Advances in Mathematics: Scientific Journal 12 (2023), no.

1, 1–16
ADV MATH ISSN: 1857-8365 (printed); 1857-8438 (electronic)
SCI JOURNAL
https://doi.org/10.37418/amsj.12.1.1

ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE

INVERSE

Anissa Elgues and Safa Menkad1

A BSTRACT. Let T ∈ B(H) be a bounded linear operator on a complex Hilbert

space H. For n ∈ N, an operator T ∈ B(H) is said to be n-normal if T n T ∗ =
T ∗ T n . In this paper we investigate a necessary and sufficient condition for the
n-normality of ST and T S, where S, T ∈ B(H). As a consequence, we generalize
Kaplansky theorem for normal operators to n-normal operators. Also, In this pa-
per, we provide new characterizations of n-normal operators by certain conditions
involving powers of Moore-Penrose inverse.

1. I NTRODUCTION AND PRELIMINARIES

let H be a complex Hilbert space and let B(H) be the algebra of all bounded
linear operators on H. For an arbitrary operator T ∈ B(H), we denote by R(T ),
N (T ) and T ∗ for the range, the null subspace and the adjoint operator of T . It is
well known that for T ∈ B(H), there is a unique factorization T = U |T |, where
1
N (U ) = N (T ) = N (|T |), U is a partial isometry, i.e. U U ∗ U = U and |T | = (T ∗ T ) 2
is the modulus of T. This factorization is called the polar decomposition of T. An
operator T ∈ B(H) is said to be nilpotent if there exists p ∈ N such that T p = 0,
1
corresponding author
2020 Mathematics Subject Classification. Primary 47A05, Secondary 47B15.
Key words and phrases. n-normal operator, Polar decomposition, Moore-Penrose inverse, EP oper-
ator, Group inverse.
Submitted: 07.12.2022; Accepted: 23.12.2022; Published: 02.01.2023.
1
2 A. Elgues and S. Menkad

normal if T T ∗ = T ∗ T and n-normal if T n T ∗ = T ∗ T n , where n ∈ N. The class

of n-normal operators was first introduced and studied by A.S. Jibril [12] as an
extension of the class of normal operators. He showed that T is n-normal if and
only if T n is normal. From the definition, it is easily seen that this class contains
nilpotent operators with nilpotency n. Also the class of n-normal operators has
been studied by many authors, mainly Alzuraiqi and Patel [2], M. Cho and B.
Nastovska [6] and Mosić and Djordjević [18].
Suppose that T, S ∈ B(H). It is known that, in general, ST is not normal.
Historical notes and several versions of the problem are investigated. Kaplansky
showed that it may be possible that ST is normal while T S is not. Indeed, in
[13] he showed that if S and ST are normal, then T S is normal if and only if T
commutes with |S|. Later, Deutsch, Gibson and Schneider [8] proved that, if T
and S are two complex square matrices, then ST and T S are normal if and only
if S ∗ ST = T SS ∗ and ST T ∗ = T ∗ T S.
In section 2 of this paper, we focus on to study the n-ormality of ST and T S.
Firstly, we present a new characterization of n-normal operators via the polar
decomposition. In particular, we show that if the partial isometry factor U of the
polar decomposition of the operator T is normal, then T is n-normal if and only
if (|T |U )n = (U |T |)n . Afterwards, we generalize Kaplansky theorem for normal
operators to n-normal operators. Also, we use the Fuglede-Putnam theorem to
prove that both T S and ST are n-normal if and only if S ∗ (ST )n = (T S)n S ∗ and
(ST )n T ∗ = T ∗ (T S)n . Finally, we provide conditions related to the factors polar
decomposition of S under which T S and ST becomes n-normal.
Now, we recall some notions that will be used in section 3 of this paper. For
T ∈ B(H), the Moore-Penrose inverse of T is the unique operator T + ∈ B(H)
which satisfies:

T T + T = T, T + T T + = T + , (T T + )∗ = T T + , (T + T )∗ = T + T.

It is well known that the Moore-Penrose inverse of T exists if and only if R(T ) is
closed. It is easy to see that R (T + ) = R (T ∗ ), T T + is the orthogonal projection
of H onto R (T ) and that T + T is the orthogonal projection of H onto R (T ∗ ). The
operator T is said to be EP operator, if R (T ) is closed and T T + = T + T . Clearly

T EP ⇐⇒ R(T ) = R(T ∗ ) ⇐⇒ N (T ) = N (T ∗ ).
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 3

Obviously, every normal operator with closed range is EP but the converse is not
true even in a finite dimensional space. For more details about on EP operators
see [5, 9, 18].
The ascent a(T ) and the descent d(T ) of T are given by
a(T ) = min{p ∈ N : N (T p ) = N (T p+1 )},
and
d(T ) = min{p ∈ N : R(T p ) = R(T p+1 )}.
In both cases the infimum of the empty set is equal to ∞. If a(T ) and d(T ) are
finite, they are equal and their common value is called the index of T and it is
denoted by ind(T ) [11, Proposition 38.3].
The group inverse of T ∈ B(H) is the unique operator T # ∈ B(H) such that

T T # T = T T # T T # = T # T T # = T # T.

T # exists if and only if ind(T ) ≤ 1 [14]. When ind(T ) = 0, the group inverse
reduces to the standard inverse, that is, T # = T −1 .
In section 3, firstly, we show that if T is n-normal operator with closed range,
then ind(T ) ≤ n, and T can be written as a direct sum of an invertible n-normal
operator and a nilpotent operator. Secondly, We prove that if T is Moore-Penrose
invertible, then T is 2n-normal if and only if its Moore-Penrose inverse is too 2n-
normal. But, this need not be true in case of (2n+1)-normal operators, as shown
in example 3.5. In [18], Mosić and Djordjević presented a number of necessary
and sufficient conditions for both Moore-penrose invertible and group invertible
elements in rings with involution to be n-normal. Motivated by them, we will give
some new equivalent conditions for the n-normaliy of Moore-Penrose invertible
operators, by omitting the assumption of the group invertibility. Finally, we show
that an n-normal operator is EP, if it is group invertible.
Now, we state some well-known results needed in the sequel.

Proposition 1.1. [2, 12] Let T ∈ B(H) and n ∈ N. T is n-normal if and only if T n
is normal.

Proposition 1.2. [2] Let T ∈ B(H) be n-normal. Then the following hold.
(a) T ∗ is n-normal.
(b) T −1 exists, then T −1 is n-normal.
4 A. Elgues and S. Menkad

(c) T m is n-normal for any positive integer m.

(d) If S ∈ B(H) is unitary equivalent to T , then S is n-normal.

The well-known Fuglede-Putnam’s Theorem is as follows:

Theorem 1.3. [19] Let T, S ∈ B(H) be normal operators and X ∈ B(H). If XS =

T X, then XS ∗ = T ∗ X.

Lemma 1.4. [17] For any T ∈ B(H) with closed range, then the following are
satisfied
(a) (T ∗ )+ = (T + )∗ .
(b) (T T ∗ )+ = (T ∗ )+ T + .
(c) (T ∗ T )+ = T + (T ∗ )+ .
(d) T + = (T ∗ T )+ T ∗ = T ∗ (T T ∗ )+ .
(e) (T ∗ )+ = T (T ∗ T )+ = (T T ∗ )+ T.
(f ) If S ∈ B(H) such that ST = T S and ST ∗ = T ∗ S, then ST + = T + S.

Lemma 1.5. [4] If T ∈ B(H) is group invertible, then

(a) T is EP if and only if R(T ) ⊂ R(T ∗ ).
(b) If S ∈ B(H), then ST = T S if and only if ST # = T # S.

2. O N THE n- NORMALITY OF PRODUCTS OF OPERATORS

Let T = U |T | be the polar decomposition of an operator T. Then It is well known

that T is normal is if and only if U is normal and U commutes with |T | [10]. The
following example shows that this need not be true in case of n-normal operators.
 
0 0 0
Example 2.1. Let T = 1 0 0 ∈ C3 . Then T 3 = 0, which implies that T is
 

1 1 0
3-normal. The canonical polar decomposition of T is T = U |T |, where
   
0 0 0 3 1 0
U = √15 2 −1 0 and |T | = √15 1 2 0
   

1 2 0 0 0 0
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 5
 
0 0 0
3 1
it is easy to verify that U = √  2 −1 0 is not normal. So U is not 3-normal
 
5 5
−4 2 0
and
 
2 −1 0
U |T | = T ̸= 4 −2 0 = |T |U.
 

0 0 0

Now, we present some new conditions under which certain operators are n-
normal.

Proposition 2.2. Let n be a positive integer and T ∈ B(H), with polar decomposition
T = U |T |. If U is normal, then the following statements are equivalent

(1) T is n-normal.
(2) (U |T |)n = (|T |U )n .
(3) U T n = T n U.

Proof.
(1) ⇒ (2) and (2) ⇒ (3) holds without the condition U is normal.
(1) ⇒ (2). Assume that T is n-normal. Then T n T ∗ = T ∗ T n . That is T n T ∗ T =
T ∗ T T n and so T n |T | = |T |T n . It follows that

(U |T |)n |T | = |T |(U |T |)n = (|T |U )n |T |.

Hence, (U |T |)n = (|T |U )n on R(|T |). Since N (|T |) = N (U ), then

(U |T |)n = (|T |U )n = 0 on N (|T |).

Therefore,

(U |T |)n = (|T |U )n on H = R(|T |) ⊕ N (|T |).

(2) ⇒ (3). Using the assumption (2), we obtain

U T n = U (U |T |)n = U (|T |U )n = (U |T |)n U = T n U.

6 A. Elgues and S. Menkad

(3) ⇒ (1). Assume that U T n = T n U . Since U is normal, by Fuglede-Putnam

theorem we get U ∗ T n = T n U ∗ . It follows that

T ∗ T n = |T |U ∗ (U |T |)n
= |T |(U |T |)n U ∗
= (|T |U )n |T |U ∗
= (|T |U )n T ∗ .

On the other hand, since U ∗ U is an orthogonal projection onto R(|T |), we have

T n T ∗ = (U |T |)n |T |U ∗
= (U |T |)n U ∗ U |T |U ∗
= U ∗ (U |T |)n U |T |U ∗
= U ∗ U (|T |U )n |T |U ∗ .
= (|T |U )n T ∗ .

Hence, T ∗ T n = T n T ∗ . So that T is n-normal. □

In [13], Kaplansky showed that if S, T ∈ B(H) such that S and ST are nor-
mal, then T S is normal if and only if T commutes with |S|. Kaplansky theorem’s
has been extended from normal operators to hyponormal operators [1], quasinor-
mal operators [3] and D-normal matrices [7]. We have the following Kaplansky
theorem’s for n-normal operators.

Theorem 2.3. Let T, S ∈ B(H) such that S is normal and ST is n-normal. Then
S ∗ ST = T S ∗ S =⇒ T S is n-normal.

Proof. Since S is normal, then by [10, Theorem 3] there exists a unitary operator
U such that
S = U |S| = |S|U.
1 1
The assumption S ∗ ST = T S ∗ S is equivalent to (S ∗ S) 2 T = T (S ∗ S) 2 . so that
|S|T = T |S|. Consequently, we have

T S = T U |S| = |S|T U = U ∗ U |S|T U = U ∗ ST U.

Therefore, T S is unitary equivalent to ST . Since ST is n-normal, according to

Proposition 1.2 (d), we conclude that T S is also n-normal. □
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 7

Remark 2.4. In Theorem 2.3, the reverse implication is false as shown by the follow-
ing example.
! !
0 0 0 I
Example 2.5. Let S = and T = in B(H ⊕ H), where P ≥ 0. Then
0 P I 0
S is normal and by a simple computation, we have (ST )2 = (T S)2 = 0. Hence ST
and T S are 2-normal, but
! !
2
0 0 0 P
S ∗ ST = ̸= = T S ∗ S.
P2 0 0 0

As a consequence of the previous theorem, we have the following result.

Corollary 2.6. Let T, S ∈ B(H) such that S is unitary. Then

ST is n-normal ⇐⇒ T S is n-normal.

Proof. =⇒ Assume that ST is n-normal. Since S is unitary, then it is normal and

S ∗ ST = T S ∗ S = T. Hence, by Theorm 2.3, ST is n-normal.

⇐=. Now we suppose that T S is n-normal. By Proposition 1.2 (a), S ∗ T ∗ is

n-normal. Moreover, since S ∗ is unitaire, by the above implication T ∗ S ∗ is too
n-normal. Therefore, ST is n-normal. □

The following result generalize Theorems 2 obtained for normal matrices in [8]
to n-normal operators on an arbitrary Hilbert space.

Theorem 2.7. Let T, S ∈ B(H) and n ∈ N. Then ST and T S are n-normal if and
only if S ∗ (ST )n = (T S)n S ∗ and (ST )n T ∗ = T ∗ (T S)n .

Proof. Assume that ST and T S are n-normal. Then (ST )n and (T S)n are normal.
Hence, From the hypothesis (ST )n S = S(T S)n and T (ST )n = (T S)n T and by
Fuglede-Putnam Theorem, we get

S ∗ (ST )n = (T S)n S ∗ and (ST )n T ∗ = T ∗ (T S)n .

Conversely, if S ∗ (ST )n = (T S)n S ∗ and (ST )n T ∗ = T ∗ (T S)n , then multiplying the

first equation by T ∗ and the second one by S ∗ we deduce that

(ST )∗ (ST )n = T ∗ (T S)n S ∗ = (ST )n (ST )∗ ,

8 A. Elgues and S. Menkad

and
(T S)n (T S)∗ = S ∗ (ST )n T ∗ = (T S)∗ (T S)n .
Therefore, ST and T S are n-normal. □

Theorem 2.8. Let T, S ∈ B(H) and let S = U |S| be the polar decompositions of S
with U is unitary. For n ∈ N. The following properties hold

(1) If T U is normal and (ST )n U = U (T S)n , then ST and T S are n-normal.

(2) If ST and T S are n-normal, then (ST )n U = U (T S)n .

Proof.
(1) Suppose that T U is normal and (ST )n U = U (T S)n .
In case n = 1, since U is unitary, then the assumption ST U = U T S is equivalent
to |S|T U = T U |S|. It follows from [8, Theorem 3], that ST and T S are normal.
Now, in case n ≥ 2, we observe that

(ST )n U = U (T S)n ⇐⇒ ST (ST )n−1 U = U T (ST )n−1 S

⇐⇒ U |S|T (ST )n−1 U = U T (ST )n−1 U |S|
⇐⇒ U ∗ U |S|T (ST )n−1 U = U ∗ U T (ST )n−1 U |S|
(2.1) ⇐⇒ |S|T (ST )n−1 U = T (ST )n−1 U |S|
(2.2) ⇐⇒ |S|2 T (ST )n−1 U = T (ST )n−1 U |S|2 .

Consequently,

(T S)n S ∗ = T (ST )n−1 SS ∗

= T (ST )n−1 U |S|2 U ∗
= |S|2 T (ST )n−1 U U ∗ (by (2.2))
= |S|2 T (ST )n−1
= |S|U ∗ U |S|T (ST )n−1
= S ∗ (ST )n .
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 9

On the other hand, we have

|S|(T S)n−1 T U = |S|T (ST )n−1 U

= T (ST )n−1 U |S| (by (2.1))
= (T S)n
= T U |S|(T S)n−1 .

Since T U is normal, then by Fuglede-Putnam theorem we obtain

(T U )∗ |S|(T S)n−1 = |S|(T S)n−1 (T U )∗ .

From this equality, it follows

(ST )n T ∗ = S(T S)n−1 T T ∗

= U |S|(T S)n−1 T U U ∗ T ∗
= U |S|(T S)n−1 (T U )∗ T U ( T U is normal)
= U (T U )∗ |S|(T S)n−1 T U
= U U ∗ T ∗ |S|(T S)n−1 T U
= T ∗ T U |S|(T S)n−1
= T ∗ (T S)n .

Hence, by Theorem 2.7, ST and T S are n-normal.

To prove (2), we also consider two cases:
Case (i): n = 1. Since ST and T S are normal, by [8, Theorem 3], |S|T U =
T U |S|. multiplying this equality by U from the left side, we get ST U = U T S.
Case (ii): n ≥ 2. By (2.2), it is enough to prove that |S|2 T (ST )n−1 U = T (ST )n−1 U |S|2 .
Sinse ST and T S are n-normal, using again the Theorem 2.7, we obtain

|S|2 T (ST )n−1 U = S ∗ ST (ST )n−1 U

= S ∗ (ST )n U
= (T S)n S ∗ U
= T (ST )n−1 SS ∗ U
= T (ST )n−1 |S ∗ |2 U
= T (ST )n−1 U |S|2 (since |S ∗ |U = U |S|).
10 A. Elgues and S. Menkad

3. M OORE -P ENROSE INVERSE AND n- NORMAL OPERATORS

It is well known that the ascent and descent of a normal operator with closed
range are finite. It is equally true for an n-normal operator with closed range.

Theorem 3.1. Let T ∈ B(H) be an n-normal. Then the following properties hold:
(1) a(t) ≤ n.
(2) If T has a closed range, then
(a) ind(T ) ≤ n.
(b) R(T k ) is closed for
" all k #≥ n and T has the following matrix rep-
T1 0
resentation T = with respect to the orthogonal sum H =
0 T2
R(T n ) ⊕ N (T n ), where T1 = T \ R(T n ) is an invertible n-normal oper-
ator and T2 is a nilpotent operator with nilpotency n.

Proof.
(1) To prove that a(t) ≤ n it will suffice to show that N (T n+1 ) ⊂ N (T n ). As the
converse inclusion is obvious it follows that N (T n ) = N (T n+1 ), which implies that
N (T k ) = N (T k+1 ), for k ≥ n. Let x ∈ N (T n+1 ). Then T n+1 (x) = 0 and so we get

T n (x) ∈ N (T ) ∩ R(T n ) ⊂ N (T n ) ∩ R(T n ) ⊂ N (T n ) ∩ R(T n ).

Since T is n-normal, T n is normal. Therefore, N (T n ) = N ((T n )∗ ). Consequently,

T n (x) ∈ N ((T n )∗ ) ∩ R(T n ) = {0}.

Hence, x ∈ N (T n ).
(2) Now, suppose that R(T ) is closed.
(a) If n = 1, then T is normal with clsed range and so ind(T ) ≤ 1.
If n ≥ 2, then (T ∗ )n T = T (T ∗ )n and R(T ) = R(T T ∗ ), since T is n-normal with
closed range. It follows that

R((T n )∗ ) = (T ∗ )n−1 R(T ∗ ) = (T ∗ )n−1 R(T ∗ T ) = R((T ∗ )n T ) = R(T (T ∗ )n ).

Then, R((T n )∗ ) ⊂ R(T ). Since T T + is the orthogonal projection onto R(T ), there-
fore T T + (T n )∗ = (T n )∗ and so T n+1 T + = T n . Thus, R(T n ) ⊂ R(T n+1 ). Also, since
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 11

the converse inclusion is obvious it follows that R(T n ) = R(T n+1 ), which implies
that N (T k ) = N (T k+1 ), for k ≥ n. So, we get d(T ) ≤ n and thus, ind(T ) ≤ n
by [11, Proposition 38.3].
(b) Finally, since ind(T ) ≤ n, according to [16, Corollary 2.2 ], we deduce that
R(T k ) is closed for all k ≥ n and
" #
T1 0
T = : R(T n ) ⊕ N (T n ) → R(T n ) ⊕ N (T n ),
0 T2
where T1 = T \ R(T n ) is an invertible operator and T2 is a nilpotent operator. Then
we have " #
n
T 0
Tn = 1 : R(T n ) ⊕ N (T n ) → R(T n ) ⊕ N (T n ).
0 0
Since T n is normal, then T1n is also normal. Therefore, T1 is n-normal. □

The following exemple shows that if T is an n-normal operator such that R(T n )
is closed, then R(T ) need not be closed.

Example 3.2. Le T : ℓ2 (N) →→ ℓ2 (N) be the unilateral weighted shift defined by

x3
T (x1 , x2 , x3 , . . .) = (0, x1 , 0, , 0, . . .).
3
Then T is 2-normal, since T 2 = 0. Hence R(T 2 ) is closed but R(T ) is not closed.

It was proved in Proposition 1.2, (b) that if T is an invertible n-normal operator,

then T −1 is also n-normal. But, what happens if we replace the inverse by the
Moore-Penrose inverse? In order to give the answer to this question, we need the
following lemma.

Lemma 3.3. Let T ∈ B(H) with closed range. If T is a nilpotent operator with
nilpotency 2, then T + is also nilpotent with nilpotency 2.

Proof. Let T = U |T | be the polar decomposition of T , where U is a partial isometry

with R(U ) = R(T ) and N (U ) = N (T ). Since T 2 = 0, R(T ) ⊂ N (T ). It follows that

R(U ) = R(T ) ⊂ N (T ) = N (U ).

Hence, U 2 = 0 and so (U ∗ )2 = 0. On the other hand, T + = |T |+ U ∗ = U ∗ |T + |.

Therefore, (T + )2 = |T |(U ∗ )2 |T + | = 0, which completes the proof. □
12 A. Elgues and S. Menkad

Proposition 3.4. Let T ∈ B(H) with closed range. Then T is 2n-normal if and only
if T + is 2n-normal. In this case (T 2n )+ = (T + )2n .

Proof. By Proposition 1.2 (c), it is suficient to consider the case that T is 2-normal.
According to theorem 2.1, T has the following operator matrix
" #
T1 0
T = : R(T 2 ) ⊕ N (T 2 ) → R(T 2 ) ⊕ N (T 2 ),
0 T2
where T1 is invertible, T12 is normal and T22 = 0. Since R(T ) is closed, we get
" # " #
−1 −2
T1 0 T1 0
T+ = and (T 2 )+ = .
0 T2+ 0 0
Therefore by Lemma 3.3 " #
−2
T 0
(T + )2 = 1 .
0 0
Hence, (T 2 )+ = (T + )2 . and T + is 2-normal, since the normality of T12 implies the
normality of T 2 . Conversely, If T + is 2-normal, then (T + )+ = T is also 2-normal.
Therefore, the proof is complete. □
Proposition 3.4 is not necessarily true if T is (2n+1)-normal. Indeed if we take
the example 2.1, T is 3-normal. Now, by simple calculations we have
   
0 1 0 0 1 −1
T + = 0 −1 1 and (T + )3 = 0 −1 1 
   

0 0 0 0 0 0
But as (T + )3 is not normal, T + is not 3-normal.
The next result provide some equivalent conditions for an operator in B(H) with
closed range to be n-normal. the condition(4) was etablished by Mosić and Djord-
jević for both Moore-penrose invertible and group invertible elements in rings with
involution [18, Lemma 3.1].

Theorem 3.5. Let T ∈ B(H) with closed range and n ∈ N. Then the following
statements are equivalent
(1) T is n-normal.
(2) T n T ∗ T = T ∗ T T n and T n T T ∗ = T T ∗ T n .
(3) T n (T ∗ T )+ = (T ∗ T )+ T n and T n (T T ∗ )+ = (T T ∗ )+ T n .
(4) T n T + = T + T n and (T ∗ )n T + = T + (T ∗ )n .
 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 13

Proof.
(1) ⇒ (2). Since T is n-normal, then T n T ∗ = T ∗ T n . So, we can easily verify that
the statement (2) holds.
(2) ⇒ (3). Assume that (2) holds. Since T ∗ T and T T ∗ are both self-adjoint, then
(T ∗ T )# = (T ∗ T )+ and (T T ∗ )# = (T T ∗ )+ . Hence from Lemma 1.5 (b), we deduce

T n (T ∗ T )+ = (T ∗ T )+ T n and T n (T T ∗ )+ = (T T ∗ )+ T n .

(3) ⇒ (4). First, we show that (T ∗ )n T + = T + (T ∗ )n . Since (T ∗ )+ = T (T ∗ T )+ , the

commutativity T n with (T ∗ T )+ implies that T n (T ∗ )+ = (T ∗ )+ T n . Thus (T ∗ )n T + =
T + (T ∗ )n , by taking an adjoint.
Now, we show that T n T + = T + T n . By Lemma 1.4, we have

T n T + = T n (T ∗ T )+ T ∗
= (T ∗ T )+ T n T ∗
= T + (T ∗ )+ T n T ∗
= T + T n (T ∗ )+ T ∗
= T + T n (T T + )∗
= T +T nT T +.

Moreover, from the hypothesis T n (T T ∗ )+ = (T T ∗ )+ T n and Lemma 1.4, we get

T + T n = T ∗ (T T ∗ )+ T n
= T ∗ T n (T T ∗ )+
= T ∗ T n (T ∗ )+ T +
= T ∗ (T ∗ )+ T n T +
= (T + T )∗ T n T +
= T +T T nT +.

Finally, T n T + = T + T n , since T + T n T T + = T + T T n T + .
(4) ⇒ (1). Assume that (4) holds. Hence T n (T ∗ )+ = (T ∗ )+ T n and T n ((T ∗ )+ )∗ =
((T ∗ )+ )∗ T n . According to Lemma 1.4 (f), we find T n T ∗ = T ∗ T n , since ((T ∗ )+ )+ =
T ∗ . Thus, T is n-normal. □
14 A. Elgues and S. Menkad

Proposition 3.6. Let T ∈ B(H) and n ∈ N. If R(T n ) is closed, then the following
statements are equivalent
(1) T is n-normal.
(2) (T n )+ T = T (T n )+ and (T n )+ T ∗ = T ∗ (T n )+ .
(3) T n is EP and (T n )+ T ∗ = T ∗ (T n )+ .

Proof.
(1) ⇒ (2). Suppose that T is n-normal. Then T n is normal and so (T n )+ =
(T n )# . Since T n T = T T n and T n T ∗ = T ∗ T n , by Lemma 1.4 (f), we deduce
(T n )+ T = T (T n )+ and (T n )+ T ∗ = T ∗ (T n )+ .
(2) ⇒ (3). From the assumption (T n )+ T = T (T n )+ , it follows (T n )+ T n =
T n (T n )+ . Hence, T is EP.
(3) ⇒ (1). Asume that (3) holds. Then (T n )+ is also EP and so

((T n )+ )# = (T n )+ )+ = T n .

Therefore, from the assumption (T n )+ T ∗ = T ∗ (T n )+ and by Lemma 1.5 (b), we

obtain T n T ∗ = T ∗ T n . Thus T is n-normal. □

Remark 3.7. It is well known that every normal operator with closed range is EP.
Hence one might expect that there is a relationship between n-normal operators and
EP operators. But in the example 2.1, T is 3-normal but it is not EP, because
   
0 0 0 1 0 0
T T + = 0 1 0 ̸= 0 1 0 = T + T.
   

0 0 1 0 0 0

Now, we provide a condition under which an n-normal operator becomes EP.

Proposition 3.8. Let T ∈ B(H) be an n-normal operator. If T is group invertible,

then T is EP.

Proof. If T is group invertible, then R(T ) is closed and d(T ) ≤ 1. This implies
R(T ) = R(T n ). On the other hand, since T is n-normal, T n is normal and so
N (T n ) = N ((T n )∗ ). By taking the orthogonal complements in this equality and
since R(T n ) and R((T n )∗ ) are closed, we deduce that

R(T n ) = R((T n )∗ ) ⊂ R(T ∗ ).

