Skip to Content
MathematicsMathematics
  • Article
  • Open Access

7 June 2026

18 Pages

Jointly A-Paranormal Tuples of Operators in Semi-Hilbertian Spaces

Department of Mathematics, College of Science, King Saud University, P.O. Box 2455, Riyadh 11451, Saudi Arabia

Abstract

This paper introduces the concept of jointly A -paranormal operator tuples acting on semi-Hilbertian spaces. We establish foundational operator inequalities and structural properties governing this new class. Specifically, we prove that joint A -paranormality is preserved under tensor products and stable under taking powers. Furthermore, we demonstrate that every commuting jointly A -hyponormal tuple is jointly A -paranormal. Since every commuting jointly A -paranormal tuple is jointly A -normaloid, we complete the hierarchy of multivariable operator classes within the semi-Hilbertian context.

1. Introduction

The study of multivariable operators on Hilbert spaces has evolved considerably [1,2,3]. With attention shifting toward specific classes of operator tuples [4,5,6,7], the properties of normal and hyponormal tuples have been rigorously established [8,9,10,11,12,13]. Other notable advancements include the analysis of spherical transforms [14,15] and the transition to joint paranormality [16,17].
While classical spaces rely on standard inner products, extending these multivariable concepts to semi-Hilbertian spaces provides a substantial generalization [18,19,20]. Employing a positive semi-definite sesquilinear form induced by a bounded positive operator A establishes a tailored framework for accommodating degenerate inner products. Initial investigations generalized geometric and spectral properties, such as the joint numerical range [21] and the joint spectral radius [22]. Subsequent efforts formalized operator classes within this setting, including jointly A -normal [16], jointly A -hyponormal [8,10], and ( A , m ) -isometric tuples [23].
To properly situate our contribution, we recall that a bounded linear operator Y on a complex Hilbert space E admitting an A -adjoint is called A -quasinormal if Y commutes with Y ♯ A Y in the A -sense (see [24]) this property places it strictly between A -normal and A -hyponormal operators. Since every commuting jointly A -hyponormal tuple is jointly A -paranormal, as we establish in Theorem 2, the jointly A -quasinormal class likewise resides above joint A -paranormality in the hierarchy. It is also worth noting that ( A , m ) -isometric tuples [23], while sharing the semi-Hilbertian setting, are defined through an entirely different mechanism—namely, the vanishing of an alternating sum of A -seminorms—and therefore do not fit directly into the hyponormal–paranormal–normaloid chain.
In the classical setting, an elegant hierarchy bridges the gap between normal and normaloid operators in the single-variable setting [25,26,27]. When Stanković [17] introduced jointly paranormal tuples, they were positioned as an intermediate class between jointly hyponormal and jointly normaloid tuples. In the semi-Hilbertian framework, recent results [8] established that every commuting jointly A -hyponormal tuple is jointly A -normaloid. However, the specific intermediate class responsible for this implication had not been identified. The principal objective of this paper is to resolve this gap by showing that joint A -paranormality is precisely the missing theoretical bridge.
To this end, we introduce the class of jointly A -paranormal tuples and rigorously establish the following chain of multivariable operator inclusions:
jointly A - hyponormal ⇒ jointly A - paranormal ⇒ jointly A - normaloid .
We emphasize that working in the semi-Hilbertian framework is not a routine adaptation of classical arguments. It introduces genuine analytical obstructions: the semi-inner product may be degenerate, the existence of the A -adjoint requires Douglas’s range inclusion theorem [28], and the identity ( Y ♯ A ) ♯ A = Π A Y Π A , where Π A denotes the orthogonal projection onto the closure of the range of A , reflects a structural degeneracy with no classical counterpart. Moreover, the absence of a Banach or C * -algebra structure prevents the direct use of standard functional calculus. Establishing the hierarchy above therefore requires new, framework-specific operator inequality techniques. Within this framework, we derive the fundamental operator inequalities and algebraic properties defining this new class, and we show in particular that joint A -paranormality is stable under taking powers and preserved under tensor products.
The paper is organized as follows. Section 2 reviews the foundational concepts of semi-Hilbertian spaces, including the role of Douglas’s range inclusion theorem in defining the A -adjoint, and recalls the relevant multivariable operator definitions. Section 3 presents our main results, detailing the fundamental operator inequalities, structural properties, and generalized log-convexity for jointly A -paranormal tuples. It further establishes stability under powers and tensor products and completes the inclusion hierarchy by proving the position of this class between jointly A -hyponormal and jointly A -normaloid tuples.

