Abstract
We study the map** properties of boundary integral operators arising when solving two-dimensional, time-harmonic waves scattered by periodic domains. For domains assumed to be at least Lipschitz regular, we propose a novel explicit representation of Sobolev spaces for quasi-periodic functions that allows for an analysis analogous to that of Helmholtz scattering by bounded objects. Except for Rayleigh-Wood frequencies, continuity and coercivity results are derived to prove wellposedness of the associated first kind boundary integral equations.
Similar content being viewed by others
References
Alber, H.D.: A quasi-periodic boundary value problem for the Laplacian and the continuation of its resolvent. Proc. R. Soc. Edinb. 82(3–4), 251–272 (1979)
Ammari, H.: Scattering of waves by thin periodic layers at high frequencies using the on-surface radiation condition method. IMA J. Appl. Math. 60(2), 199–214 (1998)
Ammari, H., Bao, G.: Coupling of finite element and boundary element methods for the scattering by periodic chiral structures. J. Comput. Math. 26(3), 261–283 (2008)
Ammari, H., He, S.: Homogenization and scattering for gratings. J. Electromagn. Waves Appl. 11(12), 1669–1683 (1997)
Ammari, H., Nédélec, J.-C.: Analysis of the diffraction from chiral gratings. In: Mathematical Modeling in Optical Science. SIAM, pp. 179–206 (2001)
Bao, G.: Finite element approximation of time harmonic waves in periodic structures. SIAM J. Numer. Anal. 32(4), 1155–1169 (1995)
Bao, G.: Variational approximation of Maxwell’s equations in biperiodic structures. SIAM J. Appl. Math. 57(2), 364–381 (1997)
Bao, G.: Recent mathematical studies in the modeling of optics and electromagnetics. J. Comput. Math. 22(2), 148–155 (2004)
Bao, G., Dobson, D.C.: On the scattering by a biperiodic structure. Proc. Am. Math. Soc. 128(9), 2715–2723 (2000)
Bao, G., Dobson, D.C., Cox, J.A.: Mathematical studies in rigorous grating theory. J. Opt. Soc. Am. A 12(5), 1029–1042 (1995)
Barnett, A., Greengard, L.: A new integral representation for quasi-periodic scattering problems in two dimensions. BIT Numer. Math. 51(1), 67–90 (2011)
Bruno, O.P., Fernandez-Lado, A.G.: Rapidly convergent quasi-periodic Green functions for scattering by arrays of cylinders–including Wood anomalies. Proc. R. Soc. A 473(2199), 20160802 (2017)
Bruno, O.P., Shipman, S.P., Turc, C., Stephanos, V.: Three-dimensional quasi-periodic shifted Green function throughout the spectrum, including Wood anomalies. Proc. R. Soc. A 473(2207), 20170242 (2017)
Chen, X., Friedman, A.: Maxwell’s equations in a periodic structure. Trans. Am. Math. Soc. 323(2), 465–507 (1991)
Cho, M.H., Barnett, A.H.: Robust fast direct integral equation solver for quasi-periodic scattering problems with a large number of layers. Opt. Express 23(2), 1775–1799 (2015)
Costabel, M.: Boundary integral operators on Lipschitz domains: Elementary results. SIAM J. Math. Anal. 19(3), 613–626 (1988)
Dobson, D.C.: A variational method for electromagnetic diffraction in biperiodic structures. ESAIM: Math. Model. Numer. Anal. 28(4), 419–439 (1994)
Dobson, D.C., Cox, J.A.: An integral equation method for biperiodic diffraction structures. Int. Conf. Appl. Theory Period. Struct. 1545, 106–114 (1991)
Dobson, D.C., Friedman, A.: The time-harmonic Maxwell equations in a doubly periodic structure. J. Math. Anal. Appl. 166(2), 507–528 (1992)
Elschner, J., Schmidt, G.: Diffraction in periodic structures and optimal design of binary gratings. Part I: direct problems and gradient formulas. Math. Methods Appl. Sci. 21(14), 1297–1342 (1998)
Jerez-Hanckes, C.: Modeling elastic and electromagnetic surface waves in piezoelectric tranducers and optical waveguides. Ph.D. thesis, École Polytechnique, Palaiseau, France (2008)
Kirsch, A.: Diffraction by periodic structures. In: Inverse Problems in Mathematical Physics. Springer, Berlin, pp. 87–102 (1993)
