goal is the derivation of the
equations that rule the evolution of the amplitude of the most rapidly amplified
modes of linear theory.
A slightly different asymptotic modelling is applied, in [121], to convective
RayleighBenard instability. The distinguishing feature of this type of instability,
with respect to asymptotic modelling, is that the area of wave vectors for amplified
modes is within a circular annulus of small thickness. Our results show that instead
of finding a unique packet of waves one finds six of them organized around the
vertices of a hexagon which may interact quadratically with each other. As a matter
of fact, one may superpose an infinity of such hexagonal figures which may evolve,
to lowest order, independently.
6.3.4 A Local Atmospheric Thermal Problem: A Triple-Deck

Viewpoint
In the framework of an atmospheric thermal problem, on a flat ground surface,

z0 0, we have (in a 2D steady case) as a typical boundary condition:
T=T 0 1 t Yx=l ; jx=l j

1 (6.39a)
with a temperature parameter
t DT0 =T 0 (6.39b)
which is assumed a small <<1 parameter, where DT is a temperature rate for
the given function Yx=l .
In (6.39a), T*(0) is the temperature (at z0 * 0) in the hydrostatic reference state
(dependent only on the vertical coordinate z0 *), and l is the local horizontal length
scale.
Far upstream, when x !/ and Y 0, we assume that we have a basic
undisturbed flow which is characterized by an Ekman layer profile4:
UEk X; z=k UG Xf1 expz=k cosz=k g; (6.40a)
4
In Section 9.2, in the framework of the quasi-hydrostatic dissipative model, the reader can find
the derivation of the geostrophic relation and also Ackerbloms model problem.
where X x0 /L and z z0 /l , with
k 1=l O sinf =n 1=2 Rel =2Rol (6.40b)
the local Reynolds and Rossby numbers being based on the local horizontal length
scale: Rel U l /n and Rol U =l =O sinf .
If in the local thermal problem we non-dimensionalize the horizontal and vertical
coordinates with l for example Yx0 =l Yx then in the dimensionless local
problem there also appears a local Boussinesq number, Bol g l /RT0 *(0), and if l
103 m; then Bol << 1, and in such a case we also have Rel >> 2Rol >>1; k
being, below, our main small parameter (see the local Eqs. 6.42ae).
Therefore, in this case we can assume:
2Rol Rel 1=a ) k Rel 1=m (6.41a)

with (0 <a < 1)
m 2 a=1 a>2 (6.41b)
As an example, if U 10m=sec; n 5m2 =sec, and f 2 O sinf

10 1=sec, the considered case, l 103 m, leads to m 5. For such a case, we
4
have the possibility of using (M is the mach number):
l U =gRT0 0=g 1=2 ) Bol =M B 1 (6.41c)
and the Boussinesq approximation is correct.

Curiously, the value m 5 is the same as the one used by Smith and co-workers
for the flow over an isolated 2D short hump in the boundary layer.
The Boussinesq stratified fluid flow is also used by Sykes (1978) as a starting
equations for the application of a triple-deck theory. When m 5 we can show that
a typical triple-deck case exists:
l =L ReL 3=8 where ReL L U =n (6.41d)
but in our LNP (1987) [17], pp. 21120, the reader can find a more general
approach. Also in [19], Sect. 31, there are various references concerning the
application of the triple-deck theory in various meteo and environmental problems.
Now, according to the Boussinesq approximation, taking into account the above
relations, we have the possibility of formulating the following dimensionless local
steady 2D thermal problem, for the local velocity components (u, w), and thermo-
dynamic perturbations y and p:

u@u=@x w@u=@z 1=g@p=@x ko5 @ 2 u=@x2 @ 2 u=@z2 (6.42a)
u@w=@x w@w=@z 1=g@ p=@z 1=gB y

(6.42b)
ko5 @ 2 u=@x2 @ 2 u=@z2
@u=@x @w=@z 0 (6.42c)
u@y=@x w@y=@z B L0 w 1=Prko5 @ 2 y=@x2 @ 2 y=@z2 (6.42d)
o y; (6.42e)
when we perform the following limiting process:
t ! 0 with M ! 0 such that t =M t 1 (6.43a)
which gives the possibility of writing a relation for the reference velocity U :
namely,
U DT gR=T0 01=2 : (6.43b)
For our above local system of Eqs. 6.42ae as boundary conditions, we first write
the usual non-slip and temperature conditions:
z 0 : u w 0; y t Yx; 0< x <1 (6.44a)
because, in dimensionless Eqs. 6.42ac, according to Boussinesq approximation,

we have :

T Tz 1 My and p pz 1 M2 p :
A second condition for the local problem is an interaction condition between the
local thermal spot and the Ekman atmospheric layer (see (6.40a, b)) far upstream:
x ! /: u ! 1 expz=k cosz=k U/ z=k (6.44b)
w p y!0 (6.44c)
and we observe also that:
if z=k !/; then u ! 1; for x ! / (6.44d)
and
if z=ko ! 0; then u z=k ; for x ! / (6.44e)
A detailed account the triple-deck theory is included in our (2002) [26], Chap. 12.
In Sect. 6.4.6 we return to the above formulated problem (6.42a)(6.42e), with
(6.44a)(6.44e) in the framework of the triple deck theory.
6.3.5 Miscellanea
We now enumerate a few particular fluid flow problems showing the ubiquitous
nature of the small parameter. A first example, illustrating the occurrence of the
small parameter in the boundary conditions, concerns high-aspect-ratio wings. Here
the small parameter is the inverse of this aspect ratio, and it occurs in the model
when one enters into the details of the no-slip condition on the wing.
A more subtle occurrence concerns the nature of the domain of flow. The best-
known example of such a situation is provided by flow in thin films, and one may
find a quite large variety of applications. Here we mention just one of them: the
coating of thin wires with a film of very viscous melt which freezes before having
time to pour away.
The second type of situation corresponds to the fact that the small parameter is
built into the particular solution one looks for without being directly apparent in the
formulation of the problem.
The third type of situation occurs when the model suited for the physical
setting is substituted by another one. Then, in general, a small parameter occurs
in the relation between the two models (for example, in the theory of fluidfluid
interfaces with material properties, and also in the so-called moving contact
line problem).
The last type of situation that we consider briefly here is when two models are
considered for the same physical phenomenon, and when the coupling between
them involves a small parameter. A very broad field of applications of the idea of
asymptotic modelling may be included under this heading.
An interesting example, it seems to me, has to do with asymptotic and numerical
simulation, when what is really done, in numerical simulation, is the substitution by
a mathematical model involving, say, partial differential equations one of them
involving only algebraic ones!
The coupling between the two above models is characterized by a small param-
eter which is the ratio of the mesh, in numerical simulations, to the same
characterized length (or time) in the continuous model.
Whenever a numerical method is chosen for some problem, it more or less
involves asymptotic modelling.
6.4 Some Key Steps for the Application of the Basic Postulate
Several key steps which are often used and are of great value in the RAM
Approach are linked with the discovery of similarity rules these being a
necessary task because, in many cases (including in various places in the preceding
chapters) one encounters a double or multiple limiting process, in which two or
more parameters approach their limits simultaneously. Therefore, one must
6.4 Some Key Steps for the Application of the Basic Postulate 139
frequently specify the relative rates of such an approach, since the order of carrying
out these several limits cannot in general be interchanged.
In unsteady fluid flow problems, when it is necessary to take into account
given initial conditions at initial time (t 0), often the passage from the full
(assumed exact) starting dimensionless equations with given initial conditions to
a limit system of simplified dimensionless model equations is, in general,
singular near t 0! This singular nature is mainly expressed by the fact that,
often via the limit passage, some partial derivatives in time (present in the full
starting dimensionless equations) disappear in the derived system of model
equations, and as a consequence, it is not possible to apply all the initial data
at t 0 this derived system of equations being invalid (inconsistent) in the
vicinity of t 0.
We know that in such a case a short-time-scale, local time, rational analysis is
necessary. In the framework of our RAM Approach, the logical and rational way for
solving the associated local/short-time problem is the consideration of an unsteady
adjustment problem, in an initial time layer near t 0, which allows us to take into
account the (long-time) effect of the transient behaviour.
This unsteady adjustment problem presents the possibility of obtaining an
answer to following question: What initial conditions can be imposed to a derived
approximate model unsteady problem, and how are these initial conditions related
to the starting given data for full (exact) starting equations?
The possibility mentioned above, however, is obviously strongly linked with the
validity of the Matched Asymptotic Expansions (MAE) technique, and mainly with
the matching process and the concept of significant degeneracy (created by Eckhaus
(1979) [109]).
This last concept, we know, provides a systematic way of determining what
should be the various stretchings, and is a quasi-systematic way of comparing the
respective weights of various terms in the equations, which measure the physical
importance of phenomena that they are likely to describe a very efficient tool for a
constructive realization of our above postulate.
In the paper by Feuillebois and Lasek [122], this concept is applied to some
problems in fluid mechanics, and we observe that significant degeneracies are, in
fact, a mathematical formalism for the principle of least degeneracy defined by Van
Dyke (see [14], Sect. 5.5).
It is here worth including some remarks inspired by the Conclusion in Paul
Germains (2000) [107]: Quite often, the modelling of stiff fluid flow problems
may be found by various empirical procedures or by an ad hoc approach. But it
seems obvious to me that the ultimate goal is to find the mathematical key which
explains not only the success, but the validity and consistency of these procedures
in practice during the numerical simulation.
Actually, it seems me that our RAM Approach is an adequate procedure for such
a full realization. In [26] the reader can find a preliminary, tentative account of our
RAM Approach to various fluid flow phenomena the present book being a new
accomplishment of this major approach.
6.4.1 Similarity Rules: Small Mach and Large Reynolds

Numbers Flow
We now illustrate the significance of similarity rules in the RAM Approach when at
least two dimensionless parameters, Mach and Reynolds, tend respectively to zero
and infinity:
Re ! 1; M ! 0 (6.45)
In such a case, as in Sect. 4.2, the combined effect of vanishing viscosity and
very low compressibility is related with three possible limiting processes:
Re fixed; M ! 0; and then Re !/ (6.46a)
M fixed; Re !/ ; and then M ! 0 (6.46b)
M ! 0; and Re !/; simultaenously (6.46c)
such that:
1=Re l Mb ; with b>0 and l O1 (6.47)
where l is the similarity parameter (fixed, of the order one), and b a positive
scalar (constant) which must be determined during the RAM Approach.
As in the case of the justification of the Boussinesq equations (4.17ad), we have
that the third (6.46c) limiting process, with the similarity rule (6.47), is the more
significant, and presents the possibility of taking into account, simultaneously, two
small but important physical effects which, via coupling, lead to a finite effect, of
the order O(1), when
b > 0 is judiciously selected during the matching.
On the other hand, limiting process (6.46c) is the more significant because, first,
the limiting process (6.46a) is rediscovered when l !/ , and secondly, the
limiting process (6.46b) is rediscovered when l ! o.
For instance, if we consider the classical Blasius problem for a slightly com-
pressible flow, with a vanishing viscosity, then a consistent RAM Approach leads to
b 4; and with l 1 ) 1=Re=M4 1; (6.48)
and for details, see Godts and Zeytounian (1990) [123].

Here we observe only that the above similarity rule, (6.48), is obtained via
matching, which is linked with a least-degeneracy principle, and presents the
possibility of obtaining a consistent, leading-order solution for the incompressible
Eulerian outer flow.
6.4.2 The Matching Principle
The essentials of the matching principle are discussed by W. Eckhaus in his Matching
Principles and Composite Expansions a short (31-pages!) preprint (NR.42) issued
by the Department of Mathematics at the University of Utrecht. Here, however, we
do not enter into detailed analysis of this paper which is undoubtedly very interesting
as a rigorous reasoning of asymptotics for applied mathematicians (see also
Eckhaus (1979) [109]) and take into account only a few extracted remarks.
In the more recent paper by Eckhaus (1994) [124], the reader can find a review of
the foundations of the two mainstream ideas in matching.
The intermediate matching and the asymptotic matching principles and the
interrelations between the two procedures are discussed. It appears, first, that Van
Dykes intuition was correct. But the necessity of overlap seems doubtful!
Any intelligent practitioner of applied analysis will find a way to correct
matching in a given problem no matter what his convictions about the intrinsic
value of the overlap hypothesis!
Matching principles are the backbone of asymptotic analysis by MAE. In
applications of our RAM Approach, by means of these rules one determines, via
the above-mentioned similarity rules, various undetermined constants.
There are, indeed, two schools of matching. One of them (originated by Kaplun
and Lagerstrom) employs intermediate variables, and the other (originated by Van
Dyke) postulates the validity of an asymptotic matching principle. Matching in
intermediate variables can be deduced from the so-called overlap hypothesis; that
is, the assumption that the (extended) domains of validity of regular (outer) and
local (inner) expansions have a non-empty interaction.
In fact, what is most important is that for a large class of fluid flow problems,
overlap implies the validity of a generalized asymptotic matching. But overlap is a
sufficient but not necessary condition for the validity of an asymptotic principle;
and an asymptotic matching principle holds while there is no overlap at all!
In Van Dyke (1975) [14], Chap. V, the reader can find a detailed consideration of
various facets of matching, which is the crucial feature of the method of MAE
involving loss of boundary conditions.
An outer expansion cannot be expected to satisfy conditions that are imposed in the
inner region (near the initial time or in the vicinity of the wall of a body). Conversely,
the inner expansion will not in general satisfy distant (at infinity) conditions.
On the other hand, the possibility of matching rests on the existence of an
overlap domain where both the inner and outer expansions are valid. The existence
of an overlap domain implies that the inner expansion of the outer expansion
should, to appropriate orders, agree with the outer expansion of the inner expansion
(see, for instance, Lagerstrom (1988) book [125]).
A more usual, a la Van Dyke, matching principle is:
The m term inner expansion of the n term outer expansion

the n term outer expansion of the m term inner expansion;
where m and n are any two integers. In practice, m is usually chosen as either n or
n + 1. This matching principle appears to suffice for any problem to which the
MAE can be successfully applied.
In 1957, Kaplun introduced the concept of a continuum of intermediate limits,
lying between the inner and outer limits and in fact, such a case was considered in
Sect. 5.5.2 above.
But although the gap between inner and outer limits has been bridged by the
intermediate problem, it is not yet apparent that there exists an overlap domain!
However, this is assured by Kapluns well-known extension theorem, which
asserts: The range of validity of the inner and outer limit extends at least slightly
into intermediate range.
Thus we can match the intermediate expansion with the outer expansion at one
end of the range and with the inner expansion at the other end.
Finally, we can formulate an intermediate matching principle: In some
overlap domain the intermediate expansion of the difference between the outer
(or inner) expansion and the intermediate expansion must vanish to the appropriate
order.
As a conclusion of this above discussion concerning the matching principle in
MAE, we now present a simple example of our practical matching rule.
We assume that near y 0, the considered model is singular when the small
parameter e ! o, with y fixed (outer limit). As a consequence, according to MAE
we introduce a local y, significant near y 0, namely:
y^ y= e; with inner limit e ! 0; with y^ fixed (6.49)
In such a case, for the function Uy; e, according to the MAE method, we
consider two asymptotic expansions:
Uy; e Uo y eU1 y e2 U2 y Oe3 ; outer (6.50a)
Uy; e U^0 y^ eU^1 y^ e2 U^2 y^ Oe3 ; inner; (6.50b)
when e ! o, respectively at y fixed and y fixed.

Here we assume that, obviously, in the framework of the method of the MAE,
the limiting values in inner expansion (6.50b) of:
U^0 /; U^1 / and U^2 / exist and are well defined: (6.51a)
On the other hand, it is also assumed that the outer expansion (6.50a) remains valid
close to y 0, according to Kapluns extension theorem, such that we can write:
U y; e U0 0 y dU0 =dyy0 y2 =2d2 U0 =dy2 y0 :::

(6.51b)
eU1 0 ey dU1 =dyy0 ::: e2 U2 0 :::
If (6.51a) is really verified, and if this is also the case for (6.51b), the matching
between the (6.50a, b) outer and inner expansions is expressed by the following
equality:
U0 0 ef^
ydU0 =dyy0 U1 0g
y2 =2d2 U0 =dy2 y0 y^dU1 =dyy0 U2 0g
e2 f^ (6.52)
Oe U^0 / eU^1 / e U^2 / Oe
3 2 3
where, in place of y, we have written e^y.

As a consequence of (6.52) we obtain the following three matching relations, up
to Oe2 :
U0 0 U^0 / (6.53a)
y^dU0 =dyy0 U1 0 U^1 / (6.53b)
y2 =2d2 U0 =dy2 y0 y^dU1 =dyy0 U2 0 U^2 /

^ (6.53c)
The relation (6.53a) is the most classical and well-known of all matching
relations, and Eckhaus (1994) [124], pp. 4358, writes:
At the first confrontation it may baffle serious students because it says [that] the regular
(outer) approximation when extended to values where it is no longer valid equals the local
(inner) approximation when extended to values where it is no longer valid . . . Intermediate
matching is based on extension theorems and on the assumption of overlap of the extended
domains of validity. Asymptotic matching principles are based on assumptions of the
structure of uniform expansions; they contain no direct reference to extension theorems
and overlap hypotheses.
Eckhaus also asks a natural question: What are the interrelations between the
two procedures? The interrelations were studied in some detail in his (1979) [109],
Chap. 3, where the following facts were established:
The existence of an overlap domain assures the validity of an asymptotic matching
principle, provided that certain explicit conditions on the structure of the regular and the
local expansions are satisfied [Theorem 3.7.1 in [109]] . . . [but] the non-existence of an
overlap domain does not preclude the validity of an asymptotic matching principle.
Finally, it appears that Van Dykes intuition was correct. As to Lagerstroms

conviction in the necessity of overlap, Eckhaus remarks that counter-examples had
already been presented by Fraenkel in 1969 and in his (1979) [109], and some
discussion is included in Lagerstrom (1988) [125] but in the latter there is no
reference to Eckhauss monograph of 1979.
6.4.3 The Least-Degeneracy Principle and the Significant

Degeneracies
As mentioned in Van Dyke [14], Sect. 5.5, a crucial step in the MAE method is the
choice of inner variables.
The guiding principles are that the inner problem should have the least possible
degeneracy, that it must include in the first approximation any essential elements
omitted in the first outer solution, and that the inner and outer solutions should match.
As an example, consider the case in Sect. 6.4.1, linked with the choice of b.
Indeed, the significant degeneracies, according to Eckhaus, are a mathematical
formalism for the principle of least degeneracy, by Van Dyke. In several papers
of Feuillebois, and Feuillebois and Lasek (see [122]), the reader can find various
applications. The method makes it possible to obtain, with the use of a computer,
the boundary layer equations for a singular perturbation problem.
The straightforward but tedious hand-calculation of the significant degeneracies
may be avoided by the use of just such a computer! However, this method is not
concerned with the full solution to some problems (in fluid mechanics), but only
with the process required for obtaining the significant degeneracies.
Usually, the class of problems considered typically involves many parameters,
for which multiple boundary layers may exist.
The application essentially provides some boundary-layer equations to solve,
together with the order of magnitude of the region in which they are valid. But it
seems that there is no proof that the solution of such equations is the required
significant approximation to the solution of the original problem.
However, there is a useful heuristic principle of correspondence (Eckhaus): If
there exists a significant approximation, then the degeneracy in the boundary layer
variable is significant.
Practically, this principle limits the field of research of the significant approxima-
tion to the solutions of the boundary-layer equations obtained as significant
degeneracies, and this principle holds for many physical problems in fluid mechanics.
In 1984, Feuillebois (see [122]) considered an interesting example which is
relative to the full Navier (NavierStokes incompressible) steady equations for a
large Reynolds number. As a results, he demonstrated the existence of the following
seven significant degeneracies:
1. The ordinary boundary layer equations.
2. Equations for an ideal fluid with high thermal conductivity.
3. Boundary layer equations with a low outer pressure gradient.
4. Thermal boundary layer equations of an ideal fluid, with a low outer pressure
gradient.
5. Equations for an ideal fluid, with a complete energy equation.
6. The momentum equation as for an ideal fluid, but the energy equation contains a
viscous term.
7. Somewhat similar to (3), but with viscous dissipation.
We observe that the cases when Pr (the Prandtl number) is low (high thermal
conductivity) may be of practical interest for liquid metals flow. An interesting
exercise is the application of the Feuillebois approach to the unsteady Navier
equations case, and also (but certainly more complicated) to NSF, large a
Reynolds number case, considered in Chap. 5.
Germain (1977) [126], pp. 8996, includes a report on his views after reading
Eckhauss papers relative to the concept of significant degeneracy.
6.4.3.1 A Last Comment Concerning the Non-standard Analysis
In Going on with asymptotics, in Guiraud (1994) [127], pp. 257307 (esp. pp.
299302), there is a short account of this topic in which the author considers that
although it deals with the foundation of mathematics, in quite an abstract way, yet
it seems to me that it has some impact on asymptotics, as I shall try to show at least
superficially.
On the other hand, Eckhaus, in his remarks concerning the non-standard analysis
(in his (1994) [124], p. 434), writes:
In the last decade a group of French mathematicians attracted attention by the use of non-
standard analysis in problems of singular perturbations. Non-standard analysis is an
invention in mathematical logic due to Robinson (see his (1966) [128]) book, in which
infinitesimally small numbers are introduced back into the analysis. Robinson attempted to
link the non-standard analysis to a concept of asymptotic expansions in a book (edited in
1975) with Lightstone [129], but without much response. For an introduction to the new
developments one can consult [130], edited in 1981, and the literature quoted there, and
more recently the Proceedings of ICIAM 91, p. 342. There is an interesting link between
these developments and the lemma given above (and concerning the extension theorems of
Kaplun introduced in 1957, and formulated in upper part of the page 432 in [124]). In the
new applications of non-standard analysis a very important role is played by the so-called
Robinsons Lemma, which is the subject of much veneration.
Eckhaus also writes:

In discussions which I had with non-standard analysts (in particular Marc and Francine
Diener at Oberwolffrach in August 1981) it became clear that the lemma quoted in Sect. 3
of [124] is a standard counterpart of Robinsons lemma. The extension theorems are very
seldom (almost never) used as an explicit tool of analysis in specific problems. An
exception is contained in my analysis of the Canard problem [131], first treated by non-
standard analysts (see [130] for references). More recent use of extension theorems, in a
related problem, is given in [132].
6.4.4 The Method of Multiple Scales (MMS)
Various problems in fluid dynamics are characterized by the presence of a small

disturbance which, because of being active over a long time, has a non-negligible
cumulation effect.
In solving such a given problem, the main approach is to combine appropriate

techniques to construct an (asymptotic) expansion which is uniformly valid over
long time intervals. A central feature of such a (MMS) method is the impossibility
of the matching (which is just the central feature of the MAE method).
Often, the inner asymptotic expansion does not tend to a defined limiting state
when the short (distorted) variable (time or space) tends to infinity. In particular, in
the case of unsteady fluid flow, the adjustment to initial state is not possible (when
the short time tends to infinity) for the outer model equations.
The fundamental characteristics of the MMS can be summarized as follows
(according to Germain [126]):
When the data of a fluid flow problem show (after careful physical inspection) that the small
parameter e is, in fact, the ratio of two scales (of time or space), the MMS consists of first
introducing two variables constructed with these scales (one of them possibly being
distorted). Next, the formal expansion of the solution in e is considered, each coefficient
of the equations being a function of two variables introduced (e.g., of time) which are
considered as independent during the entire calculation. In order to completely determine a
coefficient of this expansion, it is not enough to solve the equation where it appears for the
first time! The indeterminants which necessarily remain are chosen by making sure that the
equation in which the next term appears will lead to a solution which does not destroy, but
on the contrary, best guarantees, the validity of the approximation which is sought. From a
certain a priori knowledge of the solution, we generally assume that U depends on the
variables t and x x; y; z so that a rapid variation having a repetitive character
analogous to an oscillatory phenomenon is made to appear. This variation is itself
modulated from one moment to another and from one point of space to another.
Such a structure is described by the following dependence of the U:
U U wt; x; t; x (6.54)
where w is a so-called fast intermediary variable because this function w varies

rapidly as a function of t and x. If only dimensionless variables are used (which is
the first main operation for a realization of the RAM Approach), then this property
can be characterized by writing:
@w=@t o=e and rw l=e (6.55)
where o and jlj are of the order unity. From (6.55) we deduce two relations:
@U=@t o=e@U =@w @U =@t (6.56a)
rU 1=e@U =@w l rU (6.56b)
with analogous formulae for the higher-order derivatives.

By substituting these expressions for the derivatives into the equations
governing the considered fluid flow, the small parameter e is introduced into
these equations, even if it does not appear initially.
In a second step, we write:
U U 0 eU 1 ::: and w w0 e w1 ::: (6.57)

and then, according to considered starting equations, when both of the above
relations (6.56a, b) are taken into account, and by setting at zero the terms
proportional to the successive powers of e, a hierarchy of systems of equations is
derived for U 0 ; U 1 ; :::
The first system in this hierarchy determines U 0 at best in its dependence with
respect to w0 , but not with respect to t and x. It is usually while seeking to
determine U 1 or even other higher-order terms that the dependence of U 0
with respect to t and x is prescribed by the cancelling of the secular terms; that is, of
terms which in U 1 do not remain bounded when w0 increases indefinitely. Indeed,
if we want (6.57) to cover an interval 0(1) in variation of t and x, then because of
(6.55), this corresponds to a variation of w0 which is O1= e.
A typical example is strongly related to the description of the progressive waves
(and see Germain (1971) [133]). As noted by Germain [108], pp. 1213:
Progressive waves occur generally when a physical phenomenon is thought to be
represented by the occurrence of steep gradients in one variable only, across three-dimen-
sional manifolds, in four-dimensional spacetime, with much smoother gradients in other
directions. The mathematical structure of the representation looks like a phenomenon in
five-dimensional spacetime. In such a case the phenomenon is quantified by an n-dimen-
sional vector U. Assume that the manifold across which the gradients are steep is
Ft; x const. Then, Ut; x; F=e; e so that w F=e, is considered as a fifth variable.
There is apparently nothing in the equations which allows us to single out the dependency
of U on w. But all is changed when we add the ansatz that the proper physical solution may
be obtained as an expansion with respect to w, t and x being fixed when proceeding to the
limit of vanishing e. As a matter of fact, the method of multiple scales (MMS), through the
requirement of vanishing of secular terms, provides the way by means of which that
dependency can be figured out. A key to the existence of progressive waves is that U,
supposed to be dependent on w only, at leading order, exists as a planar wave solution ruled
by a linear system with @F=@t; @F=@x obeying a dispersion relation. That relation defines a
wave speed, and may be considered as an eikonal of the HalmiltonJakobi equation, the
solution of which is built by means of rays. The planar wave, being a solution to a linear
system, is determined only up to a scalar amplitude factor . . . This amplitude obeys an
equation which is obtained when going to higher order in the asymptotic expansion and
eliminating secular terms this amplitude/transport equation being the main result of the
application of MMS. The details depend on the particular phenomenon considered, and
there exist quite a variety of situations that may be described mathematically by such a
procedure. The small parameter e characterizes the steepness of the transversal gradient,
and if the fluid dynamics process is non-linear, and non-linearity is measured by the order of
magnitude of the amplitude, and if, furthermore, the starting equations are first-order quasi-
linear, as is the case with inviscid gas dynamics, then the amplitude obeys a partial
differential equation which is generally an inviscid Burgers equation along each ray. If
there are second-order derivatives present in the starting equations, with a small coefficient,
then the amplitude obeys a partial-differential equation which is generally Burgers. The
role of time is played by the distance along each ray, while the role of space is played by w.
One may deal with third-order derivatives, and another small parameter, yielding then the
KdV equation; and both phenomena may occur simultaneously.
Thus, when both dispersive and dissipative effects are present, the transport
equation is a BurgersKortewegde Vries (KdV) equation of the following form:
@A=@t A@A=@x n=22 @ 3 A=@x3 m@ 2 A=@x2 (6.58)

Even if the Germain presentation is new and tries to be systematic, the essential
ideas may be found, in particular, in Guiraud (1969) [134], with a method which is
essentially equivalent.
For another application of the two-scales technique, see Sect. 3.3.5, where the
derivation of the well-known non-linear model Eq. 3.61q of acoustics the KZK
(Kuznetsov, Zabolotskaya, and Khokhlov) equation was examined.
In LNP (1987) [16], pp. 1507, the reader can also find an application related to
high Strouhal (S), high Reynolds (Re), and small Mach (M) numbers, where we
assume that
SM O1 ans Re=S O1 (6.59)
the Mach number M <<1 being the main small parameter.

In such a case, the leading term u0 in u u0 M u1 :::: is the solution of the
equation for the spherical waves (which is derived from the full NSF unsteady system)
@ 2 u0 =@t2 Du0 (6.60a)
and appears a wave front
Fx t 0; and w Fx t=M (6.60b)
As a consequence, a two-scales w; x expansion, in M, is necessary for: u; r, and T.

At the leading order we obtain (from full NSF unsteady system) a linear and
homogeneous system relative to first-order derivative, @=@w, of u0 w; x, r1 w; x
and T1 w; x, with t Fx M w . As a consequence we derive the following
eikonal equation:
k2 rFx2 1 (6.61a)
and, for instance, we have as a solution for u0 w; x,
u0 w; x Aw; xk=g 1 U0 x (6.61b)
where the function U0 x is arbitrary and is determined from the compatibility

conditions linked with the second-order system of equations, which is a linear and
non-homogeneous system of equations for
@u1 w; x=@w; @r2 w; x=@w and @T2 w; x=@w
and also with the matching, initial, and boundary conditions of the considered
problem. Finally, the amplitude function Aw; x in solution (6.61b) is solution of
following transport equation:
@A=@t aA@A=@w g@A=@w dA b @ 2 A=@w2 (6.62)
where the time t (increasing with F) is a parameter which varies along the
characteristics the characteristic form of eikonal (6.61a) being
dF=dt 1 and dk=dt 0 along dx=dt k
Finally, it is interesting to mention the opinion of Van Dyke [127], p. 7

concerning our (with Guiraud) the multiple scales approach:
In fluid mechanics, however, Guiraud, working with Zeytounian, has applied multiple
scales to two quite different problems, where it is clearly the appropriate method that is
mentioned in their titles. First, in the theory of axial-flow machinery we have Application
du concept dechelles multiples a lecoulement dans une turbomachine axiale (G-Z, 1974).
Clearly, the long scale is that of the mean flow, and the short scale that of the individual
blades, the small parameter being the reciprocal of the number of blades or of stages.
Second (GZ, 1977) we have A double-scale investigation of the asymptotic structure of
rolled-up vortex sheets. This applies to the rolled-up vortex sheets from the trailing edge of
a high-aspect-ratio wing or from the leading edges of a delta wing, the small parameter
being the number of turns.
In [11], the reader can find some comments concerning these two GZ papers.
Van Dyke [127], p. 8, also writes, concerning asymptotics allied with numerics:
Eight years ago, Guiraud and Zeytounian (1986) argued persuasively that although asymp-
totic techniques will in special circumstances continue to be used to derive closed-form
solutions, their future role will be primarily as an adjunct to numerical simulation. They
illustrated this view with a number of examples from fluid dynamics, concluding with
several drawn from their own research. From casual observation I suggest that the number
of papers appearing in technical journals with the word asymptotic in their titles is
actually increasing steadily. On the other hand, the cooperation between experts on
asymptotics and numerics which I agree is desirable and inevitable-seems not yet to
have gained momentum. Meanwhile, I have for some years been experimenting with a quite
different alliance between asymptotics and the computer.
I emphasize here that in the first published GZ paper (1970), relative to the
multiple scales approach for axial-flow machinery, a decisive moment is the
appearance of a source term, in the derived leading-order model, which is propor-
tional to the jump in pressure (the difference between the two sides of one and the
same blade), orthogonal to the material blade surface in the row. This equivalent
homogeneous force occurs due to the redistribution (homogenization) of forces
exerted on the flow by the blades of the row. In fact, in 1970 we had already
conceived an homogenization technique (without our knowing it!) that presents the
possibility of replacing the effect of the blades of the row on the flow in a channel
between two consecutives blades, by an equivalent source force in model equations.
As a result of the RAM Approach, an equivalent model the through-flow is
derived, in a simple cylindrical domain without any row with blades, but with the
source force in governing model equations.
6.4.5 The Homogenization Analysis
The homogenization analysis is in fact a more sophisticated form of the MMS

approach discussed in the previous section. Seepage through a porous media for
instance: water gradually escaping by seepage through the ground is considered

one of the first examples (if we do not take into account our GZ (1971) asymptotic
analysis of the flow in an axial turbomachine) to which the method of homo-
genization was applied (Sanchez-Palencia 1974 [135]). It is a good example for
explaining the role of physical scales and the mathematical procedure, and it can be
used to illustrate the derivation of many properties of the constitutive coefficients.
Let a rigid porous medium be satured by an incompressible Newtonian fluid of
constant density r0 (see, for instance, Advances in Applied Mechanics, vol. 32
(1996) [136], pp. 28892). Driven by a steady ambient pressure gradient, the steady
flow velocity ui and pressure p in the pores are governed by the Navier equations:
@ui =@xi 0 and r0 uj @ui =@xj @p=@xi mr2 ui (6.63a,b)
On the wetted surface of the solid matrix G, there is no slip condition,
ui 0 when xi 2 G (6.63c)
For slow flows (low Reynolds number), the two terms on the right-hand side of
(6.63b), representing the pressure and the viscous force, must be dominant. Then,
both the pore pressure and the flow velocity vary according to two very different
scales: the local or microscale l characteristic of the size of pores and grains, and the
global or macroscale L imposed by the global pressure gradient. During the
homogenization analysis the main working hypothesis is the ratio of the above
two length scales:
e l=L <<1 (6.64a)
defining the velocity scale U. A moment of reflexion leads to the following

dimensionless (primed) quantities:
xi 0 xi =l; p p=P; ui 0 mL=l2 Pu0i (6.64b)
where p in (6.64b) denotes the change in pressure (because it appears only in

differential form), P/L being the magnitude of the global pressure gradient. Equa-
tion (6.63b) becomes, formally in dimensionless form:
Re uj 0 @ui 0 =@xj 0 1=e@p=@xi 0 mr2 u0i (6.64c)
where Re er0 l2 P=m2 is just the Reynolds number (micro Ul=m=r0 ).

