betaFeedback

WorksMathematics

An all-steps-at-once algorithm for time-domain wave propagation in heterogeneous and unbounded media

Abstract

We propose, analyze, and implement a convolution quadrature (CQ) algorithm for transient acoustic waves in an unbounded medium consisting of a heterogeneous domain embedded in free space, where the wave speed may be spatially variable and discontinuous. Multistep CQ, applied to the time-domain solution operator, reduces the evolution problem to a family of Helmholtz problems with damped complex frequencies, which can be solved independently in parallel, all time steps at once. The frequency problems are discretized by an overlapping coupling of finite elements (FEM) on a polygonal/polyhedral domain containing the inhomogeneity, and boundary elements (BEM) on a smooth curve enclosed by that domain, in the region where the medium is homogeneous, which accounts exactly for the unbounded exterior. Neither discretization needs to conform to the geometry of the heterogeneous domain, nor to each other. For the resulting FEM–BEM coupling we prove bounds for the solution operators of the transformed problems that are explicit in the Laplace parameter and the wave speed. These bounds yield convergence of the time discretization for A-stable multistep schemes. We also prove explicit estimates on the exponential decay of the transformed solutions away from the heterogeneity, which show that the correction provided by the BEM part is small but relevant, and cheap to compute: a few GMRES iterations suffice, and the correction is needed only over a fixed range of weakly damped frequencies, beyond which it becomes negligible and can be dropped. Numerical experiments confirm the predicted convergence rates in time and space and explore iterative (multigrid) methods for the interior FEM problems.

Introduction

Computational algorithms for simulating the propagation of time-dependent acoustic waves in heterogeneous and unbounded media are fundamental to numerous applications [8]. Three main strategies can be found in the literature: finite elements in a truncated region, with the exterior replaced by approximate non-reflecting boundary conditions (NRBCs) [20, 25, 29] or perfectly matched layers (PMLs) [7, 9]; boundary elements, which represent the unbounded exterior exactly but are restricted to piecewise homogeneous media, since a variable wave speed cannot be accommodated by boundary potentials alone [10, 23]; and couplings of finite and boundary elements, which combine both, typically with convolution quadrature (CQ) applied only to the boundary integral operators and a time-stepping scheme in the interior, as discussed below.

In this article we consider a time-domain wave propagation model problem in Rm\mathbb{R}^m, m=2,3m=2,3, involving a transient scalar wave field that propagates through a heterogeneous domain Ω\Omega, where the spatially varying wave speed may be discontinuous, and into the unbounded free-space exterior Ωc=Rm∖Ω\Omega^c=\mathbb{R}^m\setminus\Omega. The inhomogeneity is excited by an incident wave, and the resulting time-dependent wave field can be described, through the Laplace transform, by a family of frequency-domain scattering problems with complex wave numbers. CQ replaces this family by a finite set of such problems and reconstructs the full time history from their solutions, providing a robust and flexible framework for numerical simulation. Each transformed problem is solved by an overlapping coupling of the finite element method (FEM) and the boundary element method (BEM), whose construction and analysis are the subject of this paper.

Let us mention [32], where the model problem solved at each complex frequency is exactly the one considered here; the difference lies in its discretization. There, the transmission problem is recast as a volume Lippmann–Schwinger equation (see [12 Ch. 8], [39 Ch. 10]) and solved by trigonometric collocation, so that the unknown is supported throughout the inhomogeneity and the resulting operator is dense. Here, we retain the differential formulation and discretize it by finite elements, using boundary integral operators only to account for the unbounded exterior. Consequently, the analysis in [32] requires the sound speed to be globally smooth, whereas our formulation accommodates discontinuities in the sound speed across material interfaces.

