Preasymptotic error estimates of
EEM and CIP-EEM for the time-harmonic Maxwell equations with large wave number
Shuaishuai Lu
Department of Mathematics, Nanjing University, Jiangsu, 210093, P.R. China.
ssl@smail.nju.edu.cn and Haijun Wu
Department of Mathematics, Nanjing University, Jiangsu, 210093, P.R. China.
hjw@nju.edu.cn
Abstract.
Preasymptotic error estimates are derived for the linear edge element method (EEM)
and the linear -conforming interior penalty edge element method (CIP-EEM) for the time-harmonic
Maxwell equations with large wave number.
It is shown that under the mesh condition that is
sufficiently small, the errors of the solutions to both methods
are bounded by in the energy norm and in
the norm, where is the wave number and is the mesh size. Numerical tests are provided
to verify our theoretical results and to illustrate the potential of CIP-EEM in significantly reducing the
pollution effect.
This work was partially supported by the NSF of China under grants 12171238, 12261160361, and 11525103.
1. Introduction
In this paper, we consider the following time-harmonic Maxwell equations for the electric field with
the standard impedance boundary condition:
(1.1)
(1.2)
where is a bounded domain with a boundary. Here,
denotes the imaginary unit while denotes the unit outward normal to
and is the tangential component
of on . Additionally,
is the wave number, and is known as the impedance constant.
The right-hand side is related to a given current density (cf. [31]).
It is assumed that in and on (i.e. ).
The time-harmonic Maxwell equations govern the propagation of electromagnetic waves
at a specified frequency, serving as fundamental tools
for comprehending and forecasting electromagnetic field behaviors across diverse
applications such as telecommunications, radar, optics, and electromagnetic sensing.
Accurate numerical approximation of these equations is imperative for predicting
wave propagation phenomena in practical scenarios.
Since the seminal contributions of Nédélec [34, 35], the
edge element method (EEM) has garnered significant popularity and become a central tool for addressing electromagnetic field problems.
Subsequently, a substantial body of literature has emerged studying EEM for solving the Maxwell equations, we refer the readers to [31, 18, 23] and the references therein. For time-harmonic Maxwell problems with impedance boundary conditions,
Monk [31, Chapter 7] and Gatica and Meddahi [17] have investigated the convergence and error estimation of EEM under
the condition that the mesh size is sufficiently small. Both of these results are -implicit, meaning they do not discuss the
influence of on the error, and thus are applicable primarily to problems with small wave numbers.
However, in scenarios involving large wave numbers, a large number of studies of the finite element methods (FEMs) for Helmholtz equations [5, 22, 27, 28, 39, 13, 44, 43] indicate that the EEM of fixed order may also suffer from the well-known “pollution effect”, i.e., compared to the best approximation from the discrete space, the approximation ability of the solution to the EEM (with fixed order) gets worse and worse as the wave number increases.
Clearly, estimating the pollution error is significant both in theory and practice, and it has always been interesting to develop numerical methods with less pollution error. Unfortunately, rigorous pollution-error analyses for the EEM for the time-harmonic Maxwell equations are still unavailable in the literature. We recall that the term “asymptotic error estimate” refers to the error estimate without pollution error and the term “preasymptotic error estimate” refers to the estimate with non-negligible pollution effect.
The latest asymptotic error analysis for the time-harmonic Maxwell equations was conducted by Melenk and Sauter in
[29, 30], where the first type Nédélec -EEM for solving the
time-harmonic Maxwell equations with transparent boundary condition and impedance boundary condition
has been studied. Using the so-called “regularity decomposition” technique, they show
that the -EEM is pollution-free under the following
scale resolution conditions:
(1.3)
where is the mesh size, is an arbitrary positive number and is sufficiently small. The resolution
condition is natural for resolving the oscillatory
behavior of the solution, and the side constraint suppresses the
pollution effect.
The above result highlights the advantages of high-order EEM, representing a breakthrough in the
analysis of the time-harmonic Maxwell equations in the large wave number regime. But for fixed order, say second order EEM of the first type, the above two works require that is sufficiently small (see [29, Remark 4.19]) for the case of transparent boundary condition and is sufficiently small for the case of impedance boundary condition, respectively, which are too strict in .
For the special case in problem (1.1)–(1.2), Nicaise and
Tomezyk [37, 36], using - and -FEM with Lagrange elements, performed
asymptotic error analyses for a regularized Maxwell system(cf. [11, § 4.5d]).
