Home Super-Exponentially Convergent Parallel Algorithm for Eigenvalue Problems with Fractional Derivatives
Article Publicly Available

Super-Exponentially Convergent Parallel Algorithm for Eigenvalue Problems with Fractional Derivatives

  • Ihor Demkiv , Ivan P. Gavrilyuk EMAIL logo and Volodymyr L. Makarov
Published/Copyright: May 11, 2016

Abstract

A new algorithm for eigenvalue problems for linear differential operators with fractional derivatives is proposed and justified. The algorithm is based on the approximation (perturbation) of the coefficients of a part of the differential operator by piecewise constant functions where the eigenvalue problem for the last one is supposed to be simpler than the original one. Another milestone of the algorithm is the homotopy idea which results at the possibility for a given eigenpair number to compute recursively a sequence of the approximate eigenpairs. This sequence converges to the exact eigenpair with a super-exponential convergence rate. The eigenpairs can be computed in parallel for all prescribed indexes. The proposed method possesses the following principal property: its convergence rate increases together with the index of the eigenpair. Numerical examples confirm the theory.

1 Introduction

It was recognized over the last decades that realistic models of various physical phenomena can be better described with fractional calculus; see, e.g., [44, 32, 31, 50, 25, 8, 33, 47, 20], just to mention a few. A very good overview of applications of fractional calculus is given in [49, 45].

There are various definitions of fractional derivatives; see, e.g., [37, 36, 34, 38, 32, 31, 1, 49]. In [46], a fractional derivative was defined in various ways, especially through integer derivatives using the Taylor and Fourier series. For example, one can introduce fractional derivatives for periodic functions using the Fourier series and the elementary relations

DnF(sinx)=(sinx)(n)=sin(x+nπ2),DnF(cosx)=(cosx)(n)=cos(x+nπ2),n,

and replacing here the integer n by a real number. Alternatively one can use the Fourier or Laplace transform with the relations

(Dαf)(ξ)=i|α|ξα(f)(ξ),(Dαf)(ξ)=sα(f)-k=0n-1f(k)(0)sα-k-1

and declare them valid for all real values of α.

The very popular definitions of the fractional calculus are the Riemann–Liouville and Caputo fractional derivatives and integrals; see, e.g., [10].

The Riemann–Liouville integral is defined by

Iαf(x)=1Γ(α)axf(t)(x-t)α-1𝑑t,

where Γ(α) is the gamma function and a is a fixed base point. Another notation, which emphasizes the base point, is

(1.1)Dx-αaf(x)=Ixαaf(x)=1Γ(α)axf(t)(x-t)α-1𝑑t.

The following fundamental relations hold:

ddxIα+1f(x)=Iαf(x),Iα(Iβf)=Iα+βf,

the latter of which is the semigroup property. These properties make possible not only the definition of fractional integration, but also the definition of fractional differentiation, by taking enough derivatives of Iαf(x). Computing the n-th order derivative over the integral of order (n-α), the α order derivative is obtained:

DtαaRLf(t)=dndtnDt-(n-α)af(t)=dndtnItn-αaf(t)=1Γ(n-α)dndtnatf(τ)(t-τ)n-α-1𝑑τ,

where n is the nearest integer bigger than α, i.e. n-1α<n+ or n=α+1.

The Caputo fractional derivative is another option for computing fractional derivatives. It was introduced by M. Caputo in 1967 in the following way:

DtαaCf(t)=1Γ(n-α)at(t-τ)n-α-1f(n)(τ)𝑑τ,

where n-1α<n+. From these definitions, one can see that DtαaCf(n)(t)=Dtn+αaCf(t) and, for sufficient smooth f,

(1.2)DtαaCf(t)=DtαaRLf(t)-k=0n-1tk-αf(k)(a)Γ(k-α+1),

i.e., both derivatives coincide for functions f(t) with f(k)(a)=0, k=0,1,,n-1. One can also rewrite this formula as

DtαaCf(t)=DtαaRL(f(t)-k=0n-1tkf(k)(a)k!).

Both definitions differ only in the order of evaluation: whereas in the Caputo definition we first compute an ordinary derivative, then a fractional integral, in the Riemann–Liouville definition the operators are reversed. Comparing these definitions, we can see that functions which are derivable in the Caputo sense are much “fewer” than those which are derivable in the Riemann–Liouville sense.

Note that the fractional derivatives introduced above do not satisfy many properties of the classical differential calculus. But there are other definitions of fractional derivative in the literature which obey classical properties including: linearity, product rule, quotient rule, power rule, chain rule, vanishing derivatives for constant functions, Rolle’s theorem and the Mean Value Theorem; see, e.g., [22].

Using the definitions above, one can consider differential operators of fractional order, boundary and eigenvalue problems for these operators, etc., as well as various approximation methods for them; see, e.g., [31, 14, 11, 21].

The eigenvalue problem (EVP) is the problem of finding eigenpairs (eigenvalues and eigenfunctions or, in the language of mechanicians, frequencies and vibration shapes). It plays an important role in various applications concerned with vibrations and wave processes [19, 4, 41]. Popular methods such as the finite-difference method (FD), the finite element (FEM) and other variational methods, as well as spectral methods allow one to compute efficiently some lower eigenvalues only. At the same time there are applied problems requiring the computation of a great number (hundreds of thousands) of eigenvalues and eigenfunctions including eigenpairs with great indexes; see, e.g., [41, p. 273].

Over the last decade, it has been demonstrated that also eigen-oscillations of many systems in science and engineering can be modeled more accurately by employing fractional-order rather than integer-order derivatives [3, 13, 35]. In most of the fractional Sturm–Liouville formulations presented recently, the ordinary derivatives in a traditional Sturm–Liouville problem are replaced with fractional derivatives, and the resulting problems are solved using some numerical schemes such as the Adomian decomposition method [3], the fractional differential transform method [13], or using the method of the Haar wavelet operational matrix [35]. Some of the proposed algorithms are given in the literature without sufficient theoretical justification or possess the same drawbacks as the corresponding algorithms for the classical Sturm–Liouville problem. It turns out that the Sturm–Liouville problem with fractional derivatives can possess similar qualitative properties as a traditional Sturm–Liouville problem [51, 12], but some qualitative properties differ from the traditional ones.

In [9], Chen, Shen and Wang consider a spectral approximation of fractional differential equations (FDEs). A new class of generalized Jacobi functions (GJFs) is defined, which are the eigenfunctions of some fractional Jacobi-type differential operator and can serve as natural basis functions for properly designed spectral methods for FDEs. The efficient GJF Petrov–Galerkin methods for a class of prototypical fractional initial value problems (FIVPs) and fractional boundary value problems (FBVPs) of general order are constructed and analyzed.

In order to solve numerically an eigenvalue problem for fractional differential operators, we propose a new approach described below which we will refer to as the FD-method (from “functional-discrete method”, following [26, 27, 28, 5, 6]). Note that this approach for eigenvalue problems for nonlinear differential equations was applied for the first time in [28] and then continued in [15].

