Propagation of singularities for equations with C r C^{r} coefficients for r > 1 r>1

Jan Rozendaal · 2025

Read Propagation of singularities for equations with C r C^{r} coefficients for r > 1 r>1 by Jan Rozendaal (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-2510-16182.txt.

Introduction

Jan Rozendaal, writing alone, proves a result about how sharp features in waves travel through uneven media. The waves obey equations whose coefficients, the numbers that set local wave speed, are only mildly smooth. A classic law says sharp spots must move along set paths called bicharacteristics, never appearing from nowhere.

Rozendaal shows that law still holds when the coefficients have just a touch more than minimal smoothness. For one key case, even slightly less smoothness than the usual bar is enough. The proof works across curved spaces too, as long as their bending stays bounded. The result widens the range of rough media and curved spaces where wave paths can be trusted.

Catalog record

CREATORJan Rozendaal
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-16182

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 equations with C r C^{r} coefficients for r > 1 r>1 Thanks: This research was funded in part by the National Science Center, Poland, grant 2021/43/D/ST1/00667. The author is partially supported by NCN grant UMO-2023/49/B/ST1/01961. Jan Rozendaal Abstract . We observe that, for r > 1 r>1 , s s in an r r -dependent interval, p p a homogeneous pseudodifferential symbol of order m m having C r C^{r} regularity in space, and u ∈ H s + m − r ​ ( ℝ n ) u\in H^{s+m-r}(\mathbb{R}^{n}) such that p ⁡ ( x , D ) ​ u ∈ H s ​ ( ℝ n ) p(x,D)u\in H^{s}(\mathbb{R}^{n}) , each point in the H s + m − 1 H^{s+m-1} wavefront set of u u lies on a maximally extended null bicharacteristic of p p which is contained in the H s + m − 1 H^{s+m-1} wavefront set of u u . In fact, for r = 2 r=2 slightly less than C 1 , 1 C^{1,1} regularity suffices, and here the results apply to manifolds with bounded Ricci curvature. 2020 Mathematics Subject Classification Primary 58J47. Secondary 35A18, 35S50, 58J45 1. Introduction In this article we consider a version for linear equations with rough coefficients of the classical statement on propagation of singularities. 1.1. Main results Our results are formulated using pseudodifferential symbols that have limited regularity in the spatial variable. Loosely speaking, for r>0 and m\in\mathbb{R} , the class C^{r}_{*}S^{m}_{1,0} from Definition 2.3 consists of symbols that behave like an element of S^{m}_{1,0} in the fiber variable, but are elements of the Zygmund space C^{r}_{*}(\mathbb{R}^{n}) from Definition 2.1 in the spatial variable. For r\notin{{\mathbb{N}}} the latter space coincides with C^{r}(\mathbb{R}^{n}) , whereas for r\in{{\mathbb{N}}} it is a strict subset of C^{r-1,1}(\mathbb{R}^{n})\subseteq C^{r}(\mathbb{R}^{n}) . For such a symbol p and for (x,\xi)\in T^{*}\mathbb{R}^{n}\setminus o=\mathbb{R}^{2n}\setminus(\mathbb{R}^{n}\times\{0\}) , we write (1.1) \hat{p}(x,\xi):=\lim_{\lambda\to\infty}\lambda^{-m}p(x,\lambda\xi). Moreover, \Sigma_{\hat{p}}=\hat{p}^{-1}(\{0\}) is the characteristic set of \hat{p} , cf. ( 2.3 ). We refer to Section 2.1 for other basic notions from microlocal analysis which appear in the following theorem, the main result of this article. Theorem 1.1 . Let r>1 , m\in\mathbb{R} and -r<s<r . Let p\in C^{r}_{*}S^{m}_{1,0} be such that \Imag p\in C^{r}_{*}S^{m-1}_{1,0} , and such that \hat{p} is a well-defined element of C^{1}(T^{*}\mathbb{R}^{n}\setminus o) . Set (1.2) \sigma:=\begin{cases}s-r&\text{if }0<s<r,\\ \varepsilon-r&\text{if }-r<s\leq 0,\end{cases} for \varepsilon>0 . Let u\in H^{\sigma+m}(\mathbb{R}^{n}) be such that p(x,D)u\in H^{s}(\mathbb{R}^{n}) . Then \operatorname{WF}^{s+m}u\subseteq\Sigma_{\hat{p}} . Moreover, for each (x,\xi)\in\operatorname{WF}^{s+m-1}u there exists a null bicharacteristic \gamma:\mathbb{R}\to T^{*}\mathbb{R}^{n}\setminus o of \hat{p} satisfying \gamma(0)=(x,\xi) and \gamma(t)\in\operatorname{WF}^{s+m-1}u for all t\in\mathbb{R} . Theorem 1.1 follows from a microlocal statement in Theorem 3.1 (see Remark 3.2 ). The parameter \sigma in ( 1.2 ) measures the a priori regularity of u needed to leverage H^{s}(\mathbb{R}^{n}) regularity of p(x,D)u . Under the assumptions of Theorem 3.1 , the Hamilton vector field H_{\hat{p}} is continuous, and bounded on the cosphere bundle \mathbb{R}^{n}\times S^{n-1} . Moreover, \hat{p} is positively homogeneous of degree m in the fiber variable. Hence, by Peano’s theorem (see e.g. [ 25 , Proposition 1.A.1] ), orbits \gamma:\mathbb{R}\to T^{*}\mathbb{R}^{n}\setminus o of H_{\hat{p}} with specified initial data exist, and every point in the wavefront set lies on a null bicharacteristic which is completely contained in the wavefront set. As such, one might say that the wavefront set is a union of maximally extended null bicharacteristics. However, if \hat{p} is merely continuously differentiable, then orbits of H_{\hat{p}} need not be unique, and Theorem 3.1 does not preclude certain bicharacteristics from entering the wavefront set and then leaving it again. On the other hand, if \hat{p}\in C^{1,1}(T^{*}\mathbb{R}^{n}\setminus o) , then bicharacteristics are unique and one recovers the classical statement on propagation of singularities, albeit under suitable a priori regularity assumptions on u . In fact, C^{2}_{*} regularity suffices for uniqueness (see Remark 3.5 ). Of course, Theorem 1.1 implies a statement on propagation of regularity. Namely, if (x,\xi)\in T^{*}\mathbb{R}^{n}\setminus o is such that every null bicharacteristic \gamma:\mathbb{R}\to T^{*}\mathbb{R}^{n}\setminus o of \hat{p} with \gamma(0)=(x,\xi) has non-empty intersection with the complement of \operatorname{WF}^{s+m-1}u , then (x,\xi)\notin\operatorname{WF}^{s+m-1}u (see Remark 3.6 ). Moreover, as observed in Corollary 3.7 , if there exists a T>0 with the property that, for every such \gamma , there exists a t_{\gamma}\geq T such that \gamma(t_{\gamma})\notin\operatorname{WF}^{s+m-1}u , then \gamma(t)\notin\operatorname{WF}^{s+m-1}u for all such bicharacteristics and all 0\leq t\leq T . Differential operators with coefficients of C^{r}_{*}(\mathbb{R}^{n}) regularity form a concrete class to which Theorem 1.1 applies. In fact, such operators are straightforward to define on domains, and a version of Theorem 1.1 holds for differential operators on domains, Corollary 3.8 . Here only local regularity is required of p and u . We also consider second-order differential operators in divergence form, in Theorem 3.10 . Here the Sobolev interval for s is shifted. We note that, if one considers in all these results differential operators that have slightly more spatial regularity, measured using Sobolev spaces \mathcal{H}^{r,\infty}(\mathbb{R}^{n}) over the space \mathrm{bmo}(\mathbb{R}^{n}) from Definition 2.2 , then the endpoints s=-r and s=r of the Sobolev interval for s can be included, and one may let \varepsilon=0 in ( 1.2 ). For integer r , this applies in particular to operators with C^{r-1,1} regularity (see ( 2.5 )). Note that, for r=2 , our results require slightly less than C^{1,1} regularity. A concrete geometric setting in which this is relevant arises when considering wave equations on manifolds with a metric which has bounded Ricci curvature tensor. Such a metric automatically has \mathcal{H}^{2,\infty}\subseteq C^{2}_{*} regularity, and therefore Theorem 3.10 yields propagation of singularities for these metrics, as is explained in Section 4.2 . 1.2. Previous work In the smooth setting, where p\in S^{m}_{1,0}=\cap_{r>0}C^{r}_{*}S^{m}_{1,0} and \hat{p}\in C^{\infty}(T^{*}\mathbb{R}^{n}\setminus o) , Theorem 1.1 is essentially due to Hörmander (see [ 12 , 11 , 10 ] ). In this case no a priori regularity is required, and the conclusion holds for general u\in\mathcal{S}^{\prime}(\mathbb{R}^{n}) and s\in\mathbb{R} . In [ 2 ] , Bony studied propagation of singularities for nonlinear equations with rough solutions, and in the process developed and employed paradifferential calculus in a manner which has played a key role in subsequent work. Treatises of Bony’s results involving slight variations on the methods can be found in [ 14 ] and [ 22 ] . In particular, Taylor in [ 22 ] highlights the role of C^{r} regularity of the coefficients. In the setting of linear equations with rough coefficients, Taylor in [ 23 ] proved Theorem 1.1 for r=2 , but with \sigma replaced by \sigma+\delta for any \delta>0 . Results for 1<r<2 can also be found in [ 23 ] , but these are of a less concrete nature than Theorem 1.1 . For second-order hyperbolic equations and r=2 , Smith in [ 20 ] obtained the endpoint case \delta=0 , albeit under the assumption of C^{1,1} regularity. In fact, in the setting of [ 20 ] there is a distinguished time variable, and less regularity is required in this distinguished variable than in the remaining ones (see Section 4.1 ). Moreover, [ 20 ] contains an example showing that the conclusion of Theorem 1.1 fails for r=2 if \sigma is replaced by \sigma-\delta for some \delta>0 , at least under the

Read the complete text