legendre–galerkin method for the linear fredholm

Upload: isa-muslu

Post on 02-Jun-2018

212 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    1/12

    LegendreGalerkin method for the linear Fredholm

    integro-differential equations

    Mohamed Fathy a, Mohamed El-Gamel b,, M.S. El-Azab b

    a Department of Engineering Physics and Mathematics, Faculty of Engineering, Helwan University, Egyptb Department of Mathematical Sciences, Faculty of Engineering, Mansoura University, Egypt

    a r t i c l e i n f o

    Keywords:

    Galerkin

    Legendre

    Numerical

    Gaussian quadrature

    a b s t r a c t

    This paper presents a study of the performance of the Galerkin method using Legendre

    basis functions for solving linear Fredholm integro-differential problems. Convergence

    and error estimation of the method were discussed. The method is then tested on several

    examples. Numerical results are included and comparisons with other methods are made

    to confirm the efficiency and accuracy of the method.

    2014 Elsevier Inc. All rights reserved.

    1. Introduction

    The boundary value problems in terms of integro-differential equations have many practical applications such asmechanics, physics, chemistry, astronomy, biology, economics, potential theory, engineering problems[19]and electrostat-

    ics[11]. The existence and uniqueness of the solutions for these problems were discussed by Agarwal[1,2].

    The present work is motivated by the desire to obtain numerical solutions to boundary value problems for second-order

    Fredholm integro-differential equations via LegendreGalerkin method. We will consider the numerical solution of a class of

    linear Fredholm integro-differential boundary value problems in the form

    Xri0

    lixuix fx k

    Z 11

    kx; tutdt; 1:1

    ru1 rr 1u1 0; 1:2

    wherelix;fx; kx; t andux are continuous functions inL2 space, r 0; 1; 2 and k is a parameter.

    Many numerical methods have been developed for solving the integro-differential Eq.(1.1)with Fredholm type numer-

    ically. Each of these methods has its inherent advantages and disadvantages and the search for alternative, more general,

    easier and more accurate methods is a continuous and ongoing process. Next, we present a selective review of the most

    recent main methods. These methods include: Adomian decomposition [9], variational iteration method [5], homotopy anal-

    ysis method[11], Chebyshev and Taylor collocation[22],Haar wavelet[15], sinc-collocation method[17],Tau method[10],

    Taylors series expansion[8,13], hybrid Taylor and Block Pulse functions [14]and integral mean value[3].

    In recent years, a lot of attention has been devoted to the study of Legendre methods to investigate various scientific mod-

    els. Using these methods made it possible to solve differential equations of LaneEmden type [24], second and fourth order

    http://dx.doi.org/10.1016/j.amc.2014.06.057

    0096-3003/ 2014 Elsevier Inc. All rights reserved.

    Corresponding author.

    E-mail addresses:[email protected](M. Fathy), [email protected](M. El-Gamel).

    Applied Mathematics and Computation 243 (2014) 789800

    Contents lists available at ScienceDirect

    Applied Mathematics and Computation

    j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a te / a m c

    http://dx.doi.org/10.1016/j.amc.2014.06.057mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.amc.2014.06.057http://www.sciencedirect.com/science/journal/00963003http://www.elsevier.com/locate/amchttp://www.elsevier.com/locate/amchttp://www.sciencedirect.com/science/journal/00963003http://dx.doi.org/10.1016/j.amc.2014.06.057mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.amc.2014.06.057http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://crossmark.crossref.org/dialog/?doi=10.1016/j.amc.2014.06.057&domain=pdf
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    2/12

    equations[21], CahnHilliard equations with Neumann boundary conditions[28], Fredholm integral[12], Helmholtz equa-

    tion [4], second kind Volterra integral equations [29], high-order linear Fredholm integro-differential [23] and Abels integral

    equation[25].

    This paper is organized as follows: in Section2,we present the preliminaries of Legendre polynomials, new lemmas and

    theorems required for our subsequent development. In Section 3, convergence and error estimation of the method are

    described in detail. Section4 is devoted to derivation of the discrete system. Section 5 presents appropriate techniques to

    treat nonhomogeneous boundary conditions for case n 2. Section 6shows the accuracy of the proposed method using

    numerical examples and making comparisons with other methods. Section 7 gives a brief conclusion. Note that we have

    computed the numerical results by Matlab programming.