The FD-method is based on the perturbation and homotopy ideas. The perturbation in the case of ODE operators can be similar to that of the butt method (metodo dei tronconi; see, e.g., [7]), the Pruess method for BVPs [41, 39, 40] for the second-order ODEs, or of some methods for EVP from [2], where the coefficients of the differential equation are replaced by their piecewise constant approximations. This approach has been applied also to EVPs with multiple eigenvalues in [17].

The article is organized as follows. In Section 2, we describe the algorithm of the FD-method for EVPs in an abstract setting. Section 3 is devoted to the Fourier fractional derivative and some of its properties. In Section 4, we apply the FD-method for a differential equation with an integer highest derivative and a subordinated fractional derivative. Here we prove the superexponential convergence rate of our method independent of the definition of fractional derivatives in use and discuss two various algorithmic realizations: 1) solving the differential problems for the corrections of the FD-method, and 2) using a recursive procedure for the expansion coefficients of these corrections. Section 5 deals with the FD-method for a differential equation with a highest Riemann–Liouville fractional derivative, which is similar to the differential equation defining the generalized Jacobi functions (GJF). We prove the super-exponential convergence rate of our method. For the practical implementation we test the FD-algorithm directly and, besides, propose some new recursive procedure. Numerical examples are given to support the theoretical results.

2 The Homotopy-Based Method for EVPs in an Abstract Setting

Let us briefly explain the ideas of perturbation and homotopy for the eigenvalue problem

(2.1)(A+B)un-λnun=θ,

in a Hilbert space X with the scalar product (,) and with the null-element θ under the assumption that the spectrum of the operator A+B is discrete. We are looking for the eigenpair with a given fixed index n.

Let B¯ be an approximating operator for B in the sense that the eigenvalue problem

(2.2)(A+B¯)un(0)-λn(0)un(0)=θ

is “simpler” than problem (2.1).

Formally, a homotopy between two problems P1 and P2 with solutions u1 and u2 from some topological space X is defined to be a parametric problem PH(t) with a solution u(t) continuously and smoothly depending on the parameter t[0,1] and such that u(0)=u1 and u(1)=u2 (compare http://en.wikipedia.org/wiki/homotopy).

Following the homotopy idea for a given eigenpair number n, we embed our problem into the parametric family of problems

(2.3)(A+W(t))un(t)-λn(t)un(t)=θ,t[0,1]

with W(t)=B¯+tφ(B), φ(B)=B-B¯ containing both problems (2.1) and (2.2), so that we obviously have

un(0)=un(0),λn(0)=λn(0),un(1)=un,λn(1)=λn.

This suggests the idea to look for the solution of (2.3) in the form

(2.4)λn(t)=j=0λn(j)tj,un(t)=j=0un(j)tj,

where

(2.5)λn(j)=1j!djλn(t)dtj|t=0,un(j)=1j!djun(t)dtj|t=0.

Setting t=1 in (2.4), we obtain

λn=j=0λn(j),un=j=0un(j),

provided that the series in (2.4) converge for all t[0,1].

The identities given in (2.5) are not suitable for a numerical algorithm, therefore we need another way to compute the corrections λn(j),un(j) which we describe below.

Substituting (2.4) into (2.3) and matching the coefficients in front of the same powers of t, we arrive at the following recurrence sequence:

(2.6)(A+B¯)un(j+1)-λn(0)un(j+1)=Fn(j+1),j=-1,0,1,,

with Fn(0)=0 and

Fn(j+1)=Fn(j+1)(λn(1),,λn(j+1);un(0),,un(j))
=-φ(B)un(j)+p=0jλn(j+1-p)un(p)
(2.7)=λn(j+1)un(0)-φ(B)un(j)+p=1jλn(j+1-p)un(p),j=0,1,.

For the pair λn(0),un(0) corresponding to the index j=-1 we get the so-called base eigenvalue problem

(A+B¯)un(0)-λn(0)un(0)=θ,

which for simplicity is assumed to have no multiple eigenvalues, to be “simpler” than the original one, and to produce the initial data for problems (2.6), (2.7). The case of multiple eigenvalues of the base problem was studied in [17].

Problems (2.6) for higher indices j0 are solvable provided that

(Fn(j+1),un(0))=0,j=0,1,,

from where we obtain

(2.8)λn(j+1)=(φ(B)un(j),un(0)),j=0,1,.

Under this condition the general solution of the inhomogeneous equation (2.6) with the singular operator can be represented by

un(j+1)=Cun(0)+p=1,pn(Fn(j),up(0))λp(0)-λn(0)up(0)

with an arbitrary constant C. We choose the particular solution

un(j+1)=p=1,pn(Fn(j),up(0))λp(0)-λn(0)up(0)

satisfying the condition

(un(j+1),un(0))=0,j=0,1,.

The start values λn(0),un(0) for the recursion (2.6), (2.8) are the solutions of the base problem.

The truncated series

(2.9)u𝑚n=j=0mun(j),λ𝑚n=j=0mλn(j)

represent an algorithm to find the approximate solution λ𝑚n,u𝑚n (of rank m) to the solution of problem (2.1).

Below we give the error estimates of this method in the cases of a “dominated” fractional derivative and a “subordinated” fractional derivative.

3 The Fourier Fractional Derivative

The following differentiation and integration formulas can be easily proved for n, a:

(3.1)DnFsin(ax)=sin(n)(ax)=(a)nsin(ax+πn2),
(3.2)D-nFsin(ax)=n times=(a)-nsin(ax-πn2).

We can generalize these formulas in a natural way for real n and introduce the differentiation and integration operators of fractional order α (the Fourier fractional derivative) by

DαFsin(ax)=sin(α)(ax)=(a)αsin(ax+πα2),
D-αFsin(ax)=(a)-αsin(ax-πα2),

cf. [24, 23]. One can see, especially for a=nπ, that

(3.3)|DαFsin(nπx)|(nπ)α.

The functions 2sin(nπx), n=1,2,, build an orthonormal basis in the space L2,0(0,1) of functions vanishing at the ends of the interval with the norm fL2=(01f2(x)𝑑x)1/2. An arbitrary function f(x) from this space can be represented by the Fourier series

(3.4)f(x)=n=1ansin(nπx)

with an=201f(x)sin(nπx)𝑑x. The fractional derivative for such functions can be defined by

DαFf(x)=n=1an(nπ)αsin(nπx+πα2).

Given the Fourier representation (3.4), the inverse operator D-αF is defined by

D-αFf(x)=n=1an(nπ)-αsin(nπx-πα2).

