1: \LaTeXparent{nga4.tex}
2:
3: We shall collect some facts from the theory of holomorphic functions in
4: locally convex topological vector spaces ${\cal E}$ (over the complex field $%
5: {\C}$), see e.g., \cite{Di81}. Let ${\cal L}({\cal E}^n)$ be the space of
6: n-linear mappings from ${\cal E}^n$ into ${\C}$ and ${\cal L}_s({\cal E}^n)$
7: the subspace of symmetric n-linear forms. Also let ${\sl P}^n({\cal E})$
8: denote the n-homogeneous polynomials on ${\cal E}$. There is a linear
9: bijection ${\cal L}_s({\cal E}^n)\ni A\longleftrightarrow \widehat{A}\in
10: {\sl P}^n({\cal E})$. Now let ${\cal U}\subset {\cal E}$ be open and
11: consider a function $G:{\cal U}\rightarrow {\C}$.
12:
13: $G$ is said to be {\bf G-holomorphic} if for all $\theta _0\in {\cal U}$ and
14: for all $\theta \in {\cal E}$ the mapping from ${\C}$ to ${\C :}$ $\lambda
15: \rightarrow G(\theta _0+\lambda \theta )$ is holomorphic in some
16: neighborhood of zero in ${\C}$. If $G$ is G-holomorphic then there exists
17: for every $\eta \in {\cal U}$ a sequence of homogeneous polynomials $\frac
18: 1{n!}\widehat{{\rm d}^n G(\eta )}$ such that
19: $$
20: G(\theta +\eta )=\sum\limits_{n=0}^\infty \frac 1{n!}\widehat{{\rm d}%
21: ^nG(\eta )}(\theta )
22: $$
23: for all $\theta $ from some open set ${\cal V}\subset {\cal U}$. $G$ is said
24: to be {\bf holomorphic}, if for all $\eta $ in ${\cal U}$ there exists an
25: open neighborhood ${\cal V}$ of zero such that $\sum\limits_{n=0}^\infty
26: \frac 1{n!}\widehat{{\rm d}^nG(\eta )}(\theta )$ converges uniformly on $%
27: {\cal V}$ to a continuous function. We say that $G$ is holomorphic at $%
28: \theta _0$ if there is an open set ${\cal U}$ containing $\theta _0$ such
29: that $G$ is holomorphic on ${\cal U}$. The following proposition can be
30: found e.g., in \cite{Di81}.
31:
32: \begin{proposition}
33: \label{GHolLocB} $G$ is holomorphic if and only if it is G-holomorphic and
34: locally bounded.
35: \end{proposition}
36:
37: \noindent Let us explicitly consider a function holomorphic at the point $%
38: 0\in {\cal E}={\cal N}_{\Ckl }$, then
39:
40: 1) there exist $p$ and $\varepsilon >0$ such that for all $\xi _0\in {\cal N}%
41: _{\Ckl }$ with $\left| \xi _0\right| _p\leq \varepsilon $ and for all $\xi
42: \in {\cal N}_{\Ckl }$ the function of one complex variable $\lambda
43: \rightarrow G(\xi _0+\lambda \xi )$ is analytic at $0\in {\C}$, and
44:
45: 2) there exists $c>0$ such that for all $\xi \in {\cal N}_{\Ckl }$ with $%
46: \left| \xi \right| _p\leq \varepsilon $ : $\left| G(\xi )\right| \leq c$.
47:
48: \noindent As we do not want to discern between different restrictions of one
49: function, we consider germs of holomorphic functions, i.e., we identify $F$
50: and $G$ if there exists an open neighborhood ${\cal U}:0\in {\cal U}\subset
51: {\cal N}_{\Ckl }$ such that $F(\xi )=G(\xi )$ for all $\xi \in {\cal U}$.
52: Thus we define ${\rm Hol}_0({\cal N}_{\Ckl })$ as the algebra of germs of
53: functions holomorphic at zero equipped with the inductive topology given by
54: the following family of norms%
55: $$
56: {\rm n}_{p,l,\infty }(G)=\sup _{\left| \theta \right| _p\leq 2^{-l}}\left|
57: G(\theta )\right| ,\quad p,l\in \N .
58: $$
59: \bigskip
60:
61: \begin{sloppypar}
62: Let use now introduce spaces of entire functions which will be useful later.