2. Preliminary Results

Let ( E , ⟨ · , · ⟩ ) be a complex Hilbert space equipped with the norm ∥ · ∥ induced by its inner product. The C * -algebra of all bounded linear operators on E is denoted by L ( E ) . For an operator Y ∈ L ( E ) , its range, nullspace, and adjoint are denoted by Ran ( Y ) , N ( Y ) , and Y * , respectively. An operator Y is positive, denoted by Y ≥ 0 , if ⟨ Y ξ , ξ ⟩ ≥ 0 for all ξ ∈ E , where 0 denotes the zero operator. For operators X , Y , we write X ≥ A Y if X − Y ≥ A 0 .
Consider a nonzero positive operator A ∈ L ( E ) . This operator induces a positive semi-definite sesquilinear form, defined as follows:
⟨ · , · ⟩ A : E × E → C , ( ξ 1 , ξ 2 ) ↦ ⟨ A ξ 1 , ξ 2 ⟩ .
The pair ( E , ∥ · ∥ A ) is called a semi-Hilbert space, with the associated seminorm given by ∥ ξ ∥ A = ⟨ ξ , ξ ⟩ A for any ξ ∈ E . We define the unit A -sphere in E as S E A : = { ξ ∈ E ∣ ∥ ξ ∥ A = 1 } . In the special case where A = I E , this reduces to the standard unit sphere S E . It is important to note that ( E , ∥ · ∥ A ) is generally not a normed or complete space. It forms a true Hilbert space if and only if A is injective and has a closed range; that is, Ran ( A ) = Ran ( A ) ¯ .
An operator Z ∈ L ( E ) is an A -adjoint of Y ∈ L ( E ) if A Z = Y * A . Applying Douglas’s theorem [28] reveals that Y admits an A -adjoint if and only if Ran ( Y * A ) ⊆ Ran ( A ) . We denote the space of all such operators by L A ( E ) . For any Y ∈ L A ( E ) , the unique reduced solution is Y ♯ A = A + Y * A [19]. For operators in the broader space of A -bounded operators L A ( E ) , the A -seminorm is directly defined as [21]:
∥ Y ∥ A = sup ξ ∈ S E A ∥ Y ξ ∥ A = sup ξ 1 , ξ 2 ∈ S E A | ⟨ Y ξ 1 , ξ 2 ⟩ A | .
From this identity, we find ∥ Y ∥ A = ∥ Y ♯ A ∥ A for any Y ∈ L A ( E ) .
For any Y , Z ∈ L A ( E ) and λ , γ ∈ C , we have λ Y + γ Z ∈ L A ( E ) and Y Z ∈ L A ( E ) , with ( λ Y + γ Z ) ♯ A = λ ¯ Y ♯ A + γ ¯ Z ♯ A and ( Y Z ) ♯ A = Z ♯ A Y ♯ A [18,19]. By induction, for any i ∈ N , we have Y i ∈ L A ( E ) and ( Y i ) ♯ A = ( Y ♯ A ) i . A double application of the A -adjoint yields ( Y ♯ A ) ♯ A = Π A Y Π A , where Π A is the orthogonal projection onto Ran ( A ) ¯ . Due to the generalized C * -identity, it follows that ∥ Y ♯ A Y ∥ A = ∥ Y Y ♯ A ∥ A = ∥ Y ∥ A 2 .
An operator Y ∈ L A ( E ) is A -selfadjoint if A Y = Y * A . An operator U ∈ L A ( E ) is an A -isometry if ∥ U υ ∥ A = ∥ υ ∥ A for all υ ∈ E , which holds if and only if U ♯ A U = Π A [18]. Since an A -unitary operator U has both U and U ♯ A as A -isometries, it preserves the A -seminorm.
Following Guesba et al. [10], a q-tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q is jointly A -hyponormal if the operator matrix C A ( Y ) = [ Y j ♯ A , Y i ] i , j ∈ { 1 , … , q } is A ˜ -positive, where A ˜ = diag ( A , … , A ) on E q . Equivalently,
∑ l = 1 q ∑ s = 1 q ⟨ Y l ξ s , Y s ξ l ⟩ A ≥ ∑ l = 1 q Y l ♯ A ξ l A 2
for any choice of vectors ξ 1 , … , ξ q ∈ E . Note that commutativity is not required. For A = I E , this reduces to Athavale’s definition [29]. We also note that in [10], C A ( Y ) should strictly be A ˜ -positive (not merely A -positive); see [8] for details.
For any operator Y ∈ L A ( E ) , the A -spectral radius and the A -numerical radius are defined [20,30] as
r A ( Y ) : = inf i ∈ N ∥ Y i ∥ A 1 i = lim i → ∞ ∥ Y i ∥ A 1 i , ω A ( Y ) : = sup ξ ∈ S E A | ⟨ Y ξ , ξ ⟩ A | .
They satisfy the inequalities r A ( Y ) ≤ ω A ( Y ) ≤ ∥ Y ∥ A [30].
Let Y = ( Y 1 , … , Y q ) represent a q-tuple of operators in L A ( E ) q . The joint operator A -seminorm and the joint A -numerical radius are given by:
∥ Y ∥ A : = sup ξ ∈ S E A ∑ i = 1 q ∥ Y i ξ ∥ A 2 and ω A ( Y ) : = sup ξ ∈ S E A ∑ i = 1 q | ⟨ Y i ξ , ξ ⟩ A | 2 .
For a commuting q-tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q , the joint A -spectral radius is defined [16] through the limit:
r A ( Y ) : = lim i → ∞ ∑ | ν | = i , ν ∈ N q i ! ν ! Y ♯ A ν Y ν A 1 2 i ,
where Y ♯ A = ( Y 1 ♯ A , … , Y q ♯ A ) , and for any multi-index ν = ( ν 1 , … , ν q ) ∈ N q , we use the notations Y ν : = ∏ l = 1 q Y l ν l , | ν | : = ∑ i = 1 q ν i and ν ! : = ∏ i = 1 q ν i ! .
A commuting q-tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q is said to be jointly A -normaloid [4] if
r A ( Y ) = ∥ Y ∥ A .
A useful tool in our analysis is the completely positive linear map Θ Y , A : L A ( E ) → L A ( E ) , defined for a tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q by:
Z ↦ Θ Y , A ( Z ) : = ∑ l = 1 q Y l ♯ A Z Y l .
Its powers are defined inductively by Θ Y , A 0 ( Z ) = Z and Θ Y , A i ( Z ) : = Θ Y , A [ Θ Y , A i − 1 ( Z ) ] for all i ∈ N .
Baklouti et al. [16] proved that if Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q forms a commuting q-tuple, its joint A -spectral radius simplifies to the limit:
r A ( Y ) = lim i → ∞ ∥ Θ Y , A i ( I E ) ∥ A 1 2 i .
Additionally, an identity established in [16] states that for any tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q , its joint A -seminorm is given by:
∥ Y ∥ A = ∑ l = 1 q Y l ♯ A Y l A 1 2 = ∥ Θ Y , A ( I E ) ∥ A 1 2 .
Let U q represent the open unit ball within C q , governed by the standard Euclidean norm ∥ τ ∥ 2 : = ∑ i = 1 q | τ i | 2 . Following the framework in [21], we formulate the following quantity for any Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q :
∥ Y ∥ e , A = sup τ ∈ U q ∑ i = 1 q τ i Y i A .
Notice that for any commuting q-tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q , we have
r A ( Y ) ≤ max 1 2 q ∥ Y ∥ A , r A ( Y ) ≤ ω A ( Y ) ≤ ∥ Y ∥ e , A ≤ ∥ Y ∥ A
(see [21]). We note that the inequalities in (4) can simultaneously be strict even if q = 1 .

