1: \documentclass{ieot}

2: 

3: 

4: %\usepackage{epsf}

5: % (to be used by the editorial office if necessary)

6: %

7: 

8: % THEOREM Environments (Examples)------------------------------------------

9:  \newtheorem{thm}{Theorem}[section]

10:  \newtheorem{cor}[thm]{Corollary}

11:  \newtheorem{lem}[thm]{Lemma}

12:  \newtheorem{prop}[thm]{Proposition}

13:  \theoremstyle{definition}

14:  \newtheorem{defn}[thm]{Definition}

15:  \theoremstyle{remark}

16:  \newtheorem{rem}[thm]{Remark}

17:  \newtheorem{note*}[thm]{Note}

18:   \newtheorem{ex}{Example}

19:  \numberwithin{equation}{section}

20: 

21: \providecommand{\Real}{\mathop{\rm Re}\nolimits}%

22: \providecommand{\Imag}{\mathop{\rm Im}\nolimits}%

23: \providecommand{\clos}{\mathop{\rm clos}\nolimits}%

24: \providecommand{\id}{\mathop{\rm id}\nolimits}%

25: 

26: \usepackage{amssymb}

27: 

28: \begin{document}

29: % TOP MATTER

30: 

31: \title[Multivariable $\rho$-contractions]%

32:  {Multivariable $\rho$-contractions} % This is the full title of the paper

33: 

34: 

35: \author[Dmitry S. Kalyuzhny\u{\i}-Verbovetzki\u{\i}]{Dmitry S. Kalyuzhny\u{\i}-Verbovetzki\u{\i}}

36: 

37: \address{%

38:    Department of Mathematics\\

39:    The Weizmann Institute of Science\\

40:    Rehovot 76100\\

41:    Israel}

42: 

43: \email{dmitryk@wisdom.weizmann.ac.il}

44: 

45: \dedicatory{Dedicated to Israel Gohberg on his 75th birthday}

46: %----------classification, keywords, date

47: 

48: %Insert `2000 Mathematics Subject Classification' numbers here!

49: \subjclass{Primary 47A13; Secondary 47A20, 47A56}

50: 

51: \keywords{Multivariable, $\rho$-dilations, linear pencils of operators, operator radii, Agler kernels, similarity to a 1-contraction}

52: 

53: \begin{abstract}

54: We suggest a new version of the notion of $\rho$-dilation

55: ($\rho>0$) of an $N$-tuple $\mathbf{A}=(A_1,\ldots,A_N)$ of

56: bounded linear operators on a common Hilbert space. We say that

57: $\mathbf{A}$ belongs to the class $C_{\rho,N}$ if $\mathbf{A}$

58: admits a $\rho$-dilation

59: $\widetilde{\mathbf{A}}=(\widetilde{A}_1,\ldots,\widetilde{A}_N)$

60: for which

61: $\zeta\widetilde{\mathbf{A}}:=\zeta_1\widetilde{A}_1+\cdots

62: +\zeta_N\widetilde{A}_N$ is a unitary operator for each

63: $\zeta:=(\zeta_1,\ldots,\zeta_N)$ in the unit torus

64: $\mathbb{T}^N$. For $N=1$ this class coincides with the class

65: $C_\rho$ of B. Sz.-Nagy and C. Foia\c{s}. We generalize the known

66: descriptions of $C_{\rho,1}=C_\rho$ to the case of $C_{\rho,N},\

67: N>1$, using so-called Agler kernels. Also, the notion of operator

68: radii $w_\rho, \rho>0$, is generalized to the case of $N$-tuples

69: of operators, and to the case of bounded (in a certain strong

70: sense) holomorphic operator-valued functions in the open unit

71: polydisk $\mathbb{D}^N$, with preservation of all the most

72: important their properties. Finally, we show that for each

73: $\rho>1$ and $N>1$ there exists an $\mathbf{A}=(A_1,\ldots,A_N)\in

74: C_{\rho,N}$ which is not simultaneously similar to any

75: $\mathbf{T}=(T_1,\ldots,T_N)\in C_{1,N}$, however if

76: $\mathbf{A}\in C_{\rho,N}$ admits a uniform unitary

77: $\rho$-dilation then $\mathbf{A}$ is simultaneously similar to

78: some $\mathbf{T}\in C_{1,N}$.

79: \end{abstract}

80: 

81: 

82: 

83: \maketitle

84: 

85: \section{Introduction}

86: \label{intro}

87: 

88: \noindent Linear pencils of operators $L_\mathbf{A}(z):=A_0+z_1A_1+\cdots +z_NA_N$

89: on a Hilbert space which take contractive (resp., unitary or

90: $J$-unitary for some signature operator $J=J^*=J^{-1}$) values for

91: all $z=(z_1,\ldots,z_N)$ in the \emph{unit torus}

92: $\mathbb{T}^N:=\{\zeta\in\mathbb{C}^N:|\zeta_k|=1,\

93: k=1,\ldots,N\}$ serve as one of possible generalizations of a

94: single contractive (resp., unitary, $J$-unitary) operator on a

95: Hilbert space. They appear in  constructions of Agler's unitary

96: colligation and corresponding conservative (unitary) scattering

97: $N$-dimensional discrete-time linear system of Roesser type

98: \cite{Ag,BT}, and also of Fornasini--Marchesini type \cite{BSV},

99: and dissipative (contractive), conservative (unitary) or

100: $J$-conservative ($J$-unitary) scattering $N$-dimensional linear

101: systems of one more form introduced in our paper \cite{K2} and

102: studied in \cite{K2,K3,K4,K5,K6,K7,BSV}. These constructions, in

103: particular, provide the transfer function realization formulae for

104: certain classes of holomorphic functions

105: \cite{Ag,BT,K2,K4,K7,BSV}, the solutions to the Nevanlinna--Pick

106: interpolation problem \cite{AgMc,BT}, the Toeplitz corona problem

107: \cite{AgMc,BT}, and the commutant lifting problem \cite{BLTT} in

108: several variables.

109: 

110: In \cite{K3} we developed the dilation theory for multidimensional

111: linear systems, and in particular gave a necessary and sufficient

112: condition for such a system to have a conservative dilation. As a

113: special case, this gave a criterion for the existence of a

114: unitary dilation of a contractive (on $\mathbb{T}^N$) linear

115: pencil of operators on a Hilbert space. Linear pencils of

116: operators satisfying this criterion inherit the most important

117: properties of single contraction operators on a Hilbert space

118: (note that, due to \cite{K1}, not all linear pencils which take

119: contractive operator values on $\mathbb{T}^N$ satisfy this

120: criterion).

121: 

122: The purpose of the present paper is to develop the theory of

123: $\rho$-contractions in several variables in the framework of

124: ``linear pencils approach". We introduce the notion of

125: \emph{$\rho$-dilation} of an $N$-tuple

126: $\mathbf{A}=(A_1,\ldots,A_N)$ of bounded linear operators on a

127: common Hilbert space by means of a simultaneous $\rho$-dilation,

128: in the sense of B.~Sz.-Nagy and C.~Foia\c{s} \cite{SzNF1,SzNF}, of

129: the values of a homogeneous linear pencil of operators

130: $z\mathbf{A}:=\sum_{k=1}^Nz_kA_k$. The class $C_{\rho,N}$ consists

131: of those $N$-tuples of operators $\mathbf{A}=(A_1,\ldots,A_N)$

132: (\emph{$\rho$-contractions}) for which there exists a

133: $\rho$-dilation

134: $\widetilde{\mathbf{A}}=(\widetilde{A}_1,\ldots,\widetilde{A}_N)$

135: such that the operators

136: $\zeta\widetilde{\mathbf{A}}=\sum_{k=1}^N\zeta_k\widetilde{A}_k$

137: are unitary for all

138: $\zeta=(\zeta_1,\ldots,\zeta_N)\in\mathbb{T}^N$. On the one hand, this class

139: generalizes the class $C_{\rho,1}=C_\rho$ of

140: Sz.-Nagy and Foia\c{s} \cite{SzNF1,SzNF} consisting of operators

141: which admit a unitary $\rho$-dilation to the case $N>1$. On

142: the other hand, this class generalizes the class of $N$-tuples of operators $\mathbf{A}$

143: for which the associated linear pencil of operators $z\mathbf{A}$

144: admits a unitary dilation  in the sense of \cite{K3} (this

145: corresponds to $\rho=1$) to the case of $N$-tuples of operators

146: $\mathbf{A}$ which have a unitary $\rho$-dilation for $\rho\neq

147: 1$.

148: 

149: The paper is organized as follows. Section~\ref{sec:prelim} gives

150: preliminaries on $\rho$-contracti\-ons for the case $N=1$. Namely,

151: we recall the relevant definitions, the known criteria for an

152: operator to be a $\rho$-contraction, i.e., to belong to the class

153: $C_{\rho}$ of Sz.-Nagy and Foia\c{s}, the notion of operator radii

154: $w_\rho$ and their properties, and the theorem on similarity of

155: $\rho$-contractions to contractions. In Section~\ref{sec:n-rho} we

156: give the definitions of a $\rho$-dilation of an $N$-tuple of

157: operators, and of the class $C_{\rho,N}$ of $\rho$-contractions for

158: the case $N>1$, and prove a theorem which generalizes the

159: criteria of $\rho$-contractiveness to this case, as well as to the

160: case $0<\rho\neq 1$. Some properties of classes $C_{\rho,N}$ are

161: discussed. Then it is shown that the notions of a $\rho$-contraction

162: and of the corresponding class $C_{\rho,N}$, as well as the 

163: theorem just mentioned, can be extended to holomorphic functions on the \emph{open unit polydisk}

164: $\mathbb{D}^N:=\{z\in\mathbb{C}^N:|z_k|<1,\ k=1,\ldots,N\}$ that are bounded in a certain

165: strong sense,

166: though the notion of unitary $\rho$-dilation is not relevant any

167: more in this case. In Section~\ref{sec:n-rad} we define operator

168: radii $w_{\rho,N}$ of $N$-tuples of operators, and

169: operator-function radii $w_{\rho,N}^{(\infty)}$ of bounded

170: holomorphic functions on $\mathbb{D}^N,\ \rho>0$. These radii

171: generalize $w_\rho$'s and inherit all the most important

172: properties of them. In Section~\ref{sec:similar} we prove that for

173: each $\rho>1$ and $N>1$ there exists an

174: $\mathbf{A}=(A_1,\ldots,A_N)\in C_{\rho,N}$ which is not

175: simultaneously similar to any $\mathbf{T}=(T_1,\ldots,T_N)\in

176: C_{1,N}$. Then we introduce the classes $C_{\rho,N}^u,\ \rho>0$,

177: of $N$-variable $\rho$-contractions $\mathbf{A}=(A_1,\ldots,A_N)$

178: which admit a uniform unitary $\rho$-dilation. We prove that if

179: $\mathbf{A}\in C_{\rho,N}^u$ for some $\rho>1$ then $\mathbf{A}$

180: is simultaneously similar to some $\mathbf{T}\in C_{1,N}^u$. Note,

181: that since the class $C_{\rho,N}^u$ (as well as $C_{\rho,N}$)

182: increases as a function of $\rho$, for any $\rho\leq 1$ an

183: $\mathbf{A}\in C_{\rho,N}^u$ (resp., $\mathbf{A}\in C_{\rho,N}$)

184: belongs to $C_{1,N}^u$ (resp., $C_{1,N}$) itself. We show the

185: relation of our results to ones of G.~Popescu \cite{Po2} where a

186: different notion of multivariable $\rho$-contractions has been

187: introduced, and the relevant theory has been developed. The

188: classes $C_{\rho,N}^u,\ \rho>0$, which appear in

189: Section~\ref{sec:similar} in connection with the similarity

190: problem discussed there, certainly deserve a further

191: investigation.

192: 

193: \section{Preliminaries}\label{sec:prelim}

194: 

195: Let $L(\mathcal{X,Y})$ denote the Banach space of bounded linear

196: operators mapping a  Hilbert space $\mathcal{X}$ into a Hilbert

197: space $\mathcal{Y}$, and $L(\mathcal{X}):=L(\mathcal{X,X})$. For

198: $\rho>0$, an operator $\widetilde{A}\in

199: L(\widetilde{\mathcal{X}})$ is said to be a \emph{$\rho$-dilation

200: of an operator} $A\in L(\mathcal{X})$ if

201: $\widetilde{\mathcal{X}}\supset\mathcal{X}$ and

202: \begin{equation}\label{eq:rhodil}

203: A^n=\rho P_\mathcal{X}\widetilde{A}^n|\mathcal{X},\quad

204: n\in\mathbb{N},

205: \end{equation}

206:  where $P_\mathcal{X}$ denotes the orthogonal projection onto the

207: subspace $\mathcal{X}$ in $\widetilde{\mathcal{X}}$. If, moreover,

208: $\widetilde{A}$ is a unitary operator then $\widetilde{A}$ is

209: called a \emph{unitary $\rho$-dilation of $A$}. In \cite{SzNF1}

210: (see also \cite{SzNF}) B.~Sz.-Nagy and C.~Foia\c{s} introduced the

211: classes $C_\rho,\ \rho>0$, consisting of operators which admit a

212: unitary $\rho$-dilation. Due to B.~Sz.-Nagy \cite{SzN}, the class

213: $C_1$ is precisely the class of all contractions, i.e., operators

214: $A$ such that $\| A\|\leq 1$. C.~A.~Berger \cite{Be} showed that

215: the class $C_2$ is precisely the class of all operators $A\in

216: L(\mathcal{X})$, for some Hilbert space $\mathcal{X}$, which have

217: the \emph{numerical radius}

218: \[ w(A)=\sup\{ |\langle Ax,x\rangle |:\ x\in\mathcal{X},\ \| x\|

219: =1\}\] equal to at most one. Thus, the classes $C_\rho,\ \rho>0$,

220: provide a framework for simultaneous investigation of these two

221: important classes of operators.

222: 

223: Recall that the \emph{Herglotz} (or \emph{Caratheodory})

224: \emph{class} $\mathcal{H(X)}$ (respectively, the \emph{Schur

225: class} $\mathcal{S(X)}$) consists of holomorphic functions $f$ on

226: the open unit disk $\mathbb{D}$ which take values in

227: $L(\mathcal{X})$ and satisfy $\Real f(z)=f(z)+f(z)^*\succeq 0$ in

228: the sense of positive semi-definiteness of an operator (resp., $\|

229: f(z)\|\leq 1)$ for all $z\in\mathbb{D}$. Let us recall some known

230: characterizations of the classes $C_\rho$.

231: \begin{thm}\label{thm:1-char}

232: Let $A\in L(\mathcal{X})$ and $\rho>0$. The following statements

233: are equivalent:

234: \begin{description}

235:    \item[(i)] $A\in C_\rho$;

236:     \item[(ii)] the function $k_\rho^A(z,w):=\rho

237:     I_\mathcal{X}-(\rho-1)\left((zA+(wA)^*\right)+(\rho-2)(wA)^*zA$ satisfies

238: $k_\rho^A(z,z)\succeq 0$ for all $z\in\clos(\mathbb{D})$;

239:     \item[(iii)] the function

240:     $\psi_\rho^A(z):=(1-\frac{2}{\rho})I_\mathcal{X}+\frac{2}{\rho}(I_\mathcal{X}-zA)^{-1}$

241:     belongs to $\mathcal{H(X)}$;

242:     \item[(iv)] the function

243:     $\varphi_\rho^A(z):=zA\left((\varrho-1)zA-\rho I_\mathcal{X}\right)^{-1}$

244:     belongs to $\mathcal{S(X)}$.

245: \end{description}

246: \end{thm}

247: Conditions (ii) and (iii) of Theorem~\ref{thm:1-char} each

248: characterizing the class $C_\rho$ appear in \cite{SzNF1}, while

249: condition (iv) is due to C.~Davis \cite{Da}.

250: \begin{cor}\label{cor:1-char'}

251:  Condition $(${\rm ii}$)$ in Theorem~\ref{thm:1-char} can be replaced

252: by \vspace{2mm}

253: \begin{description}

254:     \item[(ii')] $k_\rho^A(C,C):=\rho I_\mathcal{X}\otimes

255:     I_{\mathcal{H}_C}-(\rho-1)(A\otimes C+

256:     (A\otimes

257:     C)^*)\\+(\rho-2)(A\otimes C)^*(A\otimes C)\succeq 0$\\ for any

258:     contraction $C$ on a Hilbert space $\mathcal{H}_C$.

259: \end{description}

260: \end{cor}

261: \begin{proof} Indeed, (ii')$\Rightarrow$(ii), hence

262: (ii')$\Rightarrow$(i). Conversely, if $A\in C_\rho\cap

263: L(\mathcal{X})$ then for any contraction $C$ on $\mathcal{H}_C$

264: one has $A\otimes C\in C_\rho$ because, by \cite{SzN}, $C$ admits

265: a unitary dilation $\widetilde{C}$, and $A$ admits a unitary

266: $\rho$-dilation $\widetilde{A}$, thus

267: $\widetilde{A}\otimes\widetilde{C}$ is a unitary $\rho$-dilation

268: of $A\otimes C$:

269: \begin{eqnarray*}

270: (A\otimes C)^n &=& A^n\otimes C^n= (\rho

271: P_\mathcal{X}\widetilde{A}^n|\mathcal{X})\otimes

272: (P_{\mathcal{H}_C}\widetilde{C}^n|\mathcal{H}_C)\\

273: &=& \rho P_{\mathcal{X\otimes

274: H}_C}(\widetilde{A}^n\otimes\widetilde{C}^n)|\mathcal{X\otimes

275: H}_C\\

276:  &=& \rho P_{\mathcal{X\otimes

277: H}_C}(\widetilde{A}\otimes\widetilde{C})^n|\mathcal{X\otimes

278: H}_C,\quad n\in\mathbb{N}.

