Abstract
In this paper, we prove that any W2,1 strong solution to second-order non-divergence form elliptic equations is locally \(W^{2,\infty }\) and piecewise C2 when the leading coefficients and data are of piecewise Dini mean oscillation and the lower-order terms are bounded. Somewhat surprisingly here the interfacial boundaries are only required to be in C1,Dini. We also derive global weak-type (1,1) estimates with respect to A1 Muckenhoupt weights. The corresponding results for the adjoint operator are established. Our estimates are independent of the distance between these surfaces of discontinuity of the coefficients
Similar content being viewed by others
References
Acquistapace, P.: On BMO regularity for linear elliptic systems. Annali di Matematica pura ed applicata. (IV) CLXI, 231–269 (1992)
Agmon, S.: The lp approach to the Dirichlet problem. I. Regularity theorems. Ann. Scuola Norm. Sup. Pisa Cl. Sci. 13(3), 405–448 (1959)
Agmon, S., Douglis, A., Nirenberg, L.: Estimates near the boundary for solutions of elliptic partial differential equations satisfying general boundary conditions. I. Comm. Pure Appl. Math. 12, 623–727 (1959)
Agmon, S., Douglis, A., Nirenberg, L.: Estimates near the boundary for solutions of elliptic partial differential equations satisfying general boundary conditions. II. Comm. Pure Appl. Math. 17, 35–92 (1964)
Ammari, H., Bonnetier, E., Triki, F., Vogelius, M.: Elliptic estimates in composite media with smooth inclusions: an integral equation approach. Ann. Sci. c. Norm. Supr. (4) 48(2), 453–495 (2015)
Bauman, P.: Positive solutions of elliptic equations in nondivergence form and their adjoints. Ark. Mat. 22(2), 153–173 (1984)
Brezis, H.: On a conjecture of J. Serrin. Atti Accad. Naz. Lincei Rend. Lincei Mat. Appl. 19(4), 335–338 (2008)
Campanato, S.: Proprietá di hôlderianitá di alcune classi di funzioni. (Italian). Ann. Scuola Norm. Sup. Pisa 17(3), 175–188 (1963)
Dong, H.: Gradient estimates for parabolic and elliptic systems from linear laminates. Arch. Rational Mech. Anal. 205, 119–149 (2012)
Dong, H.: Parabolic equations with variably partially VMO coefficients. Algebra i Analiz. 23(3), 150–174 (2011). translation in St. Petersburg Math. J. 23 (2012), no. 3, 521–539
Dong, H., Escauriaza, L., Kim, S.: On c1, C2, and weak type-(1,1) estimates for linear elliptic operators: part II. Math. Ann. 370 (1-2), 447–489 (2018)
Dong, H., Kim, D.: Higher order elliptic and parabolic systems with variably partially BMO coefficients in regular and irregular domains. J. Funct. Anal. 261, 3279–3327 (2011)
Dong, H., Kim, D.: On lp estimates for elliptic and Parabolic equations with ap weights. Trans. Amer. Math. Soc. 370(7), 5081–5130 (2018)
Dong, H., Kim, S.: On the c1, C2, and weak type-(1,1) estimates for linear elliptic operators. Comm. Partial Differential Equations 42(3), 417–435 (2017)
Dong, H., Lee, J., Kim, S.: On conormal and oblique derivarive problem for elliptic equations with Dini mean oscillation coefficients. to appear in Indiana Univ. Math. J., arXiv:1801.09836v1 [math. AP]
Dong, H., Xu, L.: Gradient estimates for divergence form elliptic systems arising from composite material. SIAM J. Math. Anal. 51(3), 2444–2478 (2019)
Escauriaza, L.: Weak type-(1,1) inequalities and regularity properties of adjoint and normalized adjoint solutions to linear nondivergence form operators with VMO coefficients. Duke Math. J. 74(1), 177–201 (1994)
Fabes, E.B., Kenig, C.E., Serapioni, R.P.: The local regularity of solutions of degenerate elliptic equations. Comm. Partial Differential Equations 7(1), 77–116 (1982)
Escauriaza, L., Montaner, S.: Some remarks on the Lp regularity of second derivatives of solutions to nondivergence elliptic equations and the Dini condition. Atti Accad. Naz. Lincei Cl. Sci. Fis. Mat. Natur. 28, 49–63 (2017)
Fabes, E., Garofalo, N., Marín-Malave, S., Salsa, S.: Fatou theorems for Some nonlinear elliptic equations. Rev. Mat. Iberoamericana 4(2), 227–251 (1988)
Fabes, E., Stroock, D.: The lp- integrability of Green’s functions and fundamental solutions for elliptic and parabolic equations. Duke. Math. J. 51(4), 997–1016 (1984)
Giaquinta, M.: Multiple Integrals in the Calculus of Variations and Nonlinear Elliptic Systems. Princeton University Press, Princeton (1983)
Krylov, N.V.: Lectures on Elliptic and Parabolic Equations in Sobolev Spaces. American Mathematical Society, Providence (2008)
Li, Y., Nirenberg, L.: Estimates for elliptic systems from composite material. Comm. Pure Appl. Math. 56, 892–925 (2003)
Li, Y., Vogelius, M.: Gradient estimates for solutions to divergence form elliptic equations with discontinuous coefficients. Arch. Rational Mech. Anal. 153, 91–151 (2000)
Lieberman, G.: Regularized distance and its applications. Pacific J. Math. 117 (2), 329–352 (1985)
Xiong, J.: C1,1 estimates for elliptic equations with partial and piecewise continuous coefficients. Methods Appl. Anal. 18(4), 373–389 (2011)
Xiong, J., Bao, J.: Sharp regularity for elliptic systems associated with transmission problems. Potential Anal. 39(2), 169–194 (2013)
Acknowledgements
This work was completed while the second author was visiting Brown University. She would like to thank the Division of Applied Mathematics at Brown University for the hospitality and the stimulating environment. The authors would like to thank Prof. Seick Kim and Yanyan Li for helpful discussions.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
H. Dong was partially supported by the NSF under agreement DMS-1600593.
L. Xu was partially supported by the China Scholarship Council (No. 201706040139).
Appendix
Appendix
In the Appendix, we give the \(W_{w}^{2,p}\)-estimate and solvability for the non-divergence form elliptic equation in C1,1 domains with the zero Dirichlet boundary condition. Consider
where λ ≥ 0, \(\mathcal {D}\in C^{1,1}\), and Lu := aijDiju + biDiu + cu. Let \(p\in (1,\infty )\) and w be an Ap weight. Denote
Now we impose the regularity assumptions on aij. Let \(\gamma _{0}=\gamma _{0}(n,p,\delta ,[w]_{A_{p}})\in (0,1)\) be a sufficiently small constant to be specified. There exists a constant r0 ∈ (0, 1) such that aij satisfy (2.4) in the interior of \(\mathcal {D}\) and are VMO near the boundary: for any \(x_{0}\in \partial \mathcal {D}\) and r ∈ (0, r0], we have
In addition, bi and c are bounded by a constant Λ. Then we have the following
Theorem A.1
Let \(p\in (1,\infty )\), w ∈ Ap, and L1 ≤ 0. There exists a sufficient small constant \(\gamma _{0}=\gamma _{0}(n,p,\delta ,[w]_{A_{p}})\in (0,1)\) such that under the above conditions, for any λ ≥ 0 and \(f\in {L_{w}^{p}}(\mathcal {D})\), there exists a unique \(u\in W_{w}^{2,p}(\mathcal {D})\) satisfying (A.1). Furthermore, there exists a constant \(C=C(n,p,\delta ,{\Lambda },\mathcal {D},[w]_{A_{p}},r_{0})\) such that
First we note that when \(\lambda \ge \lambda _{1}(n,p,\delta ,\lambda , \mathcal {D}, [w]_{A_{p}},r_{0})\), the theorem follows from the proofs of [13, Theorems 6.3 and 6.4] combined with the argument in [10, Theorem 2.5] and [23, Sections 8.5]. To deal with the case when λ ∈ [0, λ1), we need the following lemmas. The first one is a local regularity of solutions in weighted Sobolev spaces.
