Lundi 1er Octobre
Seminaire de la fédération (pas de séminaire LPT/LPTMS)
Jeudi 11 Octobre à 14h au LPTMS
Lenka Zdeborova (LPTMS)
Phase transitions in the coloring of random graphs
Average-case analysis of computational complexity presents strong
affinities with the study of disordered systems in statistical
mechanics. We consider the problem of coloring a large random graph
with a given number of colors such that no adjacent vertexes have the
same color. Using the cavity method, we present a detailed and
systematic study of the phase space of the solutions for different
connectivities and numbers of colors. We show that, for fixed number
of colors and as the connectivity increases, the set of solutions
undergoes different phase transitions similar to those happening in
the mean field theory of glasses. First, the phase space decomposes
into an exponential number of pure states, and subsequently
condensates over the largest such states. Another transition takes
place when a finite fraction of frozen variables appears inside almost
all the dominant pure states. Eventually a connectivity is reached
beyond which no solutions are available. Finally, we discuss the
algorithmic perspectives of our results.
Vendredi 19 Octobre à 14h au LPT
ANNULE POUR CAUSE DE GREVE DES TRANSPORTS !!!
Jeudi 25 Octobre à 11h au LPTMS
Cristina Toninelli (LPMA, Univ.Paris VI-VII), Marco Tarzia (LPTMS, Orsay)
Group Testing with random pools : optimal algorithms and phase transitions
The problem of Group Testing (GT) is to identify defective items
in a set of objects by testing groups of items via the query "Is there at
least one defective in the pool ?". After recalling the numerous
applications of GT, we discuss the construction of efficient algorithms
both when all tests should be performed in parallel (one-stage) and when
subsequent stages of parallel tests are allowed. Focusing on the regime of
small defective probability and large number of items, we identify optimal
two-stage algorithms based on proper random choices of the pools. On the
other hand, for one-stage algorithms we identify a non-detection/detection
phase transition. Then, we explain the connection of GT with the Hitting
Set (HS) problem, a generalization of Vertex Covering. By using the cavity
method we show that HS on random regular hyper-graphs can be either in a
replica symmetric or in a 1RSB phase. Finally, we show that the cavity
methods provides very efficient decimation based algorithms able to solve
large individual instances of the HS.
Jeudi 8 Novembre
Seminaire de la fédération (pas de séminaire LPT/LPTMS)
Jeudi 15 Novembre à 15h au LPT
Pascal Viot (LPTMC - Univ. Pierre et Marie Curie)
Gaz granulaires : de la theorie cinetique a l’analyse d’experiences
La presence de dissipation dans les interactions (collisions) entre
particules granulaires induit des proprietes tres differentes de celles
observees dans les systemes microscopiques habituels.
A faible densitee et a faible dissipation, les theories cinetiques de
type Boltzmann permettent de mettre en evidence les differentes deviations
des distributions de vitesse au comportement Gaussien ainsi que l’absence
d’equipartition de l’energie. Apres une revue de ces proprietes, nous
etudierons quelques proprietes additionnelles quand les particules sont
non-spheriques.
Dans une seconde partie, nous nous interessons a des recents dispositifs
experimentaux qui permettent de simuler des thermostats "ideaux". En
particulier, par une etude de simulation de dynamique moleculaire sur une
bicouche granulaire, nous obtenons une statistique gaussienne de la
distribution des vitesses sur plusieurs decades conformement aux resultats
experimentaux de Baxter et Olfasen pour des densitees allant au dela de la
limite des gaz granulaires et pour une vaste gamme de parametres
microscopiques ( rapport de masse des particules, coefficient de
restitution,...).
Jeudi 22 Novembre à 14h au LPTMS
Ginestra Bianconi (ICTP Trieste)
Entropy of randomized network ensembles
Randomized network ensembles are the null models of real
networks and are extensively used to compare a real system to a null
hypothesis.
In this talk we study network ensembles with the same degree
distribution, the same degree-correlations and the same community
structure of any given real network. We characterize these randomized
network ensembles by their
entropy, i.e. the normalized logarithm of the total number of networks
which are part of these ensembles.
We estimate the entropy of randomized
ensembles starting from a large set of real directed and undirected
networks. We propose entropy as an indicator to assess
the role of each structural feature in a given real network.We observe
that the ensembles with fixed scale-free degree distribution
have smaller entropy than the ensembles with homogeneous degree
distribution indicating a higher level of order in scale-free networks.