One can also define the fractional derivative of a 2π-periodic function using the exponential form of the Fourier series (see, e.g., www.xuru.org/fc/exponentials.asp):

f(x)=n=-einx2π-ππf(x)e-int𝑑tDαFf(x)=n=-(nπ)αei(nx+απ/2)2π-ππf(x)e-int𝑑t.
Remark 3.1

It is easy to see that the eigenpairs of the operator

Au=-d2udx2for all uD(A):={uH2(0,1):u(0)=0,u(1)=0},

are λk=k2π2, uk=sin(kπx) for k=1,2,. Therefore we obtain the following connection between the fractional powers Aα of this operator and the Fourier fractional derivatives on the elements (3.4):

DαFf(x)=n=1an(nπ)αsin(nπx+πα2)
=cos(πα2)n=1an(nπ)αsin(nπx)+sin(πα2)n=1an(nπ)αcos(nπx)
=cos(πα2)Aα/2f+sin(πα2)ddxDα-1Ff.

That is, we have

Aα/2=cos-1(πα2)DαF-tan(πα2)ddxDα-1F,α1.
Example 3.2

To obtain the fractional derivative of the product of functions f(x)=x and g(x)=sin(nπx), let us represent this product by the Fourier series

xsin(kπx)=p=1apsin(pπx),

where

ak=1201xsin(nπx)sin(kπx)dx={-4nk[1-(-1)n+k]π2(k2-n2)2if kn,12if k=n.

Then using the definitions (3.1) and (3.2), we have

DαF[xsin(nπx)]=k=1ak(kπ)αsin(kπx+πα2),

in particular

D1/2F[xsin(nπx)]=k=1ak(kπ)1/2sin(kπx+π4).

Let us consider the asymptotic behavior with respect to n of the Caputo and Riemann–Liouville fractional derivatives of the function sin(nπx). We have

DtαaCsin(nπt)=nπΓ(1-α)0t(t-τ)-αcos(nπτ)𝑑τ=nπΓ(1-α)φn(α)(t),

where

φn(α)(t)=t1cos(nπξ)(ξ-t)α𝑑ξ,t[0,1],α(0,1).

The asymptotic behavior of φn(α) with respect to n gives the next simple lemma.

Lemma 3.3

There exists a constant c(α) independent of n such that

n1-αmaxt[0,1]|φn(α)(t)|c(α),

where c(α)=11-α.

Proof.

By change of the variable ξ=ζ/n, we obtain

φn(α)(t)=t1cos(nπξ)(ξ-t)α𝑑ξ=1n1-αntncos(πζ)(ζ-nt)α𝑑ζ.

Further, for t[0,1], we have

n1-α|φn(α)(t)|ntndζ(ζ-nt)αdζ=(ζ-nt)-α+1-α+1|ntn=(n-nt)-α+1-α+11-α+1=c(α).

The lemma yields the estimate

(3.5)|DtαaCsin(nπt)|c(α)πΓ(1-α)nα,

which is of the same order in n as the Fourier fractional derivative (3.3). Due to (1.2) we have the same estimate for the Riemann–Liouville derivative too.

4 The Sturm–Liouville Problem with a Subordinated Fractional Derivative

Let us consider the following Sturm–Liouville problem:

(4.1){d2u(x)dx2+k(x)Dαu(x)+(λ-q(x))u(x)=0,x(0,1),u(0)=0,u(1)=0,

where Dα denotes the Fourier, Caputo or Riemann–Liouville fractional derivatives. For shortness, we will use in this section the Fourier derivative DαFu(x), since the main property in use is the asymptotic (3.3) and (3.5) which are of the same order in n for all three derivatives.

If we approximate the coefficients k(x),q(x) by the constant 0 on the whole interval (the simplest variant), then the FD-method for (4.1) consists of the following sequence of recursive problems:

(4.2){d2un(j+1)(x)dx2+λn(0)un(j+1)(x)=Fn(j+1)(x),x(0,1),un(j+1)(0)=0,un(j+1)(1)=0,j=-1,0,,

where

(4.3){Fn(j+1)(x)=-s=0jλn(j+1-s)un(s)(x)-k(x)Dαun(j)(x)+q(x)un(j)(x),x(0,1),Fn(0)(x)=0,un(0)(x)=2sin(nπx),λn(0)=(nπ)2.

The solvability condition

01Fn(j+1)(ξ)sin(nπξ)𝑑ξ=0

implies

(4.4)λn(j+1)=-01k(ξ)Dαun(j)(ξ)2sin(nπξ)𝑑ξ+01q(ξ)un(j)(ξ)2sin(nπξ)𝑑ξ.

The particular solution of problem (4.2) satisfying the orthogonality condition

01un(j+1)(ξ)sin(nπξ)𝑑ξ=0

can be represented by

(4.5)un(j+1)(x)=2p=1,pn01Fn(j+1)(ξ)sin(pπξ)𝑑ξπ2(n2-p2)sin(pπx)

and we have

Dαun(j+1)(x)=2p=1,pn01Fn(j+1)(ξ)sin(pπξ)𝑑ξπ2(n2-p2)(pπ)αsin(pπx+πα2).

Using the orthonormality of the system 2sin(pπx) for the corrections of the eigenfunction, we obtain the estimates

un(j+1)1π2(2n-1)Fn(j+1)
(4.6)1π2(2n-1){s=0j|λn(j+1-s)|un(s)+kDαun(j)+qun(j)}

and

Dαun(j+1)2π2-αmax((n-1)α2n-1,(n+1)α2n+1)Fn(j+1)
(4.7)Mn(1){s=0j|λn(j+1-s)|un(s)+kDαun(j)+qun(j)}

with

(4.8)Mn(1)=2π2-αmax((n-1)α2n-1,(n+1)α2n+1)2απ2-α(n+1)α2n-1Mn¯(1)=2α+0.5π2-αnα-1.

The corrections to the eigenvalues are estimated by

(4.9)|λn(j+1)|kDαun(j)+qun(j).

Introducing the majorants Uj, Vj and Λj by

un(j+1)Uj+1,Dαun(j+1)Vj+1,|λn(j+1)|Λj+1

and replacing the inequality signs in (4.6)–(4.9) by equal signs, we obtain the following majorant system of equations:

(4.10){Uj+1=2π2(2n-1)s=0jΛj+1-sUs,Λj+1=kVj+qUj,Vj+1=2Mn(1)s=0jΛj+1-sUs,j=0,1,,U0=1,V0=(πn)α.

A consequence of (4.10) is

(4.11)Vj+1=κVUj+1,Λj+1=κΛUj,Uj+1=κUs=0jUj-sUs,j=0,1,,

where

κV=π2(2n-1)2Mn(1),κΛ=[kπ2(2n-1)2Mn(1)+q],κU=[kMn(1)+2qπ2(2n-1)].

The last equation in (4.11) is a recurrence equation of convolution type which can be solved by the method of generating functions (see, e.g., [42, 6]). We successively obtain

Uj=(rn)j2(2j-1)!!(2j+2)!!,Vj=κΛ(rn)j2(2j-1)!!(2j+2)!!,Λj+1=κΛ(rn)j2(2j-1)!!(2j+2)!!,

where

rn=4κU=4[kMn(1)+2qπ2(2n-1)]

and rn(0,1) for n large enough.