3. Main Results

We begin by introducing the notion of jointly A -paranormal tuples.
Definition 1.
An operator q-tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q is called jointly A -paranormal if the inequality
∑ l = 1 q ∑ s = 1 q ∥ Y l Y s ξ ∥ A 2 ≥ ∑ l = 1 q ∥ Y l ξ ∥ A 2 2
holds for all unit vectors ξ ∈ S E A .
Note that if A = I E , this inequality reduces to the classical definition of jointly paranormal tuples, which was recently introduced by Stanković [17].
The following theorem provides equivalent conditions for joint A -paranormality.
Theorem 1.
For any tuple Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q , the following are equivalent:
(i) 
Y is jointly A -paranormal;
(ii) 
For all ξ ∈ S E A , the inequality holds:
∑ l = 1 q ∑ s = 1 q ∥ Y l Y s ξ ∥ A 2 ≥ ∑ l = 1 q ∥ Y l ξ ∥ A 2 2 ;
(iii) 
For all ξ ∈ S E A , we have:
⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A 2 ;
(iv) 
For all ξ ∈ E , the inequality holds:
⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A ⟨ ξ , ξ ⟩ A ≥ ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A 2 ;
(v) 
For all λ > 0 , the following operator inequality holds:
Θ Y , A 2 ( I E ) − 2 λ Θ Y , A ( I E ) + λ 2 I E ≥ A 0 .
Proof. 
Since Definition 1 aligns exactly with (ii), the equivalence between (i) and (ii) is immediate.
From the definition of Θ Y , A , it follows that
⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A = ∑ l = 1 q ⟨ Y l ♯ A Y l ξ , ξ ⟩ A = ∑ l = 1 q ⟨ Y l ξ , Y l ξ ⟩ A = ∑ l = 1 q ∥ Y l ξ ∥ A 2 .
Applying Θ Y , A twice gives
Θ Y , A 2 ( I E ) = Θ Y , A ∑ s = 1 q Y s ♯ A Y s = ∑ l = 1 q Y l ♯ A ∑ s = 1 q Y s ♯ A Y s Y l = ∑ l = 1 q ∑ s = 1 q ( Y s Y l ) ♯ A ( Y s Y l ) .
Taking the A -inner product with ξ yields
⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A = ∑ l = 1 q ∑ s = 1 q ⟨ ( Y s Y l ) ♯ A Y s Y l ξ , ξ ⟩ A = ∑ l = 1 q ∑ s = 1 q ∥ Y s Y l ξ ∥ A 2 .
Since the terms match exactly, (ii) and (iii) are equivalent.
Because ξ ∈ S E A implies ⟨ ξ , ξ ⟩ A = 1 , (iv) yields (iii). Conversely, assume (iii) holds. Since (iv) holds trivially for ∥ ξ ∥ A = 0 , we consider ξ ∈ E with ∥ ξ ∥ A ≠ 0 . Substituting the unit vector η = ξ ∥ ξ ∥ A into (iii) produces
Θ Y , A 2 ( I E ) ξ ∥ ξ ∥ A , ξ ∥ ξ ∥ A A ≥ Θ Y , A ( I E ) ξ ∥ ξ ∥ A , ξ ∥ ξ ∥ A A 2 .
Multiplying both sides by ∥ ξ ∥ A 4 leads to (iv).
For each ξ ∈ S E A , the polynomial
p ξ ( λ ) = λ 2 − 2 λ ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A + ⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A ,
has the discriminant
D ξ = 4 ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A 2 − 4 ⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A .
Since p ξ ( λ ) ≥ 0 for all λ > 0 if and only if D ξ ≤ 0 , condition (iii) ensures D ξ ≤ 0 . Given D ξ ≤ 0 , we have
⟨ ( Θ Y , A 2 ( I E ) − 2 λ Θ Y , A ( I E ) + λ 2 I E ) ξ , ξ ⟩ A ≥ 0
for all ξ ∈ S E A and all λ > 0 , which is equivalent to (v). □
The following theorem proves that commuting jointly A -hyponormal tuples are jointly A -paranormal.
Theorem 2.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a jointly A -hyponormal q-tuple of commuting operators. Then Y is jointly A -paranormal.
Proof. 
Let ξ ∈ S E A . Substituting ξ l = Y l ξ for each l ∈ { 1 , … , q } into (1) provides a suitable choice of vectors. Since the operators commute, it follows that Y l ξ s = Y l Y s ξ = Y s Y l ξ = Y s ξ l . Inserting these vectors into the left-hand side of (1) gives
∑ l = 1 q ∑ s = 1 q ⟨ Y l ξ s , Y s ξ l ⟩ A = ∑ l = 1 q ∑ s = 1 q ⟨ Y l Y s ξ , Y s Y l ξ ⟩ A = ∑ l = 1 q ∑ s = 1 q ∥ Y l Y s ξ ∥ A 2 .
Evaluating the right-hand side of (1) with these vectors yields
∑ l = 1 q Y l ♯ A Y l ξ A 2 = Θ Y , A ( I E ) ξ A 2 .
Applying the Cauchy–Schwarz inequality along with ∥ ξ ∥ A = 1 leads to
Θ Y , A ( I E ) ξ A 2 ≥ ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A 2 = ∑ l = 1 q ∥ Y l ξ ∥ A 2 2 .
Combining (6)–(8) with (1) produces
∑ l = 1 q ∑ s = 1 q ∥ Y l Y s ξ ∥ A 2 ≥ ∑ l = 1 q ∥ Y l ξ ∥ A 2 2 .
Since this holds for every ξ ∈ S E A , Theorem 1 guarantees that Y is jointly A -paranormal. □
To concretely illustrate this new class and its strict position in the established hierarchy (jointly A -hyponormal ⇒ jointly A -paranormal ⇒ jointly A -normaloid), we provide two examples showing that neither implication is reversible in general. It is worth noting that several examples in the one-dimensional case ( q = 1 ) already exist in the literature; see, for instance, [31].
The following Example presents a jointly A -paranormal tuple that is not jointly A -hyponormal.
Example 1.
Let E = C 2 and A = I E . Consider the tuple Y = ( Y 1 , Y 2 ) , where Y 1 = 0 0 1 0 and Y 2 = 0 1 0 0 . Since A = I E , the A -adjoint coincides with the standard adjoint. We first determine Θ Y , A ( I E ) :
Θ Y , A ( I E ) = Y 1 * Y 1 + Y 2 * Y 2 = I E .
Consequently, we have
Θ Y , A 2 ( I E ) = Y 1 * Θ Y , A ( I E ) Y 1 + Y 2 * Θ Y , A ( I E ) Y 2 = Y 1 * I E Y 1 + Y 2 * I E Y 2 = I E .
Thus, for each λ > 0 , we obtain
Θ Y , A 2 ( I E ) − 2 λ Θ Y , A ( I E ) + λ 2 I E = I E − 2 λ I E + λ 2 I E = ( 1 − λ ) 2 I E ≥ A 0 .
By Theorem 1, this shows that Y is jointly A -paranormal.
On the other hand, the self-commutator of Y 1 is
[ Y 1 * , Y 1 ] = Y 1 * Y 1 − Y 1 Y 1 * = 1 0 0 0 − 0 0 0 1 = 1 0 0 − 1 .
For Y to be jointly A -hyponormal, the operator matrix C A ( Y ) must be A ˜ -positive. This implies that its diagonal blocks [ Y k * , Y k ] must also be positive semi-definite. Since [ Y 1 * , Y 1 ] has a negative eigenvalue ( − 1 ), it is not positive semi-definite. Therefore, Y is not jointly A -hyponormal.
The following Example presents a commuting jointly A -normaloid tuple that is not jointly A -paranormal. More precisely, there exists a commuting jointly A -normaloid tuple ( Y 1 , Y 2 ) with Y 1 ≠ Y 2 which fails to be jointly A -paranormal.
Example 2.
Let E = C 3 and A = I E . Let Y = ( Y 1 , Y 2 ) be the tuple defined by
Y 1 = 1 0 0 0 0 1 0 0 0 a n d Y 2 = 1 0 0 0 0 0 0 0 0 .
We have Y 1 ≠ Y 2 . They commute since Y 1 Y 2 = Y 2 Y 1 = 1 0 0 0 0 0 0 0 0 . To compute the joint A -seminorm of Y , we first determine Θ Y , A ( I E ) :
Θ Y , A ( I E ) = Y 1 * Y 1 + Y 2 * Y 2 = 2 0 0 0 0 0 0 0 1 .
The joint A -seminorm of Y is given by
∥ Y ∥ A = ∥ Θ Y , A ( I E ) ∥ A 1 2 = 2 .
To find the joint A -spectral radius r A ( Y ) , we compute Θ Y , A i ( I E ) for i ≥ 2 . Applying Θ Y , A to a diagonal matrix Z = diag ( z 1 , z 2 , z 3 ) yields Θ Y , A ( Z ) = diag ( 2 z 1 , 0 , z 2 ) . Starting with Z 0 = I E , we obtain Θ Y , A ( I E ) = diag ( 2 , 0 , 1 ) and Θ Y , A i ( I E ) = diag ( 2 i , 0 , 0 ) for i ≥ 2 . By using (2), we deduce that
r A ( Y ) = lim i → ∞ ∥ Θ Y , A i ( I E ) ∥ A 1 2 i = lim i → ∞ ( 2 i ) 1 2 i = 2 .
Since r A ( Y ) = ∥ Y ∥ A = 2 , the tuple Y is jointly A -normaloid.
To see that Y is not jointly A -paranormal, we use Theorem 1 and evaluate the condition for λ = 1 . We have Θ Y , A 2 ( I E ) = diag ( 4 , 0 , 0 ) . This yields:
Θ Y , A 2 ( I E ) − 2 ( 1 ) Θ Y , A ( I E ) + ( 1 ) 2 I E = 1 0 0 0 1 0 0 0 − 1 .