Jeudi 29 Novembre à 14h au LPT
Werner Krauth (LPS - ENS)
Time correlations and "exact" sampling
In statistical physics, dynamical-systems theory and hydrodynamics, but also in
Monte Carlo simulations, time correlations describe the approach to thermal
equilibrium. In the long-time limit of diffusive systems, time correlations are
characterized through a single number, the correlation time. In this talk, we
review the crucial concept of time correlations in physics and computation, and
the tantalizing difficulty of estimating the correlation time in Monte Carlo simulations.
We then discuss an exciting reformulation of Monte Carlo algorithms, due to
Propp and Wilson, called exact sampling, where one can
prove rigorously that the Monte Carlo simulation has run sufficiently
long to generate a configuration which is completely decorrelated from the
initial condition. We will describe a general renormalization-group algorithm
which implements exact sampling for two-dimensional Ising spin glasses on large
lattices and report on the applications in three dimensions.
(C. Chanal and W. Krauth "Renormalization group approach to exact sampling"
arXiv:0707.4117 (2007))
Lundi 3 Décembre
Seminaire de la fédération
Jeudi 6 Décembre à 14h au LPTMS
Francesco Zamponi (LPT-ENS)
A first-principle theory of amorphous states of hard spheres
Hard spheres are ubiquitous in condensed matter : they have been used as
models for liquids, crystals, colloidal systems, granulars, and powders.
Packings of hard spheres are of even wider interest, as they are related
to important problems in information theory, such as digitalisation of
signals, error correcting codes, and optimization problems. In three
dimensions the densest packing of identical hard spheres has been proven
to be the FCC lattice, and it is conjectured that the closest packing is
ordered in low enough dimension. Still, amorphous packings have attracted
a lot of interest as theoretical models for glasses, because for
polydisperse colloids and granular materials the crystalline state is not
obtained in experiments for kinetic reasons. I will discuss a theory of
amorphous packings, and more generally glassy states, of hard spheres that
is based on the replica method and gives predictions on the structure and
thermodynamics of these states. In dimensions between two and six these
predictions have been successfully compared with numerical simulations. I
will also discuss the limit of large dimension where an exact solution
seem to be possible. Finally, I will discuss some possible extensions of
the theory to different systems.
Jeudi 13 Décembre à 14h au LPT
Gregory Schehr (LPT - Orsay)
Statistics of the number of zero crossings : from real roots of
Kac’s polynomials to persistence probability for the diffusion equation
We consider a class of real random polynomials, indexed by an integer d,
of large degree n and focus on the number of real roots of such random polynomials.
The probability that such polynomials have no real root in the interval
[0,1] decays as a power law n^-\theta(d) where \theta(d)>0. We show that
\theta(d) is precisely the exponent associated to
the decay of the persistence probability for the diffusion equation with
random initial conditions in space dimension d. For n
even, the probability that such polynomials have no root on the full real
axis decays as n^-2(\theta(d) + \theta(2)). We further show that the
probability that such polynomials have exactly k real roots in [0,1] has
an unusual scaling form given by n^-\tilde \varphi(k/\log n) where
\tilde \varphi(x) is a universal large deviation function.
(G. Schehr, S.N. Majumdar, Phys. Rev. Lett. 99, 060603 (2007),
arXiv:0705.2648).
Jeudi 20 Décembre à 14h au LPTMS
Guilhem Semerjian (LPT - ENS)
On message passing guided algorithms for solving constraint satisfaction
problems
The statistical mechanics studies of computer science optimization
problems have had two main outcomes. On the one hand they proposed a rich
picture of the phase diagram of random instances, and on the other they
led to new efficient heuristics (Survey Propagation) for practically
solving given instances. The theoretical understanding of these new
algorithms is not yet completely satisfactory.
In this seminar I will review general ideas about message passing guided
algorithms and present recent results on the analysis of a representant of
this class of procedures (Belief Propagation).
Joint work with Andrea Montanari and Federico Ricci-Tersenghi.
Jeudi 10 Janvier à 14h au LPT
Gerald Kneller (Centre de Biophysique Moléculaire - Orleans)
Dynamique fonctionnelle des protéines et dynamique vitreuse - études par simulation moléculaire
Différentes techniques expérimentales ont montré que la dynamique de relaxation des protéines est caractérisée par un spectre de temps de relaxation vaste, allant de la picoseconde à la seconde, voire à l’heure. La dynamique fonctionnelle qui peut être observée par spectroscopie de corrélation de fluorescence et par des études cinétiques couvre la partie « lente » du spectre, et le processus stochastique sous-jacent montre le comportement caractéristique de la dynamique brownienne fractionnaire (DBF). Par études de simulation moléculaire nous avons récemment mis en évidence que de tels processus de diffusion non-markoviens et auto-similaires dans le temps dominent également la dynamique interne des protéines à des échelles de temps beaucoup plus courtes, de l’ordre de 0.1 ps à 1 ns. Dans cet exposé seront présentés des éléments qui indiquent « l’universalité » du modèle DBF pour la dynamique de relaxation des systèmes complexes en général et des applications à la modélisation de données expérimentales.
Jeudi 17 Janvier à 14h au LPT
Sabine Klapp
(Stranski-Laboratorium for Physical und Theoretical Chemistry
and
Institute of Theoretical Physics
Technische Universitaet Berlin)
Dipolar ordering between two and three dimensions
We present computer simulation results for dipolar fluid films under strongly
coupled conditions where the bulk fluid and films of mesoscopic thicknesses
display ferromagnetic ordering. We demonstrate
that the ordering persists down to nanoscopic wall separations where the
system consists of only three monolayers. For smaller
thicknesses we observe stripe-like defects (domains) and finally the breakdown
of ferromagnetic ordering for systems close to the two-dimensional limit.
We also discuss the influence of wall boundary conditions
and additional external magnetic fields. Our results are relevant
for systems of magnetic colloids but also for the ordering behavior of thin
solid-like magnetic films.
Jeudi 24 Janvier à 14h au LPTMS
Raoul Santachiara
Interfaces critiques dans des modèles de spins : évolutions
stochastiques de Loewner-Schramm et théories conformes.
L’ évolution de Loewner-Schramm (SLE) est un processus
stochastique de croissance décrivant des courbes invariantes
conformes qui caractérisent les systèmes critiques en dimension d=2.
Cette approche représente un outil important pour la description
géométrique des phases critiques. Nous présenterons les concepts
principaux sous-jacents a cette approche et le lien entre les SLEs et
les théories conformes . Nous décrirons les interfaces SLE dans une
famille de modèles de spins qui comprend le modèle d’Ising et le modèle
de Potts a trois états.
Attention, jour, heure et salle inhabituels
Mercredi 30 Janvier au LPTMS a 11h en salle des conseils de l’IPN
Thierry Giamarchi
Creep and Depinning of domain walls
Many seemingly different macroscopic systems (magnets, ferroelectrics,
spintronic materials) have their properties controlled by the propagation
of domain walls.
For the domain walls, the competition between elasticity and disorder
leads to glassy effects quite analogous to more usual glasses, but
with also important differences. Understanding both their
static and dynamics thus poses challenging problems both from the
point of view of fundamental physics but also in view of practical
applications.
Despite important progress many questions remain. I will give some
examples and will discuss in details the question of the depinning
of the walls when subjected to an external force.
One important and still unresolved question is the effect of the
temperature on this depinning transition.
Jeudi 7 Fevrier à 14h au LPTMS
Alexander Hartmann (Computational Theoretical Physics, Institut für Physik, Universität Oldenburg, Germany)
Solution space structure of optimization problems
We study numerically the solution-space clustering properties of
random ensembles of three glassy optimization problems originating in
computational complexity. Namely we investigate the vertex-cover
problem, the number-partitioning problem and the satisfiability
problem. We use branch-and-bound type algorithms to obtain exact
solutions of these problems for moderate system sizes. For the
satisfiability problem, we also use heuristics, which allow to obtain
solutions in the "solvable part" of phase space for large systems.
Using two methods, direct neighborhood-based clustering and
hierarchical clustering, we investigate the structure of the solution
space. The main result is that the correspondence between
solution-space structure and the phase diagrams of the problems is not
unique. Namely, for vertex cover we observe a drastic change of the
solution space from large single clusters (in the replica-symmetric
phase, as obtained by previous analytical studies) to multiple
nested levels of clusters (in the replica-symmetry broken phase). In
contrast, for the number-partitioning problem, the phase space looks
always very simple, similar to a random distribution of the
lowest-energy configurations. This holds in the ``easy’’/solvable
phase as well as in the ``hard’’/unsolvable phase. Finally, for the
satisfiability problem, we observed several changes of the solution
landscape, regarding the number of clusters, their relative sizes and
regarding the fraction of frozen variables.
Jeudi 14 Fevrier à 14h au LPTMS
Koji Hukushima (Univ. Tokyo)
Monte Carlo study of some Potts (glass) models
Potts model is a generalization of the Ising model with two states to an
arbitrary number $q$ of the discrete states. Since a mean-field theory
for a Potts spin glass model has revealed the existence of a peculiar
1-step replica-symmetry-breaking (RSB) transition and a formal
similarity to structural glass, much interest has shifted toward Potts
glass models on finite-dimensional lattices and/or on sparse random
graphs. The latter at zero temperature corresponds the coloring
problem. We studied statistical-mechanical properties of some Potts
(glass) models by using finite-temperature Monte Carlo simulations.
One of the models we studied is the three-state +/-J Potts glass
model in three dimensions. We obtained a phase diagram of the model as
functions of temperature and the fraction $p$ of the antiferromagnetic
bonds. We also show some results for the coloring problem and a
mathematical puzzle, SUDOKU.
Séminaire exceptionnel organisé par le LPT
Mardi 19 Fevrier à 11h au LPT
Constantino Tsallis (CBPF, Bresil)
Mecanique statistique non extensive et generalisation du theoreme de la limite centrale
Il est bien connu que l’ entropie de Boltzmann-Gibbs-von Neumann-Shannon (BG) et la mecanique statistique
associee ont un domaine de validite extremement large. Mais il n’ est pas universel. Par exemple, beaucoup
de systemes interessants violent l’ hypothese d’ ergodicite dans des etats stationnaires ou
quasi-stationnaires. Dans le but d’ elargir quelque peu l’ applicabilite des methodes de la mecanique
statistique, une entropie plus generale, Sq, a ete propose en 1988. Pour q=1, elle reproduit l’ expression
de BG. L’ entropie de BG est additive (donc
extensive pour des systemes sans correlations, ou avec des correlations
generiques de courte portee). L’ entropie Sq est, par contre, nonadditive pour q different de l’ unite.
Neanmoins, pour une valeur speciale de q (qui depend de quelques characteristiques assez universelles du
systeme), elle peut etre extensive, donc epousant adequatement la thermodynamique classique dans la limite
de systemes tres larges. Cette situation sera illustre avec des examples physiques, incluant des systemes
fortement enchevetres quantiquement. Les fortes correlations qui demandent l’ utilisation de Sq sont
generiquement invariantes par echelle dans le regime asymptotique, et donnent lieu a des q-generalisations
des theoremes centraux de la limite (standard et celle de Levy-Gnedenko). Des applications (e.g., pour des
atomes froids dans des reseaux optiques dissipatifs, et pour des systemes Hamiltonians avec des
interactions a longue portee) seront brievement presentees.
Jeudi 13 Mars à 14h au LPT
Henri Beresticky (CAMS)
Une espece biologique peut-elle suivre un changement climatique ?
Jeudi 20 Mars à 14h au LPTMS
Julien Tailleur
(Department of Physics & Astronomy,
University of Edinburgh)
Grandes déviations hors de l’équilibre : retour à l’équilibre via un changement de variables
La théorie des grandes déviations a pour objet l’étude asymptotique d’évènements rares. Si elle
permet une reformulation naturelle de la physique statistique d’équilibre, elle prend tout son
intérêt hors de l’équilibre, où les fonctions de grandes déviations permettent une extension du
concept de potentiels thermodynamiques. Une méthode permettant le calcul de ces fonctions est donc
un des buts majeurs de la physique statistique hors de l’équilibre. Je présenterai un formalisme
analogue à celui introduit récemment par Bertini et collaborateurs (PRL 87, 040601, (2001)) visant
à remplir ce rôle.
En me basant sur un exemple traité par cette théorie, un processus d’exclusion simple dans la
limite hydrodynamique, plongé hors de l’équilibre via un contact avec deux réservoirs de densités
différentes, je montrerai qu’une transformation des densités et des courants ramènent ce système à
l’équilibre. Ce retour à l’équilibre explique la possibilité de calculer explicitement la fonction
de grandes déviations grâce à des changements de variables élémentaires. Ce traitement s’applique
également aux autres cas résolus par la ’Théorie des fluctuations macroscopiques’.
Jeudi 27 Mars à 14h au LPT
Stephane Zaleski
(Institut d’Alembert)
La fragmentation des masses liquides dans les jets d’atomiseurs
jeudi 3 Avril à 14h au LPTMS
Sergei Nechaev
(LPTMS)
On spectral density of some random hierarchical matrices
We propose the new ensemble of of hierarchical random matrices with
power-law tails
in the spectral density. The construction is based on randomization of
the Parisi-type
hierarchical block matrix $T$. We show that if the distribution on
matrix elements of $T$
is Gaussian with zero mean and exponentially decreasing variance,
$\sigma^2_\gamma=2^-\beta \gamma$ ($\gamma$ is the hierarchical
level of $T$),
then the tail of the spectral density, $\rho(\lambda)$, of Gaussian
ensemble of randomized
Parisi matrices, is $\rho(\lambda) \sim
|\lambda|^(2-\beta)/(1-\beta)$ for $\lambda\to\infty$
and $0<\beta<1$ (in the limit $N\to \infty$). For $\beta>1$ one has
termination of the the
power-law behavior. We conjecture that the similar asymptotic behavior
of $\rho(\lambda)$
remains for hierarchical random graphs with Bernoulli-type
distribution on matrix coefficients in
the $N\to\infty$ limit.
Jeudi 10 Avril à 14h au LPTMS
á preciser
Attention, heure exceptionnelle
Jeudi 17 Avril à 11h au LPT
Jorge Kurchan
(ESPCI)
Jamming versus Glass transition
Jamming and glass transitions are two phenomena whose mutual relation is not quite
clear. The `jamming transiton’ happens when a system of, say,
hard frictionles spheres is rapidly compressed and reaches an infinite pressure point.
This point is characterised by a diverging length and
soft modes, just as a critical phenomenon.
At the bottom of this criticallity is the fact that when the system jams
it does so under globally isostatic mechanical equilibrium.
On the other hand, the glass transition is also characterised by diverging dynamic (and
presumably also static) lengths. It may happen at finite temperatures and pressures, and
for soft potentials.
The question as to what, if any, is the relation between the two transitions has been
difficult to answer due to a lack of situations where both sets of theories apply
simultaneously. I shall discuss
a family of models that does precisely that : it posseses
on the one hand a jamming transition with diveging lengths and soft modes, and on the
other an ideal glass transition.
Jeudi 24 Avril au LPTMS
Dr. Bartlomiej Waclaw
(Institut fuer Theoretische Physik, Leipzig, Germany)
Finite-size effects in growing networks and counting of
metastable states in Ising spin glasses on arbitrary graphs
Many real-world networks like e.g. citation networks can be approximated
by a model where new nodes and links are consecutively added to the
network without rearranging old connections. The best known example is the
Barabasi-Albert model and its generalizations. Such networks have a
power-law distribution of degrees, that is the number of nearest
neighbors. However, for any finite network the power law cannot extend to
infinity and must have a cut-off. Since the cut-off has a direct impact on
many processes on the network like disease spreading or percolation, it is
important to know how it scales with the network size. I will
demonstrate a method which allows one to calculate these finite-size
corrections and show that the scaling exponent depends only on the
exponent in the power law.
The second part of the talk will be devoted to the problem of counting
one-spin stable states in Ising spin glasses, where couplings between
spins are taken at random from some prescribed distribution. One-spin
stable states are local energy minima such that a flip of any single spin
increases the energy. Their number N_s is known to grow exponentially with
the system size, but the rate of growth depends strongly on both the
distribution of coupling constants and the graph. Here I will assume that
couplings have a gaussian distribution with zero mean and fixed variance.
I will develop a systematic method which allows to
compute N_s for arbitrary graphs and demonstrate it on several examples,
including 2D and 3D regular lattice and their randomized versions. The
results indicate that the rate of growth is mainly determined by local
properties of the graph, on which the spin glass is defined.
Jeudi 1er et 8 Mai
Férié (pas de séminaire)
Attention, jour et heure exceptionnels !!!
Mardi 13 Mai à 11h au LPTMS
Alvaro Dominguez
(Física Teórica, Universidad de Sevilla)
Two-dimensional structures of colloids at fluid interfaces
I will briefly review the formation of structures in colloids composed
by micrometer—sized particles trapped at a fluid interface. Emphasis
will be placed on the investigation of the effective interaction
between the particles.
Recent experiments have estimulated the interest in the theoretical
research of the effective interactions which can be caused by electric
and magnetic forces, by elastic stresses (e.g., if a fluid is in a
nematic phase) and by capillary forces. I will discuss the current
status of the research and its relevance for the interpretation of the
experiments, which is actually a source of polemic.
Jeudi 15 Mai à 14h au LPT
Nicolas Brunel
(Laboratory of Neurophysics and Physiology,
Universite Paris Descartes)
Connectivite synaptique optimisant le stockage en memoire
Une des hypotheses majeures en neuroscience est que l’apprentissage et
la memoire sont dus a des modifications synaptiques dans les reseaux
de neurones du cerveau. Cette hypothese a conduit a un enorme
effort experimental pour determiner les conditions dans lesquelles les
synapses peuvent etre modifiees par l’activite des neurones pre- et
post-synaptiques, et en parallele a un effort theorique important pour
comprendre comment des `regles d’apprentissage’ deduites de ces
donnees experimentales permettent le stockage en memoire dans des
reseaux de neurones. Malgre ces efforts combines, il n’existe
cependant a l’heure actuelle que peu de comparaisons detaillees entre
theorie et experience.
La connectivite synaptique a ete etudiee en detail depuis le debut des
annees 90 par des enregistrements intracellulaires de neurones
multiples in vitro dans de nombreuses aires du cerveau (en particulier
le neocortex, le cervelet et l’hippocampe). Ces enregistrements
revelent deux caracteristiques de la connectivite qui semblent
universelles : (1) deux neurones `proches’ sont connectes avec une
probabilite de l’ordre de 10%, alors que la geometrie de leurs
axones/dendrites indique que cette probabilite pourrait en principe
etre proche de 1 ; (2) les `efficacites synaptiques’ strictement
positives ont une distribution similaire dans toutes ces regions.
Apres avoir presente ces donnees experimentales, je decrirai des
etudes theoriques recentes utilisant des techniques de physique
statistique qui permettent d’expliquer ces deux observations par un
critere d’optimalite de stockage en memoire en presence de bruit. Ces
etudes theoriques permettent aussi de reproduire la statistique des
`motifs synaptiques’ de paires et triplets de neurones dans le cortex
cerebral.
Jeudi 22 Mai à 14h au LPTMS
A. De Martino
Cellular metabolism as a constrained optimization problem
Understanding the organization of reaction fluxes in
metabolic networks from the underlying stoichiometry is a central
issue in systems biology. Methods based on mass-balance conditions
coupled with local optimization rules are able to reproduce
experimental findings for relatively simple organisms in specific
conditions. Here we define and study a constraint-based model of
metabolic networks where neither mass balance nor flux stationarity
are postulated. The relevant flux configurations optimize the global
growth of the system. Such solutions provide the correct statistics of
fluxes for the bacterium E.coli in different
environments. Comparison with random metabolic networks enables us to
clarify the role of conserved pools of metabolites in determining growth
rate and flux variability in natural networks. Finally, we are able to
connect phenomenological gene essentiality with fluxes with smaller
allowed variability in E.coli metabolism.
!!!ATTENTION heure inhabituelle !!!!!!
Jeudi 5 Juin à 14h au LPTMS
David Declercq
(ETIS/ENSEA/UCP CNRS UMR-8051)
Quelques aspects des codes correcteurs d’erreur modernes
Tout d’abord, je rappellerai les principes fondateurs des codes
correcteurs d’erreurs, ainsi que leur application à la transmission
ou au stockage de l’information. Le cadre mathématique donné par
Claude Shannon en 1948 fixe les limites à la correction d’erreur,
mais sans proposer de fonction de codage atteignant cette limite.
Les codes dits ’modernes’ peuvent s’approcher de cette limite
dans le régime asymptotique, c’est à dire quand la taille des
mots de code tend vers l’infini. Je rappelerai brièvement les
structures des turbo-codes et des codes LDPC, et surtout leur
décodeur par propagation de croyance, dont les équations de mise
à jour sont issues de l’approximation de Bethe de l’énergie.
Ensuite, je parlerai de deux sujets de recherche liées aux codes
LDPC, à savoir :
Dynamique de décodeurs quantifiés et topologies de graphes,
Estimation de la distance minimale des codes par la méthode
impulsionnelle.
!!!ATTENTION : JOUR ET HEURE EXCEPTIONNELS !!!
Lundi 9 Juin à 11h au LPTMS
Idetoshi Nishimori
Lee-Yang zeros and Griffiths singularities in 2d and 3d spin glasses
The distribution of zeros of the partition function is studied
for the two- and three-dimensional Ising spin glasses on the complex
field plane, the Lee-Yang zeros. We estimate the density of zeros
on the imaginary-field axis by an importance-sampling Monte Carlo algorithm,
which enables us to sample extremely rare events. Our results in 2d clearly
suggest that the density has an essential singularity at the origin, giving
the first evidence for Griffiths singularities in spin glasses in equilibrium.
Jeudi 12 Juin à 14h au LPT
Vincent Russier
(Institut de Chimie et des Materiaux de Paris-Est, Vitry-sur-Seine)
Interface entre liquides simples non miscibles :
étude par dynamique moléculaire et fonctionnelle de la densité.
L’interface entre deux liquides partiellement ou non miscibles intervient
dans un grand nombre de situations réelles et/ou d’applications : séparation
liquide/liquide ; membranes biologiques ; électrochimie entre électrolytes liquides.
Il s’agit en général de processus de transfert soit ionique soit d’espèces neutres,
pour lesquels la structure est un paramètre important. Du point de vue fondamental
ce systeme fait appel à un ensemble de problèmes types de la physique des liquides
aux interfaces : mouillage, ondes capillaires, coexistence de phases et présente donc
un grand intérêt en soi.
Dans le contexte d’un modèle simple de liquides (Lennard- Jones) non
miscibles, la structure à l’interface entre les liquides est décrite à partir de deux
approches complémentaires : la simulation numérique par dynamique moléculaire et la
théorie de la fonctionnelle de la densité. On montre que la structure à travers
l’interface est en partie conditionnée par la transition de "drying" que l’on observe
au delà d’une température Tdw le long de la courbe de coexistence liquide/liquide/
vapeur des liquides en présence. Le caractère de premier ordre de cette transition
est mis en évidence à partir du comportement de l’adsorption (nombre d’excès de
molécules en surface) et des tensions superficielles.
!!!ATTENTION JOUR ET HEURE EXCEPTIONNELS !!!
Lundi 16 Juin à 11h au LPT
Henk van Beijeren
(Institute for Theoretical Physics, Utrecht University)
Green-Kubo formalism for solids
Resume :
The Green-Kubo formalism yields macroscopic transport equations on the basis of
the microscopic equations of motion, by postulating the existence of a set of slowly
varying microscopic phase functions, whose dynamics allow for a closed macroscopic description.
These functions always include the long-wavelength Fourier components of energy, momentum and mass
densities. In systems with broken symmetries the order parameters describing these have to be added.
In solids these primarily are displacement fields, describing the displacements
of the atoms from their equilibrium positions.
The resulting macroscopic equations are the elastic equations describing propagation and
damping of sound and Fourier’s law of heat conduction. The Green-Kubo formalism expresses the
transport coefficients and damping constants occurring in these equations in terms of time
integrals of current-current time correlation functions.
I will present the general structure of these equations together with the
Green-Kubo expressions for transport and damping coefficients. If time allows I will consider
mode-coupling predictions for the long time behavior of the correlation functions relevant for heat
diffusion. I will discuss divergences of transport coefficients and their finite-size
renormalization.
!!!ATTENTION : JOUR ET HEURE EXCEPTIONNELS !!!
Mercredi 18 Juin à 17h au LPTMS
Martin WEIGT
(ISI Torino)
Inference of spatial details in protein-protein interactions
from multi-species sequence data
Resume :
Protein-protein interactions are at the basis of most biological
processes, and understanding the molecular basis of such interactions
is one of the outstanding challenges in systems biology. Experimental
approaches to elucidate details of protein-protein interactions
normally involve the generation of co-crystall structures ; the latter
are hard to obtain for transient interactions (interaction energies of
few kT). The rapidly growing number of sequenced genomes (ca. 500
bacterial species) is starting to offer an alternative approach : The
genetic variability of the amino-acid sequence of structurally
equivalent interacting protein pairs inside and across species allows
to identify universal molecular mechanisms. More specifically, a
mutation in the interaction surface of one protein has most likely a
deleterious effect on the interaction, and has to compensated for by
an adequate mutation in the interaction partner. Residue pairs in
contact are thus expected to be correlated between proteins.
In our work, a message passing-approach to infer global statistical
models for the amino-acid sequences of interacting protein pairs (in
statistical-physics language a disordered 21-states Potts model) is
proposed. Based purely on sequence information, this approach is able
to detect strong direct couplings of residues across proteins. For the
case of proteins involved in two-component signal transduction (where
co-crystal structures are known), the approach unveils a network of
coupled residues which, compared to the actual co-crystal structure,
are found to define with impressive accuracy the interaction surface.
Besides for structure prediction, the approach can be used to predict
sequences of unknown interaction partners up to amino-acid resolution.
This work is a collaboration with R.A. White, T. Hwa (UCSD), H.
Szurmant and J.A. Hoch (Scripps).
Jeudi 19 Juin à 14h au LPTMS
Vaclav JANIS
(Institute of Physics of the Academy of Sciences of the
Czech Republic, Praha, Czech Republic)
Replica-symmetry breaking in the mean-field theory of spin glasses : discrete vs. continuous schemes
Resume :
F. Guerra and M. Talagrand proved that the replica-symmetry breaking
construction of Parisi leads to the exact free energy of the
Sherrington-Kirkpatrick model. The proof, however, does not specify
whether hierarchies of replicas should be distributed discretely or
continuously. In our talk we discuss the genesis of both, discrete and
continuous schemes within a thermodynamic approach with enforcing
thermodynamic homogeneity of free energy. We show that hierarchic
replications of the phase space are introduced to test and reach
thermodynamic homogeneity and the number of hierarchies is fixed by
stability conditions. We demonstrate that the continuous distribution
can be obtained from the limit to infinite-many hierarchical levels of
the discrete scheme. In addition to this, we construct the Parisi
solution without demanding an infinite number of hierarchies needed
for thermodynamic stability. A natural question is then raised. How
does the Parisi continuous solution behave in situations with a stable
equilibrium state described by only a finite number of replica
hierarchies (1RSB) ?
!!!ATTENTION : JOUR ET HEURE EXCEPTIONNELS !!!