Problems of this kind have also been treated numerically using non-overlapping FEM–BEM coupling [3, 4, 19, 28, 38]. In all of these approaches, the boundary integral equation, discretized in time by CQ, is posed on the boundary of the finite element domain, which serves as both the truncation boundary and the coupling interface. The interior problem is advanced in time either by an explicit leapfrog scheme [4, 19, 3] or by the same A-stable scheme used to generate the CQ. Among the implicit schemes, we mention the trapezoidal rule in [28] and a Runge–Kutta method in [38], which is devoted to the time-dependent linear Schrödinger equation.

The method we propose and analyze here differs from these approaches in its coupling strategy and, in particular, in its geometric flexibility. The finite element domain is chosen as a simple polygonal/polyhedral domain containing the heterogeneous domain, while the boundary element problem is posed on an auxiliary curve/surface lying inside the finite element domain and surrounding the heterogeneity, in the region where the medium is homogeneous, so that the two discretizations overlap. Consequently, neither the finite element domain nor the boundary element curve/surface needs to conform to, or depend on, the geometry of the heterogeneity. This separation of the discretization geometries provides considerable flexibility in handling complex or irregular heterogeneous domains. Moreover, the auxiliary boundary element curve/surface can be chosen to be smooth, or even analytic, independently of the geometry of the heterogeneity, which is particularly advantageous for high-order BEM.

In contrast, in a non-overlapping coupling, the boundary integral operators are posed on the polygonal/polyhedral boundary of the finite element domain and therefore inherit its corners. The finite element mesh may of course be chosen to resolve the material interfaces, as we do in our numerical experiments, to preserve the accuracy of the finite elements across jumps in the speed of sound; however, this is a property of the mesh and is not a requirement of the coupling strategy; see § 8 for the alternative. This work can be viewed as a time-domain (TD) counterpart of the frequency-domain framework in [14, 16].

The advantages of our overlapping TD-FEM–BEM approach are several. First, the all-steps-at-once implementation is naturally parallel. This is an algorithmic choice rather than a choice of discretization: the same discrete solution can also be computed by marching in time, in which case one recovers a scheme of the type used in [28]; see Remark 3.1. In the all-steps-at-once formulation, however, the decoupled frequency-domain problems are independent and uniformly damped, and can therefore be solved fully in parallel.

Second, the exterior is accounted for exactly. Truncating the computational domain by means of NRBCs [29] introduces an additional approximation whose accuracy must be controlled independently, whereas the boundary element component provides an exact transparent condition on Σ\Sigma, the boundary of the FEM domain.

The price to be paid is a non-local operator. However, this price is seldom significant: unlike in classical acoustic scattering, the complex frequencies generated by the Laplace transform introduce strong damping, so that the transmitted field is negligible on Σ\Sigma for all but a small fraction of the frequencies. The coupling is therefore activated only within a fixed frequency window, whose size we quantify in this article. Even within this window, the coupled FEM–BEM system is solved by GMRES, which converges in very few iterations, typically one or two.

Finally, the interior problems associated with the transformed formulation are well suited to iterative solvers. They are coercive for every Laplace parameter ss, with positive real part, with a coercivity constant proportional to the relative damping ℜ(s)/∣s∣\Re(s)/|s|. The complex shift therefore improves the smoothing properties of standard relaxation schemes and mitigates the resonance phenomena that would otherwise arise at purely real frequencies; this is the mechanism exploited by shifted-Laplacian preconditioners for the Helmholtz equation [21], where, in contrast to our setting, the shift is introduced artificially. This makes multigrid a viable option, although, as our experiments show, not over the entire range of frequencies produced by the CQ discretization. This observation is particularly important in three dimensions, where sparse direct solvers quickly become prohibitively expensive.

On the theoretical side, this article provides bounds for the solution operators of the transformed problems that are explicit in the Laplace parameter and in the wave speed (Proposition 4.1). These bounds place the problem within the framework of [6] and yield convergence of the time discretization for BDF1, BDF2, and the trapezoidal rule, with the order of the underlying multistep method and with constants that grow at most polynomially with the final time (Theorem 2.2). We also prove exponential decay estimates for the transformed solutions away from the inhomogeneity (Proposition 4.5). These estimates quantify the frequency window in which the boundary element correction is needed (Theorem 3.2) and show that the coupled operator is a contraction outside this window (Theorem 3.3).

