Stability Analysis of Stationary Solutions for the Cahn Hilliard Equation

Similar documents
Spectral Analysis for Periodic Solutions of the Cahn Hilliard Equation on R

with deterministic and noise terms for a general non-homogeneous Cahn-Hilliard equation Modeling and Asymptotics

Asymptotic behavior near transition fronts for equations of generalized Cahn Hilliard form

Nonlinear stability of time-periodic viscous shocks. Margaret Beck Brown University

Numerical Approximation of Phase Field Models

Lecture No 1 Introduction to Diffusion equations The heat equat

Slow Modulation & Large-Time Dynamics Near Periodic Waves

A review of stability and dynamical behaviors of differential equations:

Dynamics and stochastic limit for phase transition problems. problems with noise. Dimitra Antonopoulou IACM-FORTH

NUCLEATION AND SPINODAL DECOMPOSITION IN TERNARY-COMPONENT ALLOYS

NONLOCAL DIFFUSION EQUATIONS

Nonlinear stability of semidiscrete shocks for two-sided schemes

Droplet Formation in Binary and Ternary Stochastic Systems

Step Bunching in Epitaxial Growth with Elasticity Effects

Stability, the Maslov Index, and Spatial Dynamics

Math 575-Lecture 26. KdV equation. Derivation of KdV

Pattern formation in Nikolaevskiy s equation

The Evans function and the stability of travelling waves

A sharp diffuse interface tracking method for approximating evolving interfaces

Math Partial Differential Equations 1

Some asymptotic properties of solutions for Burgers equation in L p (R)

On Multigrid for Phase Field

PARTIAL DIFFERENTIAL EQUATIONS. Lecturer: D.M.A. Stuart MT 2007

UNIQUENESS OF POSITIVE SOLUTION TO SOME COUPLED COOPERATIVE VARIATIONAL ELLIPTIC SYSTEMS

Navier-Stokes equations in thin domains with Navier friction boundary conditions

Elliptic Kirchhoff equations

Nonlinear instability of half-solitons on star graphs

Saddle Solutions of the Balanced Bistable Diffusion Equation

Hyperbolic Systems of Conservation Laws

String method for the Cahn-Hilliard dynamics

MATH 220 solution to homework 4

Presenter: Noriyoshi Fukaya

THE CAHN-HILLIARD EQUATION WITH A LOGARITHMIC POTENTIAL AND DYNAMIC BOUNDARY CONDITIONS

THE CAUCHY PROBLEM AND STEADY STATE SOLUTIONS FOR A NONLOCAL CAHN-HILLIARD EQUATION

Exponential Stability of the Traveling Fronts for a Pseudo-Para. Pseudo-Parabolic Fisher-KPP Equation

Separation of Variables in Linear PDE: One-Dimensional Problems

A REMARK ON THE GLOBAL DYNAMICS OF COMPETITIVE SYSTEMS ON ORDERED BANACH SPACES

Introduction LECTURE 1

Linear Theory of Evolution to an Unstable State

Decay profiles of a linear artificial viscosity system

Degenerate Nucleation in the Cahn Hilliard Cook Model

arxiv:math/ v1 [math.ap] 28 Oct 2005

Allen Cahn Equation in Two Spatial Dimension

Spinodal decomposition

PATTERN FORMATION IN A MIXED LOCAL AND NONLOCAL REACTION-DIFFUSION SYSTEM

Spectral stability of periodic waves in dispersive models

The stability of travelling fronts for general scalar viscous balance law

Coarsening Dynamics for the Cahn-Hilliard Equation. Jaudat Rashed

Nonlinear convective stability of travelling fronts near Turing and Hopf instabilities

Numerical Evans Function Computation Jeffrey Humpherys Brigham Young University INTRODUCTION

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games

Lecture Notes on PDEs

Weak solutions for the Cahn-Hilliard equation with degenerate mobility

Numerical methods for a fractional diffusion/anti-diffusion equation

Breakdown of Pattern Formation in Activator-Inhibitor Systems and Unfolding of a Singular Equilibrium

ERROR ANALYSIS OF STABILIZED SEMI-IMPLICIT METHOD OF ALLEN-CAHN EQUATION. Xiaofeng Yang. (Communicated by Jie Shen)

Existence of Secondary Bifurcations or Isolas for PDEs

Pattern size in Gaussian fields from spinodal decomposition

ORBITAL STABILITY OF SOLITARY WAVES FOR A 2D-BOUSSINESQ SYSTEM

Finite Difference Solution of the Heat Equation

Stochastic nonlinear Schrödinger equations and modulation of solitary waves

There is a more global concept that is related to this circle of ideas that we discuss somewhat informally. Namely, a region R R n with a (smooth)

Turning points and traveling waves in FitzHugh-Nagumo type equations

Heat kernels of some Schrödinger operators

Fission of a longitudinal strain solitary wave in a delaminated bar

Solutions of Nonlinear Planar Elliptic Problems with Triangle Symmetry

1 Assignment 1: Nonlinear dynamics (due September

Simple Explanation of Fermi Arcs in Cuprate Pseudogaps: A Motional Narrowing Phenomenon

On the split equality common fixed point problem for quasi-nonexpansive multi-valued mappings in Banach spaces

Numerical Solutions to Partial Differential Equations

Differential equations, comprehensive exam topics and sample questions

Nonlinear Wave Propagation in 1D Random Media

Understand the existence and uniqueness theorems and what they tell you about solutions to initial value problems.

Stability of Dynamical systems

On the Solvability Conditions for a Linearized Cahn-Hilliard Equation

LIFE SPAN OF BLOW-UP SOLUTIONS FOR HIGHER-ORDER SEMILINEAR PARABOLIC EQUATIONS

The Dynamics of Nucleation in Stochastic Cahn Morral Systems

VIII. Phase Transformations. Lecture 38: Nucleation and Spinodal Decomposition

Generalized transition waves and their properties

EXISTENCE OF SOLUTIONS TO THE CAHN-HILLIARD/ALLEN-CAHN EQUATION WITH DEGENERATE MOBILITY

ASYMPTOTIC STRUCTURE FOR SOLUTIONS OF THE NAVIER STOKES EQUATIONS. Tian Ma. Shouhong Wang

MATH 205C: STATIONARY PHASE LEMMA

2012 NCTS Workshop on Dynamical Systems

Math 331 Homework Assignment Chapter 7 Page 1 of 9

DYNAMICAL THEORY OF THERMODYNAMICAL PHASE TRANSITIONS

Math 273, Final Exam Solutions

Diffusive Transport Enhanced by Thermal Velocity Fluctuations

Partial Differential Equations, Winter 2015

From Boltzmann Equations to Gas Dynamics: From DiPerna-Lions to Leray

Spectral theory for magnetic Schrödinger operators and applicatio. (after Bauman-Calderer-Liu-Phillips, Pan, Helffer-Pan)

Stability for a class of nonlinear pseudo-differential equations

RANDOM PROPERTIES BENOIT PAUSADER

MATH 251 Final Examination December 16, 2014 FORM A. Name: Student Number: Section:

Integro-differential equations: Regularity theory and Pohozaev identities

On a Lyapunov Functional Relating Shortening Curves and Viscous Conservation Laws

Propagating terraces and the dynamics of front-like solutions of reaction-diffusion equations on R

THE WILLMORE FUNCTIONAL AND INSTABILITIES IN THE CAHN-HILLIARD EQUATION

Fractal Conservation Laws: Global Smooth Solutions and Vanishing Regularization

MIN-MAX VARIATIONAL PRINCIPLE AND FRONT SPEEDS IN RANDOM SHEAR FLOWS

Project Mentor(s): Dr. Evelyn Sander and Dr. Thomas Wanner

Transcription:

Stability Analysis of Stationary Solutions for the Cahn Hilliard Equation Peter Howard, Texas A&M University University of Louisville, Oct. 19, 2007 References d = 1: Commun. Math. Phys. 269 (2007) 765 808. d 2: Physica D 229 (2007) 123 165.

