International Journal of Mathematical Analysis Vol. 11, 2017, no. 18, 883-890 HIKARI Ltd, www.m-hikari.com https://doi.org/10.12988/ijma.2017.77103 Remarks on Strictly Singular Operators Ersin Kızgut Çankaya University, Institute of Natural and Applied Sciences Öğretmenler Cd No.14 06530 Ankara, Turkey Murat Yurdakul Middle East Technical University, Department of Mathematics Dumlupınar Blv No. 1 06800 Ankara, Turkey Copyright 2017 Ersin Kızgut and Murat Yurdakul. This article is distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Abstract A continuous linear operator T : E F is called strictly singular if it cannot be invertible on any infinite dimensional closed subspace of its domain. In this note we discuss sufficient conditions and consequences of the phenomenon LB(E, F ) = L s (E, F ), which means that every continuous linear bounded operator defined on E into F is strictly singular. Mathematics Subject Classification: 46A03, 46A11, 46A32, 46A45 Keywords: Strictly singular operators 1. Introduction and first results A continuous linear operator mapping a locally convex space (lcs) E into a lcs F is said to be bounded if it maps a neighborhood of the origin of E into a bounded subset of F. If E or F is a normed space, then every continuous linear operator between E and F is bounded. A continuous linear operator T : E F is called strictly singular if it fails to be invertible on any infinite dimensional closed subspace of E. On the class of Banach spaces, Kato [13] introduced strictly singular operators in connection with perturbation theory of Fredholm operators. A compact operator is strictly singular, while the converse
884 Ersin Kızgut and Murat Yurdakul is not true in general. In terms of lcs s, van Dulst first [26], [27] studied strictly singular operators on Ptak (or B-complete) spaces and generalized Hilbert spaces. Wrobel [28] characterized strictly singular operators on lcs s for the class of B r -complete spaces. L s (X, Y ) is an operator ideal, if the pair (X, Y ) belongs to the class of Banach spaces. As proved in [5], this is not the case when it belongs to the class of general lcs s. However, each of the classes of bounded strictly singular operators LB s (E, F ) and of compact operators L c (E, F ) forms an operator ideal in lcs s. By [22], for Fréchet spaces E and F, LB s (E, F ) = L s (E, F ) if F has the property (y). If E contains l 1 as a quotient, then LB s (E, F ) L s (E, F ). Lemma 1. Let E and F be lcs s where E is B r -complete. Then, L s (E, F ) forms an operator ideal. Proof. Suppose that T : E F and S : E F are strictly singular operators. Then, for any M E, by [28, Theorem 1-IV], find N M such that T N is precompact. Then find P N such that S P is precompact. The ideal property of precompact operators on lcs s yields the result. We give the following proposition as an application of operator ideal property of strictly singular operators on B r -complete lcs s. It is a generalization of [1, Problem 4.5.2], and in particular, it is also true when re-stated for bounded strictly singular operators acting on general lcs s. Proposition 1. Let E := n E i and F := m F j be lcs s where E is B r - complete. Then, T : E F is strictly singular iff each of T ij : E i F j is strictly singular for each i = 1, 2,..., n and for each j = 1, 2,..., m. Proof. Assume that each T ji is strictly singular. Let π i : E E i be the canonical projection and define ρ j : F j F by ρy j = 0 0 0 y j 0 0, for which y j is the j-th summand. Consider E π i T ji ρ j E i Fj F, and write S ji := ρ j T ji π i. Then S ji (x 1 x 2 x n ) = 0 0 0 T ji x i 0 0, where T ji is the j-th summand. By Lemma 1 and rewriting T = n m i=1 j=1 S ji, T is strictly singular. For the converse, let T L s (E, F ), and suppose that the operator T ji is not strictly singular for some i, j, and for M E, r I and s J, N rs (T M ) := sup{q s (T x) : p r (x) 1, x M}. Then by [20, Theorem 2.1], for any M E i and for some s J, N rs (T N ) > ε, for all r I. If we write M := {0} {0} M {0} {0} where M places in the i-th summand, M is a vector subspace of X. N rs (τ M) > ε, for all r I. But, that contradicts the assumption T is strictly singular. Every compact operator is clearly bounded. If we replace compactness with strict singularity, the results in [22], [29] and [25] show that a strictly singular operator T : E F between lcs s cannot be unbounded, since such an assumption leads to a contradiction with the well-known result of Bessaga, Pelczynski, and Rolewicz [2]. In Section 2 we survey the additional hypotheses
