TWMS J. App. and Eng. Math. V.13, N.1, 2023, pp.

276-284

A NEW CLASS OF NON NORMAL OPERATORS

ON HILBERT SPACES

AISSA NASLI BAKIR1 ,§

Abstract. We introduce the class of (M, k)-quasi-parahyponormal operators on a sep-

arable Hilbert space as an extension of the classes of parahyponormal and k-quasi-
parahyponormal operators given in [13]. The matrix representation of an (M, k)-quasi-
parahyponormal operator, the finite ascent, the SVEP and other several properties of
such class of operators are also presented.

Keywords: M -parahyponormal operator, k-quasi-parahyponormal operator, SVEP, fi-

nite ascent.

AMS Subject Classification(2010): 47A30, 47B47, 47B20.

1. Introduction
Let B(H) denote the Banach algebra of bounded linear operators on an infinite dimen-
sional separable complex Hilbert space H. For an operator T ∈ B(H), ker(T ) and ran(T )
denote respectively, the null space and the range of T. Recall that T ∈ B(H) is said to be
paranormal if T ?2 T 2 − 2λT ? T + λ2 ≥ 0 for all λ > 0, and k-paranormal for some integer
k ≥ 2, if kT xkk ≤ kT k xk for all unit vector x in H. An operator T is said to be quasi-class
1
A, if T ? |T |2 T ≤ T ? |T 2 |T, where |T | = (T ? T ) 2 is the module of T. Also, T is called k-quasi-
paranormal if T ∗k (T ?2 T 2 − 2λT ? T + λ2 )T k ≥ 0. This class is introduced by S. Mecheri
in [10], and it is a generalization of the class of quasi-paranormal operators, and it con-
tains the class of quasi-class A operators, see [9]. It is also shown in [4] that quasi-class
A operators satisfy Weyl’s Theorem. Authors in [3, 13] and [16] introduced the class of
parahyponormal and M -parahyponormal operators, and gave several properties for such
operators. In the present paper, we introduce a new class of (M, k)-quasi-parahyponormal
operators as a generalization of the class of k-quasi-parahyponormal operators. We give
the matrix representation, the ascent and the SVEP. Different related properties are also
proved.

1
Department of Mathematics, Laboratory of Mathematics and Application LMA, Hassiba Benbouali
University of Chlef, B.P. 78C, Ouled Fares, 02180, Chlef, Algeria.
e-mail: [email protected]; ORCID: https://orcid.org/0000-0001-6906-3307.
§ Manuscript received: November 19, 2020; accepted: January 23, 2021.
TWMS Journal of Applied and Engineering Mathematics, Vol.13, No.1 © Işık University, Department
of Mathematics, 2023; all rights reserved.
276
 AISSA NASLI BAKIR: A NEW CLASS OF NON NORMAL OPERATORS 277

2. Class of k-quasi-parahyponormal operators

Definition 2.1. [13][16] An operator T ∈ B(H) is said to be parahyponormal if
(T T ? )2 − 2λT ? T + λ2 ≥ 0
for all λ > 0.
Definition 2.2. [13] An operator T ∈ B(H) is said to be k-quasi-parahyponormal for
some integer k if
T ?k ((T T ? )2 − 2λT ? T + λ2 )T k ≥ 0
for all λ > 0.
This definition is equivalent to
kT k+1 xk2 ≤ kT k xkkT T ? T k xk
for all x ∈ H. In this section, we give some complement properties for this class of opera-
tors.

Theorem 2.1. Let S be the bilateral weighted shift defined on the usual Hilbert space `2
by Sen = αn en+1 , where (en )n is the standard basis, and (αn )n is a decreasing complex
sequence. Then, S is k-quasi-parahyponormal if and only if
|αn+k | ≤ |αn+k−1 |2
for all n.

Proof. We have
kS k+1 en k2 ≤ kS k en kkSS ? S k en k
Hence, for all n,
k
Y k−1
Y
|αn+i |2 ≤ |αn+k−1 |2 |αn+i |2
i=0 i=0
Thus,
|αn+k | ≤ |αn+k−1 |2
for all n. 
Theorem 2.2. A unitarily equivalent operator to a k-quasi-parahyponormal operator is
also k-quasi-parahyponormal.

