Moving Shape Analysis and Control
Applications to Fluid Structure Interactions

PURE AND APPLIED MATHEMATICS
A Program of Monographs, Textbooks, and Lecture Notes

EXECUTIVE EDITORS Earl J. Taft Rutgers University Piscataway, New Jersey Zuhair Nashed University of Central Florida Orlando, Florida

EDITORIAL BOARD M. S. Baouendi University of California, San Diego Jane Cronin Rutgers University Jack K. Hale Georgia Institute of Technology S. Kobayashi University of California, Berkeley Marvin Marcus University of California, Santa Barbara W. S. Massey Yale University Anil Nerode Cornell University Freddy van Oystaeyen University of Antwerp, Belgium Donald Passman University of Wisconsin, Madison Fred S. Roberts Rutgers University David L. Russell Virginia Polytechnic Institute and State University Walter Schempp Universität Siegen Mark Teply University of Wisconsin, Milwaukee

MONOGRAPHS AND TEXTBOOKS IN PURE AND APPLIED MATHEMATICS Recent Titles
I. Graham and G. Kohr, Geometric Function Theory in One and Higher Dimensions (2003) G. V. Demidenko and S. V. Uspenskii, Partial Differential Equations and Systems Not Solvable with Respect to the Highest-Order Derivative (2003) A. Kelarev, Graph Algebras and Automata (2003) A. H. Siddiqi, Applied Functional Analysis: Numerical Methods, Wavelet Methods, and Image Processing (2004) F. W. Steutel and K. van Harn, Infinite Divisibility of Probability Distributions on the Real Line (2004) G. S. Ladde and M. Sambandham, Stochastic versus Deterministic Systems of Differential Equations (2004) B. J. Gardner and R. Wiegandt, Radical Theory of Rings (2004) J. Haluska, The Mathematical Theory of Tone Systems (2004) C. Menini and F. Van Oystaeyen, Abstract Algebra: A Comprehensive Treatment (2004) E. Hansen and G. W. Walster, Global Optimization Using Interval Analysis, Second Edition, Revised and Expanded (2004) M. M. Rao, Measure Theory and Integration, Second Edition, Revised and Expanded (2004) W. J. Wickless, A First Graduate Course in Abstract Algebra (2004) R. P. Agarwal, M. Bohner, and W-T Li, Nonoscillation and Oscillation Theory for Functional Differential Equations (2004) J. Galambos and I. Simonelli, Products of Random Variables: Applications to Problems of Physics and to Arithmetical Functions (2004) Walter Ferrer and Alvaro Rittatore, Actions and Invariants of Algebraic Groups (2005) Christof Eck, Jiri Jarusek, and Miroslav Krbec, Unilateral Contact Problems: Variational Methods and Existence Theorems (2005) M. M. Rao, Conditional Measures and Applications, Second Edition (2005) A. B. Kharazishvili, Strange Functions in Real Analysis, Second Edition (2006) Vincenzo Ancona and Bernard Gaveau, Differential Forms on Singular Varieties: De Rham and Hodge Theory Simplified (2005) Santiago Alves Tavares, Generation of Multivariate Hermite Interpolating Polynomials (2005) Sergio Macías, Topics on Continua (2005) Mircea Sofonea, Weimin Han, and Meir Shillor, Analysis and Approximation of Contact Problems with Adhesion or Damage (2006) Marwan Moubachir and Jean-Paul Zolésio, Moving Shape Analysis and Control: Applications to Fluid Structure Interactions (2006)

Moving Shape Analysis and Control
Applications to Fluid Structure Interactions

Marwan Moubachir
INRIA, France

Jean-Paul Zolésio
CNRS and INRIA, France

Boca Raton London New York

Series. QA612. please access www. Marwan.copyright.(Monographs and textbooks in pure and applied mathematics . 978-750-8400. -. J.com Taylor & Francis Group is the Academic Division of Informa plc. I. No part of this book may be reprinted.crcpress. Government works Printed in the United States of America on acid-free paper 10 9 8 7 6 5 4 3 2 1 International Standard Book Number-10: 1-58488-611-0 (Hardcover) International Standard Book Number-13: 978-1-58488-611-2 (Hardcover) Library of Congress Card Number 2005053196 This book contains information obtained from authentic and highly regarded sources. Suite 300 Boca Raton. For organizations that have been granted a photocopy license by the CCC. transmitted.com (http://www.Published in 2006 by Chapman & Hall/CRC Taylor & Francis Group 6000 Broken Sound Parkway NW. FL 33487-2742 © 2006 by Taylor & Francis Group. and sources are indicated.S. including photocopying. or utilized in any form by any electronic. A wide variety of references are listed. but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. Fluid-structure interaction--Mathematics. Zolésio. Inc.24--dc22 2005053196 Visit the Taylor & Francis Web site at http://www.copyright.P. For permission to photocopy or use material electronically from this work. and are used only for identification and explanation without intent to infringe. and recording. III. p. (CCC) 222 Rosewood Drive.M68 2005 514’. MA 01923. a separate system of payment has been arranged. or in any information storage or retrieval system. and the CRC Press Web site at http://www. without written permission from the publishers. cm. LLC Chapman & Hall/CRC is an imprint of Taylor & Francis Group No claim to original U.taylorandfrancis. mechanical. reproduced. Reasonable efforts have been made to publish reliable data and information.com/) or contact the Copyright Clearance Center. Moving shape analysis and control : applications to fluid structure interactions / by Marwan Mouchabir and Jean Paul Zolesio. microfilming. Library of Congress Cataloging-in-Publication Data Mouchabir. now known or hereafter invented.com . Trademark Notice: Product or corporate names may be trademarks or registered trademarks. Title. Danvers. ISBN 1-58488-611-0 (acid-free paper) 1. Reprinted material is quoted with permission. or other means. II. Shape theory (Topology) 2.7. 278) Includes bibliographical references and index. CCC is a not-for-profit organization that provides licenses and registration for a variety of users.

Olivia et Timoth´. e e A Monique. .A Val´rie.

.

. . . . . . . . . . . . . The adjoint state and the adjoint transverse field . .2 2. . .1 Optimality conditions : case of planar parametric curves 76 3.4 2. . . . . . . . . 1. . .2 Weak convection of characteristic functions . .3 2. 3. . . . . . 3.4 Detailed overview of the book . . . . .3 Plan of the book . . . . . . . . .3. . . . . . .3. 1. .3 Tube evolution in the context of optimization problems . . . . . . . . . . 3. . . . .4. . . . . . . . .2 Parabolic version of Helly’s compactness theorem . . . . .4. . . .Contents 1 Introduction 1. The eulerian derivative and the transverse field . 3. . The inverse problem setting .3 Classical shape analysis for tubes . . . . . . 1 3 4 6 7 8 13 19 28 33 33 35 37 43 44 46 49 49 50 50 51 54 56 57 60 66 67 69 72 75 introductory example : the inverse Stefan problem The mechanical and mathematical settings . .2 Fluid-Structure interaction problems . . . . . 3. 3. . .1 Characteristics versus Eulerian flow setting . . . . . . . . . . . . . . . . . . . . . . . . . . . 1. . 1. .1 Introduction . .4. . . . . . . . . . .4. . . . . 78 ix . . . . . . . .4. . . . . 3. .4 Tube derivative concepts . .5 A first example : optimal trajectory problem . . . .3 Generation of clean open tubes .6 . . . 3. . . . . . 3. . .5 2. . . . . . . . . . . . The Eulerian partial derivative of the state . . . . . . . . . 3. . . . . . . . . . .3. flow . 3. . . . . . . . . . . . . . . .2 Optimality conditions : case of the general Eulerian setting . . . .1 The convection equation . . . . . . . . . . . . . . .5. . . . . . . . . . . . . . . . .2 Shape design of a fixed solid inside a fluid flow 1.1 Classical and moving shape analysis . . . 2 An 2. . . . . . . . 1. . . . .1 Penalization using the generalized perimeter’s time integral . . .3 Dynamical shape design of a solid inside a fluid 1. . . .5. .4 Control of an elastic solid inside a fluid flow . . . . .2. . .2 Tangential calculus for tubes . . . . . . .4. . . . . . . . . . . . The Eulerian material derivative of the state . 3 Weak evolution of sets and tube derivatives 3. . . . .2 The Galerkin approximation . . .2. . . . . . . . . . . . . . . 3. . . . . . . . . . . . 3. . . . . .1 2. . .1 Control of a fluid flow around a fixed body . . . . .4. . .

. . . . . . . . . . . . . 5. . . . . . . 4.3 Tangential Stokes formula . . . . . . . . .5. . . . .3 Iterative Scheme . . . .5. 5. . . . . . . . . . . .4. . . . . . . . . . . . .2 Solutions without step . . . 4. . . . . . . . . . . . 4. . . . . . . . . . . . .x 4 Shape differential equation and level set formulation 4. . . . . . .5 State derivative strategy . . 4. .1 The Laplace equation . 5. . . . . . . .3 Differentiability of the saddle point problem .2 A constructive algorithm . . . . .3 An asymptotic result in the 2D case . . . . . . . . . . . . . . . .1 Introduction . . . . . . 5. . .5.5.2 Function space parametrization . . . . . . . . . . . . . . . 4. . . . . . . .7 The level set formulation . . . . . . . . . . . . . . . . . .6. . . . .1 Saddle point formulation of the fluid state system . . . . . . . . .2 Classical shape differential equation setting . . . . . . . . .1 Weak formulation and solvability . . . . .2 Intrinsic tangential calculus . . . . .1 Introduction . . .3. . . .3. .5 Shape differential equation for the Laplace equation . . . . . . . . . 83 83 84 86 86 88 89 90 90 94 96 97 101 101 102 103 105 109 109 110 114 114 119 119 120 122 122 123 124 133 136 138 140 141 146 147 157 157 159 162 . . . 5. . .5. . 5. . . 5. 5. . . . . . . . . . . . . . . . . . . . . .6. .2 The shape control problem . . . .4 The asymptotic behaviour . . . 4. . . . . . . . . . . . 4. . . . . . . . . . . . . . . . . .5 Adjoint system and cost function shape derivative . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. . . 4. . . . . . . . . .6. .2 The Lagrange functional . . . . . . . . . 5.5. . . . .4 Extractor Identity . .1 Introduction . . . 5. . . .3 Shape derivative . . . . . . . . . . .8 Conclusion .7 Min-Max and function space embedding . . . . . . .7. . . .3. . . . .7. . . . . . . 5. . . . . 5 Dynamical shape control of the Navier-Stokes equations 5. 4. .7. . . 5. . . . . . . . . . . . . . . .7. 5. 5. . . . . . . . . . . . . . . . . . . . .3 The shape control problem . . . . . . . . . . . . 5. .4. . . . 4. .1 Oriented distance function . . 5. . . . .1 Non-cylindrical speed method . . .6 Min-Max and function space parametrization . . . . 5. . . . 5. . . . . . . . . . 5. . . . .4.1 Saddle point formulation of the fluid state system . . . . . .2 Problem statement . . . . .4 An example: the transverse magnetic like inverse problem . . . . . . . . . . . . . . . . . . . . . . . . .3 Elements of non-cylindrical shape calculus . . . . . . 4. . .2 The weak Piola material derivative . 4.6 Shape differential equation in Rd+1 . . . . .4 Elements of tangential calculus . . . . . . . . . .5. . . . . .5. . . . .1 An existence result for the shape differential equation 4. . . . . . . . . . . . . . 4. . . . . . . . . . . 4.7. . . . . . . . . . . . . . . . . . 5. .7. . .

. .4. . . .2 Solid adjoint system . . 215 8. . .2 Mathematical setting . . . . . . . . . .4. .3 Solid reduced order and solid weak state operator . . . . . . . . . . . . . . . . . . . .3. . .6 Eulerian vs. . . . .2 Evolution maps . . . 221 8. . . . . . . .1 Fluid adjoint system .2. .2. . . . . . . . . . . . . . . . . . . .2 Mechanical problem and main result . . . . . . . . . 7. . . . . . . . . . . . . . 220 8. . .4. . . . . . 171 6. . . . . . . . . . . . . 7. . .1 Derivatives with respect to state variables . . . . . . . .4 Inverse problem settings . . .2. . . . . . . . 163 6. 7. . . 215 8. . . . . . . . . .1 Introduction . . . . . . . .3. . . 173 6. . . . . . . . . . .2 Fluid state operator . . . . 180 6. . . . . . . . . . . . . . . . . . .3 KKT optimality conditions . . . . . . . . .3 Elements of shape calculus . . . . . . . . . . . . . . . . . . . . . . . . 7. . . . . . . . 7. . . . .5 KKT Optimality Conditions .2. . . . . 163 6. . . . . . . . . . . . . . . .2. . . . .6 Min-Max problem . . . . . .4 Fluid state operator . . . . . . . . . . . . . . . . . . . . . . . .xi 6 Tube derivative in a Lagrangian setting 163 6. . . . . . . . . . . . . 182 7 Sensitivity analysis for a simple fluid-solid interaction system 7. . . . . . . 224 . . .2 Free divergence and non-homogeneous Dirichlet boundary condition constraints . .3 Navier-Stokes equations in moving domain . . . . 222 8. . . . . . . . . . . . . . . Lagrangian non-cylindrical derivative . . .1 Transpiration boundary conditions .5 Shape optimization under state constraints . . 178 6. . . . . . . . . . . . . . . .2. . . . 7. . . . . . . . .4 Min-Max problem . . .2. . . . . 7. . . . . 175 6. . . . . . . . . . . . . . . . . .3. .4. . 220 8. .4 Unconstraint non-cylindrical shape optimization .2. 178 6. . . . . . 7. . . .3. . . 7. .1 Basic differentiation results .2. . . 7. . . . 7. . .6 Conclusion . . . . . . . . . . . . . . . . . . . .3. . .1 Introduction . .5. .5 Coupled system operator .4. . . . . . . . 165 6. . . . . . . 7. . . . .3 Well-posedness of the coupled system . . . . . . . . . . . . 185 185 185 189 190 191 191 192 193 194 195 197 197 198 212 8 Sensitivity analysis for a general fluid-structure interaction system 215 8. . . . . . . . .2 Mathematical settings . .3 Coupled system operator . . . . . . . 223 8. . . . . . . . . . . . . . . . . . . . . .1 Introduction .2.2 Optimal control of the Navier-Stokes system by moving the domain . 165 6. 7. . .1 Analysis strategy . . .4. .5. . . . . . . . . . .2 Solid adjoint system . . . . . . . 164 6. . . . . . . . . . . . . . . . .1 Solid weak state operator . . . . . . . . . . . 223 8. . .3 Comparison with the Eulerian derivative version . . . . . . . . . . . . . . . . . . . .

. . . . . . . D. . . . . . . . . . . . . . . .4 C. . . . B. . . The space of tempered distributions . . . . . . . . . . . . . . . . . . . . . . . . . A. . .4 Sobolev embedding theorems . . . . . . . . . . . . . . . . . . . . . A. . . . . .1 Classical functions . .1 Various inequalities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Completeness. .6 C. . . . . . . . . . . . . . . . . . . . D. . . . . . . . . + D. . . . . . . Fourier transform in S .3. . . .1 Spaces H m (Ω) . . . . . . . . . . . . . . . . . . . . B. . . . . . separability and reflexivity A. . . .2. . .8 Fourier transform The case of L1 functions and extension The space S (Rd ) . . . Vector valued Lebesgue spaces .12 Compact embeddings . . . . . . . . A. . D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Vector valued distributions . Fourier transform and convolution . . . . . . . . . . .5 The Dirac comb . . . . . . . . . . . A. . . . . . . . . . . . .13 Poincar´ inequalities . . . . . . . . . . B. to L2 . D. . . . . . . . . . . . D. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 The space D(Ω) .4 Derivative of the Dirac measure B. . . . . . .3 C. . B Distribution spaces B. . . A. . . .1 C. . . . . . . . . . . . . . . . . . .3 The topological dual of H s (Rd ) D. . . . . . . . . . . . . . . . . . .3 Convolution . B. . . .7 C. . . . . . . . . . .3. .3 Examples of distributions . . . . . .2. . . . . . . . . . .3. . . . . . . . C The C. . . .2 Lebesgue spaces . . . . . . .1 Measures on Ω . . . . . . . . . . B. . . Partial Fourier transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 Trace theorem in H m (Ω) . . . .6 Principal value of 1/x . . . . . .8 The space H s (Γ) .4 Density property . . . .11 The space H −m (Ω) . . . . . . . . . . . .7 Trace theorem for H m (Rd ) . . . . . . . . . D. . D. . . . . . . . .3 Smooth domains and boundary measure . . . . . . . . . . . . . . .2 C. . . D. e . . . . . . . .6 Trace theorem for H s (Rd ) . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 The Dirac measure . B. .2 The space of distributions D (Ω) . . . . . . . . . . . . . . . . . . . . . . . .2 The space H s (Rd ) . . . . . . . . . D. . . . . . . . . .2. . . . . . . . . . . . . D. . . . . . . . . . . .3. . . .5 C. . . . . . . . . . . .xii A Functional spaces and regularity of domains A. . . . . . . . . . . . B. .3. . . . . . . . . .5 Density properties . . . . . . . . . . . . . . . . .2.10 Extension theorems . . . . .2 Regular distributions . . . . . D Sobolev spaces D. . .3. . . . . . . . . . . . . . . . . . . . 239 239 241 241 242 242 243 243 247 247 249 251 251 251 252 252 253 253 255 255 257 258 259 260 261 261 263 265 265 266 267 267 269 270 270 271 272 272 274 275 275 . . . . . . . . . . .

xiii References Index 277 289 .

.

. . .3 1. . . . . . . . . . .2 2. . . . . . . . . . . . . . . . . . . . . . . . . . . Melting front propagation Non-cylindrical space-time Perturbed tube . . . . . .1 Fluid flow control . . . . . .2 1. . . .4 1. . . . . . . . . . . . . . . . . . . . . . . . . .1 1. . . . . Transverse map . Rigid solid inside a fluid flow . . . . . . . . . . . . . . . . 9 14 20 26 28 34 36 38 39 186 216 Arbitrary Euler-Lagrange map . . . . . domain . . . .List of Figures 1. . . . xv . . . . . .1 2. . . . . . . . . . . . . . . . . . . . . . . . . . .4 7. . . .5 2. . . Geometric description . . . . . . . Shape deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 2. . . . . . . . . . . . . . . . .1 8. . . . . . . . . . . . . Non-cylindrical space-time domain Shape optimization . . . . . . . . . . . . .

.

fracture and contact problems. fluid-structure interaction problems for civil transport vehicules. these differentiation results form the basis in order to analyse the first and second-order trajectory variations of any general dynamical system. Among other applications. The first two chapters give all the details of the existence and uniqueness of the flow for both cases. civil engineering constructions such as stayedcable bridges or tall towers.g. we can cite the two ways of handling the tube perimeter constraint that yield to different expressions for the surface tension. We furnish different tools that can prove appropriate in order to deal with general variational formulations involving integral functionals using either the tensorial or the joint time-space metric over non-cylindrical time-space domains also called tubes.. optimization and control.. We will also revisit the notion of shape differential equations driving the evolution of a domain subjected to certain geometrical constraints. The efficiency of these tools will be illustrated within this volume on different examples in connection with the analysis of non-cylindrical partial differential equations.. As an example. boundary tracking problems. t) for t ≥ 0. The evolution of this transverse field is governed by a dynamical transport equation involving the Lie bracket [V. The main objective is to furnish various tools to handle the motion of a moving domain on the level of its intrinsic definition. computation. In the context xvii . We shall concentrate our analysis on the Eulerian approach. Hence the evolution of a 3D domain is chosen to be described by the flow T t of a nonautonomous Eulerian velocity field V(. . .Preface Objectives and Scope of the Book This volume intends to provide a mathematical analysis of problems related to the evolution of two or three dimensional domains. These results are fundamental in order to deal with the arbitrary Eulerian-Lagrangian formulation or the level-set approach. This flow can be defined in a strong or a weak manner depending on the space and time regularity of the moving domain. This topic includes a number of engineering applications such as free surface flows. the Navier-Stokes equations for incompressible fluids in moving domains. phase changes. W]. The sensitivity of the flow mapping with respect to its associated velocity field V in the direction W is completly characterized by the transverse field Z. e. biomechanical systems. computational vision .

In case the rigid solid is replaced by an elastic solid with an arbitrary constitutive law. with respect to a fluid inflow boundary condition. The Eulerian approach is based on the introduction of the transverse field characterizing the sensitivity of the flow with respect to the velocity field perturbations. The idea is to minimize a given tracking functional related to the solid evolution. computational vision . In this case. An interesting application is the rigorous mathematical justification of transpiration boundary conditions on fixed boundary usually used by engineers as a first-order approximation of no-slip boundary conditions on a moving solid surface. the sensitivity analysis is performed using non-cylindrical identity perturbations. fluid-structure interaction problems for civil transport vehicules. Intended Audience This book will be useful for researchers or graduate students with a background in applied mathematics who are interested in the control of complex systems involving a moving boundary. biomechanical systems. Again we obtain a new fluid-structure adjoint system. The case of a rigid solid coupled with an incompressible fluid flow is based on the Eulerian approach for characterizing the partial sensitivity of the fluid with respect to the solid state variables. The end of the volume deals with boundary control of fluid-structure interaction systems. phase changes. We state the structure of the optimality conditions using either the fluid state differentiability with respect to V or the min-max derivation principle providing instantly the adjoint state problem. The next chapters aim at providing illustration of these Eulerian evolution and derivation tools for the control of systems involving fluids and solids. In the context of the level-set method. civil engineering constructions such as stayed-cable bridges or tall towers. Especially. We obtain a new fluid-structure adjoint problem that allows the computation of a given cost function gradient. we use the Lagrangian framework and the non-cylindrical identity perturbations. this shape differential equation is related to the HamiltonJacobi equation driving the optimal evolution of the level-curves.xviii of optimal control and inverse problems. fracture and contact problems. we shall consider in chapters 5 and 6 the control of a fluid described by the Navier-Stokes equations inside a 3D bounded domain through the dynamical evolution of its boundary. . Similar results are obtained using a Lagrangian description of the boundary motion. . boundary tracking problems.. More generally. This include topics such as free surface flows. this equation corresponds to the first-order optimality conditions associated to the minimization with respect to the velocity field V of a given Eulerian cost function. this book will be of great use for people who are interested .

Some parts of the book can require some knowledge of advanced mathematics that are recalled in the appendix at the end of the volume. .xix in the optimization of complex partial differential systems since it includes various tools that can be used in different contexts.

xx .

84. These new problems were formulated in terms of minimization of functionals (referred as open loop control or passive control) governed by partial differential boundary value problems where the control variable was the geometry of a given boundary part [103. 12. 110. we shall mention the work of G. he proposed a list of open problems at the French National Colloquium in Numerical Analysis. Morice [107] and also several approaches related to perturbation theory by P.R.Chapter 1 Introduction Shape Optimization was introduced around 1970 by Jean C´a [31].D Joseph [91. e.. 124. 75] and D. Lions [98]. many pioneering works were violating the validity of this assumption. 92]. the work presented in the Iowa NATO Study [85] stating the existence of optimal beams having zero cross section values. 83. 1 . At that time. 102. it was hoped that the solution of elliptic problems would be continuous with respect to the weak convergence of the coefficients. From the beginning. Indeed several classical modeling in both structural and fluid mechanics (among other fields) needed to be extended. At that time. At the same time. 34] based on the theory of distributed system control introduced by J-L. 76]. Furthermore. these research studies were mainly addressed in the context of the numerical analysis of the finite element methods. numerical approximation (including finite element methods). the future issues in the context of optimization problems. the terminology shape optimization was not connected to the structural mechanical sciences in which elasticity and optimization of the compliance played a central role. Chavent [32. Ph. it appeared that the shape controlof Boundary Value Problems (BVP) was at the crossroads of several disciplines such as PDE analysis. control and optimization theory. who une derstood. In the branch which followed the passive control approach. leading to strange results. Very soon. 7].g. Those results did not address optimization problems related to the domain but instead related to the coefficients inside the PDE. non-autonomous semi-group theory. after several engineering studies [127. Therefore. 78]. This kind of solid model is based on the assumption that the thickness undergoes only small variations. there was some independent close results concerning fluid mechanics by young researchers such as O. Pironneau [123. 35. geometry and even physics. An illustrative example concerns a very popular problem in the 80’s concerning the thickness optimization of a plate modeled by the classical Kirchoff biharmonic equation. Garabedian [74.

That paper [111] appeared but as a counterexample to the expected continuity property. The systematic use of flow mapping and intrinsic geometrythrough the fundamental role of the oriented distance function [47. The associated topology was extended by M. Micheletti introduced in parallel [105. the c role of bilipschitzian mapping was emphasized for Sobolev spaces defined in moving domainsbased on the Identity perturbation method [115. the capacity constraint for Dirichlet boundary conditions led to a fine analysis initiated in [18] and is still going on for cracks analysis. 27. applications focused more on sensitivity analysis problems than on asymptotic analysis of domains evolution. 104] a metric based on the Identity perturbation method thanks to the use of differentiable mappings. Neˇas [117]. At that period. 25. G. 26. 50] led to the revised analysis of the elastic shell theory [48. 106.2 Moving Shape Analysis It appeared that this property was not achieved by this class of problem 1 . the second author introduced [145] an asymptotic analysis for domain evolution using classical geometrical flows which are intrinsic tools for manifolds evolutions and gave existence results for the so-called shape differential equation (see also [79]). In 1972. Following the very powerful theory developed by J. The mathematical analysis of shape optimization problems began with the correct definition of derivatives of functionals and functions with respect to the domain. He showed on a one dimensionnal simple example that with weak oscillating convergence of the coefficients. 28]. At that point a main bifurcation arose with the homogenization approach [10] which up to some point was considered as a part of the Optimal Design theory. 114]. 21. in his thesis [33]. of the boundary layer theory [3] or of the manifold derivative tools [53]. 19] or Sobolev domains [50]. [52] and turns out to be the same as the one induced by the continuity along flow field deformations [147]. Concerning the large domain deformationviewpoint the previous approach led to the incremental domain evolution methods [143].M. The method of large evolution based on the flow mapping (known from 1980 as the speed method [150]) turns to be the natural setting for weak evolution of 1 Indeed. 20. . in order to study eigenvalues perturbation problems. the associated solution was converging to another problem in which the new coefficients were related to the limit of the inverse coefficients associated to the original problem [112. density perimeter [23. At that point. A. together with the choice of tangential space to the familly of shapes. 134]. At the opposite. Murat [113]. Delfour et al. 49. Chavent referred to such a result to appear in a work by F. The use of both Bounded Variation (BV) analysis and the notion of Cacciapoli sets led to the first compactness method for domain sequences and several extensions to more regular boundaries were done through the use of different concepts such as fractal boundaries. an other important bifurcation point in that theory occurred with the relaxation theory and the Special Bounded Variation (SBV) analysis which was particularly well adapted for image segmentation problems [6]. After 1975.

On the contrary. Here. we shall deal with sensitivity analysis with respect to a continuous family of shapes over a given time period. In this context. 1. the moving shape analysis deals with systems that are intrinsically defined on a moving geometry. In numbers of applications. This change of topology can be allowed by using the characteristic function of families of sets or the level set of a space-time scalar function. This way of defining the motion of domains avoids a priori the modification of the underlying topology. we shall deal with the particular problem of defining in a weak manner the convection of a characteristic function in the context of the speed method developed in [147]. This is motivated by many important applications in physics. In this case. Hence.1 Classical and moving shape analysis The object of this book is the mathematical analysis of systems involving the evolution of the geometry. we shall consider a state variable associated to a system which is a solution of a partial differential equation defined inside the moving domain over a given time period. again this topic has been already studied since it enters the classical shape analysis 2 The notion of tube (non-cylindrical evolution domains) was also independently introduced by J. we need to analyse the solvability of this non-cylindrical PDE system before going further. . differentiability and optimization of quantities related to the state of a system defined in that geometry. engineering or image processing. a domain can be made variable by considering its image by a family of diffeomorphisms parametrized by the time parameter as it happens frequently in mechanics for the evolution of continuous media. the moving shape analysis may also be referred to as a non-cylindrical shape analysis2 . the geometry is usually perturbed thanks to a map involving a scalar parameter usually referred to as a fictitious time. There exist many ways to build families of geometries. The classical shape analysis investigates the effects of perturbations of the geometry in terms of continuity. we shall concentrate on the particular framework of the present book. A first issue in this analysis is to model the evolution of the geometry.P. Aubin via the concept of abstract mutations [8]. For instance. We refer the reader to [51] for a complete review on this topic. Hence.Introduction 3 geometry allowing topological changes through the convection of either characteristic functions or oriented distance functions. This is a common topic with the classical shape analysis. In chapter 2. if we consider the geometry in a space-time configuration. After this non-exhaustive review of the context in which the shape optimization analysis emerged.

131. bridge decks . 132. In this example. . 120. 70. 65] and more recently on its controllability [66. 118. 9. 55. 62. 54. An important step was performed in [154. Here the motion of the interface between the fluid and the solid is part of the unknown of the coupled system. [96. 119. 109]. This work was generalized in [58. 67]. 128. its stability [73. 100.). 130. In this lecture note. the kinematic continuity of the velocities and the kinetic continuity of the normal stresses). .4 Moving Shape Analysis problem while introducing a perturbed state defined in the moving domain parametrized by the fictitious time parameter. 158.2 Fluid-Structure interaction problems As already mentioned above. These results were applied in order to study variational principles for an elastic solid under large displacements and for the incompressible Euler equation. 122. 38. 82. This model has been intensively studied in the last two decades on the level of its mathematical solvability [87. 88.. 138. . A general fluid-solid model consists of an elastic solid either surrounded by a fluid (aircrafts. 15.) or surrounding a fluid flow (pipelines. . reservoir tanks .g. . we refer the reader to [135. 59]. 51] for some particular results in the context of the classical shape analysis. arteries. Contrary to the last topic. The case of a compressible Newtonian fluid can be incorporated in the present framework with the price of a heavier mathematical analysis (solvability. 155] where the second author of the present book established the derivative of integrals over a moving domain with respect to its associated Eulerian velocity. non-differentiability around shocks . . e. 41]. 80. Here. 39. Early studies have been conducted in [90. 64. We also refer to the extensive litterature concerning the analysis of PDE systems defined in moving domains. Furthermore. 95. 2] for specific hyperbolic and parabolic linear problems. very few references exist for the sensitivity analysis with respect to perturbation of the evolution of the moving geometry. the con- . we will restrict ourself to viscous Newtonian incompressible fluid flows described by the Navier-Stokes equations in space dimension two or three. As an example. 43.). we shall apply the moving shape analysis on systems involving the coupling between fluids and solids. 11]. 1. It is a free boundary problem that can be solved by imposing continuity properties through the moving interface (e.g. we think to decrease the drag of a car inside the atmospheric air flow by producing specific vibrations on its body using smart materials such as piezoelectrical layers. 142. this solvability analysis has been performed in numbers of mathematical problems involving moving domains.. 126. 141. automobiles. its numerical approximation [89. Our goal is to solve inverse or control problems based on the previous general fluid-solid model. 151.

in order to recover the wind speed history. Then. we need to characterize the relationship between the drag function and the control variable on the level of its computation and its variations. Hence. Indeed the fluid state is the solution of a system of non-linear partial differential equations defined in a moving domain. it has been suggested that such a problem can be set as the inverse problem consisting in recovering the smallest upstream wind speed that leads to the worst bridge deck vibrations. As an other example. In [57]. In [108]. Here we recall that the coupled fluid-structure state is the solution of a system of partial differential equations that are coupled through continuity relations defined on the moving interface (the fluid-structure interface). Then using the tools developed in [59]. which is an Eulerian quantity. the author used a noncylindrical identity perturbation technique. it has been possible to perform in [58] the moving shape sensitivity analysis in the case of a Newtonian incompressible fluid inside a moving domain driven by the non-cylindrical Navier-Stokes equations. which is a Lagrangian quantity. this family is chosen as the design parameter. In order to characterize the sensitivity of the objective functional with respect to the control variable. we can think of the problem of aeroelastic stability of structures. the decision variable can be chosen as the upstream wind speed and the objective is to increase a functional based on the vibration amplitude history of the bridge deck during a given characteristic time period which is a function of the coupled fluidstructure state ( the wind flow and the bridge deck) which is also a function of the decision variable. allowing topology changes while using the associated level set formulation. Both authors have been dealing with such a problem in the context of the stability analysis against wind loads of bridge decks. The key point towards this sensitivity analysis is to investigate the sensitivity of the fluid state. In [13. It consists in perturbating the space-time identity operator by a family of diffeomorphisms. All the previous results use a parametrization of the moving domain based on the Lagrangian flow of a given velocity field. the authors . it is obvious that we need to characterize the sensitivity of the coupled fluid-structure state with respect to the control variable. Again. with respect to the motion of the solid. This task falls inside the moving shape analysis framework described earlier.Introduction 5 trol variable can be chosen as the electrical energy input evolution inside the piezoelectrical device and the objective is to decrease the drag which is a function of the coupled fluid-structure state ( the air and the body of the car) and this state depends on the control variable. It is a Lagrangian description of the moving geometry. The boundary of this moving domain is the solid wall. the design variable is the Eulerian velocity of the moving domain. In order to build a control law for the electrical input. In this example. which a priori does not allow topology changes but which leads to simpler sensitivity analysis results which are still comparable with the one obtained by the non-cylindrical speed method. we need to characterize the relationship between the objective functional and the decision variable on the level of its computation and its variations. 14].

6

Moving Shape Analysis

came back to the dynamical shape control of the Navier-Stokes and recovered the results obtained in [58] using the Min-Max principle allowing to avoid the state differentiation step with respect to the velocity of the domain. Now, we come back to the original problem consisting in the sensitivity analysis of the coupled fluid-structure state with respect to the control variable. Using the chain rule, the derivative of the coupled state with respect to the control variable involves the partial derivative of the fluid state with respect to the motion of the fluid-structure interface already characterized in [58, 57]. Hence, again using a Lagrangian penalization technique, already used and justified in [45, 46], it has been possible to perform in [109] the sensitivity analysis of a simple fluid-structure interaction problem involving a rigid solid within an incompressible flow of a Newtonian fluid with respect to the upstream velocity field. As already mentioned, this simple model is particularly suited for bridge deck aeroelastic stability analysis [121].

1.3

Plan of the book

The book is divided in eight chapters. Chapter 2 furnishes a simple illustration to some of the moving shape analysis results reported in the core of the lecture note. We deal with a simple inverse problem arising in phase change problems consisting in recovering the moving interface at the isothermal interface between a solid and liquid phase from measurements of the temperature on a insulated fixed part of the solid boundary. We use a least-square approach and we show how to compute the gradient of the least-square functional with respect to the velocity of the moving interface. It involves an ajoint state problem together with an adjoint transverse state, which is the novelty of the moving shape analysis compared to the classical one. In chapter 3, we consider the weak Eulerian evolution of domains through the convection, generated by a non-smooth vector field V, of measurable sets. The introduction of transverse variations enables the derivation of functionals associated to evolution tubes. We also introduce Eulerian variational formulations for the minimal curve problem. These formulations involve a geometrical adjoint state λ which is backward in time and is obtained thanks to the use of the so-called transverse field Z . In chapter 4, we recall the concept of shape differential equation developed in [145],[147]. Here, we present a simplified version and some applications in dimension 2 which enable us to reach the time asymptotic result. Furthermore, we introduce the associated level set formulation whose speed vector version was already contained in [149]. In chapter 5, we deal with a challenging problem in fluid mechanics which consists in the control of a Newtonian fluid flow thanks to the velocity evo-

Introduction

7

lution law of a moving wall. Here, the optimal control problem has to be understood as the open loop version, i.e., it consists in minimizing a given objective functional with respect to the velocity of the moving wall. This study is performed within the non-cylindrical Eulerian moving shape analysis describe in chapters 2 and 3. We focus on the use of a Lagrangian penalization formulation in order to avoid the fluid state differentiation step. In chapter 6, we introduce the Lagrangian moving shape analysis framework. It differs from the Eulerian one from the fact that the design variable is the diffeomorphism that parametrizes the moving geometry. The sensitivity analysis is simpler since it does not involve the transverse velocity field. We apply these tools in order to deal with the control of a Newtonian fluid flow thanks to the displacement evolution law of a moving wall. Chapter 7 moves to inverse problems related to fluid-structure interaction systems. Here, we consider a 2D elastic solid with rigid displacements inside the incompressible flow of a viscous Newtonian fluid. We try to recover informations about the inflow velocity field from the partial measurements of the coupled fluid-structure state. We use a least-square approach together with a Lagrangian penalization technique. We derive the structure of the gradient with respect to the inflow velocity field of a given cost function. Using the Min-Max principle, the cost function gradient reduces to the derivative of the Lagrangian with respect to the inflow velocity at the saddle point. This saddle point is solution of the first order optimality conditions. We use non-cylindrical Eulerian derivatives to compute the partial derivative of the Lagrangian functional with respect to the solid state variables, involved in the optimality system. Finally, in chapter 8 we extend the results of chapter 7, to the case of an elastic solid under large displacements inside an incompressible fluid flow. The main difference with the previous case is the use of a non-cylindrical Lagrangian shape analysis for establishing the KKT system. It forms the adjoint counterpart of the sensitivity analysis conducted in [66].

1.4

Detailed overview of the book

In order for the reader to have a simple overview of the book, we shall describe the different steps encountered while designing a complex fluid-structure interaction system. Indeed, let us consider a mechanical system that consists of a solid and a fluid interacting with each other. We would like to increase the performances of this system. These performances have to be quantitatively translated inside a cost function that we have to optimize with respect to some parameters that we will call the control variables. In the sequel, we will describe different control situations:

8

Moving Shape Analysis 1. Control of a fluid flow around a fixed body: it consists in trying to modify the fluid flow pattern around a fixed body using a boundary control which can act for example by blowing or suctioning the fluid at some part of the solid boundary. The control law will be designed in order to match some efficiency goals using the minimisation of a cost functional. 2. Shape design of a fixed solid inside a fluid flow: in this case, the control is the shape of the body. We would like to find the best shape satisfying some geometrical constraints that will optimize some cost functionals. This problem is somewhat classical in the aeronautical field, but it requires some subtle mathematical tools that we will quickly recall. 3. Dynamical shape design of a solid inside a fluid flow: the novelty compared to the last item is that the shape is moving and we are looking for the best evolution of this shape that both satisfies some geometrical constraints and optimizes some cost functionals. This is a rather natural technique in order to control a fluid flow pattern, but still its design requires some new mathematical tools that will be sketched in this introduction and more detailed in the core of this lecture note. 4. Control of an elastic solid inside a fluid flow: this is the most complex and most realistic situation where both the fluid and the solid have their own dynamics which are coupled through the fluid-solid interface. Then, we would like to control or optimize the behaviour of this coupled system thanks to boundary conditions. The mathematical analysis of this situation uses the whole framework introduced previously. This is a challenging problem, both on the mathematical point of view and on the technological side. The goal of this book is to partially answer to some issues related to this problem.

1.4.1

Control of a fluid flow around a fixed body

domain is denoted by Ωf = D \ Ω . The fluid is described by its velocity u : R+ × Ωf → R3 and its pressure p : R+ × Ωf → R. The flow is supposed to be incompressible and the state variables satisfy the classical Navier-Stokes equations, ∂t u + D u · u − ν ∆u + p = f , (0, τ ) × Ωf (1.1) div u = 0, (0, τ ) × Ωf We must endow this system with boundary conditions such as far-field Dirichlet boundary conditions on ∂D and non-slip boundary conditions on the solid

Let us consider a fixed solid obstacle Ω ⊂ R3 with boundary Γ surrounded by a viscous Newtonian fluid. We shall restrict the analysis of this system to a control volume D ∈ R3 containing the obstacle Ω (see figure 1.1). The fluid
def

Introduction boundary Γ u = u∞ , ∂D u = 0, Γ We also add initial conditions, u(t = 0) = u0 , Ωf

9

(1.2)

(1.3)

A classical issue consists in reducing the drag exerted by the fluid on the

Ωf

Γc

D Ω Γ

∂D

FIGURE 1.1: Fluid flow control

solid (e.g., car, aircraft, ship). The first step in this direction is to choose the type of control we would like to apply on this system to reach our objective. The boundary control consists in adding inside the solid a control device that modifies the boundary conditions on a part Γc ⊂ Γ of its boundary, u = g, Γc u = 0, Γ \ Γc Here the boundary control g can be either • time-independent and in this case it can be viewed as a passive control strategy consisting in modifying, e.g., the friction of the surface. • or time-dependent and in this case it can be viewed as an active control strategy consisting of a blowing/suction system that can be designed off-line (that means using an open-loop control) or on-line (that means using a feedback control law). Let us describe quickly both situations since it will be useful for the rest of the book. The objective functional A common topic in the optimization and control field of PDE systems is the choice of appropriate cost functionals, i.e., meeting both our objectives and (1.4)

10

Moving Shape Analysis

the mathematical requirements that guarantee the converge to at least one optimum parameter. This functional can depend both on the state variables (u, p) and on the control parameter g. Let us choose this function as the work needed to overcome the drag exerted by the fluid on the solid boundary, Jdrag (u, p) =
(0,τ )×Γ def

(u − u∞ ) · σ(u, p) · n dΓ dt

(1.5)
def

where σ(u, p) = −p I +ν ε(u) stands for the fluid stress tensor, ε(u) = 1 ∗ 2 (D u + D u) stands for the fluid deformation tensor and n is unit normal field. Another cost function is associated to the objective of steering the fluid along a given pattern ud during a prescribed time,
Jtrack (u, p) = 1 2
(u − ud )2 dx dt (1.6)

(0,τ )×Ωf

More generally, we can consider any cost functionals that are twice-differentiable with respect to their arguments. The control problem Here we choose to work with a control variable defined as a velocity field on the solid boundary and our objective is to minimize a given cost functional, min j(g)
g∈U

(1.7)

Our goal is now to furnish the first-order optimality conditions associated to the optimization problem 1.7. These conditions are very useful since they are the basis in order to build both a rigourous mathematical analysis and gradient-based optimization algorithms. There exists two main methods in order to derive these conditions: the first one is based on the differentiability of the state variables with respect to the control parameter and the second one relies on the existence of Lagrangian multipliers.

where j(g) = J[(u, p)(g)], U stands for an appropriate Hilbert space of the Sobolev type, typically H 1 ((0, τ ) × Γc ) and the couple (u, p)(g) stands for a weak solution of the Navier-Stokes equations associated to the control g,   ∂t u + D u · u − ν ∆u + p = f , (0, τ ) × Ωf    (0, τ ) × Ωf  div u = 0,   u = u∞ , (0, τ ) × ∂D (1.8) (0, τ ) × Γc  u = g,   c  (0, τ ) × Γ \ Γ  u = 0,   u(t = 0) = u0 , Ωf

Introduction Sensitivity

11

Let us consider a control point g ∈ U, then the cost functional j(g) is Fr´chet differentiable with respect to g [1, 71] and its directional derivative is e given by j (g), h = ∂(u,p) J[(u, p)(g)], (u , p )(g; h) (1.9)
d where (u , p )(g; h) = dg (u, p)(g) · h stands for the directional derivative of (u, p)(g) with respect to g. In the case of the functional (1.6), we have def

j (g), h =
(0,τ )×Ωf

(u(g) − ud ) · u (g; h) dx dt

Then the first-order optimality condition writes j (g), h = 0,

It can be proven by evaluating the differential quotient that (u , p )(g; h) exists in an appropriate space and that it is the solution of the linearized Navier-Stokes system,   ∂t u + D u · u + D u · u − ν ∆u + p = 0, (0, τ ) × Ωf    div u = 0, (0, τ ) × Ωf    u = 0, (0, τ ) × ∂D (1.10) (0, τ ) × Γc  u = h,   c  u = 0, (0, τ ) × Γ \ Γ    u (t = 0) = 0, Ωf ∀h ∈ U (1.11)

That means that the set of optimal controls is contained in the set of critical points for the cost function j(g). However, we would like to obtain an expression of this condition avoiding the direction h ∈ U. To this end, we introduce the adjoint variable (v, π) solution of the adjoint linearized Navier-Stokes system,   −∂t v − D v · u + ∗ D u · v − ν ∆v + π = −(u − ud ), (0, τ ) × Ωf    div v = 0, (0, τ ) × Ωf  v = 0, (0, τ ) × ∂D (1.12)   v = 0, (0, τ ) × Γ    v(t = τ ) = 0, Ωf Consequently, we are able to identify the gradient of the cost function as the trace on Γc of the adjoint normal stress tensor, i.e., j(g) = ∗ γ(0,τ )×Γc σ(v, π) · n (1.13)

This formal proof provides the basic steps needed in order to build a gradientbased optimization method associated to the control problem (1.7).

q) (1. p. min min max L(ψ. it can be proven that the min-max problem (ψ. v. q) · n) + ψ(τ ) · φ(τ ) − Ωf u0 · φ(0) (1. φ.14) [−ψ · ∂t φ + (D ψ · ψ) · φ − ν ψ · ∆φ (0. e(ψ.12). e. Hence using the differentiability of J(g).g. q. g). q) stands for the weak form of the state equations (1. Using a discretization of the parameter s leads to a standard gradient-based method such as the conjugate-gradient or the quasi-Newton method depending on the choice of A(s). (φ. q) · n) dΓ dt Ωf (0. r. g). q) = (0.q) For every control g ∈ U. Indeed let us introduce a scalar parameter s ≥ 0. . then the functional writes r g (s) + A−1 (s) J(g(s)) = 0. using a continuous iterative method. φ. (φ. r) + e(ψ. L(ψ. Then.q) min max L(ψ.12 Moving Shape Analysis where e(u.8). s ∈ (0.15) writes ∂g L(u. and a control variable g(s) that is differentiable with respect to s.8)-(1. Finally the first-order optimality for the problem (1. we get r J(g(r)) − J(g(0)) = Let us choose the control such that J(g(s)). φ.7) is equivalent to the min-max problem.. g) admits a unique saddle-point (u.11). g) = 0 (1.16) which turns out to be equivalent to (1. g).17) leads to a functional’s decrease and is referred to as a continuous gradient based optimization method. p) with respect to the control g thanks to the introduction of a Lagrangian functional that includes not only the cost functional but also the state equation. g) = J(ψ. (φ.r) (φ.15) Hence the control problem (1.r) (φ. g (s) 0 U ∗ . r) (1.τ )×Γc +ψ · + q − r div φ] + g · (σ(φ. v.τ )×∂D u∞ · (σ(φ. r. g) g∈U (ψ. r. π) which are solutions of the systems (1. r.17) J(g(r)) − J(g(0)) = − 0 | J(g(s)|2 ds That means that the control law (1. p.13).τ )×Ωf An alternative approach consists in avoiding the derivation of the fluid state (u. we can think to solve the optimality condition (1. r. p. q.U ds where A stands for an appropriate duality operator. q. π.

∂s Ts (x) = V(s.. Ωs = Ts (Ω). ε]. ∀x ∈ D C 0 ([0. where the transformation is the flow associated to a given velocity field V(s.2 Shape design of a fixed solid inside a fluid flow We again consider the situation where a fixed solid is surrounded by a fluid flow. • the speed method [145. the drag) under some perimeter. y) T−1 (.. we have been dealing with an optimization problem where the control belongs to a linear space. ε]. x) − T−1 (.R3 ) ≤C x−y R3 .e.ε].g..18) In order to carry out the sensitivity analysis of functionals depending on the shape of the solid Ω .. Ω∈A min J(Ω) (1. (s.. −1 (. R ). x∈D 3 Typically we have the following Lipschitz regularity assumptions : T(. Ts (x)). R3 ). x) ∈ (0. x) ∈ C ([0. x) − T(. These domains are the images ¯ ¯ of the original domain Ω through a given family of smooth maps 3 Ts : D → D. 125]). ε) × D Ts=0 (x) = x. ∀x ∈ D 0 3 ≤C x−y R3 C 0 ([0. Here the space of shapes is no more a linear space and the associated differential calculus becomes more tricky. Ts = I +s θ ¯ ¯ where θ : D → D.R3 ) T where C > 0. i. volume or curvature constraints. Our goal is to build gradient-based methods in order to find the optimal shape.. This optimization is an open-loop control since the shape of the obstacle is time-independent. x). x) ∈ C 1 ([0.. y) T(. 147]. we would like to solve the following problem.Introduction 13 1..e.. we introduce a family of pertubated domains Ω s ⊂ D parametrized by a scalar parameter 0 ≤ s ≤ ε. i. The speed method In the previous section.ε]. Γs = Ts (Γ) Two major classes of such mappings are given by : • the identity perturbation method ([116. The shape control consists in finding the optimal shape of the solid that reduces some objective functional (e.4.

at least for the first order terms.20)   u = 0..g.   ∂t u + D u · u − ν ∆u + p = f . Jdrag (Ω) = (0. (0. but also because it involves the solution (u. τ ) × Γ    u(t = 0) = u0 . .14 Moving Shape Analysis ¯ In order for D to be globally invariant under Ts (V) we need to impose the following viability conditions. e. τ ) × Ω  u = u∞ . (0. (0. dJ(Ω.19) This functional depends on Ω not only because it is an integral over the boundary Γ. the work to overcome the drag exerted by the fluid on the solid boundary. V) = lim s 4 which 0 J(Ωs (V)) − J(Ω) s leads. p) · ndΓ dt (1. p) of the NavierStokes system. V(s. (0.τ )×Γ (u − u∞ ) · σ(u. to the same results as the identity perturbation framework [51]. τ ) × Ω    div u = 0. Ω that depends on Ω. we choose to work in the framework of the speed method4 . τ ) × ∂D (1. x) · n(x) = 0.2: Shape deformation Let us consider the family of functionals J(Ωs ) that depends on the shapes Ωs . x ∈ ∂D Ts(V) Ω Ω Γc Γ ∂D D Γc Γ ∂D D FIGURE 1. We define the Eulerian derivative of the shape functional J(Ω) at point Ω in the direction of the vector field V ∈ V as the limit. To perform our sensitivity analysis.

we have introduced the notion of Eulerian derivative for shape functionals.D(D.D(Γ) . ε. ∀V ∈ V def In the sequel. R3 )).e. ∗ γΓ (g n). V. This result. n = 0 on ∂D ¯ refer the reader to [51] Theorem 3. V) = G(Ω). This result is easier to understand when g(Γ) is integrable over Γ.. In this case.R3 ) = g. that is to say. Indeed in this case. If this limit exists and is finite ∀ V ∈ V and the mapping V→R V → dJ(Ω. C 1 (D. a function 5 e. we shall use the notation J(Ω) = G(Ω). V · n D (Γ). G(Ω) = ∗ γΓ (g n) where g(Γ) ∈ D −k (Γ) stands for a scalar distribution and ∗ γΓ stands for the adjoint trace operator7 . def ˘ V ∈ C 0 (0. V) = Γ g V · n dΓ Basic shape derivative calculus In the previous paragraph.5 for the case of non-smooth domains. it means that the directional shape derivative can always be written as follows. called the structure theorem6 . R3 ) . Hence. Actually if J(Ω) is shape differentiable. V(0) . g ∈ L1 (Γ). div V = 0. the gradient is only supported on the boundary Γ and depends linearly on the normal vector field n. the shape gradient has the following representation. R 3 ) that we call the shape gradient such that dJ(Ω. This notion can be extended to functions defined on Banach or Hilbert spaces built on smooth domains Ω.R3 ) .g. Shape derivative structure theorem: Let J(.Introduction 15 where V is a linear space5 . is recalled as follows. in D. then the functional J(Ω) is said to be shape differentiable. V = 6 We 7 i. V) is linear and continuous.. V D(D. dJ(Ω. In the case of smooth domain.) be a differentiable shape functional at every shape Ω of class C k+1 for k ≥ 0 with shape gradient G(Ω) ∈ D(D. then its Eulerian derivative only depends on V(0) and there exists a distribution G(Ω) ∈ D(D.

dJ(Ω. then the functional J(Ω) is shape differentiable and its directional derivate writes. we define the shape ˙ derivative as y (Ω. we need to introduce the notion of material derivative on Γ. then using the Stokes formula.1 In the case where y(Ω) = Y |Ω . V) = Ω [y (Ω. it is possible to derive the Eulerian shape derivative of the following functionals. V) dΩ + Γ y(Ω)V · n dΓ REMARK 1. Hence. the Eulerian shape derivative of J(Ω) takes the following form. . V) − y(Ω) · V(0) ˙ In this case. τ . V) = lim ˙ def s→0 1 y(Ωs (V)) ◦ Ts (V) − y(Ω) s admits a limit in the Hilbert space8 H(Ω). dJ(Ω. V) + div(y(Ω) V(0))] dΩ If Ω is class C k with k ≥ 1. we get dJ(Ω. Endowed with the following definition.p (Ω)).p (Ω) or L2 (0. V) = Ω [y(Ω. its shape derivative is zero since y(Ω. ˙ dJ(Ω. V) + y(Ω) div V(0)] dΩ ˙ In order to apply the structure theorem.16 Moving Shape Analysis y ∈ H(Ω) admits a material derivative at Ω in the direction V ∈ V if the following limit y(Ω. it is useful to define the notion of shape derivative for functions. W m. if y ∈ H(Ω) admits a material derivative y(Ω. J(Ω) = Ω y(Ω) dΩ If y(Ω) is weakly shape differentiable in L1 (Ω). W m. V) = Y · V. V) ∈ H(Ω) and y · V(0) ∈ H(Ω) for all V ∈ V. V) = Ω y (Ω. Hence.g.. V) = Γ y(Ω)V · n dΓ This is a simple illustration of the structure theorem. 8 e. In the case of functionals involving integration over the boundary Γ. V) = y(Ω. where Y ∈ H(D) with Ω ⊂ D.

p (Γ). It is said that it admits a material derivative in the direction V ∈ V. V) = Γ y (Ω. V) = Γ Z|Γ − ( Z · n) n [z (Γ. As in the previous case. V) + H z(Γ) V(0) · n] dΓ where H stands for the mean curvature of Γ. V) ∈ W (Γ) and Γ y · V(0) ∈ W (Γ) for all V ∈ V. 1 z(Γs (V)) ◦ Ts (V) − z(Γ) s admits a limit in the Hilbert space W (Γ). z(Γ. Using the above definition.Introduction 17 Let z ∈ W (Γ) where W (Γ) is an Hilbert space of functions 9 defined over Γ. it is possible to transform the expression of the differential as follows.g. then the functional J(Γ) is shape differentiable and its directional derivate writes dJ(Γ. V) = Γ [z(Γ. V) = lim ˙ def s→0 J(Γ) = Γ z(Γ) dΓ If z(Γ) is weakly shape differentiable in L1 (Γ). it is possible to derive the Eulerian shape derivative of the following functional. it is also possible to introduce the notion of shape derivative for z(Γ). V) = z(Γ. . W m.2 In the case where z(Γ) = y(Ω)|Γ . V) + z(Γ) divΓ V(0)] dΓ ˙ def where stands for the tangential divergence.. V) + ( y(Ω) · n + H z(Γ) V(0) · n dΓ 9 e. If z ∈ W (γ) admits a material derivative z(Γ. dJ(Γ. V) − ˙ where Γz Γ z(Γ) divΓ V = γΓ div V − (D V · n) · n · V(0) = stands for the tangential gradient and Z is any smooth extension of z inside Ω. the Eulerian derivative takes the following form. if the following limit. Let Ω be of class C k with k ≥ 2. ˙ we define the shape derivative as z (Γ. REMARK 1. dJ(Γ. As a consequence of this definition.

τ ) × ∂D (1. +∞) (1.18 Moving Shape Analysis Application to shape design Thanks to the framework introduced previously. τ ) × Ω    div v = 0. τ ) × Γ    v(t = τ ) = 0.21) using a continuous gradient-based method. Ω The associated shape differential equation Now as in the previous section. s ∈ (0. Notably.   −∂t v − D v · u + ∗ D u · v − ν ∆v + π = −(u − ud ). (0.22)   v = 0. it is possible to build a complete sensitivity analysis of shape functionals. . (0.20) associated to the shape Ω and (v. which hold essentially when the shape gradient has some continuity properties for an ad-hoc shape topology 10 .21) is given by J(Ω) = ∗ γΓ σ(v. π) is solution of the adjoint system.23) leads to a decrease of the functional J(Ωs (V)). π) · n where (u.23) is referred to as the shape differential equation and some of its properties are studied in chapter 4 of this book. we can state the following: the shape gradient for the tracking functional J(Ω) = (0. we can choose to solve the first-order optimality equation (1. we study its solvability in the case of smooth shape functionals. That means that we write r J(Ωr (V)) − J(Ω0 ) = Then solving the equation J(Ωs (V)). The equation (1. The level-set framework In chapter 4. p) is a solution of system (1. we also relate the shape differential equation to the HamiltonJacobi equation involved in the level-set setting. τ ) × Ω  v = 0. The level-set setting consists 10 The Hausdorff-complementary topology. (0. Coming back to our optimal shape problem.τ )×Ω (u − ud )2 dx dt (1. V(s) ds 0 J(V(s)) + A−1 (s) · V(s) = 0. We also prove some results concerning the asymptotic behaviour of the solution of this equation. (0.

24) Q∈E where Q ∈ E is a smooth evolution set. and its boundary is the zero-level set. 1. The Rd+1 -approach The optimal control of moving domain is a problem which is relevant of classical (non-linear) control theory as well as of classical shape theory. min J(Q) (1. Σ = def t∈(0.3 Dynamical shape design of a solid inside a fluid flow We return to our model problem. Γs = Γs (Φ) = {x ∈ D. def def Φ(s. ε) × D → R. V(s) = −∂s Φ(s) Φ(s) Φ(s) 2 Both frameworks are equivalent if Φ(s) belongs to set of functions without steps. which means Q = def t∈(0. We show how to build without step functions and we study the shape differential equation in this setting.e. Our goal is to control this motion in order to optimize some objective functionals.We call the set Q a tube. x) > 0} Φ(s. which means that Φ(s) is different from zero almost everywhere in D. i.τ ) {t} × Ωt where Ωt is a smooth domain of R3 with boundary Γt . Our goal is to build gradient-based methods in order to find the optimal shape dynamic.4. Ωs = Ωs (Φ) = {x ∈ D. The set. but this time we consider that the shape of the solid is moving. we would like to solve the following problem. the dynamical shape control theory can . In fact on the pure theoretical level.τ ) {t} × Γt stands for the non-cylindrical lateral boundary.Introduction 19 in parametrizing the perturbed domain Ωs as the positiveness set of a scalar ¯ function Φ : (0.. x) = 0} This parametrization and the one introduced in the speed method can be linked thanks to the following identity.

As for usual mappings defined from R in some space E. the idea is to avoid the normal field ν to the lateral boundary Σ of the tube to be strictly vertical. we adopt an Eulerian viewpoint that associates 11 say there exists a tangent space. Usually we would like to minimize some cost functional. In the same way. To handle non-smooth situations. Now under simple conditions on that set G. τ j(S) = 0 J(t. When the boundary of the set Ωt is smooth enough 11 . Intuitively we would say that we require some causality in the evolution of the set Ωt . the graph G ⊂ R × E and of course any subset G is not a graph. Indeed the dynamical shape control consists in finding the optimal evolution of a spatial domain Ωt in Rd . Let us consider the mapping S : t ∈ R −→ Ωt ∈ P(Rd ) where P(Rd ) stands for the set of parts inside Rd .3: Non-cylindrical space-time domain be viewed as an application of the shape optimization theory for space-time manifolds. Tt t=τ Q0 t Ωt Σ0 Σ t=τ Q Γt t=0 Ω0 Γ 0 Rd x t=0 Ω0 Γ0 Rd x FIGURE 1. . In fact the tube Q is the graph in R×P(Rd ) ⊂ Rd+1 of the shape mapping S. any subset Q ∈ R × P(R d ) will not be a tube. S(t) ) dt Obviously. it is equivalent to the problem of finding the optimal tube Q= 0<t<τ {t} × Ωt ∈ Rd+1 . it becomes a graph.20 Moving Shape Analysis t.

Then the d+1 dimensional shape optimization analysis fully applies and the so called Shape differential Equation furnishes descent direction.24) can be transformed as V∈V min J(Q(V)) (1. τ ) × D → D. it furnishes the existence of a vector field Z∗ such that. for some α > 0. x). the minimization problem (1. t. . Γt = Tt (Γ0 ) Two major class of such mappings are given by : • the Lagrangian parametrization. x∈D In this case.e. The Rd -approach In order to carry out the sensitivity analysis of functionals depending on the def tube Q . In this case.e. ∂t Tt (x) = V(t. Tt (x)). The sensitivity analysis is then classically performed by considering horizontal vector fields ˜ Z(s. Tt = θ(t. x) ∈ (0. Wen we consider the tube Q as a subset of Rd+1 . Z(s..26) . τ ) × D Tt=0 (x) = x. x)) ∈ Rd+1 where s is the perturbation parameter of the tube.24) can be transformed as min J(Q(θ)) (1. i. Ωt = T( V )(Ω0 ). i.25) θ∈Θ • the Eulerian parametrization. the minimization problem (1. s J(Qs ) ≤ J(Q) − α Z∗ (σ) 2 dσ 0 ∀s > 0. x) = (0. t. where the transformation is the flow associated to a given velocity field V(t. We show that the existence of that field Z∗ induces the existence of a usual vector field V(t. we assume that the domains are the images of the domain Ω 0 = Ωt=0 ¯ ¯ through a given family of smooth maps Tt : D → D .. i.) ¯ ¯ where θ : (0.e. the control problem becomes a usual shape optimization problem ( as far as no real time consideration enters). Ωt = Tt (Ω0 ).Introduction 21 to each tube Q the non-empty closed convex set of speed vector fields V which transport (in a weak sense) the characteristic function of the moving domain. x) ∈ Rd which builds that tube. (t..

27). Hence. Indeed. We would like to solve the following optimization problem (ξ. noted Vξ . no uniqueness result has been obtained for this smoothness level. the Lipschitz regularity of the velocities V can be weakened using the equations satisfied by the characteristic functions ξ(t. when the field V and its divergence are simply L 1 functions. V) + F (ξ. Consequently.29) . V) equipped with a parabolic BV like topology for which the constraint (1. τ ) × D ξt=0 = χΩ . Rd )) and the divergence positive part (div V)+ ∈ L1 (0. In the non-smooth case. In this case. V).22 Existence of tubes Moving Shape Analysis In the smooth case. the new problem writes (ξ. the element Vξ is the unique (with minimal norm) vector field associated to ξ via the convection equation (1. V) (1. L∞ (D)). we can define the unique minimal norm energy element V ξ in the convex set Vξ .28) The space Uad is a space of smooth velocities. BV(D)) together with vector fields V ∈ L2 (0. we consider the set of fields W such that (ξ. For the time being.27) defines a closed subset TΩ which contains the weak closure of smooth elements This approach consists in handling characteristic functions ξ = ξ 2 which belongs to L1 (0. ∂t ξ + ξ · V = 0. We choose to adopt a different point of view inspired by the optimization problems framework. such a problem does not admit solutions and we need to add some regularization terms to ensure its solvability. x) associated to the domain Ω t (V). Actually. W) ∈ TΩ .27). Rd )) solution of problem (1. (0.V)∈TΩ TΩ = def (χΩ ◦ T−1 (V). It forms a closed convex set. V) (1. we are able to derive an existence result of solutions with initial data given in H −1/2 (D). The usual situation can be described as follows. In this case. L2 (D.27) does not make sense. For a given element (ξ. using a Galerkin approximation and some energy estimates. For a given tube ξ. V)| V ∈ Uad t inf J(ξ. τ . we shall introduce different penalization terms which furnish compactness properties of the minimizing sequences inside an ad-hoc weak topology involving bounded variation constraints. L2 (D. τ . In most situations. the notion of weak solutions associated to the convection problems (1. the correct modeling tool for shape evolution is to introduce the product space of elements (ξ = ξ 2 .V)∈TΩ inf J(ξ.27) We shall consider velocity fields such that V ∈ L1 (0. the existence of tubes follows the Cauchy-Lipschitz theory on differential equations [147. τ . τ . our final goal is to apply the weak set evolution setting to the control problem arising in various fields such as free boundary problems or image processing. 51]. V) ∈ T Ω . Let us consider a given smooth enough functional J(ξ. Then. D (1.

we are interested in differentiability properties of integrals defined over moving domains. Nevertheless. The transverse t map Tρt is the flow of a transverse field Zρ defined as follows: t Zρ = Z t (ρ. The following result states that the transverse map T ρt can be considered as a dynamical flow with respect to the perturbation variable ρ. it uses the compactness properties of tube family with bounded perimeters in R d+1 . .Introduction The penalization term F (ξ. in Ωt (V) (1.. x) ≡ Ttρ (Zρ )(x) with dx(ρ) = Z t (ρ. In particular. Rd ). This approach easily draws part of the variational properties associated to the bounded variation functions space framework. developed in [155]. V) can be chosen using several approaches: 23 • We can first consider the time-space perimeter of the lateral boundary Σ of the tube. ρ ≥ 0 dρ x(ρ = 0) = x.30) i. We shall extend these results to the case of vector fields living in L 2 ((0.31) . Tube derivative In this paragraph. τ ) × D. W) ∈ U is defined as follows. only existence results for solutions of the convection equation can be handled and the uniqueness property is lost. • We can rather consider the time integral of the spatial perimeter of the moving domain which builds the tube.e. J(Q(V)) = Q(V) f (V) dx dt The transverse map Tρt associated to two vector fields (V. is the solution of the following dynamical system : t Ttρ (Zρ ) : Ωt −→ Ωρ t t x −→ x(ρ. In this case.) = def ∂Tρt ∂ρ ◦ (Tρt )−1 (1. this method leads to heavy variational analysis developments.3 The transverse map allows us to perform sensitivity analysis on functions defined on the unperturbed domain Ω t (V). as introduced in [157]. x(ρ)). Tρt : Ωt −→ Ωρ = Ωt (V + ρW) t def x → Tt (V + ρW) ◦ Tt (V)−1 REMARK 1.

The derivative with respect to ρ at point ρ = 0 of the following composed function. J (V) · W = Futhermore. if ii) f (V) admits an Eulerian shape derivative given by the following expression. D def (1.34) then J (V) · W = Ωt (V) Ωt (V) f˙(V) · W + f (V) div Z dΩ (1. V). we set Z(t) = Zρ=0 . ∂t Z + [Z. We shall use the notation. def d ρ f˙(V) · W = f˙(V. f (V) · W = f˙(V) · W − f (V) · Z (1. let us assume that for any direction W ∈ U the following hypothesis holds. we will mainly consider derivatives of perturbed functions at point def t ρ = 0.35) Furthermore. if Ω0 is an open domain with a Lipschitzian boundary Γ 0 .33) [f (V) · W + div(f (V) Z)] dΩ (1. For a bounded measurable domain Ω0 with boundary Γ0 . τ ) × D Zt=0 = 0. W) is called the eulerian material derivative of f (V) at point V ∈ U in the direction W ∈ U. V] = W. A fundamental result lies in the fact that Z can be obtained as the solution of a linear time dynamical system depending on the vector fields (V. then J (V) · W = f (V) · WdΩ + f (V) Z · n dΓ (1. f ρ : [0. ρ0 ] → H(Ωt (V)) ρ → f (V + ρW) ◦ Tρt f˙(V. (0. V] = D Z · V − D V · Z stands for the Lie bracket of the pair (Z. The vector field Z is the unique solution of the following Cauchy problem.32) where [Z.36) Ωt (V) Γt (V) . we can state the differentiability properties of noncylindrical integrals with respect to their moving support.24 Moving Shape Analysis Since. W) ∈ U. i) f (V) admits an eulerian material derivative f˙(V) · W then J(V) is Gˆteaux differentiable at point V ∈ U and its derivative is given a by the following expression. W) = f dρ ρ=0 With the above definition.