279: \end{eqnarray*}

280: Therefore, $k_\rho^A(C,C)=k_\rho^{A\otimes C}(1,1)\succeq 0$,

281: i.e., (ii') is valid.

282: \end{proof}

283: \begin{cor}\label{cor:ac}

284:  Condition \vspace{2mm}

285: \begin{description}

286:     \item[(v)] $A\otimes C\in C_\rho$ for any

287:     contraction $C$ on a Hilbert space,

288: \end{description}\vspace{2mm}

289:  is equivalent to each of conditions $(${\rm i}$)$--$(${\rm iv}$)$ of

290: Theorem~\ref{thm:1-char}.

291: \end{cor}

292: \begin{proof}

293: See the proof of Corollary~\ref{cor:1-char'}.

294: \end{proof}

295:  Any operator $A\in C_\rho$ is

296: \emph{power-bounded}:

297: \begin{equation}\label{eq:1-pb}

298:  \| A^n\|\leq \rho,\quad n\in\mathbb{N},

299: \end{equation}

300: moreover, its \emph{spectral radius}

301: \begin{equation}\label{eq:1-sp-rad}

302: \nu(A)=\lim_{n\rightarrow+\infty}\| A^n\|^{\frac{1}{n}}

303: \end{equation}

304:  is at most

305: one. In \cite{SzNF1} an example of a power-bounded operator which

306: is not contained in any of the classes $C_\rho,\ \rho>0$, is

307: given. However, J.~A.~R.~Holbrook \cite{H1} showed that any

308: bounded linear operator $A$ with $\nu(A)\leq 1$ can be

309: approximated in the operator norm topology by elements of the

310: classes $C_\rho$. More precisely, if $C_\infty$ denotes the class

311: of bounded linear operators with spectral radius at most one, and

312: $\mathcal{X}$ is a Hilbert space, then

313: \begin{equation}\label{eq:1-clos}

314:  C_\infty\cap

315: L(\mathcal{X})=\clos\left\{\bigcup_{0<\rho<\infty}C_\rho\cap

316: L(\mathcal{X})\right\}.

317: \end{equation}

318: 

319: For a fixed Hilbert space $\mathcal{X}$, the class $C_\rho$ as a

320: function of $\rho$ increases \cite{SzNF1}:

321: \begin{equation}\label{eq:incr}

322: C_\rho\subset C_{\rho'}\ \mbox{for}\ \rho<\rho'.

323: \end{equation}

324: Moreover, it was shown by E.~Durszt \cite{Du} that $C_\rho$

325: increases strictly for $\dim\mathcal{X}\geq 2$:

326: \[ C_\rho\neq C_{\rho'}\ \mbox{for}\ \rho\neq\rho'.\]

327: \begin{prop}\label{prop:1c}

328: For $\mathcal{X}=\mathbb{C}$, the classes $C_\rho$ coincide for

329: all $\rho\geq 1$, and strictly increase for $0<\rho<1$:

330: \[ C_\rho\varsubsetneq C_{\rho'}\ \mbox{for}\ 0<\rho<\rho'\leq

331: 1.\]

332: \end{prop}

333: \begin{proof}

334: If $a\in\mathbb{C}\cong L(\mathbb{C})$ belongs to $C_\rho$ then

335: $\|a\|=|a| =\nu(a)\leq 1$. Hence $C_\rho\subset C_1$ for any

336: $\rho>0$. Since (\ref{eq:incr}) implies $C_\rho\supset C_1$ for

337: $\rho\geq 1$, we get $C_\rho=C_1$ for this case, that proves the

338: first part of this proposition.

339: 

340: For the proof of the second part, we will show that for any

341: $\varepsilon,\rho:0<\varepsilon<\rho<1$, one has

342: \begin{equation}\label{eq:a}

343:  a:=\frac{\rho}{2-\rho}\in C_\rho\backslash

344: C_{\rho-\varepsilon}.

345: \end{equation}

346:  If $0\leq\varepsilon<\rho$ then, by condition (ii) in

347: Theorem~\ref{thm:1-char}, the inclusion $a\in

348: C_{\rho-\varepsilon}$ is equivalent to

349: \[

350: \rho-\varepsilon-(\rho-\varepsilon-1)(az+\bar{a}\bar{z})+(\rho-\varepsilon-2)|az|^2\geq

351: 0,\quad z\in\clos(\mathbb{D}),\] which for $a=\frac{\rho}{2-\rho}$

352: turns into

353: \[\rho-\varepsilon-2(\rho-\varepsilon-1)\frac{\rho}{2-\rho}r\cos\theta+(\rho-\varepsilon-2)\left(\frac{\rho r}{2-\rho}\right)^2\geq

354: 0,\quad r\in [0,1],\theta\in [0,2\pi).\] Since

355: $\rho-\varepsilon-1<0$, the left-hand side of this inequality, as

356: a function of $\theta$ for a fixed $r$, has a minimum at

357: $\theta=\pi$, so the latter condition turns into

358: \[\rho-\varepsilon+2(\rho-\varepsilon-1)\frac{\rho r}{2-\rho}+(\rho-\varepsilon-2)\left(\frac{\rho r}{2-\rho}\right)^2\geq

359: 0,\quad r\in [0,1].\] The left-hand side attains its minimum at

360: $r=1$, thus the latter inequality turns into

361: \[\rho-\varepsilon+2(\rho-\varepsilon-1)\frac{\rho}{2-\rho}+(\rho-\varepsilon-2)\left(\frac{\rho}{2-\rho}\right)^2=-\frac{4\varepsilon}{(2-\rho)^2}\geq

362: 0,\] which is possible if and only if $\varepsilon=0$. Thus,

363: (\ref{eq:a}) is true.

364: \end{proof}

365: The properties of the classes $C_\rho$ become more clear due to

366: the following numerical characteristics of operators.

367: J.~A.~R.~Holbrook \cite{H1} and J.~P.~Williams \cite{W},

368: independently, introduced for any $A\in L(\mathcal{X})$ the

369: \emph{operator radii}

370: \begin{equation}\label{eq:1-rad}

371:  w_\rho(A):=\inf\{ u>0:\frac{1}{u}A\in C_\rho\}.

372: \end{equation}

373: \begin{thm}\label{thm:1-rad}

374: $w_\rho(\cdot)$ has the following properties:

375: \begin{description}

376:     \item[(i)] $w_\rho(A)<\infty$;

377:     \item[(ii)] $w_\rho(A)>0$ unless $A=0$, moreover,

378:     $w_\rho(A)\geq\frac{1}{\rho}\| A\|$;

379:     \item[(iii)] $\forall\mu\in\mathbb{C},\quad w_\rho(\mu

380:     A)=|\mu|w_\rho(A)$;

381:     \item[(iv)] $w_\rho(A)\leq 1$ if and only if $A\in C_\rho$;

382:     \item[(v)] $w_\rho(\cdot)$ is a norm on $L(\mathcal{X})$ for any $\rho:\ 0<\rho\leq 2$,

383:     and not a norm on $L(\mathcal{X}),\ \dim\mathcal{X}\geq 2$, for any

384:     $\rho>2$;

385:     \item[(vi)] $w_1(A)=\| A\|$ (of course, here $\|\cdot\|$ is the

386:     operator norm on $L(\mathcal{X})$ with respect to the

387:     Hilbert-space metric on $\mathcal{X}$);

388:     \item[(vii)] $w_2(A)=w(A)$;

389:     \item[(viii)] $w_\infty(A):=\lim\limits_{\rho\rightarrow

390:     +\infty}w_\rho(A)=\nu(A)$;

391:     \item[(ix)] $w_\rho(I_\mathcal{X})=\left\{\begin{array}{ll}

392:       1 & \mbox{for}\ \rho\geq 1, \\

393:       \frac{2}{\rho}-1 & \mbox{for}\ 0<\rho<1;

394:     \end{array}\right.$

395:     \item[(x)] if $0<\rho<\rho'$ then $w_{\rho'}(A)\leq

396:     w_\rho(A)\leq\left(\frac{2\rho'}{\rho}-1\right)w_{\rho'}(A)$,

397:     thus $w_\rho(A)$ is continuous in $\rho$ and non-increasing as

398:     $\rho$ increases;

399:     \item[(xi)] if $\| A\|=1$ and $A^2=0$ then, for any $\rho>0$,

400:     $w_\rho(A)=\frac{1}{\rho}$;

401:     \item[(xii)] if for some $\rho_0$ one has $w_{\rho_0}(A)>w_\infty(A)\

402:     (=\nu(A))$ then for any $\rho>\rho_0$ one has

403:     $w_{\rho_0}(A)>w_\rho(A)$;

404:     \item[(xiii)] $\lg w_\rho(A)$ is a convex function in $\rho,\

405:     0<\rho<+\infty$;

406:     \item[(xiv)] $w_\rho(A)$ is a convex function in $\rho,\

407:     0<\rho<+\infty$;

408:     \item[(xv)] the function $h_A(\rho):=\rho w_\rho(A)$ is

409:     non-decreasing on $[1,+\infty)$, and non-increa\-sing on

410:     $(0,1)$;

411:     \item[(xvi)] for any $\rho$ such that $0<\rho<2$ one has $\rho

412:     w_\rho(A)=(2-\rho)w_{2-\rho}(A)$, and $\lim\limits_{\rho\downarrow

413:     0}\frac{\rho}{2}w_\rho(A)=w_2(A)\ (=w(A))$;

414:     \item[(xvii)] $\forall\rho:\ 0<\rho\leq 1,\quad

415:     w_\rho(A)\geq\left(\frac{2}{\rho}-1\right)w_2(A)$;

416:     \item[(xviii)] $\forall A,B\in L(\mathcal{X}),\ \forall\rho\geq 1,\quad

417:     w_\rho(AB)\leq\rho^2w_\rho(A)w_\rho(B)$, moreover, $\rho^2$ is

418:     the best constant in this inequality for the case $\dim\mathcal{X}\geq

419:     2$;

420:     \item[(xix)] $\forall A,B\in L(\mathcal{X}),\ \forall\rho:\ 0<\rho<1,\quad

421:     w_\rho(AB)\leq(2-\rho)\rho w_\rho(A)w_\rho(B)$, moreover, $(2-\rho)\rho$ is

422:     the best constant in this inequality for the case $\dim\mathcal{X}\geq

423:     2$;

424:     \item[(xx)] $\forall\rho>0,\ \forall n\in\mathbb{N},\quad

425:     w_\rho(A^n)\leq w_\rho(A)^n$.

426: \end{description}

427: \end{thm}

428: Properties (i)--(xii), (xviii), and (xx) were proved by

429: J.~A.~R.~Holbrook \cite{H1}, properties (xiii)--(xvi) were

430: discovered by T.~Ando and K.~Nishio \cite{AnN}. Property (xix) was

431: shown by K.~Okubo and T.~Ando \cite{OAn}, and follows also from

432: (xvi) and (xviii). Finally, property (xvii) easily follows from

433: (x) and (xvi). Indeed, for $0<\rho\leq 1$ one has

434: $w_{2-\rho}(A)\geq w_2(A)$, hence $\rho

435: w_\rho(A)=(2-\rho)w_{2-\rho}(A)\geq(2-\rho)w_2(A)$, which implies

436: (xvii).

437: 

438: We have listed in Theorem~\ref{thm:1-rad} only the most important,

439: as it seems to us, properties of operator radii $w_\rho(\cdot)$.

440: Other properties of $w_\rho(\cdot)$ can be found in

441: \cite{H1,H2,FH,AnN,OAn,BaCa} and elsewhere.

442: 

443: Let us note that properties of the classes $C_\rho$ discussed

444: before Theorem~\ref{thm:1-rad}, including

445: Proposition~\ref{prop:1c}, can be deduced from properties (iv),

446: (vi)--(x) in Theorem~\ref{thm:1-rad}. Due to property (iv) in

447: Theorem~\ref{thm:1-rad}, operators from the classes $C_\rho$ are

448: called \emph{$\rho$-contractions}.

449: 

450: Any $A\in C_\rho$ satisfies the following \emph{generalized von

451: Neumann inequality} \cite{SzNF1}: for any polynomial $p$ of one

452: variable

453: \begin{equation}\label{eq:vn}

454: \| p(A)\|\leq\max_{|z|\leq 1}|\rho p(z)+(1-\rho)p(0)|.

455: \end{equation}

456: Let $A\in L(\mathcal{X}), B\in L(\mathcal{Y})$. Then $A$ is said

457: to be \emph{similar} to $B$ if there exists a bounded invertible

458: operator $S\in L(\mathcal{X,Y})$ such that

459: \begin{equation}\label{eq:1-similar}

460: A=S^{-1}BS.

461: \end{equation}

462:  B.~Sz.-Nagy and C.~Foia\c{s} proved in \cite{SzNF2}

463: (see also \cite{SzNF}) that any $A\in C_\rho$ is similar to some

464: $T\in C_1$, i.e., any $\rho$-contraction is similar to a

465: contraction.

466: 

467: To conclude this section, let us remark that the classes

468: $C_\rho$ are of continuous interest, e.g., see recent works

469: \cite{DMcCW,CaF,BaCa,NO,OS}. In \cite{Po2} the classes $C_\rho$

470: were extended to a multivariable setting; we shall discuss this generalization in

471: Section~\ref{sec:similar}.

472: 

473: \section{The classes $C_{\rho,N}$}\label{sec:n-rho}

474: Let $\rho>0$. We will say that an $N$-tuple of operators

475: $\widetilde{\mathbf{A}}=(\widetilde{A}_1,\ldots,\widetilde{A}_N)\in

476: L(\widetilde{\mathcal{X}})^N$ is a \emph{$\rho$-dilation of an

477: $N$-tuple of operators} $\mathbf{A}=(A_1,\ldots,A_N)\in

478: L(\mathcal{X})^N$ if $\widetilde{\mathcal{X}}\supset\mathcal{X}$,

479: and for any $z=(z_1,\ldots,z_N)\in\mathbb{C}^N$ the operator

480: $z\widetilde{\mathbf{A}}=\sum_{k=1}^Nz_k\widetilde{A}_k$ is a

481: $\rho$-dilation, in the sense of \cite{SzNF1}, of the operator

482: $z\mathbf{A}=\sum_{k=1}^Nz_kA_k$, i.e.,

483: \begin{equation}\label{eq:n-rhodil}

484:  (z\mathbf{A})^n=\rho

485: P_\mathcal{X}(z\widetilde{\mathbf{A}})^n|\mathcal{X},\quad

486: z\in\mathbb{C}^N,\ n\in\mathbb{N}.

487: \end{equation}

488:  These relations are equivalent to

489: \begin{equation}\label{eq:n-rhodil'}

490:  \mathbf{A}^t=\rho

491: P_\mathcal{X}\widetilde{\mathbf{A}}^t|\mathcal{X},\quad

492: t\in\mathbb{Z}^N_+:=\{\tau\in\mathbb{Z}^N:\tau_k\geq 0,\

493: k=1,\ldots,N\},

494: \end{equation}

495:  where $\mathbf{A}^t,\ t\in\mathbb{Z}^N_+$, are \emph{symmetrized

496: multi-powers of} $\mathbf{A}$:

497: \[\mathbf{A}^t:=\frac{t!}{|t|!}\sum_\sigma A_{[\sigma(1)]}\cdots

498: A_{[\sigma(|t|)]},\] and analogously for $\widetilde{\mathbf{A}}$.

499: Here for a multi-index $t=(t_1,\ldots,t_N),\

500: t!:=t_1!\cdots t_N!$ and $|t|:=t_1+\cdots+t_N$;

501: $\sigma$ runs over the set of all permutations with repetitions in a string of $|t|$ numbers from the set $\{ 1,\ldots,N\}$ such that the

502: $\kappa$-th number $[\kappa]\in\{1,\ldots,N\}$ appears in this string

503: $t_{[\kappa]}$ times. Say, if $t=(1,2,0,\ldots,0)$ then

504: \[ \mathbf{A}^t=\frac{A_1A_2^2+A_2A_1A_2+A_2^2A_1}{3}.\]

505: In the case of a commutative $N$-tuple $\mathbf{A}$ one has

506: $\mathbf{A}^t=A_1^{t_1}\cdots A_N^{t_N}$, i.e., a

507: usual multi-power. \begin{note*} Compare (\ref{eq:n-rhodil}) and

508: (\ref{eq:n-rhodil'}) with (\ref{eq:rhodil}).

509: \end{note*}

510: In the case $\rho=1$ the notion of $\rho$-dilation of an $N$-tuple

511: of operators $\mathbf{A}=(A_1,\ldots,A_N)$ coincides with the

512: notion of dilation of $\mathbf{A}$ (or corresponding linear pencil

513: $z\mathbf{A}$) as defined in \cite{K3}.

514: 

515: We will call $\widetilde{\mathbf{A}}\in

516: L(\widetilde{\mathcal{X}})^N$ a \emph{unitary $\rho$-dilation of

517: $\mathbf{A}\in L(\mathcal{X})^N$} if $\widetilde{\mathbf{A}}$ is a

518: $\rho$-dilation of $\mathbf{A}$ and for any $\zeta\in\mathbb{T}^N$

519: the operator

520: $\zeta\widetilde{\mathbf{A}}=\sum_{k=1}^N\zeta_k\widetilde{A}_k$

521: is unitary. The class of operator $N$-tuples which admit a unitary

522: $\rho$-dilation will be denoted by $C_{\rho,N}$.

