This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Semigroups of weighted composition operators on spaces of holomorphic functions

I. Chalendar and J.R. Partington

Isabelle Chalendar, Université Gustave Eiffel, LAMA, (UMR 8050), UPEM, UPEC, CNRS, F-77454, Marne-la-Vallée, France.
[email protected]

Jonathan R. Partington, School of Mathematics, University of Leeds, Leeds LS2 9JT, UK. [email protected]

Mathematics Subject Classification (2020): 30H10, 30H20, 30D05, 47B33, 47D06

1 Introduction

This paper is based on three hours of lectures given by the first author in the “Focus Program on Analytic Function Spaces and their Applications” July 1 – December 31, 2021, organized by the Fields Institute for Research in Mathematical Sciences.

The goal of this paper is to give an introduction to the properties of discrete and continuous C0C_{0}-semigroups of (weighted) composition operators on various spaces of analytic functions.

To that aim we detail the structure of semiflows of analytic functions on the open unit disc 𝔻\mathbb{D} and their generators, which provides information on the generators of continuous semigroups of composition operators on Banach spaces XX embedding in Hol(𝔻)\mathop{\rm Hol}\nolimits(\mathbb{D}), the Fréchet space of holomorphic functions on 𝔻\mathbb{D}.

An initial motivation for studying such semigroups is a better understanding of an specific universal operator. Moreover, adding a weight to a composition operator is motivated by the fact that such operators describe isometries on non-Hilbertian Hardy spaces and they appear automatically when the Banach spaces XX are replaced by Banach spaces of holomorphic functions on another domain such as the right half-plane.

Thanks to the analysis of spectral properties, we deduce the asymptotic behaviour of discrete semigroups of composition operators on various Banach spaces such as the Hardy spaces, the weighted Bergman spaces, Bloch type spaces or standard weighted Bergman space of infinite order. As a byproduct we obtain characterization of the properties of isometry and similarity to isometry, still for composition operators. We then describe the limit at infinity for continuous semigroups of composition operators.

Compactness (immediate and eventual) and analyticity of semigroups of composition operators are then considered on the Hardy space H2(𝔻)H^{2}(\mathbb{D}), even though other classes of Banach spaces may also be considered. References are given for more complete information.

Finally, we provide some perspectives for semigroups of composition operators on H2(+)H^{2}(\mathbb{C}_{+}), where +\mathbb{C}_{+} is the right-half-plane, as well as an analysis of semigroups of composition operators on the Fock space. The latter case can be treated in a complete way since the non-trivial semiflows involved are necessarily expressed using polynomials of degree one.

2 Background

2.1 Strongly continuous semigroups of operators: definition and characterization

We recall some of the standard facts about one-parameter semigroups of operators, which may be found in many places, such as [29] and [43].

Definition 2.1.

A semigroup (Tt)t0(T_{t})_{t\geq 0} of operators on a Banach space XX is a family of bounded operators satisfying:
(i) T0=IdT_{0}={\rm Id}, the identity operator, and (ii) Tt+s=TtTsT_{t+s}=T_{t}T_{s} for all s,t0s,t\geq 0.
If, in addition, it satisfies:
(iii) TtxxT_{t}x\to x as t0+t\to 0^{+} for all xXx\in X, then it is called a strongly continuous or C0C_{0}-semigroup.

A uniformly continuous semigroup (Tt)t0(T_{t})_{t\geq 0} is one satisfying TtId0\|T_{t}-{\rm Id}\|\to 0 as t0+t\to 0^{+}.

Associated with this is the notion of an infinitesimal generator, or simply generator. We define an (in general unbounded) operator AA whose domain is

D(A):={xX:limt0+Ttxxtexists},D(A):=\{x\in X:\lim_{t\to 0^{+}}\frac{T_{t}x-x}{t}\quad\hbox{exists}\},

and then

Ax:=limt0+TtxxtforxD(A).Ax:=\lim_{t\to 0^{+}}\frac{T_{t}x-x}{t}\qquad\hbox{for}\quad x\in D(A).

Moreover, the generator of a C0C_{0}-semigroup characterizes completely a semigroup, that is two C0C_{0}-semigroups are equal if and only if their generators are equal.

By the uniform boundedness principle, each C0C_{0}-semigroup is uniformly bounded on each compact interval. As a corollary, for every C0C_{0}-semigroup (Tt)t0(T_{t})_{t\geq 0}, there exists ww\in\mathbb{R} and M1M\geq 1 such that

TtMewt for all t0.\|T_{t}\|\leq Me^{wt}\mbox{ for all }t\geq 0.

Contractive C0C_{0}-semigroups are the ones for which one can take M=1M=1 and w=0w=0, whereas quasicontractive C0C_{0}-semigroups are the ones for which one can take M=1M=1 and ww is arbitrary.

The domain D(A)D(A) of the generator AA of a C0C_{0}-semigroup is always dense in XX (and moreover (A,D(A))(A,D(A)) is a closed operator). It is then natural to characterize the linear operators (A,D(A))(A,D(A)) that are the generator of C0C_{0}-semigroups. For semigroups of contractions on Hilbert spaces, Lumer and Phillips, in 1961, provided a beautiful criterion [38] (see also [29, Theorem 3.15]).

Theorem 2.2 (Lumer–Phillips).

Let (A,D(A))(A,D(A)) be a linear operator with dense domain on a Hilbert space HH. The following assertions are equivalent:

  1. (i)

    AA generates a C0C_{0}-semigroup of contractions;

  2. (ii)

    there exists λ>0\lambda>0 such that (λIdA)D(A)=H(\lambda{\rm Id}-A)D(A)=H and for all xD(A)x\in D(A),

    ReAx,x0;\mathop{\rm Re}\nolimits\langle Ax,x\rangle\leq 0;
  3. (iii)

    for all λ>0\lambda>0 we have (λIdA)D(A)=H(\lambda{\rm Id}-A)D(A)=H and for all xD(A)x\in D(A),

    ReAx,x0.\mathop{\rm Re}\nolimits\langle Ax,x\rangle\leq 0.

Since (Tt)t0(T_{t})_{t\geq 0} is a C0C_{0}-semigroup of quasicontractions on a Hilbert space HH if and only if there exists w0w\geq 0 such that (ewtTt)t0(e^{-wt}T_{t})_{t\geq 0} is a C0C_{0}-semigroup of contractions, it is then easy to deduce a version of the Lumer–Phillips Theorem for C0C_{0}-quasicontractions.

Theorem 2.3.

Let (A,D(A))(A,D(A)) be a linear operator with dense domain on a Hilbert space HH. The following assertions are equivalent:

  1. (i)

    AA generates a C0C_{0}-semigroup of quasicontractions;

  2. (ii)

    there exists λ>0\lambda>0 such that (λIdA)D(A)=H(\lambda{\rm Id}-A)D(A)=H and

    supxD(A),x1ReAx,x<;\sup_{x\in D(A),\|x\|\leq 1}\mathop{\rm Re}\nolimits\langle Ax,x\rangle<\infty;
  3. (iii)

    for all λ>0\lambda>0 we have (λIdA)D(A)=H(\lambda{\rm Id}-A)D(A)=H and

    supxD(A),x1ReAx,x<.\sup_{x\in D(A),\|x\|\leq 1}\mathop{\rm Re}\nolimits\langle Ax,x\rangle<\infty.

For generators of C0C_{0}-semigroups which are not necessarily quasicontractive, another beautiful criterion involving the growth of the resolvent is due to Hille and Yosida [29, Theorem 3.8].

Theorem 2.4.

Let AA be a linear operator defined on a linear subspace D(A)D(A) of the Banach space XX, ww\in\mathbb{R} and M>0M>0. Then AA generates a C0C_{0}-semigroup (Tt)t0(T_{t})_{t\geq 0} that satisfies

TtMewt\|T_{t}\|\leq Me^{wt}

if and only if

  • a)

    AA is closed and D(A)D(A) is dense in XX;

  • (b)

    every real λ>w\lambda>w belongs to the resolvent set of AA and for such λ\lambda and for all positive integers nn,

    (λIdA)nM(λω)n.\|(\lambda{\rm Id}-A)^{-n}\|\leq{\frac{M}{(\lambda-\omega)^{n}}}.

We shall restrict ourselves to Banach spaces XX of functions that are holomorphic on a domain Ω\Omega (usually the unit disc 𝔻\mathbb{D} but sometimes the right half-plane +\mathbb{C}_{+} or the complex plane \mathbb{C}) and satisfying the condition that point evaluations ff(z)f\mapsto f(z) are continuous for all zΩz\in\Omega. Assuming this, we have that norm convergence of a sequence (fn)(f_{n}) to ff implies local uniform convergence (uniform convergence on compact subsets of Ω\Omega).

Recall that for suitable ϕ:ΩΩ\phi:\Omega\to\Omega holomorphic, the composition operator Cϕ:XXC_{\phi}:X\to X is defined by (Cϕf)(z)=f(ϕ(z))(C_{\phi}f)(z)=f(\phi(z)), for fXf\in X and zΩz\in\Omega (assuming that CϕC_{\phi} maps XX boundedly into XX, an issue which will be discussed later). Likewise, for suitable ww holomorphic on Ω\Omega, the weighted composition operator Ww,ϕW_{w,\phi} on XX is defined by (Ww,ϕf)(z)=w(z)f(ϕ(z))(W_{w,\phi}f)(z)=w(z)f(\phi(z)).

The main theme of this paper is to study C0C_{0}-semigroups of (weighted) composition operators. However, we may also look at composition semigroups from a non-operatorial point of view (for example, as in [16]).

2.2 Analytic semiflows on a domain and models for semiflows on 𝔻\mathbb{D}

Definition 2.5.

A continuous analytic semiflow on a domain Ω\Omega is a family (ϕt)t0(\phi_{t})_{t\geq 0} of holomorphic mappings from Ω\Omega to itself satisfying:
(i) ϕ0(z)=z\phi_{0}(z)=z for all zΩz\in\Omega,
(ii) ϕt+s=ϕtϕs\phi_{t+s}=\phi_{t}\circ\phi_{s} for all s,t0s,t\geq 0, and
(iii) for all zΩz\in\Omega, the mapping tϕt(z)t\mapsto\phi_{t}(z) is continuous on [0,)[0,\infty).

Remark 2.6.

A family (ϕt)t0(\phi_{t})_{t\geq 0} of holomorphic mappings from Ω\Omega to itself satisfying only (i) and (ii) is called an algebraic semiflow. Such an algebraic semiflow (ϕt)t0(\phi_{t})_{t\geq 0} is continuous on 𝔻\mathbb{D} if and only if there exists a𝔻a\in\mathbb{D} such that limt0ϕt(a)=1\lim_{t\to 0}\phi^{\prime}_{t}(a)=1 (see [16, Thm. 8.1.16]).

In this situation there exists a unique holomorphic function G:ΩG:\Omega\to\mathbb{C} such that

ϕt(z)t=G(ϕt(z)),zΩ and t[0,).\frac{\partial\phi_{t}(z)}{\partial t}=G(\phi_{t}(z)),\,\,\,z\in\Omega\mbox{ and }t\in[0,\infty).

