On Unitary N-Dilations for Tuples of Circulant Contractions and von Neumann’s Inequality ()
1. Introduction
In 1953, Sz-Nagy [1] [2] showed that every single contraction on a Hilbert space has a unitary dilation. This is an interesting tool which can be used to prove the von Neumann inequality [3] [4] [5] [6] which states that for any contraction linear operator T on a Hilbert space the following inequality:
holds for all complex polynomials
over the unit disk, where
denotes the supremum norm of p over the unit disk. In 1963, Ando proved that every pair of commuting contractions has a simultaneous commuting dilation [7] . However, Varopoulos [8] , Parrott [9] and Crabb-Davie [10] proved that this phenomenon fails for more than three commuting contractions. In 1978, Drury [11] , in connection to his generalization of von Neumann’s inequality, and then Arveson [12] , in 1998, proved the standard commuting dilation for tuples of commuting contractions. The problem of determining if a tuple of commuting (or non-commuting) contractions admits a unitary dilation has been pursued by many authors. Over the years, several conditions that guarantee the existence of a unitary dilation for an
-tuple of commuting contractions have been studied [13] . For example: tuples of doubly commuting contractions have unitary N-dilations (
) acting on a finite dimensional space [14] .
This result has many engineering applications [15] .
In the present paper, we introduce the spectral mapping factorization of tuples of circulant matrices and its matrix version. Circulant matrices have many applications in graph theory, cryptography, physics, signal and image processing, probability, statistics, numerical analysis, algebraic coding theory and many other areas [16] . The well known results on unitary dilations of doubly commuting sets of contractions allow us to extend Sz-Nagy’s Dilation Theorem and von Neumann’s inequality to the setting of tuples of circulant contractions.
Theorem 1.1. Let
be a positive integer. Every tuple of circulant contractions has a unitary N-dilation.
Theorem 1.2. von Neumann’s inequality holds for tuples of circulant contractions.
The matrix version of the spectral mapping factorization of tuples of circulant matrices allows us to introduce a new family of completely contractive homomorphisms over the algebra of complex polynomials defined on
.
2. Preliminaries
Throughout this paper H is a Hilbert space of finite dimension n. Let
be a complex matrix. Denote by
.
2.1. Operator Norm
Definition 2.1. Let
be a unital Banach algebra. We say that
is invertible if there is an element
such that
. In this case b is unique and written
. The set
is a group under multiplication.
If a is an element of
, the spectrum of a is defined as
and its spectral radius is defined to be
Definition 2.2. Let H be a Hilbert space and let
be a linear bounded operator on H. Then the operator norm of T denoted by
is defined by
If
, then the linear bounded operator T is called a contraction. In the case, H is a finite dimensional Hilbert space, then
2.2. Complex Polynomials
Let
be the unit poly disk and let
be a complex polynomial over
. Then
Let
be the algebra of complex polynomials over
. Given
a complex polynomial over
, let us set
where the supremum is taken over the family of all n-tuples of contractions on all Hilbert spaces. It is easy to see that
is finite, since it is bounded by the sum of the absolute values of the Fourier coefficients of f, and that this quantity defines a norm on the algebra
of complex polynomials over
. For each polynomial p in
, there is always an n-tuple of contractions where this supremum is achieved. Therefore,
and
are both two normed algebras.
Let
, be complex polynomials in n variables over
. Then
2.3. Unitary Dilation
Definition 2.3. Let
and let
be a tuple of commuting contractions on H. A unitary N-dilation for
is a k-tuple of commuting unitaries
acting on a space
such that
for all
satisfying
.
Definition 2.4. A finite set
of matrices is said to be doubly commuting if
and
, for every
.
The following results will enable us to prove our main results [13] [14] .
Theorem 2.5. (Sz-Nagy-Foias). Given k doubly commuting contractions
, there is a Hilbert space K containing H and doubly commuting unitaries
so that
for all
.
Theorem 2.6. Let
be a k-tuple of doubly commuting contractions on H. Then for every
, the k-tuple
has a unitary N-dilation that acts on a space of dimension
.
In finite dimensional, every tuple of commuting contractions which has a unitary N-dilation acting on a finite dimensional Hilbert space satisfies von Neumann’s inequality [13] .
Theorem 2.7. Let
, and let
be a k-tuple of commuting contractions on H that has a unitary N-dilation acting on a finite dimensional Hilbert space K. Put m = dim K. Then there exist m points
on the k-torus
such that for every polynomial
of degree less than or equal to N,
In particular,
Now, let us turn our attention to a particular family of doubly commuting sets of matrices which have many applications in several areas such as graph theory, cryptography, physics, signal and image processing, probability, statistics, numerical analysis, algebraic coding theory and many other areas [16] [17] .