Lemma A.2
Let \(1<p\leq q<\infty \), \(z\in \overline {\mathcal {D}}\). Denote \(\mathcal {D}_{r}:=\mathcal {D}\cap B_{r}(z)\). Then if
we have
Furthermore, there exists a constant \(C=C(R,p,q,n,\delta ,{\Lambda },[w]_{A_{p}},r_{0})\) such that if (A.2) holds, then
Proof
We follow the proof of [23, Theorem 11.2.3] when w = 1. For q = p, (A.3) is obvious and (A.4) is obtained by using the method in the proof of [23, Theorem 9.4.1]. For q > p, we define
where k − 1 is the last j such that p(j) < q. Take λ sufficiently large that λ − L, as an operator acting from \(\mathring {W}_{w}^{2,p(j)}(\mathcal {D})\) onto \(L_{w}^{p(j)}(\mathcal {D})\) for \(j=0,1,\dots ,k\), is invertible. Denote
By weighted Sobolev embedding theorem, see [18, Theorem 1.3], we have
for any \(\zeta \in C_{0}^{\infty }(B_{2R}(z))\). Hence, \(g\in L_{w}^{p(1)}(\mathcal {D})\). By the choice of λ, the equation
has a solution in \(\mathring {W}_{w}^{2,p(1)}(\mathcal {D})\subset \mathring {W}_{w}^{2,p}(\mathcal {D})\) which is unique in \(\mathring {W}_{w}^{2,p}(\mathcal {D})\). We thus obtain that for j = 1,
If p(1) < q, then by repeating this argument with p(1) in place of p, we get (A.5) for j = 2. In this way we conclude (A.3).
Next we prove (A.4). By the choice of λ, for j ≥ 1 and any \(\xi ,\eta \in C_{0}^{\infty }(B_{2R}(z))\) such that η = 1 on the support of ξ, we have
where we used the weighted Sobolev embedding theorem in the last inequality. By iterating the above inequality, we obtain that for any \(\xi \in C_{0}^{\infty }(B_{3R/2}(z))\), there is an \(\eta \in C_{0}^{\infty }(B_{7R/4}(z))\) such that
Finally, recalling the conclusion for the case when p = q, we have
This yields (A.4) and the lemma is proved. □
Next we recall the resolvent operator of L − λI by
Then \(\mathcal {R}_{\lambda }\) is a bounded operator for λ ≥ λ1. The following properties of \(\mathcal {R}_{\lambda }\) for λ large play an important role in proving Lemma A.3 below.
Lemma A.3
Let the coefficients of L be infinitely differentiable, L1 ≤ 0, and λ ≥ λ1. Then
-
(1)
For any bounded f and any γ ∈ (0, 1), we have \(\mathcal {R}_{\lambda }f\in C^{1+\gamma }(\mathcal {D})\), \(\mathcal {R}_{\lambda }f=0\) on \(\partial \mathcal {D}\), and in \(\mathcal {D}\),
$$ |\mathcal{R}_{\lambda}f(x)|\leq\mathcal{R}_{\lambda}|f|(x)\leq\lambda^{-1}\sup_{x\in\mathcal{D}}|f(x)|. $$(A.6) -
(2)
There exists an integer \(m_{0}=m_{0}(n,p,\delta ,{\Lambda },\mathcal {D},[w]_{A_{p}},r_{0})\), such that for any \(f\in {L_{w}^{p}}(\mathcal {D})\), we have
$$ \sup_{x\in\mathcal{D}}|\mathcal{R}_{\lambda_{1}}^{m_{0}}f(x)|\leq C\|f\|_{{L_{w}^{p}}(\mathcal{D})}, $$(A.7)where \(C=C(n,p,\delta ,{\Lambda },\mathcal {D},[w]_{A_{p}},r_{0})>0\).
