Abstract
In this paper, we not only study asymptotical stability of a class of linear impulsive neutral delay differential equations(INDDEs), but also study stability and asymptotical stability of nonlinear INDDEs. Asymptotical stability of zero solution of linear INDDEs is studied by the properties of simple autonomous linear neutral delay differential equations(NDDEs) without impulsive perturbations. Base on this idea, numerical methods of INDDEs are constructed. The constructed numerical methods preserve asymptotical stability of linear INDDEs if corresponding methods are A-stable. Moreover, some stability and asymptotical stability criteria are established for nonlinear INDDEs, respectively. The constructed numerical methods which can preserve stability and asymptotical stability of the exact solutions under these criteria are obtained. Some numerical examples are given to confirm the theoretical results.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
Impulsive differential equations arise widely in the study of medicine, biology, economics, engineering, and so forth. In recent years, INDDEs are attracting more and more attention. In papers Anguraj and Karthikeyan (2009) and Cuevasa et al. (2009), existence uniqueness and continuous dependence of INDDEs are investigated. In papers Li and Rogovchenko (2015, 2016) and Li and Deng (2017), oscillation of first-order, second-order, even-order of INDDEs are studied, respectively. In papers Li and Deng (2017) and Li and Rogovchenko (2020), stability of first-order, third-order INDDEs are studied, respectively. Mean square exponential input-to state stability of stochastic Markovian reaction diffusion systems with impulsive perturbations has also been studied in paper Xue et al. (2023).
On the other hand, stability and asymptotical stability of numerical methods for NDDEs without impulsive perturbations have been widely studied (see Liu 1999; Enright and Hayashi 1998; Engelborghs et al. 2001; Wang and Li 2008 etc). Wang and Li (2008) studied stability and asymptotic stability of \(\theta \)-methods for nonlinear NDDEs with constant delay and proportional delay. Enright and Hayashi (1998) estabilished sufficient conditions for order of convergence results about continuous Runge–Kutta methods for NDDEs with state dependent delays. But to the best of our knowledge, up to now, there are few articles referring to stability of numerical methods for INDDEs.
The aim of this paper is to provide asymptotical stability criteria for the exact solutions and the numerical solutions of INDDEs. Applying asymptotical stability of NDDEs without impulsive perturbations, asymptotical stability criteria are obtained for the exact solutions of linear INDDEs and nonlinear INDDEs, respectively. Numerical schemes for INDDEs are constructed based on the relationship between INDDEs and NDDEs. Moreover, we proved that some numerical methods can preserve asymptotical stability of linear and nonlinear INDDEs, respectively. The rest of this paper is organized as follows. In Sect. 2, asymptotical stability of zero solution of linear INDDEs is studied by the properties of simple linear NDDEs with constant coefficients. Base on this idea, numerical methods of INDDEs are constructed. The constructed numerical methods furnished A-stable Runge–Kutta methods can preserve asymptotical stability of linear INDDEs. In Sect. 3, some stability and asymptotical stability criteria are established for nonlinear INDDEs by the properties of nonlinear NDDEs. Under these stability and asymptotical stability criteria, the constructed numerical methods furnished by implicit Euler method or 2-stage Lobatto IIIC method are stability and asymptotical stability. In Sect. 4, we provide some numerical examples to confirm our theoretical results.
2 Linear INDDEs
In this section, we consider the following scalar linear INDDE:
where \(\tau \) is a positive constant, a, b, c and \(\lambda \) are complex constants, the initial function \(\phi (t)\) is continuous differentiable on \([-\tau ,0)\), \(x'(t)\) denotes the right-hand derivative of x(t) and \(\mathbb {N}=\{0,1,2,\cdots \}\).
In the following of this paper, always assume \(\lambda \ne 1\) and \(\lambda \ne 0\). When \(\lambda =1\), INDDE (2.1) is changed into NDDE without impulsive perturbations. When \(\lambda =0\), the solution x(t) of INDDE (2.1) satisfies \(x(k\tau )=0\) for all \(k\in \mathbb {N}\).
