1. Introduction
The finite-difference time-domain (FDTD) method (Yee’s algorithm) is a widely used numerical technique for solving time-domain Maxwell’s equations of electrodynamics [
1,
2,
3]. Maxwell’s equations are discretized on a space and time-staggered rectangular grid (Yee’s lattice) using centered difference approximations. Each field component is sampled and evaluated at a particular space position, and the magnetic and electric fields are obtained at different instants of time delayed by half the sampling time step. As an explicit numerical method, the FDTD method is subject to the Courant–Friedrichs–Lewy (CFL) stability condition [
2,
3].
The accuracy and efficiency of the FDTD method can be significantly improved through a subgridding technique, where the grid is selectively refined only in certain regions of the computational domain. In the last few decades, there have been numerous investigations of subgridding and adaptive mesh refinement (AMR) algorithms [
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28,
29]. Subgridding methods reduce the computational cost and increase the computational efficiency over methods using uniform meshes when the region of interest occupies a small portion of the domain. The Huygens subgridding (HSG) technique [
26,
27,
28,
29] can reach an arbitrarily large ratio of spatial steps as large as 99. A main weakness of the subgridding technique is the late-time instability problem. Numerical studies of this issue performed in the past, such as [
4,
9,
29], indicated that the stability of the subgridding algorithm is sensitive to the interpolation method and the choice of the interpolated fields. In [
23], the spatial subgridding and temporal subgridding are handled separately in order to obtain a better stability. It has been pointed out in [
30] that the interpolation of fields in the temporal subgridding region is the main source of the late-time instability.
In this paper, we propose a new iteration-based method to stabilize the temporal subgridding FDTD algorithm. We apply separated spatial and temporal mesh interfaces so that the stability of spatial subgridding and temporal subgridding can be investigated separately. Different from other subgridding methods that focus on the interpolation or special interface treatment between coarse and fine meshes, we iterate the updating equations on the coarse–fine mesh interface a few times to obtain a stable solution for a long-term simulation. Our method is an alternative approach to improve the stability without modifying the interpolation formulas. To test the stability, we computed the eigenvalues of the system of updating equations. The maximum magnitude of the eigenvalues corresponds to the amplification factor of the system. The amplification factor of 1 or less corresponds to a stable solution. We also perform long-term simulations for a million steps to verify the stability of the solution.
The paper is organized as follows: In
Section 2, we review a couple of conventional 2D subgridding FDTD methods. In
Section 3, we present the proposed iteration-based temporal subgridding algorithm. Numerical examples to verify the stability of the proposed method and an application to the Through-the-Wall Radar (TWR) imaging are presented in
Section 4.
2. The FDTD Method and Subgridding Techniques
Consider the two-dimensional Maxwell’s equations (Faraday’s and Ampère’s laws) in the transverse-electric mode with respect to
z (
) [
3]
where
E and
H represent the electric and magnetic fields, respectively.
F/m and
H/m are the vacuum permittivity and permeability, respectively. The discretized updating equations using the Yee FDTD method can be written as [
3]
where
,
and
are the grid cell sizes in space and time, respectively.
There are two ways to handle the interface between coarse and fine meshes in the subgridding method, based on the choice of ghost values (either magnetic or electric fields). Here, we consider a coarse–fine mesh ratio of 2:1, as shown in
Figure 1.
2.1. Temporal Subgridding Algorithm Using Ghost H Fields
Referring to
Figure 2, the temporal subgridding algorithm with ghost
fields in a 2D
mode consists of the following steps:
Update the magnetic field on the coarse mesh from time to ;
Update the electric fields and on the coarse mesh from time n to ;
Interpolate the ghost magnetic field using nearby values on the coarse mesh;
Update the magnetic field on the fine mesh from time to n;
Update the electric fields and on the fine mesh from time to ;
Interpolate the ghost magnetic field using nearby values on the coarse mesh;
Update the magnetic field on the fine mesh from time n to ;
Update the electric fields and on the fine mesh from time to ;
Replace the magnetic field on the coarse mesh with the fine mesh magnetic field in the region where fine mesh overlaps coarse mesh: ;
Recalculate the electric fields and on the coarse mesh, as in step 2.
