1 Introduction

In this paper Dunford–Pettis sets and limited sets are used to characterize the classes of weak Dunford–Pettis and \(\hbox {weak}^*\) Dunford–Pettis operators. The classes of weak p-convergent and \(\hbox {weak}^*\)p-convergent operators are also studied.

Our major results are Theorems 2, 6, 14, and 18. As consequences, we obtain equivalent characterizations of Banach spaces with the Dunford–Pettis property, \(DP^*\)-property, Dunford–Pettis property of order p, and \(DP^*\)-property of order p. We generalize some results in [1, 5, 11, 15, 18].

2 Definitions and notation

Throughout this paper, X and Y will denote real Banach spaces. The unit ball of X will be denoted by \(B_X\) and \(X^*\) will denote the continuous linear dual of X. An operator \(T: X \rightarrow Y\) will be a continuous and linear function. The space of all operators from X to Y will be denoted by L(XY).

The operator T is completely continuous (or Dunford–Pettis) if T maps weakly convergent sequences to norm convergent sequences.

A Banach space X has the Dunford–Pettis property (DPP) if every weakly compact operator T with domain X is completely continuous. Equivalently, X has the DPP if and only if \(x_n^*(x_n)\rightarrow 0\) for all weakly null sequences \((x_n)\) in X and \((x_n^*)\) in \(X^*\) [8, Theorem 1]. Schur spaces, C(K) spaces, and \(L_1(\mu )\) spaces have the DPP. The reader can check [7,8,9], and [2] for a guide to the extensive classical literature dealing with the DPP, equivalent formulations of the preceding definitions, and undefined notation and terminology.

A subset A of X is called a Dunford–Pettis (DP) subset (resp. limited subset) of X if each weakly null (resp. \(w^*\) -null) sequence \((x_n^*)\) in \(X^*\) tends to 0 uniformly on A [2] (resp. [3, 24]); i.e.

$$\begin{aligned} \sup _{x \in A} |x_n^* (x) | \rightarrow 0. \end{aligned}$$

If A is a limited subset of X, then T(A) is relatively compact for any operator \(T:X\rightarrow c_0\) [3, p. 56], [24, p. 23]. The subset A of X is a DP subset of X if and only if T(A) is relatively compact whenever \(T: X \rightarrow Y\) is a weakly compact operator [2] if and only if T(A) is relatively compact whenever \(T: X \rightarrow Y\) is an operator with weakly precompact adjoint [20].

A bounded subset S of X is said to be weakly precompact provided that every sequence from S has a weakly Cauchy subsequence. Every DP subset of X is weakly precompact [2]. Since any limited set is a DP set, any limited set is weakly precompact. An operator \(T:X\rightarrow Y\) is called weakly precompact (or almost weakly compact) if \(T(B_X)\) is weakly precompact.

A Banach space X has the \(DP^*\)-property (\(DP^*P\)) if all weakly compact sets in X are limited [4, 5, 21]. The space X has the \(DP^*P\) if and only if \(x_n^*(x_n)\rightarrow 0\) for all weakly null sequences \((x_n)\) in X and \(w^*\)-null sequences \((x_n^*)\) in \(X^*\) [16]. If X has the \(DP^*P\), then it has the DPP. If X is a Schur space or if X has the DPP and the Grothendieck property (weak and \(\hbox {weak}^*\) convergence of sequences in \(X^*\) coincide), then X has the \(DP^*P\).

3 Weak Dunford–Pettis operators and \(\hbox {weak}^*\) Dunford–Pettis operators

An operator \(T:X\rightarrow Y\) is called weak Dunford–Pettis [1, p. 349] if \(\langle T(x_n), y_n^*\rangle \rightarrow 0\), whenever \((x_n)\) is a weakly null sequence in X and \((y_n^*)\) is a weakly null sequence in \(Y^*\). An operator \(T:X\rightarrow Y\) is called \(\hbox {weak}^*\)Dunford–Pettis [11] if \(\langle T(x_n), y_n^*\rangle \rightarrow 0\), whenever \((x_n)\) is a weakly null sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\).

In this section we give some characterizations of weak Dunford–Pettis and \(\hbox {weak}^*\) Dunford–Pettis operators.

Observation 1 If \(T:X\rightarrow Y\) is an operator, then \(T(B_X)\) is a DP (resp. limited) subset of Y if and only if \(T^* : Y^* \rightarrow X^*\) is completely continuous (resp. \(T^*\) is \(w^*\)-norm sequentially continuous).

To see this, note that \(T(B_X)\) is a DP (resp. limited) subset of Y if and only if

$$\begin{aligned} 0=\lim _n \sup \{|\langle y_n^*, T(x)\rangle |: x\in B_X\} = \lim _n \sup \{|\langle T^*(y_n^*), x\rangle |: x\in B_X\} =\lim _n \Vert T^*(y_n^*)\Vert \end{aligned}$$

for each weakly null (resp. \(w^*\)-null) sequence \((y_n^*)\) in \(Y^*\); that is, \(T^* : Y^* \rightarrow X^*\) is completely continuous (resp. \(T^*\) is \(w^*\)-norm sequentially continuous).

Theorem 1

[1, Theorem 5.99, p. 351] Let \(T:X\rightarrow Y\) be an operator. The following statements are equivalent:

  1. (1)

    T is a weak Dunford–Pettis operator.

  2. (2)

    T carries weakly compact subsets of X to Dunford–Pettis subsets of Y.

  3. (3)

    If \(S:Y\rightarrow Z\) is a weakly compact operator, then \(ST:X\rightarrow Z\) is completely continuous, for any Banach space Z.

We are now giving our first major result. It gives characterizations of weak Dunford–Pettis operators and generalizes [1, Theorem 5.99, p. 351].

Theorem 2

Let X and Y be Banach spaces, and let \(T:X\rightarrow Y\) be an operator. The following statements are equivalent.

  1. (1)

    T is a weak Dunford–Pettis operator.

  2. (2)

    T carries weakly precompact subsets of X to Dunford–Pettis subsets of Y.

  3. (3)

    For all Banach spaces Z, if \(S:Y\rightarrow Z\) has a weakly precompact adjoint, then \(ST:X\rightarrow Z\) is completely continuous.

  4. (4)

    If \(S:Y\rightarrow c_0\) has a weakly precompact adjoint, then \(ST:X\rightarrow c_0\) is completely continuous.

  5. (5)

    If \((x_n)\) is a weakly null sequence in X and \((y_n^*)\) is a weakly Cauchy sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

Proof

\((1) \Rightarrow (2)\) Let A be a weakly precompact subset of X. Suppose by contradiction that T(A) is not a Dunford–Pettis subset of Y. Suppose that \((y_n^*)\) is a weakly null sequence in \(Y^*\), \((x_n)\) is a sequence in A, and \(\epsilon >0\) such that \(|\langle y_n^*, T(x_n) \rangle |>\epsilon \) for all n. Without loss of generality assume that \((x_n)\) is weakly Cauchy.

