Abstract
In Linear Dynamics, the most studied class of linear operators is certainly that of weighted shifts, on the separable Banach spaces \(c_0\) and \(\ell ^p\), \(1 \le p < \infty \). Over the last decades, the intensive study of such operators has produced an incredible number of versatile, deep and beautiful results that are applicable in various areas of Mathematics; and the relationships between various important notions, especially concerning chaos and hyperbolic properties, as well as spectrum of weighted shifts, have been investigated. In this paper, we investigate the point spectrum of weighted shifts and, under some regularity hypotheses on the weight sequence, we deduce the spectrum.
Similar content being viewed by others
1 Introduction
In 1929, Birkhoff obtained an example of a linear operator having a dense orbit [5]. Some years later, the same has been shown for other two types of fundamental operators in analysis: the differentiation operators [22] and the shifts [27]. Motivated by these three examples and by the fact that density of orbits is one of the main ingredients of chaos, in the twentieth century, many mathematicians began to focus on the dynamical properties of linear operators starting from the analysis of the three ones mentioned above, thus leading to the birth of the fascinating area of Mathematics known as Linear Dynamics. Over the years, the usefulness of the Birkhoff operators, of the differentiation operators and of the shifts, has become undoubted, making them a model for understanding the behaviors of more complex operators. This is particularly illustrated by the shifts or, more generally, the weighted shifts, whose flexibility and “simplicity of action” have made them not only the most studied class of linear operators, but also an excellent tool to clearly illustrate definitions and create crucial counterexamples in Linear Dynamics. In particular, weighted shifts are known to be essential for the comprehension of some classes of linear operators, among which composition operators [11], just to cite one of the most famous and versatile examples. Hence, starting in 1995 with the characterization of hypercyclic weighted shifts [29], their dynamical properties have been so studied that, nowadays, the most of the fundamentals of Linear Dynamics (like hypercyclicity, mixing, chaos, Li-Yorke chaos, expansivity, hyperbolicity and shadowing) are completely characterized for such operators [3, 4, 10, 12]. See [1, 18] and [23] for a global overview on the subject.
As it is often the case in Operator Theory, it may be useful to know the structure of the spectrum and to impose some conditions on it, to ensure that a certain dynamical property manifests. For instance, it is known that an invertible operator T on \(\mathbb {C}^n\) is expansive if and only if the point spectrum of T does not intersect the unit circle \({{\mathbb {T}}}\) [15, Theorem 1]. Subsequently, as it is well-known that an invertible operator T on a Banach space X is hyperbolic if its spectrum does not intersect \({{\mathbb {T}}}\), it was shown in [3, Theorem D] and [24, Theorem 1] that T is uniformly expansive if and only if the same condition is satisfied by the approximate point spectrum, providing, as a corollary, that invertible hyperbolic operators are uniformly expansive. Moreover, the above requirement on the spectrum is a necessary condition to get Li-Yorke chaos [2, Corollary 6]. As hyperbolicity, also generalized hyperbolicity is defined, for an invertible operator T on a Banach space X, in terms of the spectrum of certain restrictions of the operators. These results are just a glimpse of the close relationship between the spectrum of an operator T and its dynamical properties and, also in the specific case of weighted shifts, it turns out that the spectrum plays a key role in the related theory.
In this paper we focus on weighted shifts on \(\ell ^p\)-spaces: we generalize some results, proved on the Hilbert space \(\ell ^2\), to weighted shifts defined on the Banach space \(\ell ^p\), \(1\le p <\infty .\) The present article aims, after recalling basic and fundamental results about the spectrum of shifts and weighted shifts (Sect. 2), to investigate the point spectrum of weighted shifts and, under some regularity hypotheses on the weight sequence, to deduce the spectrum (Sect. 3).
Throughout the article, as usual, \({{\mathbb {N}}}\) denotes the set of all positive integers; \({{\mathbb {Z}}}\) denotes the set of all integers; \({{\mathbb {D}}}\) and \({{\mathbb {T}}}\) are, respectively, the open unit disk and the unit circle in the complex plane \({\mathbb {C}}\).
2 Definitions and background results
In this section, we fix some terminology and recall some basic notions and results concerning the spectral theory of, first, a bounded linear operator on a complex Banach space X and, then, more specifically, of shifts [13, 19, 21]. In the sequel, \({{\mathcal {L}}}(X)\) denotes the algebra of all bounded linear operators from the complex Banach space X into itself.
2.1 Spectrum and spectral radius
Definition 2.1
Let \(T \in {{\mathcal {L}}}(X).\) The spectrum \(\sigma (T)\) of the operator T is the set
Definition 2.2
Let \(T \in {{\mathcal {L}}}(X).\) The spectral radius r(T) of the operator T is defined by
By the Gelfand formula (see, e.g., [28, Chap. 10]), the spectral radius r(T) satisfies the spectral radius formula \(r(T)=\lim _{n \rightarrow \infty } \Vert T^n \Vert ^{\frac{1}{n}}\) and hence \(r(T) \le \Vert T \Vert \), and that \(\sigma (T)\) is a non-empty compact subset of \({{\mathbb {C}}}\) contained in the ball \(\{ \lambda \in {{\mathbb {C}}} : \vert \lambda \vert \le \Vert T \Vert \}\). Moreover, if T is invertible then
and so
and hence, in this case the spectrum \(\sigma (T)\) of T lies in the annulus \(\{ \lambda \in {{\mathbb {C}}} : \frac{1}{r(T^{-1})} \le \vert \lambda \vert \le r(T) \}\).
We now recall the definitions of some subsets of the spectrum which will be used in the sequel.
Definition 2.3
Let \(T \in {{\mathcal {L}}}(X).\)
-
(1)
The set of all \(\lambda \in \sigma (T)\) such that \(T-\lambda I\) is not one-to-one is called the point spectrum of T and is denoted by \(\sigma _p(T)\). It follows that \(\sigma _p(T)\) consists of all the eigenvalues of T.