The above steps are applied recursively at each refinement level. The ghost magnetic fields are obtained by interpolation using neighboring values. All electric fields in the fine-mesh are computed using the standard Yee FDTD algorithm. On the coarse grid, the field is updated from time to , followed by the and fields, which update from time n to . In the meantime, the fine mesh solutions are updated twice: the first update of is from to n, followed by the updates of and from to ; the second update of is from n to , followed by the updates of and from to . On the coarse mesh, some magnetic fields have been updated in step 9, so we recalculate the electric fields affected by this change on the coarse–fine mesh interface in step 10.
In order to calculate the electric field on the boundary of the fine mesh, we need the coarse magnetic field to recover the missing fine region magnetic fields that are ghost values. In step 3, the ghost magnetic fields are borrowed from the coarse mesh, similar to the method in [
24]:
In step 6, the ghost magnetic fields are interpolated from the coarse mesh [
24]:
2.2. Temporal Subgridding Algorithm Using Ghost E Fields
The temporal subgridding algorithm using ghost
E fields is similar to the case using ghost
H fields, but slightly more complicated since there are two electric field components
and
in the
mode. In this case, the magnetic fields are defined at step
n and the electric fields are located at step
. Referring to
Figure 3, the algorithm can be summarized as follows:
Update the electric fields and on the coarse mesh from time to ;
Update the magnetic field on the coarse mesh from time n to ;
Interpolate the ghost electric fields and using nearby values on the coarse mesh;
Update the electric fields and on the fine mesh from time to n;
Update the magnetic field on the fine mesh from time to ;
Interpolate the ghost electric fields ( and ) using nearby values on the coarse mesh;
Update the electric fields and on the fine mesh from time n to ;
Update the magnetic field on the fine mesh from time to ;
Replace the electric fields and with the corresponding electric fields and in the region where the fine grid overlaps a coarse grid: and ;
Recalculate the magnetic field on the coarse mesh, as in step 2.
The above steps are applied recursively at each refinement level. In this case, the coarse grid electric fields
and
are updated from
to
, followed by
updates from
n to
. The fine mesh solutions are updated as follows:
and
update from
to
n, followed by the
update from
to
, then
and
update from
n to
, followed by the
update from
to
. In step 9, we replace the electric field on the coarse grid by the time-averaged electric fields on the fine grid at the coarse–fine mesh interface and recalculate the corresponding magnetic fields on the coarse mesh in step 10. The last step is to update the magnetic field by using the newly updated electric fields in step 9. In step 3, we interpolate the ghost electric fields from coarse mesh, similar to the interpolation in [
24]:
In step 6, we interpolate the ghost electric field by using the corresponding values on the coarse mesh [
24]:
3. Iteration-Based Temporal Subgridding FDTD Algorithm
In this section, we present a new temporal subgridding FDTD algorithm based on the use of an iteration, aiming at overcoming the late-time instability problem of the conventional subgridding method. As we have observed from the previous section, the conventional temporal subgridding with a coarse–fine mesh ratio of 2:1 consists of the update of the solution on the coarse mesh by one
time step and two consecutive
updates of the fine mesh solutions. The conventional temporal subgridding method is illustrated in the flow chart (a) of
Figure 4, and the iteration-based subgridding technique is illustrated in the flow chart (b) of
Figure 4.
The iteration-based subgridding FDTD algorithm with a ghost E field is summarized in the following steps:
Update and on the coarse mesh from time to ;
Update on the coarse mesh from time n to ;
Interpolate the ghost electric fields and using the nearby fields on the coarse mesh;
Update and on the fine mesh from time to n;
Update on the fine mesh from time to ;
Recalculate
on the coarse–fine mesh interface at time
n using the fine mesh values
at time
and
[
24]
Repeat steps 1 to 6 several times on the coarse–fine mesh interface;
Interpolate ghost electric fields and using the nearby coarse mesh values;
Update and on the fine mesh from time n to ;
Update on the fine mesh from time to ;
Replace the coarse grid electric fields and with the corresponding fine grid values and in the region where the fine grid overlaps the coarse grid: and ;
Recalculate on the coarse mesh, as in step 2.
