Quantitative bounds for Hölder exponents in the Krylov–Safonov and Evans–Krylov theories

Jongmyeong Kim, Se-Chan Lee · 2025

Read Quantitative bounds for Hölder exponents in the Krylov–Safonov and Evans–Krylov theories by Jongmyeong Kim, Se-Chan 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-2512-21025.txt.

Introduction

Jongmyeong Kim and Se-Chan Lee attach hard numbers to two landmark regularity theories for elliptic equations, in a favorable regime. The Krylov-Safonov and Evans-Krylov theories guarantee that solutions are Holder continuous, meaning their roughness is controlled by an exponent, but the classical proofs leave that exponent implicit. When the ellipticity ratio sits close to one, so the equation is nearly isotropic, the authors extract explicit bounds for the exponent.

For the Krylov-Safonov side they adapt the Ishii-Lions method, a coupling method built for viscosity solutions. For Evans-Krylov they use a Schauder-type perturbation, treating the nearly uniform equation as a small disturbance of a well-known one. Concrete exponents matter where the theory feeds numerical analysis or further estimates, since a guarantee without a number cannot be carried into later work.

Catalog record

CREATORJongmyeong Kim, Se-Chan Lee
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE53 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2512-21025

Browse the full library · About the Tower of Babel

Full text

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

Quantitative bounds for Hölder exponents in the Krylov–Safonov and Evans–Krylov theories Thanks: S.-C. Lee is supported by the KIAS Individual Grant (No. MG099001) at the Korea Institute for Advanced Study. Jongmyeong Kim, Se-Chan Lee Abstract . We establish quantitative bounds for Hölder exponents in the Krylov–Safonov and Evans–Krylov theories when the ellipticity ratio is close to one. Our analysis relies on the Ishii–Lions method for the Krylov–Safonov theory and a Schauder-type perturbation argument for the Evans–Krylov theory. 2020 Mathematics Subject Classification 35B65; 35D40; 35J60 1. Introduction The regularity theory for second-order elliptic equations has attracted considerable attention and has been extensively developed in the literature. In the linear uniformly elliptic setting with measurable coefficients, De Giorgi–Nash–Moser [ DG57 , Nas58 , Mos60 , Mos61 ] established the Hölder continuity of weak solutions, providing a positive answer to Hilbert’s nineteenth problem. On the other hand, the interior Hölder regularity of solutions in the non-divergence framework was achieved by Krylov–Safonov [ KS79 , KS80 ] by utilizing the ABP estimate and the Harnack inequality; see also Wang [ Wan92 ] for a related result in the parabolic setting. Since their approaches are essentially based on ellipticity rather than linearity itself, the Krylov–Safonov theory has been widely extended to the regularity theory for fully nonlinear uniformly elliptic equations; we refer to the monograph by Caffarelli–Cabre [ CC95 ] . To illustrate the issue, let 0<\lambda\leq\Lambda be ellipticity constants and let \mathcal{S}^{n} be the space of real n\times n symmetric matrices. We define Pucci’s extremal operators \mathcal{M}^{\pm}_{\lambda,\Lambda} by \mathcal{M}_{\lambda,\Lambda}^{+}(M)=\sup_{\lambda I\leq A\leq\Lambda I}\text{tr}(AM)\quad\text{and}\quad\mathcal{M}^{-}_{\lambda,\Lambda}(M)=\inf_{\lambda I\leq A\leq\Lambda I}\text{tr}(AM). The classical Krylov–Safonov theory can be stated as follows; see Section 2 for the definition of viscosity solutions. Theorem 1.1 ( [ KS79 , KS80 ] ) . If u\in C(B_{1}) is a viscosity solution of (1.1) \mathcal{M}^{+}_{\lambda,\Lambda}(D^{2}u)\geq-K\quad\text{and}\quad\mathcal{M}^{-}_{\lambda,\Lambda}(D^{2}u)\leq K\quad\text{in $B_{1}$} for some constant K\geq 0 , then u\in C^{\alpha}(\overline{B}_{1/2}) for some \alpha=\alpha(n,\lambda,\Lambda)\in(0,1) . It is noteworthy that Theorem 1.1 does not provide quantitative information on the universal Hölder exponent \alpha(n,\lambda,\Lambda) . The upper bound (in some sense) for the Hölder exponent \alpha was captured in a series of papers by Nadirashvili–Vlăduţ [ NV07 , NV08 , NV13 ] . More precisely, they proved that: for every \tau>0 , there exist a dimension n and ellipticity constants \lambda and \Lambda such that there are solutions u of ( 1.1 ) with u\notin C^{\tau} . The counterexamples proposed in their works were constructed under the assumption that the ellipticity ratio is large enough, that is, \Lambda/\lambda\gg 1 . Recently, Lee–Yun [ LY24 ] established the improved C^{1,\alpha} regularity of viscosity solutions of ( 1.1 ) when the ellipticity ratio is sufficiently close to one. Theorem 1.2 ( [ LY24 ] ) . Given \alpha\in(0,1) , there exists \varepsilon=\varepsilon(\alpha,n)>0 such that if |\Lambda/\lambda-1|<\varepsilon , then any viscosity solution u\in C(B_{1}) of ( 1.1 ) for some constant K\geq 0 belongs to C^{1,\alpha}(\overline{B}_{1/2}) . The underlying idea of this result is that the Pucci extremal operators can be viewed as a small perturbation of the Laplacian operator \Delta under the condition \Lambda/\lambda\approx 1 . Nevertheless, since the proof of Theorem 1.2 strongly relies on the contradiction argument, the consequence in [ LY24 ] is still not quantitative in the sense that it does not provide an explicit relation between the Hölder exponent \alpha and the ellipticity gap \Lambda/\lambda-1 . The first goal of this paper is to describe the dependence of the Hölder exponent, which appears in the Krylov–Safonov theory, on the ellipticity ratio in an explicit quantitative manner. Theorem 1.3 (Hölder exponent bound in the Krylov–Safonov theory) . Assume that u\in C(B_{1}) is a viscosity solution of ( 1.1 ) for some constant K\geq 0 . Given \alpha\in(0,1) , if \frac{\Lambda}{\lambda}-1<2\sqrt{\frac{1-\alpha}{n-1}}, then u\in C^{\alpha}(\overline{B}_{1/2}) with the uniform estimate \|u\|_{C^{\alpha}(\overline{B}_{1/2})}\leq C(\|u\|_{L^{\infty}(B_{1})}+K), where C>0 is a constant depending only on n , \lambda , \Lambda and \alpha . A few remarks on Theorem 1.3 are in order. First, the criterion becomes stronger as \alpha\to 1^{-} , which is consistent with the perturbative intuition that one needs \Lambda/\lambda\to 1^{+} for better regularity of solutions. Second, Mooney [ Moo19 ] captured the behavior of the Hölder exponent: \alpha\sim c(n)^{(\Lambda/\lambda)^{n-1}} for some c(n)\in(0,1) . While this asymptotic has an advantage that yields the information for general ellipticity ratio \Lambda/\lambda , it does not guarantee that \alpha\to 1^{-} as \Lambda/\lambda\to 1^{+} . We also refer to Armstrong–Silvestre–Smart [ ASS12 ] , Le [ Le20 ] and Nascimento–Teixeira [ NT23 ] for related results. Moreover, Theorem 1.3 has several simple but useful corollaries regarding fully nonlinear equations (Corollary 3.1 ) and normalized p -Laplace equations (Corollary 3.2 ). Finally, our method can be applied to deal with both elliptic and parabolic equations that hold only where the gradient is large, whose regularity theory was developed by Imbert–Silvestre [ IS16 ] in the elliptic setting. To the best of our knowledge, the Hölder regularity for viscosity solutions of the corresponding parabolic equations is new in this paper; see Theorem 3.4 and Corollary 3.5 . The proof of Theorem 1.3 is based on the Ishii–Lions method [ IL90 ] , which has played a crucial role in developing Hölder or Lipschitz regularity of viscosity solutions in different contexts. For instance, we refer to the works by Capuzzo Dolcetta–Leoni–Porretta [ CDLP10 ] for degenerate equations with superlinear Hamiltonians, Imbert–Silvestre [ IS13a ] for degenerate fully nonlinear equations and Biswas–Topp [ BT25 ] for fractional p -Laplace equations. In our setting, we consider a functional of the form \Phi(x,y)\coloneqq u(x)-u(y)-L_{1}|x-y|^{\alpha}-L_{2}|x|^{2}-L_{2}|y|^{2}, and argue by contradiction. At a maximum point (x,y)\in\overline{B}_{1/2}\times\overline{B}_{1/2} of this functional, we apply the Ishii–Jensen’s lemma [ Ish89 , Jen88 ] to find appropriate semijets (q_{x},X) at x and (q_{y},Y) at y together with a matrix inequality for X and Y . Then we exploit the matrix inequality to obtain the spectral information on X , Y and X-Y in terms of L_{1} , L_{2} and \alpha . The comparison between the Pucci extremal operators and the Laplacian operator finally gives the desired quantitative relation between \Lambda/\lambda-1 and \alpha . The second part of this paper is concerned with the Evans–Krylov theory [ Eva82 , Kry82 , Kry83 ] . We say that a fully nonlinear operator F:\mathcal{S}^{n}\to\mathbb{R} is (i) (\lambda,\Lambda) -elliptic if \mathcal{M}^{-}_{\lambda,\Lambda}(M-N)\leq F(M)-F(N)\leq\mathcal{M}^{+}_{\lambda,\Lambda}(M-N)\quad\text{for any $M,N\in\mathcal{S}^{n}$}; (ii) concave if F(\theta M+(1-\theta)N)\geq\theta F(M)+(1-\theta)F(N)\quad\text{for any $M,N\in\mathcal{S}^{n}$ and $\theta\in[0,1]$}. The classical Evans–Krylov theory reads as follows. The key ingredient of the proof is the application of the weak Harnack inequality to derive the oscillation decay of D^{2}u . Theorem 1.4 ( [ Eva82 , Kry82 , Kry83 ] ) . Let F:\mathcal{S}^{n}\to\mathbb{R} be uniformly (\lambda,\Lambda) -elliptic and concave. If u\in C(B_{1}) is a viscosity solution of F(D^{2}u)=0 in B_{1} , then u\in C^{2,\alpha}(\overline{B}_{1/2}) for some \alpha=\alpha(n,\lambda,\Lambda)\in(0,1) . As in the Krylov–Safonov theory, Theorem 1.4 does not provide a quantitative information on the Hölder exponent \alpha(n,\lambda,\Lambda) . There are various attempts to replace the concavity assumption with a weaker one; we refer to Caffarelli–Yuan [ CY00 ] , Savin [ Sav07 ] and Huang [ Hua19 ] , just to name a few. Recently, Wu–Niu [ WN23 ] obtained an improved regularity of viscosity solutions of F(D^{2}u)=0 when the ellipticity ratio \Lambda/\lambda is sufficiently close

Read the complete text