This resulting matrix has a negative eigenvalue ( − 1 ). Thus, it is not positive semi-definite. Consequently, the condition from Theorem 1 is violated. Therefore, Y is not jointly A -paranormal.
The following theorem shows that joint A -paranormality is preserved under A -unitary equivalence.
Theorem 3.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a jointly A -paranormal q-tuple and let U ∈ L A ( E ) be an A -unitary operator. Then, the tuple
Z = ( U Y 1 U ♯ A , … , U Y q U ♯ A ) ∈ L A ( E ) q
is also jointly A -paranormal.
Proof. 
Since an A -unitary operator U preserves the A -seminorm, we have
∥ U υ ∥ A = ∥ υ ∥ A
for all υ ∈ E . Applying this property to the A -adjoint further yields ∥ U ♯ A υ ∥ A = ∥ υ ∥ A .
Since Y l ∈ L A ( E ) , we have Ran ( Y l * A ) ⊆ Ran ( A ) . This ensures that N ( A ) is invariant under Y l , which implies A Y l Π A = A Y l . Evaluating the A -seminorm gives
∥ Y l Π A υ ∥ A 2 = ⟨ A Y l Π A υ , Y l Π A υ ⟩ = ⟨ A Y l υ , Y l υ ⟩ = ∥ Y l υ ∥ A 2 .
Let ξ ∈ S E A . Setting η = U ♯ A ξ produces ∥ η ∥ A = ∥ U ♯ A ξ ∥ A = ∥ ξ ∥ A = 1 . Since Y is jointly A -paranormal, we have
∑ l = 1 q ∑ s = 1 q ∥ Y l Y s η ∥ A 2 ≥ ∑ l = 1 q ∥ Y l η ∥ A 2 2 .
For any l , s ∈ { 1 , … , q } , examining the elements of Z reveals
Z l Z s ξ = ( U Y l U ♯ A ) ( U Y s U ♯ A ) ξ = U Y l Π A Y s η .
Applying the norm-preserving properties yields
∥ Z l Z s ξ ∥ A = ∥ U Y l Π A Y s η ∥ A = ∥ Y l Π A Y s η ∥ A = ∥ Y l Y s η ∥ A .
Similarly, we find
∥ Z l ξ ∥ A = ∥ U Y l U ♯ A ξ ∥ A = ∥ U Y l η ∥ A = ∥ Y l η ∥ A .
Substituting these expressions into (9) gives
∑ l = 1 q ∑ s = 1 q ∥ Z l Z s ξ ∥ A 2 ≥ ∑ l = 1 q ∥ Z l ξ ∥ A 2 2 .
Since this is true for all ξ ∈ S E A , Theorem 1 establishes that Z is jointly A -paranormal. □
Remark 1.
Unlike the classical Hilbert space setting, where A -unitary equivalence follows easily from the algebraic identity ( U Y U * ) * = U Y * U * , the semi-Hilbertian proof of Theorem 3 must carefully navigate the degeneracies inherent to this framework. The key difficulty is that the expression ( U Y l U ♯ A ) ♯ A does not simplify by a direct algebraic manipulation. Instead, the proof relies on two relations that are specific to the semi-Hilbertian structure: the identity U ♯ A U = Π A , which replaces the classical U * U = I E , and the seminorm-preservation property ∥ Y l Π A υ ∥ A = ∥ Y l υ ∥ A , which reflects the fact that N ( A ) is invariant under Y l . Neither of these has a counterpart in the classical setting, and together they constitute the essential mechanism of the argument.
Remark 2.
It is worth pointing out precisely which results in this paper require commutativity and which do not. The definition of joint A -paranormality (Definition 1), the equivalent characterizations in Theorem 1, the A -unitary invariance in Theorem 3, and the stability under powers in Theorem 8 all hold without any commutativity assumption on the tuple Y . The results that genuinely require commutativity are those involving the joint A -spectral radius, specifically Theorem 6, which establishes that every commuting jointly A -paranormal tuple is jointly A -normaloid, and Corollary 1. Both rely on the spectral radius formula of Baklouti et al. [16], which is defined exclusively for commuting tuples. The core obstruction in the noncommuting case is that the identity
r A ( Y ) = lim n → ∞ ∥ Θ Y , A n ( I E ) ∥ A 1 2 n
is no longer available, so the log-convexity argument of Theorem 6 cannot be completed. Whether the normaloid conclusion can be extended to noncommuting tuples, perhaps by replacing classical commutativity with the weaker condition of A -commutativity A Y l Y s = A Y s Y l , is a natural and interesting open problem that we leave for future investigation.
To formalize tuple exponentiation, we introduce a concatenation operation. For arbitrary operator tuples Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q and Z = ( Z 1 , … , Z s ) ∈ L A ( E ) s , we define their concatenation product as:
Y ∘ Z : = ( Y 1 Z 1 , … , Y 1 Z s , … , Y q Z 1 , … , Y q Z s ) .
We inductively define the powers of a tuple Y :
Y 1 : = Y and Y i + 1 : = Y ∘ Y i ,
for any i ∈ N .
The next result establishes the linearity and monotonicity of the operator map Θ Y , A .
Theorem 4.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q . The operator Θ Y , A is linear and monotone; specifically, for any pair of A -selfadjoint operators Z , W ∈ L A ( E ) , we have:
Z ≥ A W ⇒ Θ Y , A ( Z ) ≥ A Θ Y , A ( W ) .
Proof. 
Since the A -adjoint operation and summation are linear, the linearity of Θ Y , A is immediate. Assuming Z ≥ A W , we deduce ⟨ Z η , η ⟩ A ≥ ⟨ W η , η ⟩ A for all η ∈ E . Substituting η = Y l ξ provides
⟨ Z Y l ξ , Y l ξ ⟩ A ≥ ⟨ W Y l ξ , Y l ξ ⟩ A .
The definition of the A -adjoint implies ⟨ Y l ♯ A Z Y l ξ , ξ ⟩ A ≥ ⟨ Y l ♯ A W Y l ξ , ξ ⟩ A . Summing over l ∈ { 1 , … , q } leads to
∑ l = 1 q ⟨ Y l ♯ A Z Y l ξ , ξ ⟩ A ≥ ∑ l = 1 q ⟨ Y l ♯ A W Y l ξ , ξ ⟩ A .
Exploiting the linearity of the A -inner product yields
∑ l = 1 q Y l ♯ A Z Y l ξ , ξ A ≥ ∑ l = 1 q Y l ♯ A W Y l ξ , ξ A .
The definition of Θ Y , A then implies ⟨ Θ Y , A ( Z ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A ( W ) ξ , ξ ⟩ A . Since this holds for all ξ ∈ E , we conclude Θ Y , A ( Z ) ≥ A Θ Y , A ( W ) . □
The following theorem provides multiplicative factorization and exponentiation rules for the map Θ .
Theorem 5.
Suppose Y = ( Y 1 , … , Y q ) , Z = ( Z 1 , … , Z s ) , and Y 1 , … , Y k are operator tuples in L A ( E ) . Then the following properties hold:
(i) 
Θ Y ∘ Z , A = Θ Z , A Θ Y , A ;
(ii) 
Θ Y 1 ∘ ⋯ ∘ Y k , A = Θ Y k , A ⋯ Θ Y 1 , A ;
(iii) 
Θ Y i , A l = Θ Y , A i l for all integers i , l ∈ N .
Proof. 
For (i), the definition of Θ Y ∘ Z , A dictates that
Θ Y ∘ Z , A ( X ) = ∑ l = 1 q ∑ m = 1 s ( Y l Z m ) ♯ A X ( Y l Z m ) .
Applying ( Y l Z m ) ♯ A = Z m ♯ A Y l ♯ A yields
Θ Y ∘ Z , A ( X ) = ∑ l = 1 q ∑ m = 1 s Z m ♯ A ( Y l ♯ A X Y l ) Z m .
Rearranging the summation gives
Θ Y ∘ Z , A ( X ) = ∑ m = 1 s Z m ♯ A ∑ l = 1 q Y l ♯ A X Y l Z m = Θ Z , A ( Θ Y , A ( X ) ) .
Since X is arbitrary, it follows that Θ Y ∘ Z , A = Θ Z , A Θ Y , A .
For (ii), since (i) establishes the base case for k = 2 , assume (ii) holds for k = n . Setting W = Y 1 ∘ ⋯ ∘ Y n , we observe
Θ W ∘ Y n + 1 , A = Θ Y n + 1 , A Θ W , A .
The induction hypothesis then yields
Θ W ∘ Y n + 1 , A = Θ Y n + 1 , A ( Θ Y n , A ⋯ Θ Y 1 , A ) .
For (iii), since Y i is the i-fold concatenation of Y , applying (ii) to its l-fold concatenation yields Θ Y i , A l = Θ Y i l , A . A further application of (ii) produces Θ Y i l , A = Θ Y , A i l . □
Next, we establish the fact that every commuting jointly A -paranormal tuple is jointly A -normaloid.
Theorem 6.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a commuting jointly A -paranormal q-tuple. Then, Y is jointly A -normaloid.
Proof. 
Since (5) provides the base case n = 1 , assume (10) holds for some n ∈ N :
