Abstract
In this paper, B-spline method is developed to find an approximate solution for singular linear and non-linear higher-order differential equation. Error analysis is presented. The method is then tested on linear and nonlinear examples. The numerical results reveal that B-spline method is very efficient and accurate.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Accurate and fast numerical solution of two-point boundary value problems for ordinary differential equations is necessary in many important scientific and engineering applications, e.g. reactant concentration in a chemical reactor, boundary layer theory, control and optimization theory, and flow networks in biology, areas of astrophysics such as the theory of stellar interiors, the thermal behavior of a spherical cloud of gas, isothermal gas spheres, and the theory of thermionic currents.
The aim of this paper is to introduce B-spline method for the numerical solution of the following class of linear and non-linear singular boundary value problems:
subject to the boundary conditions
The singular boundary-value problem has arises in many branches of applied mathematics and physics such as gas dynamics, nuclear physics, chemical reactions, atomic structures, atomic calculations, study of positive radial solutions of non-linear elliptic equations etc. In recent years, seeking numerical solutions of singular differential equations has been the focus of a number of authors [2, 4–6, 8–10, 13, 14, 17, 18, 21, 23].
In recent years, a lot of attention has been devoted to the study of B-spline method to investigate various scientific models. The efficiency of the method has been formally proved by many researchers [3, 11–13, 15, 16, 20, 22, 24, 25]. Spline functions have some attractive properties. Due to the being piecewise polynomial, they can be integrated and differentiated easily. Since they have compact support, numerical methods in which spline functions are used as a basis function lead to matrix systems including band matrices. Such systems have solution algorithms with low computational cost.
The organization of the paper is as follows. In Sect. 2, we describe the basic formulation in terms of B-splines functions required for our subsequent development. Error analysis for the septic B-spline and the nonic B-spline are presented in Sect. 3. In Sect. 4, we introduce B-splines method and show how the method is used to solve linear singular higher-order boundary-value problem. Section 5 is devoted to the solution of non-linear singular higher-order boundary-value problem. Some numerical examples are presented in Sect. 6. Finally, Sect. 7 provides conclusions of the study.
2 The B-splines of \(d{\mathrm{th}}\) degree
The theory of spline functions is a very active field of approximation theory and boundary value problems, when numerical aspects are considered. In this paper we will be interested in the septic and nonic B-splines.
Consider equally spaced knots of a partition \(\Omega _{n}:0=x_{0}<x_{1}<\cdots <x_{n}=1\) with step \(h=\frac{1}{n}\), and \(x_{i}=ih\), for \(i=0,1,2,\ldots ,n\). Let \( S_{d}(\Omega _{n})\) is the space of continuously differentiable, piecewise, d-degree polynomials on \(\Omega _{n}\). That is \( S_{d}(\Omega _{n})\) is the space of the d-degree B-spline on \(\Omega _{n}\). The \( i\mathrm{th}\) B-spline basis, \( B_{i,d}(x)\), of degree d, \(i\in \mathbf Z \) is defined recursively as follows [8, 10]:
and
where \(d\ge 1\). The above relations shown in Eqs. (2.1) and (2.2) are usually referred to as the Cox-de Boor recursion formula, such that \(B_{i,d}\) are compactly supported, \(\sum _{i=-\infty }^{\infty }B_{i,d}(x)=1\), for all \( x\in \mathbf R \) and \( B_{i,d}\ge 0\). Since \(B_{0,d}(x)\) at the knot spans \((0,h),(h,2h),\ldots ,(eh,(e+1)h),\ldots ,(dh,(d+1)h)\), for \(0\le e\le d\), can be determined using the following equation,
where
3 Error analysis
3.1 Error analysis for the septic B-spline
The set of B-spline \(B_{j}(x),j=-7,-6,\ldots ,n-1\), form a basis for \(S_{7}(\Omega _{n})\). Thus we can define our septic B-spline basis in the form:
Denote by \(S_{i}=S(x_{i}), S^{(p)}_{i}=S^{(p)}(x_{i})\) for all p. Table 1 exhibits the coefficients of septic B-spline \( B_{i,7} \) and their derivatives, at the knots \( {x_{i}, i=0,1,2, \ldots ,n}\).