In this case. ∗ Actually. It is possible to define the solution of the adjoint transverse system. the adjoint field Λ is supported on the moving boundary Γt (V) and has the following structure. (0. Λ = ∗ γΣ(V) (λ n) −∂t λ − Γ λ · V − (div V)λ = f.42) .For F ∈ L2 (0.39) with f = E. Q(V)  u = u∞ . Σ(V)    u(t = 0) = u0 . −∂t Λ − D Λ · V − ∗ D V · Λ − (div V)Λ = F.4 The last identity will be of great interest while trying to prove a gradient structure result for general non-cylindrical functionals. Then in chapter 2 . (1.37) REMARK 1. (L2 (D))d ) solution of the backward dynamical system. we will establish the following adjoint identity.Introduction 25 REMARK 1.   ∂t u + D u · u − ν ∆u + p = 0. there exists a unique field Λ ∈ C 0 ([0. τ ) Λ(τ ) = 0 (1. τ .39) p is the canonical projection on Γt (V) and χΩt (V) is the characteristic function of Ωt (V) inside D. τ ]. Ω0 (1. p) is a solution of the following system. j(V) = Q(V) |u − ud |2 (1. (0. Eulerian dynamical shape control We come back to our design problem. (H 1 (D))d ).38) where λ is the unique solution of the following boundary dynamical system.40) where λ is the unique solution of problem (1. τ 0 Γt (V) τ EZ·n=− 0 Γt (V) λW · n (1. Γτ (V) (1. τ ) × ∂D   u = V. we shall deal with specific right-hand sides F of the form F(t) = γΓt (V) (f (t) n). Σ(V) λ(τ ) = 0. Q(V)    div u = 0.5 The field Λ is the dual variable associated to the transverse field Z and is a solution of the adjoint problem associated to the transverse dynamical system.41) where (u. We would like to minimize the following tracking functional with respect to V.

One of them is the use of a min-max formulation involving a Lagrangian functional coupled with a function space embedding.45) 1 f = [−(σ(v. q) (1. r.43) and λ is the adjoint transverse boundary field. we will prove in chapter 5 that this functional is differentiable and its gradient has the following form. Ωτ (V) (1. φ. j(V) = ∗ γΣ(V) − λ n − σ(v. τ ) × ∂D (1. r. Let us give here the main steps of this proof. We rewrite the minimization problem as a min-max problem. φ.47) .r)∈X×P min (φ.4: Shape optimization Using the Eulerian shape analysis framework rapidly sketched in the previous section. q) (1. Q(V)  v = 0. π) stands for the adjoint fluid state solution of the following system. Σ(V) λ(τ ) = 0.q)∈Y ×Q max LV (ψ. q) = JV (ψ. r. r) − eV (ψ. −∂t λ − λ · V − (div V) λ = f. π) · n)] · (D V · n − D u · n) + |u − ud |2 (1.46) 2 This result can be obtained using several techniques. φ. Σ(V)    v(τ ) = 0.44)   v = 0. Γτ (V) with where (v. Q(V)    div(v) = 0.26 Moving Shape Analysis Tt(V) Ω0 Ωt D D Γ0 ∂D ∂D Γt FIGURE 1. j(V) = with LV (ψ.48) (ψ. π) · n (1. (0.   −∂t v − D v · u + ∗ D u · v − ν ∆v + π = −(u − ud ). solution of the tangential dynamical system. particulary suited for non-homogeneous Dirichlet boundary problems.

p. τ . H 1 (D)) The saddle points (u.20) and to the adjoint system (1. v. W = We can state j (V). We finally use the transverse adjoint identity. X = Y = H 1 (0. H 2 (D)). π) are solutions of the first-order optimality conditions with respect to the multipliers (φ. W = Σ(V) def def def d j(V + ρW) dρ = ρ=0 d ρ L dρ (u. def In this case. The state and multiplier variables are defined on the hold-all domain D. Since the state and multiplier variables are defined in the hold-all domain D. we use a Lagrangian parametrization of the shape motion. τ ). Using the min-max derivation principle [51] recalled in chapter 5. then j (V). the functional j(θ) possesses a gradient j(θ) which is supported on the moving boundary Σ(θ) and can be represented by the following expression. This adjoint variable is only supported by the moving boundary Γ t (V) over (0.46) and where Z stands for the transverse vector field introduced previously. q) and the state variables (ψ.49) f Z · n + − σ(v. the pertubed Lagrangian L ρ only involves perturbed supports. q) · n stands for the weak fluid state operator. v.44).45) which corresponds to the adjoint system associated to the transverse dynamical system satisfied by Z. φ. π) ρ=0 (1. Σ(V) fZ·n=− Σ(V) λ W · n. r) which leads respectively to the primal system (1.Introduction and eV (ψ. p. π) · n · W with f given by equation (1. τ . P = Q = H 1 (0. The crucial point concerns the derivation with respect to the design variable V ∈ Uad . r. ∀ W ∈ Uad where λ is solution of equation (1. 1 j(θ) = ∗ γΣ(θ) ∂t E + (divΓ V) E + D E · V + E · D u + |u − ud |2 n (1. Lagrangian dynamical shape control In chapter 6.50) 2 .e. We consider a perturbation vector field W ∈ Uad with an increment parameter ρ ≥ 0.. i. we prove that for V = ∂t θ ◦ θ −1 ∈ Uad . q) = Q(V) 27 [∂t ψ + D ψ · ψ − ν∆ψ + − r] · φ − Σ(V) q div ψ Q(V ) (ψ − V) · σ(φ.

4.51)   v = 0.4 Control of an elastic solid inside a fluid flow A simple mechanical model We consider our model problem where now the solid is supposed to be a two dimensional elastically supported rigid motion.   −∂t v − D v · u + ∗ D u · v − ν ∆v + π = −(u − ud ). π) stands for the adjoint fluid state solution of the following system. we have Ωf = Tt (Ωf ). x∈D Hence. Σ(θ)    v(τ ) = 0. Ωτ 1. ∂t Tt (x) = V(t. Q(θ)  v = 0. Q(θ)    div(v) = 0. π) · n and (v. τ ) × D Tt=0 (x) = x. we only consider one degree of freedom for the structural motion : the vertical displacement d(t) e2 where e2 is the element of Cartesian basis (e1 . x) ∈ (0. Ωs = Tt (Ωs ) t t 0 0 Since we only consider one degree of freedom motion. (t. e2 ) in R2 . where the transformation is the flow associated to a given velocity field V(t. Tt (x)). As previously. we use an Eulerian parametrization. x). we write Tt(V) Ωf 0 Ωf t Ωs 0 Γs 0 ∂D D Ωs t Γs t ∂D D FIGURE 1. (0.5: Rigid solid inside a fluid flow Ωs = Ωs + d(t) e2 t 0 .28 Moving Shape Analysis where E = σ(v. For the sake of simplicity. τ ) × ∂D (1.

55) where σ(u. x) = Ext(d(t)e2 ).   u(t = 0) = u0 .56) .Introduction We set Σs ≡ ({t} × Γs ). the projected fluid loads Ff have the following expression : Ff = − Γs t σ(u. t The fluid is assumed to be a viscous incompressible Newtonian fluid. The couple (u. The solid is described by the evolution of its displacement and its velocity ˙ and the couple (d. p) satisfies the classical Navier-Stokes equations written in non-conservative form   ∂t u + D u · u − ν ∆u +   div(u) = 0. d (t = 0) = [d0 . p) = −p I +ν (D u + ∗ D u) stands for the fluid stress tensor inside Ωf . we can give an example of an appropriate flow vector field :  ˙ x ∈ Ωs  V(t. Its evolution is described by its velocity u and its pressure p. p = 0. t We complete the whole system with kinematic continuity conditions at the fluid-structure interface Γs : t def ˙ u = V = d e2 . Hence. k) stand for the structural mass and stiffness.52) V(t. ˙ d. τ ) × ∂D t where Ext is an arbitrary extension operator from Γ s into Ωf .54) where ν stands for the kinematic viscosity and u∞ is the farfield velocity field.  u = u∞ .53) where (m. Qf ≡ t 29 0<t<τ 0<t<τ ({t} × Ωf ) and Σf ≡ (0. x ∈ ∂D (1. F f is the projection of the fluid loads on Γs along the direction of motion e2 . d1 ] In our simple case. x ∈ Ωf t  V(t. d) is the solution of the following ordinary second order differential equation : ¨ m d + k d = Ff . Qf (V) Qf (V) Σf Ωf 0 (1. p) · n · e2 (1. The map Tt 0 0 is usually referred to as the Arbitrary Euler-Lagrange map. t ˙ (1. x) · n = 0. on Σ(V) (1. x) = d(t)e2 .

π) · n with (v. p. (0. p.61) 12 It . d0 .30 Moving Shape Analysis To summarize. p) · n · e2 . we prove in chapter 7 that the tracking functional is differentiable and its gradient takes the following form. p. That means that we would like to control the motion of the solid thanks to the far-field boundary conditions 12 .    div(u) = 0. (1. we get the following coupled system :  Qf (V)  ∂t u + D u · u − ν ∆u + p = 0. d. Σf   v = b e2 . d] (u∞ ) with [u.57) Optimization problem Here we choose the control variable as the far-field Dirichlet boundary condition on Σf . min j(u∞ ) u∞ ∈ U c (1. The goal is to minimize a functional depending on the state variables of the coupled fluid-structure problem. Sensitivity analysis Using the Eulerian sensitivity analysis introduced in the last section.59) This functional is a tracking functional. Qf (V)     u = u∞ . d] (u∞ ) is a weak solution of problem (1. ˙ Ω f × R2 0 (1. The purpose is to illustrate how the moving shape analysis presented in the previous section can be applied to the analysis of optimization problems for various coupled systems involving a moving interface. τ )    Γs t     u. Qf (V)    div(v) = 0. Qf (V)  v = 0. Σs (V)    v(τ ) = 0. d (t = 0) = [u0 .58) where j(u∞ ) = Ju∞ [u.57) and Ju∞ is a real functional of the following form. Ωf (V) τ can also be interpreted as an inverse problem. Ju∞ [u. j(u∞ ) = ∗ γΣf σ(v.60) (1. d] (u∞ ) = 1 2 τ 0 |d − dg |2 dt (1. π.   −∂t v − D v · u + ∗ D u · v − ν ∆v + π = 0. Σs (V)   ¨  md + kd = − σ(u. Σf    ˙ u = d e2 . d1 ] . b) solutions of the following adjoint system.

d2 ) − eu∞ (u. v. The minimization problem can be put into a Min-Max formulation involving the Lagrangian functional defined as follows. p. the reader is referred to this chapter for more details since it requires a heavier analysis. π) · n    Γs (V)  t   31 · e2 (1. b2 ) where eu∞ (u. For the sake of simpleness. Then derivating first-order optimality conditions with respect to the state variables (u. π. b2 ) = Ju∞ (u. p. d1 . d1 . we consider a more general model where the solid is supposed to be a 3D non-linear elastic solid. def def ˙ |d|2 (D v · e2 ) · e2 − ν (D v · n) · (D u · n) · n (0. d1 .60) is obtained using a bypass through the Min-Max problem. d1 . d2 . p. where the adjoint state b1 has been eliminated to the benefit of b = b2 . It involves the transverse adjoint state λ as an intermediate sensitivity multiplier that can be eventually eliminated.62).61)-(1. p. π. d2 ) stands for the global weak state operator and where the displacement d1 and the velocity d2 are considered as independent variables.62) The proof of this result is based on the Min-Max derivation principle. d2 ). 0] ˙ . d2 ) leads to the fluid and solid adjoint systems (1.Introduction    ¨  m b + k b = (d − dg ) + ∂t σ(v. p. b (τ ) = [0. A general coupled fluid-structure model In chapter 8. b1 . Lu∞ (u. d1 . b1 . The keystone is the sensitivity of the Navier-Stokes system with respect to pertubations of the solid velocity V = d2 e2 as described in the last section. (v. τ ) + s   Γt (V)       b. Hence the expression of the cost function gradient in equation (1.

.

τ ) × Γs (t. We make use of the transverse derivative concepts introduced in [154. These concepts will be illustrated on a model problem arising in free boundary systems. x) : (0. x) · ns (x) = 0. we consider the identification of a moving boundary that represents the isothermal interface between a solid phase and a liquid phase. 144].Chapter 2 An introductory example : the inverse Stefan problem In this chapter.   {t} × Γf (t) (2.1 The mechanical and mathematical settings described in Figure (2. (t. We consider the velocity field V(t. x) ∈ Σf = def t∈(0. x ∈ Γf (t). This problem is referred in the literature as the inverse Stefan problem [61. It is surrounded def     V(t. t). 2. we shall introduce various concepts that will be further developed in the core of this lecture note. x) ∈ Σs = (0. 155]. The motion of the interface Γf (t) is characterized by its scalar Eulerian normal velocity v f (x. by a solid phase Ωs (t) = D \ Ωf (t) with boundary ∂Ωs (t) = Γs ∪ Γf (t). as def We consider a fluid phase located at time t ≥ 0 in a domain Ω f (t) ⊂ D with boundary Γf (t) where D ⊂ Rd is a hold-all domain. from measurements on a fixed part of the solid boundary.1) 33 . More precisely.1) where Γs = ∂D. x) · nf (t.τ ) def  V(t. x) = v f . τ ) × D → Rd satisfying the following properties.

L ipk (V(t)) ≤ c. Tt (V) : D → Rd (2.34 Moving Shape Analysis Tt(V) Ωs 0 Ωf 0 Ωs(t) Ωf (t) Γf (t) Γf 0 Γs Γs FIGURE 2. X) : [0. τ ]. V0 (D) . k V0 (D) = def ∃ c > 0.1: Melting front propagation We consider an admissible space1 Uad for the velocity field. τ ] dt x(t = 0) = X. x(t)). ∀ t ∈ [0. t ∈ [0. X) def 1 Uad = def with k ≥ 0 and where n ` ´ k V ∈ C 0 [0. X ∈ D. dx (t) = V(t.2) X → Tt (V)(X) = xV (t. on D def y=x L ip(Dα V) for k ≥ 1 and L ip(V) = sup |V(y) − V(x)| . τ ] o with L ipk (V) = def |α|=k X n o ` ´ ¯ d V ∈ C k (D) . |y − x| .The map V ∈ Uad can be viewed as a non-autonomous velocity field {V(t) : 0 ≤ t ≤ τ } : V(t) : D → Rd def x → V(t)(x) = V(t. τ ] → Rd solution of the following Cauchy problem. We build the corresponding Lagrangian flow transformation. V · n∂D = 0. x) Let us define the field xV (t.

Qs ≡ ({t} × Ωs (t)).     0<t<τ ∂n y = f. W k+1. . τ ]. we shall consider the velocity field V as the main variable def characterizing the evolution of the melting front and define v f = V · nf on Σf .1) imply that T t (V) is a diffeomorphism2 that maps D into D.) on the fixed solid boundary Γs and it is assumed that the temperature is equal to a fixed phase transition temperature yf on the moving melting interface Γf (t) for t ∈ (0. L2 (Ωs (t)) (2. Ωs (t) = Tt (V)(Ωs ) 0 0 Γf (t) = Tt (V)(Γf ). y0 ) ∈ L2 (0. L2 (Γs ) × L2 (Ωs ).3)   y = yf . For a smooth velocity V ∈ Uad and the data (f. τ ).2 The inverse problem setting For a given evolution of the melting interface. For a given evolution of the boundary Γf (t). 60] that the conditions (2. We impose a given heat flux f (t. τ ). Σs . Hyf . τ ). In particular Ωf (t) = Tt (V)(Ωf ).   ∂t y − ∆y = 0.    y(t = 0) = y0 .4) where 1 Hϕ. τ ). we consider the solution y of the heat equation inside the solid phase. Σf . W k+1. (2.An introductory example : the inverse Stefan problem 35 It is shown in [51. τ ]. This leads to the following non-cylindrical system. there exists a unique solution [59] 0 1 y ∈ L2 (0. Ωs 0 The space-time domain Qs is called a tube. 2 ` ` ´ ´ ` ` ´ ´ ¯ d ∩ C 0 [0. we consider the solution of the heat equation (2.∞ (D) d T(V) ∈ C 1 [0. C k (D) .∞ (D) ` ` ´ ´ ¯ d T−1 (V) ∈ C 1 [0.3) and we consider its trace on the fixed boundary Σ s . τ ).Γf (t) (Ωs (t)) ∩ L∞ (0. φ = ϕ on Γ 2. Γs = Tt (V)(Γs ) 0 In the sequel. τ ]. C k (D) ` ` ´d ´ V ◦ T(V) ∈ L∞ (0.Γ (D) = def φ ∈ H 1 (D).

THEOREM 2. in O (2.8) .2: Non-cylindrical space-time domain On the mathematical viewpoint.5) It is a non-linear ill-posed inverse problem that can be solved using a leastsquare minimization problem regularized thanks to a Tikhonov zero order term. we will prove the following result. V∈Uad where y(V) stands for the solution of equation (2.7) with α > 0. The inverse Stefan problem consists in recovering the evolution of the melting front Γf (t) from the knowledge of the temperature on the fixed solid boundary Γs . we consider the functional j(V) = 1 2 τ 0 Γs |y(V) − yd |2 ds dt (2.3) and γ Σs is the zero order trace operator on Σs .Hence we look for the solution V of the following optimization problem.1 For V ∈ Uad . L2 (Γs ) . we look for V ∈ Uad such that O(V) = yd . τ ). L2 (Γs ) and the observation operator. yd ∈ O.6) min 1 O(V) − yd 2 2 O + α V 2 2 Uad (2. τ ).36 Moving Shape Analysis Tt(V) t=τ Qs 0 Qf 0 Σf 0 t t=τ Qs Qf Γf (t) Σf t=0 Γ s Ωs 0 Ωf Γ f 0 0 Rd x t=0 Γ s Ωs 0 Ωf 0 Γf 0 Rd x FIGURE 2. In the sequel. we introduce the observation space O = L2 (0. O : Uad → O def def V → O(V) = γΣs (y(V)) (2. This means that for a given temperature yd ∈ L2 (0.

−∂t λ − Γ λ · V − (div V) λ = ∂n y ∂n ϕ. j(V) = def [DV [j](V)] · W = lim def ρ→0 1 j(V + ρ W) − j(V) ρ (2.10) where the transverse adjoint state λ ∈ C 0 [0. Let us choose a perturbation direction W ∈ U ad . for the sake of simpleness. H0. Σs (2.15) ∀ φ(V) ∈ L 1 (0.      ∂n ϕ = y(V) − yd . τ ]. we have dropped the regularizing term α 2 2 V Uad .  Q(V)  ∂t ϕ + ∆ϕ = 0.9) j(V) is differentiable with respect to V and its gradient has the following structure ∗ j(V) = −∗ γΣ(V) λ n .3).   ∂n y = 0. Hence. Σ(V).  y(t = 0) = y0 .An introductory example : the inverse Stefan problem where y(V) is the solution of the following system   ∂t y − ∆y = f. τ [∂t y(V) φ(V) + 0 Ωt (V) 2 y(V) · φ(V)] dx dt = 0. Q(V). We would like to compute the directional derivative of j. Σs .13) O 2 where.14) Then the goal is to evaluate the directional derivative of the element y(V) which is a solution of the moving heat equation (2. we need to evaluate the gradient with respect to V of the functional 1 O(V) − yd 2 (2. Ωτ (V) 2.   y = yf .12)  ϕ = 0. Ω0 37 (2. H 1 (Γt (V)) is the solution of the backward dynamical system.Γt (V) (Ωt (V)) .11) and where the adjoint state ϕ is the solution of the following backward system. τ ). Σ(V)      ϕ(t = τ ) = 0. (2. Σ(V) λ(t = τ ) = 0. in Uad (2. In order to do so. we write the associated variational formulation satisfied y(V). Γτ (V) (2.3 The eulerian derivative and the transverse field A possible choice in order to solve the above minimization problem is to use a gradient based method such as the conjugate gradient method.

15). y d . it is natural to introduce the transformation between Q(V) and Q(V + ρ W). y0 ) = (0. 0) together with Ωt (V) = Ωs (t) and Γt (V) = Γf (t). Looking at equation (2. x) : Ωt (V) → Ωt (V + ρ W) ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡¡¤¡¤¡¤¡¤¡¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¡¤¡¤¡¤¡ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡£¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡£¤ £¤£¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¤ ¡£¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡£¤ ¡¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡ ¡£¤¤¡¤¡¤¡¤¡¤¡£¤¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡£¤¤¡¤¡¤¡¤¡£¤¤ ££¤¤ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡£¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡£¤ ¤£¤£¤ ¡¤£¤£¡¤¡¤¡¤¡¤¡¤£¤£¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤£¤£¡¤¡¤¡¤¡¤£¤£ ¡¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ ¡££¤¡£¤¡£¤¡£¤¡£¤¡££¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡££¤¡£¤¡£¤¡£¤¡££¤ £¤££¤ ¡£¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡¤¡£¤¡¤¡¤¡¤¡£¤ ¡££¤¤£¤¡¡¡¡¡££¤¤£¤¡¡¡¡¡¡¡¡¡££¤¤£¤¡¡¡¡££¤¤£¤ ¡¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡£¤¡¡£¤¡£¤¡£¤¡ ¤£¤££¤ ¡¡£¡£¡£¡£¡¡£¡£¡£¡£¡£¡£¡£¡£¡¡£¡£¡£¡ Q0 Ω0 def Q(V + ρ W) = Ωt (V + ρ W) = Tt (V + ρ W)(Ω0 ) FIGURE 2. A canonical choice is furnished by ¡¢ ¢ ¢ ¢ ¢ ¢ ¢ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡ ¡¢¡¡¢¡¡¢¡¡¡¢¡¡¢¡¡¢¡¡¢¡¡¡¡¡  ¢ ¢¡ ¡ ¢¡ ¡ ¢¡ ¡ ¢¡ ¢¡ ¡ ¢¡ ¡ ¢¡ ¡ ¢¡ ¡ ¢¡ ¢¡¡ ¢¡ ¢  ¢ ¢ ¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡ ¢¡¡ ¢¡ ¢ ¡¢¡¡¢¡¡¢¡¡¡¢¡¡¢¡¡¢¡¡¢¡¡¡¡¡   ¢¢¡ ¡  ¢¢¡ ¡  ¢¢¡ ¡  ¢¢¡  ¢¢¡ ¡  ¢¢¡ ¡  ¢¢¡ ¡  ¢¢¡ ¡  ¢¢¡  ¢¢¡¡  ¢¢¡  ¢¢ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡   ¢¢ ¡¢¡¡¢¡¡¢¡¡¡¢¡¡¢¡¡¢¡¡¢¡¡¡¡¡ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡ ¡¢¡¡¢¡¡¢¡¡¡¢¡¡¢¡¡¢¡¡¢¡¡¡¡¡ ¢ ¢ ¢¡¡¢ ¢ ¢¡¡¢ ¢ ¢¡¡¢ ¢ ¢¡¢ ¢ ¢¡¡¢ ¢ ¢¡¡¢ ¢ ¢¡¡¢ ¢ ¢¡¡¢ ¢ ¢¡¢ ¢ ¢¡¡¢ ¢ ¢¡¢ ¢ ¢ ¢ ¢ ¢ ¡ ¢¡¡ ¢¡¡ ¢¡¡¡ ¢¡¡ ¢¡¡ ¢¡¡ ¢¡¡¡¡¡ ¢¡¡¢¡¡¢¡¡¡¢¡¡¢¡¡¢¡¡¢¡¡¡¡¡ ¡¢ ¡¡¢ ¡¡¢ ¡¡¡¢ ¡¡¢ ¡¡¢ ¡¡¢ ¡¡¡¡¡ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡  ¢  ¢¡¢¡ ¢  ¢¡¢¡ ¢  ¢¡¢¡ ¢  ¢¡ ¢  ¢¡¢¡ ¢  ¢¡¢¡ ¢  ¢¡¢¡ ¢  ¢¡¢¡ ¢  ¢¡ ¢  ¢¡¡ ¢  ¢¡ ¢  ¢ ¡ ¢¡¡ ¢¡¡ ¢¡¡¡ ¢¡¡ ¢¡¡ ¢¡¡ ¢¡¡¡¡¡  ¢  ¢   ¢¢ ¢¡¡  ¢¢ ¢¡¡  ¢¢ ¢¡¡  ¢¢ ¢¡  ¢¢ ¢¡¡  ¢¢ ¢¡¡  ¢¢ ¢¡¡  ¢¢ ¢¡¡  ¢¢ ¢¡  ¢¢ ¢¡¡  ¢¢ ¢¡  ¢¢ ¢   ¢¢ ¢ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡ ¢ ¢ ¢ ¢ ¢ ¢ ¢ ¡ ¡¡ ¡¡ ¡¡¡ ¡¡ ¡¡ ¡¡ ¡¡¡¡¡ Γt(V) Ω0 Q(v + ρ W) ¡¦ ¦ ¦ ¦ ¦ ¦ ¡¥¡¡¥¡¡¡¥¡¡¥¡¡¥¡¡¡¥¡ ¡¦¡¡¦¡¡¡¦¡¡¦¡¡¦¡¡¡¦¡ ¥¦¥¦¡¥¡¥¦¡¥¡¥¦¡¥¦¡¥¡¥¦¡¥¡¥¦¡¥¡¥¦¡¥¦¡¥¡¥¦ ¡¦¡¥¦¡¦¡¥¦¡¥¦¡¦¡¥¦¡¦¡¥¦¡¦¡¥¦¡¥¦¡¦¡¥¦ ¡¥¦¡¡¥¦¡¡¡¥¦¡¡¥¦¡¡¥¦¡¡¡¥¦¡ ¦¡¡¦¡¡¡¦¡¡¦¡¡¦¡¡¡¦¡ ¥¦¥¦¦¡¡¥¦¥¦¦¡¡¥¦¥¦¦¡¥¦¥¦¦¡¡¥¦¥¦¦¡¡¥¦¥¦¦¡¡¥¦¥¦¦¡¥¦¥¦¦¡¡¥¦¥¦¦ ¡¦¡¡¦¡¡¡¦¡¡¦¡¡¦¡¡¡¦¡ ¡¥¡¡¥¡¡¡¥¡¡¥¡¡¥¡¡¡¥¡ ¡¥¡¡¥¡¡¡¥¡¡¥¡¡¥¡¡¡¥¡ ¡¥¡¡¥¡¡¡¥¡¡¥¡¡¥¡¡¡¥¡ ¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥¡¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥¡¥¦¥¦¥¦¥¡¡¥¦¥¦¥¦¥ ¡¦¡¡¦¡¡¡¦¡¡¦¡¡¦¡¡¡¦¡ ¡¥¡¡¥¡¡¡¥¡¡¥¡¡¥¡¡¡¥¡ ¦ ¦ ¦ ¦ ¦ ¦ ¡¥¦¡¡¥¦¡¡¡¥¦¡¡¥¦¡¡¥¦¡¡¡¥¦¡ ¡¥¦¥¡¡¥¦¥¡¡¡¥¦¥¡¡¥¦¥¡¡¥¦¥¡¡¡¥¦¥¡ ¥¦¥¦¡¡¥¦¥¦¡¡¥¦¥¦¡¥¦¥¦¡¡¥¦¥¦¡¡¥¦¥¦¡¡¥¦¥¦¡¥¦¥¦¡¡¥¦¥¦ where we have set without loss of generality. T t (ρ.cylindrical reference tube Q(V). it is clear that we need to establish how to differentiate the generic term 38 as described in Figure (2. 0. (f. x) = Tt (V + ρ W)] ◦ Tt (V)−1 (x) J(V) = Moving Shape Analysis def def 0≤t≤τ 0 τ Tt(V) Tt(V + ρ W) def def t Ωt (V) {t} × Ωt (V + ρ W) f (V) dx dt V(t) Γt(V + ρ W) V(t) + ρ W(t) ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡¡¨¡¨¡¨¡¨¡¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¡¨¡¨¡¨¡ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨ §¨§¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¨ ¡§¨§¡¨¡¨¡¨¡¨¡§¨§¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡§¨§¡¨¡¨¡¨¡§¨§ ¡¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡ ¡§¨¨¡¨¡¨¡¨¡¨¡§¨¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡§¨¨¡¨¡¨¡¨¡§¨¨ §§¨¨ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡§¨¡¨¡¨¡¨¡¨¡§¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡§¨¡¨¡¨¡¨¡§¨ ¨§¨§¨ ¡¨§¨§¡¨¡¨¡¨¡¨¡¨§¨§¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨§¨§¡¨¡¨¡¨¡¨§¨§ ¡¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ ¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨ §¨§§¨ ¡§¨§¡¨¡¨¡¨¡¨¡§¨§¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡¨¡§¨§¡¨¡¨¡¨¡§¨§ ¡§§¨¨§¨¡¡¡¡¡§§¨¨§¨¡¡¡¡¡¡¡¡¡§§¨¨§¨¡¡¡¡§§¨¨§¨ ¡¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡§¨¡¡§¨¡§¨¡§¨¡ ¨§¨§§¨ ¡¡§¡§¡§¡§¡¡§¡§¡§¡§¡§¡§¡§¡§¡¡§¡§¡§¡ .3: Perturbed tube x → T t (ρ. Since the function f (V) is defined on the non- with respect to V.3). we introduce the perturbated moving domain This family generates a perturbed tube Hence. the perturbated functional can be written as follows. To this end.

135]. it can be proven that T(V + ρ W) is continuously differentiable3 with respect to ρ and that the transverse map Tρt can be considered as the flow with respect to ρ of the transverse vector field Indeed the map Z ∈ Zad can be viewed as a non-autonomous velocity field : ` ´ def ¯ Zad = C 0 [0. τ ]. x) → Z(ρ)(t. ρ0 ]. x) as the solution of a dynamical system with respect to the parameter ρ ∈ [0. The easiest way to do so is to connect this problem to the classical shape derivative calculus handled inside the speed method framework [147. x)) = ∂ρ T t (ρ) ◦ T t (ρ)−1 (x) V(t) Q(V) Z(ρ) : (0. x)) Ωt (V+ρ W) Ωt (V) det D Tρt f (V + ρ W) ◦ Tρt dx dt t f (V + ρ W) dx dt f (V + ρ W) dx dt def V(t) + ρ W(t) .4: Transverse map = ∂ρ T(V + ρ W) ◦ T(V + ρ W)−1 (x) (t. Actually. (t. x) = Z(ρ. (C k−1 (D))d . (t.An introductory example : the inverse Stefan problem Tρt ¡¡      ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡       ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ 39 ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡             ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡             ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡             ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡             ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡ t Γt(V) ¡        ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡  ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡  ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡         ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ Γt(V + ρ W) Q(V + ρ W) ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡©¡¡¡¡¡©¡¡¡¡¡¡¡¡©¡¡¡¡¡© ©© © © © © © © © © © © © © © © © ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡¡©©¡©©¡©©¡©©¡¡©©¡©©¡©©¡©©¡©©¡©©¡©©¡¡©©¡©©¡©©¡©©¡ ¡©¡¡¡¡¡©¡¡¡¡¡¡¡¡©¡¡¡¡¡© ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ¡©©¡¡¡¡¡©©¡¡¡¡¡¡¡¡©©¡¡¡¡¡©© ©©© ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡© ¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡¡ ©© ¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡© ¡©¡¡¡¡¡©¡¡¡¡¡¡¡¡©¡¡¡¡¡© ¡©¡¡¡¡¡©¡¡¡¡¡¡¡¡©¡¡¡¡¡© ©© ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡¡©¡©¡©¡©¡¡©¡©¡©¡©¡©¡©¡©¡¡©¡©¡©¡©¡ ¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡©¡© ©©© ¡©©©¡¡¡¡¡©©©¡¡¡¡¡¡¡¡©©©¡¡¡¡¡©©© Ω0 3 in where we have performed a transport into the moving reference domain Ω t (V). This means that we need to identify a transverse velocity field that may generate the transverse map T t (ρ. J(V + ρ W) = Z(ρ. Now we shall need to differentiate the terms inside the integral with respect to ρ at point ρ = 0. τ ) × D → Rd = = def 0 0 0 τ τ τ t Tρ Ωt (V) FIGURE 2.

W) = lim f (V + ρ W) ◦ Tρt (V) − f (V) ρ→0 ρ (2. H0. t We define the field XZ (ρ. X (ρ))).40 Moving Shape Analysis We build the corresponding Lagrangian flow transformation. Using this definition. the field Z t (ρ) plays the role of the speed V(t) and the def (2. x) : [0. (t. x) → Tρ (Z)(t. we easily establish the following result. τ ).16) dρ  X (ρ = 0) = x. . τ ) × D → Rd with 0 ≤ ρ ≤ ρ0 .Γt (V) (Ωt (V)) . Hence. DEFINITION 2. x)) and we identify the transverse map T (ρ) with Tρ (Z). the following identity holds LEMMA 2. Here we define its non-cylindrical counterpart. the non-cylindrical shape derivative calculus reduces to apply the classical shape derivative calculus with the following identification : V(0) ↔ Z(t).17) In the classical shape derivative calculus.18) exists in Y(V). Tρ (Z) : (0. ρ0 ] → Rd solution of the following Cauchy problem.1 t ∂ρ det D Tρ (Zρ ) |ρ=0 = div Z(t) Compared with the classical speed method. the parameter ρ plays the role of the peuso-time t. W) in the direction W ∈ Uad if the following limit def f˙(V. (t. This allows writing the perturbated functional as follows.1 We say that a function f (V) ∈ Y(V) for V ∈ Uad admits an Eulerian material derivative f˙(V. x) = XZ (ρ. An example is furnished 1 by Y(V) = L2 (0. Let Y(V) be a given functional space defined on Q(V). ρ ∈ [0. Following this idea. x ∈ Ωt (V) (t. τ ] (2. the notion of material derivative has been introduced. τ def J(V + ρ W) = 0 Ωt (V) t t det D Tρ (Zρ ) f (V + ρ W) ◦ Tρ (Zρ ) dx dt transverse field Z(t) = Z t (ρ = 0) has the same role as V(0).   dX (ρ) = Z(ρ.

W) dx dt Ωt (V) (2. Tµ V + ρ W ) · ∂ρ Tµ (V + ρ W) dµ t + 0 W µ. This is realized thanks to the following result.20) 1. ∂ρ Tt (V + ρ W) = t 0 D(V + ρ W)(µ. V) ∈ Zad × Uad . τ ]. W ∈ Uad . τ DV [J(V)] · W = (div Z) f (V) + f˙(V. then there exists ρ0 > 0 such that the application ¯ d ]0. τ ]. We first prove the differentiability of the flow T(V + ρ W) with respect to ρ. LEMMA 2. . W) in the direction W ∈ Uad . (0. C k−1 (D) ) ρ → T(V + ρ W) is continuously differentiable and its derivative satisfies for any t ∈ [0.An introductory example : the inverse Stefan problem 41 LEMMA 2. ∂t Z + D Z · V − D V · Z = W. ρ0 [ −→ C 0 ([0. PROOF It consists of 4 steps : def (2. D REMARK 2. W) ∈ Uad × Uad .) = 0. Tµ (V + ρ W) dµ. THEOREM 2. . V] = D Z · V − D V · Z for the Lie bracket associated to (Z. we shall use the notation [Z.2 Let V ∈ Uad and f (V) ∈ Y(V) such that it admits an Eulerian material derivative f˙(V.1 In the sequel.3 Let us consider (V.2 [154. Then J(V) is differentiable with respect to V and its derivative is furnished by the following expression.19) 0 It remains to precise a specific rule to compute the transverse Z starting from the knowledge of V. 59] The transverse field Z is the unique vector in Zad such that ∂t Z + D Z · V ∈ Zad and is the solution of the following Cauchy problem. τ ) × D Z(t = 0.

Tµ (V))dµ + 0 D V(µ. Since we already have proven the unique solvability of equation (2.20) then Z ◦ T(V) belongs to Zad and satisfies equation (2.21). LEMMA 2. then the transverse field writes Z(ρ) = S(ρ) ◦ T(V + ρ W)−1 At ρ = 0. in Zad = C 1 ([0.42 def Moving Shape Analysis 2. (2. we use the above lemma to conclude.5 (i) Let us consider Z ∈ Zad such that (a) the following regularity holds ¯ ∂t Z + D Z · V ∈ C 0 ([0.4 ¯ The function S is the unique vector field. Then we prove the equivalence between equation (2.20). 4. (C k−1 (D))d ).21). (C k−1 (D))d ) (b) and it satisfies equation (2. Tµ (V)) · S(µ) dµ.21). Let us set S(ρ) = ∂ρ T(V + ρ W) . τ ]. satisfying t t def S(t) = 0 W(µ. . τ ]. (C k−1 (D))d ) (b) and it satisfies equation (2.21) 3. (ii) Let us consider S ∈ Zad solution of equation (2.20) and equation (2.21). we set S = Z ◦ T(V) We can state the following result LEMMA 2. then S ◦ T(V)−1 ∈ Zad is such that (a) the following regularity holds ¯ ∂t S ◦ T(V)−1 + D S ◦ T(V)−1 · V ∈ C 0 ([0. τ ].

An introductory example : the inverse Stefan problem 43 2.15) with respect to V. ∀ φρ ∈ Θ(ρ) where Θ(ρ) = def 1 φ ◦ (Tρ )−1 | φ ∈ L2 (0.23) div Z [∂t y φ + 0 Ωt (V) y· y· φ] + ∂t y − ˙ y · W − [Z. then we get τ 0 Ωt (V) def ∂t Tρ = −Fρ · V + (V + ρ W) ◦ Tρ −1 (yρ ◦ Tρ )Fρ ∂t Tρ (2. H0. The perturbated formulation writes τ 0 Ωt (V) Jρ (∂t yρ ) ◦ Tρ φρ ◦ Tρ + ( yρ ) ◦ Tρ · ( φρ ) ◦ Tρ dx dt = 0. The following identities are easy to check.Γt (V) (Ωt (V)) The following identities are easy to check. τ ). V] φ + y· φ− ˙ ∀φ ∈ Θ yDZ · φ− φ D Z dx dt = 0.4 The Eulerian material derivative of the state With the previous framework.22) Jρ ∂t y ρ − −1 y ρ F ρ ∂t T ρ φ + y ρ Fρ · φFρ dx dt = 0. we are ready to differentiate the variational formulation (2. def ∂ρ ∂t Tρ |ρ=0 = W − [Z. ∀φ ∈ Θ 1 where Θ = L2 (0.Γt (V) (Ωt (V)) . We set yρ = y(V + ρ W). ( yρ ) ◦ Tρ = −1 (yρ ◦ Tρ )Fρ (∂t yρ ) ◦ Tρ = ∂t (yρ ◦ Tρ ) − We set y ρ = yρ ◦ Tρ . Fρ = D Tρ and Jρ = det(Fρ ). . τ ). H0. the derivative of the variational ˙ formulation writes τ def −1 ∂ρ Fρ |ρ=0 = − D Z ∂ρ Jρ |ρ=0 = div Z ∂ρ Fρ |ρ=0 = D Z (2. V] Recalling that Tρ |ρ=0 = I and y = ∂ρ y ρ |ρ=0 .

27) exists in Y(V) and depends linearly on W. W) and def f (V. Σs (2.24)  n   ˙ = 0. ˙ Ω0 together with ∂t Z + D Z · V − D V · Z = W. τ ) × D Z(t = 0. W) = f˙(V. D (2.20). W) − f ·Z (2.26) 2. W) in the direction W ∈ Uad if it admits an Eulerian material derivative f˙(V. W) dx dt ˙ (2.5 The Eulerian partial derivative of the state Now. (0.44 Moving Shape Analysis Using the transverse dynamical system. the directional derivative of j(V) is given by the following expression τ [DV [j](V)] · W = 0 Γs y(V) − yd y(V. . we would like to identify the gradient j(V). where Z is the transverse field solution of the transverse dynamical system (2. DEFINITION 2. . To this end.) = 0. y · (∂t Z)] φ dx dt − y − (div Z) I + D Z + ∗ D Z · ∀φ ∈ Θ Performing integration by parts.2 We say that a function f (V) ∈ Y(V) for V ∈ Uad admits an Eulerian partial derivative f (V. we need to introduce several adjoint problems. Before proceeding. we shall obtain the strong formulation satisfied by the material Eulerian derivative y.25) Finally using the fact that Γs does not move. Σ(V) y    y(t = 0) = 0. it may be convenient to define the notion of Eulerian partial derivative. ˙  ˙ ˙  ∂t y − ∆y = −(div Z)∂t y + y · (∂t Z)     + div y (div Z) I −(D Z + ∗ D Z) Q(V)    ∂ y ˙ = y − (div Z) I + D Z + ∗ D Z · n. we get τ τ [∂t y φ + ˙ 0 Ωt (V) τ 0 Ωt (V) y· ˙ φ] dx dt + 0 Ωt (V) [div Z ∂t y − φ dx dt = 0.

An introductory example : the inverse Stefan problem 45 Going back to the derivation of the generic functional J(V). τ [−y(V) ∂t φ + 0 ∀ φ ∈ L (0.28) Furthermore if the boundary Γt (V) is Lipschitz. we are able to quote the following result.29) The main interest in the use of the Eulerian partial derivative is that it commutes with time and space derivatives. Ω0 Using that Z. LEMMA 2. n ds dt (2. τ ) × D Differentiation with respect to V leads to τ Ωt (V) 2 y(V) · φ] dx dt = 0. τ ) × D Integration by parts inside the last equation with φ ∈ D(Q(V)) leads to ∂t y − ∆y = 0. Then J(V) is differentiable with respect to V and its derivative is furnished by the following expression. we consider the variational formulation (2. W) dx dt + 0 Ωt (V) 0 Γt (V) f (V) Z. n = 0 on Γs . Q(V) y (t = 0) = 0. [−y ∂t φ + 0 Ωt (V) τ 0 Γs ∪Γt (V) 2 y · φ] dx dt y(V) · φ] Z. Γt (V) . Γs The Dirichlet boundary condition. W) in the direction W ∈ Uad .6 Let V ∈ Uad and f (V) ∈ Y(V) such that it admits an Eulerian partial derivative f (V.15) with φ ∈ L 2 (0. τ ) × D . then τ τ DV [J(V)] · W = f (V. the system satisfied by y . we get ∂n y = 0. n + [−y(V) ∂t φ + ∀ φ ∈ L (0. τ DV [J(V)] · W = [f (V. In order to obtain. W) + div(f Z)] dx dt 0 Ωt (V) (2. y = yf .

n . Σ(V)      ϕ(t = τ ) = 0. the derivative of j(V) writes τ [DV [j](V)] · W = 0 Γs y(V) − yd y (V. Using several integrations by parts the 4 stable ¯ by multiplication of functions in the space C k−1 (D). F · Z dx dt 0. τ . it turns that the derivative of j(V) can be written τ [DV [j](V)] · W = ∂n y ∂n ϕ Z. Ω0 Finally. Ω (V) τ Using several integrations by parts. Q(V)      ∂n y = 0.33) This expression does not allow to identify j(V) since it does not explicitly involve the vector field W. the Eulerian partial derivative of y(V) is the solution of the following linearized system. Σs (2. Γt (V) Hence.6 The adjoint state and the adjoint transverse field In order to identify the gradient j(V).  Q(V)  ∂t ϕ + ∆ϕ = 0. we introduce the following adjoint problem. Σ(V)      y (t = 0) = 0. Σs (2. n .46 differentiates as follows. Moving Shape Analysis y = −∂n y Z. Let us consider a distributed term involving the transverse field Z.      ∂n ϕ = y(V) − yd .30)  y = −∂n y Z.31) 2. n ds dt 0 Γt (V) (2.   ∂t y − ∆y = 0. . H(D)) where H(D) is an ad-hoc4 Hilbert space of functions defined in the hold-all domain D. W) ds dt (2.32)  ϕ = 0.τ D with F ∈ L2 (0.

τ . H(D)) . Γτ (V) This means that we have τ 0 Γt (V) τ f Z · n ds dt = − 0 Γt (V) λ n · W ds dt (2. the distributed term turns out to be supported by the lateral boundary. H(D)) . ∂t λ + Γ λ · V ∈ L2 (0. ∂t Λ + D Λ · V ∈ L2 (0. τ 0 D τ (2. Σ(V) (2. H(D) . where f ∈ L2 (Σ(V)). Hence. τ 0 D ∗ τ γΓt (V) f n · Z dx dt = 0 Γt (V) fZ·n Then in this case. ˘ . ˘ ` ´ ¯ def = λ ∈ C 0 [0. τ . τ 0 D τ 47 ∂t Z + D Z · V − D V · Z · Λ = τ 0 D − ∂t Λ − D Λ · V − ∗ D V · Λ − (div V) Λ · Z = 0 τ 0 D W·Λ W·Λ D for any Λ ∈ D (0. 5Λ def ad = 6 in λad ` ´ ¯ Λ ∈ C 0 [0.35) λ(t = τ ) = 0. D Hence the following adjoint identity holds. the adjoint field is also supported by Σ(V) and it can be expressed as Λ = −∗ γΓt (V) (λ n) where λ is the unique solution6 of the following adjoint tangential transverse problem −∂t λ − Γ λ · V − (div V) λ = f. −∂t Λ − D Λ · V − ∗ D V · Λ − (div V) Λ = F. τ ) × D Λ(t = τ ) = 0. τ ].34) F · Z dx dt = 0 D Λ · W dx dt In the case where F = ∗ γΓt (V) (f n).36) This identity allows to identify the gradient of j(V) with respect to V by setting f = ∂n y ∂n ϕ. τ ]. τ ) × D .An introductory example : the inverse Stefan problem following identity holds. (0. we introduce the adjoint transverse field Λ as the unique solution in an ad-hoc space5 of the following adjoint transverse dynamical problem. L2 (Γt (V)) .

.

The Sections 2 and 3 furnish the necessary mathematical framework that justify the different calculus tools described later on. ∞[. this regularity is not necessary. Rd ) ∩ def 49 . In the strong version. [150].Chapter 3 Weak evolution of sets and tube derivatives As a first reading. [21]. [145]. to be focused on Section 4 and 5. [147]. We enlarge this approach to variational problems related to the tubes evolution ( this terminology is borrowed from [8] where it is defined thanks to mutational concepts). This is why we will concentrate on the weak evolution of sets. However in most cases such as the tube deformations or other intrinsic objects. which means that the velocity field is Lipschitz.[150].g. and in [155] for the case of a non-linear viscous flow. The so-called “Speed Method” has been developed in relation with the shape optimization of systems governed by Partial Differential Equations (PDE) in e. Hence the shape differential equation has been studied in [145].. Rd )). [155] to deal with the shape differential equation and its applications. C k (Rd . In order to weaken the above regularity requirement. known as the Helly’s compactness theorem.[79] for shape functional governed by several linear classical boundary value problems (with use of the extractor estimate for the shape gradient [24]). Then for any set Ω0 ⊂ Rd . We focus on compactness results using different kinds of boundedness assumptions. 3. the transported set Ωt (V) = Tt (V)(Ω0 ) is defined at all times. These tools are usually built when the dynamic of the domains are smooth enough. This leads to existence results connected to the “parabolic version ” of the compact inclusion property from the bounded variation functions space into the integrable functions space. One of the boundedness hypotheses is based on the use of the “Density Perimeter” properties [22]. [146]. we can notice that the The weak convection of measurable sets has been introduced in [157]. it is kindly advised for someone new to this field.1 Introduction L∞ (Rd . we considered the flow mapping Tt (V) of a smooth vector field V ∈ C 0 ([0.

We shall add several results about the continuity and the compactness of the solutions. V ∈ L1 ((0.2.2 3. the following convection equation. D (3.2) . τ . .1) ξΩt = ξΩ ◦ [Tt (V)]−1 and the function ξ(t. When V is a smooth vector field.1 Weak convection of characteristic functions The convection equation def Let V be a smooth vector field in Rd and Tt (V) its associated flow mapping. div V ∈ L2 and some growth assumption on the positive part (div V) + . ∂t ξ + ξ · V = 0. e. on ∂D or any viability condition in the case of a non-smooth boundary [51]. For any measurable set Ω ⊂ Rd . can be written as ξt = ξ0 ◦ [Tt (V)]−1 and is the solution of the following convection equation.50 Moving Shape Analysis def characteristic function of the evolution domain. Many questions arise concerning the convection equation (3. n = 0.∞ (D.. ∂t ξ + ξ · V = 0. Rd ) and div V ∈ L1 ((0. whenever ξ ∈ L∞ ((0.V = div(ξV)−ξ div V does make sense as a distribution. [155]. we consider the perturbed set Ωt (V) = Tt (V)(Ω). (3.1 Notice that the term ξ. W 1. τ ) × D. the associated flow mapping is well defined for time t ∈ [0. to our mind.g. Its characteristic function satisfies the following strong transport property. ξ t = ξΩt (V) . τ ) × D ξ(t = 0) = ξΩ .(3.2) def REMARK 3. x) = ξΩt (x) satifies. 3. Nevertheless. V. τ ) × D). in a weak sense. Rd )) where D stands for a bounded hold-all domain1 containing the perturbed sets for any time t ≥ 0.2) when the vector field V is non-smooth. The incompressible situation was already introduced in [24].1) are equivalent. this is the correct 1 With the following boundary condition associated to the vector field V. τ ] and the identities (3. ξ(t = 0) = ξΩ0 This problem admits solutions when the vector field satisfies V ∈ L 2 . (0. V ∈ L1 (0. τ ) × D).

BV(D)) together with vector fields V ∈ L2 (0. L2 (D. ∂t u + u · V = f. τ . τ . Rd )) and the divergence positive part2 (div V)+ ∈ L1 (0. the notion of weak solutions associated to the convection problems (3. τ . It forms a closed convex set.3) defines a closed subset TΩ which contains the weak closure of smooth elements (ξΩ ◦ [Tt (V)]−1 .) = 0 (3. shape identification or optimization. L∞ (D)). L2 (D. defined up to a zero measure set with bottom Ω. x) dx = 0 φ(τ. a measurable non-cylindrical subset in (0. 3. 3 We say that ξ ∈ TΩ is a tube.e.Weak evolution of sets and tube derivatives 51 approach for many shape problems such as shape evolution.3) does not make sense. Rd )) solution of problem (3. fluid-structure interaction problems. τ . V) equipped with a parabolic BV like topology for which the constraint (3.2 The Galerkin approximation Let us consider the following evolution problem. the element Vξ is the unique (with minimal norm) vector field associated to ξ via the convection equation (3..2) is equivalent to the following one: τ ξ (∂t φ + V. W) ∈ TΩ . Rd ) This approach consists in handling characteristic functions ξ = ξ 2 which belongs to L1 (0. τ ] × D). For a given element (ξ. free boundary problems.14). For the time being. i. any uniqueness result has been obtained for this smoothness level. V). τ ] × D. using a Galerkin approximation and some energy estimate.4) deal with the dual evolution problem (3. we shall assume (div V)− ∈ L1 (0.2. For a given tube3 ξ. noted Vξ . L∞ (D)). the correct modeling tool for shape evolution is to introduce the product space of elements (ξ = ξ 2 .5) for such variational weak solutions.3) ∀ φ ∈ C ∞ ([0. we consider the set of fields W such that (ξ. (0.2)-(3. τ . D We shall state the following existence result. ¯ V ∈ C ∞ ([0. τ )×D. . we are able to derive an existence result of solutions with initial data given in H −1/2 (D).3). Actually. Hence. V) ∈ T Ω . τ ) × D u(t = 0) = u0 . We shall consider velocity fields such that V ∈ L1 (0. computer vision or problems where the topology of the set Ω t may be changing.2). we can define the unique minimal norm energy element V ξ in the convex set Vξ . when the field V and its divergence are simply L 1 functions. 2 To (3. 0 D φ + φ div V)dx dt + Ω φ(0. . In this case. but there exists an estimate (3. In this case. the problem (3. Actually.

1 Let V ∈ L1 (0. W −1. τ . V.R3 ) + f (s) L2 (D. τ ). x). n = 0. R3 )) with div V ∈ L1 (0.6) Using the vector notation (U m )1≤i≤m = (um )1≤i≤m . τ .L2 (D)) τ 0 ≤M u0 L2 (D) + f L1 (0. L2 (D)) ∩ C 0 ([0. τ . x) = i=1 m um (t) ei (x) i solution of the following linear ordinary differential system: [∂t um (t. x) + V(t.R3 ) s ( (div V(σ))+  )dσ ds L∞ (D) + f (σ) REMARK 3. the last system is equivi alent to the following one.τ . H − 2 (∂D)) 1 and f ∈ L1 0. x) ej (x) dx.τ . PROOF Let us consider a dense family {ek }k≥1 in L2 (D) with ek ∈ ∞ Cc (D). D um (t.52 Moving Shape Analysis PROPOSITION 3. W −1. (div V(t))+ L∞ (D) ∈ L1 (0.2 L2 (D) (3. τ ]. τ .L2 (D)) [ 1+ τ (div V(s))+ L∞ (D. there exists a constant M > 0 such that the following estimate holds true.4) admits at least one weak solution u ∈ L∞ (0. ∀ k ≥ 1. R3 )).1 (D)) Moreover. L2 (D) . 1 ≤ j ≤ m (3. u0 ∈ L2 (D). u   +  L∞ (0. τ . Then. τ . L2 (D.5) div V = (div V)+ − (div V)− . x) ] ej (x) dx = D f (t.1 (D)) together with ∂t u ∈ L1 (0. L2 (D. ∂t U m (t) + M −1 · A(t) · U m (t) = F (t) (3. equation (3. Consider the element m u (t.7) def . in L1 (0.

we have 2 D f (s) u(s) dx ≤ 2 f (s) L2 (D) um (s) L2 (D) And using Young inequality.j ]1≤i. x) um (t.τ . R3 )). τ ). x) um (s. x)|2 (div V(s. x))+ dx ds τ +2 0 D f (s.8) Then we get 1 2 ∂t |um (t)|2 ≤ 1 2 |um (t)|2 (div V(t))+ + f (t) um (t) D (3. V(t. x) dx (3. .j ]1≤i. x).j≤m = D 53 ei (x) ej (x) dx V(t. D um (t. τ .9) D D Integrating over (0.Weak evolution of sets and tube derivatives where [Mi. x) dx = − 1 2 D |um (t.6) by um (t) and by using the following identity. x). we obtain um (τ ) 2 L2 (D) ≤ um (0) 2 L2 (D) τ + 0 D |um (s. x) dx ds (3.10) Now we set ψ(s) = (div V(s.))+ We have τ τ L∞ (D) um (τ ) 2 L2 (D) ≤ um (0) 2 L2 (D) + ψ(s) um (s) 0 2 L2 (D) ds + 2 0 D f (s) u(s) Using Cauchy-Schwartz inequality.L2 (D)) L2 (D) + 0 ψ(s) + f (s) um (s) 2 L2 (D) ds . x)|2 div V(t. L2 (D. we get um (s) from which we deduce um (τ ) 2 L2 (D) L2 (D) ≤ 1 1 + um 2 2 L2 (D) ≤ um (0) 2 L2 (D) + f τ L1 (0. ei (x) ej (x) dx [Ai. We shall derive classical energy estimates. by multiplying the system (3.j≤m (t) = D Classically. this linear differential system admits a global solution when V ∈ L1 (0.

3 Tube evolution in the context of optimization problems In the previous section.12) 1 1 ∀ φ ∈ H0 (0. τ .11) Hence um remains bounded in L∞ (0. the previous existence results apply for the following evolution problem.8) can also be written in the 1 2 um (t. L2 (D)) and there exists an element u ∈ L∞ (0.3 following form D Notice that the identity (3. ∂t p(t) + div(p(t) V(t)) = g. τ . REMARK 3. this result does not guaran- . (0.L2 (D)) 1+ ψ(s) + f (s) 0 L2 (D) exp s ψ(σ) + f (σ) L2 (D) dσ ds (3.2) in the distribution sense. we have stated an existence result for weak solutions of the convection equation (3. τ ) × D p(t = 0) = p0 . x). the element u satisfies itself to the previous estimate and it can be checked that u is the solution of equation (3.τ . x) um (t.54 Moving Shape Analysis Using that u0 ∈ L2 (D).e. x)) um (t. x) dx D div(V(t. um (t. we derive: um (τ ) τ L2 (D) ≤M u0 L2 (D) + f τ L1 (0. L2 (D)) ∩ L2 (0.13) Hence when (div V)+ is turned into (div V)− . x) div V(t. τ . τ .2).14) 3. τ τ − u ∂t φ + div(φ V) dx dt = 0 D D φ(0) u0 dx + 0 D f φ dx dt φ(τ ) = 0 (3. x) dx = + (3. we get um (τ ) 2 L2 (D) ≤M u0 2 L2 (D) + f τ L1 (0.τ . i.L2 (D)) + 0 ψ(s) + f (s) L2 (D) um (s) 2 L2 (D) ds From Gronwall’s inequality. H0 (D)). L2 (D)) and a subsequence still denoted um which converges to u for the weak-* topology. D (3. Unfortunately. In the limit case.

V)∈TΩ ip inf J(ξ. In most situations. V) (3. Indeed. The space U ad is a space of smooth velocities4 .V)∈TΩ ip inf J(ξ. as introduced in [157]. k V0 (D) = def ∃c > 0. ∀ t ∈ [0. it uses the compactness properties of tubes family with bounded perimeters in Rd+1 . V) can be chosen using several approaches: • We can first consider the time-space perimeter of the lateral boundary Σ of the tube. this method leads to heavy variational analysis developments. A possible strategy is to assume different regularity levels for the velocity field in order to recover uniqueness and stability properties for initial characteristic functions [5]. In particular. we chose to adopt a different point of view inspired by the optimization problems framework. V) (3.16) The penalization term F (ξ. V0 (D) . on D def y=x |V(y) − V(x)| . This approach easily draws part of the variational properties associated to the bounded variation functions space framework. V)| V ∈ Uad t stands for the space of smooth admissible couples.Weak evolution of sets and tube derivatives 55 tee neither uniqueness nor the stability of characteristic functions. τ ] o with L ipk (V) = def |α|=k X n L ip(Dα V) for k ≥ 1 and L ip(V) = sup o ` ´ ¯ d V ∈ C k (D) . V). V) + F (ξ. This makes the mathematical setting even more complex. developed in [155]. our final goal is to apply the weak set evolution setting to the control problem arising in various fields such as free boundary problems or image processing. Then. L ipk (V(t)) ≤ c.15) where L TΩ ip = def (χΩ ◦ T−1 (V). τ ]. We would like to solve the following optimization problem L (ξ. Here. The usual situation can be described as follows: let us consider a given smooth enough functional J(ξ. |y − x| . we shall introduce different penalization terms which furnish compactness properties of the minimizing sequences inside an ad-hoc weak topology involving bounded variation constraints. the new problem writes L (ξ. We 4 Uad = def with k ≥ 0 and where n ` ´ k V ∈ C 0 [0. Consequently. • We can rather consider the time integral of the spatial perimeter of the moving domain which builds the tube. such a problem does not admit solutions and we need to add some regularization terms to ensure its solvability. V · n∂D = 0. Nevertheless.

Cc ) topology. L∞ (D)). . ξ ∈ L∞ (0. .2 conditions hold: τ 0 D A = ξ 2 = ξ.16) in order to get an existence result in some ad-hoc functional spaces endowed with specific weak topologies.17) This penalization term. Rd ) stands for the sets of Radon5 measures on D with values in Rd . will furnish the required compactness properties for the minimizing sequences of problem (3. here.16) with a particular penalization term of the following type. with the above penalization. ∀ φ ∈ L1 ((0. REMARK 3. Rd ) are relatively compact inside 0 M 1 for the weak-* σ(M 1 . given a measurable subset Ω ⊂ D. τ ) × D). 3. τ . c A useful property says that the bounded sets inside L1 (D. This drawback will be solved in the section 3. τ . Rd )) The sequence ξn → ξ in A if and only if the following (ξn − ξ) φ → 0. L∞ ). V) solving the convection equation with a given initial set Ω inside D. Rd ). we consider the following sets equipped with their respective weak topologies: DEFINITION 3. Cc (D.1 Penalization using the generalized perimeter’s time integral In this section. DEFINITION 3. we shall consider the optimization problem (3. To be more precise. we can not guarantee the generated solution set to be an open set. 0 ξ) · g → 0. Rd ). ∀ g ∈ L1 (0.Rd ) + div V(t) 2 L2 (D) dt (3. ξ ∈ L1 (0. In this case. only existence results for solutions of the convection equation can be handled and the uniqueness property is lost. τ ) × D. τ τ F (ξ.56 Moving Shape Analysis shall extend these results to the case of vector fields living in L 2 ((0.3.3. M 1 (D. Rd )) τ 0 D ( ξn − 5 This space can be defined as the dual space associated to the space (C 0 (D.1 where M 1 (D. τ . is to consider the set of pairs (ξ.3. V) = α 0 ξ(t) BV(D) dt+β 0 V(t) 2 L2 (D.4 However. involving the L1 norm of the generalized perimeter in D associated to the set characterized by ξ. The main idea.

Finally. τ ] × D). the weak formulation of problem (3.2 Parabolic version of Helly’s compactness theorem Let us define the space.1 The set TΩ is closed in A × B. ξ(t = 0) = χΩ } div V ∈ L2 (0. τ . Hence. τ . Rd )). Ω ∀ φ ∈ C ∞ ([0. τ . V) ∈ A × B. we deduce that there exists a subsequence still denoted ξn which converges in L1 for the strong topology. ∂t f ∈ L2 0.3 B = V ∈ L2 (0.19) The following compactness result holds. . Vn ) be a sequence inside TΩ converging towards (ξ. ξ · V = 0. V) turns out to be an element of TΩ .3. H −2 (D) (3.Weak evolution of sets and tube derivatives DEFINITION 3. τ .2.) = 0 is satisfied by (ξ. H −2 (D) . V) since Vn and div Vn weakly converge in L2 . the limit pair (ξ.18) PROOF Let (ξn .we get the boundedness of the following L1 (0. From the Banach-Steinhaus theorem. φ(τ. Then. BV(D) . the limit element satisfies ξ 2 = ξ. x) = 0. 6 with and fn = ξn ∈ L1 (0. 3. W = def f ∈ L1 0. τ . τ ξn (∂t φ + Vn . . L2 (D. 0 D φ − φ div Vn ) + φ(0.3). BV(D)) norm: τ 0 || ξn ||M 1 (D) dt ≤ M Using the next Proposition6 3. BV(D)) ` ´ fn = ξn = − div(ξn Vn ) + ξn div Vn ∈ L2 0. τ . V). τ . in the limit case. Let us consider the following set: TΩ = {(ξ. with ∂t ξ + THEOREM 3. L2 (D)) 57 (3.

L1 (D) → H −2 (D). τ .m L1 (0. Let us fix ε > 0. ∀ n > 0. compact injection holds. L1 (D) PROOF For d ≤ 3 the following continuous injection holds.2 page 7 in [97]. we have fn.BV(D) def + dn f L1 0.19 in [77] page 17) the following continous. for m > n. Let us consider {fn }n≥0 a uniformly bounded sequence in W.m (t0 )) H −2 (D) ≤ M. Using the result proven in the next Lemma 3. H −2 (D)) → C 0 ([0. BV(D) . if we establish the strong convergence towards zero of f n. we have the following uniform estimate fn.L1 (D)) and we get L1 (0.m in the space L1 (0. H −2 (D)) Then. τ ] Using the Lebesgue dominated convergence theorem.H −2 (D)) ≤ ε + dn fn. we deduce for f ∈ L1 0.τ .τ .m = fn − fm . BV(D) →c L1 (D). BV(D) →c L1 (D) → H −2 (D) We adapt the proof of the compactness result (Theorem 5.1 page 58 in [97]) to the non-reflexive situation.τ . using Lemma 1. we have the following embedding chain.1 (0. To .τ .τ . τ . ∀ t0 ∈ [0. τ . there exists a constant dn > 0 such that f L1 0. Then. then the proof is complete.58 Moving Shape Analysis PROPOSITION 3.BV(D)) ε 2 ≤M We can choose n such that n M ≤ fn.it will be sufficient to prove the pointwise convergence of fn.2 Uniformly bounded sets in W are sequentially relatively compact in L1 0. τ ].m L1 (0.τ .1.m 2 At this stage. τ . Furthermore using Helly’s theorem (Theorem 1. since there exists M > 0 such that fn.L1 (D) ≤n f L1 0. H −2 (D)). As L1 (D) ⊂ H −2 (D) (for d ≤ 3).m (t0 ) strongly to zero in H −2 (D).m ∈ W 1.H −2 (D) We apply this inequality with f = fn.

see also [139] page 273).m + bn.L1 (D)) ≤ + + ≤ ε 2 4 4 Then the sequence {fn }n≥0 is a Cauchy sequence in the Banach space L1 (0. H −2 (D) ≤ ε 4 LEMMA 3. Then it is a convergent sequence for the strong topology. Then.m L1 (0.m BV(D) fn. fn. we have the estimate an.m We conclude that ∀ ε > 0. φ L1 (D) ≤η φ BV(D) + cη φ H −2 (D) . there exists N ≤ 0. we deduce that a n converges strongly in H −2 (D) towards an element a ∈ H −2 (D) . we use a technique introduced by R.τ . τ . there exists a constant cη with ∀ φ ∈ BV(D).m (t) dt. L1 (D)).m (t) t0 +s t0 H −2 (D) dt ≤ 1/2 2 H −2 (D) ≤ dn s1/2 ε ≤ 4 Finally.m (t) dt Let us choose s such that t0 +s dn bn.m with t0 +s an.BV(D)) ≤ M s Again using Helly’s compactness result.1 ∀ η > 0.m = 1/s t0 fn.Weak evolution of sets and tube derivatives 59 this end. bn.1 page 59.m H −2 (D) ≤ t0 fn. ∀ m > n ≥ N. dn an.τ . ε ε ε fn.m s L1 (0. such that for m > n ≥ N .m (t) ≤ 1 fn. ∃ N ≥ 0.m = − 1 s t0 +s t0 (s − t) fn. This limit is actually zero t +s since by the same arguments we can show that 1/s t00 fn (t) dt converges in H −2 (D) for the strong topology.Temam as reported in [97] (footnote of the proof of Theorem 5. It is easy to show that the following expression holds true. Then it is a Cauchy sequence and we deduce that a = 0.m (t0 ) = an.

the set ˜ is not necessarily an open subset in ]0. REMARK 3. τ ). (3. open and clean open) in D. L2 (D)) This means that Ωt and Ωt are the same up to zero-measure sets.20) is an open set in D satisfying for a. BV(D) + cn φn H −2 (D) We introduce ψn = φn / φn ψn But also ψn L1 (D) and we derive: H −2 (D) L1 (D) ≥ η + c n ψn ψn H −2 (D) ≥η ≤ c ψn BV(D) = c . t > 0.e.4 for a. there exists φn ∈ BV(D) and cn → ∞ such that φn L1 (D) ≥ η φn BV(D) . the tube {t} × Tt (V)(Ω0 ) is open (resp.3 Generation of clean open tubes Clean open tubes DEFINITION 3. 3.e.6 Two tubes Q. we have → 0 But as ψn BV(D) = 1. If there exists a clean open tube Q such that χQ = χ Q ˜ . ˜ (t. t ∈ (0.3. which turns to be strongly convergent to zero. LEMMA 3. Then. in L2 (0.5 If the set Ωt is an open set in D for a. Ωt = int(cl(Ωt )). Then.e. τ [×D. the set ˜ A clean open tube is a set Q in ]0. t ∈ (0. for some constant c. Nethertheless when the field Q V is smooth. x) ∈ Q (3. Q are said to be equivalent if χQ = χQ . τ ) the following cleanness property: ˜ ˜ ˜ meas(∂ Ωt ) = 0. This is in contradiction with the fact that ψn L1 (D)) ≥ η. τ .60 Moving Shape Analysis PROOF Let us assume that the above estimate does not hold. there exists a subsequence strongly convergent in L1 (D) ⊂ H −2 (D). ∀ η > 0. τ [×D.2 ˜ Let Q be a measurable set in ]0. open and clean 0<t<τ open) when the initial set Ω0 is open (resp. τ [×D such that ˜ Ωt = x ∈ D.21) ˜ REMARK 3.

