Stability of average roughness, octahedrality, and strong diameter 2 properties of Banach spaces with respect to absolute sums
Rainis Haller, Johann Langemets, and Rihhard Nadel

TL;DR
This paper investigates how properties like average roughness, octahedrality, and diameter 2 are preserved or characterized in Banach spaces when forming absolute sums, providing new stability results and optimal bounds.
Contribution
It establishes how average roughness and diameter 2 properties behave under absolute sums, offering new characterizations and stability results for these geometric properties.
Findings
Absolute sums of $ ext{delta}$-average rough spaces are $ ext{delta}/N(1,1)$-average rough.
Spaces $X igoplus_p Y$ are $2^{1-1/p}$-average rough for octahedral $X,Y$ and $p eq 1, ext{infinity}$.
Diametral strong diameter 2 property is stable only for 1- and $ ext{infinity}$-sums.
Abstract
We prove that, if Banach spaces and are -average rough, then their direct sum with respect to an absolute norm is -average rough. In particular, for octahedral and and for in the space is -average rough, which is in general optimal. Another consequence is that for any in there is a Banach space which is exactly -average rough. We give a complete characterization when an absolute sum of two Banach spaces is octahedral or has the strong diameter 2 property. However, among all of the absolute sums, the diametral strong diameter 2 property is stable only for 1- and -sums.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Stability of average roughness, octahedrality, and strong diameter two properties of Banach spaces
with respect to absolute sums
Rainis Haller
,
Johann Langemets
and
Rihhard Nadel
(Date: 26.04.2013)
Abstract.
We prove that, if Banach spaces and are -average rough, then their direct sum with respect to an absolute norm is -average rough. In particular, for octahedral and and for in the space is -average rough, which is in general optimal. Another consequence is that for any in there is a Banach space which is exactly -average rough. We give a complete characterization when an absolute sum of two Banach spaces is octahedral or has the strong diameter 2 property. However, among all of the absolute sums, the diametral strong diameter 2 property is stable only for 1- and -sums.
Key words and phrases:
Average rough norm, octahedral norm, diameter 2 property, Daugavet property
2010 Mathematics Subject Classification:
Primary 46B20, 46B22
This research was supported by institutional research funding IUT20-57 of the Estonian Ministry of Education and Research.
1. Introduction
A real Banach space is said to be octahedral if, for every finite-dimensional subspace of and every , there is a norm one element such that
[TABLE]
Octahedral Banach spaces were introduced by Godefroy and Maurey [10] (see also [9]) in order to characterize Banach spaces containing an isomorphic copy of . This notion has recently been useful in studying the diameter 2 properties (see [4], [5], [11], and [12]). It is known that octahedrality is stable by taking - or -sums, and it is not stable by taking -sums for (see [11, Proposition 3.12]). More precisely, for nontrivial Banach spaces and ,
- •
if or is octahedral, then is octahedral;
- •
if and are both octahedral, then is octahedral;
- •
is not octahedral for .
We extend these results quantitatively in two directions, instead of octahedral spaces we consider more general average rough spaces, and we also consider absolute normalized norm on direct sum.
Let . A Banach space is said to be -average rough [8] if, whenever and ,
[TABLE]
Banach spaces which are -average rough are exactly the octahedral ones (see [4], [8], and [9]).
We recall that a norm on is called absolute (see [7]) if
[TABLE]
and normalized if
[TABLE]
For example, the -norm is absolute and normalized for every . If is an absolute normalized norm on (see [7, Lemmata 21.1 and 21.2]), then
- •
for all ;
- •
if with then
[TABLE]
- •
the dual norm on defined by
[TABLE]
is also absolute and normalized. Note that .
If and are Banach spaces and is an absolute normalized norm on , then we denote by the product space with respect to the norm
[TABLE]
In the special case where is the -norm, we write . Note that .
By a slice of we mean a set of the form
[TABLE]
where and . A convex combination of slices is a set of the form , where , with , and are slices of .
A dual characterization of -average roughness is well known. The dual space is -average rough if and only if the diameter of every convex combination of slices of is greater than or equal to [8, Theorem 2]. In particular, is octahedral if and only if the diameter of every convex combination of slices of is (see also [4], [9], and [11]). According to [1], the latter extreme property of a Banach space is known as the strong diameter 2 property. An important class of Banach spaces with the strong diameter 2 property and which are octahedral are the Daugavet spaces (see [1] and [4]).
In [6], it is proved that the only absolute sums which preserve the Daugavet property are the - and -sum. Surprisingly, there are many absolute norms which preserve octahedrality and the strong diameter 2 property (see Section 3).
Recently, Becerra Guerrero, López-Pérez, and Rueda Zoca introduced a sharper version of the strong diameter 2 property (see [3]). A Banach space has the diametral strong diameter 2 property if for every convex combination of relatively weakly open subsets of , for every and there is a such that
[TABLE]
By [3], Daugavet spaces have the diametral strong diameter 2 property and the diametral strong diameter 2 property implies the strong diameter 2 property. The Banach space is an example of a space with the strong diameter 2 property and failing the diametral strong diameter 2 property. As far as the authors know it is an open question posed in [3] whether there is a Banach space with the diametral strong diameter 2 property and failing the Daugavet property. Our preliminary idea to attack this question was to check whether besides - and -norm (see [3] and [13]) there are more absolute norms which preserve the diametral strong diameter 2 property. However, there are none (see Section 3).
We now describe the contents of this paper. In Section 2, we prove (see Theorem 2.4) that for -average rough Banach spaces and their absolute sum is -average rough, where is such that . In particular, we get that, for , the -sum of octahedral Banach spaces and is -average rough (see Corollary 2.6). Moreover, this number is in general the largest possible one (see Proposition 2.7). As a consequence, we obtain that for any there is a Banach space which is exactly -average rough (see Theorem 2.8). We end this section by describing when the -average roughness passes down from the absolute sum to one of the factors (see Proposition 2.11).
In Section 3, we characterize those absolute norms for which the direct sum of two octahedral Banach spaces is octahedral (see Theorem 3.2). As a consequence, we can characterize those absolute norms for which the direct sum of two separable Banach spaces with the almost Daugavet property has the almost Daugavet property (see Corollary 3.3). By duality, we can characterize the absolute norms which preserve the strong diameter 2 property (see Theorem 3.5). We end this section by proving that, similarly to the Daugavet property, among all of the absolute norms the diametral strong diameter 2 property is stable only for - and -sums (see Corollary 3.8).
2. Average roughness of absolute sums
We begin by pointing out some equivalent but sometimes more convenient formulations of average roughness, which are easily derived from the definition.
Proposition 2.1**.**
Let be a Banach space and . The following assertions are equivalent:
- (i)
* is -average rough;* 2. (ii)
whenever , , and there is a such that and
[TABLE] 3. (iii)
whenever , , and there is a such that and
[TABLE]
Remark*.*
The equivalences in Proposition 2.1 remain true if one of the following holds
- (a)
one replaces with , where and ;
- (b)
one replaces with .
The -sum of two Banach spaces inherits its -average roughness from one of the factors.
Proposition 2.2**.**
Let and be Banach spaces. If or is -average rough for some , then is also -average rough.
Proof.
We consider only the case where is -average rough. The case where is -average rough is similar. We will prove that is -average rough. Let and . By Proposition 2.1, it suffices to show that there exists such that and
[TABLE]
Since is -average rough, there is an such that and
[TABLE]
It follows that, for we have , and
[TABLE]
∎
Corollary 2.3** (see [11, Proposition 3.12]).**
If or is octahedral, then is also octahedral.
The following theorem is one of the main results in this section.
Theorem 2.4**.**
Let and be Banach spaces, an absolute normalized norm on , and be such that . If and are -average rough for some , then is -average rough.
Proof.
Assume that and are -average rough. We will prove that is -average rough. Let and . By Proposition 2.1, it suffices to show that there exists such that and
[TABLE]
Choose such that and . Denote by
[TABLE]
Note that , because . Consider first the case where and . Denote by
[TABLE]
Observe that . Since and are -average rough, by Proposition 2.1, there are and such that and
[TABLE]
and
[TABLE]
It follows that, for we have , and
[TABLE]
Consider now the case where , which means that and for all . This implies that for all . Since is -average rough, by Proposition 2.1, there exists a such that and
[TABLE]
Therefore, for we have , and
[TABLE]
The case where is similar to the case . We have thus proved that is -average rough. ∎
In particular, Theorem 2.4 applies to -norms.
Corollary 2.5**.**
If Banach spaces and are -average rough for some , then
- (a)
* is -average rough;*
- (b)
* is -average rough for .*
Corollary 2.6**.**
If Banach spaces and are octahedral and , then is -average rough.
In Corollary 2.6, we saw that if and are octahedral and , then is -average rough. We will now prove that in general is the largest possible number.
Proposition 2.7**.**
Let and be Banach spaces and . Then is not -average rough for any .
Proof.
We will prove that is not -average rough for any . Consider the elements and in , where and . It suffices to show that there is a function such that , when , and that for every and , where ,
[TABLE]
Let . Let be such that . By Maclaurin’s formula,
[TABLE]
for some . Observe that
[TABLE]
We continue by considering the cases and separately. In both cases we will use the generalized Bernoulli’s inequality, which says that for any we have .
Case I. Assume that . Since , we have
[TABLE]
Combining the estimate (2.1) with Bernoulli’s inequality we get
[TABLE]
Similarly, we obtain
[TABLE]
Therefore
[TABLE]
Thus, for , we can take
[TABLE]
Case II. Assume that . Since and , we have
[TABLE]
Combining this estimate with (2.1) and Bernoulli’s inequality, we get
[TABLE]
Similarly, we obtain
[TABLE]
Therefore
[TABLE]
Thus, for , we can take
[TABLE]
Hence is not -average rough for any . ∎
Now we are ready to show that for any there is a Banach space which is exactly -average rough.
Theorem 2.8**.**
For any there is a dual Banach space, which is -average rough and is not -average rough for any .
Proof.
If , then we can take . If , then there is a such that . Let be such that . Since is octahedral, then by Corollary 2.6 and Proposition 2.7 the Banach space is -average rough and is not -average rough for any . ∎
Remark*.*
We do not know whether a similar result to Theorem 2.8 holds for .
Theorem 2.8 and the dual caharacterization of -average rough norms (see [8, Theorem 2]) immediately implies the following.
Corollary 2.9**.**
For any there is a Banach space in which the minimal diameter of convex combination of slices is .
We end this section by describing when the -average roughness passes down from the absolute sum to one of the factors. Our results are inspired by [2, Proposition 2.5].
The following lemma is easily verified from the definitions.
Lemma 2.10**.**
Let be an absolute normalized norm on such that is an extreme point of the unit ball . Then is a strongly exposed point of , which is strongly exposed by the functional . In particular, for every there is a such that, whenever and , then .
Proposition 2.11**.**
Let and be Banach spaces and an absolute normalized norm on such that is an extreme point of . If is -average rough for some , then is -average rough.
Proof.
Assume that is -average rough. Let and . We will show that there is a such that and
[TABLE]
By Lemma 2.10, there is a such that, whenever and , then .
Consider . Since is -average rough, there is a such that and
[TABLE]
Choose with such that
[TABLE]
and
[TABLE]
Then we have
[TABLE]
which implies that
[TABLE]
It follows that and hence for all . Similarly, one obtains that and for all .
Therefore
[TABLE]
∎
Remark*.*
One can prove similarly to Proposition 2.11 that, if is an absolute normalized norm on such that is an extreme point of and is -average rough for some , then is -average rough.
Corollary 2.12**.**
If is -average rough and , then and are -average rough.
3. Octahedrality and strong diameter two properties of absolute sums
In this section, we characterize those absolute norms for which the direct sum of two octahedral Banach spaces is octahedral. In fact, there are many such norms besides the - and -norm. Since octahedrality and the strong diameter 2 property are dually connected, it follows that there are many absolute norms which preserve the strong diameter 2 property. In order to present these characterizations we will introduce the notions of positive octahedrality and the positive strong diameter 2 property. We end this section by proving that, similarly to the Daugavet property, among all of the absolute norms the diametral strong diameter 2 property is stable only for - and -sums.
We begin by recalling the following equivalent formulation of octahedrality from [11].
Proposition 3.1** (see [11, Proposition 2.2]).**
Let be a Banach space. The following assertions are equivalent:
- (i)
* is octahedral;*
- (ii)
whenever , , and , there is a such that
[TABLE]
Definition**.**
An element is positive if and . Let be an absolute normalized norm on . We say that is positively octahedral if whenever and positive there is a positive such that
[TABLE]
Remark*.*
Note that is positively octahedral if and only if there is a such that
[TABLE]
Theorem 3.2**.**
Let and be octahedral Banach spaces and an absolute normalized norm on . Then is octahedral if and only if is positively octahedral.
Proof.
Necessity. Assume that is octahedral. Let and positive . We will show that there is a positive such that
[TABLE]
Let and be such that and . Since is octahedral, there exists a such that and
[TABLE]
Take and . Then for every
[TABLE]
Sufficiency. Assume that is positively octahedral. Let be with norm one and . We will show that there is a with norm one such that
[TABLE]
Since is positively octahedral, there is a positive such that
[TABLE]
Since and are octahedral, there are and such that
[TABLE]
and
[TABLE]
Take and . It follows that and
[TABLE]
∎
Recall (see [14]) that a Banach space has the almost Daugavet property if there is a 1-norming subspace of such that
[TABLE]
holds true for every rank-one operator of the form , where and . This definition is a generalization of the well-known Daugavet property, where . In [15, Propositions 2.1 and 2.2], it is shown that if and are separable Banach spaces with the almost Daugavet property, then and have the almost Daugavet property too. Since the almost Daugavet property and octahedrality coincide for separable Banach spaces (see [14, Theorem 1.1]), we immediately get from Theorem 3.2 the following stability result for almost Daugavet spaces.
Corollary 3.3**.**
Let and be separable Banach spaces with the almost Daugavet property and an absolute normalized norm on . Then has the almost Daugavet property if and only if is positively octahedral.
In order to characterize those absolute norms which preserve the strong diameter 2 property, we introduce the following notion.
Definition**.**
Let be an absolute normalized norm on . We say that has the positive strong diameter 2 property if whenever , positive , , and with there are positive such that
[TABLE]
Remark*.*
Note that has the positive strong diameter 2 property if and only if there are such that and N\Big{(}\dfrac{1}{2}(a,1)+\dfrac{1}{2}(1,b)\Big{)}=1.
Proposition 3.4**.**
Let be an absolute normalized norm on . The space has the positive strong diameter 2 property if and only if is positively octahedral.
Proof.
Necessity. Assume that has the positive strong diameter 2 property. So there are such that and
[TABLE]
Let be such that and
[TABLE]
It implies that . Hence
[TABLE]
and
[TABLE]
Therefore is positively octahedral.
Sufficiency. Assume now that is positively octahedral. So there exist such that and
[TABLE]
Let be such that , ,
[TABLE]
and
[TABLE]
It follows that and which means that . Hence
[TABLE]
Therefore has the positive strong diameter 2 property. ∎
The duality between the strong diameter 2 property and octahedrality, Theorem 3.2, and Proposition 3.4 yield the following result, however, we prefer to give its direct proof.
Theorem 3.5**.**
Let and be Banach spaces with the strong diameter 2 property and an absolute normalized norm on . Then has the strong diameter 2 property if and only if has the positive strong diameter 2 property.
Proof.
Necessity. Assume that has the strong diameter 2 property. We will show that has the positive strong diameter 2 property. Let be positive elements in , , with , and . We will show that there are positive such that and .
Let be such that and for every . Since has the strong diameter 2 property, there are
[TABLE]
such that .
Take . Then , because
[TABLE]
and
[TABLE]
Sufficiency. We use an idea from [13]. Assume that has the positive strong diameter 2 property. Let be slices of defined by norm one functionals and scalars . Let be such that . We will show that the diameter of is 2.
Let . Consider the slices and (If , then and if , then ).
Since has the positive strong diameter 2 property, there are positive such that N\Big{(}\sum_{i=1}^{n}\lambda_{i}(a_{i},b_{i})\Big{)}>1-\delta, where satisfies for all .
It turns out that . Indeed, if and , then
[TABLE]
and
[TABLE]
Denote by
[TABLE]
Suppose that and . For every , denote by
[TABLE]
As and have the strong diameter 2 property, then there are and such that and . Take , , , and . Then , because and . Finally,
[TABLE]
Consider now the case, where or . Assume that . Since
[TABLE]
then
[TABLE]
As the diameter of is 2, there are such that
[TABLE]
Thus . Now we have
[TABLE]
∎
We now turn our attention to investigate the stability of the diametral strong diameter 2 property. From [3] and [13], we know that and have the diametral strong diameter 2 property as soon as and have the diametral strong diameter 2 property. We end this section by proving that there are no other absolute norms different from - and -norm which preserve the diametral strong diameter 2 property. Since the diametral strong diameter 2 property implies the strong diameter 2 property and the latter is stable only for absolute norms with the positive strong diameter 2 property, we can restrict our attention to them.
Consider an absolute normalized norm on , different from the -norm and -norm, such that has the positive strong diameter 2 property. Thus, for some with or , is defined by
[TABLE]
Proposition 3.6**.**
Let and be nontrivial Banach spaces and defined by (3.1). Then does not have the diametral strong diameter 2 property.
We will use the following elementary lemma.
Lemma 3.7**.**
There is a such that
[TABLE]
Proof.
Assume that . Denote by
[TABLE]
It is straightforward to show directly that the condition
[TABLE]
is equivalent to
[TABLE]
and the condition
[TABLE]
is equivalent to
[TABLE]
Note that
[TABLE]
where the first inequality is strict if and only if , and the last inequality is strict if and only if . ∎
Proof of Proposition 3.6.
By using Lemma 3.7, we choose such that
[TABLE]
Denote by
[TABLE]
Choose any . Let be such that if satisfy the conditions , , and , then
[TABLE]
Fix any and . Consider the slices and . Choose and such that and . Assuming that the Banach space has the diametral strong diameter 2 property, there exist and such that
[TABLE]
Since
[TABLE]
it follows that
[TABLE]
i.e., , which is a contradiction. ∎
Combining [3, Theorem 3.8], [13, Theorem], and Proposition 3.6, we get the following corollary.
Corollary 3.8**.**
If has the diametral strong diameter 2 property, then either or .
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] T. A. Abrahamsen, V. Lima, and O. Nygaard, Remarks on diameter 2 properties , J. Convex Anal. 20 (2013), 439–452.
- 2[2] M. D. Acosta, J. Becerra Guerrero, and G. López-Pérez, Stability results of diameter two properties , J. Convex Anal. 22 (2015), 1–17.
- 3[3] J. Becerra Guerrero, G. López-Pérez, and A. Rueda Zoca, Diametral diameter two properties in Banach spaces , ar Xiv:1509.02061 v 4.
- 4[4] J. Becerra Guerrero, G. López-Pérez, and A. Rueda Zoca, Octahedral norms and convex combination of slices in Banach spaces , J. Funct. Anal. 266 (2014), 2424–2435.
- 5[5] J. Becerra Guerrero, G. López-Pérez, and A. Rueda Zoca, Octahedral norms in spaces of operators , J. Math. Anal. Appl. 427 (2015), 171–184.
- 6[6] D. Bilik, V. Kadets, R. Shvidkoy, and D. Werner, Narrow operators and the Daugavet property for ultraproducts , Positivity 9 (2005), 45–62.
- 7[7] F. F. Bonsall and J. Duncan, Numerical Ranges II , London Mathematical Society Lecture Notes Series 10, Cambridge University Press, New York (1973).
- 8[8] R. Deville, A dual characterisation of the existence of small combinations of slices , Bull. Austral. Math. Soc. 37 (1988), 113–120.