Proof. Let T be a k-quasi-parahyponormal operator, and let A ∈ B(H) be unitarily equiv-

alent to T. Then, there exists a unitary operator U on H satisfying A = U ? T U. Hence,
A?k ((AA? )2 − 2λA? A + λ2 )Ak =
= U ? T ?k U (U ? (T T ? )2 U − 2λU ? T ? T U + λ2 )U ? T k U
= U ? T ?k (T T ? )2 T k U − 2λU ? T ?k T ? T T k U + λ2 U ? T ?k T k U
= U ? T ?k ((T T ? )2 − 2λT ? T + λ2 )T k U ≥ 0
Thus, A is k-quasi-parahyponormal. 
Theorem 2.3. Let T ∈ B(H) be a k-quasi-parahyponormal operator, and let S ∈ B(H)
be an isometry. If T commutes with S, then T S is k-quasi-parahyponormal.
278 TWMS J. APP. AND ENG. MATH. V.13, N.1, 2023

Proof. Since T is k-quasi-parahyponormal,

(T S)∗k ((T S)(T S)? )2 − 2λ(T S)? T S + λ2 )(T S)k =
= T ?k S ?k ST T ? T T ? S ? − 2λT ? T + λ2 S k T k
 

= T ?k S ?k−1 S ? S(T T ? )2 S ? S − 2λS ? T ? T S + λ2 S ? S S k−1 T k

 

= T ?k S ?k−1 (T T ? )2 − 2λT ? T + λ2 S k−1 T k

 

= S ?k−1 T ?k (T T ? )2 − 2λT ? T + λ2 T k S k−1 ≥ 0

 


Let T ∈ B(H). Denote by R(σ(T )) for the set of all rational analytic functions on the
spectrum σ(T ) of T.
Definition 2.3. [8] The operator T is said to be n-multicyclic, if there exist n (generating)
vectors x1 , x2 , .., xn in H such that
_
{g(T )xi , 1 ≤ i ≤ n , g ∈ R(σ(T ))} = H

We have then
Theorem 2.4. If T is an n-multicyclic k-quasi-parahyponormal operator, then its restric-
tion on ran(T k ) is also n-multicyclic.

Proof. Put
 
T1 T2
T =
0 T3
on the decomposition H = ran(T k ) ⊕ ker(T ?k ). Since σ(T1 ) ⊂ σ(T ) by [13, Theorem 2.3],
R(σ(T1 )) ⊂ R(σ(T )). The operator T is n-multicyclic. Then, there exist n generating
vectors x1 , x2 , .., xn ∈ H for which
_
{g(T )xi , 1 ≤ i ≤ n, g ∈ R(σ(T ))} = H

Put yi = T k xi , 1 ≤ i ≤ n. Hence,
_ _n o
{g(T1 )yi , 1 ≤ i ≤ n, g ∈ R(σ(T ))} = g(T1 )T k xi , 1 ≤ i ≤ n, g ∈ R(σ(T ))
_n o
k
= g(T )T xi , 1 ≤ i ≤ n, g ∈ R(σ(T ))
_n o
= T k g(T )xi , 1 ≤ i ≤ n, g ∈ R(σ(T ))

= ran(T k )
But
_ _
{g(T1 )yi , 1 ≤ i ≤ n, g ∈ R(σ(T ))} ⊂ {g(T1 )yi , 1 ≤ i ≤ n, g ∈ R(σ(T1 ))}

Thus,
_
ran(T k ) ⊂ {g(T1 )yi , 1 ≤ i ≤ n, g ∈ R(σ(T1 ))}

Therefore, {yi }ni=1 are n-generating vectors of T1 , and T1 is n-muticyclic. 

