Asymptotic properties of neutral type linear systems

Leonid Berezansky and Elena Braverman

arXiv: 2012.11507v1 · PDF · 21 December 2020
Categories: math.DS
Licence: CC BY 4.0 The site shows the text.

Abstract

Exponential stability and solution estimates are investigated for a delay system

\dot{x}(t) - A(t)\dot{x}(g(t))=\sum_{k=1}^m B_k(t)x(h_k(t))

of a neutral type, where A and B_k are n\times n bounded matrix functions, and g, h_k are delayed arguments. Stability tests are applicable to a wide class of linear neutral systems with time-varying coefficients and delays. In addition, explicit exponential estimates for solutions of both homogeneous and non-homogeneous neutral systems are obtained for the first time. These inequalities are not just asymptotic estimates, they are valid on every finite segment and evaluate both short- and long-term behaviour of solutions.

Items 24

Proposition

Proposition 1

311K

Let t_0\geq 0, \mu(B(t))\leq -\beta<0 and \displaystyle h \, \sup_{t\geq t_0}\|B(t)\|^2< \inf_{t\geq t_0}\,|\mu(B(t))|.

Then equation (?) is uniformly asymptotically stable.

Proposition

Proposition 2

MSDM

[KM, P. 367],[Li-Ming] Let \|A_1\|<1 and \displaystyle \mu (A_0)+\frac{\|A_0\|+\|A_1\|\|A_2\|}{1-\|A_1\|}<0.

Then system (?) is asymptotically stable.

Proposition

Proposition 3

3I8S

[KM, P. 369] If \|A_1\|+ h_2\|A_2\|<1 and

\displaystyle \mu (A_0+A_2)+\frac{\|A_0+A_2\|(\|A_1\|+h_2\|A_2\|)}{1-\|A_1\|- h_2\|A_2\|}<0

then system (?) is asymptotically stable.

Lemma

Lemma 1

ITR8

Let \|A_0\|_{[t_0,\infty)}<1, there exist an \alpha_0 <0 such that \mu (C(t)+D(t))\leq \alpha_0, Z be a fundamental matrix of system (?), and

K_0:= \frac{\|C\|_{[t_0,\infty)}+\sum_{k=0}^m \|D_k\|_{[t_0,\infty)}}{1-\|A_0\|_{[t_0,\infty)}} \left(\left\|\frac{A_0}{\mu(C+D)}\right\|_{[t_0,\infty)}+\sum_{k=0}^m \tau_k \left\|\frac{D_k}{\mu(C+D)}\right\|_{[t_0,\infty)}\right)<1.

Then, Z(t,s) satisfies

\|Z(t,s)\|\leq K, ~~t \geq s \geq t_0,

where the bound K is K= (1-K_0)^{-1}.

Used by Proof, Proof, Proof

Proof of Lemma 1

We denote z(t)=Z(t,t_0) to make expressions shorter. By definition, z satisfies (?), where the initial matrix is z(t_0)=E, and the initial functions are identically equal to the zero matrix. For an arbitrary t_1>t_0, denote J=[t_0,t_1]. By (?) and (?), we get the inequality

\|\dot{z}\|_J\leq \frac{\|C\|_{[t_0,\infty)}+\sum_{k=0}^m \|D_k\|_{[t_0,\infty)}}{1-\|A_0\|_{[t_0,\infty)}} \|z\|_J.

A fundamental matrix Y(t,s) of the ordinary differential equation

\dot{y}(t)=(C(t)+D(t))y(t)

has an exponential estimate [Coppel, Page 9]

\|Y(t,s)\|\leq e^{\int_s^t \mu (C(\xi)+D(\xi))d\xi}.

Further, we rewrite (?) with a different non-delay term

\dot{z}(t)=[C(t)+D(t)]z(t)+A_0(t)\dot{z}(g(t))-\sum_{k=0}^m D_k(t)\int_{h_k(t)}^t \dot{z}(\xi)d\xi.