Lundi 23 Juin à 14h au LPTMS
Srikanth Sastry
(JNC Bangalore)
Growing length scales, configurational entropy and dynamics in glass forming liquids
Resume :
The relationship between dynamics, a growing length scale associated
with cooperatively rearranging regions, and configurational entropy
underlie the Adam-Gibbs theory of dynamics in glass forming liquids.
While the Adam-Gibbs relation between relaxation times and
configurational entropy has been widely employed to rationalize
experimental and computer simulation data, attempts to directly study
the length scales associated with cooperatively rearranging regions
have been relatively few. Fresh impetus in this direction comes from
recent studies of spatially heterogeneous dynamics, wherein analysis
of spatial correlations of the mobility of particles allows estimation
of a dynamical length. Finite size scaling is employed in the present
work to evaluate a dynamical length scale in a model glass forming
liquid, whose relationship to relaxation times and configurational
entropy are examined. Comparison with theoretical predictions reveal
partial agreement, and yield surprises that require further
theoretical analysis to resolve. The configurational entropy is found
to determine relaxation times for all temperatures and system sizes
studied through the Adam-Gibbs relation, but the configurational
entropy of cooperatively rearranging regions grows as the temperature
is lowered, contrary to the assumptions of Adam-Gibbs theory.
!!!ATTENTION : SOUTENANCE DE THESE !!!
Vendredi 11 Juillet à 14h30 au LPT, Amphi I
Aurelien GAUTREAU
(LPT, ORSAY)
Dynamique sur reseaux complexes et dynamique des reseaux complexes
Resume :
Depuis une dizaine d’annees, les reseaux complexes font
l’objet d’etudes intensives. La caracterisation des proprietes des reseaux
statiques est desormais bien etablie. Les reseaux sont le support de processus
dynamiques. Nous en etudierons un en particulier : la propagation d’epidemie.
Nous nous interesserons aux temps d’arrivee de la maladie dans differentes
villes dans le cadre d’un modele de propagation mondial, dit modele de
metapopulation. Ensuite, nous decrirons quelques methodes pour caracteriser
empiriquement l’evolution temporelle d’un reseau. Ces mesures nouvelles
debouchent sur la proposition d’un modele d’evolution gouvernee par le trafic
pour les reseaux peses. Enfin, dans un cas limite et dans le cadre d’un modele
issu des problemes d’opinion, nous etudierons l’interaction entre les processus
dynamique sur un reseau et la topologie du reseau lui-même (reseaux adaptifs).
Jeudi 25 Septembre à 14h au LPTMS
S. Cocco
(LPS de l’Ecole Normale Superieure)
Approches inverses pour la reconstruction d’interactions effectives entre
neurones retiniens
Resume :
Les enregistrament multi-electrodes donnent acces a l’activite
simultanee d’un grand nombre de neurones pendant plusieurs heures.
Deduire des informations sur le cablage de la retine a partir de ces
donnees est un probleme important en neurobiologie.
Je presenterai deux approches inverse qui permettent d’adresser ce
probleme et d’inferer des interactions effectives entre neurones. J’ analyserai
avec ces methodes des enregistrement de l’activites
des cellules ganglionaires dans la retine de salamandre et discuterai
les resultats obtenus.
Jeudi 02 Octobre à 14h au LPT
Andrea GAMBASSI
(Max Planck Institut für Metallforschung, Stuttgart)
The Casimir effect : from quantum to critical fluctuations.
Resume :
The Casimir effect in quantum electrodynamics (QED) is perhaps the
best-known example of fluctuation-induced long-ranged force acting on
objects (conducting plates) immersed in a fluctuating medium (quantum
electromagnetic field in vacuum).
A similar effect emerges in statistical physics, where, e.g., colloidal
particles immersed in a binary liquid mixture experience an additional
force due to the classical thermal fluctuations occurring in the
surrounding medium. This Casimir-like force acquires universal features
upon approaching a critical point of the medium and becomes long-ranged at
criticality.
In turn, this universality allows the theoretical investigation of the
force via representative models and also a stringent experimental test of
the corresponding predictions.
In contrast to QED, the strength and sign of the Casimir force resulting
from critical fluctuations can be easily tailored by surface treatments
and temperature control.
The talk reviews some recent advances in the theoretical study of the
universal properties of the critical Casimir force arising in thin films.
Our predictions compare very well with the experimental results obtained
for wetting layers of fluids. We discuss how the Casimir force between a
colloidal particle and a planar wall immersed in a binary liquid mixture
has been measured with femto-Newton accuracy, comparing these experimental
results with the corresponding theoretical predictions.
Jeudi 09 Octobre à 14h au LPT
Ihor Mryglod
(Institute od Condensed Matter Physics, L’viv, Ukraine)
Resume :
Title not yet decided