 AISSA NASLI BAKIR: A NEW CLASS OF NON NORMAL OPERATORS 279

3. Class of (M, k)-quasi-parahyponormal operators

Important properties of (M, k)-quasi-parahyponormal operators are shown in this sec-
tion. In particular, the matrix representation, the finite ascent and the SVEP of (M, k)-
quasi-parahyponormal operators are presented.
Definition 3.1. [12][13] An operator T ∈ B(H) is said to be M -parahyponormal if there
exists M > 0, satisfying
M (T T ? )2 − 2λT ? T + λ2 ≥ 0
Definition 3.2. An operator T ∈ B(H) is said to be (M, k)-quasi-parahyponormal for
some integer k, if there exists M > 0, satisfying
T ?k (M (T T ? )2 − 2λT ? T + λ2 )T k ≥ 0
This definition is clearly equivalent to
√
kT k+1 xk2 ≤ M kT k xkkT T ? T k xk (1)
for all x ∈ H.
Inequality (1) shows that this class of operators is nested with respect to M, i.e., an
(M1 , k)-quasi-parahyponormal operator is (M2 , k)-quasi-parahyponormal for 0 < M1 <
M2 .
Example 3.1. Let’s consider the unilateral weighted right shift on `2 (N) defined by T en =
αn en+1 , where α1 = α3 = 21 , α2 = 14 , αn = 1, n ≥ 4, and (en ) is the satandard basis of
`2 (N). By a direct computation, we can show that T is M -quasi-parahyponormal. However,
1 √
kT e1 k2 = > M ke1 kkT T ? e1 k = 0
4
which contradicts the inequality (1). Hence, T is not M -parahyponormal.
Example (3.1) shows that the classes of (M, k)-quasi-parahyponormal operators do not
coincide.
Proposition 3.1. The class of (M, k)-quasi-parahyponormal operators is
1. Closed for the multiplication by scalars.
2. Not convex.
3. Not translation invariant.
Proof. 1. Let T be an (M, k)-quasi-parahyponormal operator, and let α be any complex
scalar. For all x ∈ H we have
√
k(αT )k+1 xk2 = |α|2k+2 kT k+1 xk2 ≤ |α|2k+2 M kT k xkkT T ? T k xk
√
= M k(αT )k xkk(αT )(αT )? (αT )k xk
Then, αT is also (M, k)-quasi-parahyponormal.
   
1 1 −1 0
2. On the other hand, operators T = and S= are M -parahyponormal
0 1 0 −1
for M ≥ 41 , in particular M = 2. However, the operator A = 12 (T + S) is not 2-
parahyponormal. Indeed, for x = (0, 1) ∈ C2 , we get
1 √
kA(0, 1)k2 = > 2k(0, 1)kkAA? (0, 1)k = 0
4
This contradicts the inequality (1). Hence, the above class is not convex.
280 TWMS J. APP. AND ENG. MATH. V.13, N.1, 2023
 
1 1
3. For T = , α = 1, β = −1, and x = (0, 1), the operator
0 1
 
0 1
R = αT + β = T − I =
0 0
verifies √
kR(0, 1)k2 = 1 > 2k(0, 1)kkRR? (0, 1)k = 0
Hence, R is not 2-parahyponormal. Thus, the considered class is not translation invariant.

With the same arguments of Theorems 2.2 and 2.3, we can state the following result
Theorem 3.1. 1. Unitarily equivalent operators to an (M, k)-quasi-parahyponormal op-
erator are also (M, k)-quasi-parahyponormal.
2. The product T S is (M, k)-quasi-parahyponormal whenever T is (M, k)-quasi-parahyponormal
and S is an isometry that commutes with T.
Remark 3.1. Assertion (1) of Theorem 3.1 is in general false if the operators are similar
to the (M, k)-quasi-parahyponormal operator and not unitarily equivalent.
Indeed, the bilateral weighted shift T defined on the Hilbert space `2 (Z) by

e , n ≤ 1 or n ≥ 3
T en = √n+1
2e3 , n=2
is M -parahyponormal for M ≥ 8, by inequality (1). In particuliar it is 8-parahyponormal,
and the operator 
en+1 , n ≤ 1 or n ≥ 3
U en = 1
3 e3 , n=2
is invertible and not unitary. Nonetheless, the operator B = U −1 T U is not 8-parahyponormal
since
 e√n+1 , n ≤ 0 or n ≥ 3


