El-Zahar et al.

Advances in Difference Equations (2020) 2020:411

https://doi.org/10.1186/s13662-020-02862-z

RESEARCH Open Access

Absolutely stable difference scheme for a

general class of singular perturbation
problems
Essam R. El-Zahar1,2* , A.M. Alotaibi3 , Abdelhalim Ebaid3 , Dumitru Baleanu4,5,6 , José Tenreiro
Machado7 and Y.S. Hamed8,9
*
Correspondence:
[email protected]; Abstract
[email protected]
1
Department of Mathematics, This paper presents an absolutely stable noniterative difference scheme for solving a
College of Science and Humanities general class of singular perturbation problems having left, right, internal, or twin
in Al-Kharj, Prince Sattam bin boundary layers. The original two-point second-order singular perturbation problem
Abdulaziz University, P.O. Box 83,
Al-Kharj, 11942, Saudi Arabia is approximated by a first-order delay differential equation with a variable deviating
2
Department of Basic Engineering argument. This delay differential equation is transformed into a three-term difference
Science, Faculty of Engineering, equation that can be solved using the Thomas algorithm. The uniqueness and
Menoufia University, Shebin El-Kom,
32511, Egypt stability analysis are discussed, showing that the method is absolutely stable. An
Full list of author information is optimal estimate for the deviating argument is obtained to take advantage of the
available at the end of the article second-order accuracy of the central finite difference method in addition to the
absolute stability property. Several problems having left, right, interior, or twin
boundary layers are considered to validate and illustrate the method. The numerical
results confirm that the deviating argument can stabilize the unstable discretized
differential equation and that the new approach is effective in solving the considered
class of singular perturbation problems.
Keywords: Singular perturbation problems; Finite difference schemes; Absolutely
stable; Boundary and interior layers

1 Introduction
Singular perturbation problems (SPPs), also known as stiff, arise very frequently in var-
ious areas of applied science, such as fluid mechanics, heat transfer, materials science,
chemical and electrical engineering, superconductor theory, and chemical reactor sys-
tems [1–9]. It is a well-known fact that SPPs exhibit stiff internal or boundary layers as the
singular perturbation parameter approaches zero [10–49]. As this happens, the stiffness
ratio increases and the classical numerical methods become inadequate for solving such
problems over standard locally uniform meshes, where nonphysical oscillations emerge
in the numerical solution due to the stability restriction on the step-size. An interesting
collection of examples of exotic numerical solutions of singular perturbation problems
obtained using inadequate numerical discretization schemes can be found in [10]. The
construction of special-purpose numerical methods for ensuring convergence regardless

© The Author(s) 2020. This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use,
sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original
author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other
third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 2 of 17

of the value of the perturbation parameter was first raised in [11, 12] using fitting factor
(stabilization term) and fitting mesh (layer adapted mesh) techniques. It is well known that
depending on the behavior of the differential operator coefficients in SPPs, the solutions
exhibit a boundary layer, twin boundary-layers, or interior layer [3, 4, 13–18]. A variety
of numerical and semianalytical numerical methods for solving different classes of SPPs
exhibiting a boundary layer, twin boundary-layers, or interior layer can be found in [18–
49]. In all of these SPP classes, a priori information about the location of the layers is used
to construct appropriate numerical methods for SPPs. And consequently, the numerical
methods developed for SPPs have different numerical integration algorithms depending
on the location of the layers. The purpose of this paper is to present a noniterative abso-
lutely stable difference scheme for solving a general class of SPPs having left, right, inter-
nal, or twin boundary layers. The method does not depend on asymptotic expansions and
needs no prior information about the location of the layers, and so it is designed to be suit-
able for the practicing engineers and applied mathematicians who need a practical tool for
solving SPPs (see, for example, [15, 38] for asymptotic techniques, designed for a similar
purpose). The method depends on replacing the original two-point second-order SPP by
a first-order approximate delay differential equation with a variable deviating argument.
This delay differential equation is transformed into a three-term difference equation that
can be solved using the Thomas algorithm [50]. The stability analysis of the method is
discussed showing that the method is absolutely stable under a certain condition on the
deviating argument whereas there is no stability restriction on the step-size. An optimal
estimate for the deviating argument is obtained to take advantage of the second-order ac-
curacy of the central finite difference method [48, 50] in addition to the absolute stability
property [19, 28, 48–50]. Several problems having left, right, interior, or twin boundary
layers are considered to validate and illustrate the method. To analyze the effect of the de-
viating argument on the solution accuracy, the maximum absolute solution error is pre-
sented in tables and figures for constant and variable deviating argument values. The nu-
merical results confirm that the deviating argument can stabilize the unstable discretized
differential equation and that the new approach is effective in solving the considered class
of singular perturbation problems.

2 Description of the method

Consider the following linear SPP:

–εy (x) + p(x)y (x) + q(x)y(x) = r(x), (1)

with BCs

y(a) = A, y(b) = B,

where 0 < ε  1, a, b, A and B are given constants, and p(x), q(x) and r(x) ∈ C 1 (x), x ∈ [a, b].
Moreover, we assume that q(x) ≥ 0 for all x ∈ [a, b]. Under these assumptions, problem
(1) has a unique solution with boundary or interior layers [13–49]. The interval [a, b] is
divided into N subintervals [xi , xi+1 ], i = 0, 1, 2, . . . , N – 1, each of length h, i.e., h = (b – a)/N
and xi = a + ih. For the sake of simplicity, we use pi = p(xi ), qi = q(xi ), ri = r(xi ), yi–1 = y(xi–1 ),
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 3 of 17

yi+1 = y(xi+1 ), yi = y (xi ). Dividing Eq. (1) by –ε and letting ω = 1/ε results in

yi – ωpi yi – ωqi yi = –ωfi . (2)

Let δi = δ(xi ) be a positive variable deviating argument (0 < δi  1). Then by using the
Taylor expansion about the point xi , we have

δi2 
y(xi – δi ) ∼
= yi – δi yi + y . (3)
2 i

By substituting Eq. (3) into Eq. (2), we have

2y(xi – δi ) – 2yi + 2δi yi – δi2 ωpi yi – δi2 ωqi yi ∼

= –δi2 ωfi . (4)

Now, applying standard forward and centered finite difference formulas for yi in Eq. (4)
results in
   
yi+1 – yi yi+1 – yi–1
2y(xi – δi ) – 2yi + 2δi – δi2 ωpi – δi2 ωqi yi ∼
= –δi2 ωfi . (5)
h 2h

Again, using the Taylor expansion and the standard backward finite difference, we obtain:

yi – yi–1
y(xi – δi ) ∼
= yi – δi . (6)
h

By substituting Eq. (6) into Eq. (5), we get the following difference equation:

Ei yi–1 + Fi yi + Gi yi+1 ∼
= Hi , i = 1, 2, . . . , N – 1, (7)

where

2 ωpi
Ei = + , (8)
hδi 2h
4
Fi = – – ωqi , (9)
hδi
2 ωpi
Gi = – , (10)
hδi 2h
Hi = –ωfi . (11)

The difference Eq. (7) and the two BCs in (1) results in a tridiagonal system that can be
easily solved using the Thomas algorithm to obtain the unknowns y1 to yN–1 .