-
(2)
The set of all \(\lambda \in \sigma (T)\) such that \(T-\lambda I\) is not Fredholm (an operator is Fredholm if its range is closed and both its kernel and its cokernel are finite-dimensional) is called the essential spectrum of T and is denoted by \(\sigma _{e}(T)\).
The essential spectrum \(\sigma _e(T)\) is a closed subset of \(\sigma (T)\) and, clearly, it is empty when the space X is finite-dimensional [20]. Moreover, it is invariant under compact perturbation, as the following result shows:
Proposition 2.4
[14, Theorem 4.1] Let \(T \in {{\mathcal {L}}}(X)\). If S is a compact operator then
We recall the following useful result:
Proposition 2.5
[19, Proposition 1.F] Let T be a bounded operator on a Banach space X . If there exists an invertible operator \(S:X \rightarrow X\) for which \(\Vert S - T\Vert < \Vert S^{-1}\Vert ^{-1}\), then T is itself invertible.
We conclude this paragraph by recalling that the above definitions of spectrum can be extended to any bounded linear operator T acting on a real Banach space X via its complexification \(T_{{\mathbb {C}}}\), in the sense that we define the spectrum of T as the set of all \(\lambda \in {\mathbb {C}}\) such that \(T_{{\mathbb {C}}} - \lambda I\) is not invertible as an operator acting on \(X_{{\mathbb {C}}}\), i.e. we define \(\sigma (T)=\sigma (T_{{\mathbb {C}}}).\)
2.2 Weighted shifts
We recall some preliminary definitions and results.
Definition 2.6
Let \(X=\ell ^p({{\mathbb {N}}})\), \(1 \le p < \infty \) or \(X=c_0({{\mathbb {N}}}).\) Let \(w=\{w_n\}_{n \in {{\mathbb {N}}}}\) be a bounded sequence of scalars, called weight sequence. Then,
-
the unilateral weighted forward shift \(F_w:X \rightarrow X\) is defined by
$$\begin{aligned}F_w( \{x_n\}_{n \in {\mathbb {N}} }) =\{w_{n-1}x_{n-1}\}_{n \in {\mathbb {N}}}, \end{aligned}$$meaning
$$\begin{aligned} F_w(\{x_1,x_2,...\})=\{0,w_1x_1,w_2x_2,...\}; \end{aligned}$$ -
the unilateral weighted backward shift \(B_w:X \rightarrow X\) is defined by
$$\begin{aligned}B_w( \{x_n\}_{n \in {{\mathbb {N}}}}) =\{w_{n+1}x_{n+1}\}_{n \in {{\mathbb {N}}}},\end{aligned}$$meaning
$$\begin{aligned} B_w(\{x_1,x_2,...\})=\{w_2x_2,w_3x_3,...\}. \end{aligned}$$
If, instead of \({\mathbb {N}}\), we consider \({{\mathbb {Z}}}\), the shift is called bilateral.
Clearly, a weighted shift (unilateral or bilateral) is injective if and only if none of the weights is zero, and a bilateral weighted shift is invertible if and only if \(\inf _{n \in {\mathbb {Z}}} \vert w_n \vert >0\). Of course, a unilateral weighted shift is never invertible.
Remark 2.7
Let \(w=\{w_n\}_{n \in {{\mathbb {Z}}}}\) be a weight sequence with \(\inf _{n \in {\mathbb {Z}}} \vert w_n \vert >0\). Obviously, \(B_w^{-1}=F_{\tilde{w}}\) and \(F_w^{-1}=B_{\tilde{w}}\), where \(\tilde{w}=\{\frac{1}{w_n}\}_{n \in {\mathbb {Z}}}\).
Remark 2.8
Let \(w=\{w_n\}_{n \in {{\mathbb {Z}}}}\), be a weight sequence. Let T be the bilateral weighted shift \(F_w\) or \(B_w\) on \(X=\ell ^p(\mathbb {Z})\), \(1 \le p < \infty \), or \(X=c_0(\mathbb {Z})\). It will be useful for the sequel to note that, for every \(n \in {{\mathbb {N}}}\),
-
(1)
\(\Vert T^n \Vert = \sup _{k \in \mathbb {Z}} \vert w_kw_{k+1} \cdots w_{k+n-1}\vert \), and
-
(2)
if T is invertible, then \(\Vert T^{-n} \Vert = \sup _{k \in \mathbb {Z}} \vert w_kw_{k+1} \cdots w_{k+n-1}\vert ^{-1}\).
Proof
We only show (1) for \(T=F_w\), as the rest follows in a similar fashion. Given \(F_w\) on \(X=\ell ^p(\mathbb {Z})\), \(1 \le p < \infty \), then
and, on \(X=c_0(\mathbb {Z})\),
Hence
By computing \(F_w^n\) at
, the reverse of the above inequality is obtained. Hence
Moreover, if \(F_w\) is invertible, an analogous computation gives
\(\square \)
Remark 2.9
By replacing \(\mathbb {Z}\) by \(\mathbb {N}\) in (1) of Remark 2.8, an analogous argument gives the norms of \(F_{w}\) and \(B_{w}\) in the unilateral case.
Proposition 2.10
[20, Proposition 1.6.15]; [30, Theorem 4]; [3, Remark 35] Let \(X=\ell ^p({{\mathbb {N}}})\), \(1 \le p < \infty \), or \(X=c_0({{\mathbb {N}}}).\) Let T be the unilateral weighted shift \(F_w\) or \(B_w\), on X. Then, the spectrum of T is the disk
Proposition 2.11
[6, Theorem 2.1]; [30, Theorem 5]; [3, Remark 35] Let \(X=\ell ^p({{\mathbb {Z}}})\), \(1 \le p < \infty \), or \(X=c_0({{\mathbb {Z}}}).\) Let T be the bilateral weighted shift \(F_w\) or \(B_w\), on X. Then, the followings hold.