Integrating from t_0 to t \leq t_1, we get:

\begin{array}{ll} z(t)= & \displaystyle Y(t,t_0)+\int_{t_0}^t Y(t,s)\mu (C(s)+D(s)) \left [\frac{A_0(s)}{\mu(C(s)+D(s))}\dot{z}(g(s)) \right. \\ & \left. \displaystyle -\sum_{k=0}^m \frac{D_k(s)}{\mu(C(s)+D(s))}\int_{h_k(s)}^s \dot{z}(\xi)d\xi\right]ds. \end{array}

Therefore, using (Proof),(?), inequality \mu (C(t)+D(t))\leq \alpha_0<0 leads to

\left\| \int_{t_0}^t Y(t,s) \mu (C(s)+D(s))~ds \right\| \leq 1

and also recalling K_0 from (1),

\|z\|_J\leq 1+\left(\left\|\frac{A_0}{\mu(C+D)}\right\|_{[t_0,\infty)} +\sum_{k=0}^m \tau_k\left\|\frac{D_k}{\mu(C+D)}\right\|_{[t_0,\infty)}\right)\|\dot{z}\|_J \leq 1+ K_0 \| z \|_J.

We get \|Z(t,t_0)\|_{[t_0,t_1]} \leq (1-K_0)^{-1}, and the expression in

the right-hand side does not depend on t_1, which implies

\displaystyle \|Z(t,t_0)\|_{[t_0,\infty)}\leq (1-K_0)^{-1}. This inequality is also valid if t_0 is replaced with s>t_0.

Hence estimate (?) is valid, which concludes the proof.

Lemma

Lemma 2

TYVS

Let \displaystyle \|A_0\|_{[t_0,\infty)}+\sum_{k=0}^m \tau_k \|D_k\|_{[t_0,\infty)}<1, \mu (C(t)+D(t))\leq \alpha_0 for some \alpha_0 <0, where D from (?) is used, Z be a fundamental matrix of (?), and the constant

\displaystyle L_0:= \displaystyle \frac{\|C+D\|_{[t_0,\infty)}}{1-\|A_0\|_{[t_0,\infty)}-\sum_{k=0}^m \tau_k \|D_k\|_{[t_0,\infty)}} \displaystyle \left(\left\|\frac{A_0}{\mu(C+D)}\right\|_{[t_0,\infty)}+\sum_{k=0}^m \tau_k \left\|\frac{D_k}{\mu(C+D)}\right\|_{[t_0,\infty)}\right)<1.

Then, Z(t,s) satisfies (?), where L= (1-L_0)^{-1} is used instead of K.

Used by Proof

Proof of Lemma 2

Again, z(t)=Z(t,t_0) satisfies (?) with the zero initial matrix-functions and z(t_0)=E. Let J=[t_0,t_1] for any t_1>t_0. Equality (?) implies

\|\dot{z}\|_J\leq \|C+D\|_{[t_0,\infty)}\|z\|_J+ \left( \|A_0\|_{[t_0,\infty)}+\sum_{k=0}^m\tau_k \|D_k\|_{[t_0,\infty)} \right) \|\dot{z}\|_J,

therefore

\|\dot{z}\|_J\leq \frac{\|C+D\|_{[t_0,\infty)}}{1-\|A_0\|_{[t_0,\infty)}-\sum_{k=0}^m\tau_k \|D_k\|_{[t_0,\infty)}}\|z\|_J.

The rest of the proof repeats the scheme for Lemma 1 and thus is omitted.

Uses Proof, Lemma 1

Theorem

Theorem 1

A250

Let \lambda and \beta be positive constants for which

\mu (P(t))\leq -\beta, \quad t \geq t_0, \quad e^{\lambda\sigma}\|A\|_{[t_0,\infty)}<1,

\begin{array}{ll} M_1:= & \displaystyle \frac{\lambda +\sum\limits_{k=1}^m e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}+\lambda e^{\lambda\sigma}\|A\|_{[t_0,\infty)}}{1-e^{\lambda\sigma}\|A\|_{[t_0,\infty)}} \vspace{2mm} \\ & \displaystyle \times \left(\left\|\frac{A}{\mu (P)}\right\|_{[t_0,\infty)} \!\!\! \!\!\! \!\!\!\! (1+\lambda\sigma)e^{\lambda\sigma} +\sum_{k=1}^m \left\|\frac{ B_k}{\mu (P)}\right\|_{[t_0,\infty)} \!\!\!\!\!\!\!\!\!\! e^{\lambda\tau_k}\tau_k \right)<1, \end{array}

where P is defined in (?).