2.1 Relations between INDDEs and NDDEs
Theorem 2.1
Assume that x(t) is the solution of (2.1) and \(y(t)=\lambda ^{\{t/\tau \}}x(t)\), \(t\in [-\tau ,\infty )\), then y(t) is the solution of the following equation without impulsive perturbations
where \(\alpha =a+\frac{\ln \lambda }{\tau }\), \(\beta =b-\frac{c\ln \lambda }{\tau }\), \(\{\frac{t}{\tau }\}=\frac{t}{\tau }-\lfloor \frac{t}{\tau }\rfloor \), \(\lfloor \cdot \rfloor \) denotes the greatest integer function and
On the other hand, if y(t) is the solution of (2.2) and \(x(t)=\lambda ^{-\{t/\tau \}}y(t)\), \(t\in [-\tau ,\infty )\), then x(t) is the solution of (2.1).
Proof
We will prove that y(t) is continuous if x(t) is the solution of (2.1). Obviously,
is continuous on \([k\tau ,(k+1)\tau )\), \(k=-1,0, 1,2,\ldots \). Therefore,
Hence \(y(0)=y(0^-)\), which implies that y(t) is continuous at \(t=0\). It follows from
that y(t) is continuous at \(t=k\tau \), \(k=1,2,\ldots \). Consequently, y(t) is continuous on \([-\tau ,\infty )\).
Next, we will prove that y(t) is the solution of (2.2) if x(t) is the solution of (2.1). For \(t\in [k\tau ,(k+1)\tau )\), \(k=0,1,2,\ldots \), we can obtain that
and
which implies that
Finally, we will prove that x(t) is the solution of (2.1) if y(t) is the solution of (2.2). For \(t\in [k\tau ,(k+1)\tau )\), \(k=0,1,2,\cdots \), we can obtain that
and
which implies that
Obviously, it follows from
that \(x(k\tau )=\lambda x(k\tau ^-)\), \(k=0,1,2,\ldots \). Hence x(t) is the solution of (2.1). \(\square \)
2.2 Asmpotical stability of linear INDDEs
By (Bellen et al. 1988, Theorem 2.1) and Theorem 2.1 of present paper, we can obtain the following results.
Theorem 2.2
Assume that
Then the solution y(t) of NDDE (2.2) tends to zero as \(t\rightarrow \infty \). Hence if
then the solution x(t) of INDDE (2.1) tends to zero as \(t\rightarrow \infty \).
Maybe some readers are interested in linear INDDE as the following form.
where \(\mathbb {Z}^+=\{1,2,\ldots \}\). In fact, on interval \([-\tau ,\tau )\), the Eq. (2.4) is a delay differential equations without impulsive perturbations. If the solution x(t), \(t\in [0,\tau )\) is seen as initial function, the Eq. (2.4) can be seen the same as (2.1) for \(t\ge \tau \). So the asymptotically stable results of Theorem 2.2 can be extended as follows.
Corollary 2.3
If
then the solution x(t) of INDDE (2.4) tends to zero as \(t\rightarrow \infty \).
In fact, all the stable and asymptotically stable results of INDDE in this paper can be extended similarly, we do not introduce in detail for concise.
2.3 Stability analysis of numerical methods for linear INDDEs
Based on the relations between INDDE (2.1) and NDDE (2.2), the numerical method for linear INDDE (2.1) can be constructed as follows
where \(t_n=nh\), \(n\in \mathbb {N}\), \(h=\frac{\tau }{m}\) and m is a positive integer. Here, the vector \(w=[w_1,w_2,\ldots ,w_s]^T\) and the matrix \(A=[a_{ij}]_{i,j=1}^s\) define a Runge–Kutta method for ODEs. For \(\forall n \in \mathbb {N}\), \(y_n\) is an approximation to \(y(t_n)\) of (2.2), \(x_n\) is an approximation to \(x(t_n)\) of (2.1), \(y_n+h\sum _{j=1}^sa_{ij}z_{n+1}^{(j)}\) is an approximation to \(y(t_n+c_ih)\), \(y_{n-m}+h\sum _{j=1}^sb_{ij}z_{n-m+1}^{(j)}\) is an approximation to \(y(t_{n-m}+c_ih)\), \(\sum _{j=1}^sc_{ij}z_{n-m+1}^{(j)}\) is an approximation to \(y'(t_{n-m}+c_ih)\), where \(c_i=\sum _{j=1}^s a_{ij}\), \(i=1,2,\ldots ,s\).
