One-sided and two-sided Green’s functions

  • Rubens de Figueiredo Camargo1Email author,

    Affiliated with

    • Ary Orozimbo Chiacchio2 and

      Affiliated with

      • Edmundo Capelas de Oliveira2

        Affiliated with

        Boundary Value Problems20132013:45

        DOI: 10.1186/1687-2770-2013-45

        Received: 16 October 2012

        Accepted: 14 February 2013

        Published: 6 March 2013

        Abstract

        We discuss the one-sided Green’s function, associated with an initial value problem and the two-sided Green’s function related to a boundary value problem. We present a specific calculation associated with a differential equation with constant coefficients. For both problems, we also present the Laplace integral transform as another methodology to calculate these Green’s functions and conclude which is the most convenient one. An incursion in the so-called fractional Green’s function is also presented. As an example, we discuss the isotropic harmonic oscillator.

        1 Introduction

        There are several methods to discuss a second-order linear partial differential equation. Among them we mention the simplest one, the method of separation of variables, and the method of integral transforms, particularly the Laplace transform, which is in many cases most convenient [1].

        On the other hand, after the method of separation of variables, we get the following general second-order linear ordinary differential equation:
        a 1 ( x ) d 2 d x 2 y ( x ) + a 2 ( x ) d d x y ( x ) + a 3 ( x ) y ( x ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equa_HTML.gif
        on the interval a < x < b http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq1_HTML.gif, whose corresponding nonhomogeneous one is given by
        a 1 ( x ) d 2 d x 2 y ( x ) + a 2 ( x ) d d x y ( x ) + a 3 ( x ) y ( x ) = F ( x ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equb_HTML.gif
        where F ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq2_HTML.gif is a forcing term. Assuming that a 1 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq3_HTML.gif is a continuously differentiable positive function on this interval and that a 2 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq4_HTML.gif and a 3 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq5_HTML.gif are continuous functions, we can write the above ordinary differential equation as follows:
        d d x [ p ( x ) d d x y ( x ) ] + q ( x ) y ( x ) = f ( x ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ1_HTML.gif
        (1)

        the so-called self-adjoint form known also as an ordinary differential equation in the Sturm-Liouville form. In this equation, p ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq6_HTML.gif and q ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq7_HTML.gif are continuous functions that are related to the coefficients a 1 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq3_HTML.gif, a 2 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq4_HTML.gif and a 3 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq5_HTML.gif. The nonhomogeneous term, f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif, is also related to F ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq2_HTML.gif [1].

        The methodology of the Laplace integral transform is adequate to discuss the one-sided Green’s functiona because the initial conditions, in general, are given in terms of the own function and the first derivative. A simple question arises when we discuss the two-sided Green’s function associated with a problem involving boundary conditions, i.e., is the Laplace transform methodology convenient to discuss this problem? The answer depends on the sort of problem we are studying, as we will see in the following sections.

        On the other hand, the fractional harmonic oscillator was discussed in a series of papers by Narahari et al. [25] where they presented the dynamic of the fractional harmonic oscillator, including also the damping, and by Tofighi [6] who discusses the intrinsic damping.

        In this paper we discuss Eq. (1) associated with an initial value problem and a boundary value problem. In both cases, we present two methodologies, the Laplace integral transform and the Green’s function methodology. After that we conclude which methodology is the most convenient one. We sum the paper up presenting the corresponding fractional case where we discuss the Green’s function associated with the fractional harmonic oscillator. Finally, we present our concluding remarks.

        2 One-sided Green’s function

        To solve the self-adjoint differential equation, we introduce the so-called one-sided Green’s function, also called influence function, a two-parameter function, denoted by G i ( x | ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq9_HTML.gif, which describes the influence of a disturbance (also known as impulse) on the value of y ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq10_HTML.gif at a point x, concentrated at the point ξ. If we know the Green’s function, as we construct below, the solution of an initial value problem composed of the self-adjoint differential equation and the initial conditions can be written as follows:
        y ( x ) = a x G i ( x | ξ ) f ( ξ ) d ξ , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ2_HTML.gif
        (2)

        where y ( a ) = 0 = y ( a ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq11_HTML.gif.