-
(a)
If T is non-invertible, then its spectrum is the disk
$$\begin{aligned}\sigma (T)=\{ \lambda \in {{\mathbb {C}}} : \vert \lambda \vert \le r(T)\}.\end{aligned}$$ -
(b)
If T is invertible, then its spectrum is the annulus
$$\begin{aligned}\sigma (T)=\left\{ \lambda \in {{\mathbb {C}}} : \frac{1}{r(T^{-1})} \le \vert \lambda \vert \le r(T)\right\} .\end{aligned}$$
The followings are well-known results which will be useful in the sequel.
Proposition 2.12
[18, Exercise 5.2.10]; [20, Proposition 1.6.14]; [30, Proposition 4] Let \(X=\ell ^p\), \(1 \le p < \infty \), or \(X=c_0.\) Let T be the unilateral (resp. bilateral) weighted shift \(F_w\) or \(B_w\). Then, T is compact if and only if \(\lim _{n \rightarrow \infty } w_n=0\) \((resp. \lim _{\vert n \vert \rightarrow \infty } w_n=0)\).
When the weights \(w=\{w_n\}_{n \in {A}}\), with \(A={\mathbb {N}}\) or \(A={\mathbb {Z}}\), are such that \(w_n =1\) for each \(n \in {A}\), then \(F_w\) and \(B_w\) reduce to the (unweighted) forward and backward shifts denoted by F and B, respectively. Clearly, \(\Vert F \Vert =\Vert B \Vert =1\), and for this simple case the spectrum and its part were completely analyzed, as partially summarized in the following result (for a detailed description of all the parts of the spectrum, see [31] and [19, Proposition 2.M], for the case \(p=2\); and [17, Corollary 3.2]; [20, Example 3.7.7]; [7, Proposition 10.2.8], for the case \(1 \le p < \infty \)).
Proposition 2.13
Let B and F be the unilateral backward and the unilateral forward shift, respectively. Then, the followings hold:
-
(1)
\(\sigma _p(F)=\emptyset \); \(\sigma _e(F)={{\mathbb {T}}}\);
-
(2)
\(\sigma _p(B)={{\mathbb {D}}};\) \(\sigma _e(B)={{\mathbb {T}}}\);
-
(3)
\(\sigma (F)=\sigma (B)=\{ \lambda \in {{\mathbb {C}}} : \vert \lambda \vert \le 1\}\).
If B and F denote the bilateral backward and the bilateral forward shift, respectively, then
-
(4)
\(\sigma _p(F)=\emptyset \); \(\sigma _e(F)={{\mathbb {T}}}\);
-
(5)
\(\sigma _p(B)=\emptyset \); \(\sigma _{e}(B)={{\mathbb {T}}}\);
-
(6)
\(\sigma (F)=\sigma (B) ={{\mathbb {T}}}\).
The definition of an adjoint operator is given, in general, for operators defined on a Hilbert space. If one tries to transfer this definition to operators on Banach spaces, it immediately appears an obstacle: the absence of an inner product. Hence, given a Banach space X, it is necessary to introduce the dual space \(X^*\), and to define a new product on \(X \times X^*\), \(\langle \cdot ,\cdot \rangle \), as in the following definition.
Definition 2.14
Let X be a Banach space and let \(X^*\) be its dual. Let \(x^* \in X^*\). Then, for each \(x \in X\), we define
Definition 2.15
(Adjoint operator) Let X be a Banach space and let \(T \in {{\mathcal {L}}}(X)\). The operator \(T^*\in {{\mathcal {L}}}(X^*)\) defined by \(T^*x^*=x^* \circ T\), that is
is the adjoint of T. When T is an operator on a Hilbert space X, then it is said to be unitary if \(T^*=T^{-1}\).
The following example will be useful in the sequel.
Example 2.16
Let \(A= {\mathbb {N}}\) or \(A={\mathbb {Z}}\). As it is well-known, the dual space of \(\ell ^p(A)\), \(1\le p < \infty \), is given by \(\ell ^q(A)\), where \(\frac{1}{p}+\frac{1}{q}=1\) for \(1< p < \infty \), and \(q=\infty \) for \(p=1\). The continuous linear functionals \(x^*\) on \(\ell ^p(A)\) are precisely the maps of the form
with \(y=\{y_n\}_{n \in A} \in \ell ^q(A)\).
Now, let \(A={\mathbb {Z}}\) and let \(F_w\) be a bilateral weighted forward shift on \( \ell ^p({{\mathbb {Z}}}) \), \(1\le p < \infty \). Then, according to Definition 2.15, its adjoint is a bilateral weighted backward shift on \( \ell ^q({{\mathbb {Z}}})\) given by
for each \(n \in {\mathbb {Z}}\) (see, for instance, [25]). In fact, given any \(x \in \ell ^p({{\mathbb {Z}}})\) and \(x^* \in \ell ^q({{\mathbb {Z}}}),\) it is
and hence \(F_w^*(e_n)=B_{\tilde{w}}(e_n)=\tilde{w}_ne_{n-1}=\overline{w_{n-1}}e_{n-1}\), for each \(n \in {\mathbb {Z}}\).
Analogously, if \(A={\mathbb {N}}\) and \(F_w\) is a unilateral weighted forward shift on \( \ell ^p({{\mathbb {N}}}) \), \(1\le p < \infty \), then its adjoint is the unilateral weighted backward shift on \( \ell ^q({{\mathbb {N}}})\) given by \(F_w^*(e_0)=0\) and \(F_w^*(e_n)=\overline{w_{n-1}}e_{n-1}\), for each \(n \ge 1\). Similarly, \(B_w^*(e_n)=\overline{w_{n+1}}e_{n+1}\), for each \(n \ge 1\).
2.3 Similarity of operators
The notion of similarity plays an important role in the theory of bounded linear operators (see [19, 20]).
Definition 2.17
(Similarity) Let X be a Banach space. Two operators \(T,S \in {{\mathcal {L}}}(X)\) are called similar if there exists an invertible operator \(W \in {{\mathcal {L}}}(X)\) such that \(S=W^{-1}TW\).