This function is called the infinitesimal generator of the analytic semiflow (ϕt)t0(\phi_{t})_{t\geq 0} on Ω\Omega and we denote by Gen(Ω)\mathop{\rm Gen}\nolimits(\Omega) the set of all infinitesimal generator of analytic semiflows on Ω\Omega.

There are several complete characterization of Gen(𝔻)\mathop{\rm Gen}\nolimits(\mathbb{D}) in [16, Chap. 10], which is discussed in more details in Section 4.2.

Analytic semiflows on 𝔻\mathbb{D} can be partitioned into two classes, depending on the localization of their Denjoy–Wolff point α\alpha, discussed below in Section 4 (see [1], [24, Chap. 2] and [16, Chap. 8]).

If α𝔻\alpha\in\mathbb{D}, by conjugating by the automorphism bαb_{\alpha}, where

bα(z):=αz1α¯z,b_{\alpha}(z):=\frac{\alpha-z}{1-\overline{\alpha}z},

we may suppose without loss of generality that α=0\alpha=0. In this case there is a semiflow model

ϕt(z)=h1(ecth(z)),\phi_{t}(z)=h^{-1}(e^{-ct}h(z)),

where cc\in\mathbb{C} with Rec0\mathop{\rm Re}\nolimits c\geq 0, and h:𝔻Ωh:\mathbb{D}\to\Omega is a conformal bijection between 𝔻\mathbb{D} and a domain Ω\Omega\subset\mathbb{C}, with h(0)=0h(0)=0 and Ω\Omega is spiral-like or star-like (if cc is real), in the sense that

ectwΩfor allwΩandt0.e^{-ct}w\in\Omega\quad\hbox{for all}\quad w\in\Omega\quad\hbox{and}\quad t\geq 0.

If α𝕋\alpha\in\mathbb{T}, then there exists a conformal map hh from 𝔻\mathbb{D} onto a domain Ω\Omega such that Ω+itΩ\Omega+it\subset\Omega for all t0t\geq 0, and there is a semiflow model

ϕt(z)=h1(h(z)+it).\phi_{t}(z)=h^{-1}(h(z)+it).

2.3 Models for analytic flows on 𝔻\mathbb{D}

This subsection relies heavily on Subsection 8.2 in [16].

Definition 2.7.

A family (ϕt)t0(\phi_{t})_{t\geq 0} of analytic selfmaps of 𝔻\mathbb{D} is a called a continuous (algebraic) flow if

  • a)

    ϕt\phi_{t} is an automorphism of 𝔻\mathbb{D} for all t0t\geq 0;

  • b)

    (ϕt)t0(\phi_{t})_{t\geq 0} is a continuous (algebraic) semiflow.

If ϕt\phi_{t} is an automorphism for all t0t\geq 0, we can introduce the notation ϕt:=ϕt1\phi_{-t}:=\phi_{t}^{-1} for all t0t\geq 0 and then observe that

  • c)

    ϕs+t=ϕsϕt\phi_{s+t}=\phi_{s}\circ\phi_{t} for all s,ts,t\in\mathbb{R};

  • d)

    for all z𝔻z\in\mathbb{D}, the mapping tϕt(z)t\mapsto\phi_{t}(z) is continuous on \mathbb{R} if (ϕt)t0(\phi_{t})_{t\geq 0} is a continuous flow.

In fact, d) is equivalent to the continuity of tϕtHol(𝔻)t\mapsto\phi_{t}\in\mathop{\rm Hol}\nolimits(\mathbb{D}) on \mathbb{R}, where Hol(𝔻)\mathop{\rm Hol}\nolimits(\mathbb{D}) is endowed with the topology of the uniform convergence on compacta of 𝔻\mathbb{D}.

Here is a characterization of continuous flows in the set of continuous semiflows [16, Thm. 8.2.4].

Theorem 2.8.

Let (ϕt)t0(\phi_{t})_{t\geq 0} be a continuous (algebraic) semiflow on 𝔻\mathbb{D}. Then it is a continuous (algebraic) flow if and only if there exists t0>0t_{0}>0 such that ϕt0\phi_{t_{0}} is an automorphism.

The following theorem [16, Thm. 8.2.6] is an explicit description of all the continuous flows on 𝔻\mathbb{D}.

Theorem 2.9.

Let (ϕt)t0(\phi_{t})_{t\geq 0} be a nontrivial continuous flow on 𝔻\mathbb{D}. Then (ϕt)t0(\phi_{t})_{t\geq 0} has one of the following three mutually exclusive forms:

  • 1)

    There exists α𝔻\alpha\in\mathbb{D} and w{0}w\in\mathbb{R}\setminus\{0\} such that

    ϕt(z)=(eiwt|α|2)z+α(1eiwt)α¯(eiwt1)z+1|α|2eiwt,t0,z𝔻.\phi_{t}(z)=\frac{(e^{-iwt}-|\alpha|^{2})z+\alpha(1-e^{-iwt})}{\overline{\alpha}(e^{-iwt}-1)z+1-|\alpha|^{2}e^{-iwt}},\,\,\,t\geq 0,\,\,\,z\in\mathbb{D}.

    Moreover it is the unique continuous flow of elliptic automorphisms for which ϕt(α)=α\phi_{t}(\alpha)=\alpha for all tt and ϕt(α)=eiwt\phi^{\prime}_{t}(\alpha)=e^{-iwt}.

  • 2)

    There exist α1,α2𝕋\alpha_{1},\alpha_{2}\in\mathbb{T}, α1α2\alpha_{1}\neq\alpha_{2} and c>0c>0 such that

    ϕt(z)=(α2α1ect)z+α1α2(ect1)(1ect)z+α2ectα1,t0,z𝔻.\phi_{t}(z)=\frac{(\alpha_{2}-\alpha_{1}e^{ct})z+\alpha_{1}\alpha_{2}(e^{ct}-1)}{(1-e^{ct})z+\alpha_{2}e^{ct}-\alpha_{1}},\,\,\,t\geq 0,\,\,\,z\in\mathbb{D}.

    Moreover it is the unique continuous flow of hyperbolic automorphisms for which ϕt(αi)=αi\phi_{t}(\alpha_{i})=\alpha_{i} for all tt (i=1,2i=1,2) and ϕt(α1)=ect\phi^{\prime}_{t}(\alpha_{1})=e^{-ct}.

  • 3)

    There exist α𝕋\alpha\in\mathbb{T} and w{0}w\in\mathbb{R}\setminus\{0\} such that

    ϕt(z)=(1iwt)z+iwαtiwα¯tz+1+iwt.\phi_{t}(z)=\frac{(1-iwt)z+iw\alpha t}{-iw\overline{\alpha}tz+1+iwt}.

    Moreover it is the unique continuous flow of parabolic automorphisms for which ϕt(α)=α\phi_{t}(\alpha)=\alpha for all tt and ϕt′′(α)=2itwα¯\phi^{\prime\prime}_{t}(\alpha)=2itw\overline{\alpha}.

2.4 C0C_{0}-semigroups of composition operators

Clearly a semiflow (ϕt)t0(\phi_{t})_{t\geq 0} induces a semigroup of composition operators (on 𝔻\mathbb{D} these are bounded, by Littlewood’s subordination theorem [37]), and the following condition gives a way of testing the strong continuity.

Proposition 2.10.

Let EE be a dense subspace of a Banach space XX and (Tt)t0(T_{t})_{t\geq 0} a semigroup of bounded operators on XX such that there exists a δ>0\delta>0 with sup0tδTt<\sup_{0\leq t\leq\delta}\|T_{t}\|<\infty. Then (Tt)(T_{t}) is a C0C_{0}-semigroup on XX if and only if limt0TtffX=0\lim_{t\to 0}\|T_{t}f-f\|_{X}=0 for all fEf\in E.

In particular, if the polynomials are dense in XX then it is enough to check that limt0TtenenX=0\lim_{t\to 0}\|T_{t}e_{n}-e_{n}\|_{X}=0 for all n=0,1,2,n=0,1,2,\ldots, where en(z)=zne_{n}(z)=z^{n}.

Proof.

Clearly, the “only if” condition holds. Conversely, if limt0TtffX=0\lim_{t\to 0}\|T_{t}f-f\|_{X}=0 for all fEf\in E, let M:=sup0tδTtM:=\sup_{0\leq t\leq\delta}\|T_{t}\|, and let ϵ>0\epsilon>0 be given and fXf\in X. We may find a pEp\in E such that fp<ϵ2(M+1)\|f-p\|<\dfrac{\epsilon}{2(M+1)}. Then

Ttff\displaystyle\|T_{t}f-f\| TtfTtp+Ttpp+pf\displaystyle\leq\|T_{t}f-T_{t}p\|+\|T_{t}p-p\|+\|p-f\|
Mfp+ϵ/2+pf<ϵ\displaystyle\leq M\|f-p\|+\epsilon/2+\|p-f\|<\epsilon

for sufficently small tt. ∎

2.5 Spaces on which semigroups are not C0C_{0}

In Proposition 2.10 we have seen a sufficient condition for a semiflow to induce a C0C_{0}-semigroup of composition operators, and in the Hardy, Dirichlet and Bergman spaces we do indeed arrive at such a semigroup.

Recently, Gallardo-Gutiérrez, Siskakis and Yakubovich [32] have shown that weighted composition operators (Wwt,ϕt)t0(W_{w_{t},\phi_{t}})_{t\geq 0} do not form a nontrivial C0C_{0}-semigroup on spaces XX satisfying HXH^{\infty}\subset X\subset\mathcal{B}, where \mathcal{B} is the Bloch space. This includes the case X=BMOAX=\rm BMOA. The proof is based on estimates for derivatives of interpolating Blaschke products.

For composition operators, and with X=HX=H^{\infty} and \mathcal{B}, this result was shown earlier by Blasco et al [15] with an argument involving the Dunford–Pettis property. For spaces between HH^{\infty} and \mathcal{B}, the result for composition operators was given by Anderson, Jovovic and Smith [2] using geometric function theory.

3 Motivation

3.1 Universal operators

Rota [45, 46] introduced the concept of a universal operator.

Definition 3.1.

An operator U(H)U\in\mathcal{L}(H) is universal if for all nonzero T(H)T\in\mathcal{L}(H) there is a closed subspace {0}\mathcal{M}\neq\{0\} of HH, an isomorphism J:HJ:\mathcal{M}\to H and a λ{0}\lambda\in\mathbb{C}\setminus\{0\} such that UU\mathcal{M}\subset\mathcal{M} and U|=J1(λT)JU_{|\mathcal{M}}=J^{-1}(\lambda T)J.

That is, a universal operator is a “model” for all T(H)T\in\mathcal{L}(H). Universal operators are of interest in the study of the invariant subspace problem, whether every operator on a separable infinite-dimensional Hilbert space has a nontrivial closed invariant subspace. This has a positive solution if and only every minimal invariant subspace of a given universal operator is finite-dimensional. We refer to [21] for more details and examples.