Θ Y , A n + 1 ( I E ) − 2 λ Θ Y , A n ( I E ) + λ 2 Θ Y , A n − 1 ( I E ) ≥ A 0 .
The monotonicity of Θ Y , A yields
Θ Y , A Θ Y , A n + 1 ( I E ) − 2 λ Θ Y , A n ( I E ) + λ 2 Θ Y , A n − 1 ( I E ) ≥ A 0 .
Simplifying this expression gives
Θ Y , A n + 2 ( I E ) − 2 λ Θ Y , A n + 1 ( I E ) + λ 2 Θ Y , A n ( I E ) ≥ A 0 .
Mathematical induction establishes (10) for all n ∈ N .
Let ξ ∈ E . Since the quadratic polynomial in λ given by
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A − 2 λ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A + λ 2 ⟨ Θ Y , A n − 1 ( I E ) ξ , ξ ⟩ A ≥ 0
is non-negative for all λ > 0 and achieves its minimum at a non-negative value, its discriminant must be non-positive. This discriminant condition implies
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n − 1 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A 2 .
We now prove ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A 1 n ≤ ⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A 1 n + 1 . Taking n = 1 as the base case, assume the statement holds for a given n. Applying (11) alongside the induction hypothesis provides
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A 2 ≤ ⟨ Θ Y , A n + 2 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ≤ ⟨ Θ Y , A n + 2 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A n n + 1 .
Since the inequality holds trivially when ⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A = 0 , we may assume this quantity is strictly positive. Rearranging algebraically yields
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A 2 − n n + 1 ≤ ⟨ Θ Y , A n + 2 ( I E ) ξ , ξ ⟩ A .
Simplifying the fractions provides
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A n + 2 n + 1 ≤ ⟨ Θ Y , A n + 2 ( I E ) ξ , ξ ⟩ A .
Raising to the power of 1 n + 2 yields
⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A 1 n + 1 ≤ ⟨ Θ Y , A n + 2 ( I E ) ξ , ξ ⟩ A 1 n + 2 .
Mathematical induction therefore establishes the fundamental chain:
⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A ≤ ⟨ Θ Y , A 2 ( I E ) ξ , ξ ⟩ A 1 2 ≤ … ≤ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A 1 n .
Since the inequality ∥ Θ Y , A n ( I E ) ∥ A ≤ ∥ Θ Y , A ( I E ) ∥ A n is known to hold, we proceed to establish the reverse inequality. Taking the supremum over all unit vectors ξ ∈ S E A on the fundamental chain provides
∥ Θ Y , A ( I E ) ∥ A ≤ ∥ Θ Y , A n ( I E ) ∥ A 1 n .
Extracting the square root and invoking (3) gives
∥ Y ∥ A ≤ ∥ Θ Y , A n ( I E ) ∥ A 1 2 n .
Taking the limit as n → ∞ and utilizing (2) produces
∥ Y ∥ A ≤ lim n → ∞ ∥ Θ Y , A n ( I E ) ∥ A 1 2 n = r A ( Y ) .
Since r A ( Y ) ≤ ∥ Y ∥ A always holds, we conclude r A ( Y ) = ∥ Y ∥ A . Thus, Y is jointly A -normaloid. □
Remark 3.
It is instructive to contrast the proof of Theorem 6 with its classical counterpart. In the standard Hilbert space setting, the fact that a paranormal operator is normaloid follows from elementary norm estimates. In the semi-Hilbertian framework, the argument is considerably more involved and proceeds through three distinct steps: First, constructing the completely positive map Θ Y , A and establishing its monotonicity (Theorem 4). Second, propagating this monotonicity inductively to derive the log-convexity inequality (11) for the sequence ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A . Third, connecting the resulting growth estimates to the joint A -spectral radius via the limit formula (2) of Baklouti et al. [16]. Each of these steps is specific to the semi-Hilbertian structure and has no direct classical analog, which illustrates why the transition to this framework demands new techniques rather than a straightforward adaptation of known arguments.
Corollary 1.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a commuting jointly A -paranormal q-tuple. Then, the following equalities hold:
r A ( Y ) = ω A ( Y ) = ∥ Y ∥ e , A = ∥ Y ∥ A .
Proof. 
Since Y is a commuting jointly A -paranormal tuple, it follows that r A ( Y ) = ∥ Y ∥ A . From (4), we deduce
r A ( Y ) ≤ ω A ( Y ) ≤ ∥ Y ∥ e , A ≤ ∥ Y ∥ A .
Since the bounds are equal, we conclude r A ( Y ) = ω A ( Y ) = ∥ Y ∥ e , A = ∥ Y ∥ A . □
Evaluating the squared A -seminorm of Y n acting on ξ yields:
∥ Y n ξ ∥ A 2 = ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A .
Since this holds for every n ∈ N and every ξ ∈ E , extracting the square roots in (12) generates the following chain:
∥ Y ξ ∥ A ≤ ∥ Y 2 ξ ∥ A 1 / 2 ≤ ⋯ ≤ ∥ Y n ξ ∥ A 1 / n ≤ ⋯ ,
for all ξ ∈ S E A .
Our next theorem establishes the invariance of the A -kernel under taking powers of the tuple.
Theorem 7.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a jointly A -paranormal q-tuple. Then,
N A ( Y ) = N A ( Y n )
for all n ∈ N , where N A ( Y ) : = { ξ ∈ E : ∥ Y l ξ ∥ A = 0 , l = 1 , … , q } .
Proof. 
For an arbitrary n ∈ N , Theorem 5(iii) implies:
N A ( Θ Y , A n ( I E ) ) = N A ( Θ Y n , A ( I E ) ) = N A ( Y n ) .
Since the log-convexity inequality (11) holds for all n ∈ N , it produces:
N A ( Y n − 1 ) ⊆ N A ( Y n ) ⊇ N A ( Y n + 1 ) ,
where Y 0 : = I . Applying these inclusions inductively confirms (13). □
The theorem below proves that the powers of a jointly A -paranormal tuple remain jointly A -paranormal.
Theorem 8.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a jointly A -paranormal q-tuple. Then, Y n is jointly A -paranormal for all n ∈ N .
Proof. 
Let n ∈ N and ξ ∈ S E A . Since ξ ∈ N A ( Y ) implies ξ ∈ N A ( Y k ) by Theorem 7, we have ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A = 0 . When ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A = 0 , the inequality holds trivially.
For ξ ∉ N A ( Y ) , we know ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A > 0 for all k ∈ N ; applying (11) then yields:
⟨ Θ Y , A k + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A k − 1 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A 2 .
Algebraic rearrangement leads to:
⟨ Θ Y , A k + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A k ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A k − 1 ( I E ) ξ , ξ ⟩ A .
Since this relation holds for every k ∈ N , iterative expansion yields the following descending chain:
⟨ Θ Y , A 2 n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A 2 n − 1 ( I E ) ξ , ξ ⟩ A ≥ ⋯ ≥ ⟨ Θ Y , A n + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n − 1 ( I E ) ξ , ξ ⟩ A ≥ ⋯ ≥ ⟨ Θ Y , A ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A 0 ( I E ) ξ , ξ ⟩ A .
Pairing terms symmetrically from the upper and lower halves yields:
⟨ Θ Y , A n + j ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + j − 1 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A j ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A j − 1 ( I E ) ξ , ξ ⟩ A
for each j ∈ { 1 , … , n } . Taking the product of these inequalities produces:
∏ j = 1 n ⟨ Θ Y , A n + j ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + j − 1 ( I E ) ξ , ξ ⟩ A ≥ ∏ j = 1 n ⟨ Θ Y , A j ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A j − 1 ( I E ) ξ , ξ ⟩ A .