In Linear Dynamics, the word “similarity” is replaced by “conjugation”, a word originally borrowed from Topological Dynamics. A special similarity on inner product spaces, i.e., on Hilbert spaces, is given by the unitarily equivalence:
Definition 2.18
(Unitarily Equivalence) Let \({\mathcal {H}}\) be a Hilbert space. Two operators \(T,S \in {{\mathcal {L}}}({{\mathcal {H}}})\) are said to be unitarily equivalent if there exists a unitary operator \(U \in {{\mathcal {L}}}({{\mathcal {H}}})\) such that \(S=U^{-1}TU\).
Note that similarity is a less severe definition than the one of unitarily equivalence. The importance of similarity and unitarily equivalence follows from the fact that they preserve many properties of an operator. It is known that similarity preserves invariant subspaces, that is, if two operators are similar, and if one has a nontrivial invariant subspace, then so does the other [19, Proposition 1.J]. Moreover, among the invariants of similarity the most important are the spectrum and the spectral radius, as the following result shows:
Proposition 2.19
[19, Proposition 2.B] Similarity preserves the spectrum and its parts, and so it preserves the spectral radius also. That is, in particular, if X is a Banach space and \(T,S \in {{\mathcal {L}}}(X)\) are two similar operators, then:
-
(1)
\(\sigma _p(T)=\sigma _p(S)\);
-
(2)
\(\sigma (T)=\sigma (S)\);
-
(3)
\(r(T)=r(S)\).
Of course, if two operators on an Hilbert space are unitarily equivalent, then they are also similar. The converse, in general, only holds for normal operators [19, Proposition 3.I]. Hence, unitarily equivalent operators on Hilbert spaces have, in particular, the same spectrum and the same point spectrum [26, Theorem A.11]. Moreover, it is well-known that unitary equivalence preserves the operator norm [19, Proposition 2.B]. The following result about unitarily equivalent weighted shifts on \(\ell ^2\) is well-known.
Proposition 2.20
[8, Proposition 6.2]; [30, Corollary 1] Let \(A= {\mathbb {N}}\) or \(A={\mathbb {Z}}\) and \(X=\ell ^2(A)\). Let T be a weighted shift \(F_w\) (resp. \(B_w\)), with weights \(w=\{w_n\}_{n \in A}\). Then, T is unitarily equivalent to the weighted shift \(F_{\tilde{w}}\) (resp. \(B_{\tilde{w}}\)), where \(\tilde{w}=\{\vert w_n \vert \}_{n \in A}.\)
As the following proposition shows, the same result holds for weighted shifts on the Banach space \(\ell ^p\), \(1\le p < \infty \), if we consider similarity instead of unitarily equivalence.
Proposition 2.21
Let \(A= {\mathbb {N}}\) or \(A={\mathbb {Z}}\) and \(X=\ell ^p(A), 1\le p < \infty .\) Let T be a weighted shift \(F_w\) (resp. \(B_w\)), with weights \(w=\{w_n\}_{n \in A}\). Then, T is similar to the weighted shift \(F_{\tilde{w}}\) (resp. \(B_{\tilde{w}}\)), where \(\tilde{w}=\{\vert w_n \vert \}_{n \in A}.\)
Proof
The proof is showed only for \(T=F_w\), as small changes give it for \(T=B_w\). Hence, let \(T=F_w\). According to Definition 2.17, we need to find an invertible operator \(W \in {{\mathcal {L}}}(X)\) with \(F_{w}=W^{-1}F_{\tilde{w}}W,\) i.e. such that
In order to do that, let us consider the operator \(T:X \rightarrow X\) given by \(T(e_n)=\gamma _n e_n,\) where \(\gamma =\{\gamma _n\}_{n \in A}\) is a bounded sequence of scalars. Then T acts as a diagonal operator, namely, \(T(\{x_n\}_{n \in A})=\{\gamma _n x_n\}_{n \in A}\).
Note that, for each \(\{x_n\}_{n \in A} \in X\),
and
Therefore, \(TF_{w}=F_{\tilde{w}}T\) if and only if the sequence \(\gamma =\{\gamma _n\}_{n \in A}\) satisfies
Therefore, choosing \(\gamma _0=1\) and taking, for \(n>0\)
and, for \(n\le 0\)
we have that such a sequence \(\{\gamma _n\}_{n \in A}\) satisfies \((\clubsuit )\), that is \(TF_{w}=F_{\tilde{w}}T\).
Moreover, note that:
-
T is linear;
-
T is bounded since, for each n, \(\vert \gamma _n \vert =1\) (as, by hypothesis, \(\vert {w}_{n-1} \vert = \vert \tilde{w}_{n-1} \vert \)), and then \(\Vert T \Vert =\sup _{n \in A}\vert \gamma _n \vert =1\);
-
T is invertible, as \(\inf _{n \in A}\vert \gamma _n\vert =1>0\). In particular, \(T^{-1}(e_n)=\frac{1}{\gamma _n}e_n\).
Taking \(W=T\) we obtain an invertible operator \(W \in {{\mathcal {L}}}(X)\) such that \(WF_{w}=F_{\tilde{w}}W,\) and hence, by Definition 2.17, \(F_{w}\) and \(F_{\tilde{w}}\) are similar. \(\square \)
3 On the spectrum of weighted shifts
Up to now, we have not made assumptions about the sign of the weights. It is well-known (see [8, p. 54]; [30, p. 56]) that if a weighted shift has a finite number of zero weights, then it is the direct sum of a finite number of finite dimensional operators and a weighted shift with nonzero weights. Hence, weighted shifts with a finite number of zero weights lead back to weighted shifts with all weights non-zero. For this reason, and by Proposition 2.20 and Proposition 2.21, from now on we only consider positive weights \(\{w_n\}_{n \in A}\).
In the previous section, Proposition 2.11 describes the spectrum of a weighted shifts with \(w=\{w_n\}_{n \in A}\), \(A={{\mathbb {N}}}\) or \(A={{\mathbb {Z}}}\), a bounded sequence of scalars. In this section, we will see more detailed results on the spectrum of \(F_w\) and \(B_w\), both in the unilateral case and in the bilateral case.