2.1 Thomas algorithm

A brief description of the Thomas algorithm [50], is presented as follows.
In this algorithm, the solution of the difference Eq. (7) can be written as

yi = Wi yi+1 + Ti , i = 1, 2, . . . , N – 1. (12)
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 4 of 17

where Wi and Ti are to be determined. From Eq. (12) we have

yi–1 = Wi–1 yi + Ti–1 , (13)

and substituting (12) and (13) into (7), we get

Gi Hi – Ei Ti–1
yi = – yi+1 + . (14)
Fi + Ei Wi–1 Fi + Ei Wi–1

Comparing (14) and (12) results in

Gi
Wi = – , i = 1, 2, . . . , N – 1, (15)
Fi + Ei Wi–1
Hi – Ei Ti–1
Ti = , i = 1, 2, . . . , N – 1. (16)
Fi + Ei Wi–1

Starting with initial conditions W0 = 0 and T0 = y0 = A, the values of Wi and Ti for

i = 1, 2, . . . , N – 1 are computed in a forward process, from Eqs. (15)–(16), and then yi
is computed in a backward process, from Eq. (12), using the BC yN = B.

3 Uniqueness and stability analysis

In this section, the uniqueness and stability condition of the proposed algorithm for the
difference Eq. (7) is analyzed.

Theorem 1 The numerical scheme Ei yi–1 + Fi yi + Gi yi+1 ∼

= Hi defined in (7)–(11) results in
a unique stable solution provided

4
δi ≤ . (17)
ω|pi |

Proof The existence of a unique solution to the tridiagonal system results from establish-
ing that the tridiagonal coefficient matrix of the system is diagonally dominant. It is clear
that Ei in (8) and Gi in (10) are nonnegative when hδ2 i + ωp2h
i
≥ 0 and hδ2 i – ωp 2h
i
≥ 0, respec-
tively. These inequalities are verified if the condition δi ≤ ω|pi | is verified. Following this
4

condition, Fi is negative and verifies the inequality |Fi | ≥ |Gi | + |Ei | = hδ4 i . Thus, under con-
dition (17), the tridiagonal coefficient matrix is diagonally dominant, and the numerical
scheme has a unique solution.
For stability analysis, suppose that a small numerical error τi–1 was introduced in the
calculation of Wi–1 so that it yields an approximate value W̃i–1 such that

W̃i–1 = Wi–1 + τi–1 . (18)

Then, from (18) and (15), we have

   
Gi Gi
τi = – – –
Fi + Ei (Wi–1 + τi–1 ) Fi + Ei Wi–1
   2 
Gi Ei τi–1 Wi E i
= –  τi–1 , (19)
(Fi + Ei (Wi–1 + τi–1 ))(Fi + Ei Wi–1 ) Gi
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 5 of 17

   
 2  Ei   Gi Ei 
   
|τi |  Wi  |τi–1 | =  |τi–1 |. (20)
Gi (Fi + Ei Wi–1 )2 

Thus, the numerical scheme defined by (7)–(11) results in a unique stable solution under
condition (17). 

Now, if we chose δi = ω|p4 i | , then for the left-end boundary layers, pi < 0, we have Ei = 0,
and the present scheme is reduced to a stable forward integration scheme. Also, on the
other hand, for the right-end boundary layers, pi > 0, we have Gi = 0, and the present
scheme is reduced to a stable backward integration scheme. Moreover, in the above two
cases, the propagation error in (20) has vanished.
Moreover, at pi = 0, for twin-boundary layers or internal layers, or at h ≤ ω p2i ∞ , we set
δi = 2h ≤ ω|p4 i | , and the present scheme is reduced to the standard central finite difference
scheme (CFD) in [48, 50]. Thus, to take advantage of the second-order accuracy of the
standard central finite difference scheme in addition to the absolute stability property, δi
can be selected according to
 
4
δi = δop = min , 2h . (21)
ω|pi |

The stability in CFD is controlled by the step-size restriction whereas the present scheme
is absolutely stable without any restrictions on the step-size.
These details are combined in the following algorithm:

3.1 Algorithm steps

Step I: compute
h = (b – a)/(N),
x = (a : h : b)
Step II: set xx = 0, index = 0, w = 1/ε
Set W = [0, 0, . . . , 0], T[0, 0, . . . , 0] where the vector length is N
Set T(1) = A, W (1) = 0
Step II: while xx ≤ b – h, compute
(i) xx = x(index);
(ii) p = p(xx); q = q(xx); r = r(xx)
(iii) δ = min(4/(w ∗ abs(p + 1E – 16)), 2h)
(iv) E(index) = 2/h + δwp/(2h)
(v) G(index) = 2/h – δwp/(2h)
(vi) F(index) = 4/h – δwq
(vii) H(index) = –δwf
(viii) index = index + 1
Step III: Do Steps (i) and (ii) for index2 = 1, 2, 3, . . . , N – 1
G(index2)
(i) Compute W (ind) = – G(index2)+E(index2)∗W (index2–1)
F(index2)∗T(index2–1)–H(index2)
(ii) Compute T(ind) = G(index2)+E(index2)∗W (index2–1)
Step IV: Compute the solution
(i) set Sol = [0, 0, . . . , 0], where the vector length is N – 2
(ii) Set Sol(N) = B
(iii) Compute Sol(index2 – 1) = W (index2 – 1) ∗ Sol(index2) + T(index2 – 1) where
index2 = N, N – 1, . . . , 3, 2, 1
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 6 of 17

This algorithm is easily adaptable in any mathematical environment. We present the

Matlab code in Appendix A.

4 Numerical results
To demonstrate the applicability of the method, we consider in the following several prob-
lems having left, right, interior, or twin boundary layers. These problems have been dis-
cussed in the literature and their approximate solutions are available for comparison.

