arXiv subjects

Sergio Davis

Publications and source records attributed to Sergio Davis.

53 records · Page 3Linked to original sources

arXiv2019

Emergence of Tsallis statistics as a consequence of invariance

For non-equilibrium systems in a steady state we present two necessary and sufficient conditions for the emergence of $q$-canonical ensembles, also known as Tsallis statistics. These conditions are invariance requirements over the definition of subsystem and environment, and over the joint rescaling of temperature and energy. Our approach is independent of, but complementary to, the notions of Tsallis non-extensive statistics and Superstatistics.

cond-mat.stat-mech↗

arXiv2017

Deterministic physical systems under uncertain initial conditions: the case of maximum entropy applied to projectile motion

The kinematics and dynamics of deterministic physical systems have been a foundation of our understanding of the world since Galileo and Newton. For real systems, however, uncertainty is largely present via external forces such as friction or lack of precise knowledge about the initial conditions of the system. In this work we focus in the latter case and describe the use of inference methodologies in solving the statistical properties of classical systems subject to uncertain initial conditions. In particular we describe the application of the formalism of Maximum Entropy (MaxEnt) inference to the problem of projectile motion given information about the average horizontal range over many realizations. By using MaxEnt we can invert the problem and use the provided information on the average range to reduce the original uncertainty in the initial conditions, while also achieving additional insights based on the shape of the posterior probabilities for the initial conditions probabilities and the projectile path distribution itself. The wide applicability of this procedure, as well as its ease of use, reveals a useful tool by which to revisit a large number of physics problems, from classrooms to frontier research.

cond-mat.stat-mech↗

arXiv2017

Bayesian statistical modelling of microcanonical melting times at the superheated regime

Homogeneous melting of superheated crystals at constant energy is a dynamical process, believed to be triggered by the accumulation of thermal vacancies and their self-diffusion. From microcanonical simulations we know that if an ideal crystal is prepared at a given kinetic energy, it takes a random time $t_w$ until the melting mechanism is actually triggered. In this work we have studied in detail the statistics of $t_w$ for melting at different energies by performing a large number of Z-method simulations and applying state-of-the-art methods of Bayesian statistical inference. By focusing on the short-time tail of the distribution function, we show that $t_w$ is actually gamma-distributed rather than exponential (as asserted in previous work), with decreasing probability near $t_w \sim 0$. We also explicitly incorporate in our model the unavoidable truncation of the distribution function due to the limited total time span of a Z method simulation. The probabilistic model presented in this work can provide some insight into the dynamical nature of the homogeneous melting process, as well as giving a well-defined practical procedure to incorporate melting times from simulation into the Z-method in order to reduce the uncertainty in the melting temperature.

cond-mat.stat-mech↗

arXiv2017

Microcanonical Monte Carlo approach for computing melting curves by atomistic simulations

We report microcanonical Monte Carlo simulations of melting and superheating of a generic, Lennard-Jones system starting from the crystalline phase. The isochoric curve, the melting temperature $T_m$ and the critical superheating temperature $T_{LS}$ obtained are in close agreement (well within the microcanonical temperature fluctuations) with standard molecular dynamics one-phase and two-phase methods. These results validate the use of microcanonical Monte Carlo to compute melting points, a method which has the advantage of only requiring the configurational degrees of freedom. Our findings show that the strict preservation of the Hamiltonian dynamics does not constitute a necessary condition to produce a realistic estimate of $T_{LS}$ and the melting point, which brings new insight on the nature of the melting transition. These results widen the use and applicability of the recently developed Z method for the determination of the melting points of materials.

cond-mat.stat-mech↗

arXiv2016

Plausible inference from the idea of estimation

The probability axioms by R. T. Cox can be regarded as the modern foundations of Bayesian inference, the idea of assigning degrees of belief to logical propositions in a manner consistent with Boolean logic. In this work it is shown that one can start from an alternative point of view, postulating the existence of an operation for \textit{estimating magnitudes given a state of knowledge}. The properties of this operation can be established through basic consistency requirements. Application of this estimation operation to the truth values of propositions in an integer representation directly yields non-negative, normalized quantities following the sum and product rules of probability.

math.PR↗

arXiv2016

Jarzynski equality in the context of maximum path entropy

In the global framework of finding an axiomatic derivation of nonequilibrium Statistical Mechanics from fundamental principles, such as the maximum path entropy -- also known as Maximum Caliber principle -- , this work proposes an alternative derivation of the well-known Jarzynski equality, a nonequilibrium identity of great importance today due to its applications to irreversible processes: biological systems (protein folding), mechanical systems, among others. This equality relates the free energy differences between two equilibrium thermodynamic states with the work performed when going between those states, through an average over a path ensemble. In this work the analysis of Jarzynski's equality will be performed using the formalism of inference over path space. This derivation highlights the wide generality of Jarzynski's original result, which could even be used in non-thermodynamical settings such as social systems, financial and ecological systems.

cond-mat.stat-mech↗

arXiv2016

Exponential Family Models from Bayes' Theorem under Expectation Constraints

It is shown that a consistent application of Bayesian updating from a prior probability density to a posterior using evidence in the form of expectation constraints leads to exactly the same results as the application of the maximum entropy principle, namely a posterior belonging to the exponential family. The Bayesian updating procedure presented in this work is not expressed as a variational principle, and does not involve the concept of entropy. Therefore it conceptually constitutes a complete alternative to entropic methods of inference.

physics.data-an↗

arXiv2016

Continuity equation for probability as a requirement of inference over paths

In this work we present the fundamental ideas of inference over paths, and show how this formalism implies the continuity equation, which is central for the derivation of the main partial differential equations that constitute non-equilibrium statistical mechanics. Equations such as the Liouville equation, Fokker-Planck equation, among others can be recovered as particular cases of the continuity equation, under different probability fluxes. We derive the continuity equation in its most general form through what we call the \bf time-slicing equation, which lays down the procedure to go from the representation in terms of a path probability functional $ρ[X()]$ to a time-dependent probability density $ρ(x; t)$. The original probability functional $ρ[X()]$ can in principle be constructed from different methods of inference; in this work we sketch an application using the maximum path entropy or maximum caliber principle.

cond-mat.stat-mech↗

arXiv2016

Temperature is not an observable in superstatistics

Superstatistics (Physica A 322, 267-275, 2003) is a formalism that attempts to explain the presence of distributions other than the Boltzmann-Gibbs distributions in Nature, typically power-law behavior, for systems out of equilibrium such as fluids under turbulence, plasmas and gravitational systems. Superstatistics postulates that those systems are found in a superposition of canonical ensembles at different temperatures. The usual interpretation is one of local thermal equilibrium (LTE) in the sense of an inhomogeneous temperature distribution in different regions of space or instants of time. Here we show that, in order for superstatistics to be internally consistent, it is impossible to define a phase-space function or observable $B(p, q)$ corresponding one-to-one to the local value of $β=1/k_B T$. Temperature then belongs to a different class of observables than the energy, which has as a phase-space function the Hamiltonian $\mathcal{H}(p, q)$. An important consequence of our proof is that, in Superstatistics, the identification of temperature with the kinetic energy is limited to the expectation of $β$ and cannot be used to measure the different temperatures in LTE or its fluctuations.

cond-mat.stat-mech↗

arXiv2016

Applications of the divergence theorem in Bayesian inference and MaxEnt

Given a probability density $P({\bf x}|{\boldsymbol λ})$, where $\bf x$ represents continuous degrees of freedom and $λ$ a set of parameters, it is possible to construct a general identity relating expectations of observable quantities, which is a generalization of the equipartition theorem in Thermodynamics. In this work we explore some of the consequences of this relation, both in the context of sampling distributions and Bayesian posteriors, and how it can be used to extract some information without the need for explicit calculation of the partition function (or the Bayesian evidence, in the case of posterior expectations). Together with the general family of fluctuation theorems it constitutes a powerful tool for Bayesian/MaxEnt problems.

cond-mat.stat-mech↗

arXiv2016

Liouville's Theorem from the Principle of Maximum Caliber in Phase Space

One of the cornerstones in non--equilibrium statistical mechanics (NESM) is Liouville's theorem, a differential equation for the phase space probability $ρ(q,p; t)$. This is usually derived considering the flow in or out of a given surface for a physical system (composed of atoms), via more or less heuristic arguments. In this work, we derive the Liouville equation as the partial differential equation governing the dynamics of the time-dependent probability $ρ(q, p; t)$ of finding a "particle" with Lagrangian $L(q, \dot{q}; t)$ in a specific point $(q, p)$ in phase space at time $t$, with $p=\partial L/\partial \dot{q}$. This derivation depends only on considerations of inference over a space of continuous paths. Because of its generality, our result is valid not only for "physical" systems but for any model depending on constrained information about position and velocity, such as time series.

physics.class-ph↗

arXiv2015

Hamiltonian formalism and path entropy maximization

Maximization of the path information entropy is a clear prescription for constructing models in non-equilibrium statistical mechanics. Here it is shown that, following this prescription under the assumption of arbitrary instantaneous constraints on position and velocity, a Lagrangian emerges which determines the most probable trajectory. Deviations from the probability maximum can be consistently described as slices in time by a Hamiltonian, according to a nonlinear Langevin equation and its associated Fokker-Planck equation. The connections unveiled between the maximization of path entropy and the Langevin/Fokker-Planck equations imply that missing information about the phase space coordinate never decreases in time, a purely information-theoretical version of the Second Law of Thermodynamics. All of these results are independent of any physical assumptions, and thus valid for any generalized coordinate as a function of time, or any other parameter. This reinforces the view that the Second Law is a fundamental property of plausible inference.

physics.data-an↗

arXiv2014

Statistical distribution of thermal vacancies close to the melting point

A detailed description of the statistical distribution of thermal vacancies near the melting point is presented, using copper as an example. As the temperature is increased, the average number of thermal vacancies generated by atoms migrating to neighboring sites also increase, according to Arrhenius' law. We present for the first time a model for the distribution of thermal vacancies, which according to our results follow a Gamma distribution. All the simulations are carried out by classical molecular dynamics and the recognition of vacancies is achieved via a recently developed algorithm. Our results could be useful in the further development of a theory explaining the mechanism of homogeneous melting, which seems to be mediated (at least in part) by the accumulation of thermal vacancies near the melting point.

cond-mat.mtrl-sci↗

arXiv2014

A Bayesian interpretation of abrupt phase transitions

The formalism used in describing the thermodynamics of abrupt (or first-order) phase transitions is reviewed as an application of maximum entropy inference. In this treatment, we show that the concepts of transition temperature, latent heat and entropy difference between phases will inevitably have an equivalent in any problem of inferring the result of a yes/no question, given information in the form of expectation values.

cond-mat.stat-mech↗

arXiv2014

A maximum entropy model for opinions in social groups

We study how the opinions of a group of individuals determine their spatial distribution and connectivity, through an agent-based model. The interaction between agents is described by a Potts-like Hamiltonian in which agents are allowed to move freely without an underlying lattice (the average network topology connecting them is determined from the parameters). This kind of model was derived using maximum entropy statistical inference under fixed expectation values of certain probabilities that (we propose) are relevant to social organization. Control parameters emerge as Lagrange multipliers of the maximum entropy problem, and they can be associated with the level of consequence between the personal beliefs and external opinions, and the tendency to socialize with peers of similar or opposing views. These parameters define a phase diagram for the social system, which we studied using Monte Carlo Metropolis simulations. Our model presents both first and second-order phase transitions, depending on the ratio between the internal consequence and the interaction with others. We have found a critical value for the level of internal consequence, below which the personal beliefs of the agents seem to be irrelevant.

physics.soc-ph↗

arXiv2014

Fitting of interatomic potentials without forces: a parallel particle swarm optimization algorithm

We present a methodology for fitting interatomic potentials to ab initio data, using the particle swarm optimization (PSO) algorithm, needing only a set of positions and energies as input. The prediction error of energies associated with the fitted parameters can be close to 1 meV/atom or lower, for reference energies having a standard deviation of about 0.5 eV/atom. We tested our method by fitting a Sutton-Chen potential for copper from \emph{ab initio} data, which is able to recover structural and dynamical properties, and obtain a better agreement of the predicted melting point versus the experimental value, as compared to the prediction of the standard Sutton-Chen parameters.

cond-mat.mtrl-sci↗

arXiv2013

Newtonian Dynamics from the principle of Maximum Caliber

The foundations of Statistical Mechanics can be recovered almost in their entirety from the Principle of Maximum Entropy. In this work we show that its non-equilibrium generalization, the Principle of Maximum Caliber (Jaynes, 1980), when applied to the unknown trajectory followed by a particle, leads to Newton's second law under two quite intuitive assumptions (the expected square displacement in one step and the spatial probability distribution of the particle are known at all times). Our derivation explicitly highlights the role of mass as an emergent measure of the fluctuations in velocity (inertia) and the origin of potential energy as a manifestation of spatial correlations. According to our findings, the application of Newton's equations is not limited to mechanical systems, and therefore could be used in modelling ecological, financial and biological systems, among others.

physics.data-an↗