Applied Mathematics and Computation 349 (2019) 304–313

Contents lists available at ScienceDirect

Applied Mathematics and Computation

journal homepage: www.elsevier.com/locate/amc

Piecewise reproducing kernel method for linear impulsive

delay differential equations with piecewise constant
arguments
Xiuying Li a, Haixia Li b, Boying Wu c,∗
a
Department of Mathematics, Changshu Institute of Technology, Suzhou, Jiangsu 215500, China
b
Department of Mathematics, Shandong Water Polytechnic, Rizhao, Shandong 276826, China
c
Department of Mathematics, Harbin Institute of Technology, Harbin, Heilongjiang 150001, China

a r t i c l e i n f o a b s t r a c t

Keywords: In this paper, we introduce a piecewise reproducing kernel method for impulsive delay
Reproducing kernel method differential equations with piecewise constant arguments. The method is an improved re-
Impulsive delay differential equations
producing kernel method. Compared with the classical reproducing kernel method, the so-
Piecewise constant arguments
lutions obtained using the present method can give good approximations for a larger time
interval. Some numerical examples are used to show the effectiveness and simplicity of
the method.
© 2018 Elsevier Inc. All rights reserved.

1. Introduction

Consider the following impulsive differential equations with piecewise constant argument

u (t ) + p(t )u(t ) + q(t )u([t − 1] ) = f (t ), t ≥ 0, t = k, k = 1, 2, . . . ,
u(k ) = ru(k ), k = 1, 2, . . . , (1.1)
u(−1 ) = A1 , u(0 ) = A0 ,
where p(t), q(t) and f(t) are continuous functions, A0 and A1 are real constants, u(k ) = u(k+ ) − u(k− ) and [ · ] denotes the
greatest integer function. Furthermore, we assume that (1.1) has a unique solution.
Impulsive differential equations have been applied successfully in ecology, population dynamic, optimal control, etc.
Therefore, impulsive differential equations have attracted much attention. It is usually impossible to obtain the exact so-
lution of such equations. Hence, it is important to develop numerical methods for solving such problems. Liu and Zeng
[1–3] proposed the linear multistep methods for impulsive differential equations and impulsive delay differential equations.
Zhang and Liang [4] obtained global superconvergence and local superconvergence of the collocation solution for linear
impulsive differential equations. Zhang [5,6] discussed the stability of Runge–Kutta methods for linear impulsive delay dif-
ferential equations and linear impulsive delay differential equations with piecewise constant arguments.
Based on the reproducing kernel theory, a method called the reproducing kernel method (RKM) was developed by Geng
and Cui [7–9]. The method has been widely applied to many fields [10–34]. However, the direct application of the RKM
to the considered impulsive delay differential equations with piecewise constant arguments can not produce accurate nu-
merical solutions. Recently, based on the proposed RKM, Geng and Qian [13,14] present the piecewise reproducing kernel

∗
Corresponding author.
E-mail addresses: [email protected] (X. Li), [email protected] (B. Wu).

https://doi.org/10.1016/j.amc.2018.12.054
0 096-30 03/© 2018 Elsevier Inc. All rights reserved.
 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313 305

method (PRKM) for singularly perturbed problems. Following the idea in [13,14], in this paper, we will present the PRKM
for impulsive delay differential equations with piecewise constant arguments.
The rest of the paper is organized as follows. In the next Section, some basic definitions and properties of the repro-
ducing kernel theory is introduced. RKM for linear differential equations is introduced in Section 3. PEKM for impulsive
differential equations with piecewise constant argument is presented in Section 4. Error analysis are introduced in Section
5. The numerical examples are provided in Section 6. Section 7 ends this paper with a brief conclusion.

2. Preliminaries

In this section, we provide the definition of solution to impulsive differential equations with piecewise constant argu-
ment, and some basic definitions and properties of the reproducing kernel Hilbert space which are useful in the following
discussion.

Definition 2.1. [35] A function u : R+ ∪ {−1} → R is said to be a solution of (1.1) if it satisfies the following conditions:
(a) u(t) is continuous for t ∈ [0, ∞) with the possible exception of the points [t] ∈ [0, ∞),
(b) u(t) is right continuous and has left-hand limit at the points [t] ∈ [0, ∞),
(c) u(t) is differentiable and satisfies u (t ) + p(t )u(t ) + q(t )u([t − 1] ) = f (t ) for any t ∈ R+ , with the possible exception
of the points [t] ∈ [0, ∞) where one-sided derivatives exist,
(d) u(k) satisfies u(k ) = ru(k ) for k ∈ N,
(e) u(t) satisfies u(−1 ) = A1 , u(0 ) = A0 .