Let \(n_1=1\) and choose \(n_2>n_1\) so that \(|\langle y_{n_2}^*, T(x_{n_1})\rangle |<\epsilon /2\). We can do this since \((y_n^*)\) is \(w^*\)-null. Continue inductively. Choose \(n_k>n_{k-1}\) so that \(|\langle y_{n_k}^*, T(x_{n_{k-1}}) \rangle |<\epsilon /2\). Since T is a weak Dunford–Pettis operator, \(\langle y^*_{n_k}, T(x_{n_k}-x_{n_{k-1}})\rangle \rightarrow 0\). However,

$$\begin{aligned} |\langle y_{n_k}^*, T(x_{n_k}-x_{n_{k-1}})\rangle |\ge |\langle y_{n_k}^*, T(x_{n_k})\rangle |-|\langle y_{n_k}^*, T(x_{n_{k-1}})\rangle |> \epsilon /2, \end{aligned}$$

a contradiction.

\((2) \Rightarrow (3)\) Let \(S:Y\rightarrow Z\) be an operator such that \(S^*:Z^*\rightarrow Y^*\) is weakly precompact. Suppose \((x_n)\) is a weakly null sequence in X. Since \(\{T(x_n):n\in {\mathbb {N}}\}\) is a Dunford–Pettis set in Y and \(S^*\) is weakly precompact, \(\{ST(x_n):n\in {\mathbb {N}}\}\) is relatively compact [20, Corollary 4]. Then \(\Vert ST(x_n)\Vert \rightarrow 0\), and thus \(ST:X\rightarrow Z\) is completely continuous.

\((3) \Rightarrow (4)\) and \((5) \Rightarrow (1)\) are obvious.

\((4) \Rightarrow (1)\) Suppose \((x_n)\) is weakly null in X and \((y_n^*)\) is weakly null in \(Y^*\). Define \(S:Y\rightarrow c_0\) by \(S(y)=(y_i^*(y))\). Then \(S^*:\ell _1\rightarrow Y^*\), \(S^*(b)=\sum b_i y_i^*\). Note that \(S^*\) maps \(B_{\ell _1}\) into the closed and absolutely convex hull of \(\{y_i^*:i\in {\mathbb {N}}\}\), which is relatively weakly compact [9, p. 51]. Then \(S^*\) is weakly compact. Hence \(ST:X\rightarrow c_0\) is completely continuous. Therefore \(\langle T(x_n), y_n^*\rangle \le \Vert ST(x_n)\Vert =\sup _i |\langle y_i^*, T(x_n)\rangle |\rightarrow 0 \), and T is weak Dunford–Pettis.

\((1) \Rightarrow (5)\) Suppose that \((x_n)\) is a weakly null sequence in X, \((y_n^*)\) is a weakly Cauchy sequence in \(Y^*\), and \(\langle y_n^*, T(x_n) \rangle \not \rightarrow 0\). Without loss of generality suppose that \(|\langle y_n^*, T(x_n) \rangle |> \epsilon \) for each \(n\in {\mathbb {N}}\), for some \(\epsilon > 0\).

Let \(n_1=1\) and choose \(n_2>n_1\) so that \(|\langle y_{n_1}^*, T(x_{n_2})\rangle |<\epsilon /2\). We can do this since \((T(x_n))\) is weakly null. Continue inductively. Choose \(n_{k+1}>n_{k}\) so that \(|\langle y_{n_k}^*, T(x_{n_{k+1}})\rangle |<\epsilon /2\). Since T is weak Dunford–Pettis, \(\langle y^*_{n_{k+1}}-y_{n_k}^*, T(x_{n_{k+1}}) \rangle \rightarrow 0\). Since

$$\begin{aligned} |\langle y^*_{n_{k+1}}-y_{n_k}^*, T(x_{n_{k+1})}\rangle |\ge |\langle y_{n_{k+1}}^*, T(x_{n_{k+1}})\rangle | -|\langle y_{n_k}^*, T(x_{n_{k+1}})\rangle |>\epsilon /2, \end{aligned}$$

we have a contradiction. \(\square \)

Corollary 3

Let X and Y be Banach spaces, and let \(T:X\rightarrow Y\) be an operator. The following statements are equivalent.

  1. (i)

    T is a weak Dunford–Pettis operator.

  2. (ii)

    For all Banach spaces Z, if \(S:Z\rightarrow X\) is a weakly precompact operator, then \(TS:Z\rightarrow Y\) has a completely continuous adjoint.

  3. (iii)

    If \(S:\ell _1\rightarrow X\) is a weakly precompact operator, then \(TS:\ell _1\rightarrow Y\) has a completely continuous adjoint.

  4. (iv)

    If \((x_n)\) is a weakly Cauchy sequence in X and \((y_n^*)\) is a weakly null sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

Proof

\(\mathrm{(i)} \Rightarrow \mathrm{(ii)}\) Let \(S:Z\rightarrow X\) be a weakly precompact operator. Then \(TS(B_{Z})\) is a Dunford–Pettis set. Thus \(TS:Z\rightarrow Y\) has a completely continuous adjoint.

\(\mathrm{(iii)} \Rightarrow \mathrm{(iv)}\) Suppose \((x_n)\) is a weakly Cauchy sequence in X and \((y_n^*)\) is a weakly null sequence in \(Y^*\). Define \(S:\ell _1\rightarrow X\) by

$$\begin{aligned} S(b)=\sum b_n x_n, \end{aligned}$$

where \(b=(b_n)\in \ell _1\). Since \(S(B_{\ell _1} )\) is contained in the closed and absolutely convex hull of \(\{ x_n : n \in {\mathbb {N}}\}\), which is weakly precompact [24, p. 27], S is weakly precompact.

By assumption, \((TS)^*=S^*T^*\) is completely continuous. Note that \(S^*(x^*)=(\langle x^*, x_i\rangle )_i\), \(x^*\in X^*\), and \(S^*T^*(y_n^*)=(\langle T^*(y_n^*), x_i\rangle )_i\). Hence

$$\begin{aligned} \langle y_n^*, T(x_n) \rangle =\langle T^*(y_n^*), x_n \rangle \le \Vert S^*T^*(y_n^*)\Vert = \sup _i |\langle T^*(y_n^*), x_i\rangle | \rightarrow 0. \end{aligned}$$

\(\mathrm{(ii)} \Rightarrow \mathrm{(iii)}\) and \(\mathrm{(iv)} \Rightarrow \mathrm{(i)}\) are obvious. \(\square \)

The following two corollaries provide equivalent characterizations of spaces with the Dunford–Pettis property.

Corollary 4

Let X be a Banach space. Then the following statements are equivalent:

  1. (i)

    X has the DPP.

  2. (ii)

    The identity operator \(i:X\rightarrow X\) is a weak Dunford–Pettis operator; that is, every weakly precompact subset of X is a Dunford–Pettis set.

  3. (iii)

    [8] Every operator \(S:X\rightarrow Z\) with weakly precompact adjoint is completely continuous, for any Banach space Z.

  4. (iv)

    [8] Every operator \(S:X\rightarrow c_0\) with weakly precompact adjoint is completely continuous.

  5. (v)

    [8] If \((x_n)\) is a weakly null sequence in X and \((x_n^*)\) is a weakly Cauchy sequence in \(X^*\), then \(x_n^*(x_n) \rightarrow 0\).

Proof

Apply Theorem 2 to the identity operator \(i:X\rightarrow X\). \(\square \)

We note that X has the DPP if and only if weakly precompact sets and DP sets coincide (since every DP set is weakly precompact [2]).

Corollary 5

Let X be a Banach space. Then the following statements are equivalent:

  1. (i)

    X has the DPP.

  2. (ii)

    [8] For all Banach spaces Z, every weakly precompact operator \(S:Z\rightarrow X\) has a completely continuous adjoint.

