Oblique derivative problems for elliptic and parabolic equations, Lecture II

Similar documents
Laplace s Equation. Chapter Mean Value Formulas

The De Giorgi-Nash-Moser Estimates

TD M1 EDP 2018 no 2 Elliptic equations: regularity, maximum principle

u( x) = g( y) ds y ( 1 ) U solves u = 0 in U; u = 0 on U. ( 3)

VISCOSITY SOLUTIONS. We follow Han and Lin, Elliptic Partial Differential Equations, 5.

Maximum Principles for Parabolic Equations

Elliptic PDEs of 2nd Order, Gilbarg and Trudinger

Lecture No 1 Introduction to Diffusion equations The heat equat

Some lecture notes for Math 6050E: PDEs, Fall 2016

Local pointwise a posteriori gradient error bounds for the Stokes equations. Stig Larsson. Heraklion, September 19, 2011 Joint work with A.

Homogeniza*ons in Perforated Domain. Ki Ahm Lee Seoul Na*onal University

Ahmed Mohammed. Harnack Inequality for Non-divergence Structure Semi-linear Elliptic Equations

Growth Theorems and Harnack Inequality for Second Order Parabolic Equations

2. Function spaces and approximation

The continuity method

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games

UNIQUENESS OF POSITIVE SOLUTION TO SOME COUPLED COOPERATIVE VARIATIONAL ELLIPTIC SYSTEMS

A GENERALIZATION OF THE FLAT CONE CONDITION FOR REGULARITY OF SOLUTIONS OF ELLIPTIC EQUATIONS

An introduction to Birkhoff normal form

Convex solutions to the mean curvature flow

NOTES ON SCHAUDER ESTIMATES. r 2 x y 2

PHASE TRANSITIONS: REGULARITY OF FLAT LEVEL SETS

HARNACK S INEQUALITY FOR COOPERATIVE WEAKLY COUPLED ELLIPTIC SYSTEMS. Ari Arapostathis

Nonlinear Diffusion in Irregular Domains

2 A Model, Harmonic Map, Problem

arxiv: v1 [math.ap] 25 Jul 2012

ξ,i = x nx i x 3 + δ ni + x n x = 0. x Dξ = x i ξ,i = x nx i x i x 3 Du = λ x λ 2 xh + x λ h Dξ,

2 Two-Point Boundary Value Problems

Stationary mean-field games Diogo A. Gomes

Regularity of flat level sets in phase transitions

Observability and measurable sets

Mixed boundary value problems for quasilinear elliptic equations

Perron method for the Dirichlet problem.

REACTION-DIFFUSION EQUATIONS FOR POPULATION DYNAMICS WITH FORCED SPEED II - CYLINDRICAL-TYPE DOMAINS. Henri Berestycki and Luca Rossi

A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION

EXISTENCE AND REGULARITY RESULTS FOR SOME NONLINEAR PARABOLIC EQUATIONS

ENGI 4430 PDEs - d Alembert Solutions Page 11.01

Symmetry Methods for Differential and Difference Equations. Peter Hydon

COMPLEX SPHERICAL WAVES AND INVERSE PROBLEMS IN UNBOUNDED DOMAINS

Obstacle Problems Involving The Fractional Laplacian

Global Maxwellians over All Space and Their Relation to Conserved Quantites of Classical Kinetic Equations

Maximum Principles for Elliptic and Parabolic Operators

A note on W 1,p estimates for quasilinear parabolic equations

The Dirichlet boundary problems for second order parabolic operators satisfying a Carleson condition

Module 7: The Laplace Equation

v( x) u( y) dy for any r > 0, B r ( x) Ω, or equivalently u( w) ds for any r > 0, B r ( x) Ω, or ( not really) equivalently if v exists, v 0.

On a general definition of transition waves and their properties

Equilibria with a nontrivial nodal set and the dynamics of parabolic equations on symmetric domains

ENGI 9420 Lecture Notes 8 - PDEs Page 8.01

u(0) = u 0, u(1) = u 1. To prove what we want we introduce a new function, where c = sup x [0,1] a(x) and ɛ 0:

The double layer potential

NONTRIVIAL SOLUTIONS TO INTEGRAL AND DIFFERENTIAL EQUATIONS

BURGESS INEQUALITY IN F p 2. Mei-Chu Chang

ENGI 9420 Lecture Notes 8 - PDEs Page 8.01

Regularity of Weak Solution to Parabolic Fractional p-laplacian

GEOMETRIC APPROACH TO CONVEX SUBDIFFERENTIAL CALCULUS October 10, Dedicated to Franco Giannessi and Diethard Pallaschke with great respect

PERTURBATION THEORY FOR NONLINEAR DIRICHLET PROBLEMS

Green s Functions and Distributions

The Helmholtz Equation

On Smoothness of Suitable Weak Solutions to the Navier-Stokes Equations

A Necessary and Sufficient Condition for the Continuity of Local Minima of Parabolic Variational Integrals with Linear Growth

Sobolev Spaces. Chapter Hölder spaces

Luis Escauriaza Universidad del País Vasco

Chapter 1. Preliminaries. The purpose of this chapter is to provide some basic background information. Linear Space. Hilbert Space.

STOP, a i+ 1 is the desired root. )f(a i) > 0. Else If f(a i+ 1. Set a i+1 = a i+ 1 and b i+1 = b Else Set a i+1 = a i and b i+1 = a i+ 1

for all subintervals I J. If the same is true for the dyadic subintervals I D J only, we will write ϕ BMO d (J). In fact, the following is true

CLASSIFICATION AND PRINCIPLE OF SUPERPOSITION FOR SECOND ORDER LINEAR PDE

MATH 6337: Homework 8 Solutions

Interior Gradient Blow-up in a Semilinear Parabolic Equation

Non-divergence Elliptic Equations of Second Order with Unbounded Drift

A BOUNDARY VALUE PROBLEM FOR MINIMAL LAGRANGIAN GRAPHS. Simon Brendle & Micah Warren. Abstract. The associated symplectic structure is given by

Existence, stability and instability for Einstein-scalar field Lichnerowicz equations by Emmanuel Hebey

Analysis in weighted spaces : preliminary version

LOWER AND UPPER SOLUTIONS TO SEMILINEAR BOUNDARY VALUE PROBLEMS: AN ABSTRACT APPROACH. Alessandro Fonda Rodica Toader. 1.

The Dirichlet boundary problems for second order parabolic operators satisfying a Carleson condition

Controllability of the linear 1D wave equation with inner moving for

SHARP BOUNDARY TRACE INEQUALITIES. 1. Introduction

arxiv: v1 [math.ap] 18 Jan 2019

MAT 257, Handout 13: December 5-7, 2011.

The purpose of this lecture is to present a few applications of conformal mappings in problems which arise in physics and engineering.

Deforming conformal metrics with negative Bakry-Émery Ricci Tensor on manifolds with boundary

INTEGRABILITY OF SUPERHARMONIC FUNCTIONS IN A JOHN DOMAIN. Hiroaki Aikawa

Point estimates for Green s matrix to boundary value problems for second order elliptic systems in a polyhedral cone

i=1 α i. Given an m-times continuously

Some remarks on the elliptic Harnack inequality

Applications of the periodic unfolding method to multi-scale problems

GLOBAL LIPSCHITZ CONTINUITY FOR MINIMA OF DEGENERATE PROBLEMS

We denote the space of distributions on Ω by D ( Ω) 2.

Multiple integrals: Sufficient conditions for a local minimum, Jacobi and Weierstrass-type conditions

Second Order Elliptic PDE

Simple Iteration, cont d

The heat equation in time dependent domains with Neumann boundary conditions

On the bang-bang property of time optimal controls for infinite dimensional linear systems

EXISTENCE OF SOLUTIONS TO ASYMPTOTICALLY PERIODIC SCHRÖDINGER EQUATIONS

Stability for parabolic quasiminimizers. Y. Fujishima, J. Habermann, J. Kinnunen and M. Masson

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation

The role of Wolff potentials in the analysis of degenerate parabolic equations

HARNACK INEQUALITY FOR NONDIVERGENT ELLIPTIC OPERATORS ON RIEMANNIAN MANIFOLDS. Seick Kim

First Boundary Value Problem for the Diffusion. Equation. I. Iterated Logarithm Test for the Boundary. Regularity and Solvability

x y More precisely, this equation means that given any ε > 0, there exists some δ > 0 such that

Transcription:

of the for elliptic and parabolic equations, Lecture II Iowa State University July 22, 2011

of the 1 2 of the

of the As a preliminary step in our further we now look at a special situation for elliptic. We begin by taking α 1, ε, and ω 0 to be positive constants with α 1 < 1 and ε < 1. We then set A = 2 1+2ε ε 4ε, and we define G : R n (0, ) R by ( y 2 ) (1+ε)/2 G(y, r) = r 2 + α1 2 + (y n ) 2 (Aω 0 r 2 ). We also introduce, for any x 1 R n, the set E(x 1, r) = {x : G(x x 1, r) < 1}.

of the Let R > 0 and suppose x n > ω 0 x and ω 0 β (1 ε)β n on Ω B(0, R). Let r (0, R) and suppose x 1 = (0,..., x1 n) satisfies the s ) x1 n (A ε6 ω 0 r 4 and E(x 1, r) B(0, R). Then η, defined by η(x) = 1 G(x x 1, r) satisfies the inequality βn β Dη ε 4 Aω 0 r on S, the subset of Ω on which G < 1.

of the To prove this estimate, we first observe that ε 6 /4 < 1 and A > 2, so A ε 6 /4 > 1. Then x n < x1 n on S because ) x1 n x n (A ε6 ω 0 r ω 0 x 4 and x < r on S. We now set N = x 1 n x n, P = x Aω 0 r r, α = ε6 4.

Then of the N 1 α A P A, (P2 + α 2 1) (1+ε)/2 + N 2 1. Combining these inequalities yields and hence ( P 1+ε + 1 α A P ) 2 1, A P 1+ε 2P A 2α A.

of the It follows that because otherwise ( 2 + 2ε 2 P A ) 1/ε P 1+ε 2P A = [APε 2] P A > [(2 + 2ε2 ) 2] P A = 2ε2 P A, and P (2/A) 1/ε = ε 4 /4, so 2ε 2 P A 2ε6 4A = 2α A.

of the To get a lower bound for N, first observe that where Now we compute where α + P ε6 4 + ε4 4 H(ε), H(s) = (1 + s 2 ) 1/s. H (s) = H(s) g(s2 ) s 2, g(σ) = 2σ ln(1 + σ). 1 + σ Since g (σ) = (1 σ)/(1 + σ) 2, it follows that g is increasing on [0, 1]. But g(0) = 0, so g is nonnegative on [0, 1] and hence H takes its maximum at s = 1.

of the Since H(1) = 2, it follows that P + α ε6 4 + ε4 2 ε4. The inequality A 2 yields so N 1 ε4 2 β Dη = 2β n (x n 1 x n ) (Aω 0 r) 2 (1 + ε)β x r 2 (P2 + α 2 1) (ε 1)/2 2 βn Aω 0 r (N 1 2 (1 ε2 )AP ε ) 2 βn Aω 0 r (1 1 2 ε4 1 + ε 4 ) = ε 4 βn Aω 0 r.

of the We are now ready to prove the corresponding estimate for the parabolic case. We now suppose that Ω satisfies an interior parabolic cone at X 0. Specifically, we assume that ω 0 β (1 ε)β n, x n < ω 0 x + ω 1 t 1/2 on SΩ Q(0, R). To simplify the arguments, we shall assume that b = 0, c = 0, and β 0 = 0. With G and η as before, we define A 0 = 1 + 64ω2 1 ε 12 ω0 2.

of the We also assume that the eigenvalues of [a ij ] are in the interval [λ, Λ] for positive constants λ Λ, and we suppose that Lu f in Ω, 0 on SΩ. (We make no assumptions about u on BΩ.) Also, we set F (r) = r n/(n+1) f n+1, with r so small that G(x x 1, r) 1 for X SΩ with x R and t ( r 2, 0) and x n 1 ) (A ε6 ω 0 r. 8

of the Then, for any numbers α 0, θ and θ 1 in (0, 1), there is a positive constant C 1 such that if u(x ) 0 for all X Ω with G(x x 1 ) < 1 and t ( r 2 /A 0, 0) u(x ) + F (r) h for all X Ω with G(x x 1 ) < θ and t = r 2 /A 0 (here h is a nonnegative constant), then u(x ) + F (r) C 4 h for all X Ω with G(x x 1 ) < θ 1 and t = (α 0 1)r 2 /A 0.

of the We now define three functions: ψ 0 = 1 θ2 α 0 (A 0 t + r 2 ) + θ 2 r 2, ψ 1 = max{ψ 0 r 2 G(x x 1, r), 0} and, for q 2 a constant to be chosen, ψ = ψ 2 1/ψ q 0. By observing that DG 2 Cψ 0 wherever ψ 1 1 2 ψ 0, we find that there are positive numbers N 1, N 2, N 3 (determined by this number C and by Λ and λ) such that Lψ ψ 1 q 0 [qn 1 1 θ 2 α 0 ξ 2 N 2 ξ + N 3 ], where ξ = ψ 1 /ψ 0. Hence, we can choose q so that Lψ 0 in Q = {X Ω : G < ψ 0, r 2 /A 0 < t < (α 0 1)r 2 /A 0 }.

of the Then, on SQ Ω, we have ψ = 0, while ψ 0 = θ 2 r 2 on BQ Ω, so ψ = (θ 2 r 2 r 2 G) 2 (θ 2 r 2 ) q (θr) 2q+4 there. Also, because β G C 2 /r (for some positive constant C 1 ) and ψ 1 r 2 in Q, β Dψ C 2 r ψ on SQ SΩ. Finally, when t = (α 0 1)r 2 /A 0, we have ψ 0 = r 2, so ψ (1 θ 2 1) 2 r 2q+4 if t = (α 0 1)r 2 /A 0 and G θ 1.

of the Now we set w = (θr) 2q 4 hψ ū and note that where Lw f in Q, β Dw w on SQ, β = w 0 on BQ, { r β on SΩ, C2 0 elsewhere. Our maximum principle gives w C 3 F (r) in Q. Rewriting this inequality for G < θ 1 and t = (α 0 1)r 2 /A 0, we see that ū C 4 h C 3 F (r), so (1 + C 3 )ū C 4 h. Hence, we have our result with C 1 = C 4 /(1 + C 3 ).

of the We are now in a position to prove a weak Harnack inequality, but it has some additional geometric complications compared to the usual one. We start by fixing a point X 0 SΩ and a positive number ρ and we write Q 1 for the intersection of Ω with the cylinder Q(X 0, ρ). Then there are constants B 1 1 and θ 1 (0, 1) and x 1 with x 1 = x 0 such that G(x x 1, B 1 ρ) < θ 1 in B(x 0, ρ) and x n 1 x n 0 ) (A ε6 ω 0 B 1 ρ. 4 We now write E 1 for the set of all X Ω on which G(x x 1, B 1 ρ) < θ 1 and t 0 (4A 0 /9)ρ 2 < t < t 0.

of the We now set h = inf Q1 u + F (ρ) and apply the estimate for each time level, noting that we can replace r in our with any convenient multiple of r. In this way, we obtain inf u + F (ρ) C(inf u + F (ρ)) E 2 Q 1 where E 2 is the set of all X Ω with G(x x 1, B 1 ρ) < θ and t 0 A 0 ρ 2 < t < t 0 (5A 0 /9)ρ 2, and θ (0, 1) is chosen so that there is a positive constant K such that d Kρ on E 2.

of the The standard weak Harnack inequality for parabolic equations then yields positive constants B 2, C, and κ such that (ρ n 2 Q 2 u κ dx ) 1/κ C(inf Q 1 u + F (ρ)) for Q 2 = Q((x 1, t 0 2A 0 ρ 2 ), B 2 ρ). We must assume for these arguments that t < t 0 (2A 0 + 4B 2 2 )ρ2 ) on BΩ, but it s possible to modify the arguments to replace this assumption with one on the behavior of u on BΩ.

of the The usual argument then gives a Hölder estimate for u near SΩ if Lu = f in Ω and Mu = 0 on SΩ. It s straightforward to modify the arguments to allow Mu = g on SΩ with g bounded (and to allow b, c, and β 0 to be nonzero), but that will take too long.

of the To prove Hölder continuity of the gradient of the solution, we introduce a different kind of for SΩ. We call any set of the form {X : T < t < 0, x ( t) 1/2 x 0 < R( t) 1/2 } (with R and T positive constants and x 0 R n ) a tusk. We say that SΩ satisfies a uniform exterior tusk at X 0 if there is a tusk such that, for each point X 0 SΩ, if the tusk is translated (so that the origin is translated to X 0 ) and then rotated in space, then the tusk is exterior to Ω. Note that the exterior tusk is weaker than the exterior parabolic cone.

of the A key fact, which we will not prove, is that, for any R and T and any positive numbers λ Λ, there are constants α > 0 and C determined only by R, T, x 0, λ, and Λ along with a function w with continuous second spatial s and continuous time exterior to the tusk (at least for t ( T, 0) satisfying the : X α w(x ) C X α, DX C X α 1, w t + a ij D ij w X α 2.

The of the Hölder gradient proceeds in a number of steps: 1 Assume that a ij and β are constant and all other coefficients are zero. of the

of the The of the Hölder gradient proceeds in a number of steps: 1 Assume that a ij and β are constant and all other coefficients are zero. 2 Show that is Hölder continuous.

of the The of the Hölder gradient proceeds in a number of steps: 1 Assume that a ij and β are constant and all other coefficients are zero. 2 Show that is Hölder continuous. 3 Prove an oscillation estimate for u P for a suitable polynomial P.

of the The of the Hölder gradient proceeds in a number of steps: 1 Assume that a ij and β are constant and all other coefficients are zero. 2 Show that is Hölder continuous. 3 Prove an oscillation estimate for u P for a suitable polynomial P. 4 Show that this oscillation estimate gives a Hölder gradient estimate.

of the The of the Hölder gradient proceeds in a number of steps: 1 Assume that a ij and β are constant and all other coefficients are zero. 2 Show that is Hölder continuous. 3 Prove an oscillation estimate for u P for a suitable polynomial P. 4 Show that this oscillation estimate gives a Hölder gradient estimate. 5 Use a perturbation argument to get the full result.

of the To be more specific, we shall look only in a parabolic neighborhood of some boundary point X 0. The neighborhood is then Q(X 0, r) for some r > 0. When a ij and β are constant, and all other coefficients are zero, then v = satisfies the s v t + a ij D ij v = 0 in Ω Q(X 0, r), v = 0 on SΩ Q(X 0, r).

of the Since SΩ satisfies a uniform exterior tusk, for each Y SΩ Q(X 0, r), there is a function w Y satisfying the preceding s (with X replaced by X Y but C independent of Y ), so we have ( ) Lv L sup v 2α w Y r α in Ω[Y, r/2], v sup v 2α w Y r α where Ω[Y, r/2] = Ω Q(Y, r/2). on PΩ[Y, r/2],

of the The maximum principle implies that in Ω[Y, r/2], and hence v sup v 2α w Y r α d(x )α v(x ) C sup v r α for X Ω[X 0, r/2]. A similar argument with v in place of v then implies that v C sup v d α r α, which says that v is Hölder continuous up to SΩ.

of the Note that the preceding result holds without assuming that β is oblique, in particular, the interior parabolic cone is not needed here. The next step will use these assumptions.