The essential spectrum, norm, and spectral radius of
abstract multiplication operators.

Over the past two decades numerous authors have studied boundedness, invertibility and compactness of multiplication operators on a variety of spaces of measurable functions on a measure space $(X,\mu)$ (or sequence spaces, when the measure is atomic). We refer to a recent preprint of J. Voigt ([7]) for some references. It seems that these authors are not aware that these multiplication operators are examples of order bounded band preserving operators (or orthomorphisms). Zaanen showed in 1977 in [9] already that the collection of all such orthomorphisms is lattice isometric to $L^{\infty}(X,\mu)$, so that a function defining a multiplication operator has to be essentially bounded and the operator norm of the multiplication operator equals the $L^{\infty}$-norm. Around that time it was also established that for an arbitrary Banach lattice $E$ the collection of band preserving operators was equal to the center $Z(E)$, where $Z(E)=\{T\in\mathcal{L}(E):|T|\leq\lambda I\mbox{ for some }\lambda\}$ and that $\|T\|=\inf\{\lambda:|T|\leq\lambda I\}$ for $T\in Z(E)$. For this and additional basic properties of the center we refer to the books [4] and [10]. In particular, Proposition 3.1.3 of [4] shows that for a Dedekind complete Banach lattice the center as defined here coincides with the collection of regular operators which commute with band projections. It was proved in [5] that $Z(E)$ is a full sub-algebra of the algebra $\mathcal{L}(E)$ of norm bounded operators, i.e., $T\in Z(E)$ is invertible in $\mathcal{L}(E)$ if and only if there exists $c>0$ such that $|T|\geq cI$. Therefore $\sigma(T)$ is equal to the essential range of $T$, when $T$ is a multiplication operator on a Banach function space. Moreover the spectral radius $r(T)=\|T\|$. The study of the essential spectrum of operators $T\in Z(E)$ was initiated in [5], where it was proved that $\sigma_{e}(T)=\sigma(T)$ for $T\in Z(E)$ with $E$ a non-atomic Banach lattice. Here the essential spectrum $\sigma_{e}(T)$ denotes the spectrum of $T+\mathcal{K}(E)$ in the Calkin algebra $\mathcal{L}(E)/\mathcal{K}(E)$. A few years later in [1] other spectra and essential spectra of operators $T\in Z(E)$ were considered and an alternative proof of $\sigma_{e}(T)=\sigma(T)$ for $T\in Z(E)$ with $E$ a non-atomic Banach lattice was given. A description of compact $T\in Z(E)$ can be derived from the more general description of compact disjointness preserving operators due to De Pagter (unpublished) and Wickstead [8].

More recently the essential norm $\|T\|_{e}$ in the Calkin algebra has been considered for certain concrete Banach sequence spaces and $L_{p}$-spaces (see [7] and [2]) for more references. The current paper was prompted by the preprint [7], which didn’t use our results from [5]. By using the term abstract multiplication operator in our title we hope that future duplication of our results by authors working on these questions for concrete function spaces will be avoided by a web search or MathScinet search. The current paper describes completely the essential spectrum and norm of an operator $T\in Z(E)$ for an arbitrary Banach lattice $E$. The paper is organized as follows. In section 2 we describe for non-atomic Banach lattices the essential spectrum and norm for an operator $T\in Z(E)$. To be able to describe the essential spectral radius for the atomic case where we might have uncountably many atoms, we describe in section 3 what we call Fréchet cluster points of a complex valued function and the Fréchet limit superior and inferior of the modulus of such a function. For the countable case this is standard textbook material, but for the uncountable case one can find some of this in General Topology textbooks, but for the convenience of the reader we have included this section. Then in section 4 the atomic case is handled. In the process of proving the results we reprove the description of a compact $T\in Z(E)$. In the final section we combine the atomic and non-atomic case to describe the essential spectrum and norm of $T\in Z(E)$ for an arbitrary Banach lattice $E$. As in [7] we associate in that case with $T\in Z(E)$ an atomic part $T_{A}$ and non-atomic part $T_{A^{d}}$, but in general (i.e., when $E$ is not Dedekind complete) we don’t have a decomposition of $T$, as these parts are not defined on all of $E$. We give in this section a concrete example to illustrate this. We conclude the paper by applying our results to prove that if $T\in Z(E)$ is essentially quasi-nilpotent, then $T$ is compact (and thus $T+\mathcal{K}(E)=0$ in the Calkin algebra).

Let $E$ be a complex Banach lattice. Let $A\subset E$. Then the disjoint complement $A^{d}$ of $A$ is defined as $A^{d}=\{x\in E:|x|\wedge|a|=0\mbox{ for all }a\in A\}$. Then $A^{dd}$ is defined as $(A^{d})^{d}$. Recall that $a\in E$ is called an atom, if $|x|\leq|a|$ implies that $x=\lambda a$ for some scalar $\lambda$, i.e., $\{a\}^{dd}=\{\lambda a:\lambda\in\mathbb{C}\}$. Now $E$ is called non-atomic if $E$ doesn’t contain any atoms. Typical examples of non-atomic Banach lattices are Banach function spaces over a non-atomic $\sigma$-finite measure space. The following theorem was proved by the author in 1980 in [5] (Theorem 1.11).

