betaFeedback

WorksMathematics

The ball fixed point property for continuous functions on locally compact spaces

Abstract

Let KK be a compact Hausdorff space and p∈Kp\in K. We prove that the Banach space C(K,p)={f∈C(K,R):f(p)=0}C(K,p)=\{f\in C(K,\mathbb R):f(p)=0\} fails the ball fixed point property (BFPP) whenever pp is non-isolated, answering a question of Avilés, Japón, Lennard, Martínez-Cervantes and Stawski. Consequently, C0(L)C_0(L) fails the BFPP for every noncompact locally compact Hausdorff space LL. Together with Barroso's characterization for C(K)C(K), this yields complete topological classifications of the BFPP for C0(L)C_0(L) and C(K,p)C(K,p). We also resolve a second question of Avilés et al., showing that the BFPP of C(K)C(K) implies that of C(K,p)C(K,p) for every compact Hausdorff space KK and every PP-point p∈Kp\in K if and only if there are no measurable cardinals.

Introduction

A Banach space XX has the ball fixed point property (BFPP) if every nonexpansive mapping from the closed unit ball of XX into itself has a fixed point. Unless otherwise specified, C(K)C(K) and C0(L)C_0(L) denote spaces of real-valued continuous functions equipped with the supremum norm.

For a compact Hausdorff space KK and p∈Kp\in K, define

C(K,p)={f∈C(K):f(p)=0}.C(K,p)=\{f\in C(K):f(p)=0\}.

Recall that p∈Kp\in K is a PP-point if every f∈C(K,p)f\in C(K,p) vanishes on a neighborhood of pp. Avilés, Japón, Lennard, Martínez-Cervantes and Stawski [1 Section 6] showed that C(K,p)C(K,p) fails BFPP if pp is not a PP-point, and asked whether the same conclusion holds for every non-isolated pp. Indeed, we show that C(K,p)C(K,p) fails the BFPP for every non-isolated point pp, regardless of whether pp is a PP-point. We also show that the converse of [1 Theorem 6.1], restricted to spaces C(K)C(K) with the BFPP, holds universally if and only if there are no measurable cardinals.

Our construction adapts the transfinite propagation technique introduced by Barroso [2]. The condition f(p)=0f(p)=0 allows us to replace the two profiles used in that work with a single decreasing profile. We also use the positive linear contraction of [2 Lemma 3.2].

The ball fixed point property in C(K,p)C(K,p)

Theorem 2.1. Let KK be a compact Hausdorff space, and suppose that p∈Kp\in K is non-isolated. Then C(K,p)C(K,p) fails the BFPP.

Lemma 2.2. Let KK be a compact Hausdorff space and p∈Kp\in K non-isolated. Then there are a limit ordinal τ\tau and an increasing family (uα)α<τ(u_\alpha)_{\alpha<\tau} in C(K,[0,1])C(K,[0,1]) such that u0=0u_0=0, each uαu_\alpha vanishes near pp, and, with Fα={uα=1}F_\alpha=\{u_\alpha=1\},

p∈⋃α<τFα.(1)p\in\bigcup_{\alpha<\tau}F_\alpha. \tag*{(1)}

Proof. Enumerate K∖{p}={xα:α<κ}K \setminus\{p\} = \{x_\alpha: \alpha< \kappa\}, where κ\kappa is an infinite cardinal, and set u0=0u_0 = 0. We proceed by transfinite recursion. Having constructed (uβ)β<α(u_\beta)_{\beta< \alpha}, define

Hα=⋃β<αsupp⁡uβ‾.H_\alpha= \bigcup_{\beta< \alpha} \overline{\operatorname{supp} u_\beta}.

Stop if p∈Hαp \in H_\alpha. Otherwise, for 0<α<κ0 < \alpha< \kappa, normality of KK yields uα∈C(K,[0,1])u_\alpha\in C(K,[0,1]) such that uα=1u_\alpha= 1 on HαH_\alpha, uαu_\alpha vanishes on a neighborhood of pp, and uα(xβ)=1u_\alpha(x_\beta) = 1 whenever α=β+1\alpha= \beta+ 1. By construction, the functions are increasing and Hα+1=supp⁡uαH_{\alpha+1} = \operatorname{supp} u_\alpha.

The recursion must stop by stage κ\kappa. Otherwise, we have that

xβ∈Fβ+1⊆Hκ(β<κ).x_\beta\in F_{\beta+1} \subseteq H_\kappa\qquad(\beta< \kappa).