Usually, \(b_{ij}=b_j(c_i)\) and \(c_{ij}=b_j'(c_i)\), where \(b_j(\theta )\), \(j=1,2,\ldots ,s\), are polynomials which define the natural continuous extension of the Runge–Kutta method, i.e. polynomials such that the approximate solution \(y_h\) defined on the whole interval of integration is given by
Put \(B=[b_{ij}]_{i,j=1}^s\) and \(C=[c_{ij}]_{i,j=1}^s\). Let \(\tilde{\alpha }=h\alpha \), \(\tilde{\beta }=h\beta \) and denote by \(\{y_n(m;\alpha ,\beta ,c)\}_{n=0}^{\infty }\) can be described by quadrupple \(\{w,A,B,C\}\). So the numerical method (2.5) for linear INDDE (2.1) can be rewritten as follows
where \(z_n\) stands for \([z_n^{(1)},z_n^{(2)},\ldots ,z_n^{(s)}]^T\) and \(e=[1,1,\ldots ,1]^T\).
A-stable Runge–Kutta methods for ODEs can be extended to asympotically stable numerical method for INDDEs. By (Bellen et al. 1988, Theorem 3.4), we can obtain the following theorem.
Theorem 2.4
Assume that the Runge–Kutta method \(\{w,A\}\) for ODE is A-stable. Then the corresponding method (2.6) for INDDE (2.1) with \(B=A\) and \(C=I\), which satisfies
Then the numerical solution \(x_n\) of (2.6) tends to zero as \(n\rightarrow \infty \).
3 Nonlinear INDDEs
In this section, we will consider the following nonlinear INDDE:
and the same equation with another initial function:
where \(\lambda \ne 1\), \(\lambda \ne 0\), \(\tau >0\), \(\phi \) and \(\tilde{\phi }\) are continuous functions on \([-\tau , 0)\), \(\lim \limits _{t\rightarrow 0^-}\phi (t)\) and \( \lim \limits _{t\rightarrow 0^-}\tilde{\phi }(t)\) exist, \(x'(t)\) denotes the right-hand derivative of x(t). Let \(\langle \cdot ,\cdot \rangle \) be a given inner product on \(\mathbb {C}^{d}\) and \(\Vert \cdot \Vert \) the corresponding norm. Assume that the function \(f: [0, \infty )\times \mathbb {C}^d\times \mathbb {C}^d\times \mathbb {C}^d\rightarrow \mathbb {C}^d\) is continuous in t and satisfies the following conditions: for arbitrary \(x, u, v, x_1, x_2, u_1, u_2, v_1, v_2 \in \mathbb {C}^d\) and \(\forall t \in [0,+\infty )\),
where R(t), \(\beta (t)\), \(\gamma (t)\) and \(\sigma (t)\) are continuous real functions and
3.1 Relations between INDDEs and NDDEs
Assume that the scalar function \(\alpha :[-\tau ,\infty )\rightarrow \mathbb {C}\) satisfies the following conditions:
-
(1)
for any \(t \in [0, \infty )\), \(\alpha (t)=\alpha (t-\tau )\);
-
(2)
\(\alpha (t)\) is infinite smooth on \([0, \tau )\);
-
(3)
\(\alpha (0)=1\) and \(\alpha (0^-)=\lambda \);
-
(4)
\(\inf _{t \in [0, \tau )} |\alpha (t)|\ge m>0\).
Theorem 3.1
Assume that x(t) is the solution of (2.1) and \(y(t)=\alpha (t)x(t),\;t\in [-\tau ,+\infty )\). Then y(t) is the solution of
where
and
On the other hand, assume that y(t) is the solution of (3.7) and \(x(t)= \frac{y(t)}{\alpha (t)},\;t\in [-\tau ,+\infty )\). Then x(t) is the solution of (2.1).
