---
title: Random Sperner Lemma Overview
url: https://www.emergentmind.com/topics/random-sperner-lemma
type: topic
---

# Random Sperner Lemma Overview

Searching arXiv for the cited papers and closely related work to ground the article.
Searching for the primary paper on the random Sperner lemma.
Random Sperner lemma is the extension of the classical Sperner lemma from finite simplicial subdivisions to the setting of random Euclidean spaces and \(L^0(\mathcal F)\)-modules. In the formulation established in "Random Sperner lemma and random Brouwer fixed point theorem" [2507.08521], the objects of classical combinatorial topology are replaced by \(L^0\)-simplexes, \(L^0\)-simplicial subdivisions, and \(\sigma\)-stable labeling functions with values in \(L^0(\mathcal F;\{0,1,\dots,n\})\). The central conclusion is the existence of a fully labeled random subsimplex obtained by finite measurable concatenation. This provides the combinatorial mechanism for a complete proof of the random Brouwer fixed point theorem in random Euclidean spaces and unifies several earlier random fixed-point frameworks [2507.08521].

## 1. Probabilistic and \(L^0\)-module framework

The ambient setting is a complete probability space \((\Omega,\mathcal F,\mathbb P)\), with random variables understood modulo almost sure equality. The algebra of real-valued random variables is denoted \(L^0(\mathcal F)\). For \(n\in\mathbb N\), the random Euclidean space \(L^0(\mathcal F;\mathbb R^n)\) is the free \(L^0(\mathcal F)\)-module of rank \(n\), consisting of equivalence classes of \(\mathbb R^n\)-valued random vectors [2507.08521].

Two topologies are used. The \((\epsilon,\lambda)\)-topology \(\mathcal T_{\epsilon,\lambda}\) is convergence in probability and is a metrizable linear topology. The locally \(L^0\)-convex topology \(\mathcal T_c\) is generated by random open balls
\[
B(x,r)=\{y:\|x-y\|<r\text{ on }\Omega\},\qquad r\in L^0_{++}(\mathcal F).
\]
Both topologies make \(L^0(\mathcal F;\mathbb R^n)\) a topological \(L^0\)-module.

The stability structure is essential. A subset \(G\subset L^0(\mathcal F;\mathbb R^n)\) is \(\sigma\)-stable if, for any sequence \(\{x_k\}\subset G\) and any countable partition \(\{A_k\}\subset\mathcal F\) of \(\Omega\), the countable concatenation \(\sum_k \tilde I_{A_k}x_k\) belongs to \(G\). A mapping \(f:G\to H\) on \(\sigma\)-stable sets is \(\sigma\)-stable if
\[
f\Big(\sum \tilde I_{A_k}x_k\Big)=\sum \tilde I_{A_k}f(x_k)
\]
for every such sequence and partition. The same framework uses \(L^0\)-convexity: \(G\subset L^0(\mathcal F;\mathbb R^n)\) is \(L^0\)-convex if, for \(x,y\in G\) and \(\lambda\in L^0(\mathcal F)\) with \(0\le \lambda\le 1\), one has \(\lambda x+(1-\lambda)y\in G\).

The module \(L^0(\mathcal F;\mathbb R^n)\) carries the \(L^0\)-norm
\[
\|x\|=\Big(\sum_{i=1}^n x_i^2\Big)^{1/2},
\]
so it is also an RN module. The measure algebra \(B_{\mathcal F}\), the Boolean algebra of measurable sets modulo null sets, provides the eventwise bookkeeping for measurable concatenations. For \(a\in B_{\mathcal F}\), \(I_a\) denotes the equivalence class of the indicator of any representative set of \(a\), and partitions in \(B_{\mathcal F}\) are written \(\{a_i\}\in p(1)\).

## 2. \(L^0\)-simplexes and random simplicial subdivisions

Let \(E\) be an \(L^0(\mathcal F)\)-module. A finite set \(\{V_0,\dots,V_n\}\subset E\) is \(L^0\)-affinely independent if either \(n=0\) or \(\{V_i-V_0:i=1,\dots,n\}\) is \(L^0\)-linearly independent. The associated \(L^0\)-simplex is
\[
S=\operatorname{conv}_{L^0}\{V_0,\dots,V_n\}
=\Big\{\sum_{i=0}^n \lambda_iV_i:\lambda_i\in L^0(\mathcal F),\ \lambda_i\ge 0,\ \sum_{i=0}^n\lambda_i=1\Big\}.
\]
In \(L^0(\mathcal F;\mathbb R^n)\), each \(x\in S\) admits a unique \(L^0\)-barycentric coordinate vector \((\lambda_0(x),\dots,\lambda_n(x))\) with \(\lambda_i(x)\in L^0_+(\mathcal F)\), \(\sum\lambda_i(x)=1\), and \(x=\sum\lambda_i(x)V_i\) [2507.08521].

A decisive structural feature is the deterministic counterpart simplex
\[
S'=\operatorname{conv}\{V_0,\dots,V_n\}\subset E,
\]
viewed in the underlying real vector space. Since \(L^0\)-affine independence implies classical affine independence, the paper uses \(S'\) to import classical combinatorics into the random setting. In particular,
\[
\operatorname{ext}(S')=\{V_0,\dots,V_n\},\qquad \operatorname{ext}_{L^0}(S)=\sigma(\operatorname{ext}(S')).
\]
Thus the random extreme points are countable concatenations of deterministic extreme points.

An \(L^0\)-simplicial subdivision of \(S\) is a finite family \(\mathcal S\) of \(n\)-vertex \(L^0\)-simplexes satisfying three conditions: \(S=\sigma\{\cup\mathcal S\}\); any two members intersect either trivially or in a common \(L^0\)-face; and the counterpart family \(\mathcal S'=\{G':G\in\mathcal S\}\) is a classical simplicial subdivision of \(S'\). This third condition is the formal bridge between random and classical geometry. The theory does not require fiberwise triangulations \(\omega\mapsto\{V_i(\omega)\}\) in \(\mathbb R^n\); instead, it proceeds through deterministic counterpart subdivisions and \(\sigma\)-stable concatenation, thereby avoiding delicate measurability issues [2507.08521].

A useful closure property under concatenation is explicit. If \(\{S^m=\operatorname{conv}_{L^0}\{Y_0^m,\dots,Y_n^m\}\}\) is a sequence of \(L^0\)-simplexes in \(S\) and \(\{a_m\}\in p(1)\), then
\[
\sum_m I_{a_m}S^m
\]
is again an \(L^0\)-simplex with vertices \(\sum_m I_{a_m}Y_i^m\), \(i=0,\dots,n\), and is contained in \(S\).

## 3. Proper \(L^0\)-labelings and the representation theorem

Let \(\mathcal S\) be an \(L^0\)-simplicial subdivision of
\[
S=\operatorname{conv}_{L^0}\{V_0,\dots,V_n\}.
\]
A labeling function is a \(\sigma\)-stable mapping
\[
\phi:\operatorname{ext}_{L^0}(\mathcal S)\to L^0(\mathcal F;\{0,1,\dots,n\}).
\]
It is proper if, for every \(y\in\operatorname{ext}_{L^0}(\mathcal S)\) and every \(i\in\{0,1,\dots,n\}\), writing \(y=\sum_{i=0}^n\lambda_i(y)V_i\),
\[
(\phi(y)=i)\wedge(\lambda_i(y)=0)=0.
\]
Equivalently, almost surely on the event \(\{\phi(y)=i\}\), one has \(\lambda_i(y)>0\). This is the random analogue of the classical Sperner boundary condition: a vertex may receive only a label corresponding to a vertex spanning the face that contains it [2507.08521].

An \(L^0\)-simplex \(G=\operatorname{conv}_{L^0}\{Y_0,\dots,Y_n\}\) is completely labeled by \(\phi\) if
\[
\bigvee_{j=0}^n(\phi(Y_j)=i)=1
\qquad\text{for every }i\in\{0,1,\dots,n\}.
\]
Thus every label appears somewhere among the vertices with probability one.

The key structural result is the representation theorem for proper \(L^0\)-labelings. If \(\operatorname{ext}(\mathcal S')=\{y_k:1\le k\le K\}\), then at deterministic extreme points the random label \(\phi(y)\) assigns values only from the deterministic face-index set
\[
\chi(y):=\{i:\alpha_i>0\},
\]
where \(y=\sum_{i=0}^n\alpha_iV_i\) with \(\alpha_i\ge 0\), \(\sum\alpha_i=1\). More precisely, \(N_{\phi(y)}\subset \chi(y)\), where \(N_{\phi(y)}\) is the set of indices with positive mass in the random label \(\phi(y)\).

The theorem then states that proper classical labeling functions \(\psi_i\) can be chosen so that the random label decomposes pointwise at deterministic extreme points, and globally there exist finitely many proper classical labeling functions \(\{\psi_m:m=1,\dots,M\}\) on the deterministic subdivision \(\mathcal S'\) together with a finite partition \(\{a_m\}\in p(1)\) such that, for every random extreme point
\[
y=\sum_{k=1}^K I_{b_k}y_k\in \operatorname{ext}_{L^0}(\mathcal S),
\]
with \(\{b_k\}\in p(1)\),
\[
\phi(y)=\sum_{k=1}^K I_{b_k}\sum_{m=1}^M I_{a_m}\psi_m(y_k).
\]
This decomposition is the essential bridge from random proper labelings to finitely many deterministic proper labelings on \(\mathcal S'\) [2507.08521]. A plausible implication is that the randomization is not carried by a pointwise random triangulation, but by measurable assembly of finitely many deterministic combinatorial patterns.

## 4. Statement and proof architecture of the random Sperner lemma

The random Sperner lemma asserts that if \(S\) is an \(n\)-vertex \(L^0\)-simplex in an \(L^0(\mathcal F)\)-module \(E\), \(\mathcal S\) is an \(L^0\)-simplicial subdivision of \(S\), and \(\phi\) is a proper \(L^0\)-labeling of \(\mathcal S\), then there exist finitely many \(n\)-vertex \(L^0\)-simplexes
\[
S_m=\operatorname{conv}_{L^0}\{Y_0^m,\dots,Y_n^m\},
\qquad Y_j^m\in \operatorname{ext}_{L^0}(\mathcal S),
\]
and a finite partition \(\{a_m\}\in p(1)\) such that the concatenated \(L^0\)-simplex
\[
\sum_{m=1}^M I_{a_m}S_m
\]
is completely labeled by \(\phi\) and satisfies, for every \(i\in\{0,1,\dots,n\}\),
\[
\phi\Big(\sum_{m=1}^M I_{a_m}Y_i^m\Big)=i.
\]
Equivalently,
\[
\sum_{m=1}^M I_{a_m}S_m
=
\operatorname{conv}_{L^0}\Big\{\sum_{m=1}^M I_{a_m}Y_0^m,\dots,\sum_{m=1}^M I_{a_m}Y_n^m\Big\}
\]
is a fully labeled random subsimplex of \(S\) [2507.08521].

The proof has three steps. First, the representation theorem expresses \(\phi\) as a finite concatenation of classical proper labeling functions \(\psi_m\) on the deterministic subdivision \(\mathcal S'\). Second, classical Sperner is applied to each \(\psi_m\), producing a completely labeled simplex \(S'_m\in\mathcal S'\). Third, these deterministic simplexes are lifted back to \(L^0\)-simplexes in \(E\) and concatenated along the partition \(\{a_m\}\). The \(\sigma\)-stability of the construction guarantees that the result is again an \(L^0\)-simplex and that the label identities are preserved.

The lemma is therefore neither a fiberwise application of classical Sperner at each \(\omega\in\Omega\) nor a purely abstract existence argument. The combinatorics is entirely housed in the deterministic counterpart subdivision \(\mathcal S'\), while randomness enters through measurable partitions in \(B_{\mathcal F}\) and the concatenation machinery. The paper emphasizes that classical parity, such as oddness of the number of fully labeled simplexes, is not required for the random conclusion; only existence is assembled measurably [2507.08521].

## 5. Application to the random Brouwer fixed point theorem

The principal application is a fixed-point theorem on \(L^0\)-simplexes. Let
\[
S=\operatorname{conv}_{L^0}\{V_0,\dots,V_n\}\subset E
\]
be an \(L^0\)-simplex in an RN module \(E\), and let \(f:S\to S\) be \(\sigma\)-stable and random sequentially continuous, equivalently \(\mathcal T_c\)-continuous on \(\sigma\)-stable sets. Then \(f\) has a fixed point [2507.08521].

The labeling used in the proof is Sperner-style. For
\[
x=\sum \lambda_i^xV_i,\qquad f(x)=\sum \mu_i^xV_i,
\]
one chooses \(\phi(x)\) to be any index \(i\) with \(\lambda_i^x>0\) and \(\lambda_i^x\ge \mu_i^x\). The paper states that such an \(i\) exists almost surely, with the proof relying on measure-algebra arguments. This produces a \(\sigma\)-stable proper labeling on any \(L^0\)-simplicial subdivision of \(S\). Applying the random Sperner lemma to successive \(L^0\)-barycentric subdivisions \(\mathcal S_m\) yields completely labeled \(L^0\)-simplexes \(S^m\) whose diameters satisfy
\[
\operatorname{diam}(\mathcal S_m)\le \Big(\frac{n}{n+1}\Big)^m\operatorname{diam}(S).
\]
Using random sequential compactness of \(S\) together with random sequential continuity of \(f\), one extracts a convergent random subsequence \(y_i^{m_k}\to y\) and \(f(y_i^{m_k})\to f(y)\). The labeling inequalities force equality of all barycentric coordinates, \(\lambda_i(y)=\mu_i(y)\), hence \(f(y)=y\).

The general random Brouwer theorem extends this from simplexes to random convex sets in \(L^0(\mathcal F;\mathbb R^d)\). If \(G\subset L^0(\mathcal F;\mathbb R^d)\) is \(\sigma\)-stable, almost surely bounded, \(L^0\)-convex, and \(\mathcal T_c\)-closed, and \(f:G\to G\) is \(\sigma\)-stable and \(\mathcal T_c\)-continuous, then \(f\) has a fixed point. The reduction chooses an \(L^0\)-simplex \(S\) containing \(G\), uses the nonexpansive, \(\sigma\)-stable random projection \(P_G\) onto \(G\), defines \(h=P_G|_S\), and sets \(\hat h=f\circ h:S\to G\subset S\). A fixed point \(z\) of \(\hat h\) lies in \(G\) and satisfies \(f(z)=z\) [2507.08521].

The same work also proves an equivalence with a random Borsuk-type statement: the random Brouwer fixed point theorem is equivalent to the nonexistence of a \(\sigma\)-stable \(\mathcal T_c\)-retraction from the random closed unit ball \(L^0(\mathcal F;\bar B_1(\mathbb R^d))\) onto the random unit sphere \(L^0(\mathcal F;\partial \bar B_1(\mathbb R^d))\).

## 6. Unification, example, and distinction from other “random Sperner” results

One of the stated aims is unification. The paper proves equivalence among the \(\mathcal T_c\) formulation above, the \(\mathcal T_{\epsilon,\lambda}\) version for \(\sigma\)-stable and \(\mathcal T_{\epsilon,\lambda}\)-continuous maps on almost surely bounded, \(\mathcal T_{\epsilon,\lambda}\)-closed, \(L^0\)-convex sets, and Ponosov’s stochastic formulation for local, continuous-in-probability operators on measurable selection sets \(G_U\) represented by Carathéodory functions [2507.08521]. It also clarifies that a \(\mathcal T_{\epsilon,\lambda}\)-continuous \(\sigma\)-stable map is automatically almost surely sequentially continuous. This suggests that the random Sperner lemma functions as a common combinatorial core behind several apparently different random fixed-point theories.

A concrete two-dimensional illustration uses
\[
S=\operatorname{conv}_{L^0}\{X_1,X_2,X_3\}\subset L^0(\mathcal F;\mathbb R^2)
\]
and the midpoints
\[
Y_{1,2}=(X_1+X_2)/2,\qquad
Y_{1,3}=(X_1+X_3)/2,\qquad
Y_{2,3}=(X_2+X_3)/2.
\]
The subdivision \(\mathcal S\) consists of the four \(L^0\)-triangles
\[
\operatorname{conv}_{L^0}\{X_1,Y_{1,2},Y_{1,3}\},\quad
\operatorname{conv}_{L^0}\{X_2,Y_{1,2},Y_{2,3}\},\quad
\operatorname{conv}_{L^0}\{X_3,Y_{1,3},Y_{2,3}\},\quad
\operatorname{conv}_{L^0}\{Y_{1,2},Y_{1,3},Y_{2,3}\}.
\]
The paper shows that this satisfies the definition of \(L^0\)-simplicial subdivision. For any proper \(L^0\)-labeling \(\phi\), the random Sperner lemma yields a finite partition \(\{a_m\}\) and random \(L^0\)-triangles \(S_m\) from \(\mathcal S\) such that \(\sum_m I_{a_m}S_m\) is completely labeled [2507.08521].

The expression “random Sperner” also appears in a different literature concerning Sperner’s theorem on antichains in the Boolean lattice. That direction studies random induced subposets of \(\mathcal P([n])\) and asks when the largest antichain is given by a middle layer intersection. In particular, "A sharp threshold for a random version of Sperner's Theorem" identifies a sharp threshold at \(p=3/4\) for that combinatorial property [2205.11630]. This is unrelated to the topological Sperner lemma on triangulations. The distinction is substantive: the \(L^0\)-module theory concerns random labelings on random simplexes and leads to random Brouwer and Borsuk results, whereas the Boolean-lattice theory concerns widths of random posets and antichain structure [2205.11630].

The main subtleties relative to classical Sperner are therefore threefold. First, labels are random variables in \(L^0(\mathcal F;\{0,\dots,n\})\), so properness is eventwise:
\[
(\phi(y)=i)\wedge(\lambda_i(y)=0)=0.
\]
Second, measurable existence of a fully labeled subsimplex relies on \(\sigma\)-stability and concatenation over partitions in \(B_{\mathcal F}\), not on fiberwise triangulations. Third, the actual simplicial combinatorics remains deterministic at the level of the counterpart subdivision \(\mathcal S'\), while the random object is assembled from finitely many deterministic labeled simplexes by measurable concatenation [2507.08521].

Source: https://www.emergentmind.com/topics/random-sperner-lemma