Then, for M_0:=(1-M_1)^{-1}, the solution of (?),(?) satisfies

\begin{array}{ll} \| x(t) \| \leq & \displaystyle M_0e^{-\lambda(t-t_0)}\left[ \| x(t_0) \| +\frac{e^{\lambda \sigma}-1}{\lambda(1-\|A\|_{[t_0,\infty)})} \|A\|_{[t_0,\infty)}\|\Psi\|_{[t_0-\sigma,t_0]}\right. \vspace{2mm} \\ & \displaystyle \left.+\sum_{k=1}^m\frac{e^{\lambda \tau_k}-1}{\lambda(1-\|A\|_{[t_0,\infty)})} \|B_k\|_{[t_0,\infty)}\|\Phi\|_{[t_0-\tau_k,t_0]}\right]+\frac{M_0}{\lambda(1-\|A\|_{[t_0,\infty)})}\|f\|_{[t_0,t]}. \end{array}

Used by Example 2

Proof of Theorem 1

We start with homogeneous system (?). Substituting x(t)=e^{-\lambda(t-t_0)}y(t) into (?), we obtain

\dot{y}(t)-A(t)e^{\lambda(t-g(t))}\dot{y}(g(t))=\lambda y(t)-\lambda e^{\lambda(t-g(t))}A(t)y(g(t)) +\sum_{k=1}^m e^{\lambda(t-h_k(t))}B_k(t)y(h_k(t)).

Equation (Proof) has the form of (?) with

A_0(t)=e^{\lambda(t-g(t))} A(t), ~~C(t)=\lambda E, ~~D_0(t)=-\lambda e^{\lambda(t-g(t))} A(t), ~~h_0(t)=g(t),

D_k(t)=e^{\lambda(t-h_k(t))}B_k(t),~k=1,\dots,m, ~~ D(t)=\sum_{k=0}^m D_k(t).

Therefore P(t)=C(t)+D(t),

\left\|\frac{A_0}{\mu (C+D)}\right\|_{[t_0,\infty)}\leq e^{\lambda\sigma}\left\|\frac{A}{\mu (P)}\right\|_{[t_0,\infty)}, \left\|\frac{D_0}{\mu (C+D)}\right\|_{[t_0,\infty)}\leq \lambda e^{\lambda\sigma}\left\|\frac{A}{\mu (P)}\right\|_{[t_0,\infty)},

\left\|\frac{D_k}{\mu (C+D)}\right\|_{[t_0,\infty)}\leq e^{\lambda\tau_k}\left\|\frac{B_k}{\mu (P)}\right\|_{[t_0,\infty)}, k=1,\dots,m,

\frac{\|C\|_{[t_0,\infty)}+\sum_{k=0}^m \|D_k\|_{[t_0,\infty)}}{1-\|A_0\|_{[t_0,\infty)}}\leq \frac{\lambda+\sum_{k=1}^m e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}+\lambda e^{\lambda\sigma}\|A\|_{[t_0,\infty)}}{1-e^{\lambda\sigma}\|A\|_{[t_0,\infty)}}.

Inequalities (1) and (?) imply the assumptions of Lemma 1, in particular, (1). Let Y(t,s) be a fundamental matrix of (Proof), then we can apply Lemma 1 to deduce \|Y(t,s)\|\leq M_0. Once X(t,s) is a fundamental matrix of system (?), it satisfies X(t,s)=e^{-\lambda(t-s)}Y(t,s). This implies the exponential estimate \|X(t,s)\| \leq M_0 e^{-\lambda(t-s)}. Let x be a solution of problem (?),(?). We use solution representation (?) and inequality (?)