Kirsch, A.: Uniqueness theorems in inverse scattering theory for periodic structures. Inverse Prob. 10, 145–152 (1994)
Kress, R.: Linear Integral Equations, vol. 82, 3rd edn. Applied Mathematical Sciences (2014)
Lai, J., Kobayashi, M., Barnett, A.: A fast and robust solver for the scattering from a layered periodic structure containing multi-particle inclusions. J. Comput. Phys. 298, 194–208 (2015)
Lechleiter, A., Nguyen, D.-L.: Volume integral equations for scattering from anisotropic diffraction gratings. Math. Methods Appl. Sci. 36(3), 262–274 (2013)
Lechleiter, A., Zhang, R.: A floquet-bloch transform based numerical method for scattering from locally perturbed periodic surfaces. SIAM J. Sci. Comput. 39(5), B819–B839 (2017)
Linton, C.M.: The green’s function for the two-dimensional Helmholtz equation in periodic domains. J. Eng. Math. 33(4), 377–401 (1998)
Liu, Y., Barnett, A.: Efficient numerical solution of acoustic scattering from doubly-periodic arrays of axisymmetric objects. J. Comput. Phys. 324, 226–245 (2016)
McLean, W.C.H.: Strongly Elliptic Systems and Boundary Integral Equations. Cambridge University Press, Cambridge (2000)
Nédélec, J.C., Starling, F.: Integral equation methods in a quasi-periodic diffraction problem for the time-harmonic Maxwell’s equations. SIAM J. Math. Anal. 22(6), 1679–1701 (1991)
Nečas, J.: Direct Methods in the Theory of Elliptic Equations. Springer, Berlin (2011)
Pestourie, R., Pérez-Arancibia, C., Lin, Z., Shin, W., Capasso, F., Johnson, S.G.: Inverse design of large-area metasurfaces. Opt. Express 26(26), 33732–33747 (2018)
Petit, R. (ed.): Electromagnetic Theory of Gratings. Springer, Berlin (1980)
Saranen, J., Vainikko, G.: Periodic Integral and Pseudodifferential Equations with Numerical Approximation. Springer, Berlin (2002)
Schmidt, G.: Boundary integral methods for periodic scattering problems. In: Around the Research of Vladimir Maz’ya II. Springer, Berlin, pp. 337–363 (2010)
Schmidt, G.: Integral equations for conical diffraction by coated grating. J. Integral Equ. Appl. 23(1), 71–112 (2011)
Shiraishi, K., Higuchi, S., Muraki, K., Yoda, H.: Silver-film subwavelength gratings for polarizers in the terahertz and mid-infrared regions. Opt. Express 24(18), 20177–20186 (2016)
Silva-Oelker, G., Aylwin, R., Jerez-Hanckes, C., Fay, P.: Quantifying the impact of random surface perturbations on reflective gratings. IEEE Trans. Antennas Propag. 66(2), 838–847 (2018)
Silva-Oelker, G., Jerez-Hanckes, C., Fay, P.: Study of W/HfO\(_2\) grating selective thermal emitters for thermophotovoltaic applications. Opt. Express 26(22), A929–A936 (2018)
Starling, F., Bonnet-Bendhia, A.-S.: Guided waves by electromagnetic gratings and non-uniqueness examples for the diffraction problem. Math. Methods Appl. Sci. 17, 305–338 (1994)
Steinbach, O.: Numerical Approximation Methods for Elliptic Boundary Value Problems. Springer, Berlin (2007)
Tartar, L.: An Introduction to Sobolev Spaces and Interpolation Spaces, vol. 3. Springer, Berlin (2007)
Zhang, B., Chandler-Wilde, S.N.: A uniqueness result for scattering by infinite rough surfaces. SIAM J. Appl. Math. 58(6), 1774–1790 (1998)
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.
This work was partially funded by Fondecyt Regular 1171491 and by grants Conicyt-PFCHA/Doctorado Nacional/2017-21171791 and 2017-21171479.
Appendices
Appendix A: Proof of Theorem 4.4
In order to prove the boundedness of the quasi-periodic Newton potential, we shall make use of the following lemma.
Lemma A.1
Let \(g\in {\mathcal {D}}({{\mathbb {R}}})\), and \(\xi \in {{\mathbb {R}}},\ \left| \xi \right| >0\). Then, for some \(C>0\), it holds that
Proof
Let \(R>0\) be so that the support of g(x) is contained in \([-R,R]\). Then,
Also, the second inequality follows by integration-by-parts:
\(\square \)
We now prove Theorem 4.4 by adapting a strategy similar to that used in [42, Theorem 6.1].
Proof of Theorem 4.4
First, consider \(f\in {\mathcal {D}}_\theta ({\mathcal {G}})\) for which the expression
holds. Since f has compact support in the \(x_2\)-direction, there exists some positive \(r \in {{\mathbb {R}}}\) such that \(f_j(x_2) = 0\) if \(\left| x_2\right| >r\), for all \(j \in {{\mathbb {Z}}}\). Fix \(R>0\) and set \(u:={\mathcal {N}}^k_\theta f\). Then, u is a quasi-periodic function on \({\mathcal {G}}\) by the quasi-periodicity of the Green’s function. Consider \(\mu \in {\mathcal {D}}({{\mathbb {R}}})\) such that \(\mu (t)=1\), for all \(t\in [0,r+R]\). We define a modified version of u as
Notice that for \(\varvec{x}\in {\mathcal {G}}^R := {\mathcal {G}}\cap \left\{ \left| x_2\right| < R\right\} \), \(u_\mu (\varvec{x}) = u(\varvec{x})\). Hence, \(u_\mu \) is an extension of u and, from the norm definition for \(H^s_\theta \left( {\mathcal {G}}^R\right) \), we find that \(\left\Vert u\right\Vert _{H^s_\theta \left( {\mathcal {G}}^R\right) }\le \left\Vert u_\mu \right\Vert _{H^s_\theta ({\mathcal {G}})}\). We now prove the boundedness of \(\left\Vert u_\mu \right\Vert _{H^s_\theta ({\mathcal {G}})}\). Since \(u_\mu \) is also \(\theta \)-quasi-periodic, it holds
Since \(\mu \) and f have compact support, we can exchange the integration order so as to write
and where we used the periodicity of \(e^{-\imath j_\theta z_1}G^k_\theta (\varvec{z},0)\). Then, replacing \(G^k_{\theta }\) by its expansion (Proposition 4.2) yields
Observe that
and consider \(j_\theta \) such that \(\beta _j\in {{\mathbb {R}}}\), i.e. \(j_\theta ^2<k^2\). From Lemma A.1, we get
Furthermore, since \(\beta _j\) is real for a finite number of j, depending only on k and \(\theta \), then for all \(j\in {{\mathbb {Z}}}\) such that \(j_\theta <k^2\), yields
Now, let us take \(j_\theta ^2>k^2\) so that \(\beta _j\) is imaginary and \(e^{\imath \beta _j\left| z_2\right| }\) decays as \(\left| z_2\right| \) increases. Since
integration-by-parts gives
By Lemma A.1, we deduce that
is bounded for all \(\xi \in {{\mathbb {R}}}\), \(j\in {{\mathbb {Z}}}\). Hence,
where C depends only on k and \(\mu \). Thus, there exists \(C>0\) depending only on \(k,k_1,\) and \(\mu \) such that for all \(s\in {{\mathbb {R}}}\),
Taking the squared \(L^2\)-norm of both sides of (34) and adding over \(j\in {{\mathbb {Z}}}\), we obtain
Since \({\mathcal {D}}_\theta ({\mathcal {G}})\) is dense in \(H^{s-2}_\theta ({\mathcal {G}})\) (cf. Proposition 2.8), the result is proven. \(\square \)
Appendix B: Regularity of Solutions and Continuity of BIOs
We extend the main results in [30, Chapter 4], introduced by Nečas [32], to the periodic case. We highlight changes needed to replicate the arguments. Our starting point is the result presented in Sect. 3. Recall that \(\theta \in [0,1)\), \(\Gamma \) a periodic curve in \({\mathcal {G}}:=[0,2\pi ]\times {{\mathbb {R}}}\) and \(\Omega \) as the open domain above \(\Gamma \) (see Fig. 2).
Lemma B.1
(Lemma 2.3 in [31], and 3.2 in [41]) Let \(u\in H^1_\theta (\Omega ^H)\) be such that
with \({\mathcal {T}}(k_1,k)\) being the DtN operator from Definition 3.6. Then, the Fourier coefficients \(u_{j}=0\) for all j in \(J^{-}_{k_1}:=\lbrace j\in {{\mathbb {Z}}}\ |\ k^2>j_{\theta }^2\rbrace \).
Proof
We proceed as in [31, Lemma 2.3],
The integral over \(\Gamma \) vanishes due to either condition: \(\gamma ^{i}_{0} u=0\) or \(\gamma ^{i}_{1} u=0\). Hence, we only need to consider the integration on \(\Gamma ^H\),
Recall the Fourier series for u and the DtN operator,
Hence,
and (35) becomes,
For \(j\not \in J^{-}_{k_1}\), we have \(\imath \beta _j\in {{\mathbb {R}}}\) and \(\overline{\imath \beta _j}-\imath \beta _j=0\). Thus,
Since \(\beta _j>0\) for all \(j\in J^{-}_{k_1}\), \(\left| u_j(H)\right| =0\), for all \(j\in J^{-}_{k_1}.\)\(\square \)
Proposition B.2
Let \(k >0\) and \(f \in {\widetilde{H}}^{-1}_{\theta }(\Omega )\) with compact support.
- (i)
Let \(g \in H^{\frac{1}{2}}_\theta (\Gamma )\) and \(k \notin K^{(\text {TM})}_{sing}\). Then, there is a unique \(u \in H^1_{\theta , {\mathrm {loc}}}(\Omega )\) that satisfies
$$\begin{aligned} \left\{ \begin{array}{l} (-\Delta -k^2)u(\varvec{x}) = f(\varvec{x})\ \text {on}\ \Omega , \\ \gamma ^i_0 u = g\ \text {on}\ \Gamma ,\\ u\text { satisfies radiation conditions at infinity}. \end{array}\right. \end{aligned}$$Moreover, the solution depends continuously on the data
$$\begin{aligned} \left\Vert u\right\Vert _{H^1_\theta (\Omega ^R)} \lesssim \left\Vert f\right\Vert _{{\widetilde{H}}^{-1}_{\theta }(\Omega )} +\left\Vert g\right\Vert _{H^{\frac{1}{2}}_\theta (\Gamma )}. \end{aligned}$$ - (ii)
Let \(w \in H^{-\frac{1}{2}}_\theta (\Gamma )\) and \(k \notin K^{(\text {TE})}_{sing}\). Then, there is a unique \(u\in H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) that satisfies
$$\begin{aligned} \left\{ \begin{array}{l} (-\Delta -k^2)u(\varvec{x}) = f(\varvec{x})\ \text {on}\ \Omega , \\ \gamma ^i_1 u = w\ \text {on}\ \Gamma ,\\ u\text { satisfies radiation conditions at infinity}. \end{array}\right. \end{aligned}$$Also, it holds
$$\begin{aligned} \left\Vert u\right\Vert _{H^1_\theta (\Omega ^R)} \lesssim \left\Vert f\right\Vert _{{\widetilde{H}}^{-1}_{\theta }(\Omega )} + \left\Vert w\right\Vert _{H^{-\frac{1}{2}}_\theta (\Gamma )}. \end{aligned}$$ - (iii)
Let \(g \in H^{\frac{1}{2}}_\theta (\Gamma )\) and \(k \notin K^{(\text {TM})}_{sing}\). Then, there is a unique \(u \in H^1_{\theta , {\mathrm {loc}}}(\Omega )\) that satisfies
$$\begin{aligned} \left\{ \begin{array}{l} (-\Delta -k^2)u(\varvec{x}) = f(\varvec{x})\ \text {on}\ \Omega , \\ \gamma ^i_0 u = g\ \text {on}\ \Gamma ,\\ u\text { satisfies the adjoint radiation condition at infinity}. \end{array}\right. \end{aligned}$$The next bound holds
$$\begin{aligned} \left\Vert u\right\Vert _{H^1_\theta (\Omega ^R)} \lesssim \left\Vert f\right\Vert _{{\widetilde{H}}^{-1}_{\theta }(\Omega )} +\left\Vert g\right\Vert _{H^{\frac{1}{2}}_\theta (\Gamma )}. \end{aligned}$$ - (iv)
Let \(w \in H^{-\frac{1}{2}}_\theta (\Gamma )\). If \(k \notin K^{(\text {TE})}_{sing}\), there is a unique \(u\in H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) that satisfies
$$\begin{aligned} \left\{ \begin{array}{l} (-\Delta -k^2)u(\varvec{x}) = f(\varvec{x})\ \text {on}\ \Omega , \\ \gamma ^i_1 u = w\ \text {on}\ \Gamma ,\\ u\text { satisfies the adjoint radiation condition at infinity}. \end{array}\right. \end{aligned}$$Moreover, the solution is bounded by the data
$$\begin{aligned} \left\Vert u\right\Vert _{H^1_\theta (\Omega ^R)} \lesssim \left\Vert f\right\Vert _{{\widetilde{H}}^{-1}_{\theta }(\Omega )} +\left\Vert w\right\Vert _{H^{-\frac{1}{2}}_\theta (\Gamma )}. \end{aligned}$$
Proof
For the standard radiation condition (Definition 3.2), items (i) and (ii) follow from the Fredholm alternative and Theorem 3.11 (see [41, Theorems 3.3 and 3.4]). The same strategy holds if the adjoint radiation condition (see Definition 3.5) is used: we just need to show that the equations in items (iii) and (iv) have the same eigenvalues as those in (i) and (ii), which follows from noticing that one can build solutions of the equations with one radiation condition from the other. \(\square \)
The last proposition motivates the definition of solution operators. We consider two different cases. Let \(k \notin K^{(\text {TM})}_{sing}\) and \(g \in H^{\frac{1}{2}}_\theta (\Gamma )\), we set
where u is the only element in \(H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) that satisfies
The corresponding adjoint version is
where v is the only element in \(H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) that satisfies
We also consider Steklov-Poincaré operators defined as
For a given domain \({\mathcal {O}}\subset {\mathcal {G}}\), \(k >0\) and a pair of functions \(u,v \in H^1_\theta ({\mathcal {O}})\), we define the following sesquilinear form:
Lemma B.3
For \(g_1\), \(g_2 \in H^{\frac{1}{2}}_\theta (\Gamma ) \) we have that
Proof
From the radiation conditions, there is an \(R>0\) such that for \(x_2 \ge R\), it holds
with \(\beta _j\) and \({\widetilde{\beta }}_j\) as in (20) and (23), respectively. Using Lemma 2.33 and the definitions of \({\mathcal {U}}_k\) and \({\mathcal {V}}_k\) leads to
with \(\Phi ^k_{\Omega ^R}\) as in (36) and \(\Gamma ^R:=\{\varvec{x}\in {\mathcal {G}}\ |\ x_2=R\}\). Subtracting these last equations, we get
In \(\Gamma ^R\) we can use the expansions given by the radiation conditions:
Then, for j such that \(\beta _j \) is a real number we have that \({\widetilde{\beta }}_j = - \beta _j\). Hence, \((\imath \overline{{\widetilde{\beta }}}_j+\imath \beta _j) = \imath (-\beta _j + \beta _j) = 0\). On the other hand, if \(\beta _j\) is pure imaginary we have that \({\widetilde{\beta }}_j = \beta _j\) and \((\imath \overline{{\widetilde{\beta }}}_j+\imath \beta _j) = \imath (-\beta _j + \beta _j) = 0\). Thus, the duality products over \(\Gamma ^R\) cancel each other out, yielding
\(\square \)
Following [30, Chapter 4], we now focus on establishing regularity properties of solutions in \(\Omega \). If u is a \(\theta \)-quasi-periodic function defined in \(\Omega \), we denote by \(u^p\) its \(\theta \)-quasi-periodic extension. For \(h\in {{\mathbb {R}}}\) with \(\left| h\right| < \pi \), we define the following estimators for the partial derivatives
The properties of \(\Delta _h^2\) are established in [30, Lemmas 4.13 to 4.15], where \(L_2({{\mathbb {R}}}^d)^m\) and \({\mathcal {D}}({{\mathbb {R}}}^d)^m\) have to be replaced by \(L^2_\theta ({\mathcal {G}})\), and \({\mathcal {D}}_\theta ({\mathcal {G}})\), respectively. There are, however, slight differences in the proofs for \(\Delta _h^1\), which are exposed when proving Lemmas B.4 and B.5.