In step 6, the magnetic field on the coarse mesh is recalculated so it can be used to update the electric field on the coarse mesh, as in step 1. Therefore, the update steps 1–6 can be viewed as an implicit system of updating equations. A classic algorithm to solve implicit problems is by using an iterative method, and it motivates us to develop the iteration-based method.
A key step of the iteration-based subgridding algorithm is step 7, where the iteration is carried out for N times. Through numerical eigenvalue tests, we study the correlation between N and the amplification factor (the largest magnitude of the eigenvalues of the system of updating equations). We found that when the number of iterations increases, the amplification factor decreases. When , the amplification factor becomes 1, up to the machine round-off error.
Note that the iteration is conducted only for the electric fields on the coarse–fine mesh interface and its neighboring magnetic fields, so it does not significantly increase the computational cost.
Using the proposed iteration-based temporal subgridding together with a regular spatial subgridding algorithm, we can obtain an iteration-based spatial–temporal subgridding method.
4. Numerical Tests
4.1. Tests for Temporal Subgridding Method
We first test the temporal subgridding methods. In the first example, we simulate a cylindrical wave in the free space. The incident source of a continuous wave with a wavelength of
is placed in the center of the domain. The domain is surrounded by PML boundaries [
3]. The temporal fine mesh is a
square zone from grid points
to
. The numerical result is shown in
Figure 5. The small rectangular box shows the boundary of the fine mesh. We can observe that there is no significant reflection or scattering after the wave crosses the interface of coarse and fine meshes.
The second test is a convergent test of the proposed temporal subgridding method, and the results are shown in
Table 1. The
norm is calculated by taking the differences between the solutions of the standard FDTD method (without subgridding) and the proposed temporal subgridding method. As observed from this table, the rate of convergence of the iteration-based subgridding algorithm maintains a second order.
In the next test, we assess the stability of the iteration-based subgridding algorithms by performing long-term simulations. We provide a random magnetic field as an initial condition and run the simulation for one million steps. The computation for this model was performed using a coarse mesh size of
. A box from grid points
to
is the region consisting of the fine mesh. We conduct a comparison between the conventional ghost-
E subgridding method and the iteration-based approach.
Figure 6 illustrates the time history of the solutions obtained from both methods at a selected location. For the conventional method, the solution starts to grow exponentially after about 800,000 steps, so it demonstrates a late-time instability. In contrast, the iteration-based approach maintains a bounded solution throughout the entire one-million-step simulation, thus avoiding late-time instability.
In the last test of this subsection, we carried out eigenvalue tests using small mesh sizes.
Figure 7 shows the plots of all eigenvalues in the complex plane on a
coarse mesh embedded with a
fine mesh. For the ghost-
E subgridding method, there are some eigenvalues outside the unit circles, so their largest magnitude is greater than 1, which results in weak growth of the solution and late-time instability. For the ghost-
H and the iteration-based subgridding methods, all eigenvalues lie on or inside the unit circle (enough close to one up to the machine round-off error of
), leading to stable solutions.
Furthermore, the largest eigenvalues (in terms of their magnitudes) are computed with various coarse grid sizes of
,
,
, and
. The fine mesh is a square region fixed at a region with
cells. Results are shown in
Table 2. We observe that the largest eigenvalue of the ghost-
E method is larger than 1. In contrast, the largest eigenvalues of the ghost-
H and the iteration-based ghost-
E subgridding methods are 1 (accurate to the machine round-off error of
).
Note that the ghost-H method works only for a 2D case so only the ghost-E method is suitable when extending the subgridding method to 3D. A future work is to extend the iteration-based method to three dimensions.
4.2. Tests for Spatial-Temporal Subgridding Method
In this subsection, we test the iteration-based spatial and temporal subgridding algorithm. Similar to the previous examples, we first test the cylindrical wave propagation where the incident source is placed at the center of the domain. The coarse mesh contains
cells. The spatial fine mesh is an
square zone from grid points
to
. The temporal fine mesh is
, two cells larger than the spatial fine mesh in all directions. On the coarse grid,
nm and the time step
is computed using a Courant number of 0.5. In the spatial–temporal fine region, the grid cells are halved, so they are
. Between the spatial and temporal fine mesh, the grid cell sizes are
. As shown in
Figure 8, the numerical results show a good agreement between the iteration-based subgridding method and the standard FDTD method.
