The Landau equation and Fisher information
Nestor Guillen, Luis Silvestre · 2025
Read The Landau equation and Fisher information by Nestor Guillen, Luis Silvestre (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-2507-05167.txt.
Introduction
Nestor Guillen and Luis Silvestre explain a breakthrough on the Landau equation in an expository note from 2025. The equation describes how particles in a plasma collide and spread. A feared outcome is blow-up, where values grow without bound in finite time. Their recent work rules out blow-up for the case of Coulomb forces, and the note lays out the ideas for a broad math audience.
The key step tracks a quantity called Fisher information along a solution. They show it does not increase over time, which blocks blow-up. A lifted helper equation, built in twice the variables, makes the proof work. For this equation it becomes a family of heat equations on a sphere. Background links round out the note.
Catalog record
| CREATOR | Nestor Guillen, Luis Silvestre |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-05167
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
The Landau equation and Fisher information Nestor Guillen, Luis Silvestre Abstract 1 Introduction The following initial value problem is of great interest in kinetic theory: to find, given an initial mass density f_{\text{in}}:\mathbb{R}^{3}\to\mathbb{R} , a function f(t,v) smooth for all times t>0 and velocities v\in\mathbb{R}^{3} , which starts at time t=0 from f_{\text{in}} and solves for t>0 \displaystyle\partial_{t}f=\bar{a}_{ij}\partial_{ij}^{2}f+f^{2}\text{ where }\bar{a}:=-D^{2}(-\Delta)^{-2}f (1) This equation is known as the homogeneous Landau equation with Coulomb potential. It is an equation with a rich structure, it has both integro-differential and nonlinear terms. The term f^{2} raises the possibility that solutions starting from a smooth initial f_{\text{in}} might develop singularities (“blow up”) in final time. In this note we explain how the blow up question for the Landau equation was ruled out through the study of the Fisher information, which turns out to be monotone in time. In [ 7 ] we show that if f is a classical solution of ( 1 ) then \displaystyle\frac{d}{dt}\int_{\mathbb{R}^{3}}\frac{|\nabla f|^{2}}{f}\;\mathrm{d}v\leq 0. This inequality is obtained via a new “lifted equation” taking place in \mathbb{R}^{6} . This lifted equation amounts to the heat equation on the sphere, a connection that was at first somewhat surprising but quite natural in hindsight. As this equation is local and linear, it is much easier to analyze than ( 1 ), and yet it encodes enough about ( 1 ) to reach conclusions about how the Fisher information evolves with time. 1.1 Origin of the Landau equation It is proper to start by reviewing the Boltzmann equation and how the Landau equation ( 1 ) arises from it. A complete introduction to the Boltzmann equation and its origins can be found in Cercignani’s book [ 5 ] . A thorough survey of the mathematical aspects of the theory and the state of the field at the turn of the century is in Villani’s review [ 17 ] . In 1872 Boltzmann [ 2 ] derived an equation to model the evolution of the statistics of particles in a monoatomic gas, it takes the form \displaystyle\partial_{t}f+v\cdot\nabla_{x}f=q(f) (2) The function f=f(t,x,v) represents the density of particles with position x and velocity v , the equation indicates how f changes with time as a result of the inertia of particles (this is captured by the term v\cdot\nabla_{x}f ) and the result of interactions between particles, such as collisions (captured by the term q(f) ). The collision operator q(f) introduced by Boltzmann should be thought of as providing a mass balance for how many particles are gaining the velocity v minus how many are losing the velocity v to end at another velocity as a result of a collision. Boltzmann showed that the collision term q(f) evaluated (t,x,v) is given by the integral (we write f(v) for f(t,x,v) , and f(w) for f(t,x,w) et cetera) \displaystyle\int_{\mathbb{R}^{3}}\int_{\mathbb{S}^{2}}(f(v_{\sigma})f(w_{\sigma})-f(v)f(w))B\;\mathrm{d}\sigma\mathrm{d}w (3) Here, \displaystyle B \displaystyle=\alpha(|v-w|)b(\sigma\cdot\tfrac{v-w}{|v-w|}) \displaystyle v_{\sigma} \displaystyle=\tfrac{1}{2}(v+w)+\tfrac{1}{2}|v-w|\sigma \displaystyle w_{\sigma} \displaystyle=\tfrac{1}{2}(v+w)-\tfrac{1}{2}|v-w|\sigma Here we will refer to \alpha:(0,\infty)\to[0,\infty) as the potential . Maxwell [ 11 ] computed \alpha(r) and b(\cos(\theta)) in terms of the force modeling interactions (collisions) between pairs of particles. For a central force given by s -th power of the inverse distance we have \displaystyle\alpha(r)=r^{\gamma},\textnormal{ where }\gamma:=\frac{s-5}{s-1} For the case of Coulomb interactions this means s=2 and so \gamma=-3 . It is well known that \displaystyle b(\sigma\cdot\sigma_{0})\approx|\sigma-\sigma_{0}|^{-2-\frac{2}{s-1}} Here \sigma_{0}=\tfrac{v-w}{|v-w|} . The integral in ( 3 ) is not finite for any non-constant f when s>2 . One way of thinking about this is that when s is close to 2 then b(\sigma\cdot\sigma_{0}) assigns most of its weight to those collisions with \sigma\approx\sigma_{0} , what are known as grazing collisions . For this reason it is interesting to consider the asymptotic regime where the contributions to the collisional integral are concentrated on directions \sigma close to \sigma_{0}=\tfrac{v-w}{|v-w|} . This is known in the literature as the grazing limit . In this regime, most of the particle interactions produce asymptotically small changes in velocity. It is a regime of physical relevance because it is used to describe particles that repel each other by Coulombic forces. The Landau collision operator is obtained from the Boltzmann collision operator in the grazing limit. Different families of collision kernels can be considered, with the common feature that they concentrate near \sigma=\sigma_{0} . The inner spherical integral in ( 3 ) is an spherical integro-differential operator applied to the function f(v)f(w) with the kernel b(\sigma\cdot\sigma_{0}) , which is a function of the angle between \sigma and \sigma_{0} . As b concentrates on small angles, the integro-differential operator converges to the Laplace-Beltrami operator on the sphere \mathbb{S}^{2} with respect to the variable \sigma , applied to the function f(v)f(w) , q_{L}(f)(v)=\int_{\mathbb{R}^{3}}\alpha(|v-w|)\Delta_{\sigma}[f(v)f(w)]\;\mathrm{d}w. (4) It is arguably not obvious that when \alpha(r)=r^{-3} , this operator is the same Landau collision operator that we wrote at the beginning of this article ( 1 ). The Landau collision operator has several equivalent formulations. It is useful to know its many faces, since they can be used for different purposes. For general \alpha(r) , the following are common representations of the Landau collision operator • Nondivergence form q_{L}(f)=\bar{a}_{ij}\partial_{ij}f+\bar{c}f, where \bar{c}=-\partial_{ij}\bar{a}_{ij} and \bar{a}_{ij}=a_{ij}\ast f , for the matrix-valued function a_{ij} given by a_{ij}(z)=\alpha(|z|)\left(|z|^{2}\delta_{ij}-z_{i}z_{j}\right) (5) • Divergence form q_{L}(f)=\partial_{i}\left(\bar{a}_{ij}\partial_{j}f-\bar{b}_{i}f\right), where \bar{a}_{ij} is as before and \bar{b}_{i}=\partial_{j}\bar{a}_{ij} . • Integral form \displaystyle q_{L}(f)= \displaystyle\partial_{i}\int_{\mathbb{R}^{3}}a_{ij}(v-w)\left(f(w)\partial_{j}f(v)-f(v)\partial_{j}f(w)\right)\mathrm{d}w. It is not too complicated to verify that the expression in ( 4 ) coincides with the integral form of the Landau collision operator. The formulas for the operator in divergence and nondivergence form are computed from this integral formulation in an attempt to make the Landau collision operator look like a classical diffusion operator. 1.2 The homogeneous regime and the blow up question The study of kinetic equations faces two main analytical difficulties. First, while the collision operator has a diffusive regularization effect, it is only with respect to the v -variable. To get a regularization in all variables we have to understand the interaction between the diffusion in v and the transport in x . This is where hypoelliptic theory applies to kinetic equations. The other difficulty is that the collision operator q(f) has a rather complex structure –it is a nonlocal and quadratic expression in f . For configurations that are homogeneous in space, namely when f(t,x,v)=f(t,v) , the transport effects in the equation vanish, provides us with an ideal situation to analyze the effects of collisions. In this “spatially homogeneous” regime the Boltzmann equation ( 2 ) simplifies to \partial_{t}f=q(f) (6) The spatially homogeneous regime, as simplified as it might be, is still of considerable interest to physicists – see for instance the discussion in Pauli’s Statistical Mechanics [ 13 ] [Chapter 1]. Mathematically, it is a necessary first step before any attempt to understand the full equation ( 2 ). The most natural idea to prove that the solution of the homogeneous Landau equation ( 6 ) stays bounded and smooth for all time is to take advantage of the regularization effect of the diffusion term in q_{L}(f) (either in divergence or non-divergence form). However, the presence of the nonlinear term f^{2} in ( 1 ) makes this approach difficult. This difficulty exists whenever the potential \alpha(r) is too singular near r=0 . The equation (