Ben = 3 2e2 , n=1
1
e3 , n=2


3
and √ √
kBe1 k2 = 18 > 8ke1 kkBB ? e1 k = 8
Theorem 3.2. Let T ∈ B(H) be an (M, k)-quasi-parahyponormal operator. If ran(T k )
is dense in H, then T is M -parahyponormal.
Proof. Let x ∈ H. By the hypothesis, there exists a sequence (xn )n in H such that x =
lim T k xn . Since T is (M, k)-quasi-parahyponormal,
n→∞
√ √
M kxkkT T ? xk = M k lim (kT k xn kkT T ? T k xn k)
n→∞
√
= lim M kT k xn kkT T ? T k xn k
n→∞
≥ lim kT k+1 xn k2
n→∞
≥ k lim T k+1 xn k2
n→∞
2
= kT xk
by the continuity of the inner product. Hence, T is M -parahyponormal. 
 AISSA NASLI BAKIR: A NEW CLASS OF NON NORMAL OPERATORS 281

Corollary 3.1. Let T be a nonzero (M, k)-quasi-parahyponormal operator but not M -

parahyponormal. Then, T admits at least a non trivial closed invariant subspace.
Proof. Suppose that T has no non trivial closed invariant subspace. Since T 6= 0, ker(T ) 6=
H and ran(T ) 6= {0} are non trivial closed invariant subspaces for T. Thus, we must
have ker(T ) = {0} and ran(T ) = H. By Theorem 3.2, T is M -parahyponormal, which
contradicts the hypothesis. 
Note that the existence of nontrivial closed invariant subspaces for operators on Hilbert
spaces remains until now, one of the hot open problems in operator theory, see [6, 11] for
further details.
Theorem 3.3. Let T ∈ B(H) be an (M, k)-quasi-parahyponormal operator. If ran(T k ) 6=
H, then T admits the matrix representation
 
T1 T2
T =
0 T3
on H = ran(T k ) ⊕ ker(T ?k ). Furthermore, T1 is M -parahyponormal, T3 is nilpotent of
order k, and σ(T ) = σ(T1 ) ∪ {0} .
Proof. Since T is (M, k)-quasi-parahyponormal,
hT ?k (M (T T ? )2 − 2λT ? T + λ2 )T k y, yi ≥ 0
for all y ∈ H. Hence,
h(M (T T ? )2 − 2λT ? T + λ2 )T k y, T k yi ≥ 0
Thus, for all x ∈ ran(T k ),
(M (T T ? )2 − 2λT ? T + λ2 )x, x = (M (T1 T1? )2 − 2λT1? T1 + λ2 )x, x ≥ 0
Consequently, T1 is M -parahyponormal. Let now P be the orthogonal projection on
ran(T k ). For all x = x1 + x2 , y = y1 + y2 ∈ H, we have
D E D E D E
T3k x2 , y2 = T k (I − P )x, (I − P )y = (I − P )x, T ∗k (I − P )y = 0

Thus, T3k = 0. Furthermore, σ(T1 ) ∪ σ(T3 ) = σ(T ) ∪ Ω, where Ω is the union of holes in
σ(T ) which happen to be a subset of σ(T1 ) ∩ σ(T3 ) by [7, Corollary 7], with σ(T1 ) ∩ σ(T3 )
has no interior point, and T3 is nilpotent. Thus, σ(T ) = σ(T1 ) ∪ {0} . 
Corollary 3.2. Let T ∈ B(H) be (M, k)-quasi-parahyponormal. If the restriction T1 =
T ran(T k ) is invertible, then T is similar to the sum of an M -parahyponormal operator
and a nilpotent operator.
Proof. Let  
T1 T2
T = on H = ran(T k ) ⊕ ker(T ∗k )
0 T3
Then, T1 is M -parahyponormal by Theorem 3.3. Since T1 is invertible, 0 ∈ / σ(T ). Hence,
σ(T1 ) ∩ σ(T3 ) = ∅. By Rosenblum’s Corollary [14], [15], there exists S ∈ B(H) for which
T1 S − ST3 = T2 . Thus,
   