Recently, Lu et al.[26] proposed an -conforming interior penalty edge element method
(CIP-EEM), which uses the same approximation space as EEM but modifies the sesquilinear form by adding a least squares
term penalizing the jump of the tangential components of the of the discrete solution at mesh interfaces.
Using the so-called “stability-error iterative improvement” technique [15], they proved that the CIP-EEM with pure imaginary penalty parameter is absolute stable (that is, stable for any and ) and obtained the
following error estimate under the conditions and :
(1.4)
where is the energy norm. This is a closely related result. However, this result does not encompass the classical EEM since cannot be zero there. For wave-number-explicit error analyses of various discontinuous Galerkin (DG) methods, we refer to [16, 14, 20].
The purpose of this paper is twofold. First, we focus on the preasymptotic error analysis of the EEM and provide the first such results in the literature for the EEM using the linear Nédélec element of the second type. To be precise, we have proved the following error estimate for the solution to the linear EEM, under the mesh condition that is sufficiently small:
(1.5)
Clearly, the first term on the right-hand side of (1.5) is of the same order as the error of the best approximation, which can be reduced by putting enough points per wavelength. However, the second term will get out of control as the wave number increases, no matter how many points are put per wavelength, as long as the number of them is fixed. Therefore, the second term is the pollution error which dominates the error bound when is large. Moreover, the following error estimate has also been proved under the same mesh condition:
(1.6)
We would like to mention that, in order to prove the error estimates (1.5)–(1.6), thanks to the recent work [9], we derive a wave-number-explicit regularity estimate (see (3.3)) for the continuous problem on the domain with only -smooth boundary. A similar estimate was proved in [25] for a domain with boundary.
Secondly, we extend the result in [26] for the CIP-EEM to a more general case, i.e., the linear CIP-EEM with complex penalty parameters . To be precise, we prove that (1.4) still holds,
if , , , and , for some positive constants and independent of and . Such an extension is meaningful since the penalty parameters that can greatly reduce the pollution error are usually close to negative real numbers (see Section 6), which is not covered by the theory given in [26]. By the way, the divergence-free constraint on the source term in [26] is dropped in this paper. Compared with the linear EEM, the linear CIP-EEM provides a candidate with enhanced stability and low pollution effects. We would like to mention that due to their easy of implementation, low-order methods are still attractive in many applications, especially for those low-regularity problems where a higher-order method cannot achieve its full convergence rate.
The key idea in our preasymptotic analysis for the EEM and CIP-EEM is to first prove stability estimates for the discrete solution , in which a delicate decomposition of (see (4.1)) and a so-called “modified duality argument” technique play important roles, and then use the obtained stability estimates to derive the desired error estimates. The “modified duality argument” technique was first used to derive preasymptotic error estimates for the FEM for Helmholtz equations with large wave numbers [44, 13]. Here we extend it to derive stability estimates for the EEM. We would like to mention that our proofs are highly non-trivial. The “stability-error iterative improvement” technique used in [26] does not work for the EEM or CIP-EEM with general parameters. We had also tried to derive preasymptotic error estimates for the EEM by mimicking the usual process (see, e.g., [31]), including using the modified duality argument to derive the error estimates directly, but failed.
The preasymptotic error analysis of higher-order EEM and CIP-EEM necessitates additional technical tools and will be investigated in another work.
The remainder of this paper is organized as follows. In Section 2, we formulate the EEM and the CIP-EEM.
Section 3 introduces some preliminary results about the stability estimates
and some error estimates of some -elliptic projections. In Section 4, we first establish
discrete stability estimates for the EEM solution.
Then we prove the preasymptotic error estimates of the EEM for the time-harmonic Maxwell problem.
In Section 5, we extend the results of preasymptotic error estimates to the linear CIP-EEM with general penalty parameters.
Finally, in Section 6, we present some numerical examples to verify our theoretical findings and the great potential of the CIP-EEM to reduce the pollution errors.
Throughout this paper, we use notations and for
the inequalities and ,
where is a positive number independent of the mesh size and the wave number ,
but the value of which can vary in different occurrences.
is a shorthand notation for the statement and .
We also denote by a generic constant which may depends on
but satisfies when .
For simplicity, we suppose , . We also assume that the domain
is strictly star-shaped with respect to a point , that is,
.
2. Formulations of EEM and CIP-EEM
In order to formulate the two methods, we first introduce some notations. For a domain with Lipschitz
boundary , we shall use the standard Sobolev space , its norm , seminorm
, and inner product. We refer to [1, 6, 31] for their definitions.
If the functions are vector-valued we shall indicate these function spaces
by boldface symbols, e.g., .
