Propagation of Singularities for the Damped Stochastic Klein-Gordon Equation
Hongyi Chen, Cheuk Yin Lee · 2025
Read Propagation of Singularities for the Damped Stochastic Klein-Gordon Equation by Hongyi Chen, Cheuk Yin Lee (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-2508-13671.txt.
Introduction
Hongyi Chen and Cheuk Yin Lee tracked how sharp features travel in a noisy wave equation in a 2025 paper. The damped stochastic Klein-Gordon equation models a vibrating field with friction and random forcing. Rare, rough events tied to the law of the iterated logarithm create singularities, points where the solution loses smoothness, and the authors ask how those points move over time.
Singularities turn out to travel exactly as they do for the plain stochastic wave equation, along paths set by the highest-order part of the operator. The proofs differ sharply from the wave case, and a lucky step helps: settling the critically damped case settles the general one. Results hint at a deeper microlocal description of random roughness.
Catalog record
| CREATOR | Hongyi Chen, Cheuk Yin Lee |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 64 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2508-13671
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Propagation of Singularities for the Damped Stochastic Klein-Gordon Equation Hongyi Chen, Cheuk Yin Lee Abstract For the 1 + 1 1+1 dimensional damped stochastic Klein-Gordon equation, we show that random singularities associated with the law of the iterated logarithm exist and propogate in the same way as the stochastic wave equation. This provides evidence for possible connections to microlocal analysis, ie. the exact regularity and singularities described in this paper should admit wavefront set type descriptions whose propagation is determined by the highest order terms of the linear operator. Despite the results being exactly the same as those of the wave equation, our proofs are significantly different than the proofs for the wave equation. Miraculously, proving our results for the critically damped equation implies them for the general equation, which significantly simplifies the problem. Even after this simplification, many important parts of the proof are significantly different than (and we think are more intuitive from the PDE viewpoint compared to) existing proofs for the wave equation. 1 Introduction We study the regularity of the one-dimensional damped stochastic Klein-Gordon equation (\mathop{}\!\mathbin{\Box}+a\partial_{t}+m^{2})u(t,x)=\dot{W}(t,x),\quad t>0,x\in\mathbb{R}, (1) interpreted in the sense of [ Wal86 ] subject to initial conditions u(0,x)=\partial_{t}u(0,x)=0 . Here a,m\in\mathbb{R} are known from the physics literature as the damping constant and mass, \mathop{}\!\mathbin{\Box}=\partial_{tt}-\partial_{xx} is the d’Alembertian in one space dimension, and \dot{W} is Gaussian space-time white noise defined on a complete probability space (\Omega,\mathscr{F},\mathbb{P}) . Let \{\mathscr{F}_{t}\}_{t\geq 0} denote the filtration generated by the noise \dot{W} . Our main results are the following Theorems. Theorem 1.1 (Law of the iterated logarithm) . For any fixed t_{0}\geq 0 and x_{0}\in\mathbb{R} , there exists a constant 0<K_{1}<\infty depending on (t_{0},x_{0}) such that \displaystyle\limsup_{h\to 0^{+}}\frac{|u(t_{0}+h,x_{0}+h)-u(t_{0},x_{0})|}{\sqrt{h\log\log(1/h)}}=K_{1}\quad\text{a.s.} (2) Theorem 1.2 (Modulus of continuity) . Fix w_{0}\geq 0 and 0<a_{1}<a_{2} , and consider the line segment \displaystyle I=\left\{(t,x)\in\mathbb{R}_{+}\times\mathbb{R}:\tfrac{t-x}{\sqrt{2}}=w_{0},\tfrac{t+x}{\sqrt{2}}\in[a_{1},a_{2}]\right\}. (3) Then, there exist constants 0<C_{1}<C_{2}<\infty depending on (w_{0},a_{1},a_{2}) such that for every sub-interval J of I , there exists a constant K_{J}\in[C_{1},C_{2}] such that \displaystyle\limsup_{h\to 0^{+}}\sup_{(t,x)\in J}\frac{|u(t+h,x+h)-u(t,x)|}{\sqrt{h\log(1/h)}}=K_{J}\quad\text{a.s.} (4) Theorem 1.1 states that at a fixed space-time point, the increments (along a characteristic direction) are locally of order \sqrt{h\log\log(1/h)} , while Theorem 1.2 asserts that the uniform modulus of continuity for the increments are of a larger order, at a logarithmic level. Therefore, Theorems 1.1 and 1.2 together suggest that there exist random space-time points at which local increments are larger than those at a fixed point. Theorem 1.4 below justifies the existence of these random singularities and shows that they propagate along the other characteristic direction. One of the ingredients for proving Theorem 1.4 is the following simultaneous upper bound for the law of the iterated logarithm. For notational convenience, we introduce a change of coordinates via a rotation by -45^{\circ} : \displaystyle(w,z)=\left(\tfrac{t-x}{\sqrt{2}},\tfrac{t+x}{\sqrt{2}}\right),\qquad(t,x)=\left(\tfrac{w+z}{\sqrt{2}},\tfrac{-w+z}{\sqrt{2}}\right). (5) The w and z directions are the characteristic directions of the wave equation. Theorem 1.3 . Fix N>0 and 0<a_{1}<a_{2} . Then, there exists a constant 0<K_{2}<\infty such that for all fixed z_{0}\in[a_{1},a_{2}] , \displaystyle\mathbb{P}\left\{\limsup_{h\to 0^{+}}\frac{|u\big(\frac{w+z_{0}}{\sqrt{2}}+h,\frac{-w+z_{0}}{\sqrt{2}}+h\big)-u\big(\frac{w+z_{0}}{\sqrt{2}},\frac{-w+z_{0}}{\sqrt{2}}\big)|}{\sqrt{h\log\log(1/h)}}\leq K_{2}\text{ for all $w\in[0,N]$}\right\}=1. (6) Theorem 1.4 (Existence and propagation of singularities) . Fix w_{0}\geq 0 , 0<z_{0}<z_{0}^{\prime} . Let t_{0}=w_{0}/\sqrt{2} . Then: (i) There exists an \mathscr{F}_{t_{0}} -measurable random variable Z=Z(\omega) such that z_{0}\leq Z\leq z_{0}^{\prime} a.s. and \displaystyle\limsup_{h\to 0^{+}}\frac{|u\big(\frac{w_{0}+Z}{\sqrt{2}}+h,\frac{-w_{0}+Z}{\sqrt{2}}+h\big)-u\big(\frac{w_{0}+Z}{\sqrt{2}},\frac{-w_{0}+Z}{\sqrt{2}}\big)|}{\sqrt{h\log\log(1/h)}}=+\infty\quad\text{a.s.} (7) (ii) If Z is any such random variable, then \displaystyle\mathbb{P}\left\{\limsup_{h\to 0^{+}}\frac{|u\big(\frac{w+Z}{\sqrt{2}}+h,\frac{-w+Z}{\sqrt{2}}+h\big)-u\big(\frac{w+Z}{\sqrt{2}},\frac{-w+Z}{\sqrt{2}}\big)|}{\sqrt{h\log\log(1/h)}}=+\infty\text{ for all $w>w_{0}$}\right\}=1. (8) Remark 1.5 . In general one can define \mathop{}\!\mathbin{\Box}=b^{2}\partial_{tt}-c^{2}\partial_{xx} , but a simple change of coordinates would allow one to produce the above results (with appropriately tilted characteristic lines) with the same calculations. We also note that all of our results can apply to the excited stochastic Klein-Gordon equation (a<0) , but with extra dependence on T the time horizon in all of the constants in the estimates. Remark 1.6 . Theorems 1.1 – 1.2 together imply that the B^{s}_{\infty,\infty}- Besov wavefront set [ DRS22 ] of u is empty if s<\frac{1}{2} and is equal to (\mathbb{R}_{+}\times\mathbb{R})\times(\mathbb{R}^{2}\setminus\{(0,0)\}) if s\geq\frac{1}{2} . With some extra work, the same statements can be said about the more familiar H^{s}- Sobolev wavefront sets [ Hin25 , Definition 7.35, Proposition 7.36] of u with the same regularity parameters. We will discuss the relevance of this in the literature review. 1.1 Literature review and motivation The (Khinchin) law of the iterated logarithm (LIL, for short) was first stated by Khinchin in 1924 [ Khi24 ] as a description of the fluctuations of a symmetric random walk. Since Brownian motion is the limit of scaled random walks, the LIL for Brownian motion [ Khi33 ] can be interpreted as both regularity and large time fluctuation statements. We will concentrate on the regularity interpretation, for which we note that it has since been generalized to stochastic differential equations and Brownian sheets [ Zim72 , OP73 ] . Importantly, local versions of the LIL with a different power of h in the denominator of ( 2 ) has been proven for fractional Brownian Motion and fractional Brownian sheets [ Ore72 , AX05 ] , which can be seen as a reflection of their regularity being higher or lower than Brownian motion. All of these have also been generalized to Gaussian processes, Gaussian random fields, and stochastic partial differential equations (SPDEs) driven by appropriate objects [ MR06 , MWX13 , LX22 , HL25 ] . Another regularity property of Brownian motions, Levy’s modulus of continuity (MC) [ Lév37 ] , is apparently at odds with the LIL since it divides by a larger denominator compared to the statement of LIL. This has also been generalized to Gaussian processes and SPDEs [ MR06 , MWX13 , LX19 , LX23 , HL25 ] . Notably, another way the LIL and MC are different is that the LIL fixes a point in parameter space while MC allows variation of the base point in parameter space. This indicates the existence of random points at which local increments are exceptionally large and fail to satisfy LIL. We will call such points LIL-singularities. Exceptional increments were first studied for the Brownian motion by Orey and Taylor [ OT74 ] who showed that the set of LIL-singularities is dense, uncountable, and has Hausdorff dimension 1, and also studied different types of singularities. Similar results are known for Brownian sheets, general Gaussian processes, and SPDEs [ Wal82 , Wal86 , KS00 , KPX00 , HK17 ] . In particular, Walsh [ Wal82 ] studied the LIL-singularities of a 2-parameter Brownian sheet and proved that singularities along an axis direction propagate perpendicularly along the other axis direction. Walsh modified his argument to show that LIL-singularities of stochastic wave equation propagate along characteristic directions [ Wal86 ] . These results have been extended to semi-fractional Brownian sheets [ BM08 ] and more general stochastic wave equations [ CN88 , LX22 ] . Interpreting LIL singularities as local statements about behavior near a point hints at possible connections to microlocal analysis [ Sat70 , Hör71 , Hör73 , Vas11 , Hin25 ] , which encompasses the sharpest propagation of singularities literature for deterministic equations. In microlocal analysis, singularities are characterized using wavefront sets. Informally, a point (x,\xi)\in\mathbb{R}^{n}\times(\mathbb{R}^{n}\setminus\{0\}) (viewed as the cotangent bundle of \mathbb{R}^{n} ) is an element of the (smooth) wavefront set of some Schwartz distribution \phi if \phi is not smooth in