3.1 The unilateral case
Proposition 3.1
Let T be a unilateral weighted shift \(F_w\) or \(B_w\) on \(\ell ^p({{\mathbb {N}}})\), \(1\le p < \infty \), with \(\{w_n\}_{n \in {\mathbb {N}}}\) a bounded sequence of positive reals. Let
Then, the spectral radius of T is
Proof
The proof is showed for \(T=F_w\), as small changes provide the case \(T=B_w\). Hence, consider \(T=F_w\). Let \(\epsilon >0\) and \(\overline{n} \in {\mathbb {N}}\) such that \(\vert w_n - w \vert < \epsilon \) for each \(n\ge \overline{n}\). Then, in particular, for each \(n> \overline{n}\),
Therefore, for each \(k \in {\mathbb {N}}\) and for each \(n> \overline{n}\),
and, hence,
Then, it follows that
If \(w=0\), then, as \(\epsilon \) is arbitrary, it follows that \(r(F_w)=0=w.\)
Now, assume that \(w>0\). From the previous computation we obtain that \(r(F_w)\le w\). We want to show that \(r(F_w)\ge w.\) Let F denote the unilateral forward shift on \(\ell ^p({{\mathbb {N}}})\), \(1\le p < \infty \), i.e.
Consider the operator \(F_{\widetilde{w}}=F_w-wF\) defined by:
Hence, \(F_{\widetilde{w}}\) is the unilateral weighted forward shift on \(\ell ^p({{\mathbb {N}}})\), \(1\le p < \infty \), with weights \(\widetilde{w}=\{w_n-w\}_{n \in {\mathbb {N}}}\). In our case, the hypothesis
implies that
which means, by Proposition 2.12, that \(F_{\widetilde{w}}\) is compact. Therefore, from Proposition 2.4 it follows that:
As \(\sigma _e(F_w)\) is a closed subset of \(\sigma (F_w)\), then it follows
Hence, the spectral radius of \(F_w\) is \(r(F_w)=w\). \(\square \)
Remark 3.2
Let \(X=\ell ^p({{\mathbb {N}}}) \), \(1\le p < \infty \), and let T denote a unilateral weighted shift \(F_w\) or \(B_w\), on X, with \(\{w_n\}_{n \in {\mathbb {N}}}\) a positive weight sequence. Let
Then, by (1) of Remark 2.8 and Proposition 3.1, together with the spectral radius formula, we get
Take \(w_{n} = 3^{{(-1)}^{n}}\). Then
but, clearly, \(\{w_{n}\}_{n \in {\mathbb {N}}}\) is not regular, with \(\varlimsup _{n\rightarrow \infty } w_{n} = 3\) and \(\varliminf _{n\rightarrow \infty } w_{n} = \frac{1}{3}\). In particular, Proposition 3.1 does not hold anymore if we replace \(\lim \) with \(\varlimsup \).
We point out that, in the Hilbert case \(p=2\), Proposition 3.1 is a consequence of [30, Proposition 15] and [8, Proposition 6.8 (a)]. As in [16, Remark 1.2] and [30, Theorem 8] for the Hilbert case \(p=2\), the point spectrum of a unilateral forward weighted shift on \(\ell ^p({{\mathbb {N}}})\), \(1\le p < \infty \), turns out to be empty, and the point spectrum of a unilateral backward weighted shift on \(\ell ^p({{\mathbb {N}}})\), \(1\le p < \infty \), satisfies
where \(\tilde{r} = \underline{\mathrm{lim}}_{n \rightarrow \infty } {\left( w_{2} \cdots w_{n+1}\right) }^{\frac{1}{n}}.\) The proofs for the general case \(1\le p < \infty \) are analogous to the ones for \(p=2\).
Proposition 3.3
Let \(F_w: \ell ^p({{\mathbb {N}}}) \rightarrow \ell ^p({{\mathbb {N}}}) \), \(1\le p < \infty \), be a unilateral weighted forward shift with \(\{w_n\}_{n \in {\mathbb {N}}}\) a bounded positive weight sequence. Then \(\sigma _p(F_w)=\emptyset \).
Proof
By contradiction, let \(\lambda \in \sigma _p(F_w)\). Then, there exists \(\{ x_n\}_{n \in {\mathbb {N}}} \in \ell ^p({\mathbb {N}})\setminus \{{0}\}\) eigenvector of \(F_w\) corresponding to the eigenvalue \(\lambda \), i.e., \(F_w(\{x_n\}_{n \in {\mathbb {N}}})=\{\lambda x_n\}_{n \in {\mathbb {N}}}\). By definition, \(F_w(\{x_n\}_{n \in {\mathbb {N}}})=\{w_{n-1}x_{n-1}\}_{n \in {\mathbb {N}}}\) and, hence, the coordinates of \(\{ x_n\}_{n \in {\mathbb {N}}}\) are such that
By hypothesis, \(w_n>0\) for each \(n \in {{\mathbb {N}}}\), therefore \(F_w\) is injective and then \(\lambda \ne 0\). Hence, it follows from \((\star )\) that \(x_n=0\) for each \(n \in {{\mathbb {N}}}\). This is a contradiction and so it must be \(\sigma _p(F_w)=\emptyset \). \(\square \)
Proposition 3.4
Let \(B_w: \ell ^p({{\mathbb {N}}}) \rightarrow \ell ^p({{\mathbb {N}}}) \), \(1\le p < \infty \), be a unilateral weighted forward shift with \(\{w_n\}_{n \in {\mathbb {N}}}\) a bounded positive weight sequence. Then
where \(\tilde{r} = \underline{\mathrm{lim}}_{n \rightarrow \infty } {\left( w_{2} \cdots w_{n+1}\right) }^{\frac{1}{n}}.\)
Proof
Let \(\lambda \in \sigma _p(B_w)\). Then, there exists \(\{ x_n\}_{n \in {\mathbb {N}}} \in \ell ^p({\mathbb {N}})\setminus \{{0}\}\) eigenvector of \(B_w\) corresponding to the eigenvalue \(\lambda \), i.e., \(B_w(\{x_n\}_{n \in {\mathbb {N}}})=\{\lambda x_n\}_{n \in {\mathbb {N}}}\). By definition, \(B_w(\{x_n\}_{n \in {\mathbb {N}}})=\{w_{n+1}x_{n+1}\}_{n \in {\mathbb {N}}}\) and, hence, the coordinates of \(\{ x_n\}_{n \in {\mathbb {N}}}\) are such that
Therefore, assuming without loss of generality \(x_{1}=1\), for each \(n \in {{\mathbb {N}}}\), we have
Then
By applying the Cauchy-Hadamard criterion, as
we obtain the conclusion. \(\square \)
3.2 The bilateral case
In this section we focus on bilateral weighted shifts. We first generalize a result on the point spectrum proved, for the Hilbert case \(p=2\), in [8, Proposition 6.8 (b)] and [30, Theorem 9], to \(1 \le p < \infty \). Then, under some regularity hypotheses on the weight sequence, we deduce the spectrum.