In particular, and for
denote the -inner product on complex-valued and spaces, respectively.
For simplicity, we shall use the shorthands:
We introduce the space
Let
where is understood as the trace from to and
is the tangential trace from to . Similarly, we will understand
as the tangential trace from to (see, e.g., [31, 29]).
Next we define the “energy” space
and introduce the following sesquilinear form on :
(2.1)
Then the variational formulation of (1.1)–(1.2) reads as:
Find such that
(2.2)
To discretize (2.2), we follow [30] to introduce a regular and quasi-uniform
triangulation of which satisfies:
(1)
The (closed) elements cover , i.e.,
.
(2)
Associated with each element is the element map, a -diffeomorphism
. The set is the reference tetrahedron. Denoting ,
there holds, with some shape-regularity constant ,
where is the Jacobian matrix of .
(3)
The intersection of two elements is only empty, a vertex, an edge, a face, or they coincide
(here, vertices, edges, and faces are the images of the corresponding
entities on the reference element ). The parameterization of common edges
or faces is compatible. That is, if two elements , share an edge (i.e.,
for edges of ) or a face
(i.e., for faces of ),
then is an affine isomorphism.
The set of all faces of is denoted by , and the set of all interior faces
by . For any , let . Denote by .
Let
be the first order Nédélec edge element space of the second type (see [31, (3.76)]):
(2.3)
where denotes the set of linear polynomials on .
The EEM for the Maxwell problem (1.1)–(1.2) reads as:
Find such that
(2.4)
To formulate the CIP-EEM, we define the jump of the
tangential components of on an interior face
:
(2.5)
where and denote the unit outward normal
to and , respectively. We introduce the following
sesquilinear form consisting of penalty terms on interior faces:
(2.6)
where are penalty parameters that are allowed to be complex-valued. Then the CIP-EEM is
constructed by simply adding to the left-hand side of EEM (2.4), which is to
find such that
(2.7)
where
(2.8)
Remark 2.1.
(a) If , then the CIP-EEM becomes the classical EEM (2.4).
(b) The CIP-EEM is a generalization to Maxwell equations of the continuous
interior penalty finite element method (CIP-FEM) for elliptic and parabolic
problems [12], in particular, the
convection-dominated problems [7] and Helmholtz equations
[44, 39, 13].
(c) The CIP-EEM was first introduced and studied in [26] but with
pure imaginary penalty parameters. This paper concerns the CIP-EEM with general
complex penalty parameters, especially real penalty parameters, which can help to
reduce the pollution errors.
(d) Compared to the discontinuous Galerkin
methods [16, 21] and hybridizable
discontinuous Galerkin method [24], the CIP-EEM involves
fewer degrees of freedom, and thus requires less computational cost.
(e) It is clear that for the exact solution , therefore,
the CIP-EEM is consistent with the Maxwell problem (2.2), and hence there
holds the following Galerkin orthogonality:
(2.9)
3. Preliminary
In this section, we list some preliminary results which will be used in the
later sections.
First of all, we list two lemmas below to give stability and regularity
estimates for the Maxwell problem (1.1)–(1.2) that are explicit with respect to the wave number.
Lemma 3.1.
Assume that is a bounded -domain and strictly star-shaped
with respect to a point and that in and on . Let be the solution to the problem (1.1)–(1.2). Then we have
(3.1)
(3.2)
(3.3)
where
and is the the surface divergence operator (see, e.g., [31, 30]).
Proof.
For the proofs of (3.1) and (3.2) we refer to [19, Theorem 3.1 and Remark 4.6].
In order to prove (3.3), we rewrite the Maxwell equations (1.1)–(1.2) as
(3.4)
(3.5)
Recall that (see, e.g., [31, (3.52)]). We have from (1.2) and (1.1) that
where we have used the trace inequality to derive the last inequality.
Then (3.3) follows by combining (3.1)–(3.2) and the above estimate. This completes the proof of the lemma.
∎
For the estimate of in (3.1), we refer to
[30, Proposition 3.6] and [19, Theorem 3.1]. The estimate of in
(3.1) and (3.2) can be obtained by using [19, Theorem 3.1 and Remark 4.6]
and mimicking the proof of [30, Proposition 3.6].
Remark 3.1.
(a) The regularity estimate (3.3) was proved in
[25, Theorem 3.4] when the domain is and in [30, Lemma 5.1] when is sufficiently smooth.
(b) [9, Theorem 1] gave a regularity estimate by assuming only -domain,
however, explicit dependence on the wave number was not considered there.