523: 

524: Let $\mathcal{C}^N$ denote the family of all $N$-tuples

525: $\mathbf{C}=(C_1,\ldots,C_N)$ of \emph{commuting strict

526: contractions} on a common Hilbert space $\mathcal{H}_\mathbf{C}$,

527: i.e., $C_kC_j=C_jC_k$ and $\| C_k\|<1$ for all $k,j\in\{

528: 1,\ldots,N\}$. An $L(\mathcal{X})$-valued function

529: \[

530: k(z,w)=\sum_{(t,s)\in\mathbb{Z}^N_+\times\mathbb{Z}^N_+}\hat{k}(t,s)\bar{w}^sz^t,\quad

531: (z,w)\in\mathbb{D}^N\times\mathbb{D}^N,\] which is holomorphic in

532: $z\in\mathbb{D}^N$ and anti-holomorphic in $w\in\mathbb{D}^N$,

533: will be called an \emph{Agler kernel} if

534: \begin{equation}\label{eq:n-calc}

535: k(\mathbf{C},\mathbf{C}):=\sum_{(t,s)\in\mathbb{Z}^N_+\times\mathbb{Z}^N_+}

536: \hat{k}(t,s)\otimes\mathbf{C}^{*s}\mathbf{C}^t\succeq 0,\quad

537: \mathbf{C}\in\mathcal{C}^N,

538: \end{equation}

539:   where the series converges in the

540: operator norm topology on $L(\mathcal{X\otimes H}_\mathbf{C})$.

541: The \emph{Agler--Herglotz class $\mathcal{AH}_N(\mathcal{X})$}

542: (resp., the \emph{Agler--Schur class

543: $\mathcal{AS}_N(\mathcal{X})$}) is the class of all

544: $L(\mathcal{X})$-valued functions $f$ holomorphic on

545: $\mathbb{D}^N$ for which $k(z,w)=f(z)+f(w)^*$ (resp.,

546: $k(z,w)=I_\mathcal{X}-f(w)^*f(z)$) is an Agler kernel. Agler

547: kernels, as well as the classes $\mathcal{AH}_N(\mathcal{X})$ and

548: $\mathcal{AS}_N(\mathcal{X})$, were defined and studied by

549: J.~Agler in \cite{Ag}. The von Neumann inequality \cite{vN}

550: implies that $\mathcal{AS}_1(\mathcal{X})=\mathcal{S(X)}$ and

551: $\mathcal{AH}_1(\mathcal{X})=\mathcal{H(X)}$.

552: \begin{rem}\label{rem:a-ker}

553:  The function $k_\rho^A(z,w)$ from condition (ii) in

554: Theorem~\ref{thm:1-char}, due to Corollary~\ref{cor:1-char'}, is

555: an Agler kernel ($N=1$).

556: \end{rem}

557: \begin{thm}\label{thm:n-char}

558: Let $\mathbf{A}\in L(\mathcal{X})^N,\ \rho>0$. The following

559: conditions are equivalent:

560: \begin{description}

561:     \item[(i)] $\mathbf{A}\in C_{\rho,N}$;

562:     \item[(ii)] the function $k_{\rho,N}^\mathbf{A}(z,w):=\rho

563: I_\mathcal{X}-(\rho-1)\left((z\mathbf{A}+(w\mathbf{A})^*\right)+(\rho-2)(w\mathbf{A})^*z\mathbf{A}$

564: is an Agler kernel on $\mathbb{D}^N\times\mathbb{D}^N$;

565:     \item[(iii)] the function

566: $\psi_{\rho,N}^\mathbf{A}(z):=(1-\frac{2}{\rho})I_\mathcal{X}+\frac{2}{\rho}(I_\mathcal{X}-z\mathbf{A})^{-1}$

567:     belongs to $\mathcal{AH}_N(\mathcal{X})$;

568:     \item[(iv)] the function

569:     $\varphi_{\rho,N}^\mathbf{A}(z):=z\mathbf{A}\left((\varrho-1)z\mathbf{A}-\rho I_\mathcal{X}\right)^{-1}$

570:     belongs to $\mathcal{AS}_N(\mathcal{X})$;

571:     \item[(v)] $\mathbf{A\otimes C}:=\sum_{k=1}^NA_k\otimes C_k\in

572:     C_\rho=C_{\rho,1}$ for all $\mathbf{C}\in\mathcal{C}^N.$

573: \end{description}

574: \end{thm}

575: \begin{rem}\label{rem:general}

576: This theorem generalizes Theorem~\ref{thm:1-char} with condition

577: (ii) replaced by condition (ii') from Corollary~\ref{cor:1-char'},

578: and added condition (v) from Corollary~\ref{cor:ac}.

579: \end{rem}

580: \begin{proof}[Proof of Theorem~\ref{thm:n-char}]

581: (i)$\Leftrightarrow$(iii). The proof of this part combines the

582: idea of B.~Sz.-Nagy and C.~Foia\c{s} \cite{SzNF1} for the proof of

583: the equivalence (i)$\Leftrightarrow$(iii) in

584: Theorem~\ref{thm:1-char} (see Remark~\ref{rem:general}) with

585: Agler's representation of functions from

586: $\mathcal{AH}_N(\mathcal{X})$ \cite{Ag}. Let

587: $\mathbf{A}=(A_1,\ldots,A_N)\in C_{\rho,N}\cap L(\mathcal{X})^N$,

588: and

589: $\widetilde{\mathbf{A}}=(\widetilde{A}_1,\ldots,\widetilde{A}_N)\in

590: L(\widetilde{\mathcal{X}})^N$ be a unitary $\rho$-dilation of

591: $\mathbf{A}$. By Corollary~4.3 in \cite{K3}, the linear function

592: $L_{\widetilde{\mathbf{A}}}(z)=z\widetilde{\mathbf{A}}$ belongs to

593: the class $\mathcal{AS}_N(\widetilde{\mathcal{X}})$. Since for any

594: $\mathbf{C}\in\mathcal{C}^N$ one has

595: $(1+\varepsilon)\mathbf{C}\in\mathcal{C}^N$ for a sufficiently

596: small $\varepsilon>0$, the operator

597: $\widetilde{\mathbf{A}}\otimes\mathbf{C}$, as well as

598: $\widetilde{\mathbf{A}}\otimes(1+\varepsilon)\mathbf{C}$, is

599: contractive. Thus, $\widetilde{\mathbf{A}}\otimes\mathbf{C}$ is a

600: strict contraction, and the series

601: \[

602: I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}+2\sum_{n=1}^\infty(\widetilde{\mathbf{A}}\otimes\mathbf{C})^n\]

603: converges in the

604: $L(\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C})$-norm to

605: \[

606: (I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}+\widetilde{\mathbf{A}}\otimes\mathbf{C})

607: (I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}-\widetilde{\mathbf{A}}\otimes\mathbf{C})^{-1}.\]

608: Moreover,

609: \begin{equation}\label{eq:pos-real}

610:     \Real[(I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}+\widetilde{\mathbf{A}}\otimes\mathbf{C})

611: (I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}-\widetilde{\mathbf{A}}\otimes\mathbf{C})^{-1}]\succeq

612: 0.

613: \end{equation}

614: Therefore,

615: \begin{eqnarray*}

616: \lefteqn{P_{\mathcal{X\otimes

617: H}_\mathbf{C}}(I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}+\widetilde{\mathbf{A}}\otimes\mathbf{C})

618: (I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}-\widetilde{\mathbf{A}}\otimes\mathbf{C})^{-1}|\mathcal{X\otimes

619: H}_\mathbf{C} }\\

620: &=&\left. P_{\mathcal{X\otimes

621: H}_\mathbf{C}}\left(I_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}+2\sum_{n=1}^\infty(\widetilde{\mathbf{A}}\otimes\mathbf{C})^n\right)\right|

622: \mathcal{X\otimes

623: H}_\mathbf{C}\\

624: &=&

625: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}+2\sum_{n=1}^\infty\sum_{|t|=n}\frac{n!}{t!}(P_\mathcal{X}\otimes

626: I_{\mathcal{H}_\mathbf{C}})(\widetilde{\mathbf{A}}^t\otimes\mathbf{C}^t)|\mathcal{X\otimes

627: H}_\mathbf{C}\\

628:  &=&

629: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}+\frac{2}{\rho}\sum_{n=1}^\infty\sum_{|t|=n}\frac{n!}{t!}\mathbf{A}^t\otimes\mathbf{C}^t

630: =

631: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}+\frac{2}{\rho}\sum_{n=1}^\infty(\mathbf{A\otimes

632: C})^n \\

633: &=&

634: (1-\frac{2}{\rho})I_{\mathcal{X}\otimes\mathcal{H}}+\frac{2}{\rho}(I_{\mathcal{X}\otimes\mathcal{H}}-\mathbf{A\otimes

635: C})^{-1}=\psi_{\rho,N}^\mathbf{A}(\mathbf{C}),

636: \end{eqnarray*}

637: and (\ref{eq:pos-real}) implies

638: $\Real\psi_{\rho,N}^\mathbf{A}(\mathbf{C})\succeq 0$. Since the

639: function

640: $(I_{\widetilde{\mathcal{X}}}+z\widetilde{\mathbf{A}})(I_{\widetilde{\mathcal{X}}}-z\widetilde{\mathbf{A}})^{-1}$

641: is well-defined and holomorphic on $\mathbb{\mathbb{D}}^N$, so is

642: \begin{equation}\label{eq:psi}

643: \psi_{\rho,N}^\mathbf{A}(z)=P_\mathcal{X}(I_{\widetilde{\mathcal{X}}}+

644: z\widetilde{\mathbf{A}})(I_{\widetilde{\mathcal{X}}}-z\widetilde{\mathbf{A}})^{-1}|\mathcal{X},\quad

645: z\in\mathbb{D}^N,

646: \end{equation}

647: and we obtain

648: $\psi_{\rho,N}^\mathbf{A}\in\mathcal{AH}_N(\mathcal{X})$.

649: 

650: Conversely, let

651: $\psi_{\rho,N}^\mathbf{A}\in\mathcal{AH}_N(\mathcal{X})$. Since

652: $\psi_{\rho,N}^\mathbf{A}(0)=I_\mathcal{X}$, according to

653: \cite{Ag}, there exist a Hilbert space

654: $\widetilde{\mathcal{X}}\supset\mathcal{X}$, its subspaces

655: $\widetilde{\mathcal{X}}_1,\ldots,\widetilde{\mathcal{X}}_N$

656: satisfying

657: $\widetilde{\mathcal{X}}=\bigoplus_{k=1}^N\widetilde{\mathcal{X}}_k$,

658: and a unitary operator $U\in L(\widetilde{\mathcal{X}})$ such that

659: \begin{equation}\label{eq:ah}

660: \psi_{\rho,N}^\mathbf{A}(z)=P_\mathcal{X}(I_{\widetilde{\mathcal{X}}}+U(z\mathbf{P}))(I_{\widetilde{\mathcal{X}}}-U(z\mathbf{P}))^{-1}|\mathcal{X},\quad

661: z\in \mathbb{D}^N,

662: \end{equation}

663:  where

664: $z\mathbf{P}:=\sum_{k=1}^Nz_kP_{\widetilde{\mathcal{X}}_k}$, i.e.,

665: we get (\ref{eq:psi}) with

666: $\widetilde{A}_k=UP_{\widetilde{\mathcal{X}}_k},\ k=1,\ldots,N$.

667: Note that for each $\zeta\in\mathbb{T}^N$ the operator

668: $\zeta\widetilde{\mathbf{A}}$ is unitary. Developing both parts of

669: (\ref{eq:ah}) into the series in homogeneous polynomials

670: convergent in the operator norm, we get

671: \[

672: I_\mathcal{X}+\frac{2}{\rho}\sum_{n=1}^\infty(z\mathbf{A})^n=I_\mathcal{X}+2\sum_{n=1}^\infty

673: P_\mathcal{X}(z\widetilde{\mathbf{A}})^n|\mathcal{X},\quad

674: z\in\mathbb{D}^N,\] that implies the relations

675: \[ (z\mathbf{A})^n=\rho

676: P_\mathcal{X}(z\widetilde{\mathbf{A}})^n|\mathcal{X},\quad

677: n\in\mathbb{N},\] for all $z\in\mathbb{D}^N$, and hence for all

678: $z\in\mathbb{C}^N$. Thus, $\widetilde{\mathbf{A}}$ is a unitary

679: $\rho$-dilation of $\mathbf{A}$, and $\mathbf{A}\in C_{\rho,N}$.

680: The equivalence (i)$\Leftrightarrow$(iii) is proved.

681: 

682:  Note that in

683: this proof we have established that each Agler representation

684: (\ref{eq:ah}) of $\psi_{\rho,N}^\mathbf{A}$ gives rise to a

685: unitary $\rho$-dilation $\widetilde{\mathbf{A}}$ of $\mathbf{A}$,

686: and vice versa. Indeed, we already showed that (\ref{eq:ah})

687: determines $\widetilde{\mathbf{A}}$. Conversely, if

688: $\widetilde{\mathbf{A}}\in L(\widetilde{\mathcal{X}})^N$ is a

689: unitary $\rho$-dilation of $\mathbf{A}$, then (\ref{eq:psi})

690: holds. Set $U:=\sum_{k=1}^N\widetilde{A}_k\in

691: L(\widetilde{\mathcal{X}})$ and

692: $\widetilde{\mathcal{X}}_k:=\widetilde{A}_k^*\widetilde{\mathcal{X}},\

693: k=1,\ldots,N$. Then $U$ is unitary, $\widetilde{\mathcal{X}}_k$ is

694: a closed subspace in $\widetilde{\mathcal{X}}$ for each

695: $k=1,\ldots,N$, the subspaces $\widetilde{\mathcal{X}}_k$ are

696: pairwise orthogonal, and

697: $\widetilde{\mathcal{X}}=\bigoplus_{k=1}^N\widetilde{\mathcal{X}}_k$

698: (see Proposition~2.4 in \cite{K2}). Thus, (\ref{eq:psi}) turns

699: into (\ref{eq:ah}).

700: 

701: (v)$\Leftrightarrow$(iv). Let (v) be true. By

702: Theorem~\ref{thm:1-char} applied for $\mathbf{A\otimes C}$ with a

703: $\mathbf{C}\in\mathcal{C}^N$, one has

704: $\varphi_{\rho}^{\mathbf{A\otimes

705: C}}\in\mathcal{S}(\mathcal{X\otimes H}_\mathbf{C})$. For

706: $\varepsilon>0$ small enough,

707: $(1+\varepsilon)\mathbf{C}\in\mathcal{C}^N$, hence

708: $\mathbf{A}\otimes(1+\varepsilon)\mathbf{C}\in C_\rho$, and

709: $\varphi_\rho^{\mathbf{A}\otimes(1+\varepsilon)\mathbf{C}}\in\mathcal{S}(\mathcal{X\otimes

710: H}_\mathbf{C})$. Thus,

711: \[ \varphi_{\rho,N}^\mathbf{A}(\mathbf{C})=\mathbf{A\otimes

712: C}((\rho-1)\mathbf{A\otimes C}-\rho

713: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}})^{-1}=\varphi_\rho^{\mathbf{A}\otimes(1+\varepsilon)\mathbf{C}}\left(\frac{1}{1+\varepsilon}\right)\]

714: is a contraction on $\mathcal{X}\otimes\mathcal{H}_\mathbf{C}$. In

715: particular, $\varphi_{\rho,N}^\mathbf{A}(z)$ is well-defined,

716: holomorphic and contractive on $\mathbb{D}^N$. Finally,

717: $\varphi_{\rho,N}^\mathbf{A}\in\mathcal{AS}_N(\mathcal{X})$.

718: 

719: Conversely, if (iv) is true then for any

720: $\mathbf{C}\in\mathcal{C}^N$:

721: \[\varphi_\rho^{\mathbf{A\otimes C}}(\lambda)=\lambda\mathbf{A\otimes

722: C}((\rho-1)\lambda\mathbf{A\otimes C}-\rho

723: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}})^{-1}=\varphi_{\rho,N}^\mathbf{A}(\lambda\mathbf{C})\]

724: is well-defined, holomorphic and contractive for

725: $\lambda\in\mathbb{D}$. Thus, $\varphi_\rho^{\mathbf{A\otimes

726: C}}\in\mathcal{S}(\mathcal{X\otimes H}_\mathbf{C})$, and by

727: Theorem~\ref{thm:1-char}, $\mathbf{A\otimes C}\in C_\rho$.

728: 

729: (v)$\Leftrightarrow$(iii) and (v)$\Leftrightarrow$(ii) are proved

730: analogously, using the following relations for

731: $\mathbf{C}\in\mathcal{C}^N,\ \lambda\in\mathbb{D}$:

732: \begin{eqnarray*}

733: \psi_{\rho,N}^\mathbf{A}(\mathbf{C})=\psi_\rho^{\mathbf{A}\otimes(1+\varepsilon)\mathbf{C}}\left(\frac{1}{1+\varepsilon}\right),

734: \quad\psi_\rho^{\mathbf{A\otimes

735: C}}(\lambda)=\psi_{\rho,N}^\mathbf{A}(\lambda\mathbf{C}),\\

736: k_{\rho,N}^\mathbf{A}(\mathbf{C,C})=k_\rho^{\mathbf{A}\otimes\mathbf{C}}(1,1),

737: \quad k_\rho^{\mathbf{A\otimes

738: C}}(\lambda,\lambda)=k_{\rho,N}^\mathbf{A}(\lambda\mathbf{C},\lambda\mathbf{C}).

739: \end{eqnarray*}

740: The proof is complete.

741: \end{proof}

742: \begin{rem}\label{rem:1-contr}

743: For the case $\rho=1$ each of conditions (ii)--(v) in

744: Theorem~\ref{thm:n-char} means that for any

745: $\mathbf{C}\in\mathcal{C}^N$ the operator $\mathbf{A\otimes C}$ is

746:  a contraction. In other words,

747:  \[ \mathbf{A}\in C_{1,N}\cap L(\mathcal{X})^N \Longleftrightarrow

748:  L_\mathbf{A}\in\mathcal{AS}_N(\mathcal{X}),\]

749:  that coincides with  in \cite[Corollary~4.3]{K3} (here $L_\mathbf{A}(z):=z\mathbf{A},\ z\in\mathbb{C}^N$).