  3. (iii)

    Every weakly precompact operator \(S:\ell _1 \rightarrow X\) has a completely continuous adjoint.

  4. (iv)

    [8] If \((x_n)\) is a weakly Cauchy sequence in X and \((x_n^*)\) is a weakly null sequence in \(X^*\), then \( x_n^*(x_n) \rightarrow 0\).

Proof

Apply Corollary 3 to the identity operator \(i:X\rightarrow X\). \(\square \)

An operator \(T:X \rightarrow Y\) is called limited if \(T(B_X)\) is limited in Y. The operator \(T:X \rightarrow Y\) is limited if and only if \(T^*:Y^* \rightarrow X^*\) is \(w^*\)-norm sequentially continuous (by Observation 1).

We are now giving our second major result. It gives a characterization of \(\hbox {weak}^*\) Dunford–Pettis operators and generalizes [11, Theorem 3.2].

Theorem 6

Let X and Y be Banach spaces and let \(T:X\rightarrow Y\) be an operator. The following statements are equivalent.

  1. (1)

    T is a \(weak^*\) Dunford–Pettis operator.

  2. (2)

    T carries weakly precompact subsets of X to limited subsets of Y.

  3. (3)

    If \(S:Z\rightarrow X\) is a weakly precompact operator, then \(TS:Z\rightarrow Y\) is limited, for any Banach space Z.

  4. (4)

    If \(S:\ell _1\rightarrow X\) is a weakly precompact operator, then \(TS:\ell _1\rightarrow Y\) is limited.

  5. (5)

    If \((x_n)\) is a weakly null sequence in X and \((y_n^*)\) is a \(w^*\)-Cauchy sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

Proof

\((1) \Rightarrow (2)\) is similar to the proof of \((1) \Rightarrow (2)\) in Theorem 2.

\((2) \Rightarrow (3)\) Suppose \(S:Z\rightarrow X\) is weakly precompact. Then \(TS(B_Z)\) is limited, and thus TS is limited.

\((3) \Rightarrow (4)\) and \((5) \Rightarrow (1)\) are obvious.

\((4) \Rightarrow (1)\) Let \((x_n)\) be a weakly null sequence in X and \((y_n^*)\) be a \(w^*\)-null sequence in \(Y^*\). Define \(S:\ell _1\rightarrow X\) by

$$\begin{aligned} S(b)=\sum b_n x_n, \end{aligned}$$

where \(b=(b_n)\in \ell _1\). Since \(S(B_{\ell _1} )\) is contained in the closed and absolutely convex hull of \(\{ x_n : n \in {\mathbb {N}}\}\), which is relatively weakly compact [9, p. 51], S is weakly compact. By assumption, TS is limited. Suppose \((e_n^*)\) denotes the unit vector basis of \(\ell _1\). Then

$$\begin{aligned} \langle y_n^*, T(x_n) \rangle =\langle y_n^*, TS(e_n^*)\rangle \rightarrow 0. \end{aligned}$$

\((1) \Rightarrow (5)\) is similar to the proof of \((1) \Rightarrow (5)\) in Theorem 2. \(\square \)

Corollary 7

Let X and Y be Banach spaces and let \(T:X\rightarrow Y\) be an operator. The following statements are equivalent.

  1. (i)

    T is a \(weak^*\) Dunford–Pettis operator.

  2. (ii)

    If \((x_n)\) is a weakly Cauchy sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

  3. (iii)

    If \(S:Y\rightarrow Z \) is an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, then \(ST:X\rightarrow Z\) is completely continuous.

  4. (iv)

    If \(S:Y\rightarrow c_0\) is an operator, then \(ST:X\rightarrow c_0\) is completely continuous.

Proof

\(\mathrm{(i)} \Rightarrow \mathrm{(ii)}\) Suppose that \((x_n)\) is a weakly Cauchy sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\). Since \((T(x_n))\) is limited in Y, \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\).

\(\mathrm{(ii)} \Rightarrow \mathrm{(iii)}\) Let \(S:Y\rightarrow Z\) be an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, but \(ST:X\rightarrow Z\) is not completely continuous. Let \((x_n)\) be weakly null in X so that \(\Vert ST(x_n)\Vert >\epsilon \), for some \(\epsilon >0\). Choose \((z_n^*)\) in \(B_{Z^*}\) so that \(\langle z_n^*, ST(x_n)\rangle > \epsilon \). Without loss of generality \((S^*(z_n^*))\) is \(w^*\)-convergent. Hence \(\langle S^*(z_n^*), T(x_n) \rangle = \langle z_n^*, ST(x_n) \rangle \rightarrow 0\), a contradiction.

\(\mathrm{(iii)} \Rightarrow \mathrm{(iv)}\) Let \(S:Y\rightarrow c_0\) be an operator. Note that \(B_{\ell _1}\), and thus \(S^*(B_{\ell _1})\) is \(w^*\)-sequentially compact (since \(c_0\) is separable). Then \(ST:X\rightarrow c_0\) is completely continuous.

\(\mathrm{(iv)} \Rightarrow \mathrm{(i)}\) Suppose \((x_n)\) is a weakly null sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\). Define \(S:Y\rightarrow c_0\) by \(S(y)=(y_i^*(y))\). Since \(ST:X\rightarrow c_0\) is completely continuous, \(\langle y_n^*, T(x_n)\rangle \le \Vert ST(x_n)\Vert \rightarrow 0\). \(\square \)

The following corollary provides a characterization of spaces with the \(DP^*P\) and generalizes [11, Corollary 3.3].

Corollary 8

Let X be a Banach space. Then the following statements are equivalent:

  1. (i)

    X has the \(DP^*P\).

  2. (ii)

    [16] The identity operator \(i:X\rightarrow X\) is a \(weak^*\) Dunford–Pettis operator ; that is, every weakly precompact subset of X is a limited set.

  3. (iii)

    [16] Every weakly precompact operator \(S:Z\rightarrow X\) is limited, for any Banach space Z.

  4. (iv)

    Every weakly precompact operator \(S:\ell _1\rightarrow X\) is limited.

  5. (v)

    [16] If \((x_n)\) is a weakly null sequence in X and \((x_n^*)\) is a \(w^*\)-Cauchy sequence in \(X^*\), then \( x_n^*(x_n) \rightarrow 0\).

Proof

Apply Theorem 6 to the identity operator \(i:X\rightarrow X\). \(\square \)

We note that X has the \(DP^*P\) if and only if weakly precompact sets and limited sets coincide (since every limited set is weakly precompact [3]).

Corollary 9

Let X be a Banach space. Then the following statements are equivalent:

  1. (i)

    X has the \(DP^*P\).

  2. (ii)

    [16] If \((x_n)\) is a weakly Cauchy sequence in X and \((x_n^*)\) is a \(w^*\)-null sequence in \(Y^*\), then \(x_n^*(x_n) \rightarrow 0\).

  3. (iii)

    [16] If \(S:X\rightarrow Z \) is an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, then S is completely continuous.

  4. (iv)

    [5] Every operator \(S:X\rightarrow c_0\) is completely continuous.

Proof

Apply Corollary 7 to the identity operator \(i:X\rightarrow X\). \(\square \)

Corollary 10

  1. (i)

    If \(Y^*\) does not contain a copy of \(\ell _1\), then every weak Dunford–Pettis operator \(T:X\rightarrow Y\) is completely continuous.