        For a fixed value of ξ, the one-sided Green’s function is the solution of the corresponding homogeneous initial value problem, i.e., for x > ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq12_HTML.gif we have the homogeneous ordinary differential equation
        d d x [ p ( x ) d d x G i ( x | ξ ) ] + q ( x ) G i ( x | ξ ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equc_HTML.gif
        and the initial conditions
        G i ( ξ | ξ ) = 0 and d d x G i ( x | ξ ) | x = ξ = 1 p ( ξ ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equd_HTML.gif

        2.1 Constant coefficients

        As a particular case, we consider the problem involving an ordinary differential equation with constant coefficients, which can represent a problem related to the classical harmonic oscillator and transmission lines, for example,
        d 2 d x 2 y ( x ) + 2 b d d x y ( x ) + ( a 2 + b 2 ) y ( x ) = f ( x ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ3_HTML.gif
        (3)

        where a and b are positive constants, and the homogeneous initial conditions are given by y ( 0 ) = 0 = y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq13_HTML.gif.

        First of all, we write the ordinary differential equation in the corresponding self-adjoint ordinary differential equation
        d d x [ e 2 b x d d x y ( x ) ] + e 2 b x ( a 2 + b 2 ) y ( x ) = e 2 b x f ( x ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Eque_HTML.gif

        where we identify p ( x ) exp ( 2 b x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq14_HTML.gif.

        Thus, the one-sided Green’s function, denoted by G i ( x | ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq15_HTML.gif, satisfies the following homogeneous ordinary differential equation for ξ fixed:
        d d x [ e 2 b x d d x G i ( x | ξ ) ] + e 2 b x ( a 2 + b 2 ) G i ( x | ξ ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equf_HTML.gif
        and the initial conditions
        G i ( ξ | ξ ) = 0 and d d x G i ( x | ξ ) | x = ξ = exp ( 2 b ξ ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equg_HTML.gif
        Two linearly independent solutions of the homogeneous ordinary differential equation are given by
        y 1 ( x ) = e b x a sin a x and y 2 ( x ) = e b x a cos a x http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equh_HTML.gif
        which furnishes for the general solution (the Green’s function)
        G i ( x | ξ ) = A ( ξ ) a e b x sin a x + B ( ξ ) a e b x cos a x , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equi_HTML.gif
        where A ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq16_HTML.gif and B ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq17_HTML.gif must be calculated by means of the initial conditions, i.e., by the following system:
        http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equj_HTML.gif
        Solving the system above, we obtain
        http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equk_HTML.gif
        which furnish for the one-sided Green’s function, for x > ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq18_HTML.gif,
        G i ( x | ξ ) = exp [ b ( x + ξ ) ] a sin [ a ( x ξ ) ] http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ4_HTML.gif
        (4)

        and, as we know, satisfies the property G i ( x | ξ ) = G i ( ξ | x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq19_HTML.gif.

        Finally, the solution of our initial value problem is given by
        y ( x ) = 1 a 0 x e b ( x ξ ) sin [ a ( x ξ ) ] f ( ξ ) d ξ . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ5_HTML.gif
        (5)

        2.2 Laplace transform

        Another way to calculate this one-sided Green’s function is by means of the methodology of Laplace transform. Multiplying Eq. (3) by the kernel of Laplace transform, exp ( s x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq20_HTML.gif with Re ( s ) > 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq21_HTML.gif and integrating from zero to infinity, we have
        s 2 F ( s ) s y ( 0 ) y ( 0 ) + 2 b [ s F ( s ) y ( 0 ) ] + ( a 2 + b 2 ) F ( s ) = G ( s ) , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equl_HTML.gif
        where
        F ( s ) = 0 e s x y ( x ) d x and G ( s ) = 0 e s x f ( x ) d x . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equm_HTML.gif
        Using the initial conditions, we obtain an algebraic equation whose solution is given by
        F ( s ) = G ( s ) ( s + b ) 2 + a 2 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equn_HTML.gif
        By means of the convolution product and the relation involving the inverse Laplace transform
        L 1 [ 1 ( s + b ) 2 + a 2 ] = exp ( b x ) a sin a x , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equo_HTML.gif
        we get
        y ( x ) = 1 a 0 x e b ( x ξ ) sin [ a ( x ξ ) ] f ( ξ ) d ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equp_HTML.gif

        which is exactly Eq. (5).