For any function g evaluated at the nodes \(x_{i}\), we define \(\Gamma \) by:
then the following recursive relations can be reduced:
and
Lemma 3.1
Let S be the septic-spline interpolation of \(y\in C^{14}[0,1]\) defined by (3.1), then the following relations hold for \(i=0,1,2,\ldots ,n\)
Proof
By substituting with Taylor series expansions of \(y_{i-7},\,y_{i-6},\,y_{i-5},\,y_{i-3},\,y_{i-2}\) and \(y_{i-1}\) about \(x_{i-4}\) in Eqs. (3.3)–(3.8), the above relations are obtained. \(\square \)
Theorem 3.1
If \(y\in C^{14}[0,1]\) and S is the septic B-spline interpolation of y defined by (3.1), then we have
Proof
Consider any function \(g\in C^{14}[0,1]\), \(\Gamma {g}\) is defined as shown in Eq. (3.2). It can be easily proved that
If we assume that
then using Eq. (3.15) yields:
Let
subtracting Eq. (3.16) from Eq. (3.14) yields:
If we assume that
then using Eq. (3.15) yields:
Let
subtracting Eq. (3.18) from Eq. (3.13) yields:
If we assume that
then using Eq. (3.15) yields:
Let
subtracting Eq. (3.20) from Eq. (3.12) yields:
If we assume that
then using Eq. (3.15) yields:
Let
subtracting Eq. (3.22) from Eq. (3.11) yields:
If we assume that
then using Eq. (3.15) yields:
Let
subtracting Eq. (3.24) from Eq. (3.10) yields:
If we assume that \(g_{i}=y^{\prime }_{i}\), then using equation (3.15) yields:
Let \(d_{1,i}=\left( S^{\prime \prime }_{i}-\left( 5040y^{\prime }{i}+1680h^{2}y^{(3)}_{i}+266h^{4}y^{(5)}_{i}+\frac{80}{3}h^{6}y^{(7)}_{i} \right) \right) \), subtracting Eq. (3.26) from Eq. (3.9) yields:
Since the coefficient matrices of the systems of equations (3.17), (3.19), (3.21), (3.23), (3.25) and ( 3.27) are diagonally dominant and their inverses are bounded then \(d_{k,i}=\text{ O }\left( h^{8}\right) ,\,k=1,2,\ldots ,6,\, i=0,1,2, \ldots ,n\), hence; the proof of all relations of the above theorem is completed. \(\square \)
3.2 Error analysis for the nonic B-spline
The set of B-spline \(B_{j}(x),j=-9,-8,\ldots ,n-1\), form a basis for \(S_{9}(\Omega _{n})\). Thus we can define our nonic B-spline basis in the form:
Table 2 exhibits the coefficients of nonic B-spline \( B_{i,9} \) and their derivatives, at the knots \( {x_{i}, i=0,1,2, \ldots ,n}\).
For any function g evaluated at the nodes \(x_{i}\), we define \(\Gamma \) by:
Then the following recursive relations can be reduced:
Lemma 3.2
Let S be the nonic B-spline interpolation of \(y\in C^{16}[0,1]\) defined by (3.28), then the following relations hold for \(i=0,1,2,\ldots ,n\)
Proof
By substituting with Taylor series expansions of \(y_{i-9},\,y_{i-8},\,y_{i-7},\,y_{i-6},\,y_{i-4},\) \(\,y_{i-3},\,y_{i-2}\) and \(y_{i-1}\) about \(x_{i-5}\) in equations (3.30)–(3.45), the above relations are obtained. \(\square \)
Theorem 3.2
If \(y\in C^{16}[0,1]\) and S is the nonic B-spline interpolation of y defined by (3.28), then we have
Proof
Consider any function \(g\in C^{16}[0,1]\), \(\Gamma {g}\) is defined as shown in Eq. (3.29). It can be easily proved that
If we assume that
then using Eq. (3.46) yields:
Let
subtracting Eq. (3.47) from Eq. (3.45) yields:
If we assume that
then using Eq. (3.46) yields:
Let
subtracting Eq. (3.49) from Eq. (3.44) yields:
If we assume that
then using Eq. (3.46) yields:
