1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
|
\chapter{Analytic Vectors}
\info{Dies ist nur ein Relikt meines Studiums analytischer Vektoren. Wird wieder entfernt, falls nicht benötigt.}
\begin{definition}{Analytic Vector for an Operator}{analytic-vector-operator}
Let $A : D(A) \to \hilb{H}$ be an unbounded linear operator in a complex Hilbert space $\hilb{H}$.
A vector $x \in \hilb{H}$ is said to be an \emph{analytic vector} for $A$,
if $x$ lies in the domain of the power $A^n$ for all $n \in \NN$, and the power series
\begin{equation*}\tag{power-series-analytic-vector}
\sum_{n=0}^{\infty} \frac{A^n x}{n!} \, z^n
\end{equation*}
has a nonzero radius of convergence.
If the power series converges for all $z \in \CC$,
we say that $x$ is an \emph{entire analytic vector} for $A$.
\end{definition}
Note that, if $x$ is analytic for $A$, then the power series
\begin{equation*}
\sum_{n=0}^{\infty} \frac{\norm{A^n x}}{n!} \, z^n
\end{equation*}
converges for all complex $z$ with $\abs{z} < t$,
that is, in the open disc with radius $t$ centered in the origin of the complex plane.
This is a well-known consequence of the convergence behavior of power series.
\begin{definition}{Analyticity of Vector-Valued Functions}{}
Let $G \subset \CC$ be open and let $\hilb{H}$ be a Hilbert space.
A function $f : G \to \hilb{H}$ is called
\begin{itemize}
\item \emph{strongly analytic} at $a \in G$, if the limit
\begin{equation*}
\lim_{z \to a} \frac{f(z) - f(a)}{z-a}
\end{equation*}
exists in norm.
\item \emph{weakly analytic} in $a \in G$, if for each $w \in \CC$ the scalar-valued function
\begin{equation*}
G \longrightarrow \CC, \quad z \longmapsto \innerp{w}{f(z)}
\end{equation*}
is analytic in $a$.
\end{itemize}
\end{definition}
\begin{lemma}{Equivalence of Weak and Strong Analyticity}{}
Let $G \subset \CC$ be open.
Then a Banach space-valued function is strongly analytic on $G$ if and only if it is weakly analytic on $G$.
\end{lemma}
\begin{myproof}
Let $X$ be a Banach space and suppose that the function $f : G \to X$ is weakly analytic.
By definition, for each $g \in X'$ the scalar valued function $g \circ f : G \to \CC$ is analytic on $G$.
Consider a point $a \in G$.
Since $G$ is open, there exists a circular contour $\gamma$ around $a$ such that $\gamma$ and its interior lie wholly inside of $G$.
By Cauchy’s Integral Formula we have
\begin{equation*}
g(f(z)) = \frac{1}{2 \pi i} \int_{\gamma} \frac{g(f(w))}{w-z} \, dw
\end{equation*}
for any $z$ in the interior of $\gamma$.
Writing
\begin{equation*}
Q(z) = \frac{f(z) - f(a)}{z - a}
\end{equation*}
for the difference quotient, we get
\begin{equation*}
g(Q(z)) = \frac{1}{2 \pi i} \int_{\gamma} \frac{g(f(w))}{(w-z)(w-a)} \, dw
\end{equation*}
and
\begin{equation*}
g \parens*{\frac{Q(z) - Q(z')}{z - z'}} = \frac{1}{2 \pi i} \int_{\gamma} \frac{g(f(w))}{(w-z)(w-z')(w-a)} \, dw
\end{equation*}
for all $z,z'$ in the interior of $\gamma$.
The family of vectors $f(w) \in X$, indexed by complex numbers $w$ on the contour $\gamma$, can be viewed as a family of bounded linear functionals $C(f(w)) : X' \to \CC$
via the canonical embedding $C : X \to X''$ of $X$ into its bidual. For every fixed $g \in X'$ the set of values $C(f(w))(g) = g(f(w))$ is bounded, because the function $g \circ f$ is continuous and the contour is compact.
In other words, the family of functionals $C(f(w))$, $w \in \gamma$, is pointwise bounded.
The Uniform Boundedness Theorem implies that there exists a constant $M > 0$ such that $\abs{g(f(w))} \le M \norm{g}$ for all $w$ on $\gamma$ and all $g \in X'$.
\begin{equation*}
\abs*{g \parens*{\frac{Q(z) - Q(z')}{z - z'}}} \le \frac{M}{2 \pi} \norm{g} \int_{\gamma} \frac{dw}{\abs{w-z}\abs{w-z'}\abs{w-a}}
\end{equation*}
If we restrict $z,z'$ to a neighborhood $N$ of $a$ that stays away from $\gamma$, then the integral on the right hand side is
bounded by a constant independent of $z$ and $z'$.
Absorbing all constants into $M' > 0$ we obtain
\begin{equation*}
\abs{g(Q(z) - Q(z'))} \le M' \norm{g} \abs{z-z'} \quad \forall z,z' \in N.
\end{equation*}
\begin{equation*}
\norm{Q(z) - Q(z')} = \sup_{\substack{g \in X'\\ \norm{g} \le 1}} \abs{g(Q(z) - Q(z'))} \le M' \norm{z - z'}.
\end{equation*}
Hence, the limit of $Q(z)$ for $z \to a$ exists by completeness of $X$.
\end{myproof}
\begin{definition}{Analytic Vector for an Unitary Group}{analytic-vector-unitary-group}
Let $\sigma : \RR \to U(\hilb{H})$ be a strongly continuous one-parameter unitary group on a complex Hilbert space $\hilb{H}$.
A vector $x \in \hilb{H}$ is said to be an \emph{analytic vector} for $\sigma$, if there exist
\begin{itemize}
\item a number $\lambda > 0$, defining a strip $I_{\lambda} = \braces{z : \abs{\Im z} < 1}$, and
\item a vector-valued function $f : I_{\lambda} \to \hilb{H}$,
\end{itemize}
with the properties that
\begin{itemize}
\item $f(t) = \sigma_t(x)$ for all $t \in \RR$,
\item $f$ is weakly analytic on $I_{\lambda}$.
\end{itemize}
In this case we write $f(z) = \sigma_z(x)$ for $z \in I_{\lambda}$.
\end{definition}
\begin{proposition}{}{}
Let $\sigma : \RR \to U(\hilb{H})$ be a strongly continuous one-parameter unitary group on a complex Hilbert space $\hilb{H}$
and let $A$ be its infinitesimal generator.
Then a vector $x \in \hilb{H}$ is analytic for $\sigma$ if and only if it is analytic for $A$.
\end{proposition}
\begin{myproof}
First, suppose that $x$ is an analytic vector for $\sigma$.
Then, there exist a number $\lambda > 0$ and a function $f : I_{\lambda} \to X$ as in \cref{definition:analytic-vector-unitary-group}.
In particular, $f$ is (strongly) analytic on the strip $I_{\lambda}$, which contains the disk $\braces{z : \abs{z} \le r}$ when $r \le \lambda$.
Hence we have Cauchy estimates
\begin{equation*}
\norm{f^{(n)}(0)} \le \frac{n!}{r^n} M \quad \forall n \in \NN,
\quad \text{where} \ M = \sup_{\abs{z} = r} \norm{f(z)}.
\end{equation*}
For real $t$ we have $f(t) = \sigma_t(x) = \exp(itA) x$ and
the mapping $t \mapsto f(t)$ is strongly differentiable with derivatives
$f^{(n)}(0) = (iA)^n x$.
This implies that the power series
\begin{equation*}
\sum_{n=0}^{\infty} \frac{\norm{A^n x}}{n!} t^n \le M \sum_{n=0}^{\infty} \frac{t^n}{r^n}
\end{equation*}
is convergent for $t \le \lambda$ by majorization. Hence $x$ is an analytic vector for the operator $A$.
Conversely, suppose that $x$ is analytic for the generator $A$ of $\sigma$.
Then, by \cref{definition:analytic-vector-operator}, $x$ lies in the domains of all powers $A^n$, $n \in \NN$, and the power series
\begin{equation*}
\sum_{n=0}^{\infty} \frac{\norm{A^n x}}{n!} z^n
\end{equation*}
has a positive radius of convergence $t>0$.
\end{myproof}
\chapterbib
\cleardoublepage
|