Similar to the previous examples, we test the stability with random initial data and run the simulation for one million steps. The computation for this model was performed using a coarse grid of
cells embedded with a
fine mesh. As shown in
Figure 9, a stable solution is obtained for the proposed method after one million steps.
Figure 10 and
Table 3 show the result of the eigenvalue test. Similar to the previous tests, the coarse mesh contains
cells, and the fine grid contains
cells.
Figure 10 shows that all the eigenvalues are located inside or on the unit circle.
Table 3 confirms that the largest eigenvalue (in terms of magnitude) is 1.0 (up to the machine round-off error of
) for all grid sizes, indicating the stability of the method.
4.3. Through-the-Wall Radar Imaging
In this example, we implement the proposed iteration-based subgridding FDTD method as the forward problem solver in the Through-the-Wall Radar (TWR) imaging application. We consider a two-dimensional rectangular room with dimensions
by
, where an object is placed in the room. The relative permittivities of the room and the object are 1 and
, respectively. We ignore the influence of the walls by setting their relative permittivity to 1. A radar is located outside the room. It sends out an electromagnetic pulse through a transmitting antenna (TX). When the signal hits an object, the reflected wave propagates back and is recorded by a receiving antenna (RX). As shown in
Figure 11, by letting the radar go around the room’s perimeter, we obtain a set of transmitted and received signals. In our simulation, we select three locations on each side of the room, so in total, we collect 12 data sets. Using these recorded data, we can reconstruct the shape of the object using the algorithm given in [
31].
In our simulations, we set the room size to be . The object in the center of the room is a circle (diameter is ) or a square (side length ) with a permittivity of . The electromagnetic incident source is a Gaussian pulse with a carrier frequency of .
For the standard FDTD method, the mesh size is , so . For the iteration-based subgridding method, we use a coarse mesh of (), embedded with a small box (1 m by 1 m) surrounding the object using a fine mesh with half grid size .
Figure 12 shows the simulation results. Green curves represent the boundaries of the objects. The open circles are the reconstructed boundary points of the object using the standard FDTD method. The red markers indicate the reconstructed points by the iteration-based subgridding method. The results of the two methods are similar for the circular object. For the squared object, the reconstructed points by the subgridding method are slightly closer to the actual boundary. In terms of computational time, the FDTD method runs for approximately 5 min, while the subgridding method takes only about 2 min. Thus, from this example, we observe that the subgridding method is more efficient.
5. Conclusions
In this paper, we develop a temporal subgridding algorithm appropriate for the FDTD method for Maxwell’s equations in two dimensions and the mode. Different from other subgridding methods, our method iterates the updating equations on the temporal interface to obtain stable solutions. It is an alternative approach to enhance the stability of the subgridding algorithm without modifying the interpolation formulas. The refinement ratio is 2:1, and the ghost value interpolation technique is applied when updating the solutions on the coarse–fine mesh interface. We numerically test the stability of the algorithm by eigenvalue tests and by performing long-term simulations. Our study shows that conventional temporal subgridding with magnetic field interpolation is stable, while if an electric field interpolation is used, it is late-time unstable since some eigenvalues have magnitudes larger than 1. The proposed iteration-based method with an electric field interpolation is stable since all eigenvalues are located on or inside the unit circle. We further use the proposed subgridding FDTD algorithm to serve as the forward problem solver in through-the-wall radar (TWR) imaging. Future work includes the extension to different refinement ratios and three dimensions, along with comparison to other subgridding methods.
Author Contributions
Methodology, P.X. and J.L.; Software, P.X. and J.L.; Validation, P.X. and J.L.; Formal analysis, P.X. and J.L.; Writing—original draft, P.X.; Writing—review & editing, P.X. and J.L. All authors have read and agreed to the published version of the manuscript.
Funding
This work was supported in part by the AFOSR Grant FA9550-16-1-0199.
Data Availability Statement
Data are contained within the article.
Acknowledgments
J.L. would like to thank the Delaware Institute for Science and Technology for summer support through the US Department of Education Title III fund.