I −S T1 0 I S
T =
0 I 0 T3 0 I
 −1  
I S I S
= (T1 ⊕ T3 )
0 I 0 I
282 TWMS J. APP. AND ENG. MATH. V.13, N.1, 2023


Definition 3.3. [1] An operator T in B(H) is said to have the Single Valued Extension
Property, briefly SVEP, at a complex number α, if for each open neighborhood V of α, the
unique analytic function f : V → H satisfying
(T − λ)f (λ) = 0
for all λ ∈ V is f ≡ 0.
Furthermore, T is said to have SVEP if T has SVEP at every complex number.
Definition 3.4. [1] For T ∈ B(H), the smallest integer m such that ker(T m ) = ker(T m+1 )
is said to be the ascent of T, and is denoted by α(T ). If no such integer exists, we shall
write α(T ) = ∞.
Definition 3.5. [1] The smallest integer m such that ran(T m ) = ran(T m+1 ) is said to be
the descent of T, and is denoted by δ(T ). If no such integer exists, we set δ(T ) = ∞.
According to [1], if α(T ) and δ(T ) are both finite, then α(T ) = δ(T ). For more details on
these notions, reader can see [1, 2] and [5]. Now, we’ll show that the considered operators
have finite ascent and SVEP.
Theorem 3.4. An (M, k)-quasi-parahyponormal operator T ∈ B(H) has finite ascent.
Proof. Let x ∈ ker(T k+1 ). Since T is (M, k)-quasi-parahyponormal operator, there exists
M > 0 such that
T ?k (M (T T ? )2 − 2λT ? T + λ2 )T k ≥ 0
Hence,
hT ?k (M (T T ? )2 T k x, xi − 2λhT ?k T ? T T k x, xi + λ2 hT ?k T k x, xi ≥ 0
for all λ > 0. Thus,
M kT T ? T k xk2 + λ2 kT k xk2 ≥ 0
for all λ > 0. Therefore, T k x = 0, which implies that x ∈ ker(T k ). This finishes the proof
since clearly ker(T k ) ⊂ ker(T k+1 ). 
Corollary 3.3. (M, k)-quasi-parahyponormal operators have SVEP at 0.
Proof. Immediate consequence of Theorem 3.4 and [1, Theorem 3.8]. 
Definition 3.6. [1] For an operator T ∈ B(H), the local resolvent set of T at a vector
x ∈ H, denoted by ρT (x), is defined to consist of complex elements z0 such that there
exists an analytic function f (z) defined in a neighborhood of z0 , with values in H, for
which (T − z)f (z) = x.
Definition 3.7. [1] The set σT (x) = C \ ρT (x) is called the local spectrum of T at x.
We’ve then the following important result
Theorem 3.5. Let  
T1 T2
T =
0 T3
be an (M, k)-quasi-parahyponormal operator with respect to the decomposition H = ran(T k )⊕
ker(T ?k ). Then, for all x = x1 + x2 ∈ H :
(a) σT3 (x2 ) ⊂ σT (x1 + x2 ).
(b) σT1 (x1 ) = σT (x1 + 0).
 AISSA NASLI BAKIR: A NEW CLASS OF NON NORMAL OPERATORS 283

Proof. a. Let z0 ∈ ρT (x1 + x2 ). Then, there exists a neighborhood U of z0 and an analytic

function f (z) defined on U, with values in H, for which
(T − z)f (z) = x, z ∈ U (2)
Let f = f1 + f2 where f1 , f2 are in the spaces O(U, ran(T k )), O(U, ker(T ?k )) respectively,
consisting of analytic functions on U with values in H, with respect to the uniform topology
[1]. Equality (2) can then be written
    
T1 − z T2 f1 (z) x1
=
0 T3 − z f2 (z) x2
Then (T3 − z)f2 (z) = x2 , z ∈ U. Hence, z0 ∈ ρT3 (x2 ). Thus, (a) holds by passing to the
complement.
b. If z1 ∈ ρT (x1 +0), then there exists a neighborhood V1 of z1 and an analytic function
g defined on V1 with values in H verifying
(T − z)f (z) = x1 + 0, z ∈ V1 (3)
Let g = g1 + g2 , where g1 ∈ O(V1 , ran(T k )), g2 ∈ O(V1 , ker(T ?k )) are as in (a). From
equation (3) we obtain
(T1 − z)g1 (z) + T2 g2 (z) = x1 and (T3 − z)g2 (z) = 0, z ∈ V1
Since T3 is nilpotent by Theorem 3.3, T3 has SVEP by [1]. Thus, g2 (z) = 0. Consequently,
(T1 − z)g1 (z) = x1 . Therefore, z1 ∈ ρT1 (x1 ), and then ρT (x1 + 0) ⊂ ρT1 (x1 ). Thus,
σT1 (x1 ) ⊂ σT (x1 + 0).
Now, if z2 ∈ ρT1 (x1 ), then, there exists a neighborhood V2 of z2 and an analytic function
h from V2 onto H, such that (T1 − z)h(z) = x1 for all z ∈ V2 . Hence,
(T − z)(h(z) + 0) = (T1 − z)h(z) = x1 = x1 + 0
Thus, z2 ∈ ρT (x1 + 0). 