(c) If , we can obtain the estimates of as follows.
Inspired by the proof of [30, Proposition 3.6], let satisfy
Let . Clearly, satisfies
(3.6)
(3.7)
We have and . Therefore, we can apply Lemma 3.1 to
obtain estimates for and then apply the triangle inequality to obtain estimates for .
In particular, the stability estimate (3.1) still holds, the regularity estimate (3.2)
holds if an additional term is added to its right-hand side, and the regularity
estimate (3.3) holds if the domain is and an additional term is added to .
Next we introduce two -elliptic projections and recall their error estimates,
which will be used in our preasymptotic error analysis for EEM. Let
(3.8)
For any , define its -elliptic projections as
(3.9)
We have the following error estimates for the projections, whose proofs can be found in [26] and are also given in the appendix for the reader’s convenience.
Lemma 3.2.
Suppose that . Then
(3.10a)
(3.10b)
(3.10c)
Let . The following lemma states that any discrete function in
has an “approximation” in , whose error can be bounded by its tangential components on .
The proof is similar to that of [21, Proposition 4.5], but simpler.
Lemma 3.3.
For any , there exists , such that
(3.11)
Proof.
We only give a sketch of the proof for the reader’s convenience. In fact,
denoting by the degrees of freedom (moments) of a function
on an edge of an element, can be defined simply by changing those
degrees of freedom of on the boundary to 0:
Then by some simple calculations, we obtain
which implies the first inequality in (3.11), which together with the inverse inequality implies
the second one. This completes the proof of the lemma.
∎
Let
and ,
where is the set of quadratic polynomials on .
Obviously
and are subspaces of and
, respectively. The following lemma gives both the Helmholtz decomposition and the discrete
Helmholtz decomposition for each .
Lemma 3.4(Helmholtz Decomposition).
For any , there exist ,
and , such that
(3.12)
(3.13)
(3.14)
Proof.
This Lemma may be proved by using [31, Theorem 3.45, Lemma 7.6].
We omit the details.
∎
4. Preasymptotic error estimates for EEM
In this section, we first establish
some stability estimates for the EEM (2.4) by using the so-called “modified duality argument”
developed for the preasymptotic analysis of FEM for Helmholtz equation [44].
Then we use the resulting stability estimates to derive preasymptotic error estimates for the EEM.
4.1. Stability estimates of EEM
The following theorem gives the stability of the EEM solution.
Theorem 4.1.
Let be the solution to (2.4),
, . There exists a constant independent
of and such that if , then
(4.1)
And as a consequence, the EEM (2.4) is well-posed.
Proof.
The proof is divided into the following steps.
Step 1. Estimating and by .
Taking in (2.4) and taking the imaginary and real parts on both sides, we may get
which imply that
and hence
(4.2)
(4.3)
Step 2. Decomposing . According to Lemma 3.3,
there exists such that
(4.4)
By Lemma 3.4, we have the following discrete Helmholtz decomposition
for :
(4.5)
where and is discrete divergence-free.
And there exists such that , ,
and
(4.6)
where we have used (4.4) to derive the last inequality. Moreover,
from [4, Theorem 2.12], we have
(4.7)
where we have used the inverse inequalities
to derive the last inequality. We have the following decomposition of :
(4.8)
The first two terms on the right-hand side can be bounded using (4.4) and
(4.6), respectively. To estimate the third term, we set
in (2.4) and using and to get
which implies
(4.9)
Therefore, using the Cauchy’s inequality and the Young’s inequality we obtain
Therefore, there exists a constant such that if , then
(4.18)
Then (4.1) follows from (4.18), (4.2) and (4.3).
This completes the proof of the theorem. ∎
Remark 4.1.
(a) This stability estimate (4.1) of the EE solution is of the same order as that
of the continuous solution (cf. (3.1) in Lemma 3.1 ) and still holds when .
(b) The modified duality argument used in Step 3 differs from the standard one only by replacing
the interpolation of the solution () to the dual problem used there by its elliptic projection
(). If the standard duality argument is used instead, only asymptotic stability estimate under
the mesh condition can be obtained. The modified duality argument was first used to
derive preasymptotic error estimates [44, 13] and preasymptotic stability estimates
[40] for FEM and CIP-FEM for the Helmholtz equation with large wave number.
4.2. Error estimates of EEM
The following theorem gives the main result of this paper.
Theorem 4.2.
Assume that is a bounded -domain and strictly star-shaped with respect to
a point and that in and on . Let be the solution
to the problem (1.1)–(1.2) and be the EEM solution to (2.4).