63: Let ${\cal E}_{2^{-l}}^k({\cal H}_{-p,\Ckl })$ denote the set of all entire
64: functions on ${\cal H}_{-p,\Ckl }$ of growth $k\in [1,2]$ and type $2^{-l},\
65: p,l\in \Z $. This is a linear space with norm
66: $$
67: {\rm n}_{p,l,k}(\varphi )=\sup _{z\in {\cal H}_{-p,\Ckkl }}\left| \varphi
68: (z)\right| \exp \left( -2^{-l}|z|_{-p}^k\right) ,\qquad \varphi \in {\cal E}%
69: _{2^{-l}}^k({\cal H}_{-p,\Ckl })
70: $$
71: The space of entire functions on ${\cal N}_{\Ckl }^{\prime }$ of growth $k$
72: and minimal type is naturally introduced by
73: $$
74: {\cal E}_{\min }^k({\cal N}_{\Ckl }^{\prime }):=\ \stackunder{p,l\in \N }{\rm pr\ lim\,}{\cal E}_{2^{-l}}^k({\cal H}_{-p,\Ckl })\ ,
75: $$
76: see e.g., \cite{Ou91}. We will also need the space of entire functions on $%
77: {\cal N}_{\Ckl }$ of growth $k$ and finite type:%
78: $$
79: {\cal E}_{\max }^k({\cal N}_{\Ckl }):=\ \stackunder{p,l\in \N }{\rm ind\ lim}%
80: {\cal E}_{2^l}^k({\cal H}_{p,\Ckl })\ .
81: $$
82: In the following we will give an equivalent description of ${\cal E}_{\min
83: }^k({\cal N}_{\Ckl }^{\prime })$ and ${\cal E}_{\max }^k({\cal N}_{\Ckl })$.
84: Cauchy's inequality and Corollary \ref{KernelCor} allow to write the Taylor
85: coefficients in a convenient form. Let $\varphi \in {\cal E}_{\min }^k({\cal %
86: N}_{\Ckl }^{\prime })$ and $z\in {\cal N}_{\Ckl }^{\prime }$, then there
87: exist kernels $\varphi ^{(n)}\in {\cal N}_{\Ckl }^{\hat \otimes n}$ such
88: that
89: $$
90: \langle z^{\otimes n},\varphi ^{(n)}\rangle =\frac 1{n!}\widehat{{\rm d}^n\varphi (0)}(z)
91: $$
92: i.e.,
93: \begin{equation}
94: \label{phi(z)}\varphi (z)=\sum_{n=0}^\infty \langle z^{\otimes n},\varphi
95: ^{(n)}\ \rangle .
96: \end{equation}
97: This representation allows to introduce a nuclear topology on ${\cal E}%
98: _{\min }^k({\cal N}_{\Ckl }^{\prime })$, see \cite{Ou91} for details. Let
99: {\rm E}$_{p,q}^\beta $ denote the space of all functions of the form (\ref
100: {phi(z)}) such that the following Hilbertian norm
101: \begin{equation}
102: \label{3StrichNorm}\lnorm \varphi \rnorm _{p,q,\beta
103: }^2:=\sum_{n=0}^\infty (n!)^{1+\beta }2^{nq}\left| \varphi ^{(n)}\right|
104: _p^2\;,\quad p,q\in \N
105: \end{equation}
106: is finite for $\beta \in [0,1]$. (By $\left| \varphi ^{(0)}\right| _p$ we
107: simply mean the complex modulus for
108: all $p$.) The space {\rm E}$_{-p-,q}^{-\beta }$ with
109: the norm $\lnorm \varphi \rnorm _{-p,-q,-\beta }$ is defined analogously.
110: \end{sloppypar}
111:
112: \begin{theorem}
113: \label{Ekminprlim}The following topological identity holds:
114: $$
115: \stackunder{p,q\in \N }{\rm pr\ lim}\;{\rm E}_{p,q}^\beta ={\cal E}_{\min
116: }^{\frac 2{1+\beta }}({\cal N}_{\Ckl }^{\prime })\quad .
117: $$
118: \end{theorem}
119:
120: The proof is an immediate consequence of the following two lemmata which
121: show that the two systems of norms are in fact equivalent.
122:
123: \begin{lemma}
124: \label{nplk3Strich}Let $\varphi \in ${\rm E}$_{p,q}^\beta $ then $\varphi
125: \in {\cal E}_{2^{-l}}^{\frac 2{1+\beta }}({\cal H}_{-p,\Ckl })$ for $l=\frac
126: q{1+\beta }$. Moreover
127: \begin{equation}
128: \label{nplk3StrichNorm}{\rm n}_{p,l,k}(\varphi )\leq \lnorm \varphi \rnorm %
129: _{p,q,\beta }\ ,\ \ k=\tfrac 2{1+\beta }\ .
