Generalized Jacobi spectral-galerkin method for nonlinear Volterra integral equations with weakly singular

Similar documents
A HYBRID SPECTRAL ELEMENT METHOD FOR FRACTIONAL TWO-POINT BOUNDARY VALUE PROBLEMS

Available online at J. Math. Comput. Sci. 4 (2014), No. 4, ISSN:

Homotopy Analysis Method for Solving Fractional Sturm-Liouville Problems

Comparison between Fourier and Corrected Fourier Series Methods

Approximating Solutions for Ginzburg Landau Equation by HPM and ADM

Research Article A Generalized Nonlinear Sum-Difference Inequality of Product Form

APPROXIMATE SOLUTION OF FRACTIONAL DIFFERENTIAL EQUATIONS WITH UNCERTAINTY

Compact Finite Difference Schemes for Solving a Class of Weakly- Singular Partial Integro-differential Equations

The Solution of the One Species Lotka-Volterra Equation Using Variational Iteration Method ABSTRACT INTRODUCTION

Numerical Solution of Parabolic Volterra Integro-Differential Equations via Backward-Euler Scheme

Extended Laguerre Polynomials

A Complex Neural Network Algorithm for Computing the Largest Real Part Eigenvalue and the corresponding Eigenvector of a Real Matrix

K3 p K2 p Kp 0 p 2 p 3 p

1. Solve by the method of undetermined coefficients and by the method of variation of parameters. (4)

ODEs II, Supplement to Lectures 6 & 7: The Jordan Normal Form: Solving Autonomous, Homogeneous Linear Systems. April 2, 2003

Mean Square Convergent Finite Difference Scheme for Stochastic Parabolic PDEs

STK4080/9080 Survival and event history analysis

A TAUBERIAN THEOREM FOR THE WEIGHTED MEAN METHOD OF SUMMABILITY

A Note on Random k-sat for Moderately Growing k

Ideal Amplifier/Attenuator. Memoryless. where k is some real constant. Integrator. System with memory

David Randall. ( )e ikx. k = u x,t. u( x,t)e ikx dx L. x L /2. Recall that the proof of (1) and (2) involves use of the orthogonality condition.

FIXED FUZZY POINT THEOREMS IN FUZZY METRIC SPACE

Supplement for SADAGRAD: Strongly Adaptive Stochastic Gradient Methods"

1 Notes on Little s Law (l = λw)

Fermat Numbers in Multinomial Coefficients

TAKA KUSANO. laculty of Science Hrosh tlnlersty 1982) (n-l) + + Pn(t)x 0, (n-l) + + Pn(t)Y f(t,y), XR R are continuous functions.

SOLVING OF THE FRACTIONAL NON-LINEAR AND LINEAR SCHRÖDINGER EQUATIONS BY HOMOTOPY PERTURBATION METHOD

FRACTIONAL VARIATIONAL ITERATION METHOD FOR TIME-FRACTIONAL NON-LINEAR FUNCTIONAL PARTIAL DIFFERENTIAL EQUATION HAVING PROPORTIONAL DELAYS

Parametric Iteration Method for Solving Linear Optimal Control Problems

Section 8 Convolution and Deconvolution

METHOD OF THE EQUIVALENT BOUNDARY CONDITIONS IN THE UNSTEADY PROBLEM FOR ELASTIC DIFFUSION LAYER

Approximately Quasi Inner Generalized Dynamics on Modules. { } t t R

EXISTENCE THEORY OF RANDOM DIFFERENTIAL EQUATIONS D. S. Palimkar

Additional Tables of Simulation Results

Math 6710, Fall 2016 Final Exam Solutions

A Novel Approach for Solving Burger s Equation

On Another Type of Transform Called Rangaig Transform

Moment Generating Function

Some Newton s Type Inequalities for Geometrically Relative Convex Functions ABSTRACT. 1. Introduction

Dynamic h-index: the Hirsch index in function of time

Some Properties of Semi-E-Convex Function and Semi-E-Convex Programming*

The Moment Approximation of the First Passage Time for the Birth Death Diffusion Process with Immigraton to a Moving Linear Barrier

A Study On (H, 1)(E, q) Product Summability Of Fourier Series And Its Conjugate Series

VARIOUS phenomena occurring in the applied sciences

If boundary values are necessary, they are called mixed initial-boundary value problems. Again, the simplest prototypes of these IV problems are:

N! AND THE GAMMA FUNCTION

An interesting result about subset sums. Nitu Kitchloo. Lior Pachter. November 27, Abstract

Notes 03 largely plagiarized by %khc

On Stability of Quintic Functional Equations in Random Normed Spaces

Some inequalities for q-polygamma function and ζ q -Riemann zeta functions

MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES. Boundary Value Problem for the Higher Order Equation with Fractional Derivative

