Semi-group of holomorphic mappings

Non-linear semi-group theory is not only of intrinsic interest, but is also important in the study of evolution problems (cf. also Evolution equation). In recent years (as of 2000) many developments have occurred, in particular in the area of non-expansive semi-groups in Banach spaces. As a rule, such semi-groups are generated by accretive operators (cf. also Accretive mapping) and can be viewed as non-linear analogues of the classical linear contraction semi-groups (cf. also Contraction semi-group). Another class of non-linear semi-groups consists of the semi-groups generated by holomorphic mappings. Such semi-groups appear in several diverse fields, including, for example, the theory of Markov stochastic branching processes, Krein spaces, the geometry of complex Banach spaces, control theory, and optimization. These semi-groups can be considered natural non-linear analogues of semi-groups generated by bounded linear operators (cf. also Semi-group of operators). These two distinct classes of non-linear semi-groups are related by the fact that holomorphic self-mappings are non-expansive with respect to Schwarz–Pick pseudo-metrics (see, for example, [a8], [a7], [a6]).

Recall that a function $h$ defined on a domain (open connected subset) $\textbf{D}$ in a complex Banach space $X$ and with values in a complex Banach space is said to be holomorphic in $\textbf{D}$ if for each $x\in\textbf{D}$ the Fréchet derivative of $h$ at $x$ (denoted by $Dh(x)$ or $h'(x)$) exists as a bounded complex-linear mapping of $X$ into the Banach space containing the values of $h$. (Cf. also Banach space of analytic functions with infinite-dimensional domains.)

If $\textbf{D}$ and $\Omega$ are domains in complex Banach spaces $X$ and $Y$, respectively, then the set of holomorphic mappings from $\textbf{D}$ into $\Omega$ is denoted by $\text{Hol}(\textbf{D},\Omega)$. The notation $\text{Hol}(\textbf{D})$ is used to denote the set $\text{Hol}(\textbf{D},\textbf{D})$ of holomorphic self-mappings of $\textbf{D}$.

A family $\{S(t)\subset\text{Hol}(\textbf{D}\}$, where $t\in(0,T)$, $T>0$, is called a (one-parameter) continuous semi-group if

\begin{equation}S(s+t)=S(s)\circ S(t),s,t,s+t\in(0,T),\end{equation}

and

\begin{equation}\lim_{t\to 0^{+}}S(t)(x)=x,x\in D,\end{equation}

where the limit is taken with respect to the strong topology of $X$.

Differentiability of semi-groups with respect to the parameter.

Let $\{S(t)\}$, $t\in(0,T)$, be a continuous semi-group defined on $\textbf{D}$. If the strong limit

\begin{equation}g(x)=\lim_{t\to 0^{+}}\frac{1}{t}(x-S(t)(x))\end{equation}

exists for each $x\in \textbf{D}$, then $g\in\text{Hol}(\textbf{D},X)$ is called the (infinitesimal) generator of the semi-group $\{S(t)\}$. In this case the semi-group $\{S(t)\}$, $t\in(0,T)$, is said to be differentiable (or generated).

For the finite-dimensional case, M. Abate proved in [a2] that each continuous semi-group of holomorphic mappings is everywhere differentiable with respect to its parameter, i.e., it is generated by a holomorphic mapping. In addition, he established a criterion for a holomorphic mapping to be a generator of a one-parameter semi-group. Earlier, for the one-dimensional case, similar facts were presented by E. Berkson and H. Porta in their study [a4] of linear $C_0$-semi-groups of composition operators on Hardy spaces. E. Vesentini investigated semi-groups of fractional-linear transformations that are isometries with respect to the infinitesimal hyperbolic metric on the unit ball of a Banach space [a18]. He used this approach to study several important problems in the theory of linear operators on indefinite metric spaces. Note that, generally speaking, such semi-groups are not everywhere differentiable in the infinite-dimensional case.

In fact, it can be shown (see for example, [a14]), that a continuous semi-group $S(t)$ of holomorphic self-mappings of a domain $\textbf{D}$ in $X$ is generated if and only if the convergence in (a2) is locally uniform on $\textbf{D}$ (cf. also Uniform convergence).

Moreover, if $\textbf{D}$ is hyperbolic (in particular, bounded; cf. also Hyperbolic metric), then $S(t)$ can be continuously extended to all of $\textbf{R}^+=[0,\infty)$ as the solution of the Cauchy problem

\begin{equation}\begin{cases}\frac{\partial u(t,x)}{\partial t}+g(u(t,x))=0,\\u(0,x)=x,\end{cases}\end{equation}

where $x\in \textbf{D}$ and $t\in \textbf{R}^+$, i.e.,

\begin{equation}u(t,x)=S(t)x,x\in \textbf{D},t\in[0,\infty)\end{equation}

(see, for example, [a12], [a13]). Thus, there is a one-to-one correspondence between locally uniformly continuous semi-groups and their generators. If $S(t)$ has a continuous extension to all of $\textbf{R}=(-\infty,\infty)$, then it is actually a one-parameter group of automorphisms of $\textbf{D}$.

Exponential and product formulas.

A holomorphic vector field

\begin{equation}T_g=g(x)\frac{\partial}{\partial x}\end{equation}

on a domain $\textbf{D}$ is determined by a holomorphic mapping $g\in\text{Hol}(\textbf{D},X)$ and can be regarded as a linear operator $T$ mapping $\text{Hol}(\textbf{D},X)$ into itself, where $T_gf\in\text{Hol}(\textbf{D},X)$ is defined by

\begin{equation}(T_gf)(x)=Df(x)g(x),x\in\textbf{D}\end{equation}

The set of all holomorphic vector fields on $\textbf{D}$ is a Lie algebra under the commutator bracket

\begin{equation}[T_g,T_h]=\bigg[g(x)\frac{\partial}{\partial x}, h(x)\frac{\partial}{\partial x}\bigg]:=\end{equation}

\begin{equation}=(Dg(x)h(x)-Dh(x)g(x))\frac{\partial}{\partial x}\end{equation}

(see, for example, [a9], [a15], [a6]).

Furthermore, each vector field (6) is locally integrable in the following sense: for each $x\in\textbf{D}$ there exist a neighbourhood $\Omega$ of $x$ and a $\delta>0$ such that the Cauchy problem (4) has a unique solution $\{u(t,x)\}\subset\textbf{D}$ defined on the set $\{|t|<\delta\}\times\Omega\subset\textbf{R}\times\textbf{D}$.

A holomorphic vector field $T_g$ defined by (6) and (7) is said to be (right) semi-complete (respectively, complete) on $\textbf{D}$ if the solution of the Cauchy problem (4) is well-defined on all of $\textbf{R}^+\times\textbf{D}$ (respectively, $\textbf{R}\times\textbf{D}$), where $\textbf{R}^+=[0,\infty)$ (respectively, $\textbf{R}=(-\infty,\infty)$).

Thus, if $\textbf{D}$ is hyperbolic, then $T_g$ is semi-complete (respectively, complete) if and only if $g$ is the generator of a one-parameter continuous semi-group (respectively, group).

On the other hand, if $\textbf{D}$ is bounded and $\widetilde{\text{Hol}}(\textbf{D},X)$ is the subspace of $\text{Hol}(\textbf{D},X)$ consisting of all $f\in\text{Hol}(\textbf{D},X)$ that are bounded on each ball strictly inside $\textbf{D}$, then a semi-group (group) $\{S(t)\}$, $t\in \textbf{R}^+$ (respectively, $t\in\textbf{R}$), induces a linear semi-group (group) $\{\mathcal{L}(t)\}$ of linear mappings $\mathcal{L}(t):\widetilde{\text{Hol}}(\textbf{D},X)\to\widetilde{\text{Hol}}(\textbf{D},X)$, defined by

\begin{equation}(\mathcal{L}(t)f)(x):=f(S(t)x),\end{equation}

where $t\in\textbf{R}^+$ ($t\in\textbf{R}$) and $x\in\textbf{D}$.

This semi-group is called the semi-group of composition operators on $\widetilde{\text{Hol}}(\textbf{D},X)$. If $\{S(t)\}$, $t\in\textbf{R}^+$ ($t\in\textbf{R}$), is $T$-continuous, (that is, differentiable), then $g\in\widetilde{\text{Hol}}(\textbf{D},X)$, $\{\mathcal{L}(t)\}$, $t\in\textbf{R}^+$ ($t\in\textbf{R}$), is also differentiable and

\begin{equation}\begin{cases}\frac{\partial\mathcal{L}(t)f}{\partial t}+T_g(\mathcal{L}(t)f)=0,\\\mathcal{L}(0)f=f,\end{cases}\end{equation}

for all $f\in\widetilde{\text{Hol}}(\textbf{D},X)$, where $g=-dS(t)/dt|_{t=0}$.

