Asymptotic Behavior of Individual Orbits of Discrete Systems

Without loss of generality we may assume that \xi_0=1. Consider \lambda in a small neighborhood of 1 in the complex plane. We will express \lambda =e^{z} with |z|< \delta_0. Choose a small \delta_0 >0 such that if |z|<\delta_0, then

Notice that if 0<|z |<\delta _0, then

Set f(z )= R(e^z ,\bar S)\bar x with |z|<\delta_0. Since 1 is a singular point of \| R(\lambda ,\bar S)\bar x \|, 0 is a singular point of f(z) in \{ |z|<\delta_0 \} . For each n\in \mathbb Z and 0<r<\delta_0, we have

If z =re^{i\varphi}, where \varphi is real, then one has

Therefore,

Consider the Laurent series of f(z ) at z =0,

where

From (Proof) it follows that

Therefore,

Letting r tend to 0 in (Proof), we come up with a_{-k}=0 for all k \ge 2. This shows that z=0 is a removable singular point (when a_{-1}=0) or a pole of first order of f(z ). This yields that the complex function g(\lambda ):= R(\lambda ,\bar S )\bar x has \lambda =1 as a removable singular point or a pole of first order. The lemma is proven.

As shown in Lemma 2.2, \xi is either a removable singular point or pole of first order. Without loss of generality we may assume that \xi_0=1 for the reader’s convenience. Then, the Laurent series of g(\lambda) is of the form

We need to show that under condition (2.3) the coefficient b_{-1}=0. In fact,

This shows \xi_0 is removable. The lemma is proven.

Consider R(\lambda ,\bar S )\bar x for all |\lambda |\not= 1. Since x is a bounded solution of (?) and \bar y=0 we have

On the other hand, the identity \lambda R(\lambda ,\bar S ) \bar x -\bar x = R(\lambda ,\bar S )\bar S \bar x gives

so,

Obviously, R(\lambda ,\bar S )\bar x is analytic on \mathbb C \backslash \Gamma. Moreover, if |\xi_1 |=1 and \xi_1\not\in \sigma (B)\cap \Gamma, (as we can easily check that \sigma (\bar B )=\sigma (B)), in a small neighborhood U(\xi_1 ) of \xi_1 we have

This shows that g(\lambda ) = R(\lambda ,\bar S )\bar x is analytically extendable to a neighborhood of \xi_1, that is, \xi_1 \not\in \sigma (x). The lemma is proven.

(i): If \sigma (x)=\emptyset, the function g(\lambda ):=R(\lambda ,\bar S)\bar x can be extended to an entire function. Using exactly the argument in the proof of Theorem 2.1 we come up with the boundedness of the complex function t(\lambda ):= (\lambda -1) R(\lambda ,\bar S)\bar x, so by Liouville’s Theorem t(\lambda ) is a constant. And thus, t(\lambda )=\lim_{\lambda\to 1} (\lambda -1)g(\lambda )= 0. The injectiveness of R(\lambda ,\bar S) for each |\lambda|\not =1 yields that \bar x=0. The converse is clear.

(ii): By (i), \bar x\not= 0, so \rho (\bar S |_{{\mathcal M}_{\bar x}})\not= \emptyset. Let \xi_0\in \rho (\bar S |_{{\mathcal M}_{\bar x}}). Then, since for |\lambda |\not= 1

it is clear that \xi_0 is a regular point of g(\lambda ).

Conversely, let \xi_0 be a regular point of g(\lambda). Without loss of generality we may assume |\xi_0 |=1, otherwise it is already in \rho (\bar S |_{{\mathcal M}_{\bar x}}). We will show that \xi_0\in \rho (\bar S |_{{\mathcal M}_{\bar x}}) by proving that the equation

has a unique solution v\in {\mathcal M}_{\bar x} for each given w\in {\mathcal M}_{\bar x}. First, we show that there is at least one solution. In fact, we note that for each n\in\mathbb Z the set of regular points of g(\lambda )=R(\lambda ,\bar S)\bar x is the same as that of \bar S^n g(\lambda )=R(\lambda ,\bar S)\bar S^n\bar x. And in turn, by the property of holomorphic functions, the set of all regular points of g(\lambda )=R(\lambda ,\bar S)\bar x must be part of that of the function k(\lambda )=R(\lambda ,\bar S)w, so k(\lambda )=R(\lambda ,\bar S)w is analytically extendable to a neighborhood of \xi_0. In particular, \lim_{\lambda \to\xi_0} k(\lambda )=v\in {\mathcal M}_{\bar x}, so

To show that equation (?) has a unique solution in {\mathcal M}_{\bar x} we can show that the homogeneous equation \xi_0 v-\bar Sv=0 has only a trivial solution in {\mathcal M}_{\bar x}. In fact, let v_0\in {\mathcal M}_{\bar x} be a solution of this equation. Then, for each |\lambda|>1, using the identity R(1 ,A)=(I-A)^{-1}=\sum_{n=0}^\infty A^n for each \|A\| <1 and \bar S^n v_0=\xi_0^n v_0 we have

Since v_0\in {\mathcal M}_{\bar x}, this function must, as above, be extendable analytically to a neighborhood of \xi_0, and this is possible only if v_0=0. Summing up, we have that \xi_0\in\rho ({\mathcal M}_{\bar x}), so the lemma is proven.

We have to show (2.8), that is, \bar x= 0, or equivalently, {\mathcal M}_{\bar x} is trivial. Suppose to the contrary that it is not. Then, by Lemma 2.7, \sigma (x)= \sigma (\bar S |_{{\mathcal M}_{\bar x}})\not= \emptyset. Since \sigma (x) is a non-empty closed subset of \Gamma and is countable, it has an isolated point, say \xi_0, so \xi_0 is an isolated singular point for g(\lambda ). By Lemma 2.2 this isolated singular point must be either a removable singular point or a pole of first order. Since \xi_0 is a pole of first order of the resolvent R(\lambda , \bar S |_{{\mathcal M}_{\bar x}}), by a well known result in Functional Analysis(We actually avoid applying Geldfand Theorem in this case.) (see e.g. [tay, Theorem 5.8 A, p. 306], or, [yos, Theorem 3, p. 229]) \xi_0 must be an eigenvalue of \bar S |_{{\mathcal M}_{\bar x}} with a non-zero eigenvector w_0. As in the proof of Lemma 2.7, (see ?)), for each |\lambda|\not= 1 we have

On the other hand, by Lemma 2.3 \xi_0 is a removable singular point for g(\lambda ), so is for R(\lambda ,\bar S)w_0. This is possible only if w_0=0, contradicting that w_0 is a non-zero vector. This proves the theorem.

This theorem is an immediate consequence of Lemma 2.5 and Theorem 2.8.

Let x(n)=B^nx_0. Then, (x(n)) is a bounded solution of (?) with (y(n))=0. Therefore, if |\lambda |>1, \lambda \in \rho (B) and \lambda \in \rho (\bar S), so by (Proof) (and the proof of Lemma 2.5),

Therefore, by Theorem 2.10, x(n)=B^nx_0 \to 0.