The Cahn Hilliard Equation We consider the Cahn Hilliard equation u t = {M(u) (F (u) κ u)}; x R d, d 1 (CH) where M C 2 (R), F C 4 (R), and we make the standard phenomenological assumptions M(u) M 0 > 0, and that F (u) has a double well form. F(u) u 1

The Physical Setting u t = {M(u) (F (u) κ u)} One setting in which the Cahn Hilliard equation arises is in the modeling of spinodal decomposition, a phenomenon in which the rapid cooling of a homogeneously mixed binary alloy causes separation to occur, resolving the mixture into regions in which one component or the other is dominant, with these regions separated by steep transition layers. More precisely, this phase separation is typically considered to take place in two steps: First, a relatively fast process occurs in which the homogeneous mixture quickly begins to separate (this is spinodal decomposition), followed by a second coarsening process in which the regions continue to grow on a relatively slower timescale. In this context, u denotes the concentration of one component of the alloy (or a convenient affine transformation of this concentration), F denotes the bulk free energy density of the alloy, M denotes the molecular mobility, and κ is a measure of interfacial energy. 2

Miller et al. Experiment with Fe Cr, 1995 3

Phenomenological Derivation Since u is a conserved density it satisfies a conservation law u t + J = 0. If we let E(u) denote the total free energy for a system in configuration u, then a standard phenomenological assumption is J = M(u) δe δu. In 1958, Cahn and Hilliard recognized that an appropriate energy is E(u) = Ω F (u) + κ 2 u 2 dx. 4

Form of the Free Energy Density For Helmholtz free energy F = U T S, F T = S. F(u) F(u) Temperature Drops u u l u r u 5

(CH): u t = {M(u) (F (u) κ u)} For any line G(u) = Au + B, we can replace F with F G, and we can also shift F with a change of variable w = u c. Without loss of generality, we can put F in the (not necessarily even) standard form: F(u) u u 1 2 u The values u 1 and u 2 are called the binodal values. This transformation places the generic homogeneous configuration at u h = 0. 6

Stability of Stationary Solutions Generally, we say that a stationary solution ū(x) is stable if for u(t, x) = ū(x) + v(t, x), with v(0, x) small in some appropriate sense, v(t, x) remains small as time evolves. Upon substitution of u into (CH) we obtain an equation for v: v t = Lv + Nv, where Lv = {M(ū) (F (ū)v κ v)}. 7

The Stationary Solutions for d = 1 In d = 1, u t = {M(u)(F (u) κu xx ) x } x. In this case, there are only three types of non-constant bounded stationary solutions ū(x) to (CH). Periodic solutions Pulse type (reversal) solutions ū(± ) = u ±, u = u + Transition waves ū(± ) = u ±, u u + The homogeneous state corresponds with a constant solution to u(t, x) u h, which is typically in the spinodal region {u : F (u) < 0}. 8

1 Periodic Solutions and Limiting Transition Wave F (u) = 1 8 u4 1 4 u2 M(u) 1, κ = 1 Equilibrium Solutions 0.8 0.6 0.4 0.2 0 0.2 0.4 0.6 0.8 1 10 8 6 4 2 0 2 4 6 8 10 x Leading eigenvalues as a function of amplitude 0.07 u=0 corresponds with constant state 0.06 Leading Eigenvalue 0.05 0.04 0.03 0.02 0.01 u=1 corresponds with transition wave 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Amplitude 9

Multiple Dimensions u t = {M(u) (F (u) κ u)} For d 2, we have a broader cast of stationary solutions. For example, in the case d = 2, some of the players are: 1. Three types of planar solutions. 2. Radial solutions ū(r) that satisfy ū (0) = 0, ū (r) < 0 for r > 0, and lim r ū(r) = u. 3. Saddle solutions, which have infemum u 1 and supremum u 2 (the binodal values), and which have the same sign as xy. 4. Doubly periodic solutions. Focus on the transition front (or kink) solution. 10

Existence and Structure of the Kink Solution Given any F in double well form, there exist precisely two kink solutions to (CH), ū(x 1 ) and ū( x 1 ), both of which are strictly monotonic, and both of which approach the binodal values u 1 < u 2 : lim ū(x 1) = u 1 ; x 1 lim ū(x 1) = u 2. x 1 + F(u) u u 1 2 u 11

The Result of Jasnow and Zia Upon linearization of (CH) about the kink solution ū(x 1 ), we obtain a perturbation equation with linear part v t = Lv := {M(ū) (F (ū)v κ v)}. The relevant eigenvalue problem is Lφ = λφ. We take the Fourier transform in the transverse variable x = (x 2, x 3,..., x d ) ( x ξ): ˆL ξ ˆφ = λ ˆφ. In the paper Crossover of Interfacial Dynamics, [Physical Review A 36 (1987) 2243 2247], David Jasnow and R. K. P. Zia showed that for M = M 0 = constant and a double Gaussian F, the leading eigenvalue λ (ξ) satisfies λ (ξ) = c ξ 3 + o( ξ 3 ). 12

The Result of Korvola, Kupiainen, and Taskinen Theorem. For the case M 1, κ = 1, and F (u) = 1 8 u4 1 4 u2, the kink solution ū(x 1 ) = tanh(x 1 /2) is stable for d 3 as follows: if ( u(0, x) ū(x 1 ) (1 + x ) r) δ, sup x R d for some δ sufficiently small, then u(t, x) = ū(x 1 ) + Aū (x 1 )φ(t, x) + v(t, x), where Here A = 1 2 R d u(0, x) ū(x 1 )dx; sup v(t, x) Ct 1 12 d 1 3. x R d φ(t, x) = (2πt 1/3 ) 1 d e iξ x t 1/3 e 1 3 ξ 3 dk. R d 1 13

The KKT Methodology The result of Korvola, Kupiainen, and Taskinen appears as Anomalous Scaling for Three-Dimensional Cahn Hilliard Fronts, Comm. Pure Appl. Math. 58 (2005) 1 39. The method is based on a renormalization group approach. The main difference between the approach of [KKT] and the current approach is the manner in which λ (ξ) is treated. In [KKT], a correction associated with λ (ξ) is added as a first order term in a timeasymptotic expansion. In the current approach, λ (ξ) is accomodated by a consideration of its effect in shifting the front location. 14

d=1 u(0,x) _ u(x) x δ(t) u(t,x) _ u(x) x d=2 Shock Layer x 2 δ (t,x ) 2 x 1 δ (t,x ) 2 15

Local Tracking Implementation General perturbations of ū(x 1 ) lead to transient shifts along the transition front, and we build a tracking method into our linearization. (Follows Howard Zumbrun ARMA 2000, Hoff-Zumbrun IUMJ 2000, Howard-Hu ARMA 2006.) Setting u(t, x) = ū(x 1 δ(t, x)) + v(t, x), where x = (x 2, x 3,..., x d ) denotes the transverse variable, we find ( t L)v = ( t L)(δ(t, x)ū x1 (x 1 )) + Q, where Lv = {M(ū) (F (ū)v κ v)} Q = Q(ū, δ, v). 16

The Spectral Analysis The relevant eigenvalue problem is Lφ = λφ. We take the Fourier transform in the transverse variable x = (x 2, x 3,..., x d ) ( x ξ): where For ξ 0, we set ϕ = D 1/2 ξ D ξ H ξ ˆφ = λ ˆφ, D ξ := x1 M(ū) x1 + ξ 2 M(ū) H ξ := κ x1 x 1 + (F (ū) + κ ξ 2 ). ˆφ, so that L ξ := D 1/2 ξ H ξ D 1/2 ξ ϕ = λϕ. The advantage of this formulation is that L ξ is self-adjoint (on D(L ξ )). 17

The Spectrum for ξ 0 1. The essential spectrum of L ξ lies entirely on the real axis and is bounded to the left of both M(u ± )F (u ± ) ξ 2 κm(u ± ) ξ 2. 2. The point spectrum of L ξ lies entirely on the real axis and is bounded to the left of κm 0 ξ 4 (recall: M(u) M 0 > 0). 3. For ξ sufficiently small, the leading eigenvalue of L ξ satisfies λ ( ξ ) c 1 ξ 3, and the remainder of the point spectrum lies to the left of c 2 ξ 2. In order to carry out the nonlinear analysis, we require a very precise understanding of the leading eigenvalue λ ( ξ ). For this, we use the Evans function. 18

The Evans Function In this setting, the Evans function is simply D(λ, ξ) = det φ 1 φ 2 φ + 1 φ + 2 φ 1 φ 1 φ 1 φ 2 φ 2 φ 2 φ + 1 φ + 1 φ + 1 φ + 2 φ + 2 φ + 2 x1 =0, where the φ ± k denote solutions of L ξφ = λφ that decay at ±. The Evans function can be analytically defined in terms of the three variables r := ξ 2 ρ ± := λ + b± r + c ± r 2 ; b ± := M(u ± )F (u ± ); c ± := κm(u ± ). b ± + 2c ± r 19

The Leading Eigenvalue For ξ sufficiently small, the leading eigenvalue of L ξ satisfies λ ( ξ ) = λ 3 ξ 3 + O( ξ 4 ), where λ 3 = 2κ(M(u ) + M(u + )) [u] 2 max{u,u + } min{u,u + } F (x) F (u )dx. 20

The Nonlinear Analysis Return now to ( t L)v = ( t L)(δ(t, x)ū x1 (x 1 ))+ Q. Let G(t, x; y) denote the Green s function associated with t L. Then v(t, x) = G(t, x; y)v 0 (y)dy + δ(t, x)ū x1 (x 1 ) R d t + G(t s, x; y) Qdyds. 0 R d There exists a splitting G = G(t, x; y) + ū x1 (x 1 )E(t, x; y) so that G L p (R d ) t d 2 (1 1 p ) ū x1 E L p (R d ) t d 1 3 (1 1 p ). 21

Choosing δ(t, x) We write t v(t, x) = G(t, x; y)v0 (y)dy + G(t s, x; y) Qdyds R d 0 R d + δ(t, x)ū x1 (x 1 ) + ū x1 (x 1 ) E(t, x; y)v 0 (y)dy R d t + ū x1 (x 1 ) E(t s, x; y) Qdyds. R d 0 We choose δ(t, x) = E(t, x; y)v 0 (y)dy R d t 0 R d E(t s, x; y) Qdyds. 22

Nonlinear Stability of Planar Transition Waves Theorem. [H. Physica D, 2007] Under the assumptions of the talk, ū is nonlinearly stable as follows: for Hölder continuous initial perturbations (u(0, x) ū(x)) C 0+γ (R d ), γ > 0, with u(0, x) ū(x 1 ) L 1 x E 0 (1 + x 1 ) 3/2, for some E 0 sufficiently small, there holds, u(t, x) ū(x 1 δ(t, x)) L p (R d ) C [ (1+t) d 2 (1 1 p ) +(1+t) (d 1 3 +1 2 )(1 1 p ) 1 6 +σ], for all 1 < p, and where δ(t, x) L p (R d 1 ) C(1 + t) d 1 3 (1 1 p )+σ. 23

Further Work One direction of further study regards the consideration of additional stationary solutions. (Classified here by nodal structure.) 1. Periodic bands x 2 + + + x 1 + + + 24

Further Work 2. Dang, Fife, and Peletier have established the existence of saddle solutions to (CH) in d = 2 (ZAMP, 1992). Also, Junping Shi (CPAM, 2002). + + 25

Further Work 3. Holzmann and Kielhöfer have established the existence of doubly periodic solutions with rectangular nodal domains (Math. Annalen 1994). + + + + + + + + + + + + + + + + + + 26

Further Work 4. Fife, Kielhöfer, Maier Paape, and Wanner have established the existence of doubly periodic solutions with nodal domains that are not rectangular (Physica D, 1997). + + + + + + + + + + + + + + + + + + 27

The case d = 1 F (u) = 1 8 u4 1 4 u2 M(u) 1, κ = 1 Equilibrium Solutions 1 0.8 0.6 0.4 0.2 0 0.2 0.4 Periodic Solutions and Limiting Transition Wave 0.6 0.8 1 10 8 6 4 2 0 2 4 6 8 10 x Leading eigenvalues as a function of amplitude 0.07 u=0 corresponds with constant state 0.06 Leading Eigenvalue 0.05 0.04 0.03 0.02 0.01 u=1 corresponds with transition wave 28 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Amplitude

The Case d = 1: u t = (M(u)(F (u) κu xx ) x ) x Let u 1 and u 2 denote the binodal values, and let u 3 and u 4 satisfy u 1 u 3 < u 4 u 2. Then there exists a bounded stationary solution to (CH) satisfying inf ū(x) = u 3; sup ū(x) = u 4 x R x R if and only if there holds F (u 4 ) [F ] [u] F (u 3 ), where [u] = (u 4 u 3 ) and [F ] = F (u 4 ) F (u 3 ). Two strict inequalities periodic solutions (including constants) One equality reversal solutions ū(± ) = u ±, u = u + Two equalities transition waves ū(± ) = u ±, u u + 29

Stability Periodic solutions are spectrally unstable, and the leading eigenvalues decrease as the amplitudes increase. Reversal solutions are spectrally unstable in the following weak sense: the eigenvalue λ = 0 has multiplicity 2. Transition front solutions are orbitally stable. 30

Coarsening Rates The curve λ max (P ) can be approximated by Langer s large P limiting equation λ max (P ) = c e P P, and the leading eigenvalue associated with u h = 0 can be used to show c = πm(0) F (0) 3 4κ e2π q κ F (0). considered the case, and they approximated the double-well potential with a double Gaussian F (u) = ν2 2 { (u K) 2 u > 0 (u + K) 2 u < 0. 31

They showed that the leading eigenvalue of ˆL ξ satisfies λ (ξ) = M 0σ 2K 2 ξ 3, where σ = + ū (x 1 ) 2 dx 1. 32

The Dispersion Relation The essential spectrum of ˆL ξ satisfies: σ ess (ˆL ξ ) (, M(u ± )( ξ 2 F (u ± ) + κ ξ 4 ], with σ ess (ˆL 0 ) (, 0]. The eigenvalue λ (0) lies in the essential spectrum. Since the linear perturbation equation asymptotically has second order diffusion, we would like to have a quadradic scaling for λ ( ξ ). 33

Numerical Study of de Mello and Filho, 2005 u x (0) = u x (1) = 0 u xxx (0) = u xxx (1) = 0 1 U(t) (a=1, b=1, ε 2 =0.001, t=1/100 x=1/50) 0.5 u(x,t) 0-0.5-1 0 0.2 0.4 0.6 0.8 1 X t=0 t=5 t t=10 t t=50 t t=100 t t=500 t Fig. 2. The time evolution of the profile of U n i for different times starting at the initial condition (t ¼ 0) till the total spinodal decomposition is attained at n ¼ 500: The size of the system is L ¼ 1: 34

Coarsening Rates In the paper, Theory of Spinodal Decomposition in Alloys [Annals of Physics 65 (1971) 53 86], James Langer suggested the following method for obtaining coarsening rates: at each time t suppose that the solution to (CH), u(t, x), is near a stationary solution ū(x), and that the difference between u and ū is determined primarily by thermal fluctuations in the material. If we adapt Langer s analysis to the current setting, we find that if P (t) denotes the period of a periodic stationary solution at time t then dp dt = 1 2 λ max(p )P, where λ max (P ) denotes the leading eigenvalue associated with a stationary periodic solution with period P. 35

The Fundamental Questions Alloys undergoing spinodal decomposition and subsequent coarsening typically exhibit changes in characteristic properities (e.g., hardness typically increases), and one fundamental question in applications regards the time scale on which these changes occur. We would like to take a slight perturbation of the constant homogenously mixed state as an initial condition, and determine the amount of coarsening that has occurred for all subsequent times. 36