750: \end{rem}

751: Let us also note that using \cite[Corollary~4.3]{K3} one can deduce (v) from (i) directly. Indeed, if $\widetilde{\mathbf{A}}\in

752: L(\widetilde{\mathcal{X}})^N$ is a unitary $\rho$-dilation of

753: $\mathbf{A}\in L(\mathcal{X})^N$ then for any

754: $\mathbf{C}\in\mathcal{C}^N$ by \cite[Corollary~4.3]{K3} the operator

755: $\widetilde{\mathbf{A}}\otimes\mathbf{C}$ is a contraction.

756: Therefore, due to \cite{SzN},

757: $\widetilde{\mathbf{A}}\otimes\mathbf{C}\in

758: L(\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C})$ has a

759: unitary dilation $U\in L(\mathcal{K}),\

760: \mathcal{K}\supset\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}$.

761: Then for any $n\in\mathbb{N}$:

762: \begin{eqnarray*}

763:  (\mathbf{A}\otimes\mathbf{C})^n &=&\rho

764: P_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}(\widetilde{\mathbf{A}}\otimes\mathbf{C})^n|\mathcal{X}\otimes\mathcal{H}_\mathbf{C} \\

765: &=&

766: \rho P_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}(

767: P_{\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C}}U^n|\widetilde{\mathcal{X}}\otimes\mathcal{H}_\mathbf{C})|

768: \mathcal{X}\otimes\mathcal{H}_\mathbf{C}\\

769: &=& \rho P_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}U^n|

770: \mathcal{X}\otimes\mathcal{H}_\mathbf{C},

771: \end{eqnarray*}

772:  i.e., $U$ is a unitary

773: $\rho$-dilation of the operator $\mathbf{A}\otimes\mathbf{C}$.

774: Thus, $\mathbf{A}\otimes\mathbf{C}\in C_{\rho}$.

775: 

776: Let us define the \emph{numerical radius of an $N$-tuple of

777: operators} $\mathbf{A}\in L(\mathcal{X})^N$ as

778: \begin{equation}\label{eq:num-rad}

779:     w^{(N)}(\mathbf{A}):=\sup_{\mathbf{C}\in\mathcal{C}^N}w(\mathbf{A}\otimes\mathbf{C}).

780: \end{equation}

781: For $N=1$, $w^{(1)}(A)=w(A)$.  Indeed,

782: \begin{eqnarray*}

783: w^{(1)}(A) &=& \sup_{\| C\|<1}w(A\otimes

784: C) \geq \sup_{0<\varepsilon<1}w(A\otimes(1-\varepsilon)I_{\mathcal{H}_C})=\sup_{0<\varepsilon<1}(1-\varepsilon)w(A)\\

785: &=& w(A);\\

786: w^{(1)}(A)

787:  &=& \sup_{\| C\|<1}w(A\otimes C) \leq  \sup_{\|

788: C\|<1}w(A)\| C\|=w(A).

789: \end{eqnarray*}

790: Here we used the properties $w(A\otimes I_\mathcal{H})=w(A)$ and

791: $w(A\otimes B) \leq w(A)\| B\|$ valid for any $A\in

792: L(\mathcal{X}),B\in L(\mathcal{H})$ (see, e.g., \cite{FH}).

793: \begin{prop}\label{prop:n-num-rad}

794: $\mathbf{A}\in C_{2,N}\Longleftrightarrow w^{(N)}(\mathbf{A})\leq

795: 1$.

796: \end{prop}

797: \begin{proof}

798: By Theorem~\ref{thm:n-char}, $\mathbf{A}\in C_{2,N}$ if and only

799: if $\mathbf{A\otimes C}\in C_2=C_{2,1}$ for any

800: $\mathbf{C}\in\mathcal{C}^N$. This, in turn, means that

801: $w(\mathbf{A\otimes C})\leq 1$ for any

802: $\mathbf{C}\in\mathcal{C}^N$ (by Berger's result mentioned in

803: Section~\ref{sec:prelim}), i.e, $w^{(N)}(\mathbf{A})\leq 1$.

804: \end{proof}

805: \begin{thm}\label{thm:pr}

806: If $\mathbf{A}\in C_{\rho,N}\cap L(\mathcal{X})^N$ for a $\rho>0$,

807: then $L_\mathbf{A}\in\rho\mathcal{AS}_N(\mathcal{X})$. For any

808: $\rho>0$ such that $\rho\neq 1$, there exists an $\mathbf{A}\in

809: L(\mathcal{X})^N$ such that

810: $L_\mathbf{A}\in\rho\mathcal{AS}_N(\mathcal{X})$ and

811: $\mathbf{A}\notin C_{\rho,N}$.

812: \end{thm}

813: \begin{proof}

814: Let $\mathbf{A}\in C_{\rho,N}\cap L(\mathcal{X})^N$ for some

815: $\rho>0$, and $\mathbf{C}\in\mathcal{C}^N$. Then $\mathbf{A}$ has

816: a unitary $\rho$-dilation $\widetilde{\mathbf{A}}\in

817: L(\widetilde{\mathcal{X}})^N$, and

818: \begin{eqnarray*}

819: \|\mathbf{A\otimes C}\| &=& \left\|\sum_{k=1}^NA_k\otimes

820: C_k\right\|=\left\|\left.\rho(P_\mathcal{X}\otimes

821: I_{\mathcal{H}_\mathbf{C}})\left(\sum_{k=1}^N\widetilde{A}_k\otimes

822: C_k\right)\right|\mathcal{X}\otimes\mathcal{H}_\mathbf{C}\right\| \\

823: &\leq &\rho\left\|\sum_{k=1}^N\widetilde{A}_k\otimes

824: C_k\right\|=\rho\left\|\widetilde{\mathbf{A}}\otimes\mathbf{C}\right\|\leq\rho

825: \end{eqnarray*}

826: (here we used again Corollary~4.3 in \cite{K3}). Thus,

827: $L_\mathbf{A}\in\rho\mathcal{AS}_N(\mathcal{X})$.

828: 

829: Now, let $0<\rho\neq 1$, and $\mathbf{A}\in L(\mathcal{X})^N$ be

830: such that

831: $\frac{1}{\rho}L_\mathbf{A}(\zeta)=\frac{1}{\rho}\zeta\mathbf{A}$

832: is a unitary operator for each $\zeta\in\mathbb{T}^N$. Then, again

833: by Corollary~4.3 in \cite{K3},

834: $L_\mathbf{A}\in\rho\mathcal{AS}_N(\mathcal{X})$. Suppose there

835: exists a unitary $\rho$-dilation $\widetilde{\mathbf{A}}\in

836: L(\widetilde{\mathcal{X}})^N$ of $\mathbf{A}$. Then for any

837: $\zeta\in\mathbb{T}^N$, $L_\mathbf{A}(\zeta)=\zeta\mathbf{A}=\rho

838: P_\mathcal{X}(\zeta\widetilde{\mathbf{A}})|\mathcal{X}$. Hence,

839: for any $\zeta\in\mathbb{T}^N$ and $x\in\mathcal{X}$,

840: \[

841: \|\zeta\widetilde{\mathbf{A}}x\|=\|x\|=\|\frac{1}{\rho}\zeta\mathbf{A}x\|=\|

842: P_\mathcal{X}(\zeta\widetilde{\mathbf{A}})x\|,\] that is possible

843: only if $\zeta\widetilde{\mathbf{A}}x\in\mathcal{X}$ for all

844: $\zeta\in\mathbb{T}^N$ and $x\in\mathcal{X}$. Therefore, for

845: $n>1$,

846: \[\rho^n\|x\|=\|(\zeta\mathbf{A})^nx\|=\|\rho

847: P_\mathcal{X}(\zeta\widetilde{\mathbf{A}})^nx\|=\rho\|(\zeta\widetilde{\mathbf{A}})^nx\|=\rho\|x\|,\]

848: that is impossible for $x\neq 0$. Thus, $\mathbf{A}\notin

849: C_{\rho,N}$.

850: \end{proof}

851: \begin{note*} Compare Theorem~\ref{thm:pr} with Remark~\ref{rem:1-contr}.

852: \end{note*}

853: The same argument as in the proof of the first part of

854: Theorem~\ref{thm:pr} shows that, for $\mathbf{A}\in C_{\rho,N}$,

855: \begin{equation}\label{eq:n-pb}

856:     \|(\mathbf{A\otimes C})^n\|\leq\rho,\quad n\in\mathbb{N},\

857:     \mathbf{C}\in\mathcal{C}^N.

858: \end{equation}

859: \begin{note*} Compare (\ref{eq:n-pb}) with (\ref{eq:1-pb}).

860: \end{note*}

861: This uniform (in $\mathbf{C}\in\mathcal{C}^N$) power-boundedness

862: of an $N$-tuple of operators $\mathbf{A}$ is, in our setting, a

863: generalization of power-boundedness of a single operator. Let us

864: define the \emph{spectral radius of an $N$-tuple of operators}

865: $\mathbf{A}\in L(\mathcal{X})^N$ as

866: \begin{equation}\label{eq:n-sp-rad}

867:     \nu^{(N)}(\mathbf{A}):=\lim_{n\rightarrow +\infty}\left(\sup_{\mathbf{C}\in\mathcal{C}^N}\|(\mathbf{A\otimes

868:     C})^n\|\right)^{\frac{1}{n}}.

869: \end{equation}

870: \begin{note*} Compare (\ref{eq:n-sp-rad}) with (\ref{eq:1-sp-rad}).

871: \end{note*}

872: In other words,

873: $\nu^{(N)}(\mathbf{A})=\nu^{(N,\infty)}(L_\mathbf{A})$, where

874: $\nu^{(N,\infty)}(f)$ is the \emph{spectral radius of an element

875: $f$ of the Banach algebra $H_N^\infty(\mathcal{X})$} consisting of

876: holomorphic $L(\mathcal{X})$-valued functions $f$ on

877: $\mathbb{D}^N$ which satisfy

878: \[ \| f\|_{\infty,N}:=\sup_{\mathbf{C}\in\mathcal{C}^N}\|

879: f(\mathbf{C})\|<\infty\] (this algebra was introduced in

880: \cite{Ag}). Here $f(\mathbf{C})$ is defined in the same manner as

881: $k(\mathbf{C,C})$ in (\ref{eq:n-calc}), i.e., for

882: \begin{eqnarray*}

883: f(z) &=& \sum_{t\in\mathbb{Z}^N_+}\hat{f}_tz^t,\quad

884: z\in\mathbb{D}^N,\\

885: f(\mathbf{C}) &:=& \sum_{t\in\mathbb{Z}^N_+}\hat{f}_t\otimes

886: \mathbf{C}^t,\quad \mathbf{C}\in\mathcal{C}^N,

887: \end{eqnarray*}

888: where the latter series converges in the

889: $L(\mathcal{X}\otimes\mathcal{H}_\mathbf{C})$-norm. For $N=1$,

890: $\nu^{(1)}(A)=\nu(A)$. Indeed,

891: \begin{eqnarray*}

892: \nu^{(1)}(A) &=& \lim_{n\rightarrow

893: +\infty}\left(\sup_{\|C\|<1}\|(A\otimes

894:     C)^n\|\right)^{\frac{1}{n}}=\lim_{n\rightarrow +\infty}\left(\sup_{\|C\|<1}\|A^n\otimes

895:     C^n\|\right)^{\frac{1}{n}}\\

896:     &=& \lim_{n\rightarrow +\infty}\left(\|A^n\|\sup_{\|C\|<1}

897:     \|C^n\|\right)^{\frac{1}{n}}=\lim_{n\rightarrow

898:     +\infty}\|A^n\|^{\frac{1}{n}}=\nu(A).

899: \end{eqnarray*}

900: \begin{rem}\label{rem:n-sp-rad}

901: For any $\mathbf{A}\in C_{\rho,N}$, by virtue of (\ref{eq:n-pb}),

902: $\nu^{(N)}(\mathbf{A})\leq 1$.

903: \end{rem}

904: \begin{thm}\label{thm:n-increase}

905: For a fixed Hilbert space $\mathcal{X}$ and any $N\geq 1$ the

906: class $C_{\rho,N}$ increases as a function of $\rho$:

907: \[ C_{\rho,N}\subset C_{\rho',N}\ \mbox{for}\ \rho<\rho'.\]

908:  Moreover, for

909: $\dim\mathcal{X}\geq 2$, $C_{\rho,N}$ increases strictly:

910: \[ C_{\rho,N}\neq C_{\rho',N}\ \mbox{for}\ \rho\neq\rho'.\]

911: For $\dim\mathcal{X}=1$ the classes $C_{\rho,N}$ coincide for all

912: $\rho\geq 1$, and strictly increase for $0<\rho<1$.

913: \end{thm}

914: \begin{proof}

915: For $N=1$ this theorem is true (see Section~\ref{sec:prelim}). For

916: $N>1$ it follows from the equivalence (i)$\Leftrightarrow$(v) in

917: Theorem~\ref{thm:n-char}.

918: \end{proof}

919: \begin{thm}\label{thm:n-vN}

920: For any $\mathbf{A}\in C_{\rho,N},\ \mathbf{C}\in\mathcal{C}^N$,

921: and a polynomial $p$ of one variable, \[\|

922: p(\mathbf{A}\otimes\mathbf{C})\|\leq\max_{|z|\leq 1}|\rho

923: p(z)+(1-\rho)p(0)|.\]

924: \end{thm}

925: \begin{proof}

926: This result follows from the generalized von Neumann inequality

927: (\ref{eq:vn}) and the equivalence (i)$\Leftrightarrow$(v) in

928: Theorem~\ref{thm:n-char}.

929: \end{proof}

930: Let us remark that results of this section on $N$-tuples of

931: operators from the classes $C_{\rho,N}$ can be extended to

932: elements of $H_N^\infty(\mathcal{X})$, though the notion of

933: unitary $\rho$-dilation no longer makes sense for this case. Define

934: $C_{\rho,N}^{(\infty)}$ as a class of functions $f\in

935: H_N^\infty(\mathcal{X})$ such that $f(\mathbf{C})\in

936: C_\rho=C_{\rho,1}$ for any $\mathbf{C}\in\mathcal{C}^N$. Then, in

937: particular, Theorem~\ref{thm:n-char} implies that $\mathbf{A}\in

938: C_{\rho,N}$ if and only if $L_\mathbf{A}\in

939: C_{\rho,N}^{(\infty)}$. The following analogue of

940: Theorem~\ref{thm:n-char} is easily obtained.

941: \begin{thm}\label{thm:n-char-bdd}

942: Let $f\in H_N^\infty(\mathcal{X})$ and $\rho>0$. The following

943: conditions are equivalent:

944: \begin{description}

945:     \item[(i)] $f\in C_{\rho,N}^{(\infty)}$;

946:     \item[(ii)] the function $k_{\rho,N}^f(z,w):=\rho

947:     I_\mathcal{X}-(\rho-1)(f(z)+(f(w)^*)+(\rho-2)f(w)^*f(z)$

948: is an Agler kernel on $\mathbb{D}^N\times\mathbb{D}^N$;

949:     \item[(iii)] the function

950:    $\psi_{\rho,N}^f(z):=(1-\frac{2}{\rho})I_\mathcal{X}+\frac{2}{\rho}(I_\mathcal{X}-f(z))^{-1}$

951:     belongs to $\mathcal{AH}_N(\mathcal{X})$;

952:     \item[(iv)] the function

953:     $\varphi_{\rho,N}^f(z):=f(z)((\varrho-1)f(z)-\rho I_\mathcal{X})^{-1}$

954:     belongs to $\mathcal{AS}_N(\mathcal{X})$.

955: \end{description}

956: \end{thm}

957: Clearly, $H_N^\infty(\mathcal{X})\cap

958: C_{1,N}^{(\infty)}=\mathcal{AS}_N(\mathcal{X})$. Set

959: \begin{equation}\label{eq:num-rad-bdd}

960: w^{(N,\infty)}(f):=\sup_{\mathbf{C}\in\mathcal{C}^N}w(f(\mathbf{C})).

961: \end{equation}

962: \begin{note*} Compare (\ref{eq:num-rad-bdd}) with (\ref{eq:num-rad}).

963: \end{note*}

964: \begin{rem}\label{rem:n-ext}

965: Proposition~\ref{prop:n-num-rad} extends directly  to the class

966: $C_{2,N}^{(\infty)}$, with $f\in H_N^\infty(\mathcal{X})$ in the

967: place of $\mathbf{A}\in L(\mathcal{X})^N$, and $w^{(N,\infty)}(f)$

968: in the place of $w^{(N)}(\mathbf{A})$. Remark~\ref{rem:n-sp-rad} extends directly 

969: to $f\in H_N^\infty(\mathcal{X})$ in the

970: place of $\mathbf{A}\in L(\mathcal{X})^N$, and

971: $\nu^{(N,\infty)}(f)$ in the place of $\nu^{(N)}(\mathbf{A})$.

972: Also, Theorems~\ref{thm:n-increase} and \ref{thm:n-vN} extend to the classes $C_{\rho,N}^{(\infty)}$.