Proof
(i) Because \(\alpha (t)\) and x(t) are continuous on \([k\tau , (k+1)\tau )\), y(t) is continuous on \([k\tau , (k+1)\tau )\), where \(k=0,1,\ldots \). Obviously, we have
and
So \(y(k\tau )= y(k\tau ^+)=y(k\tau ^-),\;k=0,1,2,\ldots \). Hence y(t) is continuous on \([-\tau , \infty )\).
Obviously, for \(t\in [k\tau , (k+1)\tau ), k=0,1,\ldots \), we can obtain that
which implies
(ii) Assume that y(t) is the solution of (3.7). For \(t\in [k\tau , (k+1)\tau ), k=0,1,\ldots \),
Obviously,
and
So \(x(k\tau )=\lambda x(k\tau ^-), k=0,1,\cdots \). Obviously, we have \(x(t)=\frac{y(t)}{\alpha (t)}=\phi (t)\), \(t\in [-\tau , 0)\). Hence x(t) is the solution of (3.1). \(\square \)
3.2 Asymptotical stability of INDDEs
Theorem 3.2
Assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)-(3.6), and \(R(t)\le 0\) for \(\forall t\ge 0\),
then
where \(\omega =\frac{\sup _{t\in [-\tau , 0)}|\alpha (t)|}{\inf _{t\in [-\tau , 0)}|\alpha (t)|}\), \(\zeta =\sup \nolimits _{t\in [-\tau , 0)}|\frac{\alpha '(t)}{\alpha (t)}|\), \(\delta \phi (s)=\Vert \phi (s-\tau )-\tilde{\phi }(s-\tau )\Vert \), \(\delta \dot{\phi }(s)=\Vert \phi '(s-\tau )-\tilde{\phi }'(s-\tau )\Vert \) and
Moreover, assume that
and
then we have
Proof
We will apply inequalities (3.3)-(3.6) to prove that the function \(F: [0, \infty )\times \mathbb {C}^d\times \mathbb {C}^d\times \mathbb {C}^d\rightarrow \mathbb {C}^d\) satisfies the following conditions respectively: for arbitrary \(y, u, v, y_1, y_2, u_1, u_2, v_1, v_2 \in \mathbb {C}^d\) and \(\forall t \in [0,+\infty )\),
where \(V(t,y,u,v,w)=F(t,y,u,F(t-\tau ,u,v,w)).\)
First of all, we will prove inequality (3.13) as follows
which implies, if \(R(t)\le 0\) for \(\forall t\),
and if \(R(t)\le \hat{R}\) and \(\hat{R}>0\) for \(\forall t\),
Second, the inequality (3.14) can be proved as follows
Third, the inequality (3.15) can be proved as follows
Finally, the inequality (3.16) can be proved as follows
By the condition (3.8) and (Wang et al. 2009, Theorem 3.1), (Wang and Li 2004, Theorem 1), we can obtain that
which implies (3.9) holds, where
Moreover, by the condition (3.10)-(3.12) and (Wang et al. 2009, Theorem 3.2), (Wang and Li 2004, Theorem 4), we can obtain that
which implies
\(\square \)
Due to the difference in (3.17) and (3.18), similar to Theorem 3.2, by (Wang et al. 2009, Theorem 3.2) or (Wang and Li 2004, Theorem 4), we can obtain the following theorem.
Theorem 3.3
Assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)-(3.6), \(R(t)\le \hat{R}\), \(\hat{R}>0\), and for \(\forall t\ge 0\),
then
where
Moreover, assume that
then we have
3.3 Special cases
In this subsection, we only introduce three special cases and some sufficient conditions for stability and asymptotical stability of (3.1) and (3.2) are obtained from them. Many different functions can be chosen as \(\alpha (t)\) (see Zhang et al. 2015), which implies many special cases can be obtained.
Special case I. Obviously, \(\alpha _1(t)=\lambda ^{\{\frac{t}{\tau }\}}\), \(t\in [-\tau , \infty )\). By Theorem 3.2, when \(\alpha (t)=\alpha _1(t)=\lambda ^{\{\frac{t}{\tau }\}}\), we can obtain the following theorem.