130: \end{equation}
131: \end{lemma}
132:
133: \TeXButton{Proof}{\proof} We look at the convergence of the series $\varphi
134: (z)=\sum_{n=0}^\infty \langle z^{\otimes n},\varphi ^{(n)}\ \rangle \ $, $%
135: z\in {\cal H}_{-p,\Ckl }\ ,\ \varphi ^{(n)}\in {\cal H}_{p,\Ckl }$ if $%
136: \sum_{n=0}^\infty (n!)^{1+\beta }2^{nq}|\varphi ^{(n)}|_p^2=\lnorm
137: \varphi \rnorm _{p,q,\beta }^2$ is finite. The following estimate holds:%
138: \begin{eqnarray*}
139: \sum_{n=0}^\infty |\langle z^{\otimes n},\varphi ^{(n)}\ \rangle | %
140: & \leq & \left( \sum_{n=0}^\infty (n!)^{1+\beta }2^{nq}|\varphi %
141: ^{(n)}|_p^2\right) ^{1/2}\left( \sum_{n=0}^\infty %
142: \frac 1{(n!)^{1+\beta }}2^{-nq}|z|_{-p}^{2n}\right) ^{1/2} %
143: \\& \leq & \lnorm \varphi \rnorm _{p,q,\beta }\cdot \left( %
144: \sum_{n=0}^\infty \left\{ \frac 1{n!}2^{-\frac{nq}{1+\beta }}%
145: |z|_{-p}^{\frac{2n}{1+\beta }}\right\} ^{1+\beta }\right) ^{1/2} %
146: \\& \leq & \lnorm \varphi \rnorm _{p,q,\beta }\left( %
147: \sum_{n=0}^\infty \frac 1{n!}2^{-\frac{nq}{1+\beta }}|z|_{-p}^{\frac{2n}{1+\beta }}\right) ^{(1+\beta )/2} \\& \leq & \lnorm \varphi \rnorm %
148: _{p,q,\beta }\exp \left( %
149: 2^{-\frac q{1+\beta }}|z|_{-p}^{\frac 2{1+\beta }}\right) .
150: \end{eqnarray*}
151: \TeXButton{End Proof}{\endproof}
152:
153: \begin{lemma}
154: \label{3Strichnplk}For any $p^{\prime },q\in \N $ there exist $p,l\in \N $
155: such that
156: $$
157: {\cal E}_{2^{-l}}^{\frac 2{1+\beta }}({\cal H}_{-p,\Ckl })\subset {\rm E}%
158: _{p^{\prime },q}^\beta
159: $$
160: i.e., there exists a constant $C>0$ such that
161: $$
162: \lnorm \varphi \rnorm _{p^{\prime },q,\beta }\leq C\;{\rm n}%
163: _{p,l,k}(\varphi ),\quad \varphi \in {\cal E}_{2^{-l}}^k({\cal H}_{-p,\Ckl %
164: }),\quad k=\tfrac 2{1+\beta }.
165: $$
166: \end{lemma}
167:
168: \TeXButton{Remark}{\remark } More precisely we will prove the following:\ If $%
169: \varphi \in {\cal E}_{2^{-l}}^k({\cal H}_{-p,\Ckl })$ then $\varphi \in $%
170: {\rm E}$_{p^{\prime },q}^\beta $ for $k=\frac 2{1+\beta }$ and $\rho
171: :=2^{q-2l/k}k^{2/k}e^2\left\| i_{p^{\prime },p}\right\| _{HS}^2<1$ (in
172: particular this requires $p^{\prime }>p$ to be such that the embedding $%
173: i_{p^{\prime },p}:{\cal H}_{p^{\prime }}\hookrightarrow {\cal H}_p$ is
174: Hilbert-Schmidt). \\Moreover the following bound holds
175: \begin{equation}
176: \label{3Normnplk}\lnorm \varphi \rnorm _{p^{\prime },q,\beta }\leq {\rm n}%
177: _{p,l,k}(\varphi )\cdot \left( 1-\rho \right) ^{-1/2}\ .
178: \end{equation}
179:
180: \TeXButton{Proof}{\proof} The assumption $\varphi \in {\cal E}_{2^{-l}}^k(%
181: {\cal H}_{-p,\Ckl })$ implies a bound of the growth of $\varphi :$%
182: $$
183: |\varphi (z)|\leq {\rm n}_{p,l,k}(\varphi )\exp (2^{-l}|z|_{-p}^k)\ .
184: $$
185: For each $\rho >0\ ,\ z\in {\cal H}_{-p,\Ckl }$ the Cauchy inequality from
186: complex analysis \cite{Di81} gives%
187: $$
188: \left| \frac 1{n!}\widehat{{\rm d}^n\varphi (0)}(z)\right| \leq {\rm n}%
189: _{p,l,k}(\varphi )\rho ^{-n}\exp (\rho ^k2^{-l})\;|z|_{-p}^n\ .