Theorem 2.1. [35] If p(t), q(t) are real constants and p = 0, f = 0, then the unique solution of (1.1) is formulated on [0, ∞) by

u(t ) = m0 ({t } )u(n ) + m1 ({t } )u(n − 1 ), t ∈ [n, n + 1 ),
m0 ( 1 )u ( n ) + m1 ( 1 )u ( n − 1 ) (2.1)
u (n + 1 ) = ,
1−r
where {t} is fractional part of t and
p −pt
m0 (t ) = e−pt , m1 (t ) = ( e − 1 ).
q
Definition 2.2. Let  be a nonempty abstract set. A function K :  ×  → R is a reproducing kernel of the Hilbert space H
if and only if

a )∀ t ∈ , K (·, t ) ∈ H
b)∀ t ∈ , ∀ ϕ ∈ H, (ϕ (· ), K (·, t )) = ϕ (t ).

The last condition is called “the reproducing property”: the value of the function ϕ at the point t is reproduced by the
inner product of ϕ ( · ) with K( · , t).
A Hilbert space which possesses a reproducing kernel is called a reproducing kernel Hilbert space (RKHS).

Definition 2.3. Let X be a nonempty set. A symmetric function K : X × X → R is called a positive definite kernel on X if

n
ci c j K ( xi , x j ) ≥ 0
i, j=1

holds for any n ∈ N, x1 , x2 , . . . , xn ∈ X, c1 , c2 , . . . , cn ∈ R.

Theorem 2.2. [36] All reproducing kernels of Hilbert spaces are positive definite, and every positive definite kernel defines a
unique reproducing kernel Hilbert space, of which it is the unique reproducing kernel.

This meas that there is one-to-one relation between RKHS and positive definite kernel.

Definition 2.4. W 3 [a, b] = {u(x ) | u (x ) is absolutely continuous, u(3) (x) ∈ L2 [a, b], u(a ) = 0}. The inner product and norm in
this space are defined as
 b
(u(y ), v(y ))3 = u(a )v(a ) + u (a )v (a ) + +u (a )v (a ) + u(3) v(3) dy (2.2)
a

and

 u 3 = (u, u )3 .
Theorem 2.3. Space W3 [a, b] is a reproducing kernel space and its reproducing kernel can be represented as

K1 (x, y ), y ≤ x,
K (x, y ) = (2.3)
K1 (y, x ), y > x,
306 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313

where K1 (x, y ) = − 120

1
(a − y )(6a4 − 3a3 (5x + 3y + 10 ) + a2 (10x2 + 5x(5y + 12 ) + y(y + 30 )) − a(10x2 (2y + 3 ) + 5x(y +
12 )y − y + 120 ) + 10x y(y + 3 ) − 5x(y3 − 24 ) + y4 ).
3 2

Proof. From Geng and Cui [7–9], it is easy to prove W3 [a, b] is a RKHS. Following we give the method to obtain its repro-
ducing kernel. Integrations by parts for (2.2) gives
 
∂ 5 K (x, b) ∂ 5 K (x, a ) ∂ 4 K (x, b)
(u(y ), K (x, y ))3 = u(b) + u ( a ) K ( x, a ) + − u ( b )
∂y 5 ∂y 5 ∂ y4
   2 
∂ K (x, a ) ∂ 4 K (x, a )  (b) ∂ K (x, b) + u (a ) ∂ K (x, a ) − ∂ K (x, a )
3 3
+u ( a ) + + u
∂y ∂ y4 ∂ y3 ∂ y2 ∂ y3
 b
∂ 6 K (x, y )
− u (y ) dy. (2.4)
a ∂ y6
Form the fact that u(y), K(x, y) ∈ W3 [a, b], we have

u (a ) = 0

and

K (x, a ) = 0. (2.5)

If
∂ 5 K (x, b) ∂ 4 K (x, b) ∂ K (x, a ) ∂ 4 K (x, a )
= 0 , = 0 , + = 0, (2.6)
∂ y5 ∂ y4 ∂y ∂ y4

∂ 3 K (x, b) ∂ 2 K (x, a ) ∂ 3 K (x, a )

= 0, − = 0, (2.7)
∂y 3 ∂ y2 ∂ y3
then (2.4) becomes
 b
∂ 6 K (x, y )
(u(y ), K (x, y ))3 = − u (y ) dy.
a ∂ y6
For ∀x ∈ [a, b], if K(x, y) also satisfies

∂ 6 K (x, y )
= −δ (y − x ), (2.8)
∂ y6
then

(u(y ), K (x, y ))3 = u(x ).

Characteristic equation of (2.8) is given by

λ6 = 0,
and therefore its characteristic value is λ = 0 whose multiplicity is 6. So, let
⎧5
⎪
⎨ ci yi , y ≤ x,
K (x, y ) = i=0
(2.9)
⎪ 5
⎩ di yi , y > x.
i=0

In view of (2.8), one obtains

∂ i K (x, x + 0 ) ∂ i K (x, x − 0 )
= , i = 0, 1, 2, . . . , 4. (2.10)
∂ yi ∂ yi
Integrating (2.8) from x − ε to x + ε with respect to y and letting ε → 0, we have

∂ 5 K (x, x + 0 ) ∂ 5 K (x, x − 0 )
− = −1. (2.11)
∂ y5 ∂ y5
From (2.5)–(2.7), (2.10) and (2.11), it is easy to determine the unknown coefficients of (2.9). 
 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313 307

3. RKM for linear differential equations

Here introduce the RKM for linear differential operator equation


Lu(t ) = u (t ) + p(t )u(t ) = g(t ), t ∈ (a, b),
(3.1)
u ( a ) = 0.

Here we only consider homogeneous initial condition u(a ) = 0 since the inhomogeneous initial conditions u(a ) = α can
be reduced to u(a ) = 0 easily. Under the assumption that (3.1) has a unique solution, we shall give the representation of
solution of (3.1) in RKHS W3 [a, b] in which every function satisfies the homogenous initial condition of (3.1).
Put ψi (t ) = Ls K (t, s )|s=ti . The orthonormal system {ψ i (t )}∞
i=1
of W3 [a, b] can be derived from Gram-Schmidt orthogonal-
ization process of {ψi (t )}∞ i=1
,


i
ψ i (t ) = βik ψk (t ), (βii > 0, i = 1, 2, . . .). (3.2)
k=1

Theorem 3.1. For (3.1), if {ti }∞

i=1
is dense in [a, b], then {ψi (t )}∞
i=1
is the complete system of W3 [a, b].

Proof. For each fixed u(t) ∈ W3 [a, b], let (u(t ), ψi (t )) = 0, (i = 1, 2, . . . ), which means that,

(u(t ), Ls K (t, s )|ti ) = Ls (u(t ), K (t, s ))|s=ti (3.3)

= Ls u(s )|s=ti = (Lu )(ti ) = 0. (3.4)

Note that {ti }∞
i=1
is dense on [a, b], hence, (Lu )(t ) = 0. It follows that u ≡ 0 from the existence of L−1 . This ends the proof. 

Theorem 3.2. If {ti }∞

i=1
is dense in [a, b] and the solution of (3.1) is unique, then the solution of (3.1) is
∞

u(t ) = A j ψ j (t ), (3.5)
j=1

j
where A j = β jl g(tl ).
l=1

Proof. Applying Theorem 3.1, it is easy to see that {ψ i (t )}∞

i=1
is the complete orthonormal basis of W3 [a, b]. By using the
reproducing property of K(r, s), we have
∞

u(t ) = (u(t ), ψ i (t ))ψ i (t )
i=1
∞ 
 i
= βik (u(t ), Ls K (t, s )|s=tk )ψ i (t )
i=1 k=1
∞ 
 i
= βik Ls (u(t ), K (t, s ))|s=tk ψ i (t )
i=1 k=1
∞ 
 i
= βik Ls u(s )|s=tk ψ i (t )
i=1 k=1
∞ 
 i
= βik g(tk )ψ i (t ) (3.6)
i=1 k=1

and the proof of the theorem is complete. 

Now, the approximate solution u(x) can be obtained by taking finite terms in the series representation of u(x) and


N
uN (t ) = A j ψ j (t ). (3.7)
j=1
∞ i
Remark: Since W3 [a, b] is a Hilbert space, it is clear that i=1 ( k=1 βik g(tk ))2 < ∞. Therefore, the sequence uN is conver-
gent in the sense of norm  · 3 .

Lemma 3.1. If u(x) ∈ W3 [a, b], then there exists a constant c such that |u(t)| ≤ cu(t)3 , |u(k ) (t )| ≤ c  u(t ) 3 , k = 1, 2.
308 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313

Proof. Since

| u(t ) |=| (u(s ), K (t, s ))3 |≤ u(s ) 3  K (t, s ) 3 ,

there exists a constant c0 such that

|u(t )| ≤ c0  u 3 .
Note that
 
 ∂ i K (t, s ) 
(i ) 
| u (t ) | = (u(s ), )3 
∂ti
 
 ∂ i K (t, s ) 
≤ u4   
∂ t i 3
≤ ci  u 3 , ( i = 0, 1 , 2 ),
where ci are constants.
Putting c = max0≤i≤m−1 {ci } and the proof of the lemma is complete. 

From above lemma, by convergence of uN (t) in the sense of norm, it is easy to obtain the following theorem.
(k )
Theorem 3.3. The approximate solution uN (t) and its derivatives uN (t ), k = 1, 2 are all uniformly convergent.

4. PRKM for impulsive differential equations with piecewise constant argument

In this section, we illustrate the PRKM for impulsive differential equations with piecewise constant argument (1.1). We
solve (1.1) by using the RKM in a piecewise fashion.
Consider (1.1) on [0, T]. We first divide [0, T] into M equidistant subintervals [t j , t j+1 ], j = 0, 1, . . . , M − 1, with t0 = 0 and
tM = T . It is required that {k : k ∈ N, k ≤ T } ⊂ {ti }M
i=0
. Denote by h the length of the subinterval, i.e., h = M T
. Then we get the
3
approximate solutions on every interval [ti−1 , ti ] by using the RKM in the RKHS W [ti−1 , ti ].
On subinterval [t0 , t1 ], (1.1) is reduced to

u (t ) + p(t )u(t ) = f (t ) − q(t )u([t − 1] )  f1 (t ), t ∈ [t0 , t1 ]
(4.1)
u ( 0 ) = A0 ,

Solving (4.1) by the RKM and taking N equidistant nodes on interval [t0 , t1 ], we obtain the approximate solution u1, N (t) of
(1.1) on [t0 , t1 ).
On subinterval [t1 , t2 ], the approximate form of (1.1) is
⎧

⎨u (t ) + p(t )u(t ) = f (t ) − q(t )u([t − 1] )  f2 (t ), t ∈ [t1 , t2 ],
⎪
u1,N (t1 − ), t1 = 1,
(4.2)
⎪u(t1 ) = α1 = u1,N (t1 − )
⎩ , t1 = 1,
1−r

A−1 , t1 = 1,
where u([t − 1] ) =
A 0 , t1 = 1.
We can also obtain the approximate solution u2, N (t) of (4.2) on the second subinterval [t1 , t2 ) by using the RKM.
In the same manner, on subinterval [ti , ti+1 ](i ≥ 2 ), (1.1) is approximated by

u (t ) + p(t )u(t ) = f (t ) − q(t )u([t − 1] )  fi+1 (t ), t ∈ [ti , ti+1 ]
(4.3)
u(ti ) = αi ,

where α i and u([t − 1] ) can be approximated by the obtained approximate solutions uj, N (t), j ≤ i. It is easy to obtain the
approximate solution ui+1,N (t ) of (1.1) on the subinterval [ti , ti+1 ) by using the RKM.
After obtaining the approximate solutions of all subintervals [ti−1 , ti ), i = 1, 2, . . . , M, these solutions are combined to
obtain the approximate solution UM, N (t) of (1.1) on the entire interval [0, T].

5. Error analysis

From Li and Wu [16], we have the following two theorems.

Theorem 5.1. If y(x) is the solution of


Ly(t ) = g(t ), t ∈ [a, b]
y (a ) = α
 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313 309

and p(t) ≥ γ > 0, then

1
|y ( x )| ≤ max{|y(a )|, max |Ly|}.
γ t∈[a,b]

Theorem 5.2. The approximation u1, N (t) to the solution of (1.1) on [t0 , t1 ] obtained by using the RKM in space W3 [t0 , t1 ]
satisfies
2
 u(t ) − u1,N (t ) ∞ = max |u(t ) − u1,N (t )| ≤ c1 h ,
t∈[t0 ,t1 ]

h
where h = N, c1 is a positive constant.

Proof. The proof follows directly from Li and Wu [15] and Theorem 5.1. 

Following we consider the error between u2, N (t) and u(t) on subinterval [t1 , t2 ].
Let u(t ) = v(t ) + α1 . Inserting it into (4.2), one obtains

Lv(t ) = f2 (t ) − p(t )α1  F2 (t ), t ∈ [t1 , t2 ],
(5.1)
v(t1 ) = 0.
Denote by v2,N (t ) be the approximate solution of (5.1) obtained by using the RKM in space W3 [t1 , t2 ], and therefore u2,N (t ) =
v2,N (t ) + u1,N (t1 ). It is easy to see that u(t )|t∈[t1 ,t2 ] is the solution of the following equation

Ly(t ) = f2 (t ), t ∈ [t1 , t2 ],
(5.2)
y(t1 ) = u(t1 ).

Introducing a new function w(t ) = y(t ) − u(t1 ), we have


Lw(t ) = f2 (t ) − p(t )u(t1 )  G2 (t ), t ∈ [t1 , t2 ],
(5.3)
w(t1 ) = 0.

Denote by w2,N (t ) be the approximate solution of (5.3) obtained by using the RKM in RKHS W3 [t1 , t2 ].

Theorem 5.3. There exists a positive constant c such that

2
 u(t ) − u2,N (t ) ∞ = max |u(t ) − u2,N (t )| ≤ c h .
t∈[t1 ,t2 ]

Proof. From Li and Wu [16], there exists a positive constant c2 such that
2
 Lv2,N − F ∞ = max |Lv2,N (t ) − F2 (t )| ≤ c2 h2 . (5.4)
t∈[t1 ,t2 ]

By using Theorem 5.2, we have

2
 F2 − G2 ∞ = max |F2 (t ) − G2 (t )| = max |u(t1 ) − α1 | ≤ c3 h , (5.5)
t∈[t1 ,t2 ] t∈[t1 ,t2 ]

where c3 is a positive constant. The combination of (5.4) and (5.5) yields that

 L(v2,N − w ) ∞ =  Lv2,N − G2 ∞
=  Lv2,N − F2 + F2 − G2 ∞
≤  Lv2,N − F2 ∞ +  F2 − G2 ∞
2
≤ ( c2 + c3 ) h .
The direct application of Theorem 5.1 gives
1 2
 v2,N − w ∞ = max |v2,N − w| ≤ ( c2 + c3 ) h . (5.6)
t∈[t1 ,t2 ] γ
Note that

 u2,N (t ) − u(t ) ∞ = v2,N (t ) + α1 − w(t ) − u(t1 ) ∞ ≤ v2,N (t ) − w(t ) ∞ +  α1 − u(t1 ) ∞ .

From (5.6) and Theorem 5.2, it follows that
2
 u(t ) − u2,N (t ) ∞ = max |u(t ) − u2,N (t )| ≤ c h ,
t∈[t1 ,t2 ]

where c is a positive constant. 

310 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313

Table 1
The absolute errors between the numerical solutions and exact solution of (6.1) at t = 10.

N The explicit Euler [6] 2-Lobatto IIIA [6] 2-Lobatto IIIC [6] Present method
−5 −7 −7
20 2.61 × 10 2.46 × 10 5.09 × 10 4.01 × 10−7
40 1.39 × 10−5 6.14 × 10−8 1.25 × 10−7 5.14 × 10−8
80 7.14 × 10−6 1.53 × 10−8 3.10 × 10−8 6.52 × 10−9
160 3.62 × 10−6 9.60 × 10−10 1.92 × 10−9 5.70 × 10−10

1.0 1.0

0.5 0.5

2 4 6 8 2 4 6 8 10

0.5 0.5

1.0 1.0

1.5 1.5

Fig. 1. Exact solution and approximate solution obtained using N = 40 in Example 6.1 (the left: Exact solution; the right: approximate solution).

Theorem 5.4. The approximate solution UM, N (t) of (1.1) satisfies

2
 UM,N (t ) − u(t ) ∞ = max |u(t ) − UM,N (t )| ≤ d h ,
t∈I

where d is a positive constant.

Proof. Similar to the proof of Theorem 5, there exists a positive constant c4 such that
2
 u(t ) − ui,N (t ) ∞ = max |u(t ) − ui,N (t )| ≤ c4 h , (i = 3, 4, . . . , M ). (5.7)
t∈[ti−1 ,ti ]

Combining (5.7) and Theorems 5.2,5.3, we get

2
 UM,N (t ) − u(t ) ∞ = max |u(t ) − UM,N (t )| ≤ d h ,
t∈I

where d is a positive constant. 

6. Numerical experiments

Example 6.1. Consider the following impulsive delay differential equations with piecewise constant arguments [6]

u (t ) − u(t ) + u([t − 1] ) = 0, t ≥ 0, t = k, k = 1, 2, . . . ,
u(k ) = 9u(k ), k = 1, 2, . . . , u(−1 ) = 1, u(0 ) = 1,

Its exact solution can be obtained by Theorem 2.2. Using the present method, taking T = 10, M = 10, the numerical
results are compared with [6] in Table 1. The exact solution and approximate solution are shown in Fig. 1. The absolute
errors between the approximate solution and exact solution are shown in Fig. 2. Table 1 and Fig. 2 show that the present
method is more accurate than the method in [6].

Example 6.2. Consider the following impulsive delay differential equations with piecewise constant arguments [6]

u (t ) + 2u(t ) + u([t − 1] ) = 0, t ≥ 0, t = k, k = 1, 2, . . . ,
u(k ) = 15 u(k ), k = 1, 2, . . . , u(−1 ) = 1, u(0 ) = 1,

Its exact solution can be obtained by Theorem 2.2. Using the present method, taking T = 10, M = 10,the numerical results
using the present method are compared with [6] in Table 2. The exact solution and approximate solution are shown in Fig. 3.
The absolute errors between the approximate solution and exact solution are shown in Fig. 4. From Table 2 and Fig. 4, we
see that the present method can produce more accurate approximate solutions.
 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313 311

8
7. 10 6 5. 10

6
6. 10 8
4. 10
6
5. 10
8
6
3. 10
4. 10

6
3. 10 2. 10 8

6
2. 10
8
1. 10
6
1. 10

2 4 6 8 10 2 4 6 8 10

Fig. 2. Absolute errors of Example 6.1 (the left: N = 40; the right: N = 160).

Table 2
The absolute errors between the numerical solutions and exact solution of (6.2) at t = 10.

N The explicit Euler [6] 2-Lobatto IIIA [6] 2-Lobatto IIIC [6] Present method

20 1.65 × 10−2 1.34 × 10−4 2.49 × 10−4 1.83 × 10−5

40 4.10 × 10−3 3.36 × 10−5 6.47 × 10−5 2.15 × 10−6
80 2.00 × 10−3 8.39 × 10−6 1.65 × 10−5 2.61 × 10−7
160 1.00 × 10−3 2.10 × 10−6 4.16 × 10−6 3.19 × 10−8

1.0 1.0

0.5 0.5

2 4 6 8 10 2 4 6 8 10

0.5 0.5

Fig. 3. Exact solution and approximate solution obtained using N = 40 in Example 6.2 (the left: Exact solution; the right: approximate solution).

6
8. 10 7
1.2 10

7
6
1. 10
6. 10
8
8. 10

6
4. 10 6. 10 8

8
4. 10
6
2. 10
8
2. 10

2 4 6 8 10 2 4 6 8 10

Fig. 4. Absolute errors of Example 6.2 (the left: N = 40; the right: N = 160).

Example 6.3. Consider the following variable coefficient inhomogeneous impulsive delay differential equations with piece-
wise constant arguments

u (t ) + 2tu(t ) + u([t − 1] ) = sin t, t ≥ 0, t = k, k = 1, 2, . . . ,
u(k ) = 12 u(k ), k = 1, 2, . . . , u(−1 ) = 1, u(0 ) = 1,
312 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313

Table 3
The numerical results of Example 6.3.

t Approximate solution U10, 40 Approximate solution U10, 80 U10,40 − U10,80

0.99 0.120252 0.120250 2.04 × 10−6

2.97 0.019182 0.019184 1.93 × 10−6
4.95 –0.104604 –0.104605 7.69 × 10−7
6.93 0.054420 0.054420 1.49 × 10−7
7.92 0.065548 0.065548 4.79 × 10−8
9.90 –0.027435 –0.027435 2.67 × 10−8

1.0 1.0

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

2 4 6 8 10 2 4 6 8 10

0.2 0.2

Fig. 5. Approximate solutions obtained for N = 40 and N = 80 of Example 6.3 (the left: N = 40; the right: N = 80).

Its exact solution is not available. Using the present method, taking T = 10, M = 10, the numerical results are given in
Table 3. The approximate solutions obtained for N = 40 and N = 80 are shown in Fig. 5.

7. Conclusion

In this paper, a new effective method is proposed for solving impulsive delay differential equations with piecewise con-
stant arguments. The comparison of numerical results show that the present method can give accurate approximate solu-
tions.

Acknowledgments

The author would like to express thanks to the unknown referees for their careful reading and helpful comments. The
work was supported by the National Natural Science Foundation of China (Grant no.11801044, No.11326237, No.11271100).

References

[1] X. Liu, Y.M. Zeng, Linear multistep methods for impulsive delay differential equations, Appl. Math. Comput. 321 (2018) 555–563.
[2] X. Liu, G.L. Zhang, M.Z. Liu, Analytic and numerical exponential asymptotic stability of nonlinear impulsive differential equations, Appl. Numer. Math.
81 (2014) 40–49.
[3] X. Liu, M.H. Song, M.Z. Liu, Linear multistep methods for impulsive differential equations, Discret. Dyn. Nat. Soc. (2012) 652928, doi:10.1155/2012/
652928.
[4] Z.H. zhang, H. Liang, Collocation methods for impulsive differential equations, Appl. Math. Comput. 228 (2014) 336–348.
[5] G.L. Zhang, M.H. Song, M.Z. Liu, Exponential stability of the exact solutions and the numerical solutions for a class of linear impulsive delay differential
equations, J. Comput. Appl. Math. 285 (2015) 32–44.
[6] G.L. Zhang, Stability of Runge–Kutta methods for linear impulsive delay differential equations with piecewise constant arguments, J. Comput. Appl.
Math. 297 (2016) 41–50.
[7] F.Z. Geng, M.G. Cui, Solving a nonlinear system of second order boundary value problems, J. Math. Anal. Appl. 327 (2007) 1167–1181.
[8] M.G. Cui, F.Z. Geng, Solving singular two-point boundary value problem in reproducing kernel space, J. Comput. Appl. Math. 205 (2007) 6–15.
[9] M.G. Cui, Y.Z. Lin, Nonlinear numerical analysis in reproducing kernel space, Nova Science Pub Inc (2009).
[10] F.Z. Geng, S.P. Qian, An optimal reproducing kernel method for linear nonlocal boundary value problems, Appl. Math. Lett. 77 (2018) 49–56.
[11] F.Z. Geng, M.G. Cui, A novel method for solving a class of singularly perturbed boundary value problems based on reproducing kernel method, Appl.
Math. Comput. 218 (2011) 4211–4215.
[12] F.Z. Geng, S.P. Qian, S. Li, A numerical method for singularly perturbed turning point problems with an interior layer, J. Comput. Appl. Math. 255
(2014) 97–105.
[13] F.Z. Geng, S.P. Qian, Piecewise reproducing kernel method for singularly perturbed delay initial value problems, Appl. Math. Lett. 37 (2014) 67–71.
[14] F.Z. Geng, Z.Q. Tang, Piecewise shooting reproducing kernel method for linear singularly perturbed boundary value problems, Appl. Math. Lett. 62
(2016) 1–8.
[15] X.Y. Li, B.Y. Wu, Error estimation for the reproducing kernel method to solve linear boundary value problems, J. Comput. Appl. Math. 243 (2013) 10–15.
[16] X.Y. Li, B.Y. Wu, A continuous method for nonlocal functional differential equations with delayed or advanced arguments, J. Math. Anal. Appl. 409
(2014) 485–493.
[17] X.Y. Li, B.Y. Wu, A numerical method for solving distributed order diffusion equations, Appl. Math. Lett. 53 (2016) 92–99.
 X. Li, H. Li and B. Wu / Applied Mathematics and Computation 349 (2019) 304–313 313

[18] X.Y. Li, B.Y. Wu, A new reproducing kernel method for variable order fractional boundary value problems for functional differential equations, J.
Comput. Appl. Math. 311 (2017) 387–393.
[19] X.Y. Li, B.Y. Wu, A new reproducing kernel collocation method for nonlocal fractional boundary value problems with non-smooth solutions, Appl.
Math. Lett. 86 (2018) 194–199.
[20] W.Y. Wang, M. Yamamoto, B. Han, Numerical method in reproducing kernel space for an inverse source problem for the fractional diffusion equation,
Inverse Probl. 29 (2013) 1–15.
[21] W.Y. Wang, B. Han, M. Yamamoto, Inverse heat problem of determining time-dependent source parameter in reproducing kernel space, Nonlinear Anal.
Real World Appl. 14 (2013) 875–887.
[22] Y. Wang, L. Su, X. Cao, X. Li, Using reproducing kernel for solving a class of singularly perturbed problems, Comput. Math. Appl. 61 (2011) 421–430.
[23] W. Jiang, Y.Z. Lin, Representation of exact solution for the time-fractional telegraph equation in the reproducing kernel space, Commun. Nonlinear Sci.
Numer. Simul. 16 (2011) 3639–3645.
[24] M. Mohammadi, R. Mokhtari, Solving the generalized regularized long wave equation on the basis of a reproducing kernel space, J. Comput. Appl.
Math. 235 (2011) 4003–4014.
[25] O.A. Arqub, M. Al-Smadi, N. Shawagfeh, Solving fredholm integrocdifferentialequations using reproducing kernel hilbert space method, Appl. Math.
Comput. 219 (2013) 8938–8948.
[26] O.A. Arqub, Numerical solutions for the robin time-fractional partial differential equations of heat and fluid flows based on the reproducing kernel
algorithm, Int. J. Numer. Method Heat 28 (2018) 828–856.
[27] O.A. Arqub, Solutions of time-fractional tricomi and keldysh equations of dirichlet functions types in hilbert space, Numer. Methods Partial Differ. Equ.
34 (2018) 1759–1780.
[28] S. Abbasbandy, B. Azarnavid, M.S. Alhuthali, A shooting reproducing kernel hilbert space method for multiple solutions of nonlinear boundary value
problems, J. Comput. Appl. Math. 279 (2015) 293–305.
[29] R. Ketabchi, R. Mokhtari, E. Babolian, Some error estimates for solving volterra integral equations by using the reproducing kernel method, J. Comput.
Appl. Math. 273 (2015) 245–250.
[30] M. Ghasemi, M. Fardi, R.K. Ghaziani, Numerical solution of nonlinear delay differential equations of fractional order in reproducing kernel hilbert
space, Appl. Math. Comput. 268 (2015) 815–831.
[31] B. Azarnavid, K. Parand, An iterative reproducing kernel method in hilbert space for the multi-point boundary value problems, J. Comput. Appl. Math.
328 (2018) 151–163.
[32] W. Jiang, H. Li, A space-time spectral collocation method for the two-dimensional variable-order fractional percolation equations, Comput. Math, Appl.
75 (2018) 3508–3520.
[33] H. Li, W. Jiang, A space-time spectral collocation method for the two-dimensional nonlinear Riesz space fractional diffusion equations, Math. Method
Appl. Sci. (2018), doi:10.1002/mma.5124.
[34] Z. Chen, L.B. Wu, Y.Z. Lin, Exact solution of a class of fractional integro-differential equations with the weakly singular kernel based on a new fractional
reproducing kernel space, Math. Method Appl. Sci. 41 (2018) 3841–3855.
[35] F. Karakoc, H. Bereketoglu, G. Seyhan, Oscillatory and periodic solutions of impulsive differential equations with piecewise constant argument, Acta
Appl. Math. 110 (2010) 499–510.
[36] N. Aronszajn, Theory of reproducing kernel, Trans. A.M.S. 168 (1950) 1–50.