A meshless method for computational electromagnetics with improved dispersion properties

Andrej Kolar-Požun, Gregor Kosec · 2025

Read A meshless method for computational electromagnetics with improved dispersion properties by Andrej Kolar-Požun, Gregor Kosec (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-2508-18205.txt.

Introduction

Andrej Kolar-Pozun and Gregor Kosec present a meshless method for computing electric and magnetic fields, in a 2025 paper. The most common method in this field works on a grid and is loved for its simplicity. The authors ask how that grid method can be freed from the grid by using local fits based on radial functions. Their goal is a method that stays simple but bends to odd shapes more easily.

Both routes they try stay fully step-by-step in time. They settle to the right answer when the right smoothing terms are added. A larger local stencil, which uses more nearby points, cuts the wave error known as dispersion. Their methods also show less direction bias in that error than the grid method. Tests back each of these claims.

Catalog record

CREATORAndrej Kolar-Požun, Gregor Kosec
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-18205

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

A meshless method for computational electromagnetics with improved dispersion properties Andrej Kolar-Požun, Gregor Kosec Abstract The finite difference time domain method is one of the most popular methods in computational electromagnetics. This work considers two possible ways of generalising it to a meshless setting by employing local radial basis function interpolation. The resulting methods turn out to be unstable, but can be stabilised by adding properly chosen hyperviscosity terms to the update equations. We demonstrate that the proposed meshless methods are convergent and can enjoy a decreased dispersion anisotropy compared to the finite difference time domain method. 1 Introduction Accurate simulations of electromagnetic waves are of great interest nowadays with one of the key drivers at the moment being the field of telecommunications, as the carrier frequencies for future communication protocols (e.g. 6G) are expected to be higher, demanding physically accurate simulations that are able to describe increasingly finer geometrical objects, such as antennas. In order to have such a description, we must solve the equations that describe the physics of electromagnetism - Maxwell’s equations. Maxwell’s equations consist of a coupled system of Partial Differential Equations (PDEs) that are too complex to be solved analytically in realistic setups and a numerical procedure must be invoked. Most of the commonly used methods in computational electromagnetics have been around for decades with several landmark papers appearing in the 1960s. In 1966, for instance, the well known Finite Difference Time Domain (FDTD) method was proposed by Yee [ 35 ] . In the same year, the Method of Moments (MoM) was popularised by Richmond [ 18 ] . The Finite Element Method (FEM), which can be traced all the way to the work of Courant in 1943 [ 4 ] , has also been applied to electromagnetics by Silvester in 1969 [ 24 ] . A comprehensive review and comparison of these can be found in any of the major textbooks on computational electromagnetics, such as [ 5 ] . Among the listed methods, the FDTD is perhaps the simplest to implement and therefore a common starting point when dealing with a given problem. Its simplicity lies in the fact that the FDTD is, in essence, just the application of the usual Finite Difference Method (FDM) on a staggered grid. This approach, however, has its limitations - it is limited to regular, grid-like geometries, making it difficult to efficiently satisfy our requirement of describing fine geometrical objects. Additionally, the underlying grid makes the numerical dispersion relation anisotropic, even if the propagation medium is isotropic. Our goal in this paper is to remedy both of these shortcomings of the FDTD by introducing a method that in a certain sense generalises it to irregular node layouts, providing a greatly increased geometrical flexibility. For that purpose, we will adopt meshless methods, which provide the means of approximation on scattered nodes without any additional structure. Meshless or meshfree methods encompass a variety of different methods, some of which are described in, for example [ 9 ] . For our work we opt for methods based on Radial Basis Functions [ 3 ] (RBFs). RBFs have first appeared in the context of scattered data interpolation in the 1970s [ 13 ] and differential operator approximation in the 1990s [ 14 ] . In the 2000s a local RBF-based operator approximation method was introduced [ 30 ] , which directly generalises the FDM, sparking an interest from the wider scientific community and resulting in a culmination of research in this field, both theoretical and experimental [ 31 , 2 ] . Having the machinery of meshless methods, a generalisation of the FDTD is then straightforward - we can simply replace the Finite Differences (FD) appearing in the FDTD by their appropriate meshless variant. We will see that there are two natural ways of doing so, resulting in two different generalisations to consider. At this point, a natural question could arise - why should one bother with a meshless variant of the FDTD, if FEM already solves Maxwell’s equations (albeit in weak form) and is known to be geometrically flexible? Our motivation for considering meshless methods over FEM is twofold. Firstly, unlike meshless methods, FEM requires an underlying triangulation, a generation of which can be tedious and cannot be fully automated, especially in higher dimensions [ 11 ] . Secondly, FEM is a fairly complex method - the attractive point of the FDTD is its simplicity and our aim is to preserve this property in the geometrically flexible version. Simplicity of the method can be of great importance when dealing with problems of increasing complexity - not only with regards to the problem geometry but also, for example, inclusion of non-homogenous or even non-linear materials. Furthermore it is also easier to build upon from an implementation perspective, in particular when it comes to parallelisation [ 5 ] . Indeed, we will see that, just like the FDTD, our proposed methods will be fully explicit. It should be mentioned that computational electromagnetics has already been studied in a meshless context. Most of existing work has been focused on weak-form meshless methods [ 20 , 7 , 6 ] , although some of the research has been done also using the Method of Fundamental Solution (MFS) [ 34 ] , which is closely related to the previously mentioned MoM method. As mentioned, however, we are interested in strong-form methods, which resemble FDTD. The latter have also been already investigated in a meshless context [ 36 , 22 , 21 ] by replacing the spatial derivatives appearing in the FDTD by their meshless variants in a similar way as we intend to. However, our proposed method differs from these by two key points, which all of the listed work has in common. Firstly, the listed work staggers the \vec{E} and \vec{H} fields on different nodes, as in the usual FDTD. This staggering can become non-trivial in a scattered setting and would require, for example, a Voronoi tessellation, which some would argue is not truly a meshless method. The second key point is much more important - despite the methods being meshless in their formulations, they seemed to have only been tested on regular node layouts, where the methods can be much better behaved, particularly in terms of stability. In our approach we will abandon the staggering of the \vec{E} and \vec{H} fields and one of our goals is to demonstrate that our methods work also on scattered, irregular nodes. The paper is structured as follows: In Section 2, the governing equations of electromagnetism are listed. In Section 3, the FDTD is briefly described, along with its previously mentioned shortcomings. In Section 4, RBF interpolation-based meshless methods are introduced. In Section 5, our proposed methods are presented and analysed. In Section 6, the dispersive properties of the proposed methods are also analysed and compared to the FDTD. We then conclude with Section 7. 2 Governing equations In electromagnetics, the relevant quantities are the electric and magnetic fields, which we denote by \vec{E} and \vec{H} , respectively. They are defined on some region of interest \Omega\subset\mathbb{R}^{3} and are generally time-dependent: \vec{E},\vec{H}:\Omega\times\mathbb{R}_{\geq 0}\to\mathbb{R}^{3}. (1) The two fields satisfy Maxwell’s curl equations, which in vacuum read: \displaystyle\frac{\partial\vec{H}}{\partial

Read the complete text