† Corresponding author. E-mail:
Several major challenges need to be faced for efficient transient multiscale electromagnetic simulations, such as flexible and robust geometric modeling schemes, efficient and stable time-stepping algorithms, etc. Fortunately, because of the versatile choices of spatial discretization and temporal integration, a discontinuous Galerkin time-domain (DGTD) method can be a very promising method of solving transient multiscale electromagnetic problems. In this paper, we present the application of a leap-frog DGTD method to the analyzing of the multiscale electromagnetic scattering problems. The uniaxial perfect matching layer (UPML) truncation of the computational domain is discussed and formulated in the leap-frog DGTD context. Numerical validations are performed in the challenging test cases demonstrating the accuracy and effectiveness of the method in solving transient multiscale electromagnetic problems compared with those of other numerical methods.
Nowadays, the world has entered into the era of stealth. However, the analysis of the electromagnetic scattering for low-observable targets, which are often multiscale (the electrically large structures having electrically small details), is still a challenging problem for numerical solvers. The typical frequency-domain method, the method of moment (MoM),[1] is a common choice to accurately deal with these problems. However, the MoM method may become computationally inefficient for wideband computations, since each frequency needs a complete resolution of a linear system of equations. The time-domain method is an attractive alternative since it employs a time-marching algorithm that permits one to find the whole frequency-domain behavior with a single simulation.
Among well-known time-domain methods, the finite-difference time-domain (FDTD)[2] method has become very popular for its versatility and power. However, it requires an orthogonal grid (structured mesh), which will need a high discretization density to capture the geometric characteristics of electrically fine structures, yet this can generate a large number of wasted unknowns in the electrically coarse domains. The sub-gridding technique[3] can alleviate this problem, but it will destroy the simple data structure of the standard FDTD scheme and greatly increase computational complexity. To overcome this limitation, the finite element time-domain (FETD) method, which is more flexible in complex geometric modeling by using an unstructured mesh, has been made[4,5] to solve Maxwell’s equations. But this method requires the solution of matrix equations either directly or iteratively at each time step. A multiscale problem usually is discretized into a great number of unknowns, and it will become expensive to perform operations with a very large sparse linear system of equations during time stepping.
The discontinuous Galerkin time-domain (DGTD) methods,[6–8] which are experiencing a fast development, are promising methods of solving multiscale problems since they possess some of the advantages of FDTD and FETD methods. On one hand, the DGTD method has most of the advantages of FDTD, such as spatial explicit algorithm, memory and computational cost only growing linearly with the number of elements, simplicity and easy parallelization.[9] Meanwhile, the perfectly matched layer (PML) truncation technique[10] can also be straightforwardly integrated into DGTD. On the other hand, the DGTD method inherently possesses the capablity of dealing with arbitrarily shaped and inhomogeneously filled objects, which is an important advantage of the conventional FETD method. The main difference is that the solution is allowed to be discontinuous across the boundaries between adjacent elements, which communicate by means of numerical fluxes.[11–13] Owing to this feature, the DGTD method is more flexible. In Ref. [8], a new three-dimensional (3D) continuous–discontinuous Galerkin finite element time-domain is proposed, it possesses the advantages of a reduced number of unknowns for the continuous Galerkin method and the block-diagonal property of the discontinuous Galerkin method. In a word, the DGTD method is much more flexible than the FDTD method in modeling complex structures and much easier than the conventional FETD method in handling multiscale problems.
There are several key steps for constructing a DGTD system: 1) determining which governing equations the DGTD method will be based on; 2) choosing the element shape and corresponding basis functions for the spatial discretization; 3) applying numerical fluxes to interfaces to stitch all elements together; and 4) selecting a time scheme based on properties of a discretized system.[14]
In this paper, we use a globally explicit DGTD method based on the leap-frog (LF) time integration scheme, to calculate the radar cross sections (RCSs) of some targets, especially including the perfect electric conductor (PEC) and dielectric coated NASA almonds. This geometry has been chosen as a challenging example of low-observation target used in the validation of numerical solvers.[15] At the same time, the electromagnetic fields are expanded into the vector basis functions,[16] which can overcome the shortcomings of node basis functions, such as spurious solutions and difficulty in treating conducting and dielectric edges. The uniaxial PML technique[17–20] is applied to the DGTD modeling. Numerical results show that this method can be competitive with MoM in terms of accuracy.
The rest of this paper is organized as follows. The theory and formulations of this method are described in detail in Section
There are mainly two kinds of equations for governing transient electromagnetic problems. Mathematically, these governing equations are equivalent, but with different discretization schemes: they differ from each other greatly in numerical properties.[21] One type is the second-order vector wave equation.[16] The reason for choosing this equation is that the first family of Nédélec elements, also known as the edge elements,[11] which are free of spurious mode, can be directly used for spatial discretization. However, it has difficulties in constructing the time-domain PML, which is a popular technique to truncate unbounded regions. Moreover, it has only one variable, unsuitable for building a DGTD system. Because of the implementation of numerical fluxes, a critical step for the DGTD method is based on both
To avoid the above difficulties, the other type, which is the first-order Maxwell’s equations with both
The DGTD method is based on a finite-element geometrical discretization of the space M into nonoverlapping elements Vm, each of which is bounded by
Applying the following vector identity and surface divergency theorem yields
Equation (
Another popular and even simpler numerical flux is called the central flux[13]
Considering that the spatial part of the fields is expanded within each element in a set of basis functions equal to the set of test functions, using a Feado–Galerkin method
(i) em and hm are column vectors varying in time with the discretized electric and magnetic field coefficients in the element m, and
(ii) Jsm is a column vector varying in time of the discretized excitations in the element m
(iii) M is the mass matrix
(iv) K is the stiffness matrix
(v) L are the flux matrices
For a DGTD system, time stepping can be performed element by element rather than solving a huge matrix system as in FETD schemes. This advantage of the DGTD method can save a large number of memories during time stepping, and furthermore, it makes parallel computation straightforward for a DGTD system.
