An asymptotic preserving and well-balanced scheme for a chemotaxis model

Σχετικά έγγραφα
Απόκριση σε Μοναδιαία Ωστική Δύναμη (Unit Impulse) Απόκριση σε Δυνάμεις Αυθαίρετα Μεταβαλλόμενες με το Χρόνο. Απόστολος Σ.

Fourier Series. MATH 211, Calculus II. J. Robert Buchanan. Spring Department of Mathematics

Jesse Maassen and Mark Lundstrom Purdue University November 25, 2013

Finite difference method for 2-D heat equation

Homework 3 Solutions

Ordinal Arithmetic: Addition, Multiplication, Exponentiation and Limit

Example Sheet 3 Solutions

D Alembert s Solution to the Wave Equation

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

Uniform Convergence of Fourier Series Michael Taylor

Homework 8 Model Solution Section

6.1. Dirac Equation. Hamiltonian. Dirac Eq.

Forced Pendulum Numerical approach


Phys460.nb Solution for the t-dependent Schrodinger s equation How did we find the solution? (not required)

SCITECH Volume 13, Issue 2 RESEARCH ORGANISATION Published online: March 29, 2018

Chapter 6: Systems of Linear Differential. be continuous functions on the interval

Solutions to Exercise Sheet 5

Graded Refractive-Index

Appendix A. Curvilinear coordinates. A.1 Lamé coefficients. Consider set of equations. ξ i = ξ i (x 1,x 2,x 3 ), i = 1,2,3

Approximation of distance between locations on earth given by latitude and longitude

1. (a) (5 points) Find the unit tangent and unit normal vectors T and N to the curve. r(t) = 3cost, 4t, 3sint

Lifting Entry (continued)

HOMEWORK 4 = G. In order to plot the stress versus the stretch we define a normalized stretch:

The Pohozaev identity for the fractional Laplacian

Matrices and Determinants

Discontinuous Hermite Collocation and Diagonally Implicit RK3 for a Brain Tumour Invasion Model

Problem Set 9 Solutions. θ + 1. θ 2 + cotθ ( ) sinθ e iφ is an eigenfunction of the ˆ L 2 operator. / θ 2. φ 2. sin 2 θ φ 2. ( ) = e iφ. = e iφ cosθ.

Other Test Constructions: Likelihood Ratio & Bayes Tests

ECE Spring Prof. David R. Jackson ECE Dept. Notes 2

ST5224: Advanced Statistical Theory II

Takeaki Yamazaki (Toyo Univ.) 山崎丈明 ( 東洋大学 ) Oct. 24, RIMS

Higher Derivative Gravity Theories

Local Approximation with Kernels

General 2 2 PT -Symmetric Matrices and Jordan Blocks 1

Section 8.3 Trigonometric Equations

12. Radon-Nikodym Theorem

Partial Differential Equations in Biology The boundary element method. March 26, 2013

Spherical Coordinates

b. Use the parametrization from (a) to compute the area of S a as S a ds. Be sure to substitute for ds!

Estimation for ARMA Processes with Stable Noise. Matt Calder & Richard A. Davis Colorado State University

The Spiral of Theodorus, Numerical Analysis, and Special Functions

From the finite to the transfinite: Λµ-terms and streams

Areas and Lengths in Polar Coordinates

ES440/ES911: CFD. Chapter 5. Solution of Linear Equation Systems

Math221: HW# 1 solutions

Practice Exam 2. Conceptual Questions. 1. State a Basic identity and then verify it. (a) Identity: Solution: One identity is csc(θ) = 1

4.6 Autoregressive Moving Average Model ARMA(1,1)

Homomorphism in Intuitionistic Fuzzy Automata

P AND P. P : actual probability. P : risk neutral probability. Realtionship: mutual absolute continuity P P. For example:

Arithmetical applications of lagrangian interpolation. Tanguy Rivoal. Institut Fourier CNRS and Université de Grenoble 1

Every set of first-order formulas is equivalent to an independent set

Introduction to the ML Estimation of ARMA processes

Second Order RLC Filters

Space-Time Symmetries

ANSWERSHEET (TOPIC = DIFFERENTIAL CALCULUS) COLLECTION #2. h 0 h h 0 h h 0 ( ) g k = g 0 + g 1 + g g 2009 =?

ΗΜΥ 220: ΣΗΜΑΤΑ ΚΑΙ ΣΥΣΤΗΜΑΤΑ Ι Ακαδημαϊκό έτος Εαρινό Εξάμηνο Κατ οίκον εργασία αρ. 2

Integrals in cylindrical, spherical coordinates (Sect. 15.7)

Solution Series 9. i=1 x i and i=1 x i.

Areas and Lengths in Polar Coordinates

1 String with massive end-points

5. Choice under Uncertainty

On the summability of divergent power series solutions for certain first-order linear PDEs Masaki HIBINO (Meijo University)

Numerical Analysis FMN011

상대론적고에너지중이온충돌에서 제트입자와관련된제동복사 박가영 인하대학교 윤진희교수님, 권민정교수님

Parametrized Surfaces

DuPont Suva 95 Refrigerant

derivation of the Laplacian from rectangular to spherical coordinates

DiracDelta. Notations. Primary definition. Specific values. General characteristics. Traditional name. Traditional notation

Durbin-Levinson recursive method

On the Galois Group of Linear Difference-Differential Equations

Solvability of Brinkman-Forchheimer equations of flow in double-diffusive convection

Exercises to Statistics of Material Fatigue No. 5

Statistical Inference I Locally most powerful tests

Derivation of Optical-Bloch Equations

EE512: Error Control Coding

( y) Partial Differential Equations

DuPont Suva. DuPont. Thermodynamic Properties of. Refrigerant (R-410A) Technical Information. refrigerants T-410A ENG

Solutions - Chapter 4

Inverse trigonometric functions & General Solution of Trigonometric Equations

k A = [k, k]( )[a 1, a 2 ] = [ka 1,ka 2 ] 4For the division of two intervals of confidence in R +

STABILITY FOR RAYLEIGH-BENARD CONVECTIVE SOLUTIONS OF THE BOLTZMANN EQUATION

A Two-Sided Laplace Inversion Algorithm with Computable Error Bounds and Its Applications in Financial Engineering

MATH423 String Theory Solutions 4. = 0 τ = f(s). (1) dτ ds = dxµ dτ f (s) (2) dτ 2 [f (s)] 2 + dxµ. dτ f (s) (3)

DuPont Suva 95 Refrigerant

The k-α-exponential Function

Section 7.6 Double and Half Angle Formulas

Lecture 34 Bootstrap confidence intervals

Acoustic Limit for the Boltzmann equation in Optimal Scaling

Econ 2110: Fall 2008 Suggested Solutions to Problem Set 8 questions or comments to Dan Fetter 1

2 Composition. Invertible Mappings

Prey-Taxis Holling-Tanner

CHAPTER 25 SOLVING EQUATIONS BY ITERATIVE METHODS

10.7 Performance of Second-Order System (Unit Step Response)

6.3 Forecasting ARMA processes

Second Order Partial Differential Equations

Sequent Calculi for the Modal µ-calculus over S5. Luca Alberucci, University of Berne. Logic Colloquium Berne, July 4th 2008

Technical Information T-9100 SI. Suva. refrigerants. Thermodynamic Properties of. Suva Refrigerant [R-410A (50/50)]

Lecture 21: Properties and robustness of LSE

Lifting Entry 2. Basic planar dynamics of motion, again Yet another equilibrium glide Hypersonic phugoid motion MARYLAND U N I V E R S I T Y O F

Transcript:

An asymptotic preserving and well-balanced scheme for a chemotaxis model Christophe Berthon jointed work with Marianne Bessemoulin, University of Nantes (France) Anaïs Crestetto, University of Nantes (France) Françoise Foucher, University of Nantes (France) Hélène Mathis, University of Nantes (France) Madison, May 25

Main motivations Hyperbolic systems of conservation law with source term t w + x f(w) = ν S(w) w Ω Steady states x f(w) = ν S(w) Manifold given by Asymptotic diffusive regime: ν Finite volume schemes Numerical approximations of the weak solutions Robustness Exact capture of (a part of) M Asymptotic preserving M = {w Ω; g(w) = }

Outline A chemotaxis model (with A. Crestetto and F. Foucher) Godunov type well-balanced scheme Model and main properties Godunov type scheme Characterization of the approximate Riemann solver Robustness and well-balanced properties Asymptotic preserving property Asymptotic convergence rate (with M. Bessemoulin and H. Mathis) relative entropies to estimate the asymptotic convergence rate A continuous estimation by C. Lattanzio and A. Tzavaras Extension to a semi-discrete scheme

A model of Chemotaxis (Ribot et al 22-24) t ρ + x ρu = t ρu + x (ρu 2 + p(ρ)) = χρ x Φ αρu t Φ D xx Φ = aρ bφ Steady states at rest u = e(ρ) χφ = cste solutions of D xx Φ bφ = aρ if ρ = : if ρ >, C < : if ρ >, C > : e(ρ) = δ χ, α, D, a, b given parameters Ω = {w R 3 ; ρ, u R, Φ } p(ρ) = δρ γ γ γ ργ + cste ( ϕ (x) = A cosh x ) ( b + B sinh x ) b, ( ϕ (x) = A cos x ) ( C + B sin x ) C ϕ p, ρ (x) = χ (ϕ (x) K), ( 2δ ϕ (x) = A cosh x ) ( C + B sinh x ) C ϕ p, ρ (x) = χ (ϕ (x) K), 2δ where A and B are some constants, C = D ( ) b aχ 2δ and ϕp = Kaχ 2δb aχ.

Asymptotic behavior Rescaling: t t/ν (long time) and α α/ν (dominant friction) Limit ν to get a diffusive regime { t ρ = x ( x p χρ x Φ) D xx Φ = bφ aρ Objectives Robustness (ρ and Φ ) Steady state preserving (well-balanced) Asymptotic preserving Godunov type strategy (CB and Chalons 25)

Godunov type scheme { t ρ + x ρu = t ρu + x (ρu 2 + p(ρ)) = χρ x Φ αρu Φ given Approximate Riemann solver: w λ L λ R w L w L w R w R λ L < < λ R HLL type solver Source term stationary contact wave x Harten-Lax-van Leer consistency condition /2 /2 w (x, t; w L, w R ) dx = /2 /2 w R (x, t; w L, w R ) dx /2 /2 w (x, t; w L, w R ) dx = 2 (w L+w R )+ t (λ L(w L w L)+λ R (w R w R ))

Because of the source term, w R stays unknown w L (t) w L t w R (x, t; w L, w R ) w R w R (t) x constant is not a natural solution /2 /2 t ( ) t w + x f(w) = S(w) dxdt /2 /2 Approximation w R (x, t; w L, w R ) dx = 2 (w L + w R ) + t f(w R (/2, t))dt + w R ( /2, t) w L w R (/2, t) w R t /2 /2 t f(w R ( /2, t))dt S(w)dxdt

As a consequence /2 /2 /2 /2 S R = t ρ R (x, t; w L, w R ) dx 2 (ρ L + ρ R ) t (ρ Ru R ρ L u L ) (ρu) R (x, t; w L, w R ) dx 2 (ρ Lu L + ρ R u R ) t (ρ Ru 2 R + p R ρ L u 2 L p L ) + ts R α t /2 /2 t /2 /2 χρ R x Φdxdt (ρu) R (x, t; w L, w R ) dxdt Approximation S R S to be defined (independently from t) Approximation /2 /2 solution of the following integral equation (ρu) R (x, t; w L, w R ) dx F( t) F( t) = 2 (ρ Lu L +ρ R u R ) t (ρ Ru 2 R +p R ρ L u 2 L p L )+ ts α t F(t)dt

Consistency conditions λ L (ρ L ρ L) + λ R (ρ R ρ R ) = ρ L u L ρ R u R λ L (ρ L u L ρ Lu L) + λ R (ρ Ru R ρ R u R ) = α t ( e α t ) ( α 2 (ρ Lu L + ρ R u R ) (ρ R u 2 R + p R ρ L u 2 L p L ) + S ) Flux continuity: ρ L u L = ρ R u R Well-balanced conditions S = χ p R p L e R e L (Φ R Φ L ) e L ρ L ρ L χϕ L = e R ρ R ρ R χϕ R

Approximate Riemann solver: w λ L λ R Φ L Φ R Φ L Φ R x { t Φ + x Ψ = aρ bφ Ψ = x Φ ρ given Harten-Lax-van Leer consistency condition /2 /2 Φ (x, t; w L, w R ) dx = /2 /2 Φ R (x, t; w L, w R ) dx Main difficulty: Evaluate Φ R /2 t /2 ( t Φ D x Ψ) dxdt = a /2 /2 t ρ R (x, t)dxdt b Φ R (x, t)dxdt Approximations /2 /2 /2 /2 ρ R (x, t)dx 2 (ρ L + ρ R ) t x Ψ(x, t)dx t (Ψ R Ψ L ) t (ρ Ru R ρ L u L )

Approximate Riemann solver: w Godunov type scheme λ L λ R Φ L Φ R Φ L Φ R x { t Φ + x Ψ = aρ bφ Ψ = x Φ ρ given Harten-Lax-van Leer consistency condition /2 /2 Φ R (x, t)dx G( t) G( t) solution of the following integral equation G( t) = 2 (Φ L + Φ R ) + D t ( ) t (Ψ R Ψ L ) + a 2 (ρ L + ρ R ) t2 2 (ρ Ru R ρ L u L ) b t G(t)dt

Godunov type scheme ( w n+ i Φ n+ i = w n i t = Φ n i t Definition of Ψ n i We impose ( ) ( )) λ i 2,R wi n w i 2,R λ i+ 2,L wi n ( ( ) ( w i+ 2,L )) λ i 2,R Φ n i Φ i 2,R λ i+ 2,L Φ n i Φ i+ 2,L Ψ n i = 2 (Φn i+ Φ n i ) E() E() is consitent with lim E() = E() = 2 b ( 2 b ) cos 2 C ( C ) 2 cos if ρ = if ρ >, C < 2 2 C ( C ) cosh if ρ >, C > to recover the steady states

Theorem Robustness (adopting a local cut-off, Chalons et al 24) ρ L,R = min(max(, ρ L,R), 2ρ HLL ) Φ L,R = min(max(, Φ L,R), 2Φ HLL ) ρ n+ i and Φ n+ i Well-balance property

Asymptotic diffusive regime Rescalling u = νu ν, t = t ν /ν and α = α ν /ν, Rescalled system given by ν t ν ρ ν + ν x (ρ ν u ν ) =, ν 2 t ν (ρ ν u ν ( ) + x ν 2 ρ ν (u ν ) 2 + p(ρ ν ) ) = χρ ν x ϕ ν α ν ρ ν u ν, ν t ν ϕ ν D xx ϕ ν = aρ ν bϕ ν, Chapman-Enskog expansion ρ ν = ρ + O(ν) ρ ν u ν = ρ u + O(ν) ϕ ν = ϕ + O(ν). Zero-order governed by ( χ t ν ρ + x α ν ρ x ϕ ) α ν xp(ρ ) D xx ϕ = bϕ aρ, ρ u = α ν xp(ρ ) + χ α ν ρ x ϕ. =,

w n+ i Φ n+ i Rescalling = w n i t = Φ n i t Asymptotic preserving scheme ( ( ) λ i 2,R wi n w i 2,R λ i+ ( )) 2,L wi n ( ( ) ( w i+ 2,L )) λ i 2,R Φ n i Φ i 2,R λ i+ 2,L Φ n i Φ i+ 2,L (u n ) = ν(u n i ) ν, t = t ν /ν and α = α ν /ν, Simplification λ R = λ L = λ λ u + p (ρ) = O() CFL restriction t ν max ( ) λi+/2 2,

In the limite ν ρ n+ i (ρu) n i =ρ n i t e (ρ n i ) ρ n i ( ( l i /2 (ρu) n i (ρu) n ) ( i + li+/2 (ρu) n i+ (ρu) n i )) ) ( ( ) + λc CF L (l i+/2 e ρ n i+ e (ρ n i ) ) ( l i+/2 e (ρ n i ) e ( ρ n i ( ) + λc CF L χ l i+/2 (ϕ n i+ ϕ n i ) l i /2 (ϕ n i ϕ n i ) = ( p(ρn i+ ) p(ρn i ) + χ ) α 2 2 ({ρ xϕ} n i /2 + {ρ xϕ} n i /2 ) + ( λc CF L ) ((ρu) n i+ 2(ρu) n i + (ρu) n i 2 2 )) ) ( + (ρu) n i (ρu) n+ ) i D ( xϕ) n i+ ( xϕ) n i 2 where we have set =b ϕn i + 2ϕn i + ϕn i+ 4 l i+/2 = a ρn i + 2ρn i + ρn i+ 4 e (ρ n i ) /ρn i + e ( ρ n i+) /ρ n i+ + (ϕ n+ i ϕ n i ) The scheme is consitent with the asymptotic regime

Numerical results Influence of γ at χ =, L = 3, =.. 3 =2 8 =3 7 25 6 2 5 5 4 3 2 5.5.5 2 2.5 3 x.5.5 2 2.5 3 x 3.5 =4.8 =5 3.6.4 2.5.2 2.5.8.6.4.5.2.5.5 2 2.5 3 x.5.5 2 2.5 3 x

Convergence rate as ν Simpler model: p-system { t τ x u = t u + x p(τ) = σu Asymptotic regime Relative entropy Rescalling t t/ϵ u ϵu ϵ α α/ϵ t τ x ū = t p( τ) = σ ū { t τ ϵ x u ϵ = ϵ 2 t u ϵ + x p(τ ϵ ) = σu ϵ η ε (τ, u τ, u) = ε2 2 (u u)2 P (τ τ) with P (τ τ) = P (τ) P (τ) p(τ)(τ τ) To satisfy t η ε (τ ε, u ε τ, u)+ ε 2 xψ(τ ε, u ε τ, u) = σ(u ε u) 2 + σ xxp(τ) p(τ ε τ)+ ε2 σ xtp(τ)(u ε u)

Lattanzio and Tzavaras estimation Introduce ϕ(t) = η(τ ε, u ε τ, u)dx Assume τ c > xx p(τ) L (Q T ) K < + xt p(τ) L 2 (Q T ) K < + R Then ϕ(t) C(ϕ() + ε 4 ) t [, T ) Objective: Recover this estimation with numerical approximations

Semi-discrete scheme p-system: semi-discrete scheme (τ i (t), u i (t)) dτ i dt = 2 (u i+ u i ) + du i dt = 2ε 2 (p(τ i+) p(τ i )) σ ε 2 u i + λ 2 (τ i+ 2τ i + τ i ) asymptotic regime: semi-discrete scheme (τ i (t), u i (t)) dτ i dt = u i = σ 2 (u i+ u i ) + 2 (p(τ i+) p(τ i )) λ 2 (u i+ 2u i + u i ) λ 2 (τ i+ 2τ i + τ i )