Caradus [17] gives a convenient sufficient condition for a Hilbert space operator to be universal.

Theorem 3.2.

Let U(H)U\in\mathcal{L}(H) be such that:
(i) dimkerU=\dim\ker U=\infty; and
(ii) UU is surjective.
Then UU is universal.

Consider now the hyperbolic automorphisms

ϕr(z)=z+r1+rz\phi_{r}(z)=\frac{z+r}{1+rz}

with r(1,1)r\in(-1,1). It is helpful to consider them using the parametrization

ψt(z):=ϕr(z)withr=1et1+et\psi_{t}(z):=\phi_{r}(z)\qquad\hbox{with}\quad r=\frac{1-e^{-t}}{1+e^{-t}}

for tt\in\mathbb{R}.

It was shown by Nordgren, Rosenthal and Wintrobe [42] that for r0r\neq 0 the operator CϕrIdC_{\phi_{r}}-{\rm Id} is universal on H2H^{2}. Of course CϕrC_{\phi_{r}} and CϕrIdC_{\phi_{r}}-{\rm Id} have the same invariant subspaces.

A simpler proof using the fact that CϕrC_{\phi_{r}} can be embedded in a C0C_{0}-group was given by Cowen and Gallardo-Gutiérrez [25].

Some work on the invariant subspaces of such operators is due to Matache [39], Mortini [41] and Gallardo-Gutiérrez and Gorkin [31].

3.2 Isometries

Another application of weighted composition operators goes back to Banach [11], who showed that every surjective isometry FF of the space C(K)C(K) of continuous complex functions on a compact metric space KK has the form

F(f)=w(fϕ),F(f)=w(f\circ\phi),

where wC(K)w\in C(K) satisfies |w|=1|w|=1 and ϕ\phi is a homeomorphism of KK.

The Hardy space H2(𝔻)H^{2}(\mathbb{D}) is a Hilbert space, and thus has many (linear) isometries; however for other pp with 1<p<1<p<\infty there are relatively few, and they are expressible as weighted composition operators. In [27] deLeeuw, Rudin, and Wermer gave a description of the isometric surjections of H1(𝔻)H^{1}(\mathbb{D}), which arise from conformal mappings of 𝔻\mathbb{D} onto 𝔻\mathbb{D}.

Moreover, Forelli [30] gave the following theorem, which does not assume surjectivity.

Theorem 3.3.

Suppose that p2p\neq 2 and that T:Hp(𝔻)Hp(𝔻)T:H^{p}(\mathbb{D})\to H^{p}(\mathbb{D}) is a linear isometry. Then there are a non-constant inner function ϕ\phi and a function FHp(𝔻)F\in H^{p}(\mathbb{D}) such that

Tf=F(fϕ)Tf=F(f\circ\phi) (1)

for fHp(𝔻)f\in H^{p}(\mathbb{D}).

3.3 Change of domain

It is well known that composition operators on the Hardy space H2(+)H^{2}(\mathbb{C}_{+}) of the right half-plane are unitarily equivalent to weighted composition operators on H2(𝔻)H^{2}(\mathbb{D}). For example the following explicit formula is given in [20].

Proposition 3.4.

Let MM denote the self-inverse bijection from 𝔻\mathbb{D} onto +\mathbb{C}_{+} given by M(z)=1z1+zM(z)=\dfrac{1-z}{1+z}, and let Ψ:++\Psi:\mathbb{C}_{+}\to\mathbb{C}_{+} be holomorphic. Then the composition operator CΨC_{\Psi} on H2(+)H^{2}(\mathbb{C}_{+}) is unitarily equivalent to the operator LΦ:H2(𝔻)H2(𝔻)L_{\Phi}:H^{2}(\mathbb{D})\to H^{2}(\mathbb{D}) defined by

LΦf(z)=1+Φ(z)1+zf(Φ(z)),L_{\Phi}f(z)=\frac{1+\Phi(z)}{1+z}f(\Phi(z)),

where Φ=MΨM\Phi=M\circ\Psi\circ M.

So for example the C0C_{0}-group (Tt)t(T_{t})_{t\in\mathbb{R}} on H2(+)H^{2}(\mathbb{C}_{+}) given by Ttg(z)=g(etz)T_{t}g(z)=g(e^{t}z) (z+)(z\in\mathbb{C}_{+}) for gH2(+)g\in H^{2}(\mathbb{C}_{+}) is unitarily equivalent to the weighted composition group (St)t(S_{t})_{t\in\mathbb{R}} given by

Stf(z)=21+z+et(1z)f(1+zet(1z)1+z+et(1z)).S_{t}f(z)=\frac{2}{1+z+e^{t}(1-z)}f\left(\frac{1+z-e^{t}(1-z)}{1+z+e^{t}(1-z)}\right).

Formulae for general domains are given, for example, in [35, Prop. 2.1]. If WW is a weighted composition operator between two Hardy–Smirnoff spaces E2(Ω1)E^{2}(\Omega_{1}) and E2(Ω2)E^{2}(\Omega_{2}), with Ω1\Omega_{1} and Ω2\Omega_{2} conformally equivalent to the disc 𝔻\mathbb{D}, then WW is unitarily equivalent to a weighted composition operator on H2(𝔻)H^{2}(\mathbb{D}). Similar formulae are given for Bergman spaces.

4 Asymptotic behaviour of TnT^{n} or TtT_{t}

4.1 The discrete unweighted case

For a fixed composition operator CϕC_{\phi} there are several possible modes of convergence for the sequence (Cϕn)n1(C^{n}_{\phi})_{n\geq 1}, some of which we now list in progressively weaker order.

  • Norm convergence. There exists an operator P(X)P\in\mathcal{L}(X) such that CϕnP0\|C_{\phi}^{n}-P\|\to 0.

  • Strong convergence. There exists an operator P(X)P\in\mathcal{L}(X) such that CϕnxPx0\|C_{\phi}^{n}x-Px\|\to 0 for all xXx\in X.

  • Weak convergence. There exists an operator P(X)P\in\mathcal{L}(X) such that CϕnxPxC_{\phi}^{n}x\to Px weakly for all xXx\in X.

In each case PP is the projection onto fix(Cϕ):={xX:Cϕx=x}\mathop{\rm fix}\nolimits(C_{\phi}):=\{x\in X:C_{\phi}x=x\} along the decomposition

X=fix(Cϕ)Im(IdCϕ)¯.X=\mathop{\rm fix}\nolimits(C_{\phi})\oplus\overline{\mathop{\rm Im}\nolimits({\rm Id}-C_{\phi})}.

The following theorem from [7] helps with the analysis.

Theorem 4.1.

Let T(X)T\in\mathcal{L}(X) with supnTn<\sup_{n}\|T^{n}\|<\infty. Then the following are equivalent.
(i) P:=limTnP:=\lim T^{n} exists and PP is a finite-rank operator.
(ii) (a) The essential spectral radius re(T)r_{e}(T) satisfies re(T)<1r_{e}(T)<1;
  (b) σp(T)𝕋{1}\sigma_{p}(T)\cap\mathbb{T}\subset\{1\}; and
  (c) if 1σ(T)1\in\sigma(T) then 11 is a pole of the resolvent (zIdT)1(z{\rm Id}-T)^{-1} of order 1.

In this case PP is the residue at 1.

We sketch the proof.

Proof.

(i) \implies (ii):
Let X1=PXX_{1}=PX, X2=(IdP)XX_{2}=({\rm Id}-P)X and Ti=TXiT_{i}=T_{X_{i}} for i=1,2i=1,2.

Then T2n(X2)0\|T_{2}^{n}\|_{\mathcal{L}(X_{2})}\to 0 as nn\to\infty, and hence r(T2)<1r(T_{2})<1.

Since σ(Y)=σ(T1)σ(T2)={1}σ(T2)\sigma(Y)=\sigma(T_{1})\cup\sigma(T_{2})=\{1\}\cup\sigma(T_{2}) and

(λIdT)1={1λ1on X1,(λIdT2)1on X2,(\lambda{\rm Id}-T)^{-1}=\begin{cases}\frac{1}{\lambda-1}&\hbox{on }X_{1},\\ (\lambda{\rm Id}-T_{2})^{-1}&\hbox{on }X_{2},\end{cases}

we see that (ii) follows.

(ii) \implies (i):
Let PP be the residue at 11, and let X1=PXX_{1}=PX, X2=(IdP)XX_{2}=({\rm Id}-P)X and Ti=TXiT_{i}=T_{X_{i}} for i=1,2i=1,2.

Then σ(T1)={1}\sigma(T_{1})=\{1\} and σ(T2)=σ(T){1}\sigma(T_{2})=\sigma(T)\setminus\{1\} by (a) and (b).

Thus r(T2)<1r(T_{2})<1 and so T2n(X2)0\|T_{2}^{n}\|_{\mathcal{L}(X_{2})}\to 0.

It follows from (c) that T1T_{1} is diagonalisable and thus T1=IdT_{1}={\rm Id}. ∎

Recall that XHol(𝔻)X\hookrightarrow\mathop{\rm Hol}\nolimits(\mathbb{D}) means that XX embeds continuously in Hol(𝔻)\mathop{\rm Hol}\nolimits(\mathbb{D}), which means that, for all λ𝔻\lambda\in\mathbb{D}, δλ:ff(λ)\delta_{\lambda}:f\mapsto f(\lambda) from XX to \mathbb{C} is bounded.

Arendt and Batty [3] have given criteria for strong convergence. More recently, from [5] we mention the following theorem.

Theorem 4.2.

Let ϕ:𝔻𝔻\phi:\mathbb{D}\to\mathbb{D} be holomorphic, Cϕ(X)C_{\phi}\in\mathcal{L}(X), where XX is a Banach space such that XHol(𝔻)X\hookrightarrow\mathop{\rm Hol}\nolimits(\mathbb{D}). Then (Cϕn)n(C_{\phi}^{n})_{n} converges uniformly if and only if re(Cϕ)<1r_{e}(C_{\phi})<1.

So it remains to study the essential spectral radius of CϕC_{\phi}. To this end we write ϕ(n)=ϕϕ n factors\phi^{(n)}=\underbrace{\phi\circ\ldots\phi}_{\hbox{ $n$ factors}}.

Denjoy–Wolff theory

A mapping ϕ:𝔻𝔻\phi:\mathbb{D}\to\mathbb{D} is an elliptic automorphism if it has the form

ϕ=ψaRθψa,\phi=\psi_{a}\circ R_{\theta}\circ\psi_{a},

where a𝔻a\in\mathbb{D} and ψa(z)=az1a¯z=ψa1(z)\psi_{a}(z)=\dfrac{a-z}{1-\overline{a}z}=\psi_{a}^{-1}(z), and Rθ(z)=eiθzR_{\theta}(z)=e^{i\theta}z with θ\theta\in\mathbb{R}.

Then the classical Denjoy–Wolff theorem states that provided ϕ\phi is not an elliptic automorphism, the sequence (ϕ(n))n(\phi^{(n)})_{n} converges uniformly to some α𝔻¯\alpha\in\overline{\mathbb{D}} on each compact subset of 𝔻\mathbb{D}.

Such an α\alpha is called the Denjoy–Wolff point of ϕ\phi and is sometimes denoted by DW(ϕ)\mathop{\rm DW}\nolimits(\phi).

Case 1. |α|=1|\alpha|=1.

Theorem 4.3.

Suppose that [z]X\mathbb{C}[z]\subset X and suppose that each LXL\in X^{\prime} with L(en)=L(e1)nL(e_{n})=L(e_{1})^{n} has the form L(f)=δz0(f):=f(z0)L(f)=\delta_{z_{0}}(f):=f(z_{0}) for some z0𝔻z_{0}\in\mathbb{D}. Then supnCϕn=\sup_{n}\|C_{\phi}^{n}\|=\infty. Therefore even weak convergence of (Cϕn)(C_{\phi}^{n}) cannot occur.

Proof.

If supnCϕn=M<\sup_{n}\|C_{\phi}^{n}\|=M<\infty then

|f(α)|=limn|f(ϕ(n)(0))|δ0Mf,|f(\alpha)|=\lim_{n\to\infty}|f(\phi^{(n)}(0))|\leq\|\delta_{0}\|M\|f\|,

so δα=δz0\delta_{\alpha}=\delta_{z_{0}} for some z0𝔻z_{0}\in\mathbb{D}, which is absurd. ∎

Case 2. |α|<1|\alpha|<1.

The natural candidate for PP is given by Pf=f(α)=δα(f)𝟏Pf=f(\alpha)=\delta_{\alpha}(f){\bf 1}, a rank-one operator.

Theorem 4.4.

For composition operators on Hp(𝔻)H^{p}(\mathbb{D}), with 1p<1\leq p<\infty, the sequence (Cϕn)(C_{\phi}^{n}) converges uniformly and strongly if and only if |α|<1|\alpha|<1 and ϕ\phi is not inner. It converges weakly if and only if |α|<1|\alpha|<1.

We proved that re(Cϕ)<1r_{e}(C_{\phi})<1 on H2H^{2} if and only if ϕ\phi is not inner and |α|<1|\alpha|<1. When ϕ\phi is inner and |α|<1|\alpha|<1 then CϕC_{\phi} is similar to an isometry (this is a necessary and sufficient condition [12]). We deduce the result for Hp(𝔻)H^{p}(\mathbb{D}) from a theorem of Shapiro [48], namely re,H2(Cϕ))2=re,Hp(Cϕ))pr_{e,H^{2}}(C_{\phi}))^{2}=r_{e,H^{p}}(C_{\phi}))^{p}. For p=1p=1 there is the inequality “>>”. Hence re,H2(Cϕ))<1r_{e,H^{2}}(C_{\phi}))<1 implies that re,Hp(Cϕ))<1r_{e,H^{p}}(C_{\phi}))<1.

We may summarise the results obtained on various Banach spaces in a table [5, 6].

Space Uniform Strong Weak
Hp(𝔻)H^{p}(\mathbb{D}) ϕ\phi not inner, |α|<1|\alpha|<1 ϕ\phi not inner, |α|<1|\alpha|<1 |α|<1|\alpha|<1
Aβp(𝔻)A^{p}_{\beta}(\mathbb{D}), β>1\beta>-1, 1p<1\leq p<\infty |α|<1|\alpha|<1 |α|<1|\alpha|<1 |α|<1|\alpha|<1
0,γ\mathcal{B}_{0},\mathcal{B}^{\gamma} |α|<1|\alpha|<1 |α|<1|\alpha|<1 |α|<1|\alpha|<1
HνpH_{\nu_{p}}^{\infty}, 0<p<0<p<\infty |α|<1|\alpha|<1 |α|<1|\alpha|<1 |α|<1|\alpha|<1

We write dAdA for the normalized Lebesgue area measure on 𝔻\mathbb{D}, i.e. dA(reiθ)=1πrdrdθdA(re^{i\theta})=\frac{1}{\pi}rdrd\theta. The standard weighted Bergman space, Aβp(𝔻)A_{\beta}^{p}(\mathbb{D}), β1\beta\geq-1, p1p\geq 1 is the space of all holomorphic functions f:𝔻f:\mathbb{D}\to\mathbb{C} such that

𝔻|f(z)|p(1|z|2)β𝑑A(z)<.\int_{\mathbb{D}}|f(z)|^{p}(1-|z|^{2})^{\beta}dA(z)<\infty.

Every AβpA_{\beta}^{p} is a Banach space when 1p<1\leq p<\infty with norm the pthp^{th} root of above integral, denoted by fAβp\|f\|_{A_{\beta}^{p}}.
The unweighted Bergman space, ApA^{p} is obtained when β=0\beta=0.
The standard Hardy space Hp(𝔻)H^{p}(\mathbb{D}) are obtained when β=1\beta=-1.

Here 0\mathcal{B}_{0} and γ\mathcal{B}_{\gamma} are variations on the Bloch space, namely,

0={fHol(𝔻):lim|z|1(1|z|2)|f(z)|=0}\mathcal{B}_{0}=\{f\in\mathop{\rm Hol}\nolimits(\mathbb{D}):\lim_{|z|\to 1}(1-|z|^{2})|f^{\prime}(z)|=0\}

and

γ={fHol(𝔻):supz𝔻(1|z|2)γ|f(z)|<}.\mathcal{B}^{\gamma}=\{f\in\mathop{\rm Hol}\nolimits(\mathbb{D}):\sup_{z\in\mathbb{D}}(1-|z|^{2})^{\gamma}|f^{\prime}(z)|<\infty\}.

For p>0p>0, the standard weighted Bergman space of infinite order, Hνp(𝔻)H_{\nu_{p}}^{\infty}(\mathbb{D}) (or HνpH_{\nu_{p}}^{\infty}), is the Banach space of all holomorphic functions f:𝔻f:\mathbb{D}\to\mathbb{C} such that

fHνp:=supz𝔻νp(z)|f(z)|<,\|f\|_{H_{\nu_{p}}^{\infty}}:=\sup_{z\in\mathbb{D}}\nu_{p}(z)|f(z)|<\infty,

with the norm as defined above, where νp(z)=(1|z|2)p\nu_{p}(z)=(1-|z|^{2})^{p}.

4.2 The continuous unweighted case

We now address the question whether we can deduce the asymptotic behaviour of a C0C_{0} semigroup (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} from properties of its generator.

As we shall explain in more detail below, for XHol(𝔻)X\hookrightarrow\mathop{\rm Hol}\nolimits(\mathbb{D}) with (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} a C0C_{0}-semigroup induced by an analytic semiflow (ϕt)t0(\phi_{t})_{t\geq 0} with

G(z)=limt0ϕt(z)zt,G(z)=\lim_{t\to 0}\dfrac{\phi_{t}(z)-z}{t},

the generator AA is given by Af=GfAf=Gf^{\prime} (fD(A)f\in D(A)) with dense domain D(A)={fX:GfX}D(A)=\{f\in X:Gf^{\prime}\in X\}.

Various well-known properties of analytic semiflows are the following, which can be found in the recent book [16].

  1. 1.

    For all t0t\geq 0, ϕt\phi_{t} is injective.

  2. 2.

    If there is a t0>0t_{0}>0 such that ϕt0\phi_{t_{0}} is an (elliptic) automorphism, then ϕt\phi_{t} is an elliptic automorphism for all t>0t>0.

  3. 3.

    For all semiflows that are not elliptic automorphisms, there is a unique α𝔻¯\alpha\in\overline{\mathbb{D}} such that limt0ϕtα,K=0\lim_{t\to 0}\|\phi_{t}-\alpha\|_{\infty,K}=0 for all compact subsets K𝔻K\subset\mathbb{D}. Such an α\alpha is called the Denjoy–Wolff point of (ϕt)t0(\phi_{t})_{t\geq 0} (see also Section 4.1).

The following classical theorem of Berkson and Porta [13] describes a semigroup in terms of its generator.

Theorem 4.5.

let (ϕt)t0(\phi_{t})_{t\geq 0} be an analytic semiflow on 𝔻\mathbb{D}. Then the generator

G(z):=limt0ϕt(z)ztG(z):=\lim_{t\to 0}\dfrac{\phi_{t}(z)-z}{t}

exists for all z𝔻z\in\mathbb{D}. Also GHol(𝔻)G\in\mathop{\rm Hol}\nolimits(\mathbb{D}) and

G(z)=(αz)(1α¯z)F(z)G(z)=(\alpha-z)(1-\overline{\alpha}z)F(z) (2)

where FHol(𝔻)F\in\mathop{\rm Hol}\nolimits(\mathbb{D}) and ReF0\mathop{\rm Re}\nolimits F\geq 0 (this implies that F0<p<1Hp(𝔻)F\in\bigcap_{0<p<1}H^{p}(\mathbb{D}) and hence has radial limits almost everywhere on 𝕋\mathbb{T}). Reciprocally, any GG of the form (2) is the generator of a semiflow.

Another characterization was given in [9, Thm. 3.9]. If one knows a priori that GH1(𝔻)G\in H^{1}(\mathbb{D}), the necessary and sufficient condition is that Re(z¯G(z))0\mathop{\rm Re}\nolimits(\overline{z}G^{*}(z))\leq 0 a.e. on 𝕋\mathbb{T}, where GG^{*} denotes the radial limit of GG. The proof relies on the following observation which can be seen as a maximum principle for harmonic functions. If h:𝔻h:\mathbb{D}\to\mathbb{C} is in H1(𝔻)H^{1}(\mathbb{D}), then h(z)=12π02πRe(eit+zeitz)f(eit)dth(z)=\frac{1}{2\pi}\int_{0}^{2\pi}\mathop{\rm Re}\nolimits\left(\frac{e^{it}+z}{e^{it}-z}\right)f^{*}(e^{it})dt, where hh^{*} stand for the radial limit of hh, and thefore if Re(h(eit))0\mathop{\rm Re}\nolimits(h^{*}(e^{it}))\leq 0 a.e. on 𝕋\mathbb{T}, then Re(h(z))0\mathop{\rm Re}\nolimits(h(z))\leq 0 for all z𝔻z\in\mathbb{D}.

We have already mentioned that FF and thus GG are always in Hp(𝔻)H^{p}(\mathbb{D}) for 0<p<10<p<1, which implies the existence of GG^{*} in Lp(𝕋)L^{p}(\mathbb{T}). Nevertheless it is not possible to improve [9, Thm. 3.9] since G(z):=z(z1z+1)G(z):=-z\left(\frac{z-1}{z+1}\right) is in Hp(𝔻)H^{p}(\mathbb{D}) for all p<1p<1, satisfies Re(z¯G(z))0\mathop{\rm Re}\nolimits(\overline{z}G^{*}(z))\leq 0 a.e. on 𝕋\mathbb{T} and is not the generator of a semiflow using (2).

Combining [16, Thm. 10.2.6] which is called “Abate’s formula” and [16, Thm. 10.2.10], we obtain the following characterization of generators.

Theorem 4.6.

