---
title: Polynomial Daugavet Property in Banach Spaces
url: https://www.emergentmind.com/topics/polynomial-daugavet-property
type: topic
---

# Polynomial Daugavet Property in Banach Spaces

Searching arXiv for recent and foundational papers on the polynomial Daugavet property.
arxiv_search.query({"search_query":"all:\"polynomial Daugavet property\" OR ti:\"Daugavet property\" AND abs:polynomial","max_results":10,"sort_by":"relevance","sort_order":"descending"})
Refining the search to include tensor-product and operator variants connected to the topic.
arxiv_search.query({"search_query":"all:\"polynomial Daugavet property\" AND (Banach OR tensor OR operator OR JB*)","max_results":10,"sort_by":"submittedDate","sort_order":"descending"})
The polynomial Daugavet property (PDP) is the requirement that every weakly compact polynomial \(P\in \mathcal{P}(X,X)\) on a Banach space \(X\) satisfy the Daugavet equation
\[
\|\mathrm{Id}_X+P\|=1+\|P\|.
\]
Because linear operators are \(1\)-homogeneous polynomials, PDP is formally stronger than the classical Daugavet property (DP). The decisive structural result is now that the two properties are equivalent for arbitrary real or complex Banach spaces: \(X\) has DP if and only if it has PDP. The proof of this equivalence replaces weak-topological arguments by arguments in the weak polynomial topology and resolves the longstanding open problem posed in the earlier literature [2507.06143].

## 1. Definition and ambient framework

For Banach spaces \(X\) and \(Y\), an \(N\)-homogeneous polynomial \(P:X\to Y\) is a map of the form \(P(x)=F(x,\dots,x)\), where \(F:X^N\to Y\) is continuous, \(N\)-linear, and symmetric. A continuous polynomial is a finite sum of homogeneous polynomials, and its norm is
\[
\|P\|=\sup\{\|P(x)\|:\|x\|\le 1\}.
\]
A rank-one polynomial is of the form \(P(x)=p(x)y_0\), with scalar-valued \(p\in\mathcal{P}(X)\) and \(y_0\in Y\). The polynomial Daugavet property is defined by requiring the Daugavet equation for every weakly compact polynomial \(P:X\to X\); equivalently, it is enough to verify the equation on rank-one-type polynomials \(x\mapsto p(x)a\) [2507.06143].

The linear antecedent is the Daugavet property: \(X\) has DP when every rank-one operator \(T:X\to X\) satisfies
\[
\|\mathrm{Id}_X+T\|=1+\|T\|,
\]
where a rank-one operator has the form \(T(x)=x^*(x)y\). In spaces with DP, the same equality extends from rank-one operators to weakly compact operators. Since linear operators are polynomials, PDP implies DP immediately; the nontrivial content is the converse implication [2507.06143].

A central topological object in the modern theory is the weak polynomial topology on \(X\), defined as the smallest topology making all scalar-valued continuous polynomials continuous. Thus a net \((x_\alpha)\) converges to \(x\) in the weak polynomial topology exactly when \(p(x_\alpha)\to p(x)\) for every \(p\in\mathcal{P}(X)\). On the bidual \(X^{**}\), the polynomial-star topology is defined by the Aron–Berner extensions \(\widehat p\) of scalar polynomials. A Goldstine-type theorem of Davie–Gamelin states that \(B_{X^{**}}\) is the polynomial-star closure of \(B_X\), and this approximation principle is a technical backbone of the modern proof of \(DP\Rightarrow PDP\) [2507.06143].

## 2. Geometric characterizations

The linear Daugavet property admits a slice-theoretic description: \(X\) has DP if and only if for every \(x\in S_X\) and every slice \(S\) of \(B_X\),
\[
\sup_{y\in S}\|x+y\|=2.
\]
Here a slice is a non-empty intersection with an open half-space. Shvidkoy’s lemma sharpens this by asserting that, if \(X\) has DP, then for every \(x\in S_X\) and \(\varepsilon>0\), the set
\[
\{y\in B_X:\|x+y\|>2-\varepsilon\}
\]
is weakly dense in \(B_X\). Equivalently, for every \(x\in S_X\) and \(y\in B_X\), there exists a net \((y_\alpha)\subset B_X\) such that \(y_\alpha\to y\) weakly and \(\|x+y_\alpha\|\to 2\) [2507.06143].

The polynomial analogue replaces slices by scalar polynomials. A standard characterization states that \(X\) has PDP if and only if, given \(x\in S_X\), \(\varepsilon>0\), and a norm-one scalar-valued polynomial \(p\in\mathcal{P}(X)\), there exist \(y\in B_X\) and \(\omega\in\mathbb K\) with \(|\omega|=1\) such that
\[
\operatorname{Re}(\omega\,p(y))>1-\varepsilon
\quad\text{and}\quad
\|x+\omega y\|>2-\varepsilon.
\]
This formulation exhibits the same near-extremal norm geometry as the linear theory, but with slices replaced by polynomial level sets [2507.06143].

The technical obstruction to deducing this directly from the linear theory is that, in infinite-dimensional Banach spaces, polynomials are generally not weakly continuous on bounded sets except in finite-type situations. Moreover, spaces with DP contain copies of \(\ell_1\), where there are polynomials that are weakly sequentially continuous but not weakly continuous on bounded sets. This prevents a direct transplantation of Shvidkoy’s weak-topological lemma to the polynomial setting and motivates the use of the weak polynomial topology instead [2507.06143].

## 3. Equivalence with the classical Daugavet property

The main theorem establishing the equivalence \(DP\iff PDP\) proceeds by upgrading Shvidkoy’s lemma from the weak topology to the weak polynomial topology. The fundamental statement is that if \(X\) has DP, then for every \(x\in S_X\) and \(y\in B_X\) there exists a net \((y_\alpha)\subset B_X\) converging to \(y\) in the weak polynomial topology such that
\[
\|x+y_\alpha\|\to 2.
\]
Equivalently, for every \(x\in S_X\) and \(\varepsilon>0\), the set
\[
\{y\in B_X:\|x+y\|>2-\varepsilon\}
\]
is dense in \(B_X\) for the weak polynomial topology [2507.06143].

The proof combines three ingredients. First, it uses the quasi-codirected point argument from Shvidkoy’s original lemma. Second, it uses the Davie–Gamelin net construction in the bidual to ensure simultaneous approximation of polynomial values through the polynomial-star topology. Third, it uses an \(\ell_1\)-type convexity estimate: if \(x,y\in B_X\) and \(\|x+y\|>2-\varepsilon\) with \(\varepsilon\in(0,1)\), then for every \(\lambda\in(0,1)\),
\[
\|x+\lambda y\|>1+\lambda-\varepsilon.
\]
This estimate allows one to pass from almost quasi-codirected pairs to aggregated averages [2507.06143].

Once the weak-polynomial version of Shvidkoy’s lemma is available, the passage to PDP is short. Given \(x\in S_X\), \(\varepsilon>0\), and \(p\in\mathcal{P}(X)\) with \(\|p\|=1\), one chooses \(y\in B_X\) and \(\omega\) with \(|\omega|=1\) so that \(\operatorname{Re}(\omega p(y))>1-\varepsilon\). The weak polynomial net \((y_\alpha)\) approximating \(y\) preserves polynomial values, so \(p(y_\alpha)\to p(y)\), while simultaneously forcing \(\|x+\omega y_\alpha\|\to 2\). For large \(\alpha\), both
\[
\operatorname{Re}(\omega p(y_\alpha))>1-\varepsilon
\quad\text{and}\quad
\|x+\omega y_\alpha\|>2-\varepsilon
\]
hold, which is exactly the geometric characterization of PDP. The converse implication \(PDP\Rightarrow DP\) is immediate because linear operators are polynomials [2507.06143].

This result solves the open problem explicitly posed in MMP10, asking whether DP and PDP coincide for all Banach spaces. It also gives a conceptual reformulation of the Daugavet phenomenon: the essential geometry of the theory is preserved when the weak topology is replaced by the weak polynomial topology, which is the natural topology for polynomial data [2507.06143].

## 4. Earlier partial results and model classes

Before the general equivalence theorem, coincidence results for DP and PDP had been established only for specific classes of spaces. The 2025 equivalence theorem situates these earlier results within a single general statement, but the partial results remain structurally significant because their proofs often use tools specific to the ambient category [2507.06143].

In complex JB\(^*\)-triples and \(C^*\)-algebras, it was proved that the Daugavet property implies the polynomial Daugavet property, and that the alternative Daugavet property implies the alternative polynomial Daugavet property. The class includes diffuse \(C^*\)-algebras and JB\(^*\)-triples without minimal tripotents. The argument is based not on weak polynomial topology, but on the strong\(^*\) topology, the sequential strong\(^*\)-continuity of scalar polynomials, and the Peirce decomposition relative to tripotents. In particular, for a complex JB\(^*\)-triple \(E\), every weakly compact polynomial \(P:E\to E\) satisfies
\[
\|\mathrm{id}_E+P\|=1+\|P\|
\]
whenever \(E\) has the Daugavet property [2206.11621].

For vector-valued uniform-algebra function spaces \(A(K,X)\), the polynomial Daugavetian index gives a quantitative formulation of PDP. For every complex Banach space \(X\) and compact Hausdorff \(K\),
\[
\mathrm{Daug}_p\big(A(K,X)\big)=\max\{\mathrm{Daug}_p(A),\mathrm{Daug}_p(X)\},
\]
where \(A\) is the base uniform algebra. Consequently,
\[
A(K,X)\text{ has the PDP}\Longleftrightarrow \text{either }A\text{ has the PDP or }X\text{ has the PDP}.
\]
In the scalar case of infinite-dimensional uniform algebras, PDP, DP, DD2P, DLD2P, property \((\mathcal D)\), and the absence of isolated points in the Shilov boundary are equivalent [2302.11153].

For \(L_1\)-preduals and spaces of Lipschitz functions, earlier equivalence results were obtained through localized Daugavet geometry. In particular, if \(X\) is an \(L_1\)-predual or \(X=\mathrm{Lip}_0(M)\) for a pointed complete metric space \(M\), then \(DP\iff PDP\). The arguments use bidual perturbation sequences, Daugavet points, \(\Delta\)-points, and a criterion that converts suitable \(c_0\)-structured perturbations in the bidual into the polynomial Daugavet property [2010.15936].

A concise summary of representative classes is given below.

| Class of spaces | PDP conclusion | Source |
|---|---|---|
| Complex JB\(^*\)-triples and \(C^*\)-algebras with DP | DP implies PDP | [2206.11621] |
| \(A(K,X)\) over a uniform algebra \(A\) | \(A(K,X)\) has PDP iff \(A\) or \(X\) has PDP | [2302.11153] |
| \(L_1\)-preduals and \(\mathrm{Lip}_0(M)\) | DP iff PDP | [2010.15936] |
| Arbitrary Banach spaces | DP iff PDP | [2507.06143] |

## 5. Daugavet centers, weak operator variants, and tensor products

The general equivalence theorem extends beyond the identity operator. A non-zero operator \(G\in L(X,Y)\) is a Daugavet center if
\[
\|G+T\|=\|G\|+\|T\|
\]
for every rank-one operator \(T\in L(X,Y)\). A non-zero polynomial \(Q\in\mathcal{P}(X,Y)\) is a polynomial Daugavet center if
\[
\|Q+P\|=\|Q\|+\|P\|
\]
for every rank-one polynomial \(P\in\mathcal{P}(X,Y)\). The 2025 theory proves that every linear Daugavet center is a polynomial Daugavet center by establishing a Shvidkoy-type lemma in the weak polynomial topology: if \(G\) is a linear Daugavet center, then for every \(x\in B_X\) and \(y\in S_Y\) there exists a net \((x_\alpha)\subset B_X\) such that \(x_\alpha\to x\) in the weak polynomial topology and \(\|y+G(x_\alpha)\|\to 2\) [2507.06143].

The same paper treats the weak operator Daugavet property (WODP) and its polynomial counterpart (PWODP). For \(x_1,\dots,x_n\in B_X\), \(\varepsilon>0\), and \(x'\in B_X\), one defines \(\mathrm{OF}(x_1,\dots,x_n;x',\varepsilon)\) as the set of those \(x\in B_X\) for which there exists \(T\in L(X,X)\) with \(\|T\|\le 1\) such that
\[
\|T(x_i)-x_i\|<\varepsilon\quad (i=1,\dots,n)
\quad\text{and}\quad
\|T(x)-x'\|<\varepsilon.
\]
WODP requires every slice of \(B_X\) to meet such sets; PWODP replaces slices by polynomial tests. The new result is that WODP implies PWODP, again via density of \(\mathrm{OF}(x_1,\dots,x_n;x',\varepsilon)\) in the weak polynomial topology [2507.06143].

These operator and polynomial variants have direct consequences for symmetric tensor products. For \(N\in\mathbb N\), the \(N\)-fold projective symmetric tensor product \(\widehat{\otimes}_{\pi,s,N}X\) is the completion of \(\bigotimes^{s,N}X\) with norm
\[
\|u\|_{\pi,s,N}:=\inf\Big\{\sum_{i=1}^n |\lambda_i|\|x_i\|^N:\ u=\sum_{i=1}^n \lambda_i\,x_i^{\otimes N}\Big\},
\]
and
\[
(\widehat{\otimes}_{\pi,s,N}X)^*\cong \mathcal{P}(^N X).
\]
It is known that WODP is stable under projective tensor products and that PWODP transfers to symmetric tensor products. Combining these facts with \(WODP\Rightarrow PWODP\), one obtains: if \(X\) has WODP, then for every \(N\in\mathbb N\), \(\widehat{\otimes}_{\pi,s,N}X\) has WODP and hence has the Daugavet property [2507.06143].

This tensor direction continues and strengthens earlier work. It was shown previously that all symmetric projective tensor products of an \(L_1\)-predual with DP, and of \(L_1(\mu,Y)\) with \(\mu\) atomless \(\sigma\)-finite, have the Daugavet property, using WODP and polynomial WODP as intermediate notions [2010.15936]. Earlier tensor-product results also established Daugavet-type geometry, octahedrality, and diameter-two phenomena for symmetric tensor products and polynomial spaces, especially for \(C(K)\) with \(K\) perfect and for spaces with the operator Daugavet property [1903.01761].

## 6. Limitations, related variants, and open questions

The modern theory clarifies the exact scope of the polynomial Daugavet property but also leaves several natural problems open. One open question is whether there exists a Bourgain-type lemma for the weak polynomial topology: namely, whether every non-empty relatively weakly polynomially open subset of \(B_X\) contains a convex combination of slices. Such a statement would provide a direct route to the weak-polynomial Shvidkoy lemma. A second difficulty is that the weak polynomial topology is not always a linear topology, so standard tools from topological vector space geometry may fail [2507.06143].

Another open problem is whether DP implies WODP in general. The 2025 results show \(WODP\Rightarrow PWODP\), but they do not establish \(DP\Rightarrow WODP\). Likewise, for non-linear Daugavet centers, it remains unknown whether a polynomial \(Q\) satisfying
\[
\|Q+T\|=\|Q\|+\|T\|
\]
for all rank-one linear operators \(T\) must necessarily be a polynomial Daugavet center, meaning that the same identity holds against rank-one polynomials [2507.06143].

The alternative polynomial Daugavet property forms a related but distinct branch. In JB\(^*\)-triples and \(C^*\)-algebras, the alternative Daugavet property implies its polynomial version for all weakly compact polynomials [2206.11621]. By contrast, the implication can fail in general Banach spaces: complex \(\ell^1\) has the alternative Daugavet property but fails the alternative polynomial Daugavet property, and in the real case both \(c_0\) and \(\ell^1\) fail the alternative polynomial property [2206.11621].

Within multilinear and bilinear Daugavet theory, slice continuity furnishes a broader framework in which weak compactness and slice structure imply Daugavet equations for bilinear maps and suggest polynomial analogues via linearization on symmetric projective tensor products. This does not by itself produce the full vector-valued polynomial Daugavet equation in general, but it clarifies why tensorial and polynomial formulations of Daugavet geometry are tightly coupled [1210.7099].

In its current form, the subject has a clear central conclusion. The polynomial Daugavet property is no longer a separate strengthening that may or may not coincide with the classical one: for Banach spaces, it is an equivalent formulation of the same underlying extremal geometry once weak-topological arguments are replaced by their weak-polynomial counterparts [2507.06143].

Source: https://www.emergentmind.com/topics/polynomial-daugavet-property