Using the methods of the above mentioned paper we can now prove the following theorem. The proof of this theorem, as well as the following theorem, depend on Lemma 1.10 of [5]. As one of the referees correctly pointed out, this lemma is incorrect as stated. Fortunately the fix is easy. In Lemma 1.9 of [5] one needs to replace $0\leq u_{n}\leq u$ by $|u_{n}|\leq u$.Then the proof of Lemma 1.9 of [5] is correct. Note that in the final step the Eberlein-Smulyan theorem is used. With this correction, Lemma 1.10 of [5] is true if we replace $0\leq u_{n}\in B$ by $u_{n}\in B$ in the statement.

Let $A$ be a non-empty set and $f:A\to\mathbb{C}$ a function. Denote by $\mathcal{F}$ the Fréchet filter on $A$, i.e., a subset $B$ of $A$ is in $\mathcal{F}$ if $B^{c}$ is a finite set.

In this case we will also say that $z$ is an $\mathcal{F}$-cluster point. Note that one can equivalently say that $z$ is a Fréchet cluster point of $f$ if and only if $z$ is a cluster point of $f(\mathcal{F})$ in the topological sense if and only if there exist distinct $a_{n}\in A$ ($n=1,2,\cdots$) such that $f(a_{n})\to z$. One can show (assuming the axiom of choice) the following proposition.

It is also known that the set of all $\mathcal{F}$-cluster points of $f$ is a closed subset of $\mathcal{C}$. Therefore we have the following proposition.

We now describe the modulus this largest and smallest $\mathcal{F}$-cluster point.

Note that these notions can be extended to unbounded function by allowing $+\infty$ or $-\infty$ as values. As in the sequential case one can prove now the following theorem.

We indicate in the next section a case, where $\limsup_{\mathcal{F}}|f|$ naturally occurs as the quotient norm of $\ell_{\infty}(A)$ by $c_{0}(A)$.

Let $E$ be an infinite dimensional (complex) Banach lattice throughout this section. Recall that $a\in E$ is called an atom, if the band generated by $a$ is one dimensional, i.e., $\{a\}^{dd}=\{\lambda a:\lambda\in\mathbb{C}\}$. Denote by $A$ the set of all positive atoms in $E$ of norm one. Then $E$ is called an atomic Banach lattice if the band $A^{dd}$ generated by $A$ equals $E$. If we denote by $P_{a}$ the band projection from $E$ onto $\{a\}^{dd}$, then every $x\in E$ can be written as an order convergent series

Note that if $A$ is countable, then we have the usual Banach sequence spaces. Now every $T\in Z(E)$ can be represented as a multiplication operator, where for each $a\in A$ there exists $\lambda_{a}\in\mathbb{C}$ such that $Ta=\lambda_{a}a$. Then, if $x=\sum_{a\in A}P_{a}x$, we have $Tx=\sum_{a\in A}\lambda_{a}P_{a}x$. The following lemma is well-known in the countable case. We include for convenience of the reader the similar proof for the uncountable case.

We now show that the expression $\limsup_{\mathcal{F}}|\lambda_{a}|$ is actually a quotient norm. Let $\ell_{\infty}(A)$ denote the Banach lattice of all bounded functions $f:A\to\mathbb{C}$ with the supremum norm and $c_{0}(A)$ the closed ideal of all $f\in\ell_{\infty}(A)$ with $\limsup_{\mathcal{F}}|\lambda_{a}|=0$. Then

Let $E$ be a (complex) Banach lattice. Denote by $A$ the set of all positive atoms in $E$ of norm one. Then the atomic part of $E$ is the band $A^{dd}$ generated by $A$ and will be denoted by $E_{A}$. The disjoint complement $A^{d}$ of $A$ will be called the non-atomic part (or continuous part) of $E$ and will be denoted by $E_{A^{d}}$. Below we shall see by means of an example that these bands in general don’t need to be projection bands (but are so of course when $E$ is Dedekind complete). Therefore we have in general that $E_{A}\oplus E_{A^{d}}$ is only a norm closed, order dense ideal in $E$. We denote the restriction of $T\in Z(E)$ to $E_{A}$ by $T_{A}$ (the atomic part of $T$) and to $E_{A^{d}}$ by $T_{A^{d}}$ (the non-atomic or continuous part of $T$). As $T$ is band preserving we have $T_{A}\in Z(E_{A})$ and $T_{A^{d}}\in Z(E_{A^{d}})$, but in general $T$ is not the direct sum of these two operators, as $T_{A}\oplus T_{A^{d}}$ is only defined on $E_{A}\oplus E_{A^{d}}$. To overcome this technical difficulty, we use the following lemma.

We conclude with an example of a Banach lattice $E$ for which the atomic part $E_{A}$ is not a projection band.

As the above example shows that in general, if $T\in Z(E)$, then $T$ is not the sum of $T_{1}$ and $T_{2}$, where $T_{A}=(T_{1})_{A}$ and $T_{A^{d}}=(T_{2})_{A^{d}}$.However that is so if only if $(T_{p})_{A}$ is compact. We now show that this is always the case for the if part of the statement.

In [3] it was observed that if $\sigma_{e}(A+B)=\sigma_{e}(A)$ for all bounded self-adjoint operators $A$ on a Hilbert space, for a given self-adjoint $B$, then $B$ is compact. The following theorem is an order analogue of this.