Abstract
Let \({{\mathcal {S}}}{{\mathcal {S}}}\) and \({{\mathcal {S}}}{{\mathcal {C}}}\) be the strictly singular and the strictly cosingular operators acting between Banach spaces, and let \(P\Phi _+\) and \(P\Phi _+\) be the perturbation classes for the upper and the lower semi-Fredholm operators. We study two classes of operators \(\Phi {\mathcal {S}}\) and \(\Phi {\mathcal {C}}\) that satisfy \({{\mathcal {S}}}{{\mathcal {S}}}\subset \Phi {\mathcal {S}}\subset P\Phi _+\) and \({{\mathcal {S}}}{{\mathcal {C}}}\subset \Phi {\mathcal {C}}\subset P\Phi _-.\) We give some conditions under which these inclusions become equalities, from which we derive some positive solutions to the perturbation classes problem for semi-Fredholm operators.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
The perturbation classes problem asks whether the perturbation classes for the upper semi-Fredholm operators \(P\Phi _+\) and the lower semi-Fredholm operators \(P\Phi _-\) coincide with the classes of strictly singular operators \({{\mathcal {S}}}{{\mathcal {S}}}\) and strictly cosingular operators \({{\mathcal {S}}}{{\mathcal {C}}},\) respectively. This problem was raised in [9] (see also [5, 19]), and it has a positive answer in some cases [11, 13,14,15, 21], but the general answer is negative in both cases [10, 8, Theorem 4.5]. However, it remains interesting to find positive answers in special cases because the definitions of \({{\mathcal {S}}}{{\mathcal {S}}}\) and \({{\mathcal {S}}}{{\mathcal {C}}}\) are intrinsic: to check that K is in one of them only involves the action of K, while to check that K is in \(P\Phi _+\) or \(P\Phi _-\) we have to study the properties of \(T+K\) for T in a large set of operators.
In this paper, we consider two classes \(\Phi {\mathcal {S}}\) and \(\Phi {\mathcal {C}}\) introduced in [2] that satisfy
We study conditions on the Banach spaces X, Y so that some of these four inclusions become equalities, and we derive new positive answers to the perturbation classes problem for semi-Fredholm operators. When \(\Phi _+(X,Y)\ne \emptyset ,\) we show that \(\Phi {\mathcal {S}}(X,Y)= P\Phi _+(X,Y)\) when \(Y\times Y\) is isomorphic to Y, \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)= \Phi {\mathcal {S}}(X,Y)\) when some quotients of X embed in Y (Theorem 3.1), and adding up we get conditions implying that \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)= P\Phi _+(X,Y)\) (Theorem 3.2). When \(\Phi _-(X,Y)\ne \emptyset \) we prove similar results (Theorems 4.1 and 4.2). We also state some questions concerning the classes \(\Phi {\mathcal {S}}\) and \(\Phi {\mathcal {C}}.\)
Notation. Along the paper, X, Y and Z denote Banach spaces and \({\mathcal {L}}(X,Y)\) is the set of bounded operators from X into Y. We write \({\mathcal {L}}(X)\) when \(X=Y.\) Given a closed subspace M of X, we denote \(J_M\) the inclusion of M into X, and \(Q_M\) the quotient map from X onto X/M. An operator \(T\in {\mathcal {L}}(X,Y)\) is an isomorphism if there exists \(c>0\) such that \(\Vert Tx\Vert \ge c\Vert x\Vert \) for every \(x\in X.\)
The operator \(T\in {\mathcal {L}}(X,Y)\) is strictly singular, and we write \(T\in {{\mathcal {S}}}{{\mathcal {S}}},\) when there is no infinite dimensional closed subspace M of X such that the restriction \(TJ_M\) is an isomorphism; and T is strictly cosingular, and we write \(T\in {{\mathcal {S}}}{{\mathcal {C}}},\) when there is no infinite codimensional closed subspace N of Y such that \(Q_NT\) is surjective. Moreover, T is upper semi-Fredholm, \(T\in \Phi _+,\) when the range R(T) is closed and the kernel N(T) is finite dimensional; T is lower semi-Fredholm, \(T\in \Phi _-,\) when R(T) is finite codimensional (hence closed); T is Fredholm, \(T\in \Phi ,\) when it is upper and lower semi-Fredholm; and T is inessential, \(T\in {{\mathcal {I}}n},\) when \(I_X-ST\in \Phi \) for each \(S\in {\mathcal {L}}(Y,X).\)
2 Preliminaries
The perturbation class \(P{\mathcal {S}}\) of a class of operators \({\mathcal {S}}\) is defined in terms of its components:
Definition 2.1
Let \({\mathcal {S}}\) denote one of the classes \(\Phi _+,\) \(\Phi _-\) or \(\Phi .\) For spaces X, Y such that \({\mathcal {S}}(X,Y)\ne \emptyset ,\)
We could define \(P{\mathcal {S}}(X,Y)= {\mathcal {L}}(X,Y)\) when \({\mathcal {S}}(X,Y)\) is empty, but this is not useful. The components of \(P\Phi \) coincide with those of the operator ideal of inessential operators \({{\mathcal {I}}n}\) when they are defined [1], but given \(S\in {\mathcal {L}}(Y,Z)\) and \(T\in {\mathcal {L}}(X,Y),\) S or T in \(P\Phi _+\) does not imply \(ST\in P\Phi _+,\) and similarly for \(P\Phi _-\) [10]. However, the following result holds, and it will be useful for us.
Proposition 2.1
Suppose that \(K\in P\Phi _+(X,Y),\) \(L\in P\Phi _-(X,Y),\) \(S\in {\mathcal {L}}(Y)\) and \(T\in {\mathcal {L}}(X).\) Then, \(SK, KT\in P\Phi _+(X,Y)\) and \(SL, LT\in P\Phi _-(X,Y).\)
Proof
Suppose that \(K\in P\Phi _+(X,Y)\) and let \(S\in {\mathcal {L}}(Y)\) and \(U\in \Phi _+(X,Y).\) If S is bijective, then \(S^{-1}U\in \Phi _+(X,Y),\) hence \(U+SK= S(S^{-1}U +K)\in \Phi _+;\) thus \(SK \in P\Phi _+(X,Y).\) In the general case, \(S=S_1+S_2\) with \(S_1,S_2\) bijective; thus, \(SK= S_1K+S_2K \in P\Phi _+(X,Y).\)
The proof of the other three results is similar. \(\square \)
3 The perturbation class for \(\Phi _+\)
Given two operators \(S,T\in {\mathcal {L}}(X,Y),\) we denote by (S, T) the operator from X into \(Y\times Y\) defined by \((S,T)x=(Sx,Tx),\) where \(Y\times Y\) is endowed with the product norm \(\Vert (y_1,y_2)\Vert _1= \Vert y_1\Vert + \Vert y_2\Vert .\)
Inspired by the results of Friedman [6], the authors of [2] defined the following class of operators.
Definition 3.1
Suppose that \(\Phi _+(X,Y) \ne \emptyset \) and let \(K\in {\mathcal {L}}(X,Y).\) We say that K is \(\Phi \)-singular, and write \(K\in \Phi {\mathcal {S}},\) when for each \(S\in {\mathcal {L}}(X,Y),\) \((S,K)\in \Phi _+\) implies \(S\in \Phi _+.\)
The definition of \(\Phi {\mathcal {S}}(X,Y)\) is similar to that of \(P\Phi _+(X,Y),\) but the former one is easier to handle because the action of S and K is decoupled when we consider (S, K) instead of \(S+K.\)
With our notation, [6, Theorems 3 and 4] can be stated as follows:
Proposition 3.1
[2, Proposition 2.2] Suppose that \(\Phi _+(X,Y)\ne \emptyset .\) Then,
Note that \({{\mathcal {S}}}{{\mathcal {S}}}\) is an operator ideal but \(P\Phi _+\) is not; \(P\Phi _+(X,Y)\) is a closed subspace of \({\mathcal {L}}(X,Y),\) and \(P\Phi _+(X)\) is an ideal of \({\mathcal {L}}(X).\)
Proposition 3.2
\(\Phi {\mathcal {S}}(X,Y)\) is closed in \({\mathcal {L}}(X,Y).\)
Proof
Let \(\{T_n\}\) be a sequence in \(\Phi {\mathcal {S}}(X,Y)\) converging to \(T\in {\mathcal {L}}(X,Y).\) Suppose that \(S\in {\mathcal {L}}(X,Y)\) and \((S,T) \in \Phi _+(X, Y \times Y).\) Note that the sequence \((S,T_n)\) converge to (S, T), because \(\Vert (S,T_n)- (S,T)\Vert = \Vert T_n-T\Vert .\)
Since \( \Phi _+(X, Y \times Y)\) is an open set, there exists a positive integer N such that \((S,T_N)\in \Phi _+(X,Y\times Y).\) Then, \(T_N\in \Phi {\mathcal {S}}(X,Y)\) implies \(S \in \Phi _+(X,Y).\) Thus \(T\in \Phi {\mathcal {S}}(X,Y).\) \(\square \)
We state some basic questions on the class \(\Phi {\mathcal {S}}.\)
Question 3.1
Suppose that \(\Phi _+(X,Y)\ne \emptyset .\)
-
(a)
Is \(\Phi {\mathcal {S}}(X,Y)\) a subspace of \({\mathcal {L}}(X,Y)?\)
-
(b)
Is \(\Phi {\mathcal {S}}\) an operator ideal?
-
(c)
Is Proposition 2.1 valid for \(\Phi {\mathcal {S}}?\)
Answering a question in [6], an example of an operator \(K\in P\Phi _+\setminus \Phi {\mathcal {S}}\) was given in [2, Example 2.3], but we do not know if the other inclusion can be strict.
Question 3.2
Suppose that \(\Phi _+(X,Y)\ne \emptyset .\) Is \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)=\Phi {\mathcal {S}}(X,Y)?\)
A negative answer to Question 3.2 would provide a new counterexample to the perturbation classes problem for \(\Phi _+.\)
Let us see that the inclusions in Proposition 3.1 become equalities in some cases.
An infinite dimensional Banach space Y is isomorphic to its square, denoted \(Y\times Y\simeq Y,\) in many cases: \(L_p(\mu )\) and \(\ell _p\) \((1\le p\le \infty ),\) \(c_0,\) and C[0, 1]. On the other hand, James’ space J and some spaces of continuous functions on a compact like \(C[0,\omega _1]\) are not isomorphic to their square, where \(\omega _1\) is the first uncountable ordinal. See [4, 20].
Theorem 3.1
Suppose that the spaces X and Y satisfy \(\Phi _+(X,Y)\ne \emptyset .\)
-
1.
If \(Y\times Y\simeq Y\), then \(\Phi {\mathcal {S}}(X,Y)= P\Phi _+(X,Y).\)
-
2.
If every infinite dimensional subspace of X has an infinite dimensional subspace N such that X/N embeds in Y, then \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)=\Phi {\mathcal {S}}(X,Y).\)
Proof
(1) Let \(U:Y\times Y\rightarrow Y\) be a bijective isomorphism and let \(V,W\in {\mathcal {L}}(Y)\) such that \(U(y_1,y_2)=Vy_1+ Wy_2.\) If \(K\in P\Phi _+(X,Y),\) for each \(S\in {\mathcal {L}}(X,Y)\) such that \((S,K)\in \Phi _+\) we have \(U(S,K)=VS+ WK\in \Phi _+.\) By Proposition 2.1, \(WK\in P\Phi _+(X,Y).\) Then, \(VS\in \Phi _+,\) hence \(S\in \Phi _+.\) Thus we conclude that \(K\in \Phi {\mathcal {S}}(X,Y).\)
(2) Let \(K\in {\mathcal {L}}(X,Y),\) \(K\notin {{\mathcal {S}}}{{\mathcal {S}}}.\) By the hypothesis, there exists an infinite dimensional subspace N of X such that \(KJ_N\) is an isomorphism, and there is an isomorphism \(U:X/N\rightarrow Y.\) Then, \(S=UQ_N\in {\mathcal {L}}(X,Y)\) is not upper semi-Fredholm. We will prove that \(K\notin \Phi {\mathcal {S}}\) by showing that \((S,K)\in \Phi _+.\)
Recall that \(\Vert Q_Nx\Vert = \mathop \textrm{dist}(x,N).\) We can choose the isomorphism U so that \(\Vert Sx\Vert =\Vert UQ_Nx\Vert \ge \mathop \textrm{dist}(x,N)\) for each \(x\in X.\) Moreover, there is a constant \(c>0\) such that \(\Vert Kn\Vert \ge c\Vert n\Vert \) for each \(n\in N.\)
Let \(x\in X\) with \(\Vert x\Vert =1,\) and let \(0<\alpha <1\) such that \(c(1-\alpha )= 2\Vert K\Vert \alpha .\)
If \(\mathop \textrm{dist}(x,N)\ge \alpha \), then \(\Vert Sx\Vert \ge \alpha .\) Otherwise, there exists \(y\in N\) such that \(\Vert x-y\Vert <\alpha ;\) hence \(\Vert y\Vert > 1-\alpha .\) Therefore,
Then, \(\Vert (S,K)x\Vert _1=\Vert Sx\Vert +\Vert Kx\Vert \ge \min \{\Vert K\Vert \alpha , \alpha \},\) hence (S, K) is an isomorphism; in particular \((S,K)\in \Phi _+,\) as we wanted to show. \(\square \)
In the known examples in which \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)\ne P\Phi _+(X,Y)\) in [8, 10], the space Y has a complemented subspace which is hereditarily indecomposable in the sense of [3, 16, 17]. Therefore, the question arises.
Question 3.3
Suppose that X and Y satisfy \(\Phi _+(X,Y)\ne \emptyset \) and \(Y\times Y\simeq Y.\)
Is \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)=P\Phi _+(X,Y)?\)
A Banach space X is subprojective if every closed infinite dimensional subspace of X contains an infinite dimensional subspace complemented in X. The spaces \(c_0,\) \(\ell _p\) (\(1\le p<\infty )\) and \(L_q(\mu )\) (\(2\le q<\infty )\) are subprojective [22]. See [7, 18] for further examples.
Corollary 3.1
Suppose that \(\Phi _+(X,Y)\ne \emptyset \) and the space X is subprojective. Then, \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)= \Phi {\mathcal {S}}(X,Y).\)
Proof
Every closed infinite dimensional subspace of X contains an infinite dimensional subspace N complemented subspace in X; thus X/N is isomorphic to the complement of N. Since \(\Phi _+(X,Y)\ne \emptyset ,\) the quotient X/N is isomorphic to a subspace of Y and we can apply Theorem 3.1. \(\square \)
The next result is a refinement of Theorem 3.1 that is proved using the previous arguments.
Theorem 3.2
Suppose that \(\Phi _+(X,Y)\ne \emptyset ,\) \(Y\times Y\) embeds in Y and every infinite dimensional subspace of X has an infinite dimensional subspace N such that X/N embeds in Y. Then, \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)=P\Phi _+(X,Y).\)
Proof
Since \(Y\times Y\) embeds in Y, there exist isomorphisms \(V,W\in {\mathcal {L}}(Y)\) such that \(R(V)\cap R(W)=\{0\}\) and \(R(V)+R(W)\) is closed. Hence, there exists \(r>0\) such that \(\Vert y_1+y_2\Vert \ge r(\Vert y_1\Vert +\Vert y_2\Vert )\) for \(y_1\in R(V)\) and \(y_2\in R(W),\) and clearly we can choose V, W so that \(r=1.\)
Let \(K\in {\mathcal {L}}(X,Y)\) with \(K\notin {{\mathcal {S}}}{{\mathcal {S}}}.\) Select an infinite dimensional subspace M of X such that \(KJ_M\) is an isomorphism, and let N be an infinite dimensional subspace of M such that there exists an isomorphism \(U:X/N\rightarrow Y.\) We can assume that \(\Vert Uz\Vert \ge z\) for each \(z\in X/N.\)
The operator \(S=VUQ_N \notin \Phi _+,\) and proceeding like in the proof of (2) in Theorem 3.1, we can show that \(S+WK\in \Phi _+.\) Then, \(WK \notin P\Phi _+,\) hence \(K \notin P\Phi _+,\) by Proposition 2.1. \(\square \)
Corollary 3.2
If X is separable, \(Y\times Y\) embeds in Y, and Y contains a copy of C[0, 1], then \({{\mathcal {S}}}{{\mathcal {S}}}(X,Y)=\Phi _+(X,Y).\)
Proof
It is well known that the space C[0, 1] contains a copy of each separable Banach space. \(\square \)
The class \(\Phi {\mathcal {S}}\) is injective in the following sense:
Proposition 3.3
Given an operator \(K \in {\mathcal {L}}(X, Y)\) and an (into) isomorphism \(L\in {\mathcal {L}}(Y,Y_0),\) if \(LK\in \Phi {\mathcal {S}}(X,Y_0)\), then \(K\in \Phi {\mathcal {S}}(X,Y).\)
Proof
Let \(K \in {\mathcal {L}}(X, Y)\) and let \(L \in {\mathcal {L}}(X,Y_0)\) be an isomorphism into \(Y_0\) such that \(LK \in \Phi {\mathcal {S}}(X,Y_0).\) Take \(S\in {\mathcal {L}}(X,Y)\) and suppose that \((S,K) \in \Phi _+(X, Y \times Y).\) Then, \((LS,LK)=(L\times L)(S,K)\in \Phi _+(X, Y_0 \times Y_0),\) where \((L\times L)\in {\mathcal {L}}(X\times X,Y_0 \times Y_0)\) is defined by \((L\times L)(x_1,x_2)=(Lx_1,Lx_2).\)
Since \(LK \in \Phi {\mathcal {S}}(X,Y_0)\) we obtain \(LS \in \Phi _+(X, Y_0).\) Therefore \(S \in \Phi _+(X, Y),\) hence \(K \in \Phi {\mathcal {S}}(X,Y).\) \(\square \)
4 The perturbation class for \(\Phi _-\)
Given two operators \(S,T\in {\mathcal {L}}(X,Y),\) we denote by [S, T] the operator from \(X\times X\) into Y defined by \([S,T](x_1,x_2) =Sx_1+Tx_2,\) where \(X\times X\) is endowed with the maximum norm \(\Vert (x_1,x_2)\Vert _\infty = \max \{\Vert y_1\Vert , \Vert y_2\Vert \}.\)
Definition 4.1
Suppose that \(\Phi _-(X,Y) \ne \emptyset \) and let \(K\in {\mathcal {L}}(X,Y).\) We say that K is \(\Phi \)-cosingular, and write \(K\in \Phi {\mathcal {C}},\) when for each \(S\in {\mathcal {L}}(X,Y),\) \([S,K]\in \Phi _-\) implies \(S\in \Phi _-.\)
Like in the case of \(\Phi {\mathcal {S}},\) the definition of \(\Phi {\mathcal {C}}(X,Y)\) is similar to that of \(P\Phi _-(X,Y),\) but the former one is easier to handle because the action of S and K is decoupled when we consider [S, K] instead of \(S+K.\)
Proposition 4.1
[2, Proposition 2.5] Suppose that \(\Phi _-(X,Y)\ne \emptyset .\) Then,
Note that \({{\mathcal {S}}}{{\mathcal {C}}}\) is an operator ideal but \(P\Phi _-\) is not; \(P\Phi _-(X,Y)\) is a closed subspace of \({\mathcal {L}}(X,Y),\) and \(P\Phi _-(X)\) is an ideal of \({\mathcal {L}}(X).\)
Proposition 4.2
\(\Phi {\mathcal {C}}(X,Y)\) is closed in \({\mathcal {L}}(X,Y).\)
Proof
Let \(\{T_n\}\) be a sequence in \(\Phi {\mathcal {C}}(X,Y)\) converging to \(T\in {\mathcal {L}}(X,Y).\) Suppose that \(S\in {\mathcal {L}}(X,Y)\) and \([S,T] \in \Phi _-(X\times X,Y).\) Note that the sequence \([S,T_n]\) converge to [S, T].
Since \(\Phi _-(X\times X,Y)\) is an open set there exists a positive integer N such that \([S,T_N] \in \Phi _-(X \times X, Y).\) Hence, \(T_N \in \Phi {\mathcal {C}}(X,Y)\) implies \(S \in \Phi _-(X,Y).\) \(\square \)
Question 4.1
Suppose that \(\Phi _-(X,Y)\ne \emptyset .\)
-
(a)
Is \(\Phi {\mathcal {C}}(X,Y)\) a subspace of \({\mathcal {L}}(X,Y)?\)
-
(b)
Is \(\Phi {\mathcal {C}}\) an operator ideal?
-
(c)
Is Proposition 2.1 valid for \(\Phi {\mathcal {C}}?\)
Answering a question in [6], an example of an operator \(K\in P\Phi _-\setminus \Phi {\mathcal {C}}\) was given in [2], but we do not know if the other inclusion can be strict.
Question 4.2
Suppose that \(\Phi _-(X,Y)\ne \emptyset .\) Is \({{\mathcal {S}}}{{\mathcal {C}}}(X,Y)=\Phi {\mathcal {C}}(X,Y)?\)
A negative answer to Question 4.2 would provide a new counterexample to the perturbation classes problem for \(\Phi _-.\)
Next we will show that the inclusions in Proposition 4.1 become equalities in some cases.
Theorem 4.1
Suppose that the spaces X and Y satisfy \(\Phi _-(X,Y) \ne \emptyset .\)
-
1.
If \(X\times X \simeq X \), then \(\Phi {\mathcal {C}}(X,Y)=P \Phi _-(X, Y).\)
-
2.
If every infinite codimensional closed subspace of Y is contained in an infinite codimensional closed subspace N which is isomorphic to a quotient of X, then \(\Phi {\mathcal {C}}(X,Y)= {{\mathcal {S}}}{{\mathcal {C}}}(X, Y).\)
Proof
(1) Let \(U:X \rightarrow X \times X\) be a bijective isomorphism and let \(U_1,U_2 \in {\mathcal {L}}(X)\) such that \(U(x)=(U_1x,U_2x).\) If \(K \in P\Phi _-(X, Y),\) for each \(S\in {\mathcal {L}}(X,Y)\) such that \([S,K] \in \Phi _-(X \times X, Y)\) we have \([S,K]U=SU_1+KU_2 \in \Phi _-(X, Y \times Y).\) By Proposition 2.1, \(KU_2 \in P\Phi _-(X, Y).\) Then, \(SU_1 \in \Phi _-(X, Y),\) hence \(S \in \Phi _-(X, Y).\) Thus we conclude that \(K \in \Phi {\mathcal {C}}(X,Y).\)
(2) Let \(K\in {\mathcal {L}}(X,Y)\) \(K\notin {{\mathcal {S}}}{{\mathcal {C}}}.\) Then, there exists an infinite codimensional closed subspace M of Y such that \(Q_NK\) is surjective. By the hypothesis, there exist an infinite codimensional closed subspace N such that \(M\subset N\) and a surjective operator \(V\in {\mathcal {L}}(X,M).\)
Observe that \(S=J_NV\in {\mathcal {L}}(X,Y)\) is not in \(\Phi _-.\) We prove that \(K\notin \Phi {\mathcal {C}}\) by showing that \([S,K] \in \Phi _-(X\times X, Y).\)
Indeed, note that \(R(S)=N.\) Moreover \(Q_NK\) surjective implies \(R(K)+N=Y.\) Since \(R([S,K])= R(S)+ R(K),\) [S, K] is surjective. \(\square \)
In the known examples in which \({{\mathcal {S}}}{{\mathcal {C}}}(X,Y)\ne P\Phi _-(X,Y)\) in [8, 10], the space X has a complemented subspace which is hereditarily indecomposable. Therefore, the question arises.
Question 4.3
Suppose that \(\Phi _-(X,Y)\ne \emptyset \) and \(X\times X\simeq X.\)
Is \({{\mathcal {S}}}{{\mathcal {C}}}(X,Y)=P\Phi _-(X,Y)?\)
A Banach space X is superprojective if each of its infinite codimensional closed subspaces is contained in some complemented infinite codimensional subspace. The spaces \(c_0,\) \(\ell _p\) \((1< p<\infty )\) and \(L_q(\mu )\) (\(1< q \le 2)\) are superprojective. See [7, 12] for further examples.
Corollary 4.1
Suppose that \(\Phi _-(X,Y) \ne \emptyset \) and the space Y is superprojective. Then, \(\Phi {\mathcal {C}}(X,Y)= {{\mathcal {S}}}{{\mathcal {C}}}(X, Y).\)
Proof
Every closed infinite codimensional subspace of Y is contained in an infinite codimensional complemented subspace N. Since \(\Phi _-(X,Y) \ne \emptyset ,\) there exists \(T\in {\mathcal {L}}(X,Y)\) with \(R(T)\supset N,\) and composing with the projection P on Y onto N we get \(R(PT)=N,\) and we can apply Theorem 4.1. \(\square \)
The following result is a refinement of the previous results in this section.
Theorem 4.2
Suppose that \(\Phi _-(X,Y) \ne \emptyset ,\) there exists a surjection from \(X \times X\) onto X, and every closed infinite codimensional subspace of Y is contained in a closed infinite codimensional subspace N which is isomorphic to a quotient of X. Then, \(P\Phi _-(X,Y)= {{\mathcal {S}}}{{\mathcal {C}}}(X,Y).\)
Proof
\(K\in {\mathcal {L}}(X,Y),\) \(K\notin {{\mathcal {S}}}{{\mathcal {C}}}.\) Then, there exists an infinite codimensional subspace N of Y such that \(Q_NK\) is surjective. By hypothesis, we can assume that there exists a surjective operator \(U:X\rightarrow N.\) Then, \(S=J_NU\in {\mathcal {L}}(X,Y)\) is not in \(\Phi _-.\) Moreover, [S, K] is surjective: \(R([S,K])= R(S)+R(K)= N+R(K)=Y,\) hence \([S,K]\in \Phi _-(X\times X,Y).\)
Let \(V:X \rightarrow X \times X\) be a surjection and let \(V_1,V_2 \in {\mathcal {L}}(X)\) such that \(V(x)=(V_1x,V_2x)\) for each \(x\in X.\) Then, \([S,K]V= SV_1+KV_2\in \Phi _-(X,Y).\) Since \(SV_1\notin \Phi _-\), we get \(KV_2\notin P\Phi _-;\) hence \(K\notin P\Phi _-\) by Proposition 2.1. \(\square \)
Corollary 4.2
If Y is separable, there exists a surjection from \(X\times X\) onto X, and X has a quotient isomorphic to \(\ell _1\), then \({{\mathcal {S}}}{{\mathcal {C}}}(X,Y)=\Phi _-(X,Y).\)
Proof
It is well known that every separable Banach space is isomorphic to a quotient of \(\ell _1.\) \(\square \)
The class \(\Phi {\mathcal {C}}(X,Y)\) is surjective in the following sense:
Proposition 4.3
Given \(K\in {\mathcal {L}}(X,Y)\) and a surjective operator \(Q\in {\mathcal {L}}(Z,X),\) if \(KQ \in \Phi {\mathcal {C}}(Z,Y)\), then \(K \in \Phi {\mathcal {C}}(X,Y).\)
Proof
Let \(S\in {\mathcal {L}}(X,Y)\) such that \([S,K] \in \Phi _-(X \times X, Y),\) and let \(Q:Z \rightarrow X\) be a surjective operator. Then, the operator \(Q\times Q\in {\mathcal {L}}(Z \times Z,X\times X)\) defined by \((Q\times Q)(a,b)=(Qa,Qb)\) is surjective. Thus, \([S,K](Q\times Q)= [SQ,KQ]\) is in \(\Phi _-(Z \times Z,Y).\) Since \(KQ \in \Phi {\mathcal {C}}(Z,Y)\), we obtain \(SQ \in \Phi _-(Z, Y),\) hence \(S \in \Phi _-(X,Y),\) and we conclude \(K \in \Phi {\mathcal {C}}(X,Y).\) \(\square \)
The dual space \((X\times X, \Vert \cdot \Vert _\infty )^*\) can be identified with \((X^*\times X^*, \Vert \cdot \Vert _1)\) in the obvious way. Hence, the conjugate operator \([S,T]^*\) can be identified with \((S^*,T^*).\) Indeed, for \(x^*\in X^*\) and \(x\in X\), we have
As a consequence, \([S,T]\in \Phi _-\) if and only if \((S^*,T^*)\in \Phi _+.\) Similarly, \((S,T)^*\) can be identified with \([S^*,T^*].\)
The following result describes the behavior of the classes of \(\Phi \)-singular and \(\Phi \)-cosingular operators under duality.
Proposition 4.4
Let \(K\in {\mathcal {L}}(X,Y).\)
-
1.
If \(K^* \in \Phi {\mathcal {S}}(Y^*, X^*)\), then \(K \in \Phi {\mathcal {C}}(X,Y).\)
-
2.
If \(K^* \in \Phi {\mathcal {C}}(Y^*, X^*)\), then \(K \in \Phi {\mathcal {S}}(X,Y).\)
Proof
(1) Let \(S\in {\mathcal {L}}(X,Y)\) such that \([S,K] \in \Phi _-(X \times X, Y).\) Then, \([S,K]^* \in \Phi _+.\) Since \([S,K]^*\equiv (S^*,K^*),\) we have \((S^*,K^*) \in \Phi _+(Y^*, X^* \times X^*),\) and from \(K^* \in \Phi {\mathcal {S}}(Y^*, X^*)\) we obtain \(S^* \in \Phi _+(Y^*, X^*);\) therefore, \(S \in \Phi _-(X, Y),\) and hence \(K \in \Phi {\mathcal {C}}(X,Y).\)
The proof of (2) is similar. \(\square \)
References
Aiena, P., González, M.: On inessential and improjective operators. Stud. Math. 131, 271–287 (1998)
Aiena, P., González, M., Martínez-Abejón, A.: Characterizations of strictly singular and strictly cosingular operators by perturbation classes. Glasg. Math. J. 54, 87–96 (2011)
Argyros, S.A., Felouzis, V.: Interpolating hereditarily indecomposable Banach spaces. J. Am. Math. Soc. 13, 243–294 (2000)
Bessaga, C., Pełczyński, A.: Banach spaces non-isomorphic to their Cartesian squares. I. Bull. Acad. Pol. Sci. Sér. Sci. Math. Astron. Phys. 8, 77–80 (1960)
Caradus, S., Pfaffenberger, W., Yood, B.: Calkin Algebras and Algebras of Operators in Banach Spaces. Lecture Notes in Pure and Applies Mathematics, M. Dekker, New York (1974)
Friedman, T.L.: Relating strictly singular operators to the condition \(X < Y \; mod \;({\cal{S} }, {\cal{T} })\) and resulting perturbations. Analysis (Munich) 22, 347–354 (2002)
Galego, E.M., González, M., Pello, J.: On subprojectivity and superprojectivity. Results Math. 71, 1191–1205 (2017)
Giménez, J., González, M., Martínez-Abejón, A.: Perturbation of semi-Fredholm operators on products of Banach spaces. J. Oper. Theory 68, 501–514 (2012)
Gohberg, I.C., Markus, A.S., Feldman, I.A.: Normally solvable operators and ideals associated with them. Bul. Akad. Štiince RSS Moldoven 10(76), 51–70 (1960). [Translation: Am. Math. Soc. Transl. (2) 61, 63–84 (1967)]
González, M.: The perturbation classes problem in Fredholm theory. J. Funct. Anal. 200, 65–70 (2003)
González, M., Martínez-Abejón, A., Salas-Brown, M.: Perturbation classes for semi-Fredholm operators on subprojective and superprojective spaces. Ann. Acad. Sci. Fenn. Math. 36, 481–491 (2011)
González, M., Pello, J.: Superprojective Banach spaces. J. Math. Anal. Appl. 437, 1140–1151 (2016)
González, M., Pello, J., Salas-Brown, M.: Perturbation classes of semi-Fredholm operators in Banach lattices. J. Math. Anal. Appl. 420, 792–800 (2014)
González, M., Pello, J., Salas-Brown, M.: The perturbation classes problem for subprojective and superprojective Banach spaces. J. Math. Anal. Appl. 489, 124191 (2020)
González, M., Salas-Brown, M.: Perturbation classes for semi-Fredholm operators in \(L_p(\mu )\)-spaces. J. Math. Anal. Appl. 370, 11–17 (2010)
Gowers, W.T., Maurey, B.: The unconditional basic sequence problem. J. Am. Math. Soc. 6, 851–874 (1993)
Gowers, W.T., Maurey, B.: Banach spaces with small spaces of operators. Math. Ann. 307, 543–568 (1997)
Oikhberg, T., Spinu, E.: Subprojective Banach spaces. J. Math. Anal. Appl. 424, 613–635 (2015)
Pietsch, A.: Operator Ideals. North-Holland, Amsterdam (1980)
Semadeni, Z.: Banach spaces non-isomorphic to their Cartesian squares. II. Bull. Acad. Pol. Sci. Sér. Sci. Math. Astron. Phys. 8, 81–84 (1960)
Weis, L.: Perturbation classes of semi-Fredholm operators. Math. Z. 178, 429–442 (1981)
Whitley, R.J.: Strictly singular operators and their conjugates. Trans. Am. Math. Soc. 113, 252–261 (1964)
Acknowledgements
Supported in part by MINCIN Project PID2019-103961.
Funding
Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors have no competing interests to declare that are relevant to the content of this article.
Additional information
Communicated by Qingxiang Xu.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
González, M., Salas-Brown, M. Two classes of operators related to the perturbation classes problem. Adv. Oper. Theory 8, 41 (2023). https://doi.org/10.1007/s43036-023-00270-0
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s43036-023-00270-0