Relative entropy: η i = η(τ i, u i τ i, u i ) dη i dt + ψ i+/2 ψ i /2 ε 2 = σ(u i u i ) 2 + p(τ i+2 ) 2p(τ i ) + p(τ i 2 ) σ (2) 2 p(τ i τ i )+ ε 2 ( ) d p(τ i+ ) p(τ i ) (u i u i ) + R τ i + R u i σ dt 2 R τ i = ε 2 λ 2 (u i+ 2u i + u i )(u i u i ) R u i = λ 2 ((p(τ i) p(τ i ))(τ i+ 2τ i + τ i ) + p (τ i )(τ i+ 2τ i + τ i )(τ i τ i )) By summation: ϕ(t) = η i(t) t ϕ(t) ϕ() = σ + ε2 σ t t (u i u i ) 2 ds + σ ( ) d p(τ i+ ) p(τ i ) (u i u i )ds + dt 2 p(τ i+2 ) 2p(τ i ) + p(τ i 2 ) (2) 2 t p(τ i τ i )ds (R τ i + R u i )ds

Sketch of the proof of the convergence rate Estimations of Lattanzio and Tzavaras σ ε 2 σ t p(τ i+2 ) 2p(τ i ) + p(τ i 2 ) (2) 2 p(τ i τ i )ds C σ D xxp(τ) ( ) d p(τ i+ ) p(τ i ) (u i u i )ds dt 2 t σ 2 t t ϕ(s)ds (u i u i ) 2 ds + C D xt p(τ) 2 ε 4 Estimation of the viscous terms t R u i ds λθ t (u i u i ) 2 ds + C(θ, D xx u 2 )ε 4 2 t t R τ i ds C( D x τ, D xx τ ) ϕ(s)ds

Theorem: Assume τ c > and ( T ( ) D tx p(τ) L 2 := d p(τ i+ ) p(τ i ) 2 /2 ds) K dt 2 D xx p(τ) L := sup sup p(τ i+2 ) 2p(τ i ) + p(τ i 2 ) t [,T ] (2) 2 K ( T ( ) ) 2 /2 τ i+ 2τ i + τ i D xx τ L := 2 ds K Then D x τ L := sup sup t [,T ] ( T D xx u L 2 := to get the convergence rate τ i+ τ i K ( ) 2 ui+ 2u i + u i 2 ds ϕ(t) ϕ() + Cε 4 + C t ϕ(s)ds ) /2 K ϕ(t) ( ϕ() + Cε 4) e CT t T

Jin-Pareschi-Toscani scheme (98) Refomulation t τ x u = Splitting scheme τ n+ i = τ n+ 2 i u n+ i u n+ 2 i t = ε 2 t u + x p(τ) = ( σ u + ( ε 2 ε 2 ) x p(τ) ) Expected result τ n+ 2 i = τ n i t u n+ 2 i = u n i t σ u n+ i + ( ε 2 ) ( G τ i+ 2 ( G u i+ 2 p(τ n+ i+ 2 ϕ n = η(τ n i, u n i τ n i, u n i ) ) G τ i 2 ) G u i 2 ) p(τ n+ i 2 ) τ n+ i+ 2 = ( τ n+ i + τ n+ ) i+ 2 ( ϕ + Cε 4) e CN n N

Numerical experiments.. Phi(eps) eps^4.. Phi(eps) eps^4 e-6 e-6 e-8 e-8 e- e- e-2 e-2 e-4 e-4 eps 2 si x < τ (x) = si x > e-6.... u (x) = δ e-6.... eps τ (x) = exp( x 2 ) + u (x) = x τ (x) Number of cells: 8 Final time of simulations: T = 2

Thanks for your attention