4.1 Left-end boundary-layer problems

Example 1 Consider the following SPBVP [2, 7, 13–15, 38]

εy (x) + y (x) – y(x) = 0, x ∈ [0, 1], (22)

with BCs y(0) = 1 and y(1) = 1. The exact solution is given by

(em2 – 1)em1 x + (1 – em1 )em2 x

y(x) = ,
em2 – em1
√ √
where m1 = (–1 + 1 + 4ε)/(2ε) and m2 = (–1 – 1 + 4ε)/(2ε).
The solution errors are listed in Table 1 for ε = 10–3 and ε = 10–4 at h = 0.001. The
solutions for ε = 10–4 using the present absolutely stable central difference (ASCD) and
the standard CFD schemes are presented in Fig. 1.

Example 2 Consider the following SPBVP [7, 13–15, 38]

εy (x) + y (x) = 1 + 2x, x ∈ [0, 1], (23)

with BCs y(0) = 0 and y(1) = 1. The exact solution is given by

(2ε – 1)(1 – e–x/ε )

y(x) = x(x + 1 – 2ε) + .
1 – e–1/ε

Table 1 The solution errors of Example 1 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.000 6.0951e–14 0.0000e+00 6.0951e–14 1.0592e–13 0.0000e+00 0.0000e+00
0.001 2.1803e–02 2.3236e–01 2.1803e–02 4.2100e–01 1.1841e–04 1.1841e–04
0.002 1.5273e–02 8.5511e–02 1.5273e–02 2.8036e–01 1.4708e–04 1.4708e–04
0.003 8.0307e–03 3.1542e–02 8.0307e–03 1.8672e–01 1.4708e–04 1.4708e–04
0.004 3.7564e–03 1.1708e–02 3.7564e–03 1.2436e–01 1.4708e–04 1.4708e–04
0.005 1.6486e–03 4.4190e–03 1.6486e–03 8.2822e–02 1.4708e–04 1.4708e–04
0.006 6.9512e–04 1.7401e–03 6.9512e–04 5.5160e–02 1.4708e–04 1.4708e–04
0.007 2.8515e–04 7.5553e–04 2.8515e–04 3.6736e–02 1.4708e–04 1.4708e–04
0.008 1.1466e–04 3.9370e–04 1.1466e–04 2.4466e–02 1.4708e–04 1.4708e–04
0.009 4.5397e–05 2.6072e–04 4.5397e–05 1.6295e–02 1.4707e–04 1.4707e–04
0.010 1.7743e–05 2.1185e–04 1.7743e–05 1.0852e–02 1.4707e–04 1.4707e–04
0.020 5.9987e–08 1.8343e–04 5.9987e–08 1.8638e–04 1.4705e–04 1.4705e–04
0.050 6.1017e–08 1.8322e–04 6.1017e–08 6.2155e–08 1.4689e–04 1.4689e–04
0.100 6.0767e–08 1.8247e–04 6.0767e–08 6.0963e–08 1.4629e–04 1.4629e–04
0.300 5.7716e–08 1.7332e–04 5.7716e–08 5.7913e–08 1.3896e–04 1.3896e–04
0.500 5.0343e–08 1.5119e–04 5.0343e–08 5.0524e–08 1.2123e–04 1.2123e–04
1.000 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 7 of 17

Figure 1 The solution of Example 1 for ε = 10–4 using the ASCD and CFD schemes

Table 2 The solution errors of Example 2 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.000 3.6835e–13 0.0000e+00 3.6835e–13 1.3312e–12 0.0000e+00 0.0000e+00
0.001 3.4477e–02 3.6814e–01 3.4477e–02 6.6658e–01 7.5381e–04 7.5381e–04
0.002 2.4176e–02 1.3606e–01 2.4176e–02 4.4436e–01 7.9840e–04 7.9840e–04
0.003 1.2725e–02 5.0684e–02 1.2725e–02 2.9624e–01 7.9760e–04 7.9760e–04
0.004 5.9580e–03 1.9275e–02 5.9580e–03 1.9749e–01 7.9680e–04 7.9680e–04
0.005 2.6175e–03 7.7195e–03 2.6175e–03 1.3166e–01 7.9600e–04 7.9600e–04
0.006 1.1048e–03 3.4678e–03 1.1048e–03 8.7774e–02 7.9520e–04 7.9520e–04
0.007 4.5373e–04 1.9031e–03 4.5373e–04 5.8516e–02 7.9440e–04 7.9440e–04
0.008 1.8268e–04 1.3268e–03 1.8268e–04 3.9011e–02 7.9360e–04 7.9360e–04
0.009 7.2459e–05 1.1142e–03 7.2459e–05 2.6007e–02 7.9280e–04 7.9280e–04
0.010 2.8408e–05 1.0353e–03 2.8408e–05 1.7338e–02 7.9200e–04 7.9200e–04
0.020 1.7708e–09 9.8000e–04 1.7708e–09 3.0067e–04 7.8400e–04 7.8400e–04
0.050 5.5511e–16 9.5000e–04 5.5511e–16 1.5680e–09 7.6000e–04 7.6000e–04
0.100 4.4409e–16 9.0000e–04 4.4409e–16 8.8818e–16 7.2000e–04 7.2000e–04
0.300 2.2204e–16 7.0000e–04 2.2204e–16 3.3307e–16 5.6000e–04 5.6000e–04
0.500 2.2204e–16 5.0000e–04 2.2204e–16 3.3307e–16 4.0000e–04 4.0000e–04
1.000 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00

The solution errors are presented in Table 2 for ε = 10–3 and ε = 10–4 at h = 0.001. The
solutions for ε = 10–4 using the ASCD and CFD schemes are depicted in Fig. 2.

