I Geometrie affine
35 pages

Découvre YouScribe en t'inscrivant gratuitement

Je m'inscris

I Geometrie affine

Découvre YouScribe en t'inscrivant gratuitement

Je m'inscris
Obtenez un accès à la bibliothèque pour le consulter en ligne
En savoir plus
35 pages
Obtenez un accès à la bibliothèque pour le consulter en ligne
En savoir plus


I. Geometrie affine 1. Definitions Espace affine = ? ? ? ensemble de points E espace vectoriel (reel) !E + application ? : { E? E ? !E (A, B) _? ??AB verifiant la relation de Chasles, plus M _? ??AM bijective ?A ? E . Consequences: ??AA = 0 , ??BA = ???AB . 2. Sous-espaces affines Definition. Propriete fondamentale: F sous-espace affine de direction !F ? !E , A ? F ? M ? F ?? ??AM ? !F Un sous-espace affine est uniquement determine par un de ses points et sa direction. Intersection de sous-espaces affines. Sous-espaces affines paralleles. 3. Repere affine, coordonnees cartesiennes Proposition : dim E = n , A0, . . . , An ? E . Proprietes equivalentes: a) A0, . . . , An ne sont pas contenus dans un sous-espace affine *= E ; b) ????A0A1, . . . , ????A0An base de !E . On dit alors que (A0, . . . , An) est un repere affine. Coordonnees d'un point M (dans ce repere) = coordonnees de ???A0M . Exemples de calculs: equations d'une droite, d'un plan. 2eme semaine 4.

  • sin? sin?

  • ap ?

  • base orthonormale convenable

  • ±1 ±1

  • f? f?

  • geometrie euclidienne

  • unique cercle



Publié par
Nombre de lectures 57
Langue English


Geography of local configurations
D. Coupier July 12, 2007
E-Mail address :david.coupier@math.univ-lille1.fr
Mail address :PeuaPliaobarotriniversitnlev´e,UiLe´1ellaL Cit´escientique,59655VilleneuvedAscqCedex,France. Telephone :33 (0)3 20 43 67 60 Fax :33 (0)3 20 43 43 02
Abstract Ad-dimensional ferromagnetic Ising model on a lattice torus is considered. As the sizenof the lattice tends to infinity, the magnetic fielda=a(n) and the pair potential depend onn bounds for the probability for local configurations to. Precise occur in a large ball are given. Under some conditions bearing on potentialsa(n) andb(n), the distance between copies of different local configurations is estimated according to their weights. Finally, a sufficient condition ensuring that a given local configuration occurs everywhere in the lattice is suggested.
Key words :Ising model, ferromagnetic interaction, FKG inequality. AMS Subject Classification :60F05, 82B20.
1 Introduction Inthetheoryofrandomgraphs,inauguratedbyErd¨osandRe´nyi[12],theappearanceof a given subgraph has been widely studied (see [4] or [22] for a general reference). In the random graph formed bynvertices, in which the edges are chosen independently with probability 0< p <1, a subgraph may occur or not according to the value ofp=p(n). In addition, under a certain condition on the probabilityp(n), its number of occurrences in the graph is asymptotically (i.e. asn+ the edges with Replacing) Poissonian. the spins of an Ising model, the notion of subgraph corresponds to the notion of what we will calllocal configuration tools coming from random; Figure 1 shows an example. Using graphs (asthreshold functionsandPoisson approximations), the study of the appearance of a given local configuration has been done in [11], [10] and [8]. In this article, this study is extended into three directions. First, the speed at which local configurations occur is precised. Moreover, when the number of copies in the graph of a given local configuration is finite, the geography of positive and negative spins surrounding one of them is described. Finally, a sufficient condition ensuring that a given local configuration is present everywhere in the graph is stated. The results obtained in these three directions are based on the same tools; the Markovian character of the measure, the control of the conditional probability for a local configuration to occur in the graph and the FKG inequality [15]. Let us consider a lattice graph in dimensiond1, with periodic boundary conditions (lattice torus). The vertex set isVn={0, . . . , n1}d integer. Thenwill be called the size The edge set, denoted byof the lattice.Enbe specified by defining the set of, will neighborsV(x) of a given vertexx: V(x) ={y6=xVn,kyxkqρ},(1) where the substraction is taken componentwise modulon,k ∙ kqstands for theLqnorm in Rd(1q≤ ∞), andρis a fixed integer. instance, the square lattice is obtained for For q=ρ= 1. Replacing theL1norm with theL From now on,norm adds the diagonals. all operations on vertices will be understood modulon particular, each vertex of the. In lattice has the same number of neighbors; we denote byVthis number. Aconfigurationis a mapping from the vertex setVnto the state space{−1,+1}. Their set is denoted byXn={−1,+1}Vnand called theconfiguration set. The Ising model is classically defined as follows (see e.g. Georgii [17] and Malyshev and Minlos [21]). Definition 1.1LetGn= (Vn, En)be the undirected graph structure with finite vertex set Vnand edge setEn. Letaandbbe two reals. TheIsing modelwith parametersaandbis the probability measureµa,bonXn={−1,+1}Vndefined by: for allσ∈ Xn, µa,b(σ) =Z1a,bexpaxXVσ(x) +bXσ(x)σ(y),(2) n{x,y}∈En where the normalizing constantZa,bis such thatPσ∈Xnµa,b(σ) 1. = 2
Following the definition of [21] p. 2, the measureµa,bdefined above is a Gibbs measure associated to potentialsaandb. Expectations relative toµa,bwill be denoted byIEa,b. In the classical presentation of statistical physics, the elements ofXnare spin configurations; each vertex ofVn Here, we shall simplyis an atom whose spin is either positive or negative. talk about positive or negative vertices instead of positive or negative spins and we shall merely denote by + andthe states +1 and parameters1. Theaandbare respectively themagnetic fieldand thepair potential. The model remaining unchanged by swapping positive and negative vertices and replacingabya, we chose to study only negative values of the magnetic fieldain order to use the FKG inequality,. Throughout the paper, the pair potentialbwill be supposed nonnegative. As the sizenof the lattice tends to infinity, the potentialsa=a(n) andb=b(n) depend onn. The case wherea(n) tends to−∞corresponds to rare positive vertices among a majority of negative ones. In order to simplify formulas, the Gibbs measureµa(n),b(n)will be merely denoted byµa,b. We are interested in the appearence in the graphGnof families of local configurations. See Section 2 for a precise definition and Figure 1 for an example. Such configurations are called “local” in the sense that the vertex set on which they are defined is fixed and does not depend onn. A local configurationηis determined by its set of positive verticesV+(η) whose cardinality and perimeter are respectively denoted byk(η) andγ(η natural idea). A (coming from [8]) consists in regarding both parametersk(η) andγ(η) through the same quantity; theweightof the local configurationη Wn(η) = exp (2a(n)k(η)2b(n)γ(η)). This notion plays a central role in our study. Indeed, the weightWn(η) represents the probabilistic cost associated to a given occurrence ofη. − − − −+ ++ + − −+ ++− − −+ +− −+
Figure 1: A local configurationηwithk(η) =|V+(η)|= 10 positive vertices and a perimeter γ(η) equals to 58, in dimensiond= 2 and on a ball of radiusr= 2 (withρ= 1 and relative to theLnorm). Proving some sharp inequalities is generally more difficult than stating only limits. In the case of random graphs, Janson et al. [20], thus Janson [19], have obtained exponential bounds for the probability of nonexistence of subgraphs. Some other useful inequalities have been suggested by Boppona and Spencer [5]. In bond percolation onZd, it is believed that, in the subcritical phase, the probability for the radius of an open cluster of being
  • Univers Univers
  • Ebooks Ebooks
  • Livres audio Livres audio
  • Presse Presse
  • Podcasts Podcasts
  • BD BD
  • Documents Documents