The analysis in this article is restricted to the time discretization. The dependence on the spatial discretization of the BEM and FEM components is more delicate, but we believe that such an analysis is attainable in a future work. The damping introduced by the Laplace transform, with ℜ(s)≥σ0>0\Re(s) \ge\sigma_0 > 0 uniformly along the quadrature contour, keeps all transformed problems coercive and avoids the resonances associated with purely real frequencies. The main difficulty is therefore to track the dependence of the relevant constants on ss, in the spirit of the estimates in § 5 and the semidiscrete analyses in [4, 28, 40].

Let us point out, however, that this damping does not by itself prevent the pollution effect [37], since a sizable portion of the quadrature frequencies has small relative damping ℜ(s)/∣s∣\Re(s)/|s|. In our experiments, however, the frequencies that carry non-negligible data are resolved by the mesh, while those that are not carry data only at the level of round-off error; see § 8.

This paper is structured as follows. In § 2, we introduce the physical model problem under consideration and the mathematical model used for its numerical approximation. We then describe the linear multistep CQ method employed for the time discretization, together with its all-steps-at-once implementation, and present the convergence results in time, relying on the well-established theory for CQ methods [6]; see also [31, 35, 40] for multistep CQ based on BDF-type schemes. In § 4, we explain how the transformed problems are addressed using our overlapping FEM–BEM framework and analyze the resulting coupled equation. A key point is the decay of the solution in the exterior of the inhomogeneity, which determines the form of the transformed problem and the role of the BEM component. This issue is examined in detail in § 5, where we also establish bounds for the continuity constants with respect to the transformed variable—precisely the type of estimates required in the preceding sections. § 8 contains an extensive set of numerical experiments illustrating the efficiency of the algorithm and providing further support for the theoretical results. Finally, § 9 summarizes the main conclusions and outlines directions for future work.

A model problem and time-discretization algorithm

Transient acoustic wave problem

We consider a model problem for transient acoustic wave propagation in Rm\mathbb{R}^{m}, m=2,3m = 2, 3, consisting of a heterogeneous domain Ω\Omega embedded in an unbounded homogeneous exterior region. The wave speed c∈L∞(Rm)c \in L^{\infty}(\mathbb{R}^{m}) is positive and may vary spatially and be discontinuous within Ω\Omega. In the exterior free space Ωc=Rm∖Ω‾\Omega^{c} = \mathbb{R}^{m} \setminus\overline{\Omega}, the wave speed is normalized to the constant cext=1c_{\mathrm{ext}} = 1.

The evolution of the total wave field in Rm\mathbb{R}^{m} is described by the scalar field ϑ=ϑ(x,t)\vartheta= \vartheta(\mathbf{x}, t), which satisfies the classical time-domain wave equation:

c−2(x)ϑtt(x,t)=Δϑ(x,t),x∈Rm, t>0.c^{-2}(\mathbf{x}) \vartheta_{tt}(\mathbf{x}, t) = \Delta\vartheta(\mathbf{x}, t), \qquad\mathbf{x} \in\mathbb{R}^{m},\ t > 0.

The total field is induced by a smooth incident field ϑinc\vartheta^{\mathrm{inc}}, which originates in the exterior region Ωc\Omega^{c}, and with support that does not intersect Ω\Omega for t≤0t \leq0. The incident field satisfies the wave equation

ϑttinc(x,t)=Δϑinc(x,t),x∈Rm.\vartheta^{\mathrm{inc}}_{tt}(\mathbf{x}, t) = \Delta\vartheta^{\mathrm{inc}}(\mathbf{x}, t), \qquad\mathbf{x} \in\mathbb{R}^{m}.

The interaction of this wave with the inhomogeneity generates a scattered and propagating field u=ϑ−ϑincu = \vartheta- \vartheta^{\mathrm{inc}}, which satisfies homogeneous initial data and the evolution equation

u∈H2([0,T],L2(Rm)),1c2u¨(t)=Δu(t)+f(t),t∈[0,T],u(0)=0,u˙(0)=0,(1)\begin{aligned} u &\in H^{2}([0,T], L^{2}(\mathbb{R}^{m})),\\ \frac{1}{c^{2}}\ddot{u}(t) &= \Delta u(t) + f(t), \qquad t \in[0,T],\\ u(0) &= 0,\qquad\dot{u}(0) = 0, \tag*{(1)} \end{aligned}

where

f(x,t):=cR(x) ϑttinc(x,t),cR(x):=1−c−2(x)f(\mathbf{x},t) := c_{\mathrm{R}}(\mathbf{x})\,\vartheta_{tt}^{\mathrm{inc}}(\mathbf{x},t), \qquad c_{\mathrm{R}}(\mathbf{x}) := 1-c^{-2}(\mathbf{x})

is supported inside Ω\Omega since c(x)=1c(\mathbf{x})=1 in the free space (i.e. for x∈Ωc\mathbf{x}\in\Omega^{c}). The function cRc_{\mathrm{R}} is the contrast function and the regularity in space of the solution uu depends on the regularity of cRc_{\mathrm{R}} as a function in Rm\mathbb{R}^{m}.

Convolution Quadrature Method

Let us introduce, for now only in a formal way, the Laplace transform:

F(s):=(Lf)(s):=∫0∞f(t)e−st dt,s∈C+:={s∈C:ℜs>0}.F(s):=(\mathcal{L}f)(s):=\int_{0}^{\infty}f(t)e^{-st}\,\mathrm{d}t,\qquad s\in\mathbb{C}_{+}:=\{s\in\mathbb{C}:\Re s>0\}.

Then U=LuU=\mathcal{L}u solves the following problem:

U(s)∈H1(Rm),s2c2U(s)−ΔU(s)=F(s).(2)\begin{aligned} U(s)&\in H^{1}(\mathbb{R}^{m}),\\ \frac{s^{2}}{c^{2}}U(s)-\Delta U(s)&=F(s). \tag*{(2)} \end{aligned}

The analysis of problem (2), together with the derivation of suitable bounds and its numerical resolution, will occupy most of this work. For now, we simply state that the problem is well posed and gives rise to a solution operator K(s)\mathrm{K}(s) such that

U=K(s)F,(3)U=\mathrm{K}(s)F, \tag*{(3)}

where K(s):L2(Ω)→H1(Rm)\mathrm{K}(s):L^{2}(\Omega)\to H^{1}(\mathbb{R}^{m}); its range is in fact contained in H2(Rm)H^{2}(\mathbb{R}^{m}). Moreover, K(s)\mathrm{K}(s) is analytic for s∈C+s\in\mathbb{C}_{+}.

Fix t>0t>0 and set f~(τ):=f(τ)χ[0,t](τ)\widetilde{f}(\tau):=f(\tau)\chi_{[0,t]}(\tau). By causality of the wave equation (the solution of (1) at time tt only depends on f∣[0,t]f|_{[0,t]}), u(t)u(t) is also the value at tt of the solution with datum f~\widetilde{f}, so that, formally,

u(t)=12πi∫μ−iRμ+iReζtK(ζ)F~(ζ) dζ=12πi∫μ−iRμ+iRK(ζ)[∫0∞f~(τ)eζ(t−τ) dτ]dζ=12πi∫μ−iRμ+iRK(ζ)[∫0tf(τ)eζ(t−τ) dτ]⏟y(t;ζ) dζ,μ>0.(4)\begin{aligned} u(t)&=\frac{1}{2\pi\mathrm{i}}\int_{\mu-\mathrm{i}\mathbb{R}}^{\mu+\mathrm{i}\mathbb{R}}e^{\zeta t}\mathrm{K}(\zeta)\widetilde{F}(\zeta)\,\mathrm{d}\zeta\\ &=\frac{1}{2\pi\mathrm{i}}\int_{\mu-\mathrm{i}\mathbb{R}}^{\mu+\mathrm{i}\mathbb{R}}\mathrm{K}(\zeta)\left[\int_{0}^{\infty}\widetilde{f}(\tau)e^{\zeta(t-\tau)}\,\mathrm{d}\tau\right]\mathrm{d}\zeta\\ &=\frac{1}{2\pi\mathrm{i}}\int_{\mu-\mathrm{i}\mathbb{R}}^{\mu+\mathrm{i}\mathbb{R}}\mathrm{K}(\zeta)\underbrace{\left[\int_{0}^{t}f(\tau)e^{\zeta(t-\tau)}\,\mathrm{d}\tau\right]}_{y(t;\zeta)}\,\mathrm{d}\zeta,\qquad\mu>0. \tag*{(4)} \end{aligned}

(The argument can be made rigorous in L2(Rm)L^{2}(\mathbb{R}^{m}): by Proposition 4.1 below, and since F~(ζ)=O(∣ζ∣−1)\widetilde{F}(\zeta)=O(|\zeta|^{-1}), the integral exists in the classical sense.) Here, y(t;ζ)y(t;\zeta), with ζ∈C+\zeta\in\mathbb{C}_{+}, is the solution of

y′−ζy=f,y(0)=0.(5)y'-\zeta y=f,\qquad y(0)=0. \tag*{(5)}

Multistep CQ methods are obtained by discretizing (4): define yκ(t;ζ)y^{\kappa}(t;\zeta) as the solution of the multistep recursion

1κ∑j=0rαjyκ(t−jκ;ζ)=∑j=0rβjf(t−jκ)+ζ∑j=0rβjyκ(t−jκ;ζ),yκ(t;ζ)=0, for t≤0.\frac{1}{\kappa}\sum_{j=0}^{r}\alpha_{j}y^{\kappa}(t-j\kappa;\zeta)=\sum_{j=0}^{r}\beta_{j}f(t-j\kappa)+\zeta\sum_{j=0}^{r}\beta_{j}y^{\kappa}(t-j\kappa;\zeta),\qquad y^{\kappa}(t;\zeta)=0,\ \text{for }t\leq0.

Applying the Laplace transform, with Yκ(s;ζ):=L(yκ(⋅;ζ))(s)Y^{\kappa}(s;\zeta):=\mathcal{L}(y^{\kappa}(\cdot;\zeta))(s) we have

Yκ(s;ζ)=1δ(e−κs)κ−ζF(s).Y^{\kappa}(s;\zeta)=\frac{1}{\frac{\delta(e^{-\kappa s})}{\kappa}-\zeta}F(s).