973: \end{rem}

974: 

975: \section{Multivariable operator and operator-function radii}\label{sec:n-rad}

976: In this section we extend the notion of operator radii $w_\rho,\

977: 0<\rho\leq\infty$, to the multivariable case, i.e., to $N$-tuples

978: of bounded linear operators and to elements of the Banach algebra

979: $H_N^\infty(\mathcal{X})$. Let $0<\rho<\infty$ and $f\in

980: H_N^\infty(\mathcal{X})$. Set

981: \[ w_{\rho,N}^{(\infty)}(f):=\inf\{ u>0:\frac{1}{u}f\in

982: C_{\rho,N}^{(\infty)}\},\] and for $\mathbf{A}\in

983: L(\mathcal{X})^N$, define

984: \[ w_{\rho,N}(\mathbf{A}):=w_{\rho,N}^{(\infty)}(L_\mathbf{A}).\]

985: Due to our remark preceding to Theorem~\ref{thm:n-char-bdd},

986: \begin{equation}\label{eq:n-rad}

987: w_{\rho,N}(\mathbf{A})=\inf\{ u>0:\frac{1}{u}\mathbf{A}\in

988: C_{\rho,N}\}.

989: \end{equation}

990: \begin{note*} Compare (\ref{eq:n-rad}) with (\ref{eq:1-rad}).

991: \end{note*}

992: Clearly, for $N=1$ and $A\in L(\mathcal{X})$,

993: $w_{\rho,1}(A)=w_\rho(A)$.

994: \begin{lem}\label{lem:w}

995: For $f\in H_N^\infty(\mathcal{X}),\ \mathbf{A}\in

996: L(\mathcal{X})^N$,

997: \begin{eqnarray}

998: w_{\rho,N}^{(\infty)}(f) &=&

999: \sup_{\mathbf{C}\in\mathcal{C}^N}w_\rho(f(\mathbf{C})),

1000: \label{eq:w-f}\\

1001: w_{\rho,N}(\mathbf{A}) &=&

1002: \sup_{\mathbf{C}\in\mathcal{C}^N}w_\rho(\mathbf{A\otimes C}).

1003: \label{eq:w-a}

1004: \end{eqnarray}

1005: \end{lem}

1006: \begin{proof}

1007: Let  $f\in H_N^\infty(\mathcal{X})$. Then for $u>0$,

1008: $\frac{1}{u}f\in C_{\rho,N}^{(\infty)}$ if and only if for any

1009: $\mathbf{C}\in\mathcal{C}^N$ one has $\frac{1}{u}f(\mathbf{C})\in

1010: C_\rho$. Therefore,

1011: \begin{eqnarray*}

1012: w_{\rho,N}^{(\infty)}(f) &=& \inf\{ u>0:\frac{1}{u}f\in

1013: C_{\rho,N}^{(\infty)}\}=\inf\{

1014: u>0:\forall\mathbf{C}\in\mathcal{C}^N,\

1015: \frac{1}{u}f(\mathbf{C})\in

1016: C_\rho\}\\

1017: &=& \sup_{\mathbf{C}\in\mathcal{C}^N}\inf\{

1018: u>0:\frac{1}{u}f(\mathbf{C})\in C_\rho\}=

1019: \sup_{\mathbf{C}\in\mathcal{C}^N}w_\rho(f(\mathbf{C})),

1020: \end{eqnarray*}

1021: i.e., (\ref{eq:w-f}) is true. Now, (\ref{eq:w-a}) follows from

1022: (\ref{eq:w-f}) and the definition of $w_{\rho,N}(\mathbf{A})$.

1023: \end{proof}

1024: \begin{thm}\label{thm:n-rad}

1025: {\bf 1.} All properties $(${\rm i}$)$--$(${\rm xx}$)$ listed in

1026: Theorem~\ref{thm:1-rad} are satisfied for

1027: $w_{\rho,N}^{(\infty)}(\cdot)$ in the place of $w_\rho(\cdot);\

1028: f,g\in H_N^\infty(\mathcal{X})$ in the place of $A,B\in

1029: L(\mathcal{X})$; $w^{(N,\infty)}(\cdot)$ in the place of

1030: $w(\cdot)$; and $\nu^{(N,\infty)}(\cdot)$ in the place of

1031: $\nu(\cdot)$.

1032: 

1033:  {\bf 2.} Properties $(${\rm i}$)$--$(${\rm

1034: xvii}$)$ listed in Theorem~\ref{thm:1-rad} are satisfied for

1035: $w_{\rho,N}(\cdot)$ in the place of $w_\rho(\cdot);\ \mathbf{A}\in

1036: L(\mathcal{X})^N$ in the place of $A\in L(\mathcal{X})$;

1037: $w^{(N)}(\cdot)$ in the place of $w(\cdot)$; and

1038: $\nu^{(N)}(\cdot)$ in the place of $\nu(\cdot)$.

1039: \end{thm}

1040: \begin{proof}

1041: {\bf 1.} Let $f\in H_N^\infty(\mathcal{X})$. Then $\|

1042: f\|_{\infty,N}=\sup_{\mathbf{C}\in\mathcal{C}^N}\|

1043: f(\mathbf{C})\|<\infty$.  By properties (vi) and (x) in

1044: Theorem~\ref{thm:1-rad}, and Lemma~\ref{lem:w}, if $0<\rho\leq 1$

1045: then

1046: \begin{eqnarray*}

1047: w_{\rho,N}^{(\infty)}(f) &=&

1048: \sup_{\mathbf{C}\in\mathcal{C}^N}w_\rho(f(\mathbf{C}))\leq\left(\frac{2}{\rho}-1\right)

1049: \sup_{\mathbf{C}\in\mathcal{C}^N}w_1(f(\mathbf{C})) \\

1050: &=&

1051: \left(\frac{2}{\rho}-1\right)\sup_{\mathbf{C}\in\mathcal{C}^N}\|

1052: f(\mathbf{C})\| 

1053: <\infty,

1054: \end{eqnarray*}

1055:  and if $\rho>1$ then

1056: \[

1057: w_{\rho,N}^{(\infty)}(f)=\sup_{\mathbf{C}\in\mathcal{C}^N}w_\rho(f(\mathbf{C}))\leq

1058: \sup_{\mathbf{C}\in\mathcal{C}^N}w_1(f(\mathbf{C}))=

1059: \sup_{\mathbf{C}\in\mathcal{C}^N}\|f(\mathbf{C})\|<\infty.\] Thus,

1060: property (i) is fulfilled.

1061: 

1062: Properties (ii)--(vii), (ix)--(xi), (xiii)--(xv), and (xvii)--(xx)

1063: easily follow from the properties in Theorem~\ref{thm:1-rad} with

1064: the same numbers, and Lemma~\ref{lem:w}.

1065: 

1066: The proof of property (viii) is an adaptation of the proof of

1067: Theorem~5.1 in \cite{H1} to our case. First of all, let us remark

1068: that property (iv) implies that if $u>w_{\rho,N}^{(\infty)}(f)$

1069: then $\frac{1}{u}f\in C_{\rho,N}^{(\infty)}$, and for any

1070: $\mathbf{C}\in\mathcal{C}^N$ one has $\frac{1}{u}f(\mathbf{C})\in

1071: C_\rho$. In particular,

1072: \[

1073: \sup_{\mathbf{C}\in\mathcal{C}^N}\left\|\left(\frac{f(\mathbf{C})}{u}\right)^n\right\|\leq\rho,\quad

1074: n\in\mathbb{N}.\] Therefore,

1075: $\nu^{(N,\infty)}\left(\frac{f}{u}\right)\leq 1$, i.e.,

1076: $\nu^{(N,\infty)}(f)\leq u$. Thus, for any $\rho>0$,

1077: $\nu^{(N,\infty)}(f)\leq w_{\rho,N}^{(\infty)}(f)$, moreover,

1078: \[\nu^{(N,\infty)}(f)\leq \lim_{\rho\rightarrow

1079: +\infty}w_{\rho,N}^{(\infty)}(f)\] (note, that due to property

1080: (x), $w_{\rho,N}^{(\infty)}(f)$ is a non-increasing and bounded

1081: from below function of $\rho$, hence it has a limit as

1082: $\rho\rightarrow +\infty$).

1083: 

1084: For the proof of the opposite inequality, let us first show that

1085: if $\nu^{(N,\infty)}(g)<1$ for some $g\in H_N^\infty(\mathcal{X})$

1086: then beginning with some $\rho_0>0$ (i.e., for all $\rho\geq

1087: \rho_0$) one has $g\in C_{\rho,N}^{(\infty)}$. Indeed, in  this 

1088:  case there exists an $s>1$ such that

1089: $\nu^{(N,\infty)}(sg)<1$. Then there exists a $B>0$ such that

1090: \[ s^n\sup_{\mathbf{C}\in\mathcal{C}^N}\| g(\mathbf{C})^n\|\leq

1091: B,\quad n\in\mathbb{N}.\] Hence, for any

1092: $\mathbf{C}\in\mathcal{C}^N$,

1093: \begin{eqnarray*}

1094: \Real\psi_{\rho,N}^g(\mathbf{C}) &=&

1095: \left(1-\frac{2}{\rho}\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}+

1096: \frac{2}{\rho}\Real(I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}-g(\mathbf{C}))^{-1}

1097: \\ &=& I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}+

1098: \frac{2}{\rho}\Real\sum_{n=1}^\infty g(\mathbf{C})^n \succeq

1099: \left(1-\frac{2}{\rho}\sum_{n=1}^\infty \|

1100: g(\mathbf{C})^n\|\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}

1101: \\

1102: &\succeq & \left(1-\frac{2}{\rho}\sum_{n=1}^\infty

1103: \frac{B}{s^n}\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}=\left(1-\frac{2B}{\rho(s-1)}\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}\succeq

1104: 0

1105: \end{eqnarray*}

1106: as soon as $\rho\geq\frac{2B}{s-1}$. Thus, by

1107: Theorem~\ref{thm:n-char-bdd}, $g\in C_{\rho,N}^{(\infty)}$ for any

1108: $\rho\geq\frac{2B}{s-1}$.

1109: 

1110: Now, if $\nu^{(N,\infty)}(f)=0$ then for any $k\in\mathbb{N}$,

1111: $\nu^{(N,\infty)}(kf)=0$. Hence, for $\rho\geq\rho_0$ we have

1112: $kf\in  C_{\rho,N}^{(\infty)}$, and by property (iv),

1113: $w_{\rho,N}^{(\infty)}(kf)\leq 1$. Thus, $$\lim_{\rho\rightarrow

1114: +\infty}w_{\rho,N}^{(\infty)}(f)\leq\frac{1}{k}$$ for any

1115: $k\in\mathbb{N}$, and $$\lim_{\rho\rightarrow

1116: +\infty}w_{\rho,N}^{(\infty)}(f)=0=\nu^{(N,\infty)}(f),$$ as

1117: required.

1118: 

1119: If $\nu^{(N,\infty)}(f)>0$ then for any $\varepsilon>0$,

1120: \[

1121: \nu^{(N,\infty)}\left(\frac{f}{(1+\varepsilon)\nu^{(N,\infty)}(f)}\right)=\frac{1}{1+\varepsilon}<1.\]

1122: Then for $\rho\geq\rho_0$,

1123: \[

1124: w_{\rho,N}^{(\infty)}\left(\frac{f}{(1+\varepsilon)\nu^{(N,\infty)}(f)}\right)\leq

1125: 1,\] hence

1126: $w_{\rho,N}^{(\infty)}(f)\leq(1+\varepsilon)\nu^{(N,\infty)}(f)$.

1127: Passing to the limit as $\rho\rightarrow+\infty$, and then as

1128: $\varepsilon\downarrow 0$, we get

1129: \[ \lim_{\rho\rightarrow

1130: +\infty}w_{\rho,N}^{(\infty)}(f)\leq\nu^{(N,\infty)}(f),\] as

1131: required. Thus, property (viii) is proved.

1132: 

1133: For the proof of property (xii), it is enough to suppose, by

1134: virtue of positive homogeneity of $w_{\rho,N}^{(\infty)}(\cdot)$

1135: and $\nu^{(N,\infty)}(\cdot)$, that for $f\in

1136: H_N^\infty(\mathcal{X})$ one has $w_{\rho_0,N}^{(\infty)}(f)=1,\

1137: \nu^{(N,\infty)}(f)<1$, and prove that for any $\rho>\rho_0$,

1138: $w_{\rho,N}^{(\infty)}(f)<1$. By Theorem~\ref{thm:n-char-bdd} and

1139: property (iv) in the present theorem,

1140: \[

1141: \frac{\rho_0}{2}\psi_{\rho_0,N}^f(z)=\left(\frac{\rho_0}{2}-1\right)I_\mathcal{X}+(I_\mathcal{X}-f(z))^{-1}\in\mathcal{AH}_N(\mathcal{X}),\]

1142: i.e., for any $\mathbf{C}\in\mathcal{C}^N$,

1143: \[

1144: \Real\left[\frac{\rho_0}{2}\psi_{\rho_0,N}^f(\mathbf{C})\right]=\Real\left[\left(\frac{\rho_0}{2}-1\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}

1145: +(I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}-f(\mathbf{C}))^{-1}\right]\succeq

1146: 0,\] and for any $\rho>\rho_0$,

1147: \[ \Real\left[\frac{\rho}{2}\psi_{\rho,N}^f(\mathbf{C})\right]=\Real\left[\left(\frac{\rho}{2}-1\right)I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}

1148: +(I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}-f(\mathbf{C}))^{-1}\right]\succeq\frac{\rho-\rho_0}{2}I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}.\]

1149: Since the resolvent $R_f(\lambda):=(\lambda

1150: I_{\mathcal{X}}-f)^{-1}$ is

1151: continuous in the $H_N^\infty(\mathcal{X})$-norm on the resolvent

1152: set of $f$, and $\nu^{(N,\infty)}(f)<1$, for $\varepsilon>0$ small

1153: enough, one has $\nu^{(N,\infty)}((1+\varepsilon)f)<1$, and

1154: \[ \Real\left[\frac{\rho}{2}\psi_{\rho,N}^{(1+\varepsilon)f}(\mathbf{C})\right]=\Real\left[\left(\frac{\rho}{2}-1\right)

1155: I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}

1156: +(I_{\mathcal{X}\otimes\mathcal{H}_\mathbf{C}}-(1+\varepsilon)f(\mathbf{C}))^{-1}\right]\succeq

1157: 0\] for any $\mathbf{C}\in\mathcal{C}^N$, i.e.,

1158: $\frac{\rho}{2}\psi_{\rho,N}^{(1+\varepsilon)f}\in\mathcal{AH}_N(\mathcal{X})$,

1159: and

1160: $\psi_{\rho,N}^{(1+\varepsilon)f}\in\mathcal{AH}_N(\mathcal{X})$.

1161: Hence, by Theorem~\ref{thm:n-char-bdd}, $(1+\varepsilon)f\in

1162: C_{\rho,N}^{(\infty)}$ which means, by property (iv), that

1163: $w_{\rho,N}^{(\infty)}((1+\varepsilon)f)\leq 1$. Thus,

1164: $w_{\rho,N}^{(\infty)}(f)\leq\frac{1}{1+\varepsilon}<1$, as

1165: required.

1166: 

1167: The first part of property (xvi) in this theorem follows from

1168: property (xvi) in Theorem~\ref{thm:1-rad}, and Lemma~\ref{lem:w}.

1169: For the proof of the second part of (xvi), we use properties (xv)

1170: and (xvi) from Theorem~\ref{thm:1-rad}, property (xv) in the

1171: present theorem, and Lemma~\ref{lem:w}:

1172: \begin{eqnarray*}

1173: \lim_{\rho\downarrow 0}\frac{\rho}{2}w_{\rho,N}^{(\infty)}(f) &=&

1174: \sup_{0<\rho<1}\left\{\frac{\rho}{2}w_{\rho,N}^{(\infty)}(f)\right\}=\sup_{0<\rho<1}\sup_{\mathbf{C}\in\mathcal{C}^N}

1175: \left\{\frac{\rho}{2}w_\rho(f(\mathbf{C}))\right\}

1176: \\

1177: &=&

1178: \sup_{\mathbf{C}\in\mathcal{C}^N}\sup_{0<\rho<1}\left\{\frac{\rho}{2}w_\rho(f(\mathbf{C}))\right\}=

1179: \sup_{\mathbf{C}\in\mathcal{C}^N}\left\{\lim_{\rho\downarrow

1180: 0}\frac{\rho}{2}w_\rho(f(\mathbf{C}))\right\} \\

1181: &=&

1182: \sup_{\mathbf{C}\in\mathcal{C}^N}w_2(f(\mathbf{C}))=w_{2,N}^{(\infty)}(f).

1183: \end{eqnarray*}

1184: The proof of property (xvi), as well as part~\textbf{1} of this

1185: theorem, is complete.

1186: 

1187: Part~{\bf 2} follows from part~{\bf 1}.

1188: \end{proof}

1189: Denote by $C_{\infty,N}^{(\infty)}$ (resp., $C_{\infty,N}$) the

1190: class of $\mathcal{C}^N$-bounded holomorphic operator valued

1191: functions on $\mathbb{D}^N$ (resp., $N$-tuples of bounded linear

1192: operators on a common Hilbert space) with spectral radius at most

1193: one.

1194: \begin{thm}\label{thm:n-clos}

1195: Let $\mathcal{X}$ be a Hilbert space. Then

1196: \begin{eqnarray}

1197: C_{\infty,N}^{(\infty)}\cap H_N^\infty(\mathcal{X})  &=&

1198: \clos\left\{\bigcup_{0<\rho<\infty}(C_{\rho,N}^{(\infty)}\cap

1199: H_N^\infty(\mathcal{X}))\right\}; \label{eq:sp-r-f} \\

1200: C_{\infty,N}\cap L(\mathcal{X})^N  &=&

1201: \clos\left\{\bigcup_{0<\rho<\infty}(C_{\rho,N}\cap

1202: L(\mathcal{X})^N)\right\}. \label{eq:sp-r-t}

1203: \end{eqnarray}

1204: \end{thm}

1205: \begin{note*}

1206: Compare (\ref{eq:sp-r-f}) and (\ref{eq:sp-r-t}) with

1207: (\ref{eq:1-clos}).