Then there exists a constant independent
of and such that when ,
the following preasymptotic error estimates hold:
Combining the above two estimates and using the triangle inequality completes the proof of the theorem.
∎
Remark 4.2.
(a) A priori error estimates of the EEM for the time-harmonic Maxwell equations with impedance
boundary conditions have been provided in [31, 17].
However, explicit dependence on the wave number is not discussed in either study.
(b) When and is , the estimates (4.19)–(4.20) still
hold if an additional term is added to (see Remark 3.1(c)).
(c) To the best of the authors’ knowledge, this theorem gives the first preasymptotic error
estimates in both and norms for the second type Nédélec linear EEM
for Maxwell’s equations, under the mesh condition that is sufficiently small.
[29, 30] proposed asymptotic error estimates for the first type
Nédélec -EEM for the Maxwell’s equations with transparent and impedance boundary
conditions, respectively, which say that the discrete solution is pollution-free if
and is sufficiently small. However, the lowest order EEM was excluded in both works.
For the second order EEM, it requires that is sufficiently small
(see [29, Remark 4.19]) for the case of transparent boundary condition or
is sufficiently small for the case of impedance boundary condition.
(d)
Our error estimation process is to prove the stability of the EEM first, and then use it to derive the error estimates.
We had tried to derive preasymptotic error estimates for the EEM mimicking the usual process (see, e.g., [31]),
including using the modified duality argument to derive the error estimates directly, but failed.
5. Preasymptotic error estimates for CIP-EEM
In this section, we show that the error estimates (4.19)–(4.20)
also hold for CIP-EEM with general penalty parameters, which can be proved by following the lines in Section 4.
We omit the similarities and point out only the differences.
The following lemma provides the continuity and coercivity of the sesquilinear form .
Lemma 5.1.
There exists a positive number such that if
and , then
(5.4a)
(5.4b)
Proof.
Since the proof of (5.4a) is straightforward by using the Cauchy-Schwarz inequality, we only prove
(5.4b).
First, for any we have
By using the local trace inequality on each element and the inverse inequality, there exists
such that
Thus, by taking and letting , we conclude that
(5.5)
Secondly, it is obvious that
(5.6)
Combining (5.5) and (5.6), we obtain (5.4b). This completes the proof of the lemma.
∎
The following lemma gives error estimates for the -elliptic projections.
Lemma 5.2.
Let . Suppose and satisfies and . Then
(5.7a)
(5.7b)
(5.7c)
Proof.
This lemma has been essentially demonstrated in [26, Theorem 4.3, Theorem 4.4].
To establish the proof of this lemma, it is sufficient to substitute [26, Lemma 4.1] with the above Lemma 5.1.
∎
Using Lemmas 5.1 and 5.2, following the same lines of the proof in
Theorem 4.1, we obtain the following stability estimate for CIP-EEM.
Theorem 5.1.
Let be the solution to (2.7), , .
There exist , such that
if , ,
, and , then
(5.8)
And as a consequence, the CIP-EEM (2.4) is well-posed.
Remark 5.1.
(a) This stability bound of the CIP-EEM solution is also of the same order as
that of the continuous solution (cf. Lemma 3.1) and holds when .
(b) When , it is proved in [26, Theorem 3.1]
that the CIP-EEM is absolutely stable (i.e. stable without any mesh constraint). Moreover,
by using a trick of “stability-error iterative improvement” developed in [15],
it is proved in [26, Theorem 4.6] that the CIP-EEM satisfies the stability estimate
under the conditions , , and
. Clearly, our result gives an improvement of that in [26]
when . When and is
large, the well-posedness of the CIP-EEM (including EEM) is still open.
The following Theorem gives preasymptotic error estimates for CIP-EEM, the proof of which is
similar to that of Theorem 4.2 and is omitted here.
Theorem 5.2.
Assume that is a bounded -domain and strictly star-shaped with respect to
a point and that in and on . Let be the solution to the
problem (1.1)–(1.2) and be the CIP-EEM solution to (2.7).
Then there exist , such that
when , ,
, and ,
the following estimates hold:
(5.9)
(5.10)
Remark 5.2.
(a) An important reason for studying the CIP-EEM is its potential to reduce the pollution
error by tuning the penalty parameters.
(c) The error estimate in the energy norm was also given in [26, Theorem 4.6]
under the conditions , , and .
Our results relax the conditions on the penalty parameters and give a new error estimate
for the CIP-EEM. Such a relaxation is meaningful in practical computations, since the penalty
parameters that can significantly reduce the pollution error are usually close to negative real numbers (see the next section).