Conflicts of Interest
The authors declare no conflicts of interest.
References
- Yee, K. Numerical solution of initial boundary value problems involving Maxwell’s equations in isotropic media. IEEE Trans. Antennas Propag. 1966, 14, 302–307. [Google Scholar]
- Taflove, A.; Brodwin, M.E. Numerical Solution of Steady-State Electromagnetic Scattering Problems Using the Time-Dependent Maxwell’S Equations. IEEE Trans. Microw. Theory Tech. 1975, 23, 623–630. [Google Scholar] [CrossRef]
- Taflove, A.; Hagness, S. Computational Electrodynamics: The Finite-Difference Time-Domain Method, 3rd ed.; Artech House: Norwood, MA, USA, 2005. [Google Scholar]
- Kim, I.; Hoeffer, W. A local mesh refinement algorithm for the time-domain finite-difference method using Maxwell’s curl equations. IEEE Trans. Microw. Theory Tech. 1990, 38, 812–815. [Google Scholar] [CrossRef]
- Zivanovic, S.; Yee, K.; Mei, K. A subgriding method for the time-domain finite-difference method to solve Maxwell’s equations. IEEE Trans. Microw. Theory Tech. 1991, 39, 471–479. [Google Scholar] [CrossRef]
- Prescott, D.T.; Shuley, N.V. A method for incorporating different sized cells into the finite-difference time-domain analysis technique. IEEE Microw. Guid. Wave Lett. 1992, 2, 434–436. [Google Scholar] [CrossRef] [PubMed]
- Shimizu, D.; Okoniewski, M.; Stuchly, M. An efficient subgriding algorithm for FDTD. In Proceedings of the 11th Annual Review of Progress in Applied Computational Electromagnetics, Monterey, CA, USA, 20–25 March 1995; Volume 2, pp. 762–766. [Google Scholar]
- Thoma, P.; Weiland, T. A consistent subgridding scheme for the finite difference time domain method. Int. J. Numer. Anal. Mod. 1996, 9, 359–374. [Google Scholar] [CrossRef]
- Okoniewski, M.; Okoniewska, E.; Stuchly, M.A. Three dimensional subgridding algorithm for FDTD. IEEE Trans. Antennas Propag. 1997, 45, 422–429. [Google Scholar] [CrossRef]
- Chevalier, M.; Luebbers, R. FDTD local grid with material traverse. IEEE Trans. Antennas Propag. 1997, 45, 411–421. [Google Scholar] [CrossRef]
- Collino, F.; Fouquet, T.; Joly, P. A space-time mesh refinement method for the 1D Maxwell’s equation. C. R. Acad. Sci. Paris 1999, 328, 263–268. [Google Scholar] [CrossRef]
- Yu, W.; Mittra, R. A new subgridding method for the finite-difference time-domain algorithm. IEEE Microw. Opt. Technol. Lett. 1999, 21, 330–333. [Google Scholar] [CrossRef]
- Wang, S.; Teixeira, F.; Lee, R.; Lee, J.F. Optimization of subgridding schemes for FDTD. IEEE Microw. Wirel. Compon. Lett. 2002, 12, 223–225. [Google Scholar] [CrossRef]
- Ahmed, I.; Chen, Z. A hybrid ADI-FDTD subgridding scheme for efficient electromagnetic computation. Int. J. Numer. Model 2004, 17, 237–249. [Google Scholar] [CrossRef]
- Vaccari, A.; Pontalti, R.; Malacarne, C.; Cristoforetti, L. A robust and efficient subgridding algorithm for finite-difference time-domain simulations of Maxwell’s equations. J. Comput. Phys. 2004, 194, 117–139. [Google Scholar] [CrossRef]
- Marrone, M.; Mittra, R. A new stable hybrid three-dimensional finite difference time domain algorithm for analyzing complex structures. IEEE Trans. Antennas Propag. 2005, 53, 1729–1737. [Google Scholar] [CrossRef]
- Donderici, B.; Teixeira, F. Improved FDTD subgridding algorithms via digital filtering and domain overriding. IEEE Trans. Antennas Propag. 2005, 53, 2938–2951. [Google Scholar] [CrossRef]