Proposition 3.5
Let \(\{w_n\}_{n \in {\mathbb {Z}}}\) be a bounded positive weight sequence. Assume that
Let \(F_w\) and \(B_w\) be the bilateral weighted forward shift and the bilateral weighted backward shift, respectively, on \(\ell ^p({{\mathbb {Z}}}) \), \(1\le p < \infty \). Then the followings hold:
-
(1)
If \(w_+ \le w_-\), then \(\{ \lambda : w_+< \vert \lambda \vert < w_-\} \subseteq \sigma _p(F_w) \subseteq \{ \lambda : w_+ \le \vert \lambda \vert \le w_-\}\).
-
(2)
If \(w_- < w_+\), then \(\sigma _p(F_w)=\emptyset \).
-
(3)
If \(w_- \le w_+\), then \(\{ \lambda : w_-< \vert \lambda \vert < w_+\} \subseteq \sigma _p(B_w) \subseteq \{ \lambda : w_- \le \vert \lambda \vert \le w_+\}\).
-
(4)
If \(w_+ < w_-\), then \(\sigma _p(B_w)=\emptyset \).
Proof
(1). We separate the two cases, Case 1.A: \(w_+ >0\) and Case 1.B: \(w_+=0\).
Case 1.A Assume \(w_+ >0\). We start by showing that \(\sigma _p(F_w) \subseteq \{ \lambda : w_+ \le \vert \lambda \vert \le w_-\}\). Let \(\lambda \in \sigma _p (F_w)\). Let \(x=\{ x_n\}_{n \in {\mathbb {Z}}}\) be an eigenvector corresponding to \(\lambda \), i.e. \(F_w(\{x_n\}_{n \in {\mathbb {Z}}})=\{\lambda x_n\}_{n \in {\mathbb {Z}}}\). By definition, \(F_w(\{x_n\}_{n \in {\mathbb {Z}}})=\{w_{n-1}x_{n-1}\}_{n \in {\mathbb {Z}}}\) and hence the coordinates of \(x=\{ x_n\}_{n \in {\mathbb {Z}}}\) are such that
Therefore, for each \(n \in {\mathbb {N}}\), we have
Then
By contradiction, assume \(\vert \lambda \vert < w_+\). Choose \(\epsilon >0 \) such that \(\vert \lambda \vert < w_+ - \epsilon \). As \(\lim _{n \rightarrow \infty } w_n=w_+\), let \(\overline{n}\) be so large that \(w_n > w_+ - \epsilon \) for each \(n\ge \overline{n}\). Then, for each \(n> \overline{n}\),
Therefore
As \(\vert \lambda \vert < w_+ - \epsilon \), then the geometric series on the right diverges: this is a contradiction. Hence, it must be \(\vert \lambda \vert \ge w_+\).
Analogously, by contradiction, assume \(\vert \lambda \vert > w_-\). Choose \(\epsilon >0\) such that \(\vert \lambda \vert > w_- + \epsilon \). As \(\lim _{n \rightarrow \infty } w_{-n}=w_-\), let \(\tilde{n}\) be so large that \(w_{-n} < w_- + \epsilon \) for each \(n\ge \tilde{n}\). Then, for each \(n> \tilde{n}\),
Therefore
As \(\vert \lambda \vert > w_- + \epsilon \), then the geometric series on the right diverges. This is a contradiction. So, it must be \(\vert \lambda \vert \le w_-\). Hence, we have just proved that, in the case \(w_+>0\), it is
Now, we prove the first inclusion in (1). That is, we show that if \(\lambda \) is such that \(w_+< \vert \lambda \vert < w_-\), then \(\lambda \) is an eigenvalue of the vector \({x}=\{x_n\}_{n \in {\mathbb {Z}}}\) defined by choosing \(x_0 \ne 0\) and \(w_{n-1} x_{n-1}=\lambda x_{n}\) for each \(n \in {{\mathbb {Z}}}\).
Indeed, choose \(\epsilon >0\) such that \(w_+ + \epsilon< \vert \lambda \vert < w_- - \epsilon \).
Let \(\overline{n}\) be so large that \(w_- - \epsilon < w_{- n}\) and \(w_{n} < w_+ + \epsilon \) for each \(n\ge \overline{n}\). Then, for each \(n> \overline{n}\),
and
Therefore
and
It follows from \((\spadesuit )\) that \(\Vert x \Vert _p < \infty \). We conclude that the vector \({x}=\{x_n\}_{n \in {\mathbb {Z}}}\) is such that
-
\(w_{n-1} x_{n-1}=\lambda x_{n}\) for each \(n \in {{\mathbb {Z}}}\);
-
\({x} \in \ell ^p({{\mathbb {Z}}})\setminus \{0\};\)
i.e. it is an eigenvector of \(\lambda \). Hence
Case 1.B Assume \(w_+=0.\) We distinguish the cases \(w_- \ne 0\) and \(w_-=0\). If \(w_- \ne 0\), and if \(\lambda \in \sigma _p(F_w)\), from the previous computation and the fact that \(\vert \lambda \vert \ge 0=w_+\), it follows that \(\sigma _p(F_w) \subseteq \{ \lambda : 0 \le \vert \lambda \vert \le w_-\}\). Moreover, a similar computation shows that \(\{ \lambda : 0< \vert \lambda \vert < w_-\} \subseteq \sigma _p(F_w)\).