Let
subtracting Eq. (3.51) from Eq. (3.43) yields:
If we assume that \(g_{i}=y^{(5)}_{i}+\frac{h^{6}}{6048}y^{(11)}_{i}-\frac{h^{8}}{34560}y^{(13)}_{i}\), then using Eq. (3.46) yields:
Let
subtracting Eq. (3.53) from Eq. (3.42) yields:
If we assume that
then using Eq. (3.46) yields:
Let
subtracting Eq. (3.55) from Eq. (3.41) yields:
If we assume that \(g_{i}=y^{(3)}_{i}-\frac{h^{8}}{172{,}800}y^{(11)}_{i}\), then using Eq. (3.15) yields:
Let
subtracting Eq. (3.57) from Eq. (3.40) yields:
If we assume that \(g_{i}=y^{\prime \prime }_{i}+\frac{h^{8}}{12{,}096{,}000}\,y^{(10)}_{i}\), then using Eq. (3.15) yields:
Let
subtracting Eq. (3.59) from Eq. (3.39) yields:
If we assume that \(g_{i}=y^{\prime }_{i}\), then using Eq. (3.47) yields:
Let
subtracting Eq. (3.61) from Eq. (3.38) yields:
Since the coefficient matrices of the systems of Eqs. (3.48), (3.50), (3.52), (3.54), (3.56), (3.58), (3.60) and (3.62) are diagonally dominant and their inverses are bounded then \(\tilde{d}_{k,i}=O(h^{10}),\,k=1(1)6,\,i=0, 1, 2, \ldots , n\), hence; the proof of all relations of this theorem is completed. \(\square \)
4 Linear singular higher-order boundary-value problems
To overcome the singularity at \(x=0\), we apply L’Hopital’s rule as x approaches zero to the terms \(\frac{k}{x}y^{\prime }(x)\) and \(\frac{k}{x^{2}}y(x)\) in Eq. (1.1) as follows [8]:
since
similarly
So the BVP in Eq. (1.1) is modified at the singular point \(x=0\) to the following form
Let the solution y(x) of the problem (1.1)–(4.1) be approximated by
where \(c_j\) are unknown real coefficients and \(B_j(x)\) are the (2r+1)-degree B-spline functions. Let \(x_0, x_1, \ldots , x_n\) be \(n + 1\) grid points in the interval [0, 1] so that \( x_i = ih, i = 0, 1, 2, \ldots , n\) where
It is required that the approximate solution satisfies the differential equation at the points \(x = x_i\), and we can easily deduce the following
Theorem 4.1
If the assumed approximate solution of the problem (1.1)–(4.1) is (4.2) then the discrete collocation system for the determination of the unknown coefficients \(\{c_j\}_{j=-d}^{n-1}\) is given by
and
and boundary conditions (1.2) can be written as
Proof
We replace each term of (1.1)–(4.1) with its corresponding approximation given by (4.3) and substituting \(x = x_i\) and applying the collocation to it. \(\square \)
Then the system in (4.4)–(4.6) takes the matrix form
i.e.
and
Notice that, \( \mathbf A _{0}\) is \( n+1\,\times \, r \) dimensional matrix, its coefficients are the coefficients of the boundary conditions equations at \( x_{0},(x=0)\), \( \mathbf A _{n}\) is \( n+1\,\times \, r \,\) dimensional matrix, its coefficients are the coefficients of the boundary conditions equations at \( x_{n},(x=1)\) and \( \mathbf A \) is a (2r+1)-diagonal matrix of order \( n+1\,\times \,n+2r+1\, \) with d non-zero bands, such that its elements have the following form:
where \( jj=i,i+1,\ldots ,n+d-1 \), and \( i=1, 2,\ldots ,n \). Now, we have a linear system of \( n+2r+1 \) equations of the \( n+d \) unknown coefficients, namely, \(\{C_j\}_{j=-d}^{n-1}\). We can obtain the coefficients of the approximate solution by solving this linear system by Q-R method.