- Donderici, B.; Teixeira, F. Domain-overriding and digital filtering for 3D FDTD subgridded simulations. IEEE Microw. Wirel. Compon. Lett. 2006, 16, 10–12. [Google Scholar] [CrossRef]
- Collino, F.; Fouquet, T.; Joly, P. Conservative space-time mesh refinement methods for the FDTD solution of Maxwell’s equations. J. Comput. Phys. 2006, 211, 9–35. [Google Scholar] [CrossRef]
- Chilton, R.; Lee, R. Conservable and provably stable FDTD subgridding. IEEE Trans. Antennas Propag. 2006, 55, 2537–2549. [Google Scholar] [CrossRef]
- Chilton, R. H-, p- and t-Refinement Strategies for the Finite-Difference-Time-Domain Method Developed via Finite-Element Principles. Ph.D. Thesis, The Ohio State University, Columbus, OH, USA, 2008. [Google Scholar]
- Venkatarayalu, N.; Lee, R.; Gan, Y.B.; Li, L.W. A stable FDTD subgridding method based on finite element formulation with hanging variables. IEEE Trans. Antennas Propag. 2007, 55, 907–915. [Google Scholar] [CrossRef]
- Xiao, K.; Pommerenke, D.; Drewniak, J. A three-dimensional FDTD subgridding algorithm with separated temporal and spatial interfaces and related stability analysis. IEEE Trans. Antennas Propag. 2007, 55, 1981–1990. [Google Scholar] [CrossRef]
- Zakharian, A.R.; Brio, M.; Dineen, C.; Moloney, J.V. Stability of 2D FDTD Algorithms with Local Mesh Refinement for Maxwell’s Equations. Commun. Math. Sci. 2006, 4, 345–374. [Google Scholar]
- Bérenger, J.P. A FDTD subgriding based on Huygens surfaces. In Proceedings of the 2005 IEEE Antennas and Propagation Society International Symposium, Washington, DC, USA, 3–8 July 2005; Volume 2, pp. 98–101. [Google Scholar]
- Berenger, J.P. A Huygens Subgridding for the FDTD Method. IEEE Trans. Antennas Propag. 2006, 54, 3797–3804. [Google Scholar]
- Berenger, J.P. Extension of the FDTD Huygens subgridding algorithm to two dimensions. IEEE Trans. Antennas Propag. 2009, 57, 3860–3867. [Google Scholar] [CrossRef]
- Berenger, J.P. Three dimensional Huygens subgridding for FDTD. In Proceedings of the 2009 IEEE Antennas and Propagation Society International Symposium, North Charleston, SC, USA, 1–5 June 2009; pp. 1–4. [Google Scholar]
- Berenger, J.P. The Huygens subgridding for the numerical solution of the Maxwell equations. J. Comput. Phys. 2011, 230, 5635–5659. [Google Scholar] [CrossRef]
- Hartley, J.; Giannopoulos, A.; Davidson, N. Switched Huygens Subgridding for the FDTD Method. IEEE Trans. Antennas Propag. 2022, 70, 6872–6882. [Google Scholar] [CrossRef]
- Charnley, M.; Wood, A. Through-the-wall radar detection analysis via numerical modeling of Maxwell’s equations. J. Comput. Phys. 2016, 313, 532–548. [Google Scholar] [CrossRef]
Figure 1.
A two-dimensional subgridding mesh with separated spatial (orange) and temporal (red) coarse–fine mesh interfaces.
Figure 1.
A two-dimensional subgridding mesh with separated spatial (orange) and temporal (red) coarse–fine mesh interfaces.
Figure 2.
Coarse and fine meshes for the subgridding algorithm in mode using ghost magnetic fields. Blue triangles denote the ghost magnetic fields on the coarse–fine mesh interface.
Figure 2.
Coarse and fine meshes for the subgridding algorithm in mode using ghost magnetic fields. Blue triangles denote the ghost magnetic fields on the coarse–fine mesh interface.
Figure 3.
Coarse and fine meshes for the subgridding algorithm in mode using ghost electric field. Blue triangles denote ghost electric fields on the coarse–fine mesh interface.
Figure 3.
Coarse and fine meshes for the subgridding algorithm in mode using ghost electric field. Blue triangles denote ghost electric fields on the coarse–fine mesh interface.
Figure 4.
Flow chart of (a) conventional and (b) iteration-based temporal subgridding methods.
Figure 4.
Flow chart of (a) conventional and (b) iteration-based temporal subgridding methods.
Figure 5.
Distribution of the magnetic field using the iteration-based temporal subgridding method. The mesh size is . The square box is the fine mesh.
Figure 5.
Distribution of the magnetic field using the iteration-based temporal subgridding method. The mesh size is . The square box is the fine mesh.
Figure 6.
Time history of the magnetic field for 1 million steps using (a) conventional and (b) iteration-based ghost-E temporal subgridding algorithms. The conventional method shows late-time instability since the solution starts to grow exponentially after approximately 800,000 steps. For the iteration-based method, the solution maintains stability throughout the entire one-million-step simulation.
Figure 6.
Time history of the magnetic field for 1 million steps using (a) conventional and (b) iteration-based ghost-E temporal subgridding algorithms. The conventional method shows late-time instability since the solution starts to grow exponentially after approximately 800,000 steps. For the iteration-based method, the solution maintains stability throughout the entire one-million-step simulation.
Figure 7.
Distribution of the eigenvalues on the complex plane for (a) Conventional ghost-E, (c) conventional ghost-H, and (d) iteration-based ghost-E temporal subgridding methods. The enlarged box of (a) is shown in (b).
Figure 7.
Distribution of the eigenvalues on the complex plane for (a) Conventional ghost-E, (c) conventional ghost-H, and (d) iteration-based ghost-E temporal subgridding methods. The enlarged box of (a) is shown in (b).
Figure 8.
Distribution of the magnetic field using (a) standard and (b) iteration-based spatial–temporal subgridding FDTD methods. The square box indicates the fine mesh.
Figure 8.
Distribution of the magnetic field using (a) standard and (b) iteration-based spatial–temporal subgridding FDTD methods. The square box indicates the fine mesh.
Figure 9.
Time history of the magnetic field () using the iteration-based spatial–temporal subgridding FDTD method for 1 million steps.
Figure 9.
Time history of the magnetic field () using the iteration-based spatial–temporal subgridding FDTD method for 1 million steps.
Figure 10.
Distribution of the eigenvalues on the complex plane for the iteration-based spatial–temporal subgridding FDTD method.
Figure 10.
Distribution of the eigenvalues on the complex plane for the iteration-based spatial–temporal subgridding FDTD method.
Figure 11.
Illustration of the Through-the-Wall Radar experiment.
Figure 11.
Illustration of the Through-the-Wall Radar experiment.
Figure 12.
Reconstructed boundary points of (a) circular and (b) squared objects using standard FDTD and iteration-based subgridding (IBS) FDTD methods. Green curves indicate the boundaries of the actual objects.
Figure 12.
Reconstructed boundary points of (a) circular and (b) squared objects using standard FDTD and iteration-based subgridding (IBS) FDTD methods. Green curves indicate the boundaries of the actual objects.
Table 1.
Convergence test of the iteration-based temporal subgridding method.
Table 1.
Convergence test of the iteration-based temporal subgridding method.
Mesh Size | Norm | Order of Convergence |
---|
| | - |
| | |
| | |
| | |
| | |
Table 2.
List of largest eigenvalues (in terms of magnitude) for three temporal subgridding methods in various mesh sizes.
Table 2.
List of largest eigenvalues (in terms of magnitude) for three temporal subgridding methods in various mesh sizes.
Method | Mesh Size | Largest Eigenvalue (Round-Off Error ) |
---|
Conventional Ghost-H | | |
| |
| |
| |
Conventional Ghost-E | | |
| |
| |
| |
Iteration-based Ghost-E | | |
| |
| |
| |
Table 3.
List of the largest eigenvalues (in terms of magnitude) of different grid sizes for the iteration-based subgridding FDTD method.
Table 3.
List of the largest eigenvalues (in terms of magnitude) of different grid sizes for the iteration-based subgridding FDTD method.
Mesh Size | Largest Eigenvalue (Round-Off Error ) |
---|
| |
| |
| |
| |
| Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).