    2. Legendre function preliminaries

    Orthogonal polynomials are widely used in applications in mathematics, mathematical physics, engineering and com-

    puter science. One of the most common orthogonal polynomial set is the Legendre polynomials. The Legendre polynomials

    Pnxsatisfy the Legendre differential equation

    1 x2y00 2xy0 nn 1y 0; 1 6 x 6 1; nP 0

    with recurrence relations

    P0n1x P0n1x 2n 1Pnx; 2:1

    xP0nx P0n1x nPnx: 2:2

    These polynomial are orthogonal on 1; 1Z 11

    PmxPnxdx2

    2n1; if m n;

    0; if m n

    ( 2:3

    and

    Z 11

    Pnxdx 2; n 0;

    0; n> 0

    2:4

    Lemma 2.1. Let n; l and N be any two positive integer numbers such that nl 6 N and l > 0, thenZ 11

    PnxP00nlxdx 0:

    Proof. Integrating the left term by parts and using Eq. (2.4)can prove the above lemma. h

    Lemma 2.2. Let n; m and N be any two integer numbers such that n P m and n; m 6 N, thenZ 11

    PnxP0mxdx 0:

    Proof. First, let n m. Integrating the left hand side in Lemma 2.2two times by parts yields

    Z 11

    PnxP0nxdx

    1

    2Pnx

    2

    1

    1 0 )

    Z 11

    PnxP0mxdx 0:0

    Second, letn > m. Using the recurrence Eq. (2.1) and (2.4), the left term inLemma 2.2can be written as follows:

    Z 11

    PnxP0mxdx

    Z 11

    Pnx2m 1Pm1x P0m2xdx

    Z 11

    PnxP0m2xdx

    Z 11

    PnxP0m4xdx

    R11

    PnxP00xdx; if m even;

    R11PnxP01xdx; if m odd(

    0:

    790 M. Fathy et al./ Applied Mathematics and Computation 243 (2014) 789800

    http://-/?-http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    3/12

    Theorem 2.1. Let n; m and N be any two integer numbers such that n ; m 6 N, then

    i

    Z 11

    P0nxPmxdx 2; if n mi;

    0; if nmi or m P n

    iiZ 1

    1

    xP0nxPmxdx

    2n2n1

    ; if m n;

    0; if n mi or m> n;

    2; if nmi

    8>:

    iii

    Z 11

    xP0nxP0mxdx

    0; if n m or m ni or n mi

    nn 1; if m ni; m> n

    mm 1; if nmi; n> m

    8>:

    iv

    Z 11

    P0nxP0mxdx

    0; if n m or m ni or nmi

    mm 1; if n mi; n> m

    nn 1; if m ni; m> n

    8>:

    where i 1; 3; 5;. . .; 2k 1 6 Nm

    Proof

    (i) Integrating the left hand side for (i) by parts yieldsZ 11

    P0nxPmxdx PnxPmx11

    Z 11

    PnxP0mxdx 1 1

    nm1

    Z 11

    PnxP0mxdx: 2:5

    Forn mi; i 1; 3; 5;. . . 6 Nm, by usingLemma (2.2), the integration(2.5)can be written as follows:Z 11

    P0nxPmxdx 2:

    As in the above case and Lemma (2.2)is considered but n mi; i 0; 2; 4;. . . 6 Nm, yields

    Z 11 P

    0

    nxPmxdx 0:

    Form P n, previous cases must be considered besidesLemma 2.2. So, the right hand side of Eq. (2.5)is equal to zero.

    (ii) CombineLemma 2.2and (i) besides recurrence Eq. (2.2). To prove (ii), the proof must be divide into four cases.

    First, letm n, by taking back recurrence Eq. (2.2)and recalling the orthogonality of the Legendre polynomials, the

    first part of the above theorem can be drawn up asZ 11

    xP0nxPmxdx n

    Z 11

    PnxPnxdx

    Z 11

    P0n1xPnxdx 2n

    2n 1:

    In the next case, let n mi; i 1; 3; 5; 6 Nm, the proof is as that in the previous case but n mi. The integration is

    equal to zero. On the other hand, Theorem 2.1(ii) can be proved in status m > n as in the previous two cases but the inte-

    gration is zero. Finally, assume n mi; i 2; 4; 6;. . . 6 Nm, by getting back to recurrence Eq. (2.2), Lemma 2.2 and

    (i). Integrating by parts yields

    Z 11

    xP0nxPmxdx Pn1xPnix11Z 1

    1Pn1xP0nixdx 2:

    (iii) The left term in the above lemma can be written by using recurrence Eq.(2.2)asZ 11

    xP0nxP0mxdx

    Z 11

    nPnxP0mxdx

    Z 11

    P0n1xP0mxdx 2:6

    form n, recalling (i),Lemma 2.1and integrating the second term in the right hand side for Eq. (2.6), the result is zero. In

    this case assume m > n andm ni, as in the first case the result is equal to nn 1. In the next status let m > n and

    m ni, the integration is equal to zero. Finally, at n > m, the recurrence can be proved as in the above cases by interchang-

    ingn by m andm insteadn.