Let G:𝔻G:\mathbb{D}\to\mathbb{C} be holomorphic function. The following assertions are equivalent:

  • (i)

    GG is the generator of an analytic semiflow on 𝔻\mathbb{D};

  • (ii)

    Re(2z¯G(z)+(1|z|2)G(z))0\mathop{\rm Re}\nolimits(2\overline{z}G(z)+(1-|z|^{2})G^{\prime}(z))\leq 0 for all z𝔻z\in\mathbb{D};

  • (iii)

    lim supz𝔻,zξRe(z¯G(z))0\limsup_{z\in\mathbb{D},z\to\xi}\mathop{\rm Re}\nolimits(\overline{z}G(z))\leq 0 for all ξ𝕋\xi\in\mathbb{T}.

Now, from results on Hp(𝔻)H^{p}(\mathbb{D}) and similar spaces in Section 4.1 we obtain the following result.

Theorem 4.7.

Let X=Hp(𝔻)X=H^{p}(\mathbb{D}) or other classical Banach spaces (e.g. Bergman, Dirichlet). Let (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} be a C0C_{0}-semigroup of composition operators on XX with generator A:fGfA:f\mapsto Gf^{\prime}. Then the semigroup converges uniformly if and only if it converges strongly, and this is if and only if GG has a zero in 𝔻\mathbb{D}, and there exists ξ𝕋\xi\in\mathbb{T} such that lim supz𝔻,zξRe(z¯G(z))<0\limsup_{z\in\mathbb{D},z\to\xi}\mathop{\rm Re}\nolimits(\overline{z}G(z))<0.

Indeed, GG has a zero in 𝔻\mathbb{D} if and only if either ϕt\phi_{t} is an elliptic automorphism for each tt or else the Denjoy–Wolff point α\alpha lies in 𝔻\mathbb{D}. The condition lim supz𝔻,zξRe(z¯G(z))<0\limsup_{z\in\mathbb{D},z\to\xi}\mathop{\rm Re}\nolimits(\overline{z}G(z))<0 implies that (ϕt)t0(\phi_{t})_{t\geq 0} is not a semiflow of automorphisms. For if (ϕt)t0(\phi_{t})_{t\geq 0} were a semigroup of automorphisms, then both GG and G-G would generate semigroups and so we would have limz𝔻,zξRe(z¯G(z))=0\lim_{z\in\mathbb{D},z\to\xi}\mathop{\rm Re}\nolimits(\overline{z}G(z))=0 everywhere on 𝕋\mathbb{T}. Note that since ϕt\phi_{t} is injective, it is not inner precisely when it is not an automorphism.

4.3 Weighted composition operators

Iterates of weighted composition operators on holomorphic function spaces XX that embed continuously in Hol(𝔻)\mathop{\rm Hol}\nolimits(\mathbb{D}) are studied in [22]. Some of the most relevant theorems here are the following, where again Ww,ϕW_{w,\phi} is the weighted composition operator fw(fϕ)f\mapsto w(f\circ\phi). The methods of proof are similar to those of [5]. We assume first that ϕ\phi is not an elliptic automorphism and that Ww,ϕW_{w,\phi} is at least power bounded.

Theorem 4.8.

With XX and Ww,ϕW_{w,\phi} as above, suppose also that α:=DW(ϕ)𝔻\alpha:=\mathop{\rm DW}\nolimits(\phi)\in\mathbb{D}. Then the sequence (Ww,ϕn)(W_{w,\phi}^{n}) converges weakly as nn\to\infty if and only if (i) |w(α)|<1|w(\alpha)|<1, or (ii) |w(α)|=1|w(\alpha)|=1 and supnWw,ϕn<\sup_{n}\|W_{w,\phi}^{n}\|<\infty.

Theorem 4.9.

Under the hypotheses of Theorem 4.8, suppose that either |w(α)|<1|w(\alpha)|<1, or (ii) |w(α)|=1|w(\alpha)|=1 and supnWw,ϕn<\sup_{n}\|W_{w,\phi}^{n}\|<\infty. Then (Ww,ϕn)(W_{w,\phi}^{n}) converges uniformly if and only if re(Ww,ϕ)<1r_{e}(W_{w,\phi})<1.

For elliptic automorphisms the story is slightly different.

Theorem 4.10.

Let ϕ\phi be an elliptic automorphism of infinite order with fixed point α𝔻\alpha\in\mathbb{D} and wA(𝔻)w\in A(\mathbb{D}) bounded away from 0 on 𝔻\mathbb{D}. Suppose that supnWw,ϕn<\sup_{n}\|W_{w,\phi}^{n}\|<\infty. Then (Ww,ϕn)(W_{w,\phi}^{n}) converges uniformly if and only if |w(α)|<1|w(\alpha)|<1.

4.4 Isometry and similarity to isometry

As a byproduct of the results in Section 4.1 we have characterizations of the properties isometry and similarity to an isometry for composition operators [5, 6].

Spaces CφC_{\varphi} isometric CφC_{\varphi} similar to an isometry
Hp(𝔻),1p<H^{p}(\mathbb{D}),1\leq p<\infty φ\varphi inner and φ(0)=0\varphi(0)=0 φ\varphi inner and b𝔻\exists b\in\mathbb{D} with φ(b)=b\varphi(b)=b
Aβp(𝔻),β>1,1p<A^{p}_{\beta}(\mathbb{D}),\beta>-1,1\leq p<\infty φ\varphi rotation φ\varphi elliptic automorphism
\mathcal{B} φ(0)=0\varphi(0)=0 and τφ=1\tau_{\varphi}^{\infty}=1 b𝔻\exists b\in\mathbb{D} with φ(b)=b\varphi(b)=b and τφ=1\tau_{\varphi}^{\infty}=1
0\mathcal{B}_{0} or α\mathcal{B}^{\alpha}, α1\alpha\neq 1 φ\varphi rotation φ\varphi elliptic automorphism
Hνp,0<p<H^{\infty}_{\nu_{p}},0<p<\infty φ\varphi rotation φ\varphi elliptic automorphism

Here

τϕ:=supz𝔻1|z|21|ϕ(z)|2|ϕ(z)|.\tau_{\phi}^{\infty}:=\sup_{z\in\mathbb{D}}\frac{1-|z|^{2}}{1-|\phi(z)|^{2}}|\phi^{\prime}(z)|.

4.5 Generators

Suppose now that (Tt)t0(T_{t})_{t\geq 0} is a C0C_{0}-semigroup of weighted composition operators Tt=wtCϕtT_{t}=w_{t}C_{\phi_{t}} acting on a suitable space of functions on a domain Ω\Omega. We have already discussed results describing generators in the unweighted case in Theorems 4.5 and 4.6, namely Af=GfAf=Gf^{\prime}, where

G(z):=limt0ϕt(z)zt.G(z):=\lim_{t\to 0}\dfrac{\phi_{t}(z)-z}{t}. (3)

The converse, as shown in work by Arendt and Chalendar [4] and Gallardo and Yakubovich [33], associates with a holomorphic function GG a semiflow satisfying u(t)=G(u(t))u^{\prime}(t)=G(u(t)), with u(0)=zu(0)=z.

Theorem 4.11.

Let XX be a function space with continuous point evaluations on a domain Ω\Omega. If either
(i) If (zn)Ω(z_{n})\subset\Omega such that znzΩ{}z_{n}\to z\in\Omega\cup\{\infty\} and limnf(zn)\lim_{n\to\infty}f(z_{n}) exists in \mathbb{C} for all fXf\in X then zΩz\in\Omega, or
(ii) Ω=𝔻\Omega=\mathbb{D} and Hol(𝔻¯)XHol(𝔻)\mathop{\rm Hol}\nolimits(\overline{\mathbb{D}})\subset X\subset\mathop{\rm Hol}\nolimits(\mathbb{D}) with continuous embeddings,
then the semigroup associated with GG is a (quasicontractive) semigroup of composition operators.

Note that Hol(𝔻¯)\mathop{\rm Hol}\nolimits(\overline{\mathbb{D}}) embeds continuously in XX means that for each ε>0\varepsilon>0, there exists a positive constant c(ε)c(\varepsilon) such that

fXc(ϵ)sup|z|1+ϵ|f(z)|,fA(D(0,1+ε)),\|f\|_{X}\leq c(\epsilon)\sup_{|z|\leq 1+\epsilon}|f(z)|,\,\,f\in A(D(0,1+\varepsilon)),

where A(D(0,1+ε))A(D(0,1+\varepsilon)) is the algebra of all holomorphic functions in D(0,1+ε)D(0,1+\varepsilon), the open disc centered at 0 and of radius 1+ε1+\varepsilon, and continuous on the closure of D(0,1+ε)D(0,1+\varepsilon).

For weighted composition operators on 𝔻\mathbb{D}, a semigroup of the form Tt=wtCϕtT_{t}=w_{t}C_{\phi_{t}} has infinitesimal generator Af=Gf+gfAf=Gf^{\prime}+gf, where now

g(z)=wt(z)tt=0g(z)=\frac{\partial w_{t}(z)}{\partial t}_{t=0}

and GG is once again given by (3). Assuming Condition (ii) in Theorem 4.11, it is shown in [32, Thm. 3.1] that every semigroup with generator of the form Af=Gf+gfAf=Gf^{\prime}+gf is a semigroup of weighted composition operators. See also [14] for a generalization of [4].

5 Compactness and analyticity

This section is mainly extracted from [10]. We restrict our study to the semigroups of composition operators on the Hardy space H2(𝔻)H^{2}(\mathbb{D}), even though some of the results can be extended to non-Hilbertian Hardy spaces or weighted Hardy spaces such as the Dirichlet space.

5.1 Immediate and eventual compactness

We recall that a semigroup (T(t))t0(T(t))_{t\geq 0} is said to be immediately compact if the operators T(t)T(t) are compact for all t>0t>0. A semigroup (T(t))t0(T(t))_{t\geq 0} is said to be eventually compact if there exists t0>0t_{0}>0 such that T(t)T(t) is compact for all tt0t\geq t_{0}. Similar definitions hold for immediately/eventually trace-class.

The following theorem [43, Chap. 2, Thm 3.3] links immediate compactness with continuity in norm.

Theorem 5.1.

Let (T(t))t0(T(t))_{t\geq 0} be a C0C_{0}-semigroup and let AA be its infinitesimal generator. Then (T(t))t0(T(t))_{t\geq 0} is immediately compact if and only if
(i) the resolvent R(λ,A)R(\lambda,A) is compact for all (or for one) λσ(A)\lambda\in\mathbb{C}\setminus\sigma(A), and
(ii) limstT(s)T(t)=0\lim_{s\to t}\|T(s)-T(t)\|=0 for all t>0t>0.

We begin with an elementary observation.

Proposition 5.2.

Suppose that for some t0>0t_{0}>0 one has |ϕt0(ξ)|=1|\phi_{t_{0}}(\xi)|=1 on a set of positive measure; then Cϕt0C_{\phi_{t_{0}}} is not compact on H2(𝔻)H^{2}(\mathbb{D}) and so the semigroup (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} is not immediately compact.

Proof.

