laa) Byiisasael Equations Lawrence C. Evans Graduate Studies in Mathematics Volume 19 ne, b {' 0-1) American Mathematical Society awCONTENTS Preface 1, Introduction 1.1. Partial differential equations 1.2. Examples 1.2.1, Single partial differential equations 1.2.2, Systems of partial differential equations 1.3, Strategies for studying PDE 1.3.1. Well-posed problems, classieal solutions 1.3.2. Weak solutions and regularity 13 LA. Overview . Typical difficulties 1.5. Problems PART I: REPRESENTATION FORMULAS FOR SOLUTIONS 2. Four Important Linear PDE 2.1. Transport equation 2.1.1, Initial-value problem, 2.1.2, Nonhomogeneous problem 2.2. Laplace's equation 16 7 7 1s 193. Fursdanentl sutution Mean-value formulas Properties of harmonie functions Green's tunction Energy methods . Heat equation He 2.3.1, Fundamental solutio 2.3.3. Properties of solutions 2.34, Energy methods 24. Wave equation 2.4.1, Solution by spherical means 2.4.2. Nonhiomogencous problem 2.4.3. Energy methods 2.5. Problems 2.6, References Nonlinear First-Order PDE 3.1, Complete integrals, envelopes 3.1.1. Complete integrals 3.1.2. New solutions from envelopes 3.2. Charactetisties 3.2.1. Derivation of characteristie ODE 3.2.2. Examples 8.2.8. Boundary conditions 3.2.4, Local solution 3.2.5. Applications 3.3. Introduction to Hamilton-Jacobi equations 3.3.1. Calculus of variations, Hamilton's ODE 3.3.2. Legendre transform, Hopf Lax formula 3.3.3. Weak solutions, uniqueness 3.4, Introduction to conservation laws 3.4.1. Shocks, entropy condition CONTENTS: 20 25 95 97 200 100 103 Lui 0 us 121 122 17 136 3 M4CONTENTS: 3.4.3, Weak solutions, uniqueness 3.4.4, Riemann’s problem 3.4.9, Long time behavior 3.5. Problems 3.6, References ae 4, Other Ways to Represent Solutions 4.2. Similarity solutions 4.2.1. Plane and traveling waves, solitons 180 4.2.2. Similarity under scaling 189 4.3, Transform methods 191 4.3.1. Fourier transform 191 4.3.2. Laplace transform 200 4.4, Converting nonlinear into linear PDB 203 4.4.1. Hopf-Cole transformation 203 4.4.2. Potential functions . ... . a ai 44.3. Hodograph and Logendre transforms 207 4.5. Asymptoties 209 45.1. Singular perturbations 209 4.5.2. Laplace's method au 4.5.8. Geometric opties, stationary phase 217 4.5.4, Homogenization 229 4.6. Power series 232 4.6.1. Noneharacteristie surfaces 232 4.6.2. Real analytic functions 237 4.6.3, Cauchy-Kovalevskaya Theorem 239 47. Problems 245 4.8. References 27 PART I: THEORY FOR LINEAR PARTIAL DIFFERENTIAL EQUATIONS 5. Sobolev Spaces 251Sel. Holder spaces 5.2. Sobolev spaces 5.2.1, Weak derivatives . “2, Deinition ot Sobolev spaces 5.2.8. Flementary properties . Approximation : He 5.3.1, Interior approximation by smooth functions 5.3.3. Global approximation by smooth functions |. Extensi Traces Sobolev inequalities 5.6.1. Gagliardo-Nirenberg-Sobolev inequality 5.0.2. Morrey’s inequality 5.6.8. General Sobolev inequalities 5.7. Compactness 5.8. Additional topies 5.8.1. Poinearé's inoqualitios 5.8.2. Difference quotients 5.8.3. Difforentiability a. 5.84. Fourier transform .9. Other spaces of functions 5.9.1. The space H-! 3.¥.2. Spaces involving time 5.10. Problems 5.11. References Second-Order Elliptic Equations 6.1, Definitions 6.1.1, Elliptic equations 6.1.2. Weak solutions 6.2, Existence of weak solutions 6.2.1, Lax-Milgram Theorem BaE ncthods CONTENTS: 289) 280) 291 295 297 298, 298, 300 305 308 309) 309) 312 au auCONTENTS: 6.2.2, Buergy estimates 2.3, Fredholm alternative 6.3. Regularity 6.3.1, Interior regularity 6.3.2. Ronndary regularity 6.4, Maximum principles . 6.4.1. Weak maximum principle 6.4.3. Harnack’s inequality 6.5. Bigenvalues and eigenfunctions 6.5.1. Bigenvalues of symmetric elliptic operators 6.5.2. Bigenvalues of nonsymmetric elliptic operators 6.6. Problems 0.7. References 7. Linear Evolution Equations 7.1. Second-order parabolic equations 7.1.1. Definitions 'xistence of woak solutions 3. Regularity 7.1.4. Maximum principles 7.2. Second-order hyperbolic equations 7.2.1. Definitions 7.2.2. Existence of weak solutions 4.28. Regularity 7.2.4, Propagation of disturbances 7.2.5. Equations in two variables 7.3. Systems of first-order hyperbolic equations 7.3.1. Definitions 7.3.2. Symmetric hyperbolic systems 7.3.3, Systems with constant coefficients 7.4, Semigroup theory 7.4.1. Definitions, elementary properties 316 319 325 326 334 34 345 352, 353, 358, 359) 364 307 308 308 369 372 377 387, 398) 398, 401 aug Als 49 422 423, 424 431 436 436xiv CONTENTS: 74.2. Geweraling contraction seinigroups 442 7.4.3. Applications M5 7.5. Problems . .. fe 9 7.6. Keterences 402 PARE Il: THEORY FOR NONLINEAR PARTIAL DIFFERENTIAL EQUATIONS 8. The Calculus of Variations 456 8.1, Introduction 456 S111, Basie ideas 456 8.1.2. First variation, Euler-Lagrange equation 457 8.1.3. Second variation 461 S.14, Systems 463 8.2. Existence of minimizers 468 8.2.1. Coercivity, lower sen 469 8.2.2. Convexity a7 8.2.3. Weak solutions of Puler-Lagrange equation 476 8.24, Systems coe cee 479 8.3. Regularity 484 8.3.1. Second derivative estimates 485 8.3.2. Remarks on higher regularity 488 8.4, Constraints 490 $8.41, Nonlinear eigenvalue probi 490 8.4.2, Unilateral constraints, variational inequalities . 494 8.4.3. Harmonie maps 498 8.4.4, Incompressibility . . aa 5u0 8.5. Critieal points 505 8.5.1, Mountain Pass Theorem . ..... 505 8.5.2. Application to semilinear elliptic PDE 510 8.6, Problems 516 8.7, References . . ae a 519 9. Nonvariational Techniques 520 9.1. Monvt ity methods 520,CONTENTS: 9.2. Fixed point arethods 9.2.1, Banach’s Fixed Point Theorem 9.2.2, Schiauder's, Schaefer's Fixed Point Theorems ¥.8. Method ot subsolutions and supersolutions 9.4. Nonexistence 94.1. Blow-up . 9.4.2. Derrick-Pohozaey identity 9.5.1. Star-shaped level sets 9.5.2. Radial symmetry 9.6. Gradient flows 9.6.1. Convex functions on Hilbert spaces 9.6.2. Subdifferentials, nonlinear semigroups 9.0.3. Applications 9.7. Problems 9.8. References 10, Hamilton-Jacobi Equations 1.1. Introduction, ¥ 10.1.1. Definitions 10.1.2. Consistency 10.2. Uniqueness 10.3. Control theory, dynamic programming 10.3.1. Introduction to control theory 1.8.2. Dynamie programming: 10.3.3. Hamilton-Jacol 10.3.4. Hopf-Lax formula revisited 104. Problems 105. References 11, Systems of Conservation Laws LLL. Introduction 11.1.1, Integral solutions 11.1.2. Traveling waves, hyperbolic systems iscoei i-Bellman equation 528 528 532 os7 BAT 5a 547 549 553 553 559 509 567 509 571 Bz 573 576 583 583 ps. 587 594 596 598 599 599 602 605CONTENTS: 11.2, Riemau’s problem 11.21, Simple waves 11.2.2, Rarefaetion waves . 11.2.3, Shock waves, contact discontinuities 11.2.4. Loeal solution of Riemann’s problem 11.3. Systems of two conservation laws 11.3.1, Riemann invariants 114. Entropy eriteria 11.4.1, Vanishing viseosity, traveling waves 114.2. Bntropy/entropy flux pairs 11.4.3. Uniqueness for a sealar conservation law 11.5. Problems 1.6. References APPENDICES: Appendix A: Notation A.1. Notation for matrices A.2. Geometric notation A.3. Notation for functions AA. Vector-valued functions A.5, Notation for estimates AS, Ls about notation Appendix B: Inequalities B.1. Convex functions B.2. Blementary inequalities 1s Frets Some cou Appendix C: Cale Cal. Boundaries . : C2. Gauss-Green Theorem C3. Polar coordinates, coarea formula, CA. Convolution and smoothing C55. Inverse Funetion Theorem C6. Implicit Punetion Theorem o12 612 615 cy 624 626 oT 634 638 642 oT os, 619, 19) 650, 651 655 656 or 67, 657, 658 663, 663, 664 665 666 669) 70CONTENTS: C7. Uniform convergence Appendix D: Linear Functional Analysis D.1, Banach spaces 11.2, Hilbert spaces 1.3. Bonnded linear operators DA, Weak convergence ee Appendix E: Measure Theory E. 1.2. Measurable functions and integration Lebesgue measure F.3, Convergence theorems for integrals EA. Differentiation F.5, Banach space-valued functions Bibliography IndexPREFACE I present in this book a wide-ranging survey of snany important topics in the theory of partial differential equations (PDE), with particular emphasis ‘on various modern approaches. I have made a huge number of editorial decisions about what to keep and what to toss out, and can only claim. that this selection seems to me about right. I of course include the ususl formulas for solutions of the usual linear PDB, but also devote large amounts of exposition 1o energy methods within Soboley spaces, to the calculus of variations, (0 conservation laws, etc. My general working principles in the writing have been these: a. PDE theory is (mostly) not restricted to two independent vari- ables. Many texts describe PDI as Il funetions of the two variables (2-,y) or (2,1) were all that matter. This emphasis seems to me misleading, as modern discoveries concerning many types of equations, both linear and nonlinear, have allowed for the rigorous treatment of these in any number of dimensions. I also find it unsatisfactory to “classify” partial differential atuaiivuis. this in parsibiv in iu vatialles, bui eiwaive ihie faine innpessivn that there is some kind of general and useful classification scheme available in general, b. Many interesting equations are nonlinear. My view is that overall we know too much about linear PDE and too little about nonlinear PDE. I have accordingly introduced nonlinear concepts early in the text and have tried hard to emphasize everywhere nonlinear snalognes of the linear theory.PREFACE. c. Understanding generalized solutions is fundamental. Many of the partial differential equations we study, especially nonlinear first-order equa- tions, do not in general possess smooth solutions. It is therefore essential to devise some kind of proper notion of generalized or weak solution. This is an important but subtle undertaking, and much of the hardest. material in this book concerns the nniqueness of appropriately defined weak solutions d. PDE theory is not a branch of functional analysis, Whereas certain classes of equations can profitably be viewed as generating abstract operators between Banach spaces, tte on an overiy abstract view- point, and consequent. ignoring of deep calculus and measure theoretic esti- mates, is ultimately limiting. e. Notation is a nightmare. I have really tried to introduce consistent jotation, which works for all the important classes of equations studied. attempt is sometimes at variance with notational conventions within a suboreo. f. Good theory is (almost) as useful as exact formulas. | incorporate this principle into the overall organization of the text, which is subdivided into three parts, roughly mimicking the historieal development of PDE the- ory itself. Part I concerns the search for explicit formulas for solutions, and Part II the abandoning of this quest in favor of general theory asserting the existence and other properties of solutions for linear equations, Part HIT is the mostly modern endeavor of fashioning general theory for important Classes of uontinear PDE. Let me also explicitly comment tere hal 1 inte ‘each section to be rigorous and complete (exceptions being the frankly heuristic treatment of asymptoties in §4.5 and an occasional reference to a research paver). ‘This means that even locally within each chapter the tovies do not necessarily progress logically from “easy” to “hard” concepts. There are many difficult proofs and computations early on, but as compensation, ‘many easier ideas later. The student should certainly omit on first reading some of the more arcane proofs. I wish next to emphasize that this is a textbook, and not a reference book. T have tried everywhere to present the essential ideas in the clearest possible settings, and therefore have almost never established sharp versions of any of the theorems. Research articles and advanced monographs, many of them fisted in the Bibliography, provide such precision aud generality. the development5x, PREFACE My goal lus rather been lu exploit, as best T ean, Ue many fuuslamental ideas of the subject within fairly simple contexts. T have greatly profited from the comments and thoughtful suggestions of many ot my colleagues, triends and students, in particular: S. Antman, J. Bang, X. Chen, A. Chorin, M. Christ, J. Cima, P, Colella, J. Cooper, M. Crandall, B. Driver, M. Feldman, M. Fitzpatrick, R. Gariepy, J. Gold- stein, D Gomes, 0 Hald, Wo Han, W. Hrusa, T Hmanen, 1 Ishii, 1 Ismael, R, Jorrard, C. Jones, B. Kawoll S. Koike, J. Lewis, TP. Lin, H. Lopes, F, Rezakhanlou, W. Schlag, D. Serre, P. Souganidis, J. Strain, W. Strauss, M. Struwe, R. Temam, B. Tvedt, JL. Vazquez, M. Weinstein, P. Wolfe, and Y. Zheng. Lespecially thank Tai-Ping Liu for many the frst draft of what is now Chapter 11 years ago writing out for me Tam extremely grateful for the suggestions and lists of mistakes from ‘earlier drafts of this book sent me by many readers, and I encourage others to sond mo their comments, at evansGmath.berkoley.edu. T have come to realize that I must be more than slightly mad to try to write a book of this length and complexity, but I am not yet crazy enough to think that I have made no mistakes. I will therefore maintain a listing of errors which come to light, and will make this accessible through the math.berkeley.edu homepage. Faye Yeager at UC Berkeley has done a really magnificent job typing and updating these notes, and Jaya Nagendra heroically typed an earlier version at the University of Maryland. My deepest thanks to both T have eon supported hy the NSF during much of the writing, most recently under grant DMS-9124342, LCE August, 1997 BerkeleyChapter 1 NTRODUCTION 1.1 Partial differential equations 1.2 Examples 1.3. Strategies for studying PDE 1.4 Overview 1.5 Problems This chapter surveys the principal theoretical issues concerning the solv- ing of partial differential equations. To follow the subsequent discussion, the reader should first of all turn to Appendix A and look over the notation presented there, particularly the multiindex notation for partial derivatives. 