Filomat 32:15 (2018), 5441–5451 Published by Faculty of Sciences and Mathematics,

https://doi.org/10.2298/FIL1815441C University of Niš, Serbia

Available at: http://www.pmf.ni.ac.rs/filomat

Remarks on n-normal Operators

Muneo Chōa , Ji Eun Leeb , Kotaro Tanahashic , Atsushi Uchiyamad

a Department of Mathematics, Kanagawa University, Hiratsuka 259-1293, Japan
b Department of Mathematics and Statistics, Sejong University, Seoul 143-747, Korea
c Department of Mathematics, Tohoku Medical and Pharmaceutical University, Sendai 981-8558, Japan
d Department of Mathematics, Yamagata University, Yamagata 990-8560, Japan

Abstract. Let T be a bounded linear operator on a complex Hilbert space and n, m ∈ N. Then T is said to
be n-normal if T∗ Tn = Tn T∗ and (n, m)-normal if T∗m Tn = Tn T∗m . In this paper, we study several
T properties of
n-normal, (n, m)-normal operators. In particular, we prove that if T is 2-normal with σ(T) (−σ(T)) ⊂ {0},
then T is polarloid. Moreover, we study subscalarity of n-normal operators. Also, we prove that if T is
(n, m)-normal, then T is decomposable and Weyl’s theorem holds for f (T), where f is an analytic function
on σ(T) which is not constant on each of the components of its domain.

1. Introduction and Motivation

Let H be a complex Hilbert space with the inner product h , i and L(H) be the set of all bounded linear
operators on H. An operator T ∈ L(H) is said to be normal if T∗ T = TT∗ , subnormal if there exists a Hilbert
space K containing H and a normal operator N on K such that NH ⊂ H and T = N|H , hyponormal if
T∗ T − TT∗ ≥ 0. An operator T is said to be scalar of order m if it admits a spectral distribution of order m,
i.e., if there is a continuous unital morphism Φ : Cm 0
(C) −→ L(H) such that Φ(z) = T, where z stands for
the identity function on C and Cm 0
(C) for the space of compactly supported functions on C, continuously
differentiable of order m (0 ≤ m ≤ ∞). An operator T is said to be subscalar of order m if it is similar to the
restriction of a scalar operator of order m to an invariant subspace. It is known that subnormal operators
are hyponormal and hyponormal operators are subscalar ([8]).
An operator T ∈ L(H) is said to have the single-valued extension property if for every open subset G of C
and any H-valued analytic function f on G such that (T − λ) f (λ) ≡ 0 on G, we have f (λ) ≡ 0 on G. For an
operator T ∈ L(H) and for x ∈ H, the local resolvent set ρT (x) of T at x is defined as the union of every open
subset G of C on which there is an analytic function f : G −→ H such that (T − λ) f (λ) ≡ x on G. The local
spectrum of T at x is given by σT (x) = C \ ρT (x). We define the local spectral subspace of an operator T ∈ L(H)
by HT (F) = {x ∈ H : σT (x) ⊂ F} for a subset F of C.

2010 Mathematics Subject Classification. Primary 47B15; Secondary 47A15

Keywords. n-normal; polaroid; subscalar
Received: 04 June 2018; Accepted: 16 August 2018
Communicated by Dragan S. Djordjević
This is partially supported by Grant-in-Aid Scientific Research No.15K04910. The second author was supported by Basic Science
Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and
Technology(2016R1A2B4007035).
Email addresses: [email protected] (Muneo Chō), [email protected];[email protected] (Ji Eun Lee),
[email protected] (Kotaro Tanahashi), [email protected] (Atsushi Uchiyama)
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5442

For an operator T ∈ L(H), the quasinilpotent part of T − λ is defined as

1
H0 (T − λ) = {x ∈ H : lim k(T − λ)n xk n = 0}.
n→∞

In general, ker((T − λ)m ) ⊂ H0 (T − λ) and H0 (T − λ) is not closed. However, if λ is an isolated point of σ(T),
then ET ({λ})H = H0 (T − λ) where Z
1
ET ({λ}) = (z − T)−1 dz
2πi ∂D
denotes the Riesz idempotent corresponding to λ with D is a closed disk centered at λ which contains no
other points of σ(T). Hence H0 (T − λ) is closed in this case.
An operator T ∈ L(H) is said to have Dunford’s property (C) if HT (F) is closed for each closed subset
F of C. An operator T ∈ L(H) is said to have the property (β) if for every open subset G of C and every
sequence fn : G → H of H-valued analytic functions such that (T − z) fn (z) converges uniformly to 0 in norm
on compact subsets of G, then fn (z) converges uniformly to 0 in norm on compact subsets of G. It is well
known that Property (β) =⇒ Dunford’s property (C) =⇒ SVEP, and the converse implications do not hold
([7, Proposision 1.2.19]). An operator T ∈ L(H) is said to be decomposable if for every open cover {U, V} of C
there are T-invariant subspaces X and Y such that H = X + Y, σ(T|X ) ⊂ U and σ(T|Y ) ⊂ V. Remark that T
is decomposable if and only if T and T∗ have the property (β) ([7, Theorem 2.5.19]).
An operator T ∈ L(H) is said to be isoloid if every isolated point of σ(T) belongs to the point spectrum of
T. Hence, hyponormal operators are isoloid ([6, Theorem 2]). Of course, there are many classes of operators
weaker than hyponormal which are isoloid. An operator T ∈ L(H) is called polaroid if every isolated point
of σ(T) is a pole of the resolvent of T. An operator T ∈ L(H) is said to be quasinilpotent if σ(T) = {0}.
In [3], S. A. Alzuraiqi and A. B. Patel introduced n-normal operators.

Definition 1.1. For n ∈ N, an operator T ∈ L(H) is said to be n-normal if

T∗ Tn = Tn T∗ . (1)

This definition seems natural. S. A. Alzraiqi and A. B. Patel proved characterizations of 2-normal, 3-normal
and n-normal operators on C2 . Also, they made several examples of n-normal operators and proved that T
is n-normal if and only if Tn is normal. Also, they proved that if T is 2-normal with the following condition
\
σ(T) (−σ(T)) = ∅, (2)

then T is subscalar. If an operator T ∈ L(H) satisfies (2), then T is invertible automatically. Recently, the
authors in [4] have studied spectral properties of an n-normal operator T satisfying the following condition
(3).
\
σ(T) (−σ(T)) ⊂ {0}. (3)

It is a little weaker assumption than this condition (2). We define (n, m)-normality as follows.
Definition 1.2. For n, m ∈ N, an operator T ∈ L(H) is said to be (n, m)-normal if

T∗m Tn = Tn T∗m .

In this paper, we study several properties of n-normal or (m, n)-normal operators. In particular, we prove
that if T is 2-normal with (3), then T is polarloid. We study subscalarity of n-normal operators. Moreover,
we show that if T is (n, m)-normal, then T is decomposable and Weyl’s theorem holds for f (T), where f is
an analytic function on σ(T) which is not constant on each of the components of its domain.
The following proposition is important in this paper.
Proposition 1.3. ([3, Proposition 2.2]) Let T ∈ L(H) and n ∈ N. Then T is n-normal if and only if Tn is normal.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5443

Therefore, we have the following result.

Theorem 1.4. Let T ∈ L(H) be n-normal. Then T has the single-valued extension property.

Proof. Since Tn is normal, it follows that Tn has the single-valued extension property. Hence T has the
single-valued extension property by [1, Theorem 2.40].

2. 2-normal Operators

In this section, we study some properties of 2-normal operators. Let M be a subspace of H. Then M is
said to be a reducing subspace for T if T(M) ⊂ M and T∗ (M) ⊂ M, that is, M is an invariant subspace for T and
T∗ .

Theorem 2.1. ([4]) Let T ∈ L(H) be 2-normal and satisfy (3). Then the following statements hold.
(i) T is isoloid and σ(T) = σa (T).
(ii) If z and w are distinct eigen-values of T and x, y ∈ H are corresponding eigen-vectors, respectively, then h x, yi = 0.
(iii) If z, w are distinct values of σa (T) and {xn }, {yn } are the sequences of unit vectors in H such that (T − z)xn → 0
and (T − w)yn → 0 (n → ∞), then lim h xn , yn i = 0.
n→∞
(iv) If z and w are distinct eigen-values of T, then ker(T − z) ⊥ ker(T − w).
2
(v) If z is a non-zero eigen-value of T, then ker(T − z) = ker(T2 − z2 ) = ker(T∗2 − z ) = ker(T∗ − z) and hence
ker(T − z) is a reducing subspace for T.

In 2012, J. T. Yuan and G. X. Ji ([10, Lemma 5.2]) proved the following Lemma.

Lemma 2.2. Let m be a positive integer, λ be an isolated point of σ(T) and E = ET ({λ}).
(i) Then the following assertions are equivalent.
(a) EH = ker((T − λ)m ).
(b) ker(E) = (T − λ)m H.
Hence λ is a pole of the resolvent of T and the order of λ is not greater than m.
(ii) If λ is a pole of the resolvent of T and the order of λ is m, then the following assertions are equivalent:
(a) E is self-adjoint.
(b) ker((T − λ)m ) ⊂ ker((T − λ)∗m ).
(c) ker((T − λ)m ) = ker((T − λ)∗m ).