190: $$
191: By polarization \cite{Di81} it follows for $z_1,\ldots ,z_n\in {\cal H}_{-p,\Ckl }$%
192: $$
193: \left| \frac 1{n!}{\rm d}^n\varphi (0)(z_1,\ldots ,z_n)\right| \leq {\rm n}%
194: _{p,l,k}(\varphi )\frac 1{n!}\left( \frac n\rho \right) ^n\exp (\rho
195: ^k2^{-l})\prod_{k=1}^n|z_k|_{-p}\ .
196: $$
197: For $p^{\prime }>p$ such that $\left\| i_{p^{\prime },p}\right\| _{HS}$ is
198: finite, an application of the kernel theorem guarantees the existence of
199: kernels $\varphi ^{(n)}\in {\cal H}_{p^{\prime },\Ckl }^{\hat \otimes n}$
200: such that
201: $$
202: \varphi (z)=\sum_{n=0}^\infty \langle z^{\hat \otimes n},\varphi ^{(n)}\
203: \rangle
204: $$
205: with the bound%
206: $$
207: \left| \varphi ^{(n)}\right| _{p^{\prime }}\leq {\rm n}_{p,l,k}(\varphi
208: )\frac 1{n!}\left( \frac n\rho \left\| i_{p^{\prime },p}\right\|
209: _{HS}\right) ^n\exp (\rho ^k\cdot 2^{-l})\ .
210: $$
211: We can optimize the bound with the choice of an $n$-dependent $\rho $.
212: Setting $\rho ^k=2^ln/k$ we obtain
213: \begin{eqnarray*}
214: \left| \varphi ^{(n)}\right| _{p^{\prime }} & \leq & {\rm n}_{p,l,k}(\varphi %
215: )\frac 1{n!}n^{n(1-1/k)}\left( \tfrac 1k2^l\right) ^{-n/k}\left\| %
216: i_{p^{\prime },p}\right\| _{HS}^ne^{n/k} %
217: \\& \leq & {\rm n}_{p,l,k}(\varphi )\;(n!)^{-1/k}\left\{ (k2^{-l})^{1/k}e %
218: \left\| i_{p^{\prime },p}\right\| _{HS}\right\} ^n\ ,
219: \end{eqnarray*}
220: where we used $n^n\leq n!\,e^n$ in the last estimate. Now choose $\beta \in
221: [0,1]$ such that $k=\frac 2{1+\beta }$ to estimate the following norm:%
222: \begin{eqnarray*}
223: \lnorm \varphi \rnorm _{p^{\prime },q,\beta }^2 %
224: & \leq & {\rm n}%
225: _{p,l,k}^2(\varphi )\sum_{n=0}^\infty (n!)^{1+\beta -\frac 2k}2^{qn}\left\{ %
226: (k2^{-l})^{1/k}e\left\| i_{p^{\prime },p}\right\| _{HS}\right\} ^{2n} %
227: \\& \leq & {\rm n}_{p,l,k}^2(\varphi )\left( 1-2^q\left\{ (k2^{-l})^{1/k}e %
228: \left\| i_{p^{\prime },p}\right\| _{HS}\right\} ^2\right) ^{-1}
229: \end{eqnarray*}
230: for sufficiently large $l$. This completes the proof.\TeXButton{End Proof}
231: {\endproof}\bigskip\
232:
233: Analogous estimates for these systems of norms also hold if $\beta ,p,q,l$
234: become negative. This implies the following theorem. For related results see
235: e.g., \cite[Prop.8.6]{Ou91}.
236:
237: \begin{theorem}
238: \label{indlimEkmax} \hfill \\If $\beta \in [0,1)$ then the following
239: topological identity holds:%
240: $$
241: \stackunder{p,q\in \N }{\rm ind\ lim}\ {\rm E}_{-p,-q}^{-\beta }={\cal E}%
242: _{\max }^{2/(1-\beta )}({\cal N}_{\Ckl }).
243: $$
244: If $\beta =1$ we have%
245: $$
246: \stackunder{p,q\in \N }{\rm ind\ lim}\ {\rm E}_{-p,-q}^{-1}={\rm Hol}_0(%
247: {\cal N}_{\Ckl })\ .
248: $$
249: \end{theorem}
250:
251: \noindent This theorem and its proof will appear in the context of section
252: \ref{Characterization}. The characterization of distributions in infinite
253: dimensional analysis is strongly related to this theorem. From this point of
254: view it is natural to postpone its proof to section \ref{Characterization}.
255: