Stable rational approximations for parabolic equation methods
Adith Ramamurti, Joseph F. Lingevitch, Jonathan C. Lighthall, Michael D. Collins · 2025
Read Stable rational approximations for parabolic equation methods by Adith Ramamurti, Joseph F. Lingevitch, Jonathan C. Lighthall, Michael D. Collins (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-2510-18622.txt.
Introduction
Adith Ramamurti and his three co-authors improve wave simulation math. The setting is layered media. One case is an elastic sheet over thick fluid. Parabolic equation methods march a wave forward step by step. Each step applies operators raised to fractional powers. Those operators are approximated by ratios of polynomials. The task is delicate. Propagating parts of the wave must survive. Fading parts must be removed. Older schemes sometimes went unstable on layered cases.
The team builds the approximations with a recent adaptive algorithm. New simulations matched reference solutions closely. Errors in transmission loss were as low as the standard method's. Often they were lower. The approach also unlocks a marching scheme called split-step Padé. That scheme now works for fluid-elastic problems. Ocean and seismic acoustics gain a wider toolbox.
Catalog record
| CREATOR | Adith Ramamurti, Joseph F. Lingevitch, Jonathan C. Lighthall, Michael D. Collins |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 59 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-18622
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Stable rational approximations for parabolic equation methods Adith Ramamurti, Joseph F. Lingevitch, Jonathan C. Lighthall, Michael D. Collins Abstract Modern parabolic equation (PE) methods for wave propagation rely on application of a variety of fractional-powered differential operators. Rational approximations of these operators need to properly map their spectra onto the complex plane, accurately handling propagating modes while annihilating evanescent ones. Standard approaches for stable and accurate rational approximations include rotating the branch cut of the operators or imposing stability constraint equations, and have yielded accurate results for wave propagation in a variety of fluid, elastic, and fluid-elastic waveguides. The stability constraint method, however, does not yield operators that are stable for all fluid-elastic waveguides, and a recent study of waveguides comprised of a thin elastic layer overlaying a thick fluid layer revealed instabilities in the approximations derived from rotated operators. In this paper, we demonstrate the applicability of a different rational approximation method, the recently-developed adaptive Antoulas-Anderson (AAA) algorithm, to simulations of wave propagation using the fluid-elastic parabolic equation. We find that simulations using operators approximated using the AAA algorithm provide excellent agreement with reference solutions, with errors in transmission loss comparable to, and often less than, that of simulations using the rotated operator method. In addition, we find that the AAA algorithm allows for the application of the split-step Padé method to fluid-elastic waveguides, which yields a large gain in computational efficiency. \catchline keywords Seismoacoustic propagation; parabolic equations; rational approximations. 1 Introduction Parabolic wave equations (one-way wave equations) are an approximate method used to obtain accurate and efficient solutions for many problems in which a solution of the full wave equation is impractical [ 1 , 2 ] . One common application of parabolic equation (PE) methods is to wave propagation in laterally-varying fluid and elastic media. In a two-dimensional layered medium parameterized by range ( r ) and depth ( z ), the parabolic (one-way) wave equation takes the form \frac{\partial u}{\partial r}=ik_{0}\left(-1+\sqrt{1+q}\right)u\,. (1) for the field u=pe^{-ik_{0}r} , where p is a scalar (fluid) or vector (elastic) field, k_{0} is the reference wavenumber, and q is a depth operator involving depth derivatives [ 2 , 3 ] . One approach for numerical solutions of this equation involves approximating the square root operator with either a Taylor series or rational approximation in q , discretizing the operator q , and solving using a Crank-Nicolson marching scheme. Another common approach, known as the split-step Padé method, involves approximating the exponential operator in the analytic solution, u(r+\Delta r)=\exp\left\{ik_{0}\Delta r\left(-1+\sqrt{1+q}\right)\right\}u(r)\,, (2) with a rational approximation in q . The operator q is then discretized on a depth grid, as in the Crank-Nicolson solution, and solved using a marching scheme [ 4 ] . The latter approach allows for significantly larger range steps, and therefore significantly increased computational efficiency, as the k_{0}\Delta r term is included in the coefficients of the rational approximation of the propagation operator. The fields are typically initialized at a finite distance r_{0} using a self-starter method [ 5 ] , which involves application of operators such as (1-iq)^{2}(1+q)^{-1/4}\exp\left\{ik_{0}r_{0}\left(-1+\sqrt{1+q}\right)\right\}\,. (3) Accurate handling of variation of the environment in the transverse direction r is done by either enforcing energy conservation [ 6 ] , the formulation of which utilizes operators of the form (1+q)^{\pm 1/4}\,, (4) or by a single-scattering calculation [ 7 , 8 ] , which uses the square root operator. Approximations of these operators, however, are not straightforward. Early attempts at solving the PE numerically approximated the square root with a Taylor expansion [ 2 ] . This, however, yields unstable behavior as the evanescent spectrum ( q<-1 ) cannot be properly represented by a real-valued polynomial. An improved approach involves using rational approximations of the square root [ 9 , 10 , 11 , 12 , 13 ] , which, when slightly modified using stability constraint equations, annihilates the evanescent spectrum [ 12 , 4 ] . This approach for stabilizing these operators, however, fails for certain cases of wave propagation in fluid and elastic waveguides [ 11 , 12 , 14 ] . To attempt to rectify this issue, a new method using rotations of the branch cut of the square root operator was introduced by Milinazzo, et al. [ 14 ] , and expanded upon by Lingevitch and Collins [ 15 ] for the exponential operator. The branch cut of the square root is rotated such that the eigenvalues of evanescent modes are forced into the upper half of the complex plane. Since its introduction, this rotated operator Padé (ROP) method has become a common method for approximating the square root operator in PE simulations. The ROP method works extremely well for the square root operator, but not as well for the exponential operator. As a result, the exponential operator is typically stabilized using a constraint equation, which works excellently for the pure-fluid case, but is unstable for some cases of propagation modeling in elastic and coupled fluid-elastic waveguides; the significant boost in computational efficiency from the split-step Padé method is therefore not available for all PE-based simulations. Nevertheless, PE simulations using the Crank-Nicolson approach are still significantly more computationally efficient than computing full-field solutions to elastic and fluid-elastic wave equations. In Appendices A and B , we describe in detail these standard methods of calculating coefficients for and stabilizing the rational approximations of the generalized PE operator, f(q)=\exp\Big\{i\sigma\left(-1+\sqrt{1+q}\right)+\delta\ln(1+q)+\nu\ln(1+cq)\Big\}\,, (5) f(q)\approx g(q)\equiv 1+\sum_{j=1}^{N}\frac{\alpha_{j}q}{1+\mu_{j}q}=\prod_{j=1}^{N}\frac{1+\gamma_{j}q}{1+\mu_{j}q}\,. (6) Recent work has extended the parabolic equation method to waveguides where a thin elastic layer overlays a fluid [ 16 , 17 ] . As initially pointed out by Wetton and Brooke [ 11 ] , the eigenvalues of the field propagating in thin elastic layers sometimes fall far below the real line on the complex plane, and can cause instabilities if not properly mapped to the upper half-plane. To stabilize these simulations (i.e. map these eigenvalues to the upper half-plane), the rotation angle of the branch cut of the square root operator must be increased to, and sometimes past, 90^{\circ} . While this works well at low frequencies, large rotations have the side effect of mapping the wavenumbers of some propagating modes in waveguides with thick fluid layers into the lower half-plane, causing instability at higher frequencies; these propagating modes, which are high-angle, grow exponentially [ 18 ] . This observation means that propagation modeling of these kinds of waveguides may be unstable (or at best inaccurate) at higher frequencies when using the ROP method to approximate the PE operators. In the decades since the previously discussed approaches were first applied to parabolic equation based wave propagation simulations and adopted as the standard methods, there have been many developments in the field of rational approximations. Among many others, these include vector fitting [ 19 , 20 , 21 ] and RKFIT [ 22 , 23 ] (see Sec. 11 of Nakatsukasa, et al. [ 24 ] for additional discussion and references). A significant advance in this area was the development of the adaptive Antoulas-Anderson (AAA) algorithm [ 24 ] , which has had significant impact in a variety of fields [ 25 ] . In this paper, we demonstrate the applicability of the AAA algorithm for rational approximation