        3 Two-sided Green’s function

        In the case that we have a two-point boundary value problem, i.e., when the boundary conditions are fixed on the extremes of the interval, we have a two-sided Green’s function, also called Green’s function, only.

        We consider the second-order ordinary differential equation in the Sturm-Liouville form
        d d x [ p ( x ) d d x y ( x ) ] + q ( x ) y ( x ) = f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equq_HTML.gif

        for x 0 < x < x 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq22_HTML.gif and the boundary conditions y ( x 0 ) = 0 = y ( x 1 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq23_HTML.gif, and comparing this problem with the corresponding one-sided Green’s function, we have put f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq24_HTML.gif in the place of f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif for convenience only.

        The two-sided Green’s function, denoted by G b ( x | ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq25_HTML.gif for each ξ, satisfies the following problem consisting of homogeneous ordinary differential equation
        d d x [ p ( x ) d d x G b ( x | ξ ) ] + q ( x ) G b ( x | ξ ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equr_HTML.gif
        for x ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq26_HTML.gif and the homogeneous boundary conditions
        G b ( x 0 | ξ ) = 0 = G b ( x 1 | ξ ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equs_HTML.gif
        This Green’s function must satisfy also the continuity
        G b ( ξ + | ξ ) = G b ( ξ | ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equt_HTML.gif
        and a jump discontinuity on the first derivative [1]
        d d x G b ( x | ξ ) | x = ξ + d d x G b ( x | ξ ) | x = ξ = 1 p ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equu_HTML.gif
        The solution y ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq10_HTML.gif can be interrelated as a displacement and be given by
        y ( x ) = x 0 x 1 G b ( x | ξ ) f ( ξ ) d ξ , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equv_HTML.gif

        where f ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq27_HTML.gif is a force per unit length. G b ( x | ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq25_HTML.gif is the displacement at x due to a force of unit magnitude concentrated at ξ. In this case, we have G b ( x | ξ ) = G b ( ξ | x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq28_HTML.gif, the so-called reciprocity law.

        3.1 Constant coefficients

        As an example, we discuss the following ordinary differential equation:
        d 2 d x 2 y ( x ) + 2 b d d x y ( x ) + ( a 2 + b 2 ) y ( x ) = f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equw_HTML.gif

        with the homogeneous boundary conditions y ( x 0 ) = 0 = y ( x 1 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq29_HTML.gif.

        First, we construct the corresponding Green’s function, as we have seen before. Two linearly independent solutions of the homogeneous ordinary differential equation are
        y 1 ( x ) = e b x a sin a x and y 2 ( x ) = e b x a cos a x . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equx_HTML.gif
        Imposing that y 1 ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq30_HTML.gif satisfies y 1 ( x 0 ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq31_HTML.gif, we can write
        y 1 ( x ) = A ( ξ ) a e b x sin [ a ( x x 0 ) ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equy_HTML.gif
        On the other hand, to satisfy y 2 ( x 1 ) = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq32_HTML.gif, we put
        y 2 ( x ) = B ( ξ ) a e b x sin [ a ( x x 1 ) ] , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equz_HTML.gif

        where A ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq16_HTML.gif and B ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq17_HTML.gif must be determined.