1208: \end{note*}

1209: \begin{proof}[Proof of Theorem~\ref{thm:n-clos}]

1210: The inclusion ``$\supset$" in (\ref{eq:sp-r-f}) and

1211: (\ref{eq:sp-r-t}) follows from Remarks \ref{rem:n-sp-rad} and

1212: \ref{rem:n-ext}, and the fact that the set of

1213: $\mathcal{C}^N$-bounded holomorphic operator-valued functions on

1214: $\mathbb{D}^N$ (resp., $N$-tuples of bounded operators) with

1215: spectral radius at most one is closed in $H_N^\infty(\mathcal{X})$

1216: (resp., $L(\mathcal{X})^N$).

1217: 

1218: To show the inclusion ``$\subset$" in (\ref{eq:sp-r-f}), observe

1219: that for $f\in C_{\infty,N}^{(\infty)}\cap

1220: H_N^\infty(\mathcal{X})$ and $0<r<1$, $\nu^{(N,\infty)}(rf)\leq

1221: r<1$. By property (viii) from Theorem~\ref{thm:n-rad}, for

1222: $\rho_0>0$ big enough, $w_{\rho_0,N}^{(\infty)}(rf)<1$, and by

1223: property (iv) from the same theorem,

1224: \[ rf\in C_{\rho_0,N}^{(\infty)}\cap

1225: H_N^\infty(\mathcal{X})\subset\clos\left\{\bigcup_{0<\rho<\infty}(C_{\rho,N}^{(\infty)}\cap

1226: H_N^\infty(\mathcal{X}))\right\}.\] Passing to the limit as

1227: $r\uparrow 1$, we get

1228: \[ f\in\clos\left\{\bigcup_{0<\rho<\infty}(C_{\rho,N}\cap

1229: H_N^\infty(\mathcal{X}))\right\},\] and the inclusion ``$\subset$"

1230: in (\ref{eq:sp-r-f}) follows. Analogously for the inclusion

1231: ``$\subset$" in (\ref{eq:sp-r-t}).

1232: \end{proof}

1233: In view of property (iv) in Theorem~\ref{thm:n-rad}, let us call

1234: the elements of the class $C_{\rho,N}$ \emph{($N$-variable)

1235: $\rho$-contractions}.

1236: 

1237: 

1238: \section{On similarity of $\rho$-contractions to $1$-contractions in several variables}\label{sec:similar}

1239: An $N$-tuple of operators $\mathbf{A}=(A_1,\ldots,A_N)\in

1240: L(\mathcal{X})^N$ is said to be \emph{simultaneously similar} to

1241: an $N$-tuple of operators $\mathbf{B}=(B_1,\ldots,B_N)\in

1242: L(\mathcal{Y})^N$ if there exists a boundedly invertible operator

1243: $S\in L(\mathcal{X,Y})$ such that

1244: \begin{equation}\label{eq:n-similar}

1245:     A_k=S^{-1}B_kS,\quad k=1,\ldots,N,

1246: \end{equation}

1247: or equivalently,

1248: \begin{equation}\label{eq:n-pencil-similar}

1249:     z\mathbf{A}=S^{-1}(z\mathbf{B})S,\quad z\in\mathbb{C}^N.

1250: \end{equation}

1251: \begin{note*}

1252: Compare (\ref{eq:n-similar}) and (\ref{eq:n-pencil-similar}) with

1253: (\ref{eq:1-similar}).

1254: \end{note*}

1255: \begin{thm}\label{thm:non-similar}

1256: For any $\rho>1$ and $N>1$, there exists an

1257: $\mathbf{A}=(A_1,\ldots,A_N)\in C_{\rho,N}$ which is not

1258: simultaneously similar to any $\mathbf{T}=(T_1,\ldots,T_N)\in

1259: C_{1,N}$.

1260: \end{thm}

1261: \begin{proof}

1262: Let $N=2$, and for any $\varepsilon\geq 0$ set

1263: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})\in

1264: L(\mathbb{C}^3)^2$, where

1265: \[ A_1^{(\varepsilon)}:=\left[\begin{array}{lll}

1266:   0 & \frac{1+\varepsilon}{\sqrt{2}} & 0 \\

1267:   0 & 0 & 0 \\

1268:   -\frac{1+\varepsilon}{\sqrt{2}} & 0 & 0

1269: \end{array}\right],\quad  A_2^{(\varepsilon)}:=\left[\begin{array}{ccc}

1270:   0 & 0 & \frac{1+\varepsilon}{\sqrt{2}} \\

1271:   \frac{1+\varepsilon}{\sqrt{2}} & 0 & 0 \\

1272:   0 & 0 & 0

1273: \end{array}\right].  \]

1274: Then for any $\varepsilon\geq 0$ and $z\in\mathbb{C}^2$,

1275: \begin{eqnarray*}

1276: z\mathbf{A}^{(\varepsilon)} &=& \left[\begin{array}{ccc}

1277:   0 & \frac{1+\varepsilon}{\sqrt{2}}z_1 & \frac{1+\varepsilon}{\sqrt{2}}z_2 \\

1278:   \frac{1+\varepsilon}{\sqrt{2}}z_2 & 0 & 0 \\

1279:   -\frac{1+\varepsilon}{\sqrt{2}}z_1 & 0 & 0

1280: \end{array}\right], \\ (z\mathbf{A}^{(\varepsilon)})^2 &=& \left[\begin{array}{ccc}

1281:   0 & 0 & 0 \\

1282:   0 & \frac{(1+\varepsilon)^2}{2}z_1z_2 & \frac{(1+\varepsilon)^2}{2}z_2^2 \\

1283:   0 & -\frac{(1+\varepsilon)^2}{2}z_1^2 & -\frac{(1+\varepsilon)^2}{2}z_1z_2

1284: \end{array}\right],\\

1285: (z\mathbf{A}^{(\varepsilon)})^3 &=&

1286: (z\mathbf{A}^{(\varepsilon)})^4=\ldots=0,

1287: \end{eqnarray*}

1288: i.e., $z\mathbf{A}^{(\varepsilon)}$ is a nilpotent operator of

1289: degree 3. Hence, for any $\rho>1$ and $z\in\mathbb{D}^2$,

1290: \begin{eqnarray*}

1291: \|\varphi_{\rho,2}^{\mathbf{A}^{(0)}}(z)\| &=& \left\|

1292: z\mathbf{A}^{(0)}((\rho-1)z\mathbf{A}^{(0)}-\rho

1293: I)^{-1}\right\|=\left\|\frac{z\mathbf{A}^{(0)}}{\rho}\left(I-\frac{\rho-1}{\rho}z\mathbf{A}^{(0)}\right)^{-1}\right\|

1294: \\

1295: &=&

1296: \left\|\frac{z\mathbf{A}^{(0)}}{\rho}+(\rho-1)\left(\frac{z\mathbf{A}^{(0)}}{\rho}\right)^2\right\|\leq\frac{1}{\rho}\|

1297: z\mathbf{A}^{(0)}\|+\frac{\rho-1}{\rho^2}\|

1298: (z\mathbf{A}^{(0)})^2\| \\

1299: &=& \frac{1}{\rho}\left\|\left[\begin{array}{ccc}

1300:   0 & \frac{z_1}{\sqrt{2}} & \frac{z_2}{\sqrt{2}} \\

1301:   \frac{z_2}{\sqrt{2}} & 0 & 0 \\

1302:   -\frac{z_1}{\sqrt{2}} & 0 & 0

1303: \end{array}\right]\right\| +\frac{\rho-1}{\rho^2}\left\|\left[\begin{array}{c}

1304:   0 \\

1305:   \frac{z_2}{\sqrt{2}} \\

1306:  -\frac{z_1}{\sqrt{2}}

1307: \end{array}\right]\left[\begin{array}{c}

1308:   0 \\

1309:   \frac{z_1}{\sqrt{2}} \\

1310:  \frac{z_2}{\sqrt{2}}

1311: \end{array}\right]^T\right\| \\

1312: &\leq & \frac{1}{\rho}

1313: +\frac{\rho-1}{\rho^2}=\frac{2\rho-1}{\rho^2}<1.

1314: \end{eqnarray*}

1315: Then, due to the von Neumann inequality in two variables

1316: \cite{An}, one has

1317: \[\|\varphi_{\rho,2}^{\mathbf{A}^{(0)}}(\mathbf{C})\|\leq\frac{2\rho-1}{\rho^2}<1,

1318: \quad \mathbf{C}\in\mathcal{C}^2,\] i.e.,

1319: $\varphi_{\rho,2}^{\mathbf{A}^{(0)}}\in\mathcal{AS}_2(\mathbb{C}^3)$.

1320: Analogously, for $\varepsilon>0$ small enough (the choice of

1321: $\varepsilon$ depends on $\rho$), one has

1322: \[\sup_{\mathbf{C}\in\mathcal{C}^2}\|\varphi_{\rho,2}^{\mathbf{A}^{(\varepsilon)}}(\mathbf{C})\|=

1323: \sup_{z\in\mathbb{D}^2}\|\varphi_{\rho,2}^{\mathbf{A}^{(\varepsilon)}}(\mathbf{z})\|\leq\frac{1+\varepsilon}{\rho}+

1324: (\rho-1)\left(\frac{1+\varepsilon}{\rho}\right)^2<1,\] i.e.,

1325: $\varphi_{\rho,2}^{\mathbf{A}^{(\varepsilon)}}\in\mathcal{AS}_2(\mathbb{C}^3)$,

1326: and by Theorem~\ref{thm:n-char}, $\mathbf{A}^{(\varepsilon)}\in

1327: C_{\rho,2}$.

1328: 

1329: Let us show now that for any $\varepsilon>0$ the pair

1330: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})$

1331: is not simultaneously similar to any pair $\mathbf{T}=(T_1,T_2)\in

1332: C_{1,2}$. Observe that

1333: \begin{eqnarray*}

1334: (A_1^{(\varepsilon)}+A_2^{(\varepsilon)})(A_1^{(\varepsilon)}-A_2^{(\varepsilon)})

1335: &=& \left[\begin{array}{ccc}

1336:   0 & \frac{1+\varepsilon}{\sqrt{2}} & \frac{1+\varepsilon}{\sqrt{2}} \\

1337:   \frac{1+\varepsilon}{\sqrt{2}} & 0 & 0 \\

1338:   -\frac{1+\varepsilon}{\sqrt{2}} & 0 & 0

1339: \end{array}\right]\left[\begin{array}{ccc}

1340:   0 & \frac{1+\varepsilon}{\sqrt{2}} & -\frac{1+\varepsilon}{\sqrt{2}} \\

1341:   -\frac{1+\varepsilon}{\sqrt{2}} & 0 & 0 \\

1342:   -\frac{1+\varepsilon}{\sqrt{2}} & 0 & 0

1343: \end{array}\right] \\

1344: &=& \left[\begin{array}{ccc}

1345:   -(1+\varepsilon)^2 & 0 & 0 \\

1346:   0 & \frac{(1+\varepsilon)^2}{2} & -\frac{(1+\varepsilon)^2}{2} \\

1347:   0 & -\frac{(1+\varepsilon)^2}{2} & \frac{(1+\varepsilon)^2}{2}

1348: \end{array}\right].

1349: \end{eqnarray*}

1350: Then

1351: \begin{equation}\label{eq:lim}

1352:     \lim_{n\rightarrow +\infty}\|

1353: [(A_1^{(\varepsilon)}+A_2^{(\varepsilon)})(A_1^{(\varepsilon)}-A_2^{(\varepsilon)})]^n\|=\infty.

1354: \end{equation}

1355: On the other hand, if

1356: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})$

1357: is simultaneously similar to some  $\mathbf{T}=(T_1,T_2)\in

1358: C_{1,2}$ then for any $n\in\mathbb{N}$ one would have

1359: \[\|

1360: [(A_1^{(\varepsilon)}+A_2^{(\varepsilon)})(A_1^{(\varepsilon)}-A_2^{(\varepsilon)})]^n\|=\|

1361: S^{-1}[(T_1+T_2)(T_1-T_2)]^nS\|\leq\| S\|\|S^{-1}\|<\infty,\]

1362: since $\| T_1\pm T_2\|\leq 1$. We get a contradiction with

1363: (\ref{eq:lim}).

1364: 

1365: Examples  of $N$-tuples of operators from $C_{\rho,N},\ \rho>1$,

1366: which are not simultaneously similar to any $\mathbf{T}\in

1367: C_{1,N}$ for the case $N>2$ can be obtained from the examples of

1368: pairs $\mathbf{A}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})$

1369: above, for sufficiently small $\varepsilon>0$, by setting zeros

1370: for the rest of operators in these $N$-tuples:

1371: $\widetilde{\mathbf{A}}^{(\varepsilon)}:=(A_1^{(\varepsilon)},A_2^{(\varepsilon)},0,\ldots,0)$.

1372: \end{proof}

1373: Let $\mathbf{A}=(A_1,\ldots,A_N)\in L(\mathcal{X})^N$. Then

1374: $\widetilde{\mathbf{A}}=(\widetilde{A}_1,\ldots,\widetilde{A}_N)\in

1375: L(\widetilde{\mathcal{X}})^N$ is called a \emph{uniform

1376: $\rho$-dilation} of $\mathbf{A}$ if

1377: $\widetilde{\mathcal{X}}\supset\mathcal{X}$ and

1378: \begin{equation}\label{eq:uniform}

1379: \forall n\in\mathbb{N},\ \forall i_1,\ldots,i_n\in\{

1380: 1,\ldots,N\},\quad A_{i_1}\cdots A_{i_n}=\rho

1381: P_\mathcal{X}\widetilde{A}_{i_1}\cdots\widetilde{A}_{i_n}|\mathcal{X},

1382: \end{equation}

1383: or equivalently,

1384: \begin{equation}\label{eq:uniform'}

1385: \forall n\in\mathbb{N},\ \forall

1386: z^{(1)},\ldots,z^{(n)}\in\mathbb{C}^N,\quad

1387: z^{(1)}\mathbf{A}\cdots z^{(n)}\mathbf{A}=\rho

1388: P_\mathcal{X}z^{(1)}\widetilde{\mathbf{A}}\cdots

1389: z^{(n)}\widetilde{\mathbf{A}}|\mathcal{X}.

1390: \end{equation}

1391: \begin{note*}

1392: Compare (\ref{eq:uniform}) and (\ref{eq:uniform'}) with

1393: (\ref{eq:n-rhodil}) and (\ref{eq:n-rhodil'}).

1394: \end{note*}

1395: Clearly, a uniform $\rho$-dilation is a $\rho$-dilation. If

1396: $\widetilde{\mathbf{A}}\in L(\widetilde{\mathcal{X}})^N$ is a

1397: uniform $\rho$-dilation of $\mathbf{A}\in L(\mathcal{X})$, and for

1398: any $\zeta\in\mathbb{T}^N$, $\zeta\widetilde{\mathbf{A}}$ is a

1399: unitary operator, then $\widetilde{\mathbf{A}}$ is called a

1400: \emph{uniform unitary $\rho$-dilation}  of $\mathbf{A}$. Denote by

1401: $C_{\rho,N}^u$ the class of $N$-tuples of operators

1402: $\mathbf{A}=(A_1,\ldots,A_N)$ on a common Hilbert space which

1403: admit a uniform unitary $\rho$-dilation. Clearly,

1404: $C_{\rho,N}^u\subset C_{\rho,N}$.

1405: \begin{thm}\label{thm:similar}

1406: Any $\mathbf{A}=(A_1,\ldots,A_N)\in C_{\rho,N}^u$ is

1407: simultaneously similar to some $\mathbf{T}=(T_1,\ldots,T_N)\in

1408: C_{1,N}^u$.

1409: \end{thm}

1410: \begin{proof}

1411: Let $\mathbf{A}=(A_1,\ldots,A_N)\in C_{\rho,N}^u\cap

1412: L(\mathcal{X})^N$, and $\mathbf{U}=(U_1,\ldots,U_N)\in

1413: L(\widetilde{\mathcal{X}})^N$ be a uniform unitary $\rho$-dilation

1414: of $\mathbf{A}$. Let $\mathcal{A}\subset

1415: L(\widetilde{\mathcal{X}})$ be the minimal $C^*$-algebra which

1416: contains the operators

1417: $I_{\widetilde{\mathcal{X}}},U_1,\ldots,U_N$, and

1418: $\mathcal{B}\subset L(\widetilde{\mathcal{X}})$ be the minimal

1419: algebra over $\mathbb{C}$ which contains the operators

1420: $U_1,\ldots,U_N$. Clearly, $\mathcal{B}\subset\mathcal{A}$. Let

1421: $\varphi:\mathcal{B}\longrightarrow L(\mathcal{X})$ be a

1422: homomorphism defined on the generators as

1423: \[\varphi: U_k\longmapsto A_k,\ k=1,\ldots,N.\]

1424: The algebra $\mathcal{B}$ consists of operators of the form

1425: \[ p(\mathbf{U})=\sum_{1\leq k\leq m,\ i_1,\ldots,i_k\in\{ 1,\ldots,N\}

1426: }\alpha_{i_1,\ldots,i_k}U_{i_1}\cdots U_{i_k},\] where

1427: $\alpha_{i_1,\ldots,i_k}\in\mathbb{C}$ for all

1428: $i_1,\ldots,i_k\in\{ 1,\ldots,N\}$. Then

1429: \begin{eqnarray*}

1430: \varphi(p(\mathbf{U})) &=&

1431: \varphi(\sum\alpha_{i_1,\ldots,i_k}U_{i_1}\cdots

1432: U_{i_k})=\sum\alpha_{i_1,\ldots,i_k}A_{i_1}\cdots

1433: A_{i_k}\\

1434: &=& p(\mathbf{A})=\rho P_\mathcal{X}p(\mathbf{U})|\mathcal{X}.

