Diffusion in Quenched Random Environments: Reviving Laplace’s First Law of Errors

Lucianno Defaveri, Eli Barkai · 2025

Read Diffusion in Quenched Random Environments: Reviving Laplace’s First Law of Errors by Lucianno Defaveri, Eli Barkai (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-05585.txt.

Introduction

Lucianno Defaveri and Eli Barkai wrote a 2025 statistical physics paper on diffusion in disordered, frozen environments, reviving an old law of Laplace. Laplace's first law of errors says error sizes fall off exponentially with their size. It was ignored for decades in favor of the Gaussian law, yet recent work found it describes tracers wandering in glassy media.

Using the trap model, where particles wait in energy wells of random depth, the authors bound the spread of a particle packet at large displacements. Both the lower and upper bounds follow Laplace like exponential laws. They connect the effect to the density of trap energies and show it is universal, though its constants shift with temperature and disorder. The paper ties a forgotten law to modern anomalous diffusion.

Catalog record

CREATORLucianno Defaveri, Eli Barkai
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-05585

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Diffusion in Quenched Random Environments: Reviving Laplace’s First Law of Errors Lucianno Defaveri, Eli Barkai Abstract Laplace’s first law of errors, which states that the frequency of an error can be represented as an exponential function of the error magnitude, was overlooked for many decades but was recently shown to describe the statistical behavior of diffusive tracers in disordered, glassy-like media. While much is known about this behavior, a key ingredient is still missing: the relationship between this observation and diffusion in a quenched random environment. We address this problem using the trap model, deriving lower and upper bounds on the particle packet for large displacements. Our results demonstrate that both bounds exhibit Laplace-like laws. We further establish a connection between the density of energy traps ρ ⁡ ( E ) \rho(E) , and the observed behavior, showing that the phenomenon is truly universal, albeit with constants that depend on temperature and the level of disorder. Introduction. Diffusion in heterogeneous random environments is observed in various scientific domains, from tracers in glasses [ 1 , 2 , 3 ] , to molecules in cellular environments [ 4 , 5 ] and charge carriers in amorphous solids [ 6 ] . The theoretical exploration of the relationship between quenched disorder and diffusion is extensive [ 7 , 8 , 9 , 10 , 11 ] . Classical theories predict a Gaussian spreading packet with a re-normalized diffusion constant in the long-time limit and under weak disorder conditions [ 8 , 12 , 13 ] . However, the emergence of exponential tails in the spreading packet of particles, as widely observed in laboratory and in simulations, revolutionized the paradigms of diffusion phenomenon [ 14 , 1 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 28 , 30 , 31 , 32 , 33 ] . In many of these systems, the disorder is not too strong, leading to normal diffusion characterized by the mean square displacement \langle x^{2}\rangle\propto t [ 16 , 19 ] . Stochastic models such as diffusing diffusivity [ 34 , 35 ] and Brownian yet non-Gaussian models [ 36 , 37 ] describe this phenomenon and its implications, including, for example, a significant impact on the rate of chemical reactions [ 38 ] . These observations are related to the work of Laplace, who formulated two laws of errors [ 39 , 40 , 41 ] . The first law states that the error distribution decays exponentially with the size of the error, while the second law asserts that the error distribution is quadratic with the error size. The first law is deeply connected to recent observations of exponential tails in diffusion processes, while the second law is intimately related to the central limit theorem and the normal distribution. Chaudhuri, Bertheir, and Kob [ 1 ] demonstrated how Laplace-like tails appear in continuous-time random walks (CTRW) [ 13 ] , promoting the idea that exponential Laplace-like tails are indeed universal [ 1 , 42 , 43 , 44 , 41 ] . However, the CTRW model is a mean-field theory in which the time between jumps are independent identically distributed random variables, and thus, mathematically, the renewal approach can be used. A natural question is whether and under what conditions Laplace-like behavior arises in systems with quenched (i.e., time-independent) disorder [ 45 , 46 ] . Despite its widespread significance, the precise mechanisms and criteria for this phenomenon in spatially quenched systems remain unresolved. For particles diffusing in a random environment, disorder appears in various forms, such as energy and entropic traps or barriers. Potential traps of differing depths induce correlations in the motion of the tracer upon re-visitation, a feature shared across other types of disorder. A major study of diffusion in these systems has focused on the long-time limit of the spreading particle packet. Novel effects, such as the transition from normal to anomalous diffusion, have been well explored [ 47 , 8 , 48 , 49 , 50 , 51 , 52 , 53 , 54 , 55 , 56 ] . Our goal is to study a different limit, relevant to experiments, where the displacement x is large but time t is intermediate. This limit allows us to focus on Laplace’s first law. In single tracer experiments, a common feature is the hopping of particles among traps. The average number of hops does not need to be large for the tracer packet’s density to exhibit universal features at large displacements. This is because large displacements correspond to many hops, leading to the emergence of universal statistical laws in the tail. Experimentally, if one waits long enough, packets of particles often transition to a normal distribution, in line with Laplace’s second law. Trap Model. We will consider a random walk in a random environment on a one-dimensional lattice with unit lattice spacing [ 7 ] . The probability of finding the particle at position x at time t is P(x,t) , where initially the particle is on the origin, x=0 . The evolution of the system is governed by a master equation with typical gain and loss terms: \frac{\partial P(x,t)}{\partial t}=-R_{x}P(x,t)+\frac{R_{x+1}}{2}P(x+1,t)+\frac{R_{x-1}}{2}P(x-1,t). (1) The transition rates R_{x}>0 are time-independent, mutually independent, identically distributed random variables drawn from a common probability density function (PDF) f(R) . This model describes a particle moving on a one-dimensional lattice, where residence times at each site are exponentially distributed with a rate R_{x} . The dynamics are non-biased, with the particle hopping to x\pm 1 with equal probability 1/2 after waiting at x . The key observable is \langle P(x,t)\rangle , where the brackets indicate an average over the quenched disorder \langle P(x,t)\rangle=\int_{0}^{\infty}...\int_{0}^{\infty}\Pi_{y=-\infty}^{\infty}f(R_{y}){\rm d}R_{y}P(x,t) . This averaging mimics a set of experiments all conducted in specific, mutually independent environments, which is a common situation in single particle tracking protocols. Ordered system. When all rates are identical and equal r , the system is ordered and Eq. ( 1 ) gives [ 7 , 57 ] : P(x,t)=\exp(-rt)I_{|x|}(rt) (2) where I_{|x|}(\cdot) is the modified Bessel function of the first kind. For large times and x\propto\sqrt{t} , the solution is Gaussian as expected from the central limit theorem. However, we are interested in a different limit, where x is large and t is fixed. Using the asymptotic limit of the modified Bessel function, we find an exponential decay with logarithmic corrections, see supplemental material (SM) [ 58 ] P(x,t)\sim\exp\left[-|x|\ln\left({2|x|\over ert}\right)\right]\frac{e^{-rt}}{\sqrt{2\pi|x|}}. (3) Rewriting, \lim_{|x|\to\infty}{-\ln P(x,t)\over|x|\ln\left({2|x|\over ert}\right)}=1 (4) so the factor \exp(-rt)/\sqrt{2\pi|x|} in Eq. ( 3 ), while important for comparison to data, contributes insignificantly to the asymptotics. Such Laplace-like tails of P(x,t) are found in full generality for CTRW models [ 1 , 45 , 42 , 43 , 46 , 44 , 41 , 59 ] . Figure 1: The log of the disorder-averaged probability \langle P(x,t)\rangle , obtained through numerical simulations, divided by x\ln[(2x)/(e\langle R\rangle t)] , illustrating the validity of Eq. ( 12 ), and hence the Laplace-like behavior. This universal law is shown for different types of disorder \rho(E) (colored symbols shown in the caption) for t=1

Read the complete text