2.4. Circulant Matrices
Let
be a finite set of complex numbers, denote by
the following Toeplitz matrix:
This matrix is called a complex circulant matrix of order m. It is possible to write this matrix as a single variable matrix polynomial in P, where P is the cyclic permutation matrix given by
Indeed,
The matrix
is a 9 × 9-complex circulant matrix. It is well known that the set of m × m-circulant matrices
is a commutative algebra. Let
be a primitive m-th root of unity. Let us denote by U the following matrix:
This matrix is called Vandermonde matrix. It is well known that this matrix has the following properties:
U is non-singular, unitary,
,
and
. It is well known that all elements of Cir(m) are simultaneously diagonalised by the same unitary matrix U [18] [19] [20] , that is, for A in Cir(m),
with
is a diagonal matrix with diagonal entries given by the ordered eigenvalues of A:
. The factorization
is called the spectral factorization of A.
3. Proof of the Main Results
In this section, we introduce the spectral mapping factorization of tuples of circulant matrices and its matrix version. We prove our main two results.
Theorem 3.1. Let A be a m × m-complex circulant matrix. Then
Proof. Let A be a m × m-complex circulant matrix. The spectral factorization of the matrix A allows us to claim that
It follows that
A simple calculation shows that
and
Therefore,
Theorem 3.2. Let A be a m × m-complex circulant matrix. Then
Proof. Let A be a m × m-complex circulant matrix. We know that
Then
Therefore,
The spectral factorization of circulant matrices [17] allows us to establish the spectral mapping factorization of tuples of circulant matrices.
Theorem 3.3. Let
be a set of m × m-complex circulant matrices. Then
.
Proof. Let
be a set of m × m-complex circulant matrices. From the spectral factorization of every
, we can say that there exist a complex m × m-unitary matrix U and m × m-diagonal matrices
such that
A simple calculation shows that
It is straightforward to see that
. Therefore,
.
Now, we are ready to establish the matrix version of the spectral mapping factorization of tuples of circulant matrices.
Theorem 3.4. Let
and let
be a set of m × m-complex circulant matrices and let
be a k × k-matrix with complex polynomials as entries. Let us denote by
and
. Then
Proof. Let
be a set of m × m-complex circulant matrices and let
be a k × k-matrix with complex polynomials as entries,
. Suppose that
and
. Theorem 3.3 allows us to claim that
. It follows that
A simple calculation shows that
Proof of Theorem 1.1
We just need to show that every tuple of circulant contractions is a doubly commuting set of contractions. Let
and let
be a k-tuple of circulant contractions. Assume that
It is clear that
Also,
Let us show that the set
is a doubly commuting set of contractions. We already know that this set is commutative. Let us observe that
Therefore, the adjoint of a complex m × m-circulant matrix is another m × m-circulant matrix. The fact that the set of m × m-circulant matrices is a commutative algebra implies that the matrices
and
, commute. Therefore, the set
is a doubly commuting set of contractions. Theorem 2.5 and Theorem 2.6 allow us to claim that, for each
, the k-tuple of circulant contractions
has a unitary N-dilation acting on a finite dimensional Hilbert space. Finally, for each
, every tuple of circulant contractions has a unitary N-dilation.
The above proof allows us to observe the following: every finite set of circulant matrices is a doubly commuting set of matrices. This enables us to prove our second main result.
First proof of Theorem 1.2
Let
be an n-tuple of circulant contractions of order m. The set
is a doubly commuting set of contractions. Theorem 2.7 allows us to claim that von Neumann’s inequality holds for this set
. Therefore,
Second proof of Theorem 1.2
Let
be a set of m × m-complex circulant contractions. Theorem 3.3 allows us to claim that
. It follows that
, for all
. Therefore,
4. Application
In this section, we construct completely contractive homomorphisms over the algebra of complex polynomials defined on
.
Theorem 4.1. Let
be a set of m × m-complex circulant contractions. Then the map
given by
is a completely contractive homomorphism.
Proof. Let
be a set of m × m-complex circulant contractions. The spectral factorization of the matrix
allows us to claim that
It follows that
A simple calculation shows that
Finally,
, since
. Suppose that
is the map given by
It is well known that the map
is a homomorphism [14] . Also,
First of all, let us show that the map
is a contractive map. Due to the fact that the elements of the set
doubly commute implies that
Therefore,
Let
and define the map
by setting
Let us show that the map
is contractive. Theorem 3.4 allows us to claim that if
we can say that
with
. Recall that
,
and
. It follows that
Finally,