  2. (ii)

    If \(B_{Y^*}\) is \(w^*\)-sequentially compact (in particular if Y is separable), then every \(\hbox {weak}^*\) Dunford–Pettis operator \(T:X\rightarrow Y \) is completely continuous.

  3. (iii)

    If X or Y has the DPP, then every operator \(T:X\rightarrow Y\) is weak Dunford–Pettis.

  4. (iv)

    If X or Y has the \(DP^*P\), then every operator \(T:X\rightarrow Y\) is \(\hbox {weak}^*\) Dunford–Pettis.

Proof

  1. (i)

    Let \(i:Y\rightarrow Y\) be the identity operator on Y. Suppose that \(T:X\rightarrow Y\) is a weak Dunford–Pettis operator. Since \(Y^*\) does not contain a copy of \(\ell _1\), \(i^*\) is weakly precompact (by Rosenthal’s \(\ell _1\) theorem). Then \(T=iT\) is completely continuous by Theorem 2.

  2. (ii)

    Let \(i:Y\rightarrow Y\) be the identity operator on Y. Suppose \(T:X\rightarrow Y\) is a \(\hbox {weak}^*\) Dunford–Pettis operator. Since \(i^*(B_{Y^*})\) is \(w^*\)-sequentially compact, \(T=iT\) is completely continuous by Corollary 7.

  3. (iii)

    Let \(T:X\rightarrow Y\) be an operator. If Y has the DPP, then the identity operator \(i:Y\rightarrow Y\) is weak Dunford–Pettis. Hence \(T=iT\) is weak Dunford–Pettis. If X has the DPP, then the identity operator \(i:X\rightarrow X\) is weak Dunford–Pettis. Hence \(T=Ti\) is weak Dunford–Pettis.

  4. (iv)

    The proof is similar to that of (iii).

\(\square \)

Clearly each completely continuous operator \(T:X\rightarrow Y \) is \(\hbox {weak}^*\) Dunford–Pettis and each \(\hbox {weak}^*\) Dunford–Pettis operator is weak Dunford–Pettis. By Corollary 10, we obtain the following result.

Corollary 11

If \(Y^*\) does not contain a copy of \(\ell _1\), then the families of completely continuous operators, \(weak^*\) Dunford–Pettis operators, and weak Dunford–Pettis operators \(T:X\rightarrow Y \) coincide.

Examples (a) Note that \(\ell _\infty \) has the \(DP^*P\) (since it has the DPP and the Grothendieck property [5]). Then the identity operator \(i:\ell _\infty \rightarrow \ell _\infty \) is \(\hbox {weak}^*\) Dunford–Pettis and is not completely continuous.

(b) A space X has the \(DP^*P\) if and only if every operator \(T:X\rightarrow c_0\) is completetely continuous [5]. Since the identity operator \(i:c_0 \rightarrow c_0\) is not completetely continuous, \(c_0\) does not have the \(DP^*P\). Thus \(i:c_0 \rightarrow c_0\) is weak Dunford–Pettis (since \(c_0\) has the DPP) and not \(\hbox {weak}^*\) Dunford–Pettis.

Corollary 12

  1. (i)

    Suppose that Y has the DPP. If \(T:X\rightarrow Y\) is an operator such that \(T^*\) is not completely continuous, then T fixes a copy of \(\ell _1\).

  2. (ii)

    Suppose that Y has the \(DP^*P\). If \(T:X\rightarrow Y\) is a non-limited operator, then T fixes a copy of \(\ell _1\).

Proof

  1. (i)

    Suppose that \(T^*\) is not completely continuous. Let \((y_n^*)\) be weakly null in \(Y^*\) so that \(\Vert T^*(y_n^*)\Vert \not \rightarrow 0\). Suppose \((x_n)\) is a sequence in \(B_X\) such that \(|\langle y_n^*, T(x_n)\rangle | >\epsilon \) for some \(\epsilon >0\). We claim that \((x_n)\) has no weakly Cauchy subsequence. If the claim is false, suppose without loss of generality that \((x_n)\) is weakly Cauchy. Since Y has the DPP, T is weak Dunford–Pettis. Then \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\) by Corollary 3. This contradiction shows that \((x_n)\) has no weakly Cauchy subsequence. By Rosenthal’s \(\ell _1\) theorem, \((x_n)\) has a subsequence equivalent to the \(\ell _1\) basis. Suppose without loss of generality that \((x_n)\) is equivalent to \((e_n^*)\), where \((e_n^*)\) denotes the basis of \(\ell _1\).

    Now, since \(|\langle y_n^*, T(x_n)\rangle | >\epsilon \) and Y has the DPP, \((T(x_n))\) has no weakly Cauchy subsequence (by Corollary 5). By Rosenthal’s \(\ell _1\) theorem, \((T(x_n))\) has a subsequence equivalent to \((e_n^*)\). Suppose without loss of generality that \((T(x_n))\) is equivalent to \((e_n^*)\). Hence T fixes a copy of \(\ell _1\).

  2. (ii)

    The proof is similar to that of (i).

\(\square \)

Corollary 12 (ii) generalizes [5, Theorem 2.3] (which states that if X and Y have the the \(DP^*P\) and \(T:X\rightarrow Y\) is a non-limited operator, then T fixes a copy of \(\ell _1\)).

A Banach space X has the Dunford–Pettis relatively compact property (DPrcP) if every DP subset of X is relatively compact [13]. Schur spaces have the DPrcP. The space X does not contain a copy of \(\ell _1\) if and only if \(X^*\) has the DPrcP [12, 13]. We note that if \(X^*\) does not contain a copy of \(\ell _1\), then \(X^{**}\), thus X, has the DPrcP [12, 13].

The space X has the Gelfand–Phillips (GP) property (or X is a Gelfand–Phillips space) if every limited subset of X is relatively compact. Schur spaces and separable spaces have the Gelfand–Phillips property [3].

An operator \(T:X\rightarrow Y\) is called Dunford–Pettis completely continuous (DPcc) if T carries weakly null and DP sequences to norm null ones [22]. An operator \(T:X\rightarrow Y\) is called limited completely continuous (lcc) if T maps weakly null limited sequences to norm null sequences [23].

The sets of all limited completely continuous, Dunford–Pettis completely continuous operators, weak Dunford Pettis, and \(\hbox {weak}^*\) Dunford Pettis operators from X to Y will be respectively denoted by LCC(XY), DPCC(XY), WDP(XY), and \(W^*DP(X,Y)\).

In the following result, we characterize Banach spaces X on which every weak (resp. \(\hbox {weak}^*\)) Dunford–Pettis operator is a DPcc (resp. lcc) operator.

Corollary 13

  1. (i)

    A Banach space X has the DPrcP if and only if \(DPCC(X, \ell _\infty )=WDP(X,\ell _\infty )\).

  2. (ii)

    A Banach space X has the GP property if and only if \(LCC(X, \ell _\infty )=W^*DP(X,\ell _\infty )\).

Proof

  1. (i)

    A Banach space X has the DPrcP if and only if \(DPCC(X, \ell _\infty )=L(X,\ell _\infty )\) [22]. Since \(\ell _\infty \) has the DPP, \(L(X,\ell _\infty )=WDP(X,\ell _\infty )\).

  2. (ii)

    A Banach space X has the GP property if and only if \(LCC(X, \ell _\infty )=L(X,\ell _\infty )\) [23]. Since \(\ell _\infty \) has the \(DP^*P\), \(L(X,\ell _\infty )=W^*DP(X,\ell _\infty )\).\(\square \)

If X has the DPrcP, then X has the GP property (since any limited set is a DP set). Thus, if X has the DPrcP, then \(L(X,\ell _\infty )=LCC(X,\ell _\infty )=DPCC(X,\ell _\infty )= WDP(X,\ell _\infty )=W^*DP(X,\ell _\infty )\).

Example We note that the identity operator \(i:\ell _\infty \rightarrow \ell _\infty \) is \(\hbox {weak}^*\) Dunford–Pettis and not lcc (since \(\ell _\infty \) does not have the GP property). Further, \(i:\ell _\infty \rightarrow \ell _\infty \) is weak Dunford–Pettis (since \(\ell _\infty \) has the DPP) and not DPcc (since \(\ell _\infty \) does not have the DPrcP).

4 Weak p-convergent operators and \(\hbox {weak}^*\)p-convergent operators

For \(1\le p<\infty \), \(p^*\) denotes the conjugate of p. If \(p=1\), we take \(c_0\) instead of \(\ell _{p^*}\). The unit vector basis of \(\ell _p\) will be denoted by \((e_n)\).

Let \(1\le p < \infty \). A sequence \((x_n)\) in X is called weakly p-summable if \((x^*(x_n))\in \ell _p\) for each \(x^*\in X^*\) [10, p. 32]. Let \(\ell _p^w(X)\) denote the set of all weakly p-summable sequences in X. The space \(\ell _p^w(X)\) is a Banach space with the norm

$$\begin{aligned} \Vert (x_n)\Vert _p^w=\sup \left\{ \left( \sum _{n=1}^\infty |\langle x^*, x_n \rangle |^p \right) ^{1/p}: x^*\in B_{X^*}\right\} \end{aligned}$$

We recall the following isometries: \(L(\ell _{p^*},X) \simeq \ell _p^w(X)\) for \(1<p<\infty \); \( L(c_0,X)\simeq \ell _p^w(X) \) for \(p=1\); that are obtained via the isometry \(T \rightarrow (T(e_n))\) [10, Proposition 2.2, p. 36].

A series \(\sum x_n\) in X is said to be weakly unconditionally convergent (wuc) if for every \(x^*\in X^*\), the series \(\sum |x^*(x_n)|\) is convergent. An operator \(T:X\rightarrow Y\) is unconditionally converging if it maps weakly unconditionally convergent series to unconditionally convergent ones.

Let \(1\le p\le \infty \). An operator \(T:X\rightarrow Y\) is called p-convergent if T maps weakly p-summable sequences into norm null sequences. The set of all p-convergent operators from X to Y is denoted by \(C_p(X,Y)\) [6].

The 1-convergent operators are precisely the unconditionally converging operators and the \(\infty \)-convergent operators are precisely the completely continuous operators. If \(p<q\), then \(C_q(X,Y)\subseteq C_p(X,Y)\).

A sequence \((x_n)\) in X is called weakly p-convergent to \(x\in X\) if the sequence \((x_{n}-x)\) is weakly p-summable [6]. Weakly \(\infty \)-convergent sequences are precisely the weakly convergent sequences.

Let \(1\le p\le \infty \). A bounded subset K of X is relatively weakly p-compact if every sequence in K has a weakly p-convergent subsequence. An operator \(T:X\rightarrow Y\) is weakly p-compact if \(T(B_X)\) is relatively weakly p-compact [6].

The set of weakly p-compact operators \(T:X\rightarrow Y\) is denoted by \(W_p(X,Y)\). If \(p<q\), then \(W_p(X,Y)\subseteq W_q(X,Y)\). A Banach space \(X\in C_p\) (resp. \(X\in W_p\)) if \(id(X)\in C_p(X,X)\) (resp. \(id(X)\in W_p(X,X)\)) [6], where id(X) is the identity map on X.

A sequence \((x_n)\) in X is called weakly p-Cauchy if \((x_{n_k}-x_{m_k})\) is weakly p-summable for any increasing sequences \((n_k)\) and \((m_k)\) of positive integers.

Every weakly p-convergent sequence is weakly p-Cauchy, and the weakly \(\infty \)-Cauchy sequences are precisely the weakly Cauchy sequences.

Let \(1\le p\le \infty \). A subset S of X is called weakly p-precompact if every sequence from S has a weakly p-Cauchy subsequence [18]. An operator \(T:X\rightarrow Y\) is called weakly p-precompact if \(T(B_X)\) is weakly p-precompact.

Let \(1\le p\le \infty \). A Banach space X has the Dunford–Pettis property of order p \((DPP_p)\) (\(1\le p\le \infty \)) if every weakly compact operator \(T:X \rightarrow Y\) is p-convergent, for any Banach space Y [6]. Equivalently, X has the \(DPP_p\) if and only if \(x_n^*(x_n)\rightarrow 0\) whenever \((x_n)\) is weakly p-summable in X and \((x_n^*)\) is weakly null in \(X^*\) [6, Proposition 3.2].

If X has the \(DPP_p\), then it has the \(DPP_q\), if \(q<p\). Also, the \(DPP_\infty \) is precisely the DPP, and every Banach space has the \(DPP_1\). C(K) spaces and \(L_1\) have the DPP, and thus the \(DPP_p\) for all p. If \(1<r<\infty \), then \(\ell _r\) has the \(DPP_p\) for \(p<r^*\). If \(1<r<\infty \), then \(L_r(\mu )\) has the \(DPP_p\) for \(p<min(2,r^*)\). Tsirelson’s space T has the \(DPP_p\) for all \(p<\infty \). Since T is reflexive, it does not have the DPP. Tsirelson’s dual space \(T^*\) does not have the \(DPP_p\), if \(p>1\) [6].

Let \(1\le p\le \infty \). A Banach space X has the \(DP^*\)-property of order p (\(DP^*P_p\)) if all weakly p-compact sets in X are limited [14]. Equivalently, X has the \(DP^*P_p\) if and only if \(x_n^*(x_n)\rightarrow 0\) whenever \((x_n)\) is weakly p-summable in X and \((x_n^*)\) is weakly null in \(X^*\) [14].

If X has the \(DP^*P_q\), then it has the \(DP^*P_p\), if \(q>p\). Further, the \(DP^*P_\infty \) is precisely the \(DP^*P\) and every Banach space has the \(DP^*P_1\). If X has the \(DP^*P\), then X has the \(DP^*P_p\), \(1 \le p \le \infty \). If X has the \(DP^*P_p\), then X has the \(DPP_p\).

Let \(1\le p<\infty \). An operator \(T:X\rightarrow Y\) is called weak p-convergent if \(\langle y_n^*,T(x_n) \rangle \rightarrow 0\) whenever \((x_n)\) is weakly p-summable in X and \((y_n^*)\) is weakly null in \(Y^*\) [15]. An operator \(T:X\rightarrow Y\) is called \(weak^*\)p-convergent if \(\langle y_n^*,T(x_n) \rangle \rightarrow 0\) whenever \((x_n)\) is weakly p-summable in X and \((y_n^*)\) is \(w^*\)-null in \(Y^*\) [15].

In the following we study weak p-convergent and \(\hbox {weak}^*\)p-convergent operators. The following result generalizes [18, Theorem 8].

Theorem 14

Let X and Y be Banach spaces, and let \(1< p<\infty \). The following statements are equivalent about an operator \(T:X\rightarrow Y\).

  1. (1)

    T is weak p-convergent.

  2. (2)

    T takes weakly p-precompact subsets of X to DP subsets of Y.

  3. (3)

    For any Banach space Z, if \(S:Y\rightarrow Z\) has a weakly precompact adjoint, then \(ST:X\rightarrow Z\) is p-convergent.

  4. (4)

    If \(S:Y\rightarrow c_0\) has a weakly precompact adjoint, then \(ST:X\rightarrow c_0\) is p-convergent.

  5. (5)

    If \((x_n)\) is a weakly p-summable sequence in X and \((y_n^*)\) is a weakly Cauchy sequence in \(Y^*\), then \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\).

Proof

\((1)\Rightarrow (2)\) Let A be a weakly p-precompact subset of X. Suppose by contradiction that T(A) is not a Dunford–Pettis subset of Y. Let \((y_n^*)\) be a weakly null sequence in \(Y^*\), and let \((x_n)\) be a sequence in A such that \(|\langle y_n^*, T(x_n) \rangle |>\epsilon \) for all n, for some \(\epsilon >0\). By passing to a subsequence, we can assume that \((x_n)\) is weakly p-Cauchy.

Let \(n_1=1\) and choose \(n_2>n_1\) so that \(|\langle y_{n_2}^*, T(x_{n_1})\rangle |<\epsilon /2\). We can do this since \((y_n^*)\) is \(w^*\)-null. Continue inductively. Choose \(n_k>n_{k-1}\) so that \(|\langle y_{n_k}^*, T(x_{n_{k-1}}) \rangle |<\epsilon /2\). Since T is weak p-convergent, \(\langle y^*_{n_k}, T(x_{n_k}-x_{n_{k-1}})\rangle \rightarrow 0\). However,

$$\begin{aligned} |\langle y_{n_k}^*, T(x_{n_k}-x_{n_{k-1}})\rangle |\ge |\langle y_{n_k}^*, T(x_{n_k})\rangle |-|\langle y_{n_k}^*, T(x_{n_{k-1}})\rangle |> \epsilon /2, \end{aligned}$$

a contradiction.

\((2)\Rightarrow (3)\) Suppose \(S:Y\rightarrow Z\) is an operator with weakly precompact adjoint. Let \((x_n)\) be a weakly p-summable sequence in X. By (2), \((T(x_n))\) is a DP subset of Y. Therefore \((ST(x_n))\) is relatively compact [20, Corollary 4]. Hence \(\Vert ST(x_n)\Vert \rightarrow 0\), and thus ST is p-convergent.

\((3) \Rightarrow (4)\) and \((5) \Rightarrow (1)\) are obvious.

\((4) \Rightarrow (1)\) Let \((x_n)\) be a weakly p-summable sequence in X and \((y_n^*)\) be a weakly null sequence in \(Y^*\). Define \(S:Y\rightarrow c_0\) by \(S(y)=(y_i^*(y))\). Then \(S^*:\ell _1\rightarrow Y^*\), \(S^*(b)=\sum b_i y_i^*\). Note that \(S^*\) maps \(B_{\ell _1}\) into the closed and absolutely convex hull of \(\{y_i^*:i\in {\mathbb {N}}\}\), which is relatively weakly compact [9, p. 51]. Then \(S^*\) is weakly compact. Hence \(ST:X\rightarrow c_0\) is p-convergent. Therefore \(\langle T(x_n), y_n^*\rangle \le \Vert ST(x_n)\Vert =\sup _i |\langle y_i^*, T(x_n)\rangle |\rightarrow 0 \), and T is weak p-convergent.

\((1) \Rightarrow (5)\) Let \((x_n)\) be a weakly p-summable sequence in X and \((y_n^*)\) be a weakly Cauchy sequence in \(Y^*\). Suppose \(\langle y_n^*, T(x_n) \rangle \not \rightarrow 0\). Without loss of generality suppose that \(|\langle y_n^*, T(x_n) \rangle |> \epsilon \) for each \(n\in {\mathbb {N}}\), for some \(\epsilon > 0\).

Let \(n_1=1\) and choose \(n_2>n_1\) so that \(|\langle y_{n_1}^*, T(x_{n_2})\rangle |<\epsilon /2\). We can do this since \((T(x_n))\) is weakly null. Continue inductively. Choose \(n_{k+1}>n_{k}\) so that \(|\langle y_{n_k}^*, T(x_{n_{k+1}})\rangle |<\epsilon /2\). Since T is a weak p-convergent operator, \(|\langle y^*_{n_{k+1}}-y_{n_k}^*, T(x_{n_{k+1}}) \rangle |\rightarrow 0\). However,

$$\begin{aligned} |\langle y^*_{n_{k+1}}-y_{n_k}^*, T(x_{n_{k+1})}\rangle |\ge |\langle y_{n_{k+1}}^*, T(x_{n_{k+1}})\rangle |-|\langle y_{n_k}^*, T(x_{n_{k+1}})\rangle |>\epsilon /2, \end{aligned}$$

and we have a contradiction. \(\square \)

Corollary 15

Let X and Y be Banach spaces, and let \(1< p<\infty \). The following statements are equivalent about an operator \(T:X\rightarrow Y\).

  1. (i)

    T is weak p-convergent.

  2. (ii)

    For every Banach space Z, if \(S:Z\rightarrow X\) is a weakly p-precompact operator, then \(TS:Z\rightarrow Y\) has a completely continuous adjoint.

  3. (iii)

    [18] If \(S:\ell _{p^*}\rightarrow X\) is an operator, then \(TS:\ell _{p^*} \rightarrow Y\) has a completely continuous adjoint.

  4. (iv)

    If \((x_n)\) is a weakly p-Cauchy sequence in X and \((y_n^*)\) is a weakly null sequence in \(Y^*\), then \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\).

Proof

\(\mathrm{(i)}\Rightarrow \mathrm{(ii)}\) Let \(S:Z\rightarrow X\) be a weakly p-precompact operator. Then \(TS(B_Z)\) is a DP set in Y. Hence \((TS)^*\) is completely continuous.

\(\mathrm{(ii)}\Rightarrow \mathrm{(iii)}\) Let \(S:\ell _{p^*}\rightarrow X\) be an operator. Since \(1< p<\infty \), \(\ell _{p^*} \in W_p\) [6]. Hence S is weakly p-compact, and thus \((TS)^*\) is completely continuous.

\(\mathrm{(iii)}\Rightarrow \mathrm{(i)}\) Let \((x_n)\) be weakly p-summable in X and let \((y_n^*)\) be weakly null in \(Y^*\). Define \(S:\ell _{p^*}\rightarrow X\) by \(S(b)=\sum b_i x_i\) [10, Proposition 2.2, p. 36]. Since \(TS(B_{\ell _{p^*}})\) is a DP set in Y, \(\langle y_n^*, TS(e_n)\rangle = \langle y_n^*, T(x_n)\rangle \rightarrow 0\).

\(\mathrm{(i)}\Rightarrow \mathrm{(iv)}\) Let \((x_n)\) be weakly p-Cauchy in X and let \((y_n^*)\) be weakly null in \(Y^*\). Since \((T(x_n))\) is a DP set in Y, \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\).

\(\mathrm{(iv)} \Rightarrow \mathrm{(i)}\) is obvious. \(\square \)

As a consequence of the previous two results we obtain the following characterizations of Banach spaces with the \(DPP_p\).

Corollary 16

[19, Theorem 1] Let \(1< p<\infty \). The following statements are equivalent about a Banach space X.

  1. (1)

    X has the \(DPP_p\).

  2. (2)

    The identity operator \(i:X\rightarrow X\) is weak p-convergent; that is, every weakly p-precompact subset of X is a Dunford–Pettis set.

  3. (3)

    Every operator \(S:X\rightarrow Z\) with weakly precompact adjoint is p-convergent, for any Banach space Z.

  4. (4)

    Every operator \(S:X\rightarrow c_0\) with weakly precompact adjoint is p-convergent.

  5. (5)

    If \((x_n)\) is a weakly p-summable sequence in X and \((x_n^*)\) is a weakly Cauchy sequence in \(X^*\), then \(x_n^*(x_n) \rightarrow 0\).

Proof

Apply Theorem 14 to the identity operator \(i:X\rightarrow X\). \(\square \)

Corollary 17

[19, Theorem 1] Let \(1< p<\infty \). The following statements are equivalent about a Banach space X.

  1. (i)

    X has the \(DPP_p\).

  2. (ii)

    For all Banach spaces Z, every weakly p-precompact operator \(S:Z\rightarrow X\) has a completely continuous adjoint.

  3. (iii)

    Every operator \(S:\ell _{p^*}\rightarrow X\) has a completely continuous adjoint.

  4. (iv)

    If \((x_n)\) is a weakly p-Cauchy sequence in X and \((x_n^*)\) is a weakly null sequence in \(X^*\), then \(x_n^*(x_n) \rightarrow 0\).

Proof

Apply Corollary 15 to the identity map \(i:X\rightarrow X\). \(\square \)

The following result generalizes [15, Theorem 2.11].

Theorem 18

Let X and Y be Banach spaces and \(T:X\rightarrow Y\) be an operator. Let \(1< p<\infty \). The following statements are equivalent.

  1. (1)

    T is \(weak^*\)p-convergent.

  2. (2)

    T carries weakly p-precompact subsets of X to limited subsets of Y.

  3. (3)

    If \(S:Z\rightarrow X\) is a weakly p-precompact operator, then \(TS:Z\rightarrow Y\) is limited, for any Banach space Z.

  4. (4)

    If \(S:\ell _{p^*}\rightarrow X\) is an operator, then \(TS:\ell _{p^*}\rightarrow Y\) is limited.

  5. (5)

    If \((x_n)\) is a weakly p-summable sequence in X and \((y_n^*)\) is a \(w^*\)-Cauchy sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

Proof

\((1)\Rightarrow (2)\) is similar to the proof of \((1) \Rightarrow (2)\) in Theorem 14.

\((2)\Rightarrow (3)\) Let \(S:Z\rightarrow X\) be a weakly p-precompact operator. Then \(TS(B_Z)\) is limited, and thus \(TS:Z\rightarrow Y\) is limited.

\((3)\Rightarrow (4)\) Let \(S:\ell _{p^*}\rightarrow X\) be an operator. Since \(1< p<\infty \), \(\ell _{p^*} \in W_p\) [6]. Hence S is weakly p-compact, and thus TS limited.

\((4)\Rightarrow (1)\) Suppose \((x_n)\) is weakly p-summable in X and \((y_n^*)\) is \(w^*\)-null in \(Y^*\). Define \(S:\ell _{p^*}\rightarrow X\) by \(S(b)=\sum b_i x_i\) [10, Proposition 2.2, p. 36]. Since \(TS(B_{\ell _{p^*}})\) is a limited set in Y, \(\langle y_n^*, TS(e_n)\rangle = \langle y_n^*, T(x_n)\rangle \rightarrow 0\).

\((1) \Rightarrow (5)\) is similar to the proof of \((1) \Rightarrow (5)\) in Theorem 14. \(\square \)

Corollary 19

Let X and Y be Banach spaces and \(T:X\rightarrow Y\) be an operator. Let \(1< p<\infty \). The following statements are equivalent.

  1. (i)

    T is \(weak^*\)p-convergent.

  2. (ii)

    If \((x_n)\) is a weakly p-Cauchy sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\), then \(\langle y_n^*, T(x_n) \rangle \rightarrow 0\).

  3. (iii)

    If \(S:Y\rightarrow Z \) is an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, then \(ST:X\rightarrow Z\) is p-convergent.

  4. (iv)

    If \(S:Y\rightarrow c_0\) is an operator, then \(ST:X\rightarrow c_0\) is p-convergent.

Proof

\(\mathrm{(i)} \Rightarrow \mathrm{(ii)}\) Suppose that \((x_n)\) is a weakly p-Cauchy sequence in X and \((y_n^*)\) is a \(w^*\)-null sequence in \(Y^*\). Since \((T(x_n))\) is limited in Y, \(\langle y_n^*, T(x_n)\rangle \rightarrow 0\).

\(\mathrm{(ii)} \Rightarrow \mathrm{(iii)}\) Let \(S:Y\rightarrow Z\) be an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, but \(ST:X\rightarrow Z\) is not p-convergent. Let \((x_n)\) be weakly p-summable in X so that \(\Vert ST(x_n)\Vert >\epsilon \), for some \(\epsilon >0\). Choose \((z_n^*)\) in \(B_{Z^*}\) so that \(\langle z_n^*, ST(x_n)\rangle > \epsilon \). Without loss of generality, \((S^*(z_n^*))\) is \(w^*\)-convergent. Then \(\langle S^*(z_n^*), T(x_n) \rangle = \langle z_n^*, ST(x_n) \rangle \rightarrow 0\), a contradiction.

\(\mathrm{(iii)} \Rightarrow \mathrm{(iv)}\) Let \(S:Y\rightarrow c_0\) be an operator. Note that \(B_{\ell _1}\), and thus \(S^*(B_{\ell _1})\) is \(w^*\)-sequentially compact. Then \(ST:X\rightarrow c_0\) is p-convergent.

\(\mathrm{(iv)} \Rightarrow \mathrm{(i)}\) Let \((x_n)\) be a weakly p-summable sequence in X and let \((y_n^*)\) be a \(w^*\)-null sequence in \(Y^*\). Define \(S:Y\rightarrow c_0\) by \(S(y)=(y_i^*(y))\). Since ST is p-convergent, \(\langle y_n^*, T(x_n)\rangle \le \Vert ST(x_n)\Vert \rightarrow 0\). \(\square \)

The following two corollaries provide equivalent characterizations of spaces with the \(DP^*P_p\).

Corollary 20

Let \(1< p<\infty \). The following statements are equivalent about a Banach space X.

  1. (i)

    X has the \(DP^*P_p\).

  2. (ii)

    [15] The identity operator \(i:X\rightarrow X\) is \(weak^*\)p-convergent; that is, every weakly p-precompact subset of X is a limited set.

  3. (iii)

    [18] Every weakly p-precompact operator \(S:Z\rightarrow X\) is limited, for any Banach space Z.

  4. (iv)

    [15] Every operator \(S:\ell _{p^*}\rightarrow X\) is limited.

  5. (v)

    [18] If \((x_n)\) is a weakly p-summable sequence in X and \((x_n^*)\) is a \(w^*\)-Cauchy sequence in \(X^*\), then \(x_n^*(x_n) \rightarrow 0\).

Proof

Apply Theorem 18 to the identity operator \(i:X\rightarrow X\). \(\square \)

Corollary 21

Let \(1< p<\infty \). The following statements are equivalent about a Banach space X.

  1. (i)

    X has the \(DP^*P_p\).

  2. (ii)

    [18] If \((x_n)\) is a weakly p-Cauchy sequence in X and \((x_n^*)\) is a \(w^*\)-null sequence in \(X^*\), then \(x_n^*(x_n) \rightarrow 0\).

  3. (iii)

    [18] If \(S:X\rightarrow Z \) is an operator such that \(S^*(B_{Z^*})\) is \(w^*\)-sequentially compact, then S is p-convergent.

  4. (iv)

    [15] Every operator \(S:X \rightarrow c_0\) is p-convergent.

Proof

Apply Corollary 19 to the identity operator \(i:X\rightarrow X\). \(\square \)

We note that an operator \(T:X\rightarrow Y\) is p-convergent if and only if T takes weakly p-precompact subsets of X into norm compact subsets of Y.

Corollary 22

Let \(1< p< \infty \).

  1. (i)

    Suppose \(S:X\rightarrow Y\) is weakly p-precompact and \(T:Y\rightarrow Z\) is an operator with weakly precompact adjoint. If Y has the \(DPP_p\), then TS is compact.

  2. (ii)

    Suppose \(S:X\rightarrow Y\) is weakly p-precompact and \(T:Y\rightarrow Z\) is an operator such that \(T^*(B_{Z^*})\) is \(w^*\)-sequentially compact. If Y has the \(DP^*P_p\), then TS is compact.

Proof

  1. (i)

    Suppose \(S:X\rightarrow Y\) is weakly p-precompact and \(T:Y\rightarrow Z\) is an operator such that \(T^*\) is weakly precompact. Since Y has the \(DPP_p\), T is p-convergent by Corollary 16. Then \(TS(B_X)\) is relatively compact, and thus TS is compact.

  2. (ii)

    The proof is similar to that of (i).

\(\square \)

Corollary 23

Let \(1< p<\infty \).

  1. (i)

    If \(Y^*\) does not contain a copy of \(\ell _1\), then every weak p-convergent operator \(T:X\rightarrow Y\) is p-convergent.

  2. (ii)

    If \(B_{Y^*}\) is \(w^*\)-sequentially compact (in particular if Y is separable), then every \(weak^*\)p-convergent operator \(T:X\rightarrow Y \) is p-convergent.

  3. (iii)

    If X or Y has the \(DPP_p\), then every operator \(T:X\rightarrow Y\) is weak p-convergent.

  4. (iv)

    If X or Y has the \(DP^*P_p\), then every operator \(T:X\rightarrow Y\) is \(weak^*\)p-convergent.

Proof

  1. (i)

    Let \(i:Y\rightarrow Y\) be the identity operator on Y. Suppose \(T:X\rightarrow Y \) is a weak p-convergent operator. By Rosenthal’s \(\ell _1\) theorem, \(i^*\) is weakly precompact. Then \(T=iT\) is p-convergent by Theorem 14.

  2. (ii)

    The proof is similar to that of (ii).

  3. (iii)

    Let \(T:X\rightarrow Y\) be an operator. If Y has the \(DPP_p\), then the identity operator \(i:Y\rightarrow Y\) is weak p-convergent. Hence \(T=iT\) is weak p-convergent. If X has the \(DPP_p\), then the identity operator \(i: X\rightarrow X\) is weak p-convergent. Hence \(T=Ti\) is weak p-convergent.

  4. (iv)

    The proof is similar to that of (iii).

\(\square \)

Clearly each p-convergent operator \(T:X\rightarrow Y \) is \(\hbox {weak}^*\)p-convergent and each \(\hbox {weak}^*\)p-convergent operator is weak p-convergent. By Corollay 23, we obtain the following result. It generalizes [15, Proposition 2.5].

Corollary 24

If \(Y^*\) does not contain a copy of \(\ell _1\), then the families of p-convergent operators, \(weak^*\)p-convergent operators, and weak p-convergent operators \(T:X\rightarrow Y \) coincide.

Let \(1\le p<\infty \). A Banach space X has the p-Gelfand–Phillips (p-GP) property (or is a p-Gelfand–Phillips space) if every limited weakly p-summable sequence in X is norm null [15].

If X has the GP property, then X has the p-GP property for any \(1\le p<\infty \). The space \(\ell _\infty \) does not have the p-GP property for any \(1\le p<\infty \) [15].

Let \(1\le p<\infty \). A space X has the p-Dunford Pettis relatively compact property (p-DPrcP) if every DP weakly p-summable sequence \((x_n)\) in X is norm null [17].

If X has the DPrcP property, then X has the p-DPrcP property for any \(1\le p<\infty \).

Corollary 25

Let \(1\le p<\infty \). If X has the p-GP (resp. the p-DPrcP) property, then the following are equivalent.

  1. (i)

    X has the \(DP^*P_p\) (resp. the \(DPP_p\)).

  2. (ii)

    \(X\in C_p\).

Proof

\((i) \Rightarrow (ii)\) We only prove the result when X has the the p-GP and the \(DP^*P_p\). The other case is similar.

Let \((x_n)\) be weakly p-summable in X. Then \((x_n)\) is limited by Corollary 20. Therefore \(\Vert x_n\Vert \rightarrow 0\), and thus \(X\in C_p\). \(\square \)

Let \(1\le p< \infty \). An operator \(T:X\rightarrow Y\) is called limited p-convergent if it carries limited weakly p-summable sequences in X to norm null ones in Y [15]. An operator \(T:X\rightarrow Y\) is called DP p-convergent if it takes DP weakly p-summable sequences to norm null sequences [17].

The sets of all limited p-convergent, DP p-convergent, weak p-convergent, and \(\hbox {weak}^*\)p-convergent operators from X to Y will be respectively denoted by \(LC_p(X,Y)\), \(DPC_p(X,Y)\), \(WC_p(X,Y)\), and \(W^*C_p(X,Y)\).

Corollary 26

Let \(1\le p<\infty \). Let X be a Banach space. The following statements hold.

  1. (i)

    X has the p-DPrcP if and only if \(WC_p(X,\ell _\infty )=DPC_p(X,\ell _\infty )\).

  2. (ii)

    X has the p-GP property if and only if \(W^*C_p(X,\ell _\infty )=LC_p(X,\ell _\infty )\).

Proof

  1. (i)

    A Banach space X has the p-DPrcP if and only if \(DPC_p(X, \ell _\infty )=L(X,\ell _\infty )\) [17]. Since \(\ell _\infty \) has the \(DPP_p\), \(L(X,\ell _\infty )=WC_p(X,\ell _\infty )\).

  2. (ii)

    A Banach space X has the p-GP if and only if \(LC_p(X, \ell _\infty )=L(X,\ell _\infty )\) [17]. Since \(\ell _\infty \) has the \(DP^*P_p\), \(L(X,\ell _\infty )=W^*C_p(X,\ell _\infty )\).

\(\square \)

Since any limited set is a DP set, any limited weakly p-summable sequence is also DP weakly p-summable. Hence if X has the p-DPrcP, then X has the p-GP property. Thus, if X has the p-DPrcP, then \(L(X,\ell _\infty )=LC_p(X,\ell _\infty )=DPC_p(X,\ell _\infty )= WC_p(X, \ell _\infty )= W^*C_p(X, \ell _\infty )\).