Example 3 Consider the following SPBVP [47]

1 
–εy (x) – y (x) = 0, x ∈ [0, 1], (24)
1+x

with BCs y(0) = 0 and y(1) = 1. The exact solution is given by

 
(1 + x)(1–1/ε) – 1
y(x) = .
(2)(1–1/ε) – 1

The solution errors are presented in Table 3 for ε = 10–3 and ε = 10–4 at h = 0.001. The
solutions for ε = 10–4 using the ASCD and CFD schemes are present in Fig. 3.
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 8 of 17

Figure 2 The solution of Example 2 for ε = 10–4 using the ASCD and CFD schemes

Table 3 The solution errors of Example 3 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.000 8.4953e–14 0.0000e+00 8.4953e–14 9.0494e–14 0.0000e+00 0.0000e+00
0.001 3.4431e–02 3.6843e–01 3.4431e–02 6.6655e–01 4.5673e–05 4.5673e–05
0.002 2.4173e–02 1.3588e–01 2.4173e–02 4.4404e–01 2.1069e–09 2.1069e–09
0.003 1.2753e–02 5.0161e–02 1.2753e–02 2.9570e–01 9.8144e–14 9.8144e–14
0.004 5.9921e–03 1.8536e–02 5.9921e–03 1.9684e–01 0.0000e+00 0.0000e+00
0.005 2.6446e–03 6.8565e–03 2.6446e–03 1.3098e–01 0.0000e+00 0.0000e+00
0.006 1.1226e–03 2.5387e–03 1.1226e–03 8.7114e–02 0.0000e+00 0.0000e+00
0.007 4.6420e–04 9.4093e–04 4.6420e–04 5.7916e–02 0.0000e+00 0.0000e+00
0.008 1.8838e–04 3.4908e–04 1.8838e–04 3.8489e–02 0.0000e+00 0.0000e+00
0.009 7.5392e–05 1.2964e–04 7.5392e–05 2.5567e–02 0.0000e+00 0.0000e+00
0.010 2.9856e–05 4.8189e–05 2.9856e–05 1.6977e–02 0.0000e+00 0.0000e+00
0.020 2.1785e–09 2.5611e–09 2.1785e–09 2.7644e–04 0.0000e+00 0.0000e+00
0.050 1.3323e–15 0.0000e+00 1.3323e–15 9.2895e–10 0.0000e+00 0.0000e+00
0.100 1.4433e–15 0.0000e+00 1.4433e–15 4.4409e–15 0.0000e+00 0.0000e+00
0.300 1.3323e–15 0.0000e+00 1.3323e–15 4.4409e–15 0.0000e+00 0.0000e+00
0.500 2.5535e–15 0.0000e+00 2.5535e–15 4.8850e–15 0.0000e+00 0.0000e+00
1.000 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00

The results in Tables 1, 2 and 3 show that at δi = 4/ω|pi | or δi = δop = 4/ω|pi |, there is no
propagation error with the Thomas algorithm (20). Also, at ε = 10–3 the ASCD is reduced
to the CFD scheme, where δop = 2h, and the results are identical. Although the ASCD
stability is verified at δi = 4/ω|pi |, selecting δi = δop takes advantage of the second-order
accuracy of the CFD in addition to the absolute stability property depending on the step-
size value. The solution obtained at ε = 10–4 using the ASCD is more accurate than that
obtained by CFD due to the absolute stability property of the ASCD, while the stability
restriction of CFD results in nonphysical oscillations in the numerical solution as illus-
trated in Figs. 1, 2 and 3. Figure 4 shows the solution error of Example 3 at ε = 10–4 using
different values of the deviating argument δi = k/ω|pi |, k = 4, 3, 2, 1. The results in Fig. 4
reveal that a more accurate solution is obtained at δi = 4/ω|pi |. Moreover, Fig. 5 depicts the
solution error of Example 3 at ε = 10–4 using constant and variable deviating arguments
δi = 4/ω|pi |∞ , δi = 4/ω|pi |, respectively. Figure 5 reveals that a more accurate solution is
obtained using a variable deviating argument δi .
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 9 of 17

Figure 3 The solution of Example 3 for ε = 10–4 using the ASCD and CFD schemes

Figure 4 The solution errors of Example 3 for ε = 10–4 and different values of δi

4.2 Right-end boundary-layer problems

Example 4 Consider the following SPBVP [2, 13]

εy (x) – y (x) = 0, x ∈ [0, 1], (25)

with BCs y(0) = 1 and y(1) = 0. The exact solution is given by

e(x–1)/ε – 1
y(x) = .
e–1/ε – 1

The solution errors are reported in Table 4 for ε = 10–3 and ε = 10–4 at h = 0.001. The
numerical solution using the ASCD and CFD schemes for ε = 10–4 are presented in Fig. 6.

Example 5 Consider the following SPBVP

εy (x) – y (x) – (1 + ε)y(x) = 0, (26)

El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 10 of 17

Figure 5 The solution errors of Example 3 for ε = 10–4 using constant and variable deviating argument δi

Table 4 The solution errors of Example 4 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.000 2.4425e–15 0.0000e+00 2.4425e–15 2.2204e–15 0.0000e+00 0.0000e+00
0.500 2.4425e–15 0.0000e+00 2.4425e–15 2.2204e–15 0.0000e+00 0.0000e+00
0.700 2.4425e–15 0.0000e+00 2.4425e–15 2.2204e–15 0.0000e+00 0.0000e+00
0.900 2.4425e–15 0.0000e+00 2.4425e–15 2.2204e–15 0.0000e+00 0.0000e+00
0.950 2.4425e–15 0.0000e+00 2.4425e–15 1.5683e–09 0.0000e+00 0.0000e+00
0.980 1.7744e–09 2.0612e–09 1.7744e–09 3.0073e–04 0.0000e+00 0.0000e+00
0.990 2.8465e–05 4.5400e–05 2.8465e–05 1.7342e–02 0.0000e+00 0.0000e+00
0.991 7.2605e–05 1.2341e–04 7.2605e–05 2.6012e–02 0.0000e+00 0.0000e+00
0.992 1.8305e–04 3.3546e–04 1.8305e–04 3.9018e–02 0.0000e+00 0.0000e+00
0.993 4.5463e–04 9.1188e–04 4.5463e–04 5.8528e–02 0.0000e+00 0.0000e+00
0.994 1.1070e–03 2.4788e–03 1.1070e–03 8.7791e–02 0.0000e+00 0.0000e+00
0.995 2.6227e–03 6.7379e–03 2.6227e–03 1.3169e–01 0.0000e+00 0.0000e+00
0.996 5.9700e–03 1.8316e–02 5.9700e–03 1.9753e–01 0.0000e+00 0.0000e+00
0.997 1.2750e–02 4.9787e–02 1.2750e–02 2.9630e–01 9.3592e–14 9.3592e–14
0.998 2.4224e–02 1.3534e–01 2.4224e–02 4.4444e–01 2.0612e–09 2.0612e–09
0.999 3.4546e–02 3.6788e–01 3.4546e–02 6.6671e–01 4.5400e–05 4.5400e–05
1.000 0.0000e+00 0.0000e+00 0.0000e+00 2.2204e–16 0.0000e+00 0.0000e+00

with BCs y(0) = 1 + exp(–(1 + ε)/ε) and y(1) = 1 + exp(–1). The exact solution is given by

y(x) = e(1+ε)(x–1)/ε + e–x .