\begin{aligned} \|x(t)\|\leq & \|X(t,t_0)\|\|x_0\| + \left. \left. \int_{t_0}^{t_0+\sigma} \|X(t,s)\| \right\| (I-S)^{-1} \right\|_{L_{\infty}[t_0,t_1] \to L_{\infty}[t_0,t_1]} \|A(s)\|\|\Psi(g(s))\|ds \\ & +\sum_{k=1}^m \left. \left. \int_{t_0}^{t_0+\tau_k} \|X(t,s)\| \right\| (I-S)^{-1} \right\|_{L_{\infty}[t_0,t_1] \to L_{\infty}[t_0,t_1]} \|B_k(s)\|\|\Phi(h_k(s))\|ds \\ \leq & M_0 e^{-\lambda(t-t_0)} \|x(t_0)\|+\frac{M_0}{\lambda\left(1-\|A\|_{[t_0,\infty)}\right)} \|A\|_{[t_0,\infty)} \left( e^{-\lambda (t-t_0-\sigma)}-e^{-\lambda(t-t_0)} \right)\|\Psi\|_{[t_0-\sigma,t_0]} \\ & +\frac{M_0}{\lambda\left(1-\|A\|_{[t_0,\infty)}\right)} \sum_{k=1}^m \|B_k\|_{[t_0,\infty)} \left( e^{-\lambda (t-t_0-\tau_k)} -e^{-\lambda(t-t_0)} \right)\|\Phi\|_{[t_0-\tau_k,t_0]}, \end{aligned}

which immediately yields (?) when f \equiv 0.

For any f in (?), we employ the above inequality for \|X(t,s)\| and

\begin{aligned} \left\| \int_{t_0}^t X(t,s) [ (I-S)^{-1}f](s)\, ds \right\| \leq \frac{M_0}{\lambda\left(1-\|A\|_{[t_0,\infty)}\right)} \|f\|_{[t_0,t]}. \end{aligned}

Uses Lemma 1

Theorem

Theorem 2

FZJ0

Let \lambda and \beta be positive constants for which

\mu (P(t))\leq -\beta, \quad t \geq t_0, \quad (1+\lambda \sigma)e^{\lambda\sigma}\|A\|_{[t_0,\infty)} +\sum_{k=1}^m \tau_k e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}<1,

\begin{array}{ll} M_2:= & \frac{\lambda +\sum\limits_{k=1}^m e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}+\lambda e^{\lambda\sigma}\|A\|_{[t_0,\infty)}} {1-(1+\lambda \sigma)e^{\lambda\sigma}\|A\|_{[t_0,\infty)} -\sum_{k=1}^m \tau_k e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}} \vspace{2mm} \\ & \displaystyle \times \left(\left\|\frac{A}{\mu (P)}\right\|_{[t_0,\infty)} \!\! \!\!\!\! (1+\lambda\sigma)e^{\lambda\sigma} +\sum_{k=1}^m \left\|\frac{ B_k}{\mu (P)}\right\|_{[t_0,\infty)} \!\!\!\!\! e^{\lambda\tau_k}\tau_k \right)<1, \end{array}

where P is defined in (?). Then the solution x of (?),(?) satisfies (?) with M_0:=(1-M_2)^{-1}.

Proof of Theorem 2

The proof of the theorem is similar to the proof of Theorem 1 if we use Lemma 2 rather than Lemma 1, (Proof) and the same notation.

Uses Lemma 2, Lemma 1, Proof

Corollary

Corollary 1

NCE7

Let \lambda and \beta be positive constants for which

\mu (P_1(t))\leq -\beta, \quad t \geq t_0,\quad e^{\lambda\sigma}\|A\|_{[t_0,\infty)}<1,

M_3:= \frac{\lambda +e^{\lambda\tau}\|B\|_{[t_0,\infty)}+\lambda e^{\lambda\sigma}\|A\|_{[t_0,\infty)}}{1-e^{\lambda\sigma}\|A\|_{[t_0,\infty)}} \left(\left\|\frac{A}{\mu (P_1)}\right\|_{[t_0,\infty)} \!\!\! \!\!\! \!\!\!\! (1+\lambda\sigma)e^{\lambda\sigma} + \left\|\frac{ B}{\mu (P_1)}\right\|_{[t_0,\infty)} \!\!\!\!\!\!\!\!\!\! e^{\lambda\tau}\tau \right)<1.