5 Non-linear singular higher-order boundary-value problems
In the case of non-linear problems, the quesilinearization technique has been used to linearize the given non-linear (1.1)–(4.1) to a sequence of a linear differential equations [10]. We choose a reasonable initial approximation for the function y(x) in \( f\left( x, y(x), y^{\prime }(x), \ldots , y^{(2r-1)}(x)\right) \), call it as \( y^{0}(x) \), and expand \( f\left( x,y(x),y^{\prime }(x),\ldots ,y^{(2r-1)}(x)\right) \) around the function \( y^{0}(x) \), then we obtain
or in general, we can write the first two-terms of the expansion for \( m= 0, 1, 2, \ldots \), ( m is the iteration index) as:
Substituting Eq. (5.1) in Eqs. (1.1) and (4.1), the non-linear differential equation will be converted to system of \( m+1\) linear differential equations written as
and
subject to the boundary conditions
Now, we seek a function y(x) that approximates the solution of Eqs. (5.2)–(5.3), which may be represented at \( n+1\) knots \({x_{i}}\). Hence, \( y(x_{i})\) and its \((2r)^\mathbf{th }\) derivatives can be written at \((m+1)^\mathbf{th }\) iteration as
where \( c_{j}^{m+1} \) are unknown real coefficients.
Theorem 5.1
If the assumed approximate solution of the problem (5.2)–(5.3) is (5.5) then the discrete collocation system for the determination of the unknown coefficients \(\{c_j\}_{j=-d}^{n-1}\) is given by
and
Proof
We replace each term of (5.2)–(5.3) with its corresponding approximation given by (5.5) and substituting \(x = x_i\) and applying the collocation to it. \(\square \)
Then the system in (5.6)–(5.7) takes the matrix form
where
and
where:
Notice that, \( \mathbf Q ^{m}_{0}\) is \( n+1\times r \) dimensional matrix, its coefficients are the coefficients of the boundary conditions at \( x_{0},(x=0)\), \( \mathbf Q ^{m}_{n}\) is \( n+1\,\times \, r \) dimensional matrix, its coefficients are the coefficients of the boundary conditions equations at \( x_{n},(x=1)\) and \( \mathbf Q ^{m}\) is a (2r+1)-diagonal matrix of order \( n+1\,\times \, n+2r+1 \) with d non-zero bands, such that its elements have the following form:
where \( zz=i,i+1,\ldots ,n+d-1 \), and \( i=1, 2,\ldots ,n \). Now, we have a linear system of \( n+2r+1 \) equations of the \( n+2r+1 \) unknown coefficients, namely, \( {c^{m}_{j}, j= - d,\ldots , n-1, m=0, 1, \ldots }\) We can obtain the coefficients of the approximate solution by solving this linear system.
6 Numerical results
We present some test examples constructed so that the analytical solution was known before-hand. The performance of the B-spline method is measured by the maximum absolute error \(E_{B-\text{ spline }}\) which is defined as
All computations were carried out using MATLAB 7.01. For the two first examples, we use B-spline of \(9^{th}\) degree, the coefficients of \( B_{i,9}\) and their derivatives, at the knots \( {x_{i}, i=0, 1, 2, \ldots , n} \) are shown in Table 2 and the two last examples, we use B-spline of \(7^{th}\) degree, the coefficients of \( B_{i,7}\) and their derivatives, at the knots \( {x_{i}, i=0, 1, 2, \ldots , n} \) are shown in Table 1.
Example 1
This is linear BVP
subject to the boundary conditions:
whose exact solution is
This problem is solved using nonic B-spline \( B_{9}(x_{i}), i=0, 1, \ldots , n\) with \(n=20\). The results are tabulated in Table 3. Also, the maximum absolute errors of the results at different n obtained by our method are tabulated in Table 4.
Example 2
This is linear BVP
subject to the boundary conditions
whose exact solution is
This problem is solved using nonic B-spline \( B_{9}(x_{i}), i=0, 1, \ldots , n\) with \(n=10\). The results are tabulated in Table 5.
Example 3
Consider the nonlinear boundary value problem
subject to the boundary conditions
whose exact solution is
This problem is solved using septic B-spline \( B_{7}(x_{i}), i=0, 1, \ldots , n\) with \(n=20\) with \(m=2\). The results are tabulated in Table 6.
Example 4
This is nonlinear BVP
subject to the boundary conditions
whose exact solution is
This problem is solved using septic B-spline \( B_{7}(x_{i}), i=0, 1, \ldots , n\) with \(n=20\) with \(m=2\). The results are tabulated in Table 7.
7 Conclusion
We presented a method for solving singular linear and nonlinear higher-order boundary value problem. This method is easy to implement and yields the desired accuracy and numerical results demonstrate this. We observed that the method works well for non-linear differential equations. Thus the proposed method is suggested as an efficient method for solving this problem.