The solution errors are presented in Table 5 for ε = 10–3 and ε = 10–4 at h = 0.001. The
numerical solution using the ASCD and CFD schemes for ε = 10–4 are given in Fig. 7.

The results in Tables 4 and 5 and Figs. 6 and 7 show that the ASCD can handle problems
with right-end boundary layers effectively as well as those with left-end boundary layers.

4.3 Internal or twin-boundary-layer problems

Example 6 Consider the following SPBVP [49] given by

εy (x) + xy (x) = 0, x ∈ [–1, 1], (27)

El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 11 of 17

Figure 6 The solution of Example 4 for ε = 10–4 using the ASCD and CFD schemes

Table 5 Solution errors of Example 5 for ε = 10–3 and ε = 10–4 at h = 0.001

xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.000 4.2411e–14 0.0000e+00 4.2411e–14 1.8874e–14 0.0000e+00 0.0000e+00
0.500 5.0469e–08 1.5141e–04 5.0469e–08 5.0537e–08 1.2125e–04 1.2125e–04
0.700 5.7848e–08 1.7354e–04 5.7848e–08 5.7926e–08 1.3898e–04 1.3898e–04
0.900 6.0894e–08 1.8267e–04 6.0894e–08 6.0976e–08 1.4631e–04 1.4631e–04
0.950 6.1142e–08 1.8341e–04 6.1142e–08 6.2717e–08 1.4690e–04 1.4690e–04
0.980 5.9470e–08 1.8361e–04 5.9470e–08 2.9484e–04 1.4706e–04 1.4706e–04
0.990 2.8120e–05 2.2859e–04 2.8120e–05 1.7169e–02 1.4709e–04 1.4709e–04
0.991 7.1893e–05 3.0595e–04 7.1893e–05 2.5779e–02 1.4709e–04 1.4709e–04
0.992 1.8153e–04 5.1643e–04 1.8153e–04 3.8708e–02 1.4709e–04 1.4709e–04
0.993 4.5140e–04 1.0892e–03 4.5140e–04 5.8119e–02 1.4709e–04 1.4709e–04
0.994 1.1003e–03 2.6476e–03 1.1003e–03 8.7266e–02 1.4709e–04 1.4709e–04
0.995 2.6096e–03 6.8880e–03 2.6096e–03 1.3103e–01 1.4709e–04 1.4709e–04
0.996 5.9461e–03 1.8426e–02 5.9461e–03 1.9674e–01 1.4709e–04 1.4709e–04
0.997 1.2712e–02 4.9822e–02 1.2712e–02 2.9541e–01 1.4709e–04 1.4709e–04
0.998 2.4176e–02 1.3525e–01 2.4176e–02 4.4356e–01 1.4709e–04 1.4709e–04
0.999 3.4512e–02 3.6770e–01 3.4512e–02 6.6605e–01 1.0174e–04 1.0174e–04
1.000 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00

with BCs y(–1) = 0 and y(1) = 2. The exact solution is given by

√
erf (x/ 2x)
y(x) = 1 + √ .
erf (1/ 2x)

The solution errors are listed in Table 6 for ε = 10–3 and ε = 10–4 at h = 0.001. The numer-
ical solutions using the ASCD and CFD schemes for ε = 10–4 are presented in Fig. 8.

Example 7 Consider the following SPBVP with twin boundary layer [40, 42]

εy (x) – 2(2x – 1)y (x) – 4y(x) = 0, x ∈ [0, 1], (28)

with BCs y(0) = 1 and y(1) = 1. The exact solution is given by

y(x) = e2x(x–1)/ε .
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 12 of 17

Figure 7 The solution of Example 5 for ε = 10–4 using the ASCD and CFD schemes

Table 6 The solution errors of Example 6 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
–1.00 2.2826e–12 0.0000e+00 2.2826e–12 6.1286e–13 0.0000e+00 1.1351e–13
–0.50 2.2826e–12 0.0000e+00 2.2826e–12 6.1286e–13 0.0000e+00 1.1351e–13
–0.10 5.3097e–06 1.5654e–03 5.3097e–06 6.1286e–13 3.0826e–71 1.1351e–13
–0.08 1.9540e–05 1.1412e–02 1.9540e–05 6.1363e–13 1.2212e–15 1.1274e–13
–0.06 4.4839e–05 5.7780e–02 4.4839e–05 5.5249e–10 1.9732e–09 5.5177e–10
–0.04 6.2362e–05 2.0590e–01 6.2362e–05 4.5925e–06 6.3342e–05 4.5925e–06
–0.02 4.6484e–05 5.2709e–01 4.6484e–05 4.0526e–04 4.5500e–02 4.0526e–04
–0.01 2.5503e–05 7.5183e–01 2.5503e–05 6.0561e–04 3.1731e–01 6.0561e–04
0.00 1.1580e–12 0.0000e+00 1.1580e–12 3.1630e–13 0.0000e+00 4.9516e–14
0.01 2.5503e–05 7.5183e–01 2.5503e–05 6.0561e–04 3.1731e–01 6.0561e–04
0.02 4.6484e–05 5.2709e–01 4.6484e–05 4.0526e–04 4.5500e–02 4.0526e–04
0.04 6.2362e–05 2.0590e–01 6.2362e–05 4.5925e–06 6.3342e–05 4.5925e–06
0.06 4.4839e–05 5.7780e–02 4.4839e–05 5.5189e–10 1.9732e–09 5.5188e–10
0.08 1.9540e–05 1.1412e–02 1.9540e–05 4.8850e–15 1.3323e–15 1.1102e–15
0.10 5.3097e–06 1.5654e–03 5.3097e–06 3.5527e–15 0.0000e+00 2.2204e–16
0.50 0.0000e+00 0.0000e+00 0.0000e+00 8.8818e–16 0.0000e+00 0.0000e+00
1.00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00

The solution errors are listed in Table 7 for ε = 10–3 and ε = 10–4 at h = 0.001. The
numerical solutions using the ASCD and CFD methods for ε = 10–4 are presented in Fig. 9.