Then for the solution x of problem (?),(?) and M_0:=(1-M_3)^{-1}, we have

\begin{array}{ll} \|x(t)\|\leq & \displaystyle M_0 e^{-\lambda(t-t_0)}\left[\|x(t_0)\| +\frac{e^{\lambda \sigma}-1}{\lambda(1-\|A\|_{[t_0,\infty)})} \|A\|_{[t_0,\infty)}\| \Psi \|_{[t_0-\sigma,t_0]}\right. \vspace{2mm} \\ & \displaystyle \left.+\frac{e^{\lambda \tau}-1}{\lambda(1-\|A\|_{[t_0,\infty)})} \|B\|_{[t_0,\infty)}\|\Phi\|_{[t_0-\tau,t_0]}\right]+\frac{M_0}{\lambda(1-\|A\|_{[t_0,\infty)})}\|f\|_{[t_0,t]}. \end{array}

Corollary

Corollary 2

Z86X

Let \lambda and \beta be positive constants for which \mu (P_2(t))\leq -\beta for t \geq t_0 and

M_4:= \left(\lambda +\sum\limits_{k=1}^m e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}\right) \sum_{k=1}^m \left\|\frac{ B_k}{\mu (P_2)}\right\|_{[t_0,\infty)} \!\!\!\!\!\!\!\!\!\! e^{\lambda\tau_k}\tau_k <1.

Then, for M_0:=(1-M_4)^{-1}, the solution x of problem (?) satisfies

\|x(t)\|\leq \displaystyle M_0e^{-\lambda(t-t_0)}\left[\|x(t_0)\| +\sum_{k=1}^m\frac{e^{\lambda \tau_k}-1}{\lambda} \|B_k\|_{[t_0,\infty)}\|\Phi\|_{[t_0-\tau_k,t_0]}\right] +\frac{M_0}{\lambda}\|f\|_{[t_0,t]}.

Theorem

Theorem 3

YZ56

Let \beta be a positive constant for which \displaystyle B(t):=\sum_{k=1}^m B_k(t), ~\mu (B(t))\leq -\beta,~\|A\|_{[t_0,\infty)}<1 and

\frac{\sum_{k=1}^m \|B_k\|_{[t_0,\infty)}}{1-\|A\|_{[t_0,\infty)}} \left(\left\|\frac{A}{\mu (B)}\right\|_{[t_0,\infty)}+\sum_{k=1}^m \tau_k\left\|\frac{B_k}{\mu (B)}\right\|_{[t_0,\infty)}\right)<1.

Then equation (?) is uniformly exponentially stable.

Used by Proof, Example 2

Theorem

Theorem 4

X0UZ

Let \beta be a positive constant for which \mu (B(t))\leq - \beta, where

\displaystyle B(t):=\sum_{k=1}^m B_k(t), ~ ~\|A\|_{[t_0,\infty)}+\sum_{k=1}^m \tau_k\|B_k\|_{[t_0,\infty)}<1

and

\left(\frac{\|B\|_{[t_0,\infty)}}{1-\|A\|_{[t_0,\infty)}-\sum_{k=1}^m \tau_k\|B_k\|_{[t_0,\infty)}} \right) \left(\left\|\frac{A}{\mu (B)}\right\|_{[t_0,\infty)}+\sum_{k=1}^m \tau_k\left\|\frac{B_k}{\mu (B)}\right\|_{[t_0,\infty)}\right)<1.

Then equation (?) is uniformly exponentially stable.

