第 55 卷 第 6 期
2020 年 12 月
JOURNAL OF SOUTHWEST JIAOTONG UNIVERSITY
Vol. 55 No. 6
Dec. 2020
ISSN: 0258-2724 DOI:10.35741/issn.0258-2724.55.6.31
Research articleMathematics
N
EW
D
ECOMPOSITIONS FOR
C
LASSES OF
O
PERATORS WITH
T
OPOLOGICAL
U
NIFORM
D
ESCENT
混合資源和相互作用模式對偏遠地區可再生能源管理的影響
Orlando García a, *, Carlos Carpintero a, José Sanabria b, Osmin Ferrer b
a Corporación Universitaria del Caribe-CECAR, Departamento de Ciencias Básicas, Sincelejo, Colombia.
ogarciam554@gmail.com, carpintero.carlos@gmail.com
b
Universidad de Sucre, Departamento de Matemáticas, Sincelejo, Colombia, jesanabri@gmail.com,
osmin.ferrer@unisucre.edu.co
Received: June 13, 2020 ▪ Review: September 8, 2020 ▪ Accepted: October 9, 2020
This article is an open-access article distributed under the terms and conditions of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0)
Abstract
The article describes a new decomposition property for operators with topological uniform descent, like Kato type operators, as well as new results on the stability of this class of operators under perturbations by operators with finite-range power based on topological descent notion, from which we can generalize many perturbation results for a large classes of operators by extending to Banach spaces known techniques on Hilbert spaces. As application of our resuts we obtain that 𝑇 is a lower semi B-Weyl
operator if and only if 𝑇 = 𝑆 + 𝐾 , where 𝑆 is a lower semi B-Browder operator and dim 𝐾𝑛(𝑋) < ∞, for
some 𝑛 ∈ ℕ. Our methods generalize to Banach spaces some results obtained by Aiena for operators acting on Hilbert spaces.
Keywords: Quasi-Fredholm Operator, Operators with Topological Uniform Descent, Operators with
Finite-range Power. 摘要 本文介紹了具有拓撲均勻下降的算子(如加藤型算子)的新分解性質,以及此類算子在拓撲 下降概念的作用下,具有有限範圍冪的算子在擾動下的穩定性的新結果。通過將希爾伯特空間上 的已知技術擴展到巴纳赫空間,可以將一類大型算子的許多攝動結果推廣化。作為我們的結果的 應用,當且僅當 T =小号+ K 時,T 是下半魏尔算符,其中小号是下半乙浏览器算子且昏暗和〖K ^ n(X)<∞} ,對於某些 n∈N。我們的方法推廣到巴纳赫空間,艾埃娜為作用於希爾伯特空間 的算子提供了一些結果. 关键词: 擬弗雷德霍尔姆算子,具有拓撲均勻下降的算子,有限範圍冪的算子
I. I
NTRODUCTIONIn the sequel, 𝑋 will be denote an infnite
dimensional normed space and𝐿(𝑋) denote the
algebra of all bounded linear operators acting on 𝑋 . For 𝑇 ∈ 𝐿(𝑋) , we denote by 𝑁(𝑇) the null space of 𝑇, 𝑅(𝑇) = 𝑇(𝑋) the range of 𝑇,𝛼(𝑇) = 𝑑𝑖𝑚 𝑁(𝑇) the nullity of 𝑇 and by 𝛽(𝑇) = 𝑐𝑜𝑑𝑖𝑚 𝑅(𝑇) = dim 𝑋 = 𝑅(𝑇) the defect of T. Given 𝑛 ∈ ℕ , we denote by 𝑇𝑛 the restriction
of𝑇 ∈ 𝐿(𝑋) onthe subspace 𝑅(𝑇𝑛) = 𝑇𝑛(𝑋).
The principal research method used in this work is the hypothetical-deductive method. In this work we employed a particular notion of topological descent, called topological descent
uniform (abbreviated TDU), to find a
descomposition property (similar to type Kato descomposition) for certain clases of operator. By using this descomposition, we stablished and proved some results which provide a moregeneral framework to able to extent results given in Hilbert spaces to a Banach spaces and we obtain interesting applications of these results for perturbation of some clases of operators.
The main results of our research are the following. We give new results on the stability of operators with topological finite descent under perturbations by operators with finite-range power and we characterized the class of semi B-Weyl operator as the following form: 𝑇 is a lower semi B-Weyl operator if and only if 𝑇 = 𝑆 + 𝐾 where 𝐾 is a lower semi B-Browder operator and
dim 𝐾𝑛(𝑋) < ∞, for some 𝑛 ∈ ℕ.
The results of our research can be potentially important for mathematicians who working in áreas such that operators theory, perturbation theory and others afine areas.
This generalizes the corresponding result of Aiena for operators on Hilbert spaces. Although the above theorems have been showed under the hypothesis that X is a Banach space ([1, Theorem 12.7] and [1, Theorem 21.3], respectively). We can see that the conclusions of these theorems remain true if the hypothesis Banach is replaced by X is a normed space. The following Theorem was proved in [1, Theorem 21.9] under the hypothesis that S compact or quasinilpotent operator. However, we can observe that this conclusion remain true if we suppose that the operator S has a finite-range power
The following Lemma was proved in [2] under the hypothesis that T is a quasiFredholm operator. However, we can observe that this conclusion remain true if we suppose that the operator T has TUD.
In the following theorem, we construct a decomposition property for operators that have TUD, similar to the Kato type operator.
As in Lemma 2.8, the condition T is a quasi-Fredholm operator can be replaced by the hypothesis T has TUD, of Theorem [2, Theorem 2.8], and the conclusion of its remains true.
The following theorem extends the results obtained in [3, Theoren 2.8] to a larger class of operators.
The following theorem was established in [4] for the particular case of Hilbert spaces. Now, we extend this to the context of Banach spaces.
II. P
RELIMINARIESDefinition 1.1: Let 𝑋 be a normed space and 𝑇 ∈ 𝐿(𝑋), then
(i)𝑇 is said to be semi-regular if 𝑅(𝑇) is
closed and 𝑁(𝑇) ⊆ 𝑅(𝑇𝑛) for every
𝑛 ∈ ℕ.
(ii)𝑇 is said to be essentially semi-regular
if 𝑅(𝑇) is closed and there exists a finite
dimensional subspace 𝐹 such that 𝑁(𝑇) ⊆
𝑅(𝑇𝑛) + 𝐹 for every𝑛 ∈ ℕ.
Theorem 1.2: Let 𝑋 be a normed space and
𝑇 ∈ 𝐿(𝑋) . The following conditionsare
equivalent:
(i)𝑇 is semi-regular,
(ii)𝑇𝑛 is semi-regular for all 𝑛 ∈ ℕ ,
(iii)𝑇𝑛 is semi-regular for some 𝑛 ∈ ℕ. Definition 1.3: Let 𝑋be a normed space. An operator 𝑇 ∈ 𝐿(𝑋) is said to beof Kato type if there exists a pair of 𝑇-invariant closed subspaces
(𝑀, 𝑁) suchthat 𝑋 = 𝑀 ⊕ 𝑁, the restriction 𝑇⎸𝑀
is semi regular and 𝑇⎸𝑁 is nilpotent.
Theorem 1.4: Let 𝑋 be a normed space. If 𝑇 ∈ 𝐿(𝑋) is essentially semi-regular then there
exist a decomposition 𝑋 = 𝑋2⊕ 𝑋2 with the
properties that 𝑇(𝑋1) ⊆ 𝑋1 , 𝑇(𝑋2) ⊆ 𝑋2 ,
dim 𝑋1< ∞ , 𝑇⎸𝑋1 is nilpotent and 𝑇⎸𝑋2 is semiregular.
Although the above theorems have been showed under the hypothesis that𝑋 is a Banach space ([1,
Theorem 12.7] and [1, Theorem 21.3],
respectively).We can see that the conclusions of
these theorems remain true if the
hypothesisBanach is replaced by 𝑋 is a normed space.The following Theorem was proved in [1,
Theorem 21.9] under the hypothesisthat
𝑆 compact or quasinilpotent operator. However, we can observe that thisconclusion remain true if we suppose that the operator S has a finite-range power.
Theorem 1.5: Let 𝑋 be a Banach space and
let 𝑇 ∈ 𝐿(𝑋) such that T isessentially
semi-regular, dim 𝑆𝑛(𝑋) for some 𝑛 ∈ ℕ and
𝑇𝑆 = 𝑆𝑇, then
III. Q
UASIF
REDHOLMO
PERATORS UNDERP
ERTURBATIONS BYO
PERATORS WITHF
INITE-R
ANGEP
OWERFor 𝑇 ∈ 𝐿(𝑋) is defined
𝜅𝑛(𝑇) =
𝑑𝑖𝑚 (𝑅(𝑇𝑛) ∩ 𝑁(𝑇))
𝑑𝑖𝑚 (𝑅(𝑇𝑛+1) ∩ 𝑁(𝑇))
In virtue of [1, Corollary 12.23], we have
𝜅𝑛(𝑇) =
𝑑𝑖𝑚 (𝑅(𝑇) + 𝑁(𝑇𝑛))
𝑑𝑖𝑚 (𝑅(𝑇) + 𝑁(𝑇𝑛+1))
Defnition 2.1: Let 𝑇 ∈ 𝐿(𝑋) . If there is
𝑑 ∈ ℕsuch that 𝜅𝑛(𝑇) = 0for𝑛 ≥ 𝑑 , then 𝑇 is
said to have uniform descent for 𝑛 ≥ 𝑑 . If in addition,𝑅(𝑇𝑛) is closed in the operator range topology of 𝑅(𝑇𝑑)for 𝑛 ≥ 𝑑 , then we say that 𝑇 has topological uniform descent (TUD for brevity) for 𝑛 ≥ 𝑑.