7 set. This allow us to build a penalization term that will generate open set solutions to the optimization problem (3. eD (QV ) = min V V∈VQ def L2 (0.Weak evolution of sets and tube derivatives Then. we use the projection theorem in the Hilbert space L2 (0. [21] for a closed set A inside D. DEFINITION 3. DEFINITION 3.L2 (D. τ .3 If the set VQ is non-empty. The ε-dilatation of the set A is given by the following Aε = def x∈A B(x. L2 (D. ξ(t = 0) = χΩ0 (3. 61 ˜ ˜ PROOF Assume that there exists two equivalent clean tubes Q and Q . τ ]. H −1/2 (D)). This notion has been introduced in order to have a control on the boundary measure of the limit set associated to a convergent family of open sets. Hence. they must be equals.22) we say that V builds the tube and we note Q = QV .5 Let Ω0 be a clean open set in D and Q a clean open def tube in ]0.e.24) admits a unique solution VQ ∈ B ∩ L2 (0.25) . ˜ t = Ω up to a measurable set Et with ˜ Then a. τ [×D with ξ = χQ ∈ C 0 ([0. this clean tube is unique. If there exists a divergence free field V ∈ B such that: ∂t ξ + ξ · V = 0. we introduce the notion of density perimeter [22]. PROOF It is easy to prove that if the set VQ is non-empty. ε) (3.Rd )) (3.6 tube Q. Rd )).23) PROPOSITION 3.τ . then the following minimization problem. QV = Q} (3. DEFINITION 3.16). L2 (D.21). then it is closed and convex in B. Rd )). We define the set of fields that build the clean open VQ = {V ∈ B. we have Ω t meas(Et ) = 0. τ ). As these two open sets satisfy the cleanness property (3. t ∈ (0. τ . Properties of the density perimeter In this section.

γ) meas(Aε ) 2ε (3. Then the convergence holds for the char-topology8 .3. Ω0 ) = 7 The µ∈Mγ (V.∞ For any smooth free divergence vector field.3 ([22]) Let us consider a family {Ωn }n≥0 of open sets in D. 1.62 Moving Shape Analysis DEFINITION 3.τ ) (3.27) Hausdorff complementary topology is given by the metric. F1 )) for any closed sets F1 . W0 (D.3. ρ(F2 . Actually we shall combine the compactness properties of the generalized perimeter with the openness control properties of the density perimeter. Θγ (V. Property 3. Y ) = sup inf 8 The x∈X y∈Y x−y char-topology is defined on the family of measurable subsets of Rd by the L2 -metric. F2 in D and where ρ(X. Ωc their associated complementary sets in D. Ωc ) 1 2 where Ω1 . we consider the following minimization problem. Ω2 are open subsets of D and Ωc .26) Property 3. Penalization using the parabolic generalized density perimeter In this section.Ω0 ) min µ ˙ 2 L2 (0. τ ]. V ∈ C 0 ([0. Let us suppose that this family converges to some subset Ω ⊂ D for the H c topology and that there exists M > 0 such that Pγ (∂Ωn ) ≤ M . Pγ (A) = sup ε∈(0. Ω2 ) = dH d (Ωc . dH c (Ω1 . F2 ) = max(ρ(F1 . we shall describe a particular type of clean tubes construction for which the set VQ is non-empty.2 ([22]) If the boundary set ∂Ω is such that Pγ (∂Ω) < ∞ then meas(∂Ω) = 0 and Ω \ ∂Ω is open in D Property 3. A2 ) = |χA1 − χA2 | Rd .3. 1 2 Furthermore dH d stands for the Hausdorff metric defined by dH d (F1 . Rd )).8 Let γ > 0.1 ([22]) The mapping Ω −→ Pγ (∂Ω) is lower semi-continuous for the Hausdorff7 complementary topology H c . the density perimeter of the set A is given by the following expression. Z dchar (A1 . F2 ).

τ ].28) When this set is empty. L2 (D. τ ) . τ Pγ (∂Ωt (V)) ≤ µ(t) ≤ 2 Pγ (∂Ω0 ) + Then. τ ).4 1. Ω0 ) = +∞.  µ(t = 0) ≤ (1 + γ) Pγ (∂Ω0 )  63 (3. we have Θ(V.e. Ω0 ) furnishes the correct estimation. we get Pγ (∂Ωt (Vn )) ≤ µ(t) ≤ 2 Pγ (∂Ω0 ) + √ τ µ(t) dt ˙ 0 τ 0 1/2 |µ(t)|2 dt ˙ Minimizing over µ ∈ Mγ (V.∞ For any V ∈ C 0 ([0. Ω0 ). we choose to set Θγ (V. Θ(Vn . τ ].. i. Ω0 ) < p ˙ as the minimizer will escape to possible variations of the function p. ∃M > 0. REMARK 3. τ ). PROPOSITION 3.5 1. the following inequality holds.30) . t ∈ (0. Ω0 ) =    µ ∈ H 1 (0. √ Pγ (∂Ωt (V)) ≤ 2 Pγ (∂Ω0 ) + τ Θ(V. W0 (D.29) PROOF By definition. τ . Ω0 ) ≤ M (3.Weak evolution of sets and tube derivatives where Mγ (V.∞ Let Vn ∈ C 0 ([0. Vn −→ V in L2 (0. p ∈ Mγ (V.31) (3.e. W0 (D.  Pγ (∂Ωt (V)) ≤ µ(t) a. using Cauchy-Schwartz inequality. Rd )) and the uniform boundedness. Ω0 )1/2 (3. PROPOSITION 3. τ ) −→ R t → Pγ (∂Ωt (V)) 2 L2 (0.τ ) is an element of H 1 (0. Rd )) with free divergence. Rd )) with the following convergence property.7 When the mapping p : (0.

τ ). Ω0 ) n→∞ (3. τ ) Using the l-s-c property of the density perimeter (Theorem 1.30) and bound estimate (3. for a. and as a consequence a. there exists a clean open tube ˜ Q= 0<t<τ ˜ {t} × Ωt such that. t ∈ (0. Ω0 ) ≤ 0 |µ(t)|2 dt ≤ lim inf Θ(Vn . Furthermore. in H c -topology . τ ) associated with Θ(Vn . Rd )) with convergence property (3.τ ). Ω0 ) ˙ n→∞ PROPOSITION 3.29). Ω0 ). χΩt (Vn ) → χΩt . τ ) for the strong topology. Using the boundedness assumption and the estimation (3. We assume that Ω0 is an open subset in D verifying Ω0 = int(cl(Ω0 )) Then. τ ) in L2 (D). By definition. still denoted µ n .32) PROOF Let µn be the unique minimizer in H 1 (0. t ∈ (0.31). Ω0 ) ≤ lim inf Θ(Vn .e.6 1.e. we have Pγ (∂Ωt (Vn )) ≤ µn a. we have Pγ (∂Ωt (V)) ≤ lim inf Pγ (∂Ωt (Vn )) ≤ µ(t) n→∞ a. τ ) But the square of the norm being also w-l-s-c. W0 (D.9 page 7 in [77]). t ∈ (0. we get √ Pγ (∂Ωt (Vn )) ≤ 2Pγ (∂Ω0 ) + τ M 1/2 C We deduce that there exists a subsequence. τ ]. this convergence holds in L2 (0.64 Then Moving Shape Analysis Θ(V.e. ˜ ˜ Ωt (Vn ) → Ωt .e. which weakly converges to an element µ ∈ H 1 (0. we have τ 0 |µ(t)|2 dt ≤ lim inf ˙ n→∞ τ τ 0 |µ˙n (t)|2 dt This leads to Θ(V. t ∈ (0.∞ Let Vn ∈ C 0 ([0.

we set Ωt = cl(ωt ) − ∂ωt . W0 (D. τ ). τ ). PROOF We have χQVn → ξQV in L2 ((0.34) ˜ Then there exists a clean open tube QV built by V with the following convergence: for a. Then. χωt = χΩt (V) for a. R )) to V with the boundedness condition: pγ (Vn . From the boundedness of Pγ (Ωt (Vn )) and using property (3. ξVn (t) → ξV (t). THEOREM 3.2 1. we also have Ωt (Vn ) → Ωt (V) in the char-topology for a.2). Then. τ ). we define the parabolic generalized density perimeter of the moving domain Ωt (V) over the time period (0. we derive that ωt (V) is an open set in D and meas(∂ωt (V)) = 0.e. τ ) × D). for the H c -topology. Ω0 ) (3. ξ(t = 0) = χΩ0 ˜ This means that Q is the unique clean tube equivalent to the limit tube QV built by V. there exists a subsequence (depending on t) which converges. to an open set ωt . τ ].Weak evolution of sets and tube derivatives 65 ˜ Moreover Ωt is the single open set such that cleanness condition (3. τ ). ˜ Hence. in H c -topology 9 The generalized perimeter of a measurable set A in D is given by Z PD (A) = χA M 1 (D. ˜ Ωt (Vn ) → Ωt . Meanwhile.9 For smooth vector fields V. we have χΩt (Vn ) → χΩt (V) in L2 (D). for almost every t ∈ (0. τ ). At each time t ∈ (0.e.33) where PD (Ωt (V)) stands for the generalized perimeter9 of Ωt (V).21) is satisfied and whose characteristic function χΩt solves the convection problem: ˜ ∂t ξ + ξ · V(t) = 0. t ∈ (0. in L2 (D). t ∈ (0.∞ Let Vn ∈ C 0 ([0. Rd )) which weakly 2 2 d L (0. τ ).3. Rd ).Rd ) = sup − div ϕ dx ∞ A ϕ ∈ Cc (D. τ . DEFINITION 3. Ω0 ) ≤ M converges in (3.e. L (D. Ω0 ) = 0 PD (Ωt (V)) dt + γ Θγ (V. ϕ ∞≤1 . τ pγ (V. t ∈ (0.

k V0 (D) = def ∃c > 0. C k (D) . V0 (D) . it is possible to perform a sensitivity analysis of the following integral. |y − x| ` ` ´ ´ ` ` ´ ´ ¯ d ∩ C 0 [0. V · n∂D = 0. L ipk (V(t)) ≤ c. Introducing the transverse field which builds the flow mapping the reference moving domain into the perturbed one. τ χ(x. ∀ t ∈ [0. we introduce alternative tools to handle the sensivity analysis of tube functionals.2). allowed us to differentiate the following Eulerian integral τ F (V)(x. t) dx dt 0 Ωt (V) with respect to V. 0 10 D with k ≥ 0 and where Uad = def n ` ´ k V ∈ C 0 [0. Ω0 )+β 0 V(t) 2 L2 (D.35) we can obtain an existence result for the optimization problem (3. t) F (χ)(x. T(V) . W k+1. if we define the following penalization term τ F (ξ. on D def y=x o |V(y) − V(x)| .4 Tube derivative concepts In this section.∞ (D) d T(V) ∈ C 1 [0.16) with the property that the associated tube solution is clean open if Ω 0 is open. 3.66 Furthermore. Moving Shape Analysis pγ (V) ≤ lim inf pγ (Vn ) n→∞ Finally. τ ]. The Eulerian setting has been presented in Chapter 2 and it is based on the introduction of a velocity field V ∈ U ad which is smooth enough 10 for its Lagrangian flow to be defined in a classical manner 11 . τ ]. A second choice consists in parametrizing the moving set thanks to the couple V. χ(V) where χ(V) is the characteristic function of the moving set Ω t (V) and is the solution of the transport equation (3. τ ]. Hence we have a parametrization of the moving set based on the couple V.Rd ) + div V(t) 2 L2 (D) dt (3. In this setting. τ ] with L ipk (V) = 11 def |α|=k X L ip(Dα V) for k ≥ 1 and L ip(V) = sup n o ` ´ ¯ d V ∈ C k (D) . V) = α pγ (V. t) dx dt.

i. j(V) = 0 Ωt (V) FΩt (V) + 0 Γt (V) fΓt (V) + Ωτ gΩτ (V) (3. V) ∈ TΩ and a vector field W such that for all s.1 Characteristics versus Eulerian flow setting Let (ξ. As recalled in section 3. A motivation for this setting is the introduction of perimeter constraints inside optimization problems for tube evolution. Hence it is possible to work in the classical ˜ def Eulerian setting with extended velocity field V = ∗ (0. let us consider heuristically the term (if it exists) ˙ ξ = ∂s ξ s |s=0 as a measure over (0. V). This measure should solve the evolution problem (3. τ dΓ dt 0 Ω(t) or the time-space perimeter. 3.Weak evolution of sets and tube derivatives 67 A third choice consists in working directly on the tube inside the time-space def Euclidian space Rd+1 = Rt × Rx . τ )×D. First.3..36) We consider a non-cylindrical functional depending on the vector field V.4) with a measure term as right hand side: ˙ ∂t ξ + τ ˙ ˙ ξ · V = − ξ · W. V) = 0 D ξ Fξ . This allows dealing with tube integrals of the following form. Furthermore it has been proved in section 3. V + sW) ∈ TΩ . |s| ≤ s1 there exists ξ s with (ξ s . We set ξ = χΩt (V) .3 that the boundedness of the first perimeter furnishes the same compactness result.e. we can either define the time integral of the space perimeter. fΓt (V) = 0 and gΩτ (V) = 0. ˜ Q(V) ˜ F (V)d Q + ˜ Σ(V) ˜ f (V)d Σ where Σ stands for the boundary of the tube Q. and we can write τ j(ξ.37) We first consider the case where the functional only involves a distributed def term defined on Q(V). dΣ Σ The boundedness of the latter characterizes the BV compactness of the tube Q inside Rd+1 . ξ(t = 0) = 0 τ (3. These perimeter constraints are associated in a variational way to the notion of surface tension which is highly required while studying free boundary problems as the one described in [153] concerning the hydrodynamical model of water waves.4.

W) = 0 D ˙ −ξ [∂t λ + div(λ V)] Using integration by parts and equation (3.41) In [154]. We can define the derivative of the functional j(V) with respect to V in the direction W as follows. def j (V. nt dΓt dt (3. τ 0 D ˙ ξF = 0 τ F Z.39) where F = ∂s F (ξ s )|s=0 . it leads to j (V. nt 0 Γt (V) = (3. We introduce the following non-cylindrical adjoint state λ: −∂t λ − div(λ V) = Fξ . W) = ∂s j(ξ s . W) = 0 D ˙ λ ∂t ξ + ˙ ξ·V =− τ λ 0 D ξ·W Integrating again by parts and using Stokes formula.36). V + s W)|s=0 (3. ∂Ωt (V) = Σ(V) def F Z. We introduce γΣ ∈ L(C 0 (]0. W) = 0 def D τ ˙ ξ Fξ + ξ F (3. nt (1 + v 2 )−1/2 dΣ (3. the trace operator on the lateral .43) where v = V. we assume that F = 0. nt . the element ξ satisfies the following identity.68 Moving Shape Analysis where FΩt (V) = Fξ |Ωt (V) and Fξ is defined in (0. τ [×D. λ(τ ) = 0 For the sake of simplicity. C 0 (Σ)). λ W. W) = 0 τ D ξ div(λ W). τ [×D). we get τ j (V. [59]. nt dΓt dt. we have τ j (V.42) ˙ We deduce that.40) j (V. τ ) × D. and we get τ (3.38) Here. as a measure on ]0. we have introduced the transverse field Z such that τ τ ∂s 0 Ωt (V+sW) F s=0 = 0 ∂Ωt (V) F Z.

Weak evolution of sets and tube derivatives

69

∗ boundary Σ of the tube Q(V) and γΣ ∈ L(M(Σ), M(]0, τ [×D)) its adjoint operator. Hence, we deduce

∗ ˙ ξ = γΣ ·

Z, nt
(1 + v 2 )

(3.44)

Comparing equation (3.41) and (3.43), we derive the following result,
τ τ

F Z, nt =
0 Γt (V) 0 Γt (V)

λ W, nt

(3.45)

In the case where we keep the final time term gΩτ , we set, gΩτ = g|Ωτ where g is defined on D. Introducing the adjoint problem, ¯ ¯ ¯ −∂t λ − div(λ V) = Fξ , λ(τ ) = g We have the following duality identity,
τ

(3.46)

g Z(τ ), nτ dΓτ (V) =
Γτ (V) 0 Γt (V)

¯ λ W(t), nt dΓt dt

(3.47)

which leads to the functional derivative,
τ

j (V, W) =
0 Γt (V)

¯ λ W, nt

3.4.2

Tangential calculus for tubes

In the last section, we have proven that the directional derivative of distributed non-cylindrical functionals involves the normal trace zt = Z, nt , on Σ(V) Let us recall the following characterization of z t , LEMMA 3.3 [58],[59]
def

(3.48)

zt ◦ Tt (V) =
t 0

t σ

W(σ), nσ ◦ Tσ (V) exp

D V(σ) · nr , nr ◦ Tr (V) dr

dσ, on Γ0

70 REMARK 3.8

Moving Shape Analysis When the vector field V is chosen in the canonical form V = V ◦ pt

the following identity holds, D V · nt = 0, and we get
t

on Γt (V)

zt ◦ Tt (V) =

0

W(σ), nσ ◦ Tσ (V) dσ, on Γ0

(3.49)

DEFINITION 3.10

Let us consider the noncylindrical gradient operator,
Σ λ, ντ

= ∂t λ +

λ·V

(3.50)

where the vector ντ = (1, V),

def

on Σ.

REMARK 3.9 The vector ντ is tangent to the lateral surface Σ since the outgoing normal field is given by 1 def ν = √ (−v, nt ) 1 + v2
with v = V, nt .
The duality identity (3.45) writes zt F √ dΣ = 1 + v2 Σ
Using the operator [
Σ Σλ Σ

(3.51)

def

W, nt λ√ dΣ 1 + v2 Σ

(3.52)

and the equation (3.40) satisfied by λ , we get
W, nt λ√ dΣ 1 + v2 Σ

zt · ντ + λ div V] √ dΣ = − 1 + v2
λ zt = 0, on ∂Σ

as Performing an integration by parts on the manifold Σ, we get the following identity, zt zt −λ divΣ √ ντ + λ div V √ 1 + v2 1 + v2
dΣ W, nt λ√ dΣ 1 + v2 Σ
(3.53)

Σ

=−

Weak evolution of sets and tube derivatives So we can prove the following proposition, PROPOSITION 3.7 − div V(t) zt +

71

zt 1 + v 2 divΣ √ ντ = W(t), nt 1 + v2

(3.54)

PROPOSITION 3.8 The normal transverse field zt satisfies the following identity, ∂t e + div(e V) − (div V) e + 1 [v (∂nt e + ∂t (e V), nt − v∂t e )] 1 + v2 1 1 − D(e V) · nt , nt = √ W, nt (3.55) 1 + v2 1 + v2

zt where e = √ . 1 + v2
PROOF We shall set E = ∗ (E1 , E2 ) = e ∗ (1, V) Hence, using classical differential operators defined on (0, τ ) × D ˜ ˜ divΣ E = divQ E − DQ E · ν, ν
def

(3.56)

˜ where E stands for an arbitrary extension of E = ∗ (E1 , E2 ) in (0, τ ) × D and divQ E = ∂t E1 + div E2 DQ E = Recalling that ν =
∗ √ 1 (−v, nt ), 1+v 2

∂t E 1 D E 1 ∂t E 2 D E 2
we get

divΣ E = divQ ∗ (E1 , E2 ) −

1 DQ ∗ (E1 , E2 ) · ∗ (−v, nt ), ∗ (−v, nt ) 1 + v2 1 ∗ = ∂t E1 + div E2 − (−v∂t E1 + ∂nt E1 , −v∂t E2 1 + v2 + D E2 · nt ) , ∗ (−v, nt ) 1 = ∂t E1 + div E2 − v 2 ∂t E1 − v∂nt E1 − v ∂t E2 · nt 1 + v2 + D E 2 · nt , nt ]

72

Moving Shape Analysis

3.4.3

Classical shape analysis for tubes

Let us consider a tube Q with lateral boundary Σ, and an horizontal perturbation leading to the perturbed tube Qs given in the following form: ˜ def • We consider an horizontal field Z = ∗ (0, Z) defined in Rd+1 = Rt × Rd x and its associated flow mapping, ˜ Ts (Z) : Rt × Rd −→ Rt × Rd x x (t, x) → (t, Ts (Z)(x) ) ˜ ˜ We designate by ZΣ the tangential trace of the vector field Z on the lateral surface Σ. From the expression of the normal field ν, we easily derive ˜ ZΣ = 1 ∗ (v zt , Z − zt nt ) 1 + v2
(3.57)

Here ν is to be understood as any extension of the normal field to a neighbourhood of the lateral boundary of the tube. For example, we can choose nt = bΩt where bΩt stands for the oriented distance to the section of the tube at time t and V can be understood as V ◦ pt where pt is the Rd projection on Γt = ∂Ωt . Mean curvature of the lateral boundary of tubes

PROPOSITION 3.9 Assume that the field V verifies for each t ∈ (0, τ ): V(t) = V(t) ◦ pt Then, on the boundary Γt (V), we can define the time-space mean curvature of the lateral boundary Σ, HΣ = divQ ν
def

1 = − √ [ ∂t V, nt − [ ( 1 + v 2 )3

Γt

Ht V, nt ] , V|Γt ] + √ 1 + v2
(3.58)

PROOF

We choose v = V, ∂t

bΩt (V) . It is obvious that
1 = √ ∂t v ( 1 + v 2 )3
b + ∂t b, V

v √ 1 + v2

But ∂t v = ∂t V,

Weak evolution of sets and tube derivatives From the definition of the oriented distance, we got ∂t bΩt (V) = − V, nt ◦ pt and ∂t bΩt (V) = − [ Then we get ∂t v = ∂t V, nt − [ On the other hand, div n √ t 1 + v2
1 1 √ , nt + √ div nt 1 + v2 1 + v2 1 Ht =− √ ε(V) · nt , nt + √ ( 1 + v 2 )3 1 + v2 =−
Γt Γt

73

(3.59)

V, nt ] ◦ pt V, nt ] , V|Γt

1 where ε(V) = 2 (D V + ∗ D V) is the deformation tensor. We consider the situation in which the field V verifies the following property:

V = V ◦ pt with pt = I −bΩt (V) from which we deduce ∂t pt = −∂t bΩt (V) bΩt (V) − bΩt (V) (∂t bΩt (V) ) bΩt (V)

(3.60)

The restriction to the boundary Γt leads to the distance bΩt (V) = 0, so the expression simplifies as follows, ∂t pt |Γt = V, nt nt and D V · nt = 0, on Γt (V)

DEFINITION 3.11 Consider f (Σ) a function defined on the lateral boundary Σ. We call the Eulerian material derivative of f (Σ) in the direction ˜ Z the following quantity, ˜ ˜ def f˙Σ (Z) = ∂s f (Σs ) ◦ Ts (Z)
s=0

(3.61)

and its associated partial Eulerian derivative is given by the following expression, ˜ def ˜ ˜ fΣ (Z) = f˙Σ (Z) − Σ f (σ), ZΣ (3.62)

74

Moving Shape Analysis Using the expression of the tangential operator Σ , we get v zt zt ˜ ˜ fΣ (Z) = f˙Σ (Z) − ∂ f − Γ f, Z − nt 2 t 1+v 1 + v2

REMARK 3.10 When f (Σ) is the restriction to the lateral boundary Σ of a function F defined over Rd+1 , we get ˜ ˜ ˜ f (Z) = F (Z)|Σ + ∂ν F Z, ν
Σ Q

˜ = ∂ν F Z, ν ˜ as FQ (Z) = 0 since F does not depend on Σ.
Shape derivative on the lateral boundary of tubes ¯ Let us now consider a function F ∈ C 1 ([0, τ ] × D). First, we assume that F is zero in the neighbourhood of t = τ so that the following derivative of the lateral boundary integral could be considered as the derivative of integral on the total boundary of the tube (as it will generate no term on the top t = τ of the tube). We set Σs = {(t, Tt (V + sW)(x)), x ∈ ∂Ω0 }

PROPOSITION 3.10 Assume the vector field V in the canonical form V(t) = V(t) ◦ pt in a neighbourhood of the lateral boundary Σ and set v = V(t), nt on Γt . The following identity holds, ∂s
Σs

F dΣ s
s=0

τ

=
0 Γt

1 √ [−v ∂t F + ∂nt F ] 1 + v2

1 +[− √ [ ∂t V, nt − Γt v, V|Γt ] ( 1 + v 2 )3 Ht +√ F Z, nt dΓt dt 1 + v2
PROOF The classical shape calculus can be applied in R t × Rd , ∂s
Σs

(3.63)

F dΣ s
s=0

=
Σ

˜ (∂ν F + HΣ F ) Z, ν dΣ

We have
and

zt ˜ Z, ν = √ 1 + v2
∂ν F = D Q F · ν 1 = √ [−v ∂t F + ∂nt F ] 1 + v2

. This implies that the right-hand side of the backward equation is zero. τ ∂s Σs dΣ s s=0 = 0 Γt 1 − √ [ ∂t V. Then.63).45). V|Γt ] + √ 1 + v2 with λ(t = τ ) = 0.9).e. by setting F = 1 in equation (3. we investigate a problem related to optimal trajectories. ∂t V. i. nt − Γt v. nt dΓt dt We introduce the adjoint field λ such that 1 −∂t λ − div(λ V) = − √ [ ∂t V. nt √ dΣ 1 + v2 1 + v2 √ We conclude using dΣ = 1 + v 2 dΓt dt. V|Γt = (1 + v 2 ) Ht 3.11 We can easily obtain the optimality condition for a minimal surface tube. nt − Γt v. V|Γt ] ( 1 + v 2 )3 Ht 1 +√ F Z. Let a be a point . nt dΣ = 0. Σ ∀W from which we deduce λ = 0 on Σ.5 A first example : optimal trajectory problem In this section. nt − ( 1 + v 2 )3 Ht +√ 1 + v2 Γt v. we deduce that the optimality condition writes λ W. using the duality identity (3. nt − ( 1 + v 2 )3 Γt Ht v. V|Γt ] Z. Let D be a compact domain in Rd with a smooth boundary ∂D. REMARK 3.Weak evolution of sets and tube derivatives Using Proposition (3. we get ∂s Σs 75 F dΣ s s=0 = Σ 1 √ [−v ∂t F + ∂nt F ] 1 + v2 1 − √ [ ∂t V.

Rd )). there exists an injective mapping γ ∈ C 1 ([0. C (D. C 1 (D. with extreme points a and x (3. the minimization problem can be written in the following terms.x gC dC ¯ Let V be a vector field in L1 (0. C 1 (D. V Ca. W) = ∂s j(V + sW)|s=0 . we consider the simple case in which the density function g C is the ¯ restriction of a given function G ∈ C 1 (D).x ) ∈ Ca. D) satisfying the following properties. Rd )) with V.. def ¯ V ∈ L1 (0.x 0 V The claim is that from any element Ca.x . J(Ca.64) For any curve Ca. Rd )). 1. ∀ (V.x . V.76 Moving Shape Analysis inside D and for any point x ∈ ∂D. we consider the family of C 1 curves which join these two points in D. C 1 (D.e. then we define the perturbed curve V 0 Ca. we need to compute the following cost function derivative. In that case.x furnishes all 1 1 ¯ the curves when V described the linear space L ((0.x in Ca.x with ¯ E = V ∈ L1 (0.x = C 1 curves Ca.1 Optimality conditions : case of planar parametric curves In order to solve the above problem. γ(1) = x Let gC ∈ L1 (Ca.x ∈ Ca. Rd )). γ(0) = a.x dCa.x V V gCa. i.x the element Ca. 1). V∈E min j(V) (3. n = 0 on the boundary ∂D. 1]). 1]. .x = T1 (V)(Ca.x the integral quantity.x ).65) where j(V) = V Ca. W) ∈ E × E First.x = γ([0. j (V. 1.x ) = Ca.5. V.x ⊂ D. Our objective is to minimize with respect to the curves Ca. n = 0 on ∂D Hence. We set Ca. 1. that is gC = G|C .x = Ca. Ca. n = 0 on ∂D 3.

67) Obviously j (V. 59]. W) = 0 [ ( G) ◦ T1 (V)).x . 1]. S1 (γ(σ)) +G(T1 (V)(γ(σ)) D S1 (γ(σ)) · τ (γ(σ)). . τ1 (T1 (γ(σ))) ] γ (σ) dσ Eliminating the parametrization γ. W) = ∂s Tt (V + sW)|s=0 We introduce We have ∂s [D T1 (V + sW) ◦ γ · γ ] (σ) 2 |s=0 = 2 D S1 (V. we get j (V. 1 (3. τ1 (T1 (V)) ] dC (3. For 0 ≤ t ≤ 1. τ (σ) = D S1 (V. Rd )). St is the solution of the following dynamical system [154. W) = 0 Ca. W) · τ. Let γ be a parametrization of the reference curve C a. D T1 (V) ◦ γ · γ from which we deduce ∂s [D T1 (V + sW) ◦ γ · γ ] (σ) |s=0 = D S1 (V. W) ◦ γ · γ (σ). 12  2  ∂t D St − D V ◦ Tt (V) · D Tt (V) · St − D V ◦ Tt (V) · D St   = D(W ◦ T (V)).Weak evolution of sets and tube derivatives 77 the parameter s of perturbations only occurs in the measure element on the 0 curves. We have 1 j(V + sW) = 0 G ◦ T1 (V + sW) ◦ γ(σ) [D T1 (V + sW) ◦ γ · γ ] (σ) dσ St (V. The Jacobian matrix D S1 is itself a solution of the following dynamical system.x while τ is the 0 unitary tangential vector to the reference curve C a. τ1 ◦ T1 (V) ◦ γ(σ) γ V where τ1 is the unitary tangential vector on the curve Ca. W) ◦ γ · γ . S1 + G(T1 (V)) D S1 · τ. W) depends linearly on W through the term S1 . t (3.x . St (t = 0) = 0 This leads to the following derivative expression. C 2 (D. ∂t St − D V(t) ◦ Tt (V) · St = W(t) ◦ Tt (V).x [ G(T1 (V)).68)    D S(t = 0) = 0 12 We shall assume in this case V ∈ C 0 ([0.66) j (V.

69)  θ(1. S1 + A(1) · · D S1 ) ◦ γ(σ) γ (σ) dσ ( θ(1). γ(σ)) = G(T1 (V)(γ(σ))) and the matrix A solves the following backward dynamical system.x we would have γ (σ) = 1.70) In this setting. σ1 ]. the derivative of the functional takes the following form. Let θ solve the following backward dynamical system. 3. developed for planar curves.68). leads to a more explicit expression in the general setting of functional derivatives.  ∗  −∂t θ − D V(t) ◦ Tt (V) · θ = 0.  2 ∗ ∗  −∂t A − D V ◦ Tt (V) · D Tt (V) · A − D V ◦ Tt (V) · A = 0.x . as we shall see in the next section.x 1 = 0 dt 0 (θ(t) W(t) ◦ Tt (V) (3. (3. Its solution (θ. 0 The vector θ and the matrix A are defined on [0.2 Optimality conditions : case of the general Eulerian setting We consider now a calculus method which shall never refer to the reference V V +sW 0 curve Ca. a.78 Moving Shape Analysis The couple (St . Let us consider the following transverse mapping.66). A) allows to clarify the linear contribution of the field W in the expression of j (V.71) + D θ(t) · · D(W ◦ Tt (V))) ◦ γ(σ) dσ] Then we get a backward calculus for the optimal field V along the reference o trajectory Ca. 1 j (V. 1] × C a. (3. . W). S1 + A(1) · · D S1 ) dC 1 = o Ca. x = γ(σ)) where the parameter σ ∈ [σ0 . DSt ) is the solution of the dynamical system (3. W) = 0 ( θ(1). In this case.  A(1.x and Ca.x parametrized by γ.5. We shall now introduce the backward adjoint dynamical system. Actually the previous Eulerian approach.72) 13 Of course σ could be chosen as the arc length of the reference curve C 0 .x with variables 13 (t. γ(σ)) = G(T1 (V))(γ(σ)) τ (γ(σ)) · ∗ τ1 (T1 (V)(γ(σ))) (3.x but only to the moving curves Ca. Tst = Tt (V + sW) ◦ Tt (V)−1 (3.

we obtain j (V.75) gC.) V where H stands for the curvature of Ca. Then we can write def 79 j(V + sW) = Cs gC s dC s Here s is understood as the shape perturbation parameter in the classical setting.73) Hence. where g C is either the trace of a distributed function defined over (0. Z(1) = Z(1. Distributed density In the very simple case where the function gC is the restriction to the curve C of a smooth function: gC = G|C .x onto C s = Ca. We recall here the very definition.x . τ ) × D or depends on the curvature V of the curve Ca.x (3. we assume that the dimension is d = 2. V (3.Z + H Z(1). . The term gC.Z = ∂n G Z(1). so that the curves can be considered as a part of the boundary of a moving set. the moving curve C s is obtained from the reference curve through the flow mapping associated to the vector field Z t .x in the direction of the vector field Z(1). We introduced the terminology transverse flow.x ) Using classical differentiation results for boundary integrals 14 . x) = Z t (0.Z = ∂s gC s ◦ Ts1 s=0 − τ gCa. we get gC.74) with Z(t. n V on Ca. This flow evolves with respect to the parameter s for a fixed time t = 1.x .x . V C s = Ts (Z t )(Ca.Weak evolution of sets and tube derivatives V V +sW This map sends the curve Ca.x .x gC.Z is the so-called boundary shape derivative of the function g(C) V on Ca.x 14 Here. W) = V Ca. n V dCa. Hence. Z(1) def Here τ g = G − ∂n G n stands for the tangential derivative of gC along the curve and is independent on the choice of the extension G of g C outside of the curve. We shall now study two specific situations. . we consider the associated speed vector Zt (s. x) = ∂ t T ◦ Tst ∂s s −1 (3. x).

79) − Λ(0). Z(1) M(D)×C 0 (D) dt (3. V] 0 1 M(D)×C 0 (D) dt M(D)×C 0 (D) M(D)×C 0 (D) = 0 −∂t Λ − D V · Λ − ∗ D Λ · V + (div V) Λ. ∂t Z + [Z. 1 j (V.76).80) Actually. 1) × D ¯ ∂t λ (3.x (3. W) = V Ca. 1) We can also observe that directly from (3. W) = 0 V Ca.76) We recall that the transverse field Z solves the following dynamical system. We introduce the adjoint state Λ solution of the following backward dynamical system. D V ∗ −∂t Λ − D V · Λ − D Λ · V + (div V) Λ = 0. (0. j (V. ∂t Z + [Z. 1 Λ. o t ∈ (0. Ca. W) = Λ(1). j (V. using (3.x dt (3. ¯ ˜ λ(1) = k. This leads to the following formula.47).78) V V ¯ V where γCa.x ) is the trace operator on the curve Ca. Z + Λ(1). we get the following functional derivative. we have 1 j (V.77) with [Z. ∗ Λ(1) = γCa.x V [∂n G + H] Z(1).81) ¯ where λ solves the backward problem.x ) (λ nt ).x V ¯ λ W(t). n dCa. (0. W M(D)×C 0 (D) dt (3.x .78). ∗ ˜ Λ(t) = γTt (V)(Ca. (0.80 Moving Shape Analysis which leads to the following derivative. Z(1) M(D)×C 0 (D) We may use the following integration by parts formula.82) . 1) × D Z(t = 0) = 0 D def (3. V] = W. D ¯ + div(λ V) = 0. nt dCa. V] = D Z · V − D V · Z.x · [(∂n G + H) n] . Z(0) As Z(0) = 0 and Λ is the solution of the adjoint system (3. it can be proven that the measure Λ is of the following form.x ∈ L(C 0 (D). 1) × D (3. W) = 0 Λ.

we choose g(y) = g (H(y). n ] Z(1). n ) This leads to j (V. n(x) − ∂τ (∂H g (a)) Z(1)(a). 1) × D ˜ ˜ λ(0) = k. we assume that the density function g depends on the curvature of the curves.Z(1) (y) + Furthermore.Z(1) = ∂H g (H(y).x k ◦ T1−t (V) nt (3. W) = V Ca. We set λ(t) = λ(1 − t) and V(t) = −V(1 − t).x ¯ V Using tangential Stokes formula [53] on Ca. n [−∂H g ∆C ( Z(1). nt dCa. it leads to ¯ ˜ λ(t) = λ(1 − t) = k ◦ T1−t (V) Then. n(x) + ∂H g (a) ∂τ Z(1)(a). n ) Z(1). D ˜ ˜ λ(t) = k ◦ Tt (V)−1 and we get Using the fact that Tt (V)−1 = Tt (Wt ) with Wt (s. n ] dCa. n(a) ¯ ¯ (3. Here.Weak evolution of sets and tube derivatives def 81 ˜ with k any extension of the function k = ∂n G + H to the domain D.x ¯ + ∂τ (∂H g (x)) Z(1)(x).x dt and V j(V) = ∗ γCa.x V y ∈ V (Ca.83) Curvature dependent density Now.x [−∆C (∂H g ) + H g + ¯ V g . W) = V Ca. (0. n dCa. n(y) Z(1).x V k ◦ T1−t (V) W(t). If div V = 0.84) .Z(1) = −∆C ( Z(1). y). n ) + (H g + ¯ V g . y) = −V(t − s. n(a) ¯ ¯ − ∂H g (x) ∂τ Z(1)(x). y). then the equation for λ turns to be a backward convection of ˜ ¯ ˜ the ending term. we get 1 j (V.x ) def ¯ y g (H(y). we get ¯ gC. we have HC. ¯ where H(y) = ∆b(x).x . then we get ˜ ˜ ˜ ∂t λ + λ · V = 0. W) = 0 V Ca. y). we have j (V. Then. y) HC.

82 Moving Shape Analysis Since the extreme points x and a are fixed. n ¯ (3.83) with k = −∆C (∂H g ) + H g + ¯ ¯ g. i.x [−∆C (∂H g ) + H g + ¯ V g .86) . V(a) = V(x) = W(a) = W(x) = 0 which implies Z(1)(a) = Z(1)(x) = 0 and the previous derivative becomes j (V.85) Eventually.x ¯ (3. n dCa. their velocities are zero.e. we get the gradient of j(V) using the expression furnished by equation (3.. n ] Z(1). W) = V Ca.

[147]. ∀t 0 (4. V(t) 0 A∗ × A k k dt.1) Classical gradient based methods allow us to control the variations of J with respect to the domain. This may be done by solving the non-linear equation for large evolution of the domain. we have 1 .1 Introduction In this chapter.2) where A is an ad-hoc duality operator. For any Ω0 in Ok and V in Vk . We consider a shape functional J which is shape differentiable in O k with respect to Vk to be specified later on. we recall the concept of shape differential equation developed in [145]. we present a simplified version and some applications in dimension 2 which enable us to reach the time asymptotic result. Usually it is a function on Γ so that we consider any extension G of g defined 1 Under some regularity assumptions which are satisfied for a large class of problems. Here. the absolute k continuity of J writes s J(Ωs (V)) − J(Ω0 ) = J(Ωt (V)) . ∗ J(Ω) = γΓ · (g n) where the shape density gradient g is a distribution on the boundary Γ. From the structure theorem for shape gradient [135]. J(Ωt (V)) + A(V(t)) = 0. Considering the problem. 83 . We denote J(Ω) its gradient. ∀s 0 (4. considered as a distribution in A∗ . V∈Vk min J(Ω(V)) we would like to elaborate a constructive method to decrease the functional following the gradient.Chapter 4 Shape differential equation and level set formulation 4.This corresponds to the well-known steepest descent method.

(4. The operator A∗ stands for the dual space of distributions of order less than k 2 It holds for shape differentiable functional whose gradient is continuous and bounded on Ok . It is endowed with 3 the Courant metric dk which is defined on the family of images of a given domain. In this case the term γ Γ · (g n) can be identified with G χ. ranging in a Sobolev space of Distributions. and Olip the set of Lipschitz open subset of D. we recall the material introduced in [145. Hence.5) Vk (I) = V ∈ C 0 (I. Ω0 ⊂ D (4. The existence of a solution for this so-called shape differential equation has been proven in [147] inside a larger setting2 . τ ) χ(0) = χΩ0 . (0. Φ(t. we simply denote Vk = Vk (R+ ). Ak ) where I is an interval of R+ which contains 0. Rd ) whose elements are C k -diffeomorphism of D. A−1 · (G(χ) χ) = 0. ∂t χ + χ. We denote T k the subset of ¯ ¯ C k (D.84 Moving Shape Analysis ∗ in a neighbourhood of the boundary Γ. We denote by Ok the set of all open C k -submanifold of D. τ ) Φ(0) = Φ0 . x) > 0}. 4. endowed with the Courant’s metric topology. ν Rd = 0. A−1 · (G(χt (Φ)) χt (Φ)) = 0. (0. 3 We refer the reader to the book [51]. 147] while solving the so-called shape differential equation. 146. We are going to recall the constructive proof of the existence of a V satisfying (4.4) where χt (Φ) = {x ∈ D. (4. . on ∂D . We are going to use the following spaces Ak = V ∈ C k (D. For I = R+ . ∂t Φ − Φ. We fix a smooth bounded hold-all D in Rd and a non-negative integer k.3) We shall see in the sequel that the previous equation can be weakened using the level set formulation where we shall solve the following equation.2 Classical shape differential equation setting In this section. the shape differential equation turns into an Hamilton-Jacobi equation for the characteristic function χ.2) and investigate the asymptotic behaviour of the method. Rd ) V.

1 [51] The inclusion (Ok+1 (Ω0 ). Although Tk is not a vector-space. for shortness.Tk ) Tk Tk For (Tk . we will write. Tk ) = sup T (s) s∈I and ¯ Vk (I) → C 1 (I. Rd )) V → [t → Tt (V)] are continuous. it corresponds to the space of Radon measures. we set Ok (Ω0 ) = REMARK 4.Shape differential equation and level set formulation k. Tk ) V → [t → Tt (V)] . C 1 (I.[150] The mapping Vk −→ C 0 (I. For any fixed domain Ω0 ∈ Ok . we have a result similar to theorem 4. LEMMA 4. PROPOSITION 4. C k (D. the following compactness result holds. Rd )) −1 V → [t → Tt (V)] T ∈ C 0 (I. dk ) is a complete metric space. C k (D. Also for a bounded universe D. dk ) is compact.2 The mapping Vk (I) → C 1 (I.1 [147]. dk ). the following continuity result holds. THEOREM 4. Ok (Ω0 )) V → [t → Ωt (V) = Tt (V)(Ω0 )] is continuous and maps bounded subsets on equicontinuous parts. THEOREM 4.1 The mappings ¯ Vk (I) → C 1 (I. In the case k = 0.1 Ω ∈ Ok ∃ T ∈ Tk . Tk ) = T T ∈ C 0 (I. For a bounded universe D. dk+1 ) → (Ok (Ω0 ). Ω = T (Ω0 ) 85 (Ok (Ω0 ).1. Tk ) + sup T (s) s∈I This space is endowed with the canonical norm C 1 (I.

A characterization involving Ok would be more elegant. s → G T (s)(Ω0 ) belongs to C 0 (I. ii) for any initial domain Ω0 . not the diffeomorphism. k+1 (ii) J are uniformly bounded on Ok+1 (respectively in R and J is shape-continuous on Ok+1 . We have the following characterization of the shape continuity. with respect to Vk+1 .3. The followings properties are equivalent i) G is shape continuous with respect to Vk (I): for any initial domain Ω0 . B).. for all V ∈ Vk (I). k Then. COROLLARY 4. THEOREM 4.1 Let G be a shape function defined on Ok with values in a fixed Banach space B. 51] for instance) that a shape functional G defined on O k with values in a fixed Banach space B is shape continuous (in the usual sense) as soon as [s → G(Ω(s))] is continuous (i. Tk ). s → G(Ωs (V)) belongs to C 0 (I. for any s 0. 4. B). for any T in C 1 (I. Let us assume the following conditions: (i) Both J and A∗ ).1 The shape control problem An existence result for the shape differential equation This section aims at proving the following theorem. Ok ). differentiable in Ok with respect to Vk . It is important to notice how easy it is to characterize the shape continuity via the space Tk . since the real objects are the domains. there exists V ∈ Vk+1 ∩ L2 (R+ .3 Let J be a shape functional.2).3 4.86 Moving Shape Analysis is surjective. Ak+1 ) and c > 0 such that. using a solution of equation (4. B)) for any [s → Ω(s)] ∈ C 0 (I. continuous and maps bounded subsets on equicontinuous parts. s J(Ωs (V)) − J(Ω0 (V)) = − V(t) 0 2 s dt = −c J(Ωt (V)) 0 2 dt . It is known (see [147. belongs to C 0 (I.e. in A∗ .

Rd ) V. H →Ak+1 →Ak (4. . i.2) exists. G is compact. an arbitrary interval I of R+ which contains 0.6) We denote A the linear and continuous duality operator from H to its dual H∗ .m = V ∈ Vk+1 sup V(s) s∈I Ak <m ⊃ GI (Bk. there exists m1 (which may depend on Ω0 ) such that for any Ω ∈ Ok (Ω0 ).2). J(Ω) It follows that GI (V)(s) H A∗ k ≤ m1 ≤ m1 A−1 L(H∗ . Provided I is compact.Shape differential equation and level set formulation 87 PROOF Provided the duality operator A of equation (4. the domain Ω0 being fixed in Ok+1 .H) = m1 The choice m = m1 is convenient. Since we assumed the shape continuity of J. LEMMA 4.. that equation (4. Let GI be the mapping defined by GI (V) : I −→ H s → −A−1 J(Ωs (V)) (4.m ) PROOF Due to the boundedness of J. and give a constructive proof of the existence of a solution of (4. a solution of this equation is convenient for the theorem. GI (V) ∈ C 0 (I.2 There exists m > 0 such that Bk. We fix κ > 1 such that the following space H= V ∈ Hκ (D. We consider. We are going to prove that GI has a fixed point. We are going to use a Sobolev space embedded in Ak to ensure the existence (and good properties) of the duality operator A. n Rd = 0 on ∂D satisfies the following embedding chain rule. H) ⊂ Vk+1 . LEMMA 4.e.2) admits at least one solution.3 The mapping GI is continuous. H).7) for V ∈ C 0 (I.

1]. G1 is a compact mapping. the image of B is para-compact in C 0 (I. By Ascoli’s theorem and the compactness of the inclusion of T k+1 (Ω0 ) in Tk (Ω0 ) (Theorem 4. Tk ) −→ C 0 (I. Tk ) V → [s → Ts (V)] G2 : C 0 (I.)(Ω0 )) G3 : C 0 (I. A∗ ) k T → J(T (. H) and (Ωn )n∈N ⊂ Ok+1 (Ω0 ) by ∀n ∈ N . Applying Leray-Schauder’s fixed point theorem. Using Corollary 4. A∗ ) −→ C 0 (I. We suppose I is compact. and so is GI . .n+1] n on D The continuity at integer points comes from equation (4. for any s 0. we deduce that GI is continuous.1. we shall build a sequence of converging domains.3. Accordingly.n+1] (V      [n.8) This so-built field V satisfies Theorem 4. We have.1).1] (V) = V.3. the continuity of G2 is equivalent to the shape continuity of the J and since the continuity of G3 is clear.   G[n. H) with G[0. a bounded subset B ⊂ C 0 (I. H∗ ) −→ C 0 (I.2 A constructive algorithm Now. H) −→ C 0 (I.2. By Theorem 4. V(t) 0 2 H dt.2).n+1] )=V Ωn (Vn ) = Ωn+1 (Vn ) = Ω0 (V0 ) = Ω Ωn+1 Ω0 [n. 4.88 PROOF Moving Shape Analysis GI can be split using the following chain rule GI = G 3 ◦ G 2 ◦ G 1 where G1 : C 0 (I. J(Ωn+1 ) − J(Ωn ) = − n+1 V(t) n 2 H n+1 dt = − J(Ωt (V) n s 2 H∗ dt (4. Tk+1 (Ω0 )). s → J(Ωn ) is bounded decreasing. H) k g → g → −A−1 g Theorem 4. Ak+1 ). Ak+1 ) is mapped by G1 on a equicontinuous part of C 0 (I.2 provides the continuity of G1 . Tk ). Since we assumed J is bounded. Let us define V ∈ C 0 (R+ .we infer that for any initial domain Ω0 there exists V in C 0 ([0. hence has a limit and so does which proves V ∈ L2 (R+ . and any n.

We denote Ωn = Ωsn (V) ⊂ Ok+1 (Ω0 ). .3. The gradient J(Ω∗ ) is not a priori defined. V(n) H = J(Ωn ) hence J(Ωn ) → 0. then Ωn → Ω∗ in Oop and J(Ωn ) → 0 and J(Ωn ) → J(Ω∗ ) PROOF The sequence (Ωn ) may be regarded as a sequence in the compact space Oop . since the L2 convergence of the speed given by this method is not sufficient in general. since the limit set has not enough regularity. Then (Ωn ) has cluster points in Oop and if Ω∗ is one of them. The sequence (Ωn ) may not be bounded in Ok (Ω0 ) or Ok+1 (Ω0 ).4 Assume that (i) the shape functional J verifies the assumptions of Theorem 4. (ii) the shape functional J is defined and continuous in Oop . Nevertheless.4 The asymptotic behaviour If V is given by Theorem 4. Ak+1 ). Ω2 ) = max x1 ∈D\Ω1 x2 ∈D \Ω2 sup inf |x1 − x2 | . Nevertheless since V satisfies (4.4) is satisfied.2).Shape differential equation and level set formulation 89 4. We denote Oop the family of all open subsets of D. we shall furnish an example of a functional where the continuity hypothesis of Lemma (4. x2 ∈D\Ω2 x1 ∈D \Ω1 sup inf |x1 − x2 | (4. (iii) the shape functional gradient J is continuous for the Hausdorff complementary topology on Ok (Ω0 ). In the next section. there exists a non-decreasing sequence (s n )n 0 such that V(sn ) → 0 since V ∈ L2 (R+ . it converges towards an open subset Ω∗ of D.9) LEMMA 4. Hence passing to a subsequence.3. In [51] it is proven to be a compact metric space for the Hausdorff-complementary metric d(Ω1 . we can use a weaker topology on the space of domains.

We consider the Dirichlet problem P(Ω. EΩ. z H−1 (Ω)×H1 (Ω) 0 + Ω 1 | z|2 2 (4.10) where |. g) 2 norm in the dual space H−1 (Ω). g) of P(Ω. Then { y(Ω. gΩ ∈ H−1 (Ω).g (y(Ω.90 Moving Shape Analysis 4.g defined on H1 (Ω) by 0 EΩ. Ω |Ω Ω. Accordingly.| denotes the euclidean norm in Rd . g) in the space H1 (Ω) endowed with the 0 norm z 2 Ω = Ω | z|2 . g) is the unique minimum of the functional EΩ. g) Ω ≤ gΩ Ω. This yields to Thus it leads to y(Ω) Ω Ω where .5.11) A mere consequence of this estimate is the following uniform boundedness result. A priori estimates An a priori estimate for solution y(Ω.∗ |Ω ∈ O} ∈ O} . g) is bounded.5 4.1 Shape differential equation for the Laplace equation The Laplace equation In this section.∗ y(Ω. g) is derived from the variational formulation of the problem : y(Ω. LEMMA 4. g)) ≤ 0 1 2 y(Ω. g) −∆y = gΩ in Ω y = 0 on Γ = ∂Ω which has a unique solution y(Ω. we are given a family g = (gΩ )Ω∈Olip such that for any Ω ∈ Olip .g (z) = − gΩ .5 Let O be a subset of Olip such that { gΩ is bounded.∗ denotes the ≤ 2 gΩ Ω. Ω.∗ (4.

For any z in H1 (Ω0 ).e. cP (Ω)−1 = inf z z L2 (Ω)d L2 (Ω) z∈H1 (Ω) 0 z=0 5 The 6 This where γs = det D Ts . i.Shape differential equation and level set formulation In the sequel. f ) with respect to the domain. Since f|Ω Ω.12) may be defined via Rayleigh-quotient. Eventually.6 There exist a constant cP > 0 such that ∀ Ω ∈ Olip . we shall use the family (f|Ω )Ω∈Olip with f ∈ L2 (D). Extending zΩ by 0 provides a z˜ in H1 (D) Ω 0 such that cP (D)−1 ≤ z˜ L2 (Ω)d . Therefore. the mapping6 s → gΩs (V) Ts is continuous from R+ to H−1 (Ω0 ) ii) For any V in Vk . the e uniform boundedness property of the solutions of P(Ω. f ) 5 will arise from the following uniform boundedness of Poincar´’s constant. there exists zΩ in H1 (Ω) 0 with cP (Ω)−1 = zΩ L2 (Ω)d . γs z ◦ Ts H−1 (Ω 0 0 0 1 s (V))×H0 (Ωs (V)) . i) The family gΩ is shape continuous. (f|Ω )Ω∈Olip ). mapping is defined as follows. accurate notation for this is P(Ω. e LEMMA 4. z H−1 (Ω )×H1 (Ω ) = gΩs (V) . for any V in Vk . cp (Ω) ≤ cP PROOF It is classical that for any Ω ∈ Olip . Thus cP (D) cP (Ω) so cP = cP (D) is Ω convenient. 0 ˙ ¸ ˙ ¸ −1 −1 gΩs (V) Ts . Strong shape continuity properties Let Ω0 be a fixed initial domain in Olip and let us assume the following. we have proven the uniform boundedness of the solutions of P(Ω. s → gΩs (V) 4 It H−1 (Ωs ) ∈ L∞ (R+ ) loc (4.∗ 91 ≤ f L2 (D) sup z∈H1 (Ω) 0 z ≤1 L2 (Ω) z Ω ≤ cP (Ω) f L2 (D) where cP (Ω) is the Poincar´’s constant4 for the domain Ω.

g (ys ) = min EΩs . 0 −1 E Ωs .g the functional defined on H1 (Ω0 ) by the following identity. g) is shape continuous: ∀ V ∈ Vk . Due to the local boundedness property (4. R ). Since ys s 2 Ω0 −1 ≤ γ(s)−1 | D(Ts )−1 | L∞ (D) s→0 s→0 ys Ωs the family (y )0≤s≤ε is uniformly bounded in Hence we can extract a subsequence (y sn )n 0 where 0 ≤ sn ≤ ε and sn → 0 that converges towards an element y ∗ . PROOF For the sake of simplicity. g)) is continuous. γ(sn )y sn H−1 (Ω0 )×H1 (Ω0 ) 0 −→ gΩ0 . g) in H1 (D) ˜ ˜ 0 Moreover. E(Ω0 .g (y(Ω0 . This is sufficient for g Ωs n Tsn . there exists ε such that ( ys Ωs )0≤s≤ε is uniformly bounded. for k 2. We have y0 = y 0 . the following properties hold.13) Since k 2 the Jacobian γsn converges towards γ(0) ≡ 1 in C (D.4 Under the last hypothesis. 0 = − g Ωs and we have Ts . weakly in H1 (Ω0 ). γ s z H−1 (Ω0 )×H1 (Ω0 ) 0 + Ω0 1∗ −1 | D Ts · 2 z|2 γs min E Ωs . y(Ωs (V).g (z) = EΩs . g) ◦ Ts −→ y(Ω0 . g) −→ y (Ω0 . these properties are still valid for k = 1 provided gΩ ∈ L2 (Ω) for any Ω ∈ Olip . i) the transported solution of P(Ω0 . g) and y s = ys ◦ Ts . Rd ) induces −1 the weak convergence of (γsn ∗ D Tsn ) · y sn towards y ∗ in L2 (Ω0 )d .g (y s ) = EΩs . g) in H1 (Ω0 ) 0 ii) the energy functional. g) is shape continuous: ∀ V ∈ Vk . the convergence of γsn (D Tsn )−1 towards I in C 0 (D. g) = EΩ0 . Using the weak-lower semi-continuity of the L2 -norm.14) Furthermore.g 1 d (4.g (z ◦ Ts ) def H1 (Ω0 ). 0 We denote E Ωs .12).g = E Ωs . y 0 s→0 H−1 (Ω0 )×H1 (Ω0 ) 0 (4. y (Ωs (V). iii) the extended solution of P(Ω0 . we have Ω0 1 | y ∗ |2 ≤ lim inf n→∞ 2 Ω0 1∗ −1 | D T sn · 2 y sn | 2 γ sn .92 Moving Shape Analysis THEOREM 4. we denote ys = y(Ωs (V).

y∗ has support in Ω0 and may be written y with y ∈ H1 (Ω0 ).g (y ∗ ) ≤ lim inf E Ωsn .g (y(Ωs . we come to EΩ. we have proven the weak-lower semi-continuity of the mapping (s.14) and the weak continuity of s → y s .15) which leads together with equation (4. g)) and equation (4. g) 2 2 Ω =− 1 gΩ . we have −1 −1 y˜n ◦ Tsn = ysn ◦ Tsn s y in H1 (D) ˜ 0 Accordingly.g (z) on R+ × H1 (Ω0 ) at (0.g (z) n→∞ n→∞ But s → E Ωs .14).g (y sn ) ≤ lim inf E Ωsn . This eventually proves that 0 y˜n s y0 in H1 (D) ˜ 0 But due to the continuity of s → EΩs . g) . which proves a weak shape-continuity for the transported solutions of P(Ω. Ω z Rd = gΩ . 0 y(Ω. The strong continuity will arise from a continuity of the norms. z) for any z. z H−1 (Ω)×H1 (Ω) 0 Choosing z = y(Ω. g). y sn y in H1 (Ω0 ) and y = y0 .g (z) is continuous for any z. Since s 0 0 the sequence (Ωsn ) converges towards Ω0 for the Hausdorff complementary topology. since ˜ 0 the boundary ∂Ω0 has non-zero capacity. Using arguments similar to the ones which established (4. y˜n s and this is sufficient for y˜n → y0 (strongly) in H1 (D) ˜ s 0 and y sn → y 0 (strongly) in H1 (Ω0 ) 0 D = y sn Ωs n → y0 ˜ D = y0 Ω0 . via the socalled compliance equality. Hence y ∗ = y 0 = y0 . g) in the above equation. there exists a y∗ in H1 (D) such that y˜n y∗ in H1 (D). g)) = − 1 y(Ω. and a compactness argument for the y s . Since the sequence ( ysn Ω ) = ( y˜n D ). y(Ω.g is written ∀ z ∈ H1 (Ω).g (y(Ω. g) 2 H−1 (Ω)×H1 (Ω) 0 (4. g)) is continuous. For any Ω in Olip .15). z) → E Ωs .g (y(Ωs .Shape differential equation and level set formulation 93 Thus. This proves that for any z 0 E Ω0 . where ˜ denote the extension to . that s → EΩs . s D with 0. the necessary (and sufficient) condition of optimality for E Ω.

It may easily be extended to a domain-dependent right-hand side [135].94 Moving Shape Analysis REMARK 4.17) where yd ∈ H1 (D) is a given target. ˙ Then the solution y(Ω) of problem P(Ω. the functional J admits an Eulerian derivative. for any φ ∈ H1 (Ω). p (4. V) ˙ in H1 (Ω) for any velocity field V ∈ Vk . for any field V ∈ Vk (k ≥ 1). J(Ω) = Ω (y(Ω) − yd )2 (4. 0 The shape gradient PROPOSITION 4. Moreover. In the case where g Ω ∈ L2 (Ω) for any Ω ∈ Olip .2 The shape control problem We would like to minimize the following shape functional. y(Ω) − yd ). ˙ Ω φ = y(Ω) .14). φ Rd 1 − [ div V(0) I −ε(V(0))] · 2 Ω + [gΩ div(V(0)) + gΩ0 .2 In this proof.16) 4. the assumption k 2 is needed to prove the convergence result (4. gΩ0 ∈ L2 (Ω0 ) and for any V inVk the mapping s → gΩs (V) ◦ Ts is strongly differentiable at s = 0 in H−1 (Ω0 ) with derivative g(Ω0 .5 Assume that for any Ω0 in Ok .18) ˜ where p is the solution of the adjoint problem P(Ω. V) = Ω (y − yd ) [div V(0)(y − yd ) − D yd · V(0)] +2 A y.2 For any domain Ω in Olip .5. f ) has a material derivative y(Ω. V) . The following differentiability result is well known when the right-hand side is fixed in H1 (D). . p − 2 div(f V) . 0 0 y(Ω.V ] φ ˙ (4. this assumption is not needed anymore. THEOREM 4. dJ(Ω.V) .

the strong material derivative of y(Ω) satisfies ∀ φ ∈ H1 (Ω) . the mapping V → dJ(Ω. We consider the element J(Ω) of A∗ given by k J(Ω) . V) = Ω [div V(0)] (y − yd )2 − 2(y − yd ) (y − D yd · V(0)) ˙ We consider the adjoint problem. 0 with Then.5. we deduce dJ(Ω. φ A (V(0)) = div V(0) I −2 ε(V(0)) div V(0)(y − yd )2 + 2(∆p) (y − D yd · V(0)) ˙ div V(0)(y − yd )2 − D yd · V(0) − 2 p · (y − yd )[div V(0)(y − yd ) − D yd · V(0)] p − 2 div(f V) . V) is linear and continuous. ˙ Ω φ =− def A (V). p y ˙ = Ω = Ω +2 A (V(0)) y .20) .19) Thanks to Theorem 4. p(Ω) Ω + 2 A (V) y(Ω) . ˜ P(Ω. we get ∂s J(Ωs ) = Ω γ (s)(y s − yd ◦ Ts )2 − 2γ(s)(y s − yd ◦ Ts )(∂s y s − D y ◦ Ts ∂s Ts ) Therefore. y(Ω) − yd ) −∆p = y − yd on Ω p=0 on Γ = ∂Ω (4. V A∗ ×Ak k = Ω (y(Ω) − yd )[div V(y(Ω) − yd ) − D yd · V] p −2 div(f V) . When the domain Ω is fixed.Shape differential equation and level set formulation 95 PROOF If Ts is the flow-mapping associated to the velocity field V. V) = Ω y. Ω φ + div(f V(0)) . y . dJ(Ω. a change of variable gives J(Ωs ) = Ωs (ys − yd )2 = Ω γs (y s − yd ◦ Ts )2 Since s → Ts is of class C 1 . (4.

p Theorem 4. ∀ Ω ∈ Olip = sup V∈Ak J(Ω) . y.V] + 2χΩ A (V) y . s → J(Ωs (V)) is continuous.3 to the Laplace equation P(Ω. there exists a constant m > 0 such that J(Ω) A∗ k ≤m y Ω Ω p Ω Ω V Ak and the uniform boundedness of y and p gives the correct result. g). p(Ω)) ˜ ˜ which is continuous from L2 (D) × H1 (D) × H1 (D) to A∗ with 0 0 k G(χΩ .6 The gradient distribution J is continuous in A∗ . . PROPOSITION 4. y (Ω).3 An asymptotic result in the 2D case We shall apply theorem 4. there exists a V ∈ Vk+1 and an open subset Ω∗ of D ⊂ R2 such that. k PROOF The gradient J(Ω) is a continuous function G(χ Ω . Shape continuity properties of the gradient THEOREM 4. − 2χΩ div(f V) . Consequently. s → p(Ωs (V)) are ˜ ˜ continuous for any V in Vk .4 provides that the mapping s → y (Ωs (V)).3 There exists a constant M > 0 such that J(Ω) PROOF J(Ω) A∗ k A∗ k ≤ M. 4.4 For any Ω0 in Ok+1 . V A∗ ×Ak k 2 ≤ div V L∞ y − yd L2 + D yd · (y − yd ) L1 V L∞ +2 A (V) L∞ y .96 Uniform boundedness Moving Shape Analysis PROPOSITION 4. p L1 + 2 f | p| L1 V L∞ Then. p) .5. V D A∗ ×Ak k = p χΩ (y − yd )[div V · (y − yd ) − D yd .

provided ( Ωn )n ˜ ˜ converges to Ω for Hausdorff topology. The general asymptotic behaviour of section 4.6 Shape differential equation in Rd+1 We come back to the optimal dynamical evolution of a geometrical domain. As an example of such a state variable. we can consider the non-cylindrical . Indeed. x)2 )−1/2 − v(t.Shape differential equation and level set formulation s 97 i) J(Ωs (V)) − J(Ω0 ) = V(t) 0 2 dt ii) for any sequence (sn )n 0 with sn → ∞. Sverak has proven in [137]. The element yQ ∈ W (Q) is a state variable which is defined as an element of some functional space built on the tube Q. Nevertheless. [19]). nt (x) where nt (x) ⊂ Rd is the unitary normal field to Γt = ∂Ωt . J(Ωsn (V)) → J(Ω∗ ) and J(Ωsn (V)) → 0. In the 2-dimensional case. say Ωt . x) = (1 + v(t. Yg ∈ H 1 ((0. τ ) × D) where D is a given bounded open domain in Rd+1 such that the cylindrical domain ]0. pointing outside the moving domain Ωt . the continuity of Ω → y (Ω) for Hausdorff-complementary ˜ topology does not hold in general. 4. where t is the time parameter. with Ωn uniformly bounded. τ [×D will contain the potential evolution of the moving tubes Q s .g. a moving domain. e. For each tube Q= {t} × Ωt ⊂ Rd+1 0≤t≤τ with lateral boundary Σ= 0≤t≤τ {t} × ∂Ωt ⊂ Rd+1 We define as previously the unitary normal field to Σ pointing outside Q.That optimality is built in the context of the minimization of some cost functional.[136] the convergence of y (Ωn ) towards y (Ω). J(Q) = 1 2 |yQ − Yg |2 dx dt def Q where Yg is a given element. Ωsn (V) → Ω∗ for Hausdorff complementary topology. ν(t.. this continuity holds under capacity constraints( [23].4 may be developed in the case d = 2 . Let us consider a cost functional of the tracking type. x).

where for a given rhs F ∈ L2 0. τ . t.τ.H0 (Ωt )) ψ∈L2 (0. ψ) = 1 2 |φ(t) ◦ Ts (Z)−1 − Yg |2 dx dt Qs Qs + + ∂t (φ(t) ◦ Ts (Z)−1 ) ∂t (ψ(t)oTs (Z)−1 ) (ψ(t) ◦ Ts (Z)−1 ) dx dt y1 ψ(0) dx Ω0 (φ(t) ◦ Ts (Z)−1 ) · − Qs F ψ(t) ◦ Ts (Z)−1 . L(s. x) ∈ Rt × Rd x we consider the associated flow mapping ˜ Ts (Z)(t. ∂tt y − ∆y = F. ψ) where the Lagrangian functional is given by the following expression. in Q (4. x) From the classical shape analysis. A special role will be played by the d dimensional vector field Z(t)(s. φ. x) The specific structure of that flow is obviously deriving from the “horizontal” ˜ character of the vector field Z itself. Ts (Z)(t. dx dt − 7 Indeed a general form for the vector field Z could induce a change in the time variable t ˜ which would create a “time extension” of the tube. That choice does not produce any gener7 ality loss in the potential deformations of the cylinder Q. we look for the best tube Q ⊂ R d+1 which minimizes the functional J. t.τ. we just consider the tube Q as a moving domain in Rd+1 and we consider an “evolution” parameter s to “follow” the virtual evolution of the tube Qs . For that purpose. L2 (D) . τ. the state variable 1 yQ ∈ L2 (0. H0 (Ωt )) satisfies the following equation. we derive the R d+1 -shape gradient associated to the shape functional J(Q) as follows: J(Qs ) = 1 1 φ∈L2 (0. x) := Z(s. . φ. x) = 0. In other words.21) The basic “Optimal Moving Domain” problem consists in finding the “best” dynamical time evolution for the moving domain Ω t in order to minimize the functional J. x) = t.98 Moving Shape Analysis wave equation.H0 (Ωt )) def min max L(s. Given a smooth enough vector field def ˜ Z(s. Z(s. t.

we can state that the eulerian directional derivative of J is only supported by the lateral boundary Σ and is given by the following expression. Now we apply. ˜ we derive the existence of a solution Z = (0. G(Q) = def ∗ γΣ (g ν ) where the density gradient g is an element of L1 (Σ) (in general a (d+1).vector measure on Σ with zero transverse order) and its expression involves.τ [×D. nt (x) g(t.Rd+1 ) Thus the “time-space” vectorial distribution is given by the following expression.Shape differential equation and level set formulation 99 Using the structure theorem. x)2 dt dΓ(x) Moreover ˜ Z(t. So we obtain s0 ˜ Z(s) 2 ds J(Qs ) ≤ J(Q) − α 0 0 H(D) for some “pivot” Hilbertian Sobolev vector space H(D) over the hold-all domain D. ˜ dJ(Q. Z) = G(Q). x) Z(t. def ˜ ˜ The construction of the tube Qs0 = Ts (Z)(Q) from the field Z solution d+1 to the previous R -shape differential equation consists in finding the R d dimensional vector field V(t.22) where A stands for an ad hoc duality positive operator as described previously. in the (d+1)-dimensional case. Z) = Notice that τ def ∂ J(Qs ) ∂s = s=0 Σ ˜ g Z · ν dΣ f dΣ = Σ 0 Γt f (t. ν(t. the Shape Differential Equation framework developed in the previous section of this chapter. x). A−1 . in the non-cylindrical wave equation case. x) = So that ˜ dJ(Q. the normal derivatives at the boundary of the state and co-state solution.Rd+1 )×D(]0. As a consequence. Z(0) D (]0.G(Ts (Z) (4. ˜ ˜ Z = 0. x) · nt (x) dΓt (x) dt That is ˜ ˜ dJ(Q. Z) to the shape differential equation.τ [×D. x) 1 + v(t. Z) = 0 τ Γt 1 1 + v(t. x)2 Z(t. x). x) whose flow mapping T t (V) builds the tube .