Used by Proof

Corollary

Corollary 3

SLBE

Let \beta be a positive constant for which \mu (B(t))\leq - \beta, where \displaystyle B(t):=\sum_{k=1}^m B_k(t), |A(t)|\leq \overline{A}, |B_k(t)|\leq \overline{B_k}, |B(t)|\leq \overline{B}, \|\overline{A}\|<1, and at least one of the following conditions holds:

\sum_{k=1}^m \|\overline{B_k}\|\left(\|\overline{A}\|+\sum_{k=1}^m \tau_k\|\overline{B_k}\|\right)<\beta (1-\|\overline{A}\|);

\|\overline{B}\|\left(\|\overline{A}\|+\sum_{k=1}^m \tau_k\|\overline{B_k}\|\right) <\beta\left(1-\|\overline{A}\|-\sum_{k=1}^m \tau_k\|\overline{B_k}\|\right).

Then equation (?) is uniformly exponentially stable.

Corollary

Corollary 4

UHA4

Let \beta be a positive constant for which \mu (B(t))\leq - \beta for any t \geq t_0, \|A\|_{[t_0,\infty)}<1,

\left(\left\|\frac{A}{\mu (B)}\right\|_{[t_0,\infty)}+\tau\left\|\frac{B}{\mu (B)}\right\|_{[t_0,\infty)}\right)\|B\|_{[t_0,\infty)}<1-\|A\|_{[t_0,\infty)}.

Then equation (?) with f\equiv 0 is uniformly exponentially stable.

Corollary

Corollary 5

ZSLZ

Let \beta be a positive constant for which \mu (B(t))\leq - \beta for any t \geq t_0, where \displaystyle B(t):=\sum_{k=1}^m B_k(t), and either
1) \displaystyle \sum_{k=1}^m \|B_k\|_{[t_0,\infty)} \sum_{k=1}^m \tau_k\left\|\frac{B_k}{\mu (B)}\right\|_{[t_0,\infty)} <1
or
2) \displaystyle \|B\|_{[t_0,\infty)} \sum_{k=1}^m \tau_k\left\|\frac{B_k}{\mu (B)}\right\|_{[t_0,\infty)} < 1- \sum_{k=1}^m \tau_k\left\|B_k\right\|_{[t_0,\infty)}
is satisfied. Then, system (?) with f\equiv 0 is uniformly exponentially stable.

Corollary

Corollary 6

4JRQ

Let \beta be a positive constant for which \mu (B(t))\leq - \beta for any t \geq t_0, and

\displaystyle \tau\left\|\frac{B}{\mu (B)}\right\|_{[t_0,\infty)} \!\!\!\! \|B\|_{[t_0,\infty)}<1.

Then system (?) is uniformly exponentially stable.

Corollary

Corollary 7

RJJ4

Assume that for some \beta>0, \displaystyle b(t):= \sum_{k=1}^m b_k(t) \leq -\beta <0,~t\geq t_0 and either

\sum_{k=1}^m\|b_k\|_{[t_0,\infty)}\left(\left\|\frac{a}{b}\right\|+\sum_{k=1}^m \tau_k \left\|\frac{b_k}{b} \right\|\right)<1-\|a\|_{[t_0,\infty)}

\|b\|_{[t_0,\infty)}\left(\left\|\frac{a}{b}\right\| +\sum_{k=1}^m \tau_k \left\|\frac{b_k}{b}\right\|\right)<1-\|a\|_{[t_0,\infty)}-\sum_{k=1}^m \tau_k \|b_k\|.

Then equation (?) is uniformly exponentially stable.

Used by Example 1

Example

Example 1

PQZQ

Consider a scalar equation with variable oscillating coefficients

\dot{x}(t)-\nu\left[0.1 \sin t ~\dot{x}(g(t))=-(1-3\cos t) x(t)-(1+3\cos t) x(h(t))\right],

where \nu>0, t-g(t)\leq \sigma, t-h(t)\leq 1=\tau. Here m=2 and

\|a\|_{[t_0,\infty)}=0.1 \nu,~\|b_1\|_{[t_0,\infty)}=\|b_2\|_{[t_0,\infty)}=4\nu,~b=b_1+b_2=-2\nu<0, \tau=1.

By (7) equation (1) is uniformly exponentially stable if \nu<\frac{1}{16.5}, by (7) the stability condition is \nu<\frac{1}{8.2}. Hence for this equation condition (7) is better than (7).