1.1. PARTIAL DIFFERENTIAL EQUATIONS A partial differential equation (PDE) is an equation involving an unknown function of two or more variables and certain of its partial derivatives. Using the notation explained in Appendix A, we can write ont symbol ically a typical PDE, as follows. Fix an integer k > 1 and let U denote an open subset of R”. DEFINITION. An expression of the form @ F(D¥u(z), D¥'u(a),...,Du(2),u(2),2)=0 (2 €U) i called a k'*-order partial differential equation, where FER™ x RY x. xX R"XRXU ORis given, and wiUoR is the unknown We solve the PDE if we find all u verifying (1), possibly only among those functions satisfying certain auxiliary boundary conditions on some part T of AU. By finding the solutions we mean, ideally, obtaining simple, explicit solutions, or, failing that, deducing the existence and other properties of solutions. DEFINITIONS. (i) The partial differential equation (1) is called linear if it has the form Y ea(z)D%u = f(z) \aise for given functions ag (\a| < k), f. This linear PDE is homogeneous if f =0. (ii) The PDE (1) is semilinear if it has the form )D*u + ap(D**u. , Du, u,z) = (iii) The PDE (1) is quasilinear if it has the form SY ag(D'1u,..., Du, u,2)D%u + ao(D*"u, lol=k (iv) The PDE (1) is fully nonlinear if it depends nonlinearly upon the highest order derivatives. +, Du,u,z) = 0. ‘A system of partial differential equations is, informally speaking, a col- lection of several PDE for several unknown functions. DEFINITION. An expression of the form (2) F(D‘u(z), D'1u(z),...,Du(z), u(e),2) =O (« €U) is called a k*"-order system of partial differential equations, where F:R™ x Rm xx R™ x R™ XU GR” is given and u:U>R™, us(u',...,u") is the unknown. Here we are supposing that the system comprises the same number m of scalar equations as unknowns (u!,...,u"). This is the most common circumstance, although other systems may have fewer or more equations than unknowns. Systems are classified in the obvious way as being linear, semilinear, ete.1.2. EXAMPLES 3 Remark. We use “PDE” as an abbreviation for both “partial differential equation” and “partial differential equations”. a 1.2, EXAMPLES There is no general theory known concerning the solvability of all partial differential equations. Such a theory is extremely unlikely to exist, given the rich variety of physical, geometric, and probabilistic phenomena which can be modeled by PDE. Instead, research focuses on various particular partial differential equations that are important for applications within and outside of mathematics, with the hope that insight from the origins of these PDE can give clues as to their solutions. Foo terest in current research. This listing is intended merely to familiarize the reader with the names and forms of varions famons PDR. To display most: clearly the mathematical structure of these equations, we have mostly set relevant physical constants to unity. We will later discuss the origin and interpretation of many of these PDE. Throughout x € U, where U is an open subset of R", and ¢ > 0. Also Du = Dru = (ur,,...,t2,) denotes the gradient of u with respect to the spatial variable x = (21,...,2n) 1.2.1. Single partial differential equations. a, Linear equations. 1. Laplace’s equation Au= ua, = i=l 2. Helmholtz’s (or eigenvalue) equation —Au = du. 3. Linear transport equation ut +3 us, =0. =I 4. Liouville’s equation w- Sus, = a1. INTRODUCTION 5. Heat (or diffusion) equation up — Au= 6. Schrédinger’s equation iu + Au =0. 7. Kolmogorou’s equation ue > Oars + You, ima = 8. Fokker-Planck equation Hh yi Ce Yew. =0. ijel =I 9. Wave equation i tye — Au = 0. 10. Telegraph equation ute + dug — Use = 0. 11. General wave equation Ouga, + > due, = ott, =I 12. Airy’s equation Up + User 13. Beam equation ty + Uses = 0.