Recognizing that these products are telescoping, simplifying them yields:
⟨ Θ Y , A 2 n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A 0 ( I E ) ξ , ξ ⟩ A .
Since ⟨ Θ Y , A 0 ( I E ) ξ , ξ ⟩ A = ⟨ I E ξ , ξ ⟩ A = 1 , we obtain:
⟨ Θ Y , A 2 n ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A 2 .
Theorem 5(iii) dictates that Θ Y , A 2 n = Θ Y n , A 2 and Θ Y , A n = Θ Y n , A . Thus, it follows that:
⟨ Θ Y n , A 2 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y n , A ( I E ) ξ , ξ ⟩ A 2 .
Since this inequality remains valid for any ξ ∈ S E A , Theorem 1 ensures that Y n is jointly A -paranormal. □
Remark 4.
The classical analog of Theorem 8 assumes the algebraic convenience of the standard adjoint identity ( Y n ) * = ( Y * ) n . In our setting, properly applying the A -adjoint to the map Θ Y , A relies intrinsically on the non-trivial factorization Θ Y i , A l = Θ Y , A i l (established in Theorem 5), without which the structural stability under powers cannot be verified.
The following corollary generalizes the log-convexity property for tuples.
Corollary 2.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E ) q be a jointly A -paranormal q-tuple, and let ξ ∈ E be arbitrary. Then,
⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A m ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n + m 2 ( I E ) ξ , ξ ⟩ A 2 ,
holds for each n ∈ N and each 0 ≤ m ≤ n , such that n ≡ 2 m .
Proof. 
While the statement is trivially true for ξ ∈ N A ( Y ) , we assume ξ ∉ N A ( Y ) . Since n − n + m 2 + 1 = n + m 2 − ( m + 1 ) , the descending chain of ratio inequalities established in Theorem 8 leads to:
⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + m 2 ( I E ) ξ , ξ ⟩ A = ⟨ Θ Y , A n ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n − 1 ( I E ) ξ , ξ ⟩ A ⋯ ⟨ Θ Y , A n + m 2 + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + m 2 ( I E ) ξ , ξ ⟩ A ≥ ⟨ Θ Y , A n + m 2 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A n + m 2 − 1 ( I E ) ξ , ξ ⟩ A ⋯ ⟨ Θ Y , A m + 1 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A m ( I E ) ξ , ξ ⟩ A = ⟨ Θ Y , A n + m 2 ( I E ) ξ , ξ ⟩ A ⟨ Θ Y , A m ( I E ) ξ , ξ ⟩ A .
Multiplying appropriately on both sides produces (14). □
In the sequel, we establish the behavior of the map Θ on tensor products of operator tuples. Let E 1 and E 2 be complex Hilbert spaces equipped with positive operators A ∈ L ( E 1 ) + and B ∈ L ( E 2 ) + . The completed Hilbert tensor product space E 1 ⊗ ¯ E 2 naturally becomes a semi-Hilbertian space with respect to the positive tensor operator A ⊗ ¯ B [18,20].
For tuples Y = ( Y 1 , … , Y q ) ∈ L A ( E 1 ) q and Z = ( Z 1 , … , Z s ) ∈ L B ( E 2 ) s , we consider the tensor extension tuples. We shall write I E 1 and I E 2 for the identity operators on E 1 and E 2 , respectively. Let I 1 = ( I E 1 , … , I E 1 ) and I 2 = ( I E 2 , … , I E 2 ) be the corresponding identity tuples. Let Y ⊗ ¯ I 2 : = ( Y 1 ⊗ ¯ I E 2 , … , Y q ⊗ ¯ I E 2 ) .
Theorem 9.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E 1 ) q . Then, for all n ∈ N and all X ∈ L A ( E 1 ) , we have:
Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B n ( X ⊗ ¯ I E 2 ) = Θ Y , A n ( X ) ⊗ ¯ Π B ,
where Π B is the orthogonal projection onto Ran ( B ) ¯ . By symmetry, an analogous identity holds for I 1 ⊗ ¯ Z .
Proof. 
The properties of the A ⊗ ¯ B -adjoint for n = 1 assert that ( Y l ⊗ ¯ I E 2 ) ♯ A ⊗ ¯ B = Y l ♯ A ⊗ ¯ Π B . Substituting this identity yields
Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B ( X ⊗ ¯ I E 2 ) = ∑ l = 1 q ( Y l ⊗ ¯ I E 2 ) ♯ A ⊗ ¯ B ( X ⊗ ¯ I E 2 ) ( Y l ⊗ ¯ I E 2 ) = ∑ l = 1 q ( Y l ♯ A ⊗ ¯ Π B ) ( X ⊗ ¯ I E 2 ) ( Y l ⊗ ¯ I E 2 ) = ∑ l = 1 q ( Y l ♯ A X Y l ) ⊗ ¯ ( Π B I E 2 I E 2 ) = ∑ l = 1 q Y l ♯ A X Y l ⊗ ¯ Π B = Θ Y , A ( X ) ⊗ ¯ Π B .
Having shown the base case n = 1 , assume the identity holds for some n ∈ N . Since Π B is an orthogonal projection, it satisfies Π B 2 = Π B . Applying the map Θ once more alongside this projection property gives
Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B n + 1 ( X ⊗ ¯ I E 2 ) = Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B Θ Y , A n ( X ) ⊗ ¯ Π B = ∑ l = 1 q ( Y l ♯ A ⊗ ¯ Π B ) Θ Y , A n ( X ) ⊗ ¯ Π B ( Y l ⊗ ¯ I E 2 ) = ∑ l = 1 q Y l ♯ A Θ Y , A n ( X ) Y l ⊗ ¯ ( Π B Π B I E 2 ) = Θ Y , A Θ Y , A n ( X ) ⊗ ¯ Π B = Θ Y , A n + 1 ( X ) ⊗ ¯ Π B .
Mathematical induction establishes the desired identity. □
We finish the paper by showing that joint A -paranormality is preserved under tensor products.
Theorem 10.
Let Y = ( Y 1 , … , Y q ) ∈ L A ( E 1 ) q be a jointly A -paranormal q-tuple. Then, the tensor tuple Y ⊗ ¯ I 2 is jointly ( A ⊗ ¯ B ) -paranormal on E 1 ⊗ ¯ E 2 .
Proof. 
Let λ > 0 . We must show that the operator W λ defined by
W λ : = Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B 2 ( I E 1 ⊗ ¯ I E 2 ) − 2 λ Θ Y ⊗ ¯ I 2 , A ⊗ ¯ B ( I E 1 ⊗ ¯ I E 2 ) + λ 2 ( I E 1 ⊗ ¯ I E 2 )
is ( A ⊗ ¯ B ) -positive. Theorem 9 provides
W λ = Θ Y , A 2 ( I E 1 ) ⊗ ¯ Π B − 2 λ Θ Y , A ( I E 1 ) ⊗ ¯ Π B + λ 2 I E 1 ⊗ ¯ I E 2 .
Let ζ ∈ E 1 ⊗ ¯ E 2 . For any operator W ∈ L A ( E 1 ) , evaluating W ⊗ ¯ Π B within the A ⊗ ¯ B form indicates ⟨ ( W ⊗ ¯ Π B ) ζ , ζ ⟩ A ⊗ ¯ B = ⟨ ( A W ⊗ ¯ B Π B ) ζ , ζ ⟩ . Since Π B is the orthogonal projection onto Ran ( B ) ¯ , it follows that B Π B = B . Utilizing this identity, we obtain
⟨ ( A W ⊗ ¯ B ) ζ , ζ ⟩ = ⟨ ( A ⊗ ¯ B ) ( W ⊗ ¯ I E 2 ) ζ , ζ ⟩ = ⟨ ( W ⊗ ¯ I E 2 ) ζ , ζ ⟩ A ⊗ ¯ B .
Applying this equivalence to the first two terms of W λ produces
⟨ W λ ζ , ζ ⟩ A ⊗ ¯ B = ⟨ ( Θ Y , A 2 ( I E 1 ) ⊗ ¯ I E 2 − 2 λ Θ Y , A ( I E 1 ) ⊗ ¯ I E 2 + λ 2 I E 1 ⊗ ¯ I E 2 ) ζ , ζ ⟩ A ⊗ ¯ B = ⟨ [ Θ Y , A 2 ( I E 1 ) − 2 λ Θ Y , A ( I E 1 ) + λ 2 I E 1 ] ⊗ ¯ I E 2 ζ , ζ ⟩ A ⊗ ¯ B .
Since Y is jointly A -paranormal, Theorem 1 implies
P λ : = Θ Y , A 2 ( I E 1 ) − 2 λ Θ Y , A ( I E 1 ) + λ 2 I E 1 ≥ A 0 .
Since the tensor product of an A -positive operator ( P λ ) and a B -positive operator ( I E 2 ) is inherently ( A ⊗ ¯ B ) -positive, it follows that ⟨ W λ ζ , ζ ⟩ A ⊗ ¯ B ≥ 0 . Since this holds for all λ > 0 , it guarantees that Y ⊗ ¯ I 2 is jointly ( A ⊗ ¯ B ) -paranormal. □
Remark 5.
The tensor product stability established in Theorem 10 diverges significantly from classical results. Because the ( A ⊗ ¯ B ) -adjoint inevitably introduces the orthogonal projection Π B via ( Y l ⊗ ¯ I E 2 ) ♯ A ⊗ ¯ B = Y l ♯ A ⊗ ¯ Π B , the proof must systematically track the propagation of this projection through all iterates of Θ. The subsequent reliance on the identity B Π B = B to finalize the positivity argument is an obstruction unique to the semi-Hilbertian framework and is entirely absent from any classical treatment.
Remark 6.
It is worth noting that even in the classical Hilbert space setting ( A = I E 1 , B = I E 2 ), if Y and Z are jointly paranormal tuples, their full tensor product Y ⊗ ¯ Z may not be jointly paranormal, even for single operators ( q = 1 ), as demonstrated by Ando [25].