In equation (1), the value of \|b_1+b_2\|_{[t_0,\infty)} is significantly less than \|b_1\|_{[t_0,\infty)}+\|b_2\|_{[t_0,\infty)}. If they are close, condition (7) is better than (7). For example, if the term of 3\cos t is omitted in the coefficients in (1), estimate in (7) becomes sharper than in (7) (\nu<\frac{1}{2.2} compared to \nu<\frac{1}{3.2}, respectively).

Uses Corollary 7

Corollary

Corollary 8

IS8L

Assume that for some \beta>0, \mu(A_0(t)+A_2(t))\leq -\beta<0, ~t\geq t_0 and at least one of the following conditions holds:

\displaystyle \|A_1\|_{[t_0,\infty)}<1,~ -\beta+\frac{(\|A_0\|_{[t_0,\infty)} +\|A_2\|_{[t_0,\infty)})(\|A_1\|_{[t_0,\infty)}+h_2\|A_2\|_{[t_0,\infty)})}{1-\|A_1\|_{[t_0,\infty)}}<0;

\displaystyle \|A_1\|_{[t_0,\infty)}+h_2\|A_2\|_{[t_0,\infty)}<1,~ -\beta+\frac{\|A_0+A_2\|_{[t_0,\infty)}(\|A_1\|_{[t_0,\infty)} +h_2\|A_2\|_{[t_0,\infty)})}{1-\|A_1\|_{[t_0,\infty)}-h_2\|A_2\|_{[t_0,\infty)}}<0.

Then equation (?) is uniformly exponentially stable.

Proof of Corollary 8

Assume first that (?) holds. We will apply Theorem 3 for m=2 to equation (?). Inequality (3) for (?) has the form

\frac{\|A_0\|_{[t_0,\infty)}+\|A_2\|_{[t_0,\infty)}}{1-\|A_1\|_{[t_0,\infty)}} \left(\left\|\frac{A_1}{\mu(A_0+A_2)}\right\|_{[t_0,\infty)} +h_2\left\|\frac{A_2}{\mu(A_0+A_2)}\right\|_{[t_0,\infty)}\right)<1,

which, due to |\mu(A_0(t)+A_2(t))| \geq \beta, yields that

\frac{\|A_0\|_{[t_0,\infty)}+\|A_2\|_{[t_0,\infty)}}{1-\|A_1\|_{[t_0,\infty)}} \left(\|A_1\|_{[t_0,\infty)}+h_2\|A_2\|_{[t_0,\infty)}\right)<\beta.

Inequality (Proof) is equivalent to the second inequality in (?).

All conditions of Theorem 3 hold for equation (?), hence this equation is uniformly exponentially stable.

The second part of the corollary follows from Theorem 4 and is justified in a similar way.

Uses Theorem 3, Theorem 4

Example

Example 2

XI8A

Consider system (?) for m=2

\dot{x}(t) - A(t)\dot{x}(g(t)) = B_1(t)x(h_1(t))+B_2(t)x(h_2(t)), ~~t \geq t_0\geq 0,

where B_1(t)=\sin^2 t~ C, B_2(t)=\cos^2 t ~C, C is a tri-diagonal and A is a variable matrix

C=\begin{pmatrix} -\alpha&\beta&0&0&\dots&0&0\\ \beta/2 &-\alpha&\beta/2&0&\dots&0&0\\ 0&\beta/2 &-\alpha&\beta/2&\dots&0&0\\ \vdots&\vdots&\vdots&\vdots&\vdots&\vdots&\vdots\\ 0&0 &0&0 &\dots &\beta&-\alpha \end{pmatrix}, ~~A=\gamma\begin{pmatrix} \cos t& \cos 2t&\cos 3t&\dots&\cos nt\\ \cos^2 t& \cos^2 2t&\cos^2 3t&\dots&\cos^2 nt\\ \vdots&\vdots&\vdots&\vdots&\vdots\\ \cos^n t& \cos^n 2t&\cos^n 3t&\dots&\cos^n nt\\ \end{pmatrix},

\alpha>0, h_1(t)=t-0.1|\sin t|, h_2(t)=t-0.1|\cos t|, t-g(t)\leq \sigma. We have for the norm \|\cdot \|_{[0,\infty)}:

\|A\|_{[0,\infty)}=n|\gamma|, ~~\tau_i=\sup_{t\geq 0} (t-h_i(t))=0.1, ~~\|B_i\|_{[0,\infty)}=\alpha+|\beta|, ~i=1,2,

B=B_1+B_2, \mu (B)=-\alpha+|\beta|. Let

Then, by Theorem 3 equation (2) is uniformly exponentially stable.