where

References

  1. [1]A. Bamberger and T. Ha Duong. Formulation variationnelle espace-temps pour le calcul par potentiel retardé de la diffraction d’une onde acoustique (i). Mathematical Methods in the Applied Sciences, 8(1):405–435, 1986.DOI
  2. [2]L. Banjai. Multistep and multistage convolution quadrature for the wave equation: algorithms and experiments. SIAM J. Sci. Comput., 32(5):2964–2994, 2010.DOI
  3. [3]L. Banjai. Implicit/explicit, BEM/FEM coupled scheme for acoustic waves with the wave equation in the second order formulation. Comput. Methods Appl. Math., 22(4):757–773, 2022.arxiv.org/abs/2203.00986
  4. [4]L. Banjai, C. Lubich, and F.-J. Sayas. Stable numerical coupling of exterior and interior problems for the wave equation. Numer. Math., 129(4):611–646, 2015.arxiv.org/abs/1309.2649
  5. [5]L. Banjai and S. Sauter. Rapid solution of the wave equation in unbounded domains. SIAM J. Numer. Anal., 47(1):227–249, 2008/09.DOI
  6. [6]L. Banjai and F.-J. Sayas. Integral Equation Methods for Evolutionary PDE: A Convolution Quadrature Approach, volume 59 of Springer Series in Computational Mathematics. Springer International Publishing, 2023.
  7. [7]J.-P. Bérenger. A perfectly matched layer for the absorption of electromagnetic waves. J. Comput. Phys., 114(2):185–200, 1994.DOI
  8. [8]D. R. Bergman. Computational Acoustics: Theory and Implementation. Wiley Series in Acoustics Noise and Vibration. John Wiley & Sons, 2018.
  9. [9]A. Bermúdez, L. Hervella-Nieto, A. Prieto, and R. Rodríguez. Perfectly matched layers for time-harmonic second order elliptic problems. Arch. Comput. Methods Eng., 17(1):77–107, 2010.DOI
  10. [10]T. Betcke, N. Salles, and W. Šmigaj. Overresolving in the Laplace domain for convolution quadrature methods. SIAM Journal on Scientific Computing, 39(1):A188–A213, 2017.arxiv.org/abs/1603.01761
  11. [11]P.-H. Cocquet and M. J. Gander. How large a shift is needed in the shifted Helmholtz preconditioner for its effective inversion by multigrid? SIAM J. Sci. Comput., 39(2):A438–A478, 2017.DOI
  12. [12]D. Colton and R. Kress. Inverse Acoustic and Electromagnetic Scattering Theory, volume 93 of Applied Mathematical Sciences. Springer, New York, third edition, 2013.
  13. [13]M. Costabel. On the limit Sobolev regularity for Dirichlet and Neumann problems on Lipschitz domains. Math. Nachr., 292(10):2165–2173, 2019.DOI
  14. [14]V. Domínguez and M. Ganesh. Analysis and application of an overlapped FEM-BEM for wave propagation in unbounded and heterogeneous media. Appl. Numer. Math., 171:76–105, 2022.
  15. [15]V. Domínguez, M. Ganesh, and F. Le Louër. Efficient simulation of time-domain acoustic wave propagation in unbounded and penetrable media. In D. de la Prida, J. Ramis, and M. Machimbarrena, editors, Proceedings of Forum Acusticum - Euronnoise 2025: 11th Convention of the European Acoustics Association, pages 5573–5580. European Acoustics Association, 2025.DOI
  16. [16]V. Domínguez, M. Ganesh, and F.-J. Sayas. An overlapping decomposition framework for wave propagation in heterogeneous and unbounded media: formulation, analysis, algorithm, and simulation. J. Comput. Phys., 403:109052, 2020.DOI
  17. [17]V. Domínguez, S.L. Lu, and F.-J. Sayas. A fully discrete Calderón calculus for two dimensional time harmonic waves. Int. J. Numer. Anal. Model., 11(2):332–345, 2014.
  18. [18]V. Domínguez, S.L. Lu, and F.-J. Sayas. A Nyström flavored Calderón calculus of order three for two dimensional waves, time-harmonic and transient. Comput. Math. Appl., 67(1):217–236, 2014.DOI
  19. [19]S. Eberle. FEM-BEM coupling for the thermoelastic wave equation with transparent boundary conditions in 3D. Z. Angew. Math. Phys., 73(4):Paper No. 163, 27, 2022.DOI