The results in Table 6 and Fig. 8 show that the ASCD and CFD lead to stable and accurate
results for the internal layer problem (26). The reason for absenting the stability restriction
of CFD in solving internal layer problems is that the internal layer occurs at a turning point
x = xin at which p(xin ) = 0. The results in Table 7 show that the solution obtained for the
twin boundary layer problem (28) at ε = 10–4 using the ASCD is more accurate than those
of the CFD and thus is due to the absolute stability property of the ASCD, while for the
CFD the stability restriction results in nonphysical oscillations in the numerical solution
as shown in Fig. 9.
Figure 10 depicts the solution error of Example 7 for ε = 10–4 using different values of
the deviating argument δi = k/ω|pi |, k = 4, 3, 2, 1. The results in Fig. 10 show that a more ac-
curate solution is obtained at δi = 4/ω|pi |. Figure 11 illustrates the solution error of Exam-
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 13 of 17

Figure 8 The solution of Example 6 for ε = 10–4 using the ASCD and CFD schemes

Table 7 The solution errors of Example 7 for ε = 10–3 and ε = 10–4 at h = 0.001
xi ε = 10–3 ε = 10–4
CFD ASCD CFD ASCD
4 4
δi = ω|pi | δi = δop δi = ω|pi | δi = δop
0.00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00 0.0000e+00
0.01 2.5175e–09 2.5175e–09 2.5175e–09 1.2904e–01 1.0226e–86 1.0226e–86
0.02 9.4549e–18 9.4549e–18 9.4549e–18 1.5959e–02 5.7090e–171 5.709e–171
0.04 4.4278e–34 4.4278e–34 4.4278e–34 2.1328e–04 0.0000e+00 0.0000e+00
0.06 4.4806e–50 1.0270e–49 4.4806e–50 2.3489e–06 0.0000e+00 0.0000e+00
0.08 6.0646e–50 1.1799e–64 6.0646e–50 2.0935e–08 0.0000e+00 0.0000e+00
0.10 6.3669e–50 6.7142e–79 6.3669e–50 1.4785e–10 0.0000e+00 0.0000e+00
0.20 8.4790e–50 1.0611e–139 8.4790e–50 2.8784e–23 0.0000e+00 0.0000e+00
0.30 1.2675e–49 3.9475e–183 1.2675e–49 2.5563e–41 0.0000e+00 0.0000e+00
0.40 2.4911e–49 3.4566e–209 2.4911e–49 2.8289e–73 0.0000e+00 0.0000e+00
0.50 2.0204e–48 7.1246e–218 2.0204e–48 8.4555e–146 0.0000e+00 0.0000e+00
0.60 2.1996e–39 3.4566e–209 2.1996e–39 1.3376e–70 0.0000e+00 0.0000e+00
0.70 4.9323e–21 3.9475e–183 4.9323e–21 2.5837e–41 0.0000e+00 0.0000e+00
0.80 3.2994e–21 1.0611e–139 3.2994e–21 2.9094e–23 0.0000e+00 0.0000e+00
0.90 2.4776e–21 6.7142e–79 2.4776e–21 1.4785e–10 0.0000e+00 0.0000e+00
0.92 2.3599e–21 1.1799e–64 2.3599e–21 2.0935e–08 0.0000e+00 0.0000e+00
0.94 2.2529e–21 1.0270e–49 2.2529e–21 2.3489e–06 0.0000e+00 0.0000e+00
0.96 2.1552e–21 4.4278e–34 2.1552e–21 2.1328e–04 0.0000e+00 0.0000e+00
0.98 9.4528e–18 9.4549e–18 9.4528e–18 1.5959e–02 5.7090e–171 5.709e–171
0.99 2.5175e–09 2.5175e–09 2.5175e–09 1.2904e–01 1.0226e–86 1.0226e–86
1.00 0.0000e+00 0.0000e+00 0.0000e+00 1.1102e–16 0.0000e+00 0.0000e+00

ple 7 for ε = 10–4 when adopting constant and variable deviating arguments δi = 4/ω|pi |∞ ,
δi = 4/ω|pi |, respectively. Figure 11 reveals that a more accurate solution is obtained using
a variable deviating argument.

5 Conclusions
This paper presented an absolutely stable noniterative central difference scheme for solv-
ing a general class of SPPs having left, right, internal, or twin boundary layers. The original
two-point second-order SPPs is approximated by a first-order delay differential equation
with a variable deviating argument. This delay differential equation was transformed into
a three-term difference equation that was solved using the Thomas algorithm. The stabil-
ity analysis of the method is discussed showing that the method is absolutely stable under
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 14 of 17

Figure 9 The solution of Example 7 for ε = 10–4 using the ASCD and CFD schemes

Figure 10 The solution errors of Example 7 at ε = 10–4 and different values of δi

Figure 11 Solution errors of Example 7 at ε = 10–4 using constant and variable deviating argument δi

a certain condition on the deviating argument without stability restriction on the step-
size. An optimal estimation for the deviating argument is obtained taking advantage of
the second-order accuracy of the central finite difference method in addition to the abso-
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 15 of 17

lute stability property. Several examples having left, right, interior, or twin boundary layers
were considered to illustrate the method. To analyze the effect of the deviating argument
on the solution accuracy, the maximum absolute error is reported in several tables and
figures for constant and variable deviating argument values. The results confirm that the
variable deviating argument results in more accurate results than the constant one. More-
over, the results reveal that the deviating argument can stabilize the unstable discretized
differential equation and that the present method is effective in solving the considered
class of SPPs.

Appendix A
Clear all, format long
Epsn=.001, a=0; b=1; w=1/Epsn;
Alpha=0 %1+exp((-1-Epsn)/Epsn)
Beta=1 %1+1/exp(1)
h0=.01
Deltamax=2*h0
JJ=3.999
Ind=0; x=a; xx=0; hh=(a:h0:b)

While xx<b-h0
Ind=Ind+1; x=hh (Ind+1);
% f=0; p=-1; q=1;
% f=-1-2*x; p=-1; q=0;
% f=0; p=-(1-x/2); q=1/2;
% f=-2/(x+1)*(log(2/(x+1))-1); p=-2; q=-2/(1+x);
% f=-x-2.9995; p=f; q=0
% f=0; p=-1; q=0
% f=0; p=1; q=0
% f=0; p=1; q=1+epsp;
% f=0; p=x; q=1;
% f=0; p=-x; q=0
% p=0; q=1; f= -cos (pi*x)^2-2*epsp^2*pi^2*cos(2*pi*x)

