Elliptic unique continuation below the Lipschitz threshold

Cole Jeznach · 2025

Read Elliptic unique continuation below the Lipschitz threshold by Cole Jeznach (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-2507-23614.txt.

Introduction

Cole Jeznach asks how rough a setting can be before a key math trait fails, in a 2025 paper. The trait is unique continuation: if a solution fades in one zone, it must fade in all. The equations here are elliptic, and their weights can be less smooth than the usual bar.

Jeznach proves the strong form still holds when the roughness obeys a stated integral test, plus mild side terms. Known counterexamples show the test is near the true edge. The edge lands close to a known class called log-Lipschitz. The result pins down, in near-final form, how much smoothness this trait needs to survive.

Catalog record

CREATORCole Jeznach
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE103 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2507-23614

Browse the full library · About the Tower of Babel

Full text

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

Elliptic unique continuation below the Lipschitz threshold Cole Jeznach Abstract . In this article, we investigate unique continuation principles for solutions u u of uniformly elliptic equations of the form − div ( A ∇ u ) = 0 -\divv(A\nabla u)=0 when A A is less regular than Lipschitz. For general matrices A A , we prove that strong unique continuation holds provided A A has modulus of continuity ω \omega satisfying the Osgood condition, ∫ 0 1 ω ​ ( t ) − 1 ​ 𝑑 t = ∞ \int_{0}^{1}\omega(t)^{-1}dt=\infty , plus some other mild hypotheses. Along with the counterexamples of [ Man98 ] , this shows that the sharp condition on A A that guarantees unique continuation is essentially that A A is log-Lipschitz. In the class of isotropic equations (i.e., A ⁡ ( x ) = a ⁡ ( x ) ​ I A(x)=a(x)I for some scalar function a a ) we show that Hölder continuity of a a of the order α ∈ ( 2 / 3 , 1 ) \alpha\in(2/3,1) is sufficient to guarantee strong unique continuation. This latter result contrasts counterexamples known for anisotropic equations, and disproves a conjecture of Miller from 1974 [ Mil74 ] . 2020 Mathematics Subject Classification Primary: 35A02, 35J15. Secondary: 35J10 1. Introduction Questions regarding unique continuation for solutions of linear PDEs are among the most fundamental. A linear partial differential operator \mathcal{L} , defined in a connected domain \Omega\subset\mathbb{R}^{n} , is said to satisfy the weak unique continuation principle if every solution u of \mathcal{L}u=0 in \Omega that vanishes on an open subset of \Omega must vanish identically: u\equiv 0 . If all solutions are L^{2}_{\mathrm{loc}}(\Omega) , then we say that the operator \mathcal{L} satisfies the strong unique continuation principle provided that whenever u vanishes to infinite order (in the L^{2} -sense) at some point x\in\Omega , i.e., (1.1) \displaystyle\lim_{r\rightarrow 0}r^{-2N}\fint_{B_{r}(x)}u^{2}\,dx=0\text{ for all }N>0, then u\equiv 0 . If the condition above fails for large N , then the vanishing order of u at x is defined as the supremum over all N\geq 0 for which the limit in ( 1.1 ) vanishes. Understanding how the size and regularity of the coefficients defining \mathcal{L} impact unique continuation remains a subtle and rich area of investigation, with many fundamental questions still open. Moreover, unique continuation principles (along with the techniques used to develop them) have found widespread applications, including in inverse problems [ KSU07 , Hab15 , CR16 ] , harmonic measure and nodal geometry [ DF88 , Log18a , Log18b , LMNN21 , Tol23 ] , quantitative uniqueness results for PDE [ ESvS03 , ESS03 , KSW15 , LMNN25 ] , and Anderson localization [ BK05 ] . The focus of this article is linear, symmetric, uniformly elliptic operators (1.2) \displaystyle\mathcal{L}u=-\divv(A\nabla u),\,\,\text{ in }B_{1}\subset\mathbb{R}^{n}. Of course when A=I is the identity, solutions of ( 1.2 ) are harmonic, and hence analytic; as such, -\Delta satisfies the strong unique continuation principle. In his pioneering work [ Car39 ] , Carleman proved that strong unique continuation also holds for operators of the form ( 1.2 ) in two dimensions, provided that A\in C^{2}(B_{1}) . The method he introduced via weighted