• 検索結果がありません。

Sperner Matroid and Sperner Map (Nonlinear Analysis and Convex Analysis)

N/A
N/A
Protected

Academic year: 2021

シェア "Sperner Matroid and Sperner Map (Nonlinear Analysis and Convex Analysis)"

Copied!
4
0
0

読み込み中.... (全文を見る)

全文

(1)

Sperner

Matroid

and Sperner

Map

By

Shyh-Nan Lee$(1)$

DepartmentofMathemathies, Chung Yuan Christian University,

Chung Li 320, Taiwan

E-mail address: [email protected]

and

Mau-Hsiang Shih

DepartmentofMathematics, National Taiwan Normal University,

88Sec.4, Ting ChouRoad, Taipei 116, Taiwan

E-mail address: [email protected]

Abstract. Areconstruction of aSperner map from aSperner matroid is illustrated. As anapplication

of this reconstruction,

we

give

anew

proofof acompletion theorem ofLovasz’s matroid version of Sperner’s

lemma.

The purpose of this note is togive asimple procedure which allows us to retrieve aSperner map ffom a

Spernermatroid.

1. Reconstruction ofSperner map from Sperner matroid

Let $K$ be atriangulation ofa$d$-simplex $a_{0}a_{1}\ldots$$a_{d}$ in aEuclidean space, and $V(K)$ the vertex set of$K$

.

Amap$\varphi$ :$V(K)arrow\{0,1, \ldots, d\}$ is said to be aSper$mer$map if for each

$i_{0}$,$i_{1}$,

$\ldots$,$i_{k}$ with $0\leq i0<i_{1}<\ldots<$

$i_{k}\leq d$and for each$v\in \mathrm{V}(\mathrm{K})\cap a_{i_{\mathrm{O}}}a_{i_{1}}\ldots$$a_{i_{k}}$, $\varphi(v)\in\{i_{0}, i_{1}, \ldots, i_{k}\}$

.

Amatroid $M$

on

$V(K)$is called aSperner

matroidover $K$iffor each $S\subset\{a_{0}, a_{1}, \ldots, a_{d}\}$ and for each$v\in V(K)\cap \mathrm{c}\mathrm{o}\mathrm{n}v(\mathrm{S})$, $v\in d_{M}(S)$, where$cmv(S)$

stands for theconvex hull of$S$and $clM\{S$) denotes the closure of$S$in $M$

.

Let $M$ be aSperner matroidover

$K$ such that $\{a\mathit{0}, a_{1}, \ldots, a_{d}\}$ forms abasis of$M$

.

Put

$F_{j}\equiv d_{M}(\{a_{0}, a_{1}, \ldots, a_{j}\})$ $(j=0,1, \ldots, d)$

.

(1) Research supported in part by theNational Science Councilofthe Republicof China

数理解析研究所講究録 1298 巻 2002 年 161-164

(2)

Let us define$\varphi$ :$V(K)arrow\{0,1, \ldots, d\}$ by setting

$\varphi(v)=0$ if $v\in F_{0}$,

$\varphi(v)=j$ if $v\in F_{j}\backslash F_{j-1}$ $(j=1,2, \ldots, d)$

.

The problem that weconsider in this paper is the following :Under what condition is$\varphi$ a Sperner map? Itis

clearly necessary that the $(d-1)$-face$a_{1}a_{2}\ldots$$a_{d}$ containsnoloops. Tosee this, if$a_{1}a_{2}\ldots$$a_{d}$ contains aloop

$v$ of$M$, then $\varphi(v)=0$ and $v\in V(K)\cap a_{1}a_{2}\ldots$$a_{d}$

.

As $\varphi$ is aSperner map, wehave $\varphi(v)\in\{1, 2, \ldots, d\}$, in

contradiction. What is perhaps surprising is that thisconditionis alsosufficient. Indeed, we have

Theorem 1. Let $M$ be a Sperner matroid

over a

triangulation $K$

of

a $d$-simplex $a_{0}a_{1}\ldots$

$a_{d}$ such

that $\{a_{0}, a_{1}, \ldots, a_{d}\}$

forms

a basis. Put $F_{j}\equiv d_{M}(\{a_{0}, a_{1}, \ldots, aj\})(j=0,1, \ldots, d)$, and let $\varphi$ : $V(K)arrow$

$\{0,1, \ldots, d\}$ be

defined

by

$\varphi(v)=0$

if

$v\in F_{0}$, $\varphi(v)=j$$if$ $v\in F_{j}\backslash F_{j-1}$ $(j=1,2, \ldots, d)$

.

Then$\varphi$ is aSpernermap

if

and only

if

the $(d-1)$

-face

$a_{1}a_{2}\ldots$$a_{d}$ contains no loops

of

$M$

.

Theproof ofTheorem 1isbased onthe following :

Lemma. Let $B$ be a basis

of

amatroidM. Suppose

(a) $S\subset T\subset B$,

(b) $X\subset B$, $X\cap T=\emptyset$,

(c) $y\in cl_{M}(T)\backslash cl_{M}(S)$

.

Then$y\in cl_{M}(T \cup X)\backslash d_{M}(S\cup X)$

.

2. Asign function

Let $K$ be atriangulation of

a

$d$-simplex $a\circ a_{1}\ldots$$a_{d}$ in aEuclidean space, $M$ aSperner matroid

over

$K$,

$B\equiv$ $(\mathrm{a}\mathrm{o}, a_{1}, \ldots, a_{d})$

an

ordered basis of$M$

.

Let$\Lambda_{B}$ : $Karrow\{-1,0, 1\}$

.

We define Ab(voVi

$\ldots$ $v_{d}$) $=1(\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{p}.-1)$

if$v_{0}\in F_{0}$ and $v_{j}\in F_{j}\backslash F_{j-1}$ $(j=1,2, \ldots, d)$, where $F_{j}\equiv \mathrm{c}1\mathrm{M}$($\{\mathrm{v}\mathrm{o},$$v_{1}$,

$\ldots$,Vj}) $(j=0,1, \ldots, d)$, and $\mathrm{f}\mathrm{a}\mathrm{i}\mathrm{j})>0$ (resp. $\det(\alpha_{ij})<0$), where

$\{$ $.\cdot.v_{d}v_{0}/\backslash =\{$ $\alpha_{00}$ $\alpha_{0d}$ . $\cdot$ . . $\cdot$ . $\alpha_{d0}$ $\alpha_{dd}$ $\{$ $a_{0}$

.

$\cdot$ . $a_{d}$

$\sum_{j=0}^{d}\alpha_{ij}=1$ $(0\leq i\leq d)$

.

$(*)$

We define$\Lambda_{B}(v_{0}v_{1}\ldots v_{d})=0$otherwise. Now, let$\varphi$ : $V(K)arrow\{0, 1, \ldots, d\}$

.

We calla$d$-simplex$v_{0}v_{1}\ldots$$v_{d}\in K$

positively (resp. negatively) completely labelled if $\varphi(vj)=j$ $(j=0,1, \ldots, d)$, and $\det(\alpha_{ij})>0$ (resp.

(3)

$\det(\alpha_{ij})<0)$, where the matrix $(\alpha_{ij})$ is given in $(*)$. A $d$-simplex of$K$ is completely labelled if it is positively

or negatively completely labelled. It is obvious that $v_{0}v_{1}\ldots$$v_{d}\in K$ is completely labelled if and only if

$\{\varphi(v_{0}), \varphi(v_{1}), \ldots, \varphi(v_{d})\}=\{0,1, \ldots, d\}$. The celebrated Sperner lemma[7] asserts thet if$\varphi$ is aSperner map

then $\#$

{

$\sigma\in K$; ais completelylabelled

}

$\equiv 1$ (mod 2). Theoriented Sperner lemma[l] states that

$\#$

{

$\sigma\in K$; $\sigma$ is positively completely labelled

}

$-\#$

{

$\sigma\in K$; $\sigma$ is negatively completely labelled

}

$=1$

.

By Thoerem 1and the oriented Sperner lemma, we have

Theorem 2. Let $K$ be a triangulation

of

a$d$-simplex$a_{0}a_{1}\ldots$$a_{d}$ in$a$ Euclidean space, and$M$ a Sperner

matroid

over

$K$ such that the $(d-1)$

-face

$a_{1}a_{2}\ldots$$a_{d}$ contains no loops

of

M.

If

$B\equiv(a_{0}, a_{1}, \ldots, ad)$ is

an

ordered basis

of

$M$, then

$\sum_{\sigma\in K}\Lambda_{B}(\sigma)=1$

.

Theorem 2wasrecentlyobtained by theauthorsin [4] with acompletely different proof. Anexample given

in [4] shows that the condition “the$(d-1)$-face$a_{\mathrm{i}}a_{2}\ldots$$a_{d}$contains noloops”cannotbedispensedwith.

By disregarding orientation inTheorem 2, we have

Theorem 3. Under the assumptions

of

Theorem 2,

$\sum|\Lambda_{B}(\sigma)|\equiv 1$ (mod 2).

$\sigma\in K$

Theorem 3is an extension of Lovdsz’s theorem. It complements Lovdsz’s theorem in two aspects :one

concerns an arbitrary matroid while the other is the assertion of oddness. Indeed,

Lov\’asz[2]

proved the

following :

Theorem 4(L. Lovdsz). Let$K$ be atriangulation

of

a$d$-simplex$a_{0}a_{1}\ldots$$a_{d}$ in $a$Euclidean space, and$M$

aSperner matroid

over

$K$ such that $M$ containsno loops.

If

$\{\mathrm{a}\mathrm{o}, a_{1}, \ldots, a_{d}\}$ is a basis

of

$M$ , then$K$ has $a$

$d$-simplex$v_{0}v_{1}\ldots$

$v_{d}$ such that $\{v_{0}, v_{1}, \ldots, v_{d}\}$ is also a basis

of

$M$

.

Finally, let

us

mentionthat it is perhaps worth developing ageneral matroid version which contains

mul-tiple balanced Sperner lemma[6], combinatorial Lefschetz fixed-point formula[5], and multiple combinatorial Stokes’ theorem [3]

(4)

References

1. A. B. Brown and S. S. Cairns, Strengthening of Sperner’s lemma applied to homology, Proc. Nat.

Acad. Sci. USA, 47(1961), 113-114.

2. L. Lovasz, Matroidsand Sperner’slemma, Europ. J. Combinatorics, 1(1980), 65-66.

3. S. N. Leeand M. H. Shih, Acounting lemma and multiple combinatorial Stokes’ theorem, Europ. J.

Combinatorics, 19(1998), 969-979.

4. S. N. Lee and M. H. Shih, Sperner matroid, Arch, der Math, toappear.

5. M. H. Shih andS. N. Lee,AcombinatorialLefschetz fixed-point formula, J. Combinatorial Theory,Ser.

A, 61(1992), 123-129.

6. M. H. Shih and S. N. Lee, Combinatorial formulae for multiple set-valued labellings, Math. Ann.

296(1993), 35-61.

7. E. Sperner, Neuer Beweis fiir dieInvarianz der Dimensionszahl und des Gebietes, Abh. Math. Sem.

Univ. Hamburg, 6(1928), 265-272

参照

関連したドキュメント

As a multidisciplinary field, financial engineering is becom- ing increasingly important in today’s economic and financial world, especially in areas such as portfolio management,

The set of families K that we shall consider includes the family of real or imaginary quadratic fields, that of real biquadratic fields, the full cyclotomic fields, their maximal

In [9], it was shown that under diffusive scaling, the random set of coalescing random walk paths with one walker starting from every point on the space-time lattice Z × Z converges

We present an introduction to the geometry of higher-order vector and covector bundles (including higher-order generalizations of the Finsler geometry and Kaluza-Klein gravity)

We present an introduction to the geometry of higher-order vector and covector bundles (including higher-order generalizations of the Finsler geometry and Kaluza-Klein gravity)

Li, “Multiple solutions and sign-changing solutions of a class of nonlinear elliptic equations with Neumann boundary condition,” Journal of Mathematical Analysis and Applications,

Matroid intersection theorem (Edmonds) Discrete separation (Frank). Fenchel-type

Nakanishi, “Exact WKB analysis and cluster algebras II: simple poles, orbifold points, and generalized cluster algebras”, arXiv:1401.7094.. 13