Efficient time-domain scattering synthesis via frequency-domain singularity subtraction
Oscar P. Bruno, Manuel A. Santana · 2025
Read Efficient time-domain scattering synthesis via frequency-domain singularity subtraction by Oscar P. Bruno, Manuel A. Santana (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-2505-06189.txt.
Introduction
Bruno and Santana introduce in 2025 a faster way to compute how waves scatter in time from trapping objects. Transform methods give exact, clean results by solving the wave equation at a set of frequencies. Traps cause near real resonances that slow that back transform badly.
Their fix subtracts the singular part in the frequency domain before transforming. The removed part is then added back by a cheap numeric step plus a large time expansion. All needed resonances and weights come from rational fits to solutions at real frequencies. An adaptive search makes sure no key resonance is missed, which keeps the method sound.
Catalog record
| CREATOR | Oscar P. Bruno, Manuel A. Santana |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 136 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-06189
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Efficient time-domain scattering synthesis via frequency-domain singularity subtraction Oscar P. Bruno, Manuel A. Santana Abstract Fourier transform-based methods enable accurate, dispersion-free simulations of time-domain scattering problems by evaluating solutions to the Helmholtz equation at a discrete set of frequencies sufficient to approximate the inverse Fourier transform. However, in the case of scattering by trapping obstacles, the Helmholtz solution exhibits nearly-real complex resonances—which significantly slows the convergence of numerical inverse transform. To address this difficulty this paper introduces a frequency-domain singularity subtraction technique that regularizes the integrand of the inverse transform and efficiently computes the singularity contribution via a combination of a straightforward and inexpensive numerical technique together with a large-time asymptotic expansion. Crucially, all relevant complex resonances and their residues are determined via rational approximation of integral equation solutions at real frequencies . An adaptive algorithm is employed to ensure that all relevant complex resonances are properly identified. 1 Introduction Recently developed “Frequency-Time Hybrid” (FTH) Fourier-transform-based methods [ 4 , 5 ] offer accurate and efficient numerical techniques for solving exterior time-domain wave scattering problems; corresponding interior problems can be tackled by such approaches as well [ 16 , 52 ] . These algorithms solve the associated Helmholtz problems at a discrete set of frequencies—typically using layer potential formulations—combined with specialized techniques such as “windowing-and-recentering” and high-frequency integration, enabling effective reconstruction of the time-domain solution via inverse Fourier transformation. This approach offers several advantages: it produces essentially dispersion-free solutions, it enables straightforward parallelization in both space and time, and it can efficiently accommodate incident fields that persist indefinitely. However, the performance of these methods is severely impacted by the presence of trapping geometries. In such cases, the Helmholtz solution exhibits nearly real complex resonances (poles), which cause extremely slow convergence in the numerical evaluation of the inverse Fourier transform. This work addresses that challenge through a frequency-domain singularity subtraction technique that regularizes the integrand and naturally transitions to a large-time asymptotic expansion expressed in terms of the complex resonances. These resonances and their residues are evaluated efficiently using only real-frequency data, by employing a novel Incidence Excitation (IE) adaptive algorithm that relies on rational approximation in the frequency variable. The resulting overall FTH-SS (Singularity-Subtraction FTH) approach enables accurate and efficient time-domain scattering simulations, for arbitrarily long time, even for highly-trapping scattering structures. Other methods, such as the Fourier transform methods [ 23 , 43 ] and the well known Convolution Quadrature (CQ) method [ 39 , 7 , 11 ] have been presented that, like the FTH methods, rely on transformation into the frequency domain. A detailed discussion of certain advantages offered by the FTH algorithms vis-à-vis other frequency-time approaches (relating, in the case of the CQ method, to time dispersion and existence of an infinite tail, and concerning increasing cost per time-step as time grows in previous Fourier transform methods), can be found in [ 4 ] and will therefore not otherwise be discussed here. The difficulties faced by the FTH method in trapping configurations can be traced to the frequency-dependent behavior of the scattering solutions. Indeed, for obstacles which are strongly trapping (see e.g. [ 3 , 33 ] for more precise definitions of the concept of trapping scatterer), the frequency domain scattering solutions typically exhibit nearly real poles as functions of frequency [ 9 ] . These complex resonances (all of which are located in the lower half-plane [ 58 ] ) manifest themselves as sharp, spike-like features along the real frequency axis [ 10 , 53 , 33 , 20 ] , whose resolution requires a dense sampling of frequency points— thereby making the numerical inversion of the Fourier transform prohibitively expensive. An additional difficulty associated with the evaluation of time domain scattering by trapping obstacles relates to the slow temporal decay of the scattered field in trapping regions. This phenomenon has been the focus of extensive analytical work, which connects the decay rate to the geometric properties of the scatterer [ 36 , 35 , 31 , 48 , 49 , 57 , 3 ] . The slow decay associated with trapping poses challenges for traditional time-domain methods such as finite difference and finite element based techniques, which require fine spatial and temporal resolution to control dispersion errors over long simulation times. For long time evaluation of the scattered field a popular alternative is to express the scattered field as an asymptotic “singularity expansion” [ 8 ] in terms of the complex resonances. Such expansions can be formally derived on the basis of evaluation of the inverse Fourier-transform via contour deformation, where for late enough times the contribution from the poles dominate. Much work has gone into proving the validity of the aforementioned singularity expansions [ 57 , 59 , 36 ] , although a rigorous justification for their validity in the trapping case has remained elusive. Nevertheless, singularity expansions have been widely used in practice to model late-time scattering phenomena [ 8 , 64 , 46 , 29 , 30 , 45 , 34 ] , although not without criticism of the formalism sometimes used [ 30 , 21 , 54 ] . The numerical examples presented in this paper—including problems involving scattering by highly-trapping obstacles—provide strong evidence for the validity and high accuracy of the singularity expansion in the asymptotic regime, regardless of the trapping character of the scatterer. As mentioned above, this contribution proposes a method for evaluation of the slowly decaying fields scattered by trapping obstacles, which relies on a certain frequency-domain singularity subtraction methodology. Subtraction of complex resonances and their residues near the real axis regularizes the integrand of the inverse Fourier-transform leading to greatly increased convergence of numerical quadrature rules; the resonance contributions can then be easily evaluated and re-incorporated to obtain the correct time-domain solution. For efficiency, the contributions from the subtracted complex resonances are computed for large times using an asymptotic numerical algorithm based on expressions that resemble the singularity expansion mentioned above (Section 4.2 ). Importantly, the singularity subtraction method does not require the validity of the singularity expansion; in any case, numerical experiments provided in this paper suggest that the singularity expansion is asymptotically valid independently of the trapping character of the scattering obstacles considered; see Section 4.3 . As noted in Remark 6 , the efficiency of the proposed method relies critically on a certain “Incident-Excitation” (IE) algorithm. In a modified form of the AAA rational-approximation approach for resonance evaluation [ 15 ] , the IE algorithm identifies the complex resonances excited by a given incident field as poles of rational approximants to the corresponding integral-equation densities, evaluated at real frequency values. Crucially, the rational approximants used in the IE method also enable the computation of field values and residues without requiring additional (and costly) boundary-integral inversions beyond those already required by the IE algorithm. This stands in clear contrast with previous methods [ 12 , 47 , 6 , 2 , 26 , 28 , 15 ] for evaluating complex resonances, which, in the present setting, necessitate the evaluation of integral resolvents over significantly larger sets of real and complex frequencies. Soon after a preliminary version [ 14 ] of the present work was made available, an alternative strategy was proposed in the