Theorem 3.4
Assume that x(t) is the solution of (2.1) and \(y_1(t)=\lambda ^{\{\frac{t}{\tau }\}}x(t),\;t\in [-\tau ,+\infty )\). Then \(y_1(t)\) is the solution of
where
and
On the other hand, assume that y(t) is the solution of (3.23) and \(x(t)=\lambda ^{-\{\frac{t}{\tau }\}}y(t),\;t\in [-\tau ,+\infty )\). Then x(t) is the solution of (2.1).
Theorem 3.5
Assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)-(3.6), and \(R(t)\le 0\) for \(\forall t\ge 0\),
then
where \(\omega _1=\max \{|\lambda |,\frac{1}{|\lambda |}\}\), \(m_1=\min \{1,|\lambda |\}\), \(\zeta _1=|\frac{\ln \lambda }{\tau }|\) and
Moreover, assume that
and
then we have
Theorem 3.6
Assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le \hat{R}\), \(\hat{R}>0\), and for \(\forall t\ge 0\),
then
where
Moreover, assume that
and
then we have
Special Case II: When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), \(\alpha _2(t)=1+(\lambda -1)\{\frac{t}{\tau }\}\), \(t\in [-\tau , \infty )\). By Theorem 3.2, when \(\alpha (t)=\alpha _2(t)\), we can obtain the following theorem.
Theorem 3.7
Assume that x(t) is the solution of (2.1) and \(y_2(t)=\alpha _2(t)x(t),\;t\in [-\tau ,+\infty )\). Then \(y_2(t)\) is the solution of
where
and
On the other hand, assume that y(t) is the solution of (3.24) and \(x(t)= \frac{y_2(t)}{1+(\lambda -1)\{\frac{t}{\tau }\}},\;t\in [-\tau ,+\infty )\). Then x(t) is the solution of (2.1).
Theorem 3.8
When \(\lambda \in \mathbb {R}\), \(\lambda >0\) and \(\lambda \ne 1\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le 0\), for \(\forall t\ge 0\),
then
where \(\omega _2=\max \{\lambda ,\frac{1}{\lambda }\}\), \(m_2=\min \{1,\lambda \}\), \(\zeta _2=\max \{|\frac{\lambda -1}{\tau }|,|\frac{\lambda -1}{\lambda \tau }|\}\) and
Moreover, assume that
and
then we have
Theorem 3.9
When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le \hat{R}\), \(\hat{R}>0\), and for \(\forall t\ge 0\),
and
then
where
Moreover, assume that
then we have
Because \(\frac{\lambda -1}{\tau +(\lambda -1)\tau \{\frac{t}{\tau }\}}\le \frac{\lambda -1}{\tau }\) for all \(\lambda >0\), \(\forall t\in \mathbb {R}\), by Theorem 3.9, we can obtain the following corollary.
Corollary 3.10
When \(\lambda \in \mathbb {R}\), \(\lambda >0\) and \(\lambda \ne 1\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le 0\), for \(\forall t\ge 0\),
then
Moreover, assume that
and
then we have
Corollary 3.11
When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le \hat{R}\), \(\hat{R}>0\), and for \(\forall t\ge 0\),
and
then
Moreover, assume that
then we have
Special Case III: When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), \(\alpha (t)=\alpha _3(t)=-\{\frac{t}{\tau }\}^2+\lambda \{\frac{t}{\tau }\}+1\), \(t\in [-\tau , \infty )\). By Theorem 3.2, we can obtain the following theorem.
Theorem 3.12
Assume that x(t) is the solution of (2.1) and \(y_3(t)=(-\{\frac{t}{\tau }\}^2+\lambda \{\frac{t}{\tau }\}+1)x(t),\;t\in [-\tau ,+\infty )\). Then \(y_3(t)\) is the solution of
where
and
On the other hand, assume that \(y_3(t)\) is the solution of (3.25) and \(x(t)= \frac{y_3(t)}{-\{\frac{t}{\tau }\}^2+\lambda \{\frac{t}{\tau }\}+1},\;t\in [-\tau ,+\infty )\). Then x(t) is the solution of (2.1).
Theorem 3.13
When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le 0\), for \(\forall t\ge 0\),
and
then
where \(m_3=\min \{1,\lambda \}\), \(\zeta _3=\sup \nolimits _{t\in [-\tau , 0)}|\frac{\alpha _3'(t)}{\alpha _3(t)}|=\max \{|\frac{\lambda -2}{\lambda \tau }|,\frac{\lambda }{\tau }\}\),
and
Moreover, assume that
and
then we have
Because \(\frac{-2\{\frac{t}{\tau }\}+\lambda }{-\{\frac{t}{\tau }\}^2\tau +\lambda \tau \{\frac{t}{\tau }\}+\tau }\le \frac{\lambda }{\tau }\) for all \(\lambda >0\), \(\forall t\in \mathbb {R}\), by Theorem 3.13, we can obtain the following corollary.
Corollary 3.14
When \(\lambda \in \mathbb {R}\) and \(\lambda >0\), assume that IDDEs (3.1) and (3.2) fulfill the inequalities (3.3)–(3.6), \(R(t)\le 0\), for \(\forall t\ge 0\),
and
then
Moreover, assume that
and
then we have
3.4 Numerical methods for nonlinear INDDEs (3.1) and (3.2)
The numerical method for nonlinear INDDE (3.1) can be constructed as follows
and
where \(Y_{n-m+1}^i=\Phi (t_{n-m+1}^i)\) if \(t_{n-m+1}^i\le 0\), \(i=1,2,\cdots ,s\),
\(t_n=nh\), \(n\in \mathbb {N}\), \(h=\frac{\tau }{m}\) and m is a positive integer.
Similarly, the numerical method for (3.2) can be constructed as follows
and
where
Theorem 3.15
Under the condition of (3.8), the constructed numerical methods (3.26)–(3.27) and (3.28)–(3.29) furnished by backward Euler method with linear interpolation (2-stage Lobatto IIIC method with linear interpolation) are stable, in the following sense,
Moreover, under the condition of Theorem 3.2, the constructed numerical method (3.26)–(3.27) and (3.28)–(3.29) furnished by backward Euler method with linear interpolation (or 2-stage Lobatto IIIC method with linear interpolation) are also asymptotically stable, that is,
Proof
By (Wang et al. 2009, Theorem 4.8), RVCRK formulae (3.26) and (3.28) furnished by backward Euler method with linear interpolation (or 2-stage Lobatto IIIC method with linear interpolation) are RN-stable, that is
Moreover, (Wang et al. 2009, Theorem 4.9), under the condition of Theorem 3.2, the RVCRK formulae (3.26) and (3.28) furnished by backward Euler method with linear interpolation (or 2-stage Lobatto IIIC method with linear interpolation) are also asymptotically stable, that is,
Because the relation between the numerical solutions INDDE and NDDE without impulsive perturbations, that is (3.27) and (3.29), the theorem holds. \(\square \)
Similar to Theorem 3.15, we can obtain the following result.
Theorem 3.16
Under the condition of (3.19), the constructed numerical methods (3.26)–(3.27) and (3.28)–(3.29) furnished by backward Euler method with linear interpolation (or 2-stage Lobatto IIIC method with linear interpolation) are stable, in the following sense
Moreover, under the condition of Theorem 3.3, the constructed numerical method (3.26)–(3.27) and (3.28)–(3.29) furnished by backward Euler method with linear interpolation (or 2-stage Lobatto IIIC method with linear interpolation) are also asymptotically stable, that is
The following results are provided to analysis the difference between the linear equations and nonlinear equations.
Remark 3.17
-
1.
The meanings of asymptotic stability of the exact solutions of linear INDDE (2.1) and nonlinear INDDE (3.1) are different. The defintions of asymptotic stability of the exact solutions of linear INDDE (2.1) and nonlinear INDDE (3.1) are provided as follows.
-
(1)
The exact solution x(t) of linear INDDE (2.1) is said to be asymptotically stable if
$$\begin{aligned} \lim _{t\rightarrow \infty }x(t)=0. \end{aligned}$$ -
(2)
For \(\forall \epsilon >0\), if there exists a constant \(\delta >0\) such that \(\Vert \Phi -\tilde{\Phi }\Vert \le \delta \) and \(\Vert \Phi '-\tilde{\Phi }'\Vert \le \delta \) imply that
$$\begin{aligned} \Vert x(t)-\tilde{x}(t)\Vert <\epsilon , \end{aligned}$$then we call the exact solution x(t) of nonlinear INDDE (3.1) and \(\tilde{x}(t)\) of (3.2) are stable.
-
(3)
The exact solution x(t) of nonlinear INDDE (3.1) and \(\tilde{x}(t)\) of (3.2) are said to be asymptotically stable, if they are stable and fulfil
$$\begin{aligned} \lim _{t\rightarrow \infty }\Vert x(t)-\tilde{x}(t)\Vert =0. \end{aligned}$$
-
(1)
-
2.
The meanings of asymptotic stability of numerical method (2.5) for linear equation (2.1) and numerical method (3.26)–(3.27) for nonlinear equation (3.1) are also different. The asymptotic stability of numerical method (2.5) and numerical method (3.26)–(3.27) are given as follows.
-
(1)
Numerical method (2.5) for linear INDDE (2.1) is said to be asymptotically stable if \(x_n\) obtained from (2.5) satisfies
$$\begin{aligned} \lim _{n\rightarrow \infty }x_n=0. \end{aligned}$$ -
(2)
For \(\forall \epsilon >0\), if there exists a constant \(\delta >0\) such that \(\Vert \Phi -\tilde{\Phi }\Vert \le \delta \) and \(\Vert \Phi '-\tilde{\Phi }'\Vert \le \delta \) imply that \(\bar{x}_n\) obtained from (3.26)–(3.27) and \(\bar{\tilde{x}}_n\) obtained from (3.28)–(3.29) satisfy
$$\begin{aligned} \Vert \bar{x}_n-\bar{\tilde{x}}_n\Vert <\epsilon , \end{aligned}$$then we call numerical methods (3.26)–(3.27) for nonlinear INDDE (3.1) and (3.28)-(3.29) for (3.2) are stable.
-
(3)
Numerical methods (3.26)–(3.27) for nonlinear INDDE (3.1) and (3.28)–(3.29) for (3.2) are said to be asymptotically stable, if they are stable and fulfil
$$\begin{aligned} \lim _{n\rightarrow \infty }\Vert \bar{x}_n-\bar{\tilde{x}}_n\Vert =0. \end{aligned}$$
-
(1)
-
3.
For linear equations, the constructed numerical methods furnished by A-stable Runge–Kutta methods can preserve asymptotic stability of the exact solutions. But, for nonlinear equations, the constructed numerical methods furnished by implicit Euler method (or 2-stage Lobatto IIIC method) can preserve asymptotic stability of the exact solutions.
4 Numerical experiments
Example 4.1
Consider the following scalar linear INDDE:
By Theorem 2.1, assume that \(y(t)= {\textrm{e}}^{-\{t\}}x(t)\), \(t\ge -1\), then x(t) is the solution of INDDE (4.1) if and only if y(t) is the solotion of the following equation:
where
By Theorem 2.2, both the solution x(t) of (4.1) and the solution y(t) of (4.2) tend to zeros as \(t\rightarrow \infty \). By Theorem 2.4, the numerical methods (2.5) for (4.1) furnished by A-stable Runge-Kutta methods are stable and asymptotically stable (see Figs. 1 and 2).
Example 4.2
Consider the following scalar nonlinear INDDE:
where a, b and \(\lambda \) are real constants. It is easy to prove that the first equation of (4.3) satisfies inequalities (3.3)–(3.6) with \(R(t)=\Re (a)\), \(\beta (t)=2|b|\), \(\gamma (t)=|b|\), \(\sigma (t)=(|a|+2)|b|\).
-
(i)
When \(a=1/10\), \(b=1/30\), \(\lambda =\frac{1}{{\textrm{e}}}\), by Theorem 3.6, the solution of (4.3) is stable and asymptotically stable. By Theorem 3.16, the numerical methods (3.26)–(3.27) for (4.3) furnished by backward Euler method (2-stage Lobatto IIIC method) are stable and asymptotically stable (See Fig. 3).
-
(ii)
When \(a=1/50\), \(b=1/50\), \(\lambda =\frac{1}{3}\), by Theorem 3.6, Theorem 3.9 or Corollary 3.11 the solution of (4.3) is stable and asymptotically stable. By Theorem 3.16, the numerical methods (3.26)–(3.27) for (4.3) furnished by backward Euler method (2-stage Lobatto IIIC method) are stable and asymptotically stable.
-
(iii)
When \(a=-50\), \(b=1/50\), \(\lambda =\textrm{e}\), by Theorem 3.5, Theorem 3.8, Corollary 3.10, Theorem 3.13 or Corollary 3.14, the solution of (4.3) is bound and asymptotically stable. By Theorem 3.15, the numerical methods (3.26)–(3.27) for (4.3) furnished by backward Euler method (2-stage Lobatto IIIC method) are stable and asymptotically stable.
Tables 1 and 2 roughly illuminate that the constructed method furnished by backward Euler method is convergent of order 1 and by 2-stage Lobatto IIIC method is convergent of order 2.
Data Availability
The datasets generated during the current study are available from the corresponding author on reasonable request.
References
Anguraj A, Karthikeyan K (2009) Existence of solutions for impulsive neutral functional differential equations with nonlocal conditions. Nonlinear Anal 70:2717–2721
Bellen A, Jackiewicz Z, Zennaro M (1988) Stability analysis of one-step methods for neutral delay-differential equations. Numer Math 52:605–619
Cuevasa C, Hernándezb E, Rabelo M (2009) The existence of solutions for impulsive neutral functional differential equations. Comput Math Appl 58:744–757
Engelborghs K, Luzyanina T, Hout KJ, Roose D (2001) Collocation methods for the computation of periodic solutions of delay differential equations. SIAM J Sci Comput 22:1593–1609
Enright WH, Hayashi H (1998) Convergence analysis of the solution of retarded and neutral delay differential equations by continuous numerical methods. SIAM J Numer Anal 35:572–585
Li XD, Deng FQ (2017) Razumikhin method for impulsive functional differential equations of neutral type. Chaos, Solitons Fractals 101:41–49
Li TX, Rogovchenko YV (2015) Oscillation of second-order neutral differential equations. Math Nachr 288:1150–1162
Li TX, Rogovchenko YV (2016) Oscillation criteria for even-order neutral differential equations. Appl Math Lett 61:35–41
Li TX, Rogovchenko YV (2020) On the asymptotic behavior of solutions to a class of third-order nonlinear neutral differential equations. Appl Math Lett 105:1–7
Liu YK (1999) Numerical solution of implicit neutral functional differential equations. SIAM J Numer Anal 36:516–528
Wang WS, Li SF (2004) Stability analysis of nonlinear delay differential equations of neutral type. Math Numer Sin 26:303–314
Wang WS, Li SF (2008) Stability analysis of \(\theta \)-methods for nonlinear neutral functional differential equations. SIAM J Sci Comput 30:2181–2205
Wang WS, Zhang Y, Li SF (2009) Stability of continuous Runge–Kutta-type methods for nonlinear neutral delay-differential equations. Appl Math Model 33:3319–3329
Xue Z, Han XX, Wu KN (2023) Mean square exponential input-to state stability of stochastic Markovian reaction diffusion systems with impulsive perturbations. J Franklin Inst 360:7085–7104
Zhang GL, Song MH, Liu MZ (2015) Asymptotical stability of the exact solutions and the numerical solutions for a class of impulsive differential equations. Appl Math Comput 258:12–21
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by Hui Liang.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work is supported by the NSF of PR China (No. 11701074).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zhang, GL., Sun, Y. & Wang, ZW. Asymptotical stability of the exact solutions and the numerical solutions for impulsive neutral differential equations. Comp. Appl. Math. 43, 8 (2024). https://doi.org/10.1007/s40314-023-02518-0
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s40314-023-02518-0