        Using the continuity at x = ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq33_HTML.gif and the jump discontinuity of the first derivative, we obtain a system of two algebraic equations involving A ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq16_HTML.gif and B ( ξ ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq17_HTML.gif whose solution is
        A ( ξ ) = sin [ a ( x 1 ξ ) ] sin [ a ( x 1 x 0 ) ] and B ( ξ ) = sin [ a ( ξ x 0 ) ] sin [ a ( x 1 x 0 ) ] http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equaa_HTML.gif
        Thus, the Green’s function is given by the following expression:
        G b ( x | ξ ) = exp [ b ( x + ξ ) ] a sin [ a ( x 1 x 0 ) ] { sin [ a ( x x 0 ) ] sin [ a ( x 1 ξ ) ] , x 0 < x < ξ , sin [ a ( ξ x 0 ) ] sin [ a ( x 1 x ) ] , ξ < x < x 1 . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equab_HTML.gif
        Using this Green’s function, the solution of the boundary value problem can be written as follows:
        y ( x ) = x 0 x 1 G b ( x | ξ ) f ( ξ ) d ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equac_HTML.gif

        3.2 Laplace transform

        As we have already said, the Laplace transform converts the differential equation with constant coefficients into an algebraic equation whose solution is
        F ( s ) = s y ( 0 ) + y ( 0 ) ( s + b ) 2 + a 2 G ( s ) ( s + b ) 2 + a 2 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equad_HTML.gif

        where F ( s ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq34_HTML.gif and G ( s ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq35_HTML.gif are the corresponding Laplace transforms of y ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq10_HTML.gif and f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif, respectively.

        Using the inverse Laplace transform and the convolution product, we get
        y ( x ) = y ( 0 ) e b x cos a x a + y ( 0 ) e b x sin a x a 1 a 0 x e b ( x ξ ) sin [ a ( x ξ ) ] f ( ξ ) d ξ . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equae_HTML.gif
        Thus, substituting the boundary conditions, we obtain two algebraic equations, a system involving y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq36_HTML.gif and y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq37_HTML.gif. Solving this algebraic system, we obtain
        y ( x ) = 1 / a sin [ a ( x 1 x 0 ) ] x 0 x e b ( x + t ) sin [ a ( x 1 x ) ] sin [ a ( x 0 t ) ] f ( t ) d t 1 / a sin [ a ( x 1 x 0 ) ] x x 1 e b ( x + t ) sin [ a ( x 1 t ) ] sin [ a ( x 0 x ) ] f ( t ) d t http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equaf_HTML.gif
        which can be rewritten as follows:
        y ( x ) = x 0 x 1 G b ( x | t ) f ( t ) d t , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equag_HTML.gif
        where the Green’s function is given by
        G b ( t | x ) = exp [ b ( x + t ) ] a sin [ a ( x 1 x 0 ) ] { sin [ a ( x 1 x ) ] sin [ a ( t x 0 ) ] , x 0 < t < x , sin [ a ( x 1 t ) ] sin [ a ( x x 0 ) ] , x < t < x 1 , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equah_HTML.gif

        which is the same expression as that obtained in Section 3.1.

        At this point we conclude that for a problem involving initial conditions, the Laplace integral transform is more convenient since y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq36_HTML.gif and y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq37_HTML.gif are known. On the other hand, i.e., for a problem involving boundary conditions, the Sturm-Liouville, as opposed to the Laplace integral transform, is more convenient in the sense that the calculation is much more simple.

        4 Fractional Green’s function

        Fractional calculus is one of the most accurate tools to refine the description of natural phenomena. The usual way to use this tool is to replace the integer-order derivatives of the partial differential equation that describes one specific phenomenon by a derivative of non-integer order. For many expected reasons, the solution of a fractional partial differential equation used to be much more complicated than the solution of the corresponding integer-order partial differential equation.

        On the other hand, many important results and generalizations were obtained using this procedure in several areas such as fluid flow, diffuse transport, electrical networks, probability, biomathematics and others [712]. Here, as a generalization to the integer case, we present a calculation associated with the so-called fractional one-sided and two-sided Green’s function relative to the fractional differential equation with constant coefficients, i.e., we obtain the fractional Green’s function for the fractional differential equation whose coefficients are constants. We discuss the problem by means of the Laplace integral transform, and as an application, we present explicitly the Green’s function associated with the fractional harmonic oscillator.