Since HκH_\kappa is closed and pp is non-isolated, this implies p∈Hκp \in H_\kappa, a contradiction. Let τ\tau be the first stopping ordinal. Since Hα+1=supp⁡uαH_{\alpha+1} = \operatorname{supp} u_\alpha does not contain pp, the ordinal τ\tau is a limit. Moreover,

supp⁡uα⊆Fα+1(α<τ).\operatorname{supp} u_\alpha\subseteq F_{\alpha+1} \qquad(\alpha< \tau).

Consequently,

p∈Hτ=⋃α<τsupp⁡uα‾⊆⋃α<τFα.p \in H_\tau= \bigcup_{\alpha< \tau} \overline{\operatorname{supp} u_\alpha} \subseteq\bigcup_{\alpha< \tau} F_\alpha.

Lemma 2.3. Let KK be a compact Hausdorff space, let τ\tau be a limit ordinal, and let (uα)α<τ(u_\alpha)_{\alpha< \tau} be an increasing family in C(K,[0,1])C(K,[0,1]). There is a positive linear contraction

S:C0([0,τ))⟶C(K),S1[0,α]=uα(α<τ).S : C_0([0,\tau)) \longrightarrow C(K), \qquad S1_{[0,\alpha]} = u_\alpha\qquad(\alpha< \tau).

If p∈Kp \in K and uα(p)=0u_\alpha(p) = 0 for every α<τ\alpha< \tau, then the range of SS is contained in C(K,p)C(K,p).

Proof. Let eα=1[0,α]e_\alpha= 1_{[0,\alpha]}. These functions are linearly independent, and their span D=span⁡{eα:α<τ}D = \operatorname{span}\{e_\alpha: \alpha< \tau\} is dense in C0([0,τ))C_0([0,\tau)) by the Stone–Weierstrass theorem. Define S:D→C(K)S : D \to C(K) by Seα=uαSe_\alpha= u_\alpha.

For α1<⋯<αn\alpha_1 < \cdots< \alpha_n, the functions

eα1,eα2−eα1,…,eαn−eαn−1e_{\alpha_1}, \qquad e_{\alpha_2} - e_{\alpha_1}, \qquad\ldots, \qquad e_{\alpha_n} - e_{\alpha_{n-1}}

are nonnegative and have pairwise disjoint supports. Their images under SS are nonnegative, so SS is positive. Moreover, every g∈span⁡{eα1,…,eαn}g \in\operatorname{span}\{e_{\alpha_1},\ldots,e_{\alpha_n}\} satisfies

−∥g∥∞eαn≤g≤∥g∥∞eαn.-\lVert g\rVert_\infty e_{\alpha_n} \leq g \leq\lVert g\rVert_\infty e_{\alpha_n}.

By positivity,

∣Sg∣≤∥g∥∞uαn≤∥g∥∞.\lvert Sg\rvert\leq\lVert g\rVert_\infty u_{\alpha_n} \leq\lVert g\rVert_\infty.

Therefore SS extends uniquely to a positive linear contraction on C0([0,τ))C_0([0,\tau)). If uα(p)=0u_\alpha(p) = 0 for every α<τ\alpha< \tau, then Sg(p)=0Sg(p) = 0 for every g∈Dg \in D, and hence for every g∈C0([0,τ))g \in C_0([0,\tau)) by continuity.

Proof of Theorem 2.1. Let (uα)α<τ(u_\alpha)_{\alpha< \tau} and SS be as in Lemmas 2.2 and 2.3, and define

C={f∈C(K,p):0≤f≤1}.\mathcal{C} = \{f \in C(K,p) : 0 \leq f \leq1\}.

For f∈Cf \in\mathcal{C} and α<τ\alpha< \tau, let

mα(f)={min⁡Fαf,Fα≠∅,1,Fα=∅.m_\alpha(f) = \begin{cases} \min_{F_\alpha} f, & F_\alpha\ne\varnothing,\\ 1, & F_\alpha= \varnothing. \end{cases}

Since the sets FαF_\alpha are increasing and p∈⋃α<τFαp \in\bigcup_{\alpha< \tau} F_\alpha, continuity of ff at pp gives

m0(f)=1,mα(f)≥mβ(f)(α≤β),inf⁡α<τmα(f)=0.(2)m_0(f) = 1, \qquad m_\alpha(f) \geq m_\beta(f) \quad(\alpha\leq\beta), \qquad\inf_{\alpha< \tau} m_\alpha(f) = 0. \tag*{(2)}

Define hf:[0,τ)→[0,1]h_f : [0,\tau) \to[0,1] by

hf(0)=1,hf(α+1)=mα(f),hf(λ)=inf⁡α<λmα(f)(3)\begin{aligned} h_f(0) &= 1,\\ h_f(\alpha+1) &= m_\alpha(f),\\ h_f(\lambda) &= \inf_{\alpha<\lambda} m_\alpha(f) \tag*{(3)} \end{aligned}

for nonzero limit ordinals λ<τ\lambda< \tau. By (2), hfh_f is decreasing, continuous in the order topology, and vanishes at infinity. Moreover, since minima over fixed compact sets and infima of bounded families are nonexpansive, we have that

∥hf−hg∥∞≤∥f−g∥∞(f,g∈C).(4)\lVert h_f-h_g\rVert_\infty\le\lVert f-g\rVert_\infty\qquad(f,g\in\mathcal{C}). \tag*{(4)}

Define T:C→CT:\mathcal{C}\to\mathcal{C} by Tf=ShfTf=Sh_f. Positivity and contractivity of SS, together with (4), imply that TT is nonexpansive. Writing eα=1[0,α]e_\alpha=1_{[0,\alpha]}, we have

hf≥hf(α)eα,Tf≥hf(α)uα,h_f \ge h_f(\alpha)e_\alpha,\qquad Tf\ge h_f(\alpha)u_\alpha,

and consequently

mα(Tf)≥hf(α)(α<τ).(5)m_\alpha(Tf)\ge h_f(\alpha)\qquad(\alpha<\tau). \tag*{(5)}

Suppose Tf=fTf=f. By (3) and (5),

mα+1(f)≥mα(f)(α+1<τ),m_{\alpha+1}(f)\ge m_\alpha(f)\qquad(\alpha+1<\tau),

and

mλ(f)≥inf⁡α<λmα(f)m_\lambda(f)\ge\inf_{\alpha<\lambda}m_\alpha(f)

for every nonzero limit ordinal λ<τ\lambda<\tau. Since m0(f)=1m_0(f)=1, transfinite induction gives

mα(f)=1(α<τ),m_\alpha(f)=1\qquad(\alpha<\tau),

contradicting (2). Hence TT has no fixed point.

Finally, define R:BC(K,p)→CR:B_{\mathcal{C}(K,p)}\to\mathcal{C} such that

Rf=f+=max⁡{f,0}.Rf=f^+=\max\{f,0\}.

Then RR is a nonexpansive retraction. So T∘RT\circ R is a fixed-point-free nonexpansive mapping on BC(K,p)B_{\mathcal{C}(K,p)} and consequently C(K,p)\mathcal{C}(K,p) fails the BFPP. □

Corollaries

Corollary 3.1. If LL is a noncompact locally compact Hausdorff space, then C0(L)C_0(L) fails the BFPP.

Proof. Let K=L∪{∞}K=L\cup\{\infty\} be the one-point compactification of LL. Since ∞\infty is non-isolated and C0(L)C_0(L) is isometric to C(K,∞)C(K,\infty), Theorem 2.1 implies that C0(L)C_0(L) fails the BFPP. □

A compact space is called extremally disconnected if the closure of every open subset is open. Barroso [2 Corollary 5.4] proved that C(K)C(K) has BFPP precisely for extremally disconnected compact spaces KK. Together with Corollary 3.1, this yields the following classification.

Corollary 3.2. For a locally compact Hausdorff space LL,

C0(L) has BFPP⟺L is compact and extremally disconnected.C_0(L)\text{ has BFPP}\quad\Longleftrightarrow\quad L\text{ is compact and extremally disconnected}.

For compact Hausdorff KK and p∈Kp\in K, equivalently,

C(K,p) has BFPP⟺p is isolated and K is extremally disconnected.C(K,p)\text{ has BFPP}\quad\Longleftrightarrow\quad p\text{ is isolated and }K\text{ is extremally disconnected}.

Proof. The characterization of C0(L)C_0(L) follows from Corollary 3.1 and [2]. For C(K,p)C(K,p), observe that C(K,p)C(K,p) is isometric to C0(K∖{p})C_0(K\setminus\{p\}). The space K∖{p}K\setminus\{p\} is compact precisely when pp is isolated, in which case it is extremally disconnected precisely when KK is extremally disconnected. □

Corollary 3.3. The following are equivalent:

  1. There are no measurable cardinals.

  2. For every compact Hausdorff space KK and every P-point p∈Kp \in K, if C(K)C(K) has the BFPP, then C(K,p)C(K,p) has the BFPP.

Proof. Suppose there are no measurable cardinals. If C(K)C(K) has the BFPP, then KK is extremally disconnected by [2], and every P-point of KK is isolated by [3 Section 4.4]. The conclusion follows from Corollary 3.2.

Conversely, suppose that a measurable cardinal κ\kappa exists. Let p∈βκp \in\beta\kappa correspond to a nonprincipal countably complete ultrafilter on κ\kappa. Then pp is a non-isolated P-point of βκ\beta\kappa. Since βκ\beta\kappa is extremally disconnected, C(βκ)C(\beta\kappa) has the BFPP, whereas C(βκ,p)C(\beta\kappa,p) fails the BFPP by Theorem 2.1. □

Corollary 3.4. Let II be a closed ideal of C(K)C(K). Then II has the BFPP if and only if

I=1UC(K)I = 1_U C(K)

for some clopen, extremally disconnected subset UU of KK.

Proof. Every closed ideal II of C(K)C(K) has the form

I={f∈C(K):f∣F=0}≅C0(K∖F)I = \{f \in C(K) : f|_F = 0\} \cong C_0(K \setminus F)

for some closed subset F⊆KF \subseteq K. Let U=K∖FU = K \setminus F. By Corollary 3.2, II has the BFPP if and only if UU is compact and extremally disconnected. Since UU is open in KK, it is compact if and only if it is clopen, in which case I=1UC(K)I = 1_U C(K). □

Corollary 3.5. Let LL be a noncompact locally compact Hausdorff space and EE a nonzero real or complex Banach space. Then the underlying real Banach space C0(L,E)C_0(L,E) fails the BFPP.

Proof. Choose e∈Ee \in E with ∥e∥=1\lVert e\rVert= 1 and a norm-one real-linear functional φ:E→R\varphi: E \to\mathbb{R} satisfying φ(e)=1\varphi(e) = 1. Define

P:C0(L,E)→C0(L),(Pf)(t)=φ(f(t)),P : C_0(L,E) \to C_0(L), \qquad(Pf)(t) = \varphi(f(t)),
J:C0(L)→C0(L,E),(Jg)(t)=g(t)e.J : C_0(L) \to C_0(L,E), \qquad(Jg)(t) = g(t)e.

Both operators are real-linear contractions, and PJ=id⁡C0(L)PJ = \operatorname{id}_{C_0(L)}.

By Corollary 3.1, there exists a fixed-point-free nonexpansive mapping A:BC0(L)→BC0(L)A : B_{C_0(L)} \to B_{C_0(L)}. Consequently,

JAP:BC0(L,E)→BC0(L,E)JAP : B_{C_0(L,E)} \to B_{C_0(L,E)}

is nonexpansive. If f=JAPff = JAPf, then

Pf=PJAPf=APf,Pf = PJAPf = APf,

contradicting the absence of fixed points of AA. □

Corollary 3.6. Every nonzero nonunital commutative complex C∗C^*-algebra fails the BFPP when regarded as a real Banach space.

Proof. By the commutative Gelfand–Naimark theorem, such an algebra is isometrically isomorphic to C0(L,C)C_0(L,\mathbb{C}) for some locally compact Hausdorff space LL. Since the algebra is nonunital, LL is noncompact, and the conclusion follows from Corollary 3.5. □

Acknowledgements

The author used OpenAI’s ChatGPT 6.1 Sol to assist with the development and review of proofs and the preparation of this manuscript. The author assumes full responsibility for the results and final content.

References

References

  1. [1]A. Avilés, M. Japón, C. J. Lennard, G. Martínez-Cervantes, and A. Stawski, The ball fixed point property in spaces of continuous functions, Rev. Real Acad. Cienc. Exactas Fis. Nat. Ser. A-Mat. 120 (2026), article 136, doi:10.1007/s13398-026-01934-1.DOI
  2. [2]C. S. Barroso, Transfinite contractive propagation and the ball fixed point property in C(K), arXiv:2609.16318 (2026).arxiv.org/abs/2609.16318
  3. [3]J. Terasawa, Extremally disconnected spaces, in Encyclopedia of General Topology, Elsevier, 2004, pp. 345–349.DOI

Paper details

vlnc:W_gzykjq3s
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

GPT-6.1 Sol (bare)

View statement
Looking for related papers…
Contents