Remarks on strictly singular operators 885 under which a bounded operator between two lcs s is automatically strictly singular. 2. Strict singularity of bounded operators In this part, the additional hypotheses for strict singularity of a bounded operator will be introduced. Our starting point will be the class of Banach spaces. Applications of [7, Lemma 2] shall yield that some of these results are meaningful on the class of lcs s. In a particular case, to obtain a characterization (Theorem 3) is possible. The most commonly known non-trivial example for L(X, Y ) = L s (X, Y ) in Banach spaces is when X = l p and Y = l q such that 1 p < q <. This phenomenon was essential in the isomorphic classification of Cartesian products of power series spaces [7], and in the problem whether the sum of two complemented subspaces is also complemented [15] (if X and Y be complemented subspaces of E then, X + Y is a complemented subspace of E if LB(X, Y ) = L s (X, Y )). A Banach space X is said to possess the Schur property (SP) if every weakly convergent sequence in X converges in norm. X is called almost reflexive, if every bounded sequence is weakly Cauchy. X is called weakly sequentially complete (wsc) if every weakly Cauchy sequence in X is weakly convergent. X is said to have the Dieudonné property (DP) if operators on X mapping weakly Cauchy sequences into weakly convergent sequences are weakly compact. A pair of Banach spaces (X, Y ) is called totally incomparable if there exists no Banach space Z which is isomorphic to a subspace of X and to a subspace of Y. A property P on a Banach space X is called hereditary if it is enjoyed by every M X. X is said to have nowhere P if it has no subspace having the property P. By L w and L v we denote the classes of weakly compact and completely continuous (or fully complete) operators, respectively. A Banach space X is said to have the Dunford-Pettis property (DPP) if L w (X, Y ) L v (X, Y ), for any Banach space Y. X is said to have the reciprocal Dunford- Pettis property (rdpp), if L v (X, Y ) L w (X, Y ). Lemma 2. Let X and Y be Banach spaces. 1. Let X be almost reflexive. Then, for any wsc Banach space Y, L(X, Y ) = L s (X, Y ). 2. Let X have DP, and let Y be wsc. Then, L(X, Y ) = L w (X, Y ). 3. Let Y be almost reflexive. Then L(X, Y ) = L v (X, Y ) implies L(X, Y ) = L s (X, Y ). 4. Let X be a Banach space with SP. Then, for every M X, l 1 M. Proof. 1. Since X is almost reflexive, if (x n ) is a bounded sequence in X, then (T x n ) has a weakly Cauchy sequence in Y. But Y is wsc, that is, every weakly Cauchy sequence converges weakly in Y. Therefore, T is weakly compact.
886 Ersin Kızgut and Murat Yurdakul 2. Let (x n ) X be weakly Cauchy, and let T L(X, Y ). Then (T x n ) is weakly Cauchy in Y. Since Y is assumed to be wsc, (T x n ) converges weakly. But X has DP, so T L w (X, Y ). 3. See [16, Theorem 1.7]. 4. Let X have the SP and suppose there exists M X not containing l 1. Then, any bounded sequence (x n ) in M, has a weakly Cauchy subsequence since M is equivalently almost reflexive. However, M inherits SP. Then the weakly Cauchy subsequence of (x k ) converges in X. Therefore, M is finite dimensional. Contradiction. Theorem 1. Let X and Y be Banach spaces. Each of the following implies L(X, Y ) = L s (X, Y ). 1. X and Y are totally incomparable. 2. X is nowhere reflexive, Y is reflexive. 3. X is nowhere reflexive, Y is quasi-reflexive. 4. X is almost reflexive and nowhere reflexive, Y is wsc (see Example 1). 5. Y has hereditary P, X has nowhere P. 6. Y is almost reflexive, X is hereditarily-l 1. 7. X has SP, Y is almost reflexive. 8. X is reflexive, Y has SP. 9. X has the hereditary DPP, Y is reflexive. 10. L(X, Y ) = L w (X, Y ) and X has DPP. 11. L(X, Y ) = L v (X, Y ) and X has rdpp. 12. X is a Grothendieck space with DPP, Y is separable. 13. X has both DP and DPP, Y is wsc (see Example 1). Proof. 1. Suppose T : X Y is a non-strictly singular operator. So find M X on which M T (M) Y. Since X and Y are totally incomparable, this is impossible. 2. See [10, Theorem b]. This result is generalized in part 5. 3. Suppose there exists a non-strictly singular operator T L(X, Y ). Then, T is an isomorphism when restricted to M X, so M is quasi-reflexive. However, by [12, Lemma 2] there exists a reflexive N M. This contradicts the assumption X is nowhere reflexive. 4. By part 1 of Lemma 2, L(X, Y ) = L w (X, Y ). Now let T : X Y which has a bounded inverse on M X. If (x n ) is a bounded sequence in M, then there exists (T x kn ) a weakly convergent subsequence of (T x n ) in Y. Hence (x kn ) is weakly convergent in M, since T has a bounded inverse on M. Thus, every bounded sequence in M has a weakly convergent subsequence in M. So M is reflexive. Contradiction. 5. For some M X suppose there exists T : X Y such that M T (M). But T (M) inherits P. Hence M has P. This contradicts X has nowhere P. Now let S : Y X be such that N S(N) X for some N Y. Since X has nowhere P, S(N) does not enjoy P. Contradiction.
Remarks on strictly singular operators 887 6. A particular case of part 5. 7. Any operator T with range Y maps bounded sequences into weakly Cauchy sequences, since Y is almost reflexive. On the other hand, any such T defined on X maps weakly Cauchy sequences into norm convergent sequences by the SP. That implies L(X, Y ) = L v (X, Y ). Hence by part 3 of Lemma 2 the result follows. 8. Let T L(X, Y ) have a bounded inverse on some M X, that is, M T (M). Since Y have SP, it also has the hereditary DPP [6]. Hence so does M. But M is reflexive. By [14, Theorem 2.1], reflexive spaces have nowhere DPP. Contradiction. 9. Let M X on which an arbitrary operator T : X Y has a bounded inverse. Then, M T (M). So, M is reflexive. Hence M cannot have DPP. Contradiction. 10. Since X has DPP, L(X, Y ) L v (X, Y ). Then, by [16, Theorem 2.3], the result follows. 11. Since X has the rdpp and L(X, Y ) = L v (X, Y ), T L v L w. By [16, Theorem 2.3], we are done. 12. By [21, Theorem 4.9], any such operator T : X Y is weakly compact. Since X has the DPP, T is completely continuous. By part 10, we reach the result. 13. By part 2 of Lemma 2, L(X, Y ) = L w (X, Y ). X possesses the DPP, so L(X, Y ) = L v (X, Y ). By [16, Theorem 2.3], the proof is completed. Example 1. Note that the non-reflexive space c 0 is almost reflexive. Suppose there exists a reflexive subspace E of c 0. Since c 0 fails SP, it is not isomorphic to any subspace of E. But this contradicts [17, Proposition 2.a.2]. The space C(K), where K is a compact Hausdorff space enjoys both DP [11], and DPP [14]. Corollary 1. Let X, Y, W, Z be Banach spaces. Then, 1. If X, Y, Z have SP and W is almost reflexive, every operator defined on (X π Y ) into W π Z is strictly singular. 2. If X and Y are reflexive spaces one of which having the approximation property, L(X, Y ) = L c (X, Y ), W and Z have SP, then every operator defined on X π Y into W ˇ ε Z is strictly singular. 3. If X is almost reflexive and Y is almost reflexive and has DPP, then every operator defined on X π Y into l 1 is strictly singular. Proof. 1. By [16, Corollary 1.6], L(W, Z ) = L c (W, Z ). So by [9, Theorem 3] we deduce W π Z is almost reflexive. On the other hand, by [23, Theorem 3.3(b)] we reach that L(X, Y ) has SP. But in [24] it is proved that L(X, Y ) (X π Y ). So (X π Y ) has SP. Therefore, Thorem 1 part 7 yields the result.
888 Ersin Kızgut and Murat Yurdakul 2. By [24, Theorem 4.21], X π Y is reflexive. By [18], SP respects injective tensor products. So W ˇ ε Z has SP. Then, part 8 of Theorem 1 finishes the proof. 3. By [6], Y has SP. Then by [16, Corollary 1.6], L(X, Y ) = L c (X, Y ). Hence, [9, Theorem 3] yields that X π Y is almost reflexive. It is clear that every operator defined from an almost reflexive space into l 1 is strictly singular. Now let us turn our attention to general class of lcs s. Let P be a class of Banach spaces having a certain hereditary property P. Then, s(p) [4] is defined by the set of lcs s E with local Banach spaces E U P for which E U is the completion of the normed space obtained by E/p 1 u (0), where U U (E) and p u its gauge functional. For instance, by s(x), we denote the class of lcs s such that each of their local Banach spaces are hereditarily l 1. A lcs E is called locally Rosenthal [3], if it can be written as a projective limit of Banach spaces each of which contains no isomorpic copy of l 1. A quasinormable lcs E admitting no isomorphic copies of l 1 is locally Rosenthal. By s(v), we denote the class of lcs s with local Banach spaces each of which having SP. A lcs E is called infra-schwartz (or Komura) if each of its local Banach spaces is reflexive. An infra-schwartz space turns out to be locally Rosenthal, as proved in [3]. In [7], it is shown that the relation L(X k, Y m ) = L s (X k, Y m ) for every k, m is sufficient for L(X, Y ) = L s (X, Y ), where X = proj lim k X k and Y = proj lim m Y m. Let λ 1 (A) (d 2 ), and λ p (A) (d 1 ) as in [8]. Then, by [30], L(λ 1 (A), λ p (A)) = LB(λ 1 (A), λ p (A)). For 1 p <, we know that λ p (A) = proj lim l p (a n ). Since l p (a n ), 1 < p < has no subspace isomorphic to l 1, L(l 1, l p ) = L s (l 1, l p ). Then, by [7, Lemma 2], L(λ 1 (A), λ p (A)) = L s (λ 1 (A), λ p (A)). Resting on the same argument, to obtain several sufficient conditions for L(E, F ) = L s (E, F ) is possible for the class of general lcs s. Theorem 2. Let E, F be lcs s. Each of the following implies LB(E, F ) = L s (E, F ). 1. E s(x) and F is locally Rosenthal. 2. E s(v) and F is a quasinormable Fréchet space. 3. E is infra-schwartz and F s(v) 4. E s(p ) and F s(p) Proof. 1. Since F is locally Rosenthal, there exists a family of Banach spaces {F m } each of which does not contain an isomorphic copy of l 1 such that F = proj lim F m. Because E s(x), there exists a family of Banach spaces {E k } such that every M k E k contains a subspace isomorphic to l 1. By part 6 of Theorem 1, any linear operator T mk : E k F m is strictly singular. Making use of [7, Lemma 2], we reach the result. 2. By [19, Theorem 6], F is locally Rosenthal. Since E s(v), by part 4 of Lemma 2, E s(x). Then, by part 1, we are done.
Remarks on strictly singular operators 889 3. Since E is infra-schwartz, any of its local Banach spaces E k is reflexive. The assumption on F completes the conditions in part 8 of Theorem 1. Combined with [7, Lemma 2], we are done. 4. Since E s(p ), one may rewrite E = proj lim k E k, where each E k has no subspace having property P. Similarly, F = proj lim m F m where each F m is hereditarily P. Hence, by part 5 of Theorem 1, L(E k, F m ) = L s (E k, F m ) for every k, m. Applying [7, Lemma 2], we obtain LB(E, F ) = L s (E, F ). Theorem 3. Let (E, F, G) be a triple of Fréchet spaces satisfying the following 1. Every subspace of E contains a subspace isomorphic to G. 2. F has no subspace isomorphic to G. Then, LB(E, F ) = L s (E, F ). Let F have continuous norm in addition. Then, (2) is also necessary if F is a Fréchet-Montel space. Proof. The sufficiency part is very similar to the proof of Theorem 2 part 4. For necessity, let E be a Fréchet space and let F be an (FM)-space admitting a continuous norm. Let any linear operator T : E F be strictly singular. Then, by [29, Proposition 1], it is bounded. Now let there exist N Y which is isomorphic to G. Then I N : N G is bounded, hence compact. Then N is finite dimensional. Contradiction. References [1] Y. A. Abramovich and C. D. Aliprantis, Problems in Operator Theory, Graduate Studies in Mathematics, American Mathematical Society, 2002. https://doi.org/10.1090/gsm/051 [2] C. Bessaga, A. Pelczynski and S. Rolewicz, On diametral approximative dimension and linear homogeneity of F-spaces, Bull. Acad. Polon. Sci., 9 (1961), 677 683. [3] C. Boyd and M. Venkova, Grothendieck space ideals and weak continuity of polynomials on locally convex spaces, Monatsch. Math., 151 (2007), 189 200. https://doi.org/10.1007/s00605-007-0483-3 [4] J. M. F. Castillo and M. A. Simões, Some problems for suggested thinking in Fréchet space theory, Extr. Math., 6 (1991), 96 114. [5] S. Dierolf, A note on strictly singular and strictly cosingular operators, Indag. Math., 84 (1981), 67 69. https://doi.org/10.1016/1385-7258(81)90018-4 [6] J. Diestel, A survey of results related to the Dunford-Pettis property, Contemp. Math., 2 (1980), 15 60. https://doi.org/10.1090/conm/002/621850 [7] P. B. Djakov, S. Önal, T. Terzio~glu and M. Yurdakul, Strictly singular operators and isomorphisms of Cartesian products of power series spaces, Arch. Math., 70 (1998), 57 65. https://doi.org/10.1007/s000130050165 [8] M. M. Dragilev, On regular basis in nuclear spaces, Math. Sbornik, 68 (1965), 153 175. [9] G. Emmanuelle, Banach spaces in which Dunford-Pettis sets are relatively compact, Arch. Math., 58 (1992), 477 485. https://doi.org/10.1007/bf01190118 [10] S. Goldberg and E. O. Thorp, On some open questions concerning strictly singular operators, Proc. Amer. Math. Soc., 14 (1963), 334 336. https://doi.org/10.1090/s0002-9939-1963-0145361-3 [11] A. Grothendieck, Sur les applications lineares faiblement compactes d espaces du type C(K), Canad. J. Math., 5 (1953), 129 173. https://doi.org/10.4153/cjm-1953-017-4
890 Ersin Kızgut and Murat Yurdakul [12] R. Herman and R. J. Whitley, An example concerning reflexivity, Studia Math., 28 (1967), 289 294. https://doi.org/10.4064/sm-28-3-289-294 [13] T. Kato, Perturbation theory for nullity, deficiency and other quantities of linear operators, J. Analyse Math., 6 (1958), 261 322. https://doi.org/10.1007/bf02790238 [14] M. C. Kester, The Dunford-Pettis Property, PhD Thesis, Oklahoma State University, 1972. [15] E. Kızgut and M. Yurdakul, Some stability results in projective tensor products, 2015. arxiv:1504.03254 [16] E. Lacey and R. J. Whitley, Conditions under which all the bounded linear maps are compact, Math. Ann., 158 (1965), 1 5. https://doi.org/10.1007/bf01370391 [17] J. Lindenstrauss and L. Tzafriri, Classical Banach Spaces I and II, Springer-Verlag, 1996. [18] F. Lust, Produits tensoriels injectifs d espaces de sidon, Colloq. Math., 32 (1975), 286 289. https://doi.org/10.4064/cm-32-2-285-289 [19] M. A. Miñarro, A characterization of quasinormable Köthe sequence spaces, Proc. Amer. Math. Soc., 123 (1995), 1207 1212. https://doi.org/10.1090/s0002-9939-1995-1227526-3 [20] C. G. Moorthy and C. T. Ramasamy, Characterizations of strictly singular and strictly discontinuous operators on locally convex spaces, Int. Journ. of Math. Analysis, 4 (2010), 1217 1224. [21] T. J. Morrison, Functional Analysis: An Introduction to Banach Space Theory, John Wiley and Sons, 2001. https://doi.org/10.1002/9781118032992 [22] S. Önal and M. Yurdakul, A note on strictly singular operators, Turk. J. Math., 15 (1991), no. 1, 42 47. [23] R. Ryan, The Dunford-Pettis property and projective tensor products, Bull. Polish. Acad. Sci. Math., 35 (1987), 785 792. [24] R. Ryan, Introduction to the Tensor Products of Banach Spaces, Springer-Verlag, 2002. https://doi.org/10.1007/978-1-4471-3903-4 [25] T. Terzioglu and M. Yurdakul, Restrictions of unbounded continuous linear operators on Fréchet spaces, Arch. Math., 46 (1986), 547 550. https://doi.org/10.1007/bf01195024 [26] D. van Dulst, Perturbation theory and strictly singular operators in locally convex spaces, Studia Math., 38 (1970), 341 372. https://doi.org/10.4064/sm-38-1-341-372 [27] D. van Dulst, On strictly singular operators, Comp. Math., 23 (1971), 169 183. [28] V. V. Wrobel, Strikt singuläre Operatoren in lokalkonvexen Räumen, Math. Nachr., 83 (1978), 127 142. https://doi.org/10.1002/mana.19780830110 [29] M. Yurdakul, A remark on a paper of J. Prada, Arch. Math., 61 (1993), 385 390. https://doi.org/10.1007/bf01201455 [30] V. Zahariuta, On the isomorphism of Cartesian products of locally convex spaces, Studia Math., 46 (1973), 201 221. https://doi.org/10.4064/sm-46-3-201-221 Received: August 11, 2017; Published: September 28, 2017