 ON THE CLASS OF n-NORMAL OPERATORS AND MOORE-PENROSE INVERSE 15

Hence, R(T ) ⊂ R(T ∗ ) and according to Lemma 1.5 (a), we conclude that T is
EP. □

A CKNOWLEDGMENT

We are very grateful to the referee for the careful reading of the paper.

R EFERENCES
[1] B. A BDELKADER , H. M ORTAD M OHAMMED : Generalizations of Kaplansky’s Theorem In-
volving Unbounded Linear Operators , Bull. Polish Acad. Sci. Math. 62 (2014), 181-186.
[2] S.A. A LZURAIQI , A.B. PATEL: On n-Normal Operators, General Mathematics Notes. 1(2)
(2010), 61-73.
[3] A. B ENALI , M.H. M ORTAD : Generalizations of Kaplansky’s theorem involving unbounded
linear operators , Bull. Pol. Acad. Sci. Math. 62(2) (2014), 181-186.
[4] E. B OASSO, V. R AKOC ĚVIC : Characterizations of EP and normal Banach algebra elements
and Banach space operators , Linear Algebra and its Applications. 5 (2011), 342353.
[5] S.L. C AMPBELL , C.D. M EYER : EP operators and generalized inverses , Canad. Math. Bull.
18 (1975), 327-33.
[6] M. C HO, B. N ACEVSKA : Spectral properties of n-normal operators , Filomat 32(14) (2018),
5063-5069.
[7] M. D ANA , R. Y OUSEF : Generalizations of some classical theorems to D-normal operators on
Hilbert spaces , Journal of Inequalities and Appl. 101 (2020), 1-9.
[8] E. D EUTSCH , P.M. G IBSON , H. S CHNEIDER : The Fuglede-Putnam theorem and normal
products of matrices , Linear Algebra Appl. 1 (1976) 52-58.
[9] D.S. D JORDJEVIC : Characterizations of normal, hyponormal and EP operators , J. Math.
Appl. 329(2) (2007), 1181-1190.
[10] T. F URUTA : On the polar decomposition of an operator , Acta Sci. Math. 46 (1983), 261-268.
[11] H. G. H EUSER : Functional Analysis , John Wiley and Sons, 1982.
[12] A.A.S. J IBRIL : On n-power normal operators , The Arabian Journal for Science and Engi-
neering, 33 (2008), 247-251.
[13] I. K APLANSKY : Products of normal operators , Duke Math. J. 20(2) (1953), 257-260.
[14] C. K ING : A note on Drazin inverses , Pacific J. Math. 70 (1977), 383-390.
[15] J.J. KOLIHA , D RAGAN D JORDJEVI ’ C AND D RAGANA C VETKOVI ’ C : Moore-Penrose inverse
in rings with involution , Linear Algebra and its Applications 33 (2007), 371-381.
[16] D.C. L AY : Spectral analysis using ascent, descent, nullity and defect , Math. Annalen. 184
(1970), 197-214.
[17] D. M OSI Ć , D.S. D JORDJEVI Ć : Moore-Penrose-invertible normal and Hermitian elements in
rings, Linear Algebra Appl. 431 (2009), 732-745.
16 A. Elgues and S. Menkad

[18] D. M OSIC , D.S.D JORDJEVIC : New characterizations of EP, generalized normal and gen-
eralized Hermitian elements in rings , Applied Mathematics and Computation 218 (2012),
6702-6710.
[19] C.R. P UTNAM : On normal operators in Hilbert spaces , American. J.Math. 73(2) (1951),
357-362.

L ABORATORY OF MATHEMATICAL TECHNIQUES ( LTM )

D EPARTMENT OF MATHEMATICS
U NIVERSITY OF BATNA 2
BATNA , ALGERIA .
Email address: [email protected]

L ABORATORY OF MATHEMATICAL TECHNIQUES ( LTM )

D EPARTMENT OF MATHEMATICS
U NIVERSITY OF BATNA 2
BATNA , ALGERIA .
Email address: [email protected]