Next we show that if T is 2-normal and satisfies (2), then T is polaroid.

Theorem 2.3. Let T ∈ L(H) be 2-normal and satisfy (3), and let λ be an isolated point of spectrum of T. Then λ is
a pole of the resolvent, that is, T is polaroid and the following statements hold.
(i) If λ = 0, then H0 (T) = ker(T2 ) = ker(T∗2 ), ET ({0}) is self-adjoint and the order of 0 is not greater than 2.
(ii) If λ , 0, then H0 (T − λ) = ker(T − λ) = ker((T − λ)∗ ), ET ({λ}) is self-adjoint and the order of λ is 1.

Proof. Let λ be an isolated point of spectrum of T.

(i) Assume that λ = 0. Since σ(T2 ) = {z2 : z ∈ σ(T)}, it follows that 0 is an isolated point of spectrum
1 1
of T2 . We want to prove that H0 (T) = H0 (T2 ). Let x ∈ H0 (T). Then kTn xk n −→ 0 and hence kT2n xk 2n =
  1
1 2 1 1
kT2n xk n −→ 0 and kT2n xk n −→ 0. Hence x ∈ H0 (T2 ). Conversely, let x ∈ H0 (T2 ). Then kT2n xk n −→ 0 and
1
 1
1 2
so kT2n xk 2n = kT2n xk n −→ 0. Since
1 2n
1 1 1
  2n+1   2n+1
kT2n+1 xk 2n+1 ≤ kTkkT2n xk ≤ kTk 2n+1 kT2n xk 2n −→ 0 (n → ∞),

it follows that x ∈ H0 (T). Hence H0 (T) = H0 (T2 ).

 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5444

Let x ∈ ET ({0}) = H0 (T) = H0 (T2 ) = ET2 ({0}). Since T2 is normal, it follows that ET2 ({0}) = ker(T2 ) =
ker(T∗2 ). Hence x ∈ ker(T2 ) and ET ({0}) ⊂ ker(T2 ). Therefore ET ({0}) = ker(T2 ) = ker(T∗2 ) and 0 is a pole of
the resolvent of T and the order of 0 is not greater than 2 by Lemma 2.2.
(ii) Next we assume λ , 0. Then λ2 is an isolated point of σ(T2 ) by [4, Lemma 2.1]. We will prove
1
H0 (T − λ) = H0 (T2 − λ2 ). Let x ∈ H0 (T − λ). Then k(T − λ)n xk n → 0. Therefore we have
1 1 1 1
k(T2 − λ2 )n xk n ≤ k(T + λ)n k n k(T − λ)n xk n ≤ kT + λkk(T − λ)n xk n −→ 0.

Hence H0 (T − λ) ⊂ H0 (T2 − λ2 ). Conversely, let x ∈ H0 (T2 − λ2 ). Since T + λ is invertible by (3), we have

1 1
k(T − λ)n xk n = k(T + λ)−n (T + λ)n (T − λ)n xk n
n on 1 1 1
≤ k (T + λ)−1 k n k(T2 − λ2 )n xk n ≤ k(T + λ)−1 kk(T2 − λ2 )n xk n −→ 0.

Hence H0 (T − λ) ⊃ H0 (T2 − λ2 ) and H0 (T − λ) = H0 (T2 − λ2 ).

Let x ∈ ET ({λ})H. Since ET ({λ})H = H0 (T − λ) = H0 (T2 − λ2 ) and T2 is normal, we have H0 (T2 − λ2 ) =
ET2 ({λ2 }) = ker(T2 − λ2 ). Hence 0 = (T2 − λ2 )x = (T + λ)(T − λ)x. Since T + λ is invertible by (3), we have
(T − λ)x = 0. Hence ET ({λ})H ⊂ ker(T − λ) and ET ({λ})H = ker(T − λ). Hence λ is a pole of the resolvent of
T and the order of λ is 1 by Lemma 2.2. Since ker(T − λ) = ker((T − λ)∗ ) by [4, Theorem 2.6], it follows that
ET ({λ}) is self-adjoint by Lemma 2.2.

Let D be a bounded open subset of C and L2 (D, H) be the Hilbert space of measurable function f : D −→ H
such that
Z  12
k f k2,D = k f (z)k2 dµ(z) < ∞,
D

where dµ is the planar Lebesgue measure. Let W 2 (D, H) be the Sobolev space with respect to ∂ and of order
2
2 whose derivatives ∂ f and ∂ f in the sense of distributions belong to L2 (D, H). The norm k f kW2 is given by
 2
 12
k f kW2 = k f k22,D + k∂ f k22,D + k∂ f k22,D for f ∈ W 2 (D, H).

Then in [3], S. A. Alzuraiqi and A. B. Patel proved the following result.

PropositionT 2.4. ([3, Theorem 2.37]) Let D be an arbitrary bounded disk in C. If T ∈ B(H) is 2-normal with (1),
that is, σ(T) (−σ(T)) = ∅, then the operator

S = z − T : W 2 (D, H) −→ W 2 (D, H)

is one to one.

We will revise this result as follows.

Theorem 2.5. Let

T D be an arbitrary bounded open disk in C. If T ∈ B(H) is 2-normal and the planar Lebesgue
measure of σ(T) (−σ(T)) is 0, then the operator

S = z − T : W 2 (D, H) −→ W 2 (D, H)

is one to one.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5445

Proof. Let f ∈ W 2 (D, H) and S f = 0. We show f = 0. Then

2
k f k2W2 = k f k22,D + k∂ f k22,D + k∂ f k22,D
Z Z Z
2
= k f (z)k dµ(z) +
2
k∂ f (z)k dµ(z) +
2
k∂ f (z)k2 dµ(z) < ∞,
D D D

and
2
kS f k2W2 = k(z − T) f k2W2 = k(z − T) f k22,D + k∂((z − T) f )k22,D + k∂ ((z − T) f )k22,D
2
= k(z − T) f k22,D + k(z − T)∂ f k22,D + k(z − T)∂ f k22,D = 0.

Hence Z
i i
k(z − T) ∂ f k22,D = k(z − T) ∂ f (z)k2 dµ(z) = 0 (i = 0, 1, 2).
D
i i
Let i be i = 0, 1, 2. Since (z − T) ∂ f (z) = 0 for z ∈ D, if z ∈ D \ σ(T), then ∂ f (z) = 0 because z − T is invertible.
This implies Z
i i
k(z − T)∗ ∂ f k22,D\σ(T) = k(z − T)∗ ∂ f (z)k2 dµ(z) = 0.
D\σ(T)

Since
Z
i i
2 2
k(z − T )∂ f k22,D = k(z2 − T2 )∂ f (z)k2 dµ(z)
D
!2 Z !2
i i
≤ sup kz + Tk k(z − T)∂ f (z)k2 dµ(z) = sup kz + Tk k(z − T)∂ f k22,D = 0,
z∈D D z∈D

i
we have (z2 − T2 ) ∂ f (z) = 0 for z ∈ D. Moreover, since T2 is normal, this implies
Z
i i
k(z2 − T2 )∗ ∂ f k22,D = k(z2 − T2 )∗ ∂ f (z)k2 dµ(z) = 0.
D

Hence
i i
0 = (z2 − T2 )∗ ∂ f (z) = (z + T)∗ (z − T)∗ ∂ f (z) for z ∈ D.
i
If z ∈ D ∩ (σ(T) \ (−σ(T))), then z + T and (z + T)∗ are invertible. Hence we obtain (z − T)∗ ∂ f (z) = 0 for
i
z ∈ D∩(σ(T) \ (−σ(T))). Since D is bounded, k∂ f k22,D < ∞ and the planar Lebesgue measure of σ(T) (−σ(T))
T
is 0, we have
Z
i i
k(z − T)∗ ∂ f k22,D = k(z − T)∗ ∂ f (z)k2 dµ(z)
D\σ(T)
Z Z
i i
+ k(z − T) ∂ f (z)k dµ(z) +
∗ 2
k(z − T)∗ ∂ f (z)k2 dµ(z)
D∩(σ(T)\(−σ(T))) D∩σ(T)∩(−σ(T))
Z
i
≤ 0 + 0 + max k(z − T)∗ k2 k∂ f (z)k2 dµ(z) = 0.
z∈D D∩σ(T)∩(−σ(T))

By [8, Proposition 2.1], we obtain k(I − P) f k2,D = 0. Hence f (z) = (P f )(z) for z ∈ D. Since S f = 0, we have
(S f )(z) = (z − T) f (z) = (z − T)(P f )(z) = 0 for z ∈ D. Since T has the single-valued extension property by
Theorem 1.4 and P f is analytic, it follows that 0 = (P f )(z) = f (z) for z ∈ D. Hence f = 0 and S is one to
one.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5446

3. n-normal Operators

Theorem 3.1. Let T ∈ L(H) be n-normal. Let σ(T) be contained in an angle < 2π/n with vertex in the origin, i.e.,
there exists θ1 ∈ [0, 2π) such that

2π
 
σ(T) ⊂ W = reiθ : 0 < r, θ1 < θ < θ1 + .
n

Then T is subscalar of order 2.

Proof. Let D be an open bounded disk such that σ(T) ⊂ D. Take an open set U such that σ(T) ⊂ U ⊂
U ⊂ D ∩ W. Let M : W 2 (D, H) → W 2 (D, H) be a multiplication operator such that (M f )(z) = z f (z) for
f ∈ W 2 (D, H) and z ∈ D. Then M is scalar of order 2 with a spectral distribution defined by Φ(φ) f = φ f
for φ ∈ C20 (C) and f ∈ W 2 (D, H). Since (z − T)W 2 (D, H) is M-invariant, it follows that S : H(D) =
W 2 (D, H)/(z − T)W 2 (D, H) → H(D) as
 
S f + (z − T)W 2 (D, H) → M f + (z − T)W 2 (D, H)

for f ∈ W 2 (D, H) is well defined and S is still scalar of order 2 with a spectral distribution
 
Φ̃(φ) f + (z − T)W 2 (D, H) = φ f + (z − T)W 2 (D, H)

for φ ∈ C20 (C) and f + (z − T)W 2 (D, H) ∈ H(D). Let V : H → H(D) be as

Vh = 1 ⊗ h + (z − T)W 2 (D, H)

for h ∈ H where (1 ⊗ h)(z) = h for z ∈ D. Then

VT = SV.

We prove that V is one to one and has dense range. Then VH is an invariant subspace of S and T = S|VH .
Hence T is subscalar of order 2.

Claim. If Vhn → 0, then hn → 0.

Let Vhn → 0. Then there exists fn ∈ W 2 (D, H) such that

2
k(z − T) fn + 1 ⊗ hn k2W2 = k(z − T) fn + 1 ⊗ hn k22,D + k(z − T)∂ f k22,D + k(z − T)∂ f k22,D → 0.

Let ζ = exp(2πi/n). Then

Z Z n 2
i i Y i
n n
k(z − T )∂ f k22,D = k(z − T )∂ fn (z)k dµ(z) =
n n 2
(ζk z − T)∂ fn (z) dµ(z)
D D k=1
n−1 2 Z
Y i
k
≤ sup (ζ z − T) k(z − T)∂ fn (z)k2 dµ(z) → 0.
z∈D k=1 D

Since Tn is normal, we have

i
k(zn − Tn )∗ ∂ f k22,D → 0.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5447

If z ∈ U, then ζk z < σ(T) for k = 1, 2, · · · , n − 1 by the assumption. Hence

Z
i i
k(z − T) ∂ fn k =
∗ 2
k(z − T)∗ ∂ fn (z)k2 dµ(z)
2,U
U
Z Y n−1 
 n−1 
∗ Y  i
= (ζk z − T)−1  (ζk z − T)∗  (z − T)∗ ∂ fn (z)dµ(z)
U k=1 k=1
n−1 2
Y  ∗ 2 i
≤ sup (ζk z − T)−1 (zn − Tn )∗ ∂ fn → 0.
k=1 z∈U 2,D

Since
Z Z
i i
k∂ fn (z)k2 dµ(z) = k(z − T)−1 (z − T)∂ fn (z)k2 dµ(z)
D\U D\U
Z
i
−1 2
≤ sup k(z − T) k k(z − T)∂ fn (z)k2 dµ(z)
z∈D\U D\U
i
≤ sup k(z − T)−1 k2 k(z − T)∂ fn k22,D → 0,
z∈D\U

we have
Z Z
i i i
k(z − T) ∂
∗
fn k22,D = k(z − T) ∂ fn (z)k dµ(z) +
∗ 2
k(z − T)∗ ∂ fn (z)k2 dµ(z)
D\U U
Z
i i
≤ sup k(z − T)∗ k2 k∂ fn (z)k2 dµ(z) + k(z − T)∗ ∂ fn k2 → 0.
2,U
z∈D\U D\U

Let P be the orthogonal projection of L2 (D, H) onto A2 (D, H). Then there exists a constant 0 < CD such that
 2

k(1 − P) fn k2,D ≤ CD k(z − T)∗ ∂ fn k2,D + k(z − T)∗ ∂ fn k2,D → 0

by Proposition 2.1 of [8]. Hence

k(z − T)P fn + 1 ⊗ hn k2,D ≤ k(z − T) fn + 1 ⊗ hn k2,D + k(z − T)(1 − P) fn k2,D

≤ k(z − T) fn + 1 ⊗ hn k2,D + sup kz − Tkk(1 − P) fn k2,D → 0.
z∈D

Hence
k(z − T)P fn + 1 ⊗ hn k∞,U = sup k(z − T)P fn (z) + hn k → 0
z∈U

by [8, Lemma 1.1]. Define Ψ : A (U, H) → H as 2

Z
1
Ψ(1) = (z − T)−1 1(z)dz
2πi ∂G

for 1 ∈ A2 (U, H) where G is an open set such that σ(T) ⊂ G ⊂ G ⊂ U and ∂G is a Jordan curve. Since