Extremal graph theory II: K t and K t,t

Online Supplement to Reactive Tabu Search in a Team-Learning Problem

Review Exercises for Chapter 9

On the Existence and Uniqueness of Solutions for Nonlinear System Modeling Three-Dimensional Viscous Stratified Flows

INTEGER INTERVAL VALUE OF NEWTON DIVIDED DIFFERENCE AND FORWARD AND BACKWARD INTERPOLATION FORMULA

Lecture 15 First Properties of the Brownian Motion

VIM for Determining Unknown Source Parameter in Parabolic Equations

The Central Limit Theorem

Averaging of Fuzzy Integral Equations

Lecture 9: Polynomial Approximations

A note on deviation inequalities on {0, 1} n. by Julio Bernués*

Basic Results in Functional Analysis

th m m m m central moment : E[( X X) ] ( X X) ( x X) f ( x)

MODIFIED ADOMIAN DECOMPOSITION METHOD FOR SOLVING RICCATI DIFFERENTIAL EQUATIONS

L-functions and Class Numbers

Boundary-to-Displacement Asymptotic Gains for Wave Systems With Kelvin-Voigt Damping

DIFFERENTIAL EQUATIONS

Prakash Chandra Rautaray 1, Ellipse 2

NEWTON METHOD FOR DETERMINING THE OPTIMAL REPLENISHMENT POLICY FOR EPQ MODEL WITH PRESENT VALUE

Samuel Sindayigaya 1, Nyongesa L. Kennedy 2, Adu A.M. Wasike 3

Existence Of Solutions For Nonlinear Fractional Differential Equation With Integral Boundary Conditions

BE.430 Tutorial: Linear Operator Theory and Eigenfunction Expansion

Research Article Legendre Wavelet Operational Matrix Method for Solution of Riccati Differential Equation

On stability of first order linear impulsive differential equations

SPECIAL FUNCTIONS are part of mathematics that covers

A Robust H Filter Design for Uncertain Nonlinear Singular Systems

SUMMATION OF INFINITE SERIES REVISITED

The Connection between the Basel Problem and a Special Integral

BEST LINEAR FORECASTS VS. BEST POSSIBLE FORECASTS

MATH 507a ASSIGNMENT 4 SOLUTIONS FALL 2018 Prof. Alexander. g (x) dx = g(b) g(0) = g(b),

Solutions to selected problems from the midterm exam Math 222 Winter 2015

Research Article Generalized Equilibrium Problem with Mixed Relaxed Monotonicity

The analysis of the method on the one variable function s limit Ke Wu

Principles of Communications Lecture 1: Signals and Systems. Chih-Wei Liu 劉志尉 National Chiao Tung University

Application of Fixed Point Theorem of Convex-power Operators to Nonlinear Volterra Type Integral Equations

F D D D D F. smoothed value of the data including Y t the most recent data.

On Numerical Solution of Boundary Integral Equations of the Plane Elasticity Theory by Singular Integral Approximation Methods

Solutions to Problems 3, Level 4

On the Differential Fractional Transformation Method of MSEIR Epidemic Model

CLOSED FORM EVALUATION OF RESTRICTED SUMS CONTAINING SQUARES OF FIBONOMIAL COEFFICIENTS

FRACTIONAL SYNCHRONIZATION OF CHAOTIC SYSTEMS WITH DIFFERENT ORDERS

ECE 570 Session 7 IC 752-E Computer Aided Engineering for Integrated Circuits. Transient analysis. Discuss time marching methods used in SPICE

B. Maddah INDE 504 Simulation 09/02/17

A Note on Prediction with Misspecified Models

A Generalized Cost Malmquist Index to the Productivities of Units with Negative Data in DEA

Fuzzy Dynamic Equations on Time Scales under Generalized Delta Derivative via Contractive-like Mapping Principles

Research Article A MOLP Method for Solving Fully Fuzzy Linear Programming with LR Fuzzy Parameters

A Continuous Space-Time Finite Element Method for an Integro-Differential Equation Modeling Dynamic Fractional Order Viscoelasticity

LINEAR APPROXIMATION OF THE BASELINE RBC MODEL JANUARY 29, 2013

Transcription:

JOURAL OF MAHEMACAL SUDY J. Mah. Sudy, Vol. x, o. x (1x), pp. 1-14 Geeralized Jacobi specral-galerki mehod for oliear Volerra iegral equaios wih weakly sigular kerels Jie She 1,, Chagao Sheg 1 ad Zhogqig Wag 3 1 School of Mahemaical Scieces, Xiame Uiversiy, Xiame, Fuia 3615, P.R. Chia. Deparme of Mahemaics, Purdue Uiversiy, Wes Lafayee, 4797-1957, USA. 3 School of Sciece, Uiversiy of Shaghai for Sciece ad echology, Shaghai, 93, P.R. Chia. Absrac. We propose a geeralized Jacobi specral-galerki mehod for he oliear Volerra iegral equaios (VEs) wih weakly sigular kerels. We esablish he exisece ad uiqueess of he umerical soluio, ad characerize he covergece of he proposed mehod uder reasoable assumpios o he olieariy. We also prese umerical resuls which are cosise wih he heoreical predicios. Key Words: Geeralized Jacobi specral-galerki mehod, oliear Volerra iegral equaios wih weakly sigular kerels, covergece aalysis. AMS Subec Classificaios: 45D5, 65L6, 41A3, 65L7 1 roducio his paper is cocered wih he umerical soluios of he oliear Volerra iegral equaios wih weakly sigular kerels: y() = f ()+Vy() := f ()+ ( s) µ K(,s)G(s,y(s))ds, := [,], (1.1) where < µ < 1, K C(D) wih D := {(,s) : s }, f C() ad G is a coiuous fucio. rece years, here has bee a icreasig ieres i sudyig VEs. he mai difficulies for dealig wih weakly sigular VEs are: (i) he iegral operaor is o-local; (ii) Correspodig auhor. Email address: she7@purdue.edu (J. She), csheg@su.xmu.edu.c (C. Sheg), zqwag@uss.edu.c (Z. Wag) hp://www.global-sci.org/aa/ 1 c 1x Global-Sciece Press

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 he soluios are usually sigular ear =. Bruer [4] ad Lubich [17] ivesigaed he smoohess properies of he exac soluios of VEs wih weakly sigular kerels. Various umerical approaches, usig he piecewise polyomial collocaio mehods ad he Ruge-Kua mehods, have bee proposed for approximaig VEs wih weakly sigular kerels [4, 5, 1, 3]. However, hese umerical mehods do o paricularly deal wih he above wo difficulies. Specral mehods are capable of providig exceedigly accurae umerical resuls wih relaively less degree of freedoms, ad have bee widely used for scieific compuaio, see, e.g., [1,,6,13,14,19,]. Sice he specral mehods are global mehods, so hey could be beer suied for o-local problems. Recely, may kids of specral collocaio mehods are proposed for solvig VEs wih smooh kerels. Li, ag ad Xu [16] iroduced a ime parallel mehod wih specral-subdomai ehaceme for VEs; Sheg, Wag ad Guo [1] preseed a mulisep specral collocaio mehod for oliear VEs; Wag ad Sheg [] also proposed a mulisep specral collocaio mehod for oliear VEs wih delays. o solve VEs wih weakly sigular kerels, may aemps have bee made o overcome he difficulies caused by he sigulariies of he soluios. Che ad ag [9,1] proposed specral collocaio mehods for weakly sigular VEs; Huag, ag ad Zhag [15] sudied he supergeomeric covergece of specral collocaio mehods for weakly sigular Volerra/Fredholm iegral equaios. hese mehods usually use orhogoal polyomials as basis fucios. Aoher approach for solvig weakly sigular VEs is o use he o polyomial sigular fucios (which reflec he sigulariies of he exac soluios) as basis fucios. For example, Bruer [3] employed a o polyomial splie collocaio mehod for VEs wih weakly sigular kerels; Cao, Herdma ad Xu [7] preseed a o polyomial sigulariy preservig collocaio mehod for VEs wih weakly sigular kerels. his paper, we develop a o polyomial specral-galerki mehod for VEs wih weakly sigular kerels. More precisely, we cosruc a specral-galerki mehod for weakly sigular VEs (1.1), usig he geeralized Jacobi fucios as basis fucios. his kid of basis fucios have bee used by Zayerouri & Kariadakis [4] ad Che, She & Wag [8] for approximaig fracioal differeial equaios. he mai sraegies ad coribuios are as follows. We propose a geeralized Jacobi specral-galerki mehod for oliear VEs wih weakly sigular kerels. he basis fucios ca be ued o mach he sigulariies of he uderlyig soluios, ad lead o a efficie implemeaio. he exisig works (cf. [9, 1]) We approximae he problem (1.1) direcly wihou ay variable rasformaios, as opposed o he approach i [9, 1] where a specral-collecio mehod is cosruced for he rasformed VEs.