1435: \end{eqnarray*}

1436: Therefore, if $p(\mathbf{U})=0$ then $\varphi(p(\mathbf{U}))=0$,

1437: and $\varphi$ is correctly defined. The homomorphism $\varphi$ is

1438: \emph{completely bounded}, i.e.,

1439: \[\|\varphi\|_{cb}:=\sup_{n\in\mathbb{N}}\|\id_n\otimes\varphi\|

1440: <\infty,\] where $\id_n$ is the identical map of the matrix

1441: algebra $\mathcal{M}_n(\mathbb{C})$ onto itself. Moreover,

1442: $\|\varphi\|_{cb}\leq\rho$. Indeed, for any $n\in\mathbb{N}$ and a

1443: polynomial $n\times n$ matrix of $N$ non-commuting variables,

1444: \[ P(\mathbf{X})=[p_{ij}(\mathbf{X})]_{i,j=1}^n=\left[\sum_{1\leq k\leq m,\ i_1,\ldots,i_k\in\{ 1,\ldots,N\}

1445: }\alpha_{i_1,\ldots,i_k}^{(ij)}X_{i_1}\cdots

1446: X_{i_k}\right]_{i,j=1}^n,\]

1447: \begin{eqnarray*}

1448: \|(\id_n\otimes\varphi)(P(\mathbf{U}))\| &=&

1449: \|(\id_n\otimes\varphi)\left([p_{ij}(\mathbf{U})]_{i,j=1}^n\right)\|

1450: =\|[\varphi(p_{ij}(\mathbf{U}))]_{i,j=1}^n\| \\

1451: &=& \|[p_{ij}(\mathbf{A})]_{i,j=1}^n\|=\|[\rho P_\mathcal{X}p_{ij}(\mathbf{U})|\mathcal{X}]_{i,j=1}^n\| \\

1452: &=& \rho\|(I_{\mathbb{C}^n}\otimes

1453: P_\mathcal{X})[p_{ij}(\mathbf{U})]_{i,j=1}^n|\mathbb{C}^n\otimes\mathcal{X}\|

1454: \\

1455: &\leq & \rho\| [p_{ij}(\mathbf{U})]_{i,j=1}^n\| = \rho\|

1456: P(\mathbf{U})\|.

1457: \end{eqnarray*}

1458: Then, by Theorem~3.1 in \cite{Pa}, there exist a Hilbert space

1459: $\mathcal{N}$, a \emph{completely contractive} homomorphism

1460: $\gamma:\mathcal{B}\longrightarrow L(\mathcal{N})$ (i.e., such

1461: that $\|\gamma\|_{cb}\leq 1$), and a boundedly invertible operator

1462: $S\in L(\mathcal{X},\mathcal{N})$ such that

1463: \[ \varphi(b)=S^{-1}\gamma(b)S,\quad b\in\mathcal{B}.\]

1464: Moreover, as was shown in the proof of Theorem~3.1 in \cite{Pa},

1465: $\gamma$ can be chosen in the form

1466: \[ \gamma(b)=P_\mathcal{N}\pi(b)|\mathcal{N},\quad b\in\mathcal{B},\]

1467: where $\pi:\mathcal{A}\longrightarrow L(\mathcal{K})$ is a

1468: $*$-homomorphism, for some Hilbert space

1469: $\mathcal{K}\supset\mathcal{N}$. In addition, it follows from

1470: Theorem~2.7 and the proof of Theorem~2.8 in \cite{Pa} that one can

1471: choose $\mathcal{K}=\mathcal{K}_1\oplus\mathcal{K}_1$, for some

1472: Hilbert space $\mathcal{K}_1$, and

1473: \[ \pi(a)=\pi_1(a)\oplus 0,\quad a\in\mathcal{A},\]

1474: where $\pi_1:\mathcal{A}\longrightarrow L(\mathcal{K}_1)$ is a

1475: unital $*$-homomorphism. Set \[ T_k:=\gamma(U_k)\in

1476: L(\mathcal{N}),\quad k=1,\ldots,N.\] Then

1477: \[ A_k=\varphi(U_k)=S^{-1}T_kS,\quad k=1,\ldots,N.\]

1478: It remains to show that $\mathbf{T}=(T_1,\ldots,T_N)\in

1479: C_{1,N}^u$. Set \[W_k:=\pi(U_k)\in L(\mathcal{K}),\quad

1480: k=1,\ldots,N.\]  Since for any $n\in\mathbb{N}$ and

1481: $i_1,\ldots,i_n\in\{ 1,\ldots,N\}$ one has

1482: \begin{eqnarray*}

1483: T_{i_1}\cdots T_{i_n} &=&

1484: \gamma(U_{i_1}\cdots

1485: U_{i_n})=P_\mathcal{N}\pi(U_{i_1}\cdots

1486: U_{i_n})|\mathcal{N}\\

1487: &=& P_\mathcal{N}W_{i_1}\cdots W_{i_n}|\mathcal{N},

1488: \end{eqnarray*}

1489: $\mathbf{W}=(W_1,\ldots,W_N)$ is a uniform $1$-dilation of

1490: $\mathbf{T}=(T_1,\ldots,T_N)$, however, still not unitary.

1491: Actually,

1492: \[ W_k=\pi(U_k)=\pi_1(U_k)\oplus 0\ (=: W_k^{(1)}\oplus 0),\quad  k=1,\ldots,N.\]

1493: Since $\pi_1$ is a unital $*$-homomorphism, and for any

1494: $\zeta\in\mathbb{T}^N$,

1495: \[

1496: (\zeta\mathbf{U})^*\zeta\mathbf{U}=I_{\widetilde{\mathcal{X}}}=\zeta\mathbf{U}(\zeta\mathbf{U})^*,\]

1497: one has, for any $\zeta\in\mathbb{T}^N$,

1498: \[

1499: (\zeta\mathbf{W}^{(1)})^*\zeta\mathbf{W}^{(1)}=I_{\mathcal{K}_1}=\zeta\mathbf{W}^{(1)}(\zeta\mathbf{W}^{(1)})^*,\]

1500: where $\mathbf{W}^{(1)}=(W_1^{(1)},\ldots,W_N^{(1)})$. Set

1501: \[ \widetilde{W}_k:=W_k^{(1)}\oplus\delta_{1k}V\in

1502: L(\mathcal{K}_1\oplus\mathcal{R}),\quad k=1,\ldots,N,\] where

1503: $\delta_{ij}$ is the Kronecker symbol, and $V$ is a unitary

1504: dilation of the zero operator on $\mathcal{K}_1$, e.g., the

1505: two-sided shift on the space

1506: $\mathcal{R}:=l^2(\mathcal{K}_1)=\bigoplus_{-\infty}^{+\infty}\mathcal{K}_1$

1507: (here we identify the space $\mathcal{K}_1$ with the subspace

1508: $\ldots\oplus\{ 0\}\oplus\{ 0\}\oplus\mathcal{K}_1\oplus\{

1509: 0\}\oplus\{ 0\}\oplus\ldots$ in $\mathcal{R}$). Then, for any

1510: $\zeta\in\mathbb{T}^N$,

1511: \[

1512: (\zeta\widetilde{\mathbf{W}})^*\zeta\widetilde{\mathbf{W}}=I_{\mathcal{K}_1\oplus\mathcal{R}}=

1513: \zeta\widetilde{\mathbf{W}}(\zeta\widetilde{\mathbf{W}})^*,\] and

1514: $\widetilde{\mathbf{W}}=(\widetilde{W}_1,\ldots,\widetilde{W}_N)\in

1515: L(\mathcal{K}_1\oplus\mathcal{R})$ is a uniform unitary

1516: $1$-dilation of $\mathbf{W}=(W_1,\ldots,W_N)$, and therefore, of

1517: $\mathbf{T}=(T_1,\ldots,T_N)$. Thus, $\mathbf{T}\in C_{1,N}^u$, as

1518: required.

1519: \end{proof}

1520: Theorem~\ref{thm:similar} is similar to the result of

1521: G.~Popescu \cite{Po2} on simultaneous similarity of

1522: $\rho$-contractions to  $1$-contractions in several variables,

1523: however his notion of multivariable $\rho$-contractions is

1524: different. Let us clarify the relation between these two results.

1525: Denote by $C_{\rho,N}^P$ (we use here this notation instead of

1526: just $C_\rho$, as in \cite{Po2}) the \emph{Popescu class} of all

1527: $N$-tuples $\mathbf{A}=(A_1,\ldots,A_N)$ of bounded linear

1528: operators on a common Hilbert space, say $\mathcal{X}$, which have

1529: a \emph{uniform isometric $\rho$-dilation}, i.e., such an

1530: $N$-tuple of operators $\mathbf{V}=(V_1,\ldots,V_N)\in

1531: L(\widetilde{\mathcal{X}})^N,\

1532: \widetilde{\mathcal{X}}\supset\mathcal{X}$, for which

1533: \begin{description}

1534:     \item[(1)] $V_k^*V_k=I_{\widetilde{\mathcal{X}}},\quad k=1,\ldots,N$;

1535: \item[(2)] $V_k^*V_j=0,\quad k\neq j$;

1536: \item[(3)] $\forall n\in\mathbb{N},\ \forall i_1,\ldots,i_n,\quad A_{i_1}\cdots A_{i_n}=\rho P_\mathcal{X}V_{i_1}\cdots

1537: V_{i_n}|\mathcal{X}.$

1538: \end{description}

1539: Condition (2) in this definition can be replaced by

1540: \begin{description}

1541: \item[(2')] $\sum\limits_{k=1}^NV_kV_k^*\preceq

1542: I_{\widetilde{\mathcal{X}}}$,

1543: \end{description}

1544: \noindent since (1)$\&$(2)$\Longleftrightarrow$(1)$\&$(2').

1545: According to \cite{Po1}, the class $C_{1,N}^P$ coincides with the

1546: class of $N$-tuples of operators $\mathbf{A}=(A_1,\ldots,A_N)$

1547: ($\in L(\mathcal{X})^N$, for some Hilbert space $\mathcal{X}$)

1548: such that \[ \sum_{k=1}^NA_kA_k^*\preceq I_\mathcal{X}.\] By

1549: Theorem~4.5 in \cite{Po2}, any $\mathbf{A}=(A_1,\ldots,A_N)\in

1550: C_{\rho,N}^P,\ \rho>0$, is simultaneously similar to some

1551: $\mathbf{T}=(T_1,\ldots,T_N)\in C_{1,N}^P$. This is a generalization of the

1552: theorem of Sz.-Nagy and Foia\c{s} \cite{SzNF2} to several

1553: variables. Theorem~\ref{thm:similar} of the present paper is a different generalization of the same result,

1554:  since our classes

1555: $C_{\rho,N}^u$ are different from Popescu's classes $C_{\rho,N}^P$

1556: for $N>1$. More precisely, the following is true.

1557: \begin{thm}\label{thm:pop}

1558: For any $N>1$ and $\rho>0$, $C_{\rho,N}^u\varsubsetneq

1559: C_{\rho,N}^P$.

1560: \end{thm}

1561: \begin{proof}

1562: Let $\mathbf{A}\in C_{\rho,N}^u\cap L(\mathcal{X})^N,\ N>1$, and

1563: $\mathbf{U}\in L(\widetilde{\mathcal{X}})^N$ be a uniform unitary

1564: $\rho$-dilation of $\mathbf{A}$. Since for any

1565: $\zeta\in\mathbb{T}^N$ the operator $\zeta\mathbf{U}$ is unitary,

1566: it follows that

1567: \[

1568: \zeta\mathbf{U}(\zeta\mathbf{U})^*=I_{\widetilde{\mathcal{X}}},\quad\zeta\in\mathbb{T}^N,\]

1569: which implies

1570: \[\sum_{k=1}^NU_kU_k^*=I_{\widetilde{\mathcal{X}}}.\]

1571: Thus, by \cite{Po1}, $\mathbf{U}\in C_{1,N}^P$. Let $\mathbf{V}\in

1572: L(\hat{\mathcal{X}})^N$ be a uniform isometric $1$-dilation of

1573: $\mathbf{U}$ in the sense of Popescu. Then for any

1574: $n\in\mathbb{N}$ and $i_1,\ldots,i_n\in\{ 1,\ldots,N\}$,

1575: \begin{eqnarray*}

1576: A_{i_1}\cdots A_{i_n} &=& \rho

1577: P_\mathcal{X}U_{i_1}\cdots U_{i_n}|\mathcal{X}=\rho 

1578: P_\mathcal{X}(P_{\widetilde{\mathcal{X}}}V_{i_1}\cdots

1579: V_{i_n}|\widetilde{\mathcal{X}})|\mathcal{X} \\

1580: &=& \rho P_\mathcal{X}V_{i_1}\cdots

1581: V_{i_n}|\mathcal{X},

1582: \end{eqnarray*}

1583: i.e., $\mathbf{V}$ is a uniform isometric $\rho$-dilation of

1584: $\mathbf{A}$ in the sense of Popescu. Thus, $\mathbf{A}\in

1585: C_{\rho,N}^P$. This proves the inclusion $C_{\rho,N}^u\subset

1586: C_{\rho,N}^P$.

1587: 

1588: Let us prove that this inclusion is proper for any $N>1$ and

1589: $\rho>0$. Firstly, consider the case $N=2$. Let $B\in

1590: L(\mathcal{X}_0)$ be any operator of the class $C_\rho$ with $\|

1591: B\|=\rho$. For example,

1592: \[ B:=\left[\begin{array}{cc}

1593:   0 & \rho \\

1594:   0 & 0

1595: \end{array}\right]\in L(\mathbb{C}^2)\]

1596: satisfies $B^2=0$ and $\| B\|=\rho$, therefore by properties (iii)

1597: and (xi) in Theorem~\ref{thm:1-rad}, $w_\rho(B)=1$, and by

1598: property (iv) in the same theorem, $B\in C_\rho$. Set

1599: $\mathcal{X}:=\mathcal{X}_0\oplus\mathcal{X}_0$,

1600: \begin{equation}\label{eq:a's}

1601:     A_1:=\left[\begin{array}{cc}

1602:   B & 0 \\

1603:   0 & 0

1604: \end{array}\right]\in L(\mathcal{X}),\quad  A_2:=\left[\begin{array}{cc}

1605:   0 & 0 \\

1606:   B & 0

1607: \end{array}\right]\in L(\mathcal{X}).

1608: \end{equation}

1609: Let $U\in L(\widetilde{\mathcal{X}}_0)$ be a unitary

1610: $\rho$-dilation of $B$. Set

1611: $\widetilde{\mathcal{X}}:=\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0\oplus\ldots$,

1612: and identify $\mathcal{X}=\mathcal{X}_0\oplus\mathcal{X}_0$ with

1613: the subspace $\mathcal{X}_0\oplus\mathcal{X}_0\oplus\{ 0\}\oplus\{

1614: 0\}\oplus\ldots$ in $\widetilde{\mathcal{X}}$. Set

1615: \[ V_1:=\left[\begin{array}{ccc}

1616:   \fbox{$\begin{array}{c}

1617:     U \\

1618:     0

1619:   \end{array}$} &  &  \\

1620:    & \fbox{$\begin{array}{c}

1621:     U \\

1622:     0

1623:   \end{array}$} &  \\

1624:    &  & \ddots \\

1625: \end{array}\right]\in L(\widetilde{\mathcal{X}}),\quad  V_2:=\left[\begin{array}{ccc}

1626:   \fbox{$\begin{array}{c}

1627:     0 \\

1628:     U

1629:   \end{array}$} &  &  \\

1630:    & \fbox{$\begin{array}{c}

1631:     0 \\

1632:     U

1633:   \end{array}$} &  \\

1634:    &  & \ddots \\

1635: \end{array}\right]\in L(\widetilde{\mathcal{X}}),\]

1636: i.e., the operators $V_1$ and $V_2$ are introduced here as

1637: infinite block-diagonal matrices with equal operator blocks

1638: $\left[\begin{array}{c}

1639:     U \\

1640:     0

1641:   \end{array}\right]\in

1642: L(\widetilde{\mathcal{X}}_0,\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0)$

1643: (resp., $\left[\begin{array}{c}

1644:     0 \\

1645:     U

1646:   \end{array}\right]\in

1647: L(\widetilde{\mathcal{X}}_0,\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0)$)

1648: on the main diagonal. We will show that the pair

1649: $\mathbf{V}=(V_1,V_2)$ is a uniform isometric $\rho$-dilation of

1650: the pair $\mathbf{A}=(A_1,A_2)$ in the sense of Popescu. First of

1651: all, observe that

1652: \[ V_1^*V_1=I_{\widetilde{\mathcal{X}}}=V_2^*V_2,\quad

1653: V_1^*V_2=V_2^*V_1=0.\] Next, the following relations hold:

1654: \begin{eqnarray*}

1655: \forall k\in\mathbb{N}, \ A_1^k= \left[\begin{array}{cc}

1656:   B^k & 0 \\

1657:   0 & 0

1658: \end{array}\right]=\left[\begin{array}{cc}

1659:   \rho P_{\mathcal{X}_0}U^k|\mathcal{X}_0 & 0 \\

1660:   0 & 0

1661: \end{array}\right] &=& \rho P_\mathcal{X}V_1^k|\mathcal{X};\\

1662: \forall k,n\in\mathbb{N}, \ \forall i_1,\ldots,i_n\in\{

1663: 1,2\},\ A_1^kA_2A_{i_1}\cdots A_{i_n} &=& 0\\

1664: &=& \rho P_\mathcal{X}V_1^kV_2V_{i_1}\cdots

1665: V_{i_n}|\mathcal{X}

1666: \end{eqnarray*}

1667: (since $A_1^kA_2=0,\

1668: P_{\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0\oplus\{

1669: 0\}\oplus\{ 0\}\oplus\ldots}V_1^kV_2=0$);