Proof
For \(f\in L^{\infty }(\mathcal {D})\), (A.6) is proved in [23, Theorem 11.2.1(3)]. To prove (A.7), we set α = n/(n − 1), p(j) = αjp, and
Notice that for j ≥ 1, we have
Therefore, by using the solvability and estimates for λ large, we have \(u_{j+1}\in \mathring {W}_{w}^{2,p}(\mathcal {D})\) and
By using Lemma A.2, we get
Hence,
By the weighted embedding theorem, we have
Iterating the above inequality yields that for j ≥ 0, we obtain
It follow from Hölder’s inequality and the definition of Ap weights that \(u_{j+1}\in W^{2,p(j)/p}(\mathcal {D})\). Then we fix a j = j(n, p) by choosing p(j) > np/2. For such j, we conclude from (A.8) and (A.9) that
which shows that (A.7) holds with m0 = j + 1. The lemma is proved. □
Next we show the solvability when λ ≥ 𝜖0 for a positive constant 𝜖0 > 0.
Lemma A.4
Let \(p\in (1,\infty )\), w ∈ Ap, 𝜖0 > 0, and L1 ≤ 0. Under the above conditions, for any λ ≥ 𝜖0 and \(u\in \mathring {W}_{w}^{2,p}(\mathcal {D})\), we have
where C > 0 depends on \(n,p,\delta ,{\Lambda },\mathcal {D},\epsilon _{0},[w]_{A_{p}}\), and r0.
Proof
We define f := λu − Lu. As noted after Theorem A.1, it suffices to prove the case when 𝜖0 ≤ λ < λ1. We follow the idea in [23, Section 11.3]. Here we list the main differences. By approximations, we may assume that the coefficients are smooth. Similar to the proof of [23, Theorem 8.5.6], we have
Therefore, it suffices to prove for ε0 ≤ λ < λ1, we have
Since
by induction on m, we have
where m ≥ 1 is any integer. We next introduce constants C1 and Mm such that
By using (A.11), 0 < λ1 − λ ≤ λ1 − 𝜖0, (A.12), (A.6), and (A.7), we obtain for m > m0,
Fixing m > m0 such that
we get (A.10). The lemma is proved. □
We now give
Proof of Theorem A.1.
By using the method of continuity, we only need to prove that for any \(u\in \mathring {W}_{w}^{2,p}(\mathcal {D})\) and λ ≥ 0, we have (A.10). To this end, without loss of generality we may assume that \(\mathcal {D}\subset B_{2R_{0}}\), where \(R_{0}=\text {diam}~{\mathcal {D}}\). We take the global barrier v0 from [23, Lemma 11.1.2]:
satisfies Lv0 ≤− 1, and v0 > 0 in \(B_{4R_{0}}\) and v0 = 0 on \(\partial B_{4R_{0}}\), where c0 > 0 is a constant to be chosen. Next we introduce a new operator \(L^{\prime }\) by
Notice that in \(\mathcal {D}\subset B_{2R_{0}}\), according to the construction of v0, we have \(L^{\prime }1\leq -\delta ^{\prime }\) for a constant \(\delta ^{\prime }>0\) depending on n, δ,Λ, and R0, provided that c0 = c0(d, δ,Λ) is sufficiently large. By using Lemma A.3 applied to \(L^{\prime \prime }:=L^{\prime }+\delta ^{\prime }\), we get
Hence, we finish the proof of the theorem. □
Finally we give the solvability of the adjoint operator of L defined by
By using a similar argument in the proof of Lemma 2.8, from Theorem A.1, we have
Corollary A.5
Let \(p\in (1,\infty )\), w ∈ Ap, and L1 ≤ 0. Assume that \(g=(g^{ij})_{i,j=1}^{n}\in {L_{w}^{p}}(\mathcal {D})\). The coefficients aij, bi, and c satisfy the same conditions as imposed in Theorem A.1. Then for any λ ≥ 0,
admits a unique adjoint solution \(u\in {L_{w}^{p}}(\mathcal {D})\). Moreover, the following estimate holds
where \(C=C(n,p,\delta ,{\Lambda },\mathcal {D},r_{0},[w]_{A_{p}})>0\).
Rights and permissions
About this article
Cite this article
Dong, H., Xu, L. Hessian Estimates for Non-divergence form Elliptic Equations Arising from Composite Materials. Potential Anal 54, 409–449 (2021). https://doi.org/10.1007/s11118-020-09832-w
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11118-020-09832-w
Keywords
- Non-divergence form elliptic equations
- Adjoint solutions
- Piecewise DMO coefficients
- Hessian estimates
- Weak-type (1,1) estimates