From the definition of the majorant sequences we obtain the accuracy estimates

un-u𝑁nk=N+1Uk2k=N+1(rn)k(2k-1)!!(2k+2)!!2κU(2N+1)!!(2N+4)!!rnN+11-rn.

It is easy to see that

n!!={i=1n22i=2n2(n2)!if n is even,i=0n-12(2i+1)=n!2n-12(n-12)!if n is odd.

Therefore,

(2N+1)!!(2N+4)!!=(2N+1)!22N+2N!(N+2)!=Γ(2N+2)22N+2Γ(N+1)Γ(N+3).

Now, the well-known Stirling’s asymptotic formula Γ(t+1)2πt(t/e)t implies

(2N+1)!!(2N+4)!!2π(2N+1)(2N+1e)2N+122N+22πN(Ne)N2π(N+2)(N+2e)N+2
122N+22πN(2N+1e)2N(2N+1e)(Ne)2N(N+2e)2
N-3/2(1+12N)2N
N-3/2.

Analogously we obtain the corresponding estimate for the eigenvalues.

Thus, we come to the following assertion.

Theorem 4.1

Let for α[0,1) the following condition be fulfilled:

rn=4[k2α+0.5π2-αnα-1+2qπ2(2n-1)]<1.

Then the FD-method for (4.1) is super-exponentially convergent with the error estimates

un-u𝑁ncN-3/2rnN+1,λn-λ𝑁ncN-3/2rnN+1,

where c is a constant independent of N.

4.1 Recursive Implementation of the Fourier Derivative

In this subsection we show that the corrections for eigenpairs can be computed without use of (4.4) and (4.5), i.e., we can avoid the computation of the integrals included. For the sake of simplicity let us consider problem (4.1) with k(x)=1 and q(x)=0 (otherwise one should expand these functions in Fourier series or approximate by a trigonometric polynomial). Substituting the Fourier representation

un(j)(x)=k=1an,k(j)2sin(kπx)

with unknown coefficients an,k(j) into the formulas (4.3), (4.4) and (4.5), we obtain the recurrence relations

λn(0)=(nπ)2,λn(1)=-(nπ)αcos(απ2),
ank(0)=δn,k,ank(1)=-2nαπ3-αsin(απ2)k[(-1)n+k-1](k2-n2)2,k=1,,,kn,
λn(j+1)=-k=1,knan,k(j)(nπ)α201sin(kπx+απ2)sin(nπx)𝑑x
=-2nπsin(απ2)k=1,kn(kπ)α[(-1)k+n-1]k2-n2ank(j),j=1,2,,
ank(j+1)=-1π2(n2-k2)(p=1jλn(j+1-p)ank(p)+2kπsin(απ2)t=1,tnant(j)(tπ)α[(-1)k+t-1]t2-k2
+ank(j)cos(απ2)(kπ)α),k=1,,,kn,j=1,2,.

Note that the pairs λn(j+1), ank(j+1) for all j=1,2, can be computed simultaneously in a loop with respect to j with an included loop with respect to k. For practical computation one can truncate the series keeping M summands. The behavior of the corrections to the eigenvalue λ5 computed with the computer algebra tool Maple for our example with α=12 and M=2048 is illustrated in Table 1.

Thus, our method of the rank 10 provides the approximation

λ510=243.9826068784193.

Table 2 demonstrates the dependence of the eigenvalue from M.

As appears from Table 1, the FD-method of rank 5 instead of 10 would provide the same accuracy.

Table 1

Correction λ5(j) vs. j for M=2048.

jλ5(j)
025π2
110π/2
20.04646
3-0.00153
40.00007
5-0.3740810-5
60.2100910-6
7-0.1237410-7
80.7541210-9
9-0.4715510-10
100.3008210-11
Table 2

The approximation λ105 vs. M.

Mλ105
8243.99009
16243.98510
32243.98375
64243.98323
128243.98297
256243.98282
512243.98272
1024243.98265
2048243.98260

4.2 Direct Implementation of the Riemann–Liouville Derivative

Let us consider the following eigenvalue problem with the Riemann–Liouville derivative:

(4.12){u′′(x)+(Dx1/20RLu)(x)+λu(x)=0,x(0,1),u(0)=0,u(1)=0,(Dx1/20RLu)(x)=1πddx0xu(t)x-t𝑑t.

We apply to problem (4.12) the simplest variant of the FD-method by setting the coefficient in front of the fractional derivative equal to zero. We obtain the base problem

{d2un(0)(x)dx2+λn(0)un(0)(x)=0,x(0,1),un(0)(0)=0,un(0)(1)=0.

The solution of the base problem is

un(0)(x)=2sin(nπx),λn(0)=(nπ)2,n=1,2,.

The next corrections are the solutions of

(4.13){d2un(1)(x)dx2+λn(0)un(1)(x)=-λn(1)un(0)(x)-Dx1/20RL(un(0)),x(0,1),un(1)(0)=0,un(1)(1)=0,Dx1/20RL(un(0))=2πn[sin(nπx)S(2nx)+cos(nπx)C(2nx)],

where S(z), C(z) are the Fresnel’s integrals; see, e.g., [43]. The solvability condition implies

λn(1)=n2π[2πS(2n)-1nC(2n)-(-1)n2n].

The general solution of (4.13) is

un(1)(x)=cn(1)un(0)(x)+12πn[S(2nx)(2xπcos(nπx)-1nsin(nπx))
-C(2nx)(2xπsin(nπx)+1ncos(nπx))+2xn]-λn(1)2π2n[-1nsin(nπx)+πxcos(nπx)]

with an arbitrary constant cn(1). We choose the particular solution satisfying the orthogonality condition

01un(1)(x)un(0)(x)𝑑x=0

and obtain

cn(1)=-116π7/2π[S(2n)42π2n+C(2n)(-42π3-2πn2)].

We find the next corrections within a guarantied accuracy by the corresponding choice of the parameter “Digits” in the computer algebra tool Maple.

To evaluate the accuracy of the results obtained, we find the exact first eigenvalue of problem (4.12) using the Laplace transform which provides the correspondence U(p)={u(x)}. The solution of the ODE (4.12) satisfying the conditions

u(0)=0,u(0)=1

is then given by

u(x)=-1(1p2+p+λ)
=-1πx+y1ey12xerf(-y1x)(y4-y1)(y3-y1)(y2-y1)-1πx+y2ey22xerf(-y2x)(y4-y2)(y3-y2)(y1-y2)
(4.14)+1πx+y3ey32xerf(-y3x)(-y4+y3)(-y3+y2)(y1-y3)-1πx+y4ey42xerf(-y4x)(-y4+y3)(y2-y4)(y1-y4),

where yi=yi(λ), i=1,,4 are the roots of the equation y4+y+λ=0, and erf(z) is the complementary error function. Setting x=1 in (4.14), we obtain a transcendent equation defining the eigenvalues of (4.12). In particular, we obtain

λ1=8.8857068923.