Lemma B.4
(Lemma 4.13 in [30]) For \(\theta \in [0,1)\), let u be a \(\theta \)-quasi-periodic function. Then, for \(i = 1,2\), it holds
- (a)
If \(\partial _i u \in L^2_\theta ({\mathcal {G}})\), then \(\left\Vert \Delta ^i_h u\right\Vert _{L^2_\theta ({\mathcal {G}})} \le \left\Vert \partial _i u\right\Vert _{L^2_\theta ({\mathcal {G}})}\) and \(\left\Vert \Delta ^i_h u - \partial _i u\right\Vert _{L^2_\theta ({\mathcal {G}})} \rightarrow 0\) as \(h \rightarrow 0\).
- (b)
If there is a constant M such that \(\left\Vert \Delta ^i_h u\right\Vert _{L^2_\theta ({\mathcal {G}})} \le M\), then, for h small, we have that \(\partial _i u \in L^2_\theta ({\mathcal {G}})\), and \(\left\Vert \partial _i u\right\Vert _{L^2_\theta ({\mathcal {G}})} \le M\).
Proof
The proof for (b) follows Lemma 4.13 in [30]. Similarly for (a) for \(i = 2\) whereas for \(i=1\), we observe that
Integrating over \({\mathcal {G}}\) yields,
where (37) follows from the periodicity of \(\left| \partial _1u(\varvec{y})\right| \). \(\square \)
Lemma B.5
(Lemma 4.15 in [30]) Let u and v belong to \(L^2_\theta ({\mathcal {G}})\), \(h \in {{\mathbb {R}}}\) such that \(\left| h\right| < \pi \). Moreover, let \(k >0 \) and \({\mathcal {O}}\subset {\mathcal {G}}\) be an open bounded set whose boundary is given by two disjoint periodic curves. Assume further that \({\text {supp}}u \subset {\mathcal {O}}\cap ({\mathcal {O}}-h\varvec{e}_2)\) and \({\text {supp}}v \subset {\mathcal {O}}\cap ({\mathcal {O}}+h\varvec{e}_2)\). Then,
- (a)
if \(u,v \in L^2_\theta ({\mathcal {O}})\), then \(\left( \Delta ^i_h u, v \right) _{L^2_\theta ({\mathcal {O}})} = - \left( u, \Delta ^i_{-h} v\right) _{L^2_\theta ({\mathcal {O}})}\), \(i=1,2\).
- (b)
if \(u,v \in H^1_\theta ({\mathcal {O}})\), then \(\Phi ^k_{{\mathcal {O}}}\left( \Delta ^i_h u,v\right) = -\Phi ^k_{{\mathcal {O}}}\left( u,\Delta ^i_{-h}v\right) \), \(i=1,2\).
Proof
For \(i=2\) the result follows verbatim from [30] whereas for \(i=1\), this is deduced directly from the definition of \(\Delta ^1_h\) and the quasi-periodicity property. \(\square \)
Theorem B.6
(Thm. 4.16 in [30]) Let \({\mathcal {O}}\subset \Omega \) be a bounded open set, whose boundary is given by two periodic curves and such that \({\overline{{\mathcal {O}}}}^{\mathcal {G}}\subset \Omega \). For \(r \ge 0\) and \(k>0\), let \(f \in H^r_\theta (\Omega )\) and \(u \in H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) be such that
Then, \(u \in H^{r+2}_\theta ({\mathcal {O}})\) and for any \(R>0\) such that \({\overline{{\mathcal {O}}}}^{\mathcal {G}}\subset \Omega ^R\), we have that
Proof
We take similar steps to those in the proof of Theorem 4.16 in [30]. Set \(r=0\) and consider a function \(\chi \in {\mathcal {D}}_\theta (\Omega ^R)\) such that \(\chi = 1\) in \({\mathcal {O}}\). Define
By direct computation, we obtain that \(\left\Vert f_1\right\Vert _{L^2_\theta (\Omega ^R)} \lesssim \left\Vert u\right\Vert _{H^1_\theta (\Omega ^R)}+\left\Vert f\right\Vert _{L^2_\theta (\Omega ^R)}\), so \(f_1 \in L^2_\theta (\Omega ^R)\). Let \(v \in H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) with null trace in \(\partial ^{\mathcal {G}}\Omega ^R\). Using (11), we have that
Also, by Lemma B.5, we have that for \(i=1,2\), if \(\overline{{\text {supp}}v}^{{\mathcal {G}}}\subset \Omega ^R\) and h is sufficiently small, it holds
Hence,
By Lemma B.4 and norm definitions, we have that
On the other hand, by the coercivity of the Helmholtz operator, we get
Taking \(v =\Delta ^i_h (\chi u)\) in (38) leads to
Here, we use the inequality \(ab \le \frac{1}{2}(\epsilon a +\epsilon ^{-1}b^2)\) for a small \(\epsilon \) to obtain
Again, by Lemma B.4, \(\left\Vert \Delta ^i_h (\chi u)\right\Vert _{L^2_\theta ({\Omega ^R}) }\lesssim \left\Vert u\right\Vert _{H^1_\theta ({\Omega ^R})}\) and, by the bound for the norm of \(f_1\), we retrieve
Finally, by recalling the norm definition on a subset \({\mathcal {O}}\subset \Omega ^R\) and Lemma B.4, it holds
The proof is then achieved by induction, analogously to that of Theorem 4.16 in [30]. \(\square \)
Now, we establish regularity results up to the boundary.
Theorem B.7
(Thm. 4.18 [30]) Assume \(\Omega \) to be a \({\mathcal {C}}^{r-1,1}\)-domain, with \(r\ge 2\). Let \({\mathcal {O}}\subset \Omega \) be a bounded subset whose boundary is composed of two periodic curves, one of them being \(\Gamma =\partial ^{{\mathcal {G}}}\Omega \). Moreover, let the wavenumber \(k>0\), \(f \in H^{r-2}_\theta (\Omega )\) and \(u \in H^1_{\theta , {\mathrm {loc}}}(\Omega )\) be such that
Then, the following bounds hold
- (i)
If \(\gamma ^i_0u \in H_\theta ^{r-\frac{1}{2}}(\Gamma )\), then \(u \in H_\theta ^{r}({\mathcal {O}})\) and
$$\begin{aligned} \left\Vert u\right\Vert _{H_\theta ^{r}({\mathcal {O}})} \lesssim \left\Vert u\right\Vert _{H_\theta ^{1}(\Omega ^R)}+\left\Vert \gamma ^i_0u\right\Vert _{H_\theta ^{r-\frac{1}{2}}(\Gamma )}+\left\Vert f\right\Vert _{H^{r-2}_\theta (\Omega ^R)}. \end{aligned}$$ - (ii)
If \(\gamma _1^iu \in H_\theta ^{r-\frac{3}{2}}(\Gamma )\), then \(u \in H_\theta ^{r}({\mathcal {O}})\) and
$$\begin{aligned} \left\Vert u\right\Vert _{H_\theta ^{r}({\mathcal {O}})} \lesssim \left\Vert u\right\Vert _{H_\theta ^{1}(\Omega ^R)}+\left\Vert \gamma ^i_1u\right\Vert _{H_\theta ^{r-\frac{3}{2}}(\Gamma )}+\left\Vert f\right\Vert _{H^{r-2}_\theta (\Omega ^R)}. \end{aligned}$$
for all \(R>0\) such that \({\mathcal {O}}\subset \Omega ^R\).