25) Also the previous differential equation can be rewritten in term of Z(t) as follows. ∂t s0 ∂ ∂ ˜ ˜ ˜ Z(t. x) = Z(s. t. the Rd -dimensional speed vector V s0 is furnished by the following expression. x)) dσ ∂t ∂t 0 By definition. Now we look for a vector field Vs0 (t. s0 Ts0 (Z(t))(x) = x + 0 Z(t)(σ. Tσ (Z)(t. x)) ∂t 0 (4. Tσ (Z(t))(x)) dσ (4..100 Moving Shape Analysis Qs0 from the initial domain Ω0 . ˜ Ts0 (Z)(t. since the Rd+1 -dimensional flow is horizontal. Tσ (Z)(t. x) such that its flow mapping Tt (Vs0 ) builds the “optimal” tubes Qs0 from the initial domain Ω0 . x) = 1. Vs0 (t. τ [×Ω0 That tube is obviously built by the vector speed V = 0. x). Since.e. i. Ts0 (Z(t))(x)) then we obtain Vs0 (t. ∀0 < t < τ (4. Tσ (Z)(t. 1. x) = ∂ Ts0 (Z(t)) Ts0 (Z(t))−1 (x) ∂t (4. x) = t. Tσ (Z(t)))(x) · ∂ Tσ (Z(t))(x)} dσ ∂t .24) Now. Ts0 (Z)−1 (t. Tt (Vs0 ) = Ts0 (Z(t))(Ω0 ).26) (4. Tσ (Z(t)))(x) ∂t + D Z(t)(σ.23) where Z(t) stands for the Rd+1 -dimensional vector field Z(t)(s. x)) + Dx Z(σ. we have ˜ Ts0 (Z)(t. by definition. x) = def ∂ ˜ ˜ Ts (Z)(t. x)))dσ we get ∂ ˜ Ts0 (Z) (t. x + 0 s0 ˜ Z(σ. Let us assume that the initial tube Q is the cylindrical domain Q =]0.27) so that ∂ [Ts0 (Z(t))](x) = ∂t s0 0 { ∂ Z(t)(σ. x)) · (Tσ (Z)(t. x) = (t. This means that the following identity holds true. t. t.

x). x) = 0 } (4. Tσ (Z(t)) ◦ Ts0 (Z(t))−1 )(x) ∂t 0 +D Z(t)(σ. x) =0 Φ(t. with the notation Ωt (Φ) = Ωt (V) (V previously defined).(g n) = G(Ω) χΩ . we have ∗ G = J(Ω) = γΓ . The singular points are those in D at which the gradient of Φ(t. x) > 0 }.[51]. x) − Φ(t. . say Φ = −1 on the boundary ∂D. Tσ (Z(t))◦Ts0 (Z(t))−1 (x)) dσ Vs0 (t. we consider the scalar shape differential equation. Γt = {x ∈ D | Φ(t. the shape differential equation writes A−1 . If V is smooth enough we get Ωt = Tt (V)(Ω0 ) In this case.28) where Φ(t. x) 2 (4.1 The level set formulation Introduction In [149]. A−1 · J(Ωt (Φ)) = 0 (4.29) is always defined.7. In order to bypass that point.[135]. From the general structure theorem for shape gradient [135]. the following Hamilton-Jacobi equation for the function Φ: ∂t Φ(t. x) = −∂t Φ(t.) vanishes.Shape differential equation and level set formulation Then.30) which would imply. We assume that no such points lie on Γt so that in a neighbourhood of the lateral boundary Σ of the tube. J(Ωt (V)) − ∂t Φ(t.7 4. x) 2 (4. 101 ∂ Z(t)(σ. They would be merely equivalent if A−1 · J(Ωt (Φ)) was proportional to Φ(t.[93] we have considered domains family parametrized as follows: Ωt = Ωt (Φ) = {x ∈ D | Φ(t.30) and (4. . x) def Φ(t.31) are not equivalent. x) = { s0 4. x) Φ(t.) is a function defined on D verifying a negative condition at the boundary of D. x) Φ(t. Tσ (Z(t))◦Ts0 (Z(t))−1 )(x))·Vσ (t. the following vector field V(t.31) Notice that the Hamilton-Jacobi versions (4. x) .

33) It leads to the following normal Hamilton-Jacobi level set equation in the whole domain D : ∂t Φ + A−1 · G Φ =0 (4.35) so-called Harold Grad Adiabatic equation of plasmas at equilibrium in the Tokomak [148]. . we need to divide by Φ . . DEFINITION 4.. derived from a method.32) on Γt . while g is a scalar distribution on the boundary. it leads to the ∂t Φ = −A−1 · g(Γt ) Γt Φ (4. for any ε > 0 and for g ∈ H −1 (D).102 Moving Shape Analysis where n is the normal field to the boundary Γ. [149]. For that reason we now focus on a class of functions Φ without step so that Φ is different from zero almost everywhere in D.). nt = − A−1 · g(Γt ) on Γt Γt Φ(t. we have introduced for the modeling of free boundary value problems arising in plasma physics 8 . with nt = Φ(t. We choose to work with the normal scalar shape differential equation obtained by setting v = V(t.) without steps.2 Solutions without step We are interested in functions Φ(t. Consider. In almost classical regular problems g turns to be a function defined on the boundary with g ∈ L2 (Γ).) following normal-level shape differential equation : def (4.34) Γt Assuming Φ is a solution to the previous normal Hamilton-Jacobi.) In terms of level set modeling. 8 The in D (4. the variational problem z ∈ arg min 1 u∈H0 (D) D 1 2 u 2 + (ε|D| − g) u dx − ε 2 D D (u(x) − u(y))+ dx dy The associated Euler-Lagrange equation writes −∆z = ε β(z) + g. .1 step t if We say that a function f defined on a set D has a meas({x ∈ D | f (x) = t }) > 0 Now we shall describe a construction of function without step.7. in order to derive a solution to the previous normal scalar shape differential equation. 4.

the velocity vector field V = A−1 · J(Ωt (Φ)) only needs to be in L2 (0. τ. following the general assumptions of the structure derivative theorem.Shape differential equation and level set formulation with β(z)(x) = meas ({y ∈ D step condition : meas ({x ∈ D REMARK 4. L2 (D)3 ) and div Vn ∈ L2 (0.e.. the cost gradient takes the following form. τ . L2 (D)3 ) together with its divergence too. in order to perform a fixed point argument inside the following iterative approximation scheme. x) − with def Φn (t.1 (D) and is non-uniquely determined.3 | 103 z(y) < z(x)}). The idea is to choose a possibly non-linear operator A powerfull enough so that its inverse A−1 would map compactly and continuously the functional space where lies the cost function gradient into the previous velocity space. when Γ is smooth enough in order to make sense. Furthermore. Indeed. . It is a scalar distribution on the manifold Γ. L∞ (D)).38) 4. i. This method can be limited to zero step functions : 1 2 u 2 z ∈ arg min 1 u∈H0 (D) D − g u dx − ε 2 (u(x))+ dx D The associated Euler-Lagrange equation writes −∆z = ε β0 (z) + g. meas ({x ∈ D in D (4.31).36) | z(x) = t}) = 0. with support on the boundary Γ. together with the extra no step condition : | z(x) = 0}) = 0 (4. we understand that in the second Hamilton-Jacobi equation (4. In almost all smooth problems.7. τ . Vn−1 = 0 J(Ωt (Φn−1 )) we only need Vn ∈ L2 (0. together with the extra no ∀ t ∈ R+ (4.3 Iterative Scheme Obviously the function χ must satisfy the constraint χ 2 = χ while the level set function Φ does not. Then.37) with β0 (z)(x) = χ{x∈D|z(x)>0} . G ∈ W 1. x). Actually. ∂t Φn (t. zero transverse order and lying in some negative Sobolev space over D. G= J(Ωt (Φ)) = G(χt (Φ)) χt (Φ) Vn−1 = A−1 · So that G is a distribution. immediately from the previous study. only its trace 9 g on the 9 g is called the density gradient.

and ζ = 0 a. we get : ζ Φ = (Φ)+ So that ζ = 1 a. (4.36) } (A0 )−1 (g) = { z solutions to (4. In the limit case. if smooth enough.8 Let Φn . Assume that Φn −→ Φ strongly in L2 (D) and Φ is without zero step : meas({x ∈ D | Φ(x) = 0 }) = 0 Then let Ωn = {x ∈ D | Φn (x) > 0 }.104 Moving Shape Analysis boundary Γ.e. We consider the following family of without step functions. Ω = {x ∈ D | Φ(x) > 0 }. z ∈ (A0 )−1 (g). in D − Ω. Then χn −→ χ strongly in L2 (D). Then ε there exists a subsequence and a limit element z such that: 1 znk −→ z weakly in H0 (D) Moreover any such element z verifies z ∈ (Aε )−1 (g) (resp. We denote by χn and χ the respective characteristic functions of those subsets. so that the sequence converges strongly in L 2 (D). with respect to the characteristic function. Since Φ has no zero step.38) } ε LEMMA 4. (Aε )−1 (g) = { z solutions to (4. (4. Φ ∈ L2 (D). it is obvious that with some continuity assumption on the gradient G(χ) (that hypothesis will be satisfied in the following example). zn ∈ (A0 )−1 (gn )). we conclude that ζ = χ. as |(Φn )+ (x) − (Φ)+ (x)| ≤ |Φn (x) − Φ(x)|. is intrinsically determined. 0 ≤ ζ ≤ 1 such that χ n weakly converges to ζ in L2 (D). PROOF Obviously. LEMMA 4.37). ε The most important result concerning such elements without zero step is the following. we have χn Φn = (Φn )+ There exists a subsequence and an element ζ. ε . At that point. the previous iterative construction Vn will converge and we will derive the existence of solutions to the Hamilton-Jacobi equation for the Level set function associated to the multivalued operators Aε and A0 .7 Let gn −→ g in H −1 (D) and zn ∈ (Aε )−1 (gn ) (resp.35). in Ω.e.

we would like to compute the derivative of J(χ) with respect to χ. J(χ) = 1/2 E def (y − yd )2 dx where E ⊂ D. ψ + k χΩ φ ψ − f ψ + χE 1/2 (φ − yd )2 ) dx (4. We introduce the usual Lagrangian functional. ψ) = ( φ.4 An example: the transverse magnetic like inverse problem As an illustration for the construction of solution to the Hamilton-Jacobi equation. We introduce the observability functional.7. The classical shape derivative Here.1 (D) We easily check the continuity of the following mapping. χ ∈ L2 (D) −→ G(χ) ∈ W 2. it is very interesting.39) 1 where p ∈ H0 (D) is the solution of the adjoint equation : −∆p + k χ p = χE ( y − yd ) The associated density gradient is given by g(Γ) = (y p)|Γ Let us consider the extension G of g inside D.40) At that point. Let Ω ⊂ D and y = y(χ) ∈ H0 (D) be the solution to the following problem.1 (D) Topological derivative versus set derivative (4. It is given by ∗ G(Ω) = γΓ (y p n) (4. φ. on that example.Shape differential equation and level set formulation 105 4. G(χ) = y p ∈ W 2. −∆y + k χ y = f where k is the contrast parameter while f is given in L 2 (D) and χ = χΩ . The analysis is greatly simplified since we consider a bounded def 1 universe D. to understand that the shape derivative of that functional J(Ω) can be relaxed to a set derivative setting which coincide in a smooth situation with the classical shape derivative and shape gradient analysis. L(Ω. we consider a simplified version of the famous transverse magnetic inverse problem.41) D .

y0 . ε in M(D) ε→0 if the saddle point (yt . defined as follows. This technique enables us to manage holes creation inside D. Following this line. pt ) continuous in H(D) ⊂ C 0 (D) We get the notion of set derivative for the set functional J(Ω). . we had G(χ) = y p. ψ) with unique saddle point (y. p). assuming d = 2 so that y. as t → 0 of measurable subsets in D ⊂ Rd verifying: Ωn → {x0 } in Hausdorff topology (meas(Ωt ) / td ) → 0 as t → 0 ¯ Then. p ∈ H 2 (D) ⊂ C 0 (D). Assume for example that Ωt is a vanishing sequence.106 Then J(Ω) = Moving Shape Analysis φ∈H 2 (D) ψ∈H 2 (D) min max L(Ω. we get : ∂ J(χ(t)){t=0} = k (yp)(x0 ) ∂t This set derivative setting works thanks to the simple fact that χ(t) → δx0 t (Dirac measure at point x0 ∈ D). φ. Such an example is considered in ([129]) for a transverse magnetic inverse shape problem in buried obstacle reconstruction. we understand that the important concept in that situation is the derivative. That derivative does make sense as soon as the right hand side derivative does. we get (see([42]) : ∂ ∂ J(χ)|{t=0} = L(χ(t). pt ) satisfies the following continuity property: ¯ t −→ (yt . ∂ J(χ(t)) = < χ(t).42) ∂t In the previous example. p0 are the solution at t = 0. Hence. considering any evolutive characteristic function χ(t). The Lagrangian L being partially concaveconvex. weakly lower semi-continuous and upper weakly semi-continuous. when it exists. the following element: χ(t) = lim ˙ (χ(t + ε) − χ(t)) . p0 ))|{t=0} ∂t ∂t where y0 . we shall consider. as t → 0. in a measure space of the mapping t → χ(t) at t = 0. G(χ(t)) > ˙ (4.

Now. we get Let yε be the solution to the following Dirichlet problem. yε (x) = 0. Ωε . ε → 0 (ln ε)2 j 1 (ε) − j 1 (0) = 2π ε→0 a1 ε lim This limit is usually referred to as the topological derivative of the functional j 1 (ε).Shape differential equation and level set formulation Hole-differentiability 107 Here we shall illustrate the various dependencies of the topological derivative in the case of the Laplace equation with Dirichlet boundary conditions. x = 1 yε (x) = − ln r + 1. . the choice of the scaling parameter aε depends not only on the type of state equation. ln ε x =r (4.   ∆yε = 0. let us choose a different functional. x = ε.  yε (x) = 1.43) and y0 (x) = 1. but also on the type of functional. we get ε j 2 (ε) − j 2 (0) = 2π ε→0 a2 ε lim Hence. Let us consider the domain Ω = x ∈ R2 together with the perforated subdomain Ωε = {x ∈ Ω | x > ε} | x <1 Obviously. j 2 (ε) = Ωε | yε (x)|2 dx 2π ln ε We find j 2 (ε) − j 2 (0) = − 1 Setting a2 = −( ln ε ). we get ε 2π → 0. Let us first consider the functional j 1 (ε) = x =1 ( ∂ yε (s))2 ds ∂n We get j 1 (ε) − j 1 (0) = 1 Setting a1 = ( ln ε )2 .

.

Despite its lack of rigorous mathematical justification in the case where the Lagrangian functional is not convex. at least formally. automobile shapes. This type of inverse problem happens frequently in the design and the control of many industrial devices such as aircraft wings. On the theoretical level. 46]. [30]. early works concerning optimal control problems for general parabolic equations written in non-cylindrical domains have been considered in [43]. [151].Chapter 5 Dynamical shape control of the Navier-Stokes equations 5. This process leads to the mathematical analysis of non-autonomous PDE’s systems. The aim of this chapter is to review several results on the dynamical shape control of the Navier-Stokes system and suggest an alternative treatment using the Min-Max principle [45. Hence the classical optimal shape optimization theory has been extended to deal with non-cylindrical domains. a new methodology to obtain Eulerian derivatives for non-cylindrical functionals has been introduced in [157]. to bypass the tedious computation of the state differentiability with respect to the shape of the moving domain.1 Introduction This chapter deals with the analysis of an inverse dynamical shape problem involving a fluid inside a moving domain. In [140]. [142]. Recently. [58]. 109 . The basic principle is to define a map sending the non-cylindrical domain into a cylindrical one. we shall show how this principle allows. [152]. satellite reservoir tanks and more generally of systems involving fluid-solid interactions. This methodology was applied in [59] to perform dynamical shape control of the non-cylindrical NavierStokes equations where the evolution of the domain is the control variable. the stabilization of structures using the variation of the domain has been addressed. [156]. [29]. cable-stayed bridges. [2]. The control variable is the shape of the moving domain. and the objective is to minimize a given cost functional that may be chosen by the designer.

Γt = Tt (Γ0 ) We set Σ ≡ ({t}×Γt ). Q(V ) (5. The quantity σ(u. Hence we define a map Tt ∈ C 1 (Ω0 ) such that Ωt = Tt (Ω0 ).1) dx = V (τ. x(τ )).   ∂t u + D u · u − ν∆u + p = 0. Γt = Tt (V )(Γ0 ) where Tt (V ) is solution the of the following dynamical system : Tt (V ) : Ω0 −→ Ω with x0 −→ x(t. T ] dτ x(τ = 0) = x0 . Without loss of generality. x0 ) ≡ Tt (V )(x0 ) (5. Ω0 where ν stands for the kinematic viscosity.1 [147] Ωt = Tt (V )(Ω0 ). as described in the following lemma : THEOREM 5. τ ∈ [0. we choose the reference configuration to be the physical configuration at initial time Ωt=0 . p) satisfies the classical Navier-Stokes equations written in nonconservative form.2) Σ(V )  u = V. The map Tt can be actually 0<t<T 0<t<T defined as the flow of a particular vector field. Its evolution is described by its velocity u and its pressure p. We introduce a diffeomorphic map sending a fixed reference domain Ω0 into the physical configuration Ωt at time t ≥ 0. in Ω0 The fluid filling Ωt is assumed to be a viscous incompressible newtonian fluid.110 Moving Shape Analysis 5. Q ≡ ({t}×Ωt ).   u(t = 0) = u0 . Q(V )   div(u) = 0. The couple (u. p) = −p I +ν(D u + ∗ D u) stands for the fluid stress tensor .2 Problem statement Let us consider a moving domain Ωt ∈ Rd .

v L2 (Σ) (5. v L2 (Σ) (5. p) = α Bu 2 2 Q(V ) + γ KV 2 2 Σ(V ) (5. H∗ ) is a general linear differential operator satisfying the following identity.   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = −α B ∗ B u.10) (5. T ) λ(T ) = 0. π) stands for the adjoint fluid state solution of the following system. ΩT and λ is the adjoint transverse boundary field. the functional j(V ) possesses a gradient j(V ) which is supported on the moving boundary Γ t (V ) and can be represented by the following expression. solution of the tangential dynamical system.8) Σ(V )  ϕ = 0. p) is a real functional of the following form : JV (u.j = ∂j ui . T . MAIN RESULT: For V ∈ U and Ω0 of class C 2 . Q(V )   div(ϕ) = 0.9) . L2 (Σ(V ))) is a general linear differential operator satisfying the following identity.7) where (ϕ. K u. v + u. ΓT (V ) with f = [−(σ(ϕ. K∗ v L2 (Σ) = KΣ u. p(V )) is a weak solution of problem (5. B ∗ v = BΣ u.2) and JV (u. This point will be solved by using the Arbitrary Lagrange-Euler (ALE) map Tt (V ) introduced previously. B u. j(V ) = −λ n − σ(ϕ. We are interested in solving the following minimization problem : V ∈U 111 min j(V ) (5. Q(V ) (5.3) where j(V ) = JV (u(V ). v L2 (Σ) + u.5) 1 where H = v ∈ L2 (0. The purpose of this chapter is to prove using several methods the following result. (0.4) where B ∈ L(H. −∂t λ − Γ λ · V − (div V ) λ = f.Dynamical shape control of the Navier-Stokes equations inside Ωt . π) · n) + α BΣ B u] · (D V · n − D u · n) + 1 α|B u|2 + γ H|K V |2 2 (5. (H0 (div. with (D u)i. p(V )) with (u(V ). π) · n + α BΣ B u + γ [−K∗ K V + KΣ K V ] (5.   ϕ(T ) = 0. Ωt (V )))d ) and K ∈ L(U.6) The main difficulty in dealing with such a minimization problem is related to the fact that integrals over the domain Ωt (V ) depend on the control variable V .

12) α u 2 2 L2 (Q(V )) + γ V 2 2 L2 (Σ(V )) (5.14) (5. B ∗ . B ∗ .15) Example 5. KΣ ) = (I.   ϕ(T ) = 0. KΣ ) = (I. −∂t λ − Γ λ · V − (div V ) λ = f. − I.112 Moving Shape Analysis Example 5. 0) This means that we consider the cost functional. JV (u. p) = Then its gradient is given by j(V ) = −λ n − σ(ϕ. − I. BΣ ) = (I. ∧n) α curl u 2 2 L2 (Q(V )) (K. K∗ . curl. − I.   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = α u. 0) (K. BΣ ) = (curl.13) Σ(V )  ϕ = 0. π) stands for the adjoint fluid state solution of the following system. p) = + γ V 2 2 L2 (Σ(V )) (5. ΓT (V ) with 1 f = −ν(D ϕ · n) · (D V · n − D u · n) + (α + γ H)|V |2 2 where (ϕ. T ) λ(T ) = 0.2 We set (B. 0) JV (u.16) . π) · n + γ V (5.1 We set (B. (0. K∗ . Q(V ) (5. ΩT (5. solution of the tangential dynamical system. Q(V )   div(ϕ) = 0.11) and λ is the adjoint transverse boundary field.

−∂t λ − Γ λ · V − (div V ) λ = f.Dynamical shape control of the Navier-Stokes equations Then its gradient is given by j(V ) = −λ n − σ(ϕ.6).   ϕ(T ) = 0. (0. In section (5. 2. π)(V ) associated to (u . p )(V ) · W . p) with respect to the design variable V . π) stands for the adjoint fluid state solution of the following system. The directional shape derivative (u . The state and multiplier spaces are chosen in order to be independent on the scalar perturbation parameter used in the computation of the derivative of the Lagrangian functional with respect to V . by using a Min-Max formulation of problem (5.5). we introduce several concepts closely related to shape optimization tools for moving domain problems. . In section (5. T ) λ(T ) = 0. solution of the tangential dynamical system. 1.3) and a transport technique. Q(V )   div(ϕ) = 0.18) Σ(V )  ϕ = 0. we choose to bypass the computation of the state shape derivative (u . π) · n + α (curl u) ∧ n + γ V 113 (5. We also recall elements of tangential calculus that will be used through this chapter.17) where (ϕ. Using the adjoint state (ϕ. we choose to prove the differentiability of the fluid state (u. This method directly furnishes the fluid state and transverse field adjoint systems and the resulting gradient j(V ).3).   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = −α ∆u. p )(V ) · W is then used to compute the directional derivative j (V ) · W of the cost functional j(V ). we are able to furnish an expression of the gradient j(V ) which is a distribution supported by the moving boundary Γt (V ). ΓT (V ) with f = [−ν D ϕ · n + α (curl u) ∧ n] · (D V · n − D u · n) 1 + α| curl u|2 + γ H|V |2 2 (5. p )(V ) and the adjoint field Λ associated to the transverse field Zt introduced in section (5. Q(V ) (5. Then we treat successively the following points. ΩT and λ is the adjoint transverse boundary field.19) In the next section.

[59]. we choose to work with domains Ωt that are images of a fixed domain Ω0 through an ALE map Tt (V ) as introduced in the first section. the reader is referred to the next chapter.1 The transverse map Tρt associated to two vector fields (V.1 Non-cylindrical speed method In this paragraph. In section (5. we recall the notion of transverse field.∞ (D))d ) that builds the support.7). W ) ∈ U is defined as follows. This technique has the advantage to transform shape calculus into classical differential calculus on vector spaces [157]. the design parameter is no more the support Ωt but rather the velocity field def V ∈ U = C 0 ([0. (W k.3 Elements of non-cylindrical shape calculus This section introduces several concepts that will be intensively used through this chapter. This means that the state and multiplier variables live in the hold-all domain D. 5. Before stating the main result of this section. In this work. Hence. This again leads to the direct computation of the fluid state and transverse field adjoints and consequently to the gradient j(V ). T ]. Transverse applications DEFINITION 5. we are interested in differentiability properties of integrals defined over moving domains. 5. we again use a Min-Max strategy coupled with a state and multiplier functional space embedding.3. For an other choice based on the non-cylindrical identity perturbation. It concerns the differential calculus of integrals defined on moving domains or boundaries with respect to their support. Hence the derivative of the Lagrangian functional with respect to V only involves terms coming from the flux variation through the moving boundary Γ t (V ).114 Moving Shape Analysis 3. Tρt : Ωt −→ Ωρ = Ωt (V + ρW ) t def x → Tt (V + ρW ) ◦ Tt (V )−1 . J1 (Ωt ) = Ωt f (Ωt ) dΩ g(Γt ) dΓ Γt J2 (Γt ) = The behaviour of J1 and J2 while perturbing their moving support highly depends on the regularity in space and time of the domains.

x) ≡ Ttρ (Zρ )(x) with d x(ρ) = Z t (ρ.) = def ∂Tρt ∂ρ ◦ (Tρt )−1 (5. D × (0. D def (5. we set Zt = Zρ=0 . A fundamental result lies in the fact that Zt can be obtained as the solution of a linear time dynamical system depending on the vector fields (V.20) i. V ] = W. The following result states that the transverse map T ρt can be considered as a dynamical flow with respect to the perturbation variable ρ. . f ρ : [0. DEFINITION 5. ρ ≥ 0 dρ x(ρ = 0) = x. x(ρ)).e.1 The transverse map allows us to perform sensitivity analysis on functions defined on the unperturbed domain Ω t (V ).. we will mainly consider derivatives of perturbed functions at point def t ρ = 0. is the solution of the following dynamical system : t Ttρ (Zρ ) : Ωt −→ Ωρ t t x −→ x(ρ. V ).3 [59] The vector field Zt is the unique solution of the following Cauchy problem. Shape derivative of non-cylindrical functionals The main theorem of this section uses the notion of a non-cylindrical material derivative that we recall here.Dynamical shape control of the Navier-Stokes equations 115 REMARK 5. V ] = DZt · V − DV · Zt stands for the Lie bracket of the pair (Zt .2 [156] t The Transverse map Tρt is the flow of a transverse field Zρ defined as follows t Zρ = Z t (ρ.22) where [Zt .21) Since. THEOREM 5. in Ωt (V ) (5. THEOREM 5.2 The derivative with respect to ρ at point ρ = 0 of the following composed function. ∂t Zt + [Zt . ρ0 ] → H(Ωt (V )) ρ → f (V + ρW ) ◦ Tρt . T ) Zt=0 = 0. W ) ∈ U.

116

Moving Shape Analysis

f˙(V ; W ) is called the non-cylindrical material derivative of f (V ) at point V ∈ U in the direction W ∈ U. We shall use the notation,
def d ρ f˙(V ) · W = f˙(V ; W ) = f dρ
ρ=0

With the above definition, we can state the differentiability properties of non-cylindrical integrals with respect to their moving support, THEOREM 5.4 [59] For a bounded measurable domain Ω0 with boundary Γ0 , let us assume that for any direction W ∈ U the following hypothesis holds, i) f (V ) admits a non-cylindrical material derivative f˙(V ) · W then J1 (.) is Gˆteaux differentiable at point V ∈ U and its derivative is given a by the following expression, J1 (V ) · W = Futhermore, if ii) f (V ) admits a non-cylindrical shape derivative given by the following expression, f (V ) · W = f˙(V ) · W − f (V ) · Zt (5.24) then J1 (V ) · W =
Ωt (V )

Ωt (V )

f˙(V ) · W + f (V ) div Zt dΩ

(5.23)

[f (V ) · W + div(f (V ) Zt )] dΩ

(5.25)

Furthermore, if Ω0 is an open domain with a Lipschitzian boundary Γ0 , then J1 (V ) · W =
Ωt (V )

f (V ) · W dΩ +

f (V ) Zt , n dΓ
Γt (V )

(5.26)

REMARK 5.2 The last identity will be of great interest while trying to prove a gradient structure result for general non-cylindrical functionals.
It is also possible to establish a similar result for integrals over moving boundaries. For that purpose, we need to define the non-cylindrical tangential material derivative, DEFINITION 5.3 The derivative with respect to ρ at point ρ = 0 of the following composed function, g ρ : [0, ρ0 ] → H(Γt (V )) ρ → g(V + ρW ) ◦ Tρt

Dynamical shape control of the Navier-Stokes equations is called the non-cylindrical material derivative of the function g(V ) ∈ H(Γt (V )) in the direction W ∈ U. We shall use the notation, g(V ) · W = g(V ; W ) = ˙ ˙
def

117

d ρ g dρ

ρ=0

This concept is involved in the differentiability property of boundary integrals, THEOREM 5.5 For a bounded measurable domain Ω0 with boundary Γ0 , let us assume that for any direction W ∈ U the following hypothesis holds, i) g(V ) admits a non-cylindrical material derivative g(V ) · W ˙ then J2 (.) is Gˆteaux differentiable at point V ∈ U and its derivative is given a by the following expression, J2 (V ) · W = Futhermore, if ii) g(V ) admits a non-cylindrical shape derivative given by the following expression, g (V ) · W = g(V ) · W − Γ g(V ) · Zt ˙ (5.28) then J2 (V ) · W =
Γt (V )

Γt (V )

[g(V ) · W + g(V ) divΓ Zt ] dΓ ˙

(5.27)

[˜ (V ) · W + H g(V ) Zt , n ] dΓ g

(5.29)

where H stands for the additive curvature (Def. (5.4)). Furthermore, if g(V ) = g (V )|Γt (V ) with g ∈ H(Ωt (V )), then ˜ ˜ J2 (V ) · W =
Γt (V )

[g (V ) · W + ( g (V ) · n + H g(V )) Zt , n ] dΓ ˜

(5.30)

Adjoint transverse field It is possible to define the solution of the adjoint transverse system, THEOREM 5.6 [58] For F ∈ L2 (0, T ; (H 1 (D))d ), there exists a unique field Λ ∈ C 0 ([0, T ]; (L2 (D))d ) solution of the backward dynamical system, −∂t Λ − D Λ · V − ∗ D V · Λ − (div V )Λ = F, (0, T ) Λ(T ) = 0 (5.31)

118

Moving Shape Analysis

REMARK 5.3 The field Λ is the dual variable associated to the transverse field Zt and is the solution of the adjoint problem associated to the transverse dynamical system.
In this chapter, we shall deal with a specific right-hand side F of the form F (t) = ∗ γΓt (V ) (f (t)n). In this case, the adjoint field Λ is supported on the moving boundary Γt (V ) and has the following structure, THEOREM 5.7 [59] For F (t) = ∗ γΓt (V ) (f (t)n), with f ∈ L2 (0, T ; L2 (Γt (V ))), the unique solution Λ of the problem is given by the following identity, Λ = (λ ◦ p) χΩt (V ) ∈ C 0 ([0, T ]; (H 1 (Γt ))d ) (5.32)

where λ ∈ C 0 ([0, T ]; H 1 (Γt )) is the unique solution of the following boundary dynamical system, −∂t λ − Γ λ · V − (div V )λ = f, (0, T ) λ(T ) = 0, Γt (V ) (5.33)

p is the canonical projection on Γt (V ) and χΩt (V ) is the characteristic function of Ωt (V ) inside D. Gradient of non-cylindrical functionals In the next sections, we will often deal with boundary integrals of the following forms,
T

K=
0 Γt (V )

E Zt , n

with E ∈ L2 (0, T ; Γt (V )) and Zt is the solution of the transverse equation (5.22). The following result allows us to eliminate the auxiliary variable Z t inside the functional K, THEOREM 5.8 [59] For any E ∈ L2 (0, T ; Γt (V )) and (V, W ) ∈ U, the following identity holds,
T 0 Γt (V ) T

E Zt , n = −

λ W, n
0 Γt (V )

(5.34)

where λ ∈ C 0 ([0, T ]; H 1 (Γt )) is the unique solution of problem (5.33) with f = E.

Dynamical shape control of the Navier-Stokes equations

119

5.4

Elements of tangential calculus

In this section, we review basic elements of differential calculus on a C k submanifold with k ≥ 2 of codimension one in Rd . The following approach avoids the use of local bases and coordinates by using the intrinsic tangential derivative.

5.4.1

Oriented distance function

Let Ω be an open domain of class C k in Rd with compact boundary Γ. We define the oriented distance function to be as follows, bΩ (x) =
where dΓ (x) = min |y − x|.
y∈Ω

dΓ (x), x ∈ Rd \ Ω −dΓ (x), x ∈ Ω

PROPOSITION 5.1 [51] Let Ω be an open domain of class C k for k ≥ 2 in Rd with compact boundary Γ. There exists a neighbourhood U (Γ) of Γ, such that b ∈ C k (U (Γ)). Furthermore, we have the following properties, i) b|Γ = n, where n stands for the unit exterior normal on Γ,

ii) D2 b : Tp(x) Γ → Tp(x) Γ coincides with the second fundamental form on Γ, where p : U (Γ) → Γ x → x − b(x) ·

b(x)

stands for the projection mapping and Tp(x) Γ stands for the tangent plane. iii) (0, β1 , . . . , βd−1 ) are the eigenvalues of D2 b associated to the eigenfunctions (n, µ1 , . . . , µd−1 ) where (βi , µi )1≤i≤d−1 are the mean curvatures and principal direction of curvatures of Γ. PROPOSITION 5.2 [51] For Γ of class C 2 , the projection mapping p is differentiable and its derivative

120 has the following properties,

Moving Shape Analysis

D p = D p = I − b · ∗ b − b D2 b D p · τ = τ, on Γ, D p · n = 0, on Γ

(5.35)

DEFINITION 5.4 [51] For Γ of class C 2 , the additive curvature H of Γ is defined as the trace of the second order fundamental form : ¯ H = Tr D2 b = ∆b = (d − 1)H, ¯ and H stands for the mean curvature of Γ. on Γ (5.36)

5.4.2

Intrinsic tangential calculus

Using arbitrary smooth extensions of functions defined on Γ to Ω ∈ R d is the most classical way of defining tangential operators. Hence the differential calculus on manifolds can be reduced to classical differential calculus in R d . In this section we recall standard formulas for differential tangential operators using arbitrary extensions. We also emphasize the particular case where the extension is of the canonical type (f ◦ p). This is the basis of a simple differential calculus in the neighbourhood of Γ. DEFINITION 5.5 For Γ of class C 2 , given any extension F ∈ C 1 (U (Γ)) of f ∈ C 1 (Γ), the tangential gradient of f is defined as
Γ

f =

def

F |Γ − (∂n F ) n

(5.37)

where ∂n F =

F · n.

PROPOSITION 5.3 [51] Assume that Γ of class C 2 is compact and f ∈ C 1 (Γ), then i) f = (P F )|Γ b· Γf =0 Γf =
Γ


def

(5.38)

where P = I − b∗ b is the orthogonal projection operator onto the tangent plane Tp(x) Γ. ii) (f ◦ p) = I −b D2 b Γ f ◦ p (f ◦ p)|Γ = Γ f

(5.39)

Dynamical shape control of the Navier-Stokes equations

121

Hence (f ◦ p) plays the role of a canonical extension in the neighbourhood U (Γ) and its gradient is tangent to the level sets of b. Consequently, we can define in an intrinsic way the tangential gradient, DEFINITION 5.6 For Γ of class C 2 and f ∈ C 1 (Γ), the tangential gradient of f is defined as
Γ

f=

(f ◦ p)|Γ

(5.40)

In the sequel, we shall use the above definition for the tangential gradient whenever the function under derivation is intrinsically defined on Γ. We now define the other classical tangential operators, DEFINITION 5.7 For Γ of class C 2 ,

i) for v ∈ (C 1 (Γ))d , and v ∈ (C 1 (U (Γ)))d an arbitrary extension, the tan˜ gential jacobian is defined as follows, DΓ v = D v |Γ − (D v · n)∗ n ˜ ˜ = D v |Γ − D v · (n ⊗ n) ˜ ˜ Furthermore, D(v ◦ p) = DΓ v ◦ p I −b D2 b
def

(5.41)

DΓ v = D(v ◦ p)|Γ

(5.42)

ii) for v ∈ (C 1 (Γ))d , and v ∈ (C 1 (U (Γ)))d an arbitrary extension, the tan˜ gential divergence is defined as follows, divΓ v = div v |Γ − (D v · n) · n ˜ ˜ Furthermore, divΓ v = div(v ◦ p)|Γ = Tr(DΓ v) (5.44)
def

(5.43)

iii) for f ∈ C 2 (Γ), and F ∈ C 2 (U (Γ)) an arbitrary extension, the tangential Laplace-Beltrami operator is defined as follows,
2 ∆Γ f = ∆F |Γ − H∂n F − ∂n F 2 with ∂n F = (D2 F · n) · n. Furthermore,

(5.45)

∆Γ f = divΓ (

Γ

f ) = ∆(f ◦ p)|Γ

(5.46)

4 For v ∈ (C 1 (Γ))d .2) and recall the associated classical solvability result.48) (5. n Rd (5.2)) with respect to the velocity field V .5 Let Γ be a C 2 -submanifold in Rd . it may be interesting to use a splitting of the function v onto a normal and a tangential component. we obtain the following identities. For the sake of simplicity. DEFINITION 5. using the weak implicit function theorem. Then.5 State derivative strategy In this section.47) vn (5. PROPOSITION 5.8 For v ∈ (C 1 (Γ))d . we introduce a weak formulation for Eq. we will prove the existence of a weak material derivative. First.1)-1 with free divergence control velocity fields.122 Moving Shape Analysis In some cases. we have DΓ v = D Γ v Γ + v n D 2 b + n · ∗ Γ Γ (5. PROPOSITION 5.4.49) (5. we will use the following tangential Stokes identity. we define the tangential component vΓ ∈ (C 1 (Γ))d and the normal component vn ∈ C 1 (Γ) such that v = v Γ + vn n Using the above definition.51) 5. we shall prove the main theorem of this chapter using an approach based on the differentiability of the solution of the Navier-Stokes system (Eq.50) divΓ v = divΓ vΓ + H vn v n = ∗ DΓ v · n + D 2 b · v Γ 5. . for E ∈ H 1 (Γ. Rd ) and ψ ∈ H 1 (Γ. R) the following identity holds E.3 Tangential Stokes formula In order to perform integration by parts on Γ. Finally. we shall only prove the main theorem in the case of example (5. Γ Γ ψ Rd + Γ (divΓ E)ψ = Γ H ψ E. (5. (5. we will be able to express the gradient of the functional j(V ). introducing the adjoint equations associated to the linearized fluid and transverse systems.

55) Ω0 (u(0) − u0 ) · v(0) ˜ div V = 0 in D. e2 (u)]. u Q(V ) (5.53) [(∂t u + D u · u + D u · V + D V · u) · v + ν D u · · D v] T (5. T . For V ∈ U and u0 ∈ H(D) such that u0 |Ω0 ∈ H(Ω0 ). Ωt (V )))d ) 1 ∂t v ∈ L2 (0. v = [0. 1 H = v ∈ L (0.5. . (H0 (Ωt (V )))d ) In the sequel. 2 1 H0 (div. div v = 0. V · n = 0 on ∂D THEOREM 5.54) − e2 (u). T .   u(0) = u0 − V (0). DEFINITION 5. v = V T 0 Ωt (V ) ∀v ∈ H (5. (H0 (div. v = [e1 (u). v = V We set U = V ∈ H 1 (0. = v∈ 1 (H0 (D))d .52) ˜ Σ(V )  u = 0. 0 Ωt (V ) F (V ) · v (5. in D. [139]. D) In order to take into account the non-homogeneous Dirichlet boundary condition on Γt (V ). V V with e1 (u). with m > 5/2. 0].Dynamical shape control of the Navier-Stokes equations 123 5.52).1 Weak formulation and solvability with F (V ) = −∂t V − D V · V + ν∆V . ˜ Ω0 v · n = 0 on ∂D div v = 0. We consider the following classical functional spaces [97]. (H m (D))d ).9 [58] We assume the domain Ω0 to be of class C 1 . eV (u). keeping in mind that ˜ the original variable is obtained by translation.9 The function u ∈ V is called a weak solution of problem (5. Q(V )  ˜  div(˜) = 0. where u ˜ ˜ satisfies the following homogeneous Dirichlet Navier-Stokes system.  ˜ ˜ ˜ ˜ u  ∂t u + D u · u + D u · V + D V · u − ν∆˜ + p = F (V ). if it satisfies the following identity. in D V = v ∈ H. T . we use the following change of variable u = u − V . we shall use the notation u instead of u. H(D) = v ∈ (L2 (D))d .

if u0 ∈ (H 2 (D))d ∪ H0 (div. W ) = lim ˙ ρ→0 1 (u(V + ρW ) ◦ Tt (V + ρW ) − u(V ) ◦ Tt (V )) ρ • Application of the weak implicit function theorem and deduction of the local differentiability of the solution u(V ) associated to the implicit equation eV (u). we need to analyse the derivability of the state u(V ) with respect to V ∈ U. LEMMA 5. There exist at least two methods in order to establish such a differentiability result: • Limit analysis of the differential quotient. We recall here how the weak implicit function theorem can be applied following the result obtained in [59]. Ωt )). u ∈ L∞ (0. u(V . if Ω is of class C 2 .124 Moving Shape Analysis 1. (H 2 (Ωt ))d ∪ H0 (div. . T .2 The weak Piola material derivative We are interested in solving the following minimization problem : V ∈U min j(V ) (5. v = 0. T . 1 1 PI : H0 (div. ∀ v ∈ H. H). and we have ∂t u ∈ H ∪ L∞ (0. it exists at least a weak solution of problem (5.52) with u ∈ H ∪ L∞ (0. 1 2. then the uniqueness of a weak solution is guaranteed. Ωt (V )) −→ H0 (div.56) with j(V ) = α 2 0 T Ωt (V ) |u(V )|2 + γ 2 T 0 Ωt (V ) |V |2 (5.1 [13] The Piola transform. 5. D) and ν is large or u0 is a small data. 1 3.5.56). In order to work with divergence free functions. H(Ωt )). we need to introduce the Piola transform that preserves the free divergence condition. Ωρ ) t v → (D Tρt · v) ◦ (Tρt )−1 is an isomorphism.57) In order to derive first-order optimality conditions for problem (5. T .

v = 0.. Ωt (V + ρ W )))d ) if and only if uρ satisfies the following identity. (H0 (div. ˆ eρ (ˆρ ). v = 0. w = 1 T 0 Ωt (V ) (∂t (D Tρt · v)) · (D Tρt · w) −(D(D Tρt · v) · (D Tρt )−1 · (∂t Tρt )) · (D Tρt · w) + (D(D Tρt · v) · v) · (D Tρt · w) +(D(D Tρt · v) · (D Tρt )−1 · ((V + ρW ) ◦ Tρt )) · (D Tρt · w) +(D((V + ρW ) ◦ Tρt ) · v) · (D Tρt · w) +ν(D(D Tρt · v) · (D Tρt )−1 ) · ·(D(D Tρt · w) · (D Tρt )−1 ) −(F (V + ρW ) ◦ Tρt ) · (D Tρt · w) eρ (v). ˆ uρ = PI (ˆρ ).e. u LEMMA 5. v = 0. u ˆ with ∀v ∈ H ˆ eρ (v). T . ∀ v ∈ Hρ 1 Hρ = v ∈ L2 (0. ∀ v ∈ Hρ with We introduce the element uρ = (D Tρt )−1 · (uρ ◦ Tρt ) defined on Ωt (V ) (i.Dynamical shape control of the Navier-Stokes equations 125 We consider the solution uρ = u(V + ρW ) defined on Ωρ of the following t implicit equation. e(V +ρW ) (u). e(V +ρW ) (u).2 The element uρ satisfies the following identity. w = 2 with F (V ) = −∂t V − D V · V + ν∆V (v(0) − u0 ) · w ˆ Ω .

This leads to the following identity. We introduce the variables (ˆ. eρ (ˆ).60) ∂t (D Tρt ◦ u)−D(D Tρt ◦ u)·(D Tρt )−1 ·∂t Tρt ˆ ˆ Using the identity. v] = (D Tρt · u) ◦ (Tρt )−1 . with T e(V +ρW ) (u). (Tρt )−1 ◦ Tρt = I we get D((Tρt )−1 ◦ Tρt ) = I D((Tρt )−1 ) ◦ Tρt · D Tρt = I .3 ∂t ((D Tρt · u)◦(Tρt )−1 )◦Tρt ˆ PROOF ∂t ((Tρt )−1 ) ◦ Tρt = −(D Tρt )−1 · ∂t Tρt = D((Tρt )−1 ) ◦ Tρt = (D Tρt )−1 (5.59) (5. v = 0 Ωρ t ∂t u + D u · u + D u · (V + ρW ) + D(V + ρW ) · u · v + ν D u · · D v − F (V + ρW ) · v with v ∈ Hρ . (D Tρt · v ) ◦ (Tρt )−1 ˆ ˆ We replace this new representation inside the state equation and we use a back transport in Ωt (V ).126 and Moving Shape Analysis ∂t Tρt = (V + ρW ) ◦ Tρt − D Tρt · V PROOF We consider the solution uρ of the perturbed state equation e(V +ρW ) = 0. v ) defined in Ω t (V ) such that u ˆ [u.58) (5. v = 1 u ˆ ∂t ((D Tρt · u) ◦ (Tρt )−1 ) ˆ + D((D Tρt · u) ◦ (Tρt )−1 ) · (V + ρW ) ˆ Q(V ) + D((D Tρt · u) ◦ (Tρt )−1 ) · (D Tρt · u) ◦ (Tρt )−1 ˆ ˆ + D(V + ρW ) · (D Tρt · u) ◦ (Tρt )−1 ◦ Tρt · (D Tρt · v ) ˆ ˆ +ν D((D Tρt · u) ◦ (Tρt )−1 ) ◦ Tρt · ·(D((D Tρt · v ) ◦ (Tρt )−1 )) ◦ Tρt ˆ ˆ −F (V + ρW ) ◦ Tρt · (D Tρt · v ) ˆ LEMMA 5.

using all the ˆ identities proven above.4 D(φ ◦ (Tρt )−1 ) ◦ Tρt = D(φ) · (D Tρt )−1 (5. we also get ∂t ((Tρt )−1 ◦ Tρt ) = 0 127 ∂t ((Tρt )−1 ) ◦ Tρt + D((Tρt )−1 ) ◦ Tρt · ∂t Tρt = 0 Using the chain rule. LEMMA 5. we deduce the expression of e 1 (ˆ.Dynamical shape control of the Navier-Stokes equations By differentiation with respect to time t. we need also the following identities. Now.62) D(V + ρW ) ◦ Tρt · (D Tρt ) = D((V + ρW ) ◦ Tρt ) We shall apply the first identity with φ = (D Tρt · u).63) ∂t (Tρt ) = ∂t (Tt (V + ρW ) ◦ Tt (V )−1 ) = ∂t (Tt (V + ρW )) ◦ Tt (V )−1 + D(Tt (V + ρW )) ◦ Tt (V )−1 · ∂t (Tt (V )−1 ) = ((V + ρW ) ◦ Tt (V + ρW )) ◦ Tt (V )−1 = (V + ρW ) ◦ Tρt − D(Tρt ) · V − D(Tt (V + ρW )) ◦ Tt (V )−1 · (D Tt−1 (V )) · ∂t (Tt (V )) ◦ Tt−1 (V ) = (V + ρW ) ◦ Tρt − D(Tt (V + ρW ) ◦ Tt (V )−1 ) · ∂t (Tt (V )) ◦ Tt−1 (V ) . we simply ρ u ˆ need to prove the following lemma in order to conclude the proof. we deduce [∂t ((D Tρt · u) ◦ (Tρt )−1 )] ◦ Tρt = [∂t (D Tρt · u) ◦ (Tρt )−1 ˆ ˆ ∂t ((Tρt )−1 ) ◦ Tρt + (D Tρt )−1 · ∂t Tρt = 0 = ∂t (D Tρt · u) + D(D Tρt · u) · ∂t ((Tρt )−1 ) ◦ Tρt ˆ ˆ = ∂t (D Tρt · u) − D(D Tρt · u) · (D Tρt )−1 · ∂t Tρt ˆ ˆ + D(D Tρt · u) ◦ (Tρt )−1 · ∂t ((Tρt )−1 )] ◦ Tρt ˆ In order to get the correct state operator.61) (5.5 ∂t Tρt = (V + ρW ) ◦ Tρt − D Tρt · V PROOF We use the definition of the Transverse map. Finally. v ). LEMMA 5. (5.

Y ∗ be two Banach spaces. and consider the following mapping. ˆ eρ (v).67) (5.1 (I. x) → ∂ρ e(ρ. THEOREM 5. I an open bounded set in R. 1 [0.65) ∀w ∈ H (5. ∀ ρ ∈ I . with s ≥ 1.10 [147] Let X. e:I ×X → Y∗ Let us assume the following hypothesis. 1 ρ→0 (F ◦ Tρt − F ) −→ ρ F · Zt (5. y is continuous. Ω0 ) ρ (ρ. recalled below. u(ρ)) = 0. X) e(ρ. In order to prove the differentiability of uρ with respect to ρ in a neighbourˆ hood of ρ = 0. a) (i) the application ρ → e(ρ. x). we cannot use the classical implicit function theorem. x). Then we shall use the weak implicit function theorem. the convergence only holds weakly in H s−1 (D). ρ0 ] × V → H∗ × H0 (div.6 [135] For any F ∈ H s (D). w = 0. x) → e(ρ.66) strongly in H s−1 (D). v) → e (v) (5. ρ0 ] → H ρ → uρ = (D Tρt )−1 · (uρ ◦ Tρt ) ˆ where uρ ∈ V is the solution of the state equation.128 Moving Shape Analysis We now consider the application. since it requires strong differentiability results in H −1 for our application. y is continuously differentiable for any y ∈ Y .64) and [0. (ρ. In the case s < 1. b) It exists u ∈ X such that u ∈ C 0. x) (ii) the application (ρ. LEMMA 5.

68) In order to apply the above theorem to Eq. ρ0 ] → R ρ → eρ (v). we need to state the following properties. LEMMA 5. (5. w = 1 t t t (∂t (D(Zρ · Tρt ) · v)) + (D(D(Zρ · Tρt ) · v) · V + (D(D(Zρ · Tρt ) · v) · v) Q(V ) t + D (D(V + ρW ) · Zρ ) ◦ Tρt + W ◦ Tρt · v − ∂ρ (F (V + ρW ) ◦ Tρt ) ·(D Tρt ·w) + (∂t (D Tρt · v)) + (D(D Tρt · v) · V + (D(D Tρt · v) · v) t +(D((V + ρW ) ◦ Tρt ) · v) − (F (V + ρW ) ◦ Tρt · (D(Zρ · Tρt ) · w) t + ν(D(D(Zρ ◦ Tρt ) · v) · (D Tρt )−1 ) · ·(D(D Tρt · w) · (D Tρt )−1 ) t − ν(D(D Tρt · v) · (D Tρt )−1 ) · D(Zρ ◦ Tρt ) · (D Tρt )−1 ) · ·(D(D Tρt · w) · (D Tρt )−1 ) t + ν(D(D Tρt · v) · (D Tρt )−1 ) · ·(D(D(Zρ ◦ Tρt ) · w) · (D Tρt )−1 ) t − ν(D(D Tρt · v) · (D Tρt )−1 ) · ·(D(D Tρt · w) · (D Tρt )−1 ) · D(Zρ ◦ Tρt ) · (D Tρt )−1 ) .Dynamical shape control of the Navier-Stokes equations c) x → e(ρ. ∂ρ eρ (v). u(ρ0 )). ˙ ∀y ∈ Y (5. w (5. 129 Then the mapping d) It exists ρ0 ∈ I such that ∂x e(ρ. x) is continuous.x(ρ0 )) ∈ ISOM(X. x) is differentiable and (ρ. u(. u(ρ0 )) · u(ρ0 ). Y ∗ ). y + ∂ρ e(ρ0 . x) → ∂x e(ρ.7 The mapping. y = 0.69) is C 1 for any (v.66).) : I → X ρ → u(ρ) is differentiable at point ρ = ρ0 for the weak topology in X and its weak derivative u(ρ) is the solution of the following linearized equation. [0. w) ∈ V ×H and its derivative is given by the following expression. ˙ ∂x e(ρ0 . x)|(ρ0 .

t ∂ρ Tρt = Zρ ◦ Tρt t ∂ρ (D Tρt )−1 = −(D Tρt )−1 · D(Zρ ◦ Tρt ) · (D Tρt )−1 LEMMA 5. ρ0 [. R3 )). w = 1 T 0 Ωt (V ) (∂t (D Tρt · v)) + (D(D Tρt · v) · V + (D(D Tρt · v) · v) + ν(D(D Tρt · v) · (D Tρt )−1 ) · ·(D(D Tρt · w) · (D Tρt )−1 ) +(D((V + ρW ) ◦ Tρt ) · v) − (F (V + ρW ) ◦ Tρt · (D Tρt · w) We use the expression of the weak state operator and the following identities. (5. v) → ∂ρ eρ (v) is weakly continuous.130 PROOF that Moving Shape Analysis We first simplify the expression of the weak state operator. the associated flow T ρt ∈ C 1 ([0. W ) ∈ V. ρ0 ] × V → H∗ (ρ.8 The mapping. using ∂t Tρt = (V + ρW ) ◦ Tρt − D Tρt · V and we get eρ (v). [0. and the weak continuity follows easily. C 2 (D. In order to apply the implicit function derivative identity. PROOF We can prove that for (V.70) . we need to express the derivative ∂ρ eρ (v) at point ρ = 0.

[0. H∗ ) (ρ. ∂v eρ (v) · δv. V → H∗ v → eρ (v) (5.9 ∂ρ eρ |ρ=0 (v). w and we use the following identities. ρ0 ] and its derivative is given by the following expression.10 The mapping. v) → ∂v eρ (v) (5. t Tρ=0 = I t Zρ |ρ=0 = Zt def LEMMA 5.71) is differentiable for any ρ ∈ [0. w = 1 (∂t (D Tρt · δv)) + D(D Tρt · δv) · V + D(D Tρt · δv) · v + D(D Tρt · v) · δv Q(V ) +(D((V + ρW ) ◦ Tρt ) · δv) +ν(D(D Tρt ·δv)·(D Tρt )−1 )··(D(D Tρt ·w)·(D Tρt )−1 ) and the mapping. w = 1 131 Q [∂t (D Zt · v) + D(D Zt · v) · V + D(D Zt · v) · v + D [D V · Zt + W ] · v] · w + [∂t v + D v · V + D v · v + D V · v] · (D Zt · w) + ν D(D Zt · v) · · D w − ν(D v · D Zt ) · · D w + ν D v · · D(D Zt · w) − ν D v · ·(D w · D Zt ) + [∂t W + D W · V + D V · W − ν∆W ] · w + (D [∂t V + D V · V − ν∆V ] · Zt ) · w + [∂t V + D V · V − ν∆V ] · (D Zt · w) PROOF We set ρ = 0 in the expression of ∂ρ eρ (v). ρ0 ] × V → L(V.Dynamical shape control of the Navier-Stokes equations LEMMA 5.72) .

12 The solution uρ ∈ V of the implicit equation. Hence the hypothesis of Th.73) is an isomorphism and its expression is furnished by the following identity. Similar a priori estimates hold for δv and the unique solvability of the linearized system is established. Moving Shape Analysis V→F δv → ∂v eρ=0 (v) · δv (5. PROOF We need the identity satisfied by uρ1 − uρ2 and we shall follow ˆ ˆ the same steps described in [56] (pp. ˆ eρ (v). for u1 and u2 solutions of the Navier-Stokes equations.9) and [139]).132 is continuous.66) and we can state the following differentiability result. it is proven that the element y = u1 − u2 satisfying the following identity. Q(V ) [(∂t y) · w + (D y · u1 ) · w + (D u2 · y) · w + (D y · V ) · w + (D V · y) · w +ν D y · · D w] = 0. (5. w = 1 Q(V ) [(∂t δv) · w + (D δv · v) · w + (D v · δv) · w + (D δv · V ) · w +(D V · δv) · w + ν D δv · · D w] PROOF This result follows from the uniqueness result for the NavierStokes system under regularity and smallness assumptions (see Th. LEMMA 5. ∀w ∈ H exists and is identically equal to the null function. 31).10) is satisfied by the Eq. Indeed. ∀w ∈ H (5.11 The mapping. LEMMA 5. (5. ∂v eρ=0 (v) · δv.74) . (5. w = 0. is Lipschitz with respect to ρ.

v = V u [(∂t u + D u · u + D u · V + D V · u) · v ˜ ˜ ˜ ˜ ˜ +2νε(˜) · ·ε(v) − F (V ) · v] u ∀v ∈ H (5. (∂t uP ) · w + (D uP · u) · w + (D u · uP ) · w + (D uP · V ) · w ˙ ˙ ˙ ˙ +(D V · uP ) · w + ν D uP · · D w = L(u.75) Q(V ) Q − [∂t u + D u · V + D u · u + D V · u] · (D Zt · w) − ν D(D Zt · u) · · D w + ν(D u · D Zt ) · · D w − ν D u · · D(D Zt · w) + ν D u · ·(D w · D Zt ) + [−∂t W − D W · V − D V · W + ν∆W ] · w − (D [∂t V + D V · V − ν∆V ] · Zt ) · w − [∂t V + D V · V − ν∆V ] · (D Zt · w) 5. w + ∂ρ eρ (ˆ)|ρ=0 .5. i. We have also characterized the linearized system satisfied by the Piola material derivative uP (V ) · W . e2 (u)]. ∂v eρ=0 (v)|v=ˆ · u˙P . w = 0. W ).. Zt .Dynamical shape control of the Navier-Stokes equations 133 THEOREM 5.76) Q(V ) . w = − [∂t (D Zt · u) + D(D Zt · u) · V + D(D Zt · u) · u + D [D V · Zt + W ] · u]·w ∀w ∈ H (5. we will identify the shape derivative ˙ u (V ) · W under some regularity assumptions. Let us consider the weak solution u of Eq.e. w ˙ ˙ with L(u. v = [0. Zt .52). v = [e1 (u). In this paragraph. V. 0]. u u which possesses the following structure. u V V with e1 (˜). we have proven that the solution u(V ) of the moving Navier-Stokes system is differentiable with respect to the velocity V . (5.11 The Piola material derivative uP = ∂ρ (ˆρ )|ρ=0 exists and is characterized by ˙ u the linear tangent equation. W ). V.3 Shape derivative In the last section. ˜ eV (˜).

u Q ˜ ˜ Σ  u = −(D u · n) Zt .4) and we get G(V )dx Ωt (V ) ·W = Ωt (V ) G (V ) · W dx + G Zt . v = V u (˜(0) − u0 ) · v(0) u (5.13 −ν Ωt ∆u · v = 2ν Ωt ε(u) · ·ε(v) − 2ν Γt ε(u) · n. (5. ˜ Ω0 (5. This definition is motivated by the following lemma. W ).12 ˙ For Ω0 of class C 2 .22)) with G = [(∂t u + D u · u + D u · V + D V · u) · v + νε(˜) · ·ε(v) − F (V ) · v] ˜ ˜ ˜ ˜ ˜ u We assume that v has a compact support. then G|Γt (V ) = 0. n . we use Th. n Γt (V ) (5. Q  ˜  div(˜ ) = 0.79) with L(V. ∀ v ∈ H 1 (div.80) ˜ ˜ PROOF d dV In order to state such a result. W ) = −∂t W − D W · V − D V · W + ν∆W − D u · W − D W · u (5.134 Moving Shape Analysis e2 (˜). (5. the shape derivative u = u − D u · Zt exists and is char˜ ˜ ˜ acterized as the solution of the following linearized system.78) THEOREM 5.81) where G (V ) · W stands for non-cylindrical shape derivative of G and Z t is the transverse vector field solution of the Transverse Equation (Eq. Ωt ) (5. v .14 G (V ) · W = [(∂t u + D u · u + D u · u + D u · V + D u · W ˜ ˜ ˜ ˜ ˜ ˜ ˜ + D W · u + D V · u ) · v + νε(˜ ) · ·ε(v) − F (V ) · W · v] ˜ ˜ u with F (V ) · W = −∂t W − D W · V − D V · W + ν∆W . LEMMA 5.77) Ω0 1 where ε(v) = 2 (D v + ∗ D v) stands for the symmetrical deviation tensor.   u (0) = 0. LEMMA 5.  ˜ ˜ ˜ ˜ ˜ ˜ u  ∂t u + D u · u + D u · u + D u · V + D V · u − ν∆˜ + p = L(V.

The boundary condition comes from the fact that the shape derivative of the condition u = 0 on Γt (V ) is given by ˜ u = − D u · Zt . we have D u|Γt = D u · (n ⊗ n) which gives u = −(D u · n) Zt . (5. ˜ ˜ REMARK 5.   ∂t u + D u · u + D u · u − ν∆u + p = 0.79). n . (5. u = W − (D u · n) Zt . on Γt (V ) ˜ The shape derivative u (V )·W of the solution u of the original non-homogeneous Dirichlet boundary problem (Eq. νε(˜ ) · ·ε(v).4 If we choose V = (V ◦ p) the canonical extension of V in Eq. Q (5.Dynamical shape control of the Navier-Stokes equations Finally we obtain d e1 (˜) · W. (5.2)) is given by the expression u (V ) · W = u (V ) · W + W ˜ (5. u = u − W and u = u − V .82) COROLLARY 5.83). on Γt (V ) (5. n . n . on Γt (V ) ˜ Since u = 0 on Γt (V ).84) . Ω0 PROOF We simply set in Eq. then we get the simpler boundary condition.79)) satisfied by the shape derivative u (V ) · W . we recover the u correct strong formulation of the linearized equation (Eq. ·v = u dV [(∂t u + D u · u + D u · u + D u · V + D u · W ˜ ˜ ˜ ˜ ˜ ˜ ˜ 135 Q(V ) + D W · u + D V · u ) · v + νε(˜ ) · ·ε(v) − F (V ) · W · v] ˜ ˜ u Using integration by parts for the term Q(V ) for any v ∈ H with compact support.   u (0) = 0.83) Σ  u = W + (D V · n − D u · n) Zt .1 The shape derivative u (V )·W of the solution u of Eq. Q   div(u ) = 0.2) exists and satisfies the following linearized problem. (5. (5.

˜ (5. Then the following identity holds. (5.6 We consider Ω0 of class C 2 .85) (5. This linearized system has been obtained independently of the system satisfied by the non-cylindrical ˙ material derivative uP (V ) · W . (5. Zt ] − [V.4 Extractor Identity In the last section.2)) in moving domain. (5. Y ] = D X · Y − D Y · X. there exists an explicit relation ˜ ˙ between the original shape u and the Piola material derivative uP (V ) · W of ˜ the shift state u = u − V .22). ˜ ˙ uP (V ) · W = u (V ) · W + [˜(V ). ∀w ∈ H . W ) ∈ U. ˜ LEMMA 5.136 Moving Shape Analysis 5. we have established the structure of the system satisfied by the non-cylindrical shape derivative u (V ) · W of the solution u(V ) of the Navier-Stokes problem in the moving domain Ωt (V ). u (V )·W stands for its shape deriva˙ tive and uP (V ) · W stands for the Piola material derivative of the shift flow ˜ u = u(V ) − V in the direction W .5. { ∂t (D u · Zt ) + D(D u · Zt ) · u + D u · (D u · Zt ) + D(D u · Zt ) · V ˜ ˜ ˜ ˜ ˜ ˜ Q(V ) + D V · (D u · Zt ) · w − D u · (D Zt · u) + D V · (D Zt · u) − D(D V · Zt ) · u · w ˜ ˜ ˜ ˜ ˜ + ∂t u + D u · V + D u · u + D V · u · (D Zt · w) − D u · W · w ˜ ˜ ˜ ˜ ˜ ˜ +ν D(D u·Zt )·· D w−ν(D u·D Zt )·· D w+ν D u·· D(D Zt ·w)−ν D u··(D w·D Zt ) ˜ ˜ ˜ ˜ + (D ∂t V + D V · V − ν∆V · Zt ) · w + ∂t V + D V · V − ν∆V · (D Zt · w)} = 0. the following identity holds for all (V. u solution of the homogeneous Navier-Stokes ˜ equations Eq.15 Let u(V ) stand for the weak solution of the non-homogeneous Navier-Stokes equations (Eq. Zt ] ˜ ˜ u where [X.52) and Zt solution of Eq.86) = u (V ) · W + [u(V ). However. Zt ] − W PROPOSITION 5. This relation can be fruitful in order to obtain an identity concerning the solution u(V ) inside Ωt (V ).

w 2. This leads to the ˜ ˜ u ˜ ˜ ˜ following identity.Dynamical shape control of the Navier-Stokes equations PROOF identity.11). w . (∂t uP ) · w + (D uP · u) · w + (D u · uP ) · w + (D uP · V ) · w ˙ ˙ ˙ ˙ +(D V · uP ) · w + ν D uP · · D w = ˙ ˙ with 2. Zt ] = uP − D u · Zt + D Zt · u. 137 We recall that the shape derivative u satisfies the following ˜ Q(V ) [∂t u + D u · u + D u · u + D u · V + D V · u ]·w+ν D u ·· D w = ˜ ˜ ˜ ˜ ˜ ˜ ˜ ˜ 1. we deduce that = L. ∀w ∈ H (5. w = Q(V ) [−∂t W − D W · V − D V · W + ν∆W − D u · W − D W · u]·w ˜ ˜ ˙ ˙ Then we set u (V ) = uP − [˜.87) with L. w with 1. w + D V · (D u · Zt ) · w − ∂t (D Zt · u) + D(D Zt · u) · u + D u · (D Zt · u) ˜ ˜ ˜ ˜ ˜ ˜ Using Theorem (5. w = − Q(V ) ∂t (D Zt · u) + D(D Zt · u) · u + D(D Zt · u) · V + D(D V · Zt ) · u ˜ ˜ ˜ ˜ ˜ + D W · u · w − [∂t u + D u · V + D u · u + D V · u] · (D Zt · w) ˜ ˜ ˜ ˜ ˜ ˜ − ν D(D Zt · u) · · D w + ν(D u · D Zt ) · · D w − ν D u · · D(D Zt · w) ˜ ˜ ˜ + ν D u · ·(D w · D Zt ) + [−∂t W − D W · V − D V · W + ν∆W ] · w ˜ − (D [∂t V + D V · V − ν∆V ] · Zt ) · w − [∂t V + D V · V − ν∆V ] · (D Zt · w) . w Q(V ) = ∂t (D u · Zt ) + D(D u · Zt ) · u + D u · (D u · Zt ) + D(D u · Zt ) · V ˜ ˜ ˜ ˜ ˜ ˜ Q(V ) +D(D Zt · u)·V +D V ·(D Zt · u) ·w + ν D(D u ·Zt )·· D w −ν D(D Zt · u)·· D w ˜ ˜ ˜ ˜ + − ∂t W − D W · V − D V · W + ν∆W − D u · W − D W · u · w ˜ ˜ 2.

we can obtain an identity only ˜ involving the 4-uplet (u.5. Let us first state a differentiability property.89) . n 2 where u (V ) · W is the solution of the shape derivative system (Eq. V.88) + Σ(V ) 1 γ V · W + (α + γ H)|V |2 Zt .138 The sequence Moving Shape Analysis [−∂t (D Zt · u) − D(D Zt · u) · u − D(D Zt · u) · V − D W · u] · w ˜ ˜ ˜ ˜ ˜ cancels and it remains the following terms. (5. j (V ).5 If we set u = u − V . the functional j(V ) is Gˆteaux differentiable at point V ∈ U a and its directional derivative has the following expression.7 For Ω0 of class C 2 .5 Adjoint system and cost function shape derivative We are now coming back to the original problem of computing the gradient of the cost function j(V ). Zt . PROPOSITION 5. ∀W ∈ U (5. 5. W ). PROOF We recall that j(V ) = α 2 T 0 Ωt (V ) |u(V )|2 + γ 2 T 0 Γt (V ) |V |2 (5. − ν D(D Zt · u) · · D w + [−∂t W − D W · V − D V · W + ν∆W ] · w ˜ Q(V ) [∂t (D u · Zt ) + D(D u · Zt ) · u + D u · (D u · Zt ) + D(D u · Zt ) · V ˜ ˜ ˜ ˜ ˜ ˜ + D V · (D u · Zt )] · w − [D u · (D Zt · u) + D V · (D Zt · u) − D(D V · Zt ) · u] · w ˜ ˜ ˜ ˜ ˜ + [∂t u + D u · V + D u · u + D V · u] · (D Zt · w) − D u · W · w ˜ ˜ ˜ ˜ ˜ ˜ +ν D(D u·Zt )·· D w−ν(D u·D Zt )·· D w+ν D u·· D(D Zt ·w)−ν D u··(D w·D Zt ) ˜ ˜ ˜ ˜ +(D [∂t V + D V · V − ν∆V ]·Zt )·w+[∂t V + D V · V − ν∆V ]·(D Zt ·w) = 0 REMARK 5. W = Q(V ) α u(V ) · u (V ) · W .83)).