6. Numerical example
In this section, we report an example to verify the theoretical findings and to demonstrate the
performance of the EEM and the CIP-EEM, in particular, the potential of the CIP-EEM to significantly
reduce the pollution errors by selecting appropriate penalty parameters.
The Nédélec’s linear edge elements of the second type are used in the numerical tests.
The linear systems resulted from edge element discretizations are solved by pardiso[38],
which is commonly used to solve large, sparse systems of linear equations.
Example.
We simulate the following three-dimensional time-harmonic Maxwell problem:
and and are chosen such that the
exact solution is given by
where is the spherical Hankel function of the first kind of order , is the
surface gradient operator (see, e.g., [31]), ,
are the spherical harmonics of order on the unit sphere (see, e.g., [10]),
, and .
In this numerical example, we triangulate the computational domain into a mesh of the type
cub6 [33] (also called Cube-VI-II in [41, Chapter 4]). Specifically, we first divide
into small cubes of the same size, then divide each small cube into small tetrahedrons as
shown in Figure 1 (left). We note that such a mesh can be simply generated
by delaunayTriangulation in MATLAB on a uniform Cartesian grid.
The penalty parameters for the CIP-EEM are simply taken from the parameters used by the CIP-FEM for the Helmholtz equation [41, Chapter 4],
but with pure imaginary perturbations to enhance stability, i.e., we set
(6.1)
according to three different kinds of interior faces as indicated in Figure 1.
We would like to remark that the real parts
of the penalty parameters are obtained by a dispersion analysis for the CIP-FEM, which is an essential tool
to understand the dispersive behavior of numerical schemes,
and it is commonly believed that the pollution errors are of the same order as the phase difference between the exact and numerical solutions [2, 3, 42, 32, 31, 8].
Such real penalty parameters can reduce phase difference of the CIP-FEM for the Helmholtz equation
on cub6-type meshes from to [41, Chapter 4], and numerical tests
there do show that the pollution errors can also be significantly reduced. Here we expect that the penalty
parameters in (6.1) can significantly reduce the pollution errors of the CIP-EEM. We postpone
the systematic dispersion analysis of CIP-EEM for Maxwell equations to a future work, and only provide
some numerical tests to confirm this expectation.
Figures 2 and 3 illustrate the relative errors and the relative
errors of the EE solutions (left), the
CIP-EE solutions (right), and the interpolations (black dot line) for , , and , respectively.
It demonstrates that when , the errors of the solutions to EEM and CIP-EEM closely match those of
the corresponding EE interpolations, implying the absence of pollution errors for small wave numbers.
Conversely, for large , the relative errors of the EE solutions decay slowly, starting from a point
considerably distant from the decaying point of the corresponding EEM interpolations. This behavior vividly
exposes the presence of pollution errors in the EEM. The CIP-EE solutions exhibit
a similar behavior to the EE solutions, but the pollution range of the former is significantly smaller
than that of the latter.
Figure 4 presents plots of relative errors
with mesh constraint for EEM and CIP-EEM with , respectively.
Note that for small wave number , the errors of EE and CIP-EE solutions closely match those
of the corresponding EE interpolations, implying that pollution errors do not manifest for small
wave numbers. For large values of , the relative errors of EE solutions deteriorate rapidly.
This behavior clearly demonstrates the impact of the pollution error in the EEM.
The CIP-EE solutions behave well till , which shows the pollution effect for this method is
significantly smaller than that of the EEM.
Denote by the interpolation onto the second-type Nédélec
edge element space . The following lemma gives the interpolation error estimates.
Lemma A.1.
We have
(A.1)
(A.2)
(A.3)
(A.4)
Proof.
For (A.1)–(A.2), we refer to [31, Theorems 8.15, 5.41]. The estimate (A.3) follows from the local trace inequality and (A.4) follows from (A.1)–(A.3). The proof is completed.
∎
Without loss of generality, we prove Lemma 3.2 only for . The proof of (3.10a) is obvious and is omitted. Next we prove (3.10b) and (3.10c).
The following lemma gives an estimate of .
Lemma A.2.
It holds
(A.5)
Proof.
Let . Similar to Lemma 3.4, we have the following decompositions (see, e.g., [31, Remark 3.46, Lemma 7.6]):
(A.6)
where , , , and , is divergence-free in , on , and there also holds
(A.7)
where we have used (3.10a) to derive the last inequality.
Next, we establish a relationship between and . Denote by
Inserting the above two inequalities into (A.27), we obtain
(A.28)
which together with Lemma A.2 completes the proof of the lemma.
∎
Finally, the proof of Lemma 3.2 follows by taking in Lemmas A.2–A.3 and using Lemma A.1. ∎
References
[1]
R. A. Adams and J. J. Fournier.
Sobolev spaces.
Elsevier, 2003.
[2]
M. Ainsworth.
Discrete dispersion relation for -version finite element
approximation at high wave number.
SIAM Journal on Numerical Analysis, 42(2):553–575, 2004.
[3]
M. Ainsworth.
Dispersive properties of high–order Nédélec/edge element
approximation of the time–harmonic Maxwell equations.
Philosophical Transactions of the Royal Society of London.
Series A: Mathematical, Physical and Engineering Sciences,
362(1816):471–491, 2004.
[4]
C. Amrouche, C. Bernardi, M. Dauge, and V. Girault.
Vector potentials in three-dimensional non-smooth domains.
Mathematical Methods in the Applied Sciences, 21(9):823–864,
1998.
[5]
I. M. Babuška and S. A. Sauter.
Is the pollution effect of the FEM avoidable for the Helmholtz
equation considering high wave numbers?
SIAM Journal on Numerical Analysis, 34(6):2392–2423, 1997.
[6]
S. Brenner and L. Scott.
The mathematical theory of finite element methods.
Springer, 2008.
[7]
E. Burman.
A unified analysis for conforming and nonconforming stabilized finite
element methods using interior penalty.
SIAM Journal on Numerical Analysis, 43(5):2012–2033, 2005.
[8]
E. Burman, H. Wu, and L. Zhu.
Linear continuous interior penalty finite element method for
Helmholtz equation with high wave number: One-dimensional analysis.
Numerical Methods for Partial Differential Equations, pages
1378–1410, 2016.
[9]
Z. Chen.
On the regularity of time-harmonic Maxwell equations with impedance
boundary conditions.
Commun. Appl. Math. Comput., 2024.
[10]
D. L. Colton, R. Kress, and R. Kress.
Inverse acoustic and electromagnetic scattering theory,
volume 93.
Springer, 1998.
[11]
M. Costabel, M. Dauge, and S. Nicaise.
Corner Singularities and Analytic Regularity for Linear Elliptic
Systems. Part I: Smooth domains.
211 pages, Feb. 2010.
[12]
J. Douglas and T. Dupont.
Interior penalty procedures for elliptic and parabolic Galerkin
methods.
In Computing Methods in Applied Sciences: Second International
Symposium December 15–19, 1975, pages 207–216. Springer, 2008.
[13]
Y. Du and H. Wu.
Preasymptotic error analysis of higher order FEM and CIP-FEM for
Helmholtz equation with high wave number.
SIAM Journal on Numerical Analysis, 53(2):782–804, 2015.
[14]
X. Feng, P. Lu, and X. Xu.
A hybridizable discontinuous Galerkin method for the time-harmonic
Maxwell equations with high wave number.
Computational Methods in Applied Mathematics, 16(3):429–445,
2016.
[15]
X. Feng and H. Wu.
-discontinuous Galerkin methods for the Helmholtz equation
with large wave number.
Math. Comp., 80(276):1997–2024, 2011.
[16]
X. Feng and H. Wu.
An absolutely stable discontinuous Galerkin method for the
indefinite time-harmonic Maxwell equations with large wave number.
SIAM Journal on Numerical Analysis, 52(5):2356–2380, 2014.
[17]
G. N. Gatica and S. Meddahi.
Finite element analysis of a time harmonic Maxwell problem with an
impedance boundary condition.
IMA Journal of Numerical Analysis, 32(2):534–552, 2012.
[18]
R. Hiptmair.
Finite elements in computational electromagnetism.
Acta Numerica, 11:237–339, 2002.
[19]
R. Hiptmair, A. Moiola, and I. Perugia.
Stability results for the time-harmonic Maxwell equations with
impedance boundary conditions.
Mathematical Models and Methods in Applied Sciences,
21:2263–2287, 2011.
[20]
R. Hiptmair, A. Moiola, and I. Perugia.
Error analysis of Trefftz-discontinuous Galerkin methods for the
time-harmonic Maxwell equations.
Mathematics of Computation, 82(281):247–268, 2013.
[21]
P. Houston, I. Perugia, A. Schneebeli, and D. Schötzau.
Interior penalty method for the indefinite time-harmonic maxwell
equations.