In other words, a holomorphic vector field $T_g$, defined by (6) and (7), and considered as a linear operator on $\widetilde{\text{Hol}}(\textbf{D},X)$, is the infinitesimal generator of the semi-group $\{\mathcal{L}(t)\}$. It is sometimes called the Lie generator. Thus, a holomorphic vector-field $T_g$ is semi-complete (respectively, complete) if and only if it is the Lie generator of a linear semi-group (respectively, group) of composition operators on $\widetilde{\text{Hol}}(\textbf{D},X)$. This follows from the observation that

\begin{equation}\mathcal{L}(t)I_{\textbf{D}}=S(t)\end{equation}

and

\begin{equation}T_gI_{\textbf{D}}=g,\end{equation}

where $I_D$ is the restriction of the identity operator to $\textbf{D}$.

Moreover, using the exponential formula representation for the linear semi-group,

(a13)

(see, for example, [a19], [a9], [a12]), one also has

(a14)

So, a locally uniformly continuous semi-group of holomorphic self-mappings can be represented in exponential form by the holomorphic vector field induced by its generator.

Another exponential representation on a hyperbolic convex domain can be given by using the so-called non-linear resolvent of .

More precisely, let be a bounded (or, more generally, hyperbolic) convex domain. Then it was shown in [a12] and [a13] that is a generator if and only if for each the mapping is a well-defined holomorphic self-mapping of .

Furthermore, if , , is any continuous family of holomorphic self-mappings of such that the limit

exists, then is a generator and the semi-group generated by can be defined by the product formula

(a15)

In particular,

(a16)

(exponential formula), where the limits in (a15) and (a16) are taken with respect to the locally uniform topology on .

Flow-invariance conditions.

Let be a convex subset of a Banach space and let be a continuous mapping on , the closure of . Then the following tangency condition of flow invariance

(a17)

is a necessary condition for the solvability of the evolution equation (a4). A result of R.H. Martin [a11] shows that if is a continuous accretive mapping on , then (a17) is also sufficient for the existence of solutions to the Cauchy problems (a4). These solutions yield a continuous semi-group of contraction mappings on .

For the class of holomorphic mappings, an analogue of Martin's theorem was given in [a3]; namely, if has a uniformly continuous extension to , then it is a semi-complete vector field if and only if it satisfies the boundary flow invariance condition (a17).

However, there are many examples of semi-complete vector fields that have no continuous extension to . In particular, if , then is semi-complete (see [a12]).

For absolutely convex domains, interior flow invariance conditions can be given in terms of their support functionals.

Let be the dual of (cf. also Duality; Adjoint space). For and , the pairing will denote . The duality mapping is defined by

for each .

If is the open unit ball in and maps into , then (a17) is equivalent to the condition

(a18)

For the Euclidean ball in , a certain condition in this direction was established by Abate [a2]. Namely, he proved that is a semi-complete vector field if and only if it satisfies the estimate

(a19)

where

For this condition becomes

(a20)

where , the open unit disc in the complex plane , and .

Despite the usefulness and simplicity of condition (a20) it is not clear how (a18) can be derived from (a20) when has a continuous extension to .

Note also that in the one-dimensional case it follows from the maximum principle for harmonic functions that (a18) implies the following interior condition:

(a21)

Conversely, it is clear that (a18) does result from (a21) if has a continuous extension to all of . It turns out that an analogue of (a21) is a necessary and sufficient condition for to be semi-complete [a1]: Let be the open unit ball in a complex Banach space . Then is a semi-complete vector field on if and only if it is bounded on each subset strictly inside and one of the following conditions holds:

a) For each there exists an such that

b) , ;

c) For each and for each ,

Furthermore, equality in one of the conditions a), b) or c) holds if and only if it holds in the other conditions and is complete.

Also, if is a semi-complete vector field, then is, in fact, complete if and only if its derivative at zero, is a conservative linear operator, i.e.,

for all and (see [a10], [a17]).

Parametric representations of generators.

It is well-known that a complete vector field on the open unit ball in a Banach space is a polynomial of degree at most (see, for example, [a15], [a6]). More precisely, has the form

(a22)

where is an element of , is a conservative operator on and is a homogeneous form of the second degree such that .

Suppose now that a complex Banach space is a so-called triple system. This is equivalent to saying that its open unit ball is a homogeneous domain, i.e., for each pair there exists a holomorphic automorphism of such that (see, for example, [a15], [a6]). Then it is well-known that for each there exists a homogeneous polynomial such that and the mapping defined by

(a23)

is a complete vector field on , which is called a transvection of (cf. also Transvection).

The cone of semi-complete vector fields on admits the decomposition

(a24)

where is the real Banach subspace of consisting of transvections and is the subcone of such that for each ,

In other words, admits a unique representation

(a25)

where is complete, and .

The natural examples of triple systems are a complex Hilbert space , the space of bounded linear operators on , and its subspaces such that if and only if (such subspaces are usually called -algebras). In the latter case the general form of transvections on is , where and is its conjugate. Thus, each semi-complete vector field on the open unit ball of a -algebra has the form

(a26)

where and .

In particular, when is the complex plane and , the open unit disc in , (a26) becomes

(a27)

where and

(a28)

In 1978 E. Berkson and H. Porta [a4], solving an entirely different problem, gave a parametric representation of generators on the unit disc in the complex plane. More precisely, if and only if for some , has the representation

with everywhere. This point is exactly the limit point of the semi-group generated by (that is, its Denjoy–Wolff point, cf. Denjoy–Wolff theorem). The Berkson–Porta formula has also been successfully exploited in other fields; for example, in the classical functional equations of E. Schröder and N.H. Abel (see [a5] and Functional equation; Schröder functional equation).

Let be a complex Hilbert space with inner product and let be its open unit ball. Let denote the Poincaré hyperbolic metric on [a7] (cf. also Poincaré model). A mapping is said to be -monotone if for each pair and positive the following condition holds: whenever and belong to . It was shown in [a13] that if is separable and is a bounded continuous mapping, then is -monotone if and only if it generates a semi-group of -non-expansive self-mappings of . Note also that -monotonicity can be equivalently described as follows:

For a bounded holomorphic mapping and for an arbitrary the latter condition is a criterion for to be semi-complete. For the one-dimensional case, if , then this condition becomes the Berkson–Porta representation of semi-complete vector fields.

References

[a1]	D. Aharonov, S. Reich, D. Shoikhet, "Flow invariance conditions for holomorphic mappings in Banach spaces" Math. Proc. Royal Irish Acad. , 99A (1999) pp. 93–104
[a2]	M. Abate, "The infinitesimal generators of semi-groups of holomorphic maps" Ann. Mat. Pura Appl. , 161 (1992) pp. 167–180
[a3]	L. Aizenberg, S. Reich, D. Shoikhet, "One-sided estimates for the existence of null points of holomorphic mappings in Banach spaces" J. Math. Anal. Appl. , 203 (1996) pp. 38–54
[a4]	E. Berkson, H. Porta, "Semi-groups of analytic functions and composition operators" Michigan Math. J. , 25 (1978) pp. 101–115
[a5]	C.C. Cowen, B.D. MacCluer, "Composition operators on spaces of analytic functions" , CRC (1995)
[a6]	S. Dineen, "The Schwartz lemma" , Clarendon Press (1989)
[a7]	K. Goebel, S. Reich, "Uniform convexity, hyperbolic geometry and nonexpansive mappings" , M. Dekker (1984)
[a8]	L.A. Harris, "Schwarz–Pick systems of pseudometrics for domains in normed linear spaces" , Advances in Holomorphy , North-Holland (1979) pp. 345–406
[a9]	J.M. Isidro, L.L. Stacho, "Holomorphic automorphism groups in Banach spaces: An elementary introduction" , North-Holland (1984)
[a10]	S.G. Krein, "Linear differential equations in Banach spaces" , Amer. Math. Soc. (1971)
[a11]	R.H. Martin Jr., "Differential equations on closed subsets of a Banach space" Trans. Amer. Math. Soc. , 179 (1973) pp. 399–414
[a12]	S. Reich, D. Shoikhet, "Generation theory for semi-groups of holomorphic mappings in Banach spaces" Abstr. Appl. Anal. , 1 (1996) pp. 1–44
[a13]	S. Reich, D. Shoikhet, "Semi-groups and generators on convex domains with the hyperbolic metric" Atti Accad. Naz. Lincei , 8 (1997) pp. 231–250
[a14]	S. Reich, D. Shoikhet, "Metric domains, holomorphic mappings and nonlinear semi-groups" Abstr. Appl. Anal. , 3 (1998) pp. 203–228
[a15]	H. Upmeier, "Jordan algebras in analysis, operator theory and quantum mechanics" , CBMS-NSF Reg. Conf. Ser. in Math. , 67 , Amer. Math. Soc. (1987)
[a16]	E. Vesentini, "semi-groups of holomorphic isometries" Adv. Math. , 65 (1987) pp. 272–306
[a17]	E. Vesentini, "Krein spaces and holomorphic isometries of Cartan domains" S. Coen (ed.) , Geometry and Complex Variables , M. Dekker (1991) pp. 409–413
[a18]	E. Vesentini, "Semi-groups of holomorphic isometries" S. Coen (ed.) , Complex Potential Theory , Kluwer Acad. Publ. (1994) pp. 475–548
[a19]	K. Yosida, "Functional analysis" , Springer (1968)