Skip to main content
Log in

On the Choice of Robin Parameters for the Optimized Schwarz Method for Domains with Non-Conforming Heterogeneities

  • Published:
Journal of Scientific Computing Aims and scope Submit manuscript

Abstract

We consider the solution of \(-\nabla \cdot (\nu (x)\nabla u)=0\) by a non-overlapping optimized Schwarz domain decomposition method, where the subdomains do not align with jumps in the coefficient \(\nu (x)\). Such a decomposition can be of interest when the jumps are geometrically complex and/or an artifact of the measured data, in which case one would often prefer a simpler decomposition that disregards the location of the discontinuities. For analysis purposes, we focus on a model problem where the diffusivity is piecewise constant, and we analyze the convergence of optimized Schwarz for the two-subdomain case. We consider using either a constant Robin parameter along the whole interface, or a parameter that is scaled proportionally to the local diffusivity. We show that the convergence rate is not robust with respect to the heterogeneity ratio when a constant Robin parameter is used; however, using a scaled Robin parameter restores robustness. We then derive optimal scaling parameter and the corresponding convergence factor. Numerical examples show that this choice also leads to robust convergence behaviour for cases not covered by the analysis.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5
Fig. 6
Fig. 7
Fig. 8

We’re sorry, something doesn't seem to be working properly.

Please try refreshing the page. If that doesn't work, please contact support so we can address the problem.

Data Availability Statement

Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.

Notes

  1. In the rest of this paper, we will use normal letters, e.g., u, to denote finite element functions or their traces on the boundary, and use bold letters, e.g., \(\mathbf {u}\), to denote the vector corresponding to u. We will often switch between normal and bold letters.

References

  1. Bennequin, D., Gander, M., Halpern, L.: A homographic best approximation problem with application to optimized Schwarz waveform relaxation. Math. Comput. 78(265), 185–223 (2009)

    Article  MathSciNet  Google Scholar 

  2. Berninger, H., Kornhuber, R., Sander, O.: Convergence behaviour of Dirichlet–Neumann and Robin methods for a nonlinear transmission problem. In: Domain decomposition methods in science and engineering XIX, pp. 87–98. Springer (2011)

  3. Berninger, H., Loisel, S., Sander, O.: The 2-Lagrange multiplier method applied to nonlinear transmission problems for the Richards equation in heterogeneous soil with cross points. SIAM J. Sci. Comput. 36(5), A2166–A2198 (2014)

    Article  MathSciNet  Google Scholar 

  4. Caetano, F., Gander, M.J., Halpern, L., Szeftel, J.: Schwarz waveform relaxation algorithms with nonlinear transmission conditions for reaction-diffusion equations. In: Domain decomposition methods in science and engineering XIX, pp. 245–252. Springer, New York (2011)

  5. Christie, M.A., Blunt, M., et al.: Tenth SPE comparative solution project: A comparison of upscaling techniques. In: SPE reservoir simulation symposium. Society of Petroleum Engineers (2001)

  6. Collino, F., Ghanemi, S., Joly, P.: Domain decomposition method for harmonic wave propagation: a general presentation. Comput. Methods Appl. Mech. Eng. 184(2–4), 171–211 (2000)

    Article  MathSciNet  Google Scholar 

  7. Dolean, V., Gander, M.J., Veneros, E.: Asymptotic analysis of optimized Schwarz methods for Maxwells equations with discontinuous coefficients. ESAIM Math. Model. Numer. Anal. 52(6), 2457–2477 (2018)

    Article  MathSciNet  Google Scholar 

  8. Dolean, V., Jolivet, P., Nataf, F.: An introduction to domain decomposition methods: algorithms, theory, and parallel implementation. SIAM (2015)

  9. Gander, M.J.: Optimized Schwarz methods. SIAM J. Numer. Anal. 44(2), 699–731 (2006)

    Article  MathSciNet  Google Scholar 

  10. Gander, M.J., Dubois, O.: Optimized Schwarz methods for a diffusion problem with discontinuous coefficient. Numer. Algorithms 69(1), 109–144 (2015)

    Article  MathSciNet  Google Scholar 

  11. Gander, M.J., Kwok, F.: Best Robin parameters for optimized Schwarz methods at cross points. SIAM J. Sci. Comput. 34(4), A1849–A1879 (2012)