4. Conclusions and Open Problems

In this paper, we introduced the class of jointly A -paranormal tuples of operators on semi-Hilbertian spaces and established its natural position within the operator hierarchy
jointly A - hyponormal ⇒ jointly A - paranormal ⇒ jointly A - normaloid ,
filling a gap that previously existed in the multivariable semi-Hilbertian theory. Along the way, we derived equivalent characterizations of joint A -paranormality via operator inequalities; the completely positive map Θ Y , A established a generalized log-convexity property for its iterates, proved stability under A -unitary equivalence and under taking powers, and showed that the class is preserved under tensor product extensions. We regard this work as a natural starting point for a broader research program in multivariable semi-Hilbertian operator theory, and we now outline what we believe are the most promising directions for future investigation. A natural and immediate next step is the introduction and systematic study of jointly A -quasinormal tuples in semi-Hilbertian spaces, where each component Y l commutes with Y l ♯ A Y l in the A -sense; this class would sit between jointly A -normal and jointly A -hyponormal tuples, further refining the hierarchy and demonstrating that the tools developed in the present paper—particularly the monotonicity of Θ Y , A established in Theorem 4—are well suited to its analysis. Several of our results rely on classical commutativity of the tuple components, and a compelling open question is whether these can be extended to the weaker and more natural condition of A -commutativity, namely, A Y l Y s = A Y s Y l for all l , s ; this is a strictly weaker requirement whenever A has a nontrivial null space, and resolving it would require, in particular, developing a joint A -spectral radius theory under this relaxed hypothesis. Finally, exploring connections with ( A , m ) -isometric tuples [23] and investigating the spectral consequences of joint A -paranormality—such as the structure of the joint A -spectrum and the existence of common invariant subspaces—are further directions that we believe deserve attention, and we hope the framework developed here will serve as a useful foundation for these and related future investigations.