π) to be the solution of the adjoint system (Eq. −∂t λ − Γ λ · V = f. We get T 0 Ωt (V ) T αu·u = − 0 Γt (V ) σ(ϕ. Q(V ) (5. (5. The expression of the directional derivative is a direct consequence of Th. Using the fluid adjoint state and the adjoint transverse field. π) in Eq. (5.   ϕ(T ) = 0.92) PROOF T 0 T Ωt (V ) We need the following identity. where (ϕ. ΓT (V ) 1 with f = −(σ(ϕ. THEOREM 5.4) and Th.90) and λ is the adjoint transverse boundary field and is the solution of the tangential dynamical system. it is possible to identify the gradient distribution associated to the functional j(V ).Dynamical shape control of the Navier-Stokes equations 139 The differentiability property is an easy consequence of the differentiability of JV (u) with respect to (u. T [∂t u + D u · u + D u · u − ν∆u + [−∂t v − D v · u + ∗ D u · v − ν∆v + T p ]v − q] u − q div u 0 T Ωt (V ) = 0 Ωt (V ) p div v 0 Ωt (V ) + 0 Γt (V ) [p n · v − νv · ∂n u + νu · ∂n v − u · q n] (5. q) = (ϕ. π) · n + γ V (5. (5.91) Σ(V )  ϕ = 0.   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = α u. ΩT (5. V ) and the shape differentiability of u(V ) with respect to V . π) · n.93) We define (ϕ. Q(V )   div(ϕ) = 0. (5. u (5.94) . π) stands for the adjoint fluid state solution of the following system. j(V ) = −λ n − σ(ϕ.13 For V ∈ U and Ω0 of class C 2 . and we set (v. the functional j(V ) possesses a gradient j(V ) which is supported on the moving boundary Γt (V ) and can be represented by the following expression.5).93).8)). π) · n) · (D V · n − D u · n) + 2 (α + γ H)|V |2 . (0. T ) λ(T ) = 0.

Avoiding the differentiation of the state equations with respect to the design variable V is of great interest for shape optimization problems. especially if . π) · n) · (D V · n − D u · n) T 0 Γt (V ) 1 + (α + γ H)|V |2 Zt . the so-called Maximum Principle.6 the formula. W = 0 Γt (V ) − (σ(ϕ. involving the adjoint of the linearized state. Actually. the tedious computation of the state differentiability is not necessary in many cases. we have considered free divergence field V . This is a consequence of a fundamental result in optimal control theory. T on Γt (V ) (5. Actually. n + 2 Then we use Th. π) · n + γV ] · W 1 E = −(σ(ϕ. u = W + (D V · n − D u · n) Zt . Furthermore.6 Min-Max and function space parametrization In the previous section. and even if the state is not differentiable. REMARK 5.140 Moving Shape Analysis We use the boundary condition on Γt (V ) for the linearized state u . π) · n) · (D V · n − D u · n) + (α + γ H)|V |2 2 and we get the correct result. n . Thus.8) with [−σ(ϕ.e. it can happen that first-order optimality conditions still hold. then this term is null and we get that 1 f = −ν(D ϕ · n) · (D V · n − D u · n) + (α + γ H)|V |2 2 5. (5. we have π(D(V − u) · n) · n = π div V |Γt using (D(V − u) · n) · n = div(V − u)|Γt − divΓ (V − u) and the fact that V − u = 0 on Γt . i. we have been using the differentiability of the fluid state with respect to the Eulerian velocity V as a sufficient condition in order to derive first-order optimality conditions.95) j (V )..

such as the divergence free condition or the nonhomogeneous Dirichlet boundary conditions. Null divergence condition The divergence free condition coming from the fact that the fluid has an homogeneous density and evolves as an incompressible flow is difficult to impose on the mathematical and numerical point of view. We suggest at least 3 possible choices to handle this condition in our Min-Max formulation. In this section. This allows us to derive the expression of the cost function gradient involving the fluid and transverse field adjoints. It can be taken into account in the state and multipliers spaces.96) is an isomorphism.Dynamical shape control of the Navier-Stokes equations 141 we deal with a moving domain system. we are concerned with the function space parametrization. but it seems to be the best approach in order to get rig- . the divergence free condition must be invariant with respect to the use of transport map during the derivation of optimality condition for the Lagrangian functional. Ω0 ) −→ H0 (div. It is well known that the Piola transform does preserve the divergence quantity. In the first part. we define the saddle point formulation of the fluid state equations and the Lagrangian functional associated to the cost functional. Indeed we have the following property: LEMMA 5. This reduces the choice of appropriate maps and indeed the ALE map Tt does not satisfy this invariance condition. Ωt ) ϕ −→ ((Jt )−1 D Tt · ϕ) ◦ Tt−1 (5.16 [13] The Piola transform 1 1 Pt : H0 (div.1 Saddle point formulation of the fluid state system In the next paragraphs. Then. This new transform introduces additional mathematical and computational efforts. 1. we shall describe how to build an appropriate Lagrangian functional that can take into account all the constraints imposed by the mechanical problem. Thus. we perform a sensitivity analysis of the Lagrangian thanks to the transverse field and the fundamental Min-Max principle. 5. In this case. which consists in transporting the different quantities defined in the perturbed moving domain back into the reference moving domain that does not depend on the perturbation parameter.6. differential calculus can be performed since the functions involved are defined in a fixed domain with respect to the perturbations.

we adopt such a strategy. its lack of rigorous mathematical justification. there exists different methods to take into account this boundary condition in a Min-Max formulation. This leads in a certain sense to a saddle point formulation or mixed formulation of the Navier-Stokes system. Let ε > 0 be a small parameter. In the sequel. ε We may work with such a modified system. keeping in minds. (5. and that the Navier-Stokes suffers from a lack of convexity while taken into account in the Lagrangian functional. on Γt (V ) (5.2)) involves an essential non-homogeneous Dirichlet boundary condition. u = V.97) Σ  u = V. the convergence of the penalized adjoint is not established. to deal with divergence free conditions in a sensitivity analysis of the moving system. We can use a lifting of the boundary conditions inside the fluid domain and define a change of variable inside the coupled system. 1. 3.142 Moving Shape Analysis orous mathematical justifications of the Lagrangian framework in the context of non-cylindrical and free boundary problems. Ω0 × R with σ ε (u) = 1 div(u) I +ν(D u + ∗ D u). It is well known that the wellposedness of such formulations is only established for the Stokes system. But still.98) Again. It has the drawback to put additional terms inside the Lagrangian functionals and to impose more regularity on the boundary conditions. at least on the mathematical computation point of view. derive the optimality conditions of the penalized Lagrangian functional and finally perform an asymptotic analysis on the adjoint and primal system. Q  ε (5. as done in Section (5. . since even for non-moving Navier-Stokes problem. it is not clear if such a procedure may actually work. One way to avoid the use of this transform is the penalization of the divergence free condition inside the Navier-Stokes system.  2 u(t = 0) = u0 . it seems to be the easiest way. A third choice is to include the divergence free condition directly into the Lagrangian functional thanks to a multiplier that may play the role of the adjoint variable associated to the primal pressure variable.5). For the time being. 2. Non-homogeneous boundary conditions The Navier-Stokes system (Eq. We may consider the new penalized system :   ∂t u + D u · u − ν∆u − 1 (div u) = 0.

(H 1 (D))2 ) We define the fluid weak state operator. In the sequel. consisting in totally transposing the Laplacian operator. We recover the boundary constraints in performing an integration by parts in the optimality conditions corresponding to the sensitivity with respect to the multipliers. T . (H 1 (D))d ) We also need test function spaces that will be useful to define Lagrange multipliers: Y (Ωt ) = def def 1 v ∈ L2 (0. Fluid state operator In this section we shall summarize the different options that we have chosen for the Lagrangian framework and define the variational state operator constraint. Following the existence result stated previously. T . T . (H 2 (Ωt ))2 ∩ (H0 (Ωt ))d ) Q = q ∈ H 1 (0. We refer to [4] for recent applications to the Navier-Stokes system. T .99) Ωt Ωt Γt Then we shall substitute inside this identity the desirable boundary conditions.7 This method has been popularized in [99] as a systematic way to study non-homogeneous linear partial differential equations. we introduce the fluid state spaces: X(Ωt ) = def u ∈ H 2 (0. We can use a very weak formulation of the state equation. REMARK 5.Dynamical shape control of the Navier-Stokes equations 143 2. These formulations are usually called very weak formulations or transposed formulations. This procedure has been already used in [135] to perform shape optimization problems for elliptic equations using Min-Max principles. (H 2 (Ωt ))d ∩ (H 1 (Ωt ))d ) def Z = p ∈ H 1 (0. eV : X × Z −→ (Y × Q)∗ . −ν∆u · φ = −ν∆φ · u + ν [u · ∂n φ − φ · ∂n u] dΓ (5. we will need to define precise state and multiplier spaces in order to endow our problem with a Lagrangian functional framework. We shall notice that these methods are still valid in the non-linear case to obtain regularity or existence results.

q) (5. q) · n) ∀ (v. v. π(V )). we avoid the computation of the state derivative with respect to V . p) ∈ X(Ωt (V )) × Z max (v. q) = 0 Ωt (V ) − u · ∂t v + (D u · u) · v − νu · ∆v T +u· + q − p div v + ΩT 0 Γt (V ) V · (σ(v. First-order optimality conditions will furnish the gradient of the original cost functional using the solution of an adjoint problem. q) ∈ Y (Ωt (V )) × Q LV (u. The KKT system will have the following structure : ∂(v. q) · (δu. better known as Karusch-Kuhn-Tucker optimality conditions.102) By using the Min-Max framework. (5. q) ∈ Y × Q u(T ) · v(T ) − Ω0 u0 · v(t = 0). v. Let us first study the saddle point problem. q) with JV (u. (u. (v. p).103).100) |u|2 + γ 2 T 0 Γt (V ) |V |2 (5. v.3) : LV (u. δq) = 0.103) Optimality Conditions In this section. p. p). Min-Max problem In this section.p) LV (u.p)∈X×Z min (v. (5. ∀ (δv. (v. we are interested in establishing the first order optimality condition for problem Eq. q) (5. This step is crucial.3) can be put in the following form: min V ∈ U min (u. δq) ∈ Y × Q → State Equations ∂(u. p. (5. q) = JV (u. we introduce the Lagrangian functional associated with Eq. q) · (δv.q)∈Y ×Q max LV (u.101) Using this functional.q) LV (u. (5. v. p) = α 2 0 T Ωt (V ) def (5. δp) ∈ X × Z → Adjoint Equations . the optimal control problem Eq. p. ∀ (δu. p) − eV (u.144 whose action is defined by : T Moving Shape Analysis eV (u. p. δp) = 0. v.2) and Eq. because it leads to the formulation of the adjoint problem satisfied by the Lagrange multipliers (ϕ(V ). p.

  −∂t ϕ − D ϕ · u + (∗ Du) · ϕ − ν∆ϕ + q = αu. q) ∈ X × Y × Q. v.105) . LV (u. ϕ. p.8 Existence and regularity results for the linearized NavierStokes adjoint problem can be found in [1. π).104) REMARK 5. 86] for the 2D case.19 For V ∈ U. Q(V )   div(ϕ) = 0. p. p. v.Dynamical shape control of the Navier-Stokes equations 145 LEMMA 5. δu = Ωt (V ) [α u · δu + δu · ∂t v − [D δu · u + D u · δu] · v + νδu · ∆v − δu · + ΩT q] δu(T ) · v(T ). (u. δp = (δp) div ϕ. ∀ δu ∈ X In order to obtain a strong formulation of the fluid adjoint problem.   ϕ(T ) = 0. p. π). v. δu = ˆ − Q(V ) [−∂t ϕ + (∗ Du) · ϕ − (Dϕ) · u − div(u) · ϕ − ν∆ϕ + −ν Σ(V ) π − αu] · δu ϕ(T ) · δu(T ) (∂n δu) · ϕ − ΩT LEMMA 5. ΩT (5. v. LV (u. q). 0 Ωt ∀ δp ∈ Z (5. we perform some integration by parts : LEMMA 5. This leads to the following fluid adjoint strong formulation.17 For V ∈ U. q) is differentiable with respect to p ∈ Z and we have T ∂p LV (u. Q(V ) Σ(V )  ϕ = 0. ϕ. There is a lack of results for the 3D case. q) ∈ Z × Y × Q. v. (p. p. These results can be easily adapted for the moving domain case.18 Q(V ) (D δu · u) · v = − Q(V ) [D v · u + div(u) · v] · δu + Σ(V ) (δu · v)(u · n) It leads to the following identity : ∂u LV (u. q) is differentiable with respect to u ∈ X and we have T 0 ∂u LV (u.

Tρt : Ωt −→ Ωρ t x → Tt (V + ρW ) ◦ Tt (V )−1 and we define the following parametrization. we perturb the moving domain Ωt (V ) by a velocity field W which generates the family of transformation Ttρ = Tt (V +ρW ).109) (Ttρ )−1 .110) min (u.6. This point can be solved using particular parametrization of the functional spaces.108) (5.q)∈Y (Ωρ )×Q t ρ 0 We need a theorem that would give the derivative of a Min-Max function with respect to a real parameter ρ ≥ 0. it is not trivial since the state and multiplier spaces X(Ωρ ) × Y (Ωρ ) depend on the perturbation parameter t t ρ. g(ρ) = j(V + ρW ) = t t L(V +ρW ) (u ◦ Rρ .3). we use the transverse map introduced in Section (5. X(Ωρ ) = u ◦ (Ttρ )−1 . v ◦ Rρ . To this aim. v. p) ∈ X(Ωt ) × Z max (v. p. with ρ ≥ 0 and the family of domains and their boundaries. u ∈ Y (Ωt (V )) This parametrization does not affect the value of the saddle point functional g(ρ). q) (5.146 Moving Shape Analysis 5.2 Function space parametrization To compute the first-order derivative of j(V ). t Y (Ωρ ) t = v◦ u ∈ X(Ωt (V )) (5. but changes the parametrization of the Lagrangian functional. In our case.p)∈X(Ωρ )×Z t (v.107) def def def (u. p. q) ∈ Y (Ωt ) × Q . Ωρ = Tt (V + ρW )(Ω0 ) t Γρ = Tt (V + ρW )(Γ0 ) t We set g(ρ) = j(V + ρW ) = min max L(V +ρW ) (u. q) (5.106) The objective of this section is to compute the following derivative : lim 1 (g(ρ) − g(0)) ρ (5.

115) . p) = V. ρ0 ] × X × Y → R def (5. x. Finally.113) (5. q) = JV +ρW (u ◦ Rρ . y) (5.Dynamical shape control of the Navier-Stokes equations t with Rρ = (Ttρ )−1 .114) y ρ ∈ Y. y) = g(ρ) G(ρ. G(ρ. x) = xρ ∈ X. v. using a fundamental identity. h(ρ) = sup y∈Y x∈X inf G(ρ. xρ . x. Then we apply it to our case of study.112) x∈X y∈Y and the sets X(ρ) = Y (ρ. y) (5. sup y∈Y G(ρ. G : [0. t 5.111) with ρ0 ≥ 0 and two topological spaces (X. we define dual functions and sets. x. y ρ ) = sup y∈Y In a similar way. q) · nρ ) ΩT u(T ) · v(T ) + Ω0 u0 · v(t = 0) ∀ (v. Y ). For each ρ ∈ I = [0. x. we are able to express the gradient j(V ) as stated in the main theorem of this chapter. we define g(ρ) = inf sup G(ρ. General theorem We a consider a functional. p. we first state a general theorem concerning the differentiability of a Min-Max problem with respect to a scalar parameter.3 Differentiability of the saddle point problem In this section.6.W T def 147 − 0 Ωρ t t t t t t − u ◦ Rρ · ∂t (v ◦ Rρ ) + (D u ◦ Rρ · u ◦ Rρ ) · v ◦ Rρ t q − p div(v ◦ Rρ ) t t t − νu ◦ Rρ · ∆(v ◦ Rρ ) + u ◦ Rρ · T − − 0 Γρ t t (V + ρW ) · (σ(v ◦ Rρ . q) ∈ Y (Ωt (V )) × Q where nρ stands for unit exterior normal of the perturbed boundary Γ ρ . We set t Lρ (u. y) (5. ρ0 ].

k) lim inf (0. x. for any sequence (ρn )n≥0 ∈ I with lim ρn = 0. i) lim xnk = x0 for the TX topology. xρ . (H4) There exists a topology TY on Y such that. there exists y 0 ∈ Y (0) n ∞ and a subsequence ρnk and for each k ≥ 1. G(ρ. i) lim ynk = y 0 for the TY topology.∞) ∂ρ G(ρ. y) (5.k) lim inf (0. x. for any sequence (ρn )n≥0 ∈ I with lim ρn = 0. y 0 ) ∀ x ∈ X(0). S(ρ) = {(x. there exists xnk ∈ X(ρnk ) such that. xnk . x∈X inf G(ρ. n ∞ ii) (ρ. x. y) = inf x∈X Finally we define the sets of saddle points. x. x. y) ≥ ∂ρ G(0. y ρ ) = h(ρ) G(ρ. y) ∈  ρ∈I X(ρ) × Y (0) n X(0) × ρ∈I Y (ρ)  (H3) There exists a topology TX on X such that.117) xρ ∈ X. g(ρ) = G(ρ. . n ∞ ∞ ii) (ρ.116) (5. y) ∀ y ∈ Y (0). y) = Moving Shape Analysis y ρ ∈ Y. ynk ) ≤ ∂ρ G(0. ρ ∈ I. y) = h(ρ)} (5. there exists ynk ∈ Y (ρnk ) such that. (H2) The partial derivative ∂ρ G(ρ. y) exists in I for all    (x. there exists x0 ∈ X(0) and a subsequence ρnk and for each k ≥ 1. x. x0 . y) ∈ X × Y. (H1) The set S(ρ) = ∅.∞) ∂ρ G(ρ.14 [40] Assume that the following hypothesis hold.118) THEOREM 5.148 and the sets Y (ρ) = X(ρ.

21 d u ◦ Rt ρ dρ ρ=0 = − D u · Zt . y) (5. y 0 ) ∈ X(0) × Y (0) is a saddle point of ∂ρ G(0.14). x. . We shall successively differentiate the distributed and the boundary integrals involved in the perturbed Lagrangian: a) Distributed terms: We set G(ρ. we need to differentiate the perturbed Lagrangian functional L(ρ). y 0 ) ∈ X(0) × Y (0) such that dg(0) = lim ρ 149 g(ρ) − g(0) = 0 ρ x∈X(0) inf sup y∈Y (0) ∂ρ G(0. def LEMMA 5. y 0 ) = sup inf ∂ρ G(0.Dynamical shape control of the Navier-Stokes equations Then there exists (x0 . (5.20 dTρt dρ dRt ρ dρ = Zt ρ=0 ρ=0 = −Zt LEMMA 5. x. x. .119) This means that (x0 .) = −u ◦ Rt · ∂t (v ◦ Rt ) + D(u ◦ Rt ) · (u ◦ Rt ) · v ◦ Rt ρ ρ ρ ρ ρ −ν(u ◦ Rt ) · ∆(v ◦ Rt ) + (u ◦ Rt ) · ρ ρ ρ q − p div(v ◦ Rt ) ρ with Rt = (Tρt )−1 . x0 . y) y∈Y (0) x∈X(0) = ∂ρ G(0. Derivative of the perturbed Lagrangian Following Th.) with respect to ρ. ρ We shall need the following lemmas in order to derivate G(ρ. y).

). we need some extra identities t corresponding to boundary shape derivates of terms involved in φ(ρ. .) and Lem.) t ρ=0 ρ=0 = − D u · Zt LEMMA 5. . n b) Boundary terms : We must now take into account the terms coming from the moving boundary Γρ . (5.)|ρ=0 = [(D u · Zt ) · ∂t v + u · (∂t (D v · Zt )) − [(D(D u · Zt )) · u + D u · (D u · Zt )] · v − (D u · u) · (D v · Zt ) +ν(D u · Zt ) · ∆v + νu · (∆(D v · Zt )) + p div(D v · Zt ) − (D u · Zt ) · PROOF (5. q] It comes easily using definition of G(ρ. d dρ G(ρ. .150 PROOF Moving Shape Analysis Using the chain rule we get D u ◦ Rt ρ Dρ = D u ◦ Rt · ρ = − Du ◦ Rt ρ ρ=0 D Rt ρ Dρ · Z (ρ. .) = (V + ρW ) · −q I +ν D(v ◦ Rt ) · nρ ρ = E(ρ) · nρ (5. x)dx ρ=0 Ωρ t = Ωt [(D u · Zt ) · ∂t v + u · (∂t (D v · Zt )) q] − [(D(D u · Zt )) · u + D u · (D u · Zt )] · v − (D u · u) · (D v · Zt ) +ν(D u · Zt ) · ∆v + νu · (∆(D v · Zt )) + p div(D v · Zt ) − (D u · Zt ) · + Γt [−u · ∂t v + (Du · u) · v − νu · ∆v + u · q − p div(v)] Zt .120) Since φ(ρ. .22 Then. . Then we set t φ(ρ. .) is defined on the boundary Γρ .20)- Then we have an expression of the derivative of distributed terms coming from the Lagrangian with respect to ρ.21). we have the following result ∂ρ G(ρ.

Γ (Zt · n) + H E. (5. n + (divΓ E) Zt .121) = Γt PROOF First. n Zt . n Then using the tangential Stokes identity from Lem. n We conclude using Ωt div E = Γt E .4). n + E. (5. we obtain the correct result. For the second identity.Dynamical shape control of the Navier-Stokes equations LEMMA 5. n − E. we need the following identities. n + (div E) Zt . n .23 [53] ∂ρ nρ |ρ=0 = nΓ = − LEMMA 5. d dρ div E(ρ) ρ=0 Ωρ t = Ωt div E + Γt (div E) Zt . n EΓt . (5. we have d dρ E(ρ). n Zt . Hence. we get d dρ E(ρ). (n ◦ Rt ) dΓ ρ Γρ t = ρ=0 Γt EΓ . nρ = div E(ρ) Γρ t Ωρ t Then we derive this quantity using Th. (5. (n ◦ Rt ) dΓ ρ = ρ=0 Γt Γρ t EΓ .5). nΓ +H E. we only need to compute the quantity EΓ . n Using Lem. To this end.23).5). n (5. we use that E(ρ). .24 d dρ E(ρ). nρ dΓ ρ=0 151 Γ (Zt · n) Γρ t = Γt E |Γt . using the Th.

Γ ρ=0 = − DΓ v · Zt = ∂ρ v|ρ=0 = 0. LEMMA 5. t − DΓ D(v ◦ Rt ) · Zρ ρ=0 ρ Γ ρ=0 ∂ρ D v ◦ Rt ◦ Tρt ρ=0 − (DΓ (D v)) · Zt ρ ∂ρ (D v) ◦ Rt · D Rt ◦ Tρt ρ=0 − (DΓ (D v)) ρ ρ ρ=0 · Zt = − D v · D Zt + D v · D(D Rt ) · ∂ρ (Tρt ) ρ = − D v · D Zt − (DΓ (D v)) · Zt − (DΓ (D v)) · Zt ρ=0 Using these results.123) .25 Moving Shape Analysis v ◦ Rt ρ PROOF Since v ◦ Rt ρ .27 EΓ = W · [−q I +ν D v] + ν V · [− D v · D Zt − DΓ (D v) · Zt ] This means that we have d dρ φ(ρ. x)dΓ ρ=0 (5. we can state the following : LEMMA 5.28 E |Γ = W · [−q I +ν D v] − ν V · [D(D v) · Zt ] (5.152 LEMMA 5. n We have also.122) Γρ t = Γt (V ) W · [−q n + ν D v · n] +ν V ·[−(D v · D Zt ) · n − (DΓ (D v) · Zt ) · n]+divΓ (V ·[−q I +ν D v]) Zt . ρ=0 = ∂ρ v ◦ Rt ◦ Tρt ρ ρ=0 LEMMA 5.26 D v ◦ Rt ρ PROOF D v ◦ Rt ρ Γ ρ=0 = − D v · D Zt − (DΓ (D v)) · Zt By definition we have Γ ρ=0 = D v ◦ Rt ρ = = = ∂ρ (D v) · D Rt ◦ Tρt ρ − (DΓ (D v)) · Zt .

d Lρ V.W dρ = ρ=0 d Jρ dρ V. LEMMA 5. q) ∈ Y (Ωt ) × Q T ρ=0 T Ωt (V ) = −α 0 u · (D u · Zt ) + 0 Γt (V ) γV ·W Zt . we have d dρ φ(ρ.29 d Jρ dρ V.W = JV.W T T ρ=0 − 0 d dρ Ωρ t G(ρ) ρ=0 − Furthermore we have.124) We shall use this expression in the sequel. ρ Lρ V.Dynamical shape control of the Navier-Stokes equations Hence.W − T 0 Ωρ t T G(ρ) − 0 Γρ t φ(ρ) u0 · v(t = 0) ∀ (v. n T + 0 Γt (V ) α 2 γ |u| + H|V |2 2 2 . n REMARK 5.9 Γt We recall that V · (D v · n) = = Ωt Ωt div(∗ D v · V ) D v · · D V + V · ∆v (5. We recall that the perturbed Lagrangian has the following form. x)dΓ ρ=0 153 Γρ t = Γt (V ) W · [−q n + ν D v · n] − ν V · [D(D v) · Zt · n] + div(V · [−q I +ν ∗ D v]) Zt .W 0 d dρ Γρ t φ(ρ) ρ=0 ∀ (v. q) ∈ Y (Ωt ) × Q − ΩT u(T ) · v(T ) + Ω0 Hence its derivative with respect to ρ at point ρ = 0 has the following expression.

15 For (u. (5. the . p. ϕ. π) saddle points of the Lagrangian functional (Eq. Indeed. THEOREM 5.100)).W dρ with T ρ=0 = −AZt − BZt − CW (5. we shall get the following expression. we would like to express the distributed term A Zt as a boundary quantity defined on the moving boundary Γt . n T CW = 0 Γt (V ) [W · [−q n + ν D v · n] − γ V · W ] The shape derivative kernel identity We shall now assume that (u. d Lρ V. n − α 2 γ |u| + H|V |2 2 2 Zt .125) AZ t = 0 Ωt (V ) αu · (D u · Zt ) + (D u · Zt ) · ∂t v − D(D u · Zt ) · u q + D u · D u · Zt · v + ν(D u · Zt ) · ∆v − (D u · Zt ) · + u·(∂t (D v·Zt ))−(D u·u)·(D v·Zt )+νu·(∆(D v·Zt ))+p div(D v·Zt ) T B Zt = 0 Γt (V ) − u · ∂t v + (Du · u) · v − νu · ∆v + u · q − p div(v) (Zt · n) − ν V · [(D(D v) · Zt ) · n] + div(V · [−q I +ν D v]) Zt . This will help us to simplify several terms involved in the derivative of LV with respect to V .154 Moving Shape Analysis Using the last identities concerning the derivative of the distributed and the boundary terms with respect to ρ. p. p. ϕ. v. π) is a saddle point of the Lagrangian functional LV . q) = (u.

q) = V T JV (u..e. p) · n) ∀ (v. v) are saddle points of the Lagrangian. q) ∈ Y × Q − ΩT u0 · v(t = 0). v.e. i. solutions of respectively the primal and adjoint fluid problem. i. T 0 Ωt (V ) 155 [α u · (D u · Zt ) + (D u · Zt ) · ∂t v − [(D(D u · Zt )) · u q] + u · (∂t (D v · Zt )) + D u · (D u · Zt )] · v + ν(D u · Zt ) · ∆v − (D u · Zt ) · T − (D u · u) · (D v · Zt ) + νu · (∆(D v · Zt )) + p div(D v · Zt ) Γt (V ) − 0 [ν V · (D(D ϕ · Zt ) · n) − (D ϕ · Zt ) · (−p n + ν(D u · n))] = 0. δu = D u · Zt δv = D v · Zt with δu(T ) = δv(T ) = δu(0) = δv(0) = 0. This leads to the following perturbation identity. ∂(u. where (u. δv) = V − T Q(V ) [−αu · δu − δu · ∂t v − u · ∂t δv + D(δu · u) · v + D(u · δu) · v q − p div(δv)] +D(u · u) · δv − ν(δu · ∆v) − ν(u · ∆δv) + δu · Γt (V ) + 0 − ν V · (D δv · n) + ν v · (D δu · n) + δv · (−p n + ν(D u · n)) [δu(T )v(T ) + u(T )δv(T )] . . We recognize immediately the distributed and boundary terms involved in the shape derivative kernel identity. p) − − 0 T 0 Ωt (V ) [−u · ∂t v + (D u · u) · v − νu · ∆v + u · T q − p div v] Γt (V ) V · (σ(v.e. δv) ∈ X(Ωt ) × Y (Ωt ) We choose specific perturbation directions. ΩT − ∀ (δu. i. q) · n) + u(T ) · v(T ) + Ω0 0 Γt (V ) v · (σ(u. ∀W ∈ U PROOF We shall use extremal conditions associated to variations with respect to (u. L2 (u. v) in the Lagrangian functional where we add a boundary integral since we consider test functions v that do not vanish on the boundary Γt (V )... p.v) L2 · (δu.Dynamical shape control of the Navier-Stokes equations following identity holds.

n + [−π div V − V · − ν V · [(D(D ϕ) · Zt ) · n] − This allows us to derive the expression of the cost function directional derivative. n (5.W dρ Σ(V ) = ρ=0 ν V · (D(D ϕ · Zt ) · n) + [−ν(D ϕ · n) · (D u · n) − νV · ∆ϕ π + ν D ϕ · · D V + ν V · ∆ϕ]) Zt . LEMMA 5. n 1 [α + γ H] |V |2 Zt . n − 1 [α + γ H] |V |2 Zt . we set (u. on Γt to simplify the remaining terms. ϕ) and we use the fact that u = V. T AZ t = 0 Γt (V ) [ν V · (D(D ϕ · Zt ) · n) − ν(D ϕ · Zt ) · (D u · n)] (5. n 2 −ν V ·[(D(D ϕ) · Zt ) · n]+[−π div V − V · T π + ν D ϕ · · D V + ν V · ∆ϕ]) Zt . n CW = 0 Γt (V ) [W · [−π n + ν D ϕ · n] − γ V · W ] We need to establish the following identity.156 Cost functional gradient Moving Shape Analysis Now.127) Then − d Lρ V. . v) = (u.30 Γt (D ϕ · Zt ) · (D u · n) = Γt (D ϕ · n) · (D u · n) Zt . n = (p div ϕ) Zt . on Γ t and ϕ = 0. T B Zt = 0 Γt (V ) [−νV · ∆ϕ + V · π] Zt . n = 0.126) REMARK 5. n 2 + [W · [−π n + ν D ϕ · n] − γ V · W ] +V · π] Zt .10 We have used that (D ϕ · Zt ) · (p n) = (D ϕ · (n ⊗ n) · Zt ) · (p n) = p ((D ϕ · n) · n) · Zt .

5.130) . n [−σ(ϕ. q) · n + γV ] · W (5. (5.  u = V. p = 0. div V = 0 in D. we use a different method based on function space embedding particulary suited for non-homogeneous Dirichlet boundary problems.129) and U = V ∈ H 1 (0. It means that the state and multiplier variables are defined in a hold-all domain D that contains the moving domain Ω t (V ) for t ∈ (0.8 157 dg(0) = Σ(V ) 1 −ν(D ϕ · n) · (D V · n − D u · n) + π div V + (α + γ H)|V |2 2 + Σ(V ) Zt .   u(t = 0) = u0 . (H m (D))d ).1 Saddle point formulation of the fluid state system We recall that we are dealing with the Navier-Stokes in a moving domain Ωt (V ) which is driven by an Eulerian velocity field V ∈ U.   ∂t u + D u · u − ν∆u +   div(u) = 0. with respect to the speed of the moving domain. Q(V ) Q(V ) Σ(V ) Ω0 (5. T . In this section. we have used a function space parametrization in order to get the gradient of a given functional related to the solution of the Navier-Stokes system in moving domain.128) Then we use Th. T ) and ∀ V ∈ U.7.8) with. 1 E = −ν(D ϕ · n) · (D V · n − D u · n) + π div V + (α + γ H)|V |2 2 and we get the correct result. V · n = 0 on ∂D (5.Dynamical shape control of the Navier-Stokes equations PROPOSITION 5.7 Min-Max and function space embedding In the previous section. 5.

  y(t = 0) = y0 . q. p. T . H 2 (D)) P = Q = H 1 (0. H 3/2 (Γt )) We are interested in the following Min-Max problem. Q(V ) (5. • The dual state (ϕ. q) · n . v. • The primal state (y. T .133) Σ(V )  ϕ = 0.131) The solution (y. π) is the solution of the fluid adjoint system. v. Q(V )   div(ϕ) = 0. This yields to the following functional. (v. p) ∈ X × P .q. (u. Q(V )   div(y) = 0. Ω0 • The multiplier satisfies the following identity.   ϕ(t = T ) = 0. v.p)∈X×P def def def def min (v. p. We introduce a Lagrange multiplier µ and a functional. H 1 (D)) M = H 1 (0. µ) = 0 Ωt (V ) [∂t u + D u · u − ν∆u + T T Ωt (V ) p] · v (u − V ) · µ − 0 q div u − 0 Γt (V ) for (u. T . q) = 0 [∂t u + D u · u − ν∆u + T p]·v− 0 Ωt (V ) q div u − 0 Γt (V ) (u − V ) · σ(v. T T Ωt (V ) EV (u.   −∂t ϕ − D ϕ · u + (∗ Du) · ϕ − ν∆ϕ + π = 0. µ) (5. µ = −q n + ν(D ϕ · n).158 Moving Shape Analysis with m > 5/2.µ)∈Y ×Q×M max EV (u. p) is a solution of the Navier-Stokes system. q.134) Then we can choose the above particular representation of the boundary Lagrange multiplier µ. T EV (u. p. p. ΩT on Γt (V ) (5. q) ∈ Y × Q and µ ∈ M with X = Y = H 1 (0. π.   ∂t y + D y · y − ν∆y + p = 0.132) Σ(V )  y = V. ϕ. λ) of this problem is characterized by the following optimality system. Q(V ) (5.

Dynamical shape control of the Navier-Stokes equations for (u.135) where j(V ) = JV (u(V ). p) ∈ X × P . q) ∈ Y × Q. (v. with σ(v. T T Ωt (V ) EV (u. p(V )) with (u(V ). This will be useful for its differentiation with respect to V . p) = α 2 T 0 Ωt (V ) |u|2 + γ 2 T 0 Γt (V ) |V |2 (5.2) and JV (u. p) is a real functional of the following form : JV (u. we may get the final expression of our saddle functional.11 The above expression of the Lagrange functional has the advantage to include only distributed terms.2 The Lagrange functional We are interested in the following minimization problem. p(V )) is a weak solution of problem (5. The following identities hold true.31 (u − V ) · (D v · n) = = Ωt (V ) 159 on Γt (V ) Γt (V ) Ωt (V ) div [∗ D v · (u − V )] [D(u − V ) · · D v + (u − V ) · ∆v] and (u − V ) · q n = = Ωt (V ) Γt (V ) Ωt (V ) div [q(u − V )] [(u − V ) · q + q div(u − V )] Using this identity. (v. p) ∈ X × P . q) ∈ Y × Q. p.136) . V ∈U min j(V ) (5. q) = 0 T [∂t u + D u · u − ν∆u + p]·v− 0 Ωt (V ) q div u + 0 Ωt (V ) [(u − V ) · q + q div(u − V ) − ν D(u − V ) · · D v − ν(u − V ) · ∆v] for (u. q) · n = −q n + ν(D ϕ · n). LEMMA 5. 5. v.7. REMARK 5.

q) (5. p. As in the previous section we perturb the tubes using a vector field W ∈ U with an increment parameter ρ ≥ 0. and state d j(V + ρW ) dρ = ρ=0 (5. p) − EV (u. p. the perturbed Lagrangian has the following form. q) = JV +ρW (u. p. q) = JV (u.137) with LV (u. we can state LEMMA 5. p. v. S(ρ) = X(ρ) × P × Y (ρ) × Q ∈ X × P × Y × Q is not a singleton since X(ρ) = u ∈ X.139) (u. q) (5. V∈U min (u. Lρ (u. v. n − 0 Ωt (V ) [−W · q − q div W + ν D W · · D v + νW · ∆v] T + 0 Γt (V ) γV · W . u|Ωρ = y(ρ) t Y (ρ) = v ∈ Y. p.q)∈Y ×Q max LV (u. v.32 T ∂V LV (u. q) dρ (5. v. v|Ωρ = ϕ(ρ) t We make the conjecture that we can bypass the min-max.p)∈X×P min (v. v. Since the functions are embedded in the hold-all domain D. p. v.q)∈Y ×Q max d ρ L (u.160 Moving Shape Analysis We may solve this problem by studying the equivalent Min-Max problem. v. q)·W = − +(u − V ) · γ −H |V |2 2 0 Γt (V ) [(∂t u + D u · u − ν∆u + p) · v − q div u α 2 |u| 2 q + q div(u − V ) − ν D(u − V ) · · D v − ν(u − V ) · ∆v − T Zt . q) The set of saddle points.138) Our main concern is the differentiation of the above functional with respect to V ∈ U. p. p) − EV +ρW (u.p)∈X×P min (v.140) ρ=0 Using non-cylindrical shape derivative framework.

q) · n Then. Q(V ) Σ(V )  ϕ = 0. p) = (y. 0) on Γt (V ) and Ωt (V ) 161 (5. p) and (v. q) = (ϕ. we get D y · · D ϕ = D y · ·(D ϕ · (n ⊗ n)) = (D y · n) · (D ϕ · n) 2 Consequently we get T (5.33 D y · · D ϕ|Γt (V ) = (D y · n) · (D ϕ · n) PROOF Using that ϕ = 0 on Γt (V ) yields to D ϕ|Γt = D ϕ · (n ⊗ n)|Γt then.143) . n + 2 [−σ(ϕ. ΩT (y.142) j (V ). LEMMA 5. W = 0 Γt (V ) − ν(D ϕ · n) · (D V · n − D u · n) + π div V T 0 Γt (V ) 1 + (α + γ H)|V |2 Zt . q) · n + γV ] · W (5. ϕ) = (V. we may state that div y|Γt = 0. T ∂V j(V ) · W = − 0 Γt (V ) − π div y + π div(y − V ) − ν D(y − V ) · · D ϕ 1 − (α + Hγ)|V |2 Zt . n + (σ(ϕ.Dynamical shape control of the Navier-Stokes equations Then we set (u.   ϕ(T ) = 0. π) with   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = α u.141) and we use that [−W · q − q div W + ν D W · · D v + νW · ∆v] = Γt (V ) W · σ(v. π) · n − γV ) · W 2 Using regularity assumptions on y and the free divergence condition on y. Q(V )   div(ϕ) = 0.

162 Moving Shape Analysis We then use theorem (5. . On the numerical point of view. In addition to the classical method based on the state derivative with respect to shape motions. Its originality lies in the fact that the domain containing the fluid is moving. We have introduced an open loop control problem based on the velocity of the moving domain with the goal of reaching a given objective related to the behaviour of the fluid. 2 5. they allow more flexible computations which can be very useful for practical purposes. as it will be shown in the next chapters.8 Conclusion In this chapter. We believe that the concepts introduced in this chapter will prove large efficiency for coupled problems involving a moving boundary.8) with 1 E = −ν(D ϕ · n) · (D V · n − D u · n) + π div V + (α + γ H)|V |2 2 and we get the correct result. Our main concern was to show how the gradient of the cost functional involved in the optimal control problem can be obtained by using non-cylindrical shape optimization concepts. Even if for the time being these methods lack from a rigorous mathematical framework. we have been dealing with a particular shape optimization problem involving the Navier-Stokes equations. an implementation of the open loop control is under study in the 2D case. we have introduced two different methods based on the Min-Max principle.

we introduce the general framework of non-cylindrical Lagrangian shape derivative and we state the structure of the gradient for general functionals under or without state constraints. The fluid is described by the Navier-Stokes equations.1 Introduction In this chapter. We also establish an equivalence result between Eulerian and Lagrangian derivatives. 163 .2 Evolution maps ˆ θ : (0.3. we state similar results as the previous chapter but with a method based directly on the Lagrangian flows which do not define a linear space. x) → θ(t. T ) × Rd −→ Rd ˆ (t.2. but has the advantage to first avoid the introduction of transverse fields and could prove appropriate to deal with fluid-solid interaction problems. 2. We deal with the state derivative with respect to the Lagrangian flow and an optimal control with a general cost function. The chapter is organised as follows. In section 6. x) Let us consider a smooth map ˆ and the image Ωθ by θ of a fixed domain Ω0 ⊂ Rd with a Lipschitzian boundary ˆ Γ0 . 1.Chapter 6 Tube derivative in a Lagrangian setting 6. 6. In section 6. We finally recover results obtained in [57] using Eulerian derivative concepts. we apply the general framework to mechanical systems involving a fluid in a moving domain.

T . Θ = def ˆ ˆ ˆ θ| (θ.164 def Moving Shape Analysis We set C k = (C k (Rd ))d and we define the space of evolution map.p (Rd )).p (Rd )) ˆ ˆ ˆ θ → φ(θ) ◦ θ . 1 ≤ p ≤ ∞.1 For θ ∈ Θ. T . W m−1. LEMMA 6. let us consider a function φ(θ) which depends on the evolution map ˆ We shall define the derivative of this function in the following way.2.p (Rd )) ˆ ˆ θ → f ◦θ ˆ ˆ is differentiable at point θ and its derivative in the direction δ θ ∈ Θ is given by ∂ ˆ ˆ ˆ [ (f ◦ θ)] · δ θ = f · δ θ. 1 ≤ p ≤ ∞.p (Rd )) (6. W m.2 Let us consider the scalar function φ ∈ L2 (0. we would like to perform the ˆ ˆ differentiation of the function θ → f ◦ θ. θ. C k ))2 (6. T .p (Rd )). ˆ 6. if p = ∞ Then the application Θ −→ L2 (0.2) ˆ ∂θ ˆ Now. W m. T ) × Rd . θ−1 ) ∈ (C 1 ([0. k ≥ m − 1 f ∈ C m (Rd ). k ≥ m − 1 f ∈ C m (Rd ). ∈ L2 (0. T ].1) ˆ ˆ REMARK 6. T . m ≥ 1. T . the set Ωθ = θ(Ω0 ) is still an open set and its ˆ boundary Γθ is Lipschitzian.1 Let us consider the scalar function f ∈ L2 (0.1 Basic differentiation results Given a function f defined on (0. if p = ∞ Let us assume that the function Θ −→ L2 (0. LEMMA 6. W m. m ≥ 1. W m.

3) with A ≡ Q ∈ ET . 165 In this section. W m−1. and J : A −→ R Q → J(Q) (6. it is possible to define derivation ˆ concepts of functions defined in Qθ or Σθ with respect to θ.6) ˆ → j(θ) = J(Q ˆ) ˆ θ θ with Qθ = ˆ 0<t<T ˆ ({t} × Ωt ) = (t.Tube derivative in a Lagrangian setting is differentiable. then the application Θ −→ L2 (0. we define the functional j = J ◦ θt such that j : Θad −→ R (6. we would like to solve the following problem. ˆ ˆ . This analysis will be performed using the Lagrangian horizontal mapping ˆ (t. W m.2 Mathematical setting Generally speaking.2. a e 6. T ).p (Rd )) ˆ ˆ ˆ φ(θ) · (δ θ ◦ θ−1 ). Q∈A min J(Q) (6.3 Elements of shape calculus Using extension of classical elements of cylindrical shape calculus using the perturbation of identity ([116].5) Q⊂D (6. we would like to perform a sensitivity analysis with respect to the shape of integrals defined on evolution sets Q ∈ E T with Q ⊂ D × (0. derivatives in the sense of Gˆteaux or Fr´chet could make sense for j( θ).p (Rd )) ˆ ˆ θ → φ(θ) ˆ is differentiable and its derivative in the direction δ θ ∈ Θ is given by [ ∂ ˆ ∂θ ˆ ˆ φ(θ)] · δ θ = [ ∂ ˆ ˆ ˆ ˆ (φ(θ) ◦ θ)] · δ θ ◦ θ−1 − ˆ ∂θ ∈ L2 (0. [101].2. θ). 6.[135]). T .4) In order to work with standard differential calculus. T . ˆ Then. θ)(Q0 ).

a. ρ0 [ → H(D) ˆ ρ → f ((I +ρδθ) ◦ θ) d dρ fρ ρ=0 ˆ is differentiable at point ρ = 0. ˆ ˆ f˙(θ) · δθ = f (θ) · δθ + ˆ f (θ) · δθ (6. if the following composed function. Under regularity conditions. ˆ DEFINITION 6.2 . t) ∈ Qθ and f (θ) · δθ = ˆ REMARK 6. We ˆ shall state a differentiability result that uses the notion of non-cylindrical material derivative. ρ0 [ → H(Qθ ) ˆ ˆ ρ → f ((I +ρδθ) ◦ θ) ◦ (I +ρδθ) d ρ dρ f ρ=0 ˆ is differentiable at point ρ = 0. (x.2 A function f (θ) ∈ H(D) admits a non-cylindrical ˆ ˆ · δθ defined over Q ˆ at point θ ∈ Θad in the direction shape derivative f (θ) θ def ˆ ˆ δ θ = δθ ◦ θ ∈ Θ.e. (x.8) We can now state the differentiability properties of non-cylindrical integrals with respect to their moving supports.166 Moving Shape Analysis Derivation of moving domain integrals In this paragraph. if the following composed function. It can be proven that the following identity holds. t) ∈ Qθ and f˙(θ) · δθ = ˆ .7) Qθ ˆ ˆ We are trying to characterize the variations of J 1 (Qθ ) with respect to θ. f ρ : [0. we are interested in the differentiability properties of integrals defined on the non-cylindrical set Q θ . ˆ J1 (Qθ ) = ˆ f (Qθ )dQθ ˆ ˆ (6. a.e.1 A function f (θ) ∈ H(Qθ ) admits a non-cylindrical ˆ ˆ ˆ material derivative f˙(θ) · δθ defined over Qθ at point θ ∈ Θad in the direction ˆ def ˆ ˆ δ θ = δθ ◦ θ ∈ Θ. ˆ DEFINITION 6. it is possible to define a different derivative that will be useful in the context of Eulerian state equations. fρ : [0. .

−vν + δθ. we get ˆ ˆ Σθ ˆ ˆ f (θ) δθ. J2 (Σθ ) = ˆ Σθ ˆ g(Σθ )dΣθ ˆ ˆ (6. δ θ).10) Furthermore. let us assume that ˆ def ˆ for any direction δ θ = δθ ◦ θ ∈ Θ the following hypothesis holds.13) For that purpose. then 2 δθ. then ˆ d ˆ J (θ) · δθ = ˆ 1 dθ Qθ ˆ Qθ ˆ ˆ ˆ f˙(θ) · δθ + f (θ) div δθ dQθ ˆ (6. ν dΣθ = ˆ T 0 Γθ ˆ ˆ f (θ) δθ.Tube derivative in a Lagrangian setting 167 THEOREM 6.12) Derivation of moving boundary integrals It is also possible to establish a similar result for integrals over moving boundaries. ν dΣθ ˆ (6. if ˆ ˆ ii) f (θ) admits a non-cylindrical shape derivative f (θ) · δθ ∈ H(Qθ ). d ˆ J (θ) · δθ = ˆ 1 dθ Futhermore. . ˆ ˆ i) f (θ) admits a non-cylindrical material derivative f˙(θ) · δθ then J1 (. if Q0 is an open domain with a Lipschitzian boundary Σ0 . we need to define the non-cylindrical tangential material derivative. then d ˆ dθ ˆ J1 (θ) · δθ = ˆ f (θ) · δθdQθ + ˆ ˆ f (θ) δθ.) is ˆ ∈ Θad and its derivative is given by the Gˆteaux differentiable at point θ a following expression. θ). nt ) 2 using that dΣθ = (1 + vν )1/2 dt dΓθ .11) Qθ ˆ Σθ ˆ REMARK 6. ν = (1 + vν )−1/2 ( 0.1 For a bounded measurable domain Ω0 with boundary Γ0 . it means that perturbations are of the type (0.3 Since we deal with horizontal transformations of the type ˆ ˆ (t.9) ˆ ˆ f (θ) · δθ + div(f (θ) δθ) dQθ ˆ (6. nθ dΓθ dt ˆ ˆ (6.

e. if the following composed function.4 ˆ is differentiable at point ρ = 0. d ˆ J (θ) · δθ = ˆ 2 dθ and if. ˆ REMARK 6.e. gρ : [0. g ρ : [0.3 A function g(θ) ∈ H(Σθ ) admits a non-cylindrical ˆ ˆ ˆ def ˆ material derivative g(θ) · δθ at point θ ∈ Θad in the direction δ θ = δθ ◦ θ ∈ Θ ˙ ˆ if the following composed function. a. ρ0 ] → H(Σθ ) ˆ ˆ ρ → g((I +ρδθ) ◦ θ) ◦ (I +ρδθ) d ρ dρ g ρ=0 is differentiable at point ρ = 0. (t. Here It can be proven that the following identity holds. (x. t) ∈ Qθ and g (θ) · δ = ˆ .2 For a bounded measurable domain Ω0 with boundary Γ0 .168 Moving Shape Analysis ˆ DEFINITION 6.15) Σθ ˆ . ˆ i) g(θ) admits a non-cylindrical material derivative g(θ) · δθ then J2 (. ρ0 [ → H(D) ˆ ρ → g((I +ρδ θ) ◦ θ) ◦ p d dρ gρ ρ=0 p stands for the projection mapping on Γθ ([51]).) is ˙ ˆ ˆ Gˆteaux differentiable at point θ ∈ Θad and its derivative is given by the a following expression. Under regularity conditions. ˆ g(θ) · δθ + g(θ) divΓ δθ dΣθ ˙ ˆ ˆ (6.4 A function g(θ) ∈ H(Σθ ) admits a non-cylindrical ˆ ˆ ˆ shape derivative g (θ) · δθ defined over Σθ at point θ ∈ Θad in the direction ˆ def ˆ δ θ = δθ ◦ θ ∈ Θ. a.14) These concepts are involved in the differentiability property of boundary integrals. let us assume that ˆ def ˆ for any direction δ θ = δθ ◦ θ ∈ Θ the following hypothesis holds. it is possible to define a different derivative that will be useful in the context of Eulerian state equations. THEOREM 6. ˆ DEFINITION 6. ˆ g(θ) · δθ = g (θ) · δθ + ˙ ˆ Γ ˆ g(θ) · δθ (6. x) ∈ Σθ and g(θ) · δθ = ˙ ˆ ˆ .

To this end. LEMMA 6.18) ˆ An adjoint identity In the remaining part of this chapter.Tube derivative in a Lagrangian setting ˆ ˆ ii) g(θ) admits a non-cylindrical shape derivative . then ˜ ˆ ˆ ˜ ˆ d ˆ J (θ) · δθ = ˆ 2 dθ Σθ ˆ g (θ) · δθ + ( ˜ ˆ Γ ˆ g (V ) · nθ + H g(θ)) δθ.3 ˆ def ˆ For any E ∈ L2 (Σθ ) and δ θ = δθ ◦ θ ∈ Θ.20) . parts formula. of the type (t.e. i. ν dΣθ ˆ (6. THEOREM 6. T ) × Ω0 . ˆ PROOF We first need to prove the following lemma.3 T 0 Γθ ˆ T T ∂t f · g = Γθ (τ ) ˆ f ·g 0 − 0 Γθ ˆ f · ∂t g + H(f · g) (Vθ · nθ ) ˆ ˆ + (f · g) · nθ (Vθ · nθ ) ˆ ˆ ˆ (6. if g(θ) = g (θ)|Σθ with g ∈ H(Qθ ). g (θ) · δθ. nt dt d Γθ (6. the following identity holds.16) ˆ Furthermore.17) ˆ REMARK 6. ˆ T 0 Γθ ˆ ˆ ˆ E∂t (δθ ◦ θ) ◦ θ−1 = T 0 Γθ ˆ −∂t E − (divΓ Vθ ) E − D E · Vθ · δθ ˆ ˆ (6.19) ˆ ˆ with Vθ = (∂t θ) ◦ θ−1 . θ) d ˆ J (θ) · δθ = ˆ 2 dθ T 0 Γθ ˆ 2 ˆ ˆ g (θ) · δθ(1 + vν )1/2 + H g(θ) δθ. ν ˜ ˆ dΣθ ˆ (6. we get horizontal.5 Using the fact that the transformations θ on Qθ are ˆ ˆ on (0. we will try to obtain several functionals with respect to the Lagrangian mapping we will need to use the following fundamental integration by the structure of ˆ θ.. then d ˆ J (θ) · δθ = ˆ 2 dθ Σθ ˆ 169 ˆ ˆ g (θ) · δθ + H g(θ) δθ.

3) with f = δθ and g = E.4 ˆ ˆ ˆ ˆ (∂t h) ◦ θ−1 = ∂t (h ◦ θ−1 ) − (D h) ◦ θ−1 ∂t (θ−1 ) ˆ ˆ ∂t (θ−1 ) = − D(θ−1 ) · V ˆ θ (6. θ LEMMA 6.5 T 0 Γθ ˆ T E · (D δθ · Vθ ) = ˆ + 0 Γθ ˆ −δθ · (D E · Vθ ) − (divΓ Vθ ) δθ · E ˆ ˆ (6. T 0 Γθ ˆ ˆ ˆ E∂t (δθ ◦ θ) ◦ θ−1 = T 0 Γθ ˆ T E · (∂t δθ) + 0 Γθ ˆ E · (D δθ · Vθ ) (6. we obtain T 0 Γθ ˆ T E · (∂t δθ) = 0 Γθ ˆ − δθ · ∂t E − (E · δθ) · nθ (Vθ · nθ ) ˆ ˆ ˆ − H(E · δθ) (Vθ · nθ ) ˆ ˆ For the second right hand side we need to establish the following lemma. we get T T ∂t 0 Γθ ˆ f ·g = 0 T Γθ ˆ (f · g)Γθ + H(f · g) (Vθ · nθ ) ˆ ˆ ˆ ∂t (f · g)|Γθ + ˆ (f · g) · nθ + H(f · g) (Vθ · nθ ) ˆ ˆ ˆ (f · g) · nθ ˆ = 0 T Γθ ˆ = 0 Γθ ˆ ∂t f · g + f · ∂ t g + +H(f · g) (Vθ · nθ ) ˆ ˆ 2 ˆ In order to use this lemma.22) (6. we use the following identities.24) This allows us to state the following.26) (E · δθ) · nθ (Vθ · nθ ) + H (E · δθ) (Vθ · nθ ) ˆ ˆ ˆ ˆ ˆ Combining these two identities. For that purpose. .23) ˆ ˆ ˆ (∂t h) ◦ θ−1 = ∂t (h ◦ θ−1 ) + D(h ◦ θ−1 ) · Vθ ˆ ˆ We apply the last identity with h = δθ ◦ θ and we obtain ˆ ˆ ∂t (δθ ◦ θ) ◦ θ−1 = ∂t δθ + D δθ · V ˆ θ (6. we need to modify the expression of E∂ t (δθ ◦ θ) ◦ ˆ−1 . we obtain the expected result.170 Moving Shape Analysis Using classical integral derivatives formulas. LEMMA 6.21) (6.25) ˆ Using lemma (6.

ˆ 1. ˙ ˆ ˆ ˆ 2. y(θ)) + δθ.4 Unconstraint non-cylindrical shape optimization ˆ j(θ) = Qθ ˆ We would like to minimize over θ ∈ Θad the following functional.Tube derivative in a Lagrangian setting 171 6. y(θ)) ˆ ˆ · y (θ) · δθ dQ ˆ ˆ θ ˆ · y (θ)δθ dΣ ˆ ˆ θ y S(θ. y(θ)) · div δθ dQ ˆ ˆ ˆ ˙ ˆ θ + Σθ ˆ ˙ ˆ ˆ S(θ. ˆ ˆ G(θ). PROPOSITION 6. the partial derivative y (θ)δθ = y(θ) · δθ − ˙ ˆ ˆ y(θ) · δθ exists in H(Qθ ). nθ dt d Γθ ˆ ˆ ˆ ˆ ˆ ˆ ˆ +HS(θ. y(θ)) · divΓ δθ dΣ ˆ ˆ ˆ ˙ ˆ θ (6. y(θ)) ˆ ˆ · y(θ)δθ + S(θ. D (R )) with support inside Qθ such that ˆ d ˆ j(θ) · δθ = ˆ dθ • using material derivative. y(θ))δθ + ˆ ˆ S (θ. ˆ G(θ). ˆ ˆ then j(. it is possible to obtain the ˆ ˆ derivative with respect to θ of the functional j(θ). y(θ))dΣθ ˆ (6. y(θ)) ˆ + 0 Γθ ˆ ˆ ˆ ˆ ˆ F (θ. y(θ))δθ + Qθ ˆ y F (θ. ˙ ˆ ˆ F (θ. y(θ))δθ + y S(θ.30) . y(θ) admits a material derivative y(θ) · δθ ∈ H(Qθ ). δθ = ˆ ˆ · y(θ)δθ + F (θ. τ → j((I +τ δθ) ◦ θ) is differentiable at τ = 0.29) • using shape partial derivative.D dt (6. ˆ ˆ 3.28) ˆ Furthermore G(θ) admits the two following representations. δ θ = Qθ ˆ ˆ ˆ F (θ. y(θ)) (6. Using the shape derivative ˆ calculus tools introduced in the previous section.2. ˆ ˆ F (θ. y(θ))dxdt + Σθ ˆ ˆ ˆ S(θ.27) with Qθ = ˆ 0<t<T ˆ ˆ ({t} × θ(Ω0 )) and y(θ) ∈ H(Qθ ). δθ D . y(θ)) + ∂n S(θ. y(θ))∂n y(θ) + Σθ ˆ T y S(θ. T .1 ˆ ˆ def ˆ Let us assume that for θ ∈ Θad and for any direction δ θ = δθ ◦ θ ∈ Θ. y(θ)) · δθ + y F (θ. y(θ)) T 0 ˆ G(θ).) is differentiable at point θ and there exists a distribution G(θ) ∈ 2 d L (0.

ˆ A∗ p.31) with A ∈ L(H(Qθ . y(θ)) · δθ d Qθ = 0. y(θ)) + y S(θ. ˆ ˆ j (θ) · δθ = L(δ. y(θ)) δθ.6 Particular case : We assume the following hypothesis: ˆ 1. p(θ)) = Σθ ˆ ˆ (p(θ)) δθ. ˆ Ay (θ) · δθ.172 Moving Shape Analysis PROOF We perform the same steps as in [135] and we use the noncylindrical calculus introduced previously. ∀ φ ∈ H(Qθ ) ˆ (6. p(θ) = L(δθ. φ). . y(θ)) δθ. y(θ)) Σθ ˆ ˆ ˆ · ψ dΣ ˆ. y(θ))∂n y(θ) ˆ ˆ ˆ + HS(θ. Σθ ˆ Σθ ˆ ˆ In this case the gradient of j(θ) is only supported on Γθ with a density gθ ˆ ˆ . ψ = Qθ ˆ y F (θ. y(θ)) ˆ ˆ · ψ d Qˆ + θ y S(θ. y(θ)) · δθ d Qθ + ˆ Σθ ˆ ˆ ˆ F (θ. ˆ From now on. p(θ)) (6. θ ∀ ψ ∈ H(Qθ ) ˆ We then use the following identity ˆ ˆ ˆ ˆ ˆ A∗ p(θ). φ = L(δθ.) ∈ H(Qθ )∗ . ˆ ˆ 3. Qθ ˆ ˆ ˆ F (θ. We define the element ˆ ˆ ˆ ˆ p(θ) ∈ H(Qθ )∗ as the solution of the following adjoint problem. p(θ)) + + Σθ ˆ Qθ ˆ ˆ ˆ F (θ. nθ ˆ ˆ ˆ ˆ ˆ ˆ ˆ S (θ. ˆ ˆ 2.32) ˆ ˆ This allows us to conclude that the derivative of j( θ) with respect to θ has the following structure. nθ ˆ REMARK 6. nθ dt dΓθ . let us assume that the partial shape derivative y (θ)δθ is the unique solution of the following linear tangent problem. y(θ)) δθ. y(θ)) · δθ + ∂n S(θ. L(δθ. y(θ)) · δθ d Σθ = ˆ ˆ ˆ s(θ. y (θ) · δθ = Ay (θ) · δθ. H(Qθ )∗ )) and L(δθ. ˆ ˆ ˆ S (θ. nθ dt d Γθ .

nθ dt d Γθ ˆ ˆ (6. y(θ)) ˆ ˆ ˆ + HS(θ. y) ˆ (6. ∈θ : e(θ. in H(D)∗ (6.33) + 6. Σθ ˆ ˆ EΓ (θ.Tube derivative in a Lagrangian setting such that the following identity holds. δθ · nθ ˆ ˆ ˆ ˆ ˆ ˆ (p(θ)) + s(θ. The design variables are the Lagrange mapping and a state variˆ ˆ able (θ. In this paragraph.35) with ˆ j(θ. y) ∈ θad × H(D) related by an abstract state variable ˆ e(θ. ˆ j (θ) · δθ = = 0 Γθ ˆ y S(θ. ˆ ∀ φ ∈ H(D) . . φ) tinuous on H(D)..37) EΩ (.. y) d Σθ ˆ (6. in H(D)∗ ˆ 8 < (θ.34) with respect to the mapping θ ∈ Θad .36) and ˆ e(θ.. y(θ)) ˆ ˆ δθ. φ) where we suppose that the applications (. Hence we would like to solve the following minimization problem. we are interested by optimization problems involving state constraints. y) = 0.34) This setting allows to avoid the derivation of the state variable y(θ) solution ˆ of the implicit equation (6.. This gradient will be involved in the first-order optimality conditions of ˆ the minimization problem for j(θ). y) d Qθ + ˆ Σθ ˆ ˆ S(θ. y) = Qθ ˆ ˆ F (θ. . y) ˆad × H(D). y.5 Shape optimization under state constraints In the previous section. y.. ˆ min j(θ. y) = 0. y(θ)) + ∂n S(θ. y(θ))∂n y(θ) T 0 T 173 ˆ g(θ). the structure of its gradient with respect to the moving domain. . This general functional involved distributed and boundary terms and a state function ˆ y(θ) ∈ H(Qθ ) that depends on the moving domain Qθ and its parametrization ˆ ˆ ˆ θ.2. φ) −→ are linear conEΓ (. φ) d Qθ + ˆ (6. φ = Qθ ˆ ˆ EΩ (θ. y). we have established for a general functional.. φ) d Σθ .