References
Agarwal, R., Chow, Y.: Finite difference methods for boundary value problems for differential equations with deviating arguments. Comput. Math. Appl. 12(11), 1143–1153 (1986)
Ali, A., Gardner, G., Gardner, L.: A collocation solution for Burgers equation using cubic B-spline finite elements. Comput. Math. Appl. Mech. Eng. 100, 325–337 (1992)
Aziz, T., Kumar, M.: A fourth-order finite-difference method based on non-uniform mesh for a class of singular two-point boundary-value problems. J. Comput. Appl. Math. 136, 337–342 (2001)
Aziz, T., Khan, A., Rashidinia, J.: A fourth-order finite-difference method based on non-uniform mesh for a class of singular two-point boundary-value problems. Appl. Math. Comput. 167, 153–166 (2005)
Baxley, J.: Some singular nonlinear boundary value problems. J. Comput. Appl. Math. 22, 463–469 (1991)
Baxley, J.: Numerical solution of singular nonlinear boundary value problems. In: Bainov, D., Covachev, V. (eds.) 3rd International Colloquium on Numerical Analysis, pp. 15–24. VSP, Utrecht (1995)
Bellman, R., Kalaba, R.: Quasilinearization and nonlinear boundary value problems. American Elsevier, New York (1965)
de Boor, C.: A practical guide to splines. Springer, New York (1978)
Caglar, N., Caglar, H.: B-spline solution of singular boundary value problems. Appl. Math. Comput. 182, 1509–1513 (2006)
Caglar, N., Caglar, H.: Fifth-degree B-spline solution for a fourth-order parabolic partial differential equations. Appl. Math. Comput. 201, 597–603 (2008)
Chawla, M., Mckee, S., Shaw, G.: Order h2 method for a singular two-point boundary-value problem. BIT 26, 318–326 (1986)
Chawla, M., Shivkumar, P.: On existence of solutions of a class of singular two-point boundary value problems. J. Comput. Appl. Math. 19, 379–388 (1987)
Chawla, M., Subramanian, R.: A new spline method for singular two point boundary value problems. J. Inst. Math. Appl. 24, 219–310 (1988)
Chawla, M., Subramanian, R., Sathi, H.: A fourth order method for a singular two point boundary value problem. BIT 28, 88–97 (1988)
El-Gamel, M., El-Shenawy, A.: The solution of a time-dependent problem by the B-spline method. J. Comput. Appl. Math. 267, 254–265 (2014)
Gustafsson, B.: A numerical method for solving singular boundary-value problems. Numer. Math. 21, 328–344 (1973)
Golub, G., Vanloan, C.: Matrix Computations, 3rd edn. The Johns Hopkins Press Ltd, London (1996)
I-Haq, F., Ul-Islam, S., Tirmizi, I.A.: Numerical technique for solution of the MRLW equation using quartic B-splines. Appl. Math. Model. 34, 4151–4160 (2010)
Jain, R.K., Jain, P.: Finite difference methods for a class of singular two-point boundary value problems. Int. J. Comput. Math. 27, 113–120 (1989)
Lakestani, M., Dehghan, M.: Numerical solutions of the generalized Kuramoto-Sivashinsky equation using B-spline functions. Appl. Math. Model. 36, 605–617 (2012)
Lee, K., Lee, Y.: Multiple positive solutions of singular two-point boundary-value problems for second-order impulsive differential equations. Appl. Math. Comput. 158, 745–759 (2004)
Mittal, R., Jain, R.: Numerical solutions of nonlinear Burgers equation with modified cubic B-splines collocation method. Appl. Math. Comput. 218(15), 7839–7855 (2012)
Russell, R., Shampine, L.: Numerical methods for singular boundary-value problems. J. Numer. Anal. 12, 13–36 (1975)
Taliaferro, S.: A nonlinear singular boundary value problem. Nonlinear Anal. Theory Methods Appl. 3, 897–904 (1979)
Usmani, R., Warsi, S.: Quintic spline solutions of boundary value problems. Comput. Math. Appl. 6, 197–203 (1980)
Acknowledgments
The authors are grateful for the referees for their valuable comments and suggestions on the original manuscript.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
El-Gamel, M., El-Shamy, N. B-spline and singular higher-order boundary value problems. SeMA 73, 287–307 (2016). https://doi.org/10.1007/s40324-016-0069-x
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s40324-016-0069-x