In particular, conditions (2) hold for \alpha=0.4, |\beta|=0.1, n|\gamma|=0.01.

Assume in addition \|A\|=n|\gamma|=0.01, \lambda=0.06, \sigma=0.1. Hence \|B_i\|_{[0,\infty)}=0.5, \mu(C)=-0.3 and calculate by Theorem 1 an estimate for solution x of the equation

\dot{x}(t) - A(t)\dot{x}(g(t)) = B_1(t)x(h_1(t))+B_2(t)x(h_2(t))+f(t), ~~t \geq t_0\geq 0,

with initial condition (?). We have

\begin{aligned} \mu (P(t))= & \mu\left(e^{\lambda (t-h_1(t))}B_1(t)+e^{\lambda (t-h_2(t))}B_2(t)-\lambda e^{\lambda (t-g(t))} A(t)+\lambda E\right) \\ & \leq \mu\left(e^{\lambda (t-h_1(t))}B_1(t)+e^{\lambda (t-h_2(t))}B_2(t)\right)+\mu\left(-\lambda e^{\lambda (t-g(t))} A(t)\right)+\mu(\lambda E). \end{aligned}

Then

\mu\left(e^{\lambda (t-h_1(t))}B_1(t)+e^{\lambda (t-h_2(t))}B_2(t)\right)=\left(e^{\lambda (t-h_1(t))}\sin^2 t+e^{\lambda (t-h_2(t))}\cos^2 t\right) \mu (C).

Since \mu (C)=-0.3<0 and

e^{\lambda (t-h_1(t))}\sin^2 t+e^{\lambda (t-h_2(t))}\cos^2 t\geq \sin^2 t+\cos^2 t=1,

we get

\mu\left(e^{\lambda (t-h_1(t))}B_1(t)+e^{\lambda (t-h_2(t))}B_2(t)\right)\leq \mu(C)=-0.3.

Next,

\mu\left(-\lambda e^{\lambda (t-g(t))} A(t)\right)\leq \lambda e^{\lambda (t-g(t))} \|A\|_{[0,\infty)}\leq 0.0006 e^{0.006},~ \mu(\lambda E)=\lambda =0.06.

Then \mu (P(t)) \leq -0.23939<0 and

M_1:= \frac{\lambda +\sum\limits_{k=1}^2 e^{\lambda\tau_k}\|B_k\|_{[t_0,\infty)}+ \lambda e^{\lambda\sigma}\|A\|_{[t_0,\infty)}}{1-e^{\lambda\sigma}\|A\|_{[t_0,\infty)}} \left(\left\|\frac{A}{\mu (P)}\right\|_{[t_0,\infty)} \!\!\! \!\!\! \!\!\!\! (1+\lambda\sigma)e^{\lambda\sigma} +\sum_{k=1}^2 \left\|\frac{ B_k}{\mu (P)}\right\|_{[t_0,\infty)} \!\!\!\!\!\!\!\!\!\! e^{\lambda\tau_k}\tau_k \right).

By numerical calculations, M_1\leq 0.5 and M_0=(1-M_1)^{-1}\leq 2. For the solution x of problem (?),(?) we have (here we omitted the norm indices)

\|x(t)\|\leq \left. \left. 2 e^{-0.06(t-t_0)}\right[ \|x(t_0)\|+0.00102 \|\Psi\| +0.102\|\Phi\|\right]+33.6 \|f\|, ~t\geq t_0\geq 0.

Uses Theorem 3, Theorem 1

Source and licence

The text on this page is from arXiv:2012.11507v1 by Leonid Berezansky and Elena Braverman, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2012/2012.11507.jsonl.