    (iv) The integration in the left hand side in (iv) can be written by using the recurrence (2.2)as

    Z 1

    1

    P0nxP0mxdx Z

    1

    1

    xP0n1xP0mxdx n 1Z

    1

    1

    Pn1xP0mxdx 2:7

    then, by using (iii) and (i) into Eq. (2.7)and substituting each case of them, (iv) is proved. h

    M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800 791

  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    4/12

    Theorem 2.2. Let n; m and N be any two integer numbers such that n ; m 6 N, then

    i

    Z 11

    P00nxPmxdx nn 1 mm 1; if nmi;

    0; if n mi or mP n

    ii

    Z 11

    xP00nxPmxdx nn 1 mm 1 2; if n mi;

    0; if nmi or mP n

    iii

    Z 11

    x2P00nxPmxdx

    2nn12n1

    ; if m n;

    nn 1 mm 1 4; if n mi 1;

    0; if nmi 1 or m> n:

    8>:

    where i 1; 3; 5;. . .; 2k 1 6 Nm 1.

    Proof

    (i) The proof must be divided into four cases. First, let n mi; i 2; 4; 6;. . .; 2k 1 6 Nm. Integrating the left hand

    side by parts twice and usingLemma 2.1produce

    Z 11 P

    00

    nxPmxdx P

    0

    nxPnix

    1

    1Z 1

    1 P

    0

    nxP

    0

    nixdx nn 1 PnxP

    0

    nix

    1

    1Z 1

    1 PnxP

    00

    nixdx

    nn 1 mm 1:

    Second, letn mi; i 1; 3; 5;. . .; 2k 1 6 Nm. As in the first case, but the value ofi is odd numbers. So, the integration

    is zero.

    Third, letm > n. By using Eq.(2.4)and integrating the left side, the result is equal to zero.

    Fourth, atm n, the value of the integration also is equal to zero by using the integration by parts.

    (ii) To prove this point, Theorem 2.1(i) andTheorem 2.1(iii) are considered. Substitute each case in each of them after

    integrating the left side of the above theorem by parts.

    (iii) Integrating the left term in the above theorem yieldsZ 11

    x2P00nxPmxdxnn 1

    2 1 1

    mn 2

    Z 11

    xP0nxPmxdx

    Z 11

    x2P0nxP0mxdx: 2:8

    Moreover, by recalling recurrence Eq.(2.2)and using it in the last term in Eq. (2.8)then expanding the result yieldsZ 11

    x2P0nxP0mxdx nm

    Z 11

    PnxPmxdxn

    Z 11

    PnxP0m1xdxm

    Z 11

    PmxP0n1xdx

    Z 11

    P0n1xP0m1xdx: 2:9

    by substituting Eq.(2.9) and Theorem 2.1(ii) into Eq. (2.8), then taking back the orthogonality of the Legendre polynomial

    (2.3)and usingTheorem 2.1(i) and Theorem 2.1(iv), the result produces four cases. By collecting the four cases, the above

    theorem can be proved. h

    3. Convergence and error estimation

    3.1. Convergence of Legendre polynomial

    Consider Fredholm integral equation of the second kind by puttingr 0 andl0x 1

    ux fx k

    Z 11

    kx; tutdt: 3:1

    If we define

    @ : X! X; @u k

    Z 11

    kx; tutdt;

    we can rewrite(3.1)as a form of

    Ik@u f:

    792 M. Fathy et al./ Applied Mathematics and Computation 243 (2014) 789800

    http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    5/12

    Theorem 3.1 [16]. Let X be a normed space, @ : X! X a compact and Ik@ be injective. Then the inverse operatorIk@1

    :X! X exist and is bounded.

    By regard to the above theorem, we conclude the uniqueness of solution of Fredholm integral equation of the second kind.