Definition 2.2: Let 𝑋 be a Banach space. An
operator 𝑇 ∈ 𝐿(𝑋) is said to bequasi-Fredholm
of degree 𝑑, and is denoted 𝑇 ∈ 𝑄𝐹(𝑑), if there
exists 𝑑 ∈ ℕsuchthat 𝜅𝑛(𝑇) = 0 for each 𝑛 ≥ 𝑑
and 𝑅(𝑇𝑑+1) is closed.
An operator is Fredholm if it is quasi-Fredholm of some degree 𝑑.
Theorem 2.3: [1, Lemma 22.17] Let 𝑋 be a
Banach space and 𝑇 ∈ 𝐿(𝑋) . If 𝑑 ∈ ℕ and
𝜅𝑛(𝑇) < ∞ for each 𝑛 ≥ 𝑑 , then following
statements are equivalent:
(i)There exists 𝑛 ≥ 𝑑 + 1 such that 𝑅(𝑇𝑛) is closed,
(ii)𝑅(𝑇𝑑+1) is closed,
(iii)𝑅(𝑇𝑛) is closed for each 𝑛 ≥ 𝑑,
(iv)𝑅(𝑇𝑖) + 𝑁(𝑇𝑗) is closed for all 𝑖, 𝑗 with 𝑖 + 𝑗 ≥ 𝑑.
From this Theorem we see easily that if 𝑇 is a quasi-Fredholm operator of somedegree 𝑑 then 𝑇
has topological uniform descent for 𝑛 ≥ 𝑑. But
the converse isnot always true as shown in [5]. Theorem 2.4: [6, Theorem 3.2] Let 𝑋 be a
Banach space. If 𝑇 is a boundedoperator with
uniform descent for 𝑛 ≥ 𝑑 on the Banach space 𝑋, the followingare equivalent:
(i)𝑇 has topological uniform descent for
𝑛 ≥ 𝑑.
(ii) There is an 𝑛 ≥ 𝑑 and a positive integer 𝑘
for which 𝑅(𝑇𝑛) + 𝑁(𝑇𝑘) isclosed in 𝑋.
Theorem 2.5: [2, Theorem 2.5] 𝑇 ∈ 𝑄𝐹(𝑑), if
and only if 𝑇∗∈ 𝑄𝐹(𝑑).
Theorem 2.6: [7, Theorem 2] An operator 𝑇 ∈ 𝐿(𝑋) is quasi-Fredholm if andonly if exists
𝑑 ∈ ℕsuch that 𝑅(𝑇𝑑) is closed and 𝑇𝑑is semi regular.
Theorem 2.7: An operator 𝑇 ∈ 𝐿(𝑋) has TUD if and only if exists 𝑑 ∈ ℕ suchthat 𝑇𝑑 is semi regular.
Proof.𝑇𝑑 is semi regular if and only if
𝑁(𝑇𝑑) = 𝑁(𝑇) ∩ 𝑅(𝑇𝑑) ⊆ 𝑅((𝑇𝑑)𝑛) , forall
𝑛 ∈ ℕ and 𝑅(𝑇𝑑) = 𝑅(𝑇𝑑+1) is closed in
𝑅(𝑇𝑑) , or equivalent 𝑁(𝑇) ∩ 𝑅(𝑇𝑑) ⊆ 𝑅(𝑇𝑛) , for all 𝑛 ≥ 𝑑 and 𝑅(𝑇𝑑+1) is closed in 𝑅(𝑇𝑑), i.e 𝑇 has topological uniformdescent.
The following Lemma was proved in [2] under the hypothesis that 𝑇 is a quasi-Fredholm operator. However, we can observe that this conclusion remain true ifwe suppose that the operator T has TUD.
Lemma 2.8: If 𝑇 ∈ 𝐿(𝑋) and T have TUD then there exists asequence {𝑁𝑗}𝑗=0
∞
of subspaces of X such that for all 𝑗 = 0,1, 2 … we have
(i)𝑁𝑗⊆ 𝑁𝑗+1,
(ii)𝑇(𝑁𝑗) ⊆ 𝑁𝑗,
(iii)𝑁𝑗⊆ 𝑁(𝑇𝑗),
(iv)𝑁𝑗∩ 𝑅(𝑇𝑑) = 0,
(v)𝑁𝑗+ 𝑅(𝑇𝑑) = 𝑁(𝑇𝑗) + 𝑅(𝑇𝑑)
Lemma 2.9: If 𝑇 ∈ 𝐿(𝑋) and T have TUD then there exists a sequence {𝑀𝑗}𝑗=0∞ of subspaces of X such that for all 𝑗 = 0,1, 2 … we have
(i)𝑀𝑗+1⊆ 𝑀𝑗, (ii)𝑇(𝑀𝑗) ⊆ 𝑀𝑗, (iii)𝑅(𝑇𝑗) ⊆ 𝑀 𝑗, (iv)𝑀𝑗+ 𝑁(𝑇𝑑) = 𝑋, (v)𝑀𝑗+ 𝑁(𝑇) = 𝑇−1(𝑀𝑗+1), (vi)𝑀𝑗∩ (𝑅(𝑇𝑗) + 𝑁(𝑇𝑑)) ⊆ 𝑅(𝑇𝑗), (vii)𝑀𝑗∩ 𝑁(𝑇𝑑) = 𝑅(𝑇𝑗) ∩ 𝑁(𝑇𝑑).
Proof. Let L be a subspace of X such that
(𝑅(𝑇) + 𝑁(𝑇𝑑)) ⊕ 𝐿 = 𝑋 and we define
𝑀𝑗= {
𝑋, 𝑖𝑓 𝑗 = 0 𝑇(𝑀𝑗−1) + 𝐿, 𝑖𝑓 𝑗 ≥ 1.
Note that 𝑀𝑗 is clearly a subspace of X, for each j.
(i) If 𝑗 = 0, we have 𝑀1= 𝑇(𝑋) + 𝐿 ⊆ 𝑋 =
𝑀0 . Suppose that 𝑀𝑗+1⊆ 𝑀𝑗 ,then 𝑀𝑗+2=
𝑇(𝑀𝑗+1) + 𝐿 ⊆ 𝑇(𝑀𝑗) + 𝐿 = 𝑀𝑗+1 . Therefore,
𝑀𝑗+1⊆ 𝑀𝑗foreach j.
(ii) If 𝑗 = 0, we have 𝑇(𝑀0) = 𝑇(𝑋) ⊆ 𝑋 =
𝑀0. Suppose that 𝑇(𝑀𝑗) ⊆ 𝑀𝑗,then 𝑇(𝑀𝑗+1) =
𝑇(𝑇(𝑀𝑗) + 𝐿) ⊆ 𝑇(𝑀𝑗+ 𝐿) = 𝑇(𝑀𝑗) ⊆
𝑇(𝑀𝑗) + 𝐿 = 𝑇(𝑀𝑗+1) .Therefore, 𝑇(𝑀𝑗) ⊆ 𝑀𝑗
(iii) If 𝑗 = 0 , we have 𝑅(𝑇0) = 𝑋 = 𝑀0.
Suppose that 𝑅(𝑇𝑗) ⊆ 𝑀𝑗, then
𝑅(𝑇𝑗+1) ⊆ 𝑇 (𝑅(𝑇𝑗)) + 𝐿 ⊆ 𝑇(𝑀
𝑗) + 𝐿 = 𝑀𝑗+1
Consequently, we obtain that 𝑅(𝑇𝑗+1) ⊆
𝑀𝑗+1, and hence 𝑅(𝑇𝑗) ⊆ 𝑀𝑗for eachj.
(iv)If 𝑗 = 0 , we have 𝑀0+ 𝑁(𝑇𝑑) = 𝑋 + 𝑁(𝑇𝑑) = 𝑋 . Suppose that 𝑀𝑗+ 𝑁(𝑇𝑑) = 𝑋 , then 𝑀𝑗+1+ 𝑁(𝑇𝑑) = 𝑇(𝑀𝑗) + 𝐿 + 𝑁(𝑇𝑑) = 𝑇(𝑀𝑗) + 𝑇 (𝑁(𝑇𝑑)) + 𝐿 + 𝑁(𝑇𝑑) = 𝑇(𝑀𝑗+ 𝑁(𝑇𝑑)) + 𝐿 + 𝑁(𝑇𝑑) = 𝑅(𝑇) + 𝑁(𝑇𝑑) + 𝐿 = 𝑋.
Therefore 𝑀𝑗+ 𝑁(𝑇𝑑) = 𝑋 for each j.
(v)Suppose that 𝑥 ∈ 𝑇−1(𝑀𝑗+1) , then
𝑇𝑥 ∈ 𝑀𝑗+1= 𝑇(𝑀𝑗) + 𝐿 and so 𝑇𝑥 = 𝑇𝑚𝑗+ 𝑙 ,
where𝑚𝑗∈ 𝑀𝑗 and 𝑙 ∈ 𝐿. In consequence𝑇(𝑥 −
𝑚𝑗) = 𝑙 , thus 𝑥 − 𝑚𝑗∈ 𝑁(𝑇) because 𝐿 ∩
𝑅(𝑇) = 0 . Hence 𝑥 ∈ 𝑁(𝑇) + 𝑀𝑗 .
Therefore𝑇−1(𝑀
𝑗+1) ⊆ 𝑀𝑗+ 𝑁(𝑇).
On the other hand, if 𝑥 ∈ 𝑀𝑗+ 𝑁(𝑇) then
𝑇𝑥 ∈ 𝑇(𝑀𝑗) and so 𝑇𝑥 ∈ 𝑇(𝑀𝑗) + 𝐿 = 𝑀𝑗+1. In
consequence 𝑥 ∈ 𝑇−1(𝑀𝑗+1).
(vi) If 𝑗 = 0 , we have 𝑀0∩ (𝑅(𝑇0) +
𝑁(𝑇𝑑)) = 𝑋 ∩ (𝑋 + 𝑁(𝑇𝑑)) = 𝑋 = 𝑅(𝑇0).
Now suppose that 𝑀𝑗∩ (𝑅(𝑇𝑗) + 𝑁(𝑇𝑑)) ⊆
𝑅(𝑇𝑗). Observe that 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) = 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) ∩ 𝑀 1∩ (𝑅(𝑇) + 𝑁(𝑇𝑑)) = 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) ∩ (𝑅(𝑇) + 𝐿) ∩ (𝑅(𝑇) + 𝑁(𝑇𝑑)) = 𝑀 𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) ∩ 𝑅(𝑇) = 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑) ∩ 𝑅(𝑇)), also 𝑇 (𝑅(𝑇𝑗) + 𝑁(𝑇𝑑+1)) = 𝑇 (𝑅(𝑇𝑗)) + 𝑇 (𝑁(𝑇𝑑+1)) = 𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑) ∩ 𝑅(𝑇). So 𝑇−1(𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑) ∩ 𝑅(𝑇)) = 𝑅(𝑇𝑗) + 𝑁(𝑇𝑑+1) + 𝑁(𝑇) = 𝑅(𝑇𝑗) + 𝑁(𝑇𝑑). Then if 𝑥 ∈ 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) = 𝑀𝑗+1∩ (𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑) ∩ 𝑅(𝑇)) we have 𝑇−1({𝑥}) ⊆ 𝑇−1(𝑀𝑗+1) ∩ 𝑇−1(𝑅(𝑇𝑗+1) + 𝑁(𝑇𝑑)) . In consequence 𝑇−1({𝑥}) ⊆ (𝑀 𝑗+ 𝑁(𝑇)) ∩ (𝑅(𝑇𝑗) + 𝑁(𝑇𝑑)). Thus 𝑇−1({𝑥}) ⊆ 𝑀 𝑗∩ (𝑅(𝑇𝑗) + 𝑁(𝑇𝑑)), hence 𝑇−1({𝑥}) ⊆ (𝑅(𝑇𝑗)) and so 𝑇(𝑇−1({𝑥})) ⊆ 𝑇(𝑅(𝑇𝑗). Therefore𝑥 ∈ 𝑅(𝑇𝑗+1).
(vii) It follows of (iii) and (vi).
In the following theorem, we construct a decomposition property for operators that have TUD, similar to the Kato type operator.
Theorem 2.10: If𝑇 ∈ 𝐿(𝑋) and T have TUD, then there exist T invariantsubspaces M and N of
X such that, 𝑋 = 𝑀 ⊕ 𝑁,𝑇⎸𝑀 is semi regular and 𝑇⎸𝑁 is nilpotent.
Proof. From Lemma 2.8 and 2.9 it follows that there are sequences of subspacesof X, {𝑁𝑗}𝑗=0
∞
and {𝑀𝑗}𝑗=0 ∞
that satisfy certain
conditions. Let 𝑀 = 𝑀𝑑 and𝑁 = 𝑁𝑑, then of (iv)
and (iii) of Lemma 2.9, and (v) of Lemma 2.8 it follows that
𝑀 + 𝑁 = 𝑀 + 𝑅(𝑇𝑑) + 𝑁
= 𝑀 + 𝑅(𝑇𝑑) + 𝑁(𝑇𝑑) = 𝑋
Also, of (iii) and (iv) of Lemma 2.8, and (v) of Lemma 2.9 it follows that
𝑀 ∩ 𝑁 = 𝑀 ∩ 𝑁 ∩ 𝑁(𝑇𝑑) = 𝑅(𝑇𝑑) ∩ 𝑁(𝑇𝑑) ∩
𝑁 = {0},
so, 𝑋 = 𝑀 ⊕ 𝑁. On the other hand, since T have TUD follows that
𝑁[(𝑇⎸𝑀)𝑑] = 𝑀 ∩ 𝑁(𝑇𝑑) = 𝑅(𝑇𝑑) ∩ 𝑁(𝑇𝑑) =
𝑅(𝑇𝑑+𝑗) ∩ 𝑁(𝑇𝑑) ⊆ 𝑅(𝑇𝑑+𝑗),
for all 𝑗 = 0,1, 2 … . Also, 𝑇𝑑(𝑋) = 𝑇𝑑(𝑀 ⊕
𝑁) = 𝑇𝑑(𝑀) = 𝑅[(𝑇⎸
𝑀)𝑑] it follows that
𝑅[(𝑇⎸𝑀)𝑑]is closed. This shows that (𝑇⎸𝑀)𝑑is
semi-regular.Hence, by Theorem 1.2, 𝑇⎸𝑀 is
semi-regular.
Theorem 2.11: Let X a Banach space, 𝑇 ∈ 𝐿(𝑋)and suppose that there existT invariant subspaces M and N of X such that 𝑋 = 𝑀 ⊕ 𝑁,
𝑇⎸𝑀 is essentiallysemi-regular and 𝑇⎸𝑁 is nilpotent, then T have TUD.
Proof. Suppose that there exist T invariant subspaces M and N of X such that 𝑋 = 𝑀 ⊕ 𝑁,
𝑇⎸𝑀 is essentially semi-regular and 𝑇⎸𝑁 is
nilpotent.Let 𝑀1 and 𝑁1 subspaces T invariant of
Msuch that 𝑇⎸𝑀1is semi-regular, 𝑇⎸𝑁1is nilpotent
and 𝑀 = 𝑀1⊕ 𝑁1 (see Theorem 1.4). If 𝑑 ∈
ℕsuch that (𝑇⎸𝑁)𝑑= 0and (𝑇⎸𝑁
1)
𝑑
= 0, then
𝑅(𝑇𝑛) = 𝑇𝑛(𝑋) = 𝑇𝑛((𝑀1⊕ 𝑁1) ⊕ 𝑁)
= 𝑇𝑛(𝑀1)
for all 𝑛 ≥ 𝑑, so 𝑅(𝑇𝑛) is closed for all 𝑛 ≥ 𝑑. Also,
𝑁(𝑇) ∩ 𝑅(𝑇𝑑) = 𝑁(𝑇) ∩ 𝑇𝑑(𝑀
1) ⊆ 𝑁(𝑇) ∩ 𝑀1
⊆ 𝑇𝑛(𝑀
1) = 𝑅(𝑇𝑛)
for all 𝑛 ≥ 𝑑, and consequently 𝑇𝑑is semi-regular.
By Theorem 2.6, we concludethat T is quasi-Fredholm.
Corollary 2.12: If X a Banach space and 𝑇 ∈ 𝐿(𝑋) is of Kato type, then T have TUD.
As in Lemma 2.8, the condition T is a quasi-Fredholm operator can be replacedby the hypothesis T has TUD, of Theorem [2, Theorem 2.8], and the conclusión of its remains true.
Theorem 2.13: If 𝑇 ∈ 𝐿(𝑋) has topological
uniform descent for 𝑛 ≥ 𝑑 and𝑁(𝑇𝑑) + 𝑅(𝑇) is
complemented, then there exist T invariant subspaces M andN of X such that, Mis closed,𝑋 =
𝑀 ⊕ 𝑁,𝑇⎸𝑀is semi-regular and 𝑇⎸𝑁is nilpotent.
Corollary 2.14: If 𝑇 ∈ 𝐿(𝑋) has topological
uniform descent for 𝑛 ≥ 𝑑 and𝑁(𝑇𝑑) + 𝑅(𝑇) is
complemented, then T is quasi-Fredholm.
The following theorem extends the results obtained in [3, Theoren 2.8] to alarger class of operators.
Theorem 2.15: If 𝑇 ∈ 𝑄𝐹(𝑑), 𝑁(𝑇𝑑) + 𝑅(𝑇)
is complemented, 𝑆 ∈ 𝐿(𝑋) suchthat
dim [𝑆𝑛(𝑋)] < ∞ for some 𝑛 ∈ ℕ and TS = ST,
then T + S is quasiFredholm.
Proof. Suppose that T is a operator quasi-Fredholm. By Theorem [2, Theorem2.8], it follows that there exist T invariant subspaces M
and N of X such that Mis closed, 𝑋 = 𝑀 ⊕
𝑁,𝑇⎸𝑀 is semi-regular and 𝑇⎸𝑁 is nilpotent. Let
𝑝, 𝑞 ∈ ℕ be such that dim [𝑆𝑝(𝑋)] < ∞ and
(𝑇⎸𝑁)𝑞= 0, then (𝑇 + 𝑆)𝑝+𝑞= ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝+𝑞 𝑘=0 𝑆𝑘 = ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=0 𝑆𝑘 + ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝+𝑞 𝑘=𝑝+1 𝑆𝑘 = 𝑇𝑝+𝑞+ ∑ 𝑐 𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘 + ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝+𝑞 𝑘=𝑝+1 𝑆𝑘 = 𝑇𝑝+𝑞 + 𝑆(∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘−1) + 𝑆𝑝( ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝+𝑞 𝑘=𝑝+1 𝑆𝑘−𝑝) Also, 𝑆 (∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘−1) (𝑀) = ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘(𝑀) ⊆ ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=𝑝 (𝑋) = ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 (𝑀 + 𝑁) = ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 (𝑀) ⊆ 𝑀 𝑆 (∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘−1) (𝑁) = ∑ 𝑐𝑝+𝑞,𝑘𝑆𝑘(𝑇𝑝+𝑞−𝑘(𝑁) 𝑝 𝑘=1 ) = 0 𝑇𝑝+𝑞(𝑀) ⊆ 𝑀 and 𝑇𝑝+𝑞(𝑁) = 0 . Then 𝑇𝑝+𝑞∈
𝐿(𝑀) is semi-regular (see Theorem 1.2) and 𝑆(∑𝑝𝑘=1𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘𝑆𝑘−1) ∈ 𝐿(𝑀) has finite-range power, from Theorem 1.5 follows
that𝑇𝑝+𝑞+ 𝑆(∑𝑝𝑘=1𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘𝑆𝑘−1) ∈ 𝐿(𝑀)
𝑇𝑝+𝑞+ 𝑆 (∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝
𝑘=1
𝑆𝑘−1) (𝑁) = 0
From Theorem 2.11 that
𝑇𝑝+𝑞+ 𝑆 (∑ 𝑐 𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘−1) is quasi-Fredholm, in consequence (𝑇 + 𝑆)𝑝+𝑞 = 𝑇𝑝+𝑞+ ∑ 𝑐 𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝 𝑘=1 𝑆𝑘 + ∑ 𝑐𝑝+𝑞,𝑘𝑇𝑝+𝑞−𝑘 𝑝+𝑞 𝑘=𝑝+1 𝑆𝑘
is quasi-Fredholm (see [8, Theorem 15]). Thus, we can conclude that T + S isquasi-Fredholm (see Theorem 1.2).
Corollary 2.16: If 𝑇 ∈ 𝐿(𝑋), has topological
uniform descent for 𝑛 ≥ 𝑑 , 𝑁(𝑇𝑑) + 𝑅(𝑇) is
complemented, 𝑆 ∈ 𝐿(𝑋) such that
dim [𝑆𝑛(𝑋)] < ∞ for some 𝑛 ∈ ℕ and TS = ST,
then T + S has topological uniform descent. By duality (see Theorem 2.5), we have. Corollary 2.17: If 𝑇 ∈ 𝑄𝐹(𝑑) , 𝑁(𝑇) ∩
𝑅(𝑇𝑑) is complemented, 𝑆 ∈ 𝐿(𝑋) such that
dim [𝑆𝑛(𝑋)] < ∞ for some 𝑛 ∈ ℕ and TS = ST,
then T + S is quasiFredholm.
Similar to Theorem 2.15, we have the following result.
Theorem 2.18: If X is a Banach space, 𝑇 ∈ 𝐿(𝑋), is a Kato type operator, and𝑆 ∈ 𝐿(𝑋), such that dim [𝑆𝑛(𝑋)] < ∞ , for some 𝑛 ∈ ℕ and TS = ST, then T + S is of Kato type operator.
IV. N
EWC
HARACTERIZATION FOR THEC
LASS OFO
PERATORSL
OWERS
EMIB-W
EYLO
PERATORSAccording [5] and [9], T is said to be semi Fredholm (resp. Fredholm,upper semi B-Fredholm, lower semi B-Fredholm), if for some
integer 𝑛 ≥ 0 therange 𝑅(𝑇𝑛) is closed and 𝑇𝑛,
viewed as an operator from the space 𝑅(𝑇𝑛) into
itself, is a semi-Fredholm (respectively,
Fredholm, upper semi-Fredholm,
lowersemi-Fredholm) operator. Analogously, 𝑇 ∈ 𝐿(𝑋) is
said to be Browder (resp.upper semi B-Browder, lower semi B-B-Browder, B-Weyl, upper semi B-Weyl,lower semi B-Weyl), if for some
integer 𝑛 ≥ 0 the range 𝑅(𝑇𝑛) is closed and 𝑇
𝑛is a Browder (resp. upper semi-Browder, lower semi -Browder, Weyl, upper semi B-Weyl, lower semi B-Weyl) operator.
Theorem 3.1: Let X be a Banach space and
let 𝑇 ∈ 𝑄𝐹(𝑑). If 𝑁(𝑇𝑑) + 𝑅(𝑇) is a
complemented subspace of X, then T is upper semi B-Weyl operator if only if there exist T invariant closed subspacesM and N of X such
that 𝑋 = 𝑀 ⊕ 𝑁 , 𝑇⎸𝑀 is upper semi Weyl
and𝑇𝑛(𝑁) = {0} for some 𝑛 ∈ ℕ.
Proof. Suppose that 𝑇 ∈ 𝑄𝐹(𝑑) is an upper
semi B-Weyl, then 𝑁(𝑇) ∩ 𝑅(𝑇𝑑) is
complemented. Since that 𝑁(𝑇𝑑) + 𝑅(𝑇) is
complemented, From [2, Theorem 2.8 and Lemma 2.7] it follows that there exist T invariant closed subspaces M and N of Xsuch that:
(i)𝑅(𝑇𝑑) is a subspace of M, (ii)𝑋 = 𝑀 ⊕ 𝑁, (iii)𝑇⎸𝑀 is semi-regular, (iv)𝑇𝑑(𝑁) = 0. Then, 𝑇𝑛(𝑋) = 𝑇𝑛(𝑀 ⊕ 𝑁) = 𝑇𝑛(𝑀) = 𝑅[(𝑇⎸ 𝑀)𝑛] , for all 𝑛 ≥ 𝑑.
Now, by [1, Corollary 12.23] and since 𝑇⎸𝑀 is semi-regular, thus we have
𝑅(𝑇𝑑) 𝑅(𝑇𝑑+1)= 𝑅[(𝑇⎸𝑀)𝑑] 𝑅[(𝑇⎸𝑀)𝑑+1] = 𝑀 𝑅(𝑇⎸𝑀) + 𝑁[(𝑇⎸𝑀)𝑑] = 𝑀 𝑅(𝑇⎸𝑀) So, 𝛽(𝑇⎸𝑀) = dim [ 𝑀 𝑅(𝑇⎸𝑀) ] = 𝑑𝑖𝑚 [𝑅(𝑇𝑅(𝑇𝑑+1𝑑))] = 𝑑𝑖𝑚 [ 𝑅(𝑇𝑛) 𝑅(𝑇𝑛+1)] = 𝛽(𝑇𝑛), for all 𝑛 ≥ 𝑑.Also, 𝛼(𝑇⎸𝑀) = dim[𝑁(𝑇) ∩ 𝑀] = 𝑑𝑖𝑚[𝑁(𝑇) ∩ 𝑁(𝑇𝑑) ∩ 𝑀] = 𝑑𝑖𝑚[𝑁(𝑇) ∩ 𝑁(𝑇𝑑) ∩ 𝑅(𝑇𝑑)] = 𝑑𝑖𝑚[𝑁(𝑇) ∩ 𝑅(𝑇𝑑)] = 𝑑𝑖𝑚[𝑁(𝑇) ∩ 𝑅(𝑇𝑛)] = 𝛼(𝑇𝑛) < ∞ For all 𝑛 ≥ 𝑑. Therefore
𝛼(𝑇⎸𝑀) − 𝛽(𝑇⎸𝑀) = 𝛼(𝑇𝑛) − 𝛽(𝑇𝑛) ≤ 0
Conversely, suppose that there there exist T invariant closed subspaces M and N of X such
that 𝑋 = 𝑀 ⊕ 𝑁, 𝑇⎸𝑀 is upper semi Weyl and
𝑇𝑛(𝑁) = {0} for some 𝑛 ∈ ℕ. Then,
𝑅(𝑇𝑛) = 𝑇𝑛(𝑋) = 𝑇𝑛(𝑀 ⊕ 𝑁) = 𝑇𝑛(𝑀) =
𝑅[(𝑇⎸𝑀)𝑛], for all 𝑛 ≥ 𝑑. It follows that 𝑅(𝑇𝑛) is closed for all 𝑛 ≥ 𝑑 , and 𝑇𝑑= 𝑇⎸𝑅(𝑇𝑑) = 𝑇⎸𝑅[(𝑇⎸
𝑀)𝑑] = (𝑇⎸𝑀)𝑑.
This shows that T is upper semi B-Weyl.
The following theorem was established in [4] for the particular case of Hilbert spaces. Now, we extend this to the context of Banach spaces.
Theorem 3.2: Let X be a Banach space and 𝑇 ∈ 𝐿(𝑋). Then T is lower semiB-Weyl, if and only if there exist 𝑆, 𝐾 ∈ 𝐿(𝑋), such that T = S +
K, S is lowersemi B-Browder and
dim [𝐾𝑛(𝑋)] < ∞, for some 𝑛 ∈ ℕ.
Proof.(⇒) This is an immediate consequence of [2, Theorem 3.4].
(⇐)Conversely, suppose that there exist 𝑆, 𝐾 ∈ 𝐿(𝑋) such that T = S + K, S is lower semi
B-Browder and dim [𝐾𝑛(𝑋)] < ∞ , for some
𝑛 ∈ ℕ. By [3, Theoren 2.8] T is lowersemi B-Browder, in consequence T is lower semi B-Weyl.