4.3 Recursive Implementation of the Riemann–Liouville Derivative

Let us consider another algorithm similar to that of Section 4.1 and based on the recurrence formulas for the expansion coefficients of the corrections.

We look for the corrections of the eigenfunctions in the form

(4.15)un(j)(x)=k=1,knan,k(j)2sin(kπx).

Note that the solvability condition 01un(j)(x)un(0)(x)𝑑x=0 is automatically satisfied. We will use the representation

(4.16)Dx1/20RL(un(j))(x)=k=1,knan,k(j)φk(x)

with

φk(x)=2kπ[sin(kπx)S(2kx)+cos(kπx)C(2kx)]
=p=1,pkμp,k2sin(pπx)+μk,ksin(kπx),

where

S(x)=0xsin(t2)𝑑t,C(x)=0xcos(t2)𝑑t

are the Fresnel integrals, and

μp,k={-22kpπ(-1)k+pC(2p)k-C(2k)p-p2+k2if pk,k2π(-C(2k)2k+2S(2k)+2kcos(kπ))if p=k.

After substitution of (4.15) and (4.16) into the expression

Fn(j+1)(x)=-p=0jλn(j+1-p)un(p)(x)-Dx1/20RL(un(j))(x)

for the right-hand side of the equations for corrections

(4.17)d2un(j+1)(x)dx2+(nπ)2u(x)=Fn(j+1)(x),

we obtain

Fn(j+1)(x)=-λn(j+1)2sin(nπx)-k=1,knfn,k(j+1)2sin(kπx),

where

fn,k(j+1)=p=1jλn(j-p+1)an,k(p)+s=1an,s(j)μs,k.

The solution of (4.17) is then given by (4.15) with

an,k(j+1)=1π2(n2-k2){p=1,pk,pn22pkan,p(j)[-C(2k)p+(-1)p+kC(2p)k]π1/2(p2-k2)
(4.18)+p=0jλn(j+1-p)an,k(p)+12πk[-C(2k)2k+(2k)3/2πS(2k)+2k(-1)k]an,k(j)}.

For the eigenvalue corrections we have

(4.19)λn(j+1)=-01Dx1/20RL(un(j))(x)un(0)(x)𝑑x=-k=1,knan,k(j)22kn[(-1)n+kC(2k)n-C(2n)k]π(-n2+k2).

The formulas (4.18), (4.19) with the initial conditions

an,k(0)=δnk,λn(0)=(nπ)2,

where δnk is the Kronecker delta, represent a recursive algorithm for the corrections which contrary to the one of Section 4.2 avoids the solution of differential problems.

In practical computations we used truncated sums with M summands instead of infinite series and then computed the N-th approximation to the eigenpair according to (2.9) (the FD-method of rank N). The corrections λ1(j), j=0,1,,10, of the series with M=32 of the FD-method for the lowest eigenvalue of problem (4.12) are given in Table 3. Thus, we have

λ101=8.88570689232811335600142,

where the first ten digits after the decimal point coincide with the exact ones.

Table 3

The corrections of the FD-method for problem (4.12)computed by truncated series with M=32 summands.

jλ1(j)
09.86960440108935861883447
1-1.01447613638170169201166
20.0297977106497561210442917
30.000748595098364490732963004
40.0000307247268252977828544031
50.15103420908383637873223210-5
60.81796708243005945524825410-7
70.47059635309877423125810510-8
80.28220382785425656531942610-9
90.17441137005565349523500010-10
100.11029424012023322837492110-11
Table 4

The FD-approximations λ𝑁n and the eigenvalue corrections λn(j) vs. M withn=3, N=10, j=10 for problem (4.12).

Mλ103λ3(10)
886.7794941481798320121360.23619830691008821940493410-14
1686.7795885027336720205576.23230346423531504360612810-14
3286.7795973153306097826733.23212622018616654289366910-14

For the third eigenvalue (n=3) and various M we obtained the results given in Table 4. The numerical results for the eigenvalue λ1 obtained by the recursive algorithm with M=32 coincide with the ones from Section 4.2. Table 5 shows the behavior of corrections for the third eigenvalue. The FD-method of the rank 10 provides the approximation

λ103=86.7795885027336720205576,

where the first nine digits after the decimal point coincide with the exact eigenvalue obtained by the Laplace transform method. One can observe the principal characteristic of the FD-method: the convergence rate increases together with the eigenvalue index.

Table 5

The FD-corrections λn(j) with M=32, n=3 for example (4.12).

jλ3(j)
088.8264396098042275695102
1-2.04972510804358999684540
20.00290212991892648225391892
3-0.0000305277301362658818404213
40.22265312612294429388022710-5
50.16488303296070080312387010-6
60.71186051820183747660544110-8
70.24413562750053069571666210-9
80.70432963793408408921734610-11
90.16361244409959548626386410-12
100.23230346423531504360612810-14

5 Application to a Jacobi-Type ODE with a Dominated Fractional Derivative

Let us consider the following problem:

(5.1){1μu(x)+q(x)u(x)+λw(x)u(x)=0,x(-1,1),u(-1)=0,I11-μxRL[Dxμ-1Cu(x)]x=1=0,1μu(x)=D1μxRL(Dxμ-1Cu(x)),

where μ(0,1), w(x)=(1-x)-μ(1+x)-μ, D1μxRL is the Riemann–Liouville derivative, Dxμ-1C the Caputo fractional derivative, and I11-μxRL the Riemann–Liouville integral (1.1). Note that the operator defined by 1μ and by the given boundary conditions is self-adjoint. This can be easily derived analogously to [9, Corollary 3.1, formula (3.25)], where one can easily show that α=s, β=-s is permissible since the Caputo and the Riemann–Liouville derivatives coincide on the functions vanishing at x=1. Besides we can set s=μ.

Approximating q(x) by the constant zero, we obtain the base problem of the FD-method with the solution

un(0)(x)=(1-x)μPn-1μ,-μ(x),λn(0)=-Γ(n+μ)Γ(n-μ),n=1,2,,

where Pnα,β(x) denotes the standard Jacobi polynomials. The eigenfunctions build an orthogonal basis in the space Lw2[-1,1] of quadratic integrable functions with weight w, and the recurrence sequence of problems for the corrections is

(5.2){L1μun(j+1)(x)+λn(0)w(x)un(j+1)(x)=-w(x)Fn(j+1)(x),x(-1,1),un(j+1)(-1)=0,I11-μxRL[Dxμ-1Cun(j+1)(x)]x=1=0,j=0,1,,

where

Fn(j+1)(x)=1w(x)[s=0jλn(j+1-s)w(x)un(s)(x)+q(x)un(j)(x)].

The solvability condition

-11w(x)Fn(j+1)(x)(x)un(0)(x)𝑑x=0

for the singular problem (5.2) implies

(5.3)λn(j+1)=-11q(x)un(j)(x)un(0)(x)𝑑x-11w(x)[un(0)(x)]2𝑑x.