    Theorem 3.2 [16]. Assume @ : Y!Y is bounded, with Y a Banach space, and assume c @ : Y! Y is one to one and onto.Further assume

    k@Pn@k ! 0 as n! 1;

    then for all sufficiently large n, say n P N, the operatorc Pn@1 exists as a bounded operator from Y to Y. Moreover, it is uni-

    formly bounded:

    supnPN

    kc Pn@1k< 1:

    For the solution ofc Pn@xn Pny;xn 2 Y and c @x y;

    xxn cc Pn@1xPnx;

    jcjkc Pn@k

    kxPnxk 6 kxxnk 6 jcjkc Pn@1kkxPnxk:

    Lemma 3.1 ([6,16]). Let xt 2Hk1; 1(a Sobolev space) and let xnt P

    ni0aiPitbe the best approximation polynomial of

    xt in the L2-norm, then

    kxt xntkL2 1;1 6 c0nkjjxtjjHk1;1

    where

    kxtkL2 1;1

    Z 11

    x2tdt

    1=2;

    kxtkHk1;1

    Xki0

    Z 11

    jxitj2dt

    !1=2:

    and c0 is a positive constant, which depends on the selected norm and is independent of x tand n.By regard to theLemma 3.1we conclude that approximation rate of Legendre polynomials isnk , alsoTheorem 3.2indi-

    cate thatkxxnk converge to zero at exactly the same speed as kxPnxk.

    3.2. Error estimation of LegendreGalerkin method

    In this section an error estimator for the LegendreGalerkin approximate solution of a Fredholm integro differential equa-

    tion is obtained[10,27,26]. Let us call enx ux unxas the error function of the Legendre approximation unxto ux,

    whereuxis the exact solution of(1.1) and (1.2). Hence,unx satisfies the following problem

    Xri0

    lixuin x k

    Z 11

    kx; tuntdt fx Hnx; 3:2

    with boundary

    run1 rr 1un1 0; 3:3

    whereHn is a perturbation term associated with unx and can be obtained by substituting unxinto the equation

    Hnx Xri0

    lixuin x k

    Z 11

    kx; tuntdtfx: 3:4

    We proceed to find an approximation en;Nxto theenx in the same way as we did before for the solution(1.1) and (1.2).

    Subtracting(3.2) and (3.3)from(1.1) and (1.2), respectively, the error function enx satisfies the equation

    Xri0

    lixein x k

    Z 11

    kx; tentdt Hnx; 3:5

    with boundary

    ren1 rr 1en1 0: 3:6

    M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800 793

    http://-/?-http://-/?-http://-/?-http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    6/12

    By solving this problem in the same way as Section3, we get the approximation en;Nx. It should be noted that in order to

    construct the Legendre approximation en;Nxto enx.

    4. The LegendreGalerkin Method

    We assume the solution of(1.1)is approximated by the finite expansion of Legendre basis function

    ux Xnj0

    cjPjx: 4:1

    The unknown coefficientscj in 4.1are determined by orthogonalizing the residual with respect to the basis functionsPrx.

    This yields the discrete system

    hl2xu00x; Prxi hl1xu

    0x; Prxi hl0xux; Prxi hfx; Prxi kZ 1

    1

    kx; tutdt; Prx

    : 4:2

    The weighted inner product h; iis taken to be

    hf;gi Z 1

    1

    fx gxdx:

    The method of approximating the integrals in(4.2)begins by integrating by parts to transfer all derivatives from u to Pr.

    Therefore, the following lemma is needed.

    Lemma 4.1. The following relations hold

    hl1xu0x; Prxi

    Z 11

    l1xPrx0uxdx; 4:3

    hl2xu00x; Prxi u

    0xl2xPrx1

    1

    Z 11

    l2xPrx00uxdx; 4:4

    hGx; Prxi Xmi0

    Gxi 2

    1 x2iP0mxi

    2; 4:5

    Z 1

    1

    Fx; tdt; Prx Xm

    i0X

    m

    l0

    xi;lFxi; tlPrxi; 4:6

    where

    xi;l xixl 4

    1 x2i1 x2lP

    0mxiP

    0mxl

    2:

    Replacing each term of(4.2) with the approximation defined in (4.3)(4.6) respectively, we obtain the following theorem.

    Theorem 4.1. If the assumed approximate solution of the boundary-value problem is (1.1), (1.2) is (4.1) then the discrete

    GalerkinLegendre system for the determination of the unknown coefficients fcjgnj0

    is given by

    Xn

    j0

    cj k

    Xm

    i0 Xm

    l0

    xixlkxi; tlPjtlPrxi u0xl2xPrx

    1

    1

    Z 11

    l2xPrx00Pjxdx

    Z 11

    l1xPrx0Pjxdx

    "

    Z 11

    l0xPjxPrxdx

    Xmq0

    xqfxqPrxqdx: 4:7

    The system(4.7)takes the matrix form

    Ac b; 4:8

    where

    A

    kh0;0e0;0 v0;0 kh1;0e1;0 v1;0 khn;0en;0 vn;0kh0;1e0;1 v0;1 kh1;1e1;1 v1;1 khn;1en;1 vn;1kh0;2e0;2 v0;2 kh1;2e1;2 v1;2 khn;2en;2 vn;2kh0;3e0;3 v0;3 kh1;3e1;3 v1;3 khn;3en;3 vn;3

    .