For the time domain integration, several approaches can be chosen. The most commonly employed one is the second-order LF[22] scheme. The basis of the LF scheme is to sample the unknown fields in a staggered way. The first-order time derivative are approximated by central difference. Average approximation is used to estimate the terms with the electric conductivity. For the two extra dissipative terms arising from the upwind flux formulation, we use the backward approximation, since an average would yield a globally implicit scheme. This fact introduces a slightly more restrictive stability condition.[13] Finally, the resulting fully explicit LFDG algorithm becomes
The flux conditions not only serve as connecting adjacent fields but also implementing boundary conditions.
1) PEC boundary conditions on a face of an element can be strictly enforced by setting the tangential electric field to be null, and tangential magnetic field to be continuous.
Perfect magnetic conducting (PMC) conditions are reciprocal of PEC
2) Regarding the absorbing boundary conditions (ABCs), the straightest one is the so-called first-order Silver–Müller (SM-ABC),[22] which can be enforced by setting
3) Incident wave conditions can also be generated in a straightforward way. Assuming that a known wave is propagating inside a total-field zone (TFZ), while outside it (scattered-field zone (SFZ)) the field is null. If
The PML plays a critical role in transient simulation of unbounded problems because of its ability to absorb the plane waves of all frequencies and incident angles. Several extensions and improvements have been made since it was proposed by Bérenger.[10,18] Here we introduce the UPML, which is regarded as an artificial anisotropic absorbing material, fulfills Maxwell equations, also attenuates the energy inside the PML without reflection. The UPML can be described in the frequency domain by using Maxwell equations
One can substitute expression (
In this study,
Because the curl terms in Eqs. (
The extension of the LF temporal integration scheme to the semi-discrete system of Eq. (
We have implemented 3D codes with the upwind numerical flux. The study of the stability of the scheme will not be addressed here, and we have derived heuristic estimations[7,22] for the maximum time steps, yielding stable schemes in each case.
In this section, we show some numerical results of scattering problems that will be illustrated to verify the accuracy and capability of the DGTD method. All simulations are performed on a PC computer with an AMD Athlon(tm) II X4 Processor 3.00 GHz and 16 GB of RAM. The incident plane wave is GAUSS pulse and expressed as
In order to verify the validity of the DGTD method, the first example is performed by a simple case of a scattering problem for a PEC sphere, of which the diameter is
The structures are illuminated with a vertically polarized plane wave (the electric field vector being along the z axis), which propagates along the negative direction of the x axis (
For verifying the capability to handle the dielectric material problems, the RCS of a dielectric sphere with the relative inductivity
From two cases above, we observe that the validity and capability to handle the PEC and dielectric scattering problems. However, the analysis of the electromagnetic scattering by low-observable targets is a challenging problem for numerical solvers. In the final example, we use the DGTD method to calculate the RCS of a typical low-observable target: the NASA almond. This geometry has been chosen as a challenging example of a low-observable target used in the validation of numerical solvers.
The NASA almond (Fig.
The structures are also illuminated with a vertically polarized plane wave (the electric field vector being along the z axis) impinging on the almond at the vertex, see Fig.
The copular bistatic RCSs of the PEC almond at 1 GHz, 2 GHz, and 3 GHz in the xoy plane and xoz plane are presented in Figs.
In this paper, a fully explicit DGTD method is developed to analyze the scattering of some typical targets, especially the NASA almonds. In the DGTD method the tetrahedral elements and their vector basis functions are used to accurately model the computational domain and electromagnetic fields, respectively. In addition, the UPML is employed to truncate the computational domain. A stable time-difference is used to discretize the equations. The RCS of several typical targets are simulated by using the DGTD method. According to the comparisons of the monostatic and bistatic RCS of the PEC and dielectric sphere for the MoM, the validity and capability of the DGTD method are demonstrated. Furthermore, the comparisons of RCS between the DGTD method and MoM for low-observable targets: the NASA almond, are given in the final example. This excellent agreement of results verifies the accuracy and effectiveness of the DGTD method of dealing with the transient multiscale electromagnetic problems.
[1] | |
[2] | |
[3] | |
[4] | |
[5] | |
[6] | |
[7] | |
[8] | |
[9] | |
[10] | |
[11] | |
[12] | |
[13] | |
[14] | |
[15] | |
[16] | |
[17] | |
[18] | |
[19] | |
[20] | |
[21] | |
[22] | |
[23] |