Using the condition

-11w(x)un(j+1)(x)un(0)(x)𝑑x=0,

we single out from the general solution (i.e. from the set of all possible solutions of (5.2)) the following particular solution:

un(j+1)(x)=p=1,pn-11Fn(j+1)(ξ)up(0)(ξ)𝑑ξλn(0)-λp(0)up(0)(x).

This equality together with the orthogonality of the system {up(0)(x)}p=1,, yields

(5.4)un(j+1)wMn(2)Fn(j+1)wMn(2)[s=1j|λn(j+1-s)|un(s)w+maxx[-1,1](|q(x)|w(x))un(j)wun(0)wun(0)w],

where

uw=(-11w(x)u2(x)𝑑x)1/2,un(0)w222μPn-1-μ,μ(x)w2=22μ2Γ(n-μ)Γ(n+μ)(2n-1)Γ2(n),
Mn(2)=[Γ(n+μ)Γ(n-μ)-Γ(n-1+μ)Γ(n-1-μ)]-1un-1(0)w=[2μn-1-μΓ(n-1+μ)Γ(n-1-μ)]-1un-1(0)w.

It was shown in [48] that, for z,

Γ(z+α)Γ(z+β)=zα-β[1+(α-β)(α-β-1)2z+o(|z|-2)],

which yields the existence of a constant c independent of n such that

Γ(n+1-μ)Γ(n+1+μ)cn-2μ,un(0)w222μ2Γ(n-μ)Γ(n+μ)(2n-1)Γ2(n).

Using [18], one can obtain

Γ(n+1-μ)Γ(n+1+μ)n-2μ(613-1)2μ,n3.

Thus, we have

(5.5)Mn(2)n1-2μun-1(0)w12μ((13-1)/6)2μn1-2μun-1(0)w12μ((13-1)/6)2μcn1/2-2μ

and can see that Mn(2) as μ0 but Mn(2)0 as n for each fixed μ>14.

Introducing the new variable

u¯n(j)(x)=un(j)(x)un(0)w,

we obtain from (5.4)

u¯n(j+1)wMn(2)s=0ju¯n(j-s)wu¯n(s)w,j=0,1,,u¯n(0)w=1.

We solve this recurrence system of inequalities of convolution type analogously as above by switching to the majorant system and using the method of generating functions; see, e.g., [6, 16, 26, 27]. Then we obtain

(5.6)u¯n(j)wun(j)wun(0)wqnj2(2j-1)!!(2j+2)!!qnj(j+1)πj,qn=4Mn(2).

Using (5.6), we have from (5.3)

|λn(j+1)|maxx[-1,1](|q(x)|w(x))qnj(j+1)πj.

Thus, we have proved the following assertion.

Theorem 5.1

Let q(x)/w(x)C[-1,1], μ>14 and qn=4Mn(2)<1. Then the FD-method converges super-exponentially with the estimates

(5.7)|λn-λn𝑚|=|j=m+1λn(j)|maxx[-1,1](|q(x)|w(x))qnm(m+1)πm11-qn,
(5.8)un-un𝑚w=j=m+1un(j)wqnm+1(m+2)π(m+1)un(0)w1-qn.

Remark 5.2

It follows from (5.5) that for μ>14 there exists such n0 that for all nn0 the assumptions of the theorem are fulfilled and the FD-method converges super-exponentially with the accuracy estimates (5.7) and (5.8). The condition μ>14 in Theorem 5.1 means that the fractional derivative should be of order greater than 12. The condition α<1 in Theorem 4.1 means that the convergence of the FD-method is guaranteed if the fractional derivative is of order less than 1, i.e., it does not dominate.

Example 5.3

Let us consider the case q(x)=w(x)x, μ=45. For n=1 the solution of the base problem is

u1(0)(x)=(1+x)4/5,λ1(0)=-Γ(9/5)Γ(1/5).

Table 6 gives the corrections λ1(j), j=0,1,,5. Thus, we have

λ51=-0.8656637788992767

and one can observe practical convergence. Nevertheless, the sufficient convergence condition does not hold since q1=28.82>1. Only beginning with n=4 we have qn<1. In order to get theoretically justified eigenvalues with smaller numbers, one should apply the general algorithm of the FD-method with a piecewise constant approximation of q(x) on a partitioning of the interval fine enough.

Table 6

The FD-corrections for (5.1) with q(x)=w(x)x, μ=4/5.

jλ1(j)
1-0.2028785620703973
2-0.7999999999999999
30.07587304555679625
40.03837812694709120
50.01714234558251600
60.005821265084716930

5.1 Recursive Implementation of the FD-Method for a Jacobi-Type Differential Operator

Now, let us show that the algorithm above can be reformulated as a recurrence algorithm with respect to the coefficients of some expansions of corrections λn(j) and un(j).

For the sake of simplicity we consider the problem

{Dxs-1{(1-x2)sD1sxu(x)}+q(x)u(x)-λu(x)=0,x(-1,1),|u(-1)|<,|u(1)|<,s(0,1).

This problem is of type (5.1) and is a particular case of the problem from [9] (the first formula on the top with α=β=0). The proof of convergence of the FD-method for this problem can be done analogously to the one of Theorem 5.1. Note that for the case of a polynomial potential q(x) and s=1, a modification of the FD-method was proposed in [29] without solving boundary value problems in each iteration. A further modification of this procedure was proposed in [30].

Analogously to the case of the Fourier fractional derivative the algorithm for the coefficients of corrections of eigenfunctions for a polynomial potential q(x) can be formulated as a recurrence procedure. But now these corrections are finite linear combinations of Legendre polynomials with the number of summands depending on the degree of the potential. For the sake of simplicity we illustrate this for q(x)=x2. The base problem in this case is

{Dxs-1{(1-x2)sD1sxun(0)(x)}-λn(0)un(0)(x)=0,x(-1,1),|un(0)(-1)|<,|un(0)(1)|<,un(0)(x)=Pn(x),λn(0)=Γ(n+s+1)Γ(n-s+1),

where Pn(x) are the Legendre polynomials. The recurrence sequence of problems for corrections of the FD-method is

(5.9){Dxs-1{(1-x2)sD1sxun(j+1)(x)}-λn(0)un(j+1)(x)=Fn(j+1)(x),x(-1,1),|un(j+1)(-1)|<,|un(j+1)(1)|<,

where

Fn(j+1)(x)=p=0jλn(j+1-p)un(p)(x)-x2un(j)(x),j=0,1,.

Using the well-known property of the Legendre polynomials

x2Pn(x)=x2n+1((n+1)Pn+1(x)+nPn-1(x))
=12n+1((n+1)(n+2)(2n+3)Pn+2(x)+((n+1)2(2n+3)+n22n-1)Pn(x)+n(n-1)2n-1Pn-2(x))
=bnPn+2(x)+cnPn(x)+dnPn-2(x),

the orthogonality condition and the mathematical induction, we can show the following representation:

(5.10)un(j)(x)=p=-j,n+2p0jan+2p(j)Pn+2p(x),an(j)=0,an(0)=1.