    .

    ....

    ..

    ....

    kh0;ne0;n v0;n kh1;ne1;n v1;n khn;nen;n vn;n

    0BBBBBBB@

    1CCCCCCCA

    4:9

    794 M. Fathy et al./ Applied Mathematics and Computation 243 (2014) 789800

    http://-/?-http://-/?-http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    7/12

    and

    ejr l2xP0jxPrx

    1

    1;

    hjrXmi0

    Xml0

    xixlkxi; tlPjtlPrxi;

    vjr

    Z 11

    l2xP00rxPjxdx 2

    Z 11

    l02xP0rxPjxdx

    Z 11

    l002xPrxPjxdxZ 1

    1

    l1xP0rxPjxdx

    Z 11

    l01xPrxPjxdxZ 1

    1

    l0xPjxPrxdx

    vjrcan be evaluated from theorems and lemmas in Section2.

    Now we have a linear system (4.8) ofn 1 equations forn 1 unknown coefficients. We can obtain the coefficient of the

    approximate solution by solving this linear system by Q-R method. The solutionc c0;. . .; cns gives the coefficients in the

    approximate LegendreGalerkin solutionux.

    Algorithm

    INPUT:n wheren 2 Z if the domain is a; b;x ab

    2 s ab

    2

    evaluate A andb

    solve Ac b

    OUTPUT: the values ofc0; c1; c2;. . .; cn s 2

    baxbaba

    end

    5. Treatment of boundary condition

    In Section3above, the development of LegendreGalerkin technique for homogeneous boundary conditions, for n 2, if

    the boundary conditions are nonhomogeneous,

    u1 c; u1 e 5:1

    then these conditions need be converted to homogeneous conditions via an interpolation by a known function. Applying the

    transformation

    yx ux 1 x

    2 c

    x 1

    2 e;

    to the problem(1.1), (1.2)yields

    X2i0

    lixyix fx k

    Z 11

    kx; tytdt

    with boundary conditions

    y1 0; y1 0;

    where

    ^fx fx e c

    2 l1x

    e cx e c2

    l0x k

    Z 11

    kx; t t 1e 1 tc

    2

    dt:

    The resulting discrete system for the coefficients n 1 in the approximate LegendreGalerkin solution

    ux Xnj0

    cjPjx 1 x

    2 c

    x 1

    2 e

    is exactly the system in(4.8), withfin that system replaced by

    ^

    f. Notice that ifc e 0, the problem reduces to the homo-geneous case.

    M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800 795

    http://-/?-http://-/?-http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    8/12

  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    9/12

    Fig. 1. (a) Comparison ofkELGk withn = 5, 10, 15, 20, 25 for example 1. (b) Legendre and exact solutions for example 1.

    Table 3

    Maximum absolute errors and maximum estimate error for Example 2 at different n.

    N LegendreGalerkinkELGj kenk

    5 5.764E05 8.700E06

    10 2.530E12 1.124E12

    15 1.776E15 2.288E15

    20 1.332E15 2.170E15

    Table 4

    Comparison of maximum absolute errors for example 2.

    LegendreGalerkin Sinc-Collocation[17] The rationalized

    Haar wavelet[15]

    2.442E15 0.4756 E06 0.0413

    Fig. 2. (a) Comparison ofkELGk withn = 5, 10, 15, 20 for example 2. (b) Legendre and exact solutions for example 2.

    M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800 797

    http://-/?-http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    10/12

    subject to the boundary conditions

    u1 u1 0;

    whose exact solution is

    ux sinpx:

    The computational results are summarized inTable 5. for different values ofn. Fig. 3(a) displays the maximum absolute

    error and maximum estimate error at differentn andFig. 3(b) bids the Legendre and exact solutions.

    Example 4 [23]. Consider the integro-differential equation

    u00 xu0 xu ex 2sinx sinx

    Z 11

    etutdt; 1< x; t< 1;

    subject to the boundary conditions

    u1 e1; u1 e;

    whose exact solution is

    ux ex:

    The computational results are summarized in Table 6 for different values ofn. Maximum absolute error is tabulated in Table 7for LegendreGalerkin together with the numerical results of S. Yalinbas [23], who used Legendre-collocation matrix

    Table 5

    Maximum absolute errors and maximum estimate error for Example 3.

    N LegendreGalerkin kELGk jjenjj

    5 6.006E01 2.919E01

    10 1.008E04 8.528E05

    15 3.282E08 2.836E08

    20 3.752E14 4.244E14

    Fig. 3. (a) Comparison ofkELGk withn = 5, 10, 15, 20 for example 3. (b) Legendre and exact solutions for example 3.

    Table 6

    Maximum absolute errors and maximum estimate error for Example 4 at different n.

    N LegendreGalerkin kELGk kenk

    5 2.136E03 6.324E03

    10 1.008E09 9.187E10

    15 8.104E15 1.297E15

    798 M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800

  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    11/12

    method to obtain the numerical solution. Fig. 4(a) shows the maximum absolute error at different n and Fig. 4(b) exhibits the

    Legendre and exact solutions.

    7. Conclusion

    This paper has discussed how the LegendreGalerkin method can be applied for obtaining solutions of integral and inte-

    gro-differential equations. The formulation and implementation of the scheme are illustrated. The proposed method was

    tested using some problems with results. This Paper discussed how the integro-differential equation with variable coeffi-

    cients can be solved using the LegendreGalerkin method. Matlab and Mathematica had been used to obtain the approxi-

    mate solution. If the problem domain is a;b, the linear transformation must be used to convert the problem domain to1; 1. The required linear transformation is s ba

    2 x ba

    2. Then, by using LegendreGalerkin technique, the approximate

    solution is converted to linear algebraic system. By solving this system, the numerical solution is obtained. Finally, By using

    inverse transformation, the final approximate solution can be produced.

    Acknowledgments

    The authors are grateful to the referees for their valuable comments and suggestions on the original manuscript.

    References

    [1]R. Agarwal, Boundary Value Problems for High Order Differential Equations, World Scientific, Singapore, 1986 .

    [2]R. Agarwal, Boundary value problems for higher order integro-differential equations non-linear analysis theory, Meth. Appl. 7 (1983) 259270 .

    [3]Z. Avazzadeh, M. Heydari, G. Loghmani, Numerical solution of Fredholm integral equations of the second kind by using integral mean value theorem,

    Appl. Math. Comput. 35 (2011) 23742383.

    [4] B. Bialecki, A. Karageorghis, Legendre Gauss spectral collocation for the Helmholtz equation on a rectangle, Numer. Algorithms 36 (2004) 203227.

    [5]N. Bildik, A. Konuralp, S. Yalinbas, Comparison of Legendre polynomial approximation and variational iteration method for the solutions of general

    linear Fredholm integro-differential equations, Appl. Math. Comput. 59 (2010) 19091917.

    [6] C. Canuto, M. Hussaini, S. Al-Shara, Spectral Methods, Springer, 2006.

    [7]H. Danfu, S. Xufeng, Numerical solution of integro-differential equations by using CAS wavelet operational matrix of integration, Appl. Math. Comput.

    194 (2007) 460466.

    [8] P. Darania, A. Ebadian, A method for the numerical solution of the integro-differential equations, Appl. Math. Comput. 188 (2007) 657668.

    [9]I. Hashim, Adomian decomposition method for solving BVPs for fourth-order integro-differential equations, J. Comput. Appl. Math. 193 (2006) 658

    664.

    [10] S. Hosseini, S. Shahmorad, Tau numerical solution of Fredholm integro-differential equations with arbitrary polynomial bases, Appl. Math. Model. 27

    (2003) 145154.

    Table 7

    Comparison of maximum absolute errors for example 4.

    LegendreGalerkin Legendre-collocation

    matrix method[23]

    8.104E15 4.600E07

    Fig. 4. (a) Comparison ofkELGk withn = 5, 10, 15 for example 4. (b) Legendre and exact solutions for example 4.

    M. Fathy et al. / Applied Mathematics and Computation 243 (2014) 789800 799

    http://refhub.elsevier.com/S0096-3003(14)00900-X/h0005http://refhub.elsevier.com/S0096-3003(14)00900-X/h0010http://refhub.elsevier.com/S0096-3003(14)00900-X/h0015http://refhub.elsevier.com/S0096-3003(14)00900-X/h0015http://refhub.elsevier.com/S0096-3003(14)00900-X/h0015http://refhub.elsevier.com/S0096-3003(14)00900-X/h0020http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0030http://refhub.elsevier.com/S0096-3003(14)00900-X/h0035http://refhub.elsevier.com/S0096-3003(14)00900-X/h0035http://refhub.elsevier.com/S0096-3003(14)00900-X/h0035http://refhub.elsevier.com/S0096-3003(14)00900-X/h0040http://refhub.elsevier.com/S0096-3003(14)00900-X/h0045http://refhub.elsevier.com/S0096-3003(14)00900-X/h0045http://refhub.elsevier.com/S0096-3003(14)00900-X/h0050http://refhub.elsevier.com/S0096-3003(14)00900-X/h0050http://refhub.elsevier.com/S0096-3003(14)00900-X/h0050http://-/?-http://refhub.elsevier.com/S0096-3003(14)00900-X/h0050http://refhub.elsevier.com/S0096-3003(14)00900-X/h0050http://refhub.elsevier.com/S0096-3003(14)00900-X/h0045http://refhub.elsevier.com/S0096-3003(14)00900-X/h0045http://refhub.elsevier.com/S0096-3003(14)00900-X/h0040http://refhub.elsevier.com/S0096-3003(14)00900-X/h0035http://refhub.elsevier.com/S0096-3003(14)00900-X/h0035http://refhub.elsevier.com/S0096-3003(14)00900-X/h0030http://refhub.elsevier.com/S0096-3003(14)00900-X/h0030http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0025http://refhub.elsevier.com/S0096-3003(14)00900-X/h0020http://refhub.elsevier.com/S0096-3003(14)00900-X/h0015http://refhub.elsevier.com/S0096-3003(14)00900-X/h0015http://refhub.elsevier.com/S0096-3003(14)00900-X/h0010http://refhub.elsevier.com/S0096-3003(14)00900-X/h0005http://refhub.elsevier.com/S0096-3003(14)00900-X/h0005http://-/?-
  • 8/10/2019 LegendreGalerkin Method for the Linear Fredholm

    12/12

    [11] H. Jaradat, O. Alsayyed, S. Al-Shara, Numerical solution of linear integro-differential equations, J. Math. Statist. 4 (2008) 250254.

    [12] A. Khater, A. Shamardan, D. Callebaut, M. Sakran, Numerical solutions of integral and integro-differential equations using legendre polynomials,

    Numer. Algorithms 46 (2007) 195218.

    [13] N. Kurta, M. Sezer, Polynomial solution of high-order linear Fredholm integro-differential equations with constant coefficients, J. Franklin Inst. 345

    (2008) 839850.

    [14] K. Maleknejad, Y. Mahmoudi, Numerical solution of linear Fredholm integral equation by using hybrid Taylor and Block Pulse functions, Appl. Math.

    Comput. 149 (2004) 799806.

    [15] K. Maleknejad, F. Mirzaee, Using rationalized Haar wavelet for solving linear integral equations, Appl. Math. Comput. 160 (2005) 579587 .

    [16] K. Maleknejad, K. Nouri, M. Yousefi, Discussion on convergence of legendre polynomial for numerical solution of integral equations, Appl. Math.

    Comput. 193 (2007) 335339.

    [17] A. Mohsen, M. El-Gamel, A sinc-collocation method for the linear fredholm integro-differential equations, Z. Angew. Math. Phys. 58 (2007) 380390.[18] J. Rashidinia, M. Zarebnia, The numerical solution of integro-differential equation by means of the sinc method, Appl. Math. Comput. 188(2007) 1124

    1130.

    [19] A. Saadatmandia, M. Dehghan, Numerical solution of the higher-order linear Fredholm integro-differential-difference equation with variable

    coefficients, Appl. Math. Comput. 59 (2010) 29963004.

    [20] A. Shahsavara, Numerical solution of linear Volterra and Fredholm integro differential equations using Haar wavelets, Math. Sci. J. 6 (2010) 8596.

    [21] J. Shen, Efficient spectral-Galerkin method i. direct solvers for second- and fourth-order equations by using legendre polynomials, SIAM J. Sci. Comput.

    15 (1994) 14891505.

    [22] S. Yalinbas, M. Sezer, A Taylor collocation method for the approximate solution of general linear Fredholm Volterra integro-difference equations with

    mixed argument, Appl. Math. Comput. 175 (2006) 675690.

    [23] S. Yalinbas, M. Sezer, H. Sorkun, Legendre polynomial solutions of high-order linear fredholm integro-differential equations, Appl. Math. Comput. 210

    (2009) 334349.

    [24] S. Yousefi, Legendre wavelets method for solving differential equations of Laneemden type, Appl. Math. Comput. 181 (2006) 14171422 .

    [25] S. Yousefi, Numerical solution of Abels integral equation by using legendre wavelets, Appl. Math. Comput. 175 (2006) 574580.