inequalities–now known as Carleman estimates– has become a ubiquitous tool for proving unique continuation results across a wide range of PDEs, well beyond the elliptic setting. Since then, the role of the smoothness condition on A has largely been clarified. When n=2 , strong unique continuation holds even for A which are merely bounded measurable [ BN55 , Sch98 ] . For dimensions n\geq 3 , improvements by [ Aro57 , Agm66 ] , and finally [ GL86 ] showed that Lipschitz regularity of A suffices to guarantee strong unique continuation. Such results are valid even when \mathcal{L} has singular lower-order terms; if \mathcal{L}u=-\divv(A\nabla u)+W\cdot\nabla u+Vu with A Lipschitz, W\in L^{n} , and V\in L^{n/2} then strong unique continuation still holds [ JK85 , Wol92 , KT01 ] . Moreover, these exponent thresholds are essentially sharp. In the other direction, Miller showed in dimensions n\geq 3 [ Mil74 ] that weak unique continuation for ( 1.2 ) may fail if we only assume A is Hölder continuous (see also the earlier example of Plis in non-divergence form [ Pli63 ] ). Later, Mandache improved the counterexample of Miller and showed that there is a matrix A which is Hölder continuous of every exponent \alpha\in(0,1) but for which \mathcal{L} does not satisfy the weak unique continuation principle. In fact, Mandache provided for each modulus of continuity \omega:[0,1]\rightarrow[0,\infty) satisfying (1.3) \displaystyle\int_{0}^{1}\dfrac{1}{\omega(t)}\,dt<\infty, an associated symmetric, uniformly elliptic matrix A with modulus of continuity \omega for which weak unique continuation fails for the operator \mathcal{L} [ Man98 ] . Taking \omega(t)=t\log(1/t)^{p} for p>1 then gives the claimed matrix A . It is worth noting that the examples of both [ Mil74 ] and [ Man98 ] are anisotropic, i.e., A(x) cannot be written as A(x)=a(x)I for any scalar function a . Unlike Carleman’s, the methods of [ GL86 , GL87 ] relied on monotonicity properties of the so-called Almgren frequency function N_{u}^{A}(r) , which for \mathcal{L}=-\Delta takes the form \displaystyle N_{u}^{I}(r)=\dfrac{r^{2}\fint_{B_{r}}\left\lvert\nabla u\right\rvert^{2}\,dx}{\fint_{\partial B_{r}}u^{2}\,d\sigma}. This frequency function, which measures in a scale-invariant way the dominant homogeneity of u in B_{r} , is known to be almost-monotone and bounded in r if A is Lipschitz. Such properties of this frequency function and its relatives have proven to be essential in quantitative unique continuation and free boundary problems. Indeed, bounds on the frequency function allow one to deduce doubling inequalities for the measure u^{2}\,dx and obtain structural information on the infinitesimal behavior of the solution u about the origin. It is impossible to list all such applications, so instead we point the reader to some of the recent results in quantitative unique continuation such as [ CNV15 , NV17a , NV17b , LM18a , LM18b , Log18a , Gal23 , CLMM24 , Fos25 ] , as well as the references [ LM20 , Spo22 , SVG24 ] for further reading. See also the recent results of [ Dav25 ] , where the author reproves elliptic unique continuation results originally obtained via the Carleman method using instead monotonicity properties of the Almgren frequency. The purpose of this article is to study the monotonicity properties of the Almgren frequency (and associated unique continuation principles) when A is less regular than Lipschitz. In the anisotropic setting, we shorten the gap between the counterexamples for unique continuation provided by [ Man98 ] and the sufficiency of the Lipschitz condition used in [ GL86 ] to prove strong unique continuation. Our main result applies to matrices A which admit a modulus of continuity \omega for which ( 1.3 ) fails (such \omega are said to satisfy the Osgood condition). Theorem 1.1 . Let u\not\equiv 0 solve (1.4) \displaystyle-\divv(A\nabla u)=0\text{ in }B_{1}\subset\mathbb{R}^{n}, where A(0)=I , A is symmetric and uniformly elliptic with constant \lambda\in(0,1) , and A has modulus of continuity \omega(t) for which \omega(t) and \phi(t)\coloneqq\omega(t)/t satisfy (1) \int_{0}^{1}\omega(t)^{-1}\,dt=\infty , (2) \int_{0}^{1}\phi(s)\,ds<\infty , (3) \phi(st)\leq C\phi(s)\phi(t) for some C>1 and all 0<s,t\leq C^{-1} . Then the vanishing order of u at 0 does not exceed N , where N is some value

Read the complete text