Numerische Mathematik, 100:485–518, 2005.
[22]
F. Ihlenburg.
Finite element analysis of acoustic scattering, volume 132 of
Applied Mathematical Sciences.
Springer-Verlag, New York, 1998.
[23]
J.-M. Jin.
Theory and computation of electromagnetic fields.
John Wiley & Sons, 2015.
[24]
P. Lu, H. Chen, and W. Qiu.
An absolutely stable -HDG method for the time-harmonic
Maxwell equations with high wave number.
Mathematics of Computation, 86(306):1553–1577, 2017.
[25]
P. Lu, Y. Wang, and X. Xu.
Regularity results for the time-harmonic Maxwell equations with
impedance boundary condition.
arXiv:1804.07856v1, 2018.
[26]
P. Lu, H. Wu, and X. Xu.
Continuous interior penalty finite element methods for the
time-harmonic Maxwell equation with high wave number.
Adv. Comput. Math., 45(5–6):3265–3291, dec 2019.
[27]
J. M. Melenk and S. Sauter.
Convergence analysis for finite element discretizations of the
Helmholtz equation with Dirichlet-to-Neumann boundary conditions.
Mathematics of Computation, 79(272):1871–1914, 2010.
[28]
J. M. Melenk and S. Sauter.
Wavenumber explicit convergence analysis for Galerkin
discretizations of the Helmholtz equation.
SIAM Journal on Numerical Analysis, 49(3):1210–1243, 2011.
[29]
J. M. Melenk and S. A. Sauter.
Wavenumber-explicit -FEM analysis for Maxwell’s equations
with transparent boundary conditions.
Foundations of Computational Mathematics, 21(1):125–241, 2020.
[30]
J. M. Melenk and S. A. Sauter.
Wavenumber-explicit -FEM analysis for Maxwell’s equations
with impedance boundary conditions.
Foundations of Computational Mathematics, 2023.
[31]
P. Monk et al.
Finite element methods for Maxwell’s equations.
Oxford University Press, 2003.
[32]
P. B. Monk and A. K. Parrott.
A dispersion analysis of finite element methods for Maxwell’s
equations.
SIAM Journal on Scientific Computing, pages 916–937, 1994.
[33]
D. J. Naylor.
Filling space with tetrahedra.
International Journal for Numerical Methods in Engineering,
44(10):1383–1395, 1999.
[34]
J. C. Nédélec.
Mixed finite elements in .
Numer. Math., 35(3):315–341, sep 1980.
[35]
J. C. Nédélec.
A new family of mixed finite elements in .
Numer. Math., 50(1):57–81, nov 1986.
[36]
S. Nicaise and J. Tomezyk.
The time-harmonic Maxwell equations with impedance boundary
conditions in polyhedral domains, pages 285–340.
De Gruyter, Berlin, Boston, 2019.
[37]
S. Nicaise and J. Tomezyk.
Convergence analysis of a -finite element approximation of the
time-harmonic Maxwell equations with impedance boundary conditions in
domains with an analytic boundary.
Numerical Methods for Partial Differential Equations,
36(6):1868–1903, 2020.
[38]
O. Schenk, K. Gärtner, W. Fichtner, and A. D. Stricker.
PARDISO: a high-performance serial and parallel sparse linear
solver in semiconductor device simulation.
Future Gener. Comput. Syst., 18:69–78, 2001.
[39]
H. Wu.
Pre-asymptotic error analysis of CIP-FEM and FEM for the
Helmholtz equation with high wave number. Part I: linear version.
IMA Journal of Numerical Analysis, 34:1266–1288, 2014.
[40]
H. Wu.
FEM and CIP-FEM for Helmholtz equation with high wave number
(in Chinese).
Mathematica Numerica Sinica, 40:191–213, 2018.
[41]
Y. Zhou.
Dispersion analysis of CIP-FEM for Helmholtz problem.
PhD thesis, Nanjing University, 2023.
[42]
Y. Zhou and H. Wu.
Dispersion analysis of CIP-FEM for the Helmholtz equation.
SIAM Journal on Numerical Analysis, 61(3):1278–1292, 2023.
[43]
B. Zhu and H. Wu.
Preasymptotic error analysis of the HDG method for Helmholtz
equation with large wave number.
Journal of scientific computing, pages 63(1–34), 2021.
[44]
L. Zhu and H. Wu.
Preasymptotic error analysis of CIP-FEM and FEM for Helmholtz
equation with high wave number. Part II: version.
SIAM Journal on Numerical Analysis, 51(3):1828–1852, 2013.