Delta=min(JJ/w/abs(p),Deltamax)
If p==0; Delta=min(JJ/w/abs(p+1e-6),Deltamax);end;

ALPHANEG(Ind)=(2/h0+Delta*w*p/2/h0);
ALPHAPOS(Ind)=(2/h0-Delta*w*p/2/h0);
ALPHAI(Ind)= -4/h0-Delta*w*q; BETA=-w*Delta*f;

if (ALPHAI)>0 |(ALPHAPOS)<0|(ALPHANEG)<0|
abs(ALPHAI)+.1<abs(ALPHAPOS)+abs(ALPHANEG),
error (’not dominant’); end
end

% Thomas algorithm in Matlab environment

subc = [ ALPHANEG’;0;0];
supc = [0;0; ALPHAPOS’];
diagc= [1; ALPHAI’;1];
Bvector = [alpha; BETA’;beta];
AAAA = spdiags ([subc diagc supc],-1:1,
length (Bvector), length (Bvector));
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 16 of 17

YSOL=AAAA\Bvector;.

Acknowledgements
The authors would like to thank the referees for their valuable comments and suggestions which helped to improve the
manuscript.

Funding
Not applicable.

Availability of data and materials

Not applicable.

Competing interests
The authors have no competing interests regarding the publication of this paper.

Authors’ contributions
All authors have equal contributions in this paper. All authors read and approved the final manuscript.

Author details
1
Department of Mathematics, College of Science and Humanities in Al-Kharj, Prince Sattam bin Abdulaziz University, P.O.
Box 83, Al-Kharj, 11942, Saudi Arabia. 2 Department of Basic Engineering Science, Faculty of Engineering, Menoufia
University, Shebin El-Kom, 32511, Egypt. 3 Department of Mathematics, Faculty of Science, University of Tabuk, P.O. Box
741, Tabuk 71491, Saudi Arabia. 4 Department of Mathematics, Cankaya University, Ankara 06530, Turkey. 5 Institute of
Space Sciences, P.O. BOX MG-23, Magurele-Bucharest RO-077125, Romania. 6 Department of Medical Research, China
Medical University Hospital, China Medical University, Taichung, Taiwan. 7 Institute of Engineering, Polytechnic of Porto,
Rua Dr. António Bernardino de Almeida, 431, 4249-015 Porto, Portugal. 8 Department of Mathematics and Statistics,
Faculty of Science, Taif University, PO Box 888, Taif, 21974, Saudi Arabia. 9 Department of Physics and Engineering
Mathematics, Faculty of Electronic Engineering, Menoufia University, Menouf 32952, Egypt.

Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Received: 15 April 2020 Accepted: 27 July 2020

References
1. Johnson, S.: Singular Perturbation Theory, Mathematical and Analytical Techniques with Applications to Engineering.
Springer, Berlin (2005)
2. Bender, C.M., Orszag, S.A.: Advanced Mathematical Methods for Scientists and Engineers. McGraw-Hill, New York
(1978)
3. Kevorkian, J., Cole, J.D.: Perturbation Methods in Applied Mathematics. Springer, New York (1981)
4. O’Malley, R.E.: Singular Perturbation Methods for Ordinary Differential Equations, vol. 89. Springer, New York (1991)
5. Roos, H.-G., Stynes, M., Tobiska, L.: Robust Numerical Methods for Singularly Perturbed Differential Equations. Springer
Ser. Comput. Math. Springer, Berlin (2008)
6. El’sgol’ts, L.E., Norkin, S.B.: Introduction to the Theory and Application of Differential Equations with Deviating
Arguments. Academic Press, New York (1973)
7. El-Zahar, E.R.: Applications of adaptive multi-step differential transform method to singular perturbation problems
arising in science and engineering. Appl. Math. Inf. Sci. 9(1), 223–232 (2015)
8. Chamkha, A.J., Rashad, A.M., EL-Zahar, E.R., EL-Mky, H.A.: Analytical and numerical investigation of Fe3 O4 –water
nanofluid flow over a moveable plane in a parallel stream with high suction. Energies 12(1), 1–18 (2019)
9. El-Zahar, E.R., Rashad, A.M., Seddek, L.F.: The impact of sinusoidal surface temperature on the natural convective flow
of a ferrofluid along a vertical plate. Mathematics 7(11), 1014 (2019). https://doi.org/10.3390/math7111014
10. Franz, S., Roos, H.G.: The capriciousness of numerical methods for singular perturbations. SIAM Rev. 53(1), 157–173
(2011)
11. Il’in, A.M.: Differencing scheme for a differential equation with a small parameter affecting the highest derivative.
Math. Notes Acad. Sci. USSR 6(2), 596–602 (1969)
12. Vulanović, R.: A uniform numerical method for quasilinear singular perturbation problems without turning points.
Computing 41(1–2), 97–106 (1989)
13. Kadalbajoo, M.K., Reddy, Y.N.: Numerical solution of singular perturbation problems via deviating arguments. Appl.
Math. Comput. 21(3), 221–232 (1987)
14. Reddy, Y.N., Reddy, K.A.: Numerical integration method for general singularly perturbed two-point boundary value
problems. Appl. Math. Comput. 133, 351–373 (2002)
15. El-Zahar, E.R.: Approximate analytical solution of singularly perturbed boundary value problems in MAPLE. AIMS
Math. 5(3), 2272–2284 (2020). https://doi.org/10.3934/math.2020150
16. Sangeetha, G., Thirupathi, P., Phaneendra, K.: Non-standard fitted operator scheme for singularly perturbed boundary
value problem. J. Math. Comput. Sci. 10(4), 793–804 (2020)
17. Melesse, W.G., Tiruneh, A.A., Derese, G.A.: Uniform hybrid difference scheme for singularly perturbed
differential-difference turning point problems exhibiting boundary layers. Abstr. Appl. Anal. 2020, Article ID 7045756
(2020)
18. Kaur, D., Kumar, V.: Numerical solution with special layer adapted meshes for singularly perturbed boundary value
problems. In: Applied Mathematical Analysis: Theory, Methods, and Applications, vol. 383. Springer, Cham (2020)
El-Zahar et al. Advances in Difference Equations (2020) 2020:411 Page 17 of 17