After substitution of (5.10) into (5.9) we arrive at the following recurrence system for the coefficients of (5.10):

(5.11)an+2p(j+1)=1λn+2p(0)-λn(0)[k=0jλn(j+1-k)an+2p(k)-(an+2p-2(j)bn+2p-2+an+2p(j)cn+2p+an+2p+2(j)dn+2p+2)]

for j=0,1, and p=-n/2,,j+1. The solvability condition for problem (5.9) yields

(5.12)λn(j+1)=an-2(j)bn-2+an+2(j)dn+2.

Now (5.11) and (5.12) build a self-contained algorithm.

For n=0 we have

λ0(0)=12,u0(0)(x)=P0(x),λ0(1)=13,u0(1)(x)=-13P2(x),λ0(2)=-245,u0(2)(x)=1126P2(x)+135P4(x),λ0(3)=11945,u0(3)(x)=-48126460P2(x)-35932340P4(x)-1693P6(x),λ0(4)=-481198450,u0(4)(x)=2369312224520P2(x)+52870311942340400P4(x)+65479604980P6(x)+119305P8(x),
λ0(27)=-0.185512925410-13.

According to our theory the sequence λ𝑁0=λ0(0)++λ0(N) converges to the exact eigenvalue

λ0=0.79839.
Remark 5.4

The algorithm described above can be generalized to the case when the coefficient r(x) in the front of the fractional derivative is not constant. In this case we cover the whole interval by a grid

ωh={x0=-1<x1<x2<<xM-1<xM=1}

with M-1 points and approximate it by a piecewise constant function r¯(x)=r(xi-1), i=1,,M. The crucial point is the solution of the base problem. To do this, we write down the general solution on each subinterval depending on two arbitrary constants. Two of these constants (on the edge subintervals) can be eliminated using the boundary conditions. Then stitching the solutions and their derivatives at the grid points, we obtain a system of 2M-2 linear homogeneous algebraic equations with 2M-2 unknowns with a matrix depending on the eigenvalue parameter λn(0). The solvability condition for this system leads to a transcendent equation with this parameter from where we obtain the eigenvalues of the base problem. Further we can iterate as usual for the FD-method exploiting the fact that the differential operator with piecewise coefficients of the problem for corrections remains the same.

References

[1] Agrawal O., Generalized variational problems and Euler–Lagrange equations, Comput. Math. Appl. 59 (2010), no. 5, 1852–1864. 10.1016/j.camwa.2009.08.029Search in Google Scholar

[2] Akulenko L. and Nesterov S., High-Precision Methods in Eigenvalue Problems and Their Applications, Chapman & Hall/CRC, Boca Raton, 2005. 10.4324/9780203401286Search in Google Scholar

[3] Al-Mdallal Q., An efficient method for solving fractional Sturm–Liouville problems, Chaos Solitons Fractals 40 (2009), 183–189. 10.1016/j.chaos.2007.07.041Search in Google Scholar

[4] Antunes P. and Ferreira R., An augmented-rbf method for solving fractional Sturm–Liouville eigenvalue problem, SIAM J. Sci. Comput. 37 (2003), no. 1, A515–A535. 10.1137/140954209Search in Google Scholar

[5] Bandyrskii B. I., Gavrilyuk I. P., Lazurchak I. I. and Makarov V. L., Functional-discrete method FD-method for matrix Sturm–Liouville problems, Comput. Methods Appl. Math. 5 (2005), no. 4, 1–25. 10.2478/cmam-2005-0017Search in Google Scholar

[6] Bandyrskij B. J., Makarov V. and Ukhanev O. L., FD-method for the Sturm–Liouville problem. Exponential convergence rate, J. Comput. Appl. Math. 85 (2000), no. 1, 1–60. Search in Google Scholar

[7] Bogoliouboff N. N. and Kryloff N. M., Sopra il metodo dei coefficienti constanti (metodo dei tronconi) per l’integrazione approssimata delle equazioni differenziali della fisica matematica, Boll. Unione Mat. Ital. 7 (1928), no. 2, 72–77. Search in Google Scholar

[8] Carpinteri A. and Mainardi F., Fractals and Fractional Calculus in Continuum Mechanics, Springer, Wien, 1997. 10.1007/978-3-7091-2664-6Search in Google Scholar

[9] Chen S., Shen J. and Wang L.-L., Generalized Jacobi functions and their applications to fractional differential equations, Math. Comp. 85 (2016), no. 300, 1603–1638. 10.1090/mcom3035Search in Google Scholar

[10] Diethelm K., The Analysis of Fractional Differential Equations, Springer, Berlin, 2010. 10.1007/978-3-642-14574-2Search in Google Scholar

[11] Diethelm K., Ford N., Freed A. and Luchko Y., Algorithms for the fractional calculus: A selection of numerical methods, Comput. Methods Appl. Mech. Engrg. 194 (2005), 743–773. 10.1016/j.cma.2004.06.006Search in Google Scholar

[12] Duan J.-S., Wang Z., Liua Y.-L. and Qiua X., Eigenvalue problems for fractional ordinary differential equations, Chaos Solitons Fractals 46 (2013), 46–53. 10.1016/j.chaos.2012.11.004Search in Google Scholar

[13] Ertürk V., Computing of eigenelements of Sturm–Liouville problems of fractional order via fractional differential transform method, Math. Comput. Appl. 16 (2011), no. 3, 712–720. 10.3390/mca16030712Search in Google Scholar

[14] Ford N. and Morgado M., Fractional boundary value problems: Analysis and numerical methods, Fract. Calc. Appl. Anal. 14 (2011), no. 4, 554–567. 10.2478/s13540-011-0034-4Search in Google Scholar

[15] Gavrilyuk I. P., Klimenko A. V., Makarov V. L. and Rossokhata N. O., Exponentially convergent parallel algorithm for nonlinear eigenvalue problems, IMA J. Numer. Anal. 27 (2007), no. 4, 818–838. 10.1093/imanum/drl042Search in Google Scholar

[16] Gavrilyuk I. P. and Makarov V., Super-exponentially convergent parallel algorithm for eigenvalue problems in Hilbert spaces, Proceedings of the 6th European Conference on Computer Systems, Kaunas University of Technology, Lithuania (2009), 86–91. Search in Google Scholar

[17] Gavrilyuk I. P., Makarov V. and Romaniuk N., Super-exponentially convergent parallel algorithm for an abstract eigenvalue problem with applications to odes, Nonl. Oscillations 18 (2015), no. 3, 332–356. Search in Google Scholar

[18] Giordano C. and Laforgia A., Inequalities and monotonicity properties for the gamma function, Math. Comput. Appl. 133 (2001), no. 1–2, 387–396. 10.1016/S0377-0427(00)00659-2Search in Google Scholar

[19] Gorenflo F. and Mainardi F., Fractional calculus: Integral and differential equations of fractional order, Fractals and Fractional Calculus in Continuum Mechanics, Springer, New York (1997), 223–276. 10.1007/978-3-7091-2664-6_5Search in Google Scholar