    Article  MathSciNet  Google Scholar 

  12. Gander, M.J., Magoulès, F., Nataf, F.: Optimized Schwarz methods without overlap for the Helmholtz equation. SIAM J. Sci. Comput. 24(1), 38–60 (2002)

    Article  MathSciNet  Google Scholar 

  13. Gander, M.J., Santugini, K.: Cross-points in domain decomposition methods with a finite element discretization. Electron. Trans. Numer. Anal. 45, 219–240 (2016)

    MathSciNet  MATH  Google Scholar 

  14. Gerardo-Giorda, L., Le Tallec, P., Nataf, F.: A Robin-Robin preconditioner for advection-diffusion equations with discontinuous coefficients. In: Herrera, I., Keyes, D.E., Widlund, O.B., Yates, R. (eds.) Proceedings of the 14th International Conference on Domain Decomposition Methods. Cocoyoc, Mexico (2002)

  15. Greer, N., Loisel, S.: The optimised Schwarz method and the two-Lagrange multiplier method for heterogeneous problems in general domains with two general subdomains. Numer. Algorithms 69(4), 737–762 (2015)

    Article  MathSciNet  Google Scholar 

  16. Gu, Y., Kwok, F.: Optimized Schwarz-based nonlinear preconditioning for elliptic PDEs. In: Domain Decomposition Methods in Science and Engineering XXV, pp. 260–267. Springer (2020)

  17. Häberlein, F., Halpern, L.: Optimized Schwarz waveform relaxation for nonlinear systems of parabolic type. In: Domain Decomposition Methods in Science and Engineering XXI, pp. 29–42. Springer (2014)

  18. Hughes, T.J.: The Finite Element Method: Linear Static and Dynamic Finite Element Analysis. Dover reprint, originally published in 1987 by Prentice-Hall, Englewood Cliffs (2000)

  19. Japhet, C., Nataf, F., Rogier, F.: The optimized order 2 method: application to convection-diffusion problems. Futur. Gener. Comput. Syst. 18(1), 17–30 (2001)

    Article  Google Scholar 

  20. Lions, P.L.: On the Schwarz alternating method. III: a variant for nonoverlapping subdomains. In: Third International Symposium on Domain Decomposition Methods for Partial Differential Equations, vol. 6, pp. 202–223. SIAM Philadelphia, PA (1990)

  21. Lui, S.: A Lions non-overlapping domain decomposition method for domains with an arbitrary interface. IMA J. Numer. Anal. 29(2), 332–349 (2009)

    Article  MathSciNet  Google Scholar 

  22. Maday, Y., Magoulès, F.: Non-overlapping additive Schwarz methods tuned to highly heterogeneous media. C. R. Math. 341(11), 701–705 (2005)

    Article  MathSciNet  Google Scholar 

  23. Maday, Y., Magoulès, F.: Optimized Schwarz methods without overlap for highly heterogeneous media. Comput. Methods Appl. Mech. Eng. 196(8), 1541–1553 (2007)

    Article  MathSciNet  Google Scholar 

  24. Nataf, F.: Absorbing boundary conditions in block Gauss–Seidel methods for convection problems. Math. Models Methods Appl. Sci. 6(04), 481–502 (1996)

    Article  MathSciNet  Google Scholar 

  25. Nataf, F., Nier, F.: Convergence rate of some domain decomposition methods for overlapping and nonoverlapping subdomains. Numer. Math. 75(3), 357–377 (1997)

    Article  MathSciNet  Google Scholar 

  26. Qin, L., Xu, X.: On a parallel Robin-type nonoverlapping domain decomposition method. SIAM J. Numer. Anal. 44(6), 2539–2558 (2006)

    Article  MathSciNet  Google Scholar 

  27. Saad, Y.: Iterative methods for sparse linear systems, vol. 82. SIAM, Philadelphia (2003)

    Book  Google Scholar 

  28. Toselli, A., Widlund, O.: Domain Decomposition Methods: Algorithms and Theory, vol. 34. Springer, New York (2005)

    Book  Google Scholar 

Download references

Acknowledgements

We thank the anonymous referee for their helpful comments, which led to a significant improvement in the quality of the paper. A portion of this research was conducted while the authors were at Hong Kong Baptist University.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Felix Kwok.