Funding

This work was supported by the Ongoing Research Funding Program (ORF-2026-802), King Saud University, Riyadh, Saudi Arabia.

Data Availability Statement

No new data were created or analyzed in this study.

Acknowledgments

The author would like to express her sincere gratitude to the referees for their valuable comments and suggestions. This work was supported by the Ongoing Research Funding Program (ORF-2026-802), King Saud University, Riyadh, Saudi Arabia. In addition, the author would like to express deep gratitude to Kais Feki (University of Sfax, Tunisia) and Hranislav Stanković (University of Niš, Serbia) for their insightful comments, valuable suggestions, and careful reading of the manuscript, all of which significantly enhanced the quality and presentation of this paper.

Conflicts of Interest

The author declares no conflicts of interest.

References

  1. Dash, A.T. Joint spectra. Stud. Math. 1973, XLV, 225–237. [Google Scholar] [CrossRef] [Scilit]
  2. Taylor, J.L. A joint spectrum for several commuting operators. J. Funct. Anal. 1970, 6, 172–191. [Google Scholar] [CrossRef] [Scilit]
  3. Vasilescu, F.H. A characterization of the joint spectrum in Hilbert spaces. Rev. Roum. Math. Pures Appl. 1977, 22, 1003–1009. [Google Scholar]
  4. Altwaijry, N.; Dragomir, S.S.; Feki, K. On A-normaloid d-tuples of operators and related questions. Quaest. Math. 2024, 47, 1305–1326. [Google Scholar] [CrossRef] [Scilit]
  5. Altwaijry, N.; Feki, K.; Stanković, H. On p-joint spectral radius of operators in Hilbert spaces and related results. Linear Multilinear Algebra 2025, 73, 985–1001. [Google Scholar] [CrossRef] [Scilit]
  6. Ahmed, O.A.M.S.; Ahmed, A.H.; Sarosh, A. (α, β)-Normal Operators in Several Variables. Math. Probl. Eng. 2022, 2022, 3020449. [Google Scholar] [CrossRef] [Scilit]
  7. Stanković, H. Spherical p-quasinorms of operator tuples. J. Math. Anal. Appl. 2025, 542, 128826. [Google Scholar] [CrossRef] [Scilit]
  8. Aljawi, S.; Feki, K.; Stanković, H. Jointly A-hyponormal m-tuple of commuting operators and related results. AIMS Math. 2024, 9, 30348–30363. [Google Scholar] [CrossRef] [Scilit]
  9. Cvetković-Ilić, D.S.; Stanković, H. On normal complements. J. Math. Anal. Appl. 2024, 535, 128216. [Google Scholar] [CrossRef] [Scilit]
  10. Guesba, M.; Beiba, E.M.O.; Ahmed, O.A.M.S. Joint A-hyponormality of operators in semi-Hilbert spaces. Linear Multilinear Algebra 2021, 69, 2888–2907. [Google Scholar] [CrossRef] [Scilit]
  11. Stanković, H. Subnormal n-th roots of matricially and spherically quasinormal pairs. Filomat 2023, 37, 5325–5331. [Google Scholar] [CrossRef] [Scilit]
  12. Stanković, H. Converse of Fuglede Theorem. Oper. Matrices 2023, 17, 705–714. [Google Scholar] [CrossRef] [Scilit]
  13. Stanković, H. Spherically quasinormal tuples: n-th root problem and hereditary properties. Complex Anal. Oper. Theory 2025, 19, 156. [Google Scholar] [CrossRef] [Scilit]
  14. Stanković, H. Spherical Mean Transform of Operator Pairs. J. Math. Anal. Appl. 2024, 530, 127743. [Google Scholar] [CrossRef] [Scilit]
  15. Stanković, H. Spherical Heinz transform of operator tuples. Z. Anal. Ihre Anwendungen 2026, 45, 139–163. [Google Scholar] [CrossRef] [Scilit]
  16. Baklouti, H.; Feki, K.; Ahmed, O.A.M.S. Joint normality of operators in semi-Hilbertian spaces. Linear Multilinear Algebra 2020, 68, 845–866. [Google Scholar] [CrossRef] [Scilit]
  17. Stanković, H. Joint paranormality: Bridging the gap between hyponormal and normaloid tuples. Linear Multilinear Algebra 2025, 73, 3004–3018. [Google Scholar] [CrossRef] [Scilit]
  18. Arias, M.L.; Corach, G.; Gonzalez, M.C. Partial isometries in semi-Hilbertian spaces. Linear Algebra Appl. 2008, 428, 1460–1475. [Google Scholar] [CrossRef] [Scilit]
  19. Arias, M.L.; Corach, G.; Gonzalez, M.C. Metric properties of projections in semi-Hilbertian spaces. Integral Equ. Oper. Theory 2008, 62, 11–28. [Google Scholar] [CrossRef] [Scilit]
  20. Saddi, A. A-Normal operators in Semi-Hilbertian spaces. Aust. J. Math. Anal. Appl. 2012, 9, 1–12. [Google Scholar]
  21. Altwaijry, N.; Feki, K.; Minculete, N. A new seminorm for d-tuples of A-bounded operators and their applications. Mathematics 2023, 11, 685. [Google Scholar] [CrossRef] [Scilit]
  22. Müller, V. On the joint spectral radius. Ann. Pol. Math. 1997, 66, 173–182. [Google Scholar] [CrossRef] [Scilit]
  23. Ghribi, S.; Jeridi, N.; Rabaoui, R. On (A,m)-isometric commuting tuples of operators on a Hilbert space. Linear Multilinear Algebra 2022, 70, 2097–2116. [Google Scholar] [CrossRef] [Scilit]
  24. Panayappan, S.; Sivamani, N. A-Quasi Normal Operators in Semi-Hilbertian Spaces. Gen. Math. Notes 2012, 10, 30–35. [Google Scholar]
  25. Ando, T. Operators with a norm condition. Acta Sci. Math. 1972, 33, 169–172. [Google Scholar]
  26. Furuta, T. On the class of paranormal operators. Proc. Jpn. Acad. 1967, 43, 594–598. [Google Scholar] [CrossRef] [Scilit]
  27. Istrǎţescu, V. On some hyponormal operators. Pac. J. Math. 1967, 22, 413–417. [Google Scholar] [CrossRef] [Scilit]
  28. Douglas, R.G. On majorization, factorization and range inclusion of operators in Hilbert space. Proc. Am. Math. Soc. 1966, 17, 413–416. [Google Scholar] [CrossRef]
  29. Athavale, A. On jointly hyponormal operators. Proc. Amer. Math. Soc. 1988, 103, 417–423. [Google Scholar] [CrossRef]
  30. Feki, K. Spectral radius of semi-Hilbertian space operators and its applications. Ann. Funct. Anal. 2020, 11, 929–946. [Google Scholar] [CrossRef] [Scilit]
  31. Furuta, T. Invitation to Linear Operators: From Matrices to Bounded Linear Operators on a Hilbert Space; CRC Press: Boca Raton, FL, USA, 2001. [Google Scholar]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Article Metrics

Citations

Article Access Statistics

Multiple requests from the same IP address are counted as one view.