        4.1 Fractional one-sided Green’s function

        Let a, b and c be real constants. We present the solution of the fractional differential equation
        http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ6_HTML.gif
        (6)

        where 1 < α 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq38_HTML.gif and 0 < β 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq39_HTML.gif and the fractional derivatives are taken in the Caputo sense [13]. We also consider y ( 0 ) = 0 = y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq13_HTML.gif as the initial conditions. In the case where α = 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq40_HTML.gif and β = 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq41_HTML.gif, we recover the results associated with the integer case, and taking a = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq42_HTML.gif and b 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq43_HTML.gif, we recover the equation associated with the fractional relaxor-oscillator as discussed in [5].

        Introducing the Laplace transform and using the initial conditions, we obtain an algebraic equation whose solution can be written as follows:
        L [ y ( x ) ] = F ( s ) a s α + b s β + c , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equai_HTML.gif
        where F ( s ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq34_HTML.gif is the Laplace transform of the f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif. This expression can be manipulated, using the geometric series, to obtain
        L [ y ( x ) ] = F ( s ) a k = 0 ( b a ) k s β k ( s α + c / a ) k + 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equaj_HTML.gif

        which is valid for | s λ c / a | < 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq44_HTML.gif.

        Using the Laplace transform of the generalized Mittag-Leffler function and its corresponding inverse [13],
        L [ t μ 1 E ν , μ ρ ( λ t ν ) ] = s ν ρ μ ( s ν λ ) ρ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equak_HTML.gif
        with Re ( s ) > 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq21_HTML.gif, Re ( μ ) > 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq45_HTML.gif, λ C http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq46_HTML.gif and | λ s ν | < 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq47_HTML.gif, we have
        y ( x ) = 1 a k = 0 ( b a ) k 0 x f ( ξ ) ( x ξ ) ( α β ) k + α 1 E α , ( α β ) k + α k + 1 [ c a ( x ξ ) α ] d ξ http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equal_HTML.gif

        which is the solution of the fractional differential equation.

        Thus, the one-sided fractional Green’s function can be written as follows:
        G ( x | ξ ) = 1 a ( x ξ ) α 1 E α , α [ c a ( x ξ ) α ] + 1 a k = 1 ( b a ) k ( x ξ ) ( α β ) k + α 1 E α , ( α β ) k + α k + 1 [ c a ( x ξ ) α ] . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equ7_HTML.gif
        (7)
        Taking b = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq48_HTML.gif, a = 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq49_HTML.gif and c = ω 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq50_HTML.gif in Eq. (7), we get
        G ( x | ξ ) = ( x ξ ) α 1 E α , α [ ω 2 ( x ξ ) α ] http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equam_HTML.gif
        which is the fractional one-sided Green’s function associated with the fractional harmonic oscillator. The one-sided Green’s function associated with the classical harmonic oscillator is recovered by introducing α = 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq40_HTML.gif in the last equation, i.e.,
        G ( x | ξ ) = ( x ξ ) E 2 , 2 [ ω 2 ( x ξ ) 2 ] , x > ξ , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equan_HTML.gif
        which can also be written as follows:
        G ( x | ξ ) = 1 ω sin [ ω ( x ξ ) ] , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equao_HTML.gif

        which is the same expression as that obtained in Eq. (4) in the case b = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq48_HTML.gif.

        To conclude this section, we plot, in Figure 1, a graphic t = x ξ × t α 1 E α , α ( t α ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq51_HTML.gif for particular values of the parameter α to compare with the sine function. This graphic is plotted using the program (MatLab R2009a). For reader interested in the Mittag-Leffler function, we suggest a recent nice book [14] and the paper [15] particularly regarding the asymptotic algebraic behavior of this function.
        http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Fig1_HTML.jpg
        Figure 1

        Graphics for f ( t ) = t α 1 E α , α ( t α ) × t http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq52_HTML.gif with t x ξ > 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq53_HTML.gif .