35) is equivalent to finding the saddle point ¯ ¯ ˆ ¯ (θ. such that the following ¯ ˆ first-order optimality conditions hold.41) ˆ max L(θ.e.174 Moving Shape Analysis In order to derive the first-order optimality conditions for problem (6. φ) · δθ + Σθ ˆ ˆ ˆ ˆ ˆ F (θ.40) (6..35) admits a local solution (θ. ¯ ¯ ˆ ¯ L(θ. ¯ Then there exists a Lagrange multiplier φ ∈ H(Qθ )∗ . y) − e(θ. solving problem (6. The local solution (θ. y. y) · δθ + EΓ (θ. δy = y F (θ. y) + EΓ (θ. y. y) + EΩ (y.39) We shall now derive the explicit form of the optimality conditions for problem (6. y) Σθ ˆ ˆ y EΓ (θ. y) Qθ ˆ ˆ · δy + · δy + y EΩ (θ. . y) ∈ Θad × H(D) PROPOSITION 6. y. φ) ˆ and ˆ Lθ . ¯ ˆ ¯ e(θ. y. ˆ ˆ ˆ L(θ.2 Let us assume the following points. Fr´chet differentiable in a neighborhood of point (θ.42) ˆ y S(θ.y) (θ. y. i. φ) ˆ · δy dQθ + ˆ · δy dΣθ (6.35). y) · δθ + EΩ (θ. ¯ ¯ ˆ ˆ min ˆ (θ. y. θ. φ) = 0 ˆ (6. y.39). Using differentiability results of j(. that the linear tangent ¯ ˆ ¯ application e ((θ. we introduce the Lagrangian functional. y ) and the functional j(θ. y ) is a regular point. ˆ Ly (θ. δ θ = ˆ ˆ ˆ F (θ. ¯ ˆ ¯ ˆ 1. φ)) · δθ. φ).). y . y . φ H(D)∗ . ¯ e ˆ θ ¯ ˆ ¯ 2. φ) + (H I +∂n ) · (S(θ. y ) = 0. φ) · δθ d Qθ ˆ + Σθ ˆ Qθ ˆ ˆ ˆ S (θ.. y). the following identity comes easily. φ) φ ∈ H(D)∗ (6. nθ d Σθ ˆ ˆ .H(D) (6.38) Consequently. y )) is surjective. Problem (6. φ) ∈ Θad × H(Qθ ) × H(Qθ )∗ of the following Min-Max problem. φ) = j(θ. y. y) is ¯ ˆ y ) ∈ Θad × H(Q ¯).

y) · δθ + EΓ (θ. . y .44) 6. Σθ ˆ δθ. φ) · δθ ¯ ¯ ¯ ¯ ˆ ¯ ˆ ¯ ˆ ¯ F (θ.Tube derivative in a Lagrangian setting 175 ¯ ˆ ¯ ¯ At a regular point (θ. φ) ¯ ˆ ·ψ+ ¯ ˆ ¯ · ψ dQ ˆ θ ¯ y EΓ (θ. φ) · 2.7 1. y . y ) (6. ˆ ˆ ˆ S (θ. y ) · δθ + EΓ (θ. nθ dΣθ . y . Lagrangian non-cylindrical derivative We shall now ask a fundamental question concerning the link between the derivative using perturbations of the identity and Eulerian derivative introduced in [59]. ¯ y F (θ. y .6 Eulerian vs. y ) + EΩ (θ. y . y . nθ dΣθ = 0 ˆ ˆ (6. y ) + EΓ (θ. φ) · δθ dΣθ = ˆ ˆ ˆ s(θ. φ) + (H I +∂n ) · (S(θ. y ) Qθ ¯ ¯ ˆ ·ψ+ ¯ y EΩ (θ. ¯ ˆ j (θ) · δθ = ¯ ¯ ¯ ¯ ¯ ¯ ˆ ¯ ˆ ¯ ˆ ¯ ˆ ¯ s(θ. φ)) δθ. y) · δθ + EΩ (θ. Qθ ˆ Let us add the following hypothesis. y ) + eΓ (θ. ˆ ˆ F (θ. φ) ¯ ˆ ¯ · ψ dΣ ˆ = 0. φ) · δθ dQθ = 0.2. y ) Then the optimality condition reduces to ¯ ˆ ˆ j (θ. ˆ ˆ Σθ ˆ In this case.43) + Σθ ¯ + Σθ ¯ ¯ ¯ ˆ ¯ +EΓ (θ. φ) + F (θ. δ θ) = Qθ ¯ ¯ ¯ ¯ ˆ ¯ ˆ ¯ F (θ. y ) + EΩ (θ. and its representation has ¯ ˆ the following form. y. φ) · δθ dQθ ˆ ¯ ¯ ¯ ˆ ¯ ˆ ¯ S (θ. there exists a Lagrange multiplier φ solution of the following linear tangent adjoint equation. y. y . y ). φ) Σθ ¯ ¯ ¯ ¯ ˆ ¯ ˆ ¯ + (H I +∂n ) · (S(θ. y . y ) · δθ + EΩ (θ. y. φ)) · δθ. the gradient is only supported on Σ θ . nθ dΣθ ˆ ˆ =0 REMARK 6. y) + eΓ (θ. θ ∀ψ ∈ H(D) + Σθ ¯ ¯ y S(θ. y .

J : A −→ R Q → J(Q) with A ≡ Q ∈ ET . Wθ = ∂t (δθ ◦ θ) ◦ θ−1 − D Vθ · δθ and the transverse field ˆ ˆ ˆ involved in the Eulerian derivative is given by Zt = δθ.45) Let us assume that this functional can be parametrized by either the Laˆ grangian mapping parametrizing the tubes in A with j = J ◦ θ such that j : Θad −→ R ˆ → j (θ) = J(Q ˆ) ˆ θ θ (6. ˆ Ω0 ˆ ˆ θ + ρθ ↓ Ωθ+ρδθ ˆ ˆ The structure of the perturbation rules is similar in the case of an Eulerian −→ ˆ θ Ωθ ˆ I +ρδθ . PROOF As shown here. we deal with a first-order perturbation of the identity inside Ω θ .176 Moving Shape Analysis Let us consider a functional J.4 ˆ The differentiability of the functional j at point θ ∈ Θad in the direction def ˆ ˆ δ θ = δθ ◦ θ ∈ Θ is equivalent to the differentiability of je with respect to V ∈ Vad in the direction W ∈ V .46) (6. in the Lagrangian mapping case.48) V → je (V ) = J(Q(V )) The following result holds true.47) or the velocity field V ∈ V building the tubes in A with j V = J ◦ Tt (V ) such that je : V −→ R (6. Q⊂D (6. ˆ ˆ ˆ ˆ δ θ = δθ ◦ θ ∈ Θ def (6. THEOREM 6. Wθ . δθ = je (Vθ ).49) ˆ ˆ ˆ ˆ with Vθ = ∂t θ ◦ θ−1 . Furthermore the respective functional derivatives are related by the following relation. ˆ j (θ).

we have ˆ ˆ ˆ ˆ ∂t (δθ ◦ θ) ◦ θ−1 = ∂t δθ + D δθ · ∂t θ ◦ θ−1 = ∂t δθ + D δθ · Vθ ˆ As stated before. there exists V ˆ = ∂t θ ◦ θ1 ∈ V such that θ = Tt (V ˆ). the derivative of the mapping T t (V ) with respect to V is d Tt (V ) · W = Zt ◦ Tt (V ) dV (6. the transverse map Tρt is the flow of a transverse velocity t field Zρ . For a given ˆ ∈ Θad .Tube derivative in a Lagrangian setting description using velocity fields. ˆ ˆ ˆ Wθ = ∂t (δθ ◦ θ) ◦ θ−1 − D Vθ ◦ δθ ˆ ˆ Indeed.53) .52) ˆ def ˆ We then set δ θ = δθ ◦ θ = Zt ◦ Tt (V ). T ) Zt=0 = 0 Using chain rule it is possible to deduce the following identity. ˆ ˆ ˆ Lagrangian map θ θ θ def ˆ ˆ Hence if we perturbed the Lagrangian map in the direction δ θ = δθ ◦ θ. we generate a velocity field perturbation Wθ given by the following expression. which means that δθ = Zt .55) (6.50) As recalled above.54) (6. (0. je (V ) · W = d d d j (Tt (V )) · W = j (Tt (V )) · ( Tt (V ) · W ) ˆ dV dV dθ (6. ∂t Zt + D Zt · V − D V · Zt = W. using the chain rule. Using the transverse equation we define Wθ such that ˆ ∂t δθ + D δθ · Vθ = Wθ + D Vθ · δθ ˆ ˆ ˆ from which we deduce that ˆ ˆ Wθ = ∂t (δθ ◦ θ) ◦ θ−1 − D Vθ · δθ ˆ ˆ (6. This velocity field is written Zt for ρ = 0 and it satisfies the following Cauchy problem.51) (6. we can identify the transverse velocity field Z t with δθ. Ω0 Tt (V + ρW ) ↓ Ω(V + ρW ) Tt (V ) 177 −→ Ω(V ) Tρt = Tt (V + ρW ) ◦ Tt (V )−1 def As proved in [59].

The ˆ couple (u. ˆ THEOREM 6.5 Let us assume that We consider a viscous incompressible newtonian fluid inside a moving domain Ωθ . H 2 (Ωθ )) ∩ L∞ (0. Using results obtained in the first section.178 Moving Shape Analysis 6. Its evolution is described by its velocity u and its pressure p. This approximation is valid around a fixed domain ˆ configuration. we deal with an optimal control problem for the Navier-Stokes ˆ system where the control is the displacement of the moving domain θ. Then.56) Σθ  u = Vθ . i. then the solution (uθ . Furthermore.3 Navier-Stokes equations in moving domain In this section. p) satisfies the classical Navier-Stokes equations written in nonconservative form. around θ = I.   ∂t u + D u · u − ν∆u + p = 0. The main result of this paragraph is the following. 6. We will first obtain a differentiability result concerning the sensitivity of the solution of Navier-Stokes equations in a moving domain with ˆ respect to the Lagrangian mapping θ ∈ Θad . pθ ) of the Navier-Stokes system (6.3.1 Transpiration boundary conditions A kinematic continuity boundary condition is satisfied on Γ θ .56) admits a mateˆ ˆ rial shape derivative in L2 (0. Ω0 m ˆ ˆ ˆ • The mapping θ ∈ Θ is such that Vθ = ∂t θ ◦ θ−1 ∈ H 1 (0. this means that ˆ ˆ ˆ the fluid velocity is equal to the velocity of the moving boundary V θ = ∂t θ◦ θ−1 ˆ on Γθ . in order to solve two questions related to the Navier-Stokes equations in a moving domain: 1. T . We recover early results obtained using Eulerian field derivative methods. T . Qθ ˆ   div(u) = 0. we shall apply the previous non-cylindrical shape derivative concepts.e. H0 (D)) with ˆ m ≥ 5/2. 2. we derive the structure of the gradient for a general cost functional. Qθ ˆ (6. ˆ ˆ   u(t = 0) = u0 . ˆ ˆ .. H 1 (Ωθ )). This result allows us to justify in a rigorous manner transpiration boundary conditions [63] in fixed domain as a first order approximation of the Navier-Stokes system in moving domain. T .

[57].   ∂t u0 + D u0 · u0 − ν∆u0 +   div(u0 ) = 0. ˆ ˆ  u = ∂t (δθ ◦ θ) ◦ θ−1 − D u · δθ. T ) Ω0 × (0. i. T ).Tube derivative in a Lagrangian setting 179 ˆ ˆ (uθ . COROLLARY 6.  0  u0 = 0. on Γθ ˆ (6. p0 ) defined in a fixed domain Ω0 × (0. p (θ) · δθ) solution of the ˆ ˆ following linear tangent system. D p0 · δθ) + (u .57) PROOF We follow the proof used in [59].1 Considering a reference flow (u0 . with θ = I. Qθ ˆ Qθ ˆ Σθ ˆ Ω0 (6.e.e. pθ ) admits a partial shape derivative (u (θ) · δθ. p ) satisfies the .   u (0) = 0. We shall detail the computation of the linearized boundary conditions. i. satisfying the following Navier-Stokes system.58) Applying this theorem in the particular case.59) The first order variation (δu.. p0 ) of problem (6. We ˆ perturbed the Lagrangian mapping with the increment ρδ θ and we get the following perturbed boundary conditions written on the fixed boundary Γ 0 . δp) = (D u0 · δθ.  u (t = 0) = u0 .59) exists and (u . d ˆ ˆ ˆ ˆ u(θ + ρδ θ) ◦ (θ + ρδ θ) dρ d ˆ ˆ u(θ + ρδ θ) dρ = ρ=0 d ˆ ˆ (∂t (θ + ρδ θ)) dρ .   ∂t u + D u · u + D u · u − ν∆u +   div(u ) = 0. leads to the derivation of transpiration boundary condition for linearized system. T ) Ω0 (6.. p ) around this flow of the solution (u0 . The only difference is that we do not need to use the transerve field Zt . We refer to [13] for the details in case of homogeneous Dirichlet boundary conditions. ρ=0 on Γ0 ρ=0 ˆ ˆ ˆ ˆ ˆ ◦ θ + (D u(θ)) ◦ θ · δ θ = ∂t δ θ ˆ ˆ ˆ ˆ ˆ (u (θ) · δθ) ◦ θ + D u(θ) · δθ ◦ θ = ∂t (δθ ◦ θ) then we get ˆ ˆ ˆ ˆ u (θ) · δθ = ∂t (δθ ◦ θ) ◦ θ−1 − D u(θ) · δθ. 0 p0 = 0. where the reference flow is ˆ considered in a fixed domain. T ) Γ0 × (0. Ω0 × (0. p = 0.

p(θ)) with (u(θ). uθ . We only need to check the boundary conditions.2 Optimal control of the Navier-Stokes system by moving the domain In this paragraph. ˆ We set θ = I and we get ˆ ˆ u = ∂t (δθ ◦ θ) ◦ θ−1 − D u0 · δθ. Γ0 × (0.3.60) ˆ ˆ  u = ∂t (δθ ◦ θ) ◦ θ−1 − (D u0 · n) δθ. We ˆ are interested in solving the following minimization problem : ˆ min j(θ) ˆ θ∈U (6.64) . In the general case. we have ˆ ˆ ˆ ˆ u (θ) · δθ = ∂t (δθ ◦ θ) ◦ θ−1 − D u(θ) · δθ. We then use D u0 · δθ = (DΓ u0 + D u0 n ⊗ n) · δθ But u0 = 0 on Γ0 .5). T )   u (0) = 0.  0 0  ∂t u + D u · u + D u · u − ν∆u + p = 0.180 Moving Shape Analysis following linear tangent system written in the reference fixed domain. v L2 (Qθ ) ˆ + u. Ω0 × (0. pθ ). we shall deal with an optimal control for the NavierStokes system. n . then DΓ u0 = 0 on Γ0 . Γ0 × (0. We choose to minimize a general cost functional based on ( θ. p) is a real functional of the following form : ˆ Jθ (u. B ∗ v L2 (Qθ ) ˆ = BΣ u. T ) Γ0 × (0. T ) on Γθ ˆ (6.61) 6.56) and Jθ (u. B u. p) = ˆ α Bu 2 2 L2 (Qθ ) ˆ + γ K Vθ ˆ 2 2 L2 (Σθ ) ˆ (6. T ) (6. p(θ)) is a weak solution of problem ˆ (6.62) ˆ ˆ ˆ ˆ ˆ where j(θ) = Jθ (u(θ). n . v L2 (Σθ ) ˆ (6.63) where B ∈ L(H. from which we deduce ˆ ˆ u = ∂t (δθ ◦ θ) ◦ θ−1 − (D u0 · n) δθ. L2 (Qθ )) is a general linear differential operator satisfying the ˆ following identity. Ω0 PROOF This a direct consequence of theorem (6. where the control is the displacement of the domain boundary ˆ ˆ ˆ Γθ . Ω0 × (0. T )   div(u ) = 0.

K∗ v L2 (Σθ ) ˆ = KΣ u.2). we perform an integration by parts. v L2 (Σθ ) ˆ + u. π) · nθ + αBΣ B u) · D u − γ(−K∗ K + KΣ K) Vθ · DΓ Vθ ˆ ˆ ˆ 1 + (6. we can state ˆ j (θ). ˆ j(θ) = − ∂t Eθ − (divΓ Vθ ) Eθ − D Eθ · Vθ ˆ ˆ ˆ ˆ ˆ − (−σ(ϕ.Tube derivative in a Lagrangian setting 181 and K ∈ L(U.   −∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + π = −α B ∗ B u. δθ L2 (Σθ ) ˆ ˆ ˆ 2 Using the adjoint operators. the functional j(θ) possesses ˆ ˆ which is supported on the moving boundary Γ ˆ and can be a gradient j(θ) θ represented by the following expression. ∂t (δ θ) ◦ θ−1 − DΓ Vθ · δθ L2 (Σθ ) ˆ ˆ ˆ 1 + αBΣ B u. ˆ THEOREM 6.1) and (6. u ˆ ˆ (−K∗ K + KΣ K) Vθ . δθ = dis − α B ∗ B u. we obtain the differentiability of j( θ) in the def ˆ ˆ perturbation direction δ θ = δθ ◦ θ ∈ Θ. π) stands for the adjoint fluid state solution of the following system. Qθ ˆ   div(ϕ) = 0.65) The main difficulty in dealing with such a minimization problem is related to ˆ the fact that integrals over the domain Ωθ depend on the control variable θ. Qθ ˆ (6. K(∂t (δ θ) ◦ θ−1 − DΓ Vθ · δθ) L2 (Σθ ) ˆ ˆ ˆ 1 2 2 + α |Bu| + γ |KVθ | nθ . ΩT and Eθ = −σ(ϕ. L2 (Σθ )) is a general linear differential operator satisfying the ˆ following identity.66) α |Bu|2 + γ |KVθ |2 nθ ˆ ˆ 2 where (ϕ.6 def ˆ ˆ ˆ For Vθ = ∂t θ ◦ θ−1 ∈ U and Ω0 of class C 2 . δθ L2 (Σθ ) ˆ ˆ ˆ ˆ 2 L2 (Qθ ) + γ ˆ . u L2 (Σθ ) + α |Bu|2 + γ |KVθ |2 nθ . ˆ   ϕ(T ) = 0. v L2 (Σθ ) ˆ (6.68) ˆ Using chain rule. δθ = α Bu. Bu L2 (Qθ ) ˆ ˆ ˆ + γ KVθ . π) · nθ + γ [−K∗ K + KΣ K] Vθ + αBΣ B Vθ ˆ ˆ ˆ ˆ PROOF (6. K u. ˆ j (θ).67) Σθ  ϕ = 0. Using theorems (6.

DΓ Vθ · δθ L2 (Σθ ) ˆ ˆ ˆ 1 2 2 + α |Bu| + γ |KVθ | nθ . ∂t (δ θ) ◦ θ−1 ˆ ˆ − γ(−K K + KΣ K) Vθ . ˆ ˆ ˆ j (θ). π) · nθ + αBΣ B u + γ (−K∗ K + KΣ K) Vθ and we use ˆ ˆ ˆ theorem (6. we deduce that ˆ ˆ u = ∂t (δθ ◦ θ) ◦ θ−1 − D u · δθ.8) and we deduce the following identity.3) and we obtain ˆ j (θ). [∂t u + D u · u + D u · u − ν∆u + p ]ϕ − π] u − π div u Qθ ˆ Qθ ˆ = Qθ ˆ [−∂t ϕ − D ϕ · u + ∗ D u · ϕ − ν∆ϕ + + Σθ ˆ p div ϕ Qθ ˆ [p n · ϕ − νϕ · ∂n u + νu · ∂n ϕ − u · π n] (6. δθ = −∂t Eθ − (divΓ Vθ ) Eθ − D Eθ · Vθ ˆ ˆ ˆ ˆ ˆ − (−σ(ϕ.83) satisfied by u . π) solution of the adjoint problem (5. δθ = −σ(ϕ. we can reciprocally deduce the expression of one of the directional derivatives from another.4). π) · nθ + αBΣ B u) · D u − γ(−K∗ K + KΣ K) Vθ · DΓ Vθ ˆ ˆ ˆ 1 + α |Bu|2 + γ |KVθ |2 nθ . Let us apply this equivalence on the optimal control treaten previously. We have established the structure of the Lagrangian derivative of the func- . u L2 (Qθ ) ˆ = − σ(ϕ.3 Comparison with the Eulerian derivative version Using theorem (6. then.3.69) Hence we get −α B ∗ B u. π) · nθ + αBΣ B u + γ(−K∗ K + KΣ K) Vθ .182 Moving Shape Analysis We now define the fluid adjoint state (ϕ. δθ L2 (Σθ ) ˆ ˆ ˆ 2 from which we deduce the expression of the gradient ˆ j( θ).70) Using the linearized system (5. u ˆ L2 (Σθ ) ˆ (6. D u · δθ ˆ ∗ L2 (Σθ ) ˆ 6. π) · nθ + αBΣ B u. L2 (Σθ ) ˆ on Σθ ˆ − −σ(ϕ. π) · nθ . δθ L2 (Σθ ) ˆ ˆ ˆ 2 We now set Eθ = −σ(ϕ.

D u · Zt ˆ − γ(−K∗ K + KΣ K) Vθ . Wθ ˆ ˆ ˆ ˆ + γ(−K K + KΣ K) Vθ . π) · nθ + αBΣ B u ˆ ˆ ˆ ˆ + γ(−K∗ K + KΣ K) Vθ . nθ ˆ ˆ 2 L2 (Σθ ) ˆ L2 (Σ(Vθ )) ˆ L2 (Σ(Vθ )) ˆ We recall that D Vθ · Zt = DΓ Vθ · Zt + (D Vθ · nθ ) Zt . nθ ˆ ˆ Using u = Vθ . Wθ ˆ ˆ ˆ ˆ + D Vθ · Zt L2 (Σ(Vθ )) ˆ ˆ − −σ(ϕ. Wθ = j (Tt (Vθ )). on Γθ we deduce that D u·Zt = DΓ Vθ ·Zt +(D u·nθ ) Zt . Zt . nθ L2 (Σ(Vθ )) ˆ ˆ ˆ ˆ ˆ 1 + α |Bu|2 + γ |KVθ |2 . nθ ˆ ˆ ˆ ˆ ˆ D u · Zt = DΓ u · Zt + (D u · nθ ) Zt . DΓ Vθ · Zt ˆ ˆ 1 + α |Bu|2 + γ |KVθ |2 . δθ = Zt and ∂t (δθ ◦ θ) ◦ θ−1 = Wθ + D Vθ · δZt . Zt . ∂t (δ θ) ◦ θ−1 ˆ ˆ − γ(−K K + KΣ K) Vθ . we get ˆ ˆ ˆ j (Tt (Vθ )). π) · nθ + αBΣ B u + γ(−K∗ K + KΣ K) Vθ . Zt . ˆ ˆ ˆ j (θ). nθ ˆ ˆ 2 L2 (Σθ ) ˆ + . π) · nθ + αBΣ B u + γ(−K∗ K + KΣ K) Vθ . Zt = −σ(ϕ. nθ ˆ ˆ ˆ ˆ ∗ L2 (Σθ ) ˆ on Γθ . Zt = −σ(ϕ.Tube derivative in a Lagrangian setting ˆ tional j(θ). DΓ Vθ · δθ L2 (Σθ ) ˆ ˆ ˆ 1 2 2 + α |Bu| + γ |KVθ | nθ . ˆ σ(ϕ. Zt = −σ(ϕ. ˆ ˆ ˆ then je (Vθ ). π) · nθ + αBΣ B u. π) · nθ − αBΣ B u · D u · nθ ˆ ˆ 1 α |Bu|2 + γ |KVθ |2 . D Vθ · nθ Zt . nθ . ˆ ˆ ˆ ˆ ˆ then we deduce j (Tt (Vθ )). δθ = −σ(ϕ. This means that D Vθ ·nθ = 0. D(Vθ − u) · nθ Zt . π) · nθ + αBΣ B u. π) · nθ + αBΣ B u + γ(−K∗ K + KΣ K) Vθ . D u · δθ ˆ ∗ L2 (Σθ ) ˆ 183 L2 (Σθ ) ˆ ˆ ˆ ˆ Let us set θ = Tt (Vθ ). nθ L2 (Σ(Vθ )) ˆ ˆ ˆ 2 We choose a canonical extension Vθ ◦p. π) · nθ + αBΣ B u. δθ L2 (Σθ ) ˆ ˆ ˆ 2 − −σ(ϕ. Wθ ˆ ˆ + L2 (Σ(Vθ )) ˆ L2 (Σ(Vθ )) ˆ −σ(ϕ.

73) ˆ ˆ ˆ ˆ ˆ ˆ where (ϕ. T ) ˆ ˆ ˆ ˆ ˆ λθ (T ) = 0. −∂t λθ − Γ λθ · Vθ − div(Vθ ) λθ = f. (6.72) . ΓT (V ) ˆ with f = σ(ϕ. π) stands for the adjoint fluid state.184 Moving Shape Analysis This is exactly the expression obtained in [57]. (0. π) · nθ + α BΣ B u + γ −K∗ K Vθ + KΣ K Vθ (6. π) · nθ − αBΣ B u · D u · nθ + ˆ ˆ then je (Vθ ) = −λθ nθ − σ(ϕ. Let us define the adjoint transverse boundary field λθ solution of the following ˆ tangential dynamical system.71) 1 α |Bu|2 + γ |KVθ |2 ˆ 2 (6. using Eulerian methods.

Chapter 7 Sensitivity analysis for a simple fluid-solid interaction system 7. we perform derivation of the Lagrangian with respect to state variables. For the sake of simplicity.2. we recall early results concerning the well-posedness of the coupled fluid-structure system. The aeroelastic system consists of an elastically supported rigid solid moving inside an incompressible fluid flow in 2-D.5. The chapter is organized in four parts : • In section 7. • In section 7. we only consider one degree of freedom for the structural motion : the vertical displacement d(t)e2 where e2 is the element of cartesian 185 .4. we introduce the notations and the mechanical system we shall deal with and we state the main result of this chapter. namely the structure of a cost functional gradient with respect to inflow boundary conditions pertubations. the keystone of the sensitivity analysis of coupled fluid-structure model is based on the moving shape analysis introduced in the previous chapters. • In section 7. The proof is given in the remaining part of the chapter.3. • In section 7. As described in the introduction of this lecture note. 7. we introduce the minimization problem and its associated Lagrangian functional.1 Introduction This chapter deals with the analysis of an inverse boundary problem arising in the study of bridge deck aeroelastic stability. This allows us to obtain the structure of the adjoint variables involved in the cost functional gradient.2 Mathematical settings We consider a two dimensionnal flexible structure in rigid motion.

1 [147] Assuming that d(.186 Moving Shape Analysis Tt(V )(. i. Ωf = Tt (V )(Ωf ). e2 . Without loss of generality. as described in the following lemma : LEMMA 7. the analysis of the coupled problem is set in Ω × (0. We consider a control volume Ω ⊂ R2 containing the solid for every time t ≥ 0. x0 ) ≡ Tt (V )(x0 ) .) is smooth enough. Hence. Qf ≡ t 0<t<T ({t} × Ωf ) and Σf ≡ Γf × (0. t 0 Ωs = Tt (V )(Ωs ) t 0 where Tt (V ) is the solution of the following dynamical system : Tt (V ) : Ω0 −→ Ω x0 −→ x(t. The structure is surrounded by a viscous fluid in the plane (e 1 . we define a map Tt ∈ C 1 (Ω0 ) such that Ωf = Tt (Ωf ).1: Arbitrary Euler-Lagrange map basis (e1 . there exists a vector field V that builds Q .e. we choose the reference configuration to be the physical configuration at initial time Ω(t = 0). we introduce a diffeomorphic map sending a fix reference domain Ω0 into the physical configuration Ω at time t ≥ 0. we write Ωs = Ωs + d(t)e2 t 0 We set Σs ≡ 0<t<T ({t} × Γs ).) Γo Ωf 0 Ωs 0 Γs 0 Γs t Γo Ωs t Ωf t V∞ V∞ FIGURE 7. Hence. e2 ). T ) t ∞ ∞ The map Tt can be actually defined as the flow of a particular vector field. In order to deal with such a coupled system. T ) where T > 0 is an arbitrary time. e3 ) in R3 . t 0 Ωs = Tt (Ωs ) t 0 Since we only consider one degree of freedom motion.

j .Sensitivity analysis for a simple fluid-solid interaction system with dx = V (τ. t The fluid is assumed to be a viscous incompressible newtonian fluid. The solid is described by the evolution of its displacement and its velocity ˙ and the couple (d. t We complete the whole system with kinematic continuity conditions at the fluid-structure interface Γs : t ˙ u = V = d e2 . we can give an example of an appropriate flow vector field :  ˙ x ∈ Ωs  V (x. Hence. p) = −p I +ν(D u + ∗ D u) stands for the fluid stress tensor inside Ωf .   u(t = 0) = u0 . k) stand for the structural mass and stiffness.2) V (x. t) = d(t)e2 . x ∈ Ωf t  V (x. d) is the solution of the following ordinary second order differential equation : ¨ m d + k d = Ff . in Ω0 187 (7. with (D u)ij = ∂j ui = ui. F f is the projection of the fluid loads on Γs along the direction of motion e2 . p) satisfies the classical Navier-Stokes equations written in non-conservative form:   ∂t u + D u · u − ν∆u +   div(u) = 0. d1 ] (7. The couple (u. ˙ d. T ] dτ x(τ = 0) = x0 . d (t = 0) = [d0 .3) In our simple case. The map Tt is 0 0 usually referred to as the Arbitrary Euler-Lagrange map.5) Γs t where σ(u. t) = Ext(d(t)e2 ). sur Σs (V ) (7.  u = u∞ . Qf (V ) Qf (V ) Σf ∞ Ωf 0 (7. x(τ )).4) where ν stands for the kinematic viscosity and u∞ is the farfield velocity field. p) · n · e2 (7.6) . τ ∈ [0. x ∈ Γf ∞ where (m. the projected fluid loads Ff have the following expression : Ff = − σ(u. Its evolution is described by its velocity u and its pressure p.1) where Ext is an arbitrary extension operator from Γ s into Ωf . t) · n = 0. p = 0. t ˙ (7.

(0. (0. T )    Γs t     u.12) ˙ −∂t λ − Γ λ · V = f.    div(u) = 0. d ) = 2 T 0 To summarize.14) Γs (V ) t Γs (V ) t . q = 0. Qf (V ) Qf (V ) Σs (V ) Σf ∞ Ωf T (7. we get the following coupled system :  Qf (V )  ∂t u + D u · u − ν∆u + p = 0. d. d. b1 . p) · n · e2 . πu∗ ) · n)|Σf + γ u∗ = 0 ∞ ∞ ∞ ∞ ∞ with (ϕ. ˙ Ω f × R2 0 (7.     ϕ(T ) = 0. π.  ∗  −∂t ϕ − D ϕ · u + ( Du) · ϕ − ν∆ϕ +   div(ϕ) = 0.  ϕ = 0. u=d Σs (V )   ¨ md + k d = −  σ(u. Γ T (d e 2 ) ˙ ˙ with f = −d b˙2 + ν (D ϕ · n) · (D u · n) − |d|2 (D ϕ · e2 ) · e2 and ˙ λ n = −b1 − m b˙2 − α(d − d2 ) e2 + g σ(φ. d0 . d.7) (7.   ϕ = b 2 e2 . j(u∗ ) = (σ(ϕu∗ . Qf (V )     u = u∞ .11) −b˙1 + k b2 = α(d − d1 ). d (u∞ )) with u. d (u∞ ) is a weak solution γ ˙ |d − d1 |2 + |d − d2 |2 + u∞ g g 2 2 Uc (7. p.8) ˙ ˙ where j(u∞ ) = Ju∞ ( u.13) (7. p. T ) g b1 (T ) = 0.188 Moving Shape Analysis Main Result: For u∞ ∈ Uc the following minimization problem. d (t = 0) = [u0 . p. Σf  ∞   ˙ e2 . d.7) and Ju∞ is a real functional of the following form : α ˙ Ju∞ ( u. Σs (d e2 ) ˙ λ(T ) = 0.9) admits at least a solution u∗ which satisfies the following necessary first-order ∞ optimality conditions.10) (7. d1 ] . (7. π) · n (7. min j(u∞ ) u∞ ∈ U c of problem (7. b2 ) are solutions of the following adjoint system.

Ωs ) > 0. with d ∈ W 1. T0 ). ∞ 0 iii) u0 ∈ (L2 (Ωf ))2 u∞ ∈ (H m (Σf ))2 m > ∞ 0 div(u0 ) = 0.2).7). [68]. [133]. R) ˙ (u.7). b2 (T ) = [0. We refer to [80] and [81] for a complete review. T ].∞ (0. T . R) ∩ C 0 ([0.1 We can eliminate the auxiliary adjoint variables (λ. there exists at least one weak solution to the initial-boundary value problem (7. we deal with nonhomogeneous Dirichlet boundary conditions at the farfield fluid boundary. π) · n · e2  g g   Γs (V )  t   ˙ |d|2 (D ϕ · e2 ) · e2 − ν (D ϕ · n) · (D u · n) · n (0. 0 (7.Sensitivity analysis for a simple fluid-solid interaction system 189 REMARK 7.15) 7. b 2 ) . 0] ˙ (7. π. a. Vd(. Ω0 are of class C 2 . (7. dans Ωf 0 u 0 = d 1 e2 . [54]. d) ∈ L2 (0.3 Well-posedness of the coupled system We are interested in recalling global existence results for weak solutions to the initial boundary value problem (7.16) then there exists a positive real time T0 = T0 (u0 . we only have one degree of freedom for the solid motion and we are dealing with two dimensional Navier-Stokes equations. T ) + s   Γt (V )       b2 . [81].1 Assume the following hypothesis : ii) a = dist(Γf . We then replace equations (7. Nevertheless. b 1 ). Hd(. One should expect some restrictions on the existence time for the solutions since depending on the data. Ω0 ) such that for any 0 T ∈ (0. T . [39].    ¨ ˙ ˙  m b2 + k b2 = α(d − d1 ) − α(d − d2 ) + ∂t σ(φ.) ) .14) by the following second order ODE. THEOREM 7. sur Γs 0 3 4 i) Ωs . the solid may vanish outside the control volume Ω.) ) ∩ L∞ (0. This problem was recently investigated by several authors [131].13) . Ωs . In our case. (7. in order to get a system only involving the adjoint variables (ϕ. T .

T . Γf .2 Using results from Fursikov et al [72]. (L2 (Γf ))2 ) ∞ ∞ 1 3 1 1 u∞ · n ∈ L2 (0. d (u∞ ) is a weak solution of problem (7. d.4 Inverse problem settings min j(u∞ ) u∞ ∈ U c (7. supp(v) ⊂ Ωf t t and satisfies the following identity : T − + 0 T 0 Ωf t ˙ u · ∂t v + md ˙ − k d Ωf 0 Ωf t [(D u · u) · v + ν D u · · D v] = md1 (0) + 1 u0 · v(0) (7. T ]. ∞ v|Ωs = · e2 . T .19) . T . ∞ v|Ωs = · e2 . p.18) We are interested in solving the following minimization problem : ˙ ˙ where j(u∞ ) = Ju∞ ( u. v · n = 0. ) ∈ (H 1 (Ω0 ))2 × R. (H 2 (Γf ))2 ) ∩ H 4 (0.7) and Ju∞ is a real functional of the following form : α ˙ Ju∞ ( u.) ) with v(T ) = ˙(T ) = 0 REMARK 7. v = 0. p. (H −1 (Γf ))2 ) ∞ ∞ 7.17) ∀ (v. T . and Vd(t) ≡ (v. d (u∞ )) with u. d. Vd(. Γf . we can relax the regularity assumption for u∞ and ask : iii) u∞ · τ ∈ L2 (0. ) ∈ (L2 (Ω0 ))2 × R. (H 2 (Γf ))2 ) ∩ H 4 (0. d. on on div(v) = 0. ) ∈ C ([0. p. supp(v) ⊂ Ωf t t div(v) = 0.190 with Moving Shape Analysis Hd(t) ≡ (v. d ) = 2 T 0 γ ˙ |d − d1 |2 + |d − d2 |2 + u∞ g g 2 2 Uc (7.

18). The divergence free condition coming from the fact that the fluid has an homogeneous density and evolves as an incompressible flow is difficult to impose on the mathematical and numerical point of view.4.2 Free divergence and non-homogeneous Dirichlet boundary condition constraints We shall now describe the way to take into account inside the Lagrangian functional several constraints associated to the coupled system. Once the full linear tangent system is defined.Sensitivity analysis for a simple fluid-solid interaction system 191 The main difficulty in dealing with such a minimization problem is related to the dependence of integrals on the unknown domain Ω f which depends also t on the control variable u∞ .4. This leads in a certain sense to a saddle point formulation or mixed formulation of the Navier-Stokes subsystem. This may be realized using a Min-Max formulation of the minimization problem (7. We shall use the following identity.20) Ωf t Ωf t Γf ∞ Γs t .1 Analysis strategy We shall focus our efforts on the derivation of first-order optimality conditions for problem (7. This point will be solved by using the ALE map Tt introduced previously. we shall use the latter choice coupled with the introduction of a transverse map in order to handle the sensitivity analysis of the Lagrangian functional with respect to variation of the fluid domain. In this chapter. it is possible to handle the derivative Du∞ (u.18) taking into account the coupled system constraint through Lagrange multipliers. D • Following [58]. p. it is possible to define an adjoint system whose solution may be involved in the computation of the objective function gradient. • An other possible choice is to try to pass through the computation of a linear tangent system and directly get the adjoint system. This involves the computation of the gradient with respect to the inflow condition u∞ of the cost function Ju∞ . We choose to include the divergence free condition directly into the Lagrangian functional thanks to a multiplier that may play the role of the adjoint variable associated to the primal pressure variable. d) · δu∞ using a back transport map into a fix domain and use the weak implicit function theorem to justify and obtain the linearized system. 7. 7. − q div u = u· q− u · (q n) − u · (q n) (7. There exist several methods to handle such a question.

3 Solid reduced order and solid weak state operator For the sake of simplicity. Σf ∞ ˙ u = d e2 .3 This kind of technique has been popularized in [99] as a systematic way to study non-homogeneous linear partial differential equations. n (7.4. This procedure has been already used in [135] and [45. since we are dealing with a moving domain.21) (7.7) involves two essential non-homogeneous Dirichlet boundary conditions. consisting in totally transposing the Laplacian operator.24) (7. −ν∆u · φ = −ν∆φ · u + ν [u · ∂n φ − φ · ∂n u] dΓ (7. These formulations are usually called very weak formulations or transposition formulations.25) Ωf T u(T ) · v(T ) − Ωf 0 u(0) · v(0) REMARK 7. ˙ ˜ ˜ Md+Kd=F (7.26) leading to the first order ordinary differential equation. Σs 0 (7.192 Moving Shape Analysis The coupled system (7. u = u ∞ . T 0 Ωf t T T Ωf t ∂t u · v = − + 0 u · ∂t v − 0 ∂Ωf t (u · v) V. ˜ d= d1 d2 = d ˙ d (7.27) . We shall also choose to transpose the time operator inside the weak formulation.23) Ωf t Ωf t Γf ∪Γs ∞ t We shall substitute inside this identity the desirable boundary conditions and we will recover the boundary constraints in performing an integration by parts inside the optimality conditions corresponding to the sensitivity with respect to the multipliers. This has to be performed very carefully. 7. We refer to [4] for recent applications on the Navier-Stokes system.22) We use a very weak formulation of the state equation. 46] to perform shape optimization problems for elliptic equations using Min-Max principles. we reduce the order of the solid governing equation by defining the global solid variable. We shall notice that these methods are still valid in the nonlinear case to obtain regularity or existence results.

we summarize the different options that we have chosen for the Lagrangian framework and define the variational state operator constraint. b2 ) = 0 T T −d1 b˙1 − d2 b1 − d0 b1 (0) + d1 (T )b1 (T ) 1 T + 0 −m d2 b˙2 + k d1 b2 − m d0 b2 (0) + m d2 (T ) b2 (T ) − 2 F b2 0 7.4. φ = 0.Sensitivity analysis for a simple fluid-solid interaction system with M= K= and F = 1 0 0m 0 −1 k 0 0 Ff 193 In the case of the coupled fluid-solid system.30) (7.31) Y1s and the load space.4 Fluid state operator In this section. on Γf ∞ 1 H0 = . 0 φ = 0. T ]) = b1 ∈ C ([0. T ]) (7. the loads depend on the fluid state variable. φ ∈ (H 1 (Ω0 ))2 . T ]) (7. Ff (u. (b1 . p) = − Γs t σ(u. p) · n · e2 We introduce the state and multiplier spaces in order to define a proper solid weak state operator. In order to deal with rigid displacement vector fields. in in Ωs 0 Ωs . s s es : X1 × X2 × L −→ (Y1s × Y2s )∗ whose action is defined by the following identity.29) (7. s X1 = d1 ∈ C 1 ([0. es (d1 . d2 . φ = 0. T ]) 1 This allows us to define a solid state operator. we introduce the following spaces : Rigid displacement spaces: H1 = def def φ ∈ (H 1 (Ω0 ))2 . Y2s = b2 ∈ C 1 ([0.28) s X2 = d2 ∈ C 1 ([0. F ). T ]) L = F ∈ C 1 ([0.

194 Moving Shape Analysis LEMMA 7. T . assuming that Ωs is connected. (H 1 (D))2 ) We also need test function spaces that will be useful to define Lagrange multipliers : Yf = def def 1 v ∈ L2 (0. p) · n) + ΩT u(T ) · v(T ) − Ω0 ∀ (v. q) · n) + Γs t 0 [us · (σ(v. v|Γs = b2 · e2 t We define the fluid weak state operator. Following the existence result stated previously. T . q) ∈ Y f × V f 7. q) · n) − (u · v) us . there exists C ∈ R2 such that 0 ϕ|Ωs = C 0 In the sequel. (v. It is now possible to couple these two operators in order to build an ad-hoc . p. we will need to define precise state and multiplier spaces in order to endow our problem with a Lagrangian functional framework. ef ∞ : X f × Z f × U s −→ (Y f × V f )∗ u whose action is defined by : ef ∞ (u. we introduce the fluid state space : Xf = def u ∈ H 2 (0.4. T . T . (H 2 (Ωf ))2 ∩ H1 ) t def Zf = p ∈ H 1 (0. (H 2 (Ωf ))2 ∩ H0 ) t Vf = f Ws = def q ∈ H 1 (0.5 Coupled system operator Our mechanical system consists of a solid part and a fluid part. (H 1 (D))2 ) (v. b2 ) ∈ Y f × Y2s .2 For ϕ ∈ H1 . q) = u T 0 T Ωf t [−u · ∂t v + (D u · u) · v − νu · ∆v + u · T Γs t q − p div v] + 0 Γf ∞ T u∞ · (σ(v. us ). These subsystems have been represented thanks to a solid and a fluid state operator. n ] u0 · v(t = 0) − 0 v · (σ(u.

b1 ) ∈ Ws × V f × Y1s 7. d2 ). b2 ) + ef ∞ (u. (v. Ff ). q) · n) − (u · v) d2 e2 . t This means that we shall choose the fluid and second solid multiplier spaces f to be the space Ws . (v. q) · n) + T 0 [(d2 e2 ) · (σ(v. b1 . q. we introduce the Lagrangian functional associated with problem (7. (v. d2 e2 ). b1 . b2 ) = Ju∞ (u. b2 ) = T es (d1 .18) : Lu∞ (u. d1 . Hence. (b1 . we choose to work with continuous test functions on Γ s . eu∞ (u. d1 . If not. p. d1 . d2 . d2 ).32) with Ju∞ (u. q. q. d1 . p. p. q. p. we have to work with the fluid constraint tensor at the fluid-solid boundary. d2 ) = α 2 T 0 def |d1 − d1 |2 + |d2 − d2 |2 + g g γ u∞ 2 2 Uc . The major point here is to notice that the fluid load F f appears in the fluid state operator and then can be coupled with the solid part thanks to the input load F of the solid operator. 2 f ∀ (v.7) and problem (7.Sensitivity analysis for a simple fluid-solid interaction system 195 coupled system operator. v. b1 . b2 ) (7. d1 . q. d2 . p. d1 · e2 ) u Ωf t = 0 T [−u · ∂t v + (D u · u) · v − νu · ∆v + u · T Γs t q − p div v] + 0 Γf ∞ u∞ · (σ(v. d2 )− eu∞ (u. which may be not convenient due to a regularity requirement. s s f eu∞ : Y f × Z f × X1 × X2 −→ (Ws × V f × Y1s )∗ whose action is defined by the following identity. To achieve this coupling.4. b2 . We define the coupled system weak state operator as follows. n ] T 0 + 0 −d1 b˙1 − d2 b1 + + ΩT −m d2 b˙2 + k d1 b2 u(T ) · v(T ) − − d0 b1 (0) + d1 (T ) b1 (T ) − 1 u0 · v(t = 0) Ω0 md0 b2 (0) + m d2 (T ) b2 (T ).6 Min-Max problem In this section. we need to decide whether or not the fluid and the solid multipliers match at the fluidsolid interface. p.

pu∞ . b1 . πu∞ . p.b2 . p.34) admits at least one saddle point for u∞ ∈ Uc . p. q. d2 . d1 . d2 ∞ .196 Moving Shape Analysis Using this functional. πu∞ . b1 ∞ . the gradient of the cost function j at point u∞ ∈ Uc is given by the following expression : j(u∞ ) = (σ(ϕu∞ . πu∞ ) · n)|Σf + γ u∞ ∞ (7.b1 )∈Ws ×V f ×Y1 max Lu∞ (u. v. d1 . b1 ) ∈ Lu∞ (u. p. b2 ) (7. b2 .18) can be put in the following form : u∞ ∈ Uc min (7. b2 ) min f s (v. v. and (uu∞ . pu∞ . d2 ) ∈ (v. b1 . We would like to apply min-max differentiability results to problem (7. b1 ) ∈ s s X f ×Z f ×X1 ×X2 f s Ws ×V f ×Y1 T T Γf ∞ = 0 (ν ∂n φu∞ − πu∞ n) · δu∞ + γ 0 Γf ∞ u∞ · δu∞ . q. The main issue is to prove that the min-max subproblem s s (u.35) PROOF Using theorem (3) from [45. b2 ).34) : < j (u∞ ). d1 ∞ . d1 . First-order optimality conditions will furnish the gradient of the original cost function using the solution of an adjoint problem. we bypass the derivation with respect to u∞ inside the min-max subproblem (7.p. d1 . we avoid the computation of the state derivative with respect to u∞ . d2 ) ∈ f s Ws ×V f ×Y1 max (v. du∞ .d2 )∈X f ×Z f ×X1 ×X2 s s X f ×Z f ×X1 ×X2 min (u. d2 . δu∞ u u u u Du∞ ∞ D = min max Lu (u. v. b2 ∞ ). q. δu∞ Du∞ ∞ (u. d1 .34). p.34).33) By using the Min-Max framework. q.2 ˜ For u∞ ∈ Uc . Reduced Gradient We assume that the conditions to apply the Min-Max principle [40] are fulfilled so we can bypass the derivation with respect to the control variable u∞ through the min-max subproblem (7. the optimal control problem (7. ϕu∞ . ˜u∞ ) the unique saddle point of b problem (7. It leads to the following result : THEOREM 7. b2 . b1 .d1 .q. q. δu∞ >= D Lu (uu∞ . φu∞ . 46]. d2 .33).

d2 . δb1 .b2 ) Lu∞ (u. q. we perform some integration by parts : LEMMA 7. δd1 . p. q. b2 ) = Ju∞ (u. Lu∞ (u. d2 . q. This step is crucial. v.p. v. b1 . q.q. we are interested in establishing the first order optimality condition for problem (7. p. b1 . δd2 ) = 0. δu = T − 0 Ωf t [−δu · ∂t v + [D δu · u + D u · δu] · v − νδu · ∆v + δu · T q] + 0 Γs t (δu · v) d2 e2 . d2 )− eu∞ (u. q. d1 . δp. d2 . p.1 Derivatives with respect to state variables Fluid adjoint system LEMMA 7. better known as Karusch-Kuhn-Tucker optimality conditions. b1 ∞ .36) The KKT system will have the following structure : ∂(v. because it leads to the formulation of the adjoint problem satisfied by the Lagrange multipliers (ϕ u∞ . d1 . b2 ) · (δv. d1 . b2 . v. p. p. q. δb2 ) = 0. q. b2 ) · (δu. δb1 ) ∈ Ws × V f × Y1s → State Equations ∂(u. d2 . d1 .3 f s s For u∞ ∈ Uc . Lu∞ (u. v. n − ΩT δu(T ) · v(T ) ∀δu ∈ X f In order to obtain a strong formulation of the fluid adjoint problem. b1 . b2 ) is differentiable with respect to u ∈ Y f and we have ∂u Lu∞ (u.5 KKT Optimality Conditions In this section. d1 . p. v. u u We recall the expression of the Lagrangian.Sensitivity analysis for a simple fluid-solid interaction system 197 7. δb2 . b1 ) ∈ Z f × X1 × X2 × Ws × V f × Y1s . b1 . v. δq. d2 ). (v. b1 . b1 .5. d1 .d2 ) Lu∞ (u. δp.34). n Ωf t Ωf t Γf ∪Γs ∞ t . b2 ∞ ). b2 ) (7. def f ∀ (δv.d1 . δq.b1 . πu∞ . s s ∀ (δu. b2 . δd2 ) ∈ X f × Z f × X1 × X2 → Adjoint Equations 7. b2 ). δd1 . (p. b1 .4 (D δu · u) · v = − [D v · u + div(u) v] · δu + (δu · v) u. d2 . p. d1 .

(u. d2 . using the reduced oder model.5. d2 . q. the Lagrangian Lu∞ (u. q. b2 ) is differentiable with respect to p ∈ Z f and we have T ∂p Lu∞ (ˆ. b1 ) ∈ X f ×Z f ×X1 ×X2 ×Ws ×V f ×Y1s . ∀δp ∈ Z f This leads to the following fluid adjoint strong formulation. b2 ). d1 . b1 . δu = ˆ − Qf [−∂t ϕ + (∗ Du) · ϕ − (Dϕ) · u − div(u) ϕ − ν∆ϕ + − Ωf T π] · δu ϕ(T ) · δu(T ) LEMMA 7. b2 ).  ∗  −∂t ϕ − D ϕ · u + ( Du) · ϕ − ν∆ϕ +   div(ϕ) = 0.  ϕ = 0.2 Solid adjoint system We recall that the ALE map is built as the flow of a vector field V that matches the solid velocity at the fluid-solid interface and is arbitrary inside the fluid domain. Furthermore. Then. LEMMA 7. d1 . x ∈ Ωf t  V (x. v. δp = u 0 Ωf t δp div ϕ. b2 ) is differentiable with respect to d1 ∈ .     ϕ(T ) = 0. π. q. p.   ϕ = b2 · e2 . x ∈ Γf ∞ (7. b2 . p.5 s s f For u∞ ∈ Uc . d1 . b1 ) ∈ X f × X1 × X2 × Ws × V f × Y1s . we have the following result. ϕ. Lu∞ (u.. d2 . b1 . b2 . v. t) = d2 · e2 . d2 . b1 . p. q. Qf Qf Σs Σf ∞ Ωf T (7. v. t) · n = 0.198 Moving Shape Analysis It leads to the following identity : ∂u Lu∞ (u. d1 . q. q = 0.6 s s f For u∞ ∈ Uc and (u. t V (x.37) 7. ϕ. t) = Ext(d2 · e2 ). the ALE map depends on d2 through V .e. d2 . d1 .38) Hence. the derivative with respect to d1 may be simpler since it does not involve a derivative with respect to the geometry. v.  x ∈ Ωs  V (x. b1 . p. p. b1 . i. b2 .

δd1 = T 0 ˙ α (d1 − d1 ) δd1 − k δd1 b2 + δd1 b1 − δd1 (T ) b1 (T ) g From which we deduce the following adjoint ODE. [156] and e in [59]. b2 ). with ρ ≥ 0.ρ = Tt (V + ρW )(Ωf ) t 0 Ωs.ρ t t Ωs −→ Ωs. We introduce a perturbation flow W associated to the perturbation δd 2 e2 . d2 .39) The derivative of the Lagrangian with respect to the solid velocity variable d 2 involves shape derivatives of domain integrals. we introduce a new map as in [157] : Tρt = Ttρ ◦ Tt−1 : Ωf −→ Ωf. d2 + ρδd2 . b1 . v . [57]. W = Ext(δd2 · e2 ) This flow generates new fluid and solid domains through the ALE map.40) . b1 . p. T t (V + ρW ). we set Ttρ = Tt (V + ρW ) The objective of this paragraph is to compute the following derivative : d L(ˆ. Then we need to introduce the concept of Transverse Field associated to a perturbation of the solid velocity. p.Sensitivity analysis for a simple fluid-solid interaction system s X1 and we have 199 ∂d1 Lu∞ (u. ϕ. (0. −b˙1 + k b2 = α(d1 − d1 ). For example. π. This point has been investigated previously by Zol´sio in [157]. t t To this end. We set Ωf. it is convenient to work with functions already defined in def Ωf. q.ρ=0 = Ωf as we proceed to the limit ρ → 0. d1 . b2 ) u ˆ dρ Transverse map and vector field Since we would like to differentiate the Lagrangian functional with respect to ρ at point ρ = 0. T ) g b1 (T ) = 0 (7. d1 .ρ = Tt (V + ρW )(Ωs ) t 0 def def For the sake of simplicity.ρ t t def def ρ=0 (7.

1 def St (.e. we only consider derivatives at point ρ = 0. ρ 0 ). Ω0 (7.) = def ∂Tρt ∂ρ ◦ (Tρt )−1 (7. in Ω def (7. .) = ∂ρ [Tt (V + ρW )] = 0 def t [D(V + ρW ) ◦ Tτ (V + ρW ) · S τ (ρ.∞ (Ω)) ρ −→ Tt (V + ρW ) is continuously differentiable and S t (ρ. [0.200 Moving Shape Analysis This map is actually the flow of a vector field following for ρ ∈ (0. ρ ≥ 0 dρ x(ρ = 0) = x. .42) t Since. Ω0 × (0. is the solution of the following dynamical system: t Ttρ (Zρ ) : Ω −→ Ω t x −→ x(ρ.43) A fundamental result furnishes the dynamical system satisfied by the vector field Zt related to the vector fields (V. THEOREM 7. THEOREM 7. we set Z t = Zρ=0 .) is the unique solution of the following Cauchy problem. ρ0 ] −→ C 0 ([0.) +W ◦ Tτ (V + ρW )dτ ] COROLLARY 7.) = S t (ρ = 0.4 The mapping.3 t The Transverse map Ttρ is the flow of a transverse field Zρ defined as follows: t Zρ = Z t (ρ. W k−1. We recall a result from [58] which might be useful for the sequel. W ). . T ) St=0 = 0. ∂t St − (D V ◦ Tt ) · St = W ◦ Tt . x(ρ)).. T ]. x) ≡ Ttρ (Zρ )(x) with d x(ρ) = (ρ. .41) i. .

n dΓ (7. T ) Zt=0 = 0. V ] = W. . Ω0 def (7.7 d dρ G(ρ) dΩ ρ=0 Ωf.1 For φ(ρ. V ] = DZt · V − DV · Zt stands for the Lie bracket of the pair (Zt . . then def ˜ ˜ φ Γ = φ | Γs + ∂ n φ Z t . We will use the framework developed in Sokolowski-Zol´sio [135]. n t def ˜ with φ |Γs = t d ˜ φ(ρ.ρ t φ(ρ) dΓ ρ=0 = Γs t φΓ + Hφ Zt . .) ◦ Tρt dρ ρ=0 then the tangential shape derivative of φ is given by the following expression. e LEMMA 7. ρ=0 Γs t .8 d dρ Γs.45) LEMMA 7. V ).46) where φΓ stands for the tangential shape derivative of φ(ρ.Sensitivity analysis for a simple fluid-solid interaction system 201 THEOREM 7.5 [59] The vector field Zt is the unique solution of the following Cauchy problem.) dρ . def ˙ φΓ = φ − Γ φ · Zt REMARK 7.ρ )). Ω0 × (0. t We recall classical definitions of shape derivative functions : DEFINITION 7. ∂t Zt + [Zt . n dΓ (7. we will need general results concerning shape derivatives of integrals over domains or boundaries.44) where [Zt . C 1 (Γs.4 we have ˜ If φ is the trace of a vector field φ defined over Ω.ρ t = Ωf t ∂ρ G(ρ) dΩ + Γs t G(ρ = 0) Zt . Shape derivatives In the sequel. the material derivat tive is given by the following expression : d ˙ φ= φ(ρ. ρ0 .) ∈ L1 (Γs ). x) ∈ C 0 (0.

ρ t −(u ◦ Rt ) · ∂t (v ◦ Rt ) + (D u ◦ Rt · u ◦ Rt ) · v ◦ Rt ρ ρ ρ ρ ρ T −νu ◦ Rt · ∆v ◦ Rt + u ◦ Rt · ρ ρ ρ T q − p div v ◦ Rt + ρ 0 Γf ∞ u∞ · (σ(v.9 d dρ Γs. b2 . q) · nρ ) ρ T T − u ◦ Rt · v ◦ Rt · ((d2 + ρδd2 ) e2 ) · nρ ρ ρ −d1 b˙1 − (d2 + ρδd2 ) b1 + Ω0 0 + 0 −m (d2 + ρδd2 ) b˙2 + k d1 b2 + ΩT u(T ) · v(T ) − u0 · v(t = 0) − d0 b1 (0) + d1 (T ) b1 (T ) − m d0 b2 (0) 1 2 f ∀ (v. b1 . This leads to the following t perturbed Lagrangian : Lρ ∞ = L(u. we have LEMMA 7.47) Derivation of the perturbed Lagrangian Thanks to the introduction of the transverse map.202 Moving Shape Analysis Following this remark. p. v) that are defined on Ωf . q) · n) + 0 Γs. q. n dΓ (7. p. d1 . . q. d1 . it is now possible to work with functions (u. . ρ We need the following lemmas in order to derivate G(ρ.) with respect to ρ. .ρ t ((d2 + ρδd2 ) e2 ) · (σ(v ◦ Rt . def with Rt = (Tρt )−1 . b2 ) u = Ju∞ (u.ρ t ˜ φ(ρ. a) Distributed terms: We set G(ρ. v. x)da ρ=0 = Γs t ˜ ˜ ˜ φ + H φ + ∂n φ Zt . We apply the previous results to the perturbed Laρ grangian functional L.) = −u ◦ Rt · ∂t (v ◦ Rt ) + D(u ◦ Rt ) · (u ◦ Rt ) · v ◦ Rt ρ ρ ρ ρ ρ −ν(u ◦ Rt ) · ∆(v ◦ Rt ) + (u ◦ Rt ) · ρ ρ ρ def q − p div(v ◦ Rt ) ρ with Rt = (Tρt )−1 . d2 + ρδd2 . b1 ) ∈ Ws × V f × Y1s + m (d2 + ρδd2 )(T ) b2 (T ). d2 + ρδd2 ) T def − 0 Ωf.

11 d u ◦ Rt ρ dρ ρ=0 = − D u · Zt PROOF Using the chain rule we get d u ◦ Rt ρ dρ = D u ◦ Rt · ρ = − D u · Zt ρ=0 D Rt ρ Dρ ρ=0 ρ=0 = − D u ◦ Rt · Z t (ρ.)|ρ=0 = [(D u · Zt ) · ∂t v + u · (∂t (D v · Zt )) − [(D(D u · Zt )) · u + D u · (D u · Zt )] · v − (D u · u) · (D v · Zt ) +ν(D u · Zt ) · ∆v + νu · (∆(D v · Zt )) + p div(D v · Zt ) − (D u · Zt ) · PROOF (7. .Sensitivity analysis for a simple fluid-solid interaction system LEMMA 7. . we have the following result ∂ρ G(ρ.11). q] It comes easily using definition of G(ρ.) ρ LEMMA 7. .10)- Then we have an expression of the derivative of distributed terms coming .12 Then.) and lemma (7.10 dTρt dρ dRt ρ dρ 203 = Zt ρ=0 ρ=0 = −Zt LEMMA 7.

.) = (V + ρW ) · σ(v ◦ Rt .204 Moving Shape Analysis from the Lagrangian with respect to ρ. n + (div E) Zt . n PROOF First. d dρ G(ρ)dΩ ρ=0 Ωf.ρ . Then we set t φ(ρ.) is defined on the boundary Γs.). . q) − u ◦ Rt · v ◦ Rt · ((d2 + ρδd2 ) e2 ) ·nρ ρ ρ ρ We set V = d2 e2 and W = δd2 e2 . d dρ div E(ρ) ρ=0 Ωf. we need some extra ident tities corresponding to boundary shape derivates of terms involved in φ(ρ.ρ . q) − u ◦ Rt · v ◦ Rt · nρ ρ ρ ρ Since φ(ρ.ρ t = Ωf t div E + Γs t (div E) Zt .ρ t div E(ρ) then we derive this quantity using lemma (7.ρ t E(ρ). .) = (d2 + ρδd2 ) e2 ) · (σ(v ◦ Rt .ρ t = Γs t E |Γt .14 d dρ E(ρ). nρ = Ωf. n . we use that Γs. n b) Boundary terms : We must now take into account the terms coming from the moving boundary Γs.13 [53] ∂ρ nρ |ρ=0 = nΓ = − LEMMA 7. . nρ dΓ ρ=0 Γ (Zt · n) Γs.7).ρ t = Ωf t [(D u · Zt ) · ∂t v + u · (∂t (D v · Zt )) q] − [(D(D u · Zt )) · u + D u · (D u · Zt )] · v − (D u · u) · (D v · Zt ) +ν(D u · Zt ) · ∆v + νu · (∆(D v · Zt )) + p div(D v · Zt ) − (D u · Zt ) · + Γt [−u · ∂t v + (Du · u) · v − νu · ∆v + u · q − p div(v)] Zt . LEMMA 7. then φ(ρ.

we have d dρ φ(ρ) dΓ ρ=0 Γs.48) Ωf t We use this expression in the sequel. ρ Lρ V. We recall that the perturbed Lagrangian has the following form.n .15 E | Γt = W ·[−q I +ν D v − u · v]+V ·[−ν D(D v · Zt ) + (D u · Zt ) · v + u · (D v · Zt )] Hence.ρ t = Γs (V ) t W · [−q I +ν D v − u · v] · n + Γs (V ) t V · [−ν D(D v · Zt ) + (D u · Zt ) · v + u · (D v · Zt )] · n div(V · [−q I +ν ∗ D v − u · v]) Zt .W − T 0 Ωf.ρ t T G(ρ) − 0 Γs.W = JV.Sensitivity analysis for a simple fluid-solid interaction system We conclude using div E = Γs t 205 Ωf t E .5 We recall that V · (D v · n) = = Ωf t Γt div(∗ D v · V ) D v · · D V + V · ∆v (7. Then we use LEMMA 7. q) ∈ Y (Ωf ) × V f t − ΩT u(T ) · v(T ) + Ω0 .ρ t φ(ρ) u0 · v(t = 0) ∀ (v. n + Γs (V t ) REMARK 7.

W T T ρ=0 − 0 d dρ . its derivative with respect to ρ at point ρ = 0 has the following expression.206 Moving Shape Analysis Hence. d Lρ V. LEMMA 7.W dρ = ρ=0 d Jρ dρ V.W dρ with T ρ=0 = −AZt − BZt − CW (7.16 0 d dρ Γs.ρ t G(ρ) ρ=0 − Furthermore we have.W T = ρ=0 0 α(d2 − d2 ) e2 · W g (7. d Lρ V. we get the following expression.ρ t φ(ρ) ∀ (v.50) AZ t = 0 Ωf (V ) t [(D u · Zt ) · ∂t v − [(D(D u · Zt )) · u q] + u · (∂t (D v · Zt )) − (D u · u) · (D v · Zt ) + νu · (∆(D v · Zt )) + p div(D v · Zt ) B Zt = T 0 Γs (V ) t + D u · (D u · Zt )] · v + ν(D u · Zt ) · ∆v − (D u · Zt ) · [−u · ∂t v + (Du · u) · v − νu · ∆v + u · q − p div(v)] (Zt · n) + V · [−ν D(D v · Zt ) + (D u · Zt ) · v + u · (D v · Zt )] · n + div(V · [−q I +ν ∗ D v − u · v]) Zt .49) Using the last identities concerning the derivative of the distributed and the boundary terms with respect to ρ. ρ=0 Ωf. q) ∈ Y (Ωf ) × V f t d Jρ dρ V. n T − 0 Γf ∞ ν u∞ · D(D v · Zt ) · n .

ϕ. ϕ. π) saddle points of the Lagrangian functional (7. we would like to express the distributed term A Zt as a boundary quantity defined on the fluid moving boundary Γs and the fixed boundary t Γf .6 For (u.32). the following identity holds. ∞ THEOREM 7. Indeed. v) in the Lagrangian functional where we have added a boundary since we consider test functions v that do not vanish on Γ f and do not match the ∞ . p. ∀ W = δd2 e2 PROOF We use extremal conditions associated to variations with respect to (u. T 0 Ωf (V ) t [(D u · Zt ) · ∂t ϕ − [(D(D u · Zt )) · u π] + u · (∂t (D ϕ · Zt )) + D u · (D u · Zt )] · ϕ + ν(D u · Zt ) · ∆ϕ − (D u · Zt ) · − (D u · u) · (D ϕ · Zt ) + νu · (∆(D ϕ · Zt )) + p div(D ϕ · Zt ) T − T 0 Γs (V ) t V · [ν D(D ϕ · Zt ) − (D u · Zt ) · ϕ − u · (D ϕ · Zt )] · n + 0 Γs (V ) t [ν (ϕ − b2 e2 ) · (D(D u · Zt ) · n) + (D ϕ · Zt ) · (−p n + ν(D u · n))] T def − 0 Γf ∞ ν u∞ · (D(D ϕ · Zt ) · n) = 0. p. This will help us to simplify several terms involved in the derivative of L with respect to V . π) is a saddle point of the Lagrangian functional L. q) · n − (u · v) n − α(d2 − d2 ) e2 g The shape derivative kernel identity We shall now assume that (u.Sensitivity analysis for a simple fluid-solid interaction system 207 T CW = 0 Γs (V ) t W · −b1 e2 − m b˙2 e2 + σ(v.

where (u. p) · n) T 0 − − ΩT 0 −d1 b˙1 − d2 b1 − Ω0 −md2 b˙2 + kd1 b2 u(T )v(T ) + u0 v(t = 0) + d0 b1 (0) − d1 (T )b1 (T ) 1 + md0 b2 (0) − md2 (T )b2 (T )∀ (v. ϕ. δv) ∈ X(Ωf ) × Y (Ωf ) t t We choose specific perturbation directions. v) are saddle points of the Lagrangian. d1 . q) · n − (u · v) n] + 0 T Γs (V ) t (v − b2 e2 ) · (σ(u.e. d1 .v) L2 · (δu. d1 . 0 ν u∞ · (D δv · n) − ΩT ∀ (δu. t L2 (u. We set (u.e.. we simplify the Lagrangian derivative at the saddle point (u. d2 .e. p. p. d2 . d2 ) T − − 0 T 0 Ωf (V ) t [−u · ∂t v + (D u · u) · v − νu · ∆v + u · T Γs (V ) t q − p div v] Γf ∞ u∞ · σ(v. ϕ) . q) · n − T 0 V · [σ(v. b2 ) ∈ Y f × V f × Y1s × Y2s 2 This leads to the following perturbation identity. δv) = T − 0 Ωf t [−δu · ∂t v − u · ∂t δv + D(δu · u) · v + D(u · δu) · v q − p div(δv)] +D(u · u) · δv − ν(δu · ∆v) − ν(u · ∆δv) + δu · T + 0 T Γs (V ) t [ν (v − b2 e2 ) · (D δu · n) + δv · (−p n + ν(D u · n))] T Γf ∞ − 0 Γs (V ) t V · [ν D(δv) − δu · v − u · δv] · n − [δu(T )v(T ) + u(T )δv(T )] . We recognize immediately the distributed and boundary terms involved in the shape derivative kernel identity. p.. b2 ). solutions of respectively the primal and adjoint fluid problem. q.208 Moving Shape Analysis solid test functions at the fluid-solid interface Γ s (V ) . b2 ) = Ju∞ (u. i. δu = D u · Zt δv = D v · Zt with δu(T ) = δv(T ) = δu(0) = δv(0) = 0.. q. v) = (u. ∂(u. i. b1 . b1 . v. i. b1 . Solid and ALE adjoint problem Using the shape derivative kernel identity.

n − ν u∞ · D(D ϕ · Zt ) · n AZ t + B Z t = 0 Γs (V ) t [ν (D ϕ · n) · (D(V − u) · n) − u · ∂t ϕ + (Du · u) · ϕ −π div V − (div V ) u · ϕ − V · (u · ϕ)] Zt . T (7. T (u · ϕ) Zt . div(V · [−π I +ν ∗ D φ − u · ϕ]) = −π div V − V · π + ν D ϕ · · D V + ν V ∆ϕ − (div V ) u · ϕ − V · (u · ϕ) (7.Sensitivity analysis for a simple fluid-solid interaction system and we get T 209 AZ t = 0 T Γs (V ) t V · [ν D(D ϕ · Zt ) − (D u · Zt ) · ϕ − u · (D ϕ · Zt )] · n − 0 Γs (V ) t [ν (ϕ − b2 e2 ) · (D(D u · Zt ) · n) + (D ϕ · Zt ) · (−p n + ν(D u · n))] T + 0 Γf ∞ ν u∞ · (D(D ϕ · Zt ) · n) We use that ϕ = b2 e2 on Γs (V ) and the following identities. n T + 0 Γf ∞ ν u∞ · (D(D ϕ · Zt ) · n) Using the following identity.51) (7. t (D ϕ · Zt ) · (p n) = (p div ϕ) Zt . n then.52) AZ t = 0 Γs (V ) t T V · [ν D(D ϕ · Zt ) − (D u · Zt ) · ϕ − u · (D ϕ · Zt )] · n Γs (V ) t − 0 [−p div ϕ + ν (D ϕ · n) · (D u · n)] Zt .53) we get B Zt = T 0 Γs (V ) t [−u · ∂t ϕ + (Du · u) · ϕ − νu · ∆ϕ + u · + − π div V − V · π − p div(ϕ)] Zt . n (D ϕ · Zt ) · (D u · n) = (D ϕ · n) · (D u · n) Zt . n + V · [−ν D(D ϕ · Zt ) + (D u · Zt ) · ϕ + u · (D ϕ · Zt )] · n π + ν D ϕ · · D V + ν V ∆ϕ T 0 Γf ∞ − (div V ) u · ϕ − V · then we get. n .

(7. T ) λ(T ) = 0.54) .210 Moving Shape Analysis We use the following identity. n We choose the velocity field V = Ext(d2 e2 ) ◦ p. we have AZ t + B Z t = T 0 Γs (V ) t −d2 b˙2 + ν (D ϕ · n) · (D u · n) − |d2 |2 (D ϕ · e2 ) · e2 Zt . This leads to t t T AZ t + B Z t = 0 Γs (V ) t ν (D ϕ · n) · (D(V − u) · n) − d2 b˙2 − π div V − (div V ) u · ϕ − d2 e2 · D ϕ · d2 e2 Zt .55) (7. (0. ΓT (V ) with f = −d2 b˙2 + ν (D ϕ · n) · (D u · n) − |d2 |2 (D ϕ · e2 ) · e2 REMARK 7. π) · n − α(d2 − d2 ) e2 g where we have used that d2 e2 · b 2 e2 · n = 0 Γs (V ) t We introduce the adjoint field λ solution of the following system. n and T CW = 0 Γs (V ) t W · −b1 e2 − m b˙2 e2 + σ(φ. u = d2 e2 on Γs . −∂t λ − Γ λ · V − λ divΓ V = f. V · (u · ϕ) = ϕ · D u · V + V · D ϕ · u and the boundary conditions. then (D V · n) · n|Γs = 0 t and div V |Γs = divΓ V + (D V · n) · n t =0 Finally. ϕ = b2 e2 on Γs .6 In our case divΓ V = 0.

Sensitivity analysis for a simple fluid-solid interaction system We recall the following property.[57] For any E(V ) ∈ L2 (Σs (V )) and (V.56) where λ ∈ C 0 ([0. Then we have T AZ t + B Z t = 0 Γs (V ) t T f Zt . LEMMA 7. T 0 Γs (V ) t T E(V ) Zt . n = − λ W.17 [59]. g ∀ W = δd2 e2 def from which we deduce the following identity. W =− 0 Γs (V ) t However. W ) ∈ U. π) · n (7.18 ∂t λn Γs (V ) t = Γs (V ) t [∂t λ + Γ λ·V] n (7.54) with t f = E. we obtain −(AZt + BZt ) = CW . H 1 (Γs )) is the unique solution of problem (7.57) Γs (V ) t Γs (V ) t We now use the following lemma. n 0 Γs (V ) t (7. 211 LEMMA 7. n λ n. the following identity holds. T ]. using the optimality condition for the Lagrangian functional. π) · n − α(d2 − d2 ) e2 · W.58) . λ n = −b1 − m b˙2 − α(d2 − d2 ) e2 + g σ(φ. This leads to T 0 Γs (V ) t T λn · W = Γs (V ) t dis 0 −b1 e2 − m b˙2 e2 + σ(φ.

This led to the derivation of first-order optimality conditions for an optimal control problem related to a tracking functional. π) · n · e2 + |d2 |2 (D ϕ · e2 ) · e2 − ν (D ϕ · n) · (D u · n) · n This concludes the proof of the main result.59) Γs (V ) t σ(φ. −b˙1 + k b2 = α(d − d1 ). This optimality system can be numerically approximated and included inside a gradient based optimization procedure. This analysis was performed using a Lagrangian functional and non-cylindrical shape derivative tools to handle perturbation with respect to the velocity of the solid. This point is under investigation following the strategy adopted for Navier-Stokes optimal control problems as in [69]. ˙ ˙ m ¨2 + k b2 = α(d − d1 ) − α(d2 − d2 ) + ∂t b g g Γs (V ) t (7. T ) g b1 (T ) = 0 Let us inject b˙1 inside the first one.212 We get ∂t Moving Shape Analysis −b1 − m b˙2 − α(d2 − d2 ) e2 + g Γs (V ) t Γs (V ) t σ(φ. (0. we have investigated sensitivity analysis for a simple 2D coupled fluid-structure system. π) · n · e2 + |d2 |2 (D ϕ · e2 ) · e2 − ν (D ϕ · n) · (D u · n) · n and we recall the other solid adjoint equation. π) · n = −d2 b˙2 + ν (D ϕ · n) · (D u · n) − |d2 |2 (D ϕ · e2 ) · e2 n This can be written as ˙ ˙ ˙ b1 + m ¨2 = −α(d2 − d2 ) + ∂t b g Γs (V ) t Γs (V ) t σ(φ. .6 Conclusion In this chapter. 7.

or change the solid equations and use a real 3D non-linear elastic model as described in the next chapter. .Sensitivity analysis for a simple fluid-solid interaction system 213 The methodology used in this chapter can be generalized to more complex fluid-structure interaction problems. We can either change the fluid model and handle compressibility as in [37].

.

This has allowed us to simply reduce the differential order by expanding the solid state and use its velocity as an explicit state variable. 8. It is surrounded by a fluid in R3 . In this case. namely the cost function gradient computation involving the solution of a linear adjoint problem. We use the classical arbitrary Euler-Lagrange (ALE) formulation. We introduce a control volume Ω ⊂ R3 containing 215 . we state the main result of this chapter. As a consequence. Then. the parametrization of the solid by its velocity is more challenging and the straightforward use of the results obtained in Chapter 7 is not obvious. We consider a solid located at time t ≥ 0 in a domain Ωs (t) ⊂ R3 with boundary Γs (t).1 Introduction In the previous chapter. The chapter is organized as follows: first of all. Finally. Here.2 Mechanical problem and main result In this section we introduce a fluid-structure interaction model. where the solid was driven by a basic second order differential equation. in the second part. particularly suited for problems involving moving boundaries. Then we have been able to use the framework developed in Chapter 5 concerning the derivative of the Navier-Stokes system with respect to the velocity of the moving domain. we choose to work with the solid displacement state variable. the situation is quite different since we are considering a more general solid model with a non-linear constitutive law. Then. we need to use the results obtained in Chapter 6 concerning the derivative of the Navier-Stokes using the non-cylindrical identity perturbations. its proof is fully developed.Chapter 8 Sensitivity analysis for a general fluid-structure interaction system 8. we introduce the mechanical problem and its mathematical description. we have performed a sensitivity analysis for a coupled fluid-solid model.

The classical conservation laws of the continuum mechanics drive the evolution of these unknowns. for the fluid velocity and pressure. in Ωf (t) where (u. The fluid state satisfies the following incompressible Navier-Stokes equations written in Eulerian conservative formulation: ∂t u + D u · u − ν∆u + p = 0. homogeneous and incompressible. (8. s Hence. the fluid evolution is restricted to the domain Ω f (t) = Ω − Ω (t).1) div u = 0. We assume the fluid Γf Ωs 0 Ωf 0 Ω Γs 0 ˆ θ(·. The notation ∂Ω stands for the boundary of Ω. In a fluid-structure interaction framework the evolution of the fluid domain Ωf (t) is induced by the structural deformation through the fluid-structure ins terface Γs (t). This approach is usually used for the solid domain Ωs (t). In addition. see figure 8. by definition Ωf (t) = Ω − Ω (t). by means of the Lagrangian formulation [94]. p) stand. In f the sequel we set Γ = ∂Ω and ∂Ωf (t) = Γf ∪ Γs (t) stands for the fluid domain boundary. respectively.1: Geometric description to be Newtonian viscous. Given a material reference configuration for the solid Ω s ⊂ Ω with boundary 0 . Indeed. p) = −p I +ν [D u + ∗ D u] with ν the kinematic viscosity of the fluid.216 Moving Shape Analysis the solid at each time t ≥ 0. Its behavior is described by its velocity and pressure. t) Γf Ωf (t) Ωs (t) Ω Γs (t) FIGURE 8.1. The elastic solid under large displacements is described by its velocity and its stress tensor. the fluid stress tensor is given by σ(u. It leads us to describe Ωf (t) according to a map acting in a fixed reference domain. in Ωf (t).

2) s The operator Ext is arbitrary defined inside Ω f . ˆ ˆs • the map θf is defined from θ|Γs . s f f ˆ ˆ θ(x0 . the pair ( θ. ∀x0 ∈ Ω0 . ˆ ˆ The solid evolution is given by its motion θ and the stress tensor field S(θ) ˆ is related to θ through an ˆ (second Piola-Kirchoff tensor [36]). Here. t) = Ext(θ|Γs )(x0 . as an arbitrary extension over domain Ω0 . T ) ∞ ∞ ˆ θ 0<t<T 0<t<T . This corresponds to the classical Lagrangian flow. t). S(θ)) satisfies the non-linear elastodynamic equations [36]: ˆ ˆ ˆ ∂tt θ − div0 [D θ · S(θ)] = f. 0 f 0 ˆ In short. t) ˆ (x0 . the 0 0 s control volume Ω = Ωf (t) ∪ Ω (t) is described by a smooth and injective map: ˆ θ : Ω × R+ −→ Ω ˆ = θ(x0 . Q ≡ ˆ θ ({t} × Ωf ) and Σf ≡ Γf × (0. t) −→ θ ˆ ˆ ˆ ˆ We set θf = θ|Ωf and θs = θ|Ωs . REMARK 8. and 0 s Ω = Ωf (t) ∪ Ω (t) stands for the actual configuration at time t > 0. Then. t). we shall only use the notation θ. The field S( θ) ˆ ˆ appropriate constitutive law. 0 ({t} × Γs ). which preserves Γf = Γf = ∂Ω. t) represents the position at time t ≥ 0 of the material 0 point x0 . Ext represents an extension operator from Γs to Ω0 such that 0 ˆs Ext(θ|Γs )|Γf = IΓf 0 0 0 REMARK 8. Then. such that [95]: 0 0 ˆ • for x0 ∈ Ωs . 0 ∀x0 ∈ Ω0 . the ALE map θ is given by ˆ ˆs θ(x0 .Sensitivity analysis for a general fluid-structure interaction system s 217 Γs . keeping in mind that it corres sponds to the solid displacement in Ω0 .2 We set Σθ ≡ ˆ We set ˆ ˆ J(θ) = det(D θ) > 0 ˆ In the sequel. θs (x0 . we take Ωf = Ω − Ω0 as the reference fluid configuration.1 We set Ω0 = Ωf ∪ Ω0 as the reference domain. t) = θs (x0 . in Ωs 0 (8.

The coupling between the solid and the fluid is realized through standard boundary conditions at the fluid-structure interface Γ s . on Γs .4) for any smooth second order tensor B. the kinematic 0 continuity of the velocity and the kinetic continuity of the stress [95]: def ˆ ˆ u = Vθ = ∂t θ ◦ θ−1 .5) ˆ ˆ ˆ ˆ D θ · S(θ) · n0 = J(θ)σ(u. Ωs 0 ˆ T (δ θ) · ·B + Ωs 0 ˆ δ θ · TA (B) = Γs 0 ˆ TΓ (B) · δ θ (8. ˆ ˆ ˆ S(θ) = λ (Tr E(θ)) I +2 µ E(θ) ˆ where λ. µ > 0 stands for the Lam´ coefficients and E( θ) is the strain tensor. We shall also assume that there exists two operators T A defined on Ωs and 0 ˆ def ˆ ˆ TΓ defined on Γs such that the linear tangent tensor T (δ θ) = S (θ) · δ θ in the 0 ˆ direction δ θ satisfies the following adjoint identity. ˆ ˆ θ (8. 0 0 we endow the fluid equations with Dirichlet boundary condition on the farfield boundary Γf : ∞ u = u∞ . on Γf ∞ .218 Moving Shape Analysis REMARK 8.3 The constitutive law for a St. Moreover. on Γs 0 where n0 stands for the unit normal vector on Γs pointing inside Ωs .3) This kind of constitutive law is often referred to as “ large displacement small strain ” model. Venant-Kirchhoff material is given by the following expression. namely. e ˆ E(θ) = 1 2 ∗ ˆ ˆ Dθ · Dθ − I (8. p)(∗ D θ)−1 · n0 .

f represents the applied body force. θ.Sensitivity analysis for a general fluid-structure interaction system 219 In summary. ˆ1 the initial displacement and solid velocity. π = 0. in Ωf . with a ALE formulation for the fluid. π) being a solution of the following backward fluid adjoint system. Main Result: For smooth u∞ ∈ Uc . ˆ θ div u = 0. π) · n|(0. is given by:                           ∂t u + D u · u − ν∆u + p = 0.    ϕ(T ) = 0. p.10) . on Γf . θ(u∞ )) (8. u0 the initial fluid velocity and θ0 .   ˆ ˆ ϕ = ψ ◦ θ−1 . θ) = 2 T 0 Ωs 0 ˆ ˆ | θ − θd | 2 (8. ˆ θ      ˆ ˆ ˆ ˆ  D θ · S(θ) · n0 = J(θ)σ(u. in Ωf .T )×Γf ∞ (8. ∂t θ = (u0 . respectively. the following cost functional. θ The principal result of this chapter reads. j(u∞ ) = σ(ϕ. p)(∗ D θ)−1 · n0 . ˆ θ u = u∞ . on Γs . θ0 . p(u∞ ). p(u∞ ). p. in Ωs . θ(u∞ )) is a weak solution of problem (8. θ1 ).6) ˆ Here.8) admits a gradient given by the following expression. Qf Qf Σs Σf ∞ Ωf T (8. ∞ (8.6) and where ˆ is a real functional of the following form J(u. in Ωf × (Ωs )2 0 0 |t=0 ˆ with (u(u∞ ).  0         ˆ ˆ ˆ ∂tt θ − div0 [D θ · S(θ)] = f.9) with (ϕ.  0         ˆ ˆ ˆ ˆ  u.   ϕ = 0. θ) α ˆ Ju∞ (u.  ∗  −∂t ϕ − D ϕ · u + D u · ϕ − ν∆ϕ +   div ϕ = 0. on Γs . the strong coupled problem.7) ˆ ˆ u = ∂t θ ◦ θ−1 . ˆ j(u∞ ) = Ju∞ (u(u∞ ).

220 Moving Shape Analysis ˆ and ψ being a solution of the following backward solid adjoint system. ∂t ψ)|t=T = (0. T . ˆ 8. π) − Vθ · ϕ · n ˆ ˆ ˆ + −(∂t E) · n + E · ˆ Γ (Vθ · n) − (divΓ Vθ ) E · n − D E · Vθ · n − E · DΓ n · Vθ ˆ ˆ ˆ on Γθ (8. (H 2 (Ωf )) t . whose action is defined by the following identity. ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ D ψ · ∗ S(θ) · n + TΓ (∗ D θ · D ψ) ◦ θ−1 [J(θ)|∗ (D θ)−1 · n|]−1 ◦ θ−1 = ˆ − Vθ · ∂t ϕ − p div ϕ − π div Vθ + ν D ϕ · · D Vθ − (div Vθ ) Vθ · ϕ ˆ ˆ ˆ ˆ ˆ − (D ϕ · Vθ ) · ϕ · n − ∗ D ϕ · σ(u. u ∈ H 2 (0. (0. v s = T 0 Ωs 0 T 0 Γs 0 ˆ ˆ ˆ [θ · ∂tt v s + D θ · S(θ) · · D v s ] + ˆ ˆ ˆ ˆ [D θ · S(θ) · n0 ] · v s ˆ T − 0 Ωs 0 f ·ˆs + v Ωs 0 ˆ ˆ v ˆ [∂t θ(T )·ˆs (T )−θ(T )·∂t v s (T )]+ Ωs 0 ˆ v ˆ [−θ1 ·ˆs (0)+θ0 ·∂t v s (0)] ˆ 8.2 Fluid state operator Xf = def We define the fluid state spaces. 0 (8. π) − (Vθ · ϕ) I. ˆ ˆ es (θ).1 Solid weak state operator ˆ ˆ es : X s −→ (Y s )∗ We define the solid state operator. T ) × Ωs 0       ˆ ˆ Ωs (ψ.2.2.11) together with the following condition.  ˆ ˆ ˆ ˆ ˆ ˆ ˆ  ∂tt ψ − D ψ · div S(θ) − D2 ψ · ·S(θ) − TA (∗ D θ · D ψ) =      ˆ ˆ α (θ − θd ). p) · n + ∗ D Vθ · σ(ϕ. 0).12) ˆ with E = σ(ϕ.

q) = u T 0 T ˆ Ωf (θ) t −u · ∂t v f + (D u · u) · v f − νu · ∆v f + u · T 0 Γs ˆ θ q − p div v f + 0 Γf ∞ u∞ · (σ(v f . p. Hence. v f = v s . (H 1 (D))2 ) ˆ ˆ (v f . This means ˆ θ f that we shall choose the fluid and solid multipliers inside the space W s . we choose to work with continuous test functions on Γ s . (v f . H 1 (D)) We also need test function spaces that will be useful to define Lagrange multipliers : Yf = def def v f ∈ L2 (0. We define the coupled system weak state operator as follows. The major point here is to notice that the kinetic continuity of the stress holds on the fluid-structure interface. θ). It is now possible to couple these two operators in order to build an ad-hoc coupled system operator. ˆ with v f = v f ◦ θ. we need to decide whether or not the fluid and the solid multipliers match at the fluid-solid interface. T . v s ) ∈ Y f × Y s . we have to work with the fluid constraint tensor at the fluidsolid boundary. These subsystems have been represented thanks to a solid and a fluid state operator. ef ∞ : X f × Z f × U s × X s −→ (Y f × V f )∗ u whose action is defined by : ˆ ef ∞ (u. If not.3 Coupled system operator Our mechanical system consists of a solid part and a fluid part.2. T . us . T . which may be not convenient due to regularity requirement. on Γs ˆ ˆ 0 Vf = f Ws = def We define the fluid weak state operator. q) · n) + T us · (σ(v f . q) ∈ Y f × V f . To achieve the coupling. 8. q) · n) − (u · v f ) us . n u(T ) · v f (T ) − def − 0 Γs ˆ θ v f · (σ(u.Sensitivity analysis for a general fluid-structure interaction system Zf = def 221 p ∈ H 1 (0. f ˆ eu∞ : Y f × Z f × X s −→ (Ws × V f )∗ . p) · n) + ΩT Ω0 u0 · v f (0) ˆ vf ∈ Y f ˆ ∀ (v f . H 2 (Ωf )) t q ∈ H 1 (0.

θ. q.14) ˆ Reduced Gradient We assume that the conditions to apply the Min-Max principle [40] are fulfilled so we can bypass the derivation with respect to the control variable u∞ through the min-max subproblem (8. (v f . θ).14). p.4 Min-Max problem In this section.q)∈Ws ×V f v max ˆ Lu∞ (u.13) with ˆ Ju∞ (u. q. v s ) = Ju∞ (u. the cost function can be put in the following form : j(u∞ ) = ˆ (u. ∂t θ ◦ θ−1 . θ. we introduce the Lagrangian functional associated with problem (8. θ) − eu∞ (u. θ). q) u T = 0 ˆ Ωf (θ) t T −u · ∂t v f + (D u · u) · v f − νu · ∆v f + u · q − p div v f + 0 Γs ˆ θ ˆ ˆ ˆ ˆ (∂t θ ◦ θ−1 ) · (σ(v f .p.2. p. p. p. ˆ ˆ ˆ ˆ ˆ ˆ ˆ eu∞ (u. n T 0 Ωs 0 T 0 Ωs 0 T + 0 Γf ∞ u∞ ·(σ(v f . p. It leads to the following result : .222 Moving Shape Analysis whose action is defined by the following identity. θ) = α 2 0 T Ωs 0 ˆ ˆ | θ − θd | 2 Using this functional. q) ∈ Ws × V f ˆ + ΩT Ω0 8. q. v f . q)·n)+ ˆ ˆ ˆ [θ·∂tt v s +D θ·S(θ)·· D v s ]− ˆ ˆ f ·ˆs v + Ωs 0 ˆ ˆ ˆ ˆ ˆ [∂t θ(T ) · v s (T ) − θ(T ) · ∂t v s (T ) − θ1 · v s (0) + θ0 · ∂t v s (0)] ˆ ˆ ˆ u(T ) · v f (T ) − u0 · v f (0) f ∀ (v f . p.ˆs . q. v s ) = es (θ). p.θ)∈X f ×Z f ×X s min f (v f . v s ) + ef ∞ (u. (v f . v s . v s ) (8.18) : def ˆ ˆ ˆ ˆ ˆ Lu∞ (u. q) · n) − (u · v f ) (∂t θ ◦ θ−1 ). v f .6) and problem (7. (v f . θ). v s ) (8.

ψ) the unique saddle point of problem (8. q. v s ) · (δv f . π) · n|(0.ˆs ) Lu∞ (u. δ θ) = 0. δq) ∈ Ws × V f → State Equations v ˆ ˆ ˆ ∂(u. the gradient of the cost function j at point u∞ ∈ Uc is given by the following expression : j(u∞ ) = σ(ϕ. δˆs ) = 0. p. θ. 46].1 ˆ ∂u Lu∞ (u. δˆs . q. θ.14). θ. p. ˆ ˆ ∀ (δu.3 KKT optimality conditions In this section. q. We recall the expression of the Lagrangian.Sensitivity analysis for a general fluid-structure interaction system 223 THEOREM 8. p. δq. p. 8. π. we bypass the derivation with respect to u∞ inside the min-max problem (8. δu = ˆ T − 0 Ωf ˆ θ −δu · ∂t v f + [D δu · u + D u · δu] · v f − νδu · ∆v f + δu · T q + 0 Γs ˆ θ ˆ ˆ (δu · v f ) ∂t θ ◦ θ−1 . δp.1 Fluid adjoint system LEMMA 8. δ θ) ∈ X f × Z f × X s → Adjoint Equations 8.14).16) The KKT system will have the following structure : ˆ ∂(vf . θ). p. ϕ. n − ΩT δu(T ) · v f (T ) ∀δu ∈ X f . p.θ) Lu∞ (u. v s ). v s ) = Ju∞ (u. v s ) ˆ ˆ (8.T )×Γf + γ [∗ K K] · u∞ ∞ (8. v f . ˆ v v f ∀ (δv f . q. p. θ.14). and (u.p. v f .1 ˆ ˆ For u∞ ∈ Uc .q. θ. δp. π.3. v f . we are interested in establishing the first order optimality condition for problem (8. because it leads to the formulation of the ˆ adjoint problem satisfied by the Lagrange multipliers (ϕ. This step is crucial. (v f . θ) − eu∞ (u. q. ψ). v f . def ˆ ˆ ˆ Lu∞ (u.15) PROOF Using theorem (3) from [45. better known as Karusch-Kuhn-Tucker optimality conditions. v s ) · (δu.

we perform some integration by parts : LEMMA 8. ψ). ϕ. Q   div ϕ = 0. p. ∀δp ∈ Z f 8. Ωf T (8. Σf  ∞   ϕ(T ) = 0. Shape derivative tools We shall need the following shape derivative formulas. θ. δu = ˆ − Qf [−∂t ϕ + ∗ D u · ϕ − D ϕ · u − (div u) ϕ − ν∆ϕ + − Ωf T π] · δu ϕ(T ) · δu(T ) LEMMA 8.224 Moving Shape Analysis In order to obtain a strong formulation of the fluid adjoint problem.  ∗ f  −∂t ϕ − D ϕ · u + D u · ϕ − ν∆ϕ + π = 0. Σs   ϕ = 0.2 This leads to the following fluid adjoint strong formulation.17) Solid adjoint system Now we shall perform the differentiation of the Lagrangian with respect to ˆ ˆ θ.3 ˆ ˆ ∂p Lu∞ (u. π. We introduce a perturbation map δ θ together with a scalar ρ ≥ 0 and the following perturbed sets. p. ϕ. def ˆ ˆ ˆ Ωf θ = θ + ρδ θ (Ωf ) = (I +ρδθ) ◦ θ (Ωf ) 0 0 ˆ θ+ρδ ˆ def ˆ ˆ ˆ Γs θ = θ + ρδ θ (Γs ) = (I +ρδθ) ◦ θ (Γs ) ˆ 0 0 θ+ρδ ˆ def ˆ ˆ with δθ = δ θ ◦ θ−1 . π. ψ). n It leads to the following identity : ˆ ˆ ∂u Lu∞ (u.  Qf  ˆ ˆ ϕ = ψ ◦ θ−1 . δp = 0 T Ωf ˆ θ δp div ϕ. θ.2 Ωf ˆ θ (D δu · u) · v f = − Ωf ˆ θ D v f · u + div(u) v f · δu + Γf ∪Γs ∞ ˆ θ (δu · v f ) u.3. .

This ˆ ˆ θ θ .) ◦ (I +ρδθ)] dρ ρ=0 then the tangential shape derivative of φ is given by the following expression.5 d dρ Γs ˆ φ(ρ) dΓ ρ=0 = Γs ˆ θ φΓ + Hφ δθ. x)da dρ Γs ˆ ˆ θ+ρδ θ = ρ=0 Γs ˆ θ ˜ ˜ ˜ φ + H φ + ∂n φ δθ.18) LEMMA 8. Γs θ ). x) ∈ C 0 ((0.Sensitivity analysis for a general fluid-structure interaction system LEMMA 8. def ˙ φΓ = φ − Γ φ · δθ REMARK 8. . ρ=0 Γs ˆ θ Following this remark. then def ˜ ˜ φΓ = φ |Γs + ∂n φ δθ.4 d dρ G(ρ) dΩ ρ=0 225 Ωf ˆ = ˆ θ+ρδ θ Ωf ˆ θ ∂ρ G(ρ)|ρ=0 dΩ + G(ρ = 0) δθ. ρ0 .6 d ˜ φ(ρ. .20) Derivation of the perturbed Lagrangian f The map I +ρδθ sends respectively the sets (Ωf . n dΓ (8. n dΓ (8.4 we have ˜ If φ is the trace of a vector field φ defined over Ω.) ∈ L1 (Γs ) ˆ θ We recall classical definitions of shape derivative functions : DEFINITION 8.) dρ . n dΓ Γs ˆ θ (8. we have LEMMA 8. . n ˆ θ def ˜ with φ |Γs = ˆ θ d ˜ φ(ρ.19) ˆ θ+ρδ θ where φΓ stands for the tangential shape derivative of φ(ρ. the material ˆ θ+ρδ ˆ derivative is given by the following expression : d ˙ φ= [φ(ρ. C 1 (Γs θ )).1 For φ(ρ. Γs ) into (Ωθ+ρδθ . ˆ ˆ ˆ ˆ ˆ θ θ+ρδ ˆ θ def Then defining its inverse Rt = (I +ρδθ)−1 we can perturb the integral supρ port inside the Lagrangian and keep functions defined on Ω f and Γs .

q. v f .7 d Rt ρ dρ = −δθ q − p div(v f ◦ Rt ) ρ ρ=0 . v s .) = − (u ◦ Rt ) · ∂t (v f ◦ Rt ) + [D(u ◦ Rt ) · u ◦ Rt ] · (v f ◦ Rt ) ρ ρ ρ ρ ρ − ν(u ◦ Rt ) · ∆(v f ◦ Rt ) + (u ◦ Rt ) · ρ ρ ρ LEMMA 8. p. def ˆ ˆ ˆ Lρ ∞ = L(u. θ. ˆ f ∀ (v f . v s ) = Ju∞ (u.ρ T + Ω0 u0 · v f (0) + 0 Ωs 0 f · vs. θ + ρδ θ) ˆ u T − 0 Ωf ˆ ˆ θ+ρδ θ − (u ◦ Rt ) · ∂t (v f ◦ Rt ) + [D(u ◦ Rt ) · u ◦ Rt ] · (v f ◦ Rt ) ρ ρ ρ ρ ρ q − p div(v f ◦ Rt ) ρ − ν(u ◦ Rt ) · ∆(v f ◦ Rt ) + (u ◦ Rt ) · ρ ρ ρ T − 0 Γs ˆ ˆ θ+ρδ θ ˆ ˆ ˆ ˆ [∂t (θ + ρδ θ) ◦ (θ + ρ δ θ)−1 ] · (σ(v f ◦ Rt . q) · nρ ) ρ T ˆ ˆ ˆ ˆ −(u◦Rt )·(v f ◦Rt ) ∂t (θ+ρδ θ)◦(θ+ρδ θ)−1 . q) ∈ Ws × V f ˆ Fluid distributed terms We set G(ρ. q)·n) − 0 Ωs 0 ˆ ˆ ˆ ˆ ˆ ˆ (θ + ρδ θ) · ∂tt v s + D(θ + ρδ θ) · S(θ + ρδ θ) · · D v s ˆ ˆ ˆ ˆ ˆ ˆ ∂t (θ + ρδ θ)(T ) · v s (T ) − (θ + ρδ θ)(T ) · ∂t v s (T ) ˆ ˆ (u ◦ Rt )(T ) · (v f ◦ Rt )(T ) ρ ρ − Ωs 0 ˆ ˆ ˆ − θ1 · v s (0) + θ0 · ∂t v s (0) − ˆ T Ωf. p. . nρ ρ ρ T − 0 Γf ∞ u∞ ·(σ(v f .226 Moving Shape Analysis leads to the following perturbed Lagrangian.

) = ∂t (θ+ρδ θ)◦(θ+ρ δ θ)−1 · σ(v f ◦Rt . d dρ div E(ρ) ρ=0 Ωf ˆ = ˆ θ+ρδ θ Ωf ˆ θ div E + Γs ˆ θ (div E) δθ.)|ρ=0 = (D u · δθ) · ∂t v f + u · ∂t (D v f · δθ) 227 +ν(D u · δθ) · ∆v f + νu · ∆(D v f · δθ) + p div(D v f · δθ) − (D u · δθ) · − [(D(D u · δθ)) · u + D u · (D u · δθ)] · v f − (D u · u) · (D v f · δθ) q Then we have an expression for the derivative of fluid distributed terms coming from the Lagrangian with respect to ρ. n Fluid boundary terms We must now take into account the terms coming from the moving boundary Γs θ . q)−(u◦Rt )·(v f ◦Rt ) ·nρ ρ ρ ρ LEMMA 8. n First. ∂ρ G(ρ.Sensitivity analysis for a general fluid-structure interaction system Then we have the following derivative. nρ = Γs ˆ ˆ θ+ρδ θ Ωf ˆ div E(ρ) ˆ θ+ρδ θ then we derive this quantity using lemma (8. Then we set ˆ θ+ρδ ˆ ˆ ˆ ˆ ˆ φ(ρ.4). n + (div E) δθ. . d dρ G(ρ)dΩ ρ=0 Ωf ˆ ˆ θ+ρδ θ = Ωf ˆ θ (D u · δθ) · ∂t v f + u · ∂t (D v f · δθ) +ν(D u · δθ) · ∆v f + νu · ∆(D v f · δθ) + p div(D v f · δθ) − (D u · δθ) · + Γs ˆ θ − [(D(D u · δθ)) · u + D u · (D u · δθ)] · v f − (D u · u) · (D v f · δθ) −u · ∂t v f + (Du · u) · v f − νu · ∆v f + u · q − p div v f q δθ. n .9 d dρ PROOF E(ρ).8 [53] ∂ρ nρ |ρ=0 = nΓ = − LEMMA 8. nρ dΓ Γs ˆ ˆ θ+ρδ θ ρ=0 Γ (δθ · n) = Γs ˆ θ E |Γt . . we use that E(ρ).

n .228 We conclude using Moving Shape Analysis div E = Γs ˆ θ Ωf ˆ θ E . d dρ φ(ρ) dΓ Γs ˆ ˆ θ+ρδ θ = ρ=0 Γs ˆ θ Wδθ · −q I +ν D v f − u · v f · n + Γs ˆ θ Vθ · −ν D(D v f · δθ) + (D u · δθ) · v f + u · (D v f · δθ) · n ˆ + Γs ˆ θ div(Vθ · −q I +ν ∗ D v f − u · v f ) δθ. LEMMA 8.21) Using the last identities.10 def d ˆ ˆ ˆ ˆ Wδθ = ∂t (θ + ρδ θ) ◦ (θ + ρ δ θ)−1 dρ ˆ ˆ where we recall that Vθ = (∂t θ) ◦ θ−1 . we shall obtain. n ˆ REMARK 8.5 Γs ˆ θ We recall that Vθ · (D v f · n) = ˆ = Ωf ˆ θ div(∗ D v f · Vθ ) ˆ D v f · · D Vθ + Vθ · ∆v f ˆ ˆ (8. ˆ def ρ=0 ˆ ˆ = ∂t (δθ ◦ θ) ◦ θ−1 − D Vθ · δθ ˆ (8.22) Ωf ˆ θ Distributed solid terms We set ˆ ˆ ˆ ˆ ˆ ˆ m(ρ) = (θ + ρδ θ) · ∂tt v s + D(θ + ρδ θ) S(θ + ρδ θ) · · D v s ˆ ˆ . LEMMA 8.11 E |Γs = Wδθ · − q I +ν D v f − u · v f + Vθ · − ν D(D v f · δθ) + (D u · δθ) · v f ˆ ˆ θ + u · (D v f · δθ) It leads to the final expression.

23) . n Γs T + Ωf T −(D u·δθ)(T )·v f (T )−u(T )·(D v f ·δθ)(T ) + Complete derivative We set (v f .ρ T LEMMA 8. v s ) = ˆ Ωs 0 T 0 Ωs 0 ˆ ˆ ˆ − θ1 · v s (0) + θ0 · ∂t v s (0) − ˆ Ω0 u0 · v f (0) − f · vs ˆ The perturbed Lagrangian has the following form. d Lρ θ ˆ θ. d ρ J dρ =α ρ=0 0 Ωs 0 T Ωf ˆ T T G(ρ) + 0 Γs ˆ ˆ θ+ρδ θ φ(ρ) + 0 Ωs 0 m(ρ) + (ρ) + 0 ˆ θ+ρδ θ T ˆ ˆ ˆ ( θ − θd ) · δ θ Using the previous identities.Sensitivity analysis for a general fluid-structure interaction system LEMMA 8. Lρ = J ρ − Then.13 d (ρ) dρ = ρ=0 Ωs 0 ˆ ˆ (∂t δ θ)(T ) · v s (T ) − δ θ(T ) · ∂t v s (T ) ˆ ˆ u(T )·v f (T ) δθ(T ). we shall furnish the derivative of the perturbed Lagrangian Lρ with respect to ρ at point ρ = 0.12 d dρ T 229 m(ρ) 0 Ωs 0 T 0 Ωs 0 ρ=0 = ˆ ˆ ˆ ˆ ˆ ˆ δ θ · ∂tt v s + D δ θ · S(θ) + D θ · S (θ) · δ θ · · D v s ˆ ˆ Final time terms We set (ρ) = Ωs 0 ˆ ˆ ˆ ˆ ∂t (θ + ρδ θ)(T ) · v s (T ) − (θ + ρδ θ)(T ) · ∂t v s (T ) ˆ ˆ + (u ◦ Rt )(T ) · (v f ◦ Rt )(T ) ρ ρ Ωf.δ ˆ dρ ρ=0 = − A Ωf + B Γ s + C Γ s + D Ωs + F Γ f 0 ˆ ˆ ∞ ˆ θ θ θ (8.

n + 0 T Γs ˆ θ + 0 T Γs ˆ θ Vθ · −ν D(D v f · δθ) + (D u · δθ) · v f + u · (D v f · δθ) · n ˆ div(Vθ · −q I +ν ∗ D v f − u · v f ) δθ. n ˆ + 0 Γs ˆ θ + Γs T T u(T ) · v f (T ) δθ(T ).230 with T Moving Shape Analysis A Ωf = ˆ θ 0 Ωf ˆ θ (D u · δθ) · ∂t v f − [(D(D u · δθ)) · u q + D u · (D u · δθ)] · v f + ν(D u · δθ) · ∆v f − (D u · δθ) · T + 0 Ωf ˆ θ u · ∂t (D v f · δθ) − (D u · u) · (D v f · δθ) + νu · ∆(D v f · δθ) − (D u · δθ)(T ) · v f (T ) − u(T ) · (D v f · δθ)(T ) + p div(D v f · δθ) + T Ωf T B Γs = ˆ θ 0 Γs ˆ θ T − u · ∂t v f + (Du · u) · v f − νu · ∆v f + u · (D Vθ · δθ) · −q I +ν D v f − u · v f · n ˆ q − p div v f δθ. n CΓs = ˆ θ 0 Γs ˆ θ ˆ ˆ (∂t δ θ) ◦ θ−1 · −q I +ν D v f − u · v f · n T D Ωs = 0 0 Ωs 0 T ˆ ˆ ˆ ˆ ˆ ˆ δ θ · ∂tt v s + D δ θ · S(θ) + D θ · S (θ) · δ θ · · D v s ˆ ˆ ˆ ˆ ˆ ( θ − θd ) · δ θ + ˆ ˆ (∂t δ θ)(T ) · v s (T ) − δ θ(T ) · ∂t v s (T ) ˆ ˆ −α 0 Ωs 0 T Ωs 0 FΓ f = − ∞ 0 Γf ∞ ν u∞ · D(D v f · δθ) · n .

where we take into account the normal stress on the moving interface by choosing test functions discontinuous . This will help us to simplify several terms involved in the derivative of L with respect to ρ.Sensitivity analysis for a general fluid-structure interaction system The shape derivative kernel identity 231 We shall now assume that (u. ΩT ˆ ˆ ∀δ θ ∈ Y s PROOF We define a new Lagrangian. LEMMA 8. π) is a saddle point of the Lagrangian functional L.14 T 0 Ωf ˆ θ [(D u · δθ) · ∂t ϕ − [(D(D u · δθ) · u) + D u · (D u · δθ)] · ϕ + ν(D u · δθ) · ∆ϕ − (D u · δθ) · π] + [u · ∂t (D ϕ · δθ) − (D u · u) · D ϕ · δθ + νu · ∆(D ϕ · δθ) + p div(D ϕ · δθ)] T + 0 Γs ˆ θ ν(ϕ − v s ◦ θ−1 ) · (D(D u · δθ) · n) + (D ϕ · δθ) · (−p n + ν D u · n) ˆ ˆ T − T 0 Γs ˆ θ Vθ · [ν D(D ϕ · δθ) − (D u · δθ) · ϕ − u · (D ϕ · δθ)] · n ˆ − 0 Γf ∞ νu∞ ·(D(D ϕ·δθ)·n)− (D u·δθ)(T )·ϕ(T )+u(T )·(D ϕ·δθ)(T ) = 0. ϕ. p.

∂(u. p. v s ) = Ju∞ (u..vf ) Ld ∞ · (δu. q. q)·n)− ˆ ˆ ˆ [θ·∂tt v s +D θ)·S(θ)·· D v s ]+ ˆ ˆ f ·ˆs v − Ωs 0 ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ [∂t θ(T ) · v s (T ) − θ(T ) · ∂t v s (T ) − θ1 · v s (0) + θ0 · ∂t v s (0)] u(T ) · v f (T ) + u0 · v f (0) − ΩT Ω0 ˆ ∀ (v f .232 Moving Shape Analysis ˆ at the fluid-structure interface. ˆ ˆ ˆ Ld ∞ (u. v f . q) ∈ Y f × Y s × V f ˆ Let us differentiate the above functional with respect to (u. p) · n) ˆ ˆ ˆ T 0 Ωs 0 T 0 Ωs 0 − 0 Γf ∞ u∞ ·(σ(v f .e. θ) ˆ u T − 0 Ωf ˆ θ −u · ∂t v f + (D u · u) · v f − νu · ∆v f + u · q − p div v f T − T 0 Γs ˆ θ Vθ · σ(v f . v). θ. δv f ) u T = 0 Ωf ˆ θ δu · ∂t v f − [(D δu · u) + (D u · δu)] · v f + νδu · ∆v f − δu · T q + 0 Ωf ˆ θ u · ∂t δv f − (D u · u) · δv f + νu · ∆δv f + p div δv f T 0 Γs ˆ θ T − 0 Γf ∞ νu∞ ·(D δv f ·n)+ ν(v f −ˆs ◦θ−1 )·(D δu·n)+δv f ·(−p n+ν D u·n) v ˆ . (v f . i. v s ) ∈ Y f × Y s . p. q) − (u · v f ) · n + (v f − v s ◦ θ−1 ) · (σ(u. v s .

Sensitivity analysis for a general fluid-structure interaction system T 233 − 0 Γs ˆ θ Vθ · ν D δv f − δu · v f − u · δv f · n ˆ − δu(T ) · v f (T ) + u(T ) · δv f (T ) ˆ ∀ (v f . we have ˆ ˆ ϕ = ψ ◦ θ−1 . ϕ. v s .ϕ. π. q) ∈ Y f × Y s × V f ˆ ΩT The first order optimality conditions corresponding to the saddle point formulation with respect to (u. u∞ (u. p) · n .vf ) Ld | ∀ (v f .θ) We choose specific perturbations. δv ) = 0. ψ) and the distributed AΩf only involves boundary terms. v s . We have T on Γs ˆ θ A Ωf + F Γ f = ∞ ˆ θ 0 Γs ˆ θ Vθ · [ν D(D ϕ · δθ) − (D u · δθ) · ϕ − u · (D ϕ · δθ)] · n ˆ T − 0 Γs ˆ θ (D ϕ · δθ) · σ(u. v f ) writes f ˆ ∂(u. Solid adjoint equation Using the shape derivative kernel identity. δu = D u · δθ. p.17). q) ∈ Y f × Y s × V f ˆ ˆ · (δu. A Ωf = ˆ θ ˆ θ δv f = D v f · δθ T 0 T Γs ˆ θ Vθ · [ν D(D ϕ · δθ) − (D u · δθ) · ϕ − u · (D ϕ · δθ)] · n ˆ − 0 Γs ˆ θ ˆ ˆ ν(ϕ − ψ ◦ θ−1 ) · (D(D u · δθ) · n) + (D ϕ · δθ) · (−p n + ν D u · n) T + 0 Γf ∞ νu∞ · (D(D ϕ · δθ) · n) Using the fluid adjoint system (8. we simplify the Lagrangian ˆ derivative at the saddle point (u.

Moving Shape Analysis div(Vθ · [−π I +ν ∗ D ϕ − u · ϕ]) = −π div Vθ − Vθ · ˆ ˆ ˆ π + ν D ϕ · · D Vθ ˆ + ν Vθ ∆ϕ − (div Vθ ) u · ϕ − (D u · Vθ ) · ϕ − (D ϕ · u) · ϕ ˆ ˆ ˆ Using the above identity. n + 0 Γs ˆ θ (D Vθ · δθ) · [−π I +ν D ϕ − u · ϕ] · n ˆ T + 0 Γs ˆ θ Vθ · [−ν D(D ϕ · δθ) + (D u · δθ) · ϕ + u · (D ϕ · δθ)] · n ˆ T + 0 Γs ˆ θ − π div Vθ − Vθ · ˆ ˆ π + ν D ϕ · · D Vθ + ν Vθ ∆ϕ − (div Vθ ) u · ϕ ˆ ˆ ˆ − (D u · Vθ ) · ϕ − (D ϕ · Vθ ) · ϕ δθ.15 The following identity holds. we get B Γs = ˆ θ T 0 Γs ˆ θ − Vθ · ∂t ϕ + (Du · Vθ ) · ϕ − νVθ · ∆ϕ + Vθ · ˆ ˆ ˆ ˆ T π − p div ϕ δθ. n Several terms can be eliminated inside the above expressions. Furthermore.234 LEMMA 8. . n + ˆ ˆ Γs T u(T ) · ϕ(T ) δθ(T ).

LEMMA 8. T 0 Γθ ˆ ˆ ˆ ∂t (δ θ) ◦ θ−1 · E = T 0 Γθ ˆ −∂t E − (divΓ Vθ ) E − D E · Vθ · δθ ˆ ˆ T + Γs ˆ θ E · δθ 0 (8. we obtain ∞ ˆ θ 235 A Ωf + B Γ s + F Γ f = ˆ ∞ ˆ θ θ T 0 Γs ˆ θ − Vθ · ∂t ϕ − p div ϕ − π div Vθ + ν D ϕ · · D Vθ ˆ ˆ ˆ − (div Vθ ) u · ϕ − (D ϕ · Vθ ) · ϕ δθ. π) − Vθ · ϕ · n ˆ We shall use the following fundamental adjoint identity.Sensitivity analysis for a general fluid-structure interaction system adding the terms AΩf + FΓf .16 For any smooth E defined in the hold-all domain D. n We now turn to the analysis of the term CΓs . ˆ θ T CΓs = ˆ θ 0 Γs ˆ θ ˆ ˆ (∂t δ θ) ◦ θ−1 · σ(ϕ. n − (D ϕ · δθ) · (−p n + ν D u · n) ˆ ˆ T + 0 Γs ˆ θ (D Vθ · δθ) · [−π I +ν D ϕ − u · ϕ] · n ˆ + Γs T u(T ) · ϕ(T ) δθ(T ).24) .

4) with B = ∗ D θ · D ψ and we get Ωs 0 ˆ ˆ T (δθ) · ·(∗ D θ · D ψ) = − Ωs 0 ˆ ˆ δθ · TA (∗ D θ · D ψ) + Γs 0 ˆ ˆ ˆ TΓ (∗ D θ · D ψ) · δ θ (8.25) .j nj T 0 Γθ ˆ ˆ ˆ ∂t (δ θ) ◦ θ−1 · (E · n) = T 0 Γθ ˆ −(∂t E) · n + E · ˆ Γ (Vθ · n) − (divΓ Vθ ) E · n ˆ T − D E · Vθ · n − E · DΓ n · Vθ · δθ + ˆ ˆ Using the previous identity. D Ωs = 0 T 0 Ωs 0 T Γs ˆ θ (E · n)(T ) · δθ(T ) ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ δ θ · ∂tt ψ + D δ θ · S(θ) + D θ · S (θ) · δ θ · · D ψ −α 0 Ωs 0 ˆ ˆ ˆ ( θ − θd ) · δ θ + Ωs 0 ˆ ˆ ˆ ˆ (∂t δ θ)(T ) · ψ(T ) − δ θ(T ) · ∂t ψ(T ) The Green formula leads to the following identity. ˆ Let us now analyse the solid distributed term. π) − (Vθ · ϕ) I and δθ(0) = 0. we get T Γs ˆ θ (E · n) · δθ 0 CΓs = ˆ θ 0 Γθ ˆ −(∂t E) · n + E · ˆ Γ (Vθ · n) − (divΓ Vθ ) E · n ˆ − D E · Vθ · n − E · DΓ n · Vθ · δθ + ˆ ˆ with E = σ(ϕ. with (E)i = Ei.236 Moving Shape Analysis Let us apply this lemma. ˆ ˆ ˆ [D δ θ · S(θ)] · · D ψ = − + Γs 0 Ωs 0 Ωs 0 ˆ ˆ ˆ ˆ ˆ D ψ · div S(θ) + D2 ψ · ·S(θ) · δ θ ˆ ˆ ˆ D ψ · ∗ S(θ) · n · δ θ ˆ ˆ We use the adjoint identity (8.

0). ∂t ψ)|t=T = (0. D Ωs = 0 T 0 Ωs 0 T 237 ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ∂tt ψ − D ψ · div S(θ) − D2 ψ · ·S(θ) − TA (∗ D θ · D ψ) − α (θ − θd ) · δ θ + 0 Γs 0 ˆ ˆ ˆ ˆ ˆ D ψ · ∗ S(θ) · n + TΓ (∗ D θ · D ψ) · δ θ + Ωs 0 ˆ ˆ (∂t δ θ)(T ) · ψ(T ) ˆ ˆ − δ θ(T ) · ∂t ψ(T ) Using the optimality condition (8. . T ) × Ωs 0       ˆ ˆ (ψ. π) − u · ϕ] · n ˆ ˆ + −(∂t E) · n + E · ˆ Γ (Vθ · n) − (divΓ Vθ ) E · n − D E · Vθ · n − E · DΓ n · Vθ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ ˆ + D ψ · ∗ S(θ) · n + TΓ (∗ D θ · D ψ) ◦ θ−1 [J(θ)|∗ (D θ)−1 · n|]−1 ◦ θ−1 ˆ with E = σ(ϕ. (0. we shall deduce that the adjoint solid state is the solution of the following backward second order system. ˆ This last identity ends the proof of the main result of this chapter.26) Furthermore. π) − (Vθ · ϕ) I.23).Sensitivity analysis for a general fluid-structure interaction system This leads to the following expression.  ˆ ˆ ˆ ˆ ˆ ˆ ˆ  ∂tt ψ − D ψ · div S(θ) − D2 ψ · ·S(θ) − TA (∗ D θ · D ψ) =      ˆ ˆ α (θ − θd ). ˆ 0 = − Vθ · ∂t ϕ − p div ϕ − π div Vθ + ν D ϕ · · D Vθ − (div Vθ ) u · ϕ ˆ ˆ ˆ ˆ − (D ϕ · Vθ ) · ϕ · n − ∗ D ϕ · σ(u. Ωs 0 (8. the following identity holds on the moving boundary Γ θ . p) · n + ∗ D Vθ · [σ(ϕ.

.

p→∞ d(xn . |α| = derivative and Dα f = def 1≤i≤d αi is the order of Dαi f = i ∂xα1 1 ∂ |α| f . C k (Ω) = def f ∈ C k−1 (Ω) : Dα f ∈ C 0 (Ω). ∀ x.1) where α = (α1 . . . A norm on a vector space E is a real-valued functional m on E such that • m(x) ≥ 0. ∀α.Appendix A Functional spaces and regularity of domains A. we can define a sequence of semi-norms dn (x) = supx∈Kn |u(x)| 1A complete normed space. ∀ x ∈ E with equality iff x = 0. ∀ x ∈ E.1 The spaces C 0 (Ω) and C k (Ω) for k ∈ N∗ are not Banach1 spaces. 239 . • m(x + y) ≤ m(x) + m(y). x αd d (A. y ∈ E c ∈ C. . |α| = k 1≤i≤d (A. However if {Kn }n∈N is a sequence of compact sets of Ω such that Ω= n∈N Kn . xp ) = 0. . .1 Classical functions Let Ω be an open subset of Rd . and we denote by C 0 (Ω) the space of continuous functions from Ω to R. • m(c x) = |c| m(x). αd ) ∈ Nd is a multi-index.2) REMARK A. A Cauchy sequence in a metric space is a sequence {xn }n≥0 which satisfies limn. For k ∈ N∗ . A normed space is a vector space which is provided with a norm. . A metric space is said to be complete in a metric space iff every Cauchy squence is convergent.

∀x. |f (y) − f (x)| ≤ c|y − x| . functions in C k (Ω) can be unbounded. A topological space is said to be metrizable if its topology can be defined by a metric. y ∈ Ω 2 A Fr´chet space is complete metrizable locally convex topological vector space. d(. we say that f is Lipschitz. ∀x. ∀x. we define B k (Ω) = f ∈ B k−1 (Ω) : Dα f ∈ B 0 (Ω). Then C 0 (Ω). ∃ c > 0. we can define on C k (Ω) a Fr´chet space structure with the e following sequence of semi-norms with an ≥ 0 such that an = 1..3) = 0≤|α|≤k x∈Ω max sup | Dα f (x)| B k (Ω) is a Banach space. let us denote by B 0 (Ω) the space of bounded continuous functions. For k ∈ N ∗ . such that the vector space operations of addition and scalar multiplication are continuous. . )-H¨lder continuous in Ω if o ∃ c > 0. If a function f is bounded and uniformly continuous 3 on Ω. y ∈ Ω When = 1. 0 ≤ |α| ≤ k. it has a unique. f is (k. Given 0 < ≤ 1. Endowed with the norm f def C k (Ω) ∀ α. Similarly for k ≥ 1. C k (Ω) ) is a Banach space. a function f is (0. i. Thus. y ∈ Ω with |x − y| < δ . | Dα f (y) − Dα f (x)| ≤ c|y − x| .2 e dk (u) = n |α|≤k dn (Dα u) Since Ω is open. v) = n∈N an n∈N In a similar way. e E is a topological vector space if E is furnished with a topology compatible with the structure of the vector space.. )-H¨lder o continuous in Ω if ∀α. ∀ 0 ≤ |α| ≤ k The space (C k (Ω). 3 A function f : Ω → R is uniformly continuous if for each ε > 0 there exists δ > 0 such that |f (x) − f (y)| < ε. continuous extension to the closure Ω.240 and a distance Moving Shape Analysis dn (u − v) 1 + dn (u − v) d(u.) is a Fr´chet space. Hence we introduce the space C k (Ω) = f ∈ B k (Ω) : Dα f uniformly continuous on Ω.e. It is locally convex if each neighborhood of the origin in E contains a convex neighborhood. |α| = k (A. .

4) C k. The space L∞ (Ω) is the set of functions on Ω that are measurable with respect to the Lebesgue measure dx and have a finite norm u L∞ (Ω) = dx −meas{x∈Ω:|u(x)|>C}=0 inf C (A.8) . for a ≥ 0.5) We note that we identify functions that are different only on a set of Lebesgue measure zero. (A. The spaces D k (Ω) for k ∈ N∗ and D(Ω) contain respectively all the functions of class C k and infinitely continuously differentiable with compact support.2.2 This identity is a consequence of the Young inequality. (Ω) the space of all (k. a b ≤ 1/p ap + 1/q bq . (Ω) We denote by C k. y ∈ Ω x=y | Dα f (y) − Dα f (x)| |y − x| (A. o def = f C k (Ω) + max 0≤|α|≤k sup x. A. (Ω) ∪ C k (Ω). b ≥ 0. (Ω) = C k. 1/p + 1/q = 1. )-H¨lder continuous functions in Ω.1 Various inequalities LEMMA A. 1 ≤ p < +∞ is the set of real functions defined on Ω that have finite norm u Lp (Ω) def 1/p Ω = |u(x)|p dx (A. 1/p + 1/q = 1. The space Lp (Ω).6) A.1 H¨lder inequality o u(x) v(x) dx ≤ u Lp (Ω) v Lq (Ω) . (Ω) is a Banach space.2 Lebesgue spaces Let Ω be any open set in Rd . Endowed with the norm f def C k. (A.7) Ω for u ∈ Lp (Ω) . REMARK A.Functional spaces and regularity of domains 241 We also define C k. v ∈ Lq (Ω).

242 REMARK A.3 inequality.

Moving Shape Analysis For p = q = 2, this inequality is referred to as the Cauchy

LEMMA A.2 Interpolation inequality u
Lr (Ω)

≤ u

α Lp (Ω)

u

1−α Lq (Ω) ,

1/r = α/p + (1 − α)/q,

(A.9)

for u ∈ Lp (Ω) ∩ Lq (Ω).

A.2.2

Completeness, separability and reflexivity

LEMMA A.3 For 1 ≤ p ≤ +∞, the space Lp (Ω) is a Banach space4 . For 1 < p < +∞ the space Lp (Ω) is a separable reflexive space and its dual space can be identified with Lq (Ω) with 1/p + 1/q = 1. The space L1 (Ω) is separable and its dual can be identified with L∞ (Ω).

A.2.3

Convolution

THEOREM A.1 Let u ∈ L1 (Rd ) and v ∈ Lp (Rd ) with 1 ≤ p ≤ ∞. Then y → u(x − y) v(y) is summable over Rd for a.e. x ∈ Rd . We define their convolution product by u v(x) = which satisfies u v ∈ Lp (Rd ) and u v
Lp

Rd

u(x − y) v(y) dy

(A.10)

≤ u

L1

v

Lp

(A.11)

The convolution can be extended to a function u ∈ L p (Rd ) thanks to the following theorem, THEOREM A.2 Young inequality Let u ∈ Lp (Rd ) and v ∈ Lq (Rd ) with 1 ≤ p ≤ ∞, 1 ≤ q ≤ ∞, 1/r = 1/p + 1/q − 1 ≥ 0 then u v ∈ Lr (Rd ) and
4 Fischer-Riez

(A.12)

u v

Lr

≤ u

Lp

v

Lq

(A.13)

theorem [17] Theorem IV. 8, p. 57

Functional spaces and regularity of domains

243

The following regularizing property will be useful to state a density property of regular functions space in Lp , THEOREM A.3 Let g ∈ D k (Rd ) for k ∈ N and u ∈ L1 (Rd ), then loc g u ∈ C k (Rd ) and Dα (g u) = (Dα g) u, |α| ≤ k (A.14)

REMARK A.4 As an application of the previous theorem, we deduce that if g ∈ D(Rd ) and u ∈ L1 (Rd ) then g u ∈ C ∞ (Rd ). loc

A.2.4

Density property

THEOREM A.4 [44] The space D(Ω) is dense in Lp (Ω) for 1 ≤ p < +∞ and every convergent sequence in D(Ω) converges in Lp (Ω). REMARK A.5 This theorem can be established using the truncation + regularization procedure introduced by Leray and Friedrichs that will be introduced in the next section.

A.3

Smooth domains and boundary measure
def

Let Ω be an open set of Rd and let its associated boundary Γ = ∂Ω be a manifold in Rd of codimension 1. The smoothness of Γ is classically characterized by introducing at each point of Γ a local diffeomorphism that locally flattens the boundary. In order to make the purpose more precise, let us introduce some notations and definitions. Let {ei }1≤i≤d be the canonical unit orthogonal basis in Rd . We use the notation ζ = (ζ , ζd ) for a point ζ ∈ Rd with ζ = (ζ1 , . . . , ζd−1 ). Let B be the unit open ball in Rd and let us define the following sets, B0 = {ζ ∈ B : ζd = 0} B+ = {ζ ∈ B : ζd > 0} , B− = {ζ ∈ B : ζd < 0} DEFINITION A.1
def def def def

(A.15)

(A.16)

Let Ω be a subset of Rd such that Γ = ∂Ω = ∅.

244

Moving Shape Analysis

a) Ω is said to be of class C k , 0 ≤ k ≤ +∞, if for each point x ∈ Γ there exist – a neighbourhood U (x) ⊂ Rd of x,

– a bijective map gx : U (x) → B with the following properties :
−1 gx ∈ C k (U (x), B), hx = gx ∈ C k (B, U (x)) def

(A.17) (A.18) (A.19)

int[Ω ∩ U (x)] = hx (B + ) Γx = Γ ∩ U (x) = hx (B0 ), B0 = gx (Γx )
def

b) Ω is said to be of class C k, , k ≥ 0, 0 < ≤ 1, if the above conditions are satisfied with a map gx ∈ C k, (U (x), B) together with its inverse hx ∈ C k, (B, U (x)). For sets of class C 1 , the unit exterior normal n to the boundary Γ can be characterized through the Jacobian matrices of g x and hx . LEMMA A.4 The outward unit normal field n(y) at point y ∈ Γx is given by the following expressions, n(y) = −
=
with (D g)i,j = ∂j gi . The family of neighbourhoods U (x) associated with all the points x of Γ is an open cover of Γ. LEMMA A.5 Let Ω ⊂ Rd be a bounded domain with compact boundary Γ. There exists a finite subcover, i.e., there exists a finite sequence of points {x i }1≤i≤m of Γ such that Γ ⊂ U1 ∪ . . . ∪ Um , where Ui = U (xi ). Furthermore there exists a partition of unity {ri }1≤i≤m associated to the family of open neighbourhoods {Ui }1≤i≤m of Γ such that : ri ∈ D(Ui ), 1≤i≤m ri (x) = 1, 0 ≤ ri (x) ≤ 1, ∀x ∈ U (A.22)

(D hx )−1 (ζ , 0) · ed , ∗ (D h )−1 (ζ , 0)e | | x d
− D gx (y)ed , |∗ D gx (y) · ed |

∀ hx (ζ , 0) = y ∈ Γx
∀ y ∈ Γx

(A.20)

(A.21)

Functional spaces and regularity of domains ¯ with U ⊂ Ui .
1≤i≤m

245

Let us define the boundary integral of a continuous function f on Γ : LEMMA A.6 For f ∈ C 0 (Γ), its integral over Γ is defined as follows f dΓ =
Γ def 1≤i≤m Γi

f ri dΓi

(A.23)

where f ri dΓi =
Γi def B0

(f ri ) ◦ hi (ζ , 0)ωi (ζ )dζ

with Γi = U (xi ) ∩ Γ and ωi = Ωxi and ωx (ζ ) = |mx (hx (ζ , 0))|| det D hx (ζ , 0)| with mx (y) = −∗ D gx (y)ed

Appendix B
Distribution spaces

B.1

Let Ω be an open set in Rd . The spaces D k (Ω) for k ∈ N∗ and D(Ω) contain respectively all the functions of class C k and infinitely continuously differentiable with compact support1 in Ω. For any open set O ⊂ Rd such that Ω ⊂ O, we define
D(Ω) = {φ|Ω : φ ∈ D(O)}
(B.1) The space D(Ω) does not reduce to the null function. Indeed let ω : R d → R defined as follows : 2 e1/(|x| −1) , |x| < 1 ω(x) = 0, |x| ≥ 1 Let Br (a) be an open ball contained in Ω, then the function ϕa,r (x) = ω( x−a ) r

The space D(Ω)

belongs to D(Ω). There exists a general procedure for the construction of functions in D(Ω) based on the previous particular function and the notion of convolution product. DEFINITION B.1 mollifier mollifier if • ρε ∈ D(Rd ), • Supp ρε ⊂ B(0, ε), • ρε ≥ 0 and
1 For

We call a family of function {ρε }ε≥0 a

Rd

ρε = 1.

f : Ω → R, the support of f is defined as follows Supp f = {x ∈ Ω : f (x) = 0}

247

ε). . • Supp ψε ⊂ Kε = ∪x∈K B(x. REMARK B. This lemma shows that the space D(Ω) is sufficiently rich and we shall see that it has density properties in numerous common functional spaces. It converges to ϕ ∈ D(Ω) if there exists a compact subset K ⊂ Ω such that Supp ϕk ⊂ K.2 with K = B(0.1 The space D(Rd ) is contained and dense in C k (Rd ) for 0 ≤ k ≤ ∞ and every convergent sequence in D(Rd ) converges in C k (Rd ).248 Moving Shape Analysis An example of a mollifier is furnished by ρε (x) = ε−d ϕ0.2 Let {ϕk }k≥0 be sequence of elements of D(Ω). we can state a useful density property for D(R d ). Then there exists a function ψε ∈ D(Rd ) such that • ψε (x) = 1.1 The proof consists of two steps : • Truncation : We prove that D k is dense in C k using the family up (x) = ψ(x/p) u(x) ∈ Dk with ψ given by lemma B. The space D(Ω) is neither a Banach space nor a Fr´chet space. THEOREM B. but the induce tive limit of Fr´chet spaces which makes it a locally convex topological vector e space. Then there exists ε0 > 0 such that uε = ρε u ∈ D(Ω) for ε < ε0 .1 (x/ε) (B.2 Generalized Urysohn Let K be a compact set in Rd and ε > 0 arbitrary. With the above definition. ∀k ≥ 0 Dα ϕk converges uniformly on K to D α ϕ as k → ∞ LEMMA B.2) LEMMA B. DEFINITION B. ∀ x ∈ K and ψε ≥ 0. Fortunately. But the corresponding inductive limit topology is not easy to manipulate.1 Let u be a function locally integrable in Ω with compact support subset of Ω. the notion of convergence of sequences will be sufficient for our purpose. 1).

∀ {ϕk }k∈N ∈ D(Ω) with lim ϕk = ϕ in D(Ω) k→∞ The space D (Ω) is a vector space and is in fact the topological dual space of D(Ω).2 If {Tk }k∈N is a sequence of elements in D (Ω) such that Tk . ϕk = T. ∀ϕ ∈ D(Ω) THEOREM B. then the linear form ϕ → lim Tk . ϕ for ϕ ∈ D(Ω) and where . In this topology. we note T. a sequence of distributions T k converges to T if and only if Tk . given any compact subset K ⊂ Ω. . which is again not metrizable. ϕ k→∞ 2 Equivalently. ∀ ϕ ∈ D(Ω) with Supp ϕ ⊂ K When the integer m can be chosen independently on K. the distribution T is said to be of finite order and its order is defined as the smallest suitable integer m. ϕ →k→∞ T. stands for the duality pairing between D (Ω) and D(Ω).3 A continuous linear form on D(Ω) is called a distribution on Ω. defined by the family of semi-norms θϕ : D (Ω) → R T → θϕ (T ) = T (ϕ) = T. As a dual space. ϕ . .2 The space of distributions D (Ω) DEFINITION B. If T ∈ D (Ω). D (Ω) is equipped with the weak-star topology σ(D (Ω). We denote by D (Ω) the space of distributions on Ω.. ∀ ϕ ∈ D(Ω) Here T continuous2 on D(Ω) means k→∞ lim T. B. D(Ω)). ϕ . there exist a constant C(K) and an integer m(K) ≥ 0 such that |T (ϕ)| ≤ C(K) |α|≤m(K) x∈Ω max sup | Dα ϕ(x)|. ϕ = T (ϕ). ϕ is convergent for all ϕ ∈ D(Ω) as k → ∞.Distribution spaces 249 • Regularization : We show that D is dense in D k using the sequence up = u ρp ∈ D with ρp = ρε=1/p and ρε a mollifying family.

REMARK B.2 The identification is made in the following manner : let T be a continuous linear form on E. then the restriction of T to D(Ω) is the distribution we are looking for.e. DEFINITION B.4 Letting T ∈ D (Ω) . ∀ ϕ ∈ D(Ω) Hence the duality pairing . Then the continuous linear forms on E are identified with distributions ( i. The differentiation of distributions (which coincides with the usual differentiation when applied to continuously differentiable functions) is the essential property which justifies the introduction of distributions. ϕ = (−1)|α| T.3 Let {Tk }k∈N ∈ D (Ω) be a convergent sequence of distributions towards T ∈ D (Ω). the derivative Dα T . ϕ = Ω f (x) ϕ(x) dx . Moving Shape Analysis When f is locally integrable on Ω. Dα ϕ . ∀ ϕ ∈ D(Ω) THEOREM B. α ∈ Nd is the distribution on Ω defined by Dα T.. elements of D (Ω)). . ∀ α ∈ Nd THEOREM B. then f can be identified with a distribution by f. Then the derivative of T has the following property k→∞ lim Dα Tk = Dα T in D (Ω). . is an extension of the scalar product of the Hilbert space (L2 (Ω). L2 (Ω) ) .. . • every convergent sequence in D(Ω) converges in E.250 belongs to D (Ω).4 Let Ω be an open set of Rd and E a locally convex topological vector space of functions on Ω satisfying : • D(Ω) is contained and dense in E.

] : L1 (Ω) → M (Ω) loc f → [f ] = µf is an injection and its image is called the set of absolutely continuous measures in Ω. Using that D(Ω) → D 0 (Ω).3.2 Regular distributions We recall that L1 (Ω) is the set of functions integrable on every compact loc set in Ω. DEFINITION B. f is called the density of µf with respect to the Lebesgue measure dx ( dµf = f (x) dx ).3 As Lp (Ω) → L1 (Ω) with 1 ≤ p ≤ +∞.3. . we define a corresponding measure loc µf (ϕ) = Ω f (x)ϕ(x) dx The application [. We usually note µ(ϕ) = µ. The space of measures is denoted M (Ω). We also say that µ(ϕ) is the integral of the function ϕ relative to the measure µ and we write µ(ϕ) = Ω ϕ(x)dµ We can extend the definition of µ(ϕ) for measurable functions that do not belong to D 0 (Ω). Furthermore µf defines a distribution and we say that this distribution is a function f . ϕ the value of the linear form µ on ϕ ∈ D 0 (Ω).Distribution spaces 251 B. every element loc of Lp (Ω) defines a distribution on Ω ( which is a measure on Ω).6 µf by Let f ∈ L1 (Ω).1 Examples of distributions Measures on Ω DEFINITION B. we can state that µ ∈ M (Ω) defines an element of D (Ω). B. It is usual to identify µf with its density f . These are called summable functions. REMARK B.5 We call every continuous linear form on the space D0 (Ω) of continuous functions with compact support a measure on Ω.3 B.

x < 0 Obviously Y ∈ L1 (R).8 Heaviside function Y :R→R x → Y (x) : Let +1.3.3 The Dirac measure Let x0 ∈ Ω.3. Dm δ. ϕ = (−1)m δ. We can then comloc pute its derivation in D (R). x > 0 0. The mapping DEFINITION B. +∞ D Y.7 Dirac measure δx0 : D0 (Ω) → R ϕ → ϕ(x0 ) defines a measure in M (Ω) called the Dirac measure at point x0 . Dm ϕ = (−1)m Dm ϕ(0) . ϕ = − then ϕ (x) dx = ϕ(0) 0 D Y = δ.252 Moving Shape Analysis B. However we shall abuse the notation and write sometimes δ(x − x0 ) for δx0 . Using integral notation.4 Derivative of the Dirac measure Let us compute the derivative of the Dirac measure in the sense of distributions. in D (R) B. Then it defines a distribution.4 The Dirac measure is the simplest example of measure which is not a function. REMARK B. we have the following properties. DEFINITION B. δx0 (ϕ) = ϕ(x0 ) = Rd ϕ(x)δ(x − x0 ) dx Rd δ(x − x0 ) dx = 1 The Dirac function can also be characterized as the derivative in the sense of distributions of a particular function.

5 In general.5 The Dirac comb The application S : D(R) → R ϕ→ k∈Z DEFINITION B. ϕ = lim ε→0 ε ϕ(x) − ϕ(−x) dx x +∞ ϕ(x) x −∞ The right hand side is referred to as the Cauchy principal value of and is denoted v. p +∞ −∞ dx +∞ ϕ(x) dx x is linear and continuous on D(R) (but not in D 0 (R)).6 Principal value of 1/x Let us consider the function. p 1 x . B. REMARK B. We thus conclude that Dm δ is not a measure for m ≥ 1. +∞ D log(|x|).Distribution spaces 253 Hence to define the distributions Dm δ. Then it defines a distribution (which is not a measure) which we denote by v. we need to consider test functions which belong at least to the class C m .3.9 ϕ(k) defines a distribution called Dirac comb. p −∞ ϕ(x) dx . x It can be checked that the mapping D(R) → R ϕ → v. a derivative in the sense of distributions of a measure on Ω defines a distribution which is not a measure except in the case where the measure is a sufficiently regular distribution. R→R x → log(|x|) By a direction computation. B. it can be stated that its derivative in the sense of distributions is given by the following identity.3.

REMARK B. 1 x and it is the derivative of the is not locally summable and there exists no associated .254 Moving Shape Analysis 1 x It is sometimes called the pseudo-function log(|x|) in the sense of distributions.6 regular distribution.

3 is defined by and its symmetriser by The translate of the function f of amplitude a ∈ Rd τa f (x) = f (x − a). DEFINITION C. ∀ ξ ∈ Rd (C. B0 (Rd )). Provided with the norm u ∞ = supx∈Rd |u(x)|.1 The Fourier transform belongs to L (L1 (Rd ).1 follows u(ξ) = (F u)(ξ) = (2π)−d/2 ˆ Rd e−i x·ξ u(x) dx . we define the Fourier co-transform ei x·ξ u(x) dx . the following properties hold 255 . v ∈ L1 (Rd ). We define the Fourier transform as def DEFINITION C. ςf (x) = f (−x). B0 (Rd ) is a Banach space.2 as follows For u ∈ L1 (Rd ). ξ REMARK C. ∀ ξ ∈ Rd (C. ∀ x ∈ Rd ∀ x ∈ Rd (C.1) THEOREM C.1 The space B0 (Rd ) stands for the space of bounded continuous functions on Rd which tend to zero at infinity.4) THEOREM C.2 For u.1 The case of L1 functions and extension to L2 Let u ∈ L1 (Rd ).2) def ¯ u(ξ) = (F u)(ξ) = (2π)−d/2 ˆ Rd DEFINITION C.3) (C.Appendix C The Fourier transform C.

ˆ u ii) Rd u(x) v(x) dx = ˆ Moving Shape Analysis F (ς u) = F u = u.3 i) If xα u ∈ L1 (Rd ) for |α| ≤ k. THEOREM C. then ξ β u ∈ L∞ and ˆ i|β| ξ β u = F (Dβ u) ˆ (C. ˆ Rd u(y) v (y) dy.3 The two last properties assert that the Fourier transform exchanges reciprocally differentiation of order |α| with multiplication by x α which suggests a kind of stability that will be established for a subspace of L1 .4 Let u ∈ L1 (Rd ) and real number λ > 0. Then the Fourier transform of the convolution product is given by the following identity F (u v) = (F u) (F v) (C. THEOREM C.9) .8) The next theorem is a fundamental result justifying the power of the Fourier transform.6 Fourier inversion ˆ Let u ∈ L1 (Rd ) such that f ∈ L1 (Rd ). We have ˆ F (f (λ. v ∈ L1 (Rd ). ˆ REMARK C. Then ξ ˆ f (x) = F (f )(x) (C.))(x) = λ−d f (ξ/λ) (C. ¯ iii) F (τa u)(ξ) = e−i a·ξ u(ξ). ˆ ς(F u) = F u.5 Let u.256 ¯ i) u = F (u) = F (¯).7) THEOREM C.6) REMARK C.2 The property (iii) expresses that the Fourier transform exchanges translation and multiplication by a complex exponential. then u ∈ C k (Rd ) and we have ˆ Dα u = (−i)|α| F (xα u) ˆ (C. THEOREM C. ˆ F (e−i a·x u) = τ−a u.5) ii) If u ∈ C k (Rd ) with Dβ u ∈ L1 (Rd ) for |β| ≤ k.

e 1 It xα D β ϕ |α|. ∀ α.7 S (Rd ) can be endowed with the structure of a Fr´chet e space with the family of semi-norms. ˆ REMARK C. ∀ α.2 The space S (Rd ) def We define the space of C ∞ functions with rapid decay at infinity1 .|β|≤k L∞ can be equivalently defined as follows n def S (Rd ) = u ∈ C ∞ (Rd ). REMARK C.8 The space D(Rd ) is contained with dense and continuous injection in the Fr´chet space S (Rd ). d This implies that f ∈ B0 (R ) which is not convenient for the applications.4 257 This result is a consequence of both the following formula F (e−a|. xα Dβ u(x) ∈ L2 (Rd ).The Fourier transform REMARK C. β ∈ Nd (C.5 The inversion formula gives f = F (g) with g = f ∈ L1 . β ∈ Nd o .11) REMARK C.6 This result is a consequence of the Fourier inversion Theorem and the identity (ii) in Theorem (C.10) and the regularizing property of convolution with Gaussian functions. Nk (ϕ) = ∀ k ∈ N. S (Rd ) = u ∈ C ∞ (Rd ). C.7 Fourier-Plancherel The Fourier transform F : L1 (Rd ) ∪ L2 (Rd ) → L2 (Rd ) can be continuously extended into a surjective isometry F : L2 (Rd ) → L2 (Rd ).2). A great improvement is the extension of the Fourier transform to square summable functions furnished by the next theorem. THEOREM C. |x|→+∞ lim xα Dβ u(x) = 0.| )(ξ) = (2 a)−d/2 e−|ξ| 2 2 /4a (C. THEOREM C.

all its derivatives belong to S (Rd ). we can identifiy elements of S (Rd ) with distributions in D (Rd ) . ϕ . (u.9 The Fourier transform F is an isometry from S into itself when S is endowed with the topology induced by L2 (Rd ). A sequence {Tk }k∈N of tempered distributions is convergent in S (Rd ) if k→∞ lim Tk . e REMARK C. ∀ ϕ ∈ S (Rd ) THEOREM C.258 Moving Shape Analysis THEOREM C.12) (C.8 Using the density of S in L2 . ϕ = T. Then the derivative of Tk has the following property k→∞ lim Dα Tk = Dα T in S (Ω). ∀ α ∈ Nd . let {Tk }k∈N ∈ S (Rd ) be a convergent sequence of tempered distributions towards T ∈ S (Ω). v)L2 = (ˆ.4).10 If T ∈ S (Rd ).10 The elements of Lp for 1 ≤ p ≤ ∞ can be identified with tempered distributions. Its inverse is given by F −1 = F and the following identities hold true.8)) and Theorem (B. v )L2 . This furnishes an alternative way of defining the Fourier transform in L 2 .9 Using the continuous and dense injection of D(R d ) → S (Rd ) (Theorem (C.4 The space S (Rd ) is defined as the space of continuous linear forms on the Fr´chet space S (Rd ). REMARK C. C. The distributions in S (Rd ) are called tempered distributions. Plancherel ˆ (C. the Fourier transform F ∈ ISOM(S ) can be extended in a unique manner as F ∈ ISOM(L 2 ).3 The space of tempered distributions DEFINITION C. Furthermore.13) REMARK C. Parseval u ˆ u L2 = u L2 .

T ) = τ−a F T .)β T ). ∀ T ∈ S (Rd ).11 Let f ∈ OM (Rd ).11 Hence the space OM leaves stable by multiplication the space S and S . F T .4 Fourier transform in S DEFINITION C. F (Dα T ) = (i · . whence the notation M for multiplicator. (ii) f T ∈ S (Rd ). i. mβ > 0. Let us consider an automorphism θ of Rd .. | Dβ ϕ| ≤ Cβ (1 + |x|)mβ THEOREM C. i.e.. T θ(x) · y = x · θ(y).)α F T . S (Rd ) if limk→∞ Tk = T.12 The Fourier transform is such that F ∈ ISOM(S (Rd )) with inverse F −1 = F and the following properties. C.6 Let T ∈ S (Rd ).The Fourier transform 259 DEFINITION C. ∀ x. S (Rd ).e. THEOREM C. y ∈ Rd . OM (Rd ) = def ϕ ∈ E (Rd ) : ∀β ∈ Nd ∃Cβ . F ϕ . then we have (i) f ϕ ∈ S (Rd ). REMARK C. iii) For β ∈ Nd .5 Let OM (Rd ) be the space of infinitely differentiable functions of slow growth in Rd . F (e−i a·. (ii) limk→∞ f Tk = f T. Dβ F (T ) = F ((−i · . i) F (τa T ) = e−i a·.14) ii) For α ∈ Nd . We denote by T θ its transpose. ϕ = T. as well as all their derivatives. def ∀ ϕ ∈ S (Rd ) (C. Its Fourier transform F T ∈ d S (R ) is the tempered distribution defined by F T. ∀ ϕ ∈ S (Rd ).

16) In the special case of distributions of compact support.17) REMARK C. then we have F (θ∗ T ) = T θ∗ F (T ) (C. Then its Fourier transform F T ∈ OM (Rd ) is given by the following identity. ϕ ◦ θ . e−i x·ξ (C.19) .18) and whose Fourier transform satisfies F (ϕ ψ) = (2π)d/2 F (ϕ) F (ψ) (C. we have the following result. C.14 Let T ∈ E (Rd ). since its support is reduced to a singleton F (δa ) = ei a·. THEOREM C.15 Let ϕ. ϕ = T. we can easily compute the Fourier transform of the Dirac measure. we define its direct image θ∗ T ∈ S (Rd ) by θ.260 Moving Shape Analysis For T ∈ S (Rd ).12 This theorem furnishes an explicit calculus of the Fourier transform of a distribution. def ∀ ϕ ∈ S (Rd ) (C. (F T )(ξ) = T (x). θ∗ T. We define their convolution product ϕ ψ ∈ S (Rd ) by (ϕ ψ)(y) = Rd ϕ(y − x) ψ(x) dx (C. when its support is compact. THEOREM C.13 Let T ∈ S (Rd ). As an application. ψ ∈ S (Rd ).15) and its inverse image T θ∗ T = | det D T θ|−1 (T θ−1 )∗ T .5 Fourier transform and convolution THEOREM C.

THEOREM C.21) F (α S) = (2π)d/2 F (α) F (S) C. we have F (T S) = (2π)d/2 F (T ) F (S) (C. b[ with values in .16 (i) F (OM (Rd )) = Oc (Rd ) and F (Oc (Rd )) = OM (Rd ). ξ) = R d e−ix·ξ ϕ(t. Its partial Fourier transform Fx with respect to x ∈ Rd is defined by (Fx ϕ)(t. REMARK C. is called the space of vectorial distributions over ]a. ∀ ϕ ∈ S (Rd ) O (Rd ) = T ∈ S (Rd ) : T REMARK C.8 The space of linear continuous mapping of D(]a. b[ an open set of R.14 By transposition we define Fx in S (Rd ) and we have Fx ∈ ISOM(S (R × Rd )). (ii) If S ∈ S (Rd ). DEFINITION C.7 Vector valued distributions Let X be a Banach space and ]a.22) −1 Furthermore Fx ∈ ISOM(S (R × Rd )) with inverse Fx = Fx . b[) into X. x) dx (C. C. T ∈ Oc (Rd ) and α ∈ OM (Rd ).7 261 We introduce the space of convolutors of S (Rd ).17 Let ϕ ∈ S (R × Rd ). b[. The measure dt stands for the Lebesgue measure on ]a. ϕ ∈ S (Rd ).6 Partial Fourier transform THEOREM C.20) (C.13 This space can be characterized as the space of distributions with rapid decay at infinity.The Fourier transform DEFINITION C.

Hence. in X (C. DEFINITION C. we can define derivatives in the sense of distributions. X) def (C. b[.15 Rd . REMARK C. X). then the application D(]a. X) = L (D(]a. ϕ = − f.262 X and we note Moving Shape Analysis D (]a. b[) → X is continuous. i ϕ . b[. b[). in X (C. b[.23) It is endowed with the topology of the uniform convergence on bounded sets in D(]a.25) . b[. and i ∈ N. REMARK C. with This definition can be extended to the higher order di di f. b[. b[) REMARK C. This definition can be extended to any open set Ω of ϕ → f.16 The application D (]a. we define its first order derivad tive in the sense of distributions as the unique element dt f ∈ D (]a. X). ϕ As in the case of scalar distributions. X) such that d d f.24) dt dt ∀ ϕ ∈ D(]a. ϕ . b[). b[). ϕ = − f. X) → D (]a. b[.17 derivatives.9 For f ∈ D (]a. i dt dt ∀ ϕ ∈ D(]a. if f ∈ D (]a. X) d f → f dt is continuous.

Y be two Banach spaces and u ∈ L1 (a.26) DEFINITION C. X). b.27) THEOREM C. X) is a Banach space if X is a Banach space.X) = dt−meas{t∈]a.18 [16] For 1 ≤ p ≤ ∞.X) def b 1/p = f (t) a p X dt < +∞ (C. b[. b[ → X t → f (t) such that (i) f is (strongly) dt-measurable on ]a.b.The Fourier transform 263 C. It is a Hilbert space if X is a Hilbert space. X) and A ∈ L (X. The measure dt stands for the Lebesgue measure on ]a. b. b[ → X t → f (t) such that (i) f is (strongly) dt-measurable on ]a. X) for 1 ≤ p < +∞ the space ]a. A u ∈ L1 (a. b[ and we set f L∞ (a.8 Vector valued Lebesgue spaces Let X be a Banach space and ]a. (ii) f is bounded almost everywhere over ]a. b.b. . (ii) f Lp (a. the following lemmas will be of great use.1 Let X.b[: f (t) inf X >C}=0 C (C. Y ). b[. b[ an open set of R. In the applications. the space Lp (a. Y ). the space of functions ]a. then 1. b. b[.11 We denote by L∞ (a. DEFINITION C. LEMMA C.10 of functions We denote by Lp (a. b.

28) 2. then b b φ. u(t) dt (C. b. |b| < ∞.264 2. LEMMA C. X) and p ≥ 1. .28) is valid for u ∈ Lp (a. b[ be an open set of R. the equality (C. For |a|. a u(t) dt = a φ. X) and φ ∈ X . Let ]a.2 1. b a Moving Shape Analysis A u(t) dt = A b a u(t) dt . b. u ∈ L1 (a.

H m (Ω) = def u ∈ D (Ω) : Dα u ∈ L2 (Ω).Appendix D Sobolev spaces D. . it is possible to define equivalently the space H m (Rd ) thanks to the Fourier transform. =  Dα u 2  L2 (Ω) def |α|≤m Ω Dα u(x)Dα v(x) dx (D.1 The space H m (Ω) endowed with the scalar product (u.2 If m1 > m2 the space H m1 (Ω) → H m2 (Ω) with continuous injection. v)H m (Ω) = and the associated norm u is a Hilbert space. We introduce the Sobolev space of integer regularity m ∈ N.2) H m (Ω) |α|≤m 1/2 (D. then (ii) The space H m (Rd ) coincides with following space u ∈ S (Rd ) : (1 + |ξ|2 )m/2 u ∈ L2 (Rd ) ˆ 265 (i) H m (Rd ) ⊂ S (Rd ). In the case Ω = Rd .1) THEOREM D.1 Spaces H m (Ω) Let Ω be an open set in Rd . |α| ≤ m (D.3 Let m ∈ N.3) THEOREM D. THEOREM D.

 m   1≤j≤m |ξj |2αj ≤ 1 + 1≤j≤m |ξj |2  ≤ C 1 + 0<|α|≤m 1≤j≤m |ξj |2αj  This theorem furnishes a new definition of the space H m (Rd ) that we shall extend to real parameter m in the next section.2 The proof of completeness uses the fact that F ∈ ISOM(S (Rd )) and the completeness of L2 (Rd ).4.1 The proof is based on the identity F (Dα u) = (iξ)α F u and the following inequality that holds ∀ 0 ≤ |α| ≤ m. H m (R ) d is equivalent to the following norm. 1/2 def R d (1 + |ξ|2 )s u(ξ) v (ξ) d ξ ˆ ˆ (D. u ∈ S (Rd ) : (1 + |ξ|2 )s/2 u(ξ) ∈ L2 (Rd ) ˆ ξ (D. The injection property is .4. Furthermore the continuous injection H s1 (Rd ) → H s2 (Rd ) holds for s1 ≥ s2 .6) THEOREM D. . the space (H s (Rd ).4) endowed with the following scalar product (u. 1/2 R (1 + |ξ|2 )m |ˆ(ξ)|2 d ξ u d REMARK D. REMARK D. D. ˆ (iii) The norm .)H s (Rd ) ) is a Hilbert space.4 For s ∈ R. (.. v)H s (Rd ) = and the associated norm.266 Moving Shape Analysis where u is the Fourier transform of u defined in section 1.5) u Hs( R d ) = R (1 + |ξ|2 )s |ˆ(ξ)|2 d xi u d (D.2 The space H s (Rd ) H s (Rd ) = def For s ∈ R we define the following Sobolev space.

. D(Rd ) is dense in H s (Rd ).4 Sobolev embedding theorems THEOREM D. D..7) The condition s > d/2 + k is optimal.6 ∀ s ∈ R.Sobolev spaces an easy consequence of the inequality (1 + ξ 2 )s2 ≤ (1 + ξ 2 )s1 . THEOREM D.7 Let s ∈ R and k ∈ N satisfying s > d/2 + k.e. (D.3 The proof is based on Theorem (C. ξ REMARK D.3 The topological dual of H s (Rd ) In this section we shall characterize the space (H s (Rd )) of continuous linear forms on H s (Rd ).1) and (1 + |ξ| 2 )s/2 ∈ L2 (Rd ) for s > d/2. We already have the following inclusions (H s (Rd )) ⊂ S (Rd ) ⊂ D (Rd ) since D(Rd ) is dense in S (Rd ) which is dense in H s (Rd ).5 The space H s (Rd ) is a normal space of distributions1 . the dual space of H s (Rd ) coincides (algebraically and topologically ) in D (Rd ) with H −s (Rd ). Then the continuous injection H s (Rd ) → B0 (Rd ) holds true.4 1 i. s1 ≥ s2 267 THEOREM D. D. REMARK D.

REMARK D. THEOREM D.1) we have F (u v) = F u F v and using w(ξ) = (1 + |ξ|2 )s ≤ 4s (1 + |ξ − η|2 )s + (1 + |η|2 )s .7 The Lebesgue indice can be deduced from homogeneity arguments.A .8) = λ−d/p v Lp |vλ |H s = λ(−d+2s)/2 |v|H s with |v|H s = def 2s 2 v Rd |ξ| |ˆ(ξ)| d ξ 1/2 . we get (1 + |ξ|2 )s |F (u v)(ξ)| ≤ C |(w |F u|) v| + C |F u| (w F v) ∈ L2 (Rd ) for s ≥ d/2. REMARK D. the Hilbert space H s (Rd ) is an algebra.8 For s > d/2. We set vλ (x) = v(λ x) and we have vλ Lp ∀2 ≤ p ≤ 2d/(d − 2s) (D.10 The continuous injection H s (Rd ) → Lp (Rd ). holds for s < d/2. THEOREM D.9 For s ∈ R.A + uH. for s ≥ 0. REMARK D.268 Moving Shape Analysis THEOREM D.8 u ∈ H s as The proof is based on a decomposition of the function u = uL. if u ∈ H s (Rd ) and ϕ ∈ S (Rd ) then ϕ u ∈ H s (Rd ). from which we deduce 1/p = 1/2−s/d.6 Again we use F (u ϕ) = F u F ϕ and the inequality (1 + |ξ|2 )s ≤ 2|s| (1 + |η|2 )s (1 + |ξ − η|2 )|s| . REMARK D.5 Using Theorem(A.

A) u) and the high frequency ˆ part uH. THEOREM D.10) .9) D.11 The following continuous injection H s (Rd ) → Lp (Rd ). we may assume that Ω is compact and use local maps (Definition(A. In this section.A = F −1 (χB(0.5) that D(Rd ) is dense in H s (Rd ) for s ∈ R.A) u). The space D(Ω) is dense in H k (Rd ).Sobolev spaces 269 with the low frequency part uL. 2 ≤ p < 2d/(d − 2) (D. we deal with the case where Ω R d . REMARK D. We use also the characterization ˆ p Lp ∞ 0 u =p λp−1 meas(|u| > λ) dλ and the Bienaym´-Tchebytchev inequality e meas(|u| > λ) ≤ λ−1 u 2 L2 THEOREM D.5 Density properties We already saw in Theorem (D. In the general case.12 Let k ∈ N∗ and Ω a domain of class C k .A = F −1 (χ B(0. holds for s = d/2.9 The proof is first performed for Rd by truncation and + regularization. We define the space m H0 (Ω) = D(Ω) H m (Ω) (D.1)) and partition of unity.

xd )|xd =0 d DEFINITION D.10 We use the partial Fourier transform (Fx u)(ξ . xd ) of u(x .6 Trace theorem for H s (Rd ) The linear application DEFINITION D. (i) The linear mapping γ0 is continuous from S (Rd ) endowed with the topology of H s (Rd ) into S (Rd−1 ) endowed with the topology of H s−1/2 (Rd−1 ).13 For s > 1/2. THEOREM D. xd = 0) stands for the trace operator of order 0.2 γj : D(Rd ) + stands for the trace operator of order j. xd ) in the direction x ∈ Rd−1 . γm : H m (Rd ) → + H m−j−1/2 (Rd−1 ) 0≤j≤m−1 u → (γj u)1≤j≤m−1 .14 The following trace mapping can be defined on D(Rd ) and extended by density + to a continuous linear surjective application.7 Trace theorem for H m (Rd ) + The linear application. D.270 Moving Shape Analysis D. → D(Rd−1 ) u → x → (γj u)(x ) = ∂xj u(x .1 γ0 : S (Rd ) → S (Rd−1 ) u → x → γ0 u(x ) = u(x . again denoted γ0 . REMARK D. Furthermore γ0 is surjective. (ii) This mapping can be extended to a continuous linear mapping from H s (Rd ) into H s−1/2 (Rd−1 ). THEOREM D.

(H s (Γ)) ≡ H −s (Γ) (D. Here we may consider a bounded domain Ω ⊂ Rd with compact boundary Γ of class C ∞ . hj )1≤j≤m nor on the associated partition of unity (rj )1≤j≤m . ζ ∈ B0 (D. but defines a class of equivalent norms. The space D(Γ) is dense in H s (Γ) for s ≥ 0. Using the space L2 (Γ) = H 0 (Γ) as the pivot space. Using the result of Lemma (A.6). + 271 D.11 The definition of the space H s (Γ) depends neither on the atlas (Uj . • Definition by localisation and diffeomorphism : We use local maps (Uj . 1≤j≤m (D. =  2  d−1 H s (R ) u H s (Γ) θj u 1≤j≤m 1/2 (D. we have the following identification.3 We define the Sobolev space of fractional order on the C ∞ manifold Γ by H s (Γ) = def u ∈ D (Γ) : θj u ∈ H s (Rd−1 ).11) DEFINITION D. 3.8 The space H s (Γ) There exists different methods to define Sobolev spaces on submanifolds of codimension 1 in Rd . 2.12) THEOREM D. 0).15 1. ζ ∈ B0 0. The space H s (Γ) provided with the norm.14) def REMARK D.Sobolev spaces m and ker γm = H0 (Rd ). hj )1≤j≤m and partition of unity {rj }1≤j≤m to define functional spaces on Γ.13) is a Hilbert space. we define the function (θj u)(ζ ) = (rj u) ◦ hj (ζ . The associated norm does depend on the choice of the atlas. .

Considered as a positive definite self-adjoint operator in L 2 (Ω) with domain D(−∆Γ ). LEMMA D.1 Let V ⊂ H m (Ω) be a dense subspace and an operator P such that . (ii) P u = u. Let m ≤ k. We may use a general procedure that may allow to define m-extension operators by density. a.9 Trace theorem in H m (Ω) THEOREM D.16) D. provided with the norm of the graph v H s (Γ) def s∈R 2 1/2 L2 (Γ) ) (D. then the mapping γm defined on D(Ω) with values in [D(Γ)]m can be extended by density to a continuous linear surjection : γm : H m (Ω) → H m−j−1/2 (Γ) 0≤j≤m−1 u → (γj u)1≤j≤m−1 m and ker γm = H0 (Ω).e. we define the space H s (Γ) = D((−∆Γ )s/2 ).272 Moving Shape Analysis • Definition as the domain of tangential operator : We consider the Laplace-Beltrami operator ∆Γ on the manifold Γ. H m (Rd )). ∀ u ∈ H m (Ω).15) =( v 2 L2 (Γ) + (−∆Γ )s/2 v (D. in Ω. D.16 Let Ω be a bounded domain in Rd with compact boundary Γ of class C k . if there exists an operator P such that (i) P ∈ L (H m (Ω).4 An open set Ω ⊂ Rd has the m-extension property for m ∈ N.10 Extension theorems DEFINITION D.

THEOREM D. we use the extension by reflexion (Babitch extension). ∀u ∈ V . we only use the extension by zero. H m (Rd )).13 For m = 0. Ω1 . xd ≥ 0 αk u(x . in Ω.18 Let Ω ⊂ Rd be a bounded open set with boundary Γ of class C k . THEOREM D. . Ω ∀ ϕ ∈ D(Ω) where [u] is the saltus of u across S. Then Ω possesses the m-extension property with 0 ≤ m ≤ k. + REMARK D. −k xd ). We conclude using Lemma (D. xd < 0 1≤k≤m where we choose (αk )1≤k≤m such that P u ∈ C m−1 (Ω) which impose continuity of derivatives with respect to xd up to order m − 1. the following identity holds vi . Ω2 ⊂ Rd and S = Ω1 ∩ Ω2 of class C m+1 such that Ω = Ω1 ∪ S ∪ Ω 2 Let u ∈ C m (Ω) such that u|Ωi ∈ D(Ωi ). LEMMA D.2). 2}. 273 Then P has a unique extension into an operator of m-extension from H m (Ω) into H m (Rd ).Sobolev spaces (i) P ∈ L (V.2 def Let Ω. This implies (−k)j αk = 1. then u ∈ H m+1 (Ω) REMARK D. (ii) P u = u. (P u)(x . a. xd ) = u(x .12 If we set vi |Ωj = ∂xi (u|Ωj ) for 1 ≤ i ≤ d and j ∈ {1.17 The open set Rd possesses the m-extension property for m ∈ N. 2} [u] ϕ · ni − u ∂xi ϕ. xd ).e. 0 ≤ j ≤ m − 1 which is a unique solvable Van der 1≤k≤m Monde system. ϕ = S i ∈ {1. For m ≥ 1.

we define the space of distributions     def Em = T ∈ D (Ω) : T = Dα fα .11 The space H −m (Ω) For m ∈ N. then H s (Ω) ≡ H m (Ω) D.19)   |α|≤m = inf fα ∈ L2 (Ω) T = |α|≤m Dα fα fα |α|≤m 2 .19 If Ω has the m-extension property with m ∈ N. H −m2 (Ω) → H −m1 (Ω).5 H s (Ω) = def Moving Shape Analysis For s ∈ R.18) THEOREM D.20 (i) The space H −m (Ω) endowed with the norm F H −m (Ω) def = sup u H m (Ω) u=0 is a Hilbert space.7 For m ∈ N. (ii) For m1 > m2 . u | u H m (Ω) endowed with the topology induced by T 2 DEFINITION D. we define the space (D. fα ∈ L2 (Ω) (D. ≤1 | F. THEOREM D. we define the space H −m (Ω) as the dual DEFINITION D.274 DEFINITION D. ∃U ∈ H s (Rd )such thatu = U |Ω def provided with the quotient norm u H s (Ω) = inf U U ∈ H s (Rd ) U |Ω = u H s (R ) d (D.6 m space of H0 (Ω).17) u ∈ D (Ω).

We define the band of Rd of width e in the direction ζ by def Be (ζ) = x ∈ Rd : a < x · ζ < b (D. D. Then the injection H m+1 (Ω) → H m (Ω) is compact.22 1 The operator −∆ + I ∈ ISOM(H0 (Ω).12 Compact embeddings THEOREM D.23 Rellich Let Ω ⊂ Rd be a bounded open set.21 The space Em coincides algebraically and topologically with H −m (Ω). ]a. THEOREM D. b[) = e.20) 1 THEOREM D.. b[⊂ R and 0 < e < ∞ such that meas(]a. THEOREM D. i. such that there .13 Poincar´ inequalities e DEFINITION D.25 Poincar´ inequality in H0 (Ω) e d Let Ω ⊂ R be an open set bounded in some direction.24 Let Ω ⊂ Rd be a bounded open set possessing the (m+1)-extension property.8 Let ζ ∈ Rd with |ζ| = 1.Sobolev spaces THEOREM D. 275 REMARK D.e. the injection m+1 m H0 (Ω) → H0 (Ω) is compact. D. H −1 (Ω)).14 This is a consequence of the Hahn-Banach theorem allowing to extend linear forms on a sub-space into the whole Banach space and the Riesz representation theorem in Hilbert spaces. Then for m ∈ N.

16 The Poincar´ constant P (Ω) = 1/λ2 where λ2 is the e first non-zero eigenvalue of the Neumann problem. 1 ∀ u ∈ H0 (Ω) (D.15 If Ω has the 1-extension property then the compact injection holds.23) . REMARK D. −∆u = λu.276 exists a band Be (ζ) with Moving Shape Analysis Ω ⊂ Be (ζ) Then u 2 L2 (Ω) ≤ e2 /2 u 2 L2 (Ω) . ∀ u ∈ H 1 (Ω) (D.21) THEOREM D. Then we have 2 u 2 L2 (Ω) − (meas Ω)−1 u(x) dx Ω ≤ P (Ω) u 2 L2 (Ω) (D.22) REMARK D. Then there exists C > 0 such that u 2 L2 (Ω) ≤C u 2 L2 (Ω) + γ0 u 2 L2 (Γ) .26 Poincar´ inequality in H 1 (Ω) e Let Ω ⊂ Rd be a bounded open set such that the following compact injection H 1 (Ω) → L2 (Ω) holds. We call Nikodym open sets the sets for which the Poincar´ e inequality holds.27 Let Ω ⊂ Rd be a bounded open set with regular boundary Γ. Ω ∂n u = 0. Γ THEOREM D.

Amann. 2004.sns. Southampton. 50:186–189. Birkh¨user Boston Inc.. R. Acquistapace. Univ.-C. Bensoussan. Paris S´r. [10] A..it/papers/luia/. [2] P. Ambrosio. Zol´sio. Flandoli. Pitman. [8] J-P. L. 1976). Nonhomogeneous Navier-Stokes equations in spaces of low regularity. CVGMT Preprint Server . 1:303–325. Pozzi.. J. Quaderni di matematica. On Some Control Problems in Fluid Mechanics. Southampton. Aguilar and J.Tools for shape evolution and morphogenesis. Coque fluide intrins`que sans approxie e mation g´om´trique. SIAM J. 1977. 171(1):63–87. Math.. and G.http://cvgmt. 1943.. Mutational and morphological analysis . Acad. 1989. On the existence of strong solutions to a coupled a fluid-structure evolution problem. Lions. 1971. Blankinship. F. Savar´. Boston. Armand. Acta Appl. [5] L. 2002. [9] H.References [1] F. A. [12] W. Sci. [7] J. G. [4] H. Math. Theoretical and Computational Fluid Dynamics. Ambrosio. [3] J. The curtain rod problem. AIAA J. 1998. C. and B. J. pages 106–115. MA. 6(1):21–52. Amer. J. Homogenization in deterministic and stochastic problems. Transport equation and Cauchy problem for BV vector fields. C. Differ. Papanicolaou. Abergel and R. IX(In print). [6] L. 29(1):89–118. 277 . London. Terreni. Bernardi. 9:1739–1745. Minimum-mass design of a plate-like structure for specified fundamental frequency. 326(11):1341– e e e 1346. A. Control Optimization. 17(1):1–40. Temam. 1990.-P. Variational equations e of Schroedinger-type in non-cylindrical domains. 2001. Equations. 1991. I Math. Variational problems in SBV and image segmentation. Fluid.-L. Math.. and G. a 1999. Beir˜o da Veiga. [11] M.-L. Aubin. Initial boundary value problems and optimal control for nonautonomous parabolic systems. Monthly. Mech. In Stochastic problems in dynamics (Sympos.

Robotique. Boundary optimization under pseudo cure vature constraint. Intrinsic geometric model for the vibration e of a constrained shell. Methods Appl. Bucur and J. Differential Equations. Zol´sio. Boisg´rault and J. Bucur and J. [16] N. Marcel Dekker.-P. 11(4):757–771. Zol´sio. Math. 1999). In Cagnol. Functional analysis. Zol´sio. RI. [23] D. Differential Equations. (4). Acad. volume 268 of Contemp. Optimisation de forme sous contrainte e capacitaire. e [19] D. Boundary variations in the Naviere e Stokes equations and Lagrangian functionals. Scuola Norm. Theory and applications.-P. Zol´sio. 2001. Un. C. e [14] S. Mat.-P. 321(3):387–390. Optimisation de forme : syst`mes nonlin´aires et e e e m´canique des fluides. Free boundary problems and density perimee ter. Boisg´rault. Cagnol and J. Sci.-P. Zol´sio. Zol´sio.-P. Shape derivative in the wave equation with e Dirichlet boundary conditions. C. Shape optimization and optimal design. J. 1994. Amer. 1995.-P. Sup. 216. [24] J.Informae tique Temps r´el. Zol´sio. Sci. 123(2):504–522. Cagnol and J. 23(11):985–1010. Math. Bucur and J. John et al. B (7).-P. [26] J. Shape analysis in membrane vibration. Bucur and J. Boll. Paris S´r. 1996. Differential Equations. J. 1995. Providence. Math. N -dimensional shape optimization under e capacitary constraint. Zol´sio. Cagnol and J. R. Existence of weak solutions for the motion of an elastic structure in an incompressible viscous fluid. 2004. Sci. J. Automatique. Sci. Ecole des Mines de Paris . Pseudo-courbure dans l’optimisation de e forme. C. [20] D. I Math. Integration I: Chapters 1–6. [25] J. 2000... 126(2):224–243. ”collection math´matiques appliqu´es pour la maitrise (in french)” edition. Zol´sio. Brezis. 2003.. [17] H. 318(9):795–800...-P. In Differential geometric methods in the control of partial differential equations (Boulder. Elements of mathematics. Bucur and J. e [21] D. Sci. Lect. Notes Pure Appl. [18] D. I Math. 1997.278 References [13] S.-P. 336(12):985–990. Zolesio. Masson. xv. Wiener’s criterion and shape continuity for the Dirichlet problem. Acad. Bucur and J.-P. . 2000. e Math. Math. [22] D. Ital. 2000. Pisa Cl. Soc. Bourbaki. R. R. Paris. Berlin: Springer. 23(4):681– 699 (1997). Ann. Acad. Proceedings of the IFIP conference. 1999. Boulakia. 158(2):175– 210. [15] M. Paris S´r. CO. PhD thesis. e e 1994. pages 23–39. 1996. 7-26..

II. Optimal Control of Complex Structures. 1974. I. Zol´sio. Ulbrich. Cannarsa. 1971. Gioan. Control Conf. Intrinsic exact model.). 2000. Numerical solution of optimal control problems governed by the compressible Navier-Stokes equations. Planchard. The damped wave equation e in a moving domain.-P. New York. R. J. Equations. Rend. Chen and J. 1990. Trltzsch (eds. In Identification of parameters in distributed systems (Joint Automat. Unie e e e versit´ Paris VI. Ulbrich.. R. C. 269:A1135–A1138. Cagnol and J. Aerospace Sci. [37] S.-H. 83:73–77. Math. Vanninathan. M. and J.G Ciarlet. Heinkenschloss. Tucsnak. 1994. Zol´sio. Michel. Sprekels. Collection de la Direction e des Etudes et de la Recherche .. Da Prato.-P. J. Partial Differ.. H.. Texas. Commun. North-Holland Studies in Mathematics and its Applications.References 279 [27] J. San Mart´ and M. C. Identification of functional parameters in partial differential equations. the equations modelling the motion of a rigid body in a viscous fluid. and J. . [29] P. Calcolo. 25(5-6):1019–1042. Existence of solutions for ın. 1974). [35] C. Body under lifting wing. A. Ghayour. Cannarsa. Sur une m´thode de r´solution du probl`me inverse dans les e e e ´quations aux d´riv´es partielles paraboliques. e [33] G. Fis.-P. Birkh¨user Verlag. [32] G. [39] C. Thomas. F. R. Analyse fonctionnelle et identification de coefficients r´partis dans les ´quations aux d´riv´es partielles. B.EDF. [30] P. Quelques r´sultats sur l’identification e e de domaines. G. Lincei.. Evolution equations in e non-cylindrical domains. Vibration of a pre-constrained elastic e thin shell. G. K. Engrs. Sci. 139:43–55. 10:207–232. 334(2):161–166. F. Math. Ser. Clarke. PhD thesis. Hoffmann and I. Atti Accad..A. Cl. Mech. 1989. G. and S.II . Sci. Sci. Modeling and regularity of the solutions. C´a. Zol´sio. K. Chavent. Naz. J. Acad. 2001. M. 85(1):1–16. Conca. Soc. Paris. Nat. Mat. Sci. [38] C. pages 31–48. Cagnol and J. Acad. Conca. [31] J. 334(3):251–256. 1961. A-B. [28] J. Lasiecka. C. Amer.-P. VIII. and M. Austin. Vibration of a pre-constrained elastic thin e shell. International Series of Numerical a Mathematics. Da Prato.S Collis. Tex. Equations.. and J. Univ. Differ. Chavent. 28:547–562. Paris. 2002. 1969. Mathematical Elasticity : Volume I. 1973. 2002. e [34] G. Acad. Zol´sio. Paris e e e S´r. Leugering. Chavent. Probl`mes e math´matiques en couplage Fluide-Structure. J. 1997. J. [36] P.

Delfour and J.-P. Delfour and J.... New York. [52] M. A boundary differential equation for e thin shells.-P. [43] G. 151(1):234–269. Advances in Design and Control . [44] R. Control Lett. 11(2):151–158. Cuer and J.-P.-P. Da Prato and J. Control of singular problem via differentiae tion of a min-max. volume 218 of Lecture Notes in Pure and Appl. [53] F.. 1988.Analysis. [41] D. Shape analysis via distance functions: e local theory. 128(1):125– 167. Zol´sio.-P. Math. Delfour and J. [47] M. Delfour and J.SIAM. Directional derivative of a minimax function. Correa and A. Differential Equations. Lecture Notes. AB. Anal. Further developments in the application e of min-max differentiability to shape sensitivity analysis. 1996. Tangential differential equations for e dynamical thin/shallow shells. Control of partial differential equations..C.C. Zol´sio. Shapes and Geometries . e Differential Calculus and Optimization.-P. Funct. and transitions (Banff. Zol´sio. Springer . Zol´sio. . Soc. Shkoller. volume 13 of CRM Proc.-P. TX. 1988. Funct.-P. . [46] M.-P. J. 26(4):834–862. Zol´sio. Anal.. 9:13–22. 1994. RI. Shape analysis via oriented distance e functions. 11(1):73–77.C.Volume 5. Delfour and J. Math. In Control of nonlinear distributed parameter systems (College Station. J. 1999). Dautray and Lions J-L. Nonlinear Anal. J. 1998. pages 45–68. Dekker. [51] M. Delfour and J. 114:108– 119. Sci. Amer. Delfour and J. An optimal control problem for a e parabolic equation in non-cylindrical domains. Shape sensitivity analysis via min max e differentiability.-P.R... Desaint and J. 1988. Zol´sio. 2004. 123(1):129–201.280 References [40] R. Providence. Seeger.-P. 1985. Differential Equations. Manifold derivative in the Laplacee Beltrami equation.C. [50] M. Zol´sio. for Rat. Zol´sio.. 1995. J. Mech. 1997. [42] M. Mathematical analysis and numerical methods for science and technology. [45] M. Coutand and S.C. [49] M. 2001.C. [48] M. 2001. Delfour and J. 119(2):426–449.C. Control Optimization. Zol´sio. On the motion of an elastic solid inside of an incompressible viscous fluid. Velocity method and Courant metric e topologies in shape analysis of partial differential equations. pages 91–123. 1990. Systems Control Lett. 1995). In Boundaries. Theory Methods Appl. Arch.C. 1989. interfaces. Zol´sio. Notes Control Inf. Anal. Syst. Zol´sio. Lect. SIAM J.

. e Automatique.. Eng. Ecole des Mines de Paris . Prog. Moutazaim. Sci. [60] R. Anal. Sensitivity analysis for an ina compressible aeroelastic system.A. R. Nonlinear Differ. Inverse Probl. 9(6-7):681–708. PhD thee e sis. 1999...A.-P. Methods Appl. Dynamical shape gradient for e the Navier-Stokes system. [55] J. Mech. 15(6):1507–1522. 146. Comput. Fern´ndez and P. M.. 1999. Math.. [65] M. Robotique. Convex Anal. M. 1999). New York. 2003. Mech. Moubachir. 1982. Le Tallec. 192(43):4837–4873. Europ´enne El´m. Halleux. 216:87–107. 2002. Moubachir. Existence of solutions for a Model of Fluid-Rigid Structure Interaction. Fern´ndez. Dziri and J. Linear stability analysis in fluida structure interaction with transpiration. 2001.-P. . [56] R. A one-phase inverse Stefan problem. Dekker. Donea. Zol´sio. for Rat. 192(43):4805–4835. 12(8):1109–1130. El Badia and F. Shape optimization and optimal design. 1995. e e [64] M. I. 1999. [62] J. [63] T. [61] A. [57] R. Formulation and mathematical analysis. pages 87–107. Paris Ser. Methods Appl. John et al.. Comput. Simonett.-P. Dziri and J.J. [66] M. Linear stability analysis in fluida structure interaction with transpiration. Giuliani. [58] R. and J. Math. J. R´v. Eulerian derivative for non-cylindrical funce tionals. Equ. and P. C. volume 216 of Lecture Notes in Pure and Appl. Notes Pure Appl.. Deriving adequate fora mulations for fluid-structure interactions problems: from ALE to tran´e spiration.A. Acad. In Shape optimization and optimal design (Cambridge. Fern´ndez and P. Mech. Le Tallec. Zol´sio. Le Tallec. Lect. 2004. Finis. II. Zol´sio. Dziri.. Methods Appl. Appl. 1999. Sci. Dziri and J. 35:183–212. 2001. [59] R. Escher and G. Engrg. Comput. Fern´ndez and M. An arbitrary LagrangianEulerian finite element method for transient dynamic fluid-structure interactions.P. 33:689–723. Cagnol. Probl`mes de fronti`re libre en fluides visqueux. S. Mech. 6(2):293–318. Numerical analysis and applications. 338:183–186. Zol´sio.Informatique Temps r´el. I. Eulerian derivative for non-cylindrical funce tionals.-P.. Models Methods Appl. Math. Esteban. and J.References 281 [54] B. Fanion. Desjardins and M. Arch. 2003. Engrg.A. 2000. Moving surfaces and abstract parabolic evolution equations. Dynamical shape control in non-cylindrical e Navier-Stokes equations. Dziri.

Garabedian and M. Miloh. Acad.-P. Glowinski and O. Monographs in Mathematics. Modelling. Schiffer. Paris. Fursikov. Friedman. M. Ser. 2000. Fursikov.282 References [67] M. Math.D. Towards the computation of minimum drag profiles in viscous laminar flow. J. An exact block-Newton algorithm a for solving fluid-structure interaction problems. Fluid-structure interaction: A theoretical ´e point of view. Zol´sio.S. 1999. Numer. INRIA. Trace theorems for threedimensional.. 2002. Optimal control of Navier-Stokes equations using Lagrange-Galerkin methods. Math. Am. Math. 1953. 35:561–587. Finis. Trans. [76] A. Convexity of domain functionals. [78] R. Nonlinear Analysis. Identities in the theory of conformal mapping. 1(2):58–66. and L. 1984.R Galper and T. Control Optimization. Soc. 1999.. Minimal surfaces and functions of bounded variation. 65:187–238. Notices Am. Shape sensitivity and large deformation of e the domain for Norton-Hoff flows. Math. Giusti. Grandmont and Y. a [80] C. and L. C. Vol. Garabedian and M. Pironneau. Gioan. Orenga.. 354(3):1079–1116. Gomez and J. Fourestey and M.. Fern´ndez and M. Schiffer. [79] N. Free boundary problems in science and technology. Boston-Basel-Stuttgart: Birkh¨user.V. e [77] E. Soc. 1976/77. R. Europ´enne El´m. Gunzburger. 80. Maday. Une m´thode d’approximations successives pour e l’identification de domaines. Math. M. 1949. a XII. [70] A. Technical report. 36(3):852–894. R. Trans. Math. Moubachir. 10(1):119–130. 9(6-7):633–653. pages 167–176. Amer. Moubachir. In Optimal control of partial differential equations (Chemnitz. 2001.. [71] A. 47(8):854–861. Hou.A. [68] F. [72] A. time-dependent solenoidal vector fields and their applications. Rev. Soc. RR4609. SIAM J. [75] P. 1998). Appl. PhD thesis. Basel. Analysis of a nonlinear fluid-structure interaction problem in velocity-displacement formulation. Universit´ de Nice. 1998. Boundary value problems and optimal boundary control for the Navier-Stokes system: The two-dimensional case. Flori and P. e . Gunzburger. 2002. Motion stability of deformable body in an ideal fluid with applications to the N spheres problem. 2:281–368. [74] P. 2003. R. 1974. volume 133 of Internat. [73] A. [69] G. Physics of Fluid. 336(8):681–686. Hou. 1992. Analyse Math. Birkh¨user. Sci.

M. [83] E. Fachbereich Mathematik. [84] E. 1967. and G. [86] M. Internat. I. J.L. Computer Methods in Applied Mechanics and Engineering. C. I-II. Methods Engrg. D. Berlin. Springer-Verlag. 1992. Zol´sio. Ala. 1972. Durham. J. C.. Math. 1981.. 43(4):752–769.Habilitation thesis.. Arch. Haug.). Starovoitov. On a motion of a solid body in a viscous fluid. Streeter. O. 1971). Global existence of weak solutions for viscous incompressible flows around a moving rigid body in three dimensions. Army Missile Command. Paris VI.I. Hinze. Sci. Arch. U. Hofmann and J. Army Research Office (Durham. Vol. 51:295–303. Joseph.E. W. Lagrangian-Eulerian finite element formulation for incompressible viscous flows. Seregin. Pironneau. Office-Durham. PhD thesis. and T. S.C.M. Math. Jaffre. [90] J. [87] K. N. Gunzburger. Optimal and instantaneous control of the instationary NavierStokes equations . editors. PhD thesis. 2(3):219–266. Technische Universitt Berlin. Redstone Arsenal. Pan. Domain perturbations: the higher order theory of infinitesimal water waves. Tartar. Two-dimensional case. L. Numer. volume 50 of NATO Advanced Study Institute Series E: Applied Sciences. The Hague. U. 1999. 1999. Appl. J. 1981.. The Lp regularity problem for the heat equation in non-cylindrical domains. S. Boundary velocity control of incompressible flow with application to viscous drag reduction. and J. 2000. and T. [93] B. C´a. pages 859–879. Ill. [92] D. Analyse num´rique de deux probl`mes de domaine optimal et e e de contrˆle ponctuel. Lectures given at the Joint C. S.K Liu. 1972. Kawohl. [88] S. Rational Mech. Joseph. 1973.J. [85] E.D Gunzburger. 2000. [89] T. o [91] D.. Optimization of distributed parameter e structures. 1974. Distributed parameter optimal design.I. 2000. In Transactions of the Seventeenth Conference of Army Mathematicians (U.P Svobodny. C.-H. Lewis. Summer . Report No. N. K. Hoffmann and V. [82] M. SIAM Journal of Control and Optimization. 9(2):633– 648. J. Math.K Zimmermann. 72–1. Lee. L. 5(2):171–184. Haug..J. 24:325– 351. H. Anal. Adv./C. Hughes. A computational method for optimal structural design. Fluid. Martinus Nijhoff Publishers.-P. Piecewise uniform structures.D.D. Optimal shape e design. 30(1):167–181. 29:329–349. Army Res. Haug and J. Hou.References 283 [81] M. volume 1740 of Lecture Notes in Mathematics.A.. Mech. N. and T.D. J. Anal. Parameter and domain dependence of eigenvalues of elliptic partial differential equations. Rational Mech.

[94] P. III. and T. M. Michel. Equ. Le Tallec and J. Fourier. Band 182. Comput.-P.. Murat. D.Doctorat `s Sciences Math´matiques.. Ann.E. Methods Appl. I. Die Grundlehren der mathematischen Wissenschaften.[C. June 1–6. M. NorthHolland. Ann. Time-dependent parabolic problems on non-cylindrical domains with inhomogeneous boundary conditions. Masmoudi. 1971.M. 4(11):1263–1297. [99] J-L. 1974.284 References School held in Tr´ia. II. Micheletti. Foundation]. Sup. [98] J-L. R. Variation d’un point de retournement par rapport au domaine. Cellina and A. Comm. 1(3):291–309. 1979. 1998. [100] G. Perturbazione dello spettro dell’operatore di Laplace. Streeter. 1987. Une m´thode de gradient pour l’identification de domaines. Lions and E. Optimal control of systems governed by partial differential equations. [105] A. Edited by A. Universit´ de Nice . Vol. Paris: Dunod. 26:683– 694. 1970.E. and J. pages 465–622. [95] P. e e e [102] B... J. 1994. Lumer and R. . Metrica per famiglie di domini limitati e propriet` a generiche degli autovalori. Quelques m´thodes de r´solution des probl`mes aux limites e e e non lin´aires. Numerical methods for nonlinear three-dimensional elasticity. F. Fondazione C. [101] M. 1972. Le Tallec. E. Amsterdam. Magenes. 8:1012– 1019.I. Sup. Non-homogeneous boundary value problems and applications.M. Ann. in relazione ad una variazione del campo. [97] J-L. Vol. [96] J-L. [103] J.I. Partial Differential Equations. [106] F. Fluid structure interaction with large structural displacements. Mignot. In Handbook of numerical analysis. Mech. 2001. Haug. Outils pour la conception optimale de formes. Lions. Lions. J. Paris: Gauthiere e Villars. Lions. e PhD thesis. 26:151–169. 2001. 7:143–182. 1972. Pisa (3). Inst. Mouro. o Ornelas. Scuola Norm. e [104] A. Springer-Verlag. Optimal design of structures with constraints on natural frequency. Engrg. 1957. Universit´ de Nice. 190(2425):3039–3067. Scuola Norm. Pisa (3). Vol. Etudes math´matiques. Schnaubelt. McCart. 1972. Micheletti. AIAA J. Sur les probl`mes mixtes pour certains syst`mes e e paraboliques dans des ouverts non cylindriques. Springer-Verlag. 1969. Puel. Evol. PhD thesis.

Nice 1975. C. Control of fluid-structure interaction phenomena. Universit´ Paris VI . e 1972. Notes Comput. e [117] J. Paris S´r. Universit´ Paris VI. Sci. 112:49–68. Sur le contrˆle par un domaine g´om´trique. 1988. 273:A708–A711. Murat. Lect. Mat. Zol´sio. Ann. (4). Computer Methods in Applied Mechanics and Engineering. Paris S´r. Ecole Nationale des Ponts et Chauss´es. numerical methods and computer systems modelling (Internat. Springer. Limit analysis of plastic structures subject to boundary vario ations. An arbitrary Lagrangian-Eulerian finite element method for interaction of fluid and rigid body. Sci. Etude de probl`mes d’optimal design. Murat and J. Zol´sio. e Tech. Morice. 1967. Truchi. [118] T. pages 454–467. 71:315–340. A-B.. Une m´thode d’optimisation de forme de domaine. Murat. Hughes. 274:A395–A398. 2004. 1976. R. application to the aeroelastic stability. PhD thesis. 95:115–138.. In Control ` e ` theory. [108] M. 1976. IRIA LABORIA. 15:63–76.-P. Pura Appl. Part 2. Murat. A two-dimensional FEM adaptive moving-node method for steady Euler Flow Simulations. [116] F. Les m´thodes directes en th´orie des ´quations elliptiques. 1992. e [112] F. Applicae tion a l’´coulement stationnaire a travers une digue poreuse. Dynamic modeling. and Math. Murat. 7th IFIP Conf. and J. Passeron. Berlin. C. Palmerio. [113] F.References 285 [107] Ph. Proc. [111] F. 1977. Simon. Sur le contrˆle d’un syst`me par les coefficients de l’op´rateur o e e ou par un domaine g´om´trique. [110] Z. Moubachir and J. Paris. Computer Methods in Applied Mechanics and Engineering. A-B. [119] B. Internal Report INRIA. Stos. Nomura and T. 1963. Contre-exemples pour divers probl`mes o` le contrˆle ine u o tervient dans les coefficients. 1975. e e e [114] F. Lecture Notes in Econom. Murat and J. In Optim. Neˇas. Arch. Acad. Un contre-exemple pour le probl`me du contrˆle dans les e o coefficients. 2002. Sympos. Acad. Cost function gradient for a coupled e fluid-solid system. [120] L. Systems. Sci. Simon. control e theory and stabilization for flexible structures: Industrial applications . Editeurs. c e e e ´ Masson et Cie. Mr´z. e [109] M. C. Rocquencourt. 54-62. Th´or`mes de non-existence pour des probl`mes de contrˆle e e e o dans les coefficients. Moubachir..-P. 1976.J. o e e Technical report. 41. 107. ´ [115] F. 1971. Mech. PhD thesis. 1974). Vol.76 015.. R.

2000). Math. Fluid Mech. Inverse Problems. Weak solutions of parabolic equations in non-cylindrical domains. Differentiation with respect to the domain in boundary value problems.M. Sur les conditions de stabilit´ d’une discontinuit´ dans un e e milieu continu.. [128] G. El´m. XII. 2002. Existence). H. 1988. Ramananjaona. Eur. 4:99–110. Finis. Savar´. Cannes. 1987. Pironneau. On optimum design in fluid mechanics. [126] A.A. Arch. Pol. [129] C... [130] R. 64:97–110. (Free falling body in a viscous incompressible fluid.. Poncin. 8:125–137. 125(6):1785–1792. 71:1–62. 2001. Z. Optimal shape design for elliptic systems. Russell.M. Anwend. [122] S. Math. Chute libre d’un solide dans un fluide visqueux incompressible. ComCon Workshop.. Simon. Propri´t´s d’une int´grale de l’´quation parabolique dans ee e e un domaine non cylindrique. and J. Special issue to celebrate Pierre Sabatier’s 65th birthday (Montpellier.. Am. Japan J. Piskorek. Funct. Finis. Farhat. Appl. [121] S. Prokert. 1973. Anal. On optimum profiles in Stokes flow. Piperno and C. El´m. Pures Appl. Lieberman. J. and G. and M. Rev. 1960. Pironneau. On evolution equations for moving domains. 1939. J. Lambert. Fluid Mech. In Stabilization of flexible structures.286 References at Aerospatiale. Proc. Starovoitov. 1987. 161(2):113–147. Rev.. Anal. Math. Numerical simulations of wind effects on ´e flexible civil engineering structures. . [132] G. M. 9(6-7):655–680. [133] D. J. Piperno and P. 1999. Acta Math. Bournet.-P. Numer. Serre. [127] H. [131] J.-E. 17(4):1087–1111. Zol´sio. D. 1997. Mech. Lesselier. 183-216. [134] J. Existence. Design of efficient partitioned procedures for ´e the transient solution of aeroelastic problems. tions for the two-dimensional motion of several rigid bodies in an incompressible viscous fluid. Montpellier/Fr. Ann. Math. Eur. Shape e reconstruction of buried obstacles by controlled evolution of a level set: from a min-max formulation to numerical experimentation. [123] O. 1997. Optim. Wei. [125] O. Springer Series in Computational Physics. 59:117–128. 2(7-8):649–687. Soc. [124] O. Global weak soluın.. Pironneau. Anal. 18(1):67–95. Tucsnak. Proc. 2001.. Parabolic problems with mixed variable lateral conditions: e an abstract approach. New York: Springer-Verlag. 1984. 1974. San Mart´ V. 76:321–351.. 1999. 1980. Ration.

. Sci. 315(5):545–549. Wave equation in time periodical domain. Numer. Introduction to shape optimization: e shape sensitivity analysis. Un r´sultat d’existence de vitesse convergente dans des e e probl`mes d’identification de domaine.References 287 [135] J. PhD thesis. e e [148] J. Proc. Acad. e e e e [141] C.Doctorat d’Etat en Math´matiques. Studies in Mathematics and its Applications edition. Zol´sio. e In Stabilization of flexible structures. 1979. Conf. Zol´sio. North-Holland.Sp´cialit´ Math´matiques. e .-P. Sokolowski and J. Tucsnak..Z. Acad. Takahashi and M.-P. Math. J. 1979. Berlin. 282-294. Complementary variational principles in elasticity and plasticity. e e Paris S´r.-P. Jablonna. and Math. 288(19):A911–A913. Navier-Stokes Equations. A855–A858. volume 16. An optimal design procedure for optimal control support. 1993. e e Universit´ de Nice . Truchi. 72(6):537–551. 2004. 1992. 1987. Optimization Theory Appl. Springer Series in Computational Mathematics. [139] R. 6(1):53–77. Acad. e e 283(11):Aiii. Identification de domaines par d´formations. C. Paris S´r. 65(2):331–362. Math. pages 7–93. 42(6):1121–1144. R. Global strong solutions for the twodimensional motion of an infinite cylinder in a viscous fluid. Zol´sio. ComCon Workshop.. PhD thesis. 1984. Springer. J. Sci. Wroclaw. [140] C. 144. Int. Zol´sio. Yang and N.. 1990. Stabilization and control of distributed systems with time-dependent spatial domains.-P. K. In Duality and complementarity in mechanics of solids (School-Conf.-P. R. Washizu. [138] T. Temam. Sver´k. Solution variationnelle d’un probl`me de valeur propre e e non lin´aire et fronti`re libre en physique des plasmas. Duality Problems of Mech. Sver´k. A-B. Sci. ˇ a [137] V. J. 1979. On optimal shape design. A-B. 1998. Truchi and J. An adjoint method for the inverse design of solidification processes with natural convection. [147] J. Zol´sio. ˇ a [136] V. 1977. [146] J. 1976)..-P. Zol´sio. I e Math. Vol. Ossolineum. 1987. C Wang.. Methods Eng. R. Stabilisation par variation du domaine. 1977). e In Convex analysis and its applications (Proc. Paris S´r. Fluid Mech. Montpellier/Fr. 1976. [142] P. Pures Appl. 1992. C. C. Universit´ de Nice . J. [144] G. Deformable Bodies. Lecture Notes in Econom. Zabaras. [145] J. pages 207–219. Muret-le-Quaire. Systems. On optimal shape design. 1988. [143] K. (9).

-P. volume 50 of NATO Adv. II (Iowa City.-P. Zol´sio.-P. 2000. volume 50 of NATO Adv. pages 372–398. Sci. II (Iowa City. 75:446–460. [152] J. E: Appl. Zol´sio. Springer. Iowa. e Scuola Norm. Shape differential equation with a non-smooth field. 1998. Sci. Vol.-P. 1994.288 References [149] J. a [156] J. Boston. Galerkine approximation for wave equation in moving e domain. e Notes Pure Appl. Domain variational formulation for free boundary probe lems. Truchi. Sci. In e Computational methods for optimal design and control (Arlington. . [153] J. Lect. Zol´sio. Sci. 1990. Zol´sio and C... pages 1152–1194. Zol´sio. Basel. [154] J. Berlin. Lect. pages 1089–1151. Vol. Nijhoff. Ann. Math. 2001. Weak shape formulation of free boundary problems. Iowa. Nijhoff. Lect. Study Inst. In Optimization of distributed parameter structures. 1980). Ser. Zol´sio. 1981. 1997). 1999. Shape analysis and weak flow. Zol´sio. 1985. Zol´sio. In e Boundary control and boundary variations (Nice. Notes Control Inf. Zol´sio. 147:191–225. [157] J. [150] J. Birkh¨user Boston.-P. 216:415–439. Weak set evolution and variational applications. Numer. Shape stabilization of wave equation. Sci. 1988. Variational formulation for incompressible Euler equation e by weak shape evolution. Systems Control Theory.-P. a [155] J. volume 133 of Internat.. Notes Cone trol Inf.-P. In Optimization of distributed parameter structures. volume 24 of Progr. 1981. VA. Sci..-P. [158] J. 21(1):11–44. pages 427–460. In Optimal control of partial differential equations (Chemnitz. MA.. Birkh¨user. The Hague. Sup. [151] J. Lect.-P. Notes Math. 1986). Ser.. Pisa Cl.. Shape stabilization of flexible structure. pages 309–323. volume 100 of Lecture Notes in Comput. The Hague.. The material derivative (or speed) method for shape ope timization.-P. Zol´sio. Study Inst. e 1740:157–341. 1998). Math. E: Appl. 1980). Ser. (4).

2. 9 Dirichlet. 218 noncylindrical. 33.Index Aeroelastic stability. 9 piezoelectrical device. 139. 1. 14 incompressible. 35. 9 feedback law. 144. 4. 51 Fluid drag. 222 tracking. 178 viability. 188. 5 Boundary control. 4 Domain deformation. 110. 14. 23. 3 perturbation. 35 noncylindrical wave. 8 free. 4. 3 strong convection. 102 Functional lagrangian. 218 kinetic continuity. 4. 3. 195. 3 tangential calculus. 50 weak convection. 72. 11 Gˆteaux differentiable. 2 mean curvature. 50 Control active. 60 weak convection. 2. 1 Condition transpiration. 19. 217 ODE. 4. 12. 120 curve. 18 Geometry additive curvature. 55 perturbation. 110 Function characteristic. 198. 77 density perimeter. 2 moving. 9. 11. 8. 199 Euler-Lagrange. 219 adjoint ODE. 5. 187 partial differential. 187. 216 289 . 17. 13. 18. 14. 10 cost gradient. 38. 9 optimal. 90 linearized Navier-Stokes. 216 noncylindrical heat. 101 Laplace. 24 a minimization. 11 Navier-Stokes. 19 value problem. 38 tube. 187. 119 moving. 61 intrinsic. 98 nonlinear elastodynamic. 10 oriented distance. 119 Interface motion. 119 without steps. 4 perimeter. 220 adjoint Navier-Stokes. 50 cost. 159. 180 passive. 178. 103 Hamilton-Jacobi. 33 kinematic continuity. 49 Equation adjoint elastodynamic.

2. 146 tangential Laplace-Beltrami. 199 Optimization transverse equation. 154. 2 Regularity lipschitz. 36 tube derivative. 16 differentiable equation. 57 adjoint transverse equation. 23. 24 moving analysis. 13. 119 transverse. 1 derivative. 10. 36 Sobolev. 34 projection. 16 eulerian material derivative. 215 fluid-structure. 17 derivative structure theorem. 47. Lebesgue dominated convergence. 121 topological derivative. 39 Model fluid-solid. 70 noncylindrical analysis. 222 Lie bracket. 41 first-order optimality. 68 transerve field. 97 eulerian derivative. 16 duality. 69. 44. 98 Index asymptotic limit. 44. 16 differentiable. 188 58 . 15. 2 special bounded variation. 136 kernel identity. 190 velocity field. 46 Tikhonov. 84 differential equation. 10. 197 transverse field. 2 tangential divergence. 158. 168 adjoint trace. 3–5 noncylindrical gradient. 121 space embedding. 68. 83 min-max. 3 observation. 25. weak Piola derivative. 36 optimization. 140. 39 Method Galerkin. 5. 165 level set. 114. 185 St Venant-Kirchoff. 3 lagrangian. 34 lagrangian multiplier. 17. 17. 15 material derivative. 10. 175 ill-posed. 4. 24 extractor identity. 58 47 adjoint transverse field. 22. 23. 101 speed. Helly. 13. 39. 120 space parametrization. 36. 133 eulerian shape derivative. 46. 14. 217 Stefan. 144. 5. 5. 88 Shape Banach-Steinhaus. 25. 18. 105 trace.290 Mapping diffeomorphism. 207 lagrangian material derivative. 66 inverse problem. 1. 105 wave. 24. 157 tangential gradient. 18. 40. 33 transverse magnetic wave. 124 221 Space regularization. 51 identity perturbation. 89 calculus. 2. 24. 195. 168 Operator lagrangian partial derivative. 194. 13 Theorem Ascoli. 73. 85 control. 73 eulerian partial derivative. 165 continuity. 3. 55 Hilbert.

5 Hausdorff. 3 BV. 62. 147 weak implicit function. 51 change. 128 Topology.Index Leray-Schauder. 22. 88 Min-Max. 18. 89 291 .

Sign up to vote on this title
UsefulNot useful