For the Hardy space, this follows since the weakly null sequence (en)n0(e_{n})_{n\geq 0} with en(z)=zne_{n}(z)=z^{n} is mapped into (ϕt0n)(\phi_{t_{0}}^{n}), which does not converge to 0 in norm. ∎

We shall now give a sufficient condition for immediate compactness of a semigroup of composition operators, in terms of the associated function GG. First, we recall a classical necessary and sufficient condition for compactness of a composition operator CϕC_{\phi} in the case when ϕ\phi is univalent [24, pp. 132, 139].

Theorem 5.3.

For ϕ:𝔻𝔻\phi:\mathbb{D}\to\mathbb{D} analytic and univalent, the composition operator CϕC_{\phi} is compact on H2(𝔻)H^{2}(\mathbb{D}) if and only if

limzξ1|z|21|ϕ(z)|2=0\lim_{z\to\xi}\frac{1-|z|^{2}}{1-|\phi(z)|^{2}}=0

for all ξ𝕋\xi\in\mathbb{T}.

The following proposition collects together standard results on trace-class composition operators [24, p. 149].

Proposition 5.4.

For ϕ:𝔻𝔻\phi:\mathbb{D}\to\mathbb{D} analytic with ϕ<1\|\phi\|_{\infty}<1, the composition operator CϕC_{\phi} is trace-class on H2(𝔻)H^{2}(\mathbb{D}).

Here is an example showing that immediate and eventual compactness are different even for semigroups of composition operators on H2(𝔻)H^{2}(\mathbb{D}).

Example 5.5.

Let hh be the Riemann map from 𝔻\mathbb{D} onto the starlike region

Ω:=𝔻{z:0<Re(z)<2 and 0<Im(z)<1},\Omega:=\mathbb{D}\cup\{z\in\mathbb{C}:0<\mathop{\rm Re}\nolimits(z)<2\mbox{ and }0<\mathop{\rm Im}\nolimits(z)<1\},

with h(0)=0h(0)=0. Since Ω\partial\Omega is a Jordan curve, the Carathéodory theorem [44, Thm 2.6, p. 24] implies that hh extends continuously to 𝕋\mathbb{T}.

Let ϕt(z)=h1(eth(z))\phi_{t}(z)=h^{-1}(e^{-t}h(z)). Note that for 0<t<log20<t<\log 2, ϕt(𝕋)\phi_{t}(\mathbb{T}) intersects 𝕋\mathbb{T} on a set of positive measure, and thus, CϕtC_{\phi_{t}} is not compact by Proposition 5.2. Moreover, for t>log2t>\log 2, ϕt<1\|\phi_{t}\|_{\infty}<1, and therefore CϕtC_{\phi_{t}} is compact (actually trace-class). Figure 1 represents the image of φt\varphi_{t} for different values of tt.

[Uncaptioned image]

Figure 1.

5.2 Compact analytic semigroups

A C0C_{0}-semigroup TT will be called analytic (or holomorphic) if there exists a sector Σθ={reiα,r+,|α|<θ}\Sigma_{\theta}=\{re^{i\alpha},r\in\mathbb{R}_{+},|\alpha|<\theta\} with θ(0,π2]\theta\in(0,\frac{\pi}{2}] and an analytic mapping T~:Σθ(X)\widetilde{T}:\Sigma_{\theta}\to\mathcal{L}(X) such that T~\widetilde{T} is an extension of TT and

supξΣθ𝔻T~(ξ)<.\sup_{\xi\in\Sigma_{\theta}\cap\mathbb{D}}\|\widetilde{T}(\xi)\|<\infty.

In both cases, the generator of TT (or T~\widetilde{T}) will be the linear operator AA defined by

D(A)={xX,limt0T(t)xxt exists}D(A)=\left\{x\in X,\lim_{\mathbb{R}\ni t\to 0}\frac{T(t)x-x}{t}\text{ exists}\right\}

and, for all xD(A)x\in D(A),

Ax=limt0T(t)xxt.Ax=\lim_{\mathbb{R}\ni t\to 0}\frac{T(t)x-x}{t}.

In the particular case of analytic semigroups, the compactness is equivalent to the compactness of the resolvent, by Theorem 5.1, since the analyticity implies the uniform continuity [29, p. 109].

Remark 5.6.

For an analytic semigroup (T(t))t0(T(t))_{t\geq 0}, being eventually compact is equivalent to being immediately compact. Indeed, consider QQ, the quotient map from the bounded linear operators on a Hilbert space (H){\mathcal{L}}(H) onto the Calkin algebra (the quotient of (H){\mathcal{L}}(H) by the compact operators). Then (QT(t))t0(QT(t))_{t\geq 0} is an analytic semigroup which vanishes for t>0t>0 large enough, and therefore vanishes identically  (this observation is attributed to W. Arendt).

We may include that remark in the following result.

Theorem 5.7.

Let G:𝔻G:\mathbb{D}\to\mathbb{C} be a holomorphic function such that the operator AA defined by Af(z)=G(z)f(z)Af(z)=G(z)f^{\prime}(z) with dense domain D(A)H2(𝔻)D(A)\subset H^{2}(\mathbb{D}) generates an analytic semigroup (T(t))t0(T(t))_{t\geq 0} of composition operators. Then the following assertions are equivalent:

  1. 1.

    (T(t))t0(T(t))_{t\geq 0} is immediately compact;

  2. 2.

    (T(t))t0(T(t))_{t\geq 0} is eventually compact;

  3. 3.

    ξ𝕋\forall\xi\in\mathbb{T}, limz𝔻,zξ|G(z)zξ|=\lim_{z\in\mathbb{D},z\to\xi}\left|\frac{G(z)}{z-\xi}\right|=\infty.

Theorem 5.8.

Let G:𝔻G:\mathbb{D}\to\mathbb{C} be a holomorphic function such that the operator AA defined by Af(z)=G(z)f(z)Af(z)=G(z)f^{\prime}(z) has dense domain D(A)H2(𝔻)D(A)\subset H^{2}(\mathbb{D}). The operator AA generates an analytic semigroup of composition operators on H2(𝔻)H^{2}(\mathbb{D}) if and only if there exists θ(0,π2)\theta\in(0,\frac{\pi}{2}) such that for all α{θ,0,θ}\alpha\in\{-\theta,0,\theta\}

lim supz𝔻,zξRe(eiαz¯G(z))0 for all ξ𝕋.\limsup_{z\in\mathbb{D},z\to\xi}\mathop{\rm Re}\nolimits(e^{i\alpha}\overline{z}G(z))\leq 0\mbox{ for all }\xi\in\mathbb{T}.

Using the semiflow model, we have the following result.

Theorem 5.9.

Let (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} be an immediately compact analytic semigroup on H2(𝔻)H^{2}(\mathbb{D}). Then the following conditions are equivalent:
1. There exists a t0>0t_{0}>0 such that ϕt0<1\|\phi_{t_{0}}\|_{\infty}<1;
2. For all t>0t>0 one has ϕt<1\|\phi_{t}\|_{\infty}<1.
Therefore, if there exists a t0>0t_{0}>0 such that ϕt0<1\|\phi_{t_{0}}\|_{\infty}<1, then (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} is immediately trace-class.

It is of interest to consider the relation between immediate compactness and analyticity for a C0C_{0}-semigroup of composition operators: this is because compactness of a semigroup (T(t))t0(T(t))_{t\geq 0} is implied by compactness of the resolvent together with norm-continuity at all points t>0t>0, as in Theorem 5.1.

Example 5.10.

Consider

G(z)=2zz1,G(z)=\frac{2z}{z-1},

Now the image of the unit disc under z¯G(z)\overline{z}G(z) is contained in the left half-plane so the operator A:fGfA:f\mapsto Gf^{\prime} generates a non-analytic C0C_{0}-semigroup of composition operators (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} on H2(𝔻)H^{2}(\mathbb{D}). On the other hand, it can be shown that CϕtC_{\phi_{t}} is compact – even trace-class – for each t>0t>0. For we have the equation

ϕt(z)eϕt(z)=e2tzez.\phi_{t}(z)e^{-\phi_{t}(z)}=e^{-2t}ze^{-z}.

Now the function zzezz\mapsto ze^{-z} is injective on 𝔻¯\overline{\mathbb{D}}; this follows from the argument principle, for the image of 𝕋\mathbb{T} is easily seen to be a simple Jordan curve. It follows that ϕt<1\|\phi_{t}\|_{\infty}<1 for all t>0t>0, and so CϕtC_{\phi_{t}} is trace-class.

Example 5.11.

The semigroup corresponding to G(z)=(1z)2G(z)=(1-z)^{2} is analytic but not immediately compact. For

ϕt(z)=(1t)z+ttz+1+t\phi_{t}(z)=\frac{(1-t)z+t}{-tz+1+t}

the Denjoy–Wolff point is 11, so the semigroup cannot be immediately compact.

The analyticity follows on calculating z¯G(z)\overline{z}G(z) for z=eiθz=e^{i\theta}. We obtain 4sin2(θ/2)-4\sin^{2}(\theta/2), which gives the result by Theorem 5.8.

6 An outlook on +\mathbb{C}_{+} and \mathbb{C}

6.1 The right halfplane +\mathbb{C}_{+}

Unlike in the case of the disc, not all composition operators on H2(+)H^{2}(\mathbb{C}+) are bounded, and there are no compact composition operators.

The following theorem was given by Elliott and Jury [28] (see also [40]). Recall that the angular derivative ϕ()\phi^{\prime}(\infty) of a self-map of +\mathbb{C}_{+} is defined by

ϕ()=limzzϕ(z)\phi^{\prime}(\infty)=\lim_{z\to\infty}\frac{z}{\phi(z)}
Theorem 6.1.

Let ϕ:++\phi:\mathbb{C}_{+}\to\mathbb{C}_{+} be holomorphic. The composition operator CϕC_{\phi} is bounded on H2(+)H^{2}(\mathbb{C}_{+}) if and only if ϕ\phi has finite angular derivative 0<λ<0<\lambda<\infty at infinity, in which case Cϕ=λ\|C_{\phi}\|=\sqrt{\lambda}. We also have for the essential norm that Cϕe=Cϕ\|C_{\phi}\|_{e}=\|C_{\phi}\| so that there are no compact composition operators on H2(+)H^{2}(\mathbb{C}_{+}).

Berkson and Porta [13] gave the following criterion for an analytic funcction GG to generate a one-parameter semiflow of analytic mappings from +\mathbb{C}_{+} into itself, namely, solutions to the initial value problem

ϕt(z)t=G(ϕt(z)),ϕ0(z)=z,\frac{\partial\phi_{t}(z)}{\partial t}=G(\phi_{t}(z)),\qquad\phi_{0}(z)=z,

namely the condition

x(ReG)xReGon+,x\frac{\partial(\mathop{\rm Re}\nolimits G)}{\partial x}\leq\mathop{\rm Re}\nolimits G\qquad\hbox{on}\quad\mathbb{C}_{+},

where as usual x=Rezx=\mathop{\rm Re}\nolimits z. In this case the semigroup with generator A:fGfA:f\mapsto Gf^{\prime} consists of composition operators.