    [26] S. Yzbas, M. Sezer, B. Kemanc, Numerical solutions of integro-differential equations and application of a population model with an improved

    Legendre method, Appl. Math. Model. 37 (2013) 20862101.

    [27] S. Shahmorad, Numerical solution of the general form linear FredholmVolterra integro-differential equations by the Tau method with an error

    estimation, Appl. Math. Comput. 167 (2005) 14181429.[28] X. Ye, The legendre collocation method for the CahnHilliard equation, Appl. Math. Model. 150 (2003) 87108.

    [29] W. Zhengsu, C. Yanping, H. Yunqing, Legendre spectral Galerkin method for second-kind Volterra integral equations, Front. Math. China 4 (2009) 181

    193.

    800 M. Fathy et al./ Applied Mathematics and Computation 243 (2014) 789800

    http://refhub.elsevier.com/S0096-3003(14)00900-X/h0055http://refhub.elsevier.com/S0096-3003(14)00900-X/h0060http://refhub.elsevier.com/S0096-3003(14)00900-X/h0060http://refhub.elsevier.com/S0096-3003(14)00900-X/h0065http://refhub.elsevier.com/S0096-3003(14)00900-X/h0065http://refhub.elsevier.com/S0096-3003(14)00900-X/h0070http://refhub.elsevier.com/S0096-3003(14)00900-X/h0070http://refhub.elsevier.com/S0096-3003(14)00900-X/h0075http://refhub.elsevier.com/S0096-3003(14)00900-X/h0080http://refhub.elsevier.com/S0096-3003(14)00900-X/h0080http://refhub.elsevier.com/S0096-3003(14)00900-X/h0085http://refhub.elsevier.com/S0096-3003(14)00900-X/h0090http://refhub.elsevier.com/S0096-3003(14)00900-X/h0090http://refhub.elsevier.com/S0096-3003(14)00900-X/h0095http://refhub.elsevier.com/S0096-3003(14)00900-X/h0095http://refhub.elsevier.com/S0096-3003(14)00900-X/h0100http://refhub.elsevier.com/S0096-3003(14)00900-X/h0105http://refhub.elsevier.com/S0096-3003(14)00900-X/h0105http://refhub.elsevier.com/S0096-3003(14)00900-X/h0105http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0120http://refhub.elsevier.com/S0096-3003(14)00900-X/h0125http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0135http://refhub.elsevier.com/S0096-3003(14)00900-X/h0135http://refhub.elsevier.com/S0096-3003(14)00900-X/h0140http://refhub.elsevier.com/S0096-3003(14)00900-X/h0145http://refhub.elsevier.com/S0096-3003(14)00900-X/h0145http://refhub.elsevier.com/S0096-3003(14)00900-X/h0145http://refhub.elsevier.com/S0096-3003(14)00900-X/h0145http://refhub.elsevier.com/S0096-3003(14)00900-X/h0145http://refhub.elsevier.com/S0096-3003(14)00900-X/h0140http://refhub.elsevier.com/S0096-3003(14)00900-X/h0135http://refhub.elsevier.com/S0096-3003(14)00900-X/h0135http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0130http://refhub.elsevier.com/S0096-3003(14)00900-X/h0125http://refhub.elsevier.com/S0096-3003(14)00900-X/h0120http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0115http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0110http://refhub.elsevier.com/S0096-3003(14)00900-X/h0105http://refhub.elsevier.com/S0096-3003(14)00900-X/h0105http://refhub.elsevier.com/S0096-3003(14)00900-X/h0100http://refhub.elsevier.com/S0096-3003(14)00900-X/h0095http://refhub.elsevier.com/S0096-3003(14)00900-X/h0095http://refhub.elsevier.com/S0096-3003(14)00900-X/h0090http://refhub.elsevier.com/S0096-3003(14)00900-X/h0090http://refhub.elsevier.com/S0096-3003(14)00900-X/h0085http://refhub.elsevier.com/S0096-3003(14)00900-X/h0080http://refhub.elsevier.com/S0096-3003(14)00900-X/h0080http://refhub.elsevier.com/S0096-3003(14)00900-X/h0075http://refhub.elsevier.com/S0096-3003(14)00900-X/h0070http://refhub.elsevier.com/S0096-3003(14)00900-X/h0070http://refhub.elsevier.com/S0096-3003(14)00900-X/h0065http://refhub.elsevier.com/S0096-3003(14)00900-X/h0065http://refhub.elsevier.com/S0096-3003(14)00900-X/h0060http://refhub.elsevier.com/S0096-3003(14)00900-X/h0060http://refhub.elsevier.com/S0096-3003(14)00900-X/h0055