        4.2 Fractional two-sided Green’s function

        Let a, b and c be real constants. We present the solution of the fractional differential equation, Eq. (6), with the homogeneous boundary conditions, y ( x 0 ) = 0 = y ( x 1 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq29_HTML.gif.

        By means of the Laplace transform, we can write
        L [ y ( x ) ] = F ( s ) + s α 1 y ( 0 ) + s α 2 y ( 0 ) a s α + b s β + c , http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equap_HTML.gif
        where F ( s ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq34_HTML.gif is the Laplace transform of the f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif. This algebraic equation can be manipulated as follows:
        L [ y ( x ) ] = F ( s ) a k = 0 ( b a ) k s β k ( s α + c / a ) k + 1 + y ( 0 ) a k = 0 ( b a ) k s β k + α 1 ( s α + c / a ) k + 1 + y ( 0 ) a k = 0 ( b a ) k s β k + α 2 ( s α + c / a ) k + 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equaq_HTML.gif
        From the inverse Laplace transform and the convolution theorem, we get
        y ( x ) = 1 a k = 0 ( b a ) k 0 x f ( x ξ ) ξ ( α β ) k + α 1 E α , ( α β ) k + α k + 1 ( c a ξ α ) d ξ + y ( 0 ) a k = 0 ( b a ) k L 1 [ s β k + α 1 ( s α + c / a ) k + 1 ] + y ( 0 ) a k = 0 ( b a ) k L 1 [ s β k + α 2 ( s α + c / a ) k + 1 ] http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equar_HTML.gif
        which can be rewritten in the following way:
        y ( x ) = 1 a k = 0 ( b a ) k 0 x f ( ξ ) ( x ξ ) ( α β ) k + α 1 E α , ( α β ) k + α k + 1 [ c a ( x ξ ) α ] d ξ + y ( 0 ) a k = 0 ( b a ) k x ( α β ) k E α , ( α β ) k + 1 k + 1 ( c a x α ) + y ( 0 ) a k = 0 ( b a ) k x ( α β ) k + 1 E α , ( α β ) k + 2 k + 1 ( c a x α ) . http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_Equas_HTML.gif

        As we have already said in Section 3, to explicitly calculate the solution, we must substitute the homogeneous boundary conditions in the last equation and determine the y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq36_HTML.gif and y ( 0 ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq37_HTML.gif. Finally, we can get the respective fractional two-sided Green’s function. We have shown this is a hard calculation. To obtain the results associated with the fractional harmonic oscillator, we introduce b = 0 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq48_HTML.gif, a = 1 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq49_HTML.gif and c = ω 2 http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq54_HTML.gif in the last equation. Remember that to obtain the respective Green’s function, we substitute f ( x ) http://static-content.springer.com/image/art%3A10.1186%2F1687-2770-2013-45/MediaObjects/13661_2012_Article_297_IEq8_HTML.gif to the corresponding delta function.

        5 Concluding remarks

        We have presented and discussed the so-called one-sided and two-sided Green’s function to study, respectively, an initial value problem and a boundary value problem. Besides that, we studied the same problems by means of the Laplace transform methodology in order to conclude which methodology was most accurate for each problem. We also obtained the fractional generalization of the one-sided and two-sided Green’s function in terms of the generalized Mittag-Leffler function.

        We conclude that for the initial value problem, the Laplace integral transform methodology is more convenient; on the other hand, for the boundary value problem, the two-sided Green’s function provides a much more simple calculation. It is important to note that in the present manuscript we did not consider the problem involving physical dimensions. This problem has been discussed in a recent paper by Inizan [16].

        A natural continuation of this work would be to study the problems involving partial differential equations with non-constant coefficients and their fractional versions, which could provide a better description of the phenomena related to those equations. Study in this direction is upcoming.

        Endnote

        a A recent historical review about George Green can be found in ref. [17].