Arvanitidis [8] showed that a necessary and sufficient condition for these composition operators to be bounded is that the non-tangential limit δ:=limzG(z)/z\delta:=\angle\lim_{z\to\infty}G(z)/z exists, in which case Cϕt=eδt/2\|C_{\phi_{t}}\|=e^{-\delta t/2} and the semigroup is quasicontractive.

In [10] it was shown that for GG holomorphic on +\mathbb{C}_{+} a necessary condition for the operator A:fGfA:f\mapsto Gf^{\prime} to generate a quasicontractive semigroup is

infz+ReG(z)Rez>,\inf_{z\in\mathbb{C}_{+}}\frac{\mathop{\rm Re}\nolimits G(z)}{\mathop{\rm Re}\nolimits z}>-\infty,

and for a contractive semigroup this infimum is non-negative.

6.2 The complex plane \mathbb{C}

For \mathbb{C} we present material on the Fock space from [23]. The Fock space is arguably one of the most important spaces of entire functions, and it is possible to give a complete answer to several questions involving (weighted) composition operators.

For 1ν<1\leq\nu<\infty the Fock space ν\mathcal{F}^{\nu} is defined to be the space of entire functions ff on \mathbb{C} such that the norm

fν:=(ν2π|f(z)|νeνz2/2𝑑m(z))1/ν\|f\|_{\nu}:=\left(\frac{\nu}{2\pi}\int_{\mathbb{C}}|f(z)|^{\nu}e^{-\nu z^{2}/2}\,dm(z)\right)^{1/\nu}

is finite. The space 2\mathcal{F}^{2} is a Hilbert space with orthonormal basis (e~n)n=0(\tilde{e}_{n})_{n=0}^{\infty} defined by

e~n(z)=znn!.\tilde{e}_{n}(z)=\frac{z^{n}}{\sqrt{n!}}.

6.2.1 Composition operators

Boundedness of composition operators on 2\mathcal{F}^{2} was characterized in [19] as follows.

Theorem 6.2.

For an entire function ϕ\phi the composition operator CϕC_{\phi} is bounded on 2\mathcal{F}^{2} if and only if either ϕ(z)=az+b\phi(z)=az+b with |a|<1|a|<1 and bb\in\mathbb{C} or ϕ(z)=az\phi(z)=az with |a|=1|a|=1. In the case where |a|<1|a|<1 the operator CϕC_{\phi} is compact.

Thus there are relatively few bounded composition operators on 2\mathcal{F}^{2} and most natural questions can be answered easily. When we come to weighted composition operators there will be much more to say.

The discussion here is based mostly on [23], although some results may also be found in [47], which concentrates on spectra and mean ergodicity.

Considering now iterates (Cϕn)n(C_{\phi}^{n})_{n} we have

Cϕnf(z)={f(anz+1an1ab)if a1,f(z)if a=1.C_{\phi}^{n}f(z)=\begin{cases}f\left(a^{n}z+\frac{1-a^{n}}{1-a}b\right)&\hbox{if }a\neq 1,\\ f(z)&\hbox{if }a=1.\end{cases}

We no wish to use Theorem 4.1 again, so we need to know whether CϕC_{\phi} is power-bounded. The following formula is due to [19] in the case ν=2\nu=2, and [26] in general.

Cϕ=exp(14|b|21|a|2)ifϕ(z)=az+bwith|a|<1.\|C_{\phi}\|=\exp\left(\frac{1}{4}\frac{|b|^{2}}{1-|a|^{2}}\right)\qquad\hbox{if}\quad\phi(z)=az+b\quad\hbox{with}\quad|a|<1.
Theorem 6.3.

The asymptotics of iteration of bounded composition operators on ν\mathcal{F}^{\nu} are as follows.

  1. 1.

    If ϕ(z)=az\phi(z)=az with |a|=1|a|=1 and a1a\neq 1, then (Cϕn)(C^{n}_{\phi}) consists of unitary operators and does not converge even weakly.

  2. 2.

    If ϕ(z)=az+b\phi(z)=az+b with |a|<1|a|<1 then (Cϕn)(C_{\phi}^{n}) converges in norm to the operator T:ff(b1a)T:f\mapsto f\left(\dfrac{b}{1-a}\right).

The proof is short, and we include it here.

Proof.

1. In the first case we have that Cϕnf(z)=f(anz)C_{\phi}^{n}f(z)=f(a^{n}z) and for f(z)=zf(z)=z there is clearly no convergence.

2. In the second case

ϕ(n)(z)=anz+1an1ab=:an+bnz, say.\phi^{(n)}(z)=a^{n}z+\frac{1-a^{n}}{1-a}b=:a_{n}+b_{n}z,\hbox{ say}.

and

|bn|21|an|2|b|2|1a|2\frac{|b_{n}|^{2}}{1-|a_{n}|^{2}}\to\frac{|b|^{2}}{|1-a|^{2}}

so CϕC_{\phi} is power bounded. Also re(Cϕ)=0r_{e}(C_{\phi})=0 since CϕC_{\phi} is compact.

For the point spectrum, suppose that λ\lambda is an eigenvalue and ff an eigenvector. Then

f(az+b)=λf(z)f(az+b)=\lambda f(z)

and so f(an+bnz)=λnf(z)f(a_{n}+b_{n}z)=\lambda^{n}f(z), which implies that

λnf(z)f(b/(1a)).\lambda^{n}f(z)\to f(b/(1-a)). (4)

This means that ff is identically zero if |λ|=1|\lambda|=1 and λ1\lambda\neq 1.

Finally if 1σ(Cϕ)1\in\sigma(C_{\phi}) then 11 is an eigenvalue since CϕC_{\phi} is compact. By (4) ff is constant, and dimker(CϕId)=1\dim\ker(C_{\phi}-{\rm Id})=1. Hence 11 is a pole of the resolvent of order at most 1. Indeed, assume that the pole order of 11 is larger than 11. Looking at the Jordan normal form of T0:=Cφ|X0T_{0}:={C_{\varphi}}_{|X_{0}}, where X0=PνX_{0}=P\mathcal{F}^{\nu} and PP the residue, we see that there exists fνf\in\mathcal{F}^{\nu} such that (IdCφ)f=1({\rm Id}-C_{\varphi})f=1_{\mathbb{C}}. Evaluating at z=b1az=\frac{b}{1-a} we obtain a contradiction. Thus the pole order is 11. It follows that PP is the projection onto ker(CφId)=1\ker(C_{\varphi}-{\rm Id})=\mathbb{C}1_{\mathbb{C}} along {fν:f(b/(1a))=0}\{f\in\mathcal{F}^{\nu}:f(b/(1-a))=0\}. Thus Pf=f(b/(1a))1Pf=f(b/(1-a))1_{\mathbb{C}}. This completes the proof that CϕnPC^{n}_{\phi}\to P is norm thanks to Theorem 4.1.

Similarly, we can characterise C0C_{0}-semigroups of bounded composition operators.

Theorem 6.4.

A C0C_{0}-semigroup (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} of bounded composition of ν\mathcal{F}^{\nu} satisfies one of the following conditions for t>0t>0.

  1. 1.

    ϕt(z)=eλtz+C(eλt1)\phi_{t}(z)=e^{\lambda t}z+C(e^{\lambda t}-1) for some λ={z:Rez<0}\lambda\in\mathbb{C}_{-}=\{z\in\mathbb{C}:\mathop{\rm Re}\nolimits z<0\} and CC\in\mathbb{C}.

  2. 2.

    ϕt(z)=eλtz\phi_{t}(z)=e^{\lambda t}z for some λi\lambda\in i\mathbb{R}.

The generator is given by

Af(z)=λ(z+C)f(z),Af(z)=\lambda(z+C)f^{\prime}(z),

where C=0C=0 in case (2).

Moreover (Cϕt)t0(C_{\phi_{t}})_{t\geq 0} converges in norm a tt\to\infty if and only if λ\lambda\in\mathbb{C}_{-}, in which case the limit TT satisfies Tf=f(C)Tf=f(-C).

Using a theorem from [4] we get:

Theorem 6.5.

Every C0C_{0}-semigroup on 2\mathcal{F}^{2} with generator of the form Af=GfAf=Gf^{\prime} for some GHol()G\in\mathop{\rm Hol}\nolimits(\mathbb{C}) is a semigroup of composition operators with generator satisfying

G(z)=az+b,with Rea<0,orG(z)=az,with ai.\begin{array}[]{rll}G(z)&=az+b,&\hbox{with }\mathop{\rm Re}\nolimits a<0,\quad\hbox{or}\\ G(z)&=az,&\hbox{with }a\in i\mathbb{R}.\end{array}

Moreover, the condition

lim sup|z|Rez¯G(z)0\limsup_{|z|\to\infty}\mathop{\rm Re}\nolimits\overline{z}G(z)\leq 0 (5)

is necessary and sufficient for such GG.

6.2.2 Weighted composition operators

For weighted composition operators Ww,ϕW_{w,\phi} we have the result of Le [36], extended by Hai and Khoi [34] as follows:

Proposition 6.6.

The weighted composition operator Ww,ϕW_{w,\phi} with ww not identically zero is bounded on ν\mathcal{F}^{\nu} if and only if both the following conditions hold:

  • wνw\in\mathcal{F}^{\nu};

  • M(w,ϕ):=supz|w(z)|2e(|ϕ(z)|2|z|2)<M(w,\phi):=\sup_{z\in\mathbb{C}}|w(z)|^{2}e^{(|\phi(z)|^{2}-|z|^{2})}<\infty.

Moreover, in this case ϕ(z)=az+b\phi(z)=az+b with |a|1|a|\leq 1. If |a|=1|a|=1 then we also have

w(z)=w(0)eb¯azw(z)=w(0)e^{-\overline{b}az} (6)

for zz\in\mathbb{C}.

See also [18] for a recent discussion of these results. Note that from [34] we have the useful inequality

M(w,ϕ)Ww,ϕM(w,ϕ)/|a|.\sqrt{M(w,\phi)}\leq\|W_{w,\phi}\|\leq\sqrt{M(w,\phi)}/|a|.

Thus for power-boundedness of Ww,ϕW_{w,\phi}, it is possible to give a complete result for |a|=1|a|=1.

Theorem 6.7.

Suppose that ϕ(z)=az+b\phi(z)=az+b with |a|=1|a|=1, then Ww,ϕW_{w,\phi} is power-bounded on ν\mathcal{F}^{\nu} if and only if |w(0)|e|b|2/2|w(0)|\leq e^{-|b|^{2}/2}.

The case |a|<1|a|<1 is apparently unsolved, although this result for non-vanishing ww (as in the case of semigroups) can be found in [18].

Theorem 6.8.