If \(w_-=0\), note that \(\{ \lambda : w_+< \vert \lambda \vert < w_-\} = \emptyset \subseteq \sigma _p(F_w).\) Moreover, if \(\lambda \in \sigma _p(F_w)\), a computation similar to above shows that \(\vert \lambda \vert =0\), i.e. \(\sigma _p(F_w) \subseteq \{0\}.\)
For the proof of (2), just note that if \(w_- < w_+\), then, arguing as above, we obtain \(\sigma _p(F_w)=\emptyset \).
(3) Assume \(w_- \le w_+\). As in (1), \(\lambda \in \sigma _p(B_w)\) if and only if there exists \(\{x_n\}_{n \in {\mathbb {Z}}} \in \ell ^{p}({\mathbb {Z}}) \setminus \{0\}\) such that \(B_w(\{x_n\}_{n \in {\mathbb {Z}}})=\{\lambda x_n\}_{n \in {\mathbb {Z}}}\). On the other hand, \(B_w(\{x_n\}_{n \in {\mathbb {Z}}})=\{w_{n+1}x_{n+1}\}_{n \in {\mathbb {Z}}}\). Hence, the components of \(x=\{ x_n\}_{n \in {\mathbb {Z}}}\) are such that
Therefore, for each \(n \in {\mathbb {N}}\) we have:
We want to determine \(\lambda \) such that \(\Vert x \Vert _p<\infty \). Arguing as in (1), with a similar computation we obtain that, if \(w_- \le w_+\) then
The implication (4) follows noting that if \(w_- > w_+\), then, by arguing as in (3), we obtain \(\sigma _p(B_w)=\emptyset \). \(\square \)
Proposition 3.6
Let \(\{w_n\}_{n \in {\mathbb {Z}}}\) be a bounded positive weight sequence. Let
Let T be the bilateral weighted shift \(F_w\) or \(B_w\), on \(X=\ell ^p({{\mathbb {Z}}}) \), \(1\le p < \infty \). Then the followings hold:
-
(a)
If T is invertible, then
$$\begin{aligned} \sigma (T)=\{ \lambda : \min \{w_-,w_+\} \le \vert \lambda \vert \le \max \{w_-,w_+\}\}.\end{aligned}$$ -
(b)
If T is not invertible, then
$$\begin{aligned}\sigma (T)=\{ \lambda : \vert \lambda \vert \le \max \{w_-,w_+\} \}.\end{aligned}$$
Proof
By using Proposition 2.11, as \(\sigma (F_w)=\sigma (B_w)\), hence without loss of generality we can consider \(T=F_w\). We may assume that \(\min \{w_-,w_+\} \ne 0,\) otherwise, by Proposition 2.12, \(F_w\) is compact and hence every nonzero \(\lambda \in \sigma (F_w)\) is an eigenvalue of \(F_w\), i.e., \(\sigma (F_w)=\{0\}\cup \sigma _p(F_w)\) [9, Proposition 7.1].
Hence, let \(\min \{w_-,w_+\} \ne 0.\) Note that, by the hypotheses, for each \(\epsilon >0\), there exist \(\overline{n}, \overline{\overline{n}} \in {{\mathbb {N}}}\) such that
and
Let \(N=\max \{ \overline{n}, \overline{\overline{n}} \}\). Note that
Hence,
Analogously,
Hence,
From the above computations it follows, by the arbitrariness of \(\epsilon >0\), that \(r(F_w) \le \max \{ w_+ , w_- \}\) and \(r(F_w^{-1})^{-1} \ge \min \{ w_+ , w_- \}\).
Hence, by using Proposition 2.11, one can get that
-
if \(F_w\) is invertible, then
$$\begin{aligned} \sigma (F_w)= \{ \lambda : r(F_w^{-1})^{-1} \le \vert \lambda \vert \le r(F_w) \} \subseteq \{ \lambda : \min \{w_-,w_+\} \le \vert \lambda \vert \le \max \{w_-,w_+\}\},\end{aligned}$$ -
if \(F_w\) is not invertible, then
$$\begin{aligned} \sigma (F_w)= \{ \lambda : \vert \lambda \vert \le r(F_w) \} \subseteq \{ \lambda : \vert \lambda \vert \le \max \{w_-,w_+\} \}.\end{aligned}$$
As \(\sigma (F_w)\) (\(\sigma (B_w)\)) is compact and \(\sigma _p(F_w) \subseteq \sigma (F_w)\) (\(\sigma _p(B_w) \subseteq \sigma (B_w)\)), then the conclusion follows by applying statements (1) and (3) of Proposition 3.5. \(\square \)
Example 3.7
As an application of the last proposition, consider the sequence \(\{w_{n}\}_{n \in {{\mathbb {Z}}}}\) defined as:
Let \(T= F_{w}\) or \(B_w\). Then, clearly, T is invertible and \(\max \{w_-,w_+\}=3\), \(\min \{w_-,w_+\}=\frac{1}{2}\). Hence, by Proposition 3.6,
We conclude with the following remark.
Remark 3.8
Let \(X=\ell ^p({{\mathbb {Z}}}) \), \(1\le p < \infty \), and let T denote a bilateral weighted shift \(F_w\) or \(B_w\), on X, with \(\{w_n\}_{n \in {\mathbb {Z}}}\) a bounded positive weight sequence. Let
and
Note that the Proposition 3.6 cannot be improved by replacing \(\min \{w_-,w_+\}\) with \(\min \{w_-^-,w_+^-\}\) and \(\max \{w_-,w_+\}\) with \(\max \{w_-^+,w_+^+\}\).