With P m2 =r0 l2 , and returning to dimensional variables, but retaining the
ordering symbol e, we obtain, as a starting equation for our homogenization
analysis, the following equation:
e2 r0 uj @ui =@xj @p=@xi e mr2 ui (6.65)

Let us assume that the geometry of the porous matrix is periodic on the micro-
scale l, although the structure may still change slowly over the macroscale L. Each
periodic cell O is a rectangular box of dimension Ol, ui and p being spatially
periodic from cell to cell. We introduce, in addition to xi also xi exi , and the
perturbation expansions
ui ui 0 eui 1 e2 ui 2 ::: (6.66a)
p p0 ep1 e2 p2 ::: (6.66b)
where ui k and pk , k 0; 1; 2; :::, are functions of xi and xi .

First, from @ui =@xi 0 we obtain
@ui 0=@xi 0 (6.67a)
@ui 1=@xi @ui 0=@xi 0 (6.67b)
@ui 2=@xi @ui 1=@xi 0 (6.67c)
and similarly, we obtain from (6.65),
0 @p0=@xi (6.68a)
0 @p0 =@xi @p1=@xi mr2 ui 0 (6.68b)
r0 uj 0 @ui 0=@xj @p1=@xi @p2=@xi mr2 ui 1 : (6.68c)
On the wetted interfaces G, the velocity vanishes; hence,
ui 0 0; ui 1 0; ui 2 0; :::: when xi 2 G: (6.69a)
In a typical O cell the flow must be periodic:
ui 0 ; ui 1 ; ui 2 ; :::p0 ; p1 ; p2 ; ::: are O periodic (6.69b)
First, from (6.68a) it is clear that:
p0 p0 xi : (6.70a)
and because of the linearity of (6.68b), ui 0 and p1 can be formally represented by
ui 0 Kij @p0 =@xj ; p1 Aj @p0 =@xj p 1 xi : (6.70b)

It then follows from (6.67a), (6.68b), (6.69a), and (6.69b) that
@Kij =@xi 0 (6.70c)
@Aj =@xi mr2 Kij dij (6.70d)
where
Kij 0 on G; and Kij ; Aj are O periodic (6.70e,f)
and (Of is the fluid volume inside the O cell)

<Aj > 1=jOj Aj dO (6.71)
Of
can be set to zero.

Equations 6.70c and 6.70d with 6.70e,f define a canonical Stokes flow bound-
ary-value problem in O cell, which can, in principle, be solved numerically for any
prescribed microstructure.
From (6.70b) we obtain, at the leading order, a well-known equation:
<ui 0 > <Kij >@p0=@xj (6.72a)
and also:
p1 n p 1 xi (6.72b)
where n is the porosity the ratio of fluid volume in the cell to the total cell volume:
n jOf j=jOj: (6.72c)
Eq. 6.72a being just the celebrated law of Darcy with <Kij > the hydraulic
conductivity.
On the other hand, the O-average of (6.67b) gives

@<ui 0 >=@xi 1=jOj @ui 1 =@xi dO 0
Of
and the interchange of volume integration with respect to xi and differentiation with
respect to xi is allowable when n constant in the above relation. Otherwise the
same is justifiable by virtue of the spatial averaging theorem (see, for instance,
[136], Sect. IIIF, pp. 3002).
By using the Gauss theorem and the boundary conditions, we see that the volume
integral vanishes; hence: @<ui 0 >=@xi 0, and this result implies, in turn, that:
@<Kij >@p0 =@xj =@xi 0: (6.73)

This equation (6.73) governs the seepage flow in a rigid porous medium on the
macroscale. Theoretical derivation in the present manner was first presented by
Snachez-Palencia in 1974.
If, now, the medium is isotropic and homogeneous on the L scale, we have
<Kij > K dij
where K is a scalar constant, and it follows from (6.73) that
@ 2 po =@xk @xk 0
With proper boundary conditions on the macroscale, p(0)can then be found. In

fact, when the Reynolds number is assumed to be finite (Re O(1)), then the
convective inertia would be on the order of (e), and (6.68b) would be replaced by
ro uj 0 @ui 0=@xj @p0=@xi @p1=@xi mr2 ui 0 (6.74a)
with
@ui 0 =@xi 0: (6.74b)
The above cell problem, (6.74a, b), is fully non-linear, and (6.73) and Darcys
law (6.72a) no longer hold.
Finally, concerning this seepage flow in rigid porous media, the reader can
find in [136] various complementary results concerning the uniqueness of the cell
boundary-value problem ((6.70c)(6.70e,f)) with zero on the right of (6.70d),
properties of hydraulic conductivity, numerical solution of the cell problem, the
effects of weak inertia, a spatial averaging theorem, and porous media with three or
more scales.
We observe that the typical steps of the homogenization theory are:
1. Identify the microscales and macroscales. This identification of scales is a
consideration of physics and is crucial to the success of the mathematical theory.
2. Introduce multiple-scale variables, in non-dimensional form of the starting fluid
flow problem, and expansions, and deduce cell boundary-value problems at
successive orders. The leading-order cell problem is homogeneous; either the
solution itself or the coefficient of the homogeneous solution are indeterminate
and independent of the microscale coordinates.
3. Use linearity and express the next-order solution in terms of the leading-order
solution and deduce an inhomogeneous cell problem.
4. Determine the solvability of the inhomogeneous cell problem mathematically,
the solvability condition for the inhomogeneous problem given that the homo-
geneous problem has a non-trivial solution!
5. Derive the equation governing the evolution of the leading-order solution (or the
coefficient of the homogeneous solution) and calculate the constitutive
coefficients from the solution of a canonical cell problem.
We have previously mentioned (for instance, in Chap. 3) that the case of low
Mach number in gas dynamics is a very difficult problem when the gas is contained
in a bounded container O, with an impermeable but eventually deformable (with
time t) wall, so that the volume occupied by the gas is a given function of time:
namely, V0(t). In such a stiff case, it is necessary to use an homogenization
technique with an infinity of microscales. Such a problem is considered in Sect. 7.2.
6.4.6 Asymptotics of the Triple-Deck Theory
Here we do not discuss in depth this triple-deck theory, as a detailed and thorough
discussion is presented in our FMIA 64 (2002) [26], Chap. 12. Some of the features
of this theory were mentioned in Sect. 6.3.4, and we now discuss the various steps
required for the derivation of model equations for viscous lower motion.
6.4.6.1 The Local Atmospheric Thermal Problem
Here I consider the formulated problem of Eqs. 6.42a6.42e with conditions

(6.44a)(6.44e).
First, if we take into account the boundary conditions on the ground z 0, it is
necessary to introduce an inner variable z such that
z z=koa ; a>1

and in this case
u koa1 z; for x !/: (6.75)
Then, from the first Eq. 6.42a, for u, we verify that
z z=koa and u koa1 ux; z;

if
and from
u=@z ::: ko52a @ 2 u=@z2

koa1 @
::: ) a 1 5 2 a ) a 2: (6.76)
As a consequence, we establish that three vertical coordinates are necessary

(z, z, z), for a rational asymptotic (triple-deck) analysis of the system of
Eqs. 6.42a6.42e: namely,
1. z, for the upper non-viscous region, where u uup ! 1, whenx ! / ;
2. z z/ko, for the middle intermediate region, where
u umid ! 1 exp(z) cos(z) U/(z), when x ! / ;
3. z z/k 2, for the lower wall viscous region, where u k ulow, and ulow ! z,
when x ! / .
Note that here we consider only the case when (m 5):
l =L ko3
For the other case,
l < ko3 L and l > ko3 L
it is necessary to apply different asymptotic analysis. The case m 6 and m 4

can be analyzed from the equations analogous to (6.42a)(6.42e). For the case
m 3 it is necessary to start from other equations, where the Boussinesq approxi-
mation does not emerge.
In particular, for m 3, we have l 104 m, and we may neglect the terms
related to the Coriolis force in the local non-Boussinesq equations.
Beginning our triple-deck analysis with the middle deck where x and z z/k
are the order one coordinates, we expand the flow variables as:
u U/
z kof umid :::
w koc wmid :::
(6.77a)
p ko2 pmid :::
y kos ymid :::
and from system of Eqs. 6.42a6.42e to find for the lowest order:
U/
z@umid =@x d U/ =zwmid 0
@umid =@x @wmid =@
z0
(6.77b)
z B ymid
@pmid =@
@ymid =@x 0:
The choice
f 1; c 1 f and s 1 (6.77c)
is necessary if we want to obtain a significant degeneracy of the Eqs. 6.42a6.42e in

the lower viscous deck region in the vicinity of z 0 near the wall.
We observe that the effects of the expansion of the boundary layer are Ok in u
and Oko2 in w. Furthermore, in the boundary layer (lower deck), if we take into
account that z z=ko2 , we necessarily have
p ko2 plow :::

and, by continuity, we obtain the form of the expansion for p in (6.77a).

Solutions for umid and wmid satisfying the upstream boundary conditions are:
umid Ax dU/ =d

z and wmid dAx=dxU/ z (6.77d)
which represent simply a vertical displacement of the streamlines through a distant

k A(x).
On the other hand, the flow in the upper non-viscous deck is driven by an outflow
from the middle deck. Far from (6.77d) we have:
Limz0 !/ wmid x;
z dAx=dx (6.77e)
In the upper deck, with z k z, as flow expansion we write:
u 1 ko2 uup :::; w; p, y ko2 wup ; pup ; yup ::: (6.78a)
and gain substitution in Eqs. 6.42a6.42e to find for the lowest order the following
equations for an inviscid motion:
@uup =@x 1=g@pup =@x 0;

@wup =@x 1=g@pup =@z 0
(6.78b)
@uup =@x @wup =@z 0;
@yup =@x B L0wup 0;
and, as a consequence we derive, for pup, a Helmholtz equation:
@ 2 =@x2 K0 2 @pup =@x 0 (6.78c)
where
K0 2 B =gL0:
In fact, according to upstream conditions we have:
Limz!o pup x; z pup x; 0 Px Limz0 !/ pmid x; z (6.78d)
but
wup x; 0 dAx=dx (6.78e)
and, as consequence of (6.77e) and matching between upper and middle decks, we
obtain, for Eq. 6.78c, the following rather strange condition for pup:
@=@x@pup =@zz0 g K0 2 dA=dx d3 A=dx3 : (6.78f)

It is obvious that the middle deck solution (6.77d) does not satisfy the no-slip
condition on z 0 a situation which is remedied by the analysis of the lower
viscous deck, near the wall, where we have the following relation:
z z=ko2 z0 =k

Now, matching with the middle (6.77a) expansion, when in U/(z), with z k z
and z is fixed, implies the following lower deck expansions:
u; w; p, y k ulow ; ko3 wlow ; ko2 plow ; ylow ::: (6.79a)
and from Eqs. 6.42a6.42e to find for the lowest order the following model
equations for viscous lower motion:
z

ulow @ulow =@x wlow @ulow =@
z B =g @ ylow =@xd z
/
1=gdPx=dx @ ulow =@
2
z ; 2
(6.79b)
@ulow =@x @wlow =@

z 0; (6.79c)
ulow @ ylow =@x wlow @ylow =@

z 1=Pr@ 2 ylow =@z2 ; (6.79d)
with the boundary conditions:
z 0 : ulow wlow 0; ylow = t Yx; 0< x <1

(6.80a)
z ! /: ulow !
z; wlow ! 0; ylow ! 0; Px and dA=dx ! 0 (6.80b)
z ! /: ulow !
z Ax; wlow !
zdA=dx; ylow ! 0 (6.80d)
after matching with the middle deck.

The specification of the above viscous problem in the lower deck is
completed by the relation (6.78f) linked with P(x) (if we take into account that
according to (6.78d) pup(x, 0) P(x)) and the function A(x).
The well-known interpretation of the strange condition (6.78f) is as follows:
The pressure P(x) driving the flow in the lower deck is itself induced in the main
stream, i.e. by the upper deck solution, as a consequence of the displacement
thickness of the lower deck transmitted through the middle deck by the passive
effect of displacement of the streamlines.
One the other hand:
This so-called strong singular self-induced coupling (a la Stewartson) arises
because the above problem to be solved in the lower viscous layer does not accept
the function P(x) as data known prior to the resolution as is the case in the
classical Prandtl boundary layer problem.
Conversely, this pressure perturbation P(x) must be calculated at the same time
as the velocity components ulow and wlow as well as the temperature perturbation
ylow. Nevertheless, it must be emphasized that the function P(x) is not completely
arbitrary, and that it is connected to the function A(x) through a relation which is
derived via the analysis of inviscid fluid motion in the upper deck via (6.78f).
The reader can find in Guirauds paper Going on with asymptotics [127],
pp. 26271, a very pertinent and personal point of view concerning the asymptotics
of the triple-deck theory. In particular, Guiraud writes:
Triple-deck theory deserves a special mention among the numerous applications of high
Reynolds numbers laminar flows. It is somewhat misleading to associate high Reynolds
numbers with laminar flow, although such an association is not completely false at the
outset. What can be said here is that, provided some caution is exercised when applying the
theory, it has proven to give very powerful results. What is unfortunate with this kind of
asymptotics is that it seems to be very esoteric, on the one hand, while it looks suspect, on
the other hand, for at least two reasons. The first is that in its simplest version the only one
reported here it is based on (ReL )1/8 being much smaller that unity; the second is that the
higher ReL , the less plausible the laminar conditions seem to be. It is outside the scope of
this paper to discuss this last issue, and we ask the reader to believe that laminar conditions
can exist (obviously within a limited range). I only intend to debunk, if possible, the
esoteric blurring which appears to have for long discouraged many scientists from investing
in it. Let me explain the main goal, assuming the reader is familiar with basic boundary
layer theory and knows the corresponding hierarchy: computation of the inviscid flow first,
followed by boundary-layer computation as such, finally (eventually) computations of
corrections, generally of order (ReL )1/2, to both. The classical theory is based on what
might called an asymptotic ansatz: namely, that a thin (of order (ReL )1/2 thick) layer of
fluid surrounds the surface of an undeformable smooth body on a O(1) scale. Each time this
asymptotic ansatz is broken, one should expect a failure of classical boundary-layer theory.
The triple-deck structure is now seen as a useful and, indeed, valuable element in
aerodynamics calculation and design. This is a substantial tribute to Stewartsons
power and foresight, and is his greatest contribution to theoretical fluid mechanics
(see, for instance, Stewartson (1974) [137]). For a pertinent comprehensive review
of progress in using the asymptotics of the triple-deck theory, see the illuminating
paper by Meyer (1983) [138].
Guiraud was fascinated by the breakdown of asymptotic theories at the leading
and trailing edges, and in his 1974 study (the first in France) in particular in his
study of separation at the trailing edge of a thin three-dimensional wing shows
that for this complicated case, three different scalings arise, and that the Stewartson
layered triple-deck structure is very well adapted.
It should be noted that the triple-deck theory was discovered by Stewartson and
Williams in England in 1969, but also independently by Neiland [139] in Soviet
Russia. Paul Germain [107] has written:
Even a genius would not have been able to build the whole of the triple-deck model without
the help of matched asymptotic expansion techniques. Triple-deck theory is now a very
important building stone in the new fluid dynamics inspired by asymptotics, and it may
be fully included within the heritage of Prandtl.
Part III
Applications of the RAM Approach to
Aerodynamics, Thermal Convection, and
Atmospheric Motions
Chapter 7
The RAM Approach in Aerodynamics
7.1 Derivation of a Through-Flow Model Problem

for Fluid Flow in an Axial Compressor
First, in Sect. 7.1.1, we again consider the simple case examined in Veuillots thesis
[140] devoted to turbomachinery fluid flow, simulated in Sect. 6.3.1; then, in Sect.
7.1.2, the more sophisticated GZ RAM Approach [141].
It is obvious that the following asymptotic theory of axial flow through a turbine,
which is likely be of considerable interest to specialists, is a fascinating application
of a complicated engineering problem using the RAM Approach with the basic
large parameter being the number of turbine blades per rotor.
7.1.1 The Veuillot Approach
We starting from the system of non-dimensional equations (6.18a)(6.18c), for u*,

w*, and G rv , and also the relations (6.17a)(6.17c) relying, u*, w, and v*,
with the functions c and Y.
First, we expand the functions u*, w*, G , and c , relative to our main small
parameter e:
(u*, w*, G ; c*) = (u o ; w o ; Go ; c o

(7.1)
e u1 ; w1 ; G1 ; c1 :::
and at zeroth-order we derive the following leading-order approximate equations

for the functions uo 1=rD @ co =@z; wo 1=rD @ co =@r, Go
r2 uo @Y=@r w o @Y=@z: namely,
@uo =@w ::::: @Y=@r @Go =@w 0 (7.2a)

162 7 The RAM Approach in Aerodynamics
@wo =@w @Y=@z @Go =@w 0; (7.2b)
@Y=@z @uo =@w @Y=@r@wo = @w 0: (7.2c)
But these three equations are not independent, since the determinant of
coefficients for @uo =@w, @wo =@w, and @Go =@w is zero! On the other hand, from
them we easily derive the following two relations:
@Go =@w r2 @Y=@r@uo =@w @Y=@z @wo =@w (7.3a)
@Go =@w f1 r2 @Y=@r2 @Y=@z2 g 0: (7.3b)
Because:
1 r2 @Y=@r2 @Y=@z2 6 0; @Go =@w 0;
and as a consequence:
@Go =@w 0; @uo =@w 0; @wo =@w 0 (7.4)
Therefore: when e tends to zero, and as a consequence of a uniform and constant

steady flow far of the row, in the leading-order the approximate limiting through-
flow in the row of an axial compressor (turbo-machine) is independent of the short
(micro)-scale w.
Therefore, it is necessary to consider in the starting Eqs. 6.18a6.18c the next
order (terms proportional to e!) In a such case we obtain, for u*1, w*1, and G1 , the
following three equations:
1=r@Go =@r 1=rD@u1 =@w @Y=@r @G1 =@w; (7.5a)
@uo =@z @wo = @r 1=D @Y =@z) @u 1 = @w

(7.5b)
@Y =@r) @w 1 = @w];
1=r) @Go =@z 1=rD@w1 =@w @Y=@z@G1 =@w: (7.5c)
From these equations, by elimination of the first-order functions, u*1, w*1, and
G1 , we obtain as a compatibility relation:
@uo =@z @wo =@r @Y=@z @Go =@r @Y=@r @Go =@z
and with
uo 1=rD @ co =@z; wo 1=rD @ co =@r

7.1 Derivation of a Through-Flow Model Problem for Fluid Flow 163
we derive an equation for the function co Lo r; z, which characterizes the

limiting through-flow in the row. Namely:
@=@rf1=rD@Lo =@rg + @ =@zf1=rD@Lo =@zg

(7.6)
@Y=@z@Go =@r @Y=@r @Go =@z:
This model equation for the function Lo is our first main rational and entirely
consistent result with the RAM Approach.
Again, due to being far upstream of the row, we have a uniform and constant
steady incompressible ro constant fluid flow: 1=2ro u2 p constant.
Then the jump,
jpj pw 1=2 pw1=2 (7.7a)
of the pressure, from blade to blade, gives
jpj e ro fu o ju1 j w o jw1 j 1=r2 Go jG1 jg O e2 (7.7b)
where |[p]| is a quantity of the order e.

But, according to Eqs. 7.5a7.5c, obviously u*1, w*1, and G1 are linear functions
of the microscale structure w.
As a consequence of (7.7b), we have to write:
Lime!0 p=ero uo @u1 =@w wo @w1 =@w

(7.8)
1=r2 Go @G1 =@w Po :
From (7.5a7.5c), we now eliminate the terms with the w derivatives in the
(7.8) relation, and express the function Po , in (7.8) simply by:
Po Duo @Go =@r wo @Go =@z (7.9a)
in the row.
Outside the row, p remains continuous, even in the presence of wakes which
are, in the considered Eulerian fluid flow, only vortex sheets (contact discontinuity
surfaces).
Finally, taking into account the periodicity in w, outside the row, we derive in
place of (7.9a) the following relation:
jpj 0 ) Po 0 (7.9b)
and
uo @Go =@r wo @Go =@z 0 ) Go Go Lo : (7.9c)

As a consequence:
In the whole outside region upstream ot the row : Go 0 (7.10a)
In downstream region outside of the row : Go Go Lo (7.10b)
From the above we can formulate the following main results relative to a
homogenized through-flow.
The velocity vector U o uo ; wo ; vo of the homogenized through-flow is
Uo 1=Dry Y ^ rLo ; (7.11)
such that the streamlines of the through-flow are obtained by the crossing of median
surfaces,
y Yr; z constant
in the inter-blade rowchannel, with the cylindrical surfaces, resulting from the
rotation around of the z-axis of the turbo-machine of meridian streamline surfaces
Lo constant
For this through-flow we have for the function Lo r; z, (see Eq. 7.6):
@=@rf1=rD@Lo =@rg @=@zf1=rD@L o =@zg

(7.12a)
@Y=@z @Go =@r @Y=@r @Go =@z;
with, as conditions (if we use (7.11) for the composante v*o):
Go 0; upstream of the row; (7.12b)
Go (r/D @Y =@r @Lo =@z - (@Y =@z) @Lo =@r], in the row; (7.12c)
Go Go Lo ; downsream of the row: (7.12d)
This axially symmetric through-flow model, which is dependent only on

coordinates r and z, introduces a fictitious force:
F Po =Dr y Y; (7.12e)
which simulates the action of the blades in the row on the turbomachinery flow.
The force F, given by (7.12e), is a memory term (a trace) which via homoge-
nization replaces (simulates) the (vanishing) effect of the blades in the row.
The first numerical applications of the above through-flow model in a turbo-

machine blade row was realized by Veuillot [140] at the ONERA and see also his
[142] paper.
7.1.2 The GZ Approach
In a more sophisticated general case, Guiraud and Zeytounian [141] consider in

1971 at the beginning, in the cylindrical coordinates r, y z, the following starting
Eulerian incompressible equation written in the matrix form:
@T=@t @R=@r @Z=@z 1=r@S=@y H=r 0; (7.13)
and make the change of coordinates from t; r; y; z to t; r; z; w as shown, in

(7.14a):
y Yt; r; z 2pek w; (7.14a)
with the idea in mind that through-flow will be independent of w whereas r will
appear as a parameter for cascade flow.
Without any approximation the flow has to be periodic in w, and we enforce this
by:
Uk t; r; z; e; w 1 Uk1 t; r; z; e; w; (7.14b)
UkN t; r; z; e; w Uk t; r; z; e; w; (7.14c)
using for convenience the index k which runs from 1 to N e 1=N <<1, the
number of blades in a row, and accordingly we assume that w is between zero and
one.
We expand, formally, Uk as powers of e, but we obviously need two such
expansions, because it is clear that the model through-flow in row is invalid near
the locus of the leading/trailing edges of the row.
The first one is a kind of outer expansion (as in (7.1) above):
Uk Uk;o e Uk;1 ::::; (7.15a)
and will fail near both ends of the row where (two) inner expansions
Uk t; r; z hr ez; e; w Uk;0 e U k;1 :::; (7.15b)
are needed. In (7.15b), z h(r) is the locus of the leading (or trailing) edge of a
row.
When the change of coordinatesfrom y to w is made in the basic matrix

form Eq. 7.13 of the inviscid and incompressible Eulerian fluid flow, we obtain
(7.16a, b):
@Gk =@w 2 p e r Lk 0 (7.16a)
with
Lk @Tk =@t @Rk =@r @Zk =@z Hk =r (7.16b)
and we observe that in matrix column Gk the following two parameters are present:
l o D=w/ ; m D=w/ t (7.16c)
and we have introduced
gk vk r uk @Y=@r r wk @Y=@z r@Y=@t (7.16d)
In (7.16c), t is a reference time, D is the diameter of the row, o is the reference

value of the angular velocity (o) of the row, and w/ is the upstream uniform axial
velocity.
Two facts should be stressed at the outset. First, if we assume axially symmetric
flow, @=@w 0, we obtain Lk 0, and it may be checked that Lk 0 is the matrix
form of axially symmetric through-flow. Second, if we use, by brute force )
e 0 in (7.16a), we do not obtain the equations of axially symmetric flow but,
rather, the highly degenerate equation @Gk =@w 0! This is somewhat strange, but
is not unexpected.
The way in which w has been defined substantiates that: when e is small,
variations in the w direction are magnified by 1= e, in comparison with variations
in t, and in the r, or z direction.
Now, substituting the basic outer expansion (7.15a) in Eq. 7.16a we derive a
hierarchy of equations. But here we write only first two:
@Gk;0 =@w 0;
(7.17a)
@Gk;1 =@w 2 p r Lk;0 0;
which consist of equations to be solved in turn.

We choose, as appropriate to the present problem, the solution of
@Gk;0 =@w 0 (7.17b)
for which uk,0, wk,0, vk,0, and pk,0 are all independent of w.
At this step we do not know the way in which these functions depend on t, r, and
z! Now, if we use the second equation of two equations (7.17a) in order to compute
uk,1,. . . and so on, we encounter a compatibility condition arising from periodicity,

which forces Lk,0 to be zero!
We have thus obtained a through-flow, axially symmetric theory. The interesting
point is that we may go a step further and produce a through-flow theory to order
e inclusively! For this, it is first necessary to define the channel between two
consecutive blades:
we w wi ; and we define
Dr; z wi we
and we then introduce an average, < >, and a jump, [ ], operation, thus:

<U> 1=D Udw; integration from we to wi (7.18a)
and
U Uwi Uwe : (7.18b)
Now, if we think of the pressure for U (in (7.18b)), then the bracketed [p] may be
viewed as: the pressure difference between the two sides of one and the same blade.
Below, the various equations shows the basic results of the GZ RAM Approach.
Up to first order in e, the average of velocity and pressure
<V1 > Vk;0 e<Vk;1 >; <p1 > pk;0 e<pk;1 > (7.19a)
satisfies, with an error of order e2 , axially symmetric through-flow equations:
DivD<V1 > O e;
@<V1 >=@t frot<V1 > 2Oez g ^ <V1 > rI 1 , (7.19b)
F1 O e2 ;
where
I 1 <p1 > 1=2j<V1 >j2 1=2O2 r2 ; O l o,

P1 1=2 p<pk;1 > e<pk;2 >, (7.19c)
S S 2 p e 1=2 wi we ; S Y y,
with
@S=@t <V1 >:rS O e2 (7.19d)
F1 1=DP1 rS ) F1 : rot F1 0: (7.19e)

Two points again need to be stressed. First, the breadth of the channel from blade
to blade, set as
Dr; s
enters in the continuity equation in an obvious way. Second, in the momentum

equation there is a source term,
1=D P1 rS F1
which is proportional to the jump in pressure and is orthogonal to S constant a

surface which is just in the middle of the channel. This force has long been known
in through-flow theory; but in fact, via a very subtle ad hoc consideration
(first published, it seems, by Chung-Hua Wu, in NACA TN 2288 (1951); see also
Wu [143] paper).
The force F(1) occurs from redistribution (homogenization) of forces acted on
the flow by the blades of the row.
The GZ [141] derivation given above is illuminating with regard to the error
involved in the approximation. To order one there is a dependency on w which may
be computed once the through-flow is known.
7.1.3 Transmission Conditions, Local Solution at the

Leading/Trailing Edges, and Matching
The above through-flow model in axial turbomachine is invalid near the locus of
leading/trailing edges of a row.
According to GZ theory [144], a local asymptotic analysis is performed by
considering the inner expansions (7.15b) and rewriting the starting matrix equation
(7.13). We obtain:
@Gk =@w 2pr@Nk =@z 2perMk 0 (7.20a)
with
Nk Zk dh=drRk (7.20b)
Mk @Tk =@t @Rk =@r Hk =r (7.20c)
and to zeroth order we obtain:
@Gk;0 =@w 2pr@Nk;0 =@z 0; (7.21)

which is, in fact, the equations of cascade flow but the configuration is that of
semi-infinite cascade flow. In [144] a detailed analysis of (7.21) is performed,
adapted to a local frame linked with the curve:
G : fz hr; y Yr; hrg
The semi-infinite cascade flow fills the gap between external (outside the row),
force-free, axially symmetric through-flow, and internal (in row) through-flow with
the source term F(1). Matching provides transmission conditions between these two
disconnected through-flows.
The necessity of such conditions appears readily as soon as any numerical
treatment of the whole through-flow in a two-row stage is attempted from
upstream to downstream (infinity) of this two-row stage!
To zeroth order these transmission conditions are rather simple and, indeed,
obvious on physical grounds: They mean that mass flow is conserved, as well as the
component of momentum parallel to the leading or trailing edge.
Local analysis has also been carried out by Guiraud and Zeytounian [144], to
first order, without a simple interpretation of the (rather complicated) result linked
with the transmission conditions!
We can consider the singular regions near the entry and exit of the row as planes
of discontinuity, if we impose the associated transmission conditions.
7.1.4 Some Complements
We now turn, briefly, to various cases concerning my work devoted with Guiraud
during 19691978, to turbomachinery fluid flows.
After the axial flow in a turbomachine, with the approximation of ideal incom-
pressible flow, has been analyzed by using an asymptotic method, assuming that the
blades are infinitely near one another [141], and in a companion paper [144] a
local study reveals the nature of the flow in their neighbourhood and leads to
a system of transmission conditions, because the partial differential equations of
the through-flow (in three different regions: upstream of a row, in a row, and
downstream of the row) must be supplemented by them in order to produce a
well-posed problem for the whole of the turbomachine (from upstream of the row to
downstream of this row) an application of the concept of multiple scales was
considered.
Namely, in [145] an asymptotic theory for the flow in an axial compressor was
considered, with the aim of devising a coupling process between the so-called
meridian through-flow and the flow around cascades. Again the small parameter
e is the inverse of the (supposed 1) number of blades per row and/or number
of stages. As a matter of fact, the cascade flow is treated as a small perturbation
of the through-flow, and has to be computed, locally, as the two-dimensional
unsteady flow around an array of couples of cascades alternately fixed and in
motion. The array is constructed by developing on a plane the section of the

compressor by a circular cylinder, and continuing, by periodicity, the pair of
cascades so obtained, at each location. The coupling between through-flow and
cascade flow is part of the analysis. It happens, incidentally, that the equations of
through-flow are obtained through an averaging process, completed on a domain of
periodicity of the array of cascades flow, while the through-flow appears locally as
an unperturbed flow for the linearized problem defining the cascade flow. The 3D
nature of the complete flow is built in by the coupling itself, as is visualized by the
occurrence of source terms in each of the two sets of equations describing through-
flow and cascade flow. This paper [145] is aimed at producing a preliminary answer
to the question of: how to devise, as rationally as possible, a way of describing the
familiar scheme of cascade flow within the computation of a mean through-flow.
The main conclusion is that the concept of cascade flow should be revisited and
reassessed as one of unsteady flow around an array of cascades.
In 1978, I published, with Guiraud, a fourth paper [146], entitled Cascade and
through-flow theories as inner and outer expansions. In this, a technique of
matched asymptotic expansions is used in order to combine two kinds of approxi-
mation. Through-flow theory forms the basis for an outer expansion, while cascade
theory forms the basis for an inner one, and matching provides boundary conditions
for both flows. It appears that for the downstream through-flow, a technique of
multiple scales is necessary at least in the vibrating case (vibrations induced by
harmonic vibrations of the blades) in order to deal with the unsteady wakes
generated by the vibrating blades, and slowly modulated downstream by the steady
part of the through-flow.
Although there are very many good papers on the theory of turbomachine flow,
we have not found any attempt analogous to that described above in [145] and
[146], and it seems difficult to comment on the relation of the work to be presented
with others. Concerning the two papers [141] and [144], we observe that in
Sirotkins paper [147] only some results are similar to ours, but the main difference
is in the approach, which is less systematic.
Our objective, with Guiraud, concerning the papers [141] and [144146] was
very modest, and we did not solve any problems nor present any results!
What we have proposed in our above-mentioned papers devoted to a rational
asymptotic description of turbomachine flows in the framework of a RAM
Approach, may be stated as follows.
Considering incompressible non-viscous fluid flow through a one-row machine,
assuming that there is a great number of blades, and, in [146], that the
corresponding cascade has a chord-to-spacing ratio of order one, we want to
show that the first few terms of an asymptotic representation of the 3D flow may
be guessed as having the form of an inner and outer multiple-scale expansion.
We confirm our guess as is usually done with problems not amenable to
mathematically rigorous analysis by an internal consistency argument: we show
that each term of the expansion, up to the order considered, may be computed by
solving well-set problems. We show the rational asymptotic process by which these
7.2 The Flow Within a Cavity Which is Changing Its Shape and Volume 171
problems may be extracted from the definition of the original 3D problem, which is a
typical RAM Approach.
For engineering applications it would have been very useful to find, as partial
problems, cascade flow theory as well as through-flow theory across a thick row.
Unfortunately, we have been unable to find any asymptotic process leading to such
a scheme. As a matter of fact, the obvious way to do so leads to only two significant
degeneracies. One is the through-flow of [141] and [144], which leaves no room for
cascade flow, and the other is the one considered in [146], which leads to cascade
flow but leaves no room for through-flow, including a thick row.
This conclusion inevitably leads to some deception, because there is no way to
embed Wus [143] technique within an asymptotic rational framework.
7.2 The Flow Within a Cavity Which is Changing Its Shape and
Volume with Time: Low Mach Number Limiting Case
Concerning the aerodynamics applications it is necessary to distinguish between

confined and unconfined flows and, especially, to elucidate the role of the Strouhal
number S (unsteadiness).
When we consider the low Mach number case, two distinguished limiting
processes emerge. One of them leads, from the full unsteady NSF equations, to
the model of incompressible flow (Navier equations) and, in the case of confined
configuration provides a dynamic interpretation of thermostatics. The other one is at
the root of acoustics.
The first limiting process (incompressible) corresponds roughly to
M ! 0 with S fixed
while the second limiting process (of acoustics) corresponds to
M ! 0 with SM O1:
Curiously enough, the acoustic model enters into the scene even in the situation
which is apparently ruled by incompressible aerodynamics.
This occurrence is due to non-uniformities: a spatial one near infinity in the case
of unconfined flow, and a temporal one for small time (in particular near time 0,
where the initial data are given) and also for high-frequency oscillations in the case
of confined flows.
In [85], the influence of these high-frequency oscillations was taken into account
by Zeytounian and Guiraud via a judicious multiple-scales technique, but with an
infinity of short acoustic scales! Here, below, we consider, as a physical situation,
the low Mach number flow within a cavity which is changing its shape and volume
with time. Such a problem presents industrial interest in the case of the compression
phase flow in an internal combustion engine (see our short notes in [86]).
We show how a limiting process corresponding to the Mach number going to
zero leads to an incompressible unsteady model flow, provided that acoustic waves
are averaged out over a great number of periods. This scheme may even describe
the case of a gas with a purely temporal variation of density due to substantial
changes of volume.
7.2.1 Formulation of the Inviscid Problem
We start from the Euler compressible dimensionless equations:
Dr=Dt rr:u 0
Du=Dt 1=gM2 rp 0
(7.22a)
DS= Dt 0
p rg expS
where D=Dt @=@t u:r, with standard notation.

The Strouhal number in the first three of these equations is assumed equal to one,
such that the reference time is t L =U , with L a typical length scale of the
cavity O(t), and U a characteristic velocity related to the motion of the wall
@Ot St.
As boundary condition we write the slip condition:
u:njSt wjSt W t; P; P 2 St (7.22b)
where the (data) velocity W(t, P) characterizes the normal displacement of the wall
S(t), and n is the unit vector normal to this wall, directed inside O (t) P being the
position point vector on the wall S(t).
But it is also necessary to take into account the conservation of the global mass
m of the cavity (a bounded domain with L as a diameter). In the dimensionless
reduced form we have:
r 1=Vt; Vt r jOtj=m ; Vt 0 1 (7.22c)
where |O (t)| is the volume of the cavity a known function of time t.

As initial conditions we write:
t 0 : u 0; p r 1 and S 0 (7.22d)
More precisely, we consider the case when the motion of the wall, S(t), is started
impulsively from rest, and in a such case,
W t; P HtWS P; all along St: (7.23a)
In this condition, the function H(t) is the Heaviside (or unit) function, such that
Limt!0 Ht 1; but initially : Ht 0 0: (7.23b)
7.2.2 The Persistence of Acoustic Oscillations
Asymptotics analysis and the RAM Approach in the formulated problem above is
not easy task, mainly due to the persistence of acoustic oscillations in the cavity
emerging for t 0 .
Therefore, if the Mach number, M, is sufficiently small, the zeroth order
approximation leads to the thermostatic isentropic evolution of the gas within the
cavity as a whole. Superimposed onto them we have acoustic oscillations which
remain undamped as long as viscosity is neglected when ones takes it into account
the Euler equations (7.22a), in place of full unsteady NSF equations.
When the NSF equations are considered (in place of Euler equations (7.22a)),
then a rather longer time O(Re1/2) is necessary in order to damp out the oscillations
but a much longer time is necessary in order that the heat exchanges can take place.
As a matter of fact, this time is O(Re)!
Indeed, some new features occur (as a consequence of the unsteadiness of the
compressible fluid flow) when one deals with internal aerodynamics. The first
concerns the leading term in the expansion of pressure which is function of time
instead of being a constant. We write:
u u0 Mu M2 u u2 :::; (7.24a)
p P0 t M2 p p2 M3 p M4 p p4 ::: (7.24b)
and an expansion similar to the one for p is valid for density r.

At the leading order, one finds that r0(t) or P0(t) belongs to a family of adiabatic
thermostatic evolution of the gas in the cavity (container-bounded domain), and are
determined from the overall conservation of mass. That is, they do not depend on
position either. Furthermore:
P0 t r0 tg and r0 t 1=Vt: (7.25)
The pair (u0, p2) belongs to a so-called quasi-incompressible model, but, as a

second peculiar feature, we find that it is perturbed by the pair u ; p , which
consists in acoustic oscillations generated during the setting-up of the motion.
Therefore, as demonstrated below, we have:
u Sn
1 Bn tsinfn t=MUn t; x; (7.26a)
r r0 tr0 =P0 t1=2 Sn

1 Bn tcosfn t=MRn t; x; (7.26b)
where
dfn t=dt P0 t=r0 t1=2 on t: (7.26c)
In (7.26c), on(t) is one of the acoustic frequencies corresponding to the shape of

the container at time t, while the pair fUn t; x; Rn t; xg serves to define the normal
mode of oscillation at frequency on(t) normalized to:

Un 2 Rn 2 dv 1
Ot
where the integral is over the bounded container.

A third peculiar feature is that the first correction due to compressibility
corresponds to the pair,
u2 ; p4 o4 t; where 4o4 t g Sn
1 Bn 2 jUn j2 : (7.26d)
The Bn are readily found as function of time t by writing down conservation of

acoustic energy. We then obtain, in particular:
r0 t1=2 Bnt Bn0: (7.26e)
On a time t ORe=M1=2 this acoustic energy is mainly damped out by

viscosity and heat conduction within a Stokes-like boundary layer (see Zeytounian
and Guiraud [85]).
Of course, turbulent mixing would be much more efficient, and the long time
persistency of acoustic oscillations is mainly a further proof that laminar mixing is
very poor!
It must be emphasized that on a laminar basis, even when the transient acoustics
has been damped out, P0(t) and r0(t) remain adiabatically related, but a much longer
time would be needed for inducing isothermal evolution.
A final feature should be pointed out. Under resonance conditions (u*, p*) gains
energy from the motion of the container, and their limiting amplitude is derived
from a fairly complicated non-linear process which is understood only in one-
dimensional situations (Chester [148], Rott [149]).
7.2.3 Derivation of an Average Continuity Equation
First we should use the time t, a slow time, and then we would bring into the
solution an infinity of fast times designed to cope with the infinity of periods of free
vibrations of the cavity O(t). Below we set U for the solution U expressed through
this variety of time-scales, and in such a case we write
@U=@t @U=@t 1=MDU (7.27)
where U/t stands for the time derivative computed when all fast times are
maintained constant, while (1/M)DU is the time derivative (with D, a differential
operator) occurring through all the fast times.
We carry such a change into the starting Euler equations (7.22a), and then
expand according to:
U U0 M U1 M2 U2 :::; with U hUi U (7.28)
where hUi is average (over all rapid oscillations and depends only of the slow time
t and space position x) of U, and U* is the fluctuating (oscillating, which depends of
all fast times) part of U. More precisely, the operation
U ) hUi (7.29a)
erases all the oscillations associated with the fast times, and obviously
DhUi 0: (7.29b)
For instance, for the fluctuating parts of u*0 and r*1, we can write, respectively,
as a (more complete) solution:
u0 Sn
1 An t Cn Bn t SnUn (7.30a)
r1 r0 tr0 t=p0 t1=2 Sn

1An t Sn Bn t CnRn (7.30b)
with
Cn cos1=Mnt and Sn sin1=Mnt (7.31a)
hCp Cqi hSp Sqi 1=2dpq;

(7.31b)
and dpq 0; if p 6 q; dpq 1; if p q;
hCni 0; hSni 0; hCp Sqi hCq Spi 0; (7.31c)
DCn dn t=dtSn; DSn dn t=dtCn; (7.31d)
where again
dfn t=dt p0 t=r0 t1=2 on ; fn 0 0: (7.32)
In (7.30a, 7.30b) Un and Rn are the normal modes of vibrations of O(t) with
eigen-frequencies on: namely,
on Rn r:Un 0; on Un rRn 0; Un :nSt 0: (7.33)
The relation (7.32) defines the scales of the fast times in relation to the speed of
sound in cavity (at the time t) and with the eigenfrequencies of the cavity at the
same time.
With (7.27), from the Euler equations (7.22a) we obtain the following equations
for the functions u, r, p, and S:
Dr M @r=@t u:rr rr:u 0

1=grp MrDu M2 r@u=@t u:ru 0
(7.34a)
DS M @S=@t u:rS 0
p rg expS;
with the slip condition
u:nSt Wt; P: (7.34b)
From the expansion (7.28), at the zero-order, from the above system (7.34a) we
derive:
Dp0 0; Dr0 0; DS0 0
which shows that r0 and S0 are independent of the fast times and, as a consequence
of the equation of state, that this is also the case for p0, which is, in fact, a function
of only the slow time t:
p0 p0 t; r0 r0 t; x; S0 S0 t; x: (7.35a)
Now, at the first order, from the equation for S, we derive the equation
DS1 @S0 =@t u0 :rS0 0
and, since S0 S0 t; x is independent of the fast time and DhS1 i 0, we have the
following average equation for S0 t; x:
@S0 =@t hu0 i:rS0 0: (7.35b)

But, close to initial time t 0, when we consider the Euler equations (7.22a)
written with the short time, t t=M, in place of the slow time, t, we use the
local-in-time asymptotic expansion
S Sa0 MSa1 MSa2 :::
where Sak Sak t; x; k 0; 1; 2; :::; and the initial condition
S 0 at t 0
We derive
@Sa0 =@t 0; @Sa1 =@t 0 ) Sa0 0; Sa1 0
and, as a consequence, from (7.35b), by continuity
S0 0: (7.35c)
On the other hand, with (7.35c) we obtain
p0 t r0 tg (7.35d)
the function r0 t being determined by the relation

r0 tdv r0 t dv ) r0 tjOtj m
Ot Ot
where jOtj is the volume of the cavity, such that

djOtj=dt W t; P ds (7.35e)
St
and m ( const) is the whole mass of the cavity, and, according to the initial
condition for the density we have jO0j m .
If, in particular, we assume that r0 t 1 (and as a consequence p0 t 1 also)
then jOtj m const. Obviously, this is not the case in the various applications!
At the first order, from the first three equations of (7.34a), with the above results,
we derive the following two equations:
D r1 dr0 =dt r0 r:u0 0, (7.36a)
1= grp1 r0 Du0 0, (7.36b)

with, from (7.34b),
u0 :nSt W t; P: (7.36c)
Since, Dh r1 i 0, from (7.36a) we derive an average (zero-order) continuity

equation:
1=r0 dr0 =dt r:hu0 i 0: (7.36d)
with (from (7.36c))
hu0 i:nSt W t; P: (7.36e)
From (7.36b), since Dhu0 i 0, we also have:
rhp1 i 0 ) hp1 i 0; p1 p1 : (7.36f)
7.2.4 Solution for the Fluctuations u0 and r1
For the fluctuations we derive, from Eqs. 7.36a, 7.36b with 7.36c, the following
acoustic-type equations with slip condition:
D r1 r0 r:u0 0; (7.37a)
1= grp1 r0 Du0 0; (7.37b)
with

u0 :n S t
0: (7.37c)
Concerning the equation for the specific entropy, we have (because S0 0):
DS1 0 ) S1 0; (7.37d)
and as a consequence, from the equation of state, for the fluctuation of the pressure,
we derive
p1 gp0 =r0 r1 : (7.37e)
As a consequence of (7.36f) we also have

hr1 i 0 ) r1 r1 (7.37f)
The solution of the two equations for u0 and r1 = r0 , obtained from (7.37a,
7.37b) when we use (7.37c), is given by (7.30a, 7.30b).
Indeed, if we use the solutions (7.30a, 7.30b) in Eqs. 7.37a and 7.37b, then:
Dr1 =r0 r:u0 Sn

1 An t Cn
Bn t Snf r0 t=p0 t1=2 dn t=dtRn r:Un g
and
p0 =r0 r r1 =r0 Du0 Sn

1 An t Sn
Bn t Cnfp0 t=r0 t1=2 rRn dn t=dtUn g
and using (7.31a7.31d) and (7.32) we determine that the right-hand side of the
above equations are quite zero.
We observe also that the eigenfunctions (the normal modes of vibrations of Ot
with eigenfrequencies on ), Un and Rn , are normalized according to:

Un 2 Rn 2 dv 1
Ot
It is now necessary to determine, from (7.34a, 7.34b), the equations for the
second-order approximation, and then derive, first, a system of two equations for
the amplitudes, An t and Bn t, which present the possibility of considering the
long time evolution of the rapid oscillations.
However, it is also necessary to derive an equation for the average value of u0 ,
which gives, with the average continuity equation (7.36d), a system of two average
equations for hu0 i and hp2 i.
7.2.5 The Second-Order Approximation
We return to system of Eqs. 7.34a with 7.34b, and consider the second-order
approximation for S2 ; p2 , and r2 . First, we obtain:
DS2 @hS1 i=@t hu0 i:rhS1 i 0
but according to (7.37d), S1 0, and also
@hS1 i=@t hu0 i:rhS1 i 0:

With zero initial condition at t 0, we have (since S1 0):
S1 0; and then : S2 0 (7.38a)
For the third-order approximation we have
DS3 @hS2 i=@t hu0 i:rhS2 i 0
and again (according to the second relation in (7.38a)) we obtain
S2 0 and S3 0 (7.38b)
Finally, from the equation of state, when we take into account that
S0 S1 S2 0
we derive the following relation between p2 and r2 :
p2 gp0 =r0 r2 1=2r0 g 1 r1 2 : (7.38c)
Now, again from the system of Eq. (7.34a), we derive two second-order
equations:
Dr2 r0 r:u1 @r1 =@t u0 :rr1 r1 r:u0 0 (7.39a)
rp2 =gr0 Du1 r1 =r0 Du0 @u0 =@t u0 :ru0 0 (7.39b)
with
u1 :nSt 0 (7.39c)
From (7.39b) we now have, first, the possibility of deriving the following
average equation for hu0 i:
@hu0 i=@t hu0 :ru0 i rhp2 i=gr0 1=r0 hr1 Du0 i 0: (7.39d)
This equation is explained below.

On the other hand, using (7.38c), we write Eqs. 7.39a, 7.39b as an inhomoge-
neous, acoustic-type, system for r2 =r0 and u1 : namely,
Dr2 =r0 r:u1 G 0,
p0 =r0 r r2 =r0 Du1 F 0, (7.40a)
u1 :nSt 0,
where
G @r1 =r0 @t r: r1 =r0 u0 1=r0 dr0 =dtr1 =r0 ; (7.40b)
F @u0 =@t u0 :ru0 g 2p0 =r0 r1 =r0 rr1 =r0 : (7.40c)
In G and F, according to (7.40b, 7.40c), we have three categories of terms:

1. The (average hGi and hFi) terms independent of the scale of the fast times.
2. The terms GL and FL which are linearly
dependent on the Cn, and Sn.
3. The terms which depend quadratically GQ and FQ on the Cn and Sn, and are
proportional to cos1=Mp t q tor sin1=Mp t q t.
As a consequence we write, in system (7.40a) for the inhomogeneous terms
G and F, the following formal representation:
G hGi p0 t=r0 t1=2 Sn

1 GnC Cn GnS Sn GQ ; (7.41a)
F hFi p0 t=r0 tSn

1 FnS Sn FnC Cn FQ ; (7.41b)
where hGi and hFi and also coefficients. GnC ; GnS ; FnC ; FnS , are determined from
(7.40b, 7.40c).
More precisely, in (7.41a, 7.41b), the terms hGi and hFi indicate the terms
independent of fast times, while in the Sn
1 we have the terms with Cn and Sn
according to (7.30a, 7.30b). On the other hand, in GQ and FQ we have the terms
proportional to
cosp q =M or sinp q =M:
Below we assume that the last quadratic terms, GQ and FQ, are not resonant
triads satisfying the relation:
jp t q tj r t; 8p; q; r (7.41c)
Thus, none of the quadratic terms can interfere with any of the terms depending
linearly on the Cn and Sn.
As a consequence of the linearity of our system (7.40a), we can, in particular,
write the solution for the fluctuations r2 =r0 and u1 , corresponding only to the
terms linearly dependent on the Cn and Sn in (7.41a, 7.41b), in the following form:
r2 =r0 Sn
1 RnC Cn RnS Sn (7.42a)
u1 p0 =r0 1=2 Sn

1 UnC Cn UnS Sn (7.42b)
and, for example, the amplitudes RnS and UnC satisfies the system:
on RnS r:UnC GnC 0
on UnC rRnS FnS 0 (7.42c)
UnC :n 0
on St:
Obviously, for RnC and UnS we obtain a similar system when in place of RnS,
UnC, GnC and FnS we write RnC, UnS, GnS and FnC.
For the existence of a solution of both these inhomogeneous systems it is
necessary to use two compatibility relations (which are, in fact,a consequence
of
the Fredholm alternative), respectively related to (GnC, FnS) and GnS ; FnC , and for
this the system (7.33), for the normal modes (Rn, Un) of vibrations of the cavity O(t)
with eigenfrequencies on, must be taken into account.
Therefore, from (7.33), after an integration by parts, it follows that

0 fon Rn r:Un RnC on Un rRn UnC gdv
Ot
(7.43a)
fon RnC r:UnC Rn on UnC rRnC Un gdv;
Ot
when we also take into account the boundary on @Ot St, the conditions:
UnC :n 0; and UnS :n 0; on St: (7.43b)
As a consequence, we derive the following compatibility condition for the

resolvability of the above, (7.42c), inhomogeneous system:

GnC Rn FnS :Un dv 0 (7.44)
Ot
Of course, a compatibility relation similar to (7.44) is verified if we write, in

place of GnC and FnS, respectively, GnS and FnC, after the use of a system similar to
(7.42c) for RnC, UnS, with GnS and FnC.
7.2.6 The Average System of Equations for the Slow Variation
With the average continuity equation (7.36d) and slip condition (7.36e), for hu0 i,
we lack sufficient information for the determination of the slow (nearly incom-
pressible) variation! Such information is derived from the average equation (7.39d).
Again, therefore, according to solution (7.30a, 7.30b), we first obtain:
hu0 :ru0 i hu0 i:rhu0 i
1=2Sn
1 An 2 Bn 2 Un :rUn (7.45a)
and
1= r0 h r1 Du0 i 1=2Sn
1 An 2 Bn 2 Rn :rRn (7.45b)
when we also make use of (7.33). From this equation we also derive the relation:
Un :rUn 1=2jrUn j2
Finally, forhu0 i we derive the following average equation of motion:
@hu0 i=@t hu0 i:rhu0 i rP 0 (7.46a)
where
P hp2 i=gr0 1=4Sn

1 An 2 Bn 2 fjUn j2 jRn j2 g (7.46b)
is a pseudo-pressure affected by the acoustic perturbations.
But, hu0 ijt0 being irrotational (according to a detailed investigation in [85],
Sect. 4 and see also Sect. 7.2.7 below), and according to the average equation
(7.46a), for hu0 i, it remains irrotational for any time t > 0:
hu0 i r: (7.47)
In such a case, the average continuity equation (7.36d) for hu0 i, with the slip
condition (7.36e) on the wall S(t), allows us to determine hu0 i due to the following
Neumann problem for potential function :
D d log r0 t=dt; (7.48a)
with
d=dnSt Wt; P: (7.48b)
In such a case, for the first term in P given by (7.46b), we write:
hp2 i=gr0 @=@t 1=2jrj2
1=4Sn
1 An 2 Bn 2 fjUn j2 jRn j2 g (7.48c)
which take into account, explicitly, the influence of the acoustics on the averaged
pressure hp2 i.
The term
1=4Sn
1 An 2 Bn 2 fjUn j2 jRn j2 g;
in (7.48c) is a trace of the acoustics, in the model problem (7.48a, 7.48b) with
(7.48c) a sequel (a memory of the acoustic oscillations) of the application of the
homogenization technique.
We observe also that as an initial condition for a hu0 i, at t 0, solution of the
average equation (7.46a) with (7.46b), according to solution (7.30a) and the starting
initial condition (7.22d), we can write:
hu0 i Sn
1 An 0Un 0; x 0 for t 0 (7.48d)
7.2.7 The Long Time Evolution of the Fast Oscillations
With the above derivation of the average system of equations for slow variation we
have eliminated only part of the secular terms in u1 and r2 . As a consequence it is
necessary to consider in detail the system of compatibility conditions (7.44) for GnC
and FnS, and similarly for GnS and FnC.
First, we consider GnC, FnS, GnS and FnC, and take into account the relations
(7.40b, 7.40c), (7.41a, 7.41b) and the solution (7.30a, 7.30b), for u*0 and r*1, with
u0 hu0 i u0 and r1 r1 .
A straightforward but technically long calculation produces the following
formulae:
GnC r0 =p0 fdBn =dtRn Bn @Rn =@t dlogr0 =dtRn r:hu0 iRn g
(7.49a)
FnS r0 =p0 fdBn =dtUn Bn @Un =@t Un :rhu0 i

(7.49b)
hu0 i:rUn g
GnS r0 =p0 fdAn =dtRn An @Rn =@t dlog r0 =dtRn

(7.49c)
r:hu0 iRn g
FnS r0 =p0 fdAn =dtUn An @Un =@t Un :rhu0 i

(7.49d)
hu0 i:rUn g
and we observe that in (7.49a) and (7.49c), according to (7.36d):
r:hu0 iRn Rn r:hu0 i hu0 i:rRn

(7.50)
hu0 i:rRn dlog r0 =dtRn
Then, from the compatibility relation (7.44) with (7.49a, 7.49b), and from a
similar (to 7.44) compatibility relation, together with (7.49c, 7.49d), we derive the
two ordinary differential equations for the amplitudes An(t) and Bn(t), taking into
account the normalization condition: namely,
dAn =dt gn tAn 0, (7.51a)
dBn =dt gn tBn 0, (7.51b)

where gn t 1=2 @=@tjUn j2 jRn j2 dv
D t

1=2 fhuo i:rjUn j2 jRn j2 gdv
D t

Un :rhuo i:Un dv
D t
which can be rewritten as

gn t 1=2dlog ro =dt Un :rhuo i:Un dv: (7.51c)
D t
This above relation is derived when we take into account that, respectively:

1=2 @=@tjUn j2 jRn j2 dv 1=2 jUn j2 jRn j2 W t; Pds
D t St
due to normalization and (7.35e), and also that

1=2 fhuo i:rjUn j2 jRn j2 gdv
D t

1=2 rfjUn j2 jRn j2 hu0 igdv
D t

1=2 jUn j2 jRn j2 r:huo idv:
D t
But:

1=2 rjUn j2 jRn j2 huo igdv
D t

1=2 jUn j2 jRn j2 Wt; Pds,
St
due to slip condition (7.36e), and

1=2 jUn j2 jRn j2 r:huo idv 1=2dlog ro =dt
D t
according to continuity equation (7.36d) and normalization condition.

At t 0 we have, as initial conditions from (7.48d),
t+ 0 : Sn>1 An 0Un 0; x huo i (7.51d)
and
Bn 0 0; n 1; 2; ::: (7.51e)
We derive the above initial conditions for An(t) and Bn(t) by applying the
starting initial conditions (7.22d) for u and r, and this gives, first, for An the
condition (7.48d), because u 0 at t 0, when we take into account the solution
(7.30a) for u*0 and also the decomposition (7.28) for U hUi U .
The value of Bn(0) 0 is related with the initial condition at t 0, for r 1,
which is compatible with the leading-order solution:
ro t 0 1; and r1 0;~

x 0:
Due to Eq. (7.51b) for Bn, obviously:
Bn t 0 for all t: (7.52a)
Concerning An(0), its values must be derived from (7.48d/7.51d), and it depends
on the value of < u0 > at t 0. On the other hand, obviously, if in condition
(7.22b) W(0, P) 0, then < u0 > is also zero at t 0, and
An 0 0
which also implies that
An t 0 is zero for all t (7.52b)
and then the oscillations are absent!

However, if the motion of the wall of the deformable in time cavity is started
impulsively from rest (or accelerated from rest to a finite velocity in a time O(M)),
then accordingly we have:
W 0 ; P 0::: but : W 0 ; P 6 0 (7.53a)
and the same holds for the averaged velocity, < u0 >.
In this case we have An 0 6 0, and as consequence:
An t is also non zero; when t

0 (7.53b)
7.2.8 Some Concluding Comments
The most important result we obtain is as follows. If the motion of the wall of the
deformable (in time) cavity, where the inviscid gas is confined, is started impul-
sively from rest, then the acoustic oscillations remain present and have a strong
effect on the pressure. Therefore, this pressure would be felt by a gauge, and would
not be related to the mean (averaged) motion. The same holds if the motion of the
wall is accelerated from rest to a finite velocity in a time O(M).
We again stress the necessity of building into the structure of the non-viscous
solution for U(u, r, p, S), when we consider the Euler equations (7.22a), a
multiplicity of times a family of fast times in contrast to Muller [150], Meister
[151], and Ali [152].
If we deal with a slightly viscous flow, when the Mach number M 1, we must
start from the full unsteady NSF equations. In such a dissipative (viscous and heat-
conducting) case, we bring into the analysis a second small parameter Re1, the
inverse of a (large) Reynolds, Re 1, number, and we must then expect that the
acoustic oscillations are damped out.
Unfortunately, a precise analytical (when a similarity rule between M and Re1
is assumed) multiple time-scale asymptotic investigation of this damping phenom-
enon appears to be even more difficult problem, and raises many questions! This
damping problem is considered mainly in the framework of the hypothesis (see
[13], pp. 148161):
Re >>1=M (7.54)
In M
uller [150], the author provides insight into the compressible NavierStokes
equations at low Mach number when slow flow is affected by acoustic effects in a
bounded domain over a long time! As an example of an application, Muller
mentions a closed piston-cylinder system in which the isentropic compression
due to a slow motion is modified by acoustic waves. Muller uses only a two-time
scale analysis, which is obviously insufficient for the elimination of the secular
terms in derived approximate systems (as has been mentioned in Sect. 7.3.5).
The results obtained recently by Ali [152] are more interesting than those
formally derived by M uller [150], in spite of the fact that in Alis paper a two-
time scale analysis is again used the Euler equations for a compressible perfect
fluid being considered on a bounded time-dependent domain Ot 2 <n , where O0
denotes the domain at the initial time t 0. The evolution of the bounded time-
dependent domain is described by a family of invertible maps:
Ft : <n ! <n (7.55a)
depending continuously on the time t, such that Ot Ft O0 for all t.

This severe assumption on the domain Ot is, nevertheless, general enough to
include a moving rigid domain, or a cylinder cut by a fixed surface and a moving
surface (piston problem), or a contractingexpanding sphere (star). In the particular

case of a moving rigid domain, Ali [152], p. 2023, writes:
Ft x x ct: (7.55b)
The map Ft has a geometric meaning and is related neither to the fluid motion
nor to the Lagrangian variables. Moreover, Ft does not need to be globally unique,
since only its restriction to a neighbourhood of the boundary @Oo characterizes the
motion of the domains boundary @Ot .
From the conclusions of Ali [152], pp. 20372038, we mention that his analysis
is not conclusive, since the theory presented is not capable of providing a full
resolution of high-frequency acoustics. Nevertheless, the representation derived, in
Sect. 6 of his paper, provides a hint of a partial theoretical comprehension of the
acoustic modes generated by the motion of the boundary.
Obviously, the main key point is that one fast time variable is not sufficient to
describe the sequence of modes produced by a generic motion of the boundary.
Thus we need to extend the Ali [152] theory to include a family of fast time
variables non-linearly related to the slow time and (eventually) to the space
variables.
It is an open question whether the number of independent fast variables for each
term of the asymptotic expansion should be increased with the order of the term.
This extension, mentioned by Ali, has a theoretical interest in itself, and is
a necessary step for the development of efficient numerical schemes for low
Mach number flows in a time-dependent bounded domain (as is the case in
a combustion problem). It was, in fact, discovered by J.-P. Guiraud and myself
30 years ago, in 1980, and it is formally realized in Sects. 7.2.17.2.6 above.
Obviously, the case when the starting equations, in place of (7.22a), are the full
unsteady NS-F equations, with
Re O1 and Pr O1 fixed; with M ! 0 (7.56)
is a more difficult problem!

Here we mention only a partial result which concerns the derivation of the
following averaged reduced system:
rhp2 i=g ro t hFi 0

r:hu1 i hGi 0 (7.57)
hHi g 1Tw thGi 0:
This average system merits careful analysis. In the first equation of (7.57),
in < F>, the Reynolds number Re is present, while in the third equation, in < H>,
the Peclet (Pe PrRe) number is present. From this set of equations a
NavierFourier-type nearly incompressible average system of equations has been
derived (see [13], pp. 149154).
A difficult problem is also the study of the viscous damping of the acoustic fast
oscillations. Obviously, the inviscid theory developed in Sect. 7.2 do not present the
possibility of investigating this damping process, and this is also the case when
Re O(1) fixed in the framework of a NavierFourier model.
On the other hand, if we deal with a slightly viscous flow (large Reynolds
number, Re 1), we must start from NSF equations, in place of the Euler
equations analysed above, and bring into the analysis a second small parameter:
e2 1=Re <<1 (7.58a)
which is the inverse of the Reynolds number.

We must then expect that the acoustic fast oscillations are damped out. Unfortu-
nately, a precise analysis of this damping phenomenon which appears, for
instance, when a general similarity rule
e2 Mb ; b> 0 (7.58b)
is assumed appears not to be an easy task, and raises many questions.

Therefore, it is first necessary to take into account an acoustic-type inhomoge-
neous system with a family of very slow times via a new operator (D):
d D U in 7:27 (7.58c)
where it is assumed that order d > M!

It appears that as a consequence of the inhomogeneity, a boundarylayer analy-
sis is necessary, which is related with a Stokes-layer of thickness
w2 e2 M: (7.58d)
The analysis of the Stokes-layer equations is rather complicated, but is necessary

for the investigation of this damping process.
A matching condition (evaluating the flux outward from the Stokes layer)
between the acoustic and Stokes-layer components of the normal velocity gives:
w dM ) d w=M Re M1=2 : (7.59)
However, further investigations are necessary if we want to understand how

viscous damping operates when this relation is not satisfied, and other points
meriting investigation include the behaviour of the Rayleigh layer.
For a deeper investigation of dissipative effects in the case of a time-dependent
cavity a problem which has practical interest in the simulation of the starting
process of a space rocket driven by a stream of gases emitted behind it when the fuel
is burned inside it is necessary to consider the similarity rule (7.58b) for large
Reynolds, Re >>1 e<<1, numbers and low Mach, M 1, numbers at least

during the starting (at t+ 0) short time interval.
Various interesting results relating to the above-mentioned combustion
problem are included in our monograph [13] devoted to low Mach numbers:
Chapter 1, pp. 1415, discusses a simple model for combustion (with various
references); Chap. 2, pp. 3233, presents a brief account of the low Mach number
theory applied to combustion (with references); and Sects. 3.3 and 3.4 in Chap. 3
deal with different non-viscous and heat-conducting models in a bounded time-
dependent domain.
Concerning the damping of acoustic oscillations by viscosity, we observe that
the Rayleigh-layer emerges, in the solution of the problem related with the damping
phenomenon, because of the conditions on the wall in Stokes-layer equations. The
investigations of the evolution of the Rayleigh-layers with time is a difficult
problem! If, on the one hand, the Stokes-layer corresponds to acoustic (oscillating)
eigenfunctions of the cavity, the Rayleigh-layer corresponds, on the other hand, to
conditions on the wall of this cavity. Moreover, the thickness of the Stokes-layer,
being given by
w M=Re1=2 (7.60)
is independent of the time and the behaviour of the Stokes-layer, for a large time,
does not have any influence on the Stokes-layer! Concerning the Rayleigh-layer,
however, its thickness grows as the square root of the time, and obviously a deeper
analysis of the interaction between these two boundary-layers, when time increase
to infinity, is required.
A last remark concerning the adaptation to the initial conditions in a time-
dependent bounded container is that our first paper,1 with Guiraud [85], includes
some preliminary results concerning this problem:
Only with the help of a multiple-scale technique, via an infinity of fast times (designed to
cope with the infinity of period of free vibration of the bounded container), do we have the
possibility of eliminating the various secular terms in derived model equations.
Unfortunately, a two-time, simple technique, with t and t t=M, is not ade-

quate, because such a technique does not provide the possibility of eliminating all
seculars terms. The main reason is that the acoustic eigenfrequencies of the
bounded container appear in the internal problem, and because the container is a
function of the slow time t (the time of the boundary velocitywall velocity related
with the deformation of the container in time), these eigenfrequencies are also
functions of the (slow) time t. More precisely, when the wall, at t+ 0, is started
impulsively from rest (t 0), the limiting case
1
This paper was subject of a communication during the 7th Colloque dAcoustique
Aerodynamique in Lyon (France), 45 November 1980 and, also, of a very fruitful discussion
with D. G. Crighton during this Lyons colloque.
M ! 0 with t fixed (7.61)
is singular near the initial time, and it is necessary to consider a local-in-time

limiting case:
M ! 0 with t fixed (7.62)
with t t=M:
In such a case, close to initial time t 0, we derive the classical equations of
acoustics and obtain the corresponding solution (see [13], Sect. 3.3.4) of the
Chapter 3. Unfortunately, this solution of the acoustic problem does not tend to a
defined limit when t tends to infinity, which shows that matching is not possible!
Chapter 8
The RAM Approach in the Benard
Convection Problem
8.1 An Introduction
During my time at the University of Lille1 from 1972 to 1996, and while living in
retirement at 12 rue Saint-Fiacre, Paris, from 1997 to 2009, I published various
papers devoted to the well-known Benard convection problem. As an introduction
to this chapter 8, I present a short account of some of these results, which at least
from my point of view seem valuable.
A first result, obtained in 1983, was published in a short note [153], where
a rigorous RAM formulation of the RayleighBenard (RB) thermal convection
problem is discussed. This result opened the way for a consistent derivation of
the second-order approximate model equations for the Benard problem of thermal
instability with non-Boussinesq effects.
In 1989, by means of a careful dimensionless analysis of the exact Benard
problem of thermal instability for a weakly expansible liquid heated from below,
as a second new result [154] I show that:
If you intend to take into account, in approximate model equations for the Benard problem,
the viscous dissipation term in the equation for the temperature, then it is necessary
to replace the classical shallow convection (RB) equations by a new set of equations, called
the deep convection DC Zeytounian equations, which contain a depth parameter.
A third result, which appears as a quantitative criterion for the valuation of

the role of the buoyancy in the Benard problem, is the following alternative [155]
obtained in 1997:
Either the buoyancy is taken into account, and in this case the free-surface deformation
effect is negligible and we rediscover the classical leading-order RayleighBenard (RB)
shallow convection, rigid-free, problem, or the free-surface deformation effect is taken
into account, and in this case at the leading-order the buoyancy does not play a significant
role in the BenardMarangoni (BM) thermocapillary instability problem.

194 8 The RAM Approach in the Benard Convection Problem
This alternative is related to the value of the reference Froude number Frd
nd =d= gd1=2 based on the thickness d of the liquid layer, magnitude of the gravity
g, and constant kinematic viscosity nd, and for:
RB model problem : Frd <<1 ) d>> nd 2 =g1=3 ; (8.1)
while for the
BM model problem : Frd 1 ) d nd 2 =g1=3 1 mm: (8.2)
On the other hand, a small effect of the viscous dissipation, in the RB model
problem, produces a complementary criterion for the thickness d: namely,
d CTd DT=g (8.3)
such that
nd 2 =g1=3 <<d CTd DT=g dSh ; (8.4)
where C(Td) is the specific heat at constant temperature Td, on z d, and DT is the
difference between the temperature Tw, on z 0, and Td.
A fourth result is my survey of 1998 [156], devoted to various facets of the BM
thermocapillary instability problem.
A fifth result is linked with my lecture notes for the Summer Course (coordinated
by M. G. Velarde and myself) held at CISM, Udine, Italy, in July 2000: Theoretical
aspects of interfacial phenomena and the Marangoni effect modelling and stability,
in [70], pp. 123190.
Finally, in 2009 I published my book Convection in Fluids: A Rational Analysis
and Asymptotic Modelling [27].
8.2 Some Unexpected Results for the Benard Problem of an

Expansible Liquid Layer Heated from Below
Consider a horizontal layer of expansible liquid and assume that an adverse

temperature gradient (bs) is maintained by heating the underside a lower horizon-
tal rigid, z 0, heated plane at temperature Tw.
The occurrence of the phenomena seems to be associated with cooling of the
liquid at its deformable (temperature-dependent) free surface, which in a conduc-
tion motionless state is at level z d. This free surface is exposed to the air, above,
at constant temperature TA and constant pressure pA.
8.2 Some Unexpected Results for the Benard Problem of an Expansible 195
A very slight excess of temperature in the heated liquid layer on z 0, above

that of the surrounding air, Tw > TA, institutes the tessellated changing structure
(according to Thompson (18811882)).
More precisely, the conduction adverse temperature gradient in liquid,
Ts z Tw bs z ) bs dTs z=dz; (8.5)
is directly determined by the difference (Tw TA), via a Newtons cooling law of
heat transfer with a conduction unit constant thermal surface conductance qs:
bs Tw TA =k=qs d (8.6)
where k is thermal conductivity of the heated liquid, and d is the thickness of the
liquid layer, both constant in a conduction motionless state.
Concerning, more precisely, the Newtons cooling law of heat transfer, written
for the basic particular case of a motionless conduction temperature Ts(z), we have:
kTd dTs z=dz + qs Td Ts z TA 0; at z d: (8.7)
the relation (8.6), for bs, being a direct consequence of (8.7).

If we introduce a conduction Biot number:
Bis Td dqs Td =kTd ; (8.8)
then for bs we obtain the following relation in place of (8.6):
bs fBis Td =1 Bis Td gTw TA =d: (8.9)
In the simplest Benard problem of a liquid heated from below, with bs, we have
four main driving effects:
1. The buoyancy directly related to the thermal shallow convection.
2. The temperature-dependent surface tension which is responsible for the
thermocapillary convection.
3. The viscous dissipation, in the equation for the temperature of the liquid layer,
which leads to a consideration of deep thermal convection.
4. The effect related to the influence of the deformable free surface.
These four main effects affect the Benard convection phenomenon, and it is
necessary, from the start of the mathematical formulation of the full Benard
problem for an expansible, viscous, and heat-conducting liquid, to take all of
them into account.
The significant interconnections between the three main facets of Benard
convection are shown in a sketch (with pecked lines) in Fig. 8.1 of our FMIA
90 [27].
RB thermal shallow convection without viscous dissipation; and temperature-

dependent surface tension.
Deep a la Zeytounian thermal convection with viscous dissipation.
Thermocapillary Marangoni convection with temperature-dependent surface
tension.
8.2.1 The Questionable Davis (1987) Upper, Free

Surface, Temperature Condition, and
the Problem of Two Biot Numbers
In the framework of a mathematical formulation of the full Benard problem, the

derivation of a consistent rational boundary condition for the temperature of the
liquid layer, T(t, x, y, z), on a deformable free surface, simulated by the dimension-
less Cartesian equation:
z 1 Zht0 ; x0 ; y0 ; (8.10)
assuming that the gravity vector g gk acts in the negative z direction and
where (a being an amplitude)
Z a=d; (8.11)
is a decisive step but a highly controversial issue!

From 1987, after the publication of the paper by Davis [157], his derived
(p. 407), upper condition was used systematically in nearly all works relative to
thin films, where the so-called Marangoni effect is taking into account (see, for
instance, our Convection in Fluids [27], Chaps. 7 and 8).
Only in 1996, in Parmentier et al. paper [158], was the problem of two Biot
numbers discussed; and in various parts of our [27] this problem is considered and
a corrected condition in place of Daviss [157] erroneous condition is derived.
It now seems appropriate to consider, first, the physical nature of the Benard
problem!
The lower heated plate temperature,
T Tw Ts z 0;
being given data, the adverse conduction temperature gradient bs appears (according
to (8.9)) as a known function of the (positive) temperature difference (Tw TA),
where TA < Tw is also a known constant temperature of the passive (motionless) air,
far above the free surface, when the conduction constant Biot number, Bis(Td), is
assumed known.
But for this it is necessary to also consider the conduction unit constant thermal
surface conductance qs(Td) as starting data for the Benard problem. If so, then
Ts z d Td Tw bs d is assumed to be determined. On the other hand, it
should also be realized that bs is always different from zero, in the framework of the
Benard convection problem heated from below!
As a consequence, the above, defined by (8.8), constant conduction Biot number,
Bis(Td), is also always different from zero. It characterizes the Benard conduction
stage, and makes it possible to determine the purely static basic temperature
gradient bs DT=d with DT Tw Td .
It is also crucial that with this above purely static basic temperature gradient
bs DT=d, in all published papers relative to thermocapillary (Marangoni) con-
vection, there are defined various dimensionless parameters e, Gr or Ra, Ma, and
Bo, and dimensionless temperature y. Therefore (the subscript d in various
temperature-dependent coefficients and dimensionless parameters or numbers is
relative to constant temperature Td):
e aTd DT expansibity parameter (8.12a)
where aTd ad is the constant (at T Td) coefficient of thermal expansion of

the liquid,
Gr e=Frd 2 Grashof number (8.12b)
or
Ra PrGrRayleigh number (8.12c)
where
Pr nd =kd Prandtl number (8.12d)
with, kd kd =rd Cd
Ma gd dDT=nd 2 rd Marangoni number (8.12e)
Bo gd=DTCd a similar to Boussinesq number (8.12f)
and, as dimensionless temperature, we introduce:
y T Td =DT (8.13)
For the convective in-motion stage, in principle Newtons cooling law (which is
a third-type boundary condition (2.42c) on a solid heated wall) can again be used, as
this is the case in almost all papers devotes to BM problems. Therefore, we write
(in dimensional quantities):
kT@T=@n qconv T TA Q0 ; at z d a ht; x; y; (8.14)
with @T=@n rT:n, where Q0 is an imposed heat flux to the environment and is
to be defined.
This condition (8.14) is, in fact, also the starting condition in Davis [157].
In (8.14) our introduced qconv is, indeed, an unknown convection heat transfer
coefficient, strongly different from the constant conduction heat transfer coeffi-
cient, qs(Td), which appears in (8.7) and (8.8).
As observed in Josephs monograph [159], the heat transfer coefficient qconv, in
the convection stage, depends in general on the free surface properties of the
expansible liquid, the unknown motion of the ambient air near the air surface,
and also the spatio-temporal structure of the temperature field. qconv is therefore
a very complicated function. Obviously, from a practical point of view, the above
upper condition (8.14) for the temperature T does not seems not have, in its general
form, any interesting perspectives in various practical applications, simply because
qconv is an unknown. Nevertheless, from a theoretical point of view it seems
preferable to derive, from (8.14), a correct upper condition for the dimensionless
temperature function y, defined by (8.13). This condition (8.14), in dimensionless
form, for y, is written, without any approximation, in the following form:
@y=@n0 Biconv fTd TA =Tw Td yg Q0 =kd bs 0;
or more precisely (n0 is dimensionless)
@y=@n0 Biconv =Bis Td f1 Bis Td yg Q0 =kd bs 0; at z0 1 Z h0 t0 ; x0 ; y0 ;

(8.15a)
with a Biot convective number:
Biconv dqconv =kd : (8.15b)
Condition (8.15a), where not only Biconv is present but also Bis(Td), is a direct
and exact consequence of (8.14), when we take into account Newtons cooling law
in the conduction stage (8.7), which leads to:
dbs Bis Td Td TA ) Td TA =Tw Td 1=Bis Td :
Only after the confusion of Biot convection, Biconv, with Biot conduction,
Bis(Td), do we obtain the (approximate?) Davis [157] reduced condition (with
Q0 0):
@y=@n0 1 Bis Td y 0; at z0 1 Z h0 t0 ; x0 ; y0 ; (8.15c)
where (in (8.15c)), in place of Bis(Td), in front of y, Davis writes a B the

meaning of which is unclear. Perhaps, with his ambiguous temperature free surface
condition, he thinks the problem too difficult to solve!
The fact is, however, that over many years (and today) many interesting papers
have been published with Daviss (8.15c) questionable upper condition for
temperature! In particular, this seems important for the papers devoted to linear
theory (see, for example, Takashimas two papers [160]1). I therefore pose a simple
question: What is the real value of these papers? Several ideas and aspects of this
two Biot problem are discussed in various chapters of my [27].
8.2.2 The Mystery of the Disappearance of Influence

of the Free Surface in the RB Leading-Order
Shallow Thermal Convection Model
First, taking into account the definition of the Grahof number, Gr (8.12b), as a ratio
of e/Frd2, when e<<1 our main (expansibility) small parameter we see that
only if Frd 2 <<1 do we have the possibility, at the leading order, to take into
account the buoyancy effect directly related to the thermal shallow convection
via Gr.
Namely, in the RB case it is necessary to consider the following limiting process:
Gr e=Frd 2 fixed; when e ! o and Frd 2 ! 0; (8.16)
simultaneously.
1
In [160] Takashima derives, from Daviss upper condition (8.15c), the following linear upper
condition:
@yL =@z0 Bis yL h0 0; at z0 1; (8.*)
where yL is the perturbation of y ( y0 + yL + . . .) relative to a steady-state solution, yO 1 z0 ,

when we assume Z<< 1, and in a such case, @y=@n0 @yL =@z0 OZ. If now we consider our
corrected upper condition (8.15a),we derive, in place of Takashimas linear upper condition (*),
for yL, the following linearized upper condition:
@yL =@z0 Biconv 0 yL h0 0; at z0 1; (8.**)
but only if we assume that Q0 kd bs f1 Biconv 0 =Bis g; and Biconv 0 constant. Another
possibility, when we assume that Q0 0, is to consider a Biconv dependent of H 1 + h0 , and
write Biconv H Biconv 1 ZLH 1 h0 , with LH 1 dBiconv H=dH. If now
Biconv 1 Bis , then at the order O() we derive the following upper (at z0 1) linearized
condition:
@yL =@z0 Bis yL h0 LH 1 h0 o; at z0 1 (8.***)
but now we have Bis in front of (yL h0 ), and Bis 6 0, which was assumed equal to a constant value
of Biconv at z0 1. In (***) we also have a third term proportional to h0 .
The mathematical dimensionless formulation of the Benard problem, heated

from below, is the subject of Chap. 4 of [27], where can be found a dimensionless
system of three dominant dimensionless equations for u0 , p, and y, where
p 1= Frd 2 fp pA =gdrd z0 1g (8.17)
which is a companion dimensionless perturbation pressure to y.

In particular, if we take into account that for the density r, at the leading order,
we can write the following approximate equation:
r rd 1 ey: (8.18a)
Then, for u0 , we obtain as a leading-order equation:
du0 =dt0 r0 p e=Frd 2 yk D0 u0 1=3r0 r0 :u0 ; (8.18b)
with
r0 :u0 edy=dt0 : (8.18c)
On the other hand, it is necessary, at the upper free surface, to take into account a
jump condition for the difference of pressure (ppA). With (8.17), in dimensionless
form, we obtain for p as an upper, free surface the following dimensionless
condition (at z0 1 h0 ):
h i
p =Frd 2 h0 t0 ; x0 ; y0 @u0i =@x0j @u0j =@x0i n0i n0j
(8.18d)
We Ma yr0k : n0 2=3 edy=dt0
where (1/2) (0 k . n0 ) is the mean curvature of the free surface.

Because Frd 2 ! 0, in (8.18b) and also in (8.18d), if we want, for e<<1, to take
into account the buoyancy effect, we see that, necessarily, we have the following
similarity rule between Z and Frd2
=Frd 2 O1; when and Frd 2 ; both ! 0 (8.19)
As a conclusion, in leading-order RB thermal shallow convection, driven by the

buoyancy force, the deformation of the upper, free surface is negligible.
In fact, at leading order we have the possibility, as free surface in the RB model
problem, of considering the plane z0 1.
But a pertinent (open) question remains. What is to be done for the determination
of the deformation h0 (t0 , x0 , y0 ) of the free surface? It seems that such an equation
for h0 (t0 , x0 , y0 ) was first discovered thanks to our RAM Approach (see [27],
pp. 106, 135).
Therefore, with the above upper, free surface (exact) condition (8.18d), written
at level z0 1 (because of the limiting process (8.19)), it is necessary to also
consider that when ! o the second term (proportional to n0 i n0 j) in (8.18d)
gives only a term:

2 @u03 =@x03 ; (8.20a)
and from a third term in (8.18d), in place of (r0k : n0 ), we derive:
@ 2 h0 =@x01 @ 2 h0 =@x02 :
2 2
(8.20b)
Finally, from the condition (8.18d) and relation (8.20b), when we assume that
(which is usually the case)
We sd d= nd 2 rd Weber number>>1; (8.21a)
such that
We We = O1 (8.21b)
we derive the following equation for the determination of the free surface deforma-
tion h0 (t0 , x0 , y0 ) in the leading-order e ! 0:
@ 2 h0 =@x01 @ 2 h0 =@x02 =We h0

2 2
(8.22)
1=We pt0 ; x0 ; y0 ; z0 1;
It should be noted that in the definition of Marangoni, Ma, and Weber. We,
numbers respectively by (8.12e) and (8.21a), the coefficients (function of Td) gd and
sd are related with a linear law for the temperature-dependent surface tension:
sT sd gd T Td with gd dsT=dTd : (8.23)
8.2.3 Influence of the Viscous Dissipation
Only an attentive examination of the dominant dimensionless energy equation, for

dimensionless temperature y, presents the possibility of understanding the influence
of the viscous dissipation phenomenon in thermal convection.
Indeed, Turcotte et al. [161] observe that the viscous dissipation term, in the
dimensionless energy equation, is linked with a dissipation number:
Di 1=2nd 2 =d 2 =DTCd

which can be rewritten in the following form (Zeytounian [155]):
Di 1=2 Bo (Frd 2 eBo=2Gr; (8.24)
where Bo is defined in (8.12f).

On the one hand, we see that when we assume that the Grashof number Gr is 0(1),
in thermal convection (buoyancy effect), according to (8.24), where always e<<1,
the constraint:
Bo 1 (8.25a)
in (8.24), leads to
Di ! 0; (8.25b)
and we derive the classical RB, thermal shallow convection model problem, where
in the energy equation for the dimensionless temperature y, the viscous dissipation
is absent (neglected at the leading-order).
In this case we derive, for the thickness dSh, in (8.4) of the liquid layer in the RB
model problem, precisely the relation:
CTd DT=g d;
mentioned in (8.4).
On the other hand, again with e<<1, Gr O(1), obviously, only the case
Bo >>1; (8.26a)
with
eBo O1 (8.26b)
allows us to take into account the viscous dissipation at leading-order in the

dominant dimensionless energy equation for dimensionless temperature y; where
the term proportional to Di* is present.
In such a case we derive the deep thermal convection model problem
(Zeytounian [154]).
Section 8.3 of [27] presents these deep thermal convection model, leading-order,
equations, together with a new (depth) parameter:
Di aTd gd=CTd : (8.26c)

8.3 The Marangoni Effect 203
8.3 The Marangoni Effect
Marangoni (1865) provided a wealth of detailed information on the effects of

variations of the potential energy of liquid surfaces and, in particular, flow arising
from variations in temperature and surfactant composition. Among the phenomena
involving Marangoni flows we observe that associated with the name of Benard
(1900), which refers to the formation of a cellular structure in a thin liquid layer
heated from below.
Chapter 7 of our recent monograph (2009) [27], presents (in the framework of
67 pages) detailed formulation, discussions, and various reflections related to
BenardMarangoni thermocapillary convection.
In fact, today, a detailed understanding of flows in thin viscous liquid films with
a deformable free surface with a temperature dependent tension is important for
a wide range of modern engineering processes. For this it is necessary to build
mathematical models that can predict the various performance of these processes,
in order to have confidence in the predictions of the derived models. For this, our
RAM Approach is obviously a well-adapted and invaluable tool!
A recent special issue of J. Engng. Maths. [162] includes various interesting
results related to the dynamics of thin liquid film. However, many of these papers
include results based on ad hoc non-rational procedures, and do not have any
relation to our RAM Approach rigorously based on the postulate formulated in
Chap. 6.
It is necessary to have in mind that in the Benard experiments (thin layer of an
expansible liquid heated from below) the influence of the free surface is present,
and obviously generates the thermocapillaryMarangoni effect.
On the other hand, in the simple Rayleigh thermal convection problem the free
surface is replaced by the plane z d. Because Lord Rayleigh, in his well-known
paper of 1916 [163], considered a liquid layer with a constant thickness d, the
MarangoniBiot effects are obviously absent in the exact formulation (a la
Rayleigh) of the thermal convection problem. The effect of the constant surface
tension is also absent, and the Weber number does not appear in Rayleighs
formulation of the thermal convection problem. As a consequence, we see that
the Rayleigh analytical problem (considered in 1916) has no relation with the
physical experimental problem considered by Benard in his various experiments
in 1900. In the Rayleigh analytical problem the main driving force, which gives a
bifurcation from a conduction motionless regime to a convective motion regime, is
the buoyancy!
Nevertheless, Rayleighs theoretical problem, leading to the well-known RB
instability problem, is a typical problem in hydrodynamic instability and
represents a transition to turbulence (chaos) in a fluid system. It is now well
known (mainly thanks to Pearson [164]) that:
Benard convective cells are primarily induced by the temperature-dependent surface
tension gradients resulting from the temperature variations along the free surface the
so-called Marangoni thermocapillary effect.
For the BenardMarangoni (BM) problem, in leading order, the equations are
those which govern an incompressible viscous liquid (a NavierFourier-type
equations) both the buoyancy force and viscous dissipation effect being neglected
but with an energy equation for y.
The main difference with the RB thermal convection model problem is the
influence of the free surface deformations, which are taken into account, and as a
consequence the upper, free surface conditions, in the BM model problem, are very
complicated, and it is just in these upper conditions that the Marangoni, Weber, and
two Biot numbers appear! (Chaps. 6 and 7 of [27] discuss a detailed derivation of
this BM model problem).
In fact, the inappropriateness of Lord Rayleighs (1916) analytical model to
Benards experiments was not adequately explained until Pearson, in 1958 [164]
showed (in a simplified ad hoc linear theory) that:
Rather than being a buoyancy-driven flow, Benard cells are a direct consequence of a
temperature-dependent surface tension.
Curiously, although Benard (as a physicist) initially assumed that surface tension
at the upper free surface of a thin layer film was an important factor in his
discovered cell formation, this idea was abandoned by Benard for some time as
the result of the work of Rayleigh in 1916 [163], where he in fact analyzed the
buoyancy-driven natural thermal convection of a layer of fluid heated from below.
Rayleigh found that if hexagonal cells are formed, the ratio of the spacing to cell
depth almost exactly equalled that measured by Benard an agreement which we
now know to have been fortuitous!
Indeed, it was the experimental work of Block [165] which put to end to the
confusion surrounding the interpretation of Benards experiments, and which
demonstrated conclusively that Benards results were not a consequence of buoy-
ancy but were induced by (temperature-dependent) surface tension. Finally, Block
concluded that:
For thin film of thickness less than 1 mm, variation in surface tension due to temperature
variations (Marangoni effect) were the cause of Benard cell formation and not buoyancy as
postulated by Rayleigh in his 1916 paper.
Only 15 years later, via an alternative [155], I presented an asymptotic rational

well-argued formulation for the Block (1956) conclusion:
Either the buoyancy is taken into account, and in such a case the free surface deformation
effect is negligible and we have the possibility to take into account in the Rayleigh-Benard
(RB), leading-order, shallow thermal convection model problem the Marangoni
thermocapillary effect only partially, or the free surface deformation effect is taken into
account, and in such a case the buoyancy does not play a significant leading-order role in
the BenardMarangoni full thermocapillary model problem.
We observe also that for a given temperature difference, the ratio between
buoyancy (Archimedean effect) and the surface tension gradient (Marangoni effect)
Gr/Ma varies with d2, and as a result (according to physicists), the Marangoni
effect dominates for small thickness of the liquid layer, and buoyancy effects
dominate for very thick liquid layers! Therefore (Bd is the dynamic Bond
number):
Gr=Ma Bd rd ad =gd gd 2 : (8.27)
Twenty-five years ago, when I first read the above sentence in Guyon, Hulin, and
Petits Hydrodynamique Physique [166] which was mainly inspired by de
Gennes various courses at the Ecole Superieure de Physique et Chimie Industrielle
in Paris I understood that these two effects should (certainly) be related to two
particular values of a single dimensionless reference parameter. A little later, I
discovered that the Grahof number Gr ad DTgd3 =nd 2 is, in fact, a ratio of two
dimensionless parameters:
expansibility parameter e ad DT
to squared Froude number Frd 2 nd =d2 =gd
For me, these two facts have been an illuminating indication that Frd2 in the ratio
Gr e= Frd 2 ; where the thickness d of the liquid layer is present, must play a
decisive role, because usually e<<1! This observation allowed me to formulate
the above-cited alternative, published in 1997 [155] (and see also our survey of
1998 [156]).
8.3.1 The Long-Wave Approach
The full BM thermocapillary problem formulated in Sect. 7.2 of [27] even in the
framework of a numerical simulation is a very difficult, awkward, and tedious
problem, mainly because of the complicated form of upper, free surface boundary
conditions.
It is clear that simplifications in a rational approach are necessary, obviating the
need for computationally expensive (in time and money) fully numerical
simulations while at the same time preserving essential elements of the physics
of the formulated BM thermocapillary convection model problem.
Among various approaches linked with this BM model problem, the formation
of long waves with respect to a very thin film layer at the surface of a falling film
is a challenging problem (which is the case for a free-falling film down a uniformly
heated vertical plane).
In a very thin film, obviously, a typical length l of the (long) waves is large
in comparison with the thickness d<< l of the thin film, so that the slope of the free
surface is always small. In such a case we have the advantage of introducing a
long-wave dimensionless parameter:
d d=l<<1 (8.28)
The essential advantage of the long-wave approximation is a drastic simplifica-

tion of the full dimensionless BM model problem, mainly because the upper, free
surface conditions are significantly simplified.
Section 7.3 of our [27] includes some BM long-wave reduced convection
model problems. These simplified model problems are derived via the introduction
of new coordinates
X dx; Y dy; Z z; (8.29a)
and new time
T dRed t; (8.29b)
where
Red Uc d=nA ; (8.29c)
Uc being determined by a similarity relation (see below) and nA nTA .

On the other hand, in place of leading-order (when e ! o) dimensionless
velocity components u0 , v0 , w0 , the following new components are also introduced:
U u0 =Red ; V v0 =Red ; W w0 =dRed (8.30a)
In place of p a
P p=Red 2 (8.30b)
and in place of y (see (8.13)), a (where in place of Td we write TA):
Y T TA =Tw TA (8.30c)
We observe also that the various coefficient functions of temperature T are fixed
at T TA, and e and Ma are defined by (Tw TA) which replace DT.
Finally, via a laborious and tedious transformation, and via the long-waves
limiting process:
d ! 0 and Red !/; such that d Red Re ; (8.31)
we derive, first (0 subscript) the following reduced system of equations for the
BM long-wave problem:
D:V0 @Wo =@Z 0; (8.32a)
DV0 =DT DP0 1=Re @ 2 V0 =@Z2 0; (8.32b)

@Po =@Z 0; (8.32c)
Pr DY0 =DT 1=Re @ 2 Y0 =@Z2 0; (8.32d)
where
D=DT @=@T U@=@X V@=@Y W@=@Z;
D @=@X, @=@Y), D2 @ 2 =@X2 @ 2 =@Y2 ; V U; V:
Equation (8.32c), @Po =@Z 0; is typically a boundary layer equation, and

the system (8.32a)(8.32d) is certainly singular near the initial time T 0, where it
is necessary to write initial data for V0, W0 and Yo the system (8.32a)(8.32d)
being, in fact, only an outer system relative to time valid far to T 0.
As boundary dimensionless conditions, we have, first:
V0 0; W0 0; and Y0 1 at Z 0: (8.33a)
If, now, Z H(T, X, Y) is the equation of the deformable surface, the kinematic
upper condition at Z H : W0 @H=@T V0 :DH; with W0 0 at Z 0, leads
to the following averaged evolution equation for H(T, X, Y):
ZH
@H=@T D: Vo dZ 0; (8.33b)
z0
which plays a central role in lubrication theory.

Above, in (8.29c), we introduced Red dUc =nA via the characteristic velocity
Uc. If, by analogy, we define a new squared Froude number, Fr2 Uc/gd, and
corresponding, modified, Weber (We) and Marangoni (Ma) number via Uc, and
then if we assume that:
Re =Fr 2 G 1; d2 We W 1 and Ma 1; (8.33c)
at leading-order we obtain, for the above Eqs. (8.32a)(8.32d), the following free
surface, at Z H, upper, reduced conditions:
P0 G =Re H 1 W D2 H Po H; (8.33d)
@Vo =@Z Re MaDYo DH@Yo =@Z; (8.33e)
@Yo =@Z Biconv Yo 0: (8.33f)
This derived simplified equations, (8.32a8.32d) with the conditions,

(8.33d8.33f), seems to be a very convenient reduced form, and it is not a bad start
as a reduced BM thermocapillary model problem, subject to numerical simulation!
8.3.2 Towards a Lubrication Equation
We observe also that as a consequence of Eq. 8.32b for V0, with (8.32a), and (8.32c)
for P0, subject to the upper condition (8.33d), we derive for V0 the following
problem (8.34), with the problem (8.35) for Y0:
ZH
@V0 =@T 1=Re @ 2 V0 =@Z2 D:V0 dZ@V0 =@Z V0 :DV0
z0
G =Re DH W DD2 H;
Z o : Vo o;
Z H : @V0 =@Z Re MaDY0 DH@Y0 =@Z ; (8.34)
where the function Y0 is the solution of the model problem:
Pr DY0 =DT 1=Re @ 2 Y0 =@Z2 0;

Y0 1 at Z 0; (8.35)
@Y0 =@Z Biconv Y0 0; at Z H:
Our monograph [27], pp. 213218, includes a more simplified case, when
Pr ! 0 which decouples problem (8.35) from problem (8.34).
Then, if in the derived decoupled problems (when Pr ! 0) we assume
Re ! 0 or : d ! o; with Red fixed; (8.36a)
and as a consequence
l>>Uc =nA d2 ; (8.36b)
assuming also that
W Re W 1 and Re Ma Ma 1; (8.36c)
we have the possibility of deriving the following (rather awkward) lubrication

equation, assuming that Biconv is a function of H denoted by B(H): namely:
@H=@T 1=3D:fH 3 W DD2 H G DH

Ma BHH 2 DH=1 HBH2 (8.37)
2
Ma H dBH=dHDH=1 HBH g 0:
3
In the case of a vanishing convective Biot number, BH ! 0, we see that if in

(8.37) the second term proportional to B(H) disappears, this is not necessarily the
case with the term dB(H)/dH, which need not be zero, and the influence of a large
Marangoni effect remains operative.
Unfortunately, usually in classical lubrication equations, if we consider a
vanishing Biot number, then the Marangoni effect also disappears! This non-
physical (from my point of view) consequence is practically always encountered
in all derived lubrication equations.
At this point it is opportune to observe that in a short paper by VanHook and
Swift [167] it is clearly mentioned that the Pearson result has two Biot numbers
(one for the conduction state and one for the perturbation), while the distinction
between the two Biot numbers has not been made in some experimental papers. A
theoretical analysis, however, should maintain the distinction!
In the unsteady one-dimensional case (T, X), taking into account that
H 1 ZhT; Xwith Z<<1; (8.38)
the linearization of (8.37) produces, at the order Z, a linear equation for the
thickness h (T, X), when B(H) ! 0:
@h=@T 1=3W @ 4 h=@X4 Ma dBH=dHH1 @ 2 h=@X2 G @ 2 h=@X2 0;

(8.39)
From (8.39) we obtain for the cut-off wave number kc (when k > kc there is a
linear instability) the relation:
kc fMa =W dBH=dHH1 G =W g1=2 : (8.40a)
Finally, we observe that in (8.39) the terms proportional to G* and W** are a
stabilizing effect in evolution of the free surface, in the Benard convection problem
of heating from below and in particular, that the thicker the film, the stronger the
gravitational stabilization.
Conversely, the term proportional to Ma*, linked with the thermocapillary (large
Marangoni) effect, has a stabilizing effect on the free surface if
dBH=dH > 0 ! (8.40b)
In particular (see the 1997 survey [168]), thermocapillary destabilization is

explained by examining the fate of an initial corrugated free surface in the linear
temperature field by a thermal condition.
Where the free surface is depressed, it lies in a region of higher temperature than
its neighbours. Therefore, if surface tension is a decreasing function of temperature,
free surface stresses drive liquid on the free surface away from the depression,
because the liquid is viscous, causing the depression to deepen further. Hydrostatic
and capillary forces cannot prevent this deepening!
8.4 From Deep to Shallow Thermal Convection Model

Problems
Thanks to a detailed analysis, we now see that the RAM Approach presents
the possibility, first, to determine the various dimensionless parameters driving
the main four physical effects which govern the mathematical formulation of the
Benard convection problem of a liquid layer heated from below and limited by
a deformable free surface from overlying ambient passive air:
e Characterizes the expansibility of the viscous liquid.
Frd The Froude number, which characterizes the thickness, d, of the liquid
layer.
Bo The Bousssinesq number, which characterizes the importance of the
viscous dissipation..
Ma The Marangoni number, which characterizes the thermocapillarity.
We The Weber number, which characterizes the effect of a constant surface
tension.
Biconv The convective Biot number, which characterizes the transfer via deform-
able free surface in the convection regime.
Then, in a second step, a dimensionless Benard problem is formulated where the

above parameters take their respective place in equations and in lower fixed
boundary and in upper deformable free surface.
Below we consider the case of thermal convection when the buoyancy is the
main driven force. For this case, the buoyancy is linked with the Grashof number,
Gr O(1).
In our RAM Approach the main working hypothesis is relative to a weakly
expansible liquid, e<<1; with
Gr e=Frd 2 O1; (8.41a)
We then see that it is necessary for Gr O(1), the condition:
Frd 2 <<1 ) d>>nd 2 =g1=3 : (8.41b)
On the other hand, the possibility of taking into account the effect of the viscous
dissipation is linked with the condition:
Di eBo O1: (8.42a)
This is possible only if the condition
Bo >>1 ) d>>CTd DT=g; (8.42b)
is satisfied.
8.4 From Deep to Shallow Thermal Convection Model Problems 211
As a consequence of the relations of (8.41a)(8.42b), we can write the following

deep thermal convection with viscous dissipation equations for the leading-order
functions:
limdeep u; y; p uD ; yD ; pD ; (8.43a)
Namely
r:uD o; (8.43b)
@uD =@t uD :ruD rpD GryD k r2 uD ;

(8.43c)
1 Di pd 1 zf@yD =@t uD :ryD g 1=Prr2 yD
1=2Di=Gr@uD i =@xj @uD j =@xi 2 ; (8.43d)
where
Di aTd gd=CTd : (8.44a)
As boundary conditions for these deep thermal convection equations we can

write:
at z 0 : uD 0 and yD 1; at z 1 : wD 0; (8.45a)
at z 1 : @ 2 wD =@z2 Ma@ 2 yD =@x2 @ 2 yD =@y2 ; (8.45b)
at z 1 : @yD =@z Biconv =BiS Td f1 BiS Td yD g 0; (8.45c)
where z x3 ; wD uD 3 uD :k:
We observe that in conditions (8.45c), written at z 1, for yD ; it seems possible
(because (8.45c) is satisfied on z 1) to identify Biconv with BiS(Td) but, in fact,
this is only a conjecture. In such a case we recover at z 1 the Davis (1987)
condition!
Now, if we consider the following (RB!) limiting process:
limDi#0 uD ; yD ; pD uRB ; yRB ; pRB (8.46a)
we find from the deep convection equations (8.43b8.43d) the usual RB equations
for the shallow thermal convection:
r:uRB 0 (8.47a)
@uRB =@t uRB :ruRB rpRB GryRB k r2 uRB ; (8.47b)
@yRB =@t uRB :ryRB 1=Prr2 yRB ; (8.47c)
For the RB equations (8.47a8.47c), the boundary conditions (8.45a8.45c) are

also true but written for, uRB, wRB and yRB .
Chapter 9
The RAM Approach in Atmospheric Motions
In Sect. 9.1 of this chapter we consider the 2D steady lee waves problem, in the
framework of a non-viscous but compressible and adiabatic fluid flow. The starting
equations are Euler two-dimensional steady equations. From these we derive a
single, rather awkward, but very convenient equation for the stream function in the
case of low Mach numbers fluid flow theory, for the application of the RAM
Approach. From this single equation for the stream function we derive a family
of model equations for the lee waves problem by considering various limiting cases.
This example very well illustrates the possibility of a theoretical investigation
before the use of the RAM Approach.
In Sect. 9.2, as an application of the RAM Approach to very difficult,
atmospheremeteo-fluid motions, we consider the derivation of a simplified (but
rather realistic) and consistent meteofluiddynamic model: namely, the low
Kibel number asymptotic model derived from the dissipative hydrostatic
equations.
In particular, our RAM Approach presents the possibility of solving the difficult
but decisive singular problem relative to initial conditions encountered by
meteorologists during their weather forecasting, and also take into account the
influence of the Ekman boundary layer near the Earths surface.
9.1 Some Models for the Lee-Waves Problem
With the dimensions, the steady two-dimensional Euler equations are written in the
following form, for the velocity components u and w, pressure p and density r all
assumed depending on the coordinates x and z:
ru @u=@x w @u=@z @p=@x 0; (9.1a)
ru @w=@x w @w=@z @p=@z g r 0; (9.1b)

214 9 The RAM Approach in Atmospheric Motions
@ru=@x @rw=@z 0; (9.1c)
u @=@x w @=@zp=rg 0: (9.1d)
We consider a mountain, and write as equation for the gravity g gk

plane:
z h Zx=l ; 1=2 x=l 1=2: (9.2a)
The slip condition gives:
w udZ=dx on z h Zx=l : (9.2b)
Again, our non-viscous and non-heat-conducting fluid is a thermally perfect gas,

and the equation of state is with dimensions:
T p=Rr, (9.3)
with R Cp Cv and g Cp =Cv .

First, from the equation of continuity (9.1c), we introduce a stream function
cx; z such that:
ru @ c=@z and rw @ c=@x: (9.4)
9.1.1 First Integrals
Using the relations (9.4), from the equation of the adiabaticity (9.1d), we derive the
following first integral
p rg P c: (9.5a)
where the function Pc is arbitrary and conservative along each streamline

c constant.
Then, from equations (9.1a, 9.1b) we obtain the well-known Bernoulli first
integral, when we exclude the pressure p:
juj2 =2 g= g 1 rg1 Pc gz Ic; (9.5b)
where the function I(c) is a second arbitrary function, also conservative along each
streamline. From these two equations (9.1a, 9.1b) we can also derive a relation for
the vorticity o:
o @u=@z @w=@x rdI=d c 1= g 1p= rdLog P=d c: (9.5c)

9.1 Some Models for the Lee-Waves Problem 215
The two arbitrary functions Ic and Pc are determined from the behaviour
conditions in the upstream unperturbed region (subscript 1), when x ! 1 and
where z is z1 the altitude of the unperturbed streamline at upstream infinity.
Therefore, at x ! 1, we assume:
u U1 z1 ; w 0; p p1 z1 ; r r1 z1 ; (9.6a)
T T1 z1 ; and S1 z1 Cv Logp1 =r1 g : (9.6b)
With (9.6a, 9.6b) we obtain, from (9.5c), the following relation:
dI=dc 1= g 1p=rdLogP=d c
(9.6c)
1=r1 fdU1 =dz1 gR=g 1U1 S1 g;
where
S1 N 2 z1 T T1 ; (9.7a)
and
N2 z1 1=T1 fdT1 =dz1 g g 1= gRg> 0: (9.7b)
9.1.2 An Equation for the Vertical Deviation dx; z
Since the functions Ic and Pc are both conservative along each streamline,
then from the Bernoulli integral (9.5b) we determine the temperature T in the
following form:
T T1 g 1=gRf1=2juj2 U2 1 z1 gz z1 g; (9.8a)
and for the density we obtain the relation
r r1 f1 T T1 =T1 g1=g1 : (9.8b)
Finally, with the above results we obtain an equation for cx; z in an awk-
ward form (derived by Zeytounian in [12], pp. 315330), although here we do not
write this equation.
For our purpose (the lee waves problem) we introduce, in place of c, the vertical
deviation of a streamline, dx; z, in the perturbed flow over a mountain, relative
to its unperturbed altitude at upstream infinity. We therefore write, for the altitude
of a perturbed streamline:
z z1 c dx; z (9.9a)
and in a such case
@c=@x r1 U1 @d=@x; @c=@z r1 U1 1 @d=@z: (9.9b,c)
In place of the slip condition (9.2b) we obtain:
dx; h Zx=l h Zx=l : (9.10)
We observe that at upstream infinity we have the relation

z1
c r1 U1 dz cz1 , z1 1 c (9.11)
o
where z1 1 c is the inverse function of cz1 .

As a consequence of relations (9.9a, 9.9b,c), for the function dx; z we derive
the following partial second-order differential equation:
@ 2 d=@x2 @ 2 d=@z2 r=r1 2 g=U2 1 N2 z1 d

1=2r=r1 2 d=dz1 LogU2 1 expS1 =Cp
1=2d=dz1 Logr1 2 U2 1 expS1 =Cp f@d=@x2 (9.12a)
2
@d=@z 2@d=@z 1g
@Log r=@x@d=@x @Log r=@z@d=@z 1:
The above equation for the deviation in altitude,
dx; z z z1 c (9.12b)
is rather awkward, but very convenient for further analysis in particular when we
assume that the upstream, constant, Mach number, M0 1 1.
For equation (9.12a), with the slip condition (9.10), we also have the following
three conditions:
dx 1; z1 0; (9.13a)
dx; z H1 0; (9.13b)
j dx 1; zj<1; (9.13c)
where H1 is the altitude of the upper level (for instance, the tropopause assumed as
a flat horizontal plane), where the streamlines are undeflected.
The last condition (9.13c) is the only possible physical one, because of the lee-
waves regime downstream of the mountain. However, in equation (9.12a) we also
have, as an unknown function, the density r, and consequently we must return to
the two relations (9.8a, 9.8b), which we transform to an relation for r in which d is
present:
r=r1 g1 1 U2 1 =2Cp T1 r1 =r2 @d=@x2 @ d=@z2

(9.14a)
2@d=@z 1 U2 1 =2Cp T1 2g=U2 1 d 1;
and we have the following upstream infinity condition:
r ! r1 z1 when x ! 1: (9.14b)
The above problem (9.12a) with (9.10), (9.13a, 9.13b, 9.13c), (9.14a), and
(9.14b) for two functions d and r, is strongly non-linear. Below we consider a
simplified case when
U1 U0 1 const (9.15a)
and
dT1 =dz1 Go 1 const; (9.15b)
such that
T1 z1 T1 01 G0 1 =T1 0z1 : (9.15c)
This linear (9.15c) distribution for T1 z1 is very well justified for the usual
meteorological situation in the troposphere, where the lee-waves regime is consid-
ered H1 being the height of the whole troposphere.
The parameter
m0 1 G0 1 H1 =T1 0 (9.15d)
is a reference parameter for the temperature profile at upstream infinity, and in

dimensionless form we have
T1 z1 =T1 0 Yz1 1 m0 1 z1 (9.15e)
where
z1 z1 =H1
is the dimensionless altitude far upstream.

9.1.3 Dimensionless Formulation
First we introduce the non-dimensional density perturbation:
v r r1 =r1 ; (9.16a)
and the non-dimensional vertical displacement of the streamline,
D d=h : (9.16b)
In place of the relation (9.12b) since far ahead of the mountain there is assumed
to be a uniform flow with velocity components U0 1 const; 0 we write:
z
z1 Boz 1=n D; z (9.16c)
Ho
where z is reduced with the vertical length scale H0, characterizing the lee-wave
process, and z1 is reduced with H1 RT1 0=g. In this case we have two ratios:
n H0 =h ; Bo H0 =H1 ; (9.17)
where n is the linearization parameter, and Bo is the Boussinesq number.

In place of the slip condition (9.10) we have the following dimensionless slip
condition:
Dx; z 1=n Zx Zx; (9.18a)
where x x=l , and the condition (9.13b) gives
Dx; z 1=Bo 0: (9.18b)
As a long-wave approximation parameter we have
e H0 =l : (9.19)
Finally, for v; defined by (9.16a), we obtain from (9.14a) the following dimen-
sionless relation:
1vg1 11= Yf1=2 g1M0 1 = n0 2 1=1v2 e2 @D=@ x2

@D=@ z2 2 n0 @ D=@ z n0 2
g1= gBo=n0 D 1=2 g1Mo 1 2 g;
(9.20a)
where
Y Yz; D 1 Bom0 1 z 1=n0 D: (9.20b)

Then, for the function Dx; z we obtain, from equation (9.12a), the following
dimensionless main equation:
Yz; Dfe2 @ 2 D=@x2 @ 2 D=@z2 1=1 ve2 @D=@x@v=@x

@D=@z@v=@z n0 @v=@zg Bo2 =gM021 S0 1 v2 D
(9.21)
n0 =2BoS0 fv2 v e2 @D=@x2 @D=@z2
2n0 @D=@zg;
In equation (9.21), S0 is the hydrostatic stability parameter:
S0 g 1=g m0 1 ; (9.22)
which characterizes the stratification of the unperturbed flow at upstream infinity.

For normal meteorological values in the troposphere, for dry air, we have the
following atmospheric values:
g 1:4;
T1 0 288 C;
m0 1 R=gG0 1 0:19037;
g 1=g m0 1 0:09534:
Conversely,
gR T1 01=2 340; 17m=sec;
and if
34m=sec U0 1 10 m=sec;
then we obtain
0:03 M0 1 0:1:
We observe that for the unknown function Dx; z, related with the lee-waves
problem over and downstream of the mountain which is the solution of equation
(9.21), we have the following conditions:
Dx; 1=n0 Zx Zx; when x 2 1=2; 1=2; (9.23a)
Dx 1; z1 0; (9.23b)
Dx; z 1=Bo 0; (9.23c)

jDx 1; zj<1; (9.23d)
In main equation (9.21) the more important parameter is
K20 S0 Bo2 =gM021 gGA G0 1 =T1 0H0 =U0 1 2 (9.24)
the DorodnitsynScorer parameter, where
GA g=Rg 1=g; (9.25a)
is the dry adiabatic temperature gradient, which plays a fundamental role when
M0 1 U0 1 =gRT1 01=2 1; but Ko2 O1: (9.25b)
An important characteristic of the modelling, via the RAM Approach, of the

above lee-waves problem, governed by the main equation (9.21), with the relations
(9.20a, 9.20b) and conditions (9.23a9.23d), is linked with the requirement that
without fail the parameter Ko2 must be bounded!
9.1.4 Four Distinguished Limiting Cases
We also observe that the strong constraint mentioned above, Ko2 O1, does not
present the possibility of considering the case:
M021 # 0; solely;
because in a such case,
Ko2 S0 Bo2 =g=M02

1 "1!
The asymptotic analysis for a such case is very complicated and deserves
another approach.
9.1.4.1 The Deep Convection Case
The first case, which is linked to the deep convection, is valid in the whole
troposphere when
Bo 01; (9.26a)
but with the following similarity rule between low Mach number and hydrostatic
stability parameter:
M0 1 1 and S0
1; (9.26b)
assumed large, such that
S0 S M021 , G0 1 GA g=RS M021 ; (9.26c)
with
S 01: (9.26d)
When the linearization parameter n0 is fixed (not considered as a small

parameter, for the case of an important elevated mountain) the limiting case
M021 # 0; with 8:26a c; and fixed g and n0 ; (9.27a)
leads, in place of the equation (9.21) and relations (9.20a, 9.20b), at the leading
order for Ddeep and vdeep ; in expansions
D Ddeep M021 D :::; v vdeep M021 v ::: ; (9.27b)
the following reduced system of two equations:
1 vdeep g1 1=f1 Bog 1=gz 1=n0 Ddeep g

(9.27c)
Bog 1=g z=f1 Bog 1=gz 1=no Ddeep ;
f1 Bog 1=g z 1=n0 Ddeep gf e2 @ 2 Ddeep =@x2 @ 2 Ddeep =@z2

1=1vdeep e2 @Ddeep =@x@vdeep =@x @Ddeep =@z@vdeep =@z (9.27d)
2
no @vdeep =@zg Bo =gS 1 vdeep Ddeep 0:
2
As boundary conditions for Ddeep, we have the full conditions (9.23a9.23d).

Combining (9.27c) and (9.27d), we can derive a single equation for the single
function, Ddeep. But this single equation is how much complex and strongly non-
linear, that we do not write a such equation here.
If we assume a complementary (linearization) constraint:
n0
1 such that : 1=n0 = n Mo21 ;
(9.27e)
and n O1 ) h U021 =gg;
then we derive a linear problem, in place of the non-linear deep convection problem
(equations (9.27c, 9.27d) with the full conditions (9.23a9.23d)), for the limit
function:
sdeep 1 g 1= gBo z1=2g1 Dl deep : (9.27f)

where Dl deep limn0"/ Ddeep.

Therefore, we obtain the following linear model equation:
e2 @ 2 sdeep =@x2 @ 2 sdeep =@z2 DBo z = 0; (9.28a)
where the coefficient D0(Boz) is given by:
D0 BozfBo=gS =1 Bog 1=gzg

(9.28b)
Bo=2g2 2g 1=1 Bog 1=gz2 :
As (linearized) conditions for the above linear equation (9.28a) we have:
sdeep x; 0 hx with x 2 1=2; 1=2;

sdeep x; 1=Bo sdeep /; z/ 0; (9.28c)

sdeep x ! /; z< / :
This above linear case, (9.28a)(9.28c), is very similar to the one considered by
Dorodnitsyn in 1950 [169]. The reader can find in Zeytounian [12], pp. 324328,
some results concerning the solution of this equation (9.28a), with (9.28b), under
the associated conditions (9.28c).
9.1.4.2 The Boussinesq Case
The second case is the Boussinesq case (considered in Chap. 4) when Bo and the
Mach number, M, are both small parameters:
Bo 1 and Mo 1 1; but So 01; (9.29a)
such that
Bo=Mo 1 B 01; , Ho Uo =gRT1 0=g;1=2 HB (9.29b)
where HB is the characteristic vertical displacement of the Boussinesq lee-waves

which is only of the order of 1 km!
9.1.4.3 The Isochoric Case
The third case is the isochoric case when g is large and the Mach number, M, is low:
Mo 1 1 such that So = 01 and Bo 01; (9.30a)

9.2 The Low Kibel Number Asymptotic Model 223
but
gMo21 M O1; , Uo 1 RT1 o1=2 : (9.30b)
9.1.4.4 The Very Thin Layer Case
The fourth case is relative to a very thin atmospheric layer, when we assume:
So Bo 1; with Bo 1; and gMo21 1; (9.31a)
such that
Bo=gMo21 1 , H0 Uo21 =g and GA G0 1 Uo21 =gT1 0: (9.31b)
For each of the above cases from the formulated full non-linear problem
equation (9.21) for the function D(x, z), with the relations (9.20a, 9.20b), (9.22), for
, Y, S0, and boundary conditions (9.23a9.23d) we derive a consistent low
Mach number model problem with Ko2 O(1)!
The fourth case is also considered in Zeytounian [12], pp. 328330. The second
(Boussinesq) case is considered here in Chap. 4, and also in [12], chapter 8. The
third (isochoric) case was considered in our thesis [2]. In [2], and also in our 1969
paper [170], the reader can find various results of computations of 2D steady lee-
waves over and downstream of several mountains. In our book [37], the 3D steady
problem is also considered, and in pp. 168170 two typical figures and some
comments are presented. Section 5.4.2, of [27], pp. 164166, concerning the
isochoric 2D steady case, contains various configurations of streamlines.
9.2 The Low Kibel Number Asymptotic Model
In the framework of an asymptotic modelling theory for the atmospheric motions, it

is first necessary to formulate a physically realistic mathematical problem written in
a dimensionless form. For this, in a coordinates frame rotating with the Earth we
consider the full NavierStokesFourier (NSF) equations for a perfect gas which
is viscous, compressible, and heat-conducting dry atmospheric air.
We take into account the Coriolis force (2Vu), the gravitational acceleration
g gk, with k the unit vector directed to the zenith (force of gravity modified by
the centrifugal force), and the effect of the radiative heat transfer.
The above velocity vector u (v, w) is the (relative) velocity vector as observed
in the Earths frame rotating with the angular velocity V O e, with
e ksinf jcosf (9.32)

The algebraic latitude of the origin-point P of the observation on the Earths

surface is f, and around P (in prediction domain D with a diameter L ) the
atmospheric flow is analyzed.
The unknown thermodynamic functions are the density of air r, the atmospheric
pressure p, and the absolute temperature T, such that the equation of state for
thermally perfect gas is taken into account (the dry atmospheric air being assumed
a trivariate baroclinic fluid):
p = R rT; (9.33)
where R is the perfect gas constant.

A main feature of the atmospheric motions is the existence of the standard
atmosphere, which is motionless and dependent only on the vertical coordinate
zS, the standard altitude directed in the opposite direction from the force of
gravity g. The standard thermodynamic functions, pS, rS, and TS, for the standard
atmosphere, satisfy the following equations:
dpS =dzS grS 0;

pS RrS TS ; (9.34)
kS dTS =dzS RTS zS 0;
where R(TS) is the radiative heat transfer in the standard atmosphere (with s as
subscript and assumed a function of TS(zS)).
Below, we work mainly with dimensionless quantities, and in particular the
thermodynamic functions, p, r, T and R(TS(zS)), are reduced relative to pS, rS, TS,
and R(TS) at the ground zS 0. The dimensionless horizontal velocity v (u, v)
and vertical velocity w are non-dimensionalized with U and eU , respectively, and
we assume that, in the hyodrostatic parameter,
e H =L ; (9.35)
the vertical length scale H is of the order of the height of the standard atmosphere
HS RTS 0=g L , assumed homogeneous, such that the Boussinesq number
(ratio of two vertical length scale),
B0 H =HS 1: (9.36)
In (9.35), L is a horizontal length scale of a domain on the Earths surface

ground, such that the influence of the Coriolis force is taken into account.
For dimensional time we write t t, and consider that the Strouhal number:
S L =U t 1 ) t L =U : (9.37a)

The (turbulent) kinematic viscosity coefficient is n n m =rS 0m=r, and
Re U L = n ; (9.37b)
is the Reynolds number, while the (turbulent) heat conduction is represented by a

Prandtl number,
Pr Cp m =k ; (9.37c)
and we set k k for the (turbulent) heat-conduction coefficient.

We will also use, as non-dimensional parameters, a Mach number,
M U =gRTS 01=2 (9.37d)
where g Cp/Cv is the ratio of specific heat capacities at constant pressure (Cp) and
at constant volume(Cv). Related with the Coriolis force we also have a Rossby
number,
1
Ro U =f L Ki; (9.37e)
f t
if we take into account (9.37a), where Ki is the Kibel number (introduced in the
Soviet Union; see, for instance, Monin [94]1 and Kibel [171]) linked with the
Rossby number, since the Strouhal number, S 1. In (9.37e),
f 2O sinf ; (9.37f)
is the Coriolis parameter, where f is a reference latitude. According to Obukhov,

the typical
p horizontal length scale for the synoptic processes is on the order of
(c gRTS 0 being the sound speed)
LOb = c =f 3000 km: (9.37g)
On the other hand, the ratio
L =LOb M=Ki; (9.37h)
is related to the horizontal compressibility of the atmosphere.

Finally, as a radiative heat transfer parameter we have
s RTS 0=gk =R; (9.37i)
1
Monins book (English translation, 1972) includes a concise introduction to physical and
mathematical thinking in meteorology. In Chapter 2, pp. 1478, Monin presents an exposition
of the Hydrodynamic Theory of Short-Range Weather Prediction.
and our Ekman number is related to
Re? e2 Re (9.37j)
by the relation
Ek? = Ki=Re? : (9.37k)
Now, for our RAM Approach below, it is very convenient to introduce the
following transformations from the spherical to the Cartesian system of coordinates
(see the relations (3.16c)):
x a0 cosf l; y a0 f f ; z r a0 (9.38a)
and for f 45 we have for the radius of the Earth, ao 6300 km.
The origin of the above right-handed curvilinear coordinates system (x, y, z)
(9.38a) lies on the Earths surface (for a flat ground, where r a0) at latitude f
and longitude l 0. Obviously, the sphericity parameter
d L =a0 ; (9.38b)
plays also an important role see the dimensionless equations (9.39a9.39d) below
and for d O(1) the equations for the atmospheric motions are more compli-
cated, as is obvious from these equations. However, in the case when d is assumed
1L a0 6300 km), we obviously have
e 1: (9.38c)
We assume, therefore, that the atmospheric motion occurs in a mid-latitude

region, distant from the equator, around some central latitude f , and sinf ,
cosf , and tanf are all of order unity.
After a careful dimensional analysis (see, for instance, [19], Chap. 2) a set of
dimensionless dominant (relative to e 1) atmospheric equations are derived,
and below our functions and timespace variables are dimensionless.
9.2.1 The Dissipative (Viscous and Non-adiabatic) NSF

atmospheric Equations
In the general case, when the Reynolds number Re is different to infinity, we derive
the following dissipative non-hydrostatic starting dominant dimensionless NSF
atmospheric equations:
rD fdvD =dt 1=Rosinf=sinf

dtanfuD =1 e dzk ^ vD g
1=gM2 1 e d zDpD
1=e2 Re@=@z m@vD =@z Oe; (9.39a)
rD fe2 dwD =dt e=Rocosf=cosf uD g

1=gM2 @pD =@z BorD (9.39b)
4=3Re@=@zm@wD =@z O e Oe d;
2
drD =dt rD f@wD =@z

1=1 e dzD vD dtanfvD
2 e dwD g 0; (9.39c)
rD dTD =dt g 1=gdpD =dt = 1=e2 RePrf@=@zk@TD =@z

Pr mg 1M2 1=1 e dzj@=@zvD =1 e dzj2
sdR=dzg O e2 ; (9.39d)
with pD rDTD.
In the above dimensionless equations (9.39a9.39d), for
UD vD ; wD ; rD ; pD ; TD with vD uD ; vD
we have, as dimensionless material time-derivative operator (with the same

notations for dimensionless time and space coordinates):
d=dt @=@ t 1=1 e dzvD :D wD @=@z (9.40a)
where, as horizontal gradient operator,
Dcosf =cosf@=@xi @=@yj: (9.40b)
such that
D:k 0: (9.40c)
Obviously, the dominant equations (9.39a9.39d), with (9.40a9.40c), are very

convenient when we assume that
d O1 but e 1 and Re
1; (9.41)
which is the case for the rather large synoptic dissipative atmospheric motions. For
equations (9.39a9.39d) it is also necessary to write boundary and initial
conditions.
Here, as boundary conditions, for the velocity (vD, wD) and temperature TD, we
write, on the flat ground,
on z = 0 : vD 0; wD 0; and k@TD =@z sR 0; (9.42a)
according to equation (9.39d) for TD.

Concerning the initial conditions, we observe that the initial (proper) data for the
four above evolution equations (9.39a9.39d) need not fit the dimensionless hydro-
static balance,
@pD =@z B0 rD 0; (9.42b)
because, in general, the dimensionless vertical velocity w need not be O(e) with
respect to the horizontal one, as is the case in the hydrostatic approximate
equations. Hence, in order to consider the most general case, we must assume
(with dimensionless quantities) that at the initial time t 0, ew is of order O(1),
and accordingly we obtain as initial conditions for the four evolution dominant
NSF equations (9.39a9.39d):
at t 0 : vD VD 0 ; ewD WD 0 ; TD TD 0 ; rD RD 0 ; (9.42c)
where data
VD 0 ; WD 0 ; TD 0 ; RD 0 ; PD 0 RD 0 TD 0 are given dissipative data:
Obviously, these data are dependent (in reality) on horizontal and vertical
coordinates in a prediction domain D with a diameter L . But it is not obvious that
these dimensionless coordinates, in data, are precisely the coordinates x, y, and z.?
This is an important problem (at least from my point of view) when we consider
the unsteady adjustment problem to hydrostatic balance in the set of a dissipative
hydrostatic, DH, approximate equations a problem which is actually still open.
Here we leave unspecified the behaviour conditions at high altitude when z " 1,
and far off in the horizontal directions. These behaviour conditions are strongly
linked to the numerical simulations, and can be changed in order to ensure the
stability of the numerical scheme used (see, for instance, the recent paper by Cullen
[172], pp. 202287).
Usually, it is necessary that:
Total energy density must decay

(9.42d)
sufficiently rapidly at infinity:
The initial-boundary value dominant, dissipative non-hydrostatic problem

(9.39a9.39d), with (9.40a9.40c) and (9.42a9.42d) seems well-posed, at least
from a fluid dynamics point of view (see, for instance the review paper by Gresho
[173], p. 52), and deserves further investigation for example, in the spirit of the
recent paper by White [172], pp. 1100.
The above formulation which is rather general for atmospheric motions
clearly shows the difficulty with these motions in the framework of a RAM
Approach. For this it is no wonder that weather prediction is a very complicated
task, mainly because it is very difficult to extract from the above general formula-
tion a model sufficiently and adequately simplified, and especially realistic, for a
simulation/forecast via a high-speed computer: What the weather will be like . . .
tomorrow evening or for the next few days?
9.2.2 Hydrostatic Limiting Processes
The above set of dimensionless dominant dissipative non-hydrostatic equations

(9.39a9.39d) are very complicated, and here we consider, for their simplification,
the main hydrostatic limiting case:
e ! o and Re ! 1; with e2 Re Re? 01; (9.43a)
assuming that d O(1), as in (9.41), which reinforces the hydrostatic constraint,

e 1. We assume that in this limiting process (9.43a) the parameters, Ro (or Ki),
M, Bo, Pr, and s are fixed and O(1).
The corresponding limiting equations below (9.45) and (9.46a9.46d) are
evidently very valuable for weather forecasting of synoptic processes (a la
Obukhov2), when (see (9.37h)):
L Lob ) such that M=Ki 01; (9.43b)
and we observe that the case of the low Mach number (M < < 1) for these synoptic
process is also not a bad idea! But with M < <1 we must also assume that Ki < < 1;
and low Kibel number approximation, according to (9.43b), is, in fact, a necessary
condition!
On the one hand, in the framework of singular perturbation problems, we see that
the hydrostatic limiting process (9.43a) is also strongly dependent on the considered
timespace region in weather prediction domain D, with the point P (x , y ) as
2
During several decades up to the early 1990s, A. M. Obukhov was the Director of the Institute of
Physics of the Atmosphere of the Academy of Sciences in Moscow actually named the Obukhov
Institute. He is a discoverer (with Kolmogorov) of the well-known (1962) KolmogorovObukhov
law 2/3 in turbulence theory.
origin, and at least is certainly singular near the initial time where the dissipative
data in (9.42c) are given. On the other hand, when Ki < < 1, in the framework of
quasi-geostrophic modelling, in the (Ekman) layer near the ground we have also a
singular behaviour.
Therefore, an asymptotic analysis shows that we can at least consider the
following two hydrostatic limiting processes linked with (9.43a) in the framework
of the non-hydrostatic dominant equations (9.39a9.39d):
LimDH 9:39a 9:39 d 9:43a; with t; x; y; z fixed; (9.44a)
and
LimDAdj 9:39a 9:39d 9:43a; with y t=e2 ; x, y,

(9.44b)
z z/e fixed]:
9.2.2.1 Dissipative Hydrostatic (DH) Large-Scale Equations
From the limiting process (9.44a) we derive the dissipative hydrostatic (DH) large-
scale, non-tangent equations for the functions
vDH uDH ; vDH ; wDH ; pDH ; rDH ; TDH

(9.45)
LimDH vD uD ; vD ; wD ; pD ; rD ; TD ;
dependent on the time-space coordinates, t, x, y, and z. Therefore, with (9.43a)

only, in place of (9.39a9.39d) we obtain the following DH model approximate
leading-order equations:
rDH fdvDH =dt 1=Kisinf=sinf d tanfuDH k ^ vDH g

1=gM2 DpDH 1=Re? @=@zm@vDH =@z; (9.46a)
@pDH =@z BorDH 0; (9.46b)
drDH =dt rDH f@wDH =@z D:vDH dtan fvDH g 0; (9.46c)
rDH dTDH =dt g 1=gdpDH =dt

1=Re? 1=Prf@=@zk@TDH =@z (9.46d)
2
Pr mg 1M j@vDH =@zj sdR=dzg;
2
with
pDH rDH TDH ; and d=dt @=@t vDH :D wDH @=@z: (9.47)
In (9.46a9.46d) the parameters d; M; Re? ; Pr; Ki; Bo; s; and sin f; and
cos f are fixed. These DH non-tangent model equations constitute a very signifi-
cant approximate system of equations for large-scale atmospheric motions d
O1 in a thin layer, such as the troposphere, around the Earths sphere.
On the flat ground, from (9.42a), we write:
vDH 0; wDH 0; and k@TDH =@z sR 0; on z 0 (9.48a)
and here again we leave unspecified the behaviour conditions at high altitude
whenz " 1, and far off, in the horizontal x and y directions.
Concerning the initial conditions for the evolution equations (9.46a, 9.46c,
9.46d), we must give, in fact, only the initial values for vDH and TDH (or pDH):
at t 0 : vDH v0 DH ; TDH T0 DH : (9.48b)
Here, the dissipative hydrostatic data vDH0, TDH0 have (strictly speaking) noth-
ing to do with the corresponding given proper dissipative non-hydrostatic initial
data VD0, TD0 in (9.42c), for the atmospheric dominant non-hydrostatic dissipative
equations (9.39a) and (9.39d).
We observe that from hydrostatic balance (9.46b), with
pDH rDH TDH ;
we also have

@logp0 DH =@z Bo=T0 DH and r0 DH p0 DH =T0 DH ; (9.48c)
and indeed, two of the initial conditions (related with the given proper dissipative
non-hydrostatic initial data WD0 and RD0) in (9.42c) have been lost during the dissi-
pative hydrostatic limiting process (9.44a) with (9.45), taking into account (9.43a).
As a consequence, a primary question arises: How are dissipative hydrostatic
data vDH0 and TDH0 related to given proper dissipative non-hydrostatic initial data
VD0, WD0, TD0 and RD0? The answer to this decisive question, for weather
forecasting, must be derived from the dissipative non-hydrostatic (where the param-
eter e is present) equations (9.39a, 9.39b), via (see (9.44b)) an unsteady adjustment
of the hydrostatic balance (9.46b) in the DH system of equations (9.46a9.46d).
The significant non-hydrostatic equations (where acoustics is present) which
govern this adjustment problem are derived below (see equations (9.50) and
(9.51a9.51d)) just through the limiting process (9.44b).
For a physical introduction to adjustment (adaptation) of meteorological
fields, see the discussion in }6 of Monin [94] a small but important book which
presents a unique account of the early evolution (up to 1968) of the dynamic and
physical bases for modelling and appropriate simulation of atmospheric motions
spanning the large spectrum of time-scales.
9.2.2.2 Adjustment to Hydrostatic Balance in the Framework of the DH

Equations
For the DH equations (9.46a9.46d) with the no-slip and temperature boundary
conditions (9.48a), singular near time t 0, it is necessary to consider a dissipative
unsteady adjustment problem of hydrostatic balance (9.46b), which is significant,
simultaneously, close to initial time and near the ground. This problem is derived
from the dissipative non-hydrostatic dominant equations (9.39a9.39d), via the
hydrostatic limiting process (9.44b), with
vDAdj ; wDAdj ; pDAdj ; rDAdj ; TDAdj

(9.49)
LimDAdj vD ; ewD ; pD ; rD ; TD ;
dependent on the timespace coordinates, y; x; y; andz:

In this case we derive the following dissipative (non-hydrostatic) adjustment
equations, which are significant close to initial time y 0 and near the ground
z 0:
rDAdj @vDAdj =@y wDAdj @vDAdj =@z 1=Re? @ 2 vDAdj =@z2 ; (9.50)
@rDAdj =@ y @rDAdj wDAdj =@z 0; (9.51a)
rDAdj @wDAdj =@ y wDAdj @wDAdj =@z 1=gM2 @pDAdj =@z

3=4Re? @ 2 wDAdj =@z2 ; (9.51b)
rDAdj @TDAdj =@ y wDAdj @TDAdj =@z g 1=g@pDAdj =@ y
wDAdj @pDAdj =@z 1=PrRe?@ 2 TDAdj =@z2
g 1M2 =Re? j@vDAdj =@zj2 4=3j@wDAdj =@zj2 ; (9.51c)
pDAdj rDAd TDAdj : (9.51d)
The above DAdj equations (9.50) and (9.51a9.51d) are derived when we
assume that the two dissipative coefficients m and k do not have a vertical structure
dependent onz! For these above DAdj (unsteady, a la Rayleigh) compressible,
viscous, and heat-conducting equations it is necessary to associate initial and
boundary conditions.
According to starting conditions (9.42c) at t 0, for (9.39a9.39d) as initial
conditions we write:
at y 0 : vDAdj VD 0 ; wDAdj WD 0 ; TDAdj TD 0 ; rDAdj RD 0 (9.52a)

At flat, thermally non-homogeneous ground, according to (9.42a), as boundary

conditions we have:
on z 0 : vDAdj 0; wDAdj 0; and @TDAdj =@z 0 (9.52b)
The question concerning the structure in altitude of the initial data in (9.52a) for
DAdj equations (9.50) and (9.51a9.51d), requires further and more detailed
investigation, because it is clear that in the dissipative case the unsteady adjustment
to hydrostatic balance raises many unsolved problems.
On the other hand, from our above RAM Approach we see that a numerical
simulation of atmospheric motions, via DH model equations (9.46a9.46d), must
be coupled with the above DAdj equations (9.50) and (9.51a9.51d), which
presents the possibility of taking into account the consistent initial conditions for
(9.46a) and (9.46d), linked with the unknown data in (9.48b).
During a numerical simulation of an atmospheric motion via an approximate
model derived asymptotically for more complete starting equations, it seems
inevitable that some partial derivatives relative to time in derived approximate
simplified models equation are lost (they disappear). In such a case, only via an
unsteady adjustment problem is it consistent, via a matching, to take into account
the influence of real data, written for the starting problem, on formulation of a well-
posed initial-boundary value problem for this derived simplified model!
9.2.3 The Dissipative Hydrostatic Equations in p-System
We now consider the dissipative hydrostatic (DH) system of equations

(9.46a9.46d), with the boundary conditions (9.48a) on z 0 and initial conditions
(9.48b, 9.48c) at t 0. For simplicity, however, we do not write the subscript DH
in these equations,
It is well known that a convenient hypoproduct of the hydrostatic approximation
is the possibility of using variables other than z as the vertical coordinate (for
example, pressure, potential temperature, and so on).
In particular, if pressure p is used as the vertical coordinate (according to
(9.46b), with Bo 1) such that
@=@z r@=@p and z Ht; x; y; p; (9.53)
where here (x, y) now denotes the horizontal coordinates on constant pressure
(isobaric) surfaces, and H(t, x, y, p) is the local height of an isobaric surface
above the flat ground surface, then, in place of DH equations (9.46a9.46d), without
any approximation nor ambiguity, we derive, for the horizontal velocity vector
v(u, v), temperature T; o dp=dt; and local height H, as a function of t, x, y, and

p, the following DH equations written in p-system:
dv=dt 1=Kisinf=sinf dtanf uk ^ v

1=gM2 DH 1=Re?@mr@v=@p=@p; (9.54a)
dT=dt g 1=gT=po 1=Re?1=Prf@rk@T=@p=@p

(9.54b)
Pr m=gg 1M2 rj@v=@pj2 sdR=dpg;
@o=@p D v d tanf v 0; (9.54c)
@H=@p T=p 0; (9.54d)
o p=T@H=@t v:DH w: (9.54e)
In these equations the operator D is the horizontal gradient on the isobaric

surface, p const, with the components cosf =cosf@=@x; @=@y; and the
material derivative operator is:
d=dt @=@t v:D o@=@p: (9.55)
In this case we assume that m; k, and R are known functions of p. The Eliasen p-
system has the advantage that the region of numerical integration has a limited
vertical extent, instead of 0 z < 1; for the original DH system
(9.46a9.46d).
One disadvantage is that the lower boundary conditions (9.48a) on flat ground
becomes a condition at the unknown isobaric surface H 0! Therefore, it is
necessary to impose, in place of boundary conditions (9.48a), the following bound-
ary conditions:
v 0; o p=T@H=@t; on H 0; (9.56a)
kp@LogT=@p sR; on H 0: (9.56b)
As a boundary condition at infinity in altitude we can assume that at the upper

end of the atmosphere, when
p 0; the total energy density decays sufficiently rapidly: (9.56c)
The DH set of equations (9.54a9.54e) in p-system, with the boundary

conditions (9.56a9.56c) and the initial conditions:
at t 0 : v v0 DH ; T T0 DH (9.56d)
can be used as a theoretical basis for the various investigations of features of

atmospheric dissipative hydrostatic motions depending on parameters Ki,
Re?; d; Pr and M2, and (from my point of view) it seems very interesting to take
into account, as a starting model, the DH model problem ((9.54a9.54e),
(9.56a9.56d)), written in p-system, for a discussion of the various approximations,
new mathematical developments, and their application to computer simulations.
For instance, in Norbury and Roulstones Large-Scale AtmosphericOcean
Dynamics [172], the first paper, by Cullen, and the second, by White, discuss
various interesting developments in the theoretical investigation of atmospheric
motions. Among other things, we observe that the smallness of the Mach number,
M, for the usual atmospheric motions, poses many unresolved difficult problems
the main reason being that from the so-called non-divergent (quasi-solenoidal)
approximation, in leading order, we obtain a very degenerate limit system of
approximate equations (see, for instance, Monin [94], }8, and in [12], the Chap. 12).
9.2.3.1 The b, l , and k Effects
In equation (9.54a) we assume that the sphericity parameter d 1; such that
sinf=sinf 1 d=tanf y Od2

tan f tan f 1 Od
since in dimensionless form, from relation for y in (9.38a), we have:
f f dy:
As a consequence we can write
1=Kisinf=sinf 1=Ki by; (9.57a)
with an error of O(d2), where (b-effect)
b d=Ki tanf : (9.57b)
In reality, (9.57b) is a similarity relation between the two small parameters Ki

and the small sphericity parameter d, when b O(1). On the other hand, when
tanf 1; for f 45 ; if we consider a low Kibel number limiting process.
A well-adapted form of the DH equations (9.39a9.39d), with b-effect, for the
low Kibel number asymptotic theory, is obtained when we assume that the follow-
ing similarity relation between the Kibel and Mach numbers, both assumed small, is
realized:
l 1=gKi=M2 (9.57c)
and we observe that l > 1 is in fact related with the Mach number, low com-
pressibility, effect, in the framework of low Kibel number flows. The relation
(9.57c) is related with (9.37h) and gives an estimation for the horizontal length
scale L .
On the other hand, the similarity relation
Ek? k Ki2 ; (9.57d)
with k O(1) as a measure of the viscous effect is motivated by the fact that
it presents the possibility of consistently deriving the so-called Ackerbloms
problem for the Ekman boundary-layer in the vicinity of the ground, p 1, in
the framework of low Kibel number flows.
9.2.3.2 Kibel Equations
In the Kibel equations (9.58a9.58d) below, the Kibel number is present but is
assumed small, such that b, l , and k are O(1) the Mach number M also being
assumed a small parameter but do not appear in (9.58a) and (9.58b) since (9.57c) is
taken into account. In these Kibel equations the term proportional to d << 1 are
neglected since b 0(1).
The dissipative coefficients are assumed constant (dimensionless m and k 1).
As consequence the starting Kibel equations, for v, T, H and o, are written in the
following form:
Kif@v=@t v:Dv o @v=@pg 1 bKi yk ^ v

l 1=KiDH k Ki2 @r @v=@p=@p; (9.58a)
Kif@T=@t v:D T o@T=@p g 1=gT=pg

1=Pr k Ki2 f@r @T=@p=@p (9.58b)
2
Pr1=l g 1Ki2 rj@v=@pj sdR=dpg;
@ o=@p D:v 0 (9.58c)
@H=@p T=p 0 (9.58d)
When Ki ! 0, the complete derivation of a consistent limit reduced model

is linked concurrently with a principal expansion and two local expansions.
A detailed derivation was performed by Guiraud and Zeytounian in [173] (see the
sketch below).
"p j I Quasi geostrophic region t; p

jt t=Ki II Local adjustment region t t=Ki;p
j II ) !I
initial condition
III Ekman BL region t;z p 1=Ki
jOKijcondition on the ground " IV Corner region t;z
j j *
j IV jOKi III z p 1=Ki
t0j j ! t
p1
For the above Kibel equations (9.58a9.58d) we have, as conditions:
v 0; o p=T@H=@t; on H 0: (9.59a)
p@Logp@H=@p=@p sR; on H 0: (9.59b)
at t 0 : v vo ; T To : (9.59d)
The above sketch presents a view of the three regions I, II, III, where the main
(QG), adjustment (Adj), and Ekman BL (Ek) expansions are considered. The corner
region IV plays a crucial role in the derivation of a second-order ageosprophic
asymptotic model.
Region II, via the unsteady adjustment process and matching, allows us to obtain
the consistent initial condition (at t 0) that must be applied to the QG single
unsteady model equation.
Region III, via the solution of the Ackerbloms problem in the steady Ekman
problem, produces (by matching) the boundary condition (at the ground p 1) that
must be supplied to the QG single non-viscous model equation.
9.2.4 The Leading-Order QG Model Problem
In considering the QG model problem it is first necessary to derive, from the above
Kibel equations (9.58a9.58d), the main (outer) QG leading-order model
equation valid in main region I, via the following limiting process:
Ki ! 0 with t; x; y; p; fixed;
(9.60a)
b; l and k being 01;
with the following (9.61a9.61d) asymptotic expansion for the functions v; H; o

and T, in the main region I:
v vQG Ki vAG ::: ; (9.61a)
H HS p Ki HQG Ki2 HAG ::: ; (9.61b)
o oQG Ki oAG ::: ; (9.61c)
T TS p Ki TQG Ki2 TAG ::: ; (9.61d)
where we have assumed that HS p and TS p are functions only of p! Although this
does not follow (concerning the dependence of time t) directly from the Kibel
equations (9.58a9.58d), it will be found to be consistent with the constancy
(relative to time t) of
d=dppdlogTS =dp sdR=dp; (9.62a)
which is a consequence of (9.58b). Of course, we have
Ts p pdHs=dp; (9.62b)
rS p p=Ts p; (9.62c)
but we do not yet know how TS p depends on p.

Then, from the boundary condition (9.59b), written for p 1, assuming that p 1
is the solution of Hp 0, it is found that (9.62b) allows us to compute Hsp, and
(9.62a) allows us to compute RTs p.
From our point of view we assume that Ts p is a given function and
g 1=gTS p dTS =dp KS p 6 0; (9.62d)
which represents data for the derived, in region I; QG model equation (see (9.65)
with (9.66) below).
In such a case, from (9.58b) we find at leading-order:
oQG 0: (9.63a)
Now, from (9.58a) we find at leading-order the well-known geostrophic balance:
k ^ vQG l0 DHQG 0 , vQG l0 k ^ DHQG ; (9.63b)
and the companion divergenceless equation for vQG ; from (9.58b)
D:vQG 0: (9.63c)
With (9.63a) from (9.58b) we derive for TQG the following equation:
@TQG =@t vQG :DTQG KS p=p oAG 0: (9.63d)
Going to higher order, we derive from the Kibel equations (9.58a, 9.58c, 9.58d)
the following three equations:
vAG k ^ @vQG =@t vQG :DvQG loDHAG b y vQG ; (9.64a)
@ oAG =@p D:vAG 0; (9.64b)
TQG p@HQG =@p; (9.64c)
From the continuity equation (9.64b), with (9.64a), and the expression for oAG
obtained through the elimination of TQG and vQG, from (9.63d), using (9.63b) as a
function of HQG, we obtain the following single quasi-geostrophic potential vortic-
ity model equation:
@LHQG =@t l0 @HQG =@x@LHQG =@y @HQG =@y@LHQG =@x
b@HQG =@x 0; (9.65)
where
LHQG l0 D2 HQG @=@pfp2 =KS p@HQG =@pg: (9.66)
We observe that the QG model equation (9.65) contains one derivation with
respect to time t, and as a consequence only one initial condition must be supplied
for HQG , via an unsteady adjustment problem, which we shall derive in the next
section 9.2.4.1. The boundary condition that must be supplied on the ground p 1,
for QG equation (9.65), will be derived below in section 9.2.4.2.
Finally, concerning the boundary conditions that must be applied at the upper
end of the atmosphere, p 0, and far off in the horizontal plane, we can again
assume that the total energy density
p2 =KS p@HQG =@p2 l0 jD2 HQG j2 (9.67)
decays sufficiently rapidly at infinity.
9.2.4.1 Adjustment to Geostrophic Balance (9.63b)
It is not difficult to verify, by trial, that the inner in time unsteady adjustment
equations to geostrophic balance (9.63b) are derived by setting
t t=Ki; (9.68a)
and applying an initial limiting process
Ki # 0; with t; x; y; and p fixed: (9.68b)
Concerning, more precisely, the introduction of a short adjustment time (like t)

in the case of the various unsteady adjustment problems and matching of meteoro-
logical fields, I would here like to exemplify my difficulty, as a young mathemati-
cian, in realizing the subtlety of arguments related to the adjustment to geostrophic
balance.
I very well remember when a 1957 graduate (PhD) student in Kibels dynamic
meteorology department at the (now) Obukhov Institute of Atmospheric Physics, in
Moscow reading Kibels newly published An Introduction to the Hydrodynamical
Methods of Short Period Weather Forecasting [171], Chap. 4, how I did not in any
way seem to understand in what manner appear, simultaneously, this adjustment
short time t ( t/Ki), in the case of the adjustment to geostrophic balance in the
framework of the low Kibel number asymptotics, and of an evolution prediction
time t both times being denoted, in Kibels book, by the same symbol, t.
Only after September 1967, while working in the Aerodynamics Department at
ONERA, did I have the possibility due to Van Dykes Perturbation Methods in
Fluid Mechanics [14] of really understanding the profound significance of these
two (inner and outer) times in the asymptotic, outerinner expansions with
matching ( t ! 1 , t ! 0) machinery.
Let us set ft for any quantity f considered as a function of t t=Ki instead of t.
First, we rewrite the Kibel equations (9.58a9.58d), with, in place of
@=@t ) 1=Ki@=@ t;
for the new functions, vt ; ot ; H t ; Tt , and then expanding
vt ; ot ; H t ; Tt vt 0 ; ot 0 ; H t 0 ; Tt 0 Ki vt 1 ; ot 1 ; H t 1 ; Tt 1
Ki2 vt 2 ; ot 2 ; Ht 2 ; Tt 2 ::: ; (9.69a)
one first finds
H t 0 ; Tt 0 HS p; TS p;
with
T S p p@HS =@p: (9.69b)

In order to find equations for vt 0 ; ot 0 ; Ht 1 ; Tt 1 we have to go to higher order:
@vt 0 =@t k ^ vt 0 lo DH t 1 0; (9.70a)
@ot 0 =@p D:vt 0 0; (9.70b)
Tt 1 p@H t 1 =@p; (9.70c)
@Tt 1 =@t KS p=p ot 0 0; (9.70d)
where KS p is given by (9.62d) as a function of p alone.

We observe that KS(p) has a suggestive interpretation-namely, if we introduce a
dimensionless specific entropy S, then with
S St 0 p OKi : dSt 0 p=dp 1=pTS pKS p: (9.71)
The above system of equations (9.70a9.70d) is the system governing the

unsteady process of adjustment to geostrophic balance (9.63b).
We observe, first, that the derivation of the unsteady adjustment local equations
significant close to initial time, in the case of the low Kibel number asymptotics, is
not a consequence of the linearization of the Kibel equations (9.58a9.58d) (as is
written in [171], p. 83).
These inner, local-in-time, unsteady adjustment (linear) equations to geostrophic
balance are rationally derived, in the general case, from full unsteady Kibel
equations (9.58a9.58d), as a significant limit when Ki ! 0, with short time t
fixed in place of time t fixed, the system of equations (9.70a9.70d), being, in fact, a
significant degeneracy of Kibel equations (9.58a9.58d) near the initial time.
From the last two equations of (9.70a and 9.70d) we derive the following
relation:

ot 0 p2 =KS p @@H t 1 =@p=@t; (9.72a)
and going back to the first two of equations of the system (9.70a and 9.70b) we find
a couple of equations for vt 0 and Ht 1 : namely,
@vt 0 =@t k ^ vt 0 lo DH t 1 0; (9.72b)
D:vt 0 @=@pfp2 =KS p@ 2 H t 1 =@t@pg 0: (9.72c)
But for the two evolution (in time t) equations (9.72b) and (9.72c), for vt 0 and
t
H 1 , it is necessary to give an initial condition for
vt 0 and for H t 1 at t 0;
according to (9.59d). Concerning vt 0 we may use the initial data, v0 , assuming:
t 0 : vt 0 v0 : (9.73a)
Concerning the initial data for H t 1 , it is necessary to assume that the data T0 (in
(9.59d)) may be set in the form
T0 TS p Ki T0 1 :::: ) t 0 : H t 1 H 0 1 ; (9.73b)
when we use the relation between T0 and H 0 : T0 p@H 0 =@p. Whenever the
data T0 p@H 0 =@p, in (9.59d), cannot be put into the above form, we must
expect that another adjustment process holds!
There is an important observation, which was known to Kibel in 1955 (see
[171]), and which concerns the way in which
limt"/ Ht 1 Ht 1 /
is related to the initial values (9.73a, 9.73b). For this, we start from equations
(9.72b), (9.72c), and first deduce the equation:
@
@=@tfk:D ^ vt 0 fp2 =KS p@H t 1 =@pgg 0: (9.74a)
@p
If we now integrate this last equation between t 0 and t / , and if we use

the geostrophic balance,
vt 0 / lo k ^ DHt 1 / ;
for limiting values of vt 0 and H t 1 , when t "/ , we obtain for H t 1 / the following
relation:
@
lo DHt 1 / fp2 =KS p@Ht 1 / =@pg
@p
(9.74b)
@
k:D ^ v0 fp2 =KS p@H 0 1 =@pg:
@p
Among other things, Kibel was able (also in 1955) to settle the main issue of the
unsteady adjustment (towards geostrophic balance (9.63b)): vt 0 and H t 1 related
with (9.72b) evolves towards the geostrophic balance (9.63b), when the short
adjustment time t tends to infinity.
As a matter of fact one has a matching relation:
limvt 0 ; H t 1 vQG t 0; x; y ; p; HQG t 0; x; y ; p: (9.74c)

t"1
Finally, we obtain from (9.74b) the initial condition that must be applied to the
main QG equation (9.65) with (9.66), in the following form:
at t 0 : LHQG k:fD ^ v0 0 g
(9.75)
@=@pfp2 =K0 p@H 0 1 =@pg;
where LHQG , in (9.75), is the same operator which appears in the QG main outer
model equation (9.65), and is given by (9.66).
Here we can observe that if it is true that to predict the field HQG, in the QG
approximation, it is sufficient to know only the initial value of HQG! But unfortu-
nately, this single initial value is related, according to (9.75), to the initial data in
(9.73a, 9.73b), which are obtained from the unknown(!) data (9.59d) prescribed for
the DH two model equations (9.54a, 9.54b), under the constraint previously men-
tioned in (9.73b).
As a consequence, the main unsteady adjustment problem, concerning the
derivation of consistent initial conditions for the quasi-geostrophic, primitive
Kibel inviscid (non-viscous and adiabatic) or dissipative hydrostatic (viscous and
non-adiabatic) model equations, is the unsteady adjustment to hydrostatic balance.
In particular, for the Kibel primitive inviscid equations (DH equations
(9.54a9.54d), when 1=Re? 0), the corresponding unsteady acoustic adjustment
problem has been considered by Guiraud and Zeytounian [106], in which it is
observed that:
The initial conditions concerning horizontal velocity and entropy for the unsteady primitive
Kibel non-viscous and adiabatic equations are simply the ones pertinent to the full Euler,
non-hydrostatic equations, but shifted vertically by an amount equal to D. (Concerning the
above-mentioned vertical shift D, see our [19], Chap. V, Sect. 19.)
We also have the possibility of deriving a new initial condition (at t 0) relative
to p 1:
HQG pTs p=KS p@HQG =@pp 1 t0

(9.76)
H 0 1 pTs p=KS p@H 0 1 =@pp 1 :
The role of this curious condition (9.76) is to serve as an initial condition for
the boundary condition on p 1 (see the derived ground condition (9.87) below)
that must be applied to the main QG, outer, equation (9.65) written for HQG.
This condition (9.87) is derived via the formulation of Ackerbloms problem in
the steady Ekman layer (region III) near the ground, and contains a time derivative.
As a consequence, (9.87) may be considered as a boundary condition for our main
QG, outer, equation (9.65), only if it is complemented with an initial condition
(9.76).
9.2.4.2 Ackerbloms Problem in the Steady Ekman BL Layer
Indeed, for the above derived main outer QG single equation (9.65) for HQG , with
the initial condition (9.75), at t 0, and behaviour (decay sufficiently rapidly)
condition (9.67) at p 0 and at infinity in the horizontal plane, it is necessary also

to write a boundary condition at the ground simulated (for low Kibel number) by
p 1.
For this within the region III (the steady Ekman layer) we introduce the new
vertical coordinate:
p p 1=Ki; (9.77a)
and in region III we consider the following inner (BL) limiting process:
Ki ! 0 with t; x; y and p fixed: (9.77b)
With (9.77a, 9.77b) we consider the following inner (BL) expansion in region
III:
v; o; H; T; r v0 ; o0 ; H0 ; T0 ; r0
Ki v1 ; o1 ; H1 ; T1 ; r1 ; (9.77c)
We then obtain from the starting Kibel equations (9.58a9.58d) the following set
of equations, for v0 ; o0 ; H0 ; T0 ; r0 ; o1 ; H1 ; T1 :
DH0 0; @ o0 =@p 0; @H0 =@p 0; (9.78a)
r0 T0 1; 1=Prko @=@pr 0 @T0 =@p 0; (9.78b)
with
at H0 0 : @T0 =@p 0: (9.78c)
Then, from matching with main QG region, when p !/ , we obtain
o0 0; limp !/ T0 T0 1; limp !/ H0 H0 1; (9.79a)
and we also have
k ^ v0 lo DH1 ko @=@p r0 1@v0 =@p 0; (9.79b)
@o1 =@p D:v0 ; (9.79c)
1=Prko @=@p r0 1@T1 =@p 0; (9.79d)
T0 1 @H1 =@p; (9.79e)

with
d v0 0;

on H1 0 :o1 1=T0 1@H1 =@t; (9.80)
b 1=T0 1@T1 =@p
sR 1;
when we assume that the main radiative transfer does not have an Ekman BL
structure.
In fact, we can assume that the flat ground in the Ekman BL layer is
characterized by:
p Pg0 Ki Pg1 ::: : (9.81)
From the above relation (9.79d) and (9.80), after a matching with the main QG
region I, we first obtain
T1 TQG;1 dT0 =dpp 1 p ; (9.82a)
where
TQG;1 TQG t; x; y; p 1 @HQG =@pp 1 :
Then, from equation (9.79d) we obtain
H1 HQG;1 T0 1p ; (9.82b)
and
H1 0 : imply : Pg0 1=T0 1HQG;1 ; (9.82c)
where HQG;1 HQG t; x; y; p 1.

Now, with above results, we can formulate a consistent Ackerbloms problem.
First we consider equation (9.79b), for v0*, and we set:
v0 vQG;1 V0 ; with vQG;1 vQG t; x; y; p 1:
From the matching with the main QG region I, when p !/ , we have:
limp !/ V0 0 and k ^ vQG;1 lo DHQG;1 0;
and we obtain, from (9.82b) for H1*, that in (9.79b):
k ^ v0 lo DH1 k ^ V0 :
Therefore, with the above results, we can formulate the following problem for
the determination of the horizontal, perturbation, velocity vector V0*:
ko @ 2 V0 =@p2 k ^ V0 0; e
V0 vQG;1 ; on p 1=T0 1HQG;1 ; j (9.83)
V0 ! 0; when p !/ :c
where we have assumed r0 1 1, because we work with dimensionless quantities.

The solution of the above Ackerbloms problem (9.83) is obtained in a standard
way:
V0 ik ^ V0 vQG;1 ik ^ vQG;1 E ; (9.84a)
where i 11=2 and
E expf1 i=2ko 1=2 p 1=T0 1HQG;1 g: (9.84b)
Now it is necessary to consider the continuity equation (9.79c), and we find for
o1* the following relation:
p
o1 D v0 dp 1=T0 1@H1 =@tp Pg (9.85a)
Pg 0
0
where
v0 RealfvQG;1 ik ^ vQG;1 1 E g: (9.85b)
A tedious computation of integral in (9.85a) with (9.85b) finally produces the

value of o1* when p !/ :
p
o1 / limp!/ o1 1=T0 1@HQG;1 =@t lo ko D2 HQG ;1 : (9.86)
Now, matching
limp!/ o1 o1 / oQG;1 ;
with the main QG region I produces the following boundary condition at the flat
ground for the QG model equation (9.65) with (9.66):
f1=T0 1@=@t 1=KS 1@=@t vQG D@=@p

p (9.87)
lo ko D2 gHQG 0; on p 1:
The above QG model problem ((9.65) with (9.66), and (9.75) and also (9.87)
with (9.76)) derived from full Kibel equations with initial and boundary conditions,
for low Kibel number, is a very appropriate application of our RAM Approach, and
provides a good example of the deconstruction of a starting problem and its
reconstruction via a main, QG, model and two local, Adj and Ek, models, by
matching for a Ki small parameter!
9.2.5 The Second-Order Ageostrophic GZ Model
For the derivation of the so-called AG model, we need to consider three local
(inner) expansions in addition to the main one. Two of them are higher
approximations of those considered previously in the framework of the QG
model derived above in section 9.2.4. From the outer approximation (9.61ab) we
first derive the main AG, second-order model equation, for HAG.
Then, from the first local (close to initial time) approximation, via a second-
order unsteady adjustment problem to the AG model equation, we obtain by
matching (between regions II and I) the initial condition, at t 0, for the AG
main equation.
Afterwards, from the second local (near-ground) approximation, via a second-
order Ackerbloms problem in steady Ekman layer, we have the possibility of
matching (between regions III and I) to derive a boundary condition, on the ground
p, for this AG main equation. This ageostrophic model is relative to the HAG
component in the main asymptotic expansion (9.61b).
In order to obtain the initial condition at t 0, for the AG main model second-
order model, the problem of adjustment to ageostrophy, in region II, must be
considered. However, so as to be able to correctly formulate this problem of
adjustment to ageostrophy, it is necessary to analyze the problem related to the
unsteady Ekman boundary layer which develops in corner region IV! In addition,
we must also elucidate the compatibility of the models between regions III and IV
by analyzing the behaviour of the unsteady Ekman boundary layer when t !/ .
It is also necessary to consider the problem of the second approximation of the
steady Ekman boundary layer, which is a necessary step for the derivation of the
boundary condition in p 1 associated with the AG main model equation
(Detailed results for this AG second-order model are included in our monograph
[12], pp. 236262).
It seems appropriate to add a final comment to this section, because it very well
shows the importance of our Approach, via the RAM, for the derivation of accurate,
reliable and consistent models. This comment concerns the so-called balance
equation which is well known in meteorological literature and often analyzed in
various ad hoc theories. This balance equation was derived by Monin in 1958 (see
Monin [94], pp. 4244 for various comments), and also by Charney in 1962 [174].
In Monin this balance equation was derived as a second-order correction with
respect to the Kibel number, Ki, within the framework of the hydrostatic
approximation. Therefore, within the framework of the present (above) low Kibel
number asymptotics analysis, we easily derive, for the second-order AG model, the
following equation:
k r ^ vAG l0 D2 HAG 2l02 f@ 2 HQG =@x@y2

(9.88)
@ 2 HQG =@x2 @ 2 HQG =@y2 g;
which may be interpreted in different ways.

Here, in our RAM Approach, equation (9.88) is obviously an explicit relation for
the vertical component of vorticity once HQG and HAG have been computed! But in
Monin and Charney, confusingly, HQG and HAG with an mysterious(!) unknown H,
(9.88) is viewed as a MongeAmpere equation for computing this unknown H,
when the horizontal velocity field (vAG) is assumed known! This curious (and
erroneous) interpretation of (9.88) producing, in reality, the possibility for
computing vAG, when HQG and HAG have been computed is obviously a direct
consequence of the absence of a rigorous logical, step-by-step, RAM Approach for
the derivation of sequential approximations. These erroneous results are present
systematically when an ad hoc approach is used for the derivation of the second-
order model, because (see the Epilogue below) it is not possible to predict the many
well balanced terms in these second-order models except via a reliable method the
RAM Approach! (The Epilogue includes, among many examples, a typical exam-
ple when the lack of such a RAM Approach often leads to very ambiguous results.
Concerning the MongeAmpere equation, see Kibel [171], in which this equation
is considered and analyzed).
9.2.6 Kibel Primitive Equations and Lee-Waves Problems as

Inner and Outer Asymptotic Models
We mentioned in Sect. 9.2.2 that in the framework of the non-hydrostatic dominant

equations (9.39a9.39d) we have the possibility of considering two limiting pro-
cesses: (9.44a) and (9.44b). In reality, when we assume that a local relief is situated
around the point P (x , y ) in the prediction domain D with a diameter L , a third
local limiting process provides the possibility of considering the classical steady
lee-waves model problem as an inner (local) model, relative to Kibel primitive
(hydrostatic and non-dissipative) equations considered as an outer model problem.
First, we consider, in place of equations (9.39a9.39d), a simplified non-dissi-
pative Re !/ without Coriolis force Ro / system of equations, which leads
to the (classical, d ! 0) dimensionless Euler atmospheric equations
rE dvE =dt 1=gM2 DpE 0; (9.89a)

drE =dt rE f@wE =@z D:vE g 0; (9.89b)
rE dTE =dt g 1=gdpE =dt 0; (9.89c)
rE fe2 dwE =dt 1=gM2 @pE =@z Bo=gM2 rE 0: (9.89d)
for vE, wE, pE rE TE , TE and rE.

From these equations, when we write the slip condition
on z 0 : wE 0; (9.90a)
in the case of a flat ground, we derive, in the hydrostatic limiting process
limP e!0 fe ! 0 with t; x; y z fixed and M O1g; (9.90b)
the well-known Kibel primitive equations,
rP dvP =dt 1=gM2 DpP 0; (9.91a)
drP =dt rP f@wP =@z D:vP g 0; (9.91b)
rP dTP =dt g 1=gdpP =dt 0; (9.91c)
@pP =@z BorP 0; (9.91d)
pP TP rP ; (9.91e)
for the limit functions,
vP ; wP ; pPr ; TP ;rP limP e!0 vE ; wE ; pE ; TE and rE ; (9.92)
when (9.90b) is realized.

However, for these same Euler equations (9.89a9.89d) we can also consider the
presence of a local relief situated around the point PO x0O ; y0O on the ground in a
local region D0 2 l0 ; m0 , inside the prediction domain D (having L as diameter).
If the local vertical length scale is h0, then the relief can be simulated by the local
dimensionless equation,
z0 h0 hx0 x0O =l0 ; y0 y0O =m0 ; with h O1; (9.93)
where z0 and x0 , y0 are physical local coordinates (with dimensions) linked with the
relief.
The above dimensionless Euler equations (9.89a9.89d), however, are written
with dimensionless coordinates: z z0 =H ; x x0 =L and y y0 =L . Rewriting
(9.93), with the dimensionless coordinates, x, y and z, we obtain, in its place, the
dimensionless equation of the relief in the following form (in place of: z 0):
z ahb x; c Z; (9.94a)
with
x x xO =e and Z y yO =e: (9.94b)
In (9.94a)
a h0 =H ; b L =l0 ; c L =m0 : (9.94c)
Corresponding the slip condition on the surface of the relief (9.94a) is:
wloc avloc :Dloc hbx; c Z on z ahb x; c Z; (9.95a)
where Dloc @=@x; @=@Z.

Obviously, from the limiting hydrostatic process (9.90b), where x and y are
fixed, which leads to the Kibel primitive equations (9.91a9.91e), we cannot take
into account, in these primitive equations, the influence of the relief linked with the
slip condition (9.95a). The reason is that
h/; / 0; (9.95b)
as a consequence of a local character of the considered relief, which is limited in D0,

inside D. As a consequence, in place of (9.90b) it is necessary to consider a new
local hydrostatic limiting process: namely,
limloc e!0 fe ! 0 with t; x; Z and z fixedg; (9.96a)
with, a, b, c, and M O(1). In such a case, for the below local limit functions, when
(9.96a, 9.96b) is realized,
limloc e!0 vE ; ewE ; pE ; TE and rE ;

(9.96b)
vloc ; wloc ; ploc ; Tloc ; rloc ;
we derive the following local steady non-hydrostatic Euler equations:
rloc vloc :Dloc vloc wloc @vloc =@z

(9.97a)
1=gM2 Dploc 0;
Dloc :rloc vloc @rloc wloc =@z 0; (9.97b)

rloc vloc :Dloc Tloc wloc @Tloc =@z

g 1=gvloc :Dloc ploc (9.97c)
wloc @ploc =@z 0;
rloc vloc :Dloc wloc wloc @wloc =@z

(9.97d)
1=gM2 @ploc =@z Borloc 0;
with
ploc rloc Tloc : (9.97e)
These local equations (9.97a9.97e) govern the steady lee-waves problem in a

non-viscous baroclinic adiabatic atmosphere over and downstream of a relief
simulated by equation (9.94a) in a local domain D0, when we use the slip condition
(9.95a) for wloc. Formal matching of
limx; Z"/ vloc ; wloc ; rloc ; ploc ; Tloc vP ; 0; rP ; pP ; TP P0 ; (9.98a)
@pP =@z BorP P0 0; (9.98b)
between the local lee-waves (9.97a9.97e) and Kibel primitive (9.91a9.91e)

model equations may be interpreted as providing lateral (in the horizontal plane)
boundary conditions at infinity for the local (inner) steady, non-hydrostatic,
dynamic lee-waves model equations (9.97a9.97e) with (9.95a), once the predic-
tion by primitive (hydrostatic) Kibel model equations (9.91a9.91e) is known at the
position P0 in D.
The set of local steady lee-waves equations (9.97a9.97e) are intensively used in
mesometeorology (as is the case in our Etudes Hydrodynamique des Phenome`nes
Mesometeorologiques [1]).
Epilogue
I hope that the reader who has become acquainted with the various parts of this
monograph now has not only une certaine idee of my main purpose, but is also
deeply convinced that there is really no better way for the derivation of significant
consistent model problems than the RAM Approach!
It might be true that the physical approach can produce valuable qualitative
analysis and a better understanding of the laws of nature related to various signifi-
cant and practical fluid-flow phenomena. Unfortunately, a purely ad hoc physical
approach would not be able to point the right way to a consistent rational obtention
of approximate simplified leading-order model problems which could be used
successfully for numerical simulation with a high-speed computer.
In particular, this is especially true because such an ad hoc approach would be
unable to provide a rational, logical method (as is the case with our RAM
Approach) for the derivation of a well-balanced associated second-order model
problem with various complementary physical effects! A well-known example is
the derivation of a second-order rational system of boundary layer compressible
and heat-conducting equations (which, it seems to me, are primarily due to
asymptotics) mentioned in the Prologue above. This extension of classical Prandtl
boundary-layer equations, by Van Dyke (1962), from the full NSF problem, has
been a crucial practical scientific contribution to the NASA program for the
atmospheric re-entry of the Space Shuttle!
Below, a typical case is considered, which is related to the modelling of the
Benard thermal convection problem (discussed in Chap. 8), up to the derivation of a
second-order model associated with the classical RB shallow thermal convection
model.
In general, the derivation of a second-order model problem, from the full
unsteady NSF, well-formulated problem is not an easy task! Usually, erroneous
results are present, systematically, when an ad hoc approach is used for the
derivation of such second-order models, because it is not possible to predict the
many, well-balanced terms in these second-order models except via a reliable
method the RAM Approach!

DOI 10.1007/978-3-642-20746-4, # Springer-Verlag Berlin Heidelberg 2012
254 Epilogue
Below, with the derivation of the second-order equation model for the Benard
convection problem, associated with the well-known RB leading-order shallow
thermal convection model, we give a typical example when the absence of such a
logical, step-by-step RAM Approach leads to very ambiguous results!
As starting, exact, Benard convection equations, we choose those derived by Hill
and Roberts in 1991. These dimensional equations, with physical quantities, are
written in the following form (see Chap. 8 for the notations):
@ui =@xi a dT=dt;
rdui =dt rfi @p=@xi @=@xj ldmm dij 2 mdij ;
adp=dt Cp r=a @ui =@xi ldii 2
2mdij dij @=@xi k@T=@xi ;
and we assume that f1 0; f2 0; f3 g, and the coefficients, l; m , and k are

assumed constant (respectively, ld ; md ; kd , as functions of constant temperature
Td ), and we write in place of @=@xj ldmm dij 2mdij the term
md fDui 1 ld =m d r@ui =@xi ; Cp T@S=@Tp , which is related to entropy
S, the specific heat at constant pressure Cp Cp T. For the derivation of model
reduced equations, from the above (non-dimensionless) convection equations, first,
Hill and Roberts consider as a limiting process (the rather exotic):
g !/ and ad ! 0 such that gad fixed and O1
which is, in fact, a bastardized, non-formalized version of our a la Boussinesq

limiting process (8.16). It is interesting to remark that Straughan (1992), p.279,
writes: The key philosophy of the Hills and Roberts paper (1991) is that typical
acceleration promoted in the fluid by variations in the density are always much less
than the acceleration of gravity!
Then, according to the above limiting process, Hills and Roberts expand the
pressure, velocity, and temperature fields, in their above dimensional equations, in
1=g ! 0, such that:
p p0 g p1 1=gp2 :::;
ui ui 1 1=gui 2 :::;
T Td T1 Td 1=gT2 Td :::
rT rd 1 ad T Td :::::
which is a rather bizarre expansion!

Epilogue 255
Now, with the above expansion, the OB equations (for a weakly expansible
liquid) are derived, via the limit:
eH=R d=Cpd gad ! 0
Thus, Hill and Roberts obtain the following reduced equations (with
dimensions!):
@ui =@xi 0;
dui 1 =dt Ra T1 di3 @p1 =@xi Dui 1 ;
dT1 =dt eH=R Td T1 u3 1 1=PrDT1 2=RaeH=R dij 1 dij 1 :
These equations are, in fact, rather similar to our deep thermal convection
equations (8.43bd) with viscous dissipation the parameter eH=R being an ana-
logue to our dissipation parameter Di, defined by (8.44a). When
eH=R ! 0;
we recover the RB model, leading-order, equations.

In our deep thermal convection equations (8.43bd), in place of pressure and
temperature, we have, respectively, the perturbations p (8.17) and y (8.13) and
Td T1 in the term proportional to eH=R u3 1 does not have an explicit form!
Straughan (1992), p. 274, presents the following form of our deep convection
equations (a system appropriate to thermal convection in a deep layer):
1=Pr@ui =@t uj @ui =@xj @ p=@xi Dui Rydi3 ;
@uk =@xk 0;
@y=@t uj @q=@xj Ru3 y mx3 Dy 2d=Rmx3 dij dij ;
where
mx3 1=1 d1 x3
p
and d is a constant which represents a depth parameter, and R Ra.
If we now consider for three functions u ui ; y and p, the following three
asymptotic expansions, relative to expansibility parameter e:
du uRB euS :::;
U u; y; p jy yRB eyS :::;
bp pRB epS :::;

256 Epilogue
then, first derived, from the full unsteady NSF convection equations, the RB,
shallow thermal, leading-order, convection equations are:
r:uRB 0;
@uRB =@t uRB :ruRB rpRB GryRB k r2 uRB ;
@ yRB =@t uRB :ryRB 1=Prr2 yRB ;
In a second step, again from the full unsteady NSF convection equations, we
derive as a second-order, linear, well-balanced system of non-homogeneous
equations for uS ; yS ; pS :
r:uS dyRB =dt;
@uS =@t uRB :ruS uS :ruRB rpS GryS k
r2 uS yRB duRB =dt 1 ld =md rdyRB =dt;
@yS =@t uRB :ryS uS :ryRB 1=Prr2 yRB
1 Gpd yRB dyRB =dt BoTd =DT yRB uRB :k
1=2Bo=Gr@uRB i =@xj @uRB j =@xi 2 ;
where Gpd 1=Cp dCp =dTd =ad , when we assume:
Cp T Cpd 1 eGpd y and aT ad 1 eAd yd ;
with
Ad dlogCp T=dT=dlogrT=dTd :
Obviously, from the above (but not obvious) second-order equations it is possi-
ble to undertake an analysis which might produce interesting results complemen-
tary to the usual (classical, a la Chandresakhar, 1961) known quantitative results for
the RB problem. Such analysis can, in particular, can produce specific complimen-
tary results obtained earlier from these RB model equations.
We see that even if an ad hoc derivation is sometimes able to produce a valuable
result at the leading-order, and even if a deficient ambiguous approach is chosen
(as in the case of Hill and Roberts), such an approach will in no way be able to
consistently derive a rational second-order approximation with well-balanced
second-order terms. This precise observation is one of the main reasons for our
present RAM Approach and for the publication of this book!
Epilogue 257
Actually, we have a certain idea of the intrinsic structure of NSF equations and,
via the presence of various non-dimensional reduced parameters in these equations
and associated initial and boundary conditions, we begin to understand thanks to
the RAM Approach the profound unity in the puzzle of partial fluid-flows with
diverse configurations. Here in Chaps. 79 we have considered only some problems
of aereodynamics, convection, and atmospheric motions; but our RAM Approach
have a large spectrum of applications in various physical, technological, and
geophysical processes. It can operate everywhere where it is possible, in analyzed
problems, to detect several dimensionless significant parameters but this is not
always an easy matter. When this is the case, then the corresponding mathematical
consistent asymptotic model appears under definite scaling relations (similarity
rules) between these parameters, which produce diverse limitations in the use of
this model! In Chap. 8 such a case was analyzed, and for the thickness, d, of a liquid
layer a strong double limitation was derived for the case of RB shallow thermal
convection (see Sect. 8.4).
This idea of obtaining mathematical models by proper scaling parameters in
multiparametric non-linear problems is presented and very well illustrated in detail
by Cercignani (kinetic theory) and Sattinger (weakly non-linear dispersive phe-
nomena) (1998). A more recent, very interesting, but rather unusual book by de
Gennes et al. (2004) will enable the reader to understand, in simple terms, some
mundane questions affecting our daily lives questions that have often come to the
fore during our many interactions with industry (capillarity and wetting phenom-
ena, drops, bubbles, pearls, waves, and so on). The strategy in this book is to
sacrifice scientific rigour by an impressionistic approach (a la de Gennes) based
primarily on qualitative arguments, which makes it possible to grasp things more
clearly and to envisage novel situations. For me, this book is a collection of
problems well suited to a RAM Approach if, under the physics, we are able to
discover the main significant small (or large) parameters.
I hope that the points of view, linked with the RAM Approach emphasized
throughout this book devoted to the modelling of NSF equations, have been well
presented, and that some readers (at least) will be convinced that this technique
provides very powerful tools for the derivation of consistent rational not-stiff
models as an aid to numerical simulation. It is clear that at the present time a gap
still exists between the Rational Asymptotic Modelling (RAM) Approach and
High-Speed-Computer Numerical (H-S-CN) Simulation. In many cases the results
of numerical computations (which are often fascinating) do not correspond satis-
factorily with experimental/laboratory visualizations often because of the absence
of criteria by which the limit of validity of used models is known. Rarely in
scientific publications do we encounter the theoretical treatment of a full
unsteady fluid-flow problem directly inspired by the technology, and the work of
producing a simple description of this fluid flow, via a model problem that can be
used to explain it by using numerical simulation, is lacking (see, for instance, Sect.
6.1, footnote 2, page 119). The crucial problem of initial conditions, which strongly
influence the subsequent formation and evolution of the fluid flow, is usually
overlooked!
258 Epilogue
We are convinced that: The more computing and numerical algorithmic pro-
cesses, via high-speed computers, becomes efficient, the more will be the need for
conceptually consistent techniques capable of unravelling stiff fluid-flow problems.
The Rational Asymptotic Modelling (RAM) Approach, among others(?), proves to
be a (more!) efficient tool. This (rather optimistic) statement was, during 197080,
a guideline for me when I was working on the asymptotic modelling of atmospheric
flows!
Modelling of atmospheric flows especially via approximate and consistent
well-balanced models for global, short-range, and local weather-forecasting
involves vast and complex applied mathematics as is seen clearly in Sect. 9.2
above, in the framework of the very particular QG asymptotic model. During the
twenty-year period of my scientific activity at the University of Lille-I which was
devoted mainly to the modelling of fluid flow phenomena (see my monograph
(2002)), I published two books (1990 and 1991) and several survey and review
papers (1976, 1982, 1983, 1985, 1991) concerning asymptotic models of atmo-
spheric motions. In this scientific endeavour I was strongly influenced by my own
conception of meteorology (doubtless an unconscious inheritance from my
muscovite scientific period from 1957 to 1966, with I. A. Kibel) as a fluid dynamics
discipline which lies in a privileged area for the application of the RAM Approach!
My purpose (which was certainly rather naive) was, during the 1990s, to initiate a
process which does not seem to have sufficiently attracted the attention of scientists.
Namely, the use of methods of formal asymptotic analysis for carrying out asymp-
totic modelling; that is, for building approximate consistent theoretical models
based on various meteo situations and atmospheric motions for the use in weather
forecasting! Conceptually, thanks to such an asymptotic (outerinner) approach,
where the matching plays a central role: the problem of taking into account
the atmospheric initial data (using weather observations) for the used, reduced
(filtered!) model is then well-posed!1
A recent operational approach at Meteo-France, working with three coupled
models Arpege-Aladin, for a numerical weather prediction model for Central
Europe, and also Arome as a regional prediction model is conceptually very
1
In the case of a hydrostatic large horizontal-scale numerical weather prediction model, this
initialization is consistently solved thanks to the derivation, near the (weather prediction) initial
time, of a system of local unsteady equations (with a short time t t=e, where e is the small
hydrostatic parameter (Sect. 9.2) in the framework of a well-posed initial-value (a la Cauchy)
problem. It might be true that adjustment to the state of static equilibrium is brought about (by the
generation and scattering of internal acoustic waves) during only a few minutes in all (see Monin
(1972), Section 6, and also our Meteorological Fluid Dynamics (1991), Chapter 5). Unfortunately,
the theoretical analysis (for an adiabatic atmospheric motion) by Guiraud and Zeytounian (1982),
and also the numerical simulation by Outrebon (1981), show that: The two sets of initial
conditions for full adiabatic, Euler, atmospheric equations, and the so-called Kibel primitive
equations, subject to hydrostatic approximation are merely shifted vertically by the amount of
vertical displacement during the whole process (during t "/ and matching) of the vertical, one-
dimensional, unsteady adjustment, atmospheric adiabatic, motion; this vertical shift being a quite
significant phenomenon (Outrebon (1981)).
Epilogue 259
different from our, a la Zeytounian RAM Approach, based on asymptotics with

matching. This is unfortunate, because the Arpege model, which is a pilot model
for the Aladin model is, in fact, as large synoptic hydrostatic model, derived from
the hydrostatic limiting process (Sect. 9.2.2): e ! 0. In such a case we see that the
problem concerning the consistent determination of the associated initial conditions
for this Arpege model, in the framework of the RAM Approach, is linked with an
unsteady adjustment problem where, in place of the time t, it is necessary to
introduce a short time t t=e. Relative to t, the acoustic-type model equations,
for this unsteady adjustment problem, are derived from the full starting, non-
hydrostatic, dimensionless equations for atmospheric motions.
In place of this RAM Approach via the unsteady adjustment problem and
matching, t !/ , t 0 Meteo-France uses a statistical approach inspired
mainly by Kalnays (2002) book, which covers methods for numerical modelling,
data assimilation (for the determination of initial conditions using weather
observations), and predictability, and includes a discussion of equations of motion
and their approximations.
Obviously, the actual meteo philosophy is completely different from our RAM
Approach, which is specifically a fluid dynamics approach using asymptotics. But
perhaps this meteo philosophy is well-adapted to awkward practical weather-
forecasting meteo machinery, which every day, amidst more serious business,
shows, as entertainment, what the weather will be like tomorrow or for the next
few days! Of course, weather forecasting is also an important business. . . but I
think that atmospheric motions, as a fluid dynamics problem, poses, for our
environment, numerous very interesting and specific challenges. And obviously,
the applications of the RAM Approach for these specific terrestrial problems
undoubtedly deserves particular attention!
Cited references
Hills, R., Roberts, P.: Stab. Appl. Anal. Continuous Media. 1, 205212 (1991)
Straughan, B.: The Energy Method, Stability, and Nonlinear Convection. Appl. Math. Sci, vol. 91,
2nd edn. Springer, New York (2004)
Chandrasekhar, S.: Hydrodynamic and Hydromagnetic Stability. Clarendon, Oxford (1961)
Cercignani, C., Sattinger, D.H.: Scaling Limits and Models in Physical Processes. DMV Seminar,
Band 28, Birkhauser (1998)
de Gennes, P.-G., Brochard-Wyart, F., Quere, D.: Capillarity and Wetting phenomena, Drops,
Bubbles, Pearls, Waves. Springer, New York (2004)
Zeytounian, R.Kh.: Asymptotic Modelling of Fluid Flow Phenomena. FMIA, vol. 64. Kluwer,
Dordrecht (2002)
Zeytounian, R.Kh.: Asymptotic Modeling of Atmospheric Flow. Springer, Berlin/Heidelberg
(1990)
RKh, Zeytounian: Meteorological Fluid Dynamics. Lecture Notes in Physics, m5. Springer,
Heidelberg (1991)
RKh, Zeytounian: La Meteorologie du point de vue du Mecanicien des Fluides. Fluid Dynamics
Transactions, vol. 8, pp. 289352. Polish Academy of Sciences, Warszawa (1976)
260 Epilogue
RKh, Zeytounian: Asymptotic phenomena in meteorology. In: Miller, J.J.H. (ed.) Proceedings of
the BAIL II Conference, pp. 124138. Boole, Dublin (1982)
Zeytounian, R.Kh.: Modeles asymptotiques pour les ecoulements de fluides pesants en rotation
Application a la Meteorologie. Conference generale invitee au 6e Congres de Mecanique,
pp. 3.13.42. Lyon, 59 Sep 1983
Zeytounian, R. Kh.: Asymptotically Consistent Models for Atmospheric Flows. CISM, Von
Karman session, pp. 121. Udine (Italy) Oct 1983
RKh, Zeytounian: Recent advances in asymptotic modelling of tangent atmospheric motions. Int.
J. Eng. Sci. 23, 12391288 (1985)
Zeytounian, R.Kh.: Fluid dynamics limits of meteorological equations. Adv. Mech. 14(3), 386
(1991)
Kalnay, E.: Atmospheric Modeling, Data Assimilation and Predictability. Cambridge University
Press, Cambridge (2002)
References
1. Zeytounian, R.Kh.: Etudes Hydrodynamique des Phenomenes Mesometeorologiques. Cours

a lEcole de la Meteorologie, pp. 163. Direction de la Meteorologie Nationale, Paris (1968)
2. Zeytounian, R.Kh.: Etude des Phenomenes dondes dans les Ecoulements Stationnaires dun
Fluide Stratifie non Visqueux. Publication ONERA N 126, 94 pages, Chatillon, fevrier 1969.
English version: Study of Wave Phenomena in the Steady Flow of an Inviscid Stratified Fluid.
Royal Aircraft Establishment, Library Translation N 1404 (1969)
3. Landau, L.D., Lifschitz, E.M.: Fluid Mechanics. Pergamon, London (1959)
4. Yih, C.-S.: Stratified Flows. Academic, London (1980)
5. Batchelor, G.K.: An Introduction to Fluid Dynamics. Cambridge University Press,
Cambridge (1988)
6. Zeytounian, R.Kh.: Three-dimensional rotational motion in a compressible fluid. Atmo-
spheric and Oceanic Physics. vol. 2, p. 61. Feb 1966. English version of a Russian edition
7. Long, R.R.: Some Aspects of the Flow of Stratified Fluids: III. Continuous Density Gradients.
Tellus, vol. 7, N 3, (1955)
8. Boussinesq, J.: Theorie Analytique de la Chaleur, vol. 2. Gauthier- Villars, Paris (1903)
9. Zeytounian, R.Kh.: A rigorous derivation of the equations of the compressible viscous fluid
motion with gravity at low mach number. Archives of Mechanics/Archiwum Mechaniki
Stosowanej 26(3), 499509 (1974)
10. Zeytounian, R.Kh.: On the foundations of the Boussinesq approximation applicable to
atmospheric motions. Izv Atmos Ocean Phys 39(Suppl. 1), S1S14 (2003). English version
of a Russian edition
11. Guiraud, J.-P., Zeytounian, R. Kh.: Asymptotic modelling of fluid flows. General features.
J. Theor. Appl. Mech. Special Issue 1986, 124 (1986)
12. Zeytounian, R.Kh.: Asymptotic Modelling of Atmospheric Flows. Springer, Berlin/
Heidelberg (1990)
13. Zeytounian, R.Kh.: Topics in Hyposonic Flow Theory. LNP, vol. 672. Springer, Berlin/
Heidelberg (2006)
14. Van Dyke, M.: Perturbation Methods in Fluid Mechanics (Annotated edition of Perturbation
Methods in Fluid Mechanics. Academic, New York/London, 1964). The Parabolic, Stanford
(1975)
15. Zeytounian, R.Kh.: Les Modeles Asymptotiques de la Mecanique des Fluides I. LNP,
vol. 245. Springer, Berlin/Heidelberg (1986)
16. Zeytounian, R.Kh.: Les Modeles Asymptotiques de la Mecanique des Fluides II. LNP,
vol. 276. Springer, Berlin/Heidelberg (1987)
17. Zeytounian, R.Kh.: Modelisation asymptotique en mecanique des fluides newtoniens.
Mathematiques et Applications, SMAI, vol. 15. Springer, Berlin/Heidelberg (1994)

262 References
18. Zeytounian, R.Kh.: La Meteorologie du point de vue du Mecanicien des Fluides. Fluid
Dynamics Transactions, vol. 8, pp. 289352. Polish Academy of Sciences, Warszawa (1976)
19. Zeytounian, R.Kh.: Meteorological Fluid Dynamics. Lecture Notes in Physics, vol. m5.
Springer, Heidelberg (1991)
20. Holton, J.R.: An Introduction to Dynamic Meteorology. Academic, New York (1979)
21. Gill, A.E.: AtmosphereOcean Dynamics. Academic, New York (1982)
22. Haltiner, G.J., Williams, R.T.: Numerical Prediction and Dynamic Meteorology. John Wiley,
New York (1980)
23. Pedlosky, J.: Geophysical Fluid Dynamics. Springer, New York (1987)
24. Yang, H.: Wave Packets and Their Bifurcations in Geophysical Fluid Dynamics. Appl. Math.
Sci., vol. 85. Springer, New York (1990)
25. Zeytounian, R.Kh.: Recent advances in asymptotic modelling of tangent atmospheric
motions. Int. J. Eng. Sci. 23, 12391288 (1985)
26. Zeytounian, R.Kh.: Asymptotic Modelling of Fluid Flow Phenomena. FMIA, vol. 64,
pp. xii545. Kluwer Academic, Dordrecht (2002)
27. Zeytounian, R.Kh.: Convection in Fluids: A Rational Analysis and Asymptotic Modelling.
FMIA, vol. 90, pp. xv396. Springer, Dordrecht (2009)
28. Birkhoff, G.: Numerical fluid dynamics. SIAM Rev. 25(1), 136 (1983)
29. Zeytounian, R.Kh.: A historical survey of some mathematical aspects of Newtonian fluid
flows. Appl. Mech. Rev. 54(6), 525562 (2001)
30. Cercignani, C., Illner, R., Pulvirenti, M.: The Mathematical Theory of Dilute Gas. AMS,
vol. 106. Springer, New York (1994)
31. Serrin, J.: Mathematical Principles of Classical Fluid Mechanics. Handbuch der Physik,
vol. VIII/1, pp. 125263. Springer, Berlin (1959)
32. Newton, I.: Philosophiae Naturalis Principia Mathematica. Joseph Streater, London (1687)
33. Euler, L.: Principes generaux du mouvement des fluides. Hist. Acad, pp. 274315. Berlin
(1755)
34. Bernoulli, J.: Hydraulics (1737)
35. Bernoulli, D.: Hydrodynamica. Argentorati (1738)
36. Le Rond, J., DAlembert.: Traite de lequilibre et du Mouvement des Fuides (1744)
37. Zeytounian, R.Kh.: Theory and Applications of Non-viscous Fluid Flows. Springer, Berlin/
Heidelberg (2002)
38. Birch, T.: The Works of the Honourable Robert Boyle, vol. 5. A. Miller, London (1744)
39. Kirchhoff, G.: Ann. der. Phys. 134, 177193 (1868)
40. Navier, C.-L.: Sur les lois des mouvements des fluides, en ayant egard a ladhesion des
molecules. Ann. de Chimie 19, 244260 (1821)
41. Poisson, S. D.: Memoire sur les equations generales de lequilibre et du mouvement des corps
solides elastiques et des fluides. J. Ecole Polytechnique. 13 cahier 20, 1174 (1831)
42. Truesdell, C.: The Mechanical Foundations of Elasticity and Fluid Dynamics. Gordon and
Breach, New York (1966)
43. Cauchy, A.: Ex. de math. 3. Oeuvres (2) 8, 195226. 22752, 25377 (1828)
44. Truesdell, C.: J. Rat. Mech. Anal. 1, 125 (1952)
45. Truesdell, C.: Amer. Math. Monthly. 60, 445 (1953)
46. Hamel, G.: Math. Ann. 66, 350 (1908)
47. Zeytounian, R.Kh.: Theory and Applications of Viscous Fluid Flows. Springer, Berlin/
Heidelberg (2004)
48. Stokes, G.G.: On the theories of the internal friction of fluids in motion, and of the
equilibrium and motion of elastic solids. Trans. Camb. Philos. Soc 8, 287319 (1845)
49. Saint-Venant, A.J.B. (de): C. R. Acad. Sci., Paris 17, 12401243 (1843)
50. Kazhikov, A.: Some new statements for initial boundary value problems for NavierStokes
equations of viscous gas. In: Galdi, G.P., Malek, J., Necas, J. (eds.) Progress in Theoretical
and Computational Fluid Mechanics, pp. 3372. Longman Scientific and Technical,
New York (1994)
References 263
51. Lions, P.L.: Mathematical Topics in Fluid Mechanics, Volume 2: Compressible Models.
Oxford Lecture Series in Mathematics and its Applications. Clarendon, Oxford (1998)
52. Leray, J.: La vie des sciences. Comptes Rendus Acad. Sci. Serie Generale 11, 287290 (1994)
53. Gibbs, J.W.: On the equilibrium of heterogeneous substances. Trans. Conn. Acad. 3, 108248
and 343524 (1875)
54. Fourier, J.: Memoire danalyse sur le mouvement de la chaleur dans les fluides. Memoires de
lAcademie Royale des Sciences de lInstitut de France (2), vol. 12, (1833), 507530 (with
editorial changes), pp. 595616 of Oeuvres de Fourier, Tome 2 (1890)
55. Poincare, H.: Thermodynamique. Gauthier-Villars, Paris (1892)
56. Truesdell, C., Muncaster, R.G.: Fundamental of Maxwells Kinetic Theory of a Simple
Monoatomic Gas. Academic, New York (1980)
57. Kreiss, H.O.: Initial boundary value problems for hyperbolic equations. Comm. Pure Appl.
Math. 23, 277298 (1970)
58. Kreiss, H.O.: Initial Boundary Value Problems for Hyperbolic Equations. Lecture Notes in
Mathematics, vol. 363, pp. 6474. Springer, Berlin (1974)
59. Belov, Yu.Ya., Yanenko, N.N.: Influence of viscosity on the smoothness of solutions of
incompletely parabolic system. Math. Notes 10, 480483 (1971)
60. Oliger, J., Sundstr om, A.: Theoretical and practical aspects of some initial boundary value
problems in fluid dynamics. SIAM J. Appl. Math. 35, 419446 (1978)
61. Majda, A.: Compressible Fluid Flow and Systems of Conservation Laws in Several Space
Variables. AMS, vol. 53. Springer, New York (1984)
62. Kreiss, H.O., Lorenz, J.: Initial-Boundary Value Problems and the NavierStokes Equations.
Academic press, Boston (1989)
63. Greenspan, H.P.: The Theory of Rotating Fluids. Cambridge University Press, Cambridge
(1968)
64. Nayfeh, A.H.: Perturbation Methods. Wiley, New York (1973)
65. Beirao da Veiga, H.: Singular limits in compressible fluid dynamics. Arch. Ration. Mech.
Anal. 128, 313327 (1994)
66. Gresho, P.M.: Advances in Applied Mechanics. vol. 28, pp. 45140 (1992)
67. Gustafsson, B., Sundstr om, A.: Incompletely parabolic problem in fluid dynamics. SIAM
J. Appl. Math. 35, 343357 (1978)
68. Viviand, H., Veuillot, J.P.: Publication ONERA, N 19784 (1978)
69. Viviand, H.: Publication ONERA, N 19834 (1983)
70. Velarde, M.G., Zeytounian, R.Kh.: Interfacial Phenomena and the Marangoni Effect. CISM
Courses and Lectures, vol. 428. Springer, Vienna/New York (2002)
71. Whitham, G.B.: Linear and Nonlinear waves. Wiley-InterScience, New York (1974)
72. Zeytounian, R. Kh.: Nonlinear long waves on water and solutions. Physics-Uspekhi, English
edition of the Russian Uspekhi Fizicheskikh Nauk, 38(12), 13331381 (1995)
73. Dobrokhotov, S.Yu., Safarevich, A.I.: On the behavior of an incompressible fluid velocity
field at infinity. Fluid Dynamics, 31(4), 511514 (1996)
74. Serrin, J.: On the uniqueness of compressible fluid motion. Arch. Ration Mech. Anal.
3, 271288 (1959)
75. Batchelor, G. K.: J. Fluid. Mech. 1, 177190 (1956)
76. Wood, W.W.: Boundary layers whose streamline are closed. J. Fluid Mech. 2, 7787 (1957)
77. Guiraud, J.-P., Zeytounian, R.Kh. (eds.): Asymptotic modelling of fluid flows.
J. de Mecanique Th. et Appl. Special Issue 1986, 313 pages (1986)
78. Doering, C.R., Gibbon, J.D.: Applied Analysis of the NavierStokes Equations. Cambridge
University Press, Cambridge (1995)
79. Zeytounian, R.Kh.: Well-posedness of problems in fluid dynamics (a fluid-dynamical point of
view). Russian Math. Surv 54(3), 479564 (1999)
80. Riley, N.: J. Inst. Math. Appl. 3, 419454 (1967)
81. Strikwerda, J. C.: Comm. Pure Appl. Math. 30, 797822 (1977)
264 References
82. Valli, A.: Mathematical results for compressible flows. In: Rodrigues, J.F., Sequeira, A. (eds.)
Mathematical Topics in Fluid Mechanics. Longman Scientific, New York (1992)
83. Temam, R.: Some development on NavierStokes equations in the second half of the
twentieth century. In: Pier, J.P. (ed.) Development of Mathematics 19502000,
pp. 10491106. Birkauser, Basel (2000)
84. Leray, J.: Etude de diverses equations integrales non lineaires et de quelques problemes que
pose lhydrodynamique. J. Math. Pures Appl. 12, 182 (1933), and see also: Acta Math.,
vol. 63, 193248, 1934
85. Zeytounian, R.Kh., Guiraud, J.-P.: Evolution dOndes Acoustiques dans une Enceinte et
Concept dIncompressibilite. Publications IRMA, vol. 2, fasc. 2, n 4, IV1 a IV35. Lille
(1980)
86. Guiraud, J.-P., Zeytounian, R. Kh.: Proceedings of the BAIL I Conference, pp. 297300.
Boole Press, Dublin, June 1980. See also: Zeytounian, R.Kh., Guiraud, J.-P.: Comptes rendus
Acad. Sci., Paris, 290, B, 7578 (1980)
87. Wilcox, C.: Scattering Theory for the dAlembert Equation in Exterior Domain. Lecture
Notes in Mathematics, vol. 442. Springer, Berlin (1975)
88. Zeytounian, R. Kh.: Int. J. Eng. Sci. 38 (17), 19831992 (2000)
89. Stokes, G. G.: Trans. Cambridge Phil. Soc., vol. ix (8), Papers iii, 1, 1851
90. Oseen, C.W.: Arkiv f ur Mat., Astr. och Fysik, vol. 6, N 29, 1910; 7, N 1 (1911)
91. Lagerstrom, P. A.: Laminar Flow Theory, In: Moore, F.K. (ed.) Theory of Laminar Flows,
pp. 20285. Princeton University Press, Princeton, Section B (1964)
92. Coulouvrat, F.: J. Acoustique. 5, 321359 (1992)
93. Kevorkian, J., Cole, J.D.: Multi Scale and Singular Perturbation Methods. AMS, vol. 114.
Springer, New York (1996)
94. Monin, A.S.: Weather Forecasting as a Problem in Physics. MIT, Cambridge (1972). English
translation
95. Miles, J. W.: J. Fluid Mech. 33(4), 803814 (1968)
96. Kozhevnikov, V.N.: Isv. Akad. Nauk SSSR, Atmosph. Ocean Phys. 4(1), 1627 (1968)
97. Guiraud, J.-P., Zeytounian, R. Kh.: Geophys. Astrophys. Fluid Dyn. 12(1/2), 6172 (1979)
98. Zeytounian, R.Kh.: Comptes rendus Acad. Sci., Paris, 290(Serie A):567569 (1980)
99. Howarth, L. N.: Q. J. Mech. Appl. Math. 4 (2), 157165 (1951)
100. Rozhdestvenskii, B.L., Yanenko, N.N.: Quasi-linear Systems of Equations and their
Applications to Gas Dynamics. Nauka, Moscow (1968) (Russian edition)
101. Stewartson, K.: The Theory of Laminar Boundary-Layers in Compressible Fluids. Clarendon,
Oxford (1964)
102. Oleinik, O.A., Samokhin, V.N.: Mathematical Models in Boundary Layer Theory. Chapman
& Hall/CRC, Boca Raton (1999)
103. Van Dyke, M.: Nineteenth-century roots of the boundary-layer idea. SIAM Rev. 36, 415424
(1994)
104. Antonsev, S.N., Kazhikhov, A.V., Monakhov, V.N.: Boundary Value Problems in Mechanics
of Nonhomogeneous Fluids. Nauka, Novosibirsk (1983)
105. Solonnikov, V.A., Kazhikhov, A.V.: Existence theorems for the equations of motion a
compressible viscous fluid. Ann Rev. Fluid Mech. 13, 7995 (1981)
106. Guiraud, J.-P., Zeytounian, R.Kh.: A note on the adjustment to hydrostatic balance. Tellus
34(1), 5054 (1982)
107. Germain, P.: The New Mechanics of Fluids of Ludwig Prandtl. In: Ludwig Prandtl ein
Fuhrer in der Stromungslehre. Meier, G. (ed) pp. 3140. Vieweg, Braunschweig (2000)
108. Germain, P.: My discovery of mechanics. In: Maugin, G.A., Drouot, R., Sidoroff, F. (eds.)
Continuum Thermomechanics. Paul Germains Anniversary Volume, pp. 127. Kluwer
Academic Publishers, Dordrecht (2000)
109. Eckhaus, W.: Asymptotic Analysis of Singular Perturbations. North-Holland, Amsterdam
(1979)
References 265
110. Temam, R.: NavierStoles Equations, a talk at the Special Session. On the Diverse Mathe-
matical Legacy of Jean Leray. American Math. Soc., Austin, Texas, 8 Oct 1999
111. Chemin, J.-Y. Jean Leray et NavierStokes. In Jean Leray 19061998, pp. 7182. No. special
de la Gazette des Mathematiciens, SMF, Paris (2000)
112. Ladyzhenskaya, O.A.: The Mathematical Theory of Viscous Incompressible Flow. Gos Izdat
Fiz-Mat Lit., Moskva, 1961 (in Russian). English translation by Gordon and Breach,
New York (1963)
113. Quartapelle, L.: Numerical Solution of the Incompressible Navier Stokes Equations.
Birkhauser, Basel (1993)
114. Feireisl, E.: Dynamic of Viscous Compressible Fluids. Oxford Science, Oxford (2004)
115. Novotny, A., Straskraba, I.: Introduction to the Mathematical Theory of Compressible Flow.
Oxford Lecture Series in Mathematics and its Applications, Oxford (2004)
116. Guiraud, J.-P., Zeytounian, Kh: A double-scale investigation of the asymptotic structure of
rolled-up vortex sheets. J. Fluid Mech. 79(part 1), 93112 (1977)
117. Guiraud, J.P., RKh, Zeytounian: Vortex sheets and concentrated vorticity: a variation on the
theme of asymptotic modelling in fluid mechanics. In: Hormung, H.G., M uller, E.A. (eds.)
Proceeding of the colloquium Vortex Motion. Vieweg and Son, Braunschweig (1982)
118. Guiraud, J.-P.: Ann. de Physique France. 5, 3359 (1980)
119. Stewartson, K., Stuart, J. T.: J. Fluid Mech. 48 (3), 529545 (1971)
120. Guiraud, J.-P., Zeytounian, R. Kh.: J. de Mecanique. 17 (3), 337402 (1978)
121. De Coninck, F., Guiraud, J.-P., Zeytounian, R. Kh.: Q. J. Mech. Appl. Math. 36 (1), 118
(1983)
122. Feuillebois, F., Lasek, A.: Une methode de calcul des degenerescences significatives
et quelques applications en mecanique des fluides. J. Theor. Appl. Mech. N special 1986
(Guiraud, J.-P., Zeytounian, R. Kh., eds) 295313 (1986)
123. Godts, S., Zeytounian, R. Kh.: ZAAM. 70 (1), 6769 (1990)
124. Eckhaus, W.: Fundamental concepts of matching. SIAM Rev. 36(3), 431439 (1994)
125. Lagerstrom, P.A.: Matched Asymptotic Expansions Ideas and Techniques. AMS, vol. 76.
Springer, New York (1988)
126. Germain, P.: Methodes asymptotiques en mecanique des fluides. In: Balian, R., Peube, J.L.
(eds.) Fluid Dynamics, les Houches Lectures, pp. 3147. Gordon and Breach, New York/
Paris (1977)
127. Guiraud, J.-P.: In: Asymptotic modelling in fluid mechanics. Bois, P.A., Deriat, E., Gatignol,
R., Rigolot, A. (eds.). Lecture Notes in Physics, vol. 442, pp. 257307. Springer, Berlin/
Heidelberg (1995)
128. Robinson, A.: Nonstandard Analysis. North-Holland, Amsterdam (1966)
129. Lightstone, A.H., Robinson, A.: Non-Archimedean Fields and Asymptotic Expansions.
North-Holland, Amsterdam (1975)
130. Lutz, R., Goze, M.: Nonstandard analysis. A practical guide with applications. Lecture Notes
in Mathematics, vol. 881. Springer, Berlin/Heidelberg/New York (1981)
131. Eckhaus, W.: In: Lecture Notes in Mathematics. 985, Verhulst, F. (ed.), Springer, Berlin
(1983)
132. Braaksma, B.: Phantom ducks and models of excitability. J. Dynam Differ Equat 4, 485513
(1992)
133. Germain, P.: Progressive Waves. 14th Ludwig Prandtl Memorial Lecture (Mannheim, 16
April 1971). Sonderdruck aus dem Jahrbuch 1971 der DGLR
134. Guiraud, J.-P.: Focalisation dans les ondes courtes non lineaires. Application au bruit
balistique de focalisation. Colloque AGARD. ONERA, TP 686 (1969)
135. Sanchez-Palencia, E.: Comportement local et macroscopic dun type de milieux physiques
heterogenes. Int. Eng. Sci. 12, 331351 (1974)
136. Mei, C.C., Auriault, J.-L., Nig, C.O.: Some application of the homogenization theory. Adv.
Appl. Mech 32, 277348 (1996)
266 References
137. Stewartson, K.: Multistructural boundary layers on flat plates and related bodies. Adv. Appl.
Mech 14, 145239 (1974)
138. Meyer, R.E.: A view of the triple deck. SIAM J. Appl. Math. 43(3), 639663 (1983)
139. Neiland, V.Ya.: Towards a theory of separation of the laminar boundary layer in a supersonic
stream. Isv. Akad. Nauk. SSSR. Mekh. Zhidkosti i Gaza 4, 3335 (1969)
140. Veuillot, J.-P. La Recherche Aerospatiale. vol. 6, p. 313, ONERA (1971)
141. Guiraud, J.-P., Zeytounian, R.Kh.: La Recherche Aerospatiale. vol. 2, pp. 6587, ONERA
(1971)
142. Veuillot, J.-P.: Calculation of the Quasi Three-dimensional Flow in a Turbomachine Blade
Row. Trans. ASME, J. Eng. Power, 76-GT-56, 110 (1976)
143. Wu, Ch-Hua.: Trans. ASME, 74(8), 1363 (1952)
144. Guiraud, J.-P., Zeytounian, R.Kh.: La Recherche Aerospatiale. vol. 5, pp. 237256, ONERA
(1971)
145. Guiraud, J.-P., Zeytounian, R.Kh.: Application du concept dechelles multiples a
lecoulement dans une turbomachine axiale. Int. J. Eng. Sci. 12, 311330 (1974)
146. Guiraud, J.-P., Zeytounian, R.Kh.: Cascade and through-flow theories as inner and outer
expansions. Int. J. Eng. Sci. 16, 515526 (1978)
147. Sirotkin, Ya.A.: Fluid Dynamics (Translation of Izv. A. N. SSSR, Mechanika Zhidkosti i
Gaza), vol. 2, pp. 64 (1967)
148. Chester, W.: J. Fluid Mech. 18, 44 (1964)
149. Rott, N.: Rimrot and Tabarrok (ICTAM, 163), North Holland (1980)
150. Muller, B.: Low Mach number asymptotics of the NavierStokes equations. J. Eng. Math.
34, 97109 (1998)
151. Meister, A.: Asymptotic single and multiple scale expansions in the low mach number limit.
SIAM J. Appl. Math. 60, 256271 (1999)
152. Ali, G.: Low mach number flow in time-dependent domain. SIAM J. Appl. Math. 63(6),
20202041 (2003)
153. Zeytounian, R.Kh.: C. R. Acad. Sci., Paris, serie I, 297, 271274 (1983)
154. Zeytounian, R. Kh.: Int. J. Eng. Sci. 27(11), 13611366 (1989)
155. Zeytounian, R. Kh.: Int. J. Engng. Sci. 35(5), 455466 (1997)
156. Zeytounian, R. Kh.: The BenardMarangoni thermocapillary instability problem. Uspekhi
Fizicheskikh Nauk, 41(3), 241267 (1998) (English edition)
157. Davis, S. H.: Ann. Rev. Fluid Mech. 19, 403435 (1975)
158. Parmentier, P.M., Regnier, V.C., Leblond, G.: Nonlinear analysis of coupled gravitational
and capillary thermoconvection in thin fluid layers. Phys. Rev. E 54(1), 411423 (1996)
159. Joseph, D.S.: Stability of Fluid Motions, vol. II. Springer, Heidelberg (1976)
160. Takashima, M.: J. Phys. Soc. Japan, 50(8), 27452750 and 27512756 (1981)
161. Turcotte et al., J. Fluid Mech.64, 369 (1974)
162. Crowley, D.G., Lawrence, C.J., Wilson, S.K.(eds.): The Dynamics of Thin Liquid Film.
J. Eng. Maths. Special Issue, 50(23) (2004)
163. Rayleigh, L: On convection currents in horizontal layer of fluid when the higher temperature
is on the underside. Philos. Mag., Ser. 32(6), 529546 (1916)
164. Pearson, J.R.A.: On convection cells induced by surface tension. J. Fluid Mech. 4, 489 (1958)
165. Block, M.J.: Surface tension as the cause of Benard cells and surface deformation in a liquid
film. Nature 178, 650651 (1956)
166. Guyon, E., Hulin, J. P., Petit, L.: Hydrodynamique Physique. Inter Edition/Editions du
CNRS, Paris 1991. English edition (with Mitescu, C.D.): Physical Hydrodynamics, Oxford
University Press (2001)
167. VanHook, S. J., Swift J. B.: Phys. Rev. E. 56(4), 48974898 (1997)
168. Oron, A., Davis, S.H., Bankoff, S.G.: Long-scale evolution of thin liquid films. Rev. Mod.
Phys. 69(3), 931980 (1997)
169. Dorodnitsyn, A. A.: Trudy Tsentral Inst. Prognoz. 21(48), 335 (1950)
References 267
170. Zeytounian, R.Kh.: Criterion for filtering parasite solutions in numerical computations. Phys.
Fluids. 12(Suppl. II), II46II50 (1969)
171. Kibel, I.A.: An Introduction to the Hydrodynamical Methods of Short Period Weather
Forecasting. Pergamon, New York, 1963. Translated from the Russian edition, Moscow
(1957)
172. Norbury, J., Roulstone, I. (eds.): Large-Scale AtmosphericOcean Dynamics. I. Analytical
Methods and Numerical Models. II. Geometric Methods and Models. Cambridge University
Press, Cambridge (2002)
173. Guiraud, J.-P., Zeytounian, R.Kh.: Asymptotic derivation of the ageostrophic model for
atmospheric hydrostatic flows. Geophys. Astrophys. Fluid Dyn. 15((3 + 4)), 283295 (1980)
174. Charney, J.G.: Proceedings of International Symposium Numerical Weather Prediction,
p. 131. Tokyo, Meteorological Society of Japan (1962)
Index
A Atmospheric thermal problem, 135137

Ackerbloms problem, 236, 243, 245 Atmospheric values, 219
Acoustic/gas dynamics equations, 111 Average, 175
Acoustics, 171 continuity equation, 174178
eigenfrequencies, 190 system, 187
equations, 6869, 7677
limiting process, 93, 99
oscillations, 173 B
Reynolds number, 60, 76 Balance equation, 247
waves are filtered, 67 Baroclinic motion, 32
Ad hoc procedures, 119 Basic postulate, 117
Adiabatic case, 82 Basic small parameter, 124
Adiabatic flow, 24 Behaviour conditions, 91, 228, 231, 233
Adjustment, 38 Benard experiments, 203
to geostrophic balance, 240 Benard heated-from-below thermal convection
to hydrostatic balance, 232233 problem, 71, 193
of meteorological fields, 231 BenardMarangoni thermocapillary
problem to Boussinesq approximate convection, 203
equations, 9294 Benard problem, 43, 195
Adjustment (Adj), 237 Bernoulli first integral, 214
Aerodynamics, 60 Bifurcation-type equations, 132
Ageosprophic asymptotic model, 237 Biot convective number, 198
adiabatic equation, 108 Biot numbers, 43
a la Ladyzhenskaya result, 122 Blade skeletons, 125
a la NSF-dominant, 76 Blade-to-blade channel, 126
a la Stewartson, 157158 Blasius problem, 140
a la Van Dyke, matching, 141 Block conclusion, 204
Algebraic latitude, 224 BL Prandtl concept, near t0!, 98
Alis paper, 187 BL Prandtl equations, 27, 103104
Amplitude function, 148 BM long-wave problem, 206
Angular velocity, 55, 223 BM model problem, 194
Asymptotic Body shape, 4445
acoustic expansions, 108 Boltzmann equation, 19
expansions, 65, 73, 77, 79, 86, 104 Bottom boundary condition, 44
rational approach, 10, 84 Boundary conditions, 40, 228, 234
asymptotics of rolled vortex sheets, 128129 Boussinesq approximation, 7, 81, 136
Atmospheric flows, 9, 89, 223 Boussinesq case, 222

270 Index
Boussinesq model equations, 8688 Coriolis parameter, 55, 225

Boussinesq limiting process, 86 Corner Ekman region, 237
Boussinesq number, 46, 53, 83, 197, 218, 224 Corner small fourth Reyleigh region, 99
Boussinesq similarity rule, 84 Creeping flows, 60
Boussinesq number in Benard problem, 210 Curious condition, 243
Bousssinesq number in lee-waves problem, Cut-off wave number, 209
218
Bulk-viscous fluids, 40
BurgersKortewegde Vries (KdV) D
equation, 147 Damping, acoustic oscillations, 190
Damping phenomenon, 187
Darcys law, 153
C Daviss condition, 196
Canard problem, 145 Deconstruction, 65, 247
Canonical Stokes flow, 152 Deep convection case, 220222
CauchyPoisson law, 30 Deep (Zeytounian) convection equations, 85
Cauchy initial value problem, 36, 58 Deep convection model, 210211
Cauchys principle, 28 Deep thermal convection model problem, 55,
Cauchy stress tensor, 20 196, 202
Characteristic surface, 42 Deformation tensor, 20
ClausiusDuhem inequalities, 31 2D equation of Long, 5, 90, 220
Closed system, 120 Derivation-justification of Boussinesq
Closeness parameter, 128, 129 equations, 86
Coefficient of cubical expansion, 54 Deviation in altitude, 216
Companion dimensionless perturbation DH equations written in p-system, 234
pressure, 200 Dimensionless
Companion simplified leading-order altitude far upstream, 217
Boussinesq equation, 87 equations, 124
Compatibility Euler equations, 85
condition, 182 formulation, 218220
relation, 113 main equation, 218
Compressible Rayleigh problem, 109 NSF equations, 100
Compressible Stokes equations, 74 quantities, 82
Compression phase of an internal combustion temperature, 197
engine, 69, 190 Dissipative (non-hydrostatic) adjustment
Concentrated baroclinicity, 127 equations, 232
Concept of matching, 65, 141 Dissipative data, 229230
Concept of significant degeneracy, 139, 144 Dissipative hydrostatic data, 230231
Condition at infinity, 45, 91 Dissipative non-hydrostatic, 226
Condition, temperature on the wall, 47, 72 Distribution function, 19
Conductance, 195 Divergence, free flows, 36, 87
Conduction Biot number, 195 DorodnitsynScorer parameter, 6, 220
Conductive fluids, 40 Double limiting process, 73, 138
Conservation of energy, 33 Double scale, 92
Consistency of the model, 118 Downstream from the mountain, 83
Consistent initial conditions, 243 Dry adiabatic temperature gradient, 220
Consistent upper condition, 92 Dry atmospheric air, 224
Constitutive equations, 20 3D system, 76
Convection, 196 Dynamic condition, 44
Convective Biot number, 208210
Convective Rayleigh-Benard instability, 135
Cooling of liquid with deformable surface, 194 E
Core of a vortex sheet, 127 Earths radius, 55
Coriolis force, 55, 155, 223 Eikonal equation, 148
Index 271
Ekman BL, 237 Force of Archimedes, 46

Ekman boundary-layer, 56, 236 Formal expansion, 128
Ekman number, 56, 226 Four distinguished limiting cases, 220223
Elastic fluid, 22, 24 Fourier equation, 57
Eliasen p-system, 234 Fouriers law, 33
Elimination of secular terms, 133 Four regions structure for NSF equations, 98
Emergency situation (sudden rise in Four time scales of evolution, 134
temperature locally on the wall), 98 Free-boundary problem, 36
Emergency-temperature-accident Free surface, 194
phenomena, 110 Fictitious pressure perturbation, 67
Energy equation, 33 Froude number (Fr), 6, 53, 61, 83, 210
Equations Full NSF non-dimensional equation, 52
for an expansible liquid, 70 Full NSF problem, 59
of continuity, 21
of continuum mechanics, 20
of Davey, Hocking and Stewarston, 134 G
for determination of free surface Gas dynamics problem, 44
deformation, 201 Gas is contained in a container O, 66
for slow variation, 182184 General models, 51
for specific entropy, 33 General slip condition, 47
of state, 23, 69 Geophysical nature application, 60
for thickness, 209 Geostrophic balance, 238
Equivalent source force, 149 Gradient operator, 83
Euler compressible dimensionless Grashof (Gr) number, 54, 61, 197
equations, 172 Gravitational acceleration, 223
Euler compressible equations, 63 Gravitational force, 21
Euler equations, 105 G-Z approach, 165168
Euler equations, for atmosphere, 82 GZ functional equation, 48
Eulerian fluid flow equations, 22
Euler inviscid compressible, adiabatic
equations, 103 H
Euler non-dissipative system, 3 Heal conductivity coefficient, 42
Euler process, 63 Heat conduction, 225
Expansibility parameter, 54, 61, 69 Heat flux vector, 20
Expansibity parameter, 197 Heat transfer, 23
Expansible liquid, 53 Helmholtz equation, 91, 156
Expansion for the pressure, 173 High Reynolds numbers, 98
Extension theorem, 142 Homogenization theory, 60, 149, 153, 168,
182184
Homogenized through-flow, 164
F Horizontal compressibility, 225
Far-field equation, 78 Hydraulic conductivity, 152
Far-field similarity, 45 Hydrodynamic stability, 9, 130, 134
Fast variable, 128 Hydrostatic
Fictitious force, 164 approximation, 56, 83
Fifth intermediate matching region, 112 balance, 60, 231
Filtering of acoustic fast waves, 26 constraint, 229
First integrals, 90, 214215 limiting case, 229
First key step, 82 limiting process, 249
First law, 23 parameter, 224
Flat ground, 56, 231, 233 Hydrostatic stability" parameter, 219, 220
Flow in turbomachinery row, 59 Hyperbolicparabolic equations, 57, 106,
Flow within cavity, 171191 120, 121
Fluid flow strongly viscous, 57 Hyperbolic systems, 35, 63
272 Index
Hypersonic similarity, 44, 45 Kibel number, 56, 60

Hyposonic similarity, 46, 60 Kibel number approximation, 229
Kibel primitive equations, 249
Kinematic condition, 44
I Kinematic viscosity, 225
Incompletely parabolic NSF system to Kinetic theory of gases, 47
hyperbolic Euler system, 65 Knudsen (Kn) number, 19, 47
Incompletely parabolic system, 57, 106, 121 Knudsen (Kn) sub-layer, 47
Incompressibility constraint, 41 Kuznetsov, Zabolotskaya, and Khokhlov
"Incompressible" Navier limit equations, 89 (KZK) equation, 76, 78, 79, 148
Incompressible process, 4, 171
Infinity
in altitude, 72, 234 L
of short acoustic scales, 171172, 174 Ladyzhenskaya theorem, 122123
Influence Landau equation, 133
of acoustics, averaged pressure, 183 LandauStuart equation, 131134
of free surface, 199201 Laplace elliptic equation, 37
Initial-boundary value Laplace equation, 3637, 44
dominant, 229 Large number of blades, 59
problems, 35 Large Reynolds number, 59, 140
Initial conditions, 36, 37, 69, 94, 101, 107, 110, Large-scale equations, 230231
228, 232, 234, 241 Latitude, 55, 56
Inner Boussinesq problem, 92 Law of Darcy, 152
Inner equations, 105, 241 Laws of mechanics, 21
Inner (BL) expansion, 244 Leading-order equation, 87
Inner expansions, 165 Leading-order QG model problem, 237247
Inner vertical coordinate, 104 Least-degeneracy principle, 144145
Interface, 43 Lee-waves problem, 5, 213223
Interfacial curvature, 43 Lerays first rigorous work, 122
Intermediate-matching, 141 Lerays pioneer papers, 5859
Intermediate-matching model equations, 112 Less significant limit system, 90
Internal consistency argument, 170 Limiting case for lee-waves, 220223
Internal short-wavelength gravity waves, 92 Limiting processes, 63, 86, 99, 107, 140, 199
Inviscid Euler fluid flows, 59 Linear acoustics system of equations, 68
Isentropic fluid flow, 59 Linear Helmholtz equation, 7, 91
Isentropicity, 34 Linearization, 218
Isobaric fluid flow, 31 Linear model equation for lee-waves, 222
Isobaric surface, 234 Local acoustic limiting process, 67
Isochoric case, 222223 Local atmospheric thermal problem, 154158
Isochoricity, 24 Local expansions, 236
Isochoric system, 4, 25 Local hydrostatic limiting process, 250
Local-in-time equations, 241
J Local-in-time limiting case, 191
Jump condition, 200 Local Reynolds and Rossby numbers, 136
Justification of Boussinesq model equations, 86 Local sound speed for perfect gas, 42, 45
Justification problem, 3 Local steady non-hydrostatic Euler
equations, 250
Local thermal problem, 136
K Local time, 93
Kaden problem, 129 Longitude, 56
Kapluns, 142 Long problem, 91, 220
Key steps, 6, 84, 138 Long time evolution, 184186
Kibel equations, 236237 Long-wave approximation, 55
Index 273
Long-wave approximation parameter, 218 Navier existence and smoothness

Long-wave dimensionless parameter, 205 problems, 123
Long-wave parameter, 55 NavierFourier-type equations, 204
Long-waves limiting process, 206 Navier incompressible equation, 27
Lower wall viscous region, 155 Navier main limiting process, 65
Low Kibel number, 61, 223251 Near-field close to acoustic source, 78
Low Mach (hyposonic), 89 Neumann boundary condition, 26
Low-Mach asymptotic, 39 Newtons cooling law of heat transfer, 43, 195
Low Mach numbers, 9 Non-conductive fluids, 40
Low Reynolds number, 60, 71 Non-degeneracy of the density, 58
Low Rossby model, 46 Non-dimensional form, 52
Low Rossby number, 60 Non-homogeneous (isochoric) fluid flow, 25
Low squared Froude number, 70 Non-hydrostatic dissipative equations, 231
Lubrication equation, 60, 208209 Non-linear, 91
Non-linear acoustics, 60, 76
Non-standard analysis, 145
M Non-tangent model equations, 231
Mach number, 19, 53, 83, 216, 225 Non-uniform validity, near t 0, 6263
Mach number low compressibility, Non-viscous equation, 108
effect, 236 Normal-mode decomposition, 132
Macroscale, 153 Normal velocity, 42
Main adjustment, 237 No-slip boundary condition, 27, 40, 47, 58
Main equation for lee-waves, 220 NSF equations, 1, 3435, 97
Main third key step, 86 NSF equations as a closed system, 120124
Mangler and Webers solution, 129
Marangoni convection, 55
Marangoni effect, 203209 O
Marangoni number, 43, 197, 210 OberbeckBoussinesq model equations, 7071
Matching, 73, 75, 93, 156, 157, 244246, 251 OberbeckBoussinesq shallow thermal
condition, 108, 114 convection equations, 5455
principle, 141 Open boundary, 42
process, 69 Oscillatory viscous flows, 57
relations, 105, 110, 242 Oseen limiting process, 72
Material motion, 52 Oseen pressure, 72
Mathematically rigorous results, 5859 Oseen region, 73
Mathematical topics in fluid dynamics, 49 Oseen time-space variables, 72
Mathematics, 118 Outer equation, 105
Mean curvature of the free surface, 200 Outer expansion, 165
Meteorological/oceanographic applications, 60 Outer-inner expansions with matching
Method of multiple scales (MMS), 145149 machinery, 240
Microscale structure, 151, 163 Outflow from middle deck, 156
Middle intermediate region, 154 Over all rapid oscillations, 175
MMAE, 39 Overlap, 141
MongeAmpere equation, 248251 Over-specification, 48
Motion of the wall, start impulsively from
rest, 172
Multiple boundary layers, 144 P
Multiple limiting process, 138 Parabolic approximation, 57, 77, 120
Multiple-scale method, 79, 133, 149, 171172 Partial second-order differential equation, 216
Peclet number, 54
Perfect gas, 24
N Periodicity, 163
Nabla, 21 Persistence of acoustic oscillations, 173174
Navier equation, 26 Perturbation of pressure, 70
274 Index
Plane approximation, 56 layer, 189, 190

Poisson equation, 26 model problem, 194
Porous media, 60 Reconstruction, 247
Possibility of matching, 141 Reduced BM thermocapillary model
Postulates, 29 problem, 207
Postulates of Gibbs, 32 Reduced system of equations, 206
Potential f, 44 Reference frequency, 61
Potential function, 37 Reference hydrostatic state, 83
Practical matching rule, 142 Regular coupling condition, 53, 106
PrandtlBatchelor condition, 47, 48 Reynolds number, 19, 53, 131, 189
PrandtlBatchelor region, 99 Right-handed curvilinear coordinates
PrandtlBatchelor unsteady equations, 104 system, 56
Prandtl boundary layer equations, 48, 64 Rigorous investigations, 119
Prandtl equation, 105 Rigorous mathematical results, 32, 39
Prandtl local limiting process, 64 Rigorous mathematical theory, 35
Prandtl number, 53, 197, 225 Robinsons Lemma, 145
Principal expansion, 236 Role of buoyancy, 193
Problem of Biot numbers, 196 Rolled-up vortex sheet, 127130
Problem of classical acoustics, 93 Rolled vortex sheets, 9, 127
Projection operators, 132 Rossby number, 56, 225
Pseudo-unsteady method, 42 Rotating flow, 60
p-system, 233237
S
Q Second first integral, 91
Quasi-geostrophic modelling, 230 Second-order Ageostrophic G-Z Model,
Quasi-geostrophic potential vorticity model 247248
equation, 239 Second-order approximation, 179182
Quasi-incompressible model, 173 Second-order consistent equations, 89
Quasi incompressible system, 67 Seepage flow in rigid porous media, 153
Quasi-non-divergent limit equation, 89 Short adjustment time, 240
Short times, 67, 107, 241
Short-time-scale, 139
R Signalling problem, 37
Radiative heat transfer parameter, 225 Significant degeneracies, 144145
Radius of the earth, 226 Significant degeneracy near the initial time, 241
RAM Approach, 7, 54, 59, 61, 64, 77 Similarity parameter, 54, 84
Rate-of-deformation tensor, 35 Similarity relations, 53, 73, 78, 235, 236
Rate of temperature fluctuation, 74 Similarity rules, 60, 138, 140, 200
Rational Asymptotic Modelling Approach Singularity, Prandtl boundary layer, 97
(RAMA), 9, 51, 118 Singular limit, 26
in fluid dynamics, 10 Sixth problem of the millennium, 123
process, 102 Slenderness parameter, 129
Ratio of specific heat capacities, 225 Slightly compressible flow, 140
Ratio of the above two length scales, 150 Slightly frictional, 59
Rayleigh analytical problem, 203 Slip boundary condition, 27, 40
Rayleigh Slip condition, 63, 65, 91, 105, 172
number, 54, 197 Small Mach number, 60, 140
problem, 99, 110 Small parameter, 128, 131, 138, 189
thermal instability problem, 43 Solid wall, 72
RayleighBenard (RB), 54, 193 Solution of the Ackerbloms problem, 124, 246
convection, 55, 61 Solvability, 35
instability problem, 203 Sommerfeld radiation condition, 92
Index 275
Source signal, 61 Thermally conducting fluid, 43

Source term, 169 Thermally non-homogeneous ground, 233
Spatial averaging theorem, 152 Thermally perfect gas, 23, 82
Specific entropy, 4, 24, 129 Thermal spot, 110
Specific heat, 23 Thermocapillary, 196
Specific internal energy, 70 Thermodynamic functions, 82
Specifying equation, 22 Thermodynamic perturbations, 85
Specifying relation, 57 Thermodynamic pressure, 33
Speed of sound, 82 Thickness d of the liquid layer, 194
Spherical coordinates, 56 Thin film Benard-Marangoni free-surface
Sphericity parameter, 56, 226 problem, 43
Stabilizing effect, 209 Thin film Marangoni problem, 44
Standard thermodynamic functions, 224 Three limiting cases, 84
Starting matrix equation, 168 Three regions structure, 237
State relations, 34 Through-flow model, 127, 161171
Steady 2D case, 90 Time derivative, 52
Steady Ekman layer, 243, 244 Trailing vortex sheets, 127
Steady lee-waves at infinity downstream Transient behaviour, 69
of the mountain, 91 Transient unsteady-state phase, 132
Steady lee-waves problem, 85 Transmission conditions, 168169
Steichen equation, 45 Transonic, 44
Stiff problems, 59, 117 Transonic similarity, 45
Stokesian fluid, 29 Triple-deck theory, 137, 154158
Stokes and Ossen model equations, 60 Trivariate fluid, 32
Stokes equation, 72 Trivariate liquid, 54
Stokes flow, 72 Tropopause, 216
Stokes idea of fluidity, 29 Troposphere, 217
Stokes-layer of thickness, 189, 190 Turbomachinery fluid flows, 124127
Stokes-like boundary layer, 174 Turbulence, 59, 123
Stokes limiting process, 71 Turbulent flow, 60
Stokes paradox, 72 Two hydrostatic limiting processes, 230
Stokes pressure, 71 Two stream functions, 4
Stokes relation, 30
Stratification, 8586
Strongly viscous, 57, 124 U
Strong singular self-induced coupling, Ultra-thin film, 55
157158 Unbalanced equation, 86, 87
Strouhal number, 52, 172, 224, 225 Unconditionally stable, 131
Subsonic flow, 42, 44 Unconfined perturbations, 134135
Supersonic flow, 42, 44 Underspecified boundary conditions, 48
Suspensions, 60 Unified theory, 135
Symmetric through-flow equations, 167 Unsteady adjustment, 38, 69, 139
Synoptic processes, 225, 229 equations, 107
inviscid problem, 111
T problem, 37
Temperature Unsteady dimensional equations, 108
condition, 58 Unsteady equations, 108
field on the wall, 58 Unsteady fluid flow problems, 139
parameter, 60 Unsteady Oseen equation, 73
Thcrmocapillary, 55 Unsteady waves on the water, 44
Thennocapillary destabilization, 209 Upper non-viscous region, 154
Thermal convection, 54, 71 Usual meteorological situation, 217
276 Index
V W
Vaisala internal frequency, 8586 Wakes, 163
Validity, NSF equations, 75 Water wave motion, 37
Vanishing effect of the blades in the row, 164 Wave number, 61
Velocity, 21, 25 Weak solution, 124
Vertical component of vorticity, 248 Weak form of the no-slip condition on the
Vertical deviation, 215217 moving wall, 47
Vertical deviation of streamline, 215216 Weakly expansible liquid, 70, 193, 210
Vertical displacement of the streamlines, 156 Weather prediction domain, 229230
Very thin layer case, 223 Weber number, 210
Veuillot approach, 161165 Well-posed, 35
Vicinity of z 0 on the wall, 6263 Well-posed initial boundary value
Viscous dissipation, 35, 202, 210 problem, 120, 233
Viscous effect, 236 Well-posedness, 120, 121
Viscous lower motion, 157
Vortex flows, 9
Vortex sheets, 127, 163 Z
Vorticity, 25, 214 Zeytounian deep-thermal convection, 55

Radyadour Kh. Zeytounian Auth. Navier-Stokes-Fourier Equations A Rational Asymptotic Modelling Point of View PDF

Uploaded by

Copyright:

Available Formats

Radyadour Kh. Zeytounian Auth. Navier-Stokes-Fourier Equations A Rational Asymptotic Modelling Point of View PDF

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

Radyadour Kh. Zeytounian Auth. Navier-Stokes-Fourier Equations A Rational Asymptotic Modelling Point of View PDF

Uploaded by

Copyright:

Available Formats

NavierStokesFourier Equations

ISBN 978-3-642-20745-7 e-ISBN 978-3-642-20746-4

# Springer-Verlag Berlin Heidelberg 2012

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

The last Chap. 9 is devoted to atmospheric motions. First, we derive for 2D

model is considered, and a global quasi-geostrophic (GQG) model is derived from

asymptotic, techniques could be used to advantage to simplify problem to the extent

This requirement being, in fact, a strong condition for a consistent application of

Let me close this Prologue by a remark:

Kaplun, S.: ZAMP. 5, 11136 (1954)

1 NSF Equations and Modelling: A French Touch . . . . . . . . . . . . . . 1

Part I NavierStokesFourier Equations: General Main Models

2 Newtonian Fluid Dynamics as a Mathematical Physical Science . . . 19

3 From NSF Equations to General Main Model Equations . . . . . . . 51

4 A Typical RAM Approach: Boussinesq Model Equations . . . . . . . . 81

Part II A Sketch of a Mathematical Theory of the RAM Approach

5 The Structure of Unsteady NSF Equations at Large

5.4.2 Acoustic and Rayleigh Problems Near

6 The Mathematics of the RAM Approach . . . . . . . . . . . . . . . . . . . . 117

Part III Applications of the RAM Approach to Aerodynamics, Thermal

7 The RAM Approach in Aerodynamics . . . . . . . . . . . . . . . . . . . . . . 161

7.2.1 Formulation of the Inviscid Problem . . . . . . . . . . . . . . . . 172

8 The RAM Approach in the Benard Convection Problem . . . . . . . . 193

9 The RAM Approach in Atmospheric Motions . . . . . . . . . . . . . . . . . 213

There are two mathematicalphysical descriptions of fluid dynamics.1 The first of

@f =@t x:rf 1=KnQf ; f ;

where f t; x; x is precisely the density probability of finding a molecule at the

where M U =c is the constant Mach (dimensionless) number, based on the

R. Kh. Zeytounian, NavierStokesFourier Equations, 19

where c 1; x1 ; x2 ; x3 ; jxj2 is a five-component (the so-called collisional

2.1 From Newton to Euler

r Du=Dt rg internal force per unit volume: (2.1)

A second part of Newtons Principia is related to the conservation of mass: To

D=Dt @=@t u:r;

are @=@xi ; i 1; 2; 3, where the time is denoted by t, and x x1 ; x2 ; x3 is the

Dr=Dt rr:u 0: (2.3)

This (compressible) equation of continuity (2.3) remains unaltered when viscos-

2.1.1 Eulerian Elastic Fluid

In reality, fluid dynamics was first envisaged as a systematic mathematical

Bernoullis Hydrodynamica (1738) [35], and also in DAlembert Traite de

rDu=Dt rg rp: (2.4)

2.1.2 From Adiabaticity to Isochoricity

Cv DT=Dt pD=Dt1=r 0; (2.6)

S R=g 1logp=rg const; (2.9)

D=Dt logp=rg 0 ) DS=Dt 0: (2.10)

g tends to infinity incompressible limit case; (2.11a)

such that R O(1), and that in such a case,

Cv tends to zero but Cp  R O1; (2.11b)

As a consequence, for a Eulerian incompressible but non-homogeneous

r Du=Dt rg rp; (2.12a)

This isochoric system of three Eqs. 2.12a2.12c is very well investigated in

2.2 Navier Viscous Incompressible, Constant Density Equations

The equation of motion of a viscous and incompressible homogeneous (with a

DuN =Dt 1=r rp g n DuN ; (2.13a)

where n is the constant kinematic viscosity, and D  r2 is the Laplace operator

Because n multiplies the derivative of highest order in Navier equation (2.13a),

DuE =Dt 1=r rp g O: (2.14a)

Dp r fr : uE : ruE  r:gg;

S R=g 1logp=rg const; (2.9)

Cv tends to zero but Cp R O1; (2.11b)

where n is the constant kinematic viscosity, and D r2 is the Laplace operator

Dp r fr : uE : ruE r:gg;

@p=@n r fr : uE : ruE g : n;

r Dui =Dt @=@xj mr@ui =@xj @uj =@xi

rCv DT=Dt p r u r k rT F; (2.34c)

F 2mTrace Du2 lr u2 ; (2.35b)

Trace Du2 Du : Du 1=4@ui =@xj @uj =@xi 2 ; (2.36)

u n U > 0 and n D t 0; (2.44a)