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 3 he res of his paper is orgaized as follows. Secio, we prese he geeralized Jacobi specral-galerki mehod for oliear VEs (1.1). Some useful lemmas for he covergece aalysis are provided i Secio 3. he exisece, uiqueess ad covergece of he geeralized Jacobi specral-galerki mehod are give i Secio 4. We prese i Secio 5 umerical experimes, which cofirm he heoreical expecaios. Some cocludig remarks are give i he fial secio. he geeralized Jacobi specral-galerki mehod his secio, we shall propose a specral-galerki mehod usig geeralized Jacobi fucios as basis fucios for problem (1.1). o his ed, we firs iroduce he shifed Jacobi polyomials ad he shifed geeralized Jacobi fucios o he ierval..1 he shifed Jacobi polyomials o. For α,β > 1, le J (α,β) (x), x Λ := ( 1,1) be he sadard Jacobi polyomial of degree, ad deoe he weigh fucio χ (α,β) (x) = (1 x) α (1+x) β. he se of Jacobi polyomials is a complee L χ (α,β) (Λ)-orhogoal sysem, i.e., Λ J (α,β) m (x)j (α,β) (x)χ (α,β) (x)dx = γ (α,β) m δ m,, (.1) where δ m, is he Kroecker fucio, ad γm α,β = α+β+1 Γ(α+1)Γ(β+1), Γ(α+ β+) m =, α+β+1 Γ(m+α+1)Γ(m+ β+1), (m+α+ β+1) m!γ(m+α+ β+1) m 1. (.) paricular, J (α,β) (x) =1. he shifed Jacobi polyomial of degree is defied by Clearly, he se of { J (α,β) fucio χ (α,β) J (α,β) () = J (α,β) ( 1),. (.3) ()} is a complee L ()-orhogoal sysem wih he weigh χ (α,β) () = ( ) α β. fac, by (.1) ad (.3) we kow ha J (α,β) m () J (α,β) ()χ (α,β) ()d = ( ) α+β+1γ (α,β) m δ m,. (.4) For ay ieger, we deoe by {x (α,β),ω (α,β) } = he odes ad he correspodig Chrisoffel umbers of he sadard Jacobi-Gauss ierpolaio o he ierval Λ. Le

4 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 P () be he se of polyomials of degree a mos o he ierval, ad (α,β) shifed Jacobi-Gauss quadraure odes o he ierval, be he (α,β) = (x(α,β) +1),. (.5) Due o he propery of he sadard Jacobi-Gauss quadraure, i follows ha for ay φ() P +1 (), φ()χ (α,β) ()d = ( ) 1 α+β+1 φ( 1 (x+1))χ(α,β) (x)dx = ( ) α+β+1 φ( (x(α,β) +1))ω (α,β) (.6) = = ( ) α+β+1 φ( (α,β) )ω (α,β). = By (.4) ad (.6), we furher obai ha for ay m+ +1, = J (α,β) m ( (α,β) ) J (α,β) ( (α,β) )ω (α,β) = γ (α,β) m δ m,. (.7). he shifed geeralized Jacobi fucios o. he shifed geeralized Jacobi fucio of degree is defied by (cf. [8]) P (α,β) (α,β) () := β J (), α,β > 1,. (.8) Le F (β) () be he fiie-dimesioal fracioal polyomial space (cf. [8]) F (β) () := {β ψ() : ψ() P ()} =spa{p (α,β) : }. (.9) Due o (.4) ad (.8), i is clear ha he se of {P (α,β) orhogoal sysem wih he weigh fucio χ (α, β) P (α,β) m ()P (α,β) ()χ (α, β) ()d = = (α,β) β J m J (α,β) m () J (α,β) () J (α,β) ()} is a complee L ()- χ (α, β) (), amely, ()χ (α, β) ()d ()χ (α,β) ()d = ( ) α+β+1γ (α,β) m δ m,. Because of (.6), i follows ha for ay ϕ() = β φ() wih φ() P +1 (), (.1) ϕ()χ (α, β) ()d = φ()χ (α,β) ()d = ( ) α+β+1 = = ( ) α+β+1 = ( (α,β) ) β ϕ( (α,β) )ω (α,β). φ( (α,β) )ω (α,β) (.11)

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 5 ex, le (u,v) χ (α, β) ad v χ (α, β) be he ier produc ad orm of space L (), re- χ (α, β) specively. We also iroduce he followig discree ier produc ad orm o he ierval, u,v (α, β) χ = ( ) α+β+1 = v,χ (α, β) = v,v 1 χ (α, β) haks o (.11), for ay φ,ψ F (β) (), (φ,ψ) χ (α, β) = φ,ψ χ (α, β). ( (α,β) ) β u( (α,β), φ χ (α, β) )v( (α,β) )ω (α,β), (.1) = φ (α, β),χ. (.13).3 he geeralized Jacobi specral-galerki mehod for problem (1.1) o describe he specral-galerki scheme for problem (1.1), we firs rasform he iegral ierval [,] o usig he rasformaio: he he equaio (1.1) becomes s = τ, τ. (.14) y() = f ()+Vy() = f ()+ ( ) 1 µ ( τ) µ K (, τ ) ( τ G,y( τ )) dτ. (.15) he geeralized Jacobi specral-galerki scheme is o seek Y() F (1 µ) (), such ha (Y,ϕ) χ = ( f,ϕ) χ +(VY,ϕ) χ, ϕ F (1 µ) (). (.16) We ow describe a umerical implemeaio for (.16). o his ed, we se Y() = m= y m P ( µ,1 µ) m (). (.17) Subsiuig (.17) io (.16) ad akig ϕ = P ( µ,1 µ) (), we obai ha for, m= Se y m (P ( µ,1 µ) m,p ( µ,1 µ) ) χ = ( f,p ( µ,1 µ) ) χ y = (y,,y ), A = (a m ),m, a m = (P ( µ,1 µ) m f = ( f,p ( µ,1 µ),p ( µ,1 µ) v (y) = (VY,P ( µ,1 µ) ) χ = ( ) χ, f = ( f,, f ), +(VY,P ( µ,1 µ) ) µγ ( µ,1 µ) m δ m,, ) χ, v(y) = (v,,v ). ) χ. (.18) (.19)

6 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 he, he sysem (.18) becomes Ay = f+v(y). (.) acual compuaio, we use he quadraure formula (.1) o approximae he erms f ad v, amely, f f,p ( µ,1 µ) χ ( ) µ = = ( ( µ,1 µ) ) µ f ( ( µ,1 µ) )P ( µ,1 µ) ( ( µ,1 µ) )ω ( µ,1 µ), (.1) ad v (y) µ 3 3µ i,= G( ( µ,1 µ) i ( ( µ,1 µ) i ( µ,) ) µ 1 K( ( µ,1 µ) i /,Y( ( µ,1 µ) i, ( µ,1 µ) i ( µ,) ( µ,) /) /))P ( µ,1 µ) ( ( µ,1 µ) i )ω ( µ,1 µ) i ω ( µ,). (.) his is a (oliear) implici scheme, which ca be solved, for isace, by he ewo ieraive mehod. 3 Some useful lemmas his secio, we prese some useful lemmas. For his purpose, we firs recall he defiiios of he fracioal iegrals ad fracioal derivaives i he sese of Riema-Liouville (see, e.g., [11, 18]). Defiiio 3.1. (Fracioal iegrals ad derivaives). For ρ R +, he lef ad righ fracioal iegrals are respecively defied as axu(x) ρ = 1 x Γ(ρ) a u(y) (x y) 1 ρ dy, x > a; x ρ b u(x) = 1 b Γ(ρ) x u(y) dy, x < b, (3.1) (y x) 1 ρ where Γ( ) is he usual Gamma fucio. For s [k 1,k) wih k, he lef-sided Riema-Liouville fracioal derivaive of order s is defied by ad s xu(x) = 1 Γ(k s) d k x dx k a u(y) dy, x (a,b), (3.) (x y) s k+1 ad he righ-sided Riema-Liouville fracioal derivaive of order s is defied by xdb s ( 1)k u(x) = Γ(k s) d k b dx k x u(y) dy, x (a,b). (3.3) (y x) s k+1

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 7 is clear ha for ay k, ad k x = D k, xd k b = ( 1)k D k, where D k := dk dx k. hus, we ca defie he Riema-Liouville fracioal derivaives as adxu(x) s = D k ax k s u(x), xdb s = ( 1)k D k x k s u(x). Accordig o heorem.14 of [11], we have ha for ay absoluely iegrable fucio u ad real s, adxa s xu(x) s = u(x), xdbx s b s u(x) = u(x), x (a,b). (3.4) ad ex, le F (β) (Λ) := {(1+x)β ψ(x) : ψ(x) P (Λ)} =spa{(1+x) β J (α,β) (x) : }, B r α,β (Λ) := {u(x) : u L χ (α, β) (Λ), 1 D β+l x u L χ (α+β+l,l) (Λ) for l r}, r. Deoe by c a geeric posiive cosa idepede of, ad he soluios of y() ad Y(). Accordig o heorem 4.3 of [8], we have Lemma 3.1. Le α > 1, β >, for ay u B α,β r (Λ) wih ieger r, we have where π (α,β) π (α,β) u u L χ (α, β) (Λ) c (β+r) 1 Dx β+r u L (Λ), (3.5) (α+β+r,r) is he sadard L (Λ)-orhogoal proecio upo F (β) χ (α, β) (Λ), defied by Λ Similarly, we defie ( π (α,β) u(x) u(x))ψ(x)χ(α, β) (x)dx =, ψ F (β) (Λ). (3.6) B r α,β () := {v() : v L χ (α, β) H r α,β () := {v() : v L χ (α, β) Deoe by π (α,β) he L χ (α, β) (), D β+l (), D β+l b χ v L () χ (α+β+l,l) for l r}, r, () for l r}, r. v L χ (α, β) ()-orhogoal proecio upo F (β) (), (π (α,β) v v,φ) χ (α, β) By Lemma 3.1, we obai he followig resuls. =, φ F (β) (). (3.7)

8 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 Lemma 3.. Le α > 1, β >. For ay v Bα,β r () wih ieger r, we have π (α,β) paricular, if v Hα,β r (), he Proof. Se u(x) := v() π (α,β) v v χ (α, β) v v χ (α, β) c β (β+r) D β+r c r (β+r) D β+r. Sice π(α,β) = v() = (x+1) (x+1) i he variable x, ad hece by he defiiios (3.6) ad (3.7), π (α,β) v (α+β+r,r) χ. (3.8) v (α, β) χ. (3.9) ad π(α,β) u(x) belog o F (β) (Λ) v() = = (x+1) π(α,β) u(x). (3.1) he above wih (3.5) ad (3.) yields ( π (α,β) ) α β+1 v v = χ (α, β) Λ c α β+1 (β+r) c β (β+r) ( π (α,β) u(x) u(x)) (1 x) α (1+x) β dx ( 1Dx β+r u(x) ) (1 x) α+β+r (1+x) r dx Λ (D β+r v() ) ( ) α+β+r r d. (3.11) his leads o he resul (3.8). Furhermore, (D β+r v() ) ( ) α+β+r r d ( (D )(β+r) β+r v() ) ( ) α β d. his leads o he resul (3.9). 4 Exisece, uiqueess ad error esimae his secio, we firs verify he exisece ad uiqueess of he soluio of (.16), ad he we aalyze ad characerize he covergece of scheme (.16) uder reasoable assumpios o he olieariy. heorem 4.1. Assume ha K(,s) C(D) ad G saisfies he followig Lipschiz codiio: he, for sufficiely small such ha he equaio (.16) possesses a uique soluio. G(s,y 1 ) G(s,y ) γ y 1 y, γ. (4.1) c µ β <1, (4.)

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 9 Proof. We firs prove he exisece. Cosider he followig ieraio process: (Y (m),ϕ) χ = ( f,ϕ) χ +(VY (m 1),ϕ) χ, ϕ F (1 µ) (). (4.3) Accordig o he defiiio (3.7) of he proecio operaor π ( µ,1 µ), we kow from (4.3) ha Y (m) = π ( µ,1 µ) ( f +VY (m 1) ). (4.4) ex, le Ỹ (m) =Y (m) Y (m 1). he, by (4.4) we ge Ỹ (m) = π ( µ,1 µ) ( VY (m 1) VY (m )). his, alog wih he proecio heorem, implies Ỹ (m) χ = π ( µ,1 µ) ( VY (m 1) VY (m )) χ VY (m 1) VY (m ). χ Furher, by (4.5), (4.1), (.1), he Cauchy-Schwarz iequaliy ad usig he raformaio = s+ (τ+1)( s), we ge ha Ỹ (m) χ ( ( s) µ K(,s) ( G(s,Y (m 1) (s)) G(s,Y (m ) (s)) ) ( ) ds) µ µ 1 d [ c ( s) µ ds ( s) µ( Ỹ (m 1) (s) ) ds ]( ) µ µ 1 d c ( ) µ ( s) µ( Ỹ (m 1) (s) ) dsd c (Ỹ(m 1) (s) ) ( ) µ ( s) µ dds s c (Ỹ(m 1) (s) ) 1 (1 τ) µ (1+τ) µ ( s) 1 µ dτds 1 c (Ỹ(m 1) (s) ) ( s) 1 µ ds c µ (Ỹ(m 1) (s) ) ( s) µ s µ 1 ds = c µ Ỹ (m 1) χ. hus, if c µ β < 1, he Ỹ (m) χ as m. his implies he exisece of soluio of (.16). is easy o prove he uiqueess of soluio of (.16). heorem 4.. Assume ha K(,s) C(D), y B µ,1 µ r () wih ieger r, G fulfills he Lipschiz codiio (4.1) ad is sufficiely small saisfyig he codiio (4.). he, here holds y Y χ paricular, if y H µ,1 µ r (), he y Y χ c µ 1 µ r 1 D µ+r+1 c r µ r 1 D µ+r+1 (4.5) (4.6) y (1 µ+r,r) χ. (4.7) y χ. (4.8)

1 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 Proof. By (.16) we kow Y = π ( µ,1 µ) ( ) f +VY. (4.9) Subracig (4.9) from (1.1) yields Usig (1.1) agai, we obai y Y = f π ( µ,1 µ) f π ( µ,1 µ) f = y π ( µ,1 µ) A combiaio of he previous wo equaliies leads o y Y = y π ( µ,1 µ) his, ogeher wih he proecio heorem, gives f +Vy π ( µ,1 µ) VY. y+π ( µ,1 µ) Vy Vy. y+π ( µ,1 µ) ( ) Vy VY. (4.1) y Y χ y π ( µ,1 µ) y π ( µ,1 µ) y χ y χ + π ( µ,1 µ) ( ) Vy VY χ + Vy VY χ. (4.11) ex, by a argume similar o (4.6) we deduce ha Vy VY χ c µ y Y. χ (4.1) Moreover, by (3.8) we ge ha for y B µ,1 µ r () wih ieger r, y π ( µ,1 µ) y χ c µ 1 µ r 1 D µ+r+1 y (1 µ+r,r) χ. (4.13) herefore, by (4.11) - (4.13) ad (4.), we obai he desired resul (4.7). Fially, by (3.9) ad a similar argume, we derive he resul (4.8). Remark 4.1. he Lipschiz codiio (4.1) appears o be ecessary for our covergece aalysis. However, some umerical experimes below show ha, eve if he Lipschiz codiio is o saisfied, he scheme is sill coverge. 5 umerical Resuls his secio, we prese some umerical resuls o illusrae he efficiecy of he geeralized Jacobi specral-galerki mehod.

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 11 5.1 Liear problem Cosider firs he liear VE wih weakly sigular kerel (cf. [9]): y() = f () ( s).35 y(s)ds, [,6], (5.1) where f () = 3.6 + 4.5 B(4.6,.65) ad B(, ) is he Bea fucio defied by 1 ( 1 ) x+y 1γ B(x,y) = x 1 (1 ) y 1 (y 1,x 1) d =. he exac soluio is y() = 3.6. Figure 5.1, we lis he discree L ()-errors χ ad he maximum errors of (5.1), hey idicae he algebraic covergece. fac, a direc compuaio shows ha y B µ,1 µ r () wih µ =.35 ad r = 6. Hece, accordig o (4.7), we ca expec a covergece rae for he L ()-orm o be of he order r µ+1=6.65. χ ()-orm ploed i Figure 5.1 is abou 8.3. he observed covergece rae for he L χ able 5.1 below, we compare he maximum errors of our algorihm wih ha of he collocaio mehod suggesed i [9] (see able 1 of [9]). We observe ha our mehod provides more accurae umerical resuls. Figure 5.1: he umerical errors of (5.1). Figure 5.: he umerical errors of (5.). able 5.1: A compariso of L ()-errors for (5.1). 4 6 8 Ref. [9] 6.7887e+1.4594e-1 1.337e- 1.95e-3 Our mehod 6.5755e+ 3.174e-3 1.1957e-4 1.4431e-5 1 1 14 16 Ref. [9] 4.486e-4 1.3478e-4 4.8583e-5 1.998e-5 Our mehod.8175e-6 7.733e-7.795e-7 8.48e-8

1 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 Cosider ex he liear VE wih weakly sigular kerel: y() = f ()+ ( s) 1/3 y(s)ds, [,1]. (5.) We choose f such ha he soluio y of (5.) is give by y()= /3 cos(). ca be verified readily ha y B µ,1 µ () wih µ = 1/3, so heorem 4. predics ha he errors of he geeralized Jacobi specral approximaio will decrease faser ha ay algebraic rae. Figure 5., we lis he discree L ()-errors ad he maximum errors of (5.). χ We observe ha he umerical errors decay expoeially as icreases. 5. oliear problem Cosider firs he oliear VE wih weakly sigular kerel: y() = f ()+ ( s) 1/ exp ( s 1/ y(s)/ ) ds, [,1]. (5.3) We choose f such ha he soluio y of (5.3) is give by y() = 1/ l(+e). Clearly, he exac soluio y B 1/,1/ (). However, he Lipschiz codiio (4.1) is o saisfied for problem (5.3). Figure 5.3, we plo he discree L ()-errors wih µ=1/ ad he maximum errors of (5.3). is show ha he umerical errors decay expoeially as icreases. his χ meas ha our algorihm is sill valid for problem (5.3), eve if he Lipschiz codiio (4.1) is o saisfied. Figure 5.3: he umerical errors of (5.3). Figure 5.4: he umerical errors of (5.4). Cosider ex he oliear VE wih weakly sigular kerels: y() = exp()+ 4 3 3/ ( s) 1/ exp( s)y (s)ds, [,1], (5.4) wih he exac soluio y() = exp(). Clearly, he exac soluio y B 1/,1/ (). Moreover, he Lipschiz codiio (4.1) is o saisfied for problem (5.4). Figure 5.4, we lis he discree L ()-errors wih µ = 1/ ad he maximum χ errors of (5.4). hey also idicae ha he umerical errors decay expoeially as icreases.

J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 13 6 Cocludig Remarks his paper, we proposed a geeralized Jacobi specral-galerki mehod for he oliear VEs wih weakly sigular kerel. his mehod ca be implemeed efficiely. We showed he exisece ad uiqueess of he umerical soluio ad proved is covergece rae uder reasoable assumpios o he olieariy. umerical experimes demosrae ha he proposed mehod are very effecive for dealig wih liear ad oliear VEs. Ackowledgmes he work of J.S. is suppored i par by SF of Chia gras 1137198 ad 1141111; ad he work of Z.W. is suppored i par by SF of Chia gra 111715 ad he Research Fud for Docoral Program of Higher Educaio of Chia (gra 13317116). Refereces [1] C. Berardi ad Y. Maday, Specral Mehod, i Hadbook of umerical Aalysis, par 5, edied by P. G. Ciarle ad J. L. Lios, orh-hollad, 1997. [] J. P. Boyd, Chebyshev ad Fourier Specral Mehods, Spriger-Verlag, Berli, 1989. [3] H. Bruer, opolyomial splie collocaio for Volerra equaios wih weakly sigular kerels, SAM J. umer. Aal., (1983), 116-1119. [4] H. Bruer, Collocaio mehods for Volerra egral ad Relaed Fucioal Differeial Equaios, Cambridge Uiversiy Press, Cambridge, 4. [5] H. Bruer, A. Pedas ad G. Vaiikko, he piecewise polyomial collocaio mehod for oliear weakly sigular Volerra equaios, Mah. Comp., 7 (1999), 179-195. [6] C. Cauo, M. Y. Hussaii, A. Quareroi ad. A. Zag, Specral Mehods: Fudameals i Sigle Domais, Spriger-Verlag, Berli, 6. [7] Y. Cao,. Herdma ad Y. Xu, A hybrid collocaio mehod for Volerra iegral equaios wih weakly sigular kerels, SAM J. umer. Aal., 41 (3), 364-381. [8] S. Che, J. She ad L. Wag, Geeralized Jacobi Fucios ad heir Applicaios o Fracioal Differeial Equaios, submied. [9] Y. Che ad. ag, Specral mehods for weakly sigular Volerra iegral equaios wih smooh soluios, J. Compu. Appl. Mah., 33 (9), 938-95. [1] Y. Che ad. ag, Covergece aalysis of he Jacobi specral-collocaio mehods for Volerra iegral equaios wih a weakly sigular kerel. Mah. Comp., 79 (1), 147-167. [11] K. Diehelm. he Aalysis of Fracioal Differeial Equaios, Lecure oes i Mah., Vol. 4. Spriger, Berli, 1. [1]. Diogo ad P. Lima, Supercovergece of collocaio mehods for a class of weakly sigular Volerra iegral equaios. J. Compu. Appl. Mah., 18 (8), 37-316. [13] D. Fuaro, Polyomial Approximaios of Differeial Equaios, Spriger-Verlag, 199. [14] B. Guo, Specral Mehods ad heir Applicaios, World Scieific, Sigapore, 1998.

14 J. She, C. Sheg ad Z. Wag / J. Mah. Sudy, x (1x), pp. 1-14 [15] C. Huag,. ag ad Z. Zhag, Supergeomeric covergece of specral collocaio mehods for weakly sigular Volerra ad Fredholm iegral equaios wih smooh soluios, J. Compu. Mah., 9 (11), 698-719. [16] X. Li,. ag ad C. Xu, Parallel i ime algorihm wih specral-subdomai ehaceme for Volerra iegral equaios, SAM J. umer. Aal., 51 (13), 1735-1756. [17] C. Lubich, Ruge-Kua heory for Volerra ad Abel iegral equaios of he secod kid, Mah. Comp., 41 (1983), 87-1. [18]. Podluby, Fracioal Differeial Equaios, volume 198 of Mahemaics i Sciece ad Egieerig. Academic Press c., Sa Diego, CA, 1999. [19] J. She ad. ag, Specral ad High-order Mehods wih Applicaios, Sciece Press, Beiig, 6. [] J. She,. ag ad L. Wag, Specral Mehods: Algorihms, Aalysis ad Applicaios, Spriger Series i Compuaioal Mahemaics, Vol. 41, Spriger-Verlag, Berli, Heidelberg, 11. [1] C. Sheg, Z. Wag ad B. Guo, A mulisep Legedre-Gauss specral collocaio mehod for oliear Volerra iegral equaios, SAM J. umer. Aal., 5 (14), 1953-198. [] Z. Wag ad C. Sheg, A hp-specral collocaio mehod for oliear Volerra iegral equaios wih vaishig variable delays, Mah. Comp., i press. [3] Q. Wu, O graded meshes for weakly sigular Volerra iegral equaios wih oscillaory rigoomeric kerels, J. Compu. Appl. Mah., 63 (14), 37-376. [4] M. Zayerouri ad G.E. Kariadakis. Fracioal Surm-Liouville eige-problems: heory ad umerical approximaio. J. Compu. Phys., 5:495 517, 13.