Ethics declarations

Funding

This work is partially supported by the NSERC Discovery Grant (Canada, Ref. No. RGPIN-2021-02595), the Research Grants Council of the Hong Kong SAR, China (Project No. HKBU 12301817, HKBU 12301018) and the National Natural Science Foundation of China (Young Scientist Fund No. 11501483)

Author contribution

This research was conducted as part of the PhD dissertation of Yaguang Gu, which was completed under the supervision of Felix Kwok at Hong Kong Baptist University. The first draft of the manuscript was written by Yaguang Gu, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Conflicts of Interest/Competing Interests

Not Applicable

Availability of data and material

Not Applicable

Code availability

Not Applicable

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix: Useful Sobolev-Type Estimates

Appendix: Useful Sobolev-Type Estimates

Lemma 3

([28], Lemma B.5) Let \(\phi \) be a basis function associated with a node of an element \(K \subset \varOmega \subset \mathbb {R}^n\), Then there exist constants c and C, independent of h, such that

$$\begin{aligned} c h^{n} \le \Vert \phi \Vert _{L^{2}(K)}^{2} \le C h^{n}. \end{aligned}$$

Consequently, for any \(u_\varGamma \), the trace of a finite element function on \(\varGamma \), there exist constants \(c_1\) and \(C_1\), independent of h and H, such that

$$\begin{aligned} c_1 h^{n-1} \mathbf {u}_\varGamma ^{T} \mathbf {u}_\varGamma \le \Vert u_\varGamma \Vert _{L^2(\varGamma )}^{2} \le C_1 h^{n-1} \mathbf {u}_\varGamma ^{T} \mathbf {u}_\varGamma . \end{aligned}$$

Lemma 4

([28], Lemma A.14 and Corollary A.15) Let \(\varOmega \) be a Lipschitz continuous domain with diameter H. Let \(\varGamma \in \partial \varOmega \) have nonvanishing measure. If \(u\in H^1(\varOmega )\) vanishes on \(\varGamma \), then there exists a constant C, independent of H, such that

$$\begin{aligned} \Vert {u} \Vert _{L^2(\varOmega )}^2 \le C^2 H^2 \Big \vert {u}\Big \vert _{H^1(\varOmega )}^2. \end{aligned}$$

Moreover, if we define

$$\begin{aligned} \Vert {u} \Vert _{H_s^{1}(\varOmega )}^{2}:=\Big \vert {u}\Big \vert _{H^{1}(\varOmega )}^{2}+\frac{1}{H^{2}}\Vert {u} \Vert _{L^{2}(\varOmega )}^{2} \end{aligned}$$

as the scaled \(H^1\)-norm, then we obtain the following estimate:

$$\begin{aligned} \Vert {u} \Vert _{H_s^{1}(\varOmega )}^{2}\le (1+C^2)\Big \vert {u}\Big \vert _{H^{1}(\varOmega )}^{2}. \end{aligned}$$

Lemma 5

Let \(u_\varGamma \) be the trace of a finite element function on \(\varGamma \), and let \(\mathbf {u}_\varGamma \) be the vector corresponding to \(u_\varGamma \). Let \(A_1\) be the stiffness matrix associated with the standard Laplacian bilinear form, and \(S_1\) be its Schur complement after the interior nodes have been eliminated. Then there exist constants c and C, independent of h and H, such that the following sharp estimate holds:

$$\begin{aligned} cH^{-1}\Vert {u_\varGamma } \Vert _{L^2(\varGamma )}^2 \le \mathbf {u}_\varGamma ^TS_1\mathbf {u}_\varGamma \le C h^{-1}\Vert {u_\varGamma } \Vert _{L^2(\varGamma )}^2, \end{aligned}$$
(52)

where h is the diameter of the finite element, and H is the diameter of the subdomain. Consequently, there exists a constant \(C'\), independent of h and H, such that

$$\begin{aligned} \kappa (S_1)<C'\frac{H}{h}. \end{aligned}$$
(53)

Proof

The proof is based on Lemma 4.11 in [28], where the authors have derived the upper bound. For the lower bound, they have also proved that

$$\begin{aligned} H\Vert {u} \Vert _{L^{2}(\varGamma )}^{2} \le H^{2}\Vert u\Vert _{H^{1 / 2}(\varGamma )}^{2} \le C_{t}^{2}\left( H^{2}\Big \vert {u}\Big \vert _{H^{1}(\varOmega _{i})}^{2}+\Vert {u} \Vert _{L^{2}(\varOmega _{i})}^{2}\right) , \end{aligned}$$

where \(C_t\) is a constant independent of h and H. Simplification of the above inequality gives

$$\begin{aligned} H^{-1}\Vert {u} \Vert _{L^{2}(\varGamma )}^{2} \le C_{t}^{2}\left( \;\Big \vert {u}\Big \vert _{H^{1}(\varOmega _{i})}^{2}+\frac{1}{H^2}\Vert {u} \Vert _{L^{2}(\varOmega _{i})}^{2}\right) =C_{t}^{2}\Vert {u} \Vert _{H_s^1(\varOmega _i)}^2, \end{aligned}$$

where we have used the definition of the scaled \(H^1\)-norm (see also (4.4) in [28]). By Lemma 4, we obtain

$$\begin{aligned} H^{-1}\Vert {u} \Vert _{L^{2}(\varGamma )}^{2} \le C \Big \vert {u}\Big \vert _{H^1(\varOmega _i)}^2. \end{aligned}$$
(54)

Since \(\mathbf {u}_\varGamma ^TS_1\mathbf {u}_\varGamma = \Big \vert {u}\Big \vert _{H^1(\varOmega _i)}^2\), actually we have finished the proof for the lower bound.

The condition number estimate is then a direct result from the spectral estimates. Note that \(\kappa (S_1)\) is none other than the condition number of the Schur complement matrix for the Laplace operator. \(\square \)

Lemma 6

([28], Lemma 4.10) Let \(\varOmega \) be a domain, and let a finite element trace function \(u_\varGamma \) be given on the boundary \(\varGamma =\partial \varOmega \). Let \(\mathbf {u}=\mathscr {H}_1(\mathbf {u}_\varGamma )\) be the discrete 1-harmonic extension of \(u_\varGamma \) into \(\varOmega \). Then there exist constants c and C, independent of h and H, such that

$$\begin{aligned} c\left| u_{\varGamma }\right| _{H^{1 / 2}(\varGamma )}^{2} \le |u|_{H^{1}\left( \varOmega _{1}\right) }^{2} \le C\left| u_{\varGamma }\right| _{H^{1 / 2}(\varGamma )}^{2}. \end{aligned}$$
(55)

Lemma 7

([11], Lemma 4.5) Let \(\varOmega \) be a domain, and let a finite element trace function \(u_\varGamma \) be given on the boundary \(\varGamma =\partial \varOmega \). Then there exist a constant C, independent of h and H, such that

$$\begin{aligned} \Big \vert {u_{\varGamma }}\Big \vert _{H^{1/2}(\varGamma )}^{2} \le \frac{C}{h}\Vert {u_{\varGamma }} \Vert _{L^{2}(\varGamma )}^{2}. \end{aligned}$$
(56)

Lemma 8

([28], Lemma A.17) Let \(\varOmega \) be a bounded Lipschitz continuous polygon or polyhedron with diameter H. Then for any \(u\in H^{1/2}(\partial \varOmega )\) such that the measure of the set \(\{x\in \partial \varOmega ; u(x)=0.\}\) is non-zero, there exists a constant C, independent of h, such that

$$\begin{aligned} \Vert {u} \Vert _{L^2(\partial \varOmega )}^2\le CH\Big \vert {u}\Big \vert _{H^{1/2}(\partial \varOmega )}^2. \end{aligned}$$
(57)

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Gu, Y., Kwok, F. On the Choice of Robin Parameters for the Optimized Schwarz Method for Domains with Non-Conforming Heterogeneities. J Sci Comput 89, 5 (2021). https://doi.org/10.1007/s10915-021-01617-6

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1007/s10915-021-01617-6

Keywords

Mathematics Subject Classification

Navigation