Proof
We bound the derivative \(\partial _1 u\) as in Theorem B.6 while bounds for \(\partial _2 u\) may be obtained from the boundary value problem:
The remainder of the proof follows that of [30, Theorem 4.18], requiring only minor modifications to the periodic setting. \(\square \)
Corollary B.8
(Thm. 4.21 [30]) Assume that \(\Omega \) is a \({\mathcal {C}}^{r-1,1}\)-domain and \(r\ge 2\). Then, for \(k \notin K^{(\text {TM})}_{sing}\), we have that
- (i)
For \(0\le s \le r-1\),
$$\begin{aligned} {\mathcal {U}}_k : H^{s+\frac{1}{2}}_\theta (\Gamma ) \rightarrow H^{s+1}_{\theta ,{\mathrm {loc}}}(\Omega ), \quad {\mathcal {V}}_k : H^{s+\frac{1}{2}}_\theta (\Gamma ) \rightarrow H^{s+1}_{\theta ,{\mathrm {loc}}}(\Omega ). \end{aligned}$$ - (ii)
For \(-r+1\le s \le r-1\),
$$\begin{aligned} \gamma _1^i {\mathcal {U}}_k : H^{s+\frac{1}{2}}_\theta (\Gamma ) \rightarrow H^{s-\frac{1}{2}}_{\theta }(\Gamma ),\quad \gamma _1^i {\mathcal {V}}_k : H^{s+\frac{1}{2}}_\theta (\Gamma ) \rightarrow H^{s-\frac{1}{2}}_{\theta }(\Gamma ). \end{aligned}$$
Proof
We begin by proving (i). The case \(s=0\) is direct from Proposition B.2, while the result for \(s=r+1\) follows from Theorem B.7. For \(0<s<r-1\), the result is derived by interpolation [30, Appendix B]—interpolation of quasi-periodic spaces in the boundary \(\Gamma \) follows from their definition, inducing an isomorphism to regular Sobolev spaces on closed boundaries [24, Chapter 8]). For (ii), the result for positive s is deduced by similar arguments as those used for (i) whereas the result for \(s<0\) is due to the duality pairing in Lemma B.3. \(\square \)
Theorem B.9
(Theorem 4.24 in [30]) Assume \(\Omega \) to be Lipschitz. Let \(k>0\), \(f \in L^2_\theta (\Omega )\) and \(u \in H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) such that
If \(\gamma _0^iu \in H^1_\theta (\Gamma ) \) then \(\gamma _1^iu \in L_\theta ^2(\Gamma )\) and, for R such that \(\Gamma \subset {\mathcal {G}}^R\), we have that
Proof
First, we assume that \(u \in H^2_{\theta ,{\mathrm {loc}}}(\Omega )\) and, following the proof for [30, Theorem 4.24], it can be shown that
Now consider a bounded open set \({\mathcal {O}}\subset \Omega ^R\) such that \({\overline{{\mathcal {O}}}}^{\mathcal {G}}\subset \Omega ^R\), with \(\partial ^{\mathcal {G}}{\mathcal {O}}\) composed of two periodic curves, one of them being \(\Gamma ^R\). By Theorem 2.29 and the definition of the Neumann trace for smooth functions, we have that
and, for \(0<\epsilon <\frac{1}{2}\), it holds
Then, by Theorem B.6, we derive
We now take \(u\in H^1_{\theta , {\mathrm {loc}}}(\Omega )\) and assume that \(\Gamma \) can be parametrized as \((x,\zeta (x))\) with \(x \in (0,2\pi )\). Consider a sequence of smooth functions \(\{\zeta _n\}_{n \in {{\mathbb {N}}}}\) such that
Then, \(\Omega = \{\varvec{x}\in {\mathcal {G}}\ |\ x_2 > \zeta (x)\}\). Define
Following the proof of [30, Theorem 4.24], let us define \(g(\varvec{x}):= \gamma _0^i u(x_1,\zeta (x_1))\), where the trace is taken over \(\Gamma \). Finally, we consider \(\lambda >k^2\), \(R>0\) such that \(\Gamma \subset {\mathcal {G}}^R\) and a sequence \(\{u_n\}_{n\in {{\mathbb {N}}}}\) where each \(u_n\in H^1_\theta (\Omega _n)\) satisfies
The elements of the sequence \(\{u_n\}_{n \in {{\mathbb {N}}}}\) are well defined in \(H^1_\theta (\Omega _{n}^R)\) since the domain is bounded and the associated operator is elliptic. Since \(\Gamma _n\) is smooth, by Theorem B.7, we have that each \(u_n\) belongs to \(H^2_\theta (\Omega _{n}^R)\) for all \(n\in {{\mathbb {N}}}\). Hence, we can use the result for elements of \(H^2_{\theta , {\mathrm {loc}}}(\Omega )\). To conclude, we need to show that a proper extension of \(u_n\) converges to u in \(H^1_\theta (\Omega ^R)\), which is done in [30, Theorem 4.24] for regular Sobolev spaces and extended to quasi-periodic Sobolev spaces with only minor modifications. \(\square \)
Corollary B.10
(Theorem 4.25 in [30]) Assume \(\Omega \) to be Lipschitz. Let \(k \notin K^{(\text {TM})}_{sing}\). For \(\left| s\right| \le \frac{1}{2}\), it holds
Proof
The case \(s=0\) is given by Proposition B.2, \(s=\frac{1}{2}\) is given by Theorem B.9, and \(s=-\frac{1}{2}\) is obtained by the duality relation in Lemma B.3. For all other \(\left| s\right| <\frac{1}{2}\), the result follows by interpolation. \(\square \)
In order to prove the map** properties of the double layer potential, we need one more auxiliary result. For \(k>0\), we denote by \({\mathcal {U}}^-_k\) the solution operator in \(\Omega ^- := {\mathcal {G}}\setminus {\overline{\Omega }}^{\mathcal {G}}\), and \({\mathcal {U}}^+_k := {\mathcal {U}}_k\). Given \(\lambda \in {{\mathbb {R}}}\) such that \(k^2-\lambda >0\), we set \({\mathcal {U}}^\pm _{k,\lambda } := {\mathcal {U}}^\pm _{\sqrt{k^2-\lambda }}\).