1
kΨ(1)k ≤ max k(z − T)−1 kk1k∞,U `(∂G)
2π z∈∂G

for 1 ∈ A2 (U, H) where `(∂G) denotes the length of ∂G and

(z − T)P fn + 1 ⊗ hn ∈ A2 (U, H),

 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5448

we have
Z
1
Ψ((z − T)P fn + 1 ⊗ hn ) = (z − T)−1 (z − T)P fn (z) + hn dz


2πi ∂G
Z 
1 
= P fn (z) + (z − T)−1 hn dz = 0 + hn → 0.
2πi ∂G

Corollary 3.2. Under the same hypothesis as in Theorem 3.1, if σ(T) has nonempty interior, then T has a nontrivial
invariant subspace.

Proof. By the hypothesis, T is subscalar of order 2 from Theorem 3.1. Since σ(T) has nonempty interior, we
get this result from [5, Theorem 2.1].

In [9], C.R. Putnam proved that if T is hyponormal, then

πkT∗ T − TT∗ k ≤ m(σ(T))

where m is the Lebesque measure in the complex plane. This is well known as Putnam’s inequality.

Lemma 3.3. Let T ∈ L(H) be n-normal and let M ⊂ H be an invariant subspace for T. Then the following
assertions hold.
(i) (T|M )n is subnormal.
(ii) Let σ(T|M ) = {λ}. Then T|M = λ if λ , 0 and (T|M )n = 0 if λ = 0.

Proof. (i) Since (T|M )n = Tn |M and Tn is normal, (T|M )n is subnormal.

(ii) Suppose λ = 0. Then (T|M )n is subnormal by (1) and

σ ((T|M )n ) = {zn |z ∈ σ(T|M )} = {0}.

It follows that (T|M )n = 0 by Putnam’s inequality.

Suppose λ , 0. Then (T|M )n is subnormal and σ ((T|M )n ) = {λn }. It follows that (T|M )n = λn by Putnam’s
inequality. Since σ(T|M ) = {λ} and
 n−1 
Y 
0 = (T|M )n − λn =  (T|M − λζk ) (T|M − λ),
k=1

we have
 n−1 −1
Y 
T|M − λ =  (T|M − λζ ) · 0 = 0,
 k

k=1

where ζ = exp(2πi/n).

Definition 3.4. Let λ ∈ σ(T) be arbitrary, n ∈ N and ζ := exp(2πi/n). We say that T has property (n) at λ if

λζk < σ(T) for k = 1, · · · , n − 1.

Remark. We do not need the assumption that λ is an isolated point of σ(T) in the following theorem.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5449

Theorem 3.5. Let T ∈ L(H) be n-normal. Then the following assertions hold.
(i) H0 (T) = H0 (Tn ) = ker(Tn ) = ker(T∗n ).
(ii-1) If λ , 0, then H0 (T − λ) = ker(T − λ).
(ii-2) If λ , 0 and T has property (n) at λ, then H0 (T − λ) = ker(T − λ) = ker((T − λ)∗ ).

Proof. (i) Since Tn is normal, we have H0 (T) ⊂ H0 (Tn ) = ker(Tn ) = ker(T∗n ). It is known that ker(Tn ) ⊂ H0 (T).
Hence H0 (T) = H0 (Tn ) = ker(Tn ) = ker(T∗n ).
(ii-1) We claim H0 (T − λ) ⊂ H0 (Tn − λn ).
Let x ∈ H0 (T − λ) and ζ = exp(2πi/n). Then
1 1
k(Tn − λn )m xk m = k(T − λζ)m (T − λζ2 )m · · · (T − λζm−1 )m (T − λ)m xk m
1
≤ kT − λζkkT − λζ2 k · · · kT − λζm−1 kk(T − λ)m xk m −→ 0 (m → ∞).

Hence x ∈ H0 (Tn − λn ).
Since Tn is normal, H0 (Tn − λn ) = ker(Tn − λn ). Put M = ker(Tn − λn ). Then M is an invariant subspace
of T and σ ((T|M )n ) = σ(Tn |M ) = {λn }. Hence σ(T|M ) ⊂ {λ, λζ, · · · , λζn−1 }. Put σ(T|M ) = {µ1 , · · · , µr } with
µi , µ j (i , j) and µni = λn for i = 1, 2, · · · , r. Let Fi be the Riesz idempotent corresponding to µi ∈ σ(T|M ).
 
Then Fi F j = 0 (i , j), F1 + · · · + Fr = IM , σ (T|M )|Fi M = σ T|Fi M = {µi } and σ T|(IM −Fi )M = σ(T|M ) \ {µi } for i =

1, 2, · · · , r. This shows that T|Fi M = µi for i = 1, 2, · · · , r by Lemma 3.3. Put C = (kF1 k + kF2 k + · · · + kFr k)−1 > 0.
Since

kxk = kF1 x + F2 x + · · · + Fr xk ≤ kF1 xk + kF2 xk + · · · + kFr xk

≤ (kF1 k + kF2 k + · · · + kFr k) kxk,

we have
kxk ≥ C(kF1 xk + kF2 xk + · · · + kFr xk) for all x ∈ M.
Let 0 , x ∈ H0 (T − λ) ⊂ M. Then
 r  n1
1  X1 
k(T − λ)n xk = k(T|M − λ)n xk ≥ C
n nkFk (T|M − λ)n xk
k=1
 r  n1  r  n1
 X   X 
= C k(T|M − λ)n Fk xk = C k(T|Fk M − λ)n Fk xk
k=1 k=1
 r  n1
1 
X  1 1
= C n  |µk − λ| kFk xk ≥ |µk − λ|C n kFk xk n .
n


k=1

By letting n → ∞, it follows that Fk x = 0 for all k such as µk , λ. Hence if there does not exist k such that
µk = λ, then x = F1 x + F2 x + · · · + Fr x = 0 which is a contradiction. Hence there exists a unique number
k0 ∈ {1, · · · , r} such that µk0 = λ and Fk0 x = x. Hence x ∈ Fk0 M = ker(T|Fk0 M − λ) ⊂ ker(T − λ). Hence
H0 (T − λ) ⊂ ker(T − λ). Since the converse inclusion is clear, we have H0 (T − λ) = ker(T − λ).
(ii-2) Let T have property (n) at λ. Since Tn is normal, we have

H0 (T − λ) = ker(T − λ) ⊂ H0 (Tn − λn ) = ker(Tn − λn ) = ker((Tn − λn )∗ ).

Conversely, let y ∈ H0 (Tn − λn ) = ker(Tn − λn ) = ker((Tn − λn )∗ ). Then (Tn − λn )y = 0 and (Tn − λn )∗ y = 0.

Since λζk < σ(T) for k = 1, · · · , n − 1, it follows that
 n−1 −1
Y 
(T − λ)y =  (T|M − λζ ) (Tn − λn )y = 0
 k

k=1
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5450

and  n−1 −1

Y 
(T − λ) y =  (T|M − λζ )  (Tn − λn )∗ y = 0.
∗  k ∗

k=1

Hence H0 (T−λ) = ker(T−λ) = ker(T −λ ) = ker((Tn −λn )∗ ) ⊂ ker((T−λ)∗ ). Since ker((T−λ)∗ ) ⊂ ker((Tn −λn )∗ )
n n

is clear, we have

H0 (T − λ) = ker(T − λ) = ker((T − λ)∗ ) = ker(Tn − λn ) = ker((Tn − λn )∗ ).

Theorem 3.6. Let T ∈ L(H) be n-normal. Then T is isoloid and polaroid.

Moreover, let λ be an isolated point of the spectrum of T. Then λ is a pole of the resolvent and following statements
hold.
(i) If λ = 0, then ET ({0})H = H0 (T) = H0 (Tn ) = ker(Tn ) = ker(T∗n ), ET ({0}) is self-adjoint and the order of 0 is not
greater than n.
(ii) If λ , 0, then ET ({λ})H = H0 (T − λ) = ker(T − λ) and the order of λ is 1.

Proof. (i) Assume that 0 is an isolated point of σ(T). Since H0 (T) = H0 (Tn ) = ker(Tn ) = ker(T∗n ) by Theorem
3.5, we have ET ({0})H = H0 (T) = H0 (Tn ) = ker(Tn ) = ker(T∗n ). Hence 0 is a pole of the resolvent of T, ET ({0})
is self-adjoint and the order of pole is not greater than n by Lemma 2.2.
(ii) Next we assume λ is a nonzero isolated point of σ(T). Since H0 (T − λ) = ker(T − λ) by Theorem 3.5, we
have ET ({λ})H = H0 (T − λ) = ker(T − λ). Hence λ is a pole of the resolvent of T and the order of pole is 1
by Lemma 2.2.

4. (n, m)-normal Operators

Definition 4.1. For n, m ∈ N, an operator T ∈ L(H) is said to be (n, m)-normal if

T∗m Tn = Tn T∗m .

From the definition, it is clear that T is (n, m)-normal if and only if T∗ is (m, n)-normal. Moreover, if Tn is
normal, then T is (n, m)-normal for every m. Indeed, since Tn is normal and Tm · Tn = Tn · Tm , it follows from
Fuglede theorem that T∗m · Tn = Tn · T∗m . Hence T is (n, m)-normal. From [4], we restate the properties of
(m, n)-normal operators.
Lemma 4.2. Let T ∈ L(H) be (n, m)-normal. Then the following statements hold.
(i) T∗ is (m, n)-normal.
(ii) If T−1 exists, then T−1 is (n, m)-normal.
(iii) If S ∈ L(H) is unitary equivalent to T, then S is (n, m)-normal.
(iv) If M is a closed subspace of H which reduces T, then T|M is (n, m)-normal on M.
(v) If T is (n, m)-normal, then Tk is normal where k is the least common multiple of n and m.
(vi) If T is quasi-nilpotent, then T is nilpotent.
Proof. The proofs of the statements of (i), (ii), (iii), and (iv) are clearly holds by the definition.
(v) Let k := n · j and k := m · `. Since T is (n, m)-normal, it follows that

` j
z }| { z }| {
T∗k Tk = T∗m · · · T∗m · Tn · · · Tn = Tn · · · Tn · T∗m · · · T∗m = Tk T∗k ,

which means that Tk is normal.

(vi) If T is quasi-nilpotent, i.e., σ(T) = {0}, then σ(Tk ) = {0} for every k ∈ N. Let k0 be the least common
multiple of n and m. Then Tk0 is normal by Lemma 4.2 (v). Hence Tk0 = 0.
 M. Chō et al. / Filomat 32:15 (2018), 5441–5451 5451

Corollary 4.3. Let T ∈ L(H) be (n, m)-normal. Then T is isoloid and polaroid.
Moreover, let λ be an isolated point of the spectrum of T. Then λ is a pole of the resolvent and following statements
hold.
(i) If λ = 0, then H0 (T) = ET ({0})H = ker(Tnm ) = ker(T∗nm ), ET ({0}) is self-adjoint and the order of 0 is not greater
than n.
(ii) If λ , 0, then H0 (T − λ) = ET ({λ})H = ker(T − λ) and the order of λ is 1.
Proof. Since Tnm is normal by Lemma 4.2, we have these results from Theorem 3.6.
We say that Weyl’s theorem holds for T if

σ(T) \ π00 (T) = σw (T), or equivalently, σ(T) \ σw (T) = π00 (T),

where σw (T) = {λ ∈ C : T − λ is not Weyl}, π00 (T) = {λ ∈ iso (σ(T)) : 0 < dim ker(T − λ) < ∞}, and iso (σ(T))
denotes the set of all isolated points of σ(T).

Theorem 4.4. Let T ∈ L(H) be (n, m)-normal. Then the following statements hold.
(i) T is decomposable.
(ii) If f is an analytic function on σ(T) which is not constant on each of the components of its domain, then Weyl’s
theorem holds for f (T).

Proof. (i) Since Tnm , T∗nm are normal by Lemma 4.2, it follows Tnm is decomposable. Hence T is decomposable
by [7, Theorem 3.3.9].
(ii) Since T is polaroid by Theorem 3.6 or Corollary 4.3 and T has the single-valued extension property
by Theorem 1.4, it follows that Weyl’s theorem holds for f (T) by [2, Theorem 3.14].

References
[1] P. Aiena, Fredholm and local spectral theory with applications to multipliers, Kluwer Academic Publishers, Dordrecht, 2004.
[2] P. Aiena, E. Aponte and E. Balzan E, Weyl type theorems for left and right polaroid operators, Integ. Eq. Op. Th. 66, (2010), 1–20.
[3] S. A. Alzuraiqi and A. B. Patel, On n-normal operators, General Math. Notes 1, (2010), 61–73.
[4] M. Chō and B. Načevska Nastovska, Spectral properties of n-normal operators, to appear in Filomat.
[5] J. Eschmeier, Invariant subspaces for operators with Bishop’s property (β) and thick spectrum, J. Funct. Anal. 94, (1990), 196–222.
[6] J. G. Stampfli, Hyponormal operators, Pacific J. Math. 12, (1962), 1453–1458.
[7] K. Laursen and M. Neumann, An introduction to local spectral theory, Clarendon Press, Oxford, 2000.
[8] M. Putinar, Hyponormal operators are subscalar, J. Op. Th. 12, (1984), 385–395.
[9] C. R. Putnam, An inequality for the area of hyponormal spectra, Math. Z. 116, (1970), 323–330.
[10] J. T. Yuan and G. X. Ji, On (n, k)-quasiparanormal operators, Studia Math. 209, (2012), 289–301.