To see this, take for example
Then, by Remark 2.8 and by the spectral radius formula,
Clearly, \(\{w_{-n}\}_{n \in {\mathbb {N}}}\) and \(\{w_n\}_{n \in {\mathbb {N}}}\) are not regular, with \(w_-^-=\frac{1}{2}\), \(w_-^+=2\), \(w_+^-=\frac{1}{3}\), \(w_+^+=3\), and hence \(\max \{w_-^+,w_+^+\}=3\) and \(\min \{w_+^-,w_-^-\}=\frac{1}{3}.\)
References
Bayart, F., Matheron, É.: Dynamics of Linear Operators, vol. 179. Cambridge University Press, Cambridge (2009)
Bermúdez, T., Bonilla, A., Martínez-Giménez, F., Peris, A.: Li–Yorke and distributionally chaotic operators. J. Math. Anal. Appl. 373, 83–93 (2011)
Bernardes, N.C., Jr., Cirilo, P.R., Darji, U.B., Messaoudi, A., Pujals, E.R.: Expansivity and shadowing in linear dynamics. J. Math. Anal. Appl. 461, 796–816 (2018)
Bernardes, N.C., Jr., Darji, U.B., Pires, B.: Li–Yorke chaos for composition operators on \(L^p\)-spaces. Monatsh. Math. 191, 13–35 (2020)
Birkhoff, G.D.: Démonstration d’un théorème élémentaire sur les fonctions entières. C. R. Acad. Sci. Paris Ser. I Math. 189, 473–475 (1929)
Bourhim, A.: Spectrum of bilateral shifts with operator-valued weights. Proc. Am. Math. Soc. 134, 2131–2137 (2006)
Cheng, R., Mashreghi, J., Ross, W.T.: Function Theory and \(\ell ^p\) Spaces, vol. 75. American Mathematical Society, Washington (2020)
Conway, J.B.: The Theory of Subnormal Operator. American Mathematical Society, Washington (1991)
Conway, J.B.: A Course in Functional Analysis, vol. 96. Springer, New York (2019)
D’Aniello, E., Darji, U.B., Maiuriello, M.: Generalized hyperbolicity and shadowing in \(L^p\) spaces. J. Differ. Equ. 298, 68–94 (2021)
D’Aniello, E., Darji, U.B., Maiuriello, M.: Shift-like operators on \(L^p(X)\). J. Math. Anal. Appl. 515, 126393, 13 (2022)
Darji, U.B., Pires, B.: Chaos and frequent hypercyclicity for composition operators. Proc. Edinb. Math. Soc. (2) 64, 513–531 (2021)
Dowson, H.R.: Spectral Theory of Linear Operators, vol. 12 of London Mathematical Society Monographs. Academic Press, Inc. [Harcourt Brace Jovanovich, Publishers], London, New York (1978)
Edmunds, D., Evans, W.: Spectral Theory and Differential Operators. Clarendon Press, Oxford (1987)
Eisenberg, M., Hedlund, J.H.: Expansive automorphisms of Banach spaces. Pacific J. Math. 34, 647–656 (1970)
Faour, L., Khalil, R.: On the spectrum of weighted operator shifts. Houston J. Math. 13, 549–556 (1987)
Gohberg, S.G.: Israel and M. Kaashoek, Basic Classes of Linear Operators. Birkhäuser, Switzerland (2012)
Grosse-Erdmann, K.-G., Peris Manguillot, A.: Linear Chaos. Universitext. Springer, London (2011)
Kubrusly, C.S.: Spectral Theory of Bounded Linear Operators. Birkhäuser, Switzerland (2020)
Laursen, K.B., Neumann, M.M.: An Introduction to Local Spectral Theory, vol. 20 of London Mathematical Society Monographs. New Series. The Clarendon Press, Oxford University Press, New York (2000)
Lay, D.C.: Spectral properties of generalized inverses of linear operators. SIAM J. Appl. Math. 29, 103–109 (1975)
MacLane, G.R.: Sequences of derivatives and normal families. J. Anal. Math. 2, 72–87 (1952)
Maiuriello, M.: Dynamics of Linear Operators. Aracne Editore, Rome (2022). ISBN: 9791221801316
Mazur, M.: Hyperbolicity, expansivity and shadowing for the class of normal operators. Funct. Differ. Equ. 7, 147–156 (2000)
Miller, T.L., Miller, V.G., Neumann, M.M.: Local spectral properties of weighted shifts. J. Oper. Theory 51(1), 71–88 (2004)
Mukherjea, A., Pothoven, K.: Real and Functional Analysis. Mathematical Concepts and Methods in Science and Engineering, vol. 6. Plenum Press, New York, London (1978)
Rolewicz, S.: On orbits of elements. Stud. Math. 32, 17–22 (1969)
Rudin, W.: Functional Analysis, 2nd edn. McGraw-Hill Book Co., New York (1991)
Salas, H.N.: Hypercyclic weighted shifts. Trans. Am. Math. Soc. 347, 993–1004 (1995)
Shields, A.L.: Weighted shift operators and analytic function theory. Top. Oper. Theory (C. Pearcy Ed.) 13, 49–128 (1974)
Sudo, T.: The spectrum theory for weighted shift operators and their \(C^*\)-algebras. Bull. Fac. Sci. Univ. Ryukyus 1–31 (2018)
Acknowledgements
This research has been partially supported by the “Gruppo Nazionale per l’Analisi Matematica, la Probabilità e le loro Applicazioni dell’Istituto Nazionale di Alta Matematica F. Severi” and by the project Vain-Hopes within the program Valere of Università degli Studi della Campania “Luigi Vanvitelli”. It has also been partially accomplished within the UMI Group TAA “Approximation Theory and Applications”.
Funding
Open access funding provided by Università degli Studi della Campania Luigi Vanvitelli within the CRUI-CARE Agreement.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access 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/.
About this article
Cite this article
D’Aniello, E., Maiuriello, M. On the spectrum of weighted shifts. Rev. Real Acad. Cienc. Exactas Fis. Nat. Ser. A-Mat. 117, 4 (2023). https://doi.org/10.1007/s13398-022-01328-z
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s13398-022-01328-z