19. El-Zahar, E.R., Machado, J.T., Ebaid, A.: A new generalized Taylor-like explicit method for stiff ordinary differential
equations. Mathematics 7(12), 1154 (2019). https://doi.org/10.3390/math7121154
20. Rakmaiah, S., Phaneendra, K.: Numerical solution of singularly perturbed boundary value problems with twin
boundary layers using exponential fitted scheme. Commun. Math. Appl. 10(4), 797–807 (2019)
21. Phaneendra, K., Lalu, M.: Gaussian quadrature for two-point singularly perturbed boundary value problems with
exponential fitting. Commun. Math. Appl. 10(3), 447–467 (2019)
22. Dubey, R.K., Gupta, V.: A mesh refinement algorithm for singularly perturbed boundary and interior layer problems.
Int. J. Comput. Methods 1950024 (2019). https://doi.org/10.1142/S0219876219500245
23. Geng, F., Tang, Z., Zhou, Y.: Reproducing kernel method for singularly perturbed one-dimensional initial-boundary
value problems with exponential initial layers. Qual. Theory Dyn. Syst. 17(1), 177–187 (2018)
24. Wu, L., Ni, M., Lu, H.: Internal layer solution of singularly perturbed optimal control problem with integral boundary
condition. Qual. Theory Dyn. Syst. 17(1), 49–66 (2018)
25. Mukherjee, K.: Parameter-uniform improved hybrid numerical scheme for singularly perturbed problems with
interior layers. Math. Model. Anal. 23(2), 167–189 (2018)
26. Geng, F.Z., Qian, S.P.: A new numerical method for singularly perturbed turning point problems with two boundary
layers based on reproducing kernel method. Calcolo 54(2), 515–526 (2017)
27. Sharp, N., Trummer, M.: A spectral collocation method for systems of singularly perturbed boundary value problems.
Proc. Comput. Sci. 108, 725–734 (2017)
28. Cen, Z., Le, A., Xu, A.: A high-order finite difference scheme for a singularly perturbed reaction-diffusion problem with
an interior layer. Adv. Differ. Equ. 2017(1), 202 (2017)
29. Geetha, N., Tamilselvan, A., Subburayan, V.: Parameter uniform numerical method for third order singularly perturbed
turning point problems exhibiting boundary layers. Int. J. Appl. Comput. Math. 2(3), 349–364 (2016)
30. Volkov, V., Lukyanenko, D., Nefedov, N.: Asymptotic-numerical method for the location and dynamics of internal
layers in singular perturbed parabolic problems. In: International Conference on Numerical Analysis and Its
Applications, pp. 721–729. Springer, Cham (2016)
31. El-Zahar, E.R.: Piecewise approximate analytical solutions of high order singular perturbation problems with a
discontinuous source term. Int. J. Differ. Equ. 2016, Article ID 1015634 (2016). https://doi.org/10.1155/2016/1015634
32. Atay, M.T., Cengizci, S., Eryilmaz, A.: SCEM approach for singularly perturbed linear turning mid-point problems with
an interior layer. New Trends Math. Sci. 4(1), 115–124 (2016)
33. Phaneendra, K., Rakmaiah, S., Reddy, M.C.K.: Computational method for singularly perturbed boundary value
problems with dual boundary layer. Proc. Eng. 127, 370–376 (2015)
34. Phaneendra, K., Rakmaiah, S., Reddy, M.C.K.: Numerical treatment of singular perturbation problems exhibiting dual
boundary layers. Ain Shams Eng. J. 6(3), 1121–1127 (2015)
35. O’Riordan, E., Quinn, J.: A linearised singularly perturbed convection-diffusion problem with an interior layer. Appl.
Numer. Math. 98, 1–17 (2015)
36. Geng, F.Z., Qian, S.P., Li, S.: A numerical method for singularly perturbed turning point problems with an interior layer.
J. Comput. Appl. Math. 255, 97–105 (2014)
37. Shen, J., Han, M.: Canard solution and its asymptotic approximation in a second-order nonlinear singularly perturbed
boundary value problem with a turning point. Commun. Nonlinear Sci. Numer. Simul. 19(8), 2632–2643 (2014)
38. El-Zahar, E.R., EL-Kabeir, S.M.M.: A new method for solving singularly perturbed boundary value problems. Appl.
Math. Inf. Sci. 7(3), 927–938 (2013)
39. El-Zahar, E.R.: Approximate analytical solutions of singularly perturbed fourth-order boundary value problems using
differential transform method. J. King Saud Univ., Sci. 25(3), 257–265 (2013)
40. Geng, F.Z., Qian, S.P.: Reproducing kernel method for singularly perturbed turning point problems having twin
boundary layers. Appl. Math. Lett. 26(10), 998–1004 (2013)
41. Liu, C.S.: The Lie-group shooting method for solving nonlinear singularly perturbed boundary value problems.
Commun. Nonlinear Sci. Numer. Simul. 17(4), 1506–1521 (2012)
42. Kadalbajoo, M.K., Arora, P., Gupta, V.: Collocation method using artificial viscosity for solving stiff singularly perturbed
turning point problem having twin boundary layers. Comput. Math. Appl. 61(6), 1595–1607 (2011)
43. Attili, B.S.: Numerical treatment of singularly perturbed two point boundary value problems exhibiting boundary
layers. Commun. Nonlinear Sci. Numer. Simul. 16(9), 3504–3511 (2011)
44. Wang, Y., Su, L., Cao, X., Li, X.: Using reproducing kernel for solving a class of singularly perturbed problems. Comput.
Math. Appl. 61(2), 421–430 (2011)
45. Kumar, M., Mishra, H.K., Singh, P.: A boundary value approach for a class of linear singularly perturbed boundary value
problems. Adv. Eng. Softw. 40(4), 298–304 (2009)
46. Kumar, M., Singh, P., Mishra, H.K.: An initial-value technique for singularly perturbed boundary value problems via
cubic spline. Int. J. Comput. Methods Eng. Sci. Mech. 8(6), 419–427 (2007)
47. Qiu, Y., Sloan, D.M., Tng, T.: Numerical solution of singularly perturbed two-point boundary value problem using
equidistribution: analysis of convergence. Appl. Math. Comput. 116, 121–143 (2000)
48. Greenspan, D., Cassulli, V., Horova, I.: Numerical analysis for applied mathematics, science and engineering. Acta Appl.
Math. 37(3), 313–314 (1994)
49. Lopez, L.: Stability of a three-point scheme for linear second-order singularly perturbed BVPs with turning points.
Appl. Math. Comput. 52(2–3), 279–300 (1992)
50. Burden, R.L., Faires, J.D.: Numerical Analysis. PWS. Kent Publishing Co., Boston (1989)