4. Conclusion
We’ve shown certain fundamental properties of the given class of operators in the present
article. The ascent, the matrix representation, the restriction on invariant subspaces and
the n-multicyclicity as well as other considerable properties are established.
Acknowledgement. I would like to present my warm thanks to the referees for their
careful and valuable reading of the manuscript. My gratitude is also expressed to the
DGRSDT in Algeria for financing the Laboratory of Mathematics and Application LMA
in my University.

References
[1] Aiena, P., (2004), Fredholm and Local Spectral Theory with Applications to Multipliers, Kluwer Aca-
demic Publishers.
[2] Caradus, S. R., (1966), Operators with finite ascent and descent, Pac. J. Math., 18, No. 3, pp. 437-449.
[3] Chennappan, N. Karthikeyan, S., (2000), Paranormal composition operators, Indian J. Pure Appl.
Math 31, pp. 591-601.
[4] B. P. Duggal, B. P., Jeon, I. H., Kim, I. H., (2006), On Weyl’s theorem for quasi-class A operators, J.
Kor. Math. Soc., 43, No. 4, pp. 899-909.
[5] Finch, J. K., (1975), The Single Valued Extension Property On a Banach Space, Pac. J. Math. 58, No.
1, pp. 61-69.
[6] Foias, C., Pearcy, C., (2005), On the hyperinvariant subspace problem, J. Funct. Anal, 219, pp. 134-142.
284 TWMS J. APP. AND ENG. MATH. V.13, N.1, 2023

[7] Han, J. K., Lee, H. Y., Lee, W. Y., (1999), Invertible completions of 2 × 2 upper triangular operator
matrices, Proc. Amer. Math. Soc., 128, pp. 119-123.
[8] Herrero, D. A., (1978), On multicyclic operators, Int. equa. Oper. Th., 1, pp. 57-102.
[9] Mecheri, S., (2013), On quasi-class A operators, Anal.Tint. Ifice. Ale. Univ. Matematica, Tomul LIX,
f.1, DOI: 10.2478/v10157-012-0020-0, pp. 163-172.
[10] Mecheri, S., (2012), Bishop’s property β and Riesz idempotent for k-quasi-paranormal operators,
Banach J. Math. Anal., 6, No. 1, pp. 147-154.
[11] Mecheri, S., (2018), Hyperinvariant Subspace Problem for Some Classes of Operators, Mediterr. J.
Math., 15:185.
[12] Panayappan, S., Senthilkumar, D., (2002), Parahyponormal and M -*-paranormal composition oper-
ators, Acta Ciencia Indica, XXVIII, No. 4, pp. 611-614.
[13] Parvatham S., Senthilkumar, D., (2020), Spectral properties of k-quasi-parahyponormal operators,
Int. J. Math. Trends Tech., 66, pp. 73-76.
[14] Rosenblum, M., (1969), The operator equation BX − XA = Q with self adjoint A and B, Proc. Amer.
Math. Soc, 20, pp. 115-120.
[15] Schweinsberg, A. R., (1982), The operator equation AX − XB = C with normal A and B, Pacific J.
Math., 102, No.2, pp. 447-453.
[16] Veluchamy, T., (2010), Posinormal Composition Operators on Weighted Hardy space, Int. Math.
Forum, 5, No. 24, pp. 1195-1205.

Dr. Aissa Nasli Bakir is currently an associate professor in the Department of

Mathematics, Hassiba Benbouali University of Chlef, Algeria. He obtained his PhD
from Abdelhamid Ibn Badis University of Mostaganem in Algeria in 2010. His area
of interest includes functional analysis and operator theory.