Lemma B.11
Let \(\Omega \) be Lipschitz and set \(k >0\). Then, there exists \(\lambda \in {{\mathbb {R}}}\) such that \(k^2-\lambda >0\) and \({\mathcal {U}}^+_{k,\lambda }\) as well as \({\mathcal {U}}^-_{k,\lambda }\) are well defined in \(H^{\frac{1}{2}}_\theta (\Gamma )\). For \(\left| s\right| <\frac{1}{2}\), we also have that
Proof
Since the eigenvalues of the problem in \(\Omega \) and \(\Omega ^-\) are numerable we can find \(\lambda \) such that \(k^2-\lambda >0\) and \(\left| \theta +j\right| \ne \sqrt{k^2-\lambda }\), for every \(j \in {{\mathbb {Z}}}\). Then, the following sets of equations
are satisfied by only one element of \(H^1_{\theta ,{\mathrm {loc}}}(\Omega )\) and \(H^1_{\theta ,{\mathrm {loc}}}(\Omega ^-)\), respectively. Then, consider w defined as
Thanks to the properties of the solution operators, we have
By Theorem 4.9, it holds
with \([\gamma _1 w]:=\gamma ^i_1w-\gamma ^e_1w\). Then, by the continuity of the single layer potential and Corollary B.10, it holds
Thus, we can conclude that
from where the result follows. \(\square \)
We define operators \({\mathcal {V}}^\pm _{k,\lambda }\) in a similar fashion to \({\mathcal {U}}^\pm _{k,\lambda }\) by using the adjoint radiation condition (cf. Definition 3.5) and repeating the steps presented above. It is easy to check that both operators have the same properties.
Proof of Theorem 4.10
Results for \({\mathsf {S}}{\mathsf {L}}_\theta ^{k}\) and \({{\mathsf {V}}}_\theta ^{k}\) can be established directly from their definitions and Theorems 2.29 and 4.7. Now, consider \(\eta \), \(\mu \in {\mathcal {D}}_\theta (\Gamma )\) and let \(\lambda \in {{\mathbb {R}}}\) be such that \(k^2-\lambda >0\) and \({\mathcal {V}}^+_{k,\lambda }\) is well defined. By the map** properties of \({\mathsf {S}}{\mathsf {L}}_\theta ^k\), we have that
Applying Lemma 2.33 leads to
On the other hand, since \((-\Delta - k^2+\lambda ) {\mathcal {V}}^+_{k,\lambda } \mu = 0\) in \(\Omega ^R\), Green’s formula yields
As the single layer potential satisfies the radiation condition in Definition 3.2 (cf. Proposition 4.2) and \({\mathcal {V}}^+_{k,\lambda } \mu \) the adjoint version (Definition 3.5), we get
by the same arguments as in the proof of Lemma B.3. Then,
The first term in the right-hand side can be bounded as
where the last inequality follows from the continuity of \(\gamma _0^i {\mathsf {S}}{\mathsf {L}}_\theta ^k = {{\mathsf {V}}}_\theta ^k\) and Corollaries B.10 or B.8 depending on whether \(\Gamma \) is Lipschitz or smoother, respectively. For the second term, it holds
where the last inequality is due to the continuity of \({\mathsf {S}}{\mathsf {L}}^k_\theta \), and Lemma B.11. Map** properties for \(\gamma _1^i{\mathsf {S}}{\mathsf {L}}_\theta ^k = {{{\mathsf {K}}}^k_\theta }'\) are obtained by density arguments.
For the double layer potential and its traces, pick \(g\in {\mathcal {D}}_\theta (\Gamma )\) and use the representation formula in Theorem 4.9—with \({\mathcal {U}}^+_{k,\lambda } g\) extended by zero to \(\Omega ^-\)—to obtain
Thus, we obtain the estimate
By Lemma B.11, the map** properties of \({\mathsf {S}}{\mathsf {L}}_\theta ^k\) and Theorem 4.4, we obtain
Finally, by using Corollary B.10, we have that
Bounds for the norms in \({\mathcal {G}}\setminus {\overline{\Omega }}^{\mathcal {G}}\) are derived by using \({\mathcal {U}}_{k,\lambda }^-\) and repeating the same procedure. The continuity of \(\gamma _0^i \mathrm {D}{\mathsf {L}}^\theta _k = {{\mathsf {K}}}^k_\theta \) is direct from the trace continuity in Theorem 2.29. The Neumann trace can be estimated as follows
The first term on the right-hand side is bounded by Corollary B.10 whereas the second one is bounded by the continuity of \({{{\mathsf {K}}}_\theta ^k}^\prime \). The last term is bounded by the continuity of the Neumann trace, that of the Newton potential and Corollary B.10. \(\square \)
Rights and permissions
About this article
Cite this article
Aylwin, R., Jerez-Hanckes, C. & Pinto, J. On the Properties of Quasi-periodic Boundary Integral Operators for the Helmholtz Equation. Integr. Equ. Oper. Theory 92, 17 (2020). https://doi.org/10.1007/s00020-020-2572-9
Received:
Revised:
Published:
DOI: https://doi.org/10.1007/s00020-020-2572-9