        Declarations

        Acknowledgements

        We are grateful to Prof. J. Vaz Jr., Dr. J. Emílio Maiorino and Dr. E. Contharteze Grigoletto for several useful discussions. Besides that, we are thankful to the referees for several important suggestions which improved this article a lot.

        Authors’ Affiliations

        (1)
        Departamento de Matemática, Faculdade de Ciências
        (2)
        Departamento de Matemática, Imecc, Unicamp

        References

        1. Capelas de Oliveira E: Special Functions and Applications. 2nd edition. Editora Livraria da Física, São Paulo; 2012. (in Portughese)
        2. Clarke T, Narahari Achar BN, Hanneken JW: Mittag-Leffler functions and transmission lines. J. Mol. Liq. 2004, 114: 159-163. 10.1016/j.molliq.2004.02.014View Article
        3. Narahari Achar BN, Hanneken JW, Clarke T: Response characteristics of a fractional oscillator. Physica A 2002, 309: 275-288. 10.1016/S0378-4371(02)00609-XView Article
        4. Narahari Achar BN, Hanneken JW, Clarke T: Damping characteristic of a fractional oscillator. Physica A 2004, 339: 311-319. 10.1016/j.physa.2004.03.030MathSciNetView Article
        5. Narahari Achar BN, Hanneken JW, et al.: Dynamic response of the fractional relaxor-oscillator to a harmonic driving force. In Advances in Fractional Calculus: Theoretical Developments and Applications in Physics and Engineering. Edited by: Sabatier J. Springer, Heidelberg; 2007.
        6. Tofighi A: The intrinsic damping of the fractional oscillator. Physica A 2003, 329: 29-34. 10.1016/S0378-4371(03)00598-3View Article
        7. Capelas de Oliveira E, Mainardi F, Vaz J Jr.: Models based on Mittag-Leffler functions for anomalous relaxation in dielectrics. Eur. Phys. J. Spec. Top. 2011, 193: 161-171. 10.1140/epjst/e2011-01388-0View Article
        8. Capelas de Oliveira E, Vaz J Jr.: Tunneling in fractional quantum mechanics. J. Phys. A, Math. Theor. 2011., 44: Article ID 185303
        9. Figueiredo Camargo R, Charnet R, Capelas de Oliveira E: On the fractional Green function. J. Math. Phys. 2009., 50: Article ID 043514
        10. Kilbas AA, Srivastava HM, Trujillo JJ Mathematics Studies 204. In Theory and Applications of Fractional Differential Equations. Elsevier, Amsterdam; 2006. Edited by Jan Van Mill
        11. Klafter J, Lim SC, Metzler R (Eds): Fractional Dynamics, Recent Advances. World Scientific, Singapore; 2011.
        12. De Paoli AL, Rocca MC: The fractionary Schrödinger equation, Green functions and ultradistribution. Phys. A, Stat. Mech. Appl. 2013, 392: 111-122. 10.1016/j.physa.2012.08.017MathSciNetView Article
        13. Figueiredo Camargo R, Chiacchio AO, Capelas de Oliveira E: Differentiation to fractional orders and the fractional telegraph equation. J. Math. Phys. 2008., 49: Article ID 033505
        14. Mainardi F: Fractional Calculus and Waves in Linear Viscoelasticity. Imperial College Press, London; 2010.View Article
        15. Gorenflo R, Mainardi F: Fractional calculus, integral and differential equations of fractional order. In Fractals and Fractional Calculus in Continuum Mechanics. Edited by: Carpinteri A, Mainardi F. Springer, Wien; 1997:223-276. (E-print arxiv:0805.3823), in particular, pp. 247-252View Article
        16. Inizan P: Homogeneous fractional embeddings. J. Math. Phys. 2008., 49: Article ID 082901
        17. Challis L, Sheard F: The Green of Green function. Phys. Today 2003, 56: 41-46.View Article

        Copyright

        © Camargo et al.; licensee Springer. 2013

        This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://​creativecommons.​org/​licenses/​by/​2.​0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.