  20. [20]B. Engquist and A. Majda. Absorbing boundary conditions for the numerical simulation of waves. Math. Comp., 31(139):629–651, 1977.
  21. [21]Y. A. Erlangga, C. W. Oosterlee, and C. Vuik. A novel multigrid based preconditioner for heterogeneous Helmholtz problems. SIAM J. Sci. Comput., 27(4):1471–1492, 2006.
  22. [22]M. J. Gander, I. G. Graham, and E. A. Spence. Applying GMRES to the Helmholtz equation with shifted Laplacian preconditioning: what is the largest shift for which wavenumber-independent convergence is guaranteed? Numer. Math., 131(3):567–614, 2015.DOI
  23. [23]M. Ganesh and F. Le Louër. A high-order algorithm for time-domain scattering in three dimensions. Adv. Comput. Math., 49(4):Paper No. 46, 43, 2023.DOI
  24. [24]C. Geuzaine and J.-F. Remacle. Gmsh: a 3-D finite element mesh generator with built-in pre- and post-processing facilities. Internat. J. Numer. Methods Engrg., 79(11):1309–1331, 2009.DOI
  25. [25]D. Givoli. Numerical Methods for Problems in Infinite Domains, volume 33 of Studies in Applied Mechanics. Elsevier, Amsterdam, 1992.
  26. [26]P. Grisvard. Elliptic Problems in Nonsmooth Domains. Pitman Advanced Publishing Program, Boston, 1985.
  27. [27]E. Hairer, C. Lubich, and M. Schlichte. Fast numerical solution of nonlinear Volterra convolution equations. SIAM J. Sci. Statist. Comput., 6(3):532–541, 1985.DOI
  28. [28]M.E. Hassell and F.-J. Sayas. A fully discrete BEM-FEM scheme for transient acoustic waves. Comput. Methods Appl. Mech. Engrg., 309:106–130, 2016.DOI
  29. [29]A. B. Hölter, M. Lemke, S. Weinzierl, and L. Stein. A tutorial on non-reflective characteristic boundary conditions for direct and adjoint time-domain acoustic simulations. Journal of Theoretical and Computational Acoustics, 2026.DOI
  30. [30]R. Kress. Linear Integral Equations, volume 82 of Applied Mathematical Sciences. Springer, New York, third edition, 2014.
  31. [31]F. Laliena and F.-J. Sayas. Theoretical aspects of the application of convolution quadrature to scattering problems. Numerische Mathematik, 112(4):637–678, 2009.DOI
  32. [32]A. Lechleiter and P. Monk. The time-domain Lippmann-Schwinger equation and convolution quadrature. Numer. Methods Partial Differential Equations, 31(2):517–540, 2015.arxiv.org/abs/1407.7563
  33. [33]C. Lubich. Convolution quadrature and discretized operational calculus. I. Numerische Mathematik, 52(2):129–145, 1988.DOI
  34. [34]C. Lubich. Convolution quadrature and discretized operational calculus. II. Numerische Mathematik, 52(4):413–425, 1988.DOI
  35. [35]C. Lubich. On the multistep time discretization of linear initial-boundary value problems and their boundary integral equations. Numer. Math., 67(3):365–389, 1994.
  36. [36]W. McLean. Strongly elliptic systems and boundary integral equations. Cambridge University Press, 2000.
  37. [37]J. M. Melenk and S. Sauter. Wavenumber explicit convergence analysis for Galerkin discretizations of the Helmholtz equation. SIAM J. Numer. Anal., 49(3):1210–1243, 2011.DOI
  38. [38]J.M. Melenk and A. Rieder. Runge-Kutta convolution quadrature and FEM-BEM coupling for the time-dependent linear Schrödinger equation. J. Integral Equations Appl., 29(1):189–250, 2017.DOI
  39. [39]J. Saranen and G. Vainikko. Periodic integral and pseudodifferential equations with numerical approximation. Springer Monographs in Mathematics. Springer-Verlag, Berlin, 2002.DOI
  40. [40]F.-J. Sayas. Retarded potentials and time domain boundary integral equations, volume 50 of Springer Series in Computational Mathematics. Springer, Cham, 2016.DOI

Paper details

vlnc:W_ynuh7ay8
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Claude (Anthropic)

View statement
Looking for related papers…
Contents