1670: \begin{eqnarray*}

1671:  A_2= \left[\begin{array}{cc}

1672:   0 & 0 \\

1673:   B & 0

1674: \end{array}\right]=\left[\begin{array}{cc}

1675:   0 & 0 \\

1676:   \rho P_{\mathcal{X}_0}U|\mathcal{X}_0 & 0

1677: \end{array}\right] &=& \rho P_\mathcal{X}V_2|\mathcal{X};\\

1678: \forall k,n\in\mathbb{N}, \ \forall i_1,\ldots,i_n\in\{

1679: 1,2\},\ A_2^{k+1}A_{i_1}\cdots A_{i_n} &=& 0\\

1680: &=& \rho P_\mathcal{X}V_2^{k+1}V_{i_1}\cdots

1681: V_{i_n}|\mathcal{X}

1682: \end{eqnarray*}

1683: (since $A_2^2=0,\

1684: P_{\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0\oplus\{

1685: 0\}\oplus\{ 0\}\oplus\ldots}V_2^2=0$);

1686: \begin{eqnarray*}

1687: \forall k\in\mathbb{N}, \ A_2A_1^k = \left[\begin{array}{cc}

1688:   0 & 0 \\

1689:   B^{k+1} & 0

1690: \end{array}\right] &=& \left[\begin{array}{cc}

1691:   0 & 0 \\

1692:   \rho P_{\mathcal{X}_0}U^{k+1}|\mathcal{X}_0 & 0

1693: \end{array}\right] = \rho P_\mathcal{X}V_2V_1^k|\mathcal{X};\\

1694: \forall k\in\mathbb{N},\ A_2A_1^kA_2=0 &=&

1695: P_\mathcal{X}V_2V_1^kV_2|\mathcal{X};\\

1696: \forall k,n\in\mathbb{N},\ \forall i_1,\ldots,i_n\in\{

1697: 1,2\}, &\ &  A_2A_1^kA_2A_{i_1}\cdots A_{i_n}=0\\

1698: &=& \rho P_\mathcal{X}V_2V_1^kV_2V_{i_1}\cdots

1699: V_{i_n}|\mathcal{X}

1700: \end{eqnarray*}

1701: (since $A_1^kA_2=0,\

1702: P_{\widetilde{\mathcal{X}}_0\oplus\widetilde{\mathcal{X}}_0\oplus\{

1703: 0\}\oplus\{ 0\}\oplus\ldots}V_2V_1^kV_2=0$). Finally, we get

1704: \[ \forall n\in\mathbb{N},\ \forall i_1,\ldots,i_n\in\{

1705: 1,2\}, \   A_{i_1}\cdots A_{i_n}=\rho

1706: P_\mathcal{X}V_{i_1}\cdots V_{i_n}|\mathcal{X}.\]

1707: Thus, $\mathbf{V}$ is a uniform isometric $\rho$-dilation of

1708: $\mathbf{A}$ in the sense of Popescu. However, for any

1709: $\zeta\in\mathbb{T}^N$,

1710: \[ \|\zeta\mathbf{A}\|=\left\|\left[\begin{array}{cc}

1711:   \zeta_1B & 0 \\

1712:   \zeta_2B & 0

1713: \end{array}\right]\right\|=\sqrt{2}\| B\|=\sqrt{2}\rho>\rho.\]

1714: Therefore, $\zeta\mathbf{A}\notin C_\rho$ for all

1715: $\zeta\in\mathbb{T}^N$. We obtain $\mathbf{A}\in

1716: C_{\rho,2}^P\backslash C_{\rho,2}^u$ (moreover, $\mathbf{A}\notin

1717: C_{\rho,2}$).

1718: 

1719: For the case $N>2$ (and any $\rho>0$) an analogous example of

1720: $\widetilde{\mathbf{A}}\in C_{\rho,N}^P\backslash C_{\rho,N}^u$ is

1721: easily obtained from the previous one, by setting zeros for the

1722: rest of operators in the $N$-tuple, i.e.,

1723: $\widetilde{\mathbf{A}}:=(A_1,A_2,0,\ldots,0)$, where $A_1$ and

1724: $A_2$ are defined in (\ref{eq:a's}). In this case the construction

1725: of a uniform isometric $\rho$-dilation of $\mathbf{A}$ in the

1726: sense of Popescu should be slightly changed (we leave this to a

1727: reader as an easy exercise).

1728: \end{proof}

1729: \begin{rem}

1730: The pair

1731: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})$

1732: constructed in Theorem~\ref{thm:non-similar} doesn't belong to the

1733: class $C_{\rho,2}^u$ for any $\varepsilon>0$ and $\rho>1$. Indeed,

1734: we have shown in Theorem~\ref{thm:non-similar} that

1735: $\mathbf{A}^{(\varepsilon)}$ is not simultaneously similar to any

1736: $\mathbf{T}=(T_1,T_2)\in C_{1,2}$, not speaking of $\mathbf{T}\in

1737: C_{1,2}^u$. Thus, by Theorem~\ref{thm:pop},

1738: $\mathbf{A}^{(\varepsilon)}\notin C_{\rho,2}^u$. This can be shown

1739: also by the following estimate: if $\mathbf{A}^{(\varepsilon)}\in

1740: C_{\rho,2}^u$ for some $\varepsilon>0$ and $\rho>1$, then there

1741: exists a uniform unitary $\rho$-dilation

1742: $\mathbf{U}^{(\varepsilon)}=(U_1^{(\varepsilon)},U_2^{(\varepsilon)})$

1743: of

1744: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})$,

1745: and for any $n\in\mathbb{N}$,

1746: \begin{eqnarray*}

1747: \|

1748: [(A_1^{(\varepsilon)}+A_2^{(\varepsilon)})(A_1^{(\varepsilon)}-A_2^{(\varepsilon)})]^n\|

1749: &=& \|\rho

1750: P_\mathcal{X}[(U_1^{(\varepsilon)}+U_2^{(\varepsilon)})(U_1^{(\varepsilon)}-U_2^{(\varepsilon)})]^n|\mathcal{X}\|

1751: \\

1752: &\leq& \rho\|

1753: [(U_1^{(\varepsilon)}+U_2^{(\varepsilon)})(U_1^{(\varepsilon)}-U_2^{(\varepsilon)})]^n\|=\rho<\infty.

1754: \end{eqnarray*}

1755: This contradicts to (\ref{eq:lim}). Thus, for each $\rho>1$ we

1756: obtain for $\varepsilon>0$ small enough,

1757: $\mathbf{A}^{(\varepsilon)}=(A_1^{(\varepsilon)},A_2^{(\varepsilon)})\in

1758: C_{\rho,2}\backslash C_{\rho,2}^u$, as well as

1759: $\widetilde{\mathbf{A}}:=(A_1,A_2,0,\ldots,0)\in

1760: C_{\rho,N}\backslash C_{\rho,N}^u$.

1761: \end{rem}

1762: 

1763: \begin{thebibliography}{99}% Replace 9 by 99 if 10 or more references

1764: %

1765: % Please note the use of "\and" between author names below

1766: %

1767: \bibitem{Ag}

1768:  {J. Agler},

1769:  `On the representation of certain holomorphic functions defined on a

1770:  polydisc', in Topics in Operator Theory: Ernst D.~Hellinger Memorial

1771:   Volume (L.~de~Branges, I.~Gohberg, and J.~Rovnyak, eds.),

1772:  {\em  Oper.\ Theory\ Adv.\ Appl.\ }48 (1990) 47--66 (Birkh\"auser-Verlag,

1773:  Basel).

1774:  %

1775: \bibitem{AgMc}

1776: {J.~Agler \and J.~E.~McCarthy}, `Nevanlinna--Pick

1777: interpolation on the bidisk', {\em  J.\ Reine\ Angew.\ Math.\ }

1778: 506 (1999) 191--204.

1779: %

1780: \bibitem{An}

1781: {T.~Ando}, `On a pair of commutative contractions', {\em

1782: Acta\ Sci.\ Math.\ (Szeged)\ } 24 (1963) 88--90.

1783: %

1784: \bibitem{AnN}

1785: {T.~Ando \and K.~Nishio}, `Convexity properties of

1786: operator radii associated with unitary $\rho$-dilations', {\em

1787: Michigan\ Math.\ J.\ } 20 (1973) 303--307.

1788: %

1789: \bibitem{BaCa}

1790: {C.~Badea \and G.~Cassier}, `Constrained von Neumann

1791: inequalities', {\em  Adv.\ Math.\ } 166\ no.~2 (2002) 260--297.

1792: %

1793: \bibitem{BLTT}

1794: {J.~A.~Ball, W.~S.~Li, D.~Timotin \and T.~T.~Trent}, `A

1795: commutant lifting theorem on the polydisc', {\em Indiana\ Univ.\

1796: Math.\ J.\ } 48\ no.~2 (1999) 653--675.

1797: %

1798: \bibitem{BSV}

1799: {J.~A.~Ball, C.~Sadosky \and V.~Vinnikov}, `Conservative

1800: input-state-output systems with evolution on a multidimensional

1801: integer lattice', preprint.

1802: %

1803: \bibitem{BT}

1804: {J.~A.~Ball \and T.~T.~Trent}, `Unitary colligations,

1805: reproducing kernel Hilbert spaces, and Nevanlinna-Pick

1806: interpolation in several variables', {\em J.\ Funct.\ Anal.\ }

1807: 157\ no.~1 (1998) 1--61.

1808: %

1809: \bibitem{Be}

1810: {C.~A.~Berger}, `A strange dilation theorem', {\em

1811: Notices\ Amer.\ Math.\ Soc.\ } 12 (1965) 590.

1812: %

1813: \bibitem{CaF}

1814: {G.~Cassier \and T.~Fack}, `Contractions in von Neumann

1815: algebras', {\em  J.\ Funct.\ Anal.\ } 135\ no.~2 (1996) 297--338.

1816: %

1817: \bibitem{Da}

1818: {C.~Davis}, `The shell of a Hilbert-space operator', {\em

1819: Acta\ Sci.\ Math.\ (Szeged)\ }  29 (1968) 69--86.

1820: %

1821: \bibitem{DMcCW}

1822: {M.~A.~Dritschel, S.~McCullough \and H.~J.~Woerdeman},

1823: `Model theory for $\rho$-contractions, $\rho\le2$', {\em  J.\

1824: Operator\ Theory\ } 41\ no.~2 (1999) 321--350.

1825: %

1826: \bibitem{Du}

1827: {E.~Durszt}, `On unitary $\rho$-dilations of operators',

1828: {\em Acta\ Sci.\ Math.\ (Szeged)\ } 27 (1966) 247--250.

1829: %

1830: \bibitem{FH}

1831: {C.~K.~Fong \and J.~A.~R.~Holbrook}, `Unitarily invariant

1832: operator norms', {\em  Canad.\ J.\ Math.\ } 35\ no.~2 (1983)

1833: 274--299.

1834: %

1835: \bibitem{H1}

1836: {J.~A.~R.~Holbrook}, `On the power-bounded operators of

1837: Sz.-Nagy and Foia\c{s}', {\em Acta\ Sci.\ Math.\ (Szeged)\ } 29

1838: (1968) 299--310.

1839: %

1840: \bibitem{H2}

1841: {J.~A.~R.~Holbrook}, `Inequalities governing the operator

1842: radii associated with unitary $\rho $-dilations', {\em  Michigan\

1843: Math.\ J.\ } 18 (1971) 149--159.

1844: %

1845: \bibitem{K2}

1846: {D.~S.~Kalyuzhniy}, `Multiparametric dissipative linear

1847: stationary dynamical scattering systems: discrete case', {\em J.\

1848: Operator\ Theory} 43\ no.~2 (2000) 427--460.

1849: %

1850: \bibitem{K3}

1851: {D.~S.~Kalyuzhniy}, `Multiparametric dissipative linear

1852: stationary dynamical scattering systems: discrete case. II.

1853: Existence of conservative dilations', {\em Integral\ Equations\

1854: Operator\ Theory} 36\ no.~1 (2000) 107--120.

1855: %

1856: \bibitem{K4}

1857: {D.~S.~Kalyuzhniy}, `On the notions of dilation,

1858:   controllability, observability, and minimality in the theory of

1859:   dissipative scattering linear nD systems', in Proceedings CD of the

1860:    Fourteenth International Symposium of Mathematical Theory of

1861: Networks and Systems (MTNS), June 19--23, 2000, Perpignan, France

1862: (A.~El~Jai and M.~Fliess, Eds.), or \verb

1863: http://www.univ-perp.fr/mtns2000/articles/SI13_3.pdf/.

1864: %

1865: \bibitem{K6}

1866: {D.~S.~Kalyuzhniy-Verbovetzky}, `Cascade connections of

1867: linear systems and factorizations of holomorphic operator

1868: functions around a multiple zero in  several variables', {\em

1869: Math.\ Rep.\ (Bucur.)} 3(53)\ no.~4 (2001) 323--332.

1870: %

1871: \bibitem{K7}

1872: {D.~S.~Kalyuzhniy-Verbovetzky}, `On $J$-conservative

1873: scattering system realizations in several variables', {\em

1874: Integral\ Equations\ Operator\ Theory } 43\ no.~4 (2002) 450--465.

1875: %

1876: \bibitem{K1}

1877: {D.~S.~Kalyuzhny\u\i}, `The von Neumann inequality for

1878: linear matrix functions of several variables', {\em Mat.\ Zametki}

1879: 64\ no.~2 (1998) 218--223 (Russian). English transl. in {\em

1880: Math.\ Notes } 64\ no.~1--2 (1998) 186--189 (1999).

1881: %

1882: \bibitem{K5}

1883: {D.~S.~Kalyuzhny\u\i-Verbovetski\u\i}, `Cascade

1884: connections of multiparameter linear systems and the conservative

1885: realization of a decomposable inner operator function on the

1886: bidisk', {\em Mat.\ Stud. } 15\ no.~1 (2001) 65--76 (Russian).

1887: %

1888: \bibitem{NO}

1889: {T.~Nakazi \and K.~Okubo}, `$\rho$-contraction and

1890: $2\times 2$ matrix', {\em  Linear\ Algebra\ Appl.\ } 283\ no.~1-3

1891: (1998) 165--169.

1892: %

1893: \bibitem{vN}

1894: {J.~von Neumann}, `Eine Spektraltheorie f\"{u}r

1895: allgemeine Operatoren eines unit\"{a}ren Raumes', {\em Math.\

1896: Nachr. } 4 (1951) 258--281 (German).

1897: %

1898: \bibitem{OAn}

1899: {K.~Okubo \and T.~Ando}, `Operator radii of commuting

1900: products', {\em  Proc.\ Amer.\ Math.\ Soc.\ } 56\ no.~1 (1976)

1901: 203--210.

1902: %

1903: \bibitem{OS}

1904: {K.~Okubo \and I.~Spitkovsky}, `On the characterization

1905: of $2\times 2$ $\rho$-contraction matrices', {\em  Linear\

1906: Algebra\ Appl.\ } 325\ no.~1-3 (2001) 177--189.

1907: %

1908: \bibitem{Pa}

1909: {V.~I.~Paulsen}, `Every completely polynomially bounded

1910: operator is similar to a contraction', {\em J.\ Funct.\ Anal.\ }

1911: 55\ no.~1 (1984) 1--17.

1912: %

1913: \bibitem{Po1}

1914: {G.~Popescu}, `Isometric dilations for infinite sequences

1915: of noncommuting operators', {\em  Trans.\ Amer.\ Math.\ Soc.\ }

1916: 316\ no.~2 (1989) 523--536.

1917: %

1918: \bibitem{Po2}

1919: {G.~Popescu}, `Positive-definite functions on free

1920: semigroups', {\em  Canad.\ J.\ Math.\ } 48\ no.~4 (1996) 887--896.

1921: %

1922: \bibitem{SzN}

1923: {B.~Sz.-Nagy}, `Sur les contractions de l'espace de

1924: Hilbert', {\em  Acta\ Sci.\ Math.\ (Szeged)\ } 15 (1953) 87--92

1925: (French).

1926: %

1927: \bibitem{SzNF1}

1928: {B.~Sz.-Nagy \and C. Foia\c{s}}, `On certain classes of

1929: power-bounded operators in Hilbert space', {\em  Acta\ Sci.\

1930: Math.\ (Szeged)\ } 27 (1966) 17--25.

1931: %

1932: \bibitem{SzNF2}

1933: {B.~Sz.-Nagy \and C. Foia\c{s}}, `Similitude des

1934: op\'{e}rateurs de class ${\mathcal C}\sb{\rho }$ \`{a} des

1935: contractions', {\em C.\ R.\ Acad.\ Sci.\ Paris\ S\'{e}r.\ A-B\ }

1936: 264 (1967) A1063--A1065 (French).

1937: %

1938: \bibitem{SzNF}

1939:  {B.~Sz.-Nagy \and C. Foia\c{s}},

1940:  {\em Harmonic analysis of operators on Hilbert space}

1941:  (North-Holland, Amsterdam--London, 1970).

1942: %

1943: \bibitem{W}

1944: {J.~P.~Williams}, `Schwarz norms for operators', {\em

1945: Pacific\ J.\ Math.\ } 24 (1968) 181--188.

1946: %

1947: 

1948: \end{thebibliography}

1949: 

1950: \subsection*{Acknowledgements}

1951: I am grateful for the hospitality of the Universities of Leeds and

1952: Newcastle upon Tyne where a part of this work was carried out

1953: during my visits under the International Short Visit Scheme of the

1954: LMS (grant no. 5620). I wish to thank also Dr. Michael Dritschel

1955: from the University of Newcastle upon Tyne for useful discussions.

1956: 

1957: ----------------------------

1958: \end{document}

1959: % ------------------------------------------------------------------------

1960: