Numerical Approximation of Hyperbolic

Systems of Conservation Laws, 2nd

edition (Applied Mathematical Sciences
series) Edwige Godlewski
Visit to download the full and correct content document:
https://ebookmeta.com/product/numerical-approximation-of-hyperbolic-systems-of-co
nservation-laws-2nd-edition-applied-mathematical-sciences-series-edwige-godlewski/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...

Analysis of Legal Argumentation Documents A

Computational Argumentation Approach Translational
Systems Sciences 29 Hayato Hirata

https://ebookmeta.com/product/analysis-of-legal-argumentation-
documents-a-computational-argumentation-approach-translational-
systems-sciences-29-hayato-hirata/

Diophantine Approximation and the Geometry of Limit

Sets in Gromov Hyperbolic Metric Spaces 1st Edition
Lior Fishman

https://ebookmeta.com/product/diophantine-approximation-and-the-
geometry-of-limit-sets-in-gromov-hyperbolic-metric-spaces-1st-
edition-lior-fishman/

Elliptic Regularity Theory by Approximation Methods

London Mathematical Society Lecture Note Series Edgard
A. Pimentel

https://ebookmeta.com/product/elliptic-regularity-theory-by-
approximation-methods-london-mathematical-society-lecture-note-
series-edgard-a-pimentel/

Mathematical Modeling the Life Sciences: Numerical

Recipes in Python and MATLAB 1st Edition N. G. Cogan

https://ebookmeta.com/product/mathematical-modeling-the-life-
sciences-numerical-recipes-in-python-and-matlab-1st-edition-n-g-
cogan/
Stochastic Approximation A Dynamical Systems Viewpoint
2nd Edition Vivek S Borkar

https://ebookmeta.com/product/stochastic-approximation-a-
dynamical-systems-viewpoint-2nd-edition-vivek-s-borkar/

B Series Algebraic Analysis of Numerical Methods John C

Butcher

https://ebookmeta.com/product/b-series-algebraic-analysis-of-
numerical-methods-john-c-butcher/

Weighted Polynomial Approximation and Numerical Methods

for Integral Equations 1st Edition Peter Junghanns

https://ebookmeta.com/product/weighted-polynomial-approximation-
and-numerical-methods-for-integral-equations-1st-edition-peter-
junghanns/

Lattice Rules Numerical Integration Approximation and

Discrepancy Josef Dick Peter Kritzer Friedrich
Pillichshammer

https://ebookmeta.com/product/lattice-rules-numerical-
integration-approximation-and-discrepancy-josef-dick-peter-
kritzer-friedrich-pillichshammer/

Conservation of Architectural Heritage 2nd Edition

Antonella Versaci

https://ebookmeta.com/product/conservation-of-architectural-
heritage-2nd-edition-antonella-versaci/
Applied Mathematical Sciences

Edwige Godlewski
Pierre-Arnaud Raviart

Numerical
Approximation
of Hyperbolic Systems
of Conservation Laws
Second Edition
Applied Mathematical Sciences

Volume 118

Series Editors
Anthony Bloch, Department of Mathematics, University of Michigan, Ann Arbor,
MI, USA
abloch@umich.edu
C. L. Epstein, Department of Mathematics, University of Pennsylvania,
Philadelphia, PA, USA
cle@math.upenn.edu
Alain Goriely, Department of Mathematics, University of Oxford, Oxford, UK
goriely@maths.ox.ac.uk
Leslie Greengard, New York University, New York, NY, USA
Greengard@cims.nyu.edu

Advisory Editors
J. Bell, Center for Computational Sciences and Engineering, Lawrence Berkeley
National Laboratory, Berkeley, CA, USA
P. Constantin, Department of Mathematics, Princeton University, Princeton, NJ,
USA
R. Durrett, Department of Mathematics, Duke University, Durham, CA, USA
R. Kohn, Courant Institute of Mathematical Sciences, New York University,
New York, NY, USA
R. Pego, Department of Mathematical Sciences, Carnegie Mellon University,
Pittsburgh, PA, USA
L. Ryzhik, Department of Mathematics, Stanford University, Stanford, CA, USA
A. Singer, Department of Mathematics, Princeton University, Princeton, NJ, USA
A. Stevens, Department of Applied Mathematics, University of Münster, Münster,
Germany
S. Wright, Computer Sciences Department, University of Wisconsin, Madison, WI,
USA

Founding Editors
F. John, New York University, New York, NY, USA
J. P. LaSalle, Brown University, Providence, RI, USA
L. Sirovich, Brown University, Providence, RI, USA
The mathematization of all sciences, the fading of traditional scientific boundaries,
the impact of computer technology, the growing importance of computer modeling
and the necessity of scientific planning all create the need both in education and
research for books that are introductory to and abreast of these developments. The
purpose of this series is to provide such books, suitable for the user of mathematics,
the mathematician interested in applications, and the student scientist. In particular,
this series will provide an outlet for topics of immediate interest because of the nov-
elty of its treatment of an application or of mathematics being applied or lying close
to applications. These books should be accessible to readers versed in mathematics
or science and engineering, and will feature a lively tutorial style, a focus on topics
of current interest, and present clear exposition of broad appeal. A compliment to
the Applied Mathematical Sciences series is the Texts in Applied Mathematics se-
ries, which publishes textbooks suitable for advanced undergraduate and beginning
graduate courses.

More information about this series at http://www.springer.com/series/34

Edwige Godlewski • Pierre-Arnaud Raviart

Numerical Approximation
of Hyperbolic Systems of
Conservation Laws
Second Edition
Edwige Godlewski Pierre-Arnaud Raviart
Laboratoire Jacques-Louis Lions Laboratoire Jacques-Louis Lions
Sorbonne University Sorbonne University
Paris, France Paris, France

ISSN 0066-5452 ISSN 2196-968X (electronic)

Applied Mathematical Sciences
ISBN 978-1-0716-1342-9 ISBN 978-1-0716-1344-3 (eBook)
https://doi.org/10.1007/978-1-0716-1344-3

Mathematics Subject Classification: 35L65, 35L67, 65M06, 65M08, 65M12, 76Nxx, 35L50, 35L60,
35Q35, 65Mxx, 35Q20, 35Q86, 76P05, 76W05, 80A32

© Springer Science+Business Media, LLC, part of Springer Nature 1996, 2021

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, 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.
The publisher, the authors, and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Science+Business Media, LLC,
part of Springer Nature
The registered company address is: 1 New York Plaza, New York, NY 10004, U.S.A.
Preface to the Second Edition

There was an obvious need to complete the first edition of this textbook
with the treatment of source terms. Thus, a new chapter (Chap. VII) has
been added, which also provides a few important principles concerning non-
conservative systems that are naturally introduced with the derivation of
well-balanced or asymptotic preserving schemes. Note that most theoretical
results are only referred to since it is out of scope to give detailed proofs;
these may be tricky and are often quite technical.
We took the opportunity of this second edition to include more examples
in the introduction chapter (now Chap. I), such as MHD, shallow water, and
flow in a nozzle, and to give some insights on multiphase flow models; this last
subject deserves a much longer treatment. Then we thought it is important
to emphasize the change of frame from Eulerian to Lagrangian coordinates
and the specificity of fluid systems. Additionally, the low Mach limit has been
addressed in the chapter devoted to multidimensional systems (now Chap. V)
with the final section introducing all Mach schemes.
For 25 years, there has been a tremendous lot of work dedicated to the
numerical approximation of hyperbolic systems, among which we choose to
introduce the relaxation approach, now at the end of Chap. IV and the case
of discontinuous fluxes, and interface coupling, a topic covered in Chap. VII.
Both subjects are treated in some specific outlines.
Then, some complements may be found here and there, such as recalling
some results of our earlier publication at the beginning of Chap. IV, or more
examples of systems of two equations in Chap. II.
We must finally confess that it took us some time to complete the work
of this second edition, for different reasons. In fact, most of this work was
achieved several years ago, which may explain why only few very recent results
are presented, some of them are just mentioned in the notes at the end of
each chapter, to give a hint and provide references where the subject is more
thoroughly treated.

v
Preface to the First Edition

This work is devoted to the theory and approximation of nonlinear hyper-

bolic systems of conservation laws in one or two space variables. It follows
directly a previous publication on hyperbolic systems of conservation laws by
the same authors, and we shall make frequent references to Godlewski and
Raviart (1991) (hereafter noted G.R.), though the present volume can be read
independently. This earlier publication, apart from a first chapter, especially
covered the scalar case. Thus, we shall detail here neither the mathematical
theory of multidimensional scalar conservation laws nor their approximation
in the one-dimensional case by finite-difference conservative schemes, both of
which were treated in G.R., but we shall mostly consider systems. The the-
ory for systems is in fact much more difficult and not at all completed. This
explains why we shall mainly concentrate on some theoretical aspects that
are needed in the applications, such as the solution of the Riemann problem,
with occasional insights into more sophisticated problems.
The present book is divided into six chapters, including an introductory
chapter1 . For the reader’s convenience, we shall resume in this Introduction
the notions that are necessary for a self-sufficient understanding of this book
–the main definitions of hyperbolicity, weak solutions, and entropy– present
the practical examples that will be thoroughly developed in the following
chapters, and recall the main results concerning the scalar case.
Chapter I is devoted to the resolution of the Riemann problem for a general
hyperbolic system in one space dimension, introducing the classical notions
of Riemann invariants and simple waves, the rarefaction and shock curves,
and characteristics and entropy conditions. The theory is then applied to the
p-system.
In Chap. II, we make a closer study of the one-dimensional system of gas
dynamics. We solve the Riemann problem in detail and then present the

1 The numbering of the chapters has changed in the second edition, the Introduction is
now Chap. I. Hence in what follows, Chap. I refers to what is now Chap. II and so on.
vii
viii Preface to the First Edition

simplest models of reacting flow, first the Chapman-Jouguet theory and then
the Z.N.D. model for detonation.
After this theoretical approach, we go into the numerical approximation of
hyperbolic systems by conservative finite-difference methods. The most usual
schemes for one-dimensional systems are developed in Chap. III, with special
emphasis on the application to gas dynamics. The last section begins with a
short account on the kinetic theory so as to introduce kinetic schemes.
Chapter IV is devoted to the study of finite volume methods for bidimen-
sional systems, preceded by some theoretical considerations on multidimen-
sional systems.
For the sake of completeness, we could not avoid the problem of boundary
conditions. Chapter V is but an introduction to the complex theory and
presents some numerical boundary treatment.
The authors wish to thank R. Abgrall, F. Coquel, F. Dubois, and particu-
larly T. Gallouet, B. Perthame, and D. Serre, from whom they learned a great
deal and who answered willingly and most amiably their many questions.
They owe thanks to the SMAI reading committee and to the reviewers,
who made very valuable suggestions.
The first author is grateful to all her colleagues who encouraged her in
completing this huge work, especially to H. Le Dret and F. Murat for so
often giving her their time, and to L. Ruprecht for her kind and competent
assistance in the retyping of the final manuscript; such friendly help was
invaluable.

Paris, France E. Godlewski and P.-A. Raviart

September 1995
Contents

I Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 Definitions and Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Fluid Systems in Eulerian and Lagrangian Frames . . . . . . . . . . 6
3 Some Averaged Models: Shallow Water, Flow in a Duct, and
Two-Phase Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
4 Weak Solutions of Systems of Conservation Laws . . . . . . . . . . . 27
4.1 Characteristics in the Scalar One-Dimensional Case . . . 27
4.2 Weak Solutions: The Rankine-Hugoniot Condition . . . . 30
4.3 Example of Nonuniqueness of Weak Solutions . . . . . . . . 35
5 Entropy Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.1 A Mathematical Notion of Entropy . . . . . . . . . . . . . . . . . 37
5.2 The Vanishing Viscosity Method . . . . . . . . . . . . . . . . . . . 44
5.3 Existence and Uniqueness of the Entropy Solution in
the Scalar Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52

II Nonlinear Hyperbolic Systems in One Space Dimension . . 55

1 Linear Hyperbolic Systems with Constant Coefficients . . . . . . . 55
2 The Nonlinear Case, Definitions and Examples . . . . . . . . . . . . . 58
2.1 Change of Variables, Change of Frame . . . . . . . . . . . . . . 60
2.2 The Gas Dynamics Equations . . . . . . . . . . . . . . . . . . . . . . 66
2.3 Ideal MHD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3 Simple Waves and Riemann Invariants . . . . . . . . . . . . . . . . . . . . 80
3.1 Rarefaction Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
3.2 Riemann Invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
4 Shock Waves and Contact Discontinuities . . . . . . . . . . . . . . . . . . 92
5 Characteristic Curves and Entropy Conditions . . . . . . . . . . . . . 103
5.1 Characteristic Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.2 The Lax Entropy Conditions . . . . . . . . . . . . . . . . . . . . . . . 107
5.3 Other Entropy Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 110

ix
x Contents

6 Solution of the Riemann Problem . . . . . . . . . . . . . . . . . . . . . . . . . 116

7 Examples of Systems of Two Equations . . . . . . . . . . . . . . . . . . . . 120
7.1 The Case of a Linear or a Linearly Degenerate
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2 The Riemann Problem for the p-System . . . . . . . . . . . . . 122
7.3 The Riemann Problem for the Barotropic
Euler System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137

III Gas Dynamics and Reacting Flows . . . . . . . . . . . . . . . . . . . . . . . 141

1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
1.1 Properties of the Physical Entropy . . . . . . . . . . . . . . . . . . 141
1.2 Ideal Gases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
2 Entropy Satisfying Shock Conditions . . . . . . . . . . . . . . . . . . . . . . 153
3 Solution of the Riemann Problem . . . . . . . . . . . . . . . . . . . . . . . . . 171
4 Reacting Flows: The Chapman-Jouguet Theory . . . . . . . . . . . . 188
5 Reacting Flows: The Z.N.D. Model for Detonations . . . . . . . . . 207
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

IV Finite Volume Schemes for One-Dimensional Systems . . . . . 215

1 Generalities on Finite Volume Methods for Systems . . . . . . . . . 215
1.1 Extension of Scalar Schemes to Systems: Some
Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
1.2 L2 Stability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230
1.3 Dissipation and Dispersion . . . . . . . . . . . . . . . . . . . . . . . . 232
2 Godunov’s Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236
2.1 Godunov’s Method for Systems . . . . . . . . . . . . . . . . . . . . 236
2.2 The Gas Dynamics Equations in a Moving Frame . . . . 240
2.3 Godunov’s Method in Lagrangian Coordinates . . . . . . . 242
2.4 Godunov’s Method in Eulerian Coordinates
(Direct Method) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 245
2.5 Godunov’s Method in Eulerian Coordinates
(Lagrangian Step + Projection) . . . . . . . . . . . . . . . . . . . . 246
2.6 Godunov’s Method in a Moving Grid . . . . . . . . . . . . . . . 249
3 Godunov-Type Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
3.1 Approximate Riemann Solvers and Godunov-Type
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
3.2 Roe’s Method and Variants . . . . . . . . . . . . . . . . . . . . . . . . 259
3.3 The H.L.L. Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 269
3.4 Osher’s Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
4 Roe-Type Methods for the Gas Dynamics System . . . . . . . . . . 283
4.1 Roe’s Method for the Gas Dynamics Equations: (I)
The Ideal Gas Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
4.2 Roe’s Method for the Gas Dynamics Equations: (II)
The “Real Gas” Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
Contents xi

4.3 A Roe-Type Linearization Based on Shock Curve

Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
4.4 Another Roe-Type Linearization Associated with a
Path . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 303
4.5 The Case of the Gas Dynamics System in Lagrangian
Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
5 Flux Vector Splitting Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
5.1 General Formulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320
5.2 Application to the Gas Dynamics Equations: (I)
Steger and Warming’s Approach . . . . . . . . . . . . . . . . . . . 322
5.3 Application to the Gas Dynamics Equations: (II) Van
Leer’s Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
6 Van Leer’s Second-Order Method . . . . . . . . . . . . . . . . . . . . . . . . . 329
6.1 Van Leer’s Method for Systems . . . . . . . . . . . . . . . . . . . . 329
6.2 Solution of the Generalized Riemann Problem . . . . . . . . 333
6.3 The G.R.P. for the Gas Dynamics Equations in
Lagrangian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
6.4 Use of the G.R.P. in van Leer’s Method . . . . . . . . . . . . . 345
7 Kinetic Schemes for the Euler Equations . . . . . . . . . . . . . . . . . . 354
7.1 The Boltzmann Equation . . . . . . . . . . . . . . . . . . . . . . . . . . 354
7.2 The B.G.K. Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
7.3 The Kinetic Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368
7.4 Some Extensions of the Kinetic Approach . . . . . . . . . . . 388
8 Relaxation Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
8.1 Introduction to Relaxation . . . . . . . . . . . . . . . . . . . . . . . . 394
8.2 Model Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
8.3 A Relaxation Scheme for the Euler System . . . . . . . . . . 407
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 420

V The Case of Multidimensional Systems . . . . . . . . . . . . . . . . . . . 425

1 Generalities on Multidimensional Hyperbolic Systems . . . . . . . 425
1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 425
1.2 Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 428
1.3 Simple Plane Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433
1.4 Shock Waves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 437
2 The Gas Dynamics Equations in Two Space Dimensions . . . . . 439
2.1 Entropy and Entropy Variables . . . . . . . . . . . . . . . . . . . . 440
2.2 Invariance of the Euler Equations . . . . . . . . . . . . . . . . . . 443
2.3 Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 450
2.4 Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
2.5 Plane Wave Solutions: Self-Similar Solutions . . . . . . . . . 460
3 Multidimensional Finite Difference Schemes . . . . . . . . . . . . . . . . 468
3.1 Direct Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 468
3.2 Dimensional Splitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 480
4 Finite-Volume Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 487
4.1 Definition of the Finite-Volume Method . . . . . . . . . . . . . 488
xii Contents

4.2 General Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499

4.3 Usual Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 517
5 Second-Order Finite-Volume Schemes . . . . . . . . . . . . . . . . . . . . . 533
5.1 MUSCL-Type Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 533
5.2 Other Approaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 546
6 An Introduction to All-Mach Schemes for the System of Gas
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 547
6.1 The Low Mach Limit of the System of Gas
Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 548
6.2 Asymptotic Analysis of the Semi-Discrete
Roe Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 552
6.3 An All-Mach Semi-Discrete Roe Scheme . . . . . . . . . . . . . 561
6.4 Asymptotic Analysis of the Semi-Discrete
HLL Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 568
6.5 An All-Mach Semi-Discrete HLL Scheme . . . . . . . . . . . . 574
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 578

VI An Introduction to Boundary Conditions . . . . . . . . . . . . . . . . . 581

1 The Initial Boundary Value Problem in the
Linear Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 581
1.1 Scalar Advection Equations . . . . . . . . . . . . . . . . . . . . . . . . 582
1.2 One-Dimensional Linear Systems. Linearization . . . . . . 587
1.3 Multidimensional Linear Systems . . . . . . . . . . . . . . . . . . . 590
2 The Nonlinear Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599
2.1 Nonlinear Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 599
2.2 Nonlinear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 602
3 Gas Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 606
3.1 Fluid Boundary (Linearized Approach) . . . . . . . . . . . . . . 607
3.2 Solid or Rigid Wall Boundary . . . . . . . . . . . . . . . . . . . . . . 610
4 Absorbing Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 610
5 Numerical Treatment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 618
5.1 Finite Difference Schemes . . . . . . . . . . . . . . . . . . . . . . . . . 618
5.2 Finite Volume Approach . . . . . . . . . . . . . . . . . . . . . . . . . . 621
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 625

VII Source Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627

1 Introduction to Source Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 627
1.1 Some General Considerations for Systems with
Source Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 628
1.2 Simple Examples of Source Terms in the Scalar Case . . 629
1.3 Numerical Treatment of Source Terms . . . . . . . . . . . . . . 632
1.4 Examples of Systems with Source Terms . . . . . . . . . . . . 639
2 Systems with Geometric Source Terms . . . . . . . . . . . . . . . . . . . . 643
2.1 Nonconservative Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 644
2.2 Stationary Waves and Resonance . . . . . . . . . . . . . . . . . . . 650
Contents xiii

2.3 Case of a Nozzle with Discontinuous Section . . . . . . . . . 656

2.4 The Example of the Shallow Water System . . . . . . . . . . 662
3 Specific Numerical Treatment of Source Terms . . . . . . . . . . . . . 665
3.1 Some Numerical Considerations for Flow in a Nozzle . . 665
3.2 Preserving Equilibria, Well-Balanced Schemes . . . . . . . . 667
3.3 Schemes for the Shallow Water System . . . . . . . . . . . . . . 675
4 Simple Approximate Riemann Solvers . . . . . . . . . . . . . . . . . . . . . 679
4.1 Definition of Simple Approximate Riemann Solvers . . . 679
4.2 Well-Balanced Simple Schemes . . . . . . . . . . . . . . . . . . . . . 682
4.3 Simple Approximate Riemann Solvers in Lagrangian
or Eulerian Coordinates . . . . . . . . . . . . . . . . . . . . . . . . . . . 685
4.4 The Example of the Gas Dynamics Equations with
Gravity and Friction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 688
4.5 Link with Relaxation Schemes . . . . . . . . . . . . . . . . . . . . . 697
5 Stiff Source Terms, Asymptotic Preserving Numerical
Schemes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 705
5.2 Some Simple Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 707
5.3 Derivation of an AP Scheme for the Linear Model . . . . 711
5.4 Euler System with Gravity and Friction . . . . . . . . . . . . . 721
6 Interface Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 731
6.1 Introduction to Interface Coupling . . . . . . . . . . . . . . . . . . 731
6.2 The Interface Coupling Condition . . . . . . . . . . . . . . . . . . 734
6.3 Numerical Coupling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 744
Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 746

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 749

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 831
I
Introduction

1 Definitions and Examples

In this section, we present the general form of systems of conservation laws in

several space variables, and we give some important examples of such systems
that arise in continuum physics.
Let Ω be an open subset of Rp , and let fj , 1 ≤ j ≤ d, be d smooth functions
from Ω into Rp ; the general form of a system of conservation laws in several
space variables is

∂u  ∂
d
+ fj (u) = 0, x = (x1 , . . . , xd ) ∈ Rd , t > 0, (1.1)
∂t j=1
∂xj

where ⎛ ⎞
u1
⎜ ⎟
u = ⎝ ... ⎠
up
is a vector-valued function from Rd × [0, +∞[ into Ω. The set Ω is called the
set of states and the functions
⎛ ⎞
f1j
⎜ .. ⎟
fj = ⎝ . ⎠
fpj

are called the flux functions. One says that system (1.1) is written in
conservative form.
In the sequel, we will also write the system (1.1) in the form

∂u
+ ∇ · f (u) = 0
∂t
where f stands for the matrix-valued function

© Springer Science+Business Media, LLC, part of Springer Nature 2021 1

E. Godlewski, P.-A. Raviart, Numerical Approximation of Hyperbolic
Systems of Conservation Laws, Applied Mathematical Sciences 118,
https://doi.org/10.1007/978-1-0716-1344-3_1
2 I. Introduction

f = (fij )1≤i≤p,1≤j≤d
and ∇· is the divergence operator (we will equivalently use the notation div)

d
∂
∇·f = fj .
j=1
∂x j

Formally, the system (1.1) expresses the conservation of the p quan-

tities u1 , . . . , up . In fact, let D be an arbitrary domain of Rd , and let
n = (n1 , . . . , nd )T be the outward unit normal to the boundary ∂D of D.
Then, it follows from (1.1) that

d 
d
u dx + fj (u) nj dS = 0.
dt D j=1 ∂D

This balance equation has now a very natural meaning: the time variation of
D
u dx is equal to the losses through the boundary ∂D.
In all the following, we shall be concerned with the study of hyperbolic
systems of conservation laws, which we define in the following way. For all
j = 1, . . . , d, let
∂fij
Aj (u) = (u)
∂uk 1≤i,k≤p

be the Jacobian matrix of fj (u); the system (1.1) is called hyperbolic if, for
any u ∈ Ω and any ω = (ω1 , . . . , ωd ) ∈ Rd , ω = 0, the matrix


d
A(u, ω) = ωj Aj (u)
j=1

has p real eigenvalues λ1 (u, ω) ≤ λ2 (u, ω) ≤ . . . ≤ λp (u, ω) and p linearly

independent corresponding eigenvectors r1 (u, ω), . . . , rp (u, ω), i.e.,

A(u, ω) rk (u, ω) = λk (u, ω) rk (u, ω), 1 ≤ k ≤ p.

If, in addition, the eigenvalues λk (u, ω) are all distinct, the system (1.1) is
called strictly hyperbolic.
In fact, little is known about systems in more than one space variable unless
they are symmetrizable, i.e., there exists for all u ∈ Ω a symmetric positive-
definite matrix A0 (u) smoothly varying with u such that the matrices

A0 (u)Aj (u), 1≤j≤d

are symmetric. Symmetrizable systems of conservation laws are clearly hy-

perbolic. Note that most of the systems of conservation laws that arise in
practice are symmetrizable; this is a consequence of the existence of an en-
tropy function (Godunov-Mock theorem [543], see Theorem 5.1 below).
1 Definitions and Examples 3

For such systems, we shall study the Cauchy problem, or initial value prob-
lem (IVP): find a function u : (x, t) ∈ Rd × [0, ∞[→ u(x, t) ∈ Ω that is a
solution of (1.1) satisfying the initial condition

u(x, 0) = u0 (x), x ∈ Rd , (1.2)

where u0 : Rd → Ω is a given function. The initial boundary value problem

(I.B.V.P.) will be considered in Chap. VI. One aim of this introduction is to
make precise in which sense (1.1), (1.2) is to be taken.
In the one-dimensional case, when u0 has the following particular form,

u , x<0
u0 (x) = (1.3)
ur , x > 0,

with constant states u , ur , this Cauchy problem is called the (one-dimensional)

Riemann problem (see Chap. V, Remark 2.8, for the definition of a 2-D Rie-
mann problem).
In the scalar case (i.e., p = 1), the simplest example of a nonlinear conser-
vation law is given by Burgers’ equation.
Example 1.1. The Burgers’ equation. The scalar parabolic equation

∂u ∂u ∂2u
+u −ν =0 (1.4a)
∂t ∂x ∂x2
was introduced in particular by Burgers as the simplest differential model
for a fluid flow and is therefore often called the (viscous) Burgers’ equation.
Though very simple, this equation can be regarded as a model for decaying
free turbulence (see Cole [325]). A number of authors have developed the
asymptotic theory of Navier-Stokes equations in terms of Burgers’ equation,
and it is thus often used in numerical tests [1059]. Burgers studied the limit
equation when ν tends to zero, which we write in conservation form
 
∂u ∂ u2
+ = 0. (1.4b)
∂t ∂x 2
Equation (1.4b) is the inviscid Burgers’ equation (or Burgers’ equation with-
out viscosity), which, for brevity, we shall simply call from now on Burg-
ers’ equation. It occurs in particular in wave theory to depict the distortion
of waveform in simple waves (see Lighthill [800, Sec. 2.9], Whitham [1188,
Sec. 2.8]).
We shall see that Burgers’ equation possesses all the features of a scalar
convex equation
∂u ∂
+ f (u) = 0, (1.5)
∂t ∂x
where f : R → R is a convex smooth function. In particular, the Cauchy
problem for Burgers’ equation may have discontinuous weak solutions even
for a smooth initial function u0 , and the solution of the Riemann problem
4 I. Introduction

is either a shock propagating or a rarefaction wave (see Fig. 4.2), both being
kinds of waves that are involved in the solution of the Riemann problem for
a system.
Remark 1.1. It will be very easy to derive the properties of Eq. (1.5) when the
flux f is concave, from the convex case. An example of a concave flux is illus-
trated by the classical LWR traffic model (for Lighthill-Whitham-Richards)
for which
ρ
f (ρ) = umax ρ(1 − ),
ρmax
where ρ ∈ [0, ρmax ] measures the density of cars and umax is a maximum
velocity. 
Finally, it is worth mentioning that the Cauchy problem for (1.4a) has an
explicit solution, obtained using the Cole-Hopf transform
ϕx
u = −2ν .
ϕ
It eliminates the nonlinear term and transforms (1.4a) into the heat equation

∂ϕ ∂2ϕ
= ν 2,
∂t ∂x
for which explicit expressions of the solution are known. For details, we refer
to the original papers of Hopf [630], Cole [325], and Whitham (1974, Chap-
ter 4) [1188] where a thorough study of Eq. (1.4a) (including limit as ν → 0
and shock structure) can be found. 

Example 1.2. The Buckley-Leverett equation. In petroleum engineering, the

Buckley-Leverett equation is a one-dimensional model for a two-phase flow,
an oil and water mixture, in a porous medium. We assume that the porosity
is constant (taken to be 1) and we ignore capillarity and gravity effects. Then
the reduced water saturation s satisfies and Eq. (1.5) with flux

kr,w (s)/μw
f (s) = , (1.6)
kr,w (s)/μw + kr,o (s)/μo

where μ denotes the viscosity (supposed to be constant) and kr (s) the relative
permeability of a fluid (the indices w and o refer to water and oil, respec-
tively); the set of states is the interval [0, 1]. A typical example is obtained
by taking kr,w = s2 , kr,o (s) = (1 − s)2 , so that f writes

s2
f (s) = ,
s2 + (1 − s)2 μw /μo

and it is an increasing, S-shaped (nonconvex) function (see [1147] for a pos-

sible extension of this model and references therein). 
1 Definitions and Examples 5

Let us next introduce some fundamental examples of the systems of con-

servation laws that we shall study in the following chapters.
Example 1.3. The p-system. A model for one-dimensional isentropic gas
dynamics in Lagrangian coordinates is given by the following system of two
equations: ⎧
⎪ ∂v ∂u
⎨ − =0
∂t ∂x (1.7)
⎪
⎩ ∂u + ∂ p(v) = 0,
∂t ∂x
where v is the specific volume, u is the velocity, and the pressure p = p(v) is
a given function of v. For a polytropic isentropic ideal gas, we have

p(v) = Av −γ

for some constants A = A(s) > 0 (depending on the constant entropy) and
γ ≥ 1.
System (1.7) is of the form

∂w ∂
+ f (w) = 0,
∂t ∂x
where
   
v −u
w= , f (w) = , Ω = {(v, u) ∈ R2 ; v > 0}.
u p(v)

The Jacobian matrix of f is easily computed

 
0 −1
A(w) = ,
p (v) 0

and the system (1.7) is strictly hyperbolic provided that we assume p (v) < 0.
In that case, the matrix A(w) (which depends only on v) has indeed two real
distinct eigenvalues
 
λ1 = − (−p (v)) < λ2 = (−p (v)).

Besides the scalar case, the p-system (1.7) is the simplest nontrivial ex-
ample of a nonlinear system of conservation laws. Note that any nonlinear
wave equation  
∂2g ∂ ∂g
− σ =0
∂t2 ∂x ∂x
can be put in the form (1.7) by setting

∂g ∂g
u= , v= , p(v) = −σ(v),
∂t ∂x
as one can easily check. 
6 I. Introduction

2 Fluid Systems in Eulerian and Lagrangian Frames

System (1.7) is the simplest model for a fluid system in Lagrangian coordi-
nates. We now consider the full system of gas dynamics, an example that
appears to be fundamental in the applications, without assuming the flow is
isentropic, and in the general multidimensional case.
Example 2.1. The equations of gas dynamics in Eulerian coordinates. In Eule-
rian coordinates, the Euler equations for a compressible inviscid fluid (where
we neglect heat conduction) can be written in the following conservative form:
⎧
⎪ ∂ρ  3 ∂
⎪
⎪ + (ρuj ) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪ j=1 j
⎪
⎨ ∂  3 ∂
(ρui ) + (ρui uj + p δij ) = 0, 1 ≤ i ≤ 3, (2.1)
⎪
⎪ ∂t j=1 ∂xj
⎪
⎪
⎪
⎪ 
⎪
⎪ ∂ 3 ∂
⎩ (ρe) + ((ρe + p)uj ) = 0.
∂t j=1 ∂x j

In (2.1), ρ is the density of the fluid, u = (u1 , u2 , u3 )T the velocity, p the

pressure, ε the specific (i.e., per unit mass) internal energy, and e = ε + |u|2
2

the specific total energy. Equations (2.1) express respectively the laws of
conservation of mass, momentum, and total energy for the fluid.
Note that (2.1) may be equivalently written
⎧
⎪ ∂ρ
⎪
⎪ + ∇ · (ρu) = 0,
⎪
⎪ ∂t
⎪
⎪
⎨
∂
(ρu) + ∇ · (ρu ⊗ u + pI) = 0,
⎪ ∂t
⎪
⎪
⎪
⎪
⎪
⎪
⎩ ∂ (ρe) + ∇ · ((ρe + p)u) = 0
∂t
where I is the d × d identity matrix and u ⊗ u denotes the matrix-valued
function (u ⊗ u)ij = ui uj . More generally, for v ∈ Rp , w ∈ Rq , we define the
p × q matrix v ⊗ w by

(v ⊗ w)ij = vi wj , 1 ≤ i ≤ p, 1 ≤ j ≤ q.

Note also that

∇ · (pI) = ∇p
where ∇ is the gradient operator.
We need to add to (2.1) an equation for p. We shall see in Chap. V that
Galilean invariance requires that p does not depend on u. The “equation of
state” can be taken in the form
2 Fluid Systems in Eulerian and Lagrangian Frames 7

p = p(ρ, ε),

e.g., for a polytropic ideal gas, the equation of state is given by

p = (γ − 1)ρ ε, γ > 1.

Now, setting
qi = ρ u i , 1 ≤ i ≤ 3, E = ρ e,
the system (2.1) can be put into the general framework of system (1.1) if we
take
⎛ ⎞ ⎛ ⎞
ρ q1
⎜q1 ⎟ ⎜ p + q12 /ρ ⎟
⎜ ⎟ ⎜ ⎟
U=⎜ ⎟ ⎜
⎜q2 ⎟ , f1 (U) = ⎜ q1 q2 /ρ ⎟ ,
⎟ (2.2a)
⎝q3 ⎠ ⎝ q1 q3 /ρ ⎠
E (E + p)q1 /ρ
⎛ ⎞ ⎛ ⎞
q2 q3
⎜ q1 q2 /ρ ⎟ ⎜ q1 q3 /ρ ⎟
⎜ ⎟ ⎜ ⎟
f2 (U) = ⎜
⎜ p + q 2
2 /ρ ⎟
⎟ , f 3 (U) = ⎜ q2 q3 /ρ ⎟
⎜ ⎟ (2.2b)
⎝ q2 q3 /ρ ⎠ ⎝ p + q32 /ρ ⎠
(E + p)q2 /ρ (E + p)q3 /ρ

with  
E |q|2
p = p ρ, − 2
ρ 2ρ
and if the set of states is

|q|2
Ω= (ρ, q = (q1 , q2 , q3 ), E); ρ > 0, q ∈ R , E −
3
>0 .
2ρ

We shall see that in the one-dimensional case (resp. bidimensional), the

system (2.2a), (2.2b) of the gas dynamics equations with usual equation of
state is indeed a strictly hyperbolic (resp. hyperbolic) symmetrizable nonlin-
ear system of conservation laws.
In many applications, we do not have to solve the full system but only
a somewhat reduced system of equations. On the one hand, one can often
suppose, for instance, that the flow is barotropic so that the equation of state
reduces to
p = p(ρ),
and it suffices to solve the system of equations for conservation of mass and
momentum. On the other hand, if we assume that the flow has some symme-
try, one can reduce the number of space variables. For instance, assuming slab
symmetry, the gas dynamics equations become (using obvious notations)
8 I. Introduction
⎧
⎪
⎪ ∂ρ ∂
⎪
⎪ + (ρu) = 0,
⎪
⎪ ∂t ∂x
⎨
∂ ∂
(ρu) + (ρu2 + p) = 0, (2.3)
⎪
⎪ ∂t ∂x
⎪
⎪
⎪ ∂
⎪ ∂
⎩ (ρe) + ((ρe + p)u) = 0.
∂t ∂x
Again, choosing ρ, q = ρu, and E = ρe as dependent variables, (2.3) fits into
our framework with d = 1, p = 3.
More generally, if we only assume that the flow is invariant with respect
to (x2 , x3 ), the gas dynamics equations become with obvious notations
⎧
⎪ ∂ρ ∂
⎪
⎪ + (ρu) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ (ρu) + (ρu2 + p) = 0,
⎪
⎪ ∂t ∂x
⎪
⎨
∂ ∂
(ρv) + (ρuv) = 0, (2.4)
⎪ ∂t
⎪ ∂x
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ (ρw) + (ρuw) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪
⎩ ∂ (ρe) + ∂ ((ρe + p)u) = 0.
∂t ∂x
Again (2.4) fits into our framework with d = 1, p = 5. 

Example 2.2. The equations of ideal magnetohydrodynamics (MHD) in Eule-

rian coordinates. The MHD equations model the flow of a conducting fluid
in the presence of a magnetic field. The ideal MHD equations are obtained
when we neglect displacement current, electrostatic forces, viscosity, and heat
conduction. They read using convenient units
⎧
⎪ ∂ρ
⎪
⎪ + ∇ · (ρu) = 0,
⎪
⎪ ∂t
⎪
⎪
⎪
⎪
⎪
⎪ ∂ ∗ 1
⎨ ∂t (ρu) + ∇ · (ρu ⊗ u + p I − μ B ⊗ B) = 0,
⎪
(2.5)
⎪
⎪ ∂B
⎪
⎪ + ∇ · (B ⊗ u − u ⊗ B) = 0,
⎪
⎪ ∂t
⎪
⎪
⎪
⎪
⎪
⎪ ∂ 1
⎩ (ρe∗ ) + ∇ · ((ρe∗ + p∗ )u − (u · B)B) = 0
∂t μ
with the additional requirement

∇ · B = 0. (2.6)
2 Fluid Systems in Eulerian and Lagrangian Frames 9

In (2.5), B = (B1 , B2 , B3 )T is the magnetic field; μ is the resistivity of the

1
fluid (supposed to be constant), p∗ = p + |B|2 is the total pressure, sum of
2μ
1 1
the static pressure p and the magnetic pressure |B|2 ; and e∗ = ε + |u|2 +
2μ 2
1
|B| is the total specific energy.
2
2μρ
It is a simple matter to check that

∇ · (∇ · (B ⊗ u − u ⊗ B)) = 0

so that, by the third equation (2.5), the constraint (2.6) is automatically

satisfied as soon it is satisfied at time t = 0. In fact, the third equation (2.5)
is usually written in the form
∂B
+ ∇ × (B × u) = 0
∂t
which is equivalent under the constraint (2.6) (the expression ∇ × (B × u) is
also classically noted curl(B ∧ u)).
Again the ideal MHD system can be put into the framework of system (1.1)
if we set
⎛ ⎞
q1
⎛ ⎞ ⎜ 1 ⎟
ρ ⎜ p + q12 /ρ + p∗ − B12 ⎟
⎜ μ ⎟
⎜ q1 ⎟ ⎜ 1 ⎟
⎜ ⎟ ⎜ q1 q2 /ρ − B1 B2 ⎟
⎜ q2 ⎟ ⎜ ⎟
⎜ ⎟ ⎜ μ ⎟
⎜ q3 ⎟ ⎜ 1 ⎟
⎜ ⎟
U = ⎜ ⎟ , f1 (U) = ⎜ ⎜ q1 q3 /ρ − B1 B3 ⎟,...
B μ ⎟
⎜ 1⎟ ⎜ ⎟
⎜ B2 ⎟ ⎜ 0 ⎟
⎜ ⎟ ⎜ ⎟
⎝ B3 ⎠ ⎜ B2 u1 − B1 u2 ⎟
⎜ ⎟
E∗ ⎜ B3 u1 − B1 u3 ⎟
⎝ 1 ⎠
∗ ∗
(E + p )q1 /ρ − (q · B)B1
μρ
with
E∗ |q|2 |B|2 |B|2
p∗ = p(ρ, − 2 − )+ .
ρ 2ρ 2μρ 2μ
The set of states Ω is then defined as

∗ ∗|q|2 |B|2
Ω= (ρ, q, B, E ); ρ > 0, q ∈ R , B ∈ R , E −
3 3
− )>0 .
2ρ 2μ

If we assume that the flow is invariant with respect to (x2 , x3 ), the ideal
MHD equations read (using obvious notations)
10 I. Introduction
⎧
⎪ ∂ρ ∂
⎪
⎪ + (ρu) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ (ρu) + (ρu2 + p∗ ) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂ ∂ Bx By
⎪
⎪ (ρv) + (ρuv − ) = 0,
⎪
⎪
⎪
⎪
∂t ∂x μ
⎪
⎪ ∂ ∂ Bx Bz
⎨ (ρw) + (ρuw − ) = 0,
∂t ∂x μ (2.7)
⎪
⎪
⎪
⎪ ∂By ∂
⎪
⎪ + (By u − Bx v) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂Bz ∂
⎪
⎪ + (Bz u − Bx w) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂ ∂ Bx
⎪
⎪ (ρe∗ ) + ((ρe∗ + p∗ )u − (Bx u + By v + Bz w)) = 0,
⎪
⎪ ∂t ∂x μ
⎪
⎩
Bx = constant.

We shall show in the next chapter that the ideal MHD system is indeed
hyperbolic (but not strictly hyperbolic) and symmetrizable in the whole set
of states Ω. 
More generally, a number of nonlinear hyperbolic systems of conservation
laws arising in fluid mechanics can be written in the form
⎧
⎪ ∂ρ
⎪
⎨ ∂t + ∇ · (ρu) = 0,
(2.8)
⎪
⎪
⎩ ∂ (ρΦ) + ∇ · (ρΦ ⊗ u + g(ρ, Φ)) = ρs(ρ, Φ)
∂t
where ⎛ ⎞
Φ1
⎜ ⎟
Φ = ⎝ ... ⎠ , g = (gij )1≤i≤p−1,1≤j≤3 .
Φp−1
In (2.8), ρ is again the mass density of the fluid, u is its velocity while ρΦ
stands for the other conservative variables, and s is a source term. Note that,
in (2.8), we have distinguished the mass conservation equation from the other
conservation laws. Indeed, the equations of gas dynamics and the ideal MHD
equations fit into the framework of such a general system (2.8) with
   
u pI
Φ= , g(ρ, Φ) = (2.9)
e pu

and
2 Fluid Systems in Eulerian and Lagrangian Frames 11
⎛ ⎞
⎛ ⎞ 1
u p∗ I − B⊗B
⎜ μ ⎟
⎜ ⎟ ⎜ ⎟
⎜ ⎟ ⎜ −u ⊗ B ⎟
Φ = ⎜B⎟ , g(ρ, Φ) = ⎜ ⎟ (2.10)
⎝ρ⎠ ⎜ ⎟
⎝ 1 ⎠
e∗ p∗ u − (u · B)B
μ
respectively, and s = 0.
Let us now write the system (2.8) using Lagrangian coordinates. Consider
the differential system
dx
= u(x, t) (2.11)
dt
and, for all ξ ∈ R3 , we denote by t → x(ξ, t) the solution of (2.11) that
satisfies the initial condition

x(ξ, 0) = ξ. (2.12)

Then (ξ = (ξ1 , ξ2 , ξ3 ), t) are the Lagrangian coordinates associated with the

velocity field u. If we set
∂xi
J(ξ, t) = det (ξ, t) , (2.13)
∂ξj

it is a standard exercise to show that

∂J
(ξ, t) = J(ξ, t)(∇ · u)(x(ξ, t), t). (2.14)
∂t
Now, given a function ϕ = ϕ(x, t) expressed in Eulerian coordinates, we
denote by ϕ = ϕ(ξ, t) this function expressed in Lagrangian coordinates, i.e.,

ϕ(ξ, t) = ϕ(x(ξ, t), t).

Then, using (2.14), an easy computation shows that

∂ϕ  ∂
3
∂ ∂ϕ
(ϕ J) = J + ∇ · (ϕu) = J + (ϕuj ) .
∂t ∂t ∂t j=1
∂xj

Hence, the system (2.8) becomes

⎧
⎪ ∂
⎨ (ρ J) = 0,
∂t (2.15)
⎪
⎩ ∂ (ρΦ J) + J ∇ · g = 0.
∂t
Let us again assume that the flow is invariant with respect to (x2 , x3 ) so
that (2.8) reads
12 I. Introduction
⎧
⎪ ∂ρ ∂
⎨ + (ρu),
∂t ∂x (2.16)
⎪
⎩ ∂ (ρΦ) + ∂ (ρΦu + g) = Jρs
∂t ∂x
and (2.15) becomes
⎧
⎪
⎪ ∂
⎨ (ρ J) = 0,
∂t
(2.17)
⎪
⎪
⎩ ∂ (ρΦ J) + J ∂g = Jρs.
∂t ∂x
Since
∂ ∂x ∂ ∂
= =J ,
∂ξ ∂ξ ∂x ∂x
we obtain, suppressing the bars for simplicity
⎧
⎪
⎪ ∂
⎨ (ρ J) = 0,
∂t
(2.18)
⎪
⎪ ∂ ∂g
⎩ (ρΦ J) + J = Jρs, g = g(ρ, Φ).
∂t ∂ξ

Let us put this system in a more classical form. The first equation (of mass
conservation) gives
ρ J = ρ0 ,
where ρ0 (ξ) = ρ(ξ, 0). If we introduce the specific volume
1
τ= ,
ρ
we get
J = ρ0 τ
so that the one-dimensional analogue of (2.14)

∂J ∂u ∂u
=J =
∂t ∂x ∂ξ
becomes
∂τ ∂u
ρ0 − = 0,
∂t ∂ξ
and the second equation (2.15) gives

∂Φ ∂g
ρ0 + = ρ0 s.
∂t ∂ξ
Finally, we introduce a mass variable m such that

dm = ρ0 dξ.
2 Fluid Systems in Eulerian and Lagrangian Frames 13

Then the system (2.16) written in Lagrangian coordinates takes the simple
form ⎧
⎪ ∂τ ∂u
⎪
⎨ ∂t − ∂m = 0,
(2.19)
⎪ ∂Φ
⎪ ∂g
⎩ + = s,
∂t ∂m
with s = s(1/τ, Φ).
Remark 2.1. Let us have a look at the two-dimensional case. We now
write (2.8) in the form (without source for simplicity)
⎧
⎪
⎪ ∂ρ ∂ ∂
⎨ + (ρu) + (ρv) = 0,
∂t ∂x ∂y
⎪
⎪ ∂ ∂ ∂
⎩ (ρΦ) + (ρuΦ + g(ρ, Φ)) + (ρvΦ + h(ρ, Φ)) = 0
∂t ∂x ∂y

with obvious notations, and it becomes (see (2.15))

⎧
⎪ ∂
⎪
⎨ ∂t (ρ J) = 0,
⎪
⎪ ∂ ∂g ∂h
⎩ (ρΦ J) + J ( + ) = 0.
∂t ∂x ∂y

In order to obtain a more useful form, denoting by (ξ, η, t) the Lagrangian

∂g ∂h
coordinates, we express the derivatives and in terms of derivatives wrt.
∂x ∂y
(ξ, η). We use the fact that the 2×2 Jacobian matrix of the map (x, y) → (ξ, η)
is the inverse of the Jacobian matrix of the map (ξ, η) → (x, y) which is easily
computed, and together with the definition (2.13) of J, we get

∂g ∂h ∂g ∂ξ ∂g ∂η ∂h ∂ξ ∂h ∂η
J( + ) = J( + + + )
∂x ∂y ∂ξ ∂x ∂η ∂x ∂ξ ∂y ∂η ∂y

∂g ∂y ∂g ∂y ∂h ∂x ∂h ∂x
= − − + )
∂ξ ∂η ∂η ∂ξ ∂ξ ∂η ∂η ∂ξ
∂ ∂y ∂x ∂ ∂y ∂x
= (g −h )+ (−g + h ).
∂ξ ∂η ∂η ∂η ∂ξ ∂ξ
∂x ∂y ∂x ∂y
Setting a = ,b = ,c = ,d = , we have J = ad − bc, and we
∂ξ ∂ξ ∂η ∂η
notice that the system in Lagrangian coordinates can be equivalently written
(dropping the bars)
14 I. Introduction
⎧ ∂τ ∂ ∂
⎪
⎪ ρ0 − (ud − vc) − (−ub + va) = 0,
⎪
⎪ ∂t ∂ξ ∂η
⎪
⎪
⎪
⎪ ∂Φ ∂ ∂
⎪
⎪
⎨ ρ0 ∂t + ∂ξ (gd − hc) + ∂η (−gb + ha) = 0,
   
⎪
⎪ ∂ a ∂ u
⎪
⎪ − = 0,
⎪
⎪ ∂t b ∂ξ v
⎪
⎪    
⎪
⎪
⎩ ∂ c − ∂ u = 0.
∂t d ∂η v

The three-dimensional case can be handled in a similar way. For a deeper

analysis, we refer to [639] and [416, 419]. 

Example 2.3. The equations of gas dynamics in Lagrangian coordinates.

If we choose in (2.19)
⎛ ⎞ ⎛ ⎞
u p
⎜v ⎟ ⎜0⎟
Φ=⎜ ⎟ ⎜ ⎟
⎝w⎠ , g = ⎝ 0 ⎠
e pu

we obtain the system of gas dynamics (2.4) which reads in Lagrangian coor-
dinates ⎧
⎪ ∂τ ∂u
⎪
⎪ − = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪
⎪ ∂u ∂p
⎪
⎪ + = 0,
⎪
⎪ ∂t ∂m
⎪
⎨
∂v
= 0, (2.20)
⎪ ∂t
⎪
⎪
⎪
⎪
⎪ ∂w
⎪
⎪ = 0,
⎪
⎪ ∂t
⎪
⎪
⎪
⎪
⎩ ∂e + ∂(pu) = 0
∂t ∂m
2
with p = p(τ, ε) = p(τ, e − u2 ). By setting
⎛ ⎞ ⎛ ⎞
τ −u
⎜u⎟ ⎜ p ⎟  
⎜ ⎟ ⎜ ⎟ |u|2
V= ⎜ ⎟ ⎜ ⎟
⎜ v ⎟ , f (V) = ⎜ 0 ⎟ , Ω = V; τ > 0, u = (u, v, w) ∈ R , e− 2 > 0 ,
3

⎝w ⎠ ⎝ 0 ⎠
e pu

we obtain a nonlinear system of conservation laws

∂V ∂
+ f (V) = 0 (2.21)
∂t ∂m
that again will be seen to be strictly hyperbolic and symmetrizable under
natural physical assumptions. On the other hand, the system (2.3) written
2 Fluid Systems in Eulerian and Lagrangian Frames 15

in Lagrangian coordinates is of the form (2.21) with

⎛ ⎞ ⎛ ⎞
τ −u
V = ⎝u⎠ , f (V) = ⎝ p ⎠ .
e pu

If we assume in addition that the flow is barotropic, the equation of state

reduces to p = p(τ ). Again, it suffices to solve the system of equations of
conservation of mass and momentum
⎧
⎪ ∂τ ∂u
⎨ − = 0,
∂t ∂m
⎪
⎩ ∂u + ∂p = 0.
∂t ∂m
We obtain the p-system (1.7) but with different notations. 

Example 2.4. The ideal MHD equations in Lagrangian coordinates.

If we choose in (2.19) the pair (Φ, g(ρ, Φ)) as
⎛ ⎞
p∗
⎛ ⎞ ⎜ Bx ⎟
u ⎜ − By ⎟
⎜ v ⎟ ⎜ μ ⎟
⎜ ⎟ ⎜ ⎟
⎜ w ⎟ ⎜ B x ⎟
⎜ − B ⎟
Φ=⎜ ⎜τ By ⎟
⎟, g = ⎜ μ
z
⎟,
⎜ ⎟ ⎜ −B ⎟
⎝ τ Bz ⎠ ⎜ x v ⎟
⎜ −Bx w ⎟
⎜ ⎟
e ⎝ Bx ⎠
∗
p u− (Bx u + By v + Bz w)
μ
we obtain the ideal MHD equations which become in Lagrangian coordinates
⎧
⎪
⎪
∂τ
−
∂u
⎪
⎪ = 0,
⎪
⎪ ∂t ∂m
⎪
⎪ ∗
⎪
⎪ ∂u ∂p
⎪
⎪ + = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪
⎪ ∂v ∂ Bx
⎪
⎪ − ( By ) = 0,
⎪
⎪ ∂t ∂m μ
⎪
⎨ ∂w ∂ Bx
− ( Bz ) = 0, (2.22)
⎪
⎪ ∂t ∂m μ
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ (τ By ) − (Bx v) = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪
⎪ ∂ ∂
⎪
⎪ (τ Bz ) − (Bx w) = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪
⎪ ∂e∗ ∂ Bx
⎪
⎩ + (p∗ u − (Bx u + By v + Bz w)) = 0
∂t ∂m μ
where
16 I. Introduction

|B|2 1 τ |B|2
p∗ = p(τ, ε) + , e∗ = ε + (u2 + v 2 + w2 ) + .
2μ 2 2μ

Again this system is of the form (2.21) 

In fact, in many applications, we need to consider systems of conservation
laws with source terms

∂u  ∂
d
+ fj (u) = s(u), x = (x1 , . . . , xd ) ∈ Rd , t > 0, (2.23)
∂t j=1
∂xj

where s is a smooth function from Ω into Rp , s may also depend on x.

This source term may be present in different situations; it may be linked to
external forces, such as gravity, and to the presence of chemical reaction; or
it may have a geometric nature. Let us give below several examples of such
situations.
Example 2.5. The system of gas dynamics with gravity and friction.
If we take into account both gravity and friction terms, the Euler system
of gas dynamics becomes
⎧
⎪ ∂ρ
⎪
⎪ + ∇ · (ρu) = 0,
⎪
⎪ ∂t
⎪
⎪
⎨
∂
(ρu) + ∇ · (ρu ⊗ u + pI) = ρ(g − αϕ(u)), (2.24)
⎪
⎪ ∂t
⎪
⎪
⎪
⎪
⎪
⎩ ∂ (ρe) + ∇ · ((ρe + p)u) = ρ(g · u − αψ(u))
∂t
where g stands for the gravity force and α > 0 is a friction constant coefficient.
The functions ϕ and ψ are typically of the forms

ϕ(u) = |u|χ u, ψ(u) = a|u|χ+2

where χ = 0 for a linear friction, χ = 1 for a quadratic one, and a > 0 is a

constant.
Using the notations of Example 2.1, we have here
⎛ ⎞
0
⎜ ⎟
⎜ ⎟
⎜ ρ(g − αϕ( q )) ⎟
s(U) = ⎜ ρ ⎟ .
⎜ ⎟
⎝ q ⎠
g · q − αρψ( )
ρ
Assuming slab symmetry, we obtain the system
2 Fluid Systems in Eulerian and Lagrangian Frames 17
⎧
⎪
⎪ ∂ρ ∂
⎪
⎪ + (ρu) = 0,
⎪
⎪ ∂t ∂x
⎨
∂ ∂
(ρu) + (ρu2 + p) = ρ(g − αϕ(u)), (2.25)
⎪
⎪ ∂t ∂x
⎪
⎪
⎪ ∂
⎪ ∂
⎩ (ρe) + ((ρe + p)u) = ρ(gu − αψ(u)).
∂t ∂x
Passing in Lagrangian coordinates, (2.25) becomes
⎧
⎪
⎪ ∂τ ∂u
⎪
⎪ − = 0,
⎪
⎪ ∂t ∂m
⎨
∂u ∂u
+ = g − αϕ(u), (2.26)
⎪
⎪ ∂t ∂m
⎪
⎪
⎪ ∂e
⎪ ∂
⎩ + (pu) = gu − αψ(u)).
∂t ∂m


Example 2.6. The equations of reacting gas flows. Consider the more complex
example of a system of reacting gas flows. We assume that the system is
multicomponent, consisting of n reacting species. If we neglect again viscosity
and heat conduction, the equations of the flow are obtained by adjoining
to the Euler system of gas dynamics the equations governing the chemical
reactions between species
∂
(ρYi ) + ∇ · (ρYi u) = mi ωi , 1 ≤ i ≤ n − 1. (2.27)
∂t
In (2.27), Yi is the mass fraction; mi is the molar mass, also termed the
molecular weight, of the ith species; and ωi is the molar production rate of
the ith species with
n n
Yi = 1, mi ωi = 0, (2.28)
i=1 i=1

so that we have replaced the equation of conservation of the nth species by

the total mass conservation which is part of the Euler system.
In order to close the system of equations, we need an equation of state which
can be taken in the form

ε = ε(ρ, p, Y), Y = (Y1 , . . . , Yn−1 )T (2.29)

and expressions for the production rates

mi ωi = mi ωi (ρ, p, Y), 1 ≤ i ≤ n − 1. (2.30)

For example, if we suppose that each species behaves as a polytropic ideal

gas with adiabatic exponent γi , we have
18 I. Introduction


n
1 p
ε= Yi h0i +
i=1
γ(Y) − 1 ρ

where h0i is the enthalpy of formation of the species i at 00 K and

n
Yi Cp,i

γ(Y) = ni=1
i=1 i Cv,i
Y

Cv,i (resp. Cp,i ) denotes its specific heat at constant volume (resp. at con-
stant pressure). For more details on ideal gases, see Chap. III, Sect. 1.2. If
Cp,i
in addition γi = is independent of i, i.e., all the species have the same
Cv,i
adiabatic exponent γ, we obtain γ(Y) = γ.
Now, going back to the general case, we set

ρr = (m1 ω1 , . . . , mn−1 ωn−1 )T

and      
U fj 0
W= , gj = , 1 ≤ j ≤ 3, s=
ρY qj Y ρr
where we have used the notations (2.2a), (2.2b) of Example 2.1. We obtain
the system
∂W
+∇·g =s
∂t
which is indeed of the form (2.23) (note that we may write the flux with
ρY
Y= , i.e., in terms of the conservative variables).

Assuming slab symmetry, the equations of reacting gas flows become
⎧
⎪
⎪
∂ρ ∂
⎪
⎪ + (ρu) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪ ∂ ∂
⎨ (ρu) + (ρu2 + p) = 0,
∂t ∂x (2.31)
⎪
⎪ ∂ ∂
⎪
⎪ (ρe) + ((ρe + p)u) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎪ ∂
⎪ ∂
⎩ (ρY) + (ρYu) = ρr.
∂t ∂x
In Lagrangian coordinates, the last equation (2.31) gives

∂
(ρY J) = Jρr
∂t
or equivalently by the mass conservation equation
2 Fluid Systems in Eulerian and Lagrangian Frames 19

∂Y
= r.
∂t
Hence, (2.31) gives in Lagrangian coordinates (dropping the bars)
⎧
⎪ ∂τ ∂u
⎪
⎪ − = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪ ∂u
⎪ ∂p
⎪
⎨ + = 0,
∂t ∂m
(2.32)
⎪
⎪ ∂e ∂(pu)
⎪
⎪ + = 0,
⎪
⎪ ∂t ∂m
⎪
⎪
⎪
⎪
⎩ ∂Y = r.
∂t
We next restrict ourselves to the simple case where we consider only two
species, the unburnt gas and the burnt gas. We denote by z the mass fraction
of the burnt gas so that 1 − z is the mass fraction of the unburnt gas. The
simplest model is obtained by assuming that on the one hand the unburnt
gas is converted to the burnt gas through an irreversible chemical reaction
with a production rate
r = r(ρ, p, z) (2.33)
and on the other hand that both unburnt and burnt gases behave as poly-
tropic ideal gases with the same adiabatic exponent γ so that the equation
of state reads
1 p
ε(ρ, p, z) = zh0b + (1 − z)h0u + .
γ−1ρ
Observe that h0b < h0u for an exothermic reaction. In this simple case, (2.27)
resumes to
∂
(ρz) + ∇ · (ρzu) = ρr.
∂t
If we assume that the kinetics of the chemical reaction obeys an Arrhenius
mechanism, we have
 
E∗
r = K(1 − z) exp − ,
RT

where E ∗ is the activation energy and T is the temperature. In practice, a

simpler expression for r is often used, for instance,

r = K(1 − z)α ,

with α = 1
2 or 1. 
One can even consider systems of conservation laws more general than
(2.23), namely, systems of the form
20 I. Introduction

∂u
+ ∇ · f (x, t, u) = s(x, t, u), x ∈ Rd , t > 0, (2.34)
∂t
where the flux function f = (fj )1≤j≤d and the source term s are smooth
functions from Rd × [0, +∞[×Ω into (Rp )d and Rp , respectively.

3 Some Averaged Models: Shallow Water, Flow in a

Duct, and Two-Phase Flow

The examples we consider now are obtained by averaging a given complete

fluid model, the incompressible Navier-Stokes system on a column of water
for the Saint-Venant system, respectively the Euler system on a section for
the flow in a duct (or a nozzle). The corresponding models necessarily result
from some simplifying assumption, in order that the averaged quantities give
a good description of the flow. We will see that the averaging procedure
leads to systems with a geometric source term (depending on x), so that they
do not benefit from a useful Lagrangian description. We begin by stating a
classical lemma (Leibniz integral rule), which is involved in the derivation
of the following models and explains why the averaging procedure leads in
general to nonconservative terms.
Lemma 3.1
Let α, β (resp. ϕ) be sufficiently smooth real functions on R2 (resp. R3 ). We
have the identity, for i = 1, 2,
β(x1 ,x2 ) β(x1 ,x2 )
∂ ∂
ϕ(x1 , x2 , z)dz = ϕ(x1 , x2 , z)dz
α(x1 ,x2 ) ∂xi ∂xi α(x1 ,x2 )
∂β ∂α
− ϕ(x1 , x2 , β(x1 , x2 )) − ϕ(x1 , x2 , α(x1 , x2 )) .
∂xi ∂xi
Example 3.1. The Saint-Venant system. It models shallow water with topog-
raphy and reads
⎧
⎪ ∂h
⎪
⎨ ∂t + ∇ · (hu) = 0,
(3.1)
⎪
⎪ ∂ g
⎩ (hu) + ∇ · (hu ⊗ u) + ∇h = −gh∇Z,
2
∂t 2
where h = h(x, y, t) denotes the water height, u(x, y, t) = (u, v)T (x, y, t) is
the flow velocity, and Z = Z(x, y) characterizes the topography. This system,
also called the Saint-Venant system [87], writes as the barotropic system of
gas dynamics with a source term; the first equation expresses the conservation
of mass of a column of water, and in the second equation, the pressure law
g
is taken as p(h) = h2 .
2
3 Some Averaged Models: Shallow Water, Flow in a Duct, and Two-Phase Flow 21

System (3.1) is obtained by averaging the incompressible Euler system

with free boundary on a vertical column of water, from a bottom z = Z(x, y)
to the free surface, say z = η(x, y, t), with kinematic boundary conditions.
η
Thus, h = η − Z and the velocity is the average u = h1 Z u(x, y, z, t)dz,
and we have dropped the bars for simplicity. The hydrostatic approximation
H
amounts to introduce a small parameter  = where H and L are two char-
L
acteristic dimensions along the vertical and horizontal direction, respectively,
to rescale the equations and neglect the vertical acceleration of the fluid (of
order O(2 )) in the rescaled Euler system. This assumption yields that the
pressure is simply the hydrostatic pressure ρ0 gh, where g is the gravitational
constant and ρ0 the (constant) fluid density usually taken equal to 1 for wa-
ter. The advantage of the averaging procedure is to transform a problem with
a free surface into a problem where the height of water becomes a dependent
variable, and this explains why it leads to a compressible flow model. Note
that the source term is a byproduct of the integration step, resulting from
Lemma 3.1 (we do not go into the details of the computation; see [512] and
the references therein).
Let us introduce the energy,
1
E(h, u, Z) = (h|u|2 + gh2 ) + gZh.
2
Then smooth solutions of (3.1) can be shown to satisfy the energy conserva-
tion equation
 g 
∂t E + div (u E + h2 ) = 0.
2
We will see that this means that system (3.1) admits a mathematical en-
tropy, which is the physical energy; in case of a discontinuous solution, it
becomes an inequality. Let us also mention that this energy conservation
equation can be obtained directly from the averaging of the energy equation
in the incompressible Navier-Stokes system together with the shallow water
approximation.
If we assume that the flow is one-dimensional, in the direction x, we get
⎧
⎪ ∂h ∂
⎪
⎨ + (hu) = 0,
∂t ∂x
(3.2)
⎪
⎪
⎩ ∂ (hu) + ∂ (hu2 + g h2 ) = −ghZ  (x).
∂t ∂x 2
A more complete model, including viscosity and friction, can be derived start-
ing from the Navier-Stokes system together with the hydrostatic assumption;
we refer to [512]. We will give more references concerning extended shallow
water models in Chap. VII concerning source terms. 

Example 3.2. Flow in a duct with variable cross section. We now give another
example of a system for which the source term comes from an averaging
22 I. Introduction

procedure (which means here by integrating on a section) together with some

symmetry assumption. We consider the flow in a duct with variable section (a
nozzle) aligned with the x axis, with cross section S(x) smoothly varying with
space, and we note the area |S(x)| = A(x). By symmetry, we can consider a
2d duct (in the plane z = 0), limited by a curve with equation y = A(x)/2
(resp. y = −A(x)/2) for the upper (resp. lower part).
In order to derive a simplified PDE model, we start by integrating the
system (2.1) on a small streamtube, say ω, around the x axis, between x and
x + δx, with section S(ξ), ξ ∈ [x, x + δx]. First we define for any quantity ϕ
its average over a section S(x) noted ϕ by

A(x)ϕ(x, t) = ϕ(x, y, t)dy.

S(x)

We consider the integral of the first equation of (1.7) and get for the first
term
x+δx
∂t ρ(x, t)dx = ∂t ( ρ(x, t)dx) = ∂t A(ξ)ρ(ξ, t)dξ
ω ω x

where ρ(x, t) is the average density over the section S(x). Then we write

∇ · (ρu)dx = ρu.ndσ.
ω ∂ω

Now, the surface ∂ω is made of the two parallel sections, say Sx (resp. Sx+δx )
with outward normal −e1 (resp. e1 ), where e1 denotes the unit vector along
the x axis, and with length A(x) (resp. A(x + δx)) and a lateral surface which
we denote Σ, then
= + + .
∂ω Sx Sx+δx Σ

We assume that the flow propagates in a streamtube; hence, u · n = 0 on

Σ and the corresponding integral vanishes. The two remaining integrals are
easily computed, because u · e1 = u. Indeed, for any quantity ϕ, we have

ϕu.n dσ + ϕu.ndσ = −A(x)ϕu(x, t) + A(x + δx)ϕu(x + δx, t).

Sx Sx+δx

Thus, dividing by δx and letting δx → 0, we get an equation for the averaged

density
∂t (Aρ) + ∂x (Aρu) = 0.
Similarly, we integrate the first momentum conservation equation, which
we can write
∂t (ρu) + ∇ · (ρuu + pe1 ) = 0.
3 Some Averaged Models: Shallow Water, Flow in a Duct, and Two-Phase Flow 23

Following the same lines, we obtain the terms ∂t ρu + ∂x (ρu2 + p). However,
there is now an added contribution coming from the pressure term p e1 inte-
grated on the lateral surface Σ. The tangent vector to the upper part of Σ is
(1, A (x)/2)T and the normal to the upper part (resp. lower) is (−A (x)/2, 1)T
(resp. (−1, −A (x)/2)T ), and we add the equal contributions of the upper and
lower parts. Hence, if n denotes the unit normal vector (obtained by normal-
izing the normal vector (−A (x)/2, 1)T ) and dσ is the arc length measure
along Σ, we get
x+δx
1
p(x, t) e1 · n dσ = p(ξ, A(ξ)/2, t) + p(ξ, −A(ξ)/2, t) A (ξ)dξ.
Σ 2 x

Thus, dividing by δx and letting δx → 0, we get a term A (x)p, where p

denotes the mean pressure p(x, t) = 12 (p(x, A(x)/2, t) + p(x, −A(x)/2, t) on
Σ. Thus, averaging the momentum conservation equation yields

∂t (Aρu) + ∂x (A(ρu2 + p)) − pA (x) = 0.

In the same way, integrating the second momentum conservation equation,

∂t (ρv) + ∇ · (ρvu + pe2 ) = 0,

yields
∂t (Aρv) + ∂x (A(ρuv)) − δp = 0,
with δp(x, t) = 12 (p(x, A(x)/2, t) − p(x, −A(x)/2, t). Now, if we add the as-
sumption that the flow is symmetric, the average velocity v vanishes, similarly
ρv = 0, and also the term corresponding to the pressure Σ p(x, t) e2 · n dσ =
0, we can then ignore the second momentum equation.
The energy equation comes as the other ones since we integrate the term
∇·((e+p)u), and again, only a term with u.n is involved. This term vanishes
on Σ and is easily computed on the cross sections S(x), S(x + δx).
Let us assume last that both pressure terms p and p can be replaced by
p(ρ, ε), and ρu2 by (ρu)2 /ρ̄, then dropping the bars for simplicity, we get the
system ⎧
⎪ ∂ ∂
⎪
⎪ (Aρ) + (Au) = 0,
⎪
⎪ ∂t ∂t
⎪
⎪
⎨
∂ ∂
(Aρu) + (A(ρu2 + p)) = pA (x), (3.3)
⎪
⎪ ∂t ∂x
⎪
⎪
⎪
⎪
⎪
⎩ ∂ (Aρe) + ∂ (A(e + p)u) = 0,
∂t ∂x
which also looks like the Euler system, provided we define a surfacic density
r = Aρ, with a source term which is generally called a geometric source term
(when A is not constant). However, we have to look more precisely at the
pressure law given in terms of conservative variables. Assuming for simplicity
24 I. Introduction

that the fluid is barotropic, system (3.3) gives

⎧
⎪ ∂ ∂
⎪
⎨ ∂t (Aρ) + ∂x (Aρu) = 0,
⎪ ∂
⎪ ∂
⎩ (Aρu) + (A(ρu2 + p(ρ))) = p(ρ)A (x),
∂t ∂x
which now looks like the barotropic Euler system with a source. Then the
pressure term can be given in terms of r = Aρ, P (r) = Ap(r/A). However,
this definition of P introduces a flux depending on x which does not simplify
the study. In fact, the system can be written equivalently replacing A (x)
∂
by A and completed by an equation which says that A does not vary
∂x
with time. This approach is now classical for a geometric source term (which
thus does not depend on t) or some systems of conservation laws with flux
depending on x. It yields here
⎧
⎪ ∂ ∂
⎪
⎪ (Aρ) + (Aρu) = 0,
⎪
⎪ ∂t ∂x
⎪
⎪
⎨
∂ ∂ ∂
(Aρu) + (A(ρu2 + p())) − p(ρ) A = 0, (3.4)
⎪
⎪ ∂t ∂x ∂x
⎪
⎪
⎪
⎪
⎪
⎩ ∂ A = 0.
∂t
System (3.4) is a first example of a system with a nonconservative product
∂
since, when A is not constant, the term p(ρ) A cannot be put in divergence
∂x
form. 
Similarly, system (3.2) can be transformed and completed by an equation
∂
Z = 0, and it provides another example of a system with a nonconservative
∂t
term. We shall come again on this approach later on in Chap. VII and study
the corresponding nonconservative systems in more details.
Example 3.3. Simple models of two-phase flow. In the context of two-phase
flows, a similar averaging approach leads to a model which may be justified
for a stratified flow. Assuming the flow evolves between two plates and con-
sists of two inviscid fields separated by a smooth single-valued surface, the
averaging procedure allows to reduce a two-dimensional (stratified) flow to
a one-dimensional model. Denoting by z = α(x, t) the interface between the
two fluids, one integrates the 2D-Euler system on [0, α] (resp. [α, 1]) for fluid 1
(resp. fluid 2), with equation of state p1 (ρ, s) (resp. p2 (ρ, s)). In this approach,
α plays the role of a volume fraction. In the following, for ease of notation,
we will set α1 = α, α2 = 1 − α. The resulting system has six equations, but
it involves interface quantities, and more modeling assumptions are needed
to close the system, in particular assumptions concerning the values of the
3 Some Averaged Models: Shallow Water, Flow in a Duct, and Two-Phase Flow 25

velocity and pressure at the interface, among which some simplifications are
usually done to lead to a tractable model. We do not go into details for which
we refer to [965, 1073], and give only some simple examples.
Assume for simplicity that the flow is isentropic. We first consider the
case where the two fields have an equal pressure p. It leads to a four-equation
model, which writes, with k = 1, 2 and α1 + α2 = 1,
⎧
⎪ ∂ ∂
⎪
⎨ ∂t (αk ρk ) + ∂x (αk ρk uk ) = 0,
(3.5)
⎪
⎪
⎩ ∂ (αk ρk uk ) + ∂ (αk ρk u2 ) + αk ∂ p + Δp ∂ αk = Mk ,
k
∂t ∂x ∂x ∂x
with equations of state ρi = ρi (p) and where we have noted Δp = p − pI , a
pressure default. Here pI denotes an interfacial pressure and Δp is some given
function of the state u = (αρ1 , (1 − α)ρ2 , αρ1 u1 , (1 − α)ρ2 u2 )T . In the right-
hand side, the Mk ’s are some source terms, satisfying M1 + M2 = 0, which
we assume do not contain derivatives of the variables: they model interfacial
forces (such as interface drag which is proportional to the relative velocity
ur = u1 − u2 ) or external forces (such as gravity). When Δp = 0, we get the
so-called one-pressure model. Note that the second equation in (3.5) can also
be written
∂ ∂ ∂
(αk ρk uk ) + (αk (ρk u2k + p)) − pI α k = Mk ,
∂t ∂x ∂x
which clearly resembles the second equation obtained in the preceding exam-
ple modeling the flow in a nozzle, with now αk (x, t) unknown and depending
also on t. System (3.5) is nonconservative, and it can be put in the form

∂ ∂
u + A(u) u = 0.
∂t ∂x
However, the resulting model may fail to be hyperbolic, in the sense that
A(u) does not necessarily have real eigenvalues for all physically acceptable
quantities. In particular, the one pressure model (Δp = 0) is not hyper-
bolic in general (not in the range of relative velocities in common physical
applications), so that, usely, one builds some “pressure correction” term Δp
proportional to u2r , where again ur = u1 − u2 is the relative velocity [896],
or a correction term involving derivatives (see [1130]). This system, though
simple, gives an example where the computation of the eigenvalues, which
are solution of a fourth-degree polynomial, cannot be achieved explicitly, and
the result is obtained by perturbation, using some Taylor expansions.
A related model keeps two different pressures, leading to the so-called five-
equation two-pressure model which writes, ignoring the external forces and
transfer terms between the phases,
26 I. Introduction
⎧
⎪ ∂ ∂
⎪
⎪ (αk ρk ) + (αk ρk uk ) = 0, k = 1, 2,
⎪
⎪ ∂t ∂x
⎪
⎪
⎨
∂ ∂ ∂
(αk ρk uk ) + (αk (ρk u2k + pk )) − pI αk = 0, k = 1, 2, (3.6)
⎪ ∂t
⎪ ∂x ∂x
⎪
⎪
⎪
⎪
⎪
⎩ ∂ α1 + uI ∂ α1 = δ,
∂t ∂x
with equations of state pi = pi (ρi ) and where uI , pI are the interfacial veloc-
ity and pressure which are supposed to be known functions of the state. A
frequent choice is uI = u2 , pI = p1 ; the model is often called the (isentropic)
model of Baer-Nunziato. The eigenvalues are explicitly computed [965], and
the system is hyperbolic (outside resonance). In the last equation, when the
source term δ is taken proportional to the relative pressure p1 − p2 , with a
coefficient which can be stiff, one speaks of a pressure relaxation term. At
least formally, the limit for zero relaxation time forces the equality of pres-
sures p1 = p2 and gives the four-equation two-fluid model (3.5); for the full
seven-equation model with energy, a hyperbolic five-equation reduced model
is derived in [890] via an asymptotic analysis for zero relaxation time (see
also [16]).
An equilibrium mixture model corresponds to the assumption that the two
fluids have equal velocity, and we have only one total momentum equation,
obtained by adding the momentum equations of each phase; it writes in the
barotropic case as a three-equation model
⎧
⎪ ∂ ∂
⎪
⎨ ∂t (αk ρk ) + ∂x (αk ρk u) = 0, k = 1, 2,
(3.7)
⎪
⎪
⎩ ∂ (ρu) + ∂ (ρu2 + p) = 0,
∂t ∂x

where ρ = k=1,2 αk ρk is the total density, and the system is closed with
equations of state for each fluid pk (ρk ) and equating the pressure p = p1 (ρ1 ) =
p2 (ρ2 ). We may equivalently replace one equation of conservation of the mass
of one phase by the sum of the two equations which writes
∂ ∂
ρ+ ρu = 0.
∂t ∂x
The resulting system (3.7) is hyperbolic, and the eigenvalues are explicit [279].
For modeling a phase transition, one can consider a relaxation source term
of the form ± 1ε (g1 − g2 ), gi denoting the chemical potential of phase i, that
can be added in the right-hand side of each equation of conservation of mass
(the sum of the term is zero so as to keep the conservation of total mass).
Then the limit system as ε → 0 is the (barotropic) Euler system (conservation
of total mass and momentum) with a pressure law which takes into account
the transition between the two phases; we refer to [238] (see also [602] for the
system with energy).
4 Weak Solutions of Systems of Conservation Laws 27

It is just an insight into the extreme variety of models for two-phase flow,
which, according to the modeling assumptions, differ by many aspects, in
particular the number of equations, the presence of nonconservative products
[1010], not mentioning the source terms, and also for which the hyperbolic
feature is not an obvious issue (see [346, 364, 693], and also [511] for mul-
tiphase flow modeling via Hamilton’s principle). Though we did not go into
details on the modeling assumptions, because we focus on giving different
examples of systems for which a general numerical approach is possible, it
should be emphasized that a given model is valid only in some flow configu-
ration and may fail to reproduce other configurations (see [965, 1073], for a
recent survey on multi-fluid models [195]. 
For the sake of simplicity, we now essentially concentrate in the following
on systems of the form (1.1) and occasionally on systems of the form (2.23)
since they cover a large range of physical situations. The treatment of source
terms will be given some consideration later on, in a special chapter.

4 Weak Solutions of Systems of Conservation Laws

Let us go back to the Cauchy problem (1.1), (1.2) for a general system of
conservation laws. We shall say that a function u : Rd × [0, ∞[−→ Ω is a clas-
sical solution of (1.1), (1.2) if u is a C 1 function that satisfies Eqs. (1.1), (1.2)
pointwise.
An essential feature of this problem is that there do not exist in general
classical solutions of (1.1), (1.2) beyond some finite time interval, even when
the initial condition u0 is a very smooth function. Let us illustrate this fact
by studying the simple case of a one-dimensional scalar equation.

4.1 Characteristics in the Scalar One-Dimensional Case

Thus, assume p = d = 1 and let f : R → R be a C 1 function. We consider

the problem
∂u ∂
+ f (u) = 0, x ∈ R, t > 0, (4.1)
∂t ∂x
u(x, 0) = u0 (x), x ∈ R, (4.2)
and set
a(u) = f  (u). (4.3)
Let u be a classical solution of (4.1) so that we can write (4.1) in noncon-
servative form
∂u ∂u
+ a(u) = 0,
∂t ∂x
28 I. Introduction

and let us introduce the characteristic curves associated with (4.1). They are
defined as the integral curves of the differential equation
dx
= a(u(x(t), t)). (4.4)
dt

Proposition 4.1
Assume that u is a smooth solution of (4.1). The characteristic curves (4.4)
are straight lines along which u is constant.

Proof. Consider a characteristic curve passing through the point (x0 , 0), i.e.,
a solution of the ordinary differential system
⎧
⎨ dx = a(u(x(t), t))
dt
⎩
x(0) = x0 .

It exists at least on a small time interval [0, t0 [. Along such a curve, u is

constant since
d ∂u ∂u dx
u(x(t), t) = (x(t), t) + (x(t), t) (t)
dt ∂t ∂x dt
∂u ∂u
= + a(u) (x(t), t) = 0.
∂t ∂x
Therefore, it follows from (4.4) that the characteristic curves are straight
lines whose constant slopes depend on the initial data, and the characteristic
straight line passing through the point (x0 , 0) is defined by the equation

x = x0 + t a(u0 (x0 )). (4.5)

This important property gives a way to construct smooth solutions.

One sets
u(x, t) = u0 (x0 ), (4.6)
where x0 is solution of (4.5). This is the so-called method of characteristics
(see example 4.1). 
Let us next assume that there exist two points x1 < x2 such that
1 1
m1 = < m2 = .
a(u0 (x1 )) a(u0 (x2 ))

Then, the characteristics C1 and C2 drawn from the points (x1 , 0) and (x2 , 0),
respectively, have slopes m1 and m2 and intersect necessarily at some point P .
At this point P , the solution u should take both values u0 (x1 ) and u0 (x2 ),
which is clearly impossible. Hence, the solution u cannot be continuous at the
point P . Note that this phenomenon is independent of the smoothness of the
4 Weak Solutions of Systems of Conservation Laws 29

t=1

x
0 1
u

u(., 1)
u(., 0)
x
0

Fig. 4.1 Method of characteristics for Burgers’ equation

functions u0 and f . Indeed, using (4.5), we see that the two characteristics
intersect at time t if

t(a(u0 (x1 )) − a(u0 (x2 ))) = x2 − x1 .

Thus, unless the function x → a(u0 (x)) is monotone increasing, in which case
this equation has no positive solution t, we cannot define a classical solution
u for all time t > 0 (see Fig. 4.1). Moreover, one can determine the critical
time T ∗ up to which a classical solution exists and can be constructed by the
method of characteristics; T ∗ is given by
1 d
T∗ = − , α = min a(u0 (y)).
min(α, 0) y∈R dy

The multidimensional case will be considered in Chap. V, Sect. 1.2.

Example 4.1. We want to solve the following Cauchy problem for Burgers’
equation (1.4b) with the initial condition
⎧
⎪
⎨ 1, x ≤ 0,
u(x, 0) = u0 (x) = 1 − x, 0 ≤ x ≤ 1,
⎪
⎩
0, x > 1.
30 I. Introduction

By using the method of characteristics, we can solve up to the time when the
characteristics intersect. We already know by (4.5) that the characteristic
passing through the point (x0 , 0) is given by

x = x(x0 , t) = x0 + tu0 (x0 )

so that ⎧
⎪
⎪ x0 + t, x0 ≤ 0,
⎨
x(x0 , t) = x0 + t(1 − x0 ), 0 ≤ x0 ≤ 1,
⎪
⎪
⎩
x0 , x0 ≥ 1.
For t < 1, the characteristics do not intersect (see Fig. 4.1). Hence, given a
point (x, t) with t < 1, we draw the (backward) characteristic passing through
this point, and we determine the corresponding point x0
⎧
⎪ x − t, if x ≤ t < 1,
⎪
⎪
⎨
x−t
x0 = , if t ≤ x ≤ 1,
⎪
⎪ 1−t
⎪
⎩
x, if x ≥ 1.

It consists of a front moving to the right and steepening until it becomes a

“shock”(see Fig. 4.1). This discontinuity corresponds to the fact that at time
t = 1 the characteristics intersect. 
In short, using the method of characteristics (involving the implicit func-
tion theorem), one can prove that for u smooth enough a classical solution
of (4.1), (4.2) exists in a small time interval. On the other hand, we have
seen that in the nonlinear case a (u) = 0, discontinuities may develop after
a finite time. The above considerations lead us to introduce weak solutions
(which are indeed weaker than the classical solutions!).

4.2 Weak Solutions: The Rankine-Hugoniot Condition

Consider the Cauchy problem (1.1), (1.2), and assume u0 ∈ L∞ d p

loc (R ) , where
L∞loc is the space of locally bounded measurable functions; we want to state
precisely in which sense (1.1), (1.2) is to be taken. Let C10 (Rd × [0, +∞[)
denote the space of C 1 functions ϕ with compact support in Rd × [0, +∞[
(which means that ϕ is the restriction to Rd × [0, +∞[ of a C 1 function
with compact support in an open set containing Rd × [0, +∞[). We begin by
noticing that if u is C 1 and ϕ ∈ C10 (Rd × [0, +∞[)p , we obtain by Green’s
theorem (or integration by parts)
4 Weak Solutions of Systems of Conservation Laws 31

∞  ∂u d
∂ 
− + fj (u) · ϕ dx dt
0 Rd ∂t j=1
∂xj
∞  ∂ϕ 
d
∂ϕ 
= u· + fj (u) · dx dt + u(x, 0) · ϕ(x, 0) dx,
0 Rd ∂t j=1
∂xj Rd

where the dot · denotes the Euclidean inner product on Rp . Thus, any classical
solution u of (1.1), (1.2) satisfies ∀ϕ ∈ C10 (Rd × [0, +∞[)p

∞  ∂ϕ 
d
∂ϕ 
u· + fj (u) · dx dt + u0 (x) · ϕ(x, 0) dx = 0. (4.7)
0 Rd ∂t j=1 ∂xj Rd

Next, we remark that (4.7) makes sense if u ∈ L∞

loc (R × [0, +∞[) . Hence,
d p

we introduce the following definition.

Definition 4.1
Assume that u0 ∈ L∞ d p ∞
loc (R ) . A function u ∈ Lloc (R × [0, +∞[) is called
d p

a weak solution of the Cauchy problem (1.1), (1.2) if u(x, t) ∈ Ω a.e. and
satisfies (4.7) for any function ϕ ∈ C10 (Rd × [0, +∞[)p .
By construction, a classical solution of problem (1.1), (1.2) is also a weak
solution. Conversely, by choosing ϕ in C∞ d p ∞
0 (R ×]0, ∞[) , where C0 (R ×]0,
d
∞
+∞[) is the space of C functions with compact support in R ×]0, ∞[, we
d

obtain that any weak solution u satisfies (1.1) in the sense of distributions on
Rd ×]0, ∞[. Moreover, if u happens to be a C 1 function, then it is a classical
solution. Indeed, let ϕ ∈ C10 (Rd ×]0, +∞[)p ; integrating (4.7) by parts gives
∞  ∂u d
∂ 
+ fj (u) · ϕ dx dt = 0,
0 Rd ∂t j=1
∂xj
so that (1.1) holds pointwise.
Next, if we multiply (1.1) by a test function ϕ ∈ C10 (Rd × [0, +∞[)p ,
integrate by parts, and compare with (4.7), we obtain

(u(x, 0) − u0 (x)) · ϕ(x, 0) dx = 0,

Rd

which yields (1.2) pointwise.

We shall now consider solutions of (1.1) in the sense of distributions that
are only piecewise smooth and therefore admit discontinuities.
First, we notice that the above argument shows that any distributional
solution u is a classical solution of (1.1) in any domain where u is C 1 . We
restrict the study to a particular type of discontinuous function; in fact, this
is not restrictive for the examples that we shall encounter (if one regards the
structure of BV (bounded variation) functions, this simplification is not too
unreasonable (see, for instance, DiPerna [423])). For the sake of brevity, we
32 I. Introduction

say that a function u is “piecewise C 1 ” if there exist a finite number of smooth

orientable surfaces Σ in the (t, x)-space outside of which u is a C 1 function
and across which u has a jump discontinuity. Given a surface of discontinuity
Σ, we denote by n = (nt , nx1 , nx2 , . . . , nxd )T (= 0) a normal vector to Σ and
by u+ and u− the limits of u on each side of Σ,

u± (x, t) = lim u((x, t) ± εn).

ε→0,ε>0

In the one-dimensional case, we assume in addition that any line of discon-

tinuity Σ has a parametrization of the form (t, ξ(t)), where ξ : t → ξ(t) is a
C 1 function from some time interval (t1 , t2 ) into R, and we set in the same
way
u± (ξ(t), t) = lim u(ξ(t) ± ε, t).
ε→0,ε>0

We now show that even in the frame of “piecewise C 1 ” functions, not

every discontinuity is admissible. The values of u and f (u) on each side of Σ
are linked by a relation that is hidden in the equations, as results from the
following theorem.
Theorem 4.1
Let u : Rd × [0, +∞[→ Ω be a piecewise C 1 function (in the above sense).
Then, u is a solution of (1.1) in the sense of distributions on Rd ×]0, +∞[ if
and only if the two following conditions are satisfied:
(i) u is a classical solution of (1.1) in the domains where u is C 1 ;
(ii) u satisfies the jump condition


d
(u+ − u− )nt + (fj (u+ ) − fj (u− )) nxj = 0 (4.8)
j=1

along the surfaces of discontinuity.

The jump relation (4.8) is known as the Rankine-Hugoniot condition.

Proof. Suppose that u is a piecewise C 1 function, solution of (1.1) in the sense

of distributions on Rd ×]0, +∞[. We have already observed that u satisfies
property (i).
Now, let Σ be a surface of discontinuity of u, M a point of Σ, and D a small
ball centered at M (for simplicity, we assume that Σ ∩ D is the only surface
of discontinuity of u in D). We denote by D± the two open components of
D on each side of Σ. Let ϕ ∈ C∞ p
0 (D) . We write

 ∂ϕ 
d
∂ϕ 
0= u· + fj (u) · dx dt = + .
D ∂t j=1
∂xj D+ D−
Another random document with
no related content on Scribd:
contentedly in the thought that Heaven had given me, if only for one
little hour, a lover so loyal, true and brave. Ludovic, my love, my poor
starved heart thanks God for you.”
For an instant the word was at his lips which would have told her his
secret, for, surely, the opportunity was apt. Perhaps it was a feeling
that, in a higher sense, in that atmosphere so fully charged with
tenderness and love, the cold shock of the announcement would be
unfitting; perhaps, too, his sensitive, innate chivalry made him shrink
from taking advantage of that supreme moment. The very certainty
that the stroke must win held him back from making it. Anyhow it
passed, and when rapture allowed him speech it was of a still more
urgent matter, their escape. She told him it was for that she had
risked the message.
“The Baron does not say so, but I know I am destined for Krell. And
once there,” she shuddered, “I may say farewell to my hopes and to
my liberty, except on terms which are now forever impossible.”
He understood, and signified it by a kiss.
“There is no reason, I hope,” he said, “why we should not push on
again for Beroldstein. The longest and worst part of the way has
been travelled, and the end of our journey is now not so far off. With
a couple of hours’ start we could laugh at pursuit, and need not fear
the high roads to-night.”
“Then let us go, dearest,” she urged.
He smiled at the eagerness he loved. “Everything is arranged,” he
replied. “Ompertz is waiting with horses, and will ride with us. I fear,
though, we must leave Countess Minna behind this time. But she is
now safe from this fellow.”
A look of disappointment clouded Ruperta’s face. “Rollmar will visit
my sins on poor Minna’s head.”
“Her penance shall be of short duration, I promise you that,” Ludovic
assured her confidently. “She shall join you in a very few days.
Rollmar is too sensible to take a foolish and futile revenge. Indeed, it
is best; more, it is necessary. We have no horse for her.”
“And Minna hates riding, if you had. Well then, we must leave her. It
is easier now,” she added, with a loving look of confidence.
In a very few minutes preparations for the escape and the journey
were made. Ludovic extinguished the light, and, cautiously opening
the door, crept out, leading the way along the narrow passage, and
down the winding stairs, descending to the outer door by which his
guide had admitted him to the castle. No one was to be seen; the
door was unlocked; they passed out, and crossed an angular court-
yard to a massive stone door set in the outer wall. This, as Ludovic’s
conductor had shown him, was left merely bolted on the inside; at a
strong pull it swung slowly open, and they found themselves in a
passage cut through the rock and leading out into the wood.
Ludovic put his arm round Ruperta to help her along the rough path.
“Now for our faithful Ompertz and the horses,” he said
encouragingly. “He is near at hand. Another hour, dearest, will see
us miles away from this hateful place.”
They were now at the end of the cutting. It was with a delicious
sense of freshness and liberty that Ruperta felt the wind through the
trees blowing on her face. Her lover’s strong arm was round her—in
a few minutes the enemies of her happiness were to be given the
slip. There was just light enough to see the path; a stronger blast of
wind came through the wood, deadening the sound of another rush.
More quickly than they could realize it, they were surrounded by half
a dozen men who had suddenly sprung from their ambush. Before
Ludovic could put his hand to a weapon, he was seized by four
strong fellows, who held his arms firmly, and began to drag him back
to the castle. Ruperta, with all her spirit, was powerless to render him
any help. She herself had been captured by two men who, with less
violence, but equally insistent force, kept her from following.
But the dashing of her hopes, the sickening sense of the Count’s
treachery, made her desperate and reckless. She struggled furiously
with her captors, two tall, evil-looking ruffians who had, however,
evidently had orders to treat her with as much respect as their object
permitted. This was to take her back to the castle by another
entrance; but they found it not so easy. Ruperta resisted vigorously,
then, remembering that Ompertz might be near, she began calling
for help. It was but a faint hope, but, to her joy, she heard an
answering call which was followed by the welcome appearance of
the great dashing swashbuckler, who came through the wood with a
leap and uplifted sword, a very fury to the rescue.
Evidently the men thought so, for it was with no very confident air
that one of them released his hold on Ruperta, and, drawing his
sword, stood before her to keep Ompertz off. A dog might as well
have tried the repel the spring of an attacking lion. With a mighty
sweep his sword was sent flying among the trees, and it was only by
a smart backward spring that he cheated the soldier’s blade of its
second blow.
At the same moment Ruperta found herself free, her other captor
thinking less of his charge than of his skin, which was, indeed, just
then in jeopardy of damage. She quickly told her rescuer what had
happened. He just checked an oath of angry disappointment.
“I told him what to expect,” he said, savagely rueful. “But we both
hoped I might prove a false prophet. Oh,”—he set his teeth
ominously—“oh, for five minutes alone with this precious Count! He
should never tell another lie while I lived, or he.”
Ruperta entreated him to follow her lover and free him. He felt the
urgency of the move, yet hesitated.
“I dare not leave you, Princess, and if we go together”—he gave a
shrug—“I am only one to defend you against this gang of bandits. It
were better to see you into safety first.”
But she would not hear of abandoning Ludovic while there was a
chance of rescue. She too would go back; she had no fear.
Ompertz saw the true courage in her eyes, and no longer opposed
her wish. The two men had skulked away; they were scarcely worth
consideration now. The soldier gave his hand to Ruperta, and, sword
in the other, led her quickly along the passage to the stone door. It
was closed and fast bolted; the men had clearly taken their prisoner
through, and now had him safely lodged. Ompertz gave a kick at the
unyielding barrier.
“No hope of opening that fellow from outside,” he remarked, with a
baffled shake of the head. “And, Highness, let me tell you the sooner
for your sake we get out of this ugly trap the better. We should not
have a chance if these rascals took it into their heads to drop a few
lumps of rock down on ours.”
Although Ruperta had little fear of that awkward contingency, she
recognized the futility of staying there. Her heart was full of
indignation and a terrible anxiety for her lover. But hers was a nature
which rage and fear simply stirred into action; she would never bow
to the inevitable or confess herself beaten.
“Yes. Come back with me quickly,” she said, with sudden resolution.
Ompertz glanced at her and knew that the move was not prompted
by fear, at least for herself. They hurried back along the passage of
rock and into the wood.
“The horses are close by,” Ompertz said, in a tone of doubtful
suggestion.
“That is well; we may want them,” Ruperta replied, and he saw that
she had in her mind a plan of action.
“The Chancellor brought men—soldiers—with him? How many?”
“About eighty.”
“They are near?”
“Hard by, in the forest beyond the valley.”
“That is well,” she said. “I can trust myself to them. I am their
Princess. It is only their leaders who are so vilely treacherous.”
Ompertz looked a little dubious. “If they were all like me, Princess,
you might trust them to the death.”
“And you think I cannot rely upon them to protect me against the
false hearts and lying tongues of the cowards who threaten us? At
least I will try them.”
There was a rustling in the wood, and Count Irromar stood before
them.
 CHAPTER XXVII
AN UNWISE MERCY

“YOU have taken an unfair advantage, Princess, of my willingness to

serve you,” he said, with a dark smile.
“I am again, as I might have expected, the victim of your treachery,”
Ruperta retorted, full of scornful anger.
He made a deprecating gesture. “You must blame me no more now.
The business is out of my hands. The treatment of which you may
complain is not mine. I am no longer a free agent.”
His meaning was as obvious as was its falsehood. Ompertz took a
step forward.
“Free agent or not, Count,” he said bluffly, “I shall make bold to hold
you responsible for the outrage suffered by Lieutenant von Bertheim
at the hands of your men. I was just wishing for an interview with
you.”
The Count was eyeing him full of stern malignity. “And having
chanced upon it, what do you want to say, my fine fellow?” he asked
contemptuously.
The ugly look on the soldier’s face deepened. “Only this,” he
answered threateningly. “That unless you give an instant order for
our friend’s release, this fine fellow will take upon himself to run you
through, and that without delay.”
A streak of moonlight falling through the trees showed a smile of
ineffable scorn on the Count’s strong face. It also glinted on the
barrel of a pistol which he suddenly presented full at the soldier’s
breast.
“Silence, you dog!” he commanded. “You need a lesson in the
manners befitting a lady’s presence. If you speak another word it will
be your last.”
Ruperta sprang between them. “Count, if you harm this man your life
shall pay for it. I swear. I have power that may astonish you before
long. Yes; I will have you hanged if you do not instantly release the
Lieutenant.”
“You are quite mistaken, Princess,” he replied seriously. “The
Lieutenant is not my prisoner.”
“You liar,” she cried, beside herself with indignation at the way he
was playing with her. “You will tell me next he is not in your house, in
your keeping.”
“It is true enough,” he replied coolly. “But I have no power to release
him. Perhaps you have, Highness.”
The sneer was worthy of him; he had come to hate this woman
whom he might not love.
“We shall see,” she returned. “You refuse?”
“I fear I must—even at the risk of the penalty which your Highness
has foreshadowed.”
“Very well, then,” she said. “You shall see how I will keep my word.
Come, Captain.”
She turned to Ompertz and prepared to move away.
“Permit me to escort you back to his Excellency,” Irromar said. “He
charged me to look after you, and my responsibility is strict.”
“Your responsibility!” she echoed scornfully. “Surely, Count, you have
forfeited any claim to that I will never enter your abominable den
again.”
“It is most unfortunate,” he replied, with a somewhat mocking show
of apology, “that I should have to bear the brunt and odium of your
Chancellor’s actions. Surely, Princess,” he continued, as though
urged merely by his innate love of setting his actions in a false light,
“you must be aware that it was a risky thing to attempt to continue
your elopement under the Baron’s very eye; an eye which looks not
too favourably on the Lieutenant’s pretensions. I should certainly
have warned you against any such mad attempt, had I not thought
that your good sense made it unnecessary.”
Ruperta turned from him, disdainfully impatient. “I cannot discuss the
matter with you, Count, especially as I have good reasons for
believing no word you say.”
He gave a shrug. “It is most unfortunate, I must repeat, this
persistence in imagining my ill-will. As for your interest in the
Lieutenant’s welfare, I can only refer you to Baron Rollmar, to whom
it is now my duty to conduct you.”
He advanced to her with outstretched hand. She shrank from him.
Ompertz whispered a word to her as he fell back a pace. These
movements altered the relative positions of the three. Ruperta had
scarcely caught the soldier’s whisper, but she was quick-witted
enough to divine his intention. She suffered Irromar to lay his hand
on her arm. It gave her an excuse for struggling—to make a sudden
clutch at the hand which held the pistol. Simultaneously Ompertz
gave a swift spring, and, as Ruperta’s hold hampered the Count from
turning to meet his attack, seized him from behind and got his arm
tightly round his neck.
Irromar was a very Hercules, but now he was taken at a
disadvantage, and Ompertz was of strength far above the average. It
was a fierce joy to him to find his muscles round that lying throat,
and in a very few seconds he had the Count half-throttled on the
ground. Then the pistol was wrested away, and their enemy lay at
their mercy.
“Now let me put an end to the villain,” Ompertz gasped, as with
fingers gripping the Count’s throat and knee pressing on his chest he
held out his hand for the pistol.
 “Ruperta’s hold hampered the Count from turning to meet
his attack.”
Page 276.
But Ruperta refused. Perhaps the livid, distorted face showed her
too vividly the horror of such a midnight deed, and obscured the
sense of expediency.
“No,” she objected. “We cannot. He must not die here—like this.”
“Then you give Lieutenant von Bertheim’s life for his,” Ompertz
urged, bitterly baulked. “In Heaven’s name, let me put a bullet
through his lying brain, and do a good deed for once.”
But she would not consent. “If he swears on his honour that he will
release the Lieutenant, his life shall be spared,” she said.
Ompertz groaned at the throwing away of this chance. “His honour!
You will repent it if you trust to that,” he said, as he tightened his grip
on the Count’s throat, since he might not shoot him.
But Ruperta saw his intention, and insisted that he should relax his
hold. “You hear, Count?” she said.
“I swear,” he gasped.
“Of course he swears,” growled Ompertz.
For some moments Irromar lay panting; the soldier looking down on
him with a grim hankering that was almost comic. Suddenly, from a
position in which most men would have been helpless, the Count,
who seemed one compact mass of muscle, contrived by a
convulsive effort to throw himself on his side, and a desperate
struggle began. The suddenness of the effort had taken Ompertz by
surprise, and so at some disadvantage. Still, he welcomed the
renewed struggle, since it gave him an excuse for shooting. But
once, when he might have fired with deadly effect, he hesitated
through fear of hitting Ruperta who had seized one of the Count’s
arms, and then, when he did fire, the bullet seemed to take no effect
at all. With an exclamation of disappointment, he dropped the pistol,
and set himself to grapple in deadly earnest with his formidable
adversary. But great as was his strength, it was pitted now against
one of the strongest sets of muscles in Europe. Little by little the
Count got the advantage, he was a skilful wrestler and knew all the
tricks of that art, so that not even Ruperta’s weight hanging on to his
arm made the struggle evenly balanced. Before long he was able to
force Ompertz backwards and, by a dexterous twist, to spring clear
of him. It was only just in time, for Ruperta had taken Ompertz’
sword, and was only hesitating to use it from fear of striking the
wrong man as they swayed and turned in their desperate encounter.
Now the Count was free. “Quick! the sword!” Ompertz cried, as he
recovered his balance and sprang to her for the weapon. There was
a loud laugh of mockery, and, almost before Ompertz had turned to
rush after him, the Count had disappeared in the darkness. Sword in
hand, the soldier followed as best he could, only to be brought up
very soon by the manifest hopelessness of the pursuit and the fear
of missing the Princess. To her he returned, baffled and fuming.
“I said you would regret it, Highness,” was his reproachful greeting.
She was pale and trembling slightly from the excitement. “It cannot
be helped,” she replied, with a touch of authority. “I am sorry for your
sake, but I could not have the man, whatever his crimes, done to
death like that.”
“He has the devil in him,” Ompertz exclaimed wrathfully. “Now
between him and the Chancellor, who has the infernal touch too, I
fear, you may say good-bye to the chance of getting the Lieutenant
free. And I had my prayer answered and my fingers round that
villain’s throat. It was wicked to fling away the chance.”
“Yes, I am sorry now,” Ruperta agreed, showing not half the intense
regret she felt. “But I am not going to submit myself tamely as a
victim to these outrages and false dealings. I am going to
Beroldstein.”
“You, Princess? To Beroldstein?”
“Alone,” she answered resolutely. “I will appeal to the King of Drax-
Beroldstein, since the Duke of Waldavia, my own father, cannot help
me.”
“But the King of Drax-Beroldstein,” Ompertz objected, “is not
Ludovic, but Ferdinand.”
“So much the better,” she returned. “It makes my task less
disagreeable and scarcely more doubtful.”
He recognized the hideous complications which made her plan so
hopeless, yet he saw no sufficient reason for breaking his pledge of
secrecy. After all, Ludovic’s release was the great thing to try for; in
the interests of that, the less known of his identity the better.
“I may go with you, Princess? The horses——”
“No,” she replied. “I should like your escort, but cannot take you
hence. It will be something for me to know that one trusty heart is left
near Ludovic. But I fear. What can you do for my Ludovic against
those cruel villains, the Count and Rollmar?” She turned away in an
access of heart-chilling despair, then next moment had recovered
herself.
“Come, let us not lose another instant,” she said resolutely. “You
must find me an escort among the soldiers. Surely there are some
who will run this risk for their Princess, for any woman, indeed, who
is in such a dire strait as I.”
He told her of certain good fellows there whose acquaintance he had
made in the guard-room, and who, he was sure, would be ready to
risk their lives in this service for her.
“If all goes well, they shall not be losers for standing by me in my
extremity. At least they are human; Rollmar is a fiend.”
They came to the three horses—bitter suggestion of their failure—
mounted, and made their way towards the spot where the men were
encamped. Ompertz’s thoughts were divided between admiration for
this courageous girl and sadness at the thought of how small was
her chance of success.
But the affair, he told himself, was too difficult for his poor brain; he
could see no light through the darkness; only hope that chance, after
leaving them so terribly in the lurch, might once again stand their
friend and accomplish what seemed beyond the scope of every
imaginable plan.
By a difficult path they arrived presently, after many a hindrance from
wood and rock, within a stone’s throw of where the troops lay
encamped. Leaving Ruperta in a place of safety, or, at least, in
concealment, Ompertz went forward to find his men for the purpose.
Half an hour later he, with many misgivings, had taken leave of the
Princess who, with an escort of three stout fellows, started off
through the forest to strike the nearest point of the main road to
Beroldstein. Ruperta had supplemented Ompertz’s explanation by
an appeal to the men to stand by her in her distress. She knew, she
said, the risk her escort would be running; how those who guarded
her flight would do so at the peril of their lives, and she would accept
no service that, with this knowledge, was not freely given. But
Ompertz, a shrewd judge of, at any rate, certain characters, had
made no mistake in choosing the men. Their records were not,
perhaps, of the best repute, but they were three staunch dare-devils,
who would think no more of giving up their prospects and lives at a
word from the Princess than of passing their mug of beer to a thirsty
comrade. They had instantly and heartily sworn to see her through
her long ride, or give their lives in her service, and she felt she need
have no fear of their failing her. So they set off.
The first part of the journey was slow and difficult enough; however,
one of the men knew the country and was confident that they could
not lose their way. Nevertheless, the darkness of the forest
hampered their progress, but, with the dawn, the track, too, grew
lighter as the party emerged upon a hilly stretch of heath.
“We are now but a mile from the great road,” said the man who knew
the way.
They could push on now at a smart pace; time, Ruperta felt, was
everything, and all through the long hours of darkness her
impatience had been torture. It was not many minutes before the
broad coach-road came in sight beyond a belt of woodland which
fringed it. Just before they reached it, hastening over the grassy
road, one of the men, who was riding a few paces ahead, held up a
warning hand.
As they reined up, the ring of horses’ hoofs fell upon their ears. The
man quickly threw himself from the saddle and crept forward to the
corner whence he could get a view of the road. Next instant he came
rushing back, motioning them to turn aside among the trees.
“Horsemen coming fast! Quick! They may be after Her Highness.
Quick, under the trees!”
They had scarcely taken cover, when the other party rode by at a
quick pace. Four men, with a fifth at their head, riding in haste and
looking neither to the right nor left. The figure of the leader was
unmistakable.
“It is Count Irromar,” Ruperta exclaimed under her breath. “In pursuit
of me.”
She was wrong. It was the Count, but he was not in search of her.
He was riding post-haste to Beroldstein on business of his own.
 CHAPTER XXVIII
AT THE USURPER’S COURT

IT was with considerable surprise that King Ferdinand of Drax-

Beroldstein, as yet scarcely settled comfortably into his snatched
dignity, heard that the notorious law-defier and outlaw, Count
Irromar, was at the palace, asking for a private audience on business
of the utmost importance. Had the King been a strong man, or one
who felt his position unassailable, he would probably have handed
the noble brigand over to his officers of justice, congratulating
himself on getting the most troublesome and dangerous of his
subjects so cheaply in his power. But Ferdinand was neither. He was
a weak man who had been unable to resist the chance, urged upon
him by designing favourites, to seize a crown which for the moment
seemed to be left without a wearer, and, having put it on his head,
was now trembling inwardly at his own temerity. He could afford to
despise no man, and his only strength came not from within, but was
forced on him by circumstances from without. It was almost a weak
man’s strength of desperation; no one can be so strong by fits and
starts as your thoroughly feeble character who dare not show his
weakness.
Then there was the haunting mystery of Ludwig’s disappearance. At
every waking moment, Ferdinand told himself that his cousin was
surely dead, but in his dreams, he was alive and seeking retribution.
In spite of the assurances of all his friends and flatterers, Ferdinand
found himself doubting every one, from his ministers to the soldiery.
He dreaded to read in every new-comer’s face the solution of the
mystery, the end of his day. Still, he had cast his die, the boats were
burned behind him—foolishly, he told himself, since he might, by
constituting himself regent, have grasped the power clean-handed—
and now, as it was, there seemed nothing for it but to assume a
resolution which he had not, and to keep by force what treachery
had won. It had all seemed so easy and desirable, this pursuit of
power, this scheming for a throne, in the days of preparation; when
suddenly the coup had to be made, and responsibility to be
assumed, it was not so pleasant.
Doubtless it was a shrewd knowledge of the usurper’s character that
gave Irromar confidence to put his head into the lion’s mouth. At the
same time, he was well armed, both for attack and defence, with the
knowledge he held.
On receiving the somewhat astounding message, Ferdinand
hesitated. His first impulse was that of the bully; to order the arrest of
this formidable outlaw. Then his chronic feeling of insecurity
prompted him to hear what the visitor had to communicate. Such a
man had not come boldly there without good reason, and he could
easily be arrested after the interview. Accordingly, he gave orders for
a guard to be in readiness and for the Count to be admitted to an
audience.
With an affectation of homage which scarcely concealed his bold
confidence, Irromar entered the royal presence, and, having bowed
low, stood before the Usurper in the easy fearlessness of conscious
power. Ferdinand had a set frown on his sharp, gambler’s face; he
might as well have thought to melt a rock by frowning at it, as
thereby to intimidate the strong, reckless nature confronting him.
Perhaps he felt this, as, with an effort at self-assertion, he bid the
Count say what had brought him thither.
“I have come on a matter which is for your Majesty’s ear alone,” was
the sturdy reply.
Ferdinand affected to hesitate, then motioned his curious circle to a
distance. “Now speak out, Count, and briefly.”
But Irromar dropped his strong vibrating voice almost to a whisper,
as he bent forward to the King. “It is of your Majesty’s cousin, Prince
Ludwig, that I have come here to speak!”
He watched closely the effect of his words, and saw nothing but a
curious, indefinable expression flash across his hearer’s face. But it
was enough. And although Ferdinand’s next remark was made in a
tone of studied indifference, Count Irromar knew that the hit was
more than a touch.
“Well? You know, perhaps, what has become of him? His fate?”
Irromar bowed assent. “He is at this moment in my power: a prisoner
in my castle in the Teufelswald.”
If the news gave Ferdinand an uncomfortable thrill, he did not show
it. The pale face, with its stiff yellow moustache and beard, remained
impassive. Only, in the eyes there was a light of fierce concern.
Perhaps, after all, the knowledge that one phase of his uncertainty
was at an end came as a relief to him.
“Well?” Ferdinand had now to use his cunning; he would let
suggestions come from the other side.
“I thought,” the Count answered readily, “that the information might
be of vital interest to your Majesty.”
“In what way, Count?”
“It is not for me to dictate the use your Majesty should make of it.”
His guard was good; it would have to be drawn out and weakened.
“And yet I dare be sworn,” Ferdinand returned, with his cunning
smile, “that you had a use for it in your mind, or you would hardly
have ventured hither.”
Irromar understood the invitation. “Perhaps, sire, a use which may
be to the advantage of both of us,” he replied coolly.
Ferdinand was leaning sideways in his chair, with his hand playing at
his sparse beard; it was a demeanour of sly reserve. “We should like
to have your views, Count, as to this double advantage,” he said.
“Certainly, sire,” Irromar replied, playing his part with every outward
sign of deference. “You will, perhaps, graciously pardon me if I
express them too bluntly; but the position and opportunity are critical,
and plain speaking fits them best.”
Ferdinand gave a quick, impatient nod of authority, and the Count
proceeded.
“The Prince, is, as I have said, my prisoner, secretly hidden away
where no man, unless I choose, can ever find him. He fell into my
hands by an accident, and the fact is practically a secret which need
never be known, save to those whose interest would be to ignore it.
To all intents, he is dead and buried. It is for your Majesty to say
whether he shall ever come to life again.”
He paused. “Go on,” Ferdinand said curtly.
“As to your Majesty’s interest and wishes in the matter,” Irromar
continued, in the same tone of guarded deference, which yet
seemed to mock as it flattered, “I do not presume to make a
suggestion, or anticipate what may be in your Majesty’s mind. All
that I wish to put forward is my hearty willingness to serve you, sire,
in this matter. And, that you may trust me.”
Ferdinand, revolving keenly the crisis, smiled with a purposeful scorn
which hid the inner working of his mind. “Confidence in Count
Irromar is a somewhat unreasonable demand, methinks,” he
observed.
“Without a guarantee, yes?” was the ready rejoinder. “It suggests the
second and minor advantage of the situation; that which affects my
poor self.”
“Ah?” Ferdinand was indifferently curious. Perhaps he felt he could,
if expedient, secure that guarantee without the Count’s active co-
operation.
“The very disrepute of my antecedents,” Irromar went on, with the
confidence arising from a strong position, “is, although it naturally
appears to the contrary, the very guarantee for my liberty. Your
Majesty is justly incredulous; but let me explain away the apparent
absurdity. In a word, I am sick of my present outlawry, legal and
moral. My one great desire is to rehabilitate myself, to take up once
more the position to which I was born, and which, in my hot-headed
madness, I chose to throw away. There is but one hand from which I
can hope to receive back what I have squandered, the good name,
the noble position; but one countenance to which I can look for
pardon and favour. If once that hand is held out, that countenance
turned favourably towards me, am I likely to reject that royal
generosity and return to my dog’s life? Now, sire, have I made my
meaning plain?”
“You have—quite plain,” Ferdinand answered. Then he paused, his
manner seeming to command silence on the other’s part as well.
Once or twice he glanced sharply at the Count’s face, that strong,
keenly determined face. He was scheming rapidly, vaguely,
uncomfortably. The crisis for which he had been preparing himself
was, now that it had suddenly arisen, rather more than he could
confidently meet. And his discomposure was due less to the urgency
of the situation than to the manner of its announcement, and, above
all, to the man who set it so boldly before him. For during the whole
interview he had been oppressed and irritated by the sense of his
inferiority to the Count, an inferiority none the less galling in that it
was of evil; such better qualities as they may have possessed did
not enter into the question. This man’s personality and character
were dominant; their owner looked down from a higher plane of evil
upon the weak tool of political intriguers, seated uneasily on his
stolen throne.
But, apart from purely personal considerations, the manifest
superiority forced this question upon Ferdinand. Would it be wise for
him to put himself in the power of this resolute, cunning spirit? The
Count’s argument was plausible enough, but what deep scheme
might not lurk behind it? Had Irromar shown himself a weaker man,
Ferdinand would probably have employed him to put his awkward
cousin out of the way, and then taken the obvious means of securing
his ever-lasting silence. But, somehow, as he looked at his visitor
and mentally gauged him, he could not see in him an easy victim.
Still, for the moment, power was on the King’s side, only he must,
indeed, be careful how he let it slip away. At any rate, the matter was
too difficult for an off-hand decision; he would take counsel with a
more astute mind than his own; as it was, he and this master-spirit
were unevenly matched. And in the meantime he would gratify and
avenge his wounded vanity by showing his power.
So, with a deepening frown, he at length broke the tense pause.
“You are a bold man, Count, to come here and make this proposition
to us. For what may have prompted you to this temerity, the wild life
you have led may, perhaps, be responsible.”
Both men gave a smile, and the Count’s produced the effect which
the King’s vainly intended.
“Nobody,” Ferdinand continued, “but yourself would have conceived
so bold a step. No one in any but our position would have seemed to
invite it.”
“Your Majesty will hardly blame me for seizing a chance so
momentous to both,” Irromar returned, bluffly.
“At least,” Ferdinand replied, guardedly, “we cannot blame you for
hastening to impart to us news so important. That may weigh with us
in the view we shall take in our judgment of you.”
The Count was quick enough to see the line Ferdinand was taking,
and, with the impetuosity of a strong, impatient nature, he set about
brushing aside the barrier of shuffling behind which the King was
entrenching himself.
“There is scarcely time or room for the question of judgment to come
in, sire,” he said, emphatically. “I am a man of action, accustomed to
go straightway to the point at issue. This matter clearly admits of no
temporizing. Your Majesty’s judgment of me is at the moment of little
consequence. My all-important quality is that I am the jailer of the
one person in the world whose condition must supremely affect your
Majesty’s welfare.”
“That,” replied Ferdinand, with a purposeful show of scorn, “is a
matter upon which we do not invite your opinion. The King of Drax-
Beroldstein must not be dictated to by the outlaw of the Teufelswald.”
The Count flushed purple. “The King——,” he began hotly, then
checked the words at his lips. Doubtless he saw Ferdinand’s object
in provoking him, and resolved to meet him at his own game. “I
should be the last man to presume to usurp the functions of your
Majesty’s advisers,” he said, with a significant smile, “or interfere,
unbidden, with aught that concerns you. I fear that already, in my
zeal, I may have been guilty of officiousness. Is it, then, your
pleasure, sire, that I set Prince Ludwig free?”
Ferdinand had settled his course, and, that once accomplished,
could keep to it firmly enough. “That,” he answered, with an
assumption of dignity, “is a question for our advisers. It is not to be
determined in a moment, certainly not at the suggestion of Count
Irromar. We are not unmindful of your zeal, Count, and shall take it
into consideration in dealing with you. But for the moment we must,
as you will understand, at least make a show of doing our duty. You
have set our laws at defiance, you have been the very scourge of a
wide district of our kingdom. You”—and here a peculiar sneering
smile spread over his face—“you, who have taken upon yourself so
boldly to advise us, will recognize that we cannot afford to reward
your long list of black deeds with immediate tokens of our favour. It
would raise an easy and hideous suspicion. It would at once brand
us as our cousin’s murderer. No! Policy of State must stand before
all things, and that policy demands your arrest.”
All through the speech Irromar’s face had been growing darker, and
at the last word he made a swift gesture of rage.
“Arrest? Your Majesty is joking!”
It was all he could say, but there was clearly no jest in Ferdinand’s
crafty face as he signed to the group that, in scarcely veiled curiosity,
stood apart. He had given his orders, and the men were ready. At a
word from an alert official, Count Irromar, inwardly raging, and
frowning threats, found himself surrounded and a prisoner.
“Your Majesty,” he cried darkly, “will do well to consider this step you
are taking.”
Ferdinand waved his hand with a gesture of dismissal. “We will see
you again, Count; you understand?” he said significantly, as he rose
and walked away.