[20] Hilfer R., Applications of Fractional Calculus in Physics, World Scientific, Hackensack, 2000. 10.1142/3779Search in Google Scholar

[21] Jovanovic B., Vulkov L. and Delic A., Boundary value problems for fractional pde and their numerical approximation, Numerical Analysis and its Applications (NAA 2012), Lecture Notes in Comput. Sci. 8236, Springer, Berlin (2013), 38–49. 10.1007/978-3-642-41515-9_4Search in Google Scholar

[22] Katugampola U., A new fractional derivative with classical properties, preprint 2014, http://arxiv.org/abs/1410.6535. Search in Google Scholar

[23] Liouville J., Memoire: Sur le calcul des differentielles àindices quelconques, J. Ècole Polytech. 13 (1832), 71–162. Search in Google Scholar

[24] Liouville J., Memoire sur quelques questions de gèometrie et de mècanique, et sur un noveau gentre pour resoudre ces questions, J. Ecole Polytech. 13 (1832), 1–69. Search in Google Scholar

[25] Mainardi F., Fractional Calculus and Waves in Linear Viscoelasticity: An Introduction to Mathematical Models, World Scientific, Hackensack, 2010. 10.1142/p614Search in Google Scholar

[26] Makarov V., On the functional-discrete method of arbitrary accuracy order for Sturm–Liouville problems with piece-wise constant coefficients, Dokl. Akad. Nauk 320 (1991), 34–39. Search in Google Scholar

[27] Makarov V., FD-method: Exponential convergence rate, Comput. Methods Appl. Math. 1 (1997), no. 82, 69–74. Search in Google Scholar

[28] Makarov V., Functional-discrete method of solving eigenvalue problems for nonlinear differential equations, Dopov. Nats. Akad. Nauk Ukr. Mat. Prirodozn. Tekh. Nauki 2008 (2008), no. 8, 16–22. Search in Google Scholar

[29] Makarov V. and Klymenko Y. V., Application of the FD-method to the solution of the Sturm-Liouville problem with coefficients of special form, Ukrainian Math. J. 59 (2007), no. 8, 1264–1273. 10.1007/s11253-007-0086-0Search in Google Scholar

[30] Makarov V. and Romanyuk N., New properties of the FD-method and its application to the Sturm–Liouville problems, Dopov. Nats. Akad. Nauk Ukr. Mat. Prirodozn. Tekh. Nauki 2014 (2014), no. 2, 26–31. 10.15407/dopovidi2014.02.026Search in Google Scholar

[31] Malinowska A., Odzijewicz T. and Torres D., Advanced Methods in the Fractional Calculus of Variations, Springer, Berlin, 2015. 10.1007/978-3-319-14756-7Search in Google Scholar

[32] Malinowska A. and Torres D., Introduction to the Fractional Calculus of Variations, Imperial College Press, London, 2012. 10.1142/p871Search in Google Scholar

[33] Mendes R. and Vazquez L., The dynamical nature of a backlash system with and without fluid friction, Nonlinear Dynam. 47 (2007), 363–366. 10.1007/s11071-006-9035-ySearch in Google Scholar

[34] Miller K. S. and Ross B., An Introduction to the Fractional Calculus and Fractional Differential Equations, John Wiley & Sons, New York, 1993. Search in Google Scholar

[35] Neamaty A., Darzi R., Zaree S. and Mohammadzadeh B., Haar wavelet operational matrix of fractional order integration and its application for eigenvalues of fractional Sturm–Liouville problem, World Appl. Sci. J. 16 (2012), no. 12, 1668–1672. Search in Google Scholar

[36] Oldham K. B. and Spanier J., The Fractional Calculus, Academic Press, London, 1974. Search in Google Scholar

[37] Oliveira E. C. and Machado J. A. T., A review of definitions for fractional derivatives and integral, Math. Probl. Eng. 2014 (2014), 1–6. 10.1155/2014/238459Search in Google Scholar

[38] Podlubny I., Geometric and physical interpretation of fractional integration and fractional differentiation, Fract. Calc. Appl. Anal. 5 (2002), 367–386. Search in Google Scholar

[39] Pruess S., Estimating the eigenvalues of Sturm–Liouville problems by approximating the differential equation, SIAM J. Numer. Anal. 10 (1973), 55–68. 10.1137/0710008Search in Google Scholar

[40] Pruess S. and Fulton T., Mathematical software for Sturm–Liouville problems, ACM Trans. Math. Software 19 (1993), no. 3, 360–376. 10.1145/155743.155791Search in Google Scholar

[41] Pryce J., Numerical Solution of Sturm–Liouville Problems, Oxford University Press, Oxford, 1994. Search in Google Scholar

[42] Reingold E., Nievergelt J. and Deo N., Combinatorial Algorithms: Theory and Practice, Prentice Hall, Englewood Cliffs, 1977. Search in Google Scholar

[43] Remmert R. and Schumacher G., Funktionentheorie, Springer, Berlin, 2007. Search in Google Scholar

[44] Saha Ray S., Fractional Calculus with Applications for Nuclear Reactor Dynamics, CRC Press, Boca Raton, 2016. 10.1201/b18684Search in Google Scholar

[45] Samko S. G., Kilbas A. A. and Marichev O. I., Fractional Integrals and Derivatives, Theory and Applications, Gordon and Breach, New York, 1993. Search in Google Scholar

[46] Tarasov V., Fractional derivative as fractional power of derivative, Int. J. Math. 18 (2007), no. 3, 281–299. 10.1142/S0129167X07004102Search in Google Scholar

[47] Tarasov V., Fractional Dynamics: Application of Fractional Calculus to Dynamics of Particles, Fields and Media, Springer, Berlin, 2010. 10.1007/978-3-642-14003-7Search in Google Scholar

[48] Tricomi F. and Erdèlyi A., The asymptotic expansion of a ratio of gamma functions, Pacific J. Math. 1 (1951), no. 1, 133–142. 10.2140/pjm.1951.1.133Search in Google Scholar

[49] Vasil’ev V. V. and Simak L. A., Fractional Calculus and Approximation Methods in Modelling of Dynamical Systems, National Academy of Sciences of Ukraine, Kiev, 2008. Search in Google Scholar

[50] Zaslavsky G., Hamiltonian Chaos and Fractional Dynamics, Oxford University Press, Oxford, 2008. Search in Google Scholar

[51] Zayernouri M. and Karniadakis G. E., Fractional Sturm–Liouville eigenproblems: Theory and numerical approximation, J. Comput. Phys. 252 (2013), 495–517. 10.1016/j.jcp.2013.06.031Search in Google Scholar

Received: 2016-2-25
Revised: 2016-4-20
Accepted: 2016-4-21
Published Online: 2016-5-11
Published in Print: 2016-10-1

© 2016 by De Gruyter

Downloaded on 2.11.2025 from https://www.degruyterbrill.com/document/doi/10.1515/cmam-2016-0018/html
Scroll to top button