For ϕ(z)=az+b\phi(z)=az+b with |a|<1|a|<1 and ww nonvanishing the operator Ww,ϕW_{w,\phi} is bounded on ν\mathcal{F}^{\nu} if and only if

w(z)=ep+qz+rz2w(z)=e^{p+qz+rz^{2}}

and either:
(i) |r|<β/2|r|<\beta/2, in which case Ww,ϕW_{w,\phi} is compact on ν\mathcal{F}^{\nu}, or
(ii) |r|=β/2|r|=\beta/2 and, with t=q+b¯at=q+\overline{b}a, one has either t=0t=0 or else t0t\neq 0 and r=β2t2|t|2r=-\frac{\beta}{2}\frac{t^{2}}{|t|^{2}}. In case (ii) Ww,ϕW_{w,\phi} is not compact on ν\mathcal{F}^{\nu}.

Finally, we can use this to characterise C0C_{0}-semigroups on the Fock space [23].

Theorem 6.9.

A C0C_{0}-semigroup Ttf(z)=wt(z)f(ϕt(z))T_{t}f(z)=w_{t}(z)f(\phi_{t}(z)) (t0t\geq 0) of weighted composition operators on the Fock space ν\mathcal{F}^{\nu} has one of the following two expressions:

  1. 1.

    ϕt(z)=exp(λt)z+C(exp(λt)1)\phi_{t}(z)=\exp(\lambda t)z+C(\exp(\lambda t)-1) for some λ\lambda\in\mathbb{C}_{-} and CC\in\mathbb{C}; in which case wt=ept+qtz+rtz2w_{t}=e^{p_{t}+q_{t}z+r_{t}z^{2}}, where explicit formulae for ptp_{t}, qtq_{t} and rtr_{t} can be given.

  2. 2.

    ϕt(z)=exp(λt)z+C(exp(λt)1)\phi_{t}(z)=\exp(\lambda t)z+C(\exp(\lambda t)-1) for some λi\lambda\in i\mathbb{R} and CC\in\mathbb{C}, in which case wt(z)=wt(0)exp(C¯(exp(λt)1)z)w_{t}(z)=w_{t}(0)\exp(\overline{C}(\exp(\lambda t)-1)z) and moreover wt(0)=eμte|C|2(eλt1)w_{t}(0)=e^{\mu t}e^{|C|^{2}(e^{\lambda t}-1)} for some μ\mu\in\mathbb{C}.

References

  • [1] M. Abate. Iteration theory of holomorphic maps on taut manifolds. Commenda di Rende (Italy): Mediterranean Press, 1989.
  • [2] A. Anderson, M. Jovovic, and W. Smith. Composition semigroups on BMOA and HH^{\infty}. J. Math. Anal. Appl., 449(1):843–852, 2017.
  • [3] W. Arendt and C. J. K. Batty. Tauberian theorems and stability of one-parameter semigroups. Trans. Amer. Math. Soc., 306(2):837–852, 1988.
  • [4] W. Arendt and I. Chalendar. Generators of semigroups on Banach spaces inducing holomorphic semiflows. Israel J. Math., 229(1):165–179, 2019.
  • [5] W. Arendt, I. Chalendar, M. Kumar, and S. Srivastava. Asymptotic behaviour of the powers of composition operators on Banach spaces of holomorphic functions. Indiana Univ. Math. J., 67(4):1571–1595, 2018.
  • [6] W. Arendt, I. Chalendar, M. Kumar, and S. Srivastava. Powers of composition operators: asymptotic behaviour on Bergman, Dirichlet and Bloch spaces. J. Aust. Math. Soc., 108(3):289–320, 2020.
  • [7] W. Arendt, A. Grabosch, G. Greiner, U. Groh, H. P. Lotz, U. Moustakas, R. Nagel, F. Neubrander, and U. Schlotterbeck. One-parameter semigroups of positive operators, volume 1184 of Lecture Notes in Mathematics. Springer-Verlag, Berlin, 1986.
  • [8] A. G. Arvanitidis. Semigroups of composition operators on Hardy spaces of the half-plane. Acta Sci. Math. (Szeged), 81(1-2):293–308, 2015.
  • [9] C. Avicou, I. Chalendar, and J. R. Partington. A class of quasicontractive semigroups acting on Hardy and Dirichlet space. J. Evol. Equ., 15(3):647–665, 2015.
  • [10] C. Avicou, I. Chalendar, and J. R. Partington. Analyticity and compactness of semigroups of composition operators. J. Math. Anal. Appl., 437(1):545–560, 2016.
  • [11] S. Banach. Theory of linear operations, volume 38 of North-Holland Mathematical Library. North-Holland Publishing Co., Amsterdam, 1987. Translated from the French by F. Jellett, With comments by A. Pełczyński and Cz. Bessaga.
  • [12] F. Bayart. Similarity to an isometry of a composition operator. Proc. Amer. Math. Soc., 131(6):1789–1791 (electronic), 2003.
  • [13] E. Berkson and H. Porta. Semigroups of analytic functions and composition operators. Michigan Math. J., 25(1):101–115, 1978.
  • [14] E. Bernard. Weighted composition semigroups on Banach spaces of holomorphic functions. Preprint, 2022.
  • [15] O. Blasco, M.D. Contreras, S. Díaz-Madrigal, J. Martínez, M. Papadimitrakis, and A.G. Siskakis. Semigroups of composition operators and integral operators in spaces of analytic functions. Ann. Acad. Sci. Fenn. Math., 38(1):67–89, 2013.
  • [16] F. Bracci, M. D. Contreras, and S. Díaz-Madrigal. Continuous semigroups of holomorphic self-maps of the unit disc. Springer Monographs in Mathematics. Springer, Cham, 2020.
  • [17] S. R. Caradus. Universal operators and invariant subspaces. Proc. Amer. Math. Soc., 23:526–527, 1969.
  • [18] T. Carroll and C. Gilmore. Weighted composition operators on Fock spaces and their dynamics. J. Math. Anal. Appl., 502(1):125234, 2021.
  • [19] B. J. Carswell, B. D. MacCluer, and A. Schuster. Composition operators on the Fock space. Acta Sci. Math. (Szeged), 69(3-4):871–887, 2003.
  • [20] I. Chalendar and J. R. Partington. On the structure of invariant subspaces for isometric composition operators on H2(𝔻)H^{2}(\mathbb{D}) and H2(+)H^{2}(\mathbb{C}_{+}). Arch. Math. (Basel), 81(2):193–207, 2003.
  • [21] I. Chalendar and J. R. Partington. Modern approaches to the invariant-subspace problem, volume 188 of Cambridge Tracts in Mathematics. Cambridge University Press, Cambridge, 2011.
  • [22] I. Chalendar and J. R. Partington. Weighted composition operators: isometries and asymptotic behaviour. J. Operator Theory, 86(1):189–201, 2021.
  • [23] I. Chalendar and J.R. Partington. Weighted composition operators on the Fock space: iteration and semigroups. Preprint, 2021, https://arxiv.org/abs/2106.00427.
  • [24] C. C. Cowen and B. D. MacCluer. Composition operators on spaces of analytic functions. Studies in Advanced Mathematics. CRC Press, Boca Raton, FL, 1995.
  • [25] C.C. Cowen and E.A. Gallardo-Gutiérrez. A new proof of a Nordgren, Rosenthal and Wintrobe theorem on universal operators. In Problems and recent methods in operator theory, volume 687 of Contemp. Math., pages 97–102. Amer. Math. Soc., Providence, RI, 2017.
  • [26] J. Dai. The norm of composition operators on the Fock space. Complex Var. Elliptic Equ., 64(9):1608–1616, 2019.
  • [27] K. de Leeuw, W. Rudin, and J. Wermer. The isometries of some function spaces. Proc. Amer. Math. Soc., 11:694–698, 1960.
  • [28] S. Elliott and M. T. Jury. Composition operators on Hardy spaces of a half-plane. Bull. Lond. Math. Soc., 44(3):489–495, 2012.
  • [29] K.-J. Engel and R. Nagel. A short course on operator semigroups. Universitext. Springer, New York, 2006.
  • [30] F. Forelli. The isometries of HpH^{p}. Canad. J. Math., 16:721–728, 1964.
  • [31] E.A. Gallardo-Gutiérrez and P. Gorkin. Minimal invariant subspaces for composition operators. J. Math. Pures Appl. (9), 95(3):245–259, 2011.
  • [32] E.A. Gallardo-Gutiérrez, A.G. Siskakis, and D.V. Yakubovich. Generators of C0C_{0}-semigroups of weighted composition operators. Preprint, 2021, https://arxiv.org/abs/2110.05247.
  • [33] E.A. Gallardo-Gutiérrez and D.V. Yakubovich. On generators of C0C_{0}-semigroups of composition operators. Israel J. Math., 229(1):487–500, 2019.
  • [34] P. V. Hai and L. H. Khoi. Boundedness and compactness of weighted composition operators on Fock spaces p()\mathcal{F}^{p}(\mathbb{C}). Acta Math. Vietnam., 41(3):531–537, 2016.
  • [35] R. Kumar and J. R. Partington. Weighted composition operators on Hardy and Bergman spaces. In Recent advances in operator theory, operator algebras, and their applications, volume 153 of Oper. Theory Adv. Appl., pages 157–167. Birkhäuser, Basel, 2005.
  • [36] T. Le. Normal and isometric weighted composition operators on the Fock space. Bull. Lond. Math. Soc., 46(4):847–856, 2014.
  • [37] J. E. Littlewood. On inequalities in the theory of functions. Proc. London Math. Soc. (2), 23:481–519, 1925.
  • [38] G. Lumer and R. S. Phillips. Dissipative operators in a Banach space. Pac. J. Math., 11:679–698, 1961.
  • [39] V. Matache. On the minimal invariant subspaces of the hyperbolic composition operator. Proc. Amer. Math. Soc., 119(3):837–841, 1993.
  • [40] V. Matache. Weighted composition operators on H2H^{2} and applications. Complex Anal. Oper. Theory, 2(1):169–197, 2008.
  • [41] R. Mortini. Cyclic subspaces and eigenvectors of the hyperbolic composition operator. In Travaux mathématiques, Fasc. VII, Sém. Math. Luxembourg, pages 69–79. Centre Univ. Luxembourg, Luxembourg, 1995.
  • [42] E. Nordgren, P. Rosenthal, and F. S. Wintrobe. Invertible composition operators on HpH^{p}. J. Funct. Anal., 73(2):324–344, 1987.
  • [43] A. Pazy. Semigroups of linear operators and applications to partial differential equations, volume 44 of Applied Mathematical Sciences. Springer-Verlag, New York, 1983.
  • [44] C. Pommerenke. Boundary behaviour of conformal maps, volume 299. Springer Grundlehren, 1993.
  • [45] G. C. Rota. Note on the invariant subspaces of linear operators. Rend. Circ. Mat. Palermo (2), 8:182–184, 1959.
  • [46] G.-C. Rota. On models for linear operators. Comm. Pure Appl. Math., 13:469–472, 1960.
  • [47] W. Seyoum and T. Mengestie. Spectrums and uniform mean ergodicity of weighted composition operators on Fock spaces. Bull. Malays. Math. Sci. Soc., 45(1):455–481, 2022.
  • [48] J. H. Shapiro. The essential norm of a composition operator. Ann. Math. (2), 125:375–404, 1987.