Lab session: numerical simulations of sponateous polarization

Similar documents
Appendix B. The Finite Difference Scheme

2 Finite difference basics

High resolution entropy stable scheme for shallow water equations

Additional Codes using Finite Difference Method. 1 HJB Equation for Consumption-Saving Problem Without Uncertainty

Linear, affine, and convex sets and hulls In the sequel, unless otherwise specified, X will denote a real vector space.

FTCS Solution to the Heat Equation

Lecture 21: Numerical methods for pricing American type derivatives

Lecture 10 Support Vector Machines II

MATH 5630: Discrete Time-Space Model Hung Phan, UMass Lowell March 1, 2018

NON-CENTRAL 7-POINT FORMULA IN THE METHOD OF LINES FOR PARABOLIC AND BURGERS' EQUATIONS

Chapter 4 The Wave Equation

PHYS 705: Classical Mechanics. Calculus of Variations II

Numerical Transient Heat Conduction Experiment

Formal solvers of the RT equation

APPENDIX A Some Linear Algebra

Numerical Heat and Mass Transfer

Chapter 12. Ordinary Differential Equation Boundary Value (BV) Problems

SUPPLEMENTARY INFORMATION

Module 1 : The equation of continuity. Lecture 1: Equation of Continuity

4DVAR, according to the name, is a four-dimensional variational method.

Week3, Chapter 4. Position and Displacement. Motion in Two Dimensions. Instantaneous Velocity. Average Velocity

Difference Equations

Modified Mass Matrices and Positivity Preservation for Hyperbolic and Parabolic PDEs

n α j x j = 0 j=1 has a nontrivial solution. Here A is the n k matrix whose jth column is the vector for all t j=0

Queens College, CUNY, Department of Computer Science Numerical Methods CSCI 361 / 761 Spring 2018 Instructor: Dr. Sateesh Mane.

Gaussian Mixture Models

Chapter 4: Root Finding

CHAPTER 5 NUMERICAL EVALUATION OF DYNAMIC RESPONSE

Introduction to Vapor/Liquid Equilibrium, part 2. Raoult s Law:

Problem adapted reduced models based on Reaction-Diffusion Manifolds (REDIMs)

Parametric fractional imputation for missing data analysis. Jae Kwang Kim Survey Working Group Seminar March 29, 2010

Lectures - Week 4 Matrix norms, Conditioning, Vector Spaces, Linear Independence, Spanning sets and Basis, Null space and Range of a Matrix

Maximum Likelihood Estimation of Binary Dependent Variables Models: Probit and Logit. 1. General Formulation of Binary Dependent Variables Models

Limited Dependent Variables

CME 302: NUMERICAL LINEAR ALGEBRA FALL 2005/06 LECTURE 13

( ) [ ( k) ( k) ( x) ( ) ( ) ( ) [ ] ξ [ ] [ ] [ ] ( )( ) i ( ) ( )( ) 2! ( ) = ( ) 3 Interpolation. Polynomial Approximation.

(2mn, m 2 n 2, m 2 + n 2 )

Modelli Clamfim Equazioni differenziali 22 settembre 2016

CALCULUS CLASSROOM CAPSULES

THE VIBRATIONS OF MOLECULES II THE CARBON DIOXIDE MOLECULE Student Instructions

8.592J: Solutions for Assignment 7 Spring 2005

The Finite Element Method

Differentiating Gaussian Processes

3.1 Expectation of Functions of Several Random Variables. )' be a k-dimensional discrete or continuous random vector, with joint PMF p (, E X E X1 E X

DUE: WEDS FEB 21ST 2018

Modelli Clamfim Equazione del Calore Lezione ottobre 2014

Canonical transformations

Lecture 5.8 Flux Vector Splitting

Linear Approximation with Regularization and Moving Least Squares

Math1110 (Spring 2009) Prelim 3 - Solutions

Inner Product. Euclidean Space. Orthonormal Basis. Orthogonal

Chapter 5. Solution of System of Linear Equations. Module No. 6. Solution of Inconsistent and Ill Conditioned Systems

Lecture 17 : Stochastic Processes II

Convex Optimization. Optimality conditions. (EE227BT: UC Berkeley) Lecture 9 (Optimality; Conic duality) 9/25/14. Laurent El Ghaoui.

Marginal Effects in Probit Models: Interpretation and Testing. 1. Interpreting Probit Coefficients

New Method for Solving Poisson Equation. on Irregular Domains

Errors for Linear Systems

Normally, in one phase reservoir simulation we would deal with one of the following fluid systems:

APPROXIMATE PRICES OF BASKET AND ASIAN OPTIONS DUPONT OLIVIER. Premia 14

1-Dimensional Advection-Diffusion Finite Difference Model Due to a Flow under Propagating Solitary Wave

Salmon: Lectures on partial differential equations. Consider the general linear, second-order PDE in the form. ,x 2

Modelli Clamfim Equazioni differenziali 7 ottobre 2013

MMA and GCMMA two methods for nonlinear optimization

Homogenization of reaction-diffusion processes in a two-component porous medium with a non-linear flux-condition on the interface

Primer on High-Order Moment Estimators

COS 521: Advanced Algorithms Game Theory and Linear Programming

A Computational Viewpoint on Classical Density Functional Theory

PART 8. Partial Differential Equations PDEs

REAL ANALYSIS I HOMEWORK 1

Solution of the Navier-Stokes Equations

Computational Astrophysics

Beyond Zudilin s Conjectured q-analog of Schmidt s problem

Amplification and Relaxation of Electron Spin Polarization in Semiconductor Devices

Airflow and Contaminant Simulation with CONTAM

LECTURE 9 CANONICAL CORRELATION ANALYSIS

Complete subgraphs in multipartite graphs

The equation of motion of a dynamical system is given by a set of differential equations. That is (1)

Nice plotting of proteins II

Inductance Calculation for Conductors of Arbitrary Shape

Lecture 12: Discrete Laplacian

Physics 207: Lecture 20. Today s Agenda Homework for Monday

1 The Sidrauski model

8 Derivation of Network Rate Equations from Single- Cell Conductance Equations

A new Approach for Solving Linear Ordinary Differential Equations

Feature Selection: Part 1

Bézier curves. Michael S. Floater. September 10, These notes provide an introduction to Bézier curves. i=0

(Online First)A Lattice Boltzmann Scheme for Diffusion Equation in Spherical Coordinate

Transfer Functions. Convenient representation of a linear, dynamic model. A transfer function (TF) relates one input and one output: ( ) system

coordinates. Then, the position vectors are described by

Lecture 2: Numerical Methods for Differentiations and Integrations

C4B Machine Learning Answers II. = σ(z) (1 σ(z)) 1 1 e z. e z = σ(1 σ) (1 + e z )

Numerical Simulation of One-Dimensional Wave Equation by Non-Polynomial Quintic Spline

Maximum Likelihood Estimation of Binary Dependent Variables Models: Probit and Logit. 1. General Formulation of Binary Dependent Variables Models

Lecture 12. Modeling of Turbulent Combustion

Finding The Rightmost Eigenvalues of Large Sparse Non-Symmetric Parameterized Eigenvalue Problem

Computational Geodynamics: Advection equations and the art of numerical modeling

Application of B-Spline to Numerical Solution of a System of Singularly Perturbed Problems

The Multiple Classical Linear Regression Model (CLRM): Specification and Assumptions. 1. Introduction

1. Introduction. The present work is devoted to the numerical approximations of weak solutions of the Euler equations: (1.1)

A Kinetic Model for the formation of Swarms with nonlinear interactions

Transcription:

Lab sesson: numercal smulatons of sponateous polarzaton Emerc Boun & Vncent Calvez CNRS, ENS Lyon, France CIMPA, Hammamet, March 2012

Spontaneous cell polarzaton: the 1D case The Hawkns-Voturez model for spontaneous polarzaton n 1D reads as follows: t ρ(t, x) = xx ρ(t, x) + ρ(t, 0) x ρ(t, x), t > 0, x (0, + ). The equaton s a transport-dffuson equaton. Numercal ssues: blow-up, non-trval steady states, self-smlar decay.

Behavour of solutons Theorem There s a nce and smple dchotomy: If M < 1, the soluton s global n tme. It converges towards a (unque) self-smlar profle G M : lm ρ(t, x) 1 ( ) x t L G M = 0. t 1 t + If M = 1 there s a one-parameter famly of steady states: ν α (x) = α exp( αx). Convergence holds + the frst moment s conserved: α 1 = x>0 ρ0 (x) dx. If M > 1 and x > 0 x ρ 0 (x) 0, the soluton blows up n fnte tme.

Contents 1- Dffuson equaton 2- The dffuson-transport equaton 3- The dffuson-transport equaton n self-smlar varables 4- Cluster formaton

Contents 1- Dffuson equaton 2- The dffuson-transport equaton 3- The dffuson-transport equaton n self-smlar varables 4- Cluster formaton

The dffuson equaton n the half-lne We begn wth the dffuson equaton t ρ(t, x) = xx ρ(t, x), x > 0 + Neumann boundary condton at x = 0 We rewrte the equaton n dvergence form, t ρ(t, x) + x F = 0, The flux F s gven by Fck s law: F = x ρ. The boundary condton at {x = 0} s F (0) = 0 (no-flux boundary condton).

Tme-Space dscretzaton We dscretze the functon ρ(t, x) on a regular grd [0 : t : T ] [0 : x : L], t s the tme step, T s the fnal tme of computaton, x s the space step, L s the length of the numercal nterval.

Strategy for computng ρ(t, x) numercally We replace the equaton t ρ(t, x) + x F = 0 wth the numercal dscretzaton ρ(t + t, x) ρ(t, x) t We ntroduce + F (t, x + x/2) F (t, x x/2) x ρ n = ρ(n t, x) F n = F (n t, x ± x/2). ± 1 2 The equaton (1) rewrtes at (t, x) = (t n, x ), = 0. (1) ρ n t + F n F n + 1 1 2 2 x = 0

Boundary condtons The boundary condton F (0) = 0 reads F 1 2 For = 1 the equaton (1) reads We have smlarly for = Nx = 0. 1 ρ n F n 0 1 1+ + 1 2 = 0 t x Nx ρ n Nx t + 0 F n Nx 1 2 x = 0

We have F = x ρ, therefore The dffuson equaton ρ(t, x + x) ρ(t, x) F (t, x + x/2) = x The scheme wrtes fnally ρ n t + 1 ( ρn +1 ρn x x + ρn ρ n ) 1 = 0, x Remark. It concdes wth the classcal fnte dfference scheme for the heat equaton ( = 1 2 t ) x 2 ρ n + t x 2 ρn +1 + t x 2 ρn 1,

Explct scheme for the dffuson equaton Left boundary condton (zero-flux) when = 1 ρ n t + 1 ( ρn +1 ρn x x ) + 0 = 0, Rght boundary condton (zero-flux) when = Nx ρ n t + 1 ( 0 + ρn ρ n ) 1 = 0, x x Flux of the densty when 2 Nx 1 ρ n t + 1 ( ρn +1 ρn x x + ρn ρ n ) 1 = 0, x

Explct scheme for the dffuson equaton Left boundary condton (zero-flux) when = 1 ρ n t + 1 ( ρn +1 ρn x x ) + 0 = 0, Rght boundary condton (zero-flux) when = Nx ρ n t + 1 ( 0 + ρn ρ n ) 1 = 0, x x Flux of the densty when 2 Nx 1 ρ n t + 1 ( ρn +1 ρn x x + ρn ρ n ) 1 = 0, x

Explct scheme for the dffuson equaton Left boundary condton (zero-flux) when = 1 ρ n t + 1 ( ρn +1 ρn x x ) + 0 = 0, Rght boundary condton (zero-flux) when = Nx ρ n t + 1 ( 0 + ρn ρ n ) 1 = 0, x x Flux of the densty when 2 Nx 1 ρ n t + 1 ( ρn +1 ρn x x + ρn ρ n ) 1 = 0, x

Structure of the sclab code 1/4 Space and tme grd of resoluton. %% Space dscretsaton L = 10; dx = 0.1; x = [0:dx:L]; Nx = length(x); %% Tme dscretsaton T = 10; dt = dx^2/4; t = [0:dt:T]; Nt = length(t);

Structure of the sclab code 2/4 Intal data ρ 0 : we start wth a gaussan %% Intal condton for the densty of molecules rho_0 sgma = 1; rho0 = exp(-x.^2/sgma^2); Z = sum(rho0*dx); rho0 = (M/Z)*rho0; rho1 = rho0;

Structure of the sclab code 3/4 Tme loop to update the value of the densty ρ at each tme step n for n = 1:Nt rho0 = rho1;... end

Structure of the sclab code 4/4 Space loop to update the value of the densty ρ at each pont of the grd for n = 1:Nt end end rho0 = rho1; for = 1:Nx f == 1 %% Left boundary condton rho1() =... elsef == Nx %% Rght boundary condton rho1() =... else rho1() =... end

The maxmum prncple Recall the numercal scheme for the dffuson equaton ( = 1 2 t ) x 2 ρ n + t x 2 ρn +1 + t x 2 ρn 1, Observaton. s a lnear combnaton of ρ n 1, ρn and ρ n +1. In order to guarantee the maxmum prncple, t s necessary to mpose a condton between t and x : 2 t x 2 < 1. CFL condton. In ths case, s a convex combnaton of ρ n 1, ρn and ρ n +1. Therefore, = 1... Nx mn ρ 0 j ρ n max j j ρ 0 j

Conservaton of mass The conservaton of mass s automatc when the numercal scheme s wrtten wth the flux formulaton F n F n + 1 1 2 2 ρ n + t x To see ths, smply sum up the equaton: 1 t Nx =1 x 1 t Nx =1 = 0 ρ n x + F n Nx+ 1 2 F n 1 2 = 0 The no-flux boundary condton F n 1 = F n Nx+ 1 2 2 x = ρ n x = = 0 guarantees ρ 0 x.

Contents 1- Dffuson equaton 2- The dffuson-transport equaton 3- The dffuson-transport equaton n self-smlar varables 4- Cluster formaton

The dffuson-transport equaton We swtch to the one-dmensonal polarzaton equaton. t ρ(t, x) x (µ(t)ρ(t, x)) = xx ρ(t, x) + Neumann boundary condton at x = 0 + nonlnear couplng va µ(t) = ρ(t, 0). Agan, we rewrte the equaton n dvergence form, t ρ(t, x) + x F = 0, The flux F s gven by: F = x ρ µρ. The boundary condton at {x = 0} s F (0) = 0 (no-flux boundary condton). We assume that the speed µ(t) s nonnegatve µ(t) 0.

Strategy for computng ρ(t, x) numercally We replace the equaton t ρ(t, x) + x F = 0 wth the numercal dscretzaton ρ n F n F n + + 1 1 2 2 = 0 t x F n + 1 2 = ρn +1 ρn x }{{} Fck slaw + µ n ρ n. + 1 2 }{{} transport Important queston. How to nterpolate the value ρ n + 1 2 = ρ(t, x + x/2)?

The correct choce for ρ n + 1 2 Recall the key assumpton µ n 0. Choce 1. ρ n + 1 2 Choce 2. ρ n + 1 2 Choce 3. ρ n + 1 2 = ρ n = = for transport only ( 1 + µ n t ) ρ n µ n t x x ρn 1, = ρ n +1 [Upwnd scheme] ( 1 µ n t x ) ρ n + µ n t x ρn +1, = 1 2 (ρn + ρ n +1 ) [Centered scheme] = ρ n + µ n t 2 x ρn +1 µ n t 2 x ρn 1, Only the second choce preserves the maxmum prncple, under the condton µ n t x < 1.

Combnaton of the two fluxes We eventually get ρ n t + 1 ( ρn +1 ρn x x µ n ρ n +1 + ρn ρ n 1 x ) + µ n ρ n = 0, wth sutable boundary condtons for = 1 and = Nx.

Contents 1- Dffuson equaton 2- The dffuson-transport equaton 3- The dffuson-transport equaton n self-smlar varables 4- Cluster formaton

Self-smlar rescalng Rewrte the densty ρ n self-smlar varables, ( 1 ρ(t, x) = u log ) x 1 + 2t, 1 + 2t 1 + 2t The equaton for u(τ, y) s the followng, ) τ u(τ, y) y (yu(τ, y)+v(τ)u(τ, y) = yy u(τ, y), v(τ) = u(τ, 0). We get an addtonal drft term y (yu(τ, y)) due to the change of frame (t, x) (τ, y).

Long-tme asymptotcs In the sub-crtcal regme, M < 1, the soluton u(τ, y) converges towards a unque statonary state G, gven by G(y) = α exp ( αy y 2 ). 2 The parameter α s fxed by the conservaton of mass G(y) dy = u(τ, y) dy = ρ(t, x) dx = M. y>0 y>0 x>0 Exercse. Perform the numercal smulatons for u(τ, y), soluton to the polarzaton equaton n self-smlar varables (the only dfference s the addtonal drft term). Compare the long-tme behavour wth the expected value for G.

Contents 1- Dffuson equaton 2- The dffuson-transport equaton 3- The dffuson-transport equaton n self-smlar varables 4- Cluster formaton

A one-dmensonal model for cluster formaton The smplest model for bacteral chemotaxs s the followng coupled system for the cell densty ρ(t, x) and the chemcal concentraton S(t, x). { t ρ(t, x) + x (ρ(t, x) u(t, x)) = xx ρ(t, x), x R xx S(t, x) = ρ(t, x) u(t, x) = vφ (v x S(t, x)) dv v V For smplcty we set V = ( 1, 1). The functon φ depends on ndvdual features of the bactera (e.g. the way they react to the chemcal sgnal).

Cluster formaton Exercse. Perform the numercal smulatons for the couple (ρ, S) for dfferent choces of functon φ: φ(y ) = Y,.e. the Keller-Segel model, φ(y ) = sgn (Y ),.e. a step functon: t s a good model for chemotaxs n bactera populatons.

References V. Calvez, N. Meuner and R. Voturez, C. R. Math. Acad. Sc. Pars (2010) J. Saragost, V. Calvez, N